SFP Han of Fir Pro Eng 5 TH Edi

free ebooks ==> www.ebook777.
com
Morgan J. Hurley
Editor-in-Chief
SFPE Handbook
of Fire Protection
Engineering
Fifth Edition
www.ebook777.com
free ebooks ==> www.ebook777.com
SFPE Handbook of Fire Protection

Engineering
www.ebook777.com
Morgan J. Hurley
Editor-in-Chief
SFPE Handbook
of Fire Protection
Engineering
Fifth Edition
Editor-in-Chief Editors
Morgan J. Hurley, P.E., FSFPE Daniel Gottuk, Ph.D., P.E.
Aon Fire Protection Engineering Hughes Associates
Greenbelt, MD, USA
John R. Hall Jr., Ph.D.
National Fire Protection Association
Kazunori Harada, Dr. Eng.
Kyoto University
Erica Kuligowski, Ph.D.
National Institute of Standards and Technology
Milosh Puchovsky, P.E., FSFPE
Worcester Polytechnic Institute
José Torero, Ph.D.
The University of Queensland
John M. Watts Jr., Ph.D., FSFPE
The Fire Safety Institute
Christopher Wieczorek, Ph.D.
FM Global
ISBN 978-1-4939-2564-3 ISBN 978-1-4939-2565-0 (eBook)

DOI 10.1007/978-1-4939-2565-0
Library of Congress Control Number: 2015953225
Springer New York Heidelberg Dordrecht London

# Society of Fire Protection Engineers 2016
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or
part of the material is concerned, specifically the rights of translation, reprinting, reuse of
illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way,
and transmission or information storage and retrieval, electronic adaptation, computer software,
or by similar or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are
exempt from the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in
this book are believed to be true and accurate at the date of publication. Neither the publisher nor
the authors or the editors give a warranty, express or implied, with respect to the material
contained herein or for any errors or omissions that may have been made.
Printed on acid-free paper
Springer Science+Business Media LLC New York is part of Springer Science+Business Media
(www.springer.com)
www.ebook777.com
This edition of the SFPE Handbook of Fire Protection

Engineering is dedicated to Philip DiNenno, who was the editor
in chief for the first four editions of this handbook. In the
mid-1980s, Phil DiNenno, Jack Watts, Doug Walton, Craig
Beyler, and Dick Custer had an idea to create a collection
of calculation methods for fire protection engineering.
From this idea emerged the SFPE Handbook of Fire Protection
Engineering, which was first published in 1988.
No other single event had as significant an impact on

establishing the profession of fire protection engineering
as the publication of this handbook. As Vyto Babrauskas said:
“The field [of fire protection engineering] has made very
gratifying progress in these last four decades. . .. The most
remarkable positive achievement I think has been the SFPE
Handbook, published first in 1988. . . . [W]ith the publication
of the first edition of the SFPE Handbook in 1988, all of a
sudden we could properly describe this as a science-based
profession.” [Babrauskas, V. “Some Neglected Areas in Fire
Safety Engineering,” Fire Science and Technology
Vol. 32 No. 1 (2013) pp. 35–48.]
When they began creating the first edition, Phil and his
colleagues had no model other than handbooks used in other
professions. Phil contributed the leadership, vision, and moti-
vation necessary to develop the handbook, and he did so using
entirely volunteer resources. This would be an incredible
accomplishment for anyone. Phil did it before he turned 35.
www.ebook777.com
Foreword
This edition marks a passing of the torch for the SFPE Handbook of Fire
Protection Engineering. All of the editors of the prior editions except for two
(Jack Watts and John Hall) have retired, and a new editorial team has taken
their place. Additionally, Springer has assumed the role of publisher begin-
ning with this edition.
For the first four editions, the SFPE Handbook of Fire Protection Engi-
neering was published by the National Fire Protection Association. The
Society of Fire Protection Engineers owes a debt of gratitude to NFPA.
Without their encouragement and confidence, this handbook might never
have existed.
With a new editorial team emerge many changes. The chapters relevant to
human behavior in fire have been significantly refocused and augmented. The
fundamental engineering chapters have been revised to provide a better
foundation for the chapters that follow. Many new chapters related to fire
protection system selection and design have been added. The chapters
associated with fire resistance design have been modified to reflect advances
over the last decade. And, this edition includes several new chapters pertinent
to industrial fire protection.
The editors owe a debt of gratitude to those whom they follow. Continuing
a successful endeavor is much easier than launching it.
Acknowledgment of Past Authors
Name 1st edition 2nd edition 3rd edition 4th edition

Ahrens, Martha J. ✓ ✓
Alpert, Ronald L. ✓ ✓
Atreya, Arvind ✓ ✓ ✓ ✓
Babrauskas, Vytenis ✓ ✓ ✓ ✓
Back III, Gerard G. ✓ ✓
Barry, Thomas F. ✓ ✓ ✓
Beck, Vaughan R. ✓
Beever, Paula F. ✓ ✓
Beller, Douglas K. ✓ ✓
Berlin, Geoffrey N. ✓
(continued)
vii
viii Foreword
(continued)
Beyler, Craig ✓ ✓ ✓ ✓
Bowen, Jr., Jacob Van ✓
Bryan, John L. ✓ ✓ ✓ ✓
Buchanan, Andy ✓ ✓
Bukowski, Richard W. ✓
Carpenter, Douglas ✓
Chang, Jeremy ✓
Chapman, Robert E. ✓ ✓
Charters, David ✓
Cooper, Leonard Y. ✓ ✓ ✓
Cox, Geoff ✓
Croce, Paul A. ✓ ✓
Custer, Richard L. P. ✓ ✓ ✓
Delichatsios, M. A. ✓ ✓
DiNenno, Philip J. ✓ ✓ ✓
Dodd, F. J. ✓
Donegan, H. A. ✓ ✓ ✓
Drysdale, D. D. ✓ ✓ ✓ ✓
Emmons, Howard W. ✓ ✓ ✓ ✓
Evans, David D. ✓ ✓
Fahy, Rita F. ✓ ✓
Fitzgerald, Robert W. ✓ ✓ ✓ ✓
Fleischmann, Charles ✓ ✓ ✓ ✓
Fleming, Russell P. ✓ ✓ ✓ ✓
Franssen, Jean-Marc ✓
Frantzich, Håkan ✓
Friedman, Raymond ✓ ✓ ✓ ✓
Gottuk, Daniel T. ✓ ✓ ✓
Grant, Casey C. ✓ ✓ ✓ ✓
Gray, Brian ✓ ✓
Griffiths, John ✓
Gwynne, Steven ✓
Hadjisophocleous, George ✓
Hall, Jr., John R. ✓ ✓ ✓ ✓
Harmathy, Tibor Z. ✓ ✓ ✓ ✓
Hasemi, Yuji ✓
Heskestad, Gunnar ✓ ✓ ✓ ✓
Hickey, Harry E. ✓ ✓ ✓ ✓
Hinkley, Peter L. ✓ ✓
Hirschler, Marcelo M. ✓ ✓ ✓
Hu, Yu-Shu ✓
Hurley, Morgan ✓
Hyslop, J. S. ✓
Iwankiw, Nestor ✓
Jaluria, Yogesh ✓ ✓
Janssens, M. ✓ ✓ ✓
Jin, Tadahisa ✓ ✓
Joglar, Francisco ✓ ✓
Johnson, Peter ✓
(continued)
www.ebook777.com
Foreword ix
(continued)
Kandola, B. S. ✓ ✓ ✓ ✓
Kanury, A. Murty ✓ ✓ ✓
Kersken-Bradley, Marita ✓
Klote, John H. ✓ ✓ ✓ ✓
Kodur, V. K. R. ✓ ✓
Kuligowski, Erica ✓
Kumar, Suresh ✓
Lattimer, Brian Y. ✓ ✓
Lee, K. Y. ✓ ✓ ✓ ✓
Lie, T. T. ✓ ✓ ✓
MacLennan, Hamish A. ✓ ✓
Mawhinney, Jack R. ✓ ✓
McCaffrey, Bernard ✓ ✓
McGrattan, Kevin ✓
Meacham, Brian J. ✓ ✓ ✓
Mehaffey, Jim ✓
Miles, Stewart ✓
Milke, James A. ✓ ✓ ✓ ✓
Modarres, Mohammad ✓ ✓
Morgan, Alexander ✓
Mowrer, Frederick W. ✓ ✓
Mudan, Krishna S. ✓ ✓
Mulholland, George W. ✓ ✓ ✓ ✓
Nelson, Harold E. ✓ ✓ ✓
Notarianni, Kathy A. ✓ ✓
Nowlen, Steven ✓
Ohlemiller, T. J. ✓ ✓ ✓ ✓
Parry, Gareth ✓
Pauls, Jake ✓ ✓
Phillips, William G. B. ✓ ✓ ✓
Proulx, Guylène ✓ ✓
Purser, David A. ✓ ✓ ✓ ✓
Quintiere, James G. ✓ ✓ ✓ ✓
Ramachandran, G. ✓ ✓ ✓ ✓
Roby, R. J. ✓
Rockett, John A. ✓ ✓ ✓
Rosenbaum, Eric ✓
Salisbury, Matthew ✓
Scheffey, Joseph L. ✓ ✓ ✓
Schifiliti, Robert P. ✓ ✓ ✓ ✓
Simmons, Robert F. ✓ ✓ ✓
Siu, Nathan ✓ ✓
Stretton, A. J. ✓ ✓ ✓ ✓
Stroup, David W. ✓
Tanaka, Takeyoshi ✓
Tewarson, Archibald ✓ ✓ ✓ ✓
Thomas, Ian ✓
Thomas, Philip H. ✓ ✓ ✓ ✓
(continued)
x Foreword
(continued)
Tien, C. L. ✓ ✓ ✓ ✓
Titus, John J. ✓ ✓ ✓ ✓
Torero, José ✓
Walton, William D. ✓ ✓ ✓ ✓
Watts, Jr., John M. ✓ ✓ ✓ ✓
White, Derek A. ✓ ✓
White, Robert H. ✓ ✓ ✓ ✓
Wickström, Ulf ✓
Wolski, Armin ✓
Wood, Christopher ✓
Yung, David ✓
Zalosh, Robert G. ✓ ✓ ✓ ✓
www.ebook777.com
Metrication
The editors of the SFPE Handbook of Fire Protection Engineering have

worked toward the expanded use of SI units for this fifth edition. In some
instances, however, US customary units have been retained. For example,
when equations, correlations, or design methodologies have input variables
or constants that have been developed from data originally in US customary
units, those units are retained. This is also the case for certain tables, charts,
and nomographs. Where equations employing US customary units are used
in worked examples, the results are presented as SI units as well.
xi
www.ebook777.com
Contents
Volume I
1 Introduction to Fluid Mechanics . . . . . . . . . . . . . . . . . . . . 1

Bart Merci
2 Conduction of Heat in Solids . . . . . . . . . . . . . . . . . . . . . . . 25
Ofodike A. Ezekoye
3 Convection Heat Transfer . . . . . . . . . . . . . . . . . . . . . . . . . 53
Arvind Atreya
4 Radiation Heat Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . 102
Revised by C. Lautenberger, Original chapter
authored by C.L. Tien, K.Y. Lee, and A.J. Stretton
5 Thermochemistry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
D.D. Drysdale
6 Chemical Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Raymond Friedman
7 Thermal Decomposition of Polymeric Materials . . . . . . . . 167
Artur Witkowski, Anna A. Stec, and T. Richard Hull
8 Structural Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
Luke A. Bisby
9 Properties of Building Materials . . . . . . . . . . . . . . . . . . . . 277
V.K.R. Kodur and T.Z. Harmathy
10 Chemical Kinetics and Fire . . . . . . . . . . . . . . . . . . . . . . . . 325
Gregory T. Linteris and John F. Griffiths
11 Diffusion Flames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
Ali S. Rangwala
12 Fundamentals of Premixed Flames . . . . . . . . . . . . . . . . . . 373
Grunde Jomaas
13 Fire Plumes, Flame Height, and Air Entrainment . . . . . . . 396
Gunnar Heskestad
xiii
xiv Contents
14 Ceiling Jet Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429

Ronald L. Alpert
15 Vent Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455
Takeyoshi Tanaka
16 Effect of Combustion Conditions on Species
Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 486
Daniel T. Gottuk and Brian Y. Lattimer
17 Flammability Limits of Premixed and Diffusion
Flames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 529
Craig Beyler
18 Ignition of Liquids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 554
D.D. Drysdale
19 Smoldering Combustion . . . . . . . . . . . . . . . . . . . . . . . . . . 581
Guillermo Rein
20 Spontaneous Combustion and Self-Heating . . . . . . . . . . . . 604
Brian F. Gray
21 Flaming Ignition of Solid Fuels . . . . . . . . . . . . . . . . . . . . . 633
José Torero
22 Electrical Fires . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 662
Vytenis Babrauskas
23 Surface Flame Spread . . . . . . . . . . . . . . . . . . . . . . . . . . . . 705
Yuji Hasemi
24 Smoke Characterization and Damage Potentials . . . . . . . . 724
Jeffrey S. Newman, Geary G. Yee, and Paul Su
25 Heat Transfer from Fires to Surfaces . . . . . . . . . . . . . . . . 745
Brian Y. Lattimer
26 Heat Release Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 799
Vytenis Babrauskas
27 Calorimetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 905
Marc Janssens
28 The Cone Calorimeter . . . . . . . . . . . . . . . . . . . . . . . . . . . . 952
Vytenis Babrauskas
29 Compartment Fire Modeling . . . . . . . . . . . . . . . . . . . . . . . 981
James G. Quintiere and Colleen A. Wade
30 Estimating Temperatures in Compartment Fires . . . . . . . 996
William D. Walton, Philip H. Thomas,
and Yoshifumi Ohmiya
31 Zone Computer Fire Models for Enclosures . . . . . . . . . . . 1024
William D. Walton, Douglas J. Carpenter,
and Christopher B. Wood
www.ebook777.com
Contents xv
32 Modeling Fires Using Computational Fluid

Dynamics (CFD) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1034
Kevin McGrattan and Stewart Miles
33 Enclosure Smoke Filling and Fire-Generated
Environmental Conditions . . . . . . . . . . . . . . . . . . . . . . . . . 1066
Frederick W. Mowrer
34 Methods for Predicting Temperatures
in Fire-Exposed Structures . . . . . . . . . . . . . . . . . . . . . . . . 1102
Ulf Wickström
35 Fire Load Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1131
Mario Fontana, Jochen Kohler, Katharina Fischer,
and Gianluca De Sanctis
36 Combustion Characteristics of Materials
and Generation of Fire Products . . . . . . . . . . . . . . . . . . . . 1143
Mohammed M. Khan, Archibald Tewarson,
and Marcos Chaos
Volume II
37 Performance-Based Design . . . . . . . . . . . . . . . . . . . . . . . . 1233
Morgan J. Hurley and Eric R. Rosenbaum
38 Fire Scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1262
George V. Hadjisophocleous and Jim R. Mehaffey
39 Engineering Considerations for Fire Protection
System Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1289
Milosh Puchovsky and Craig Hofmeister
40 Design of Detection Systems . . . . . . . . . . . . . . . . . . . . . . . 1314
Robert P. Schifiliti, Richard L.P. Custer,
and Brian J. Meacham
41 Hydraulics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1378
Kenneth E. Isman
42 Automatic Sprinkler System Calculations . . . . . . . . . . . . . 1423
Russell P. Fleming
43 Halon Design Calculations . . . . . . . . . . . . . . . . . . . . . . . . . 1450
Casey C. Grant
44 Clean Agent Total Flooding Fire
Extinguishing Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1483
Philip J. DiNenno and Eric W. Forssell
45 Carbon Dioxide Systems . . . . . . . . . . . . . . . . . . . . . . . . . . 1531
Jeff Harrington and Joseph A. Senecal
46 Water Mist Fire Suppression Systems . . . . . . . . . . . . . . . . 1587
Jack R. Mawhinney and Gerard G. Back III
xvi Contents
47 Foam Agents and AFFF System Design

Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1646
Joseph L. Scheffey
48 Foam System Calculations . . . . . . . . . . . . . . . . . . . . . . . . . 1707
Hamid R. Bahadori
49 Considerations for Coordinating and Interfacing
Fire Protection and Life Safety Systems . . . . . . . . . . . . . . 1740
David Jacoby, David LeBlanc, Jeffrey Tubbs,
and Andrew Woodward
50 Smoke Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1785
John H. Klote
51 Smoke Control by Mechanical Exhaust
or Natural Venting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1824
James A. Milke
52 Structural Fire Engineering of Building
Assemblies and Frames . . . . . . . . . . . . . . . . . . . . . . . . . . . 1863
Jean-Marc Franssen and Nestor Iwankiw
53 Analytical Methods for Determining Fire Resistance
of Steel Members . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1909
James A. Milke
of Concrete Members . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1949
Charles Fleischmann, Andy Buchanan, and Anthony Abu
of Timber Members . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1979
Robert H. White
56 Egress Concepts and Design Approaches . . . . . . . . . . . . . . 2012
Richard W. Bukowski and Jeffrey S. Tubbs
57 Selecting Scenarios for Deterministic Fire Safety
Engineering Analysis: Life Safety for Occupants . . . . . . . 2047
Daniel Nilsson and Rita Fahy
58 Human Behavior in Fire . . . . . . . . . . . . . . . . . . . . . . . . . . 2070
Erica D. Kuligowski
59 Employing the Hydraulic Model in Assessing
Emergency Movement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2115
Steven M.V. Gwynne and Eric R. Rosenbaum
60 Computer Evacuation Models for Buildings . . . . . . . . . . . 2152
Erica D. Kuligowski
61 Visibility and Human Behavior in Fire Smoke . . . . . . . . . 2181
Tokiyoshi Yamada and Yuki Akizuki
www.ebook777.com
Contents xvii
62 Combustion Toxicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2207

David A. Purser
Volume III
63 Assessment of Hazards to Occupants from Smoke,
Toxic Gases, and Heat . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2308
David A. Purser and Jamie L. McAllister
64 Engineering Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2429
S.M.V. Gwynne and K.E. Boyce
65 Liquid Fuel Fires . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2552
D.T. Gottuk and D.A. White
66 Fire Hazard Calculations for Large, Open
Hydrocarbon Fires . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2591
Craig L. Beyler
67 Vapor Clouds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2664
Nicolas F. Ponchaut, Francesco Colella,
and Kevin C. Marr
68 Effects of Thermal Radiation on People:
Predicting 1st and 2nd Degree Skin Burns . . . . . . . . . . . . 2705
Christopher J. Wieczorek and Nicholas A. Dembsey
69 Flammable Gas and Vapor Explosions . . . . . . . . . . . . . . . 2738
Robert Zalosh
70 Dust Explosions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2766
Robert Zalosh
71 BLEVES and Fireballs . . . . . . . . . . . . . . . . . . . . . . . . . . . 2792
Alfonso Ibarreta, Hubert Biteau, and Jason Sutula
72 Introduction to Fire Risk Analysis . . . . . . . . . . . . . . . . . . . 2817
John M. Watts Jr. and John R. Hall Jr.
73 Probability and Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . 2827
John R. Hall Jr. and Francisco Joglar
74 Reliability, Availability, and Maintainability . . . . . . . . . . . 2875
Francisco Joglar
75 Building Fire Risk Analysis . . . . . . . . . . . . . . . . . . . . . . . . 2941
Brian J. Meacham, David Charters, Peter Johnson,
and Matthew Salisbury
76 Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2992
Kathy A. Notarianni and Gareth W. Parry
77 Decision Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3048
H.A. Donegan
xviii Contents
78 Data for Engineering Analysis . . . . . . . . . . . . . . . . . . . . . . 3073

Marty Ahrens and John R. Hall Jr.
79 Measuring Consequences in Economic Terms . . . . . . . . . . 3098
G. Ramachandran and John R. Hall Jr.
80 Computer Simulation for Fire Risk Analysis . . . . . . . . . . . 3117
William G.B. Phillips and Rita F. Fahy
Revised by Douglas K. Beller
81 Engineering Economics . . . . . . . . . . . . . . . . . . . . . . . . . . . 3137
John M. Watts, Jr. and Robert E. Chapman
82 Fire Risk Indexing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3158
John M. Watts Jr.
83 Risk-Informed Industrial Fire Protection Engineering . . . 3183
Thomas F. Barry
84 Product Fire Risk Analysis . . . . . . . . . . . . . . . . . . . . . . . . 3211
John R. Hall Jr.
85 Health Care Application of Quantitative Fire
Risk Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3226
Håkan Frantzich
86 The Building Envelope: Fire Spread, Construction
Features and Loss Examples . . . . . . . . . . . . . . . . . . . . . . . 3242
Daniel J. O’Connor
87 Wildland Fires . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3283
Albert Simeoni
88 Fires in Vehicle Tunnels . . . . . . . . . . . . . . . . . . . . . . . . . . 3303
Ricky Carvel and Haukur Ingason
89 Fire Risk Analysis for Nuclear Power Plants . . . . . . . . . . . 3326
Nathan O. Siu, Nicholas Melly, Steven P. Nowlen,
and Mardy Kazarians
90 Fire Risk in Mass Transportation . . . . . . . . . . . . . . . . . . . 3370
Armin Wolski and Jarrod Alston
Appendix 1 Conversion Factors . . . . . . . . . . . . . . . . . . . . . . . 3397

Appendix 2 Thermophysical Property Data . . . . . . . . . . . . . . 3425
Appendix 3 Fuel Properties and Combustion Data . . . . . . . . 3437
Appendix 4 Configuration Factors . . . . . . . . . . . . . . . . . . . . . 3476
Appendix 5 Piping Properties . . . . . . . . . . . . . . . . . . . . . . . . 3483
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3493
www.ebook777.com
Introduction to Fluid Mechanics

1
Bart Merci
Fluid Properties certain time is defined as in Equation 1.1, taking

the local limit for a small volume.
In this section, a number of fluid properties are In an incompressible flow, the density does
defined. An implicit assumption in the classical not vary. In general, liquids can be considered
fluid mechanics is the ‘continuum hypothesis’, ‘incompressible’. In gases, the density can vary
implying that we treat fluids as continuous due to variations in pressure or temperature (see
media, not as an ensemble of individual below: ideal gas law).
molecules [1]. This is justified in ‘normal’ The reciprocal of density is the ‘specific vol-
circumstances. This way, the fluid and flow ume’ (m3/kg).
quantities are continuous and local quantities to
be interpreted as averages over a volume V*
Viscosity
which is very small (but still very large when
compared to distances between molecules).
Fluids can flow. The viscosity is the fluid prop-
This assumption allows to define local fluid and
erty that indicates how easily molecules can
flow properties (e.g. velocity vectors). The con-
move with respect to each other. Fluid particles
tinuum hypothesis is adopted here.
with different velocity have the tendency to
A fluid can be a liquid or a gas (vapour).
evolve to the same common velocity, through
exchange of momentum. In other words, fluid
layers with different velocities exert a shear
Density
stress τ onto each other. Most technically rele-
vant fluids are ‘Newtonian’: the shear stress
The mass density is the amount of fluid mass
increases linearly with the strain rate
inside a volume:
(or velocity gradient):
m
ρ¼ : ð1:1Þ dv
V τ¼μ : ð1:2Þ
dy
Its unit is kg/m3.
In a variable density flow, the density can vary The unit of τ is Pa (¼ N/m2).
in space and time and the local density at a The proportionality factor, relating the veloc-
ity gradient to the shear stress, is the dynamic
viscosity μ (unit: Pa.s).
In gases, μ typically increases with tempera-
B. Merci (*)
Department of Flow, Heat and Combustion Mechanics, ture, whereas in liquids it decreases with increas-
Ghent University, Ghent, Belgium ing temperature.
M.J. Hurley (ed.), SFPE Handbook of Fire Protection Engineering, 1

DOI 10.1007/978-1-4939-2565-0_1, # Society of Fire Protection Engineers 2016
2 B. Merci
Sometimes, the kinematic viscosity is used: This is Fourier’s law. The minus sign indicates
μ that the heat flux is always from high temperature
ν¼ : ð1:3Þ to low temperature.
ρ
The unit of the conduction coefficient (k or λ)
Its unit is m2/s. is W/(m.K).
The shear stress, Equation 1.2, causes friction The conduction coefficient, specific heat and
losses in case of flow. The higher the viscosity, density can be combined to obtain the thermal
the larger the flow losses become for the same diffusivity:
velocity gradient. In other words, the resistance
k
of the fluid against (imposed) flow increases with α¼ ð1:6Þ
ρc
increasing viscosity.
The viscosity of a fluid is never zero. The
important implication is that, whenever there is The unit of α is m2/s.
a solid boundary, this boundary always exerts an
influence on the flow field (e.g. causing the
development of a boundary layer). Diffusion Coefficient
In a mixture of fluids (see below), one species

Specific Heat can diffuse in the mixture due to concentration
gradients of that species in the mixture. It is
The specific heat or thermal capacity, c, is the common practice to apply Fick’s law for many
amount of energy required to cause a temperature flows:
rise of 1 K (or 1 C) in 1 kg of the fluid. Its unit is
!
J/(kg.K). Jk ¼ ρDk ∇Y: ð1:7Þ
In gases the value of the specific heat depends
on the circumstances under which the energy is The diffusion coefficient D thus provides the
supplied. If the pressure is kept constant, the relation between the diffusion flux Jk (kg/(m2s))
notation is cp. If the volume is kept constant, of species k and the spatial gradient of the local
the notation is cv. The difference between the mass fraction Yk (i.e. the amount of mass of
two values is called the gas constant R (also in species k per kg mixture) of that species. The
J/(kg.K)): minus sign expresses that the diffusion flux is
c p ¼ cV þ R: ð1:4Þ always from higher concentration to lower
concentration.
The unit of D is m2/s.
For liquids and solids, cp cv.
Conduction Coefficient Dimensionless Groups of Fluid

Properties
The conduction coefficient expresses how easily
heat flows inside a material. Its value indicates By combining the fluid properties, dimension-
the heat flux per unit area (W/m2) related to a less groups can be constructed. Indeed, the
spatial temperature gradient (K/m): units of ν, α and D are the same (m2/s). Physi-
! cally, the interpretation is that ν tries to make
q_ ¼ k∇T ¼ λ∇T: ð1:5Þ the velocity field uniform inside a fluid
www.ebook777.com
1 Introduction to Fluid Mechanics 3
(through exchange of momentum), α tries to Internal Energy

make the temperature field uniform (through
heat exchange by conduction) and D tries to The local motion of molecules in a fluid is related
make the concentration field in a mixture to the internal energy (e or u, with unit J/kg). This
homogeneous (through concentration gradient is a measure for the thermal energy.
driven diffusion).
The resulting dimensionless groups read:
– The Prandtl number: Enthalpy
ν μc p μc p
Pr ¼ ¼ ¼ : ð1:8Þ The quantity (static) enthalpy (h, with unit J/kg)
α λ k
is related to the internal energy through addition
– The Schmidt number: of pressure, divided by mass density:
ν p p
Sc ¼ : ð1:9Þ h¼uþ ¼eþ : ð1:11Þ
D ρ ρ
– The Lewis number:
α Entropy
Le ¼ : ð1:10Þ
D
The entropy is a measure for the disorder in the
Clearly, these numbers are connected:
fluid. It is related to the second law of thermody-
Le ¼ Sc:Pr1 .
namics. This quantity is typically not particularly
It is important to note that the dimensionless
relevant for fire related issues.
numbers Equations 1.8, 1.9, and 1.10 are still
fluid properties, not flow properties.
As long as no mixtures are considered, the
Prandtl number is the most relevant dimension- Equation of State
less fluid property, when heat transfer is an issue.
Liquids
In liquids, the density is essentially constant,

State Properties
relatively very weakly dependent on pressure
and temperature. Yet, the general expression
State properties describe the state of the fluid, not
that provides the equation of state defines the
the material properties of the fluid.
relation between density, temperature and
pressure:
Pressure ρ ¼ f ð p; T Þ ð1:12Þ
The pressure (p) can be defined as the normal

force per unit area at a certain point. The unit is Gases: Ideal Gas Law
Pa. Pressure differences are the driving force for
fluid flows. In gases, it is common practice to specify Equa-
tion 1.12 as the ‘ideal gas law’:
Temperature p ¼ ρRT ð1:13Þ
The unit of temperature (T) is Kelvin (K). The For fire related flows, this is justified. Most gases
temperature must not be confused with heat (the behave as air would do and air behaves as an
unit of which is Joule, J). ideal gas (with the exception of extremely low or
4 B. Merci
high pressure or temperature, but this is not rele-

vant for real-life fire applications). The gas con- Conservation Equations
stant R (J/(kg.K)) has been introduced in
Equation 1.4 and the temperature T is expressed Figure 1.1 visualises a streamline through a sur-
in Kelvin (K). face of a (control) volume. This concept will be
used to develop the conservation equations in the
integral formulation. A streamline is defined
such that locally the velocity vector is tangent
Mixtures to the streamline. A collection of streamlines is
called a stream tube.
In fire related flows, the fluid can be a mixture.
Obvious examples are smoke or flames. A dis-
tinction must be made between chemical and Conservation of Mass—Continuity
physical issues. If toxicity is an issue, chemical Equation
aspects are important. As long as the flow itself is
concerned, the physical behaviour of many gas- Conservation of mass expresses the following
eous mixtures resembles very much the principle:
behaviour of hot air. One reason is that the spe- The amount of mass that flows into a station-
cies most often encountered, have comparable ary volume per unit time, equals the outflow of
diffusivities (with the important exception of mass per unit time out of that same volume plus
hydrogen, which has a much higher diffusivity). the amount of mass accumulation per unit time in
Another reason is that typically by far mixtures that same volume.
in fire related flows consist mainly of air. Mathematically, this is formulated as follows:
As a consequence, the simplification is made • The net outflow per unit time is given by a
very commonly to treat a mixture of hot gases as closed surface integral over the entire area of
hot air, applying the ideal gas law (Equation 1.13) the manifold ∂V, enclosing the volume V:
with the gas constant for air and using the (tem- ðð
!! !
perature dependent) viscosity for hot air. There- ρ v :n dA; in this expression, v is the local
fore, mixtures of gases do not receive much ∂V
!
attention when fluid mechanics aspects are con- velocity vector at a certain position on ∂V, n
sidered in case of fire. the local normal vector on the surface (i.e. the
Yet, a few definitions are introduced here. The vector with length equal to 1, locally perpen-
mass fraction Yi of species i is the ratio of the dicular to the surface and pointed outward)
local amount of mass of species i to the local and dA the area of an infinitesimal element
amount of mass of mixture. It is therefore a
non-dimensional quantity. Conservation of mass
leads to the statement that, everywhere in physi-
cal space, the sum of all mass fractions of all
X
i¼1 Y i ¼ 1.
N
species equals unity: dA
Using the notion of mass fractions, the v
θ
fluid properties of mixtures can be determined n
from the fluid properties of their constituent
species. E.g. the specific heat becomes Fig. 1.1 Streamline through a surface. Notation: dA is
X the area of an infinitesimal part of the surface; n is the
c¼ N
i¼1 Y i ci . normal vector, with length equal to unity, perpendicular to
Also state properties can be defined as such. dA and pointing ‘outward’ of the control volume, spanned
X
E.g. static enthalpy becomes: h ¼ N
i¼1 Y i hi .
by the surface; v is the local flow velocity vector at
position dA; θ is the angle between vectors n and v
www.ebook777.com
!!
on the surface; note that the inner product v :n A2, v2, ρ2
!!
> 0 for outflow, while v :n < 0 for inflow; A1, v1, ρ1
• The accumulation of mass per unit time is
obtained from a derivation with respect to
time of the integral of the mass density over
∂
the entire volume: ∂t ∭ ρdV.
V
The conservation of mass thus reads: Fig. 1.2 Illustration of conservation of mass for steady
ðð flow (Equation 1.15) through a pipe expansion. Dashed
∂ !! lines: boundary of control volume. Bold vectors: normal
∭ ρdV þ ρ v :n dA ¼ 0: ð1:14Þ
∂t V vectors (unity length, perpendicular to surface and
∂V pointing outward). The other vectors indicate velocity
vectors
This equation is also called the continuity
equation.
An important simplification is found in the The integral in Equation 1.14 in fact refers to
case of permanent (or ‘steady’) motion. In that the total net mass flow rate (kg/s) through a
case, the time derivative disappears in surface with area A:
Equation 1.14: ðð
ðð !!
m_ ¼ ρ v :n dA ð1:17Þ
!!
ρ v :n dA ¼ 0: ð1:15Þ
A
∂V
If the mass density is not included, the total net
A further simplification concerns incompressible volume flow rate (m3/s) through a surface with
fluids (e.g. water in a pipe under normal area A is found:
conditions). In that case, density does not change, ðð
!!
so that not only Equation 1.15 applies, but it V_ ¼ v :n dA: ð1:18Þ
further simplifies to read:
A
ðð
!!
v :n dA ¼ 0: ð1:16Þ Expression (1.14) can also be formulated in dif-
∂V
ferential form, applying Green’s theorem:
A very simple illustration of Equation 1.15 is !

provided on the basis of Fig. 1.2. There is no ∂ρ
þ ∇: ρ v ¼ 0: ð1:19Þ
flow through the solid boundaries (solid lines in ∂t
ðð
!!
Fig. 1.2), so the only contributions to ρ v :n dA The symbol ∇ is the divergence operator:

∂V ! ∂! ∂ ! ∂! ! ! !
∇: v ¼ 1x þ 1 y þ 1 z : vx 1 x þ v y 1 y þ vz 1 z
stem from surfaces 1 and 2. In surface 1, the ∂x ∂y ∂z
velocity vector is pointing inward, while the ∂vx ∂v y ∂vz
¼ þ þ :
normal vector is by definition pointing outward, ∂x ∂y ∂z
so the contribution (under the simplified assump- ð1:20Þ
tion of uniform flow through the cross-section)
!
becomes: ρ1 v1 A1 . On surface 2, the velocity In Equation 1.20, 1 x is the notation for the unity
and the normal vectors are pointing outward, vector, i.e. a vector with length equal to unity, in
leading to: þρ2 v2 A2 . Equation 1.15 thus provides: the x-direction.
ρ1 v1 A1 þ ρ2 v2 A2 ¼ 0 ! ρ1 v1 A1 ¼ ρ2 v2 A2 . In Expression (1.15), for steady flow, reads in
case of incompressible flow (Equation 1.16) this differential form:
further simplifies to: v1 A1 ¼ v2 A2 .
6 B. Merci
!
∇: ρ v ¼ 0; ð1:21Þ σyy (y+dy)
y
τyx (y+dy)
while expression (1.16), for incompressible
fluids, becomes:
! σxx (x) τxy (x+dx)
∇: v ¼ 0: ð1:22Þ dy
τxy (x) σxx (x+dx)

This shows that the velocity field for any flow of
an incompressible fluid is ‘divergence free’, or dx
‘solenoidal’. τyx (y)

σyy (y)
x
Total Momentum
Fig. 1.3 Definition of normal stresses and shear stresses
(2D)
Now the integral formulation for the conserva-
tion of total momentum is discussed. Figure 1.1
The total force consists of:
again serves as the basic sketch.
• Surface forces:
Conservation of total momentum refers to the
– Pressure (Pa);
expression of Newton’s second law, applied to
– Viscous stresses (Pa);
flows. The net change in momentum of a system
• Body forces:
per unit time in a certain sense and direction
– Gravity (N);
equals the net force on that system in that sense
– Others (not relevant for fire related flows).
and direction.
These forces are discussed now, in differential
Expressed for a stationary volume, this
formulation:
becomes:
8
The total force onto a stationary volume > ∂σ xx ∂τxy ∂τxz
>
> F ¼ þ þ þ ρgx
equals the sum of the net outflow of momentum > tot, x
> ∂x ∂y ∂z
>
>
per unit time out of that same volume plus the <
∂τ yx ∂σ yy ∂τ yz
accumulation of momentum per unit time in that Ftot, y ¼ þ þ þ ρg y ð1:25Þ
>
> ∂x ∂y ∂z
same volume. >
>
>
>
The local amount of momentum per unit vol- : Ftot, z ¼ ∂τzx þ ∂τzy þ ∂σ zz þ ρgz
>
! ∂x ∂y ∂z
ume is ρ v (kg/(m2s)). Newton’s second law thus
reads: The final terms in Equation 1.25 refer to
the gravity acceleration vector, multiplied with
ðð
∂ ! ! !! ! the local mass density. Figure 1.3 shows how the
∭ ρ v dV þ ρ v v :n dA ¼ F tot : ð1:23Þ
∂t V normal stresses and shear stresses are defined.
∂V The shear stresses are found from Stokes’ law:
Note that Equation 1.19 is a vector equation,
∂vx ∂v y
i.e. the equation is valid for each component/ τxy ¼ τ yx ¼ μ þ
∂y ∂x
direction individually.
For a permanent (or ‘steady’) motion, expres- ∂vx ∂vz
τxz ¼ τzx ¼ μ þ ð1:26Þ
sion (1.19) simplifies to: ∂z ∂x

ðð ∂vz ∂v y
! !! ! τ yz ¼ τzy ¼ μ þ :
ρ v v : n dA ¼ F tot : ð1:24Þ ∂y ∂z
∂V
www.ebook777.com

The shear stresses are thus proportional to the 2 ∂vx 1 !
σ xx ¼ p þ μ ∇: v
dynamic viscosity and the local velocity 3 ∂x 3
gradients.
2 ∂v y 1 !
The normal stresses contain contributions σ yy ¼ p þ μ ∇: v ð1:27Þ
3 ∂y 3
from stresses due to fluid dilatation (for variable
density flows only) and pressure: 2 ∂vz 1 !
σ zz ¼ p þ μ ∇: v :
3 ∂z 3
The above results in the Navier–Stokes

equations:
8
> ∂ ∂vx ∂vx ∂vx ∂p 2 ∂ ∂vx 1 ! ∂τxy ∂τxz
>
> ðρvx Þ þ ρvx þ ρv y þ ρvz ¼ þ μ ∇: v þ þ þ ρgx
>
> ∂t ∂x ∂y ∂z ∂x 3 ∂x ∂x 3 ∂y ∂z
>
>
>
<
∂ ∂v y ∂v y ∂v y ∂p 2 ∂ ∂v y 1 ! ∂τxy ∂τ yz
ρv y þ ρvx þ ρv y þ ρvz ¼ þ μ ∇: v þ þ þ ρg y
>
> ∂t ∂x ∂y ∂z ∂y 3 ∂y ∂y 3 ∂x ∂z
>
>
>
>∂
>
> ∂vz ∂vz ∂vz ∂p 2 ∂ ∂vz 1 ! ∂τxz ∂τ yz
: ðρvz Þ þ ρvx þ ρv y þ ρvz ¼ þ μ ∇: v þ þ þ ρgz
∂t ∂x ∂y ∂z ∂z 3 ∂z ∂z 3 ∂x ∂y
ð1:28Þ
Note that the presence of the gravity force is The change (per unit time) of the total internal
essential in order to account for the Archimedes energy of a system equals the sum of the heat
force. This is essential for buoyancy-driven added (per unit time) to the system and the work
forces, which is important in the context of fire. (per unit time) exerted onto that system.
Also note that pressure gradients (or pressure The total internal energy consists of:
differences) are the driving force for flows, not • Static internal energy e (J/kg) or ρe (J/m3);
the absolute pressure level. • Kinetic energy ρv2/2 (J/m3).
The mathematical formulation of the first law
of thermodynamics for a stationary open system
Energy
can be found in many textbooks (e.g. [2–9]).
It reads:
Conservation of energy refers to the first law of
thermodynamics:
ð !
!
∂ 1 1 !! !! ! ! !
ρe þ ρv2 dV ¼ ∮ ρe þ ρv2 v :n dS ∮ p v :n dS þ ∮ τ : v :n dS
∂t 2 ∂V 2 ∂V ∂V
V
ð ð
!! !!
þ ρg : v dV þ ρSh dV ∮ q :n dS ð1:29Þ
∂V
V V
The terms on the right hand side are: sign convection (see previous sections: the
• First term: Net inflow of total internal energy normal vector is pointing outward).
into the control volume (‘convection’); the • Second term: Work of the flow against pres-
minus sign is necessary to comply with the sure. This is work from a force (pressure),
8 B. Merci
! !
exerted onto the surface. The work by the
so that with g ¼ g1 no minus sign is
pressure onto the flow is positive for inflow y
and negative for outflow, which explains the required in this term (if the y-direction is
minus sign. positive vertically upward).
• Third term: Work by the viscous stresses. This • Fifth term: Volumetric source term of heat /
is work from a force (viscous stresses, internal energy (e.g. radiation). This term can
Equations 1.26 and 1.27, exerted onto the be positive or negative.
surface. With the sign conventions used • Final term: Net incoming flux of heat/internal
(Fig. 1.3 and outward pointing normal vector), energy (e.g. conduction). The flux with the
this is a term with a plus sign. flow cannot be added to this term (as it is
• Fourth term: Work by gravity. This is work by already included in the convection term).
a volume force, exerted inside the volume. The energy equation can also be formulated,
This work is positive for a downward flow, using enthalpy Equation 1.11:
ð ð !
!
∂ 1 ∂ 1 !! ! ! !
ρh þ ρv2 dV ¼ pdV ∮ ρh þ ρv2 v :n dS þ ∮ τ : v :n dS
∂t 2 ∂t ∂CV 2 ∂CV
CV CV
ð ð
!! !!
þ ρ g : v dV þ ρSh dV ∮ q :n dS: ð1:30Þ
∂CV
CV CV
In differential formulation, this reads:
!
∂ 1 2 1 2 ! ∂p
!
!! XN
!! !
ρh þ ρv þ ∇: ρ h þ v v ¼ þ ∇: τ : v þ ρSh þ ρ Y i g : v i ∇:q ð1:31Þ
∂t 2 2 ∂t i¼1
The (static) enthalpy is the mass-weighted sum combustion. Combustion reactions transform
of the enthalpies of species i: chemically bound enthalpy into sensible
enthalpy and as such cause a temperature rise,
X
N
but the sum of sensible and chemical enthalpy
h¼ Y i hi : ð1:32Þ
i¼1 does not change locally. If the energy equation is
expressed in terms of temperature (or sensible
The enthalpy hi is the sum of a reference enthalpy), a source term due to the combustion
enthalpy (the chemical standard formation heat release rate does appear.
enthalpy of species i) and a ‘sensible’ (thermal) The final term in Equation 1.31 reads:
enthalpy [5–9]. For ideal gases this reads:
!
ðT ! X
N
hi ðT Þ ¼ hre f , i þ c p, i ðT ÞdT; ð1:33Þ ∇: q ¼ ∇:ðλ∇T Þ ∇: ρ hi Di ∇Y i þ D:E:

T re f i¼1
ð1:34Þ
with cp,i the specific heat of species i, defined
above.
The abbreviation ‘D.E.’ stands for the ‘Dufour
Note that in Equation 1.31, expressed in terms
effect’, i.e. and additional enthalpy flux due to
of enthalpy, the source term ρSh contains
species concentration differences. This effect is
e.g. radiation, but not a heat release rate due to
www.ebook777.com
ignored in fire related flows. The first terms in low-Mach number flows (note: this is not true for
Equation 1.34 refer to Fourier’s law for heat explosions). The time derivative of pressure can
conduction, Equation 1.5. The middle terms often be ignored. Also the work done by gravity,
refer to an enthalpy flux due to diffusion, using by the viscous shear stresses and by the normal
Fick’s law, Equation 1.7. stresses becomes very small and the kinetic
The general expression, Equation 1.31, can energy is negligible. Using Equations 1.8 and
often be simplified. Many fire-induced flows are 1.9, the energy equation becomes:
!
! XN
∂ μ 1 1
ðρhÞ þ ∇: ρh v ¼ ∇: ∇h þ μ hi ∇Y i þ ρSh ð1:35Þ
∂t Pr i¼1
Sci Pr
For unity Lewis number (Lei ¼ 1 for all i, Buoyancy

Equation 1.10) fluids, this further simplifies to:
! μ The main relevance of the fundamental law of
∂
ðρhÞ þ ∇: ρh v ¼ ∇: ∇h þ ρSh : hydrostatics, Equation 1.37, lies in the fact that in
∂t Pr many fire related flows, buoyancy plays a domi-
ð1:36Þ nant role. This can be learnt from the
Navier–Stokes equations, Equation 1.28, combin-
ing the forces due to pressure gradients and grav-
! !
Hydrostatics ity. In the vertical direction (still with g ¼ g1 y),
using Equation 1.38, the resulting force per unit
Hydrostatics area reads:
dp
From the general Navier–Stokes equations ρg ¼ ðρamb ρÞg: ð1:40Þ
(1.28), the basic law for hydrostatics is immedi- dy
ately recovered. Indeed, setting all velocities in a In the process of getting to expression (1.40),
certain environment equal to zero, the only terms the implicit assumption is made that pressure
remaining are: differences in the horizontal directions are
! small.
∇ p ¼ ρamb g : ð1:37Þ
Equation 1.40 reveals that the driving force in
Equation 1.37 is valid at any time (in the absence situations where buoyancy dominates, stems
! ! from density differences, in the presence of a
of motion). For the special case where g ¼ g1 y ,
gravity field. This is known as Archimedes’
with g ¼ 9.81 m/s2, Equation 1.37 reads (in the law. Note that, since gravity acts in the vertical
y-direction): direction only, buoyancy forces by definition
dp also act in the vertical direction only.
¼ ρamb g ð1:38Þ For small density differences, the approxima-
dy
tion ρ ρamb is typically made in the
Note that Equation 1.37 in such circumstances Navier–Stokes equations, except that the differ-
also implies that pressure does not vary in the ence ðρamb ρÞ is accounted for in combination
horizontal directions. with gravity (Equation 1.40). This is called
Equation 1.38 can be integrated: Boussinesq’s approximation. In the context of
small density differences, expression (1.40) can
p ¼ pre f ρamb g y yre f : ð1:39Þ be developed further, using a Taylor series
10 B. Merci
temperature correction. Using the thermal volu-

∂ρ
expansion: ρ ¼ ρðT; pÞ ) ρ ρamb þ ∂T metric expansion coefficient:
p

∂ρ
ðT T amb Þþ ∂ p ð p pamb Þ. Typically the 1 ∂ρ
T β¼ ; ð1:41Þ
pressure correction is much smaller than the ρ ∂T p
the Archimedes force becomes:
ðρamb ρÞg ¼ ρamb βðT T amb Þg , if βðT T 1 Þ 1: ð1:42Þ
The basic expression is thus Equation 1.40, based number, which is the ratio of inertial forces to
on density differences, while Equation 1.42 is viscous forces:
only valid for small enough temperature
ρuL uL
differences. Re ¼ ¼ : ð1:43Þ
μ ν
The viscous forces tend to damp the inherent

Scaling Laws—Dimensionless Flow instabilities in the non-linear convection terms
Numbers in the Navier–Stokes equations, while these
instabilities can evolve towards fully-developed
In this section, starting from the governing turbulence for large enough Reynolds number.
equations, some scaling laws and non- This is addressed in the next section.
dimensional flow numbers are introduced. The When buoyancy is dominant, the
proportionality ρuL Δρg leads to the Froude
2
characteristic length scale is L, the characteristic
velocity is u. number, which is the ratio of inertial forces to
the Archimedes force:
ρu2
Dimensionless Flow Numbers Fr ¼ : ð1:44Þ
ΔρgL
Examination of the terms in the Navier–Stokes

In the fire community, this is often simplified to:
equations, Equations 1.28 and 1.26, leads to the
following proportionalities: ρut ρuL ΔLp Δρg
2
u2
Fr ¼ : ð1:45Þ
μLu2 . Several non-dimensional flow numbers gL
can be derived now, as follows. The importance Expression (1.44) resembles the underlying
of each of the numbers mentioned, depends on physics more than Equation 1.45. On the other
the importance of the corresponding terms in the hand, the difference between expressions (1.44)
Navier–Stokes equations. The convection term/ and (1.45) is no more than a numerical factor,
inertia term is always important, as it depending on the densities at hand. Moreover, in
characterizes the flow. Depending on the flow many experiments it is much more straightfor-
configuration, one or more terms are in competi- ward to measure velocities than mass densities,
tion with (or determine) the inertia term (or thus so that it is easier to characterize the experimen-
the flow). This is explained next. tal set-up through formulation (1.45). This
When the viscous stresses prevail, the explains why the use of Equation 1.45 is popular
ρu2
proportionality L μLu2 leads to the Reynolds in diagrams and correlations.
www.ebook777.com
If large (imposed) pressure differences occur, certain configuration, is through the use of differ-
sometimes the Euler number comes into play, ent fluids. Indeed, assume that the fluid does not
through ρuL ΔLp:
2
change (and that the densities do not change).
Then preservation of Re reveals that:
Δp Re1 ¼ Re2 ) u1νL1 ¼ u2νL2 ) u2 ¼ uL1 L2 1 . Preserva-
Eu ¼ : ð1:46Þ
u2 tion of the Froude number (still with the assump-
In fire related flows, this is often not relevant. tion that densities do not change) leads to:
qffiffiffiffi
u2 u2
In buoyancy driven flows, applying Fr 1 ¼ Fr 2 ) gL11 ¼ gL22 ) u2 ¼ u1 LL21 . Clearly,
Boussinesq’s hypothesis, the driving force
this is inconsistent with the requirement, stem-
(Equation 1.42) can also be made dimensionless
ming from the preservation of the Reynolds num-
as:
ber. Both numbers can be preserved if, starting
L3 gβΔT from the requirement for preservation of the
Ra ¼ : ð1:47Þ Froude number, the fluid’s viscosity is modified
αν
such that also the Reynolds number is preserved.
This is the Rayleigh number. Alternatively, the This is not straightforward.
Grashof number can be used: Fortunately, both the Reynolds number and
the Froude number have the property that, as
L3 gβΔT
Gr ¼ : ð1:48Þ soon as they are large enough, their actual value
ν2
becomes irrelevant. In other words, as soon as
The relation between the two is: Ra ¼ Gr:Pr, they are sufficiently high, the qualification ‘high’
with the Prandtl number as defined in Equa- is sufficient, not the exact number. This is due to
tion 1.8. The Grashof number can be interpreted turbulence, overwhelming molecular phenomena
as a ratio of buoyancy forces (with Boussinesq’s (see next section). This can also be understood
approximation) to the viscous forces. This is intuitively. The Reynolds number is the ratio of
relevant in boundary layers (see below). inertia to viscous damping forces. Either the
damping force is strong enough to overcome
the inherent instabilities in the non-linear con-
Scaling vection terms in the Navier–Stokes equations
(laminar flow), almost strong enough (transi-
In this section, scaling is briefly discussed in the tional flow) or not strong enough (turbulent
context of fluid mechanics. As such, only the flow). When turbulence is fully developed, the
momentum equation is considered, albeit that at strength of the viscous stress becomes irrelevant,
the end of this section, some remarks are i.e. the true value of the Reynolds number
formulated on the fire heat release rate (using becomes irrelevant. For the Froude number, it is
the energy equation) and the study of unsteady most instructive to examine expression (1.44).
phenomena (using the mass conservation equa- The driving force for buoyancy is in the denomi-
tion). As a consequence, no comments are nator. If density differences become small, buoy-
formulated on e.g. convective heat transfer or ancy becomes irrelevant and the Froude number
conduction through solids, nor on radiation. For is high. As such, high values of the Froude num-
an extensive discussion on scaling, the reader is ber implies that buoyancy is not important and
referred to [10, 11]. thus that the error is small when the Froude
The main non-dimensional numbers in number is not preserved (as long as it stays
low-Mach number flows are the Reynolds num- sufficiently high).
ber Equation 1.43 and the Froude number Equa- Knowing this, it is instructive to examine the
tion 1.44 (or Equation 1.45). Firstly, it is order or magnitude of Reynolds number and
mentioned that the only way to preserve both Froude number in fire related flows. Indeed, if
numbers when scaling (up or down) a flow in a one of the numbers can be expected to be high,
12 B. Merci
that number need not be preserved in scaling. temporal evolution of quantities (e.g. tempera-
Typical dimensions are in the order of 1 m: ture) depends on the dimensions of the configu-
pffiffiffi
L ¼ OðmÞ. Typical velocities are in the order of ration as t L. This is relevant when unsteady
1 m/s: u ¼ Oðm=sÞ. Densities are in the order of phenomena are studied.
1 kg/m3: ρ ¼ Oðkg=m3 Þ. The dynamic viscosity in
gases is in the order of 106 Pa.s: μ ¼

O 106 Pa:s . Using these numbers, the Reynolds
Turbulence
number Equation 1.43 is: Re ¼ O 1:1:1 106
¼
6 There are numerous text books on turbulence and
O 10 , while the Froude number Equation 1.44
1:1 turbulent flows, e.g. [12, 13]. Only some intro-
is: Fr ¼ O 1:10:1 ¼ Oð0:1Þ. Obviously, these are
ductory comments are presented here.
rough order of magnitude analyses, but it is clear
that in fire related flows, the choice will be made to
preserve the Froude number, not the Reynolds
Reynolds Number
number, when scaling is applied.
The energy equation also provides informa-
In the previous section it has been mentioned that
tion regarding scaling laws. The simplified for-
the Reynolds number Equation 1.43 is the ratio
mulation (1.36) can be used for fire-related flows.
on inertia to viscous forces. It is well-known that
Yet, temperatures are very important in fire
the convection term in the Navier–Stokes
related flows, so the energy equation should be
equations (1.28) is inherently unstable and that
interpreted in terms of sensible enthalpy, in
the flow becomes turbulent when the viscous
which case the fire heat release rate ( Q, _ in W)
forces are not strong enough to damp the
comes into play. Knowing that, in terms of instabilities, i.e. when the Reynolds number
dimensions, (sensible) enthalpy differences can becomes sufficiently high. Below a certain
be re-written as the product of specific heat and threshold number, the flow remains ‘laminar’.
temperature differences, Equation 1.36 leads to There is no sudden change from ‘laminar’ to
ρc ΔT ρc ΔTu
the following proportionalities: pt pL ‘turbulent’: there is a ‘transition’ zone in
Q_ between.
L3
kΔT
L2
.
This reveals that: Care must be taken in the definition of this
‘critical’ Reynolds number, in the sense that the
Q_ uρc p ΔTL2 : ð1:49Þ length scale must be defined. In flows over
flat plates, it is common practice to use the dis-
It is common practice to scale configurations such tance from the leading edge and Rec is in
that the temperatures remain the same. This also the order of 500.000. In pipe flows, it is
implies that densities do not change (if the same common practice to use the pipe diameter as
fluid is applied). As has just been explained, the characteristic length scale and Rec is in the
Froude number Equation 1.44 is preserved, so order of 2.000.
pffiffiffi
that the velocity scales as L . As a conse- It is important to stress that the Reynolds
quence, the fire heat release rate scales as: number is a flow property, not a fluid property.
pffiffiffiffiffi 2 Turbulence is typically defined on the basis of
Q_ 1 L1 L
¼ pffiffiffiffiffi 12 ) Q_ L5=2 : ð1:50Þ a number of properties [13]:
_
Q2 L2 L2 • Randomness: there are fluctuations in the
flow;
Finally, it is noteworthy that the conservation of
• Three-dimensionality: even if the mean flow
mass, Equation 1.19, reveals that:
is 2D or axisymmetric, the vortices or ‘eddies’
t L=u: ð1:51Þ are always three-dimensional;
• There is a wide range of length scales and
Applying Froude scaling, the velocity scales as time scales in the flow. The largest scales are
pffiffiffi
L, so that expression (1.51) reveals that the determined by the configuration at hand,
www.ebook777.com
while the smallest scales are determined by ‘same’ turbulent flow are made, repetitive
the Reynolds number. The smallest scales can measurements of the quantity are made at the
easily be 10,000 times smaller than the largest same location, and the average value of the
scales. measurements is determined. In a simplified
• Turbulent mixing is very effective. manner, though, one can think of this procedure
• There is a lot of diffusion and dissipation. as a time averaging, where the averaging period
Turbulence dies out quickly if not sustained Δt is sufficiently long, compared to the largest
by velocity gradients in the mean flow. turbulent time scales, but sufficiently short com-
• There is vortex stretching, transferring energy pared to time scales associated with possible
from the mean flow to turbulent fluctuations. variations in the mean flow:
It is instructive to briefly explain the
ðt ðt
randomness in the flow. Indeed, knowing that 1 1
vx ðtÞ ¼ vx ðtÞdt; T ðtÞ ¼ T ðtÞdt:
the Navier–Stokes equations (Equation 1.28) Δt Δt
are deterministic, one may pose the question tΔt tΔt
how it is possible that randomness occurs when ð1:52Þ

applying deterministic boundary and initial
It is clear that this is only possible if the turbulent
conditions. The reason is that there are always
time scales are short, compared to time scales in
small fluctuations, i.e. the boundary and initial
the mean flow. The ‘integral’ turbulent time scale
conditions are never known with infinite preci-
is typically less than 1 s, so in many fire related
sion. Due to the unstable convection terms in the
flows this concept of Reynolds averaging is
Navier–Stokes equations, turbulent flows are
possible.
extremely sensitive to details and this creates
Using Equation 1.52, the instantaneous value
randomness in the instantaneous flow fields.
can be expressed as the sum of the (Reynolds)
This makes it impossible to make long-term
averaged value and the instantaneous fluctuation
predictions of instantaneous turbulent flow fields
around that value:
and explains why turbulent flows are tackled in
simulation through statistical approaches (see vx ðtÞ ¼ vx ðtÞ þ v,x ðtÞ; T ðtÞ ¼ T ðtÞ þ T , ðtÞ:
below). Obviously, the mean flow can still be
ð1:53Þ
deterministic (see below).
Note that:
Reynolds Averaging v,x ðtÞ ¼ 0; T , ðtÞ ¼ 0; vx ðtÞ
As mentioned in the previous section, the ¼ vx ðtÞ; T ðtÞ ¼ T ðtÞ: ð1:54Þ

fluctuations in a turbulent flow make a direct
analysis through the Navier–Stokes equations Applying this averaging technique to the conser-
(Equation 1.28) impossible. Therefore, a statis- vation equations (1.19), (1.28) and (1.36), the
tical approach is adopted. The primary interest equations are obtained for the Reynolds-
is often the mean flow. To that purpose, the averaged quantities. They are very similar to
Navier–Stokes equations are averaged. The the instantaneous equations, but some additional
concept of Reynolds averaging is explained terms appear:
first. • Reynolds stresses in the momentum
Consider a turbulent flow. Measuring a veloc- equations;
ity component (or e.g. a temperature) at a certain • Turbulent heat fluxes in the energy equation.
location will then yield a fluctuating signal, as This is explained next. For the sake of ease,
explained. One can now determine the ‘average’ the energy equation is simplified here: it is
of that signal. The true definition of a Reynolds expressed in terms of temperature and no chemi-
average [12, 13] is that many realizations of the cal reactions, nor radiation, are considered. The
14 B. Merci
averaging of the chemical and radiative source Equations 1.28 and 1.36. The mean value of the
terms is a separate problem, not addressed here. product is not equal to the product of the mean
The additional terms appear as a consequence values:
of the presence of products in the instantaneous
, , , , , ,
vx v y ¼ ðvx þ v,x Þ v y þ v,y ¼ vx v y þ vx v y þ vx v y þ vx v y ¼ vx v y þ vx v y
ð1:55Þ
vx T ¼ ðvx þ v,x Þ T þ T , ¼ vx T þ vx T , þ v,x T þ v,x T , ¼ vx T þ v,x T , :
Simplifying further to a steady boundary layer the difference directions are statistically
flow of an incompressible fluid over a flat plate correlated. This is the case, explained from
without external pressure gradient, the main Fig. 1.4, showing a situation in a flow with a
remaining dominant terms are: mean velocity gradient. The discussion is given
here for the top left eddy, but it prevails for all
∂vx ∂v y
þ ¼0 eddies. At the left side of the top left eddy, the
∂x ∂y instantaneous motion is downward, as indicated
by the arrow. Knowing that the mean velocity in
∂vx ∂vx ∂ ∂vx ∂
vx þ vy ¼ ν ðv,x v,y Þ the vertical direction equals zero, this implies
∂x ∂y ∂y ∂y ∂y
that v’ < 0. In its downward motion, the eddy
ð1:56Þ brings along fluid with a higher (mean) velocity
in the horizontal direction into a region with
∂T ∂T ∂ ∂T ∂ , ,
vx þ vy ¼ α vyT lower (mean) velocity. Thus, the impact is a
∂x ∂y ∂y ∂y ∂y
local increase in horizontal velocity, in other
words u’ > 0. Clearly, from a statistical point
Clearly, the final terms, stemming from turbu- of view the velocity fluctuations in both
lence, are similar in nature to the molecular vis- directions are correlated, in such a manner that
cous stresses and the molecular thermal diffusion u0 v0 < 0. At the right side of the top left eddy, the
terms. The main question is now what the turbu- instantaneous motion is upward (v’ > 0) and
lent correlations look like. Indeed, terms like v,x v,y (in the mean) lower horizontal velocity is
are only non-zero if the velocity fluctuations in brought into a region with (in the mean) higher
Fig. 1.4 Sketch of U

turbulent fluctuations
(eddies) in a flow with a
mean velocity gradient v’ > 0
v’ < 0 v’ < 0
v’ > 0 v’ > 0
v’ < 0
www.ebook777.com
horizontal velocity, causing u’ < 0. Thus, again the Kolmogorov scales. Indeed, at those scales,
u0 v0 < 0. A similar reasoning can be built up for viscous damping ‘kills’ turbulence, i.e. dissipates
the temperature fluxes. the turbulent kinetic energy into heat.
Additionally, it is clear that the fluctuations, The notion of energy cascade, introduced by
caused by the turbulent eddy motion, will be Richardson, is worth mentioning. The basic
larger as the mean velocity (or temperature) mechanism is as follows:
gradients are larger. • Energy is taken from the mean flow and trans-
The above led to the following ‘eddy viscos- ferred to kinetic energy of turbulent eddies;
ity’ modeling concept, introduced by this occurs around the integral scales;
Boussinesq: • The turbulent eddies break up, transferring
their energy to the eddies of smaller scale;
∂vx only little energy is dissipated in this break-
v,x v,y ¼ νt ð1:57Þ
∂y up process;
• The break-up process of eddies continues
In other words, a ‘turbulent’ or ‘eddy’ viscosity
(‘cascade process’) until the eddies become
is simply added to the molecular viscosity in
so small that they cannot survive the damping
Equation 1.56. This reflects the physical obser-
action of viscosity anymore;
vation that momentum transfer increases in tur-
• The dissipation takes place at the smallest
bulent flows through the turbulent motion of
turbulence scales.
eddies. These cause ‘large scale’ momentum
It is important to appreciate that, whereas the
transfer.
dissipation takes place at the smallest scales, the
Similarly, this concept can be applied to the
dissipation rate is determined by the production
heat fluxes:
rate of turbulence from the mean flow in the
∂T energy containing range (in equilibrium
v,y T , ¼ αt ð1:58Þ conditions).
∂y
This phenomenology is reflected in the choice
In other words, the addition of the turbulent for turbulence modeling in CFD (Computational
thermal diffusivity to the molecular thermal dif- Fluid Dynamics). One extreme approach is not to
fusivity reflects the physical observation that heat model turbulence, i.e. to completely resolve all
transfer increases in turbulent flows through the turbulent motions, down to the smallest scales.
turbulent motion of eddies. These cause ‘large Knowing that these small scales can easily be in
scale’ heat transfer. the order of 0.1 mm or less, and realizing that the
computational mesh needs to be sufficiently fine
to resolve the smallest eddies, it is immediately
Turbulence Modeling clear that this approach is not feasible in typical
fire related flow simulations, where dimensions
As mentioned above, there is always a wide are in the order of 1 m. Worse than that, in
range of length scales and time scales in turbu- addition to unacceptable computing time and
lent flows. The higher the Reynolds number, the memory requirements, most of the time and
wider this range, because the smallest scales memory would be devoted to simulating the
become smaller and smaller. smallest scales [13], whereas the primary interest
The largest turbulence scales are called the is typically in the large scale flow phenomena
‘integral’ scales. The smallest scales are called (or in the mean flow).
the ‘Kolmogorov’ scales. A detailed discussion The other extreme is RANS (Reynolds-
of the spectrum is considered beyond the scope Averaged Navier–Stokes) turbulence modeling.
of this section, but it is important to appreciate In this approach, Reynolds averaging (see previ-
that most of the turbulent kinetic energy is in the ous section) is applied and all turbulent motions,
integral scale range (‘energy containing range’), i.e. the entire turbulent spectrum, are modeled.
while turbulence is dissipated at scales around Only the mean flow is resolved. The k-ε model
16 B. Merci
belongs to this class of models. The advantages equations itself is modified, the results inevitably
of the RANS approach are clear: the computa- change. This is not the case in RANS
tional mesh only needs to be fine enough to simulations, where the results become indepen-
resolve the mean flows; the time step dent of the mesh applied, provided it is fine
(in transient calculations) can be chosen on the enough.
basis of mean flow phenomena; one immediately For more discussion on turbulence modeling,
gets a solution for the mean flow. There are in the context of reacting flows, the reader is also
major disadvantages, though. Firstly, all turbu- referred to [14].
lence is modeled. Knowing that the largest tur-
bulent scales are configuration dependent, it
cannot be expected that a single RANS model
Boundary Layers—External Flows
can deal with arbitrary configurations in a reli-
able manner. Second, in fire related flows large
In section “Scaling”, it was mentioned that the
scale flow unsteadiness often plays an important
absolute value of the Reynolds number Equa-
role, e.g. in the entrainment process of air into
tion 1.43 becomes irrelevant as Re becomes
flames or smoke. Such unsteadiness is not cap-
high. In other words, the flow can be considered
tured in (unsteady) RANS and must be modeled.
‘inviscid’, i.e. μ ¼ 0. Stated in another manner:
Again, being configuration dependent, RANS
viscosity becomes irrelevant in the
models cannot be expected to be as accurate as
Navier–Stokes equations. This, however, is only
approaches where this unsteadiness is resolved.
true in the absence of solid boundaries. Indeed,
This explains the popularity of the LES
since viscosity is never really equal to zero, there
(Large-Eddy Simulations) technique in CFD for
is a ‘no-slip’ boundary condition at any solid
fire related flows. In this technique, the large
boundary: due to the viscous forces, the fluid
scale eddies are resolved and only the effect of
locally takes the velocity of the solid boundary,
the small scale eddies is modeled. This technique
at the solid boundary. In fire-related flows, the
offers the advantage of resolving the large-scale
solid boundaries typically stand still, so that the
flow unsteadiness (and buoyancy effects). Also,
no-slip boundary condition implies that the fluid
the unacceptable fineness of the computational
velocity equals zero.
mesh as required in DNS is avoided. Yet, there is
In fire related flows, boundary layers appear
a very important caveat. Indeed, in order to guar-
as ‘external’ flows or in ‘internal flows’. ‘Inter-
antee the quality of LES results, 80 % of the
nal’ flows are discussed in the next section.
turbulent kinetic energy must be resolved
Examples of fire related boundary layers in
[13]. It is common practice to use the computa-
external flows are: the flow over surfaces (hori-
tional mesh as filter in the LES approach, i.e. the
zontal or vertical) with e.g. flame spread, flow of
size of the computational mesh cells determines
smoke underneath a ceiling, atmospheric bound-
the size of the eddies still resolved. In many CFD
ary layers in forest fires, etc. An ‘external flow’ is
simulations performed on today’s computers, the
a flow where a ‘free stream velocity’ can be
mesh size is in the order of 10 cm or more. Very
defined, i.e. a velocity that is not affected by the
often, it cannot be guaranteed that 80 % of the
presence of the solid boundary.
turbulent kinetic energy is effectively resolved,
Both the boundary layer flow itself and the
so that care must be taken in the interpretation of
corresponding (convective) heat transfer can be
the CFD results. In other words, blind belief in
of importance in the context of fire.
the exactness of under-resolved LES must be
Consider first the situation of a flow with free
avoided. Also, it must be understood that if the
stream velocity U1 over a smooth flat plate,
computational mesh is used as filter for the
without external pressure gradient, schematically
instantaneous Navier–Stokes equations, as is
shown in Fig. 1.5. At the flat plate, U ¼ 0 (no slip
common practice in the fire safety science com-
boundary condition), whereas ‘sufficiently far
munity, no grid independent results can be
away’ from the plate U ¼ U1. The notion
expected from LES. Indeed, as the filter of the
www.ebook777.com
Fig. 1.5 Schematic representation of flow over a flat plate (Source: http://www.cortana.com)
‘sufficiently far away’ is related to the thickness Using x as characteristic distance, the follow-
of the boundary layer, which can be defined as: ing Reynolds number can be defined:
y ¼ δ : vx ¼ 0:99U 1 : ð1:59Þ U1 x
Rex ¼ : ð1:61Þ
ν
In words: the boundary layer thickness δ is the
distance from the plate where the velocity equals The viscous shear stress at the plate then
99 % of the free stream velocity. Other measures, becomes:
such as displacement thickness and momentum
∂vx U1
thickness, can also be used to characterize the τs ¼ μ ρν
∂y y¼0 δ
boundary layer thickness, but this is not essential
for the present discussion. 1=2
U1 x
Two flow regions can be defined: ρU 21 : ð1:62Þ
ν
• y < δ: Strong velocity gradients and viscous
shear stresses; This can be expressed in a non-dimensional man-
• y > δ: Negligible velocity gradients and vis- ner, by introducing the friction coefficient:
cous shear stresses.
From an order of magnitude analysis, in the τs, x U 1 x 1=2
C f,x ¼1 2 ¼ Re1=2 :
assumption that δ x, with x the distance from 2ρU 1
ν x
the leading edge of the flat plate, and in the ð1:63Þ

assumption of laminar flow in the boundary
layer, it can be shown that the boundary layer The ‘Blasius’ solution for laminar boundary
thickness grows as: layers over smooth flat plates indeed yields:
1=2 δlam ¼ 4:92xRe1=2 ; C f , x:lam ¼ 0:664Re1=2
μx x x
δlam : ð1:60Þ
ρU 1 ð1:64Þ
In words: the laminar boundary layer thickness However, as mentioned, there are inherent
grows with the square root of the distance from instabilities in the convection terms in the
the leading edge. It is thicker as the kinematic Navier–Stokes equations. These instabilities are
viscosity is higher. The latter shows that the damped near the flat plate, primarily due to the
influence region of the flat plate is larger for blocking effect and the viscous forces, so that
fluids with higher viscosity. turbulent vortices (eddies) cannot develop.
18 B. Merci
rffiffiffiffi
However, as the laminar boundary layer thick- τs
u* ¼ ; ð1:67Þ
ness grows with the distance from the leading ρ
edge (Equation 1.60), turbulence can start to
develop. There is a critical Reynolds number and the non-dimensional distance from the solid
Rex,crit (Equation 1.61) beyond which there is boundary, expressed in ‘viscous’ units:
transition from laminar to turbulent flow. For a yu*
smooth flat plate, Rex,crit is in the order of yþ ¼ : ð1:68Þ
ν
500.000.
As mentioned in the previous section, the Three regions can be distinguished inside the
momentum (and heat) transfer strongly increase boundary layer:
in turbulent motions, as compared to the aligned • Laminar (or ‘viscous’) sub-layer, y+ < 5:
laminar flow, since momentum (and heat) are very close to the solid boundary, all turbu-
transferred on a larger scale through the turbulent lence is damped (due to blocking effect and
eddies. As a result, the surface friction (and heat viscous forces) and the flow is essentially
transfer) increase and the boundary layer laminar. The velocity increases linearly with
becomes thicker. It can be shown that: the distance from the solid boundary.
• Logarithmic layer, 30 < y+ < 300: the
τs
δturb ¼ 0:37xRe1=5 ; C f , x, turb ¼ 1 2 motion is turbulent and there is a logarithmic
x
2ρU 1 relation between the mean velocity and the
¼ 0:0592Re1=5
x : ð1:65Þ distance from the solid boundary.
• Buffer layer: 5 < y+ < 30: transitional region
Thus, a turbulent boundary layer grows more between the laminar sub-layer and the loga-
rapidly than a laminar boundary layer. rithmic layer.
Before discussing the turbulent boundary It must be stressed that the discussion above
layer in more detail, it is worth mentioning that, refers to smooth surfaces. Roughness on a sur-
very similar to boundary layers at the level of face will affect the transition to turbulence and
velocities, thermal boundary layers can be the turbulent boundary layer structures. This can
defined. Indeed, the thermal diffusivity α plays be important, e.g. when the wind load on
the same role for heat transfer as the kinematic buildings is considered in built environment or
viscosity ν does for momentum transfer, as men- when wind effects are considered in the context
tioned before. The thermal boundary layer thick- of e.g. forest fires. This, however, is considered
ness is defined as: beyond the scope of the present chapter.
y ¼ δT : T T s ¼ 0:99ðT 1 T s Þ; ð1:66Þ
with Ts the surface temperature at the flat plate. Internal Flows—Flows in Pipes—
The Prandtl number (Equation 1.8) then Pressure Losses
determines whether the thermal boundary layer
is thicker or not than the flow boundary layer: A major difference from the previous section on
• Pr ¼ 1: δ ¼ δT; external flows, is that in internal flows the notion
• Pr < 1: δ < δT; example: air; ‘free stream velocity’ does not exist. In fully
• Pr > 1: δ > δT; example: water. developed flow conditions, the flow is entirely
Now the turbulent layer is discussed in more affected by the presence of the solid boundary
detail. Figure 1.6 presents a profile as measured and, consequently, by the fluid’s viscosity. The
in a pipe (which is in fact an internal flow, see discussion is based here on flows through pipes,
next section), but the boundary layer near the since pipes are a common configuration (e.g. water
solid boundary is very similar. through pipes for sprinklers or water hoses in fire
The results are expressed in a nondimensional service intervention). Some comments are
manner, introducing the friction velocity: formulated for flows through ducts in the end.
www.ebook777.com
35
Laminar Transition
30 sublayer buffer zone Turbulent core
25
20
u λu u yu
— log10 ——* —= 5.75 log10 ——*+ 5.5
u* v u* v
15
u yu*
— = —–
10 u* v
Nikuradse
5 Reichardt
1 2 3 4 5
yu
log10 ——*
v
Fig. 1.6 Turbulent velocity profiles as measured in a pipe: different regions in boundary layer
When the entrance region of a smooth pipe is When the cross-section is not round, the diam-
considered, a boundary layer develops from the eter D is replaced by the hydraulic diameter Dh,
solid boundary, very similar to what has been defines as four times the cross-sectional area
described in the previous section. However, divided by the cross-section perimeter:
since this boundary layer grows on the entire
surface, there is a point where the entire cross- 4A
section is covered by a ‘boundary layer’. This Dh ¼ : ð1:70Þ
P
point determines the ‘entrance length’. From that
point onward, the boundary layers do not evolve It is straightforward to show that for fully devel-
and the flow becomes fully developed. oped laminar flows, the following expressions
Depending on the Reynolds number, the flow is hold (with R ¼ D/2 the radius of the pipe and r
again laminar or turbulent. It is clear that the dis- the radial distance from the pipe symmetry axis):
tance from the entrance is not a useful characteris- • Parabolic velocity profile:
tic length, since in fully developed flow conditions,
the velocity profiles are independent of that dis- u r 2
¼2 1 ; ð1:71Þ
tance. Clearly, the pipe diameter is a useful quan- Um R
tity. At the same time, there is no free stream • Wall shear stress (friction):
velocity. A mean velocity Um can be computed
from the volume flow rate and the cross-sectional ∂u Um
τs, lam ¼ μ ¼ 4μ ; ð1:72Þ
area. Thus, the Reynolds number is now defined as: ∂r r¼R R
Um D • Friction factor:
Re ¼ : ð1:69Þ
ν
The critical Reynolds number beyond which the τs, lam 16
f lam ¼ 1 2 ¼ ; ð1:73Þ
flow becomes turbulent is around Recrit ¼ 2300. 2ρU m
ReD
20 B. Merci
64
Laminar flow f = —
R
0.1
Laminar Critical Transition
0.09
flow zone zone
0.08 Complete turbulence, rough pipes
0.07 0.05
0.04
0.06
0.03
0.05 0.02
(Lv 2/D2g)
Friction factor f = —————
0.015
0.04
∈
hL
Relative roughness —
D
0.01
0.008
Rcr
0.006
0.03
0.004
0.025
0.002
0.02 0.001
∈ (ft) ∈(mm) 0.0008
0.0006
Riveted steel 0.003–0.03 0.9–9.0
0.0004
0.015 Concrete 0.001–0.01 0.3–3.0
Wood stove 0.0006–0.03 0.18–0.9 0.0002
Cast iron 0.00085 0.25
Galvanized iron 0.0005 0.15 0.0001
Asphalted cast iron 0.0004 0.12
Smooth pipes 0.000,05
0.01 Commercial steel 0.000,001
or wrought iron 0.00015 0.046
0.009 Drawn tubing 0.000005 0.0015
0.000,005
0.008 0.000,01
7 9 103 2(103) 3 4 5 6 7 9 104 2(104) 3 4 5 6 7 9 105 2(105) 3 4 5 6 7 9106 2(106) 3 4 5 67 9 107 2(107) 3 4 5 67 9 108
vD
Reynolds number R = —–
v
Fig. 1.7 Moody diagram
• Pressure loss over a distance L in the pipe: pressure loss is still linear with the length of
the pipe, but proportional to the mean velocity
1 4L
Δ plam ¼ f lam ρU 2m : ð1:74Þ squared and inversely proportional with the
2 D pipe diameter.
Note that the pressure loss is linear with the In duct flows, essentially the same reasoning
length of the pipe and, with Equation 1.73, linear holds. The major difference is in the values for
with the mean velocity and inversely propor- the friction factor f, important to estimate pres-
tional with the square of the pipe’s diameter. sure losses. Secondary flows appear in the
Recall that this is only true for laminar flows. corners of duct, transporting momentum from
For turbulent flows, the expressions become the center to the corners and leading to a relative
more complex. The velocity profile can be increase in velocity near the corners.
approximated as: It is instructive to quantify pressure losses for
internal flows as:
u r 1=7 Um
¼ 1 , Umax ¼ : ð1:75Þ
U max R 0:817 1
Δ pL ¼ CL ρU 2m : ð1:76Þ
2
Expression (1.74) still holds for the pressure
losses, but the friction factor is no longer The loss coefficient CL must be defined,
obtained from Equation 1.73. The Moody dia- depending on the situation (geometry and flow
gram [14] (Fig. 1.7) reveals that, for large type—laminar/turbulent). All pressure losses
enough Reynolds number, the friction factor must be accounted for in the design. This holds
is determined by the relative roughness of the for e.g. the design of the piping system for
pipe, independent of the Reynolds number. As sprinklers (i.e. what pump must be chosen) or
a consequence, for turbulent flows, the the design of a smoke extraction system
www.ebook777.com
(i.e. what extraction fans are required to over-

V
come all pressure losses, including the ones in
1
the exhaust system).
Some examples are briefly mentioned here: Static pressure hole
• Straight sections: Moody diagram, see above.
• Curves/bends: CL is determined by the total
angle and the radius of the bend
(e.g. CL ¼ 0.14 for an angle of 90o with radius
2D, but it is about 1.2 for the same angle of
90o but with radius ¼ 0, i.e. a sharp bend).
Curves and bends are always important to
consider in calculations of pressure losses.
2
• Sudden pipe expansion: CL ¼ 1 AA12 . A (p1 – p)
special case concerns the flow into a large

space, i.e. A2 ! 1. Then CL ¼ 1.
• Sudden pipe constriction: the flow is Fig. 1.8 Sketch of Pitot tube
constricted and then widens again behind the
constriction. A good estimate for a sudden
principle of a Pitot tube. The flow is stagnated.
constriction is CL ¼ 0.5, while CL goes
By measuring the pressure increase cause by this
down to 0 for a very gentle constriction.
stagnation, the velocity can be computed. Indeed,
• Flows through openings: CL primarily
applying Bernoulli’s Equation 1.77 at constant
depends on the edges of the opening. The
height z yields:
most typical situation is that the edges are sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Sharp. In that case CL typically varies 1 1 2ð ptot pstat Þ
between the values CL ¼ 0.4 and CL ¼ 0.7. ptot þ ρ02 ¼ pstat þ ρv2 ) v ¼ :
2 2 ρ
ð1:78Þ
Application Example 2: Venturi Flowrate
Bernoulli Equation
Meter Figure 1.9 shows the basic principle of a
Venturi meter. It is essentially a converging
The Bernoulli equation is of fundamental impor-
cone, from which the flowrate through a pipe
tance. The equation, valid on any streamline, has
can be calculated. Indeed, applying Bernoulli’s
been developed for incompressible flows (liquids).
Equation 1.77 at constant height z yields:
Yet, it can also be applied to low-Mach number
flows where the density hardly changes (gas flows 1 1
in fire related flows, as long as the density along the p1 þ ρv21 ¼ p2 þ ρv22 : ð1:79Þ
2 2
streamline does not vary strongly).
With the notation now that z is the height Conservation of mass allows elimination of v1:
(i.e. the z-direction is vertically upward), v1 ¼ v2 AA21 . Insertion in Equation 1.80 and
Bernoulli’s equation reads: introducing V_ ¼ v2 A2 yields:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
p þ ρv2 þ ρgz ¼ const: ð1:77Þ 2A21 ð p1 p2 Þ
2 V_ ¼ A2 : ð1:80Þ
ρ A21 A22
A few application examples are briefly
mentioned. Application Example 3: Flow Out of a Large
Tank Figure 1.10 sketches the situation. A
Application Example 1: Velocity Measurement large tank is considered, so that the liquid surface
with a Pitot Tube Figure 1.8 shows the basic can be approximated as standing still, i.e. v1 ¼ 0.
22 B. Merci
sffiffiffiffiffiffiffiffiffi
2Δ p
V_ ¼ Cd A : ð1:82Þ
P1, V1 P2, V2
ρ
This can be interpreted as a ‘correction’ to the

cross-section area that is effectively used for
outflow (or inflow). For an orifice, the value of
Cd is around 0.6 for e.g. open doors or windows,
going up to about 0.7 for flows through small
gaps [10].
Finally, it is mentioned that, when pressure
losses are considered (see previous section),
Manometer Bernoulli’s equation can be extended to:
1 1
p1 þ ρv21 þ ρgz1 ¼ p2 þ ρv22 þ ρgz2
2 2
Fig. 1.9 Sketch of Venturi meter þ Δ pL, 12 : ð1:83Þ
In Equation 1.84, the final term reflects the pres-

p1V1 sure loss between points 1 and 2 on the
1 streamline.
z1 p2,V2 Wind
z2 2 Wind is an important factor in fire protection

engineering. An obvious example is the effect
Datum of wind on the development of forest fires,
where convection strongly affects the direction
Fig. 1.10 Outflow out of a large tank
and speed of fire spread. Another examples
concerns smoke and heat control (SHC) in case
Approximating p1 ¼ p2 ¼ patm, the flow is of fire inside a building, where wind will exert a
generated by the gravity force: pressure load onto the building. The distribution
1 2 pffiffiffiffiffiffiffi of the load (positive and negative) affects the
ρv2 ¼ ρgz ) v2 ¼ 2gz: ð1:81Þ performance of the SHC system. The wind can
2
also induce internal flows into the building,
The assumption p1 ¼ p2 ¼ patm is a reasonable depending on leakages or open windows or
assumption if the tank is open and if the liquid doors.
density is much higher than the density of air (the It is common practice to consider steady wind
latter is practically always fulfilled). conditions. Clearly, wind gusts can have an
Application Example 4: Flow Through impact in the course of the fire. This is not con-
an Orifice In the previous example, the pressure sidered in the discussion below.
driving the fluid out of the tank stems from gravity.
In general, the pressure difference (Δp) over an
opening determines the flow through the opening. Natural Wind Characteristics
From the (mean) velocity and the cross-sectional
area (A) of the opening, the volume flow rate A classical method to characterize the boundary
through that opening can be computed. A dis- layer, when there is no detailed information of
charge coefficient Cd is introduced, though: surface roughness, is the use of a power law:
www.ebook777.com
Fig. 1.11 Schematic

representation of wind
flow, interacting with a tall –
building
Plan Rear Eddy Wind

+ –
Side Wake
α
z under-pressure on all other sides (including
vðzÞ ¼ v zre f : ð1:84Þ the roof). This is illustrated in Fig. 1.11. This
zre f
pressure load distribution affects the performance
The exponent α is given the value α ¼ 1/7 for of SHC systems and can cause internal flows
‘smooth’ surfaces. This is related to turbulent (e.g. through open windows or doors), as men-
boundary layer velocity profiles. Other values tioned. However, it must be stressed that
apply for ‘rough’ surfaces, i.e. when ‘obstacles’ Fig. 1.11 is a strong simplification of reality. Not
such as trees or buildings disturb the boundary only is the wind assumed steady and perpendicular
layer. An alternative option is then the use of a to one side of the building, it is also not supposed to
log law expression: be affected by the environment. In reality, tall
buildings are situated in a built environment, so
v* z
vðzÞ ¼ ln : ð1:85Þ the oncoming wind profile need not obey
κ zo
expressions like (1.85) or (1.86) and need not be
In expression (1.86), v* is the friction velocity unidirectional. For obvious reasons, the direction
Equation 1.67, κ is the von Karman coefficient of the oncoming wind, even if not disturbed by the
(κ ¼ 0.4) and zo is the aerodynamic roughness environment, varies in time, depending on atmo-
length. The reader is referred to specialized liter- spheric pressure distributions. Finally, modern
ature for more details (e.g. [15]). buildings are not necessarily rectangular in shape.
All these factors indicate the need for either small-
scale wind tunnel experiments or extensive CFD
Interaction of Wind with Buildings studies. For the time being, since it is necessary to
consider many wind directions and velocities and
It is well-known that wind, impinging in a perpen- the (built) environment can be complex and hard
dicular direction onto a rectangular building, to characterize as boundary conditions in
causes over-pressure on the windward side and CFD simulations, wind tunnel experiments seem
24 B. Merci

preferable. The model scale can then be built on a γ Ratio of specific heats γ ¼
Cp
Cv
table that can be turned around in the wind tunnel
to examine various angles of oncoming wind. ν Specific volume (m3.kg1)
To finalize this section, it is recalled that ρ Density (kg.m3)
wind-induced over-pressure and under-pressure
are proportional to the wind velocity squared. Subscripts
This is to be expected from Bernoulli’s equa-
tion 1.77. Clearly, the flow does not stagnate atm Atmospheric
entirely over the entire surface and thus pressure gas Gas
coefficients are introduced: liq Liquid
1
Δ pw ¼ Cw ρamb v2w : ð1:86Þ
2 References
The wind coefficient can be positive (over- 1. G.K. Batchelor (1967) An introduction to fluid dynam-
pressure at the windward side) or negative ics, Cambridge University Press.
(under-pressure). In e.g. [16] more info is found 2. H.D. Baehr (1978) Thermodynamik, Springer Verlag.
3. A. Bejan (1993) Heat transfer, John Wiley and Sons.
on this topic. 4. W.M. Rohsenhow, J.P. Hartnett and E.N. Ganić
(1985), Handbook of Heat Transfer Fundamentals
(2nd ed.), McGraw–Hill Book Company.
Nomenclature 5. K.K. Kuo (1986) Principles of combustion, John
Wiley and Sons.
6. T. Poinsot and D. Veynante (2001) Theoretical and
Av Vent area (m2) numerical combustion, Edwards.
Cp Specific heat at constant pressure (kJ. 7. P.A. Libby and F.A. Williams (1980) Turbulent
kg1.K1) reacting flows, Springer Verlag.
Cv 8. N. Peters (2000) Turbulent combustion, Cambridge
Specific heat at constant volume (kJ.
University Press.
kg1.K1) 9. G. Cox (1995) Combustion fundamentals of fire, Aca-
D Fireball diameter (m) demic Press.
Eav Total expansion energy (kJ) 10. B. Karlsson and J.G. Quintiere (2000) Enclosure fire
L Latent heat of vaporization (kJ.kg1) dynamics, CRC Press.
11. J.G. Quintiere (2006) Fundamentals of fire phenom-
m Flammable mass (kg) ena, John Wiley and Sons.
Mliq Liquid mass (kg) 12. H. Tennekes and J.L. Lumley (1972) A first course in
P Pressure (Pa) turbulence, MIT Press
R 13. S.B. Pope (2000) Turbulent flows, Cambridge Univer-
Distance from explosive material (m)
sity Press.
s Specific entropy (kJ.kg1.K1) 14. L.F. Moody (1944), “Friction factors for pipe flow”,
T Temperature (K) Transactions of the ASME 66 (8): 671–684.
td Duration of the fireball (s) 15. B. Blocken and J. Carmeliet (2004) “Pedestrian wind
u Internal energy (kJ.kg1) environment around buildings: Literature review and
practical examples.” Journal of Thermal Envelope
V Vessel volume (m3) and Building Science 28(2): 107-159.
WTNT Equivalent mass of TNT (kg) 16. J.H. Klote and J.A. Milke (2002) Principles of smoke
Xf Mass fraction of the initial liquid mass management, American Society of Heating,
Refrigerating & Air-Conditioning Engineers, Inc.
that flashes to vapor
Xg Mass fraction of the initial vapor mass
that does not condense during expansion Bart Merci is a Professor at Ghent University (Belgium).
He is Head of the research unit “Combustion, Fire and
Fire Safety”. Having completed a PhD (Ghent University,
Greek Symbols 2000) on turbulence modeling in CFD simulations of non-
premixed combustion, he is an expert in fluid mechanics
β Fraction of energy released converted into aspects in reacting flows, more particularly related to fire
the blast wave. and smoke dynamics.
www.ebook777.com
Conduction of Heat in Solids

2
Ofodike A. Ezekoye
Introduction scenarios. The treatise of Carslaw and Jaeger [1]

covers most solutions for conduction phenomena.
Heat transfer is an area of thermal engineering Other useful texts that discuss conduction phe-
that focuses on the transport, exchange, and nomena are readily available [2, 3]. It is useful to
redistribution of thermal energy. The three build up this discussion by first identifying where
modes or ways that heat can be transferred have conduction heat transfer ties into overall energy
been termed conduction, convection, and radia- conservation and energy transfer.
tion. In this chapter, the basic physics associated
with conduction heat transfer will be presented,
and it will be shown through examples how the Energy Conservation
tools and analysis typically used for conduction
problems can be applied to design and analysis The fundamental laws that allow us to analyze
when fire occurs. and predict fire phenomena are often termed
Conduction heat transfer only occurs in a conservation laws. Conservation laws are essen-
medium. This is a distinction between conduction tially balance equations that allow us to model
and radiation, which does not require a medium. how variables that describe the physical world
The medium or state of matter in which conduc- dynamically evolve. In fire systems, we typically
tion takes place can be a gas, liquid, or solid. The model the physical world using mass conserva-
distinction between conduction and convection tion, momentum conservation, energy conserva-
heat transfer is associated with whether the tion, and chemical species conservation. For this
medium has some ordered flow or bulk motion. chapter, we are interested in describing how heat
Heat transfer, when there is a mass averaged is transferred in media that are not deforming
velocity, is termed convection. Heat transfer that (i.e., are in rigid body motion with no unbalanced
takes place in a stationary frame of reference is forces) or reacting (fixed chemical species and
called conduction. More details will be presented mass). We do assume, however, that the medium
on the mechanisms that allow heat transfer to can possibly have heat transferred to it either
occur in a stationary medium as we proceed through interactions with its surrounding or
through this discussion. Solutions will be through some other energy input into it. Also,
provided for selected configurations and we assume that the medium may have different
amounts of thermal energy stored within it at
different locations. To more precisely describe
O.A. Ezekoye (*)
energy transfer processes, we rely on the first law
Department of Mechanical Engineering, University
of Texas at Austin, ETC 7.130, MS C2200, of thermodynamics. The energy conservation
TX 78712, Austin principle is the basis for heat transfer.

26 O.A. Ezekoye
Thermodynamic Properties necessary to define the heat transfer rate in terms

of temperature differences. The empirical law
The first law of thermodynamics is a statement of defining the heat transfer rate to a body immersed
energy conservation [4, 5]. It states that the in a fluid is called Newton’s law of cooling.
change in energy for an identifiable set of matter When Newton’s law of cooling is used, the heat
can only result from heat transferred across the transfer rate to the body is Q_ net, in ¼ hAðT T 1 Þ
material’s boundary or work done either by or on If we apply Newton’s law of cooling to the first
the material. The thermodynamic property or law, we arrive at a result called the lumped
energy function that best describes the molecu- thermal approximation in conduction analysis.
lar, atomic, electronic, and nuclear energy of a
material is the internal energy. In terms of the
internal energy, U, the first law is Lumped Thermal Analysis
dU
¼ Q_ net, in þ W_ net, in Briefly, the lumped thermal approximation
dt
allows one to model the overall transient thermal
U is the internal energy, Q is heat added to the response of a body at some initial temperature
system and W is work done on the system. subjected to either a change to the external fluid
The total internal energy U is a system temperature or as a result of some local heating
integrated value that represents the total thermal within the object. The validity of this approxima-
energy of the material system of interest. We can tion will be discussed in more detail in later
describe the local internal energy in terms of a mass sections. For the purposes of this discussion, we
specific internal energy, u, that is simply the total will say that the approximation is valid when the
internal energy, U, for a region of matter divided by time scales for internal energy transfer and
the mass of that region. The internal energy, like subsequent homogenization of the temperature
any other thermodynamic variable can be defined field within an object are much smaller than the
in terms of other thermodynamic variables. There time scales for energy transfer from the surface
is an approximation used in thermodynamics that of the body to an external thermal reservoir. In
states that the internal energy for an incompressible short, the lumped thermal approximation is rea-
material can be specified in terms of the tempera- sonable when temperature differences within a
ture. The thermodynamic property specific heat body are relatively small when compared to tem-
capacity at constant volume, cv, relates perature differences between the surface of the
differential changes in the mass specific internal body and a characteristic temperature of the exte-
energy to differential changes in temperature. rior fluid. It can be shown that a nondimensional
heat transfer parameter called the Biot number
du (Bi) which represents the ratio of the internal
cv ¼ and mcv dT ¼ dU
dT conductive resistance to the external convective
The mass is defined as the product of density and resistance should be small for the lumped ther-
volume. The control-mass statement of the first mal approximation to be valid. A mathematical
law for a case with no net work done becomes: statement of the energy equation in the lumped
approximation is (Fig. 2.1):
dT
ρVc ¼ Q_ net, in dT hc A
dt ¼ ðT T 1 Þ
dt ρVc
This form of the first law neither provides infor-
mation about spatial variations in energy within This first order ordinary differential equation
the medium nor describes how energy is trans- can be integrated and one form of the solution is:
ferred. Experience tells us that the heat transfer
T Te
into some identifiable mass element likely ¼ eðhc A=ρVcÞt ¼ et=tc
depends on temperature differences. It will be T0 Te
www.ebook777.com
2 Conduction of Heat in Solids 27
heat transfer coefficient is proportional to the

fluid velocity h ¼ Cu1=2 . This results in:
tACT u1=2 ρLc

¼ ¼ RTI
T0 Te C
ln
T ACT T e
This combination of parameters is the well

Fig. 2.1 Schematic showing convective flow over an known response time index for sprinklers.
object that will be analyzed using a lumped thermal
approximation
Fourier’s Law of Conduction
In the above, T0 is the initial temperature of the
body and Te is the external fluid temperature As previously noted, the lumped approximation
surrounding the object. There is a characteristic does not allow one to predict the spatial variation
time in the problem defined as: of temperature within a body. In some sense, it
provides an average or lumped temperature
ρVc
tc ¼ response. To be able to predict the spatial varia-
hA tion of temperature, it is necessary to introduce
The characteristic time provides an estimate another physical law that models how heat is
of the time required for the nondimensional tem- transported when temperatures differences exist
perature to relax to its steady value. This rela- within a body. We expect heat to flow across a
tively simple solution is useful in characterizing body in proportion to the temperature difference
a large number of important problems in fire across the body, and perhaps inversely related to
systems [6]. the distance across the body. Fourier’s law states
that the heat flux is proportional to the tempera-
Example 1 The lumped thermal approximation ture gradient (the spatial derivative of the tem-
is frequently used to analyze the response of a perature). For a one dimensional homogeneous
sprinkler head as it activates due to a change in and isotropic object this reduces to the simple
the environment temperature because of a fire. A expression:
sprinkler head fuse can be modeled as a cylinder
dT
of diameter 4 mm and length 12 mm. The density q} ¼ k
dx
can be approximated as being 1000 kg/m3. The
specific heat capacity is approximately 1 kJ/kgK. We use the notation q00 to indicate a heat transfer
The heat transfer coefficient of the smoke gases rate per unit area. The proportionality between
is 20 W/m2K. If the smoke gases are 200 C and the heat flux and the spatial derivative of temper-
the fuse is initially at 20 C, how long will it take ature is the thermal conductivity.
for the fuse to open if the activation temperature
is 80 C?
The solution is arrived at from inverting: Thermal Conductivity
T Te
¼ eðhc A=ρVcÞt ¼ et=tc For materials like air, water, glass, and copper,
T0 Te the thermal conductivity is isotropic (i.e., does
ρLc T0 Te not depend on orientation), but it has a tempera-
tACT ¼ ln ture dependence. Under conditions in which the
hc T ACT T e
overall thermal conductivity difference across
For the values that we specified, we find that the body is small relative to the any particular
the fuse opens in 243 s. Chapter 3, shows that the value of the thermal conductivity in the body, we
28 O.A. Ezekoye
2 3
can consider k to be essentially a constant. In fire kxx 0 0
applications, this is often not the case, but for the k¼4 0 k yy 0 5
sake of analysis we will often use this approxi- 0 0 kzz
mation when generating analytical solutions.
There are materials for which the thermal con-
For metals, many crystalline solids, many amor-
ductivity depends both on the local temperature
phous solids, liquids, and gases, the conduction
and also on the orientation. In contrast to isotropic
process is considered to take place in an isotropic
materials for which there is no directional effect,
medium. For such materials, the thermal conduc-
anisotropic materials have this directional depen-
tivity can vary spatially and with temperature,
dence. The most commonly encountered aniso-
but does not have an orientation effect.
tropic material in fire applications is wood. The
2 3
grain structure of wood is the source of the anisot- 1 0 0
ropy. Practically, we would find that for the same k ¼ kðT ðx; yÞÞ4 0 1 0 5
temperature difference across a given thickness 0 0 1
of wood, the heat transfer rate depends on
whether this temperature difference is aligned
with the grains or aligned perpendicular to the Homogeneous Systems
grains. Of course, as one heats wood, there are Most obvious in gases, it is known that random
also chemical changes to the wood. So, the ther- molecular motion transfers heat from hot
mal conductivity depends on the temperature, molecules to cooler ones. For solids other wave
composition, and orientation. For a simple analy- like effects are important. There is a relatively
sis, the effects of decomposition are often simple theory that describes the physics of ther-
neglected for the initial ignition process. mal conductivity. Conduction heat transfer can
Again, Fourier’s law states that the heat flux be thought of in terms of a carrier particle with a
!00
vector, q , is proportional to the temperature gra- characteristic velocity and characteristic length
dient, where the proportionality constant is the scale over which it acts. The development of this
thermal conductivity, k. In general, k is a second perspective of thermal conductivity, based on the
order tensor and has different values depending on properties of notional particles is described by
the face and orientation of a differential volume Kaviany [7, 8]. In some sense, this description is
[1–3]. For a general anisotropic material a simple generalization of the kinetic theory
2 3 description of thermal conductivity for gases.
00
kxx kxy kxz For gases, we understand that the kinetic theory
!
q ¼ 4 k yx k yy k yz 5∇T of gases describes how k varies in terms of a
kzx kzy kzz characteristic gas velocity, u, the number density
of molecules, n, the mean free path, l, and the
This suggests that the component of the heat
molecular internal energy described by the
flux vector in the x-direction depends on all
molecular mass and heat capacity (mc).
components of the temperature gradient.

∂T ∂T ∂T 1
qx ¼ kxx þ kxy þ kxz k ffi mcv nul
∂x ∂x ∂x 3
For some materials that are frequently dealt with In the following table adapted from Kaviany [7, 8],
in fire analyses, such as wood, there is some the characteristic parameters for various types of
simplification in the dependence of thermal con- conduction systems are provided (Table 2.1).
ductivity on orientation. Laminates like wood are
said to be orthotropic. For an orthotropic mate- Examples of Homogenous Materials
rial, the off-diagonal elements of the thermal In fire analysis, most solid materials are
conductivity tensor are zero and the diagonal approximated as being homogeneous. Examples
elements are not equal to each other. of homogeneous systems in fire applications are
www.ebook777.com
Table 2.1 Characteristic quantities used in microscale carrier model of conduction

Microscale carrier Fluid particle (random Phonon (quantal lattice vibration) Electron
motion)
Regimes Dilute gases Acoustic phonon Free electrons and valence
and optical phonon electrons
Mean free path Interparticle spacing Lattice dimension Lattice dimension
Carrier concentration Fluid density Solid density Free electron density
Carrier speed Thermal speed Speed of sound Electron drift velocity
Fig. 2.2 Range of thermal

conductivities for different
materials (Adapted
from [2])
simple polymeric materials, metals, and various resistance models. For a mixed medium that
types of insulating materials (Fig. 2.2). is comprised of several different conducting
elements, the parallel approximation provides
Composite Systems an upper bound on an effective thermal conduc-
Treatment of composite material thermal conductivity, while the series approximation provides a
tivity is somewhat more complicated than the lower bound.
treatment for homogeneous materials. With
increased use of composite materials like polymer Examples of Composite Materials
impregnated concrete as structural components, it Examples of composite materials include many
is useful to discuss how to construct an effective types of insulating materials in which at least two
thermal conductivity for such materials. The key types of materials are mixed in various mass
to constructing an effective thermal conductivity fractions. The mass or volume fractions of the
is to develop a meaningful way to average the constituents can then be used along with their
thermal properties for the system. A representa- individual conductivities to define an effective
tive averaging volume is the term used to describe conductivity for the system. Various mixing
the volume over which one can meaningfully rules have been developed for the effective ther-
average the properties of the composite in order mal conductivity. Gebhart [9] discusses a general
to properly thermally characterize the material. way of classifying the effective thermal conduc-
The simplest treatments of composite media ther- tivity of a binary system comprised of a matrix
mal conductivity use either series or parallel material a and added material b as:
30 O.A. Ezekoye

ke kb Li Fig. 2.3 Schematic of
¼ f ; Φb ; ; Bi : differential volume in
ka ka L which steady one Q Q
dimensional heat equation x x+Δx
Depending on the ratio of the thermal
is developed
conductivities, the ratios of the characteristic
lengths of the a and b segments within the medium,
x x + Δx
and the relative volumetric ratios, different
correlations exist for the effective conductivity. To develop the one dimensional conduction
Kaviany [7] presents a correlation for the model, we consider an elemental volume, ΔV,
effective conductivity for random porous solids located between spatial locations x and x + Δx
(e.g., continuous solid and fluid phases) as might for a heat transfer process that is in steady state.
occur for a wound insulation material, We can apply the first law of thermodynamics to
0:2800:757 logðεÞ0:057 logðks =k f Þ the elemental volume and consider a case in
hk i ks which there is no internal generation (Fig. 2.3).
¼
kf kf
Q = Q = Constant
which is valid for fluid porosity (volume frac- x x+Δx
tion) in the range of 0:2 < ε < 0:6. Q = q″A = −kA
dT
dx
d dT
Heat Equation Formulations kA =0
dx dx
The heat equation is the name given to the differ- Application of Fourier’s law leads to an
ential equation that models heat conduction in energy equation specified in terms of temperature
materials. The heat equation is most generally gradients defined within the solid. The solution
developed in a three dimensional, unsteady can be found by simple integration.
form. Depending on the scenario of interest, it If the thermal conductivity, k, is nearly con-
is not always necessary to solve the full formula- stant over the temperature range of interest to the
tion of the heat equation. By formulating an problem, then we see that a very simple relation-
appropriate reduced form of the heat equation, ship holds between the temperature difference
one can generally compute an accurate represen- across the solid, the thermal conductivity, and
tation of the temperature profile and heat flux the thickness of the solid. It is apparent that an
distribution in the material. In the following analogy holds between this form and Ohm’s
sections, several reduced model forms for the law, where the heat transfer rate is identified
heat equation will be discussed. as a current, the temperature difference, ΔT
is identified as a potential change, and L/kA is
identified as a generalized resistance.
Steady One Dimensional Models
Under conditions in which there is a primary heat

Cylindrical Shells
This same type of analysis can be formulated for
transfer direction, it is appropriate to formulate a
cylindrical shells. The difference in the analysis
one-dimensional form of the heat equation. Fur-
is that the cylindrical shell has variable surface
ther, when the time scale for changes in boundary
area (Fig. 2.4).
conditions and sources are large relative to the
Applying Fourier’s law over concentric cylin-
time scale over which the thermal system
drical elements yields
equilibrates, the analysis can be treated as being
steady. A discussion of how to define the time to
dT
equilibrate in conduction systems will follow in a Q_ ¼ Aq ¼ 2πrL k
dr
later section.
www.ebook777.com
Similar to the development for the planar slab surface. At least one end of the rod is assumed
geometry, an effective resistance can be defined to be fixed at a temperature different from the
for the cylindrical system. Integrating the equa- environmental fluid temperature. For one dimen-
tion twice yields: sional heat transfer to be valid, the length of the
fin divided by the diameter should be large and a
2πkLðT 1 T 2 Þ
Q_ ¼ Biot number Bi ¼ hD/k for the fin should be
lnðr 2 =r 1 Þ small. In a fire scenario, a fully exposed beam
We can extract a resistance from this expression might be modeled as being a fin [10]. Develop-
to be: ment of the pin fin equation begins with a power
balance on a differential section of the fin, as
lnðr 2 =r 1 Þ shown below (Fig. 2.5).
R¼
2πkL One dimensional analysis (radially lumped) is
valid when d/L<<1 and Bi<<1.
The power balance on the differential element
Fin Approximation coupled with performing a limiting process as Δx
The fin approximation refers to one dimensional approaches zero results in the pin fin equation
conduction analysis where heat transfer has a shown below.
predominant direction, there is no transverse tem-
perature gradient, and the heat transfer in the d2 T
kAc h c Pð T T 1 Þ ¼ 0
transverse direction is simply defined through dx2
Newton’s law of cooling. The simplest example
of use of the fin approximation is in the develop- This second order ordinary differential equation
ment of the pin fin model. A pin fin is slender rod requires two boundary conditions (BCs). The
of length L and diameter D with convective heat base temperature is often specified as known
transfer taking place over most of the rod’s (i.e., T(x ¼ 0)). Typical BCs at x ¼ L are:
(a) Known tip temperature or Dirichlet condition
(e.g., T(L) ¼ 20 C)
(b) Known tip heat flux or Neumann condition
(e.g., q00 (x ¼ L) ¼ 0)
(c) Convection tip or Robin’s condition (e.g.,

kdT
dx x¼L ¼ hðT ðx ¼ LÞ T 1 Þ
Solutions and examples of use of the fin
approximation will be provided in a later section.
It is, however, useful to discuss one limiting case
solution for the pin fin. A pin fin is said to be
Fig. 2.4 Schematic of cylindrical shell in which one semi-infinite if the effect of the imposed temper-
dimensional, steady cylindrical formulation of heat equa- ature at the fin base does not affect the tempera-
tion is developed ture distribution over the entire length of the fin.
Fig. 2.5 Development of pin fin equation for cylindrical rod

32 O.A. Ezekoye
Simple 1D Composite Systems Example 2 A composite wall for a furnace is

The simplest composite systems can be thought made of two materials, an insulating material
of in two limiting cases. One might imagine with thermal conductivity kA and an exterior
configurations with two different conductors skin with thermal conductivity kB. Within the
either in a series or in a parallel configuration. furnace there is an internal heat transfer coeffi-
For the simpler case, in which the conductors are cient hc,i and internal temperature Ti. Outside of
in series, it is useful to discuss the nature of the this wall there is a heat transfer coefficient hc,o
interface between the two conductors. The and external temperature To. We can calculate
idealized interface between surfaces is considered the total heat flux (heat transfer rate per unit area)
to be perfect contact in which the interface and also the intermediate temperatures (e.g., the
temperatures are the same on both surfaces. In interface temperature between the two materials)
contrast, an imperfect contact is said to have a using a simple resistance analogy (Fig. 2.7).
contact resistance and there is a temperature dis- By identifying a simple conductive resistance
continuity at the interface between the two for the slabs we see that the heat transfer rate can
surfaces (Fig. 2.6). In the equation shown below be specified as
hi is known as the interface conductance with the
Ti T1 T1 T2 T2 T3 T3 T0
same units as the heat transfer coefficient (W/m2K) Q_ ¼ ¼ ¼ ¼
1=hc, i A LA =kA A LB =kB A 1=hC, o A

dT dT
kA ¼ hi ðT u T s Þ ¼ kB The heat transfer rate can be thought of as being
dx u dx s proportional to a potential difference (i.e., the
Fig. 2.6 Comparison of

perfect and imperfect
thermal contact showing
contact resistance
temperature jump for q’’ dT dT
imperfect contact A − kA = − kB
dx u dx s
B
kA>kB
Q = hi A(Tu − Ts)
Fig. 2.7 Circuit analogy

for conduction in two slabs
kA kB
hc,i , Ti hc,o , To
Ti T1 T2 T3 To
1/hc,i A LA/kAA LB/kBA 1/hc,o A
www.ebook777.com
temperature drop) and inversely proportional to a or spherical coordinate systems depending on the
resistance. We can use: particular type of problem of interest.
_ i ¼ ΔT i
QR Boundary Condition Approximations
X X
Q_ Ri ¼ ΔT i and Assumptions
Boundary conditions are one of the three types
The sum of the individual potential drops equals that were previously identified in the fin discus-
the total potential difference across the system. sion: prescribed temperature also known as
Dividing by the sum of the resistances, we get: Dirichlet conditions, prescribed flux also known
as Neumann conditions, or mixed flux and tem-
Ti To
Q_ ¼ X which on a per unit area basis becomes perature also known as Robins conditions. For
Ri Neumann type conditions, a so-called compati-
00 Ti To bility condition must hold for a steady formula-
q ¼ X 00
Ri tion to be valid. In practice, this means that the
next heat flux into the object must be zero to
For a case with Ti ¼ 1000 C, To ¼ 300 C, ensure that a steady solution exists. The three
hci ¼ 30 W/(m2K), hco ¼ 10 W/(m2K), LA ¼ boundary conditions are stated mathematically
2 cm, LB ¼ 0.2 cm, kA ¼ 1 W/(mK), kB ¼ below.
20 W/(mK) we find that the equivalent resistance
Tjx¼L ¼ T s
00
X 00 Km2
REQ ¼ Ri ¼ 0:153
W ∂T
k ¼ qs
00
The heat flux is q ¼ 1000300 ∂x x¼L
0:153 ¼ 4:56 kW=m .
2
Check the magnitudes of all the resistance

∂T
elements. If any of these is particularly small k ¼ hc Tjx¼L T e
∂x x¼L
compared to the others, we can sometimes
neglect that effect in our analysis.
Transient One Dimensional Models

Steady-Multidimensional Models
For formulations in which transient effects must
For problems in which the heat transfer processes be included, there are conditions in which there is
internally equilibrate on time scales that are a predominant heat transfer direction. When
shorter than the times over which external time there is a preferred direction for heat flow, the
dependent processes and/or thermal forcing thermal forcing can either be from one of the
occur, the problem can be formulated as being boundaries or from an internal source. For most
steady. There are conditions in which there is not fire problems, the forcing will occur from a
a dominant or preferred direction for heat flow boundary or face. While all physical problems
for which a one dimensional formulation is inap- have some finite characteristic length, we will
propriate. For such problems a multidimensional discuss an approximation that specifies the extent
model is required. The governing equation in this of the domain to be semi-infinite.
formulation is simply:
000
. Thermally Thick and Thin Approximations
∇2 T ¼ Q_ v k We can summarize studies of one dimensional
000
unsteady conduction using a map of solution
Where Q_ v is a volumetric source term. approximations in a Fourier number (Fo) and
Problems can be set up in Cartesian or cylindrical Biot number (Bi) space. The Fourier number
34 O.A. Ezekoye
physically represents the thermal penetration

thickness, δ, divided by the geometric dimension
of the system, L.
2
αt δ
Fo ¼ ¼
L2 L
As previously noted, the Biot number can be

interpreted as a ratio of the conductive resistance
to the convective resistance or as the ratio of the
temperature differences in the solid to the driving
convective temperature difference. Recall that a
lumped thermal approximation is valid when
Bi<<1. Fig. 2.8 Map identifying different conduction solution
The Biot number defined in terms of the true regimes as described by Biot and Fourier numbers
length scale of the object can be considered to be (Adapted from [12])
the global Biot number for the scenario. A local
Biot number can be constructed using the ther- Transient Multidimensional Models
mal diffusion length as the characteristic length.
In the following equation, it can be shown that For transient multidimensional cases, we rely on
the local Biot number is specified as the product the full solution of the heat equation. For simple
of the global Biot number and the square root of geometries, we can use analytical methods to
the Fourier number [11]. construct the solutions. With increased geomet-
rical complexity, computational methods are
hR hδ R hδ 1=2 required. In later sections, both analytical and
Bi ¼ ¼ ¼ Fo
k k δ k computational techniques for solving these
problems will be detailed.
hδ Tw To
¼ Bi Fo1=2 ¼ ∂T ∂ ∂T ∂ ∂T
k T1 Tw ρc ¼ k þ k
∂t ∂x ∂x ∂y ∂y
This local Biot number can then be interpreted as
∂ ∂T
þ Q_v
000
an estimate of the temperature difference between þ k
∂z ∂z
the wall and interior temperatures relative to
difference between the wall and fluid reservoir
temperatures. Because it is a local Biot number
and defined in terms of the thermal penetration Analytical Solutions and Examples
depth, it also provides insight into what might be
considered to be a semi-infinite domain. For small Despite the increased accessibility and power of
Fourier numbers, the thermal diffusion distance δ computing devices, analytical models continue
is smaller than the global or geometric length of to be important in analyzing and characterizing
the object. The map below can be used to charac- conduction phenomena in fire systems. Analyti-
terize the various domains for which different cal solutions are essential in the verification of
types of approximations are valid. The diagonal computational models and are often used to pro-
line in the upper left hand quadrant represents the vide back-of-the-envelope engineering guidance.
demarcation between a convective boundary con- In this section, solutions will be provided for
dition and fixed surface temperature on a semi- some of conduction formulations that have been
infinite body (Fig. 2.8). discussed.
www.ebook777.com
Steady-One Dimensional Examples the insulated pipe through convection and

radiation. There is a convective heat transfer
Critical Thickness of Insulation coefficient hc,o and the radiative transfer process
For the cylindrical system, we noted that the one is approximated using a radiation heat transfer
dimensional steady form of the heat equation is: coefficient. We can define an overall conductance
for the system using the convective resistances
d dT and also the conductive resistance that we just
rk ¼0
dr dr developed for cylindrical systems.
Integrating twice, we derive the temperature Ti To
Q_ ¼ UAðT i T o Þ ¼
solution. 1=UA
T1 T lnðr=r 1 Þ The overall conductance is modeled as

¼
T 1 T 2 lnðr 2 =r 1 Þ 1 ðr2 =r 1 Þ
¼ 2πr11Lhc, i þ ln2πk ðr3 =r2 Þ
þ ln2πk þ 2πr 1
UA AL ðhc, o þhr, o Þ
BL 3L
The heat transfer rate is given by: where r1 is the inner radius of the pipe, r2 is the
outer radius of the pipe, and r3 is the outer radius
2πkLðT 1 T 2 Þ
Q_ ¼
of the insulation. We can neglect the internal
lnðr 2 =r 1 Þ convection resistance and the conduction resis-
tance of the pipe relative to the external convec-
A resistance can be defined as: tion resistance and the conduction resistance of
the insulation when:
lnðr 2 =r 1 Þ
R¼
2πkL hc , o r 3 hc, o r 3 lnðr 2 =r 1 Þ
<< 1 and
h c, i r 1 kA
Often in fire scenarios there is a desire to mini-
mize heat transfer to a pipe or rod from the high hc, o r 3 lnðr 3 =r 2 Þ
<<
temperature environment. Insulating the pipe or kB
rod is one approach to minimize the heat transfer For such a case, we find that the heat transfer
rate. The effect of insulating a pipe is explored in rate is related to the driving temperature differ-
this discussion. A schematic of the system is ence as:
shown below:
Ti To Ti To
hc,i Q_ ¼ ¼
R lnðr o =r i Þ=2πLk þ 1=2πLr o ho
While we might assume that the addition of an

hc,o and hr,o insulating layer will always result in a decrease
in the heat transfer rate, we can easily see that the
denominator has a minimum value when there
has been addition of insulation. This minimum
kA kB value of resistance means that the heat transfer
rate increases with the addition of some insula-
Consider that an internal heat transfer coeffi- tion. Taking the derivative of the denominator
cient exists because of the fluid flow within a and identifying the extremum shows that when
pipe. The pipe is made of material A and has a the insulating layer has a radius of
thermal conductivity, kA. Insulation wrapped
around the pipe has thermal conductivity, kB. k
r 0 ¼ r cr ¼
The fluid flowing in the pipe is at a temperature ho
Ti. External to the insulated pipe is an environ-
ment at temperature To that interacts with the heat transfer has a maximum value.
36 O.A. Ezekoye
Example 3 An exposed water sprinkler line is increased the conductive resistance by the insu-
within a compartment on fire. The fire products lation, we decreased the external convective
are at a temperature of 1000 C. We can assume resistance by increasing the area. The total resis-
that the heat transfer coefficient between the fire tance has decreased from 0.354 Km/W to
products and the pipe is 30 W/(m2K). The pipe 0.33 Km/W. The choice of insulator matters.
has a diameter of 5 cm, a thickness of 5 mm, and The same thickness insulation, but with a thermal
a thermal conductivity of 30 W/(mK). At one conductivity of 0.1 W/(mK) increases the equiv-
section of the pipe, the water has a mean temper- alent resistance to 0.59 Km/W and decreases the
ature of 30 C and an internal heat transfer coef- total heat transfer rate to 1.6 kW/m.
ficient of 100 W/(m2K). We can calculate the
heat flux at this section of pipe using a thermal Fin Model of a Beam Extending Between
resistance circuit. Two Walls
Frequently, engineers must evaluate the thermal
response of beams and columns that are affected
by high temperature gases associated with a fire
[10]. Imagine that two walls of a compartment
ho, To
are at temperatures Tw and a high temperature
hi, Ti
gas flows around a fully exposed beam. We can
evaluate the thermal response of the beam using
a fin model (Fig. 2.9).
Earlier, we derived the constant cross-
sectional area fin equation:
00 ðT i T o ÞAi Ti To d2 T
q i Ai ¼ X 00 ¼ kAc h c Pð T T 1 Þ ¼ 0
Ri ln ro
ri
dx2
1
hi Ai þ 2πkL þ ho1Ao
Solutions to this equation are either in terms of
Note that the radial location of the heat flux must exponential functions or hyperbolic sine and
be specified as the heat flux is different at the cosine. It is convenient to define an excess tem-
outer edge of the pipe from its value at the inner perature in terms of the difference between the
edge. The total heat transfer rate is 2.7 kW/m of fin temperature and the fluid temperature
pipe. The heat flux at the inner wall is 9.7 kW/m2. (Fig. 2.10).
For an insulator with thermal conductivity of
1 W/(mK) and the same external heat transfer d2 θ
m2 θ ¼ 0
coefficient of 30 W/(m2K), the critical radius dx2
r cr ¼ hko is 0.03 m or 3 cm. Thus, adding 1 cm of θ ¼ T T1
an insulator with a thermal conductivity of 1 W/ m2 ¼ hc P=kAc
(mK) to the pipe increases the heat transfer rate.
We would have:
00 ðT i T o ÞAi 2L
qi Ai ¼ X 00
Ri
T To
¼ i
rp ro
ln ln
ri rp x
1
hi Ai þ 2πkL þ 2πkL þ ho1Ao
The total heat transfer rate increases to Fig. 2.9 Schematic of beam convectively heated
2.9 kW/m. As previously noted, while we between two walls
www.ebook777.com
T(x) hF, TF hC, TC
x−
x+
Fig. 2.11 Schematic diagram of a bar extending between

two domains with different temperatures and different
−L L x heat transfer coefficients
Fig. 2.10 Temperature profiles for suspended beam

example showing effect of the gas to wall temperature
At the interface of the fire flow and the cold flow,
difference on the temperature profile the bar temperature is continuous and can be
specified to be an interface temperature TI.
For this case in which the walls are at the same Also, the heat flux is continuous at the interface.
temperature we arrive at a solution: Since the thermal conductivity does not change,
we simply write this as:
coshðmxÞ
T ¼ T 1 þ ðT w T 1 Þ
coshðmLÞ ∂T ∂T
¼
∂xþ ∂x
If we had defined the origin for the x coordinate 1=2 1=2
as being one of the walls, and if the parameter ðT I T C, 1 Þ hkAA Pc ¼ ðT I T F, 1 Þ hkAF Pc
mL>>1, the exponential solution is appropriate
and decays to zero at large x values. mF T F, 1 þ mA T C, 1
TI ¼
mF þ mA
θ
¼ emx
θb For a case in which the heat transfer coefficient is
the same on both sides of the interface, the inter-
Example 4 An unprotected round steel bar face temperature is 310 C and is a simple aver-
extends across two compartments. The bar has a age of the fire side gas temperature and the cold
diameter of 5 cm. One compartment is fully compartment air temperature. If the heat transfer
involved in fire with gas temperatures of 600 C coefficients had been different, the interface tem-
around the bar. The heat transfer coefficient is perature is the weighted average of the two
30 W/(m2K). The compartment temperature of temperatures as shown above.
the room that the bar extends into is at 20 C. For
convenience, we assume that the void space also Flame Temperature Thermocouple
has a air temperature of 20 C and that the heat Measurement
transfer coefficients are also 30 W/(m2K) in the Consider a thermocouple that is modeled as a rod
void space and in the nonfire compartment. We of diameter D, thermal conductivity k, and length
can estimate the temperature of the bar within the L with half of the thermocouple length in air and
void space to see if it might pose an ignition the other half inserted into a pool fire flame. It is
hazard using fin analysis. useful to determine if there is a conduction error
For a sufficiently long bar we can use the in the thermocouple. There are several possible
semi-infinite assumption which states that the fin models that can be used to illustrate this
temperature distribution in the fin has an expo- effect. Here, we take an idealized scenario in
nential variation (Fig. 2.11). which the thermocouple wire is modeled as
θC ðxÞ T ðxþ Þ T C, 1 þ θF ðxÞ being a fin with insulated ends in the hot fluid
¼ ¼ emx and and also in the cold fluid. In reality, there is a
θ C, B T I T C, 1 θ F, B
convective end condition in the hot region (i.e., at
T ðx Þ T F, 1 þ
¼ ¼ emx the thermocouple junction) and the termination
T I T F, 1 point in the cold region is often very far away
38 O.A. Ezekoye
Fig. 2.12 Fin suspended Ts

across free surface in fluids
with different temperatures
and heat transfer
coefficients
∂2T ∂2T
+ =0
To ∂x2 ∂y2 To
To
Fig. 2.13 Plate in which Laplace equation is to be solved

to determine 2D temperature profile
from the point where the thermocouple is of 60 W/(mK), and a length of 6 cm equally split
inserted into the flame. The question posed here between the hot and cold fluids, we find that the
is whether the temperature at the end placed into interface temperature is 1150 K and that the tip
the hot fluid (i.e., the flame) reaches the hot fluid temperature is 1751 K. This represents an
temperature (in which case, the thermocouple is approximately 12 % error in the predicted
measuring the correct temperature) or whether freestream temperature.
the conduction losses to the cold side are
affecting the measurement. The insulated end
solution for a rod of diameter D, thermal conduc- Steady Multidimensional Example
tivity k, and length L suspended between two
fluids/flows is There are many heat transfer systems for which
there is a need to generate multidimensional
θA T A T A1 cosh½mA ðLA xA Þ solutions. Often it is appropriate to model a
¼ ¼
θA0 T A0 T A1 coshðmA LA Þ two dimensional temperature variation for
If the fin is not semi-infinite (i.e., mL is not large geometries in which two characteristic lengths
relative to unity), we need to use the full solution are of comparable magnitude and the third char-
to get the interface temperature. The interface acteristic length is significantly longer. For such
temperature (Fig. 2.12) is geometries, if one is interested in specifying the
temperature variation at a cross-section of the
T A1 mA tanhðmA LA Þ þ T B1 mB tanhðmB LB Þ geometry at lengths far from the boundaries of
T A0 ¼
½mA tanhðmA LA Þ þ mB tanhðmB LB Þ the long direction, a two dimensional approxima-
tion often proves to be valid. In this section, we
As we saw earlier, if the heat transfer coefficient discuss the separation of variables approach to
is the same on the hot and cold sides, we get an solving such problems.
interface temperature that is the simple mean
value of the hot and cold temperatures. The Separation of Variables Applied to Two
error in the measured end temperature Dimensional Fin

T A ðLÞ T A1 1 T B1 1
¼ 1 Example 5 A very long rectangular bar has three
T A1 2 T A1 coshðmA LA Þ
sides maintained at temperature 20 C and one
For a case with an air temperature of 300 K and a side at temperature 120 C (Fig. 2.13). Because
flame temperature of 2000 K, if the thermocou- the bar is very long, we can neglect the axial heat
ple has a diameter of 2 mm, a heat transfer transfer problem and focus our attention on the
coefficient of 30 W/m2K, thermal conductivity heat transfer processes at some intermediate slice
www.ebook777.com
within the bar. We are simplifying this three in a weighted integral sense. There is a deep
dimensional problem into a two dimensional relationship between the process in separation
problem. Further, we can define a new tempera- of variables and the theory of Fourier series as
ture, sometimes called the excess temperature, well as many computational techniques for solv-
defined as the temperature subtracting off some ing differential equations. One simple starting
reference value. For this problem, it is conve- point for separation of variables solution is to
nient to think of the 20 C temperature as a assume a solution form of
reference value.
We define an excess temperature to be θðx; yÞ ¼ XðxÞY ð yÞ
θðx; yÞ ¼ T ðx; yÞ 20. In terms of the excess Upon substituting this solution form into the
temperature, we have the following boundary partial differential equation, one obtains two
conditions: 2
separable ordinary differential equations: X1 ddxX2
x ¼ 0, 0 < y < b; θ¼0 2
¼ Y1 dd yY2 ¼ λ2
y ¼ 0, 0 < x < a; θ¼0 The solution of the X(x) equation yields the
x ¼ a, 0 < y < b; θ¼0 trigonometric functions, while the solution of the
y ¼ b, 0 < x < a; θ ¼ 100
Y(y) equation yields hyperbolic trigonometric
We use a standard technique for the solution of functions sinh and cosh. Substitution of this
finite domain partial differential equations called form of solution into the PDE results in unique
separation of variables to solve for the tempera- choices for the parameters λ and also presents an
ture distribution in the plate section. Separation opportunity to define the so-called Fourier
of variables relies on expanding the dependent coefficients for the problem (Fig. 2.14).
function θ(x,y) in terms of an appropriate set of
basis functions. For this Cartesian coordinates X1
2½1 ð1Þn
θðx; yÞ ¼ 100
example, the basis functions turn out to be sine
n¼1
nπ sinhðnπb=aÞ
and cosine functions. The members of the family
of functions that constitute the set of basis nπx nπy
sin sinh
functions are said to be orthogonal to each other a a
Fig. 2.14 Separation of

0 0.9 0.9
0.1
00.02.1
0.503.4
0.7.8
.5
variables solution (θ(x, y)/ 0.9

.
0.8
0.4 0
60
0.8 0.6
100) for 2D plate 0.7
0.2
conduction problem 0.6 0.7

0.6
50 0.5 0.5
3
0.
0.4
0.3 0.4
y dimension (mm)
40
0.
2
0.1
0.3
0.2
0.1
30 0.2
20 0.1 0.1
10
0
0 10 20 30 40 50 60
x dimension (mm)
40 O.A. Ezekoye
Transient Lumped Examples with Time

Dependent Forcing
Often the transient response of a thermal system

must be evaluated. The lumped thermal approxi-
mation is valid when the internal conduction
resistance is small relative to the external con-
vective resistance.
Fig. 2.15 Schematic showing convective flow over an
Internal conduction resistance L=ks A hc L object that will be analyzed using a lumped thermal
ffi ¼ approximation
External convection resistance 1=hc A ks
Imagine a cold spherical heat sensor emersed in

an initially cold gas for which the gas tempera-
ture is a linearly increasing function of time. It is
useful to understand how the heat sensor’s tem-
perature will vary with time. Assume that
Bi ¼ 0.1 and Fo ~ 10. T(r,t ¼ 0) ¼ To.
The model used to characterize the tempera-
ture variation of the sphere is the same lumped
thermal model previously discussed. The differ-
ence is that because the reservoir temperature is
time varying, the temperature solution is no lon-
ger a simple exponential function (Fig. 2.15).
dT
ρVc ¼ hc AðT T 1 Þ
dt
with T 1 ðtÞ ¼ T o þ at
dT hc A
¼ ðT T 1 Þ
dt ρVc
Fig. 2.16 Solution for lumped approximation problem in
The solution (Fig. 2.16) with θðtÞ ¼ T T o is which fluid temperature is a linearly increasing function
h t i of time
θðtÞ ¼ aτ exp 1 þ at
τ
h ti parameterized by a new independent variable. For
¼ a t τ 1 exp an ordinary differential equation (e.g., a thermally
τ
lumped system analysis), application of the
We see in Fig. 2.16 that there is a time lag Laplace transform yields an algebraic equation
between the increasing temperature of the fluid that can be solved for the transformed dependent
and the increasing temperature of the detector. variable. The time dependent form is retrieved
Additionally, we see that the detector tempera- using an inversion integral called the Bromwich
ture at any given time is always lower than the integral. In the equation below, the Laplace trans-
fluid temperature. form integral operator is applied to a general tem-
perature function that depends on time and spatial
Laplace Transform Methods location. The definition of the Laplace transform is
The Laplace transform is one of several integral shown below.
transform methods that can be used in conduction
analysis [12]. For time dependent functions, the ð
1
Laplace transform maps the time dependent LðT Þ ¼ T^ ðx; sÞ ¼ T ðx; tÞest dt
derivative terms into algebraic terms that are 0
www.ebook777.com
An example of the use of the Laplace transform solution for a step change. A simple case to
is provided below. Consider the lumped analysis present the ideas of this concept is the problem
system that we have been using in which there is a step change in the external
temperature of a reservoir in contact with a
dT hc A lumped thermal system. We know the solution
¼ ðT T 1 Þ
dt ρVc from earlier sections.
Straightforward application of the transform
yields: dT hc A
¼ ðT T 1 Þ
dt ρVc
hc A ^ T1
sT^ ðsÞ T ð0Þ ¼ T ðsÞ Graphically, a series of external temperature step
ρVc s
changes can be looked at as follows:
Because the initial condition shows up in this
transformation, it is often convenient to use
superposition to force the initial condition to be
zero. For such a case we have,
T2
hc A ^ θ1
sθ^ ðsÞ ¼ θ ðsÞ
ρVc s
T1
Solving for the transformed temperature variable
To
yields:
θ
θ^ ðsÞ ¼ 1 If we focus on the initial step change, this is
τc s s þ τ1
c no different from the original problem that we
Using any one of a number of inversion tables or considered. The offset in time can be addressed
online calculators, we retrieve the inverse of this using an offset in time. Simply by creating a new
function. time variable, e.g., s ¼ tt1, we create an offset
time variable. To explore what is meant by a step

t change in the external temperature, we formally
θðtÞ ¼ θ1 1 exp
τc formulate the mathematical equation using the
Heaviside step function.
The result above is equivalent to the result shown
earlier. The power of the Laplace transform dT hc A
method is that it can be applied to arbitrarily ¼ ðT T 0 ðT 1 T 0 ÞH s ðt t1 Þ
dt ρVc
complex differential equations. The challenge
ðT 2 T 1 ÞH s ðt t2 ÞÞ
of the method has typically been generating the
inverse transform. With increasing accessibility If we now define a relative temperature TT0,
of symbolic mathematical software tools, with we arrive at the equation
some freely available online such as Wolfram
Alpha, this particular challenge is no longer dθ hc A
quite as severe. ¼ ðθ Δθ1 Hs ðt t1 Þ Δθ2 H s ðt t2 ÞÞ
dt ρVc
Duhamel Integral Methods To solve this equation, it is convenient to

The Duhamel integral is a method for generating define a series of linear transformations and
a solution for an arbitrarily complex time depen- use superposition to solve this individual
dent forcing of a conduction system using the equations.
42 O.A. Ezekoye
θ ¼FþG inverse analysis is the determination of the heat

dF dG hc A flux from a temperature measurement. The gener-
þ ¼ ðF þ G Δθ1 H s ðt t1 Þ alization of the Duhamel form states that the time
dt dt ρVc
dependent variation of temperature can be defined
Δθ2 H s ðt t2 ÞÞ
in terms of an integral of the product of the step
dF hc A response solution, which is time dependent, and
¼ ðF Δθ1 H s ðt t1 ÞÞ
dt ρVc the time derivative of the unsteady effect.
dG hc A
¼ ðG Δθ2 H s ðt t2 ÞÞ
dt ρVc
Transient Semi-infinite (Thermally
Because θ(0) ¼ 0, we can enforce that both F Thick) One-Dimensional Examples
and G are also equal to zero and have solutions of
form: For thermally thick problems in which the ther-
mal penetration wave never reaches the back
FðtÞ ¼ Δθ1 1 eðtt1 Þ=tc side, analytical solutions are available for a
range of boundary conditions. These solutions
The solution for the excess temperature is then:
are generally specified in terms of tabulated
functions [13]. Useful approximate solutions
θðtÞ ¼ Δθ1 1 eðtt1 Þ=tc
can also be developed for these problems using

integral approximations and the scaling
þ Δθ2 1 eðtt2 Þ=tc
properties of diffusive transport (Fig. 2.17).
For a series of step changes, we can generalize For the thermally thick conduction problem,
this superposition of solutions to generate: the solution is often described as being for a
semi-infinite domain (Fo<<1). Using either a
X Δθi
θ ðt Þ ¼ 1 eðtτi Þ=tc Δτi Laplace transform solution approach or any num-
Δτi ber of other approaches, one can arrive at the
ð
dθ
t temperature variation for a constant surface tem-
¼ 1 eðtτÞ=tc dτ perature boundary condition.
dτ
ð
2 η u2
0
T T0
We immediately see that for the case in which θ¼ ¼ 1 1=2 e du
Ts T0 π 0
the freestream temperature is changing as a linear !
function of time T 1 ðtÞ ¼ T o þ at x
¼ erfc
We get ð4αtÞ1=2
ðt We see that the solution is defined using the
θðtÞ ¼ a 1 eðtτÞ=tc dτ complementary error function (erfc). The surface
0 heat flux variation with time is:

¼ at aet=tc tc et=tc 1 " #

2 η2 1
¼ at atc 1 et=tc ¼ a t tc 1 et=tc qs ¼ kðT s T 0 Þ 1=2 e
π ð4αtÞ1=2 η¼0
which is the same result that we arrived at using a k ðT s T 0 Þ
much simpler analysis in an earlier section. The ¼
ðπαtÞ1=2
power of the Duhamel formulation is evident in
problems like the semi-infinite slab problem for Interestingly, an approximation for the heat flux
which the solution is expressed in terms of the that depends on the thermal penetration depth
error function. One interesting application of the does a reasonable job of predicting the overall
Duhamel method that we will use in the section on trends.
www.ebook777.com
Fig. 2.17 Governing equation for semi-infinite model of conduction in a slab and associated temperature profiles
k ðT s T 0 Þ k ðT s T 0 Þ The Newton’s law of cooling convective bound-

qs ffi ¼ pffiffiffiffi
δðtÞ αt ary condition case is analytically quite challeng-
ing, but can be reduced to a very simple form
In the approximation, we have used the sing the
pffiffiffiffi using the type of scaling previously described
simplified scaling law δ αt (Fig. 2.18).
A similar approximation can be used for the We note that the internal heat flux can be
constant surface heat flux boundary condition. approximated using a diffusion model and must
We recognize that the approximate variation is be balanced by the Newton’s law of cooling
as follows: convection term.
pffiffiffiffi pffi
q αt q t kðT s ðtÞ T 0 Þ
Ts ¼ T0 þ s ¼ T 0 þ psffiffiffiffiffiffiffi qs ffi ¼ hðT e T s Þ
k kρc δ
The surface temperature has t1/2 dependence In solving for the wall temperature we arrive at:
which we can compare to exact analytical result
pffiffiffi
" # h αt
T s ðtÞ T 0
qs 4αt 1=2 x2 =4αt x ¼ k pffiffiffi
T ðx; tÞ ¼ T 0 þ e x erfc ðT e T 0 Þ 1 þ h kαt
k π ð4αtÞ1=2
which at
" At long time, we get constant surface tempera-
1=2 #
q 4αt ture. We can compare this to the exact solution
x ¼ 0 is T s ¼ T 0 þ s
k π (Fig. 2.19):
44 O.A. Ezekoye
Fig. 2.18 Schematic of semi-infinite model of conduc-

tion in a slab with convective boundary conditions
Fig. 2.20 Schematic of one dimensional conduction in a

1 slab with convective boundary conditions
f(x)
g(x) thermal behavior of such systems. The starting

point for constructing such solutions is a one
0.5 dimensional transient problem. As an example,
for the slab/wall geometry for which there is
convective heating on two sides, there is a sepa-
ration of variables series solution for the prob-
lem. For relatively long times as is defined by the
0 Fourier number (Fo>1), the series solution
0 2 4
converges with a small number of terms. Physi-
x
cally, the temperature profile (i.e., the solution)
Fig. 2.19 Comparison of approximate and exact solution becomes smoother with increasing time. Because
of semi-infinite model of conduction in a slab with con- so few terms are required to represent the tem-
vective boundary conditions perature field evolution, it is possible to define
the solution in terms of a simple product of a time
dependent function and a spatially varying func-
Ts T0 2 hc
¼ 1 eðhc =kÞ αt erfc ðαtÞ1=2 tion (Fig. 2.20). The governing equation, initial
Te T0 k condition and boundary conditions for the one
dimensional problem are:
Transient Multidimensional Examples ∂T ∂2 T
¼α 2
∂t ∂x
Multidimensional transient problems represent a
t ¼ 0 : T ¼ T0
class of problems that pose sufficient analytical
complexity that we often revert to computational ∂T
x¼0: ¼0
solutions for these systems. For some simple ∂x
geometries, however, there are tabulated results ∂T
x¼L: k ¼ hð T T 1 Þ
that are useful for determining the overall ∂x
www.ebook777.com
The solution of the one dimensional problem is:

Overview of Computational Issues
T T1 θ X1 in Conduction
¼ ¼
T 0 T 1 θ0 n¼0
For complex geometries or for problems with
sin ðλn LÞ nonlinear effects and multiple physical processes,

λn L þ sin ðλn LÞ cos ðλn LÞ analytical solutions are usually not easily found.
For the majority of practical heat transfer pro-
eλn αt
2
cos ðλn xÞ
cesses in fire, a fire engineer will resort to the
use of computational tools to characterize the
The eigenvalues of the problem can be written as:
system. Olenick and Carpenter [14] summarize
λn L ¼ Bi cot ðλn LÞ. Also, the argument of
the key features of a number of computational
the exponential function can be written as:λ2n αt ¼
tools available for simulating thermal response
ðλn LÞ2 Lαt2 ¼ ðλn LÞ2 Fo. This suggests that the associated with fire endurance testing. These
solution is defined by the following parameters: tools use a variety of numerical discretization
techniques. In the following section, we discuss
x the types of computational techniques most often
Bi, Fo,
L used for fire problems and present simple
For long enough times, the one term expansion examples for some of these cases to help clarify
generates [2]: what underlying steps are taken in some of the
tools and codes described in [14]. Useful refer-
T T1 θ θ θCL ence texts for numerical solution of the differen-
¼ ¼
T 0 T 1 θ0 θCL θ0 tial equations associated with conduction are Shih
¼ CðBiÞ f ðtÞgðxÞ [15] and Ferziger [16].
To start, we define a very simple conduction/
The centerline temperature varies only with time, diffusion problem with a source (S).
while the temperature at any location normalized
by the centerline temperature is essentially a d2 ϕ
Γ þS¼0
function of the spatial location. dx2
It can be easily shown that the temperature The analytical solution can be easily found by
within a block can be found as the product of the integrating this equation twice. For the case of a
temperatures found from three separate one constant source we get:
dimensional solutions (Fig. 2.21):
S 2
θðt; x; y; xÞ ¼ Pðx; tÞPð y; tÞPðz; tÞ ϕ ¼ C2 þ C1 x x
2Γ
Consider the case with boundary conditions:
h,T∞
dϕ
¼ 0 and ϕðLÞ ¼ 0
dx x¼0
h,T∞ C1 is zero by the adiabatic condition, and C2 ¼ 2Γ

S
such that;
S
ϕ¼ L x2
2Γ
Fig. 2.21 Schematic diagram of cube with external con- We will compare this solution to those generated
vective heating by the numerical solutions.
46 O.A. Ezekoye
Finite Difference Approximations Symbolically, we can arrange the system of alge-

braic equations into a form shown below.
The finite difference technique is used to solve
differential equations by expanding the depen- ai1 ϕi1 ai ϕi þ aiþ1 ϕiþ1 ¼ b
dent variable, at specified positions in the Γ Γ 2Γ
ai1 ¼ , aiþ1 ¼ , ai ¼ , b ¼ S Δx
computational space into a Taylor series, appro- Δx Δx Δx
priately adding and subtracting terms from other
For the case in which Γ ¼ 0:2, 0 < x < L,
series terms together to generate approximations
and S ¼ 4
for the derivative terms in the underlying differ-
We will use N ¼ 5 grid points,
ential equation. As an example, the dependent
Δx ¼ 0:25L for L ¼ 1
variable is expanded below in a Taylor expansion
at locations 1 and 3 (Fig. 2.22). ai1 ¼ 1, aiþ1 ¼ 1, ai ¼ 2, b ¼ 1

∂ϕ 1 ∂2 ϕ 2 32 3
ϕ1 ¼ ϕ2 Δx þ Δx2 þ O Δx3 2 2 ϕ1
∂x 2 2 ∂x 2
2
6 1 76 ϕ2 7
6 2 1 76 7
6 1 2 1 76 ϕ3 7
∂ϕ 1 ∂2 ϕ 6 76 7
4 1 2 1 54 ϕ4 5
ϕ3 ¼ ϕ2 þ Δx þ Δx2 þ O Δx3
∂x 2 2 ∂x 2
2
1 ϕ5
2 3 2 3 2 3
By subtracting the two series from each other, we 1 ϕ1 8
6 1 7 6 ϕ2 7 6 7:5 7
arrive at one representation for the derivatives at 6 7 6 7 6 7
location 2. ¼6 7
6 1 7 with solution 6 7 6 7
6 ϕ3 7 ¼ 6 6 7
4 1 5 4 ϕ4 5 4 3:5 5

∂ϕ ϕ ϕ1 0 ϕ5 0
¼ 3 þ O Δx2
∂x 2 2Δx

∂2 ϕ ϕ 2ϕ2 þ ϕ1 Which is exactly the same as the answer
¼ 3 þ O Δx2
∂x 22 Δx 2
provided by the exact solution.
Substituting the approximations into the original
differential equation then provides an algebraic Finite Volume Approximation
rule by which to evaluate the values of the depen-
dent variable at the prespecified (grid) points. An alternative way of generating algebraic
equations representing the differential equation
∂2 ϕ ϕ3 2ϕ2 þ ϕ1
Γ 2 ¼ Γ is the finite volume technique. The finite volume
∂x 2 Δx2
technique is closely related to Galerkin finite
And the source term can be approximated as element techniques. Both are members of a fam-
S2 ¼ Sð ϕ 2 Þ ily of solution techniques called method of
We see that the final form of the algebraic weighted residuals. For both types of solution
system of equations is techniques, an integrated form of the underlying
differential equation is used to develop the alge-
ϕ3 2ϕ2 þ ϕ1
Γ ¼ Sðϕ2 Þ braic equations. In both types of techniques, one
Δx2
weights the equation before integrating. Because
the weight in the control volume method is just
equal to unity, the process to developing the
algebraic equations has a simple physical inter-
pretation. The finite volume technique is some-
times called the control volume technique or the
Fig. 2.22 Schematic of finite difference stencil subdomain method [17] (Fig. 2.23).
www.ebook777.com
2 32 3 2 3
2 2 ϕ1 1
6 1 2 1 76 ϕ2 7 6 1 7
6 76 7 6 7
6 1 2 1 76 ϕ3 7 ¼ 6 1 7
6 76 7 6 7
4 1 2 1 54 ϕ4 5 4 1 5
1 ϕ5 0
This system of linear equations, written in matrix

form has a banded structure and can be solved
using a number of linear algebraic solution
strategies. For tridiagonal matrices, an efficient
Fig. 2.23 Schematic of finite volume cells
algorithm exists (Thomas algorithm) for solving
the system [18]. The solution is:
2 3 2 3
Consider the diffusion equation in conserva- ϕ1 8
6 ϕ2 7 6 7:5 7
tive form. 6 7 6 7
6 ϕ3 7 ¼ 6 6 7
6 7 6 7
4 ϕ4 5 4 3:5 5
d dϕ
Γ þS¼0 ϕ5 0
dx dx
The solution is exactly the same as the answer
provided by the exact solution.
The equation is integrated over the control
volume
ðe
d dϕ Finite Element Approximations
Γ þ S dx ¼ 0 to become
dx dx
w
This discussion is on Galerkin finite element
ðe techniques which are a subclass of the method
dϕ dϕ of weighted residuals (MWR) [15,
Γ Γ þ S dx ¼ 0
dx e dx w 17]. The method explicitly identifies an approxi-
w
mate solution to the governing equation, for
A linear profile assumption is made between cell which the approximate solution is generated
centroids for ϕ. Assume S varies linearly over the from a broader class of functions than the true
control volume solution resides in. There is an approximation, ϕ,e
to ϕ that produces a residual in the solution of the
Γe ðϕE ϕP Þ Γw ðϕP ϕW Þ diffusion equation when the approximation is
þ S Δx ¼ 0
Δxe Δxw used.
Collecting terms and casting into an algebraic
e
d2 ϕ
equation yields: Γ þS¼R
dx2
aE ϕ E þ a P ϕ P þ a W ϕ W ¼ b In MWR, the weighted average of the residual is
zero, rather than the residual. The weight func-
Γe Γw 2Γw
aE ¼ , aW ¼ , aP ¼ , b ¼ S Δx tion is called W(x).
Δxe Δxw Δxw
ð
Again, for the case in which Γ ¼ 0:25, 0 < x RðxÞW ðxÞdx ¼ 0
< L, and S ¼ 4 domain
We will use N ¼ 5 grid points,
Δx ¼ 0:25L for L ¼ 1 The approximating function is assumed to be
expandable in a set of linearly independent
ai1 ¼ 1, aiþ1 ¼ 1, ai ¼ 2, b ¼ 1 “basis functions” Nj(x).
48 O.A. Ezekoye
Fig. 2.24 Triangular basis functions and schematic of overlapping functions
X
e¼
ϕ ϕ j N j ðxÞ conditions, initial conditions, etc. The use of the
term inverse relates to the idea that the analysis
In Galerkin FEM the weight function is specified inverts the normal rules of causality in order to
to be the same as the basis function. find an answer that best agrees with observations.
Wj(x) ¼ Nj(x). In conduction problems, typical examples of use
ð of inverse analysis are determining the heat flux
RðxÞW ðxÞdx ¼ 0 on a slab given a limited number of interior
domain
temperature measurement or identifying material
properties that allow a model to best match
Effectively, we are forcing the residual to be measured temperature data. There are many
orthogonal to the basis functions. Properties of approaches to generating an inverse solution,
the basis function in terms of ease of evaluation, and some of these techniques are essentially
accuracy etc. suggest the use of polynomials. optimization methods. One might summarize
An example of basis functions is shown below optimization based analysis as the search for
(Fig. 2.24). parameters that when used in a forward formula-
The weighted residual becomes: tion of the problem of interest generate solutions
ðb ! whose deviation from experimental data is
e
d2 ϕ
Γ 2 þ S N j ðxÞdx ¼ 0 quantified., There are, however, specialized
a dx techniques that have been developed for inverse
analysis that have not been specifically devel-
While ϕ should be twice differentiable, the oped from an optimization perspective. These
“weak form” allows a wider class of functions techniques generally require that the problem be
to be used. We integrate by parts. We get an formulated in an inverse sense. For problems
equation in terms of integrals of basis functions formulated in an inverse sense, there ultimately
and coefficients of the unknown function at nodal becomes an ill posed mathematical problem. Ill
points. posed problems [19] are those for which the
Aϕ jþ1 þ Bϕ j þ Cϕ j1 þ D j ¼ 0 solution is not unique or does not smoothly
change relative to small changes in an input
parameter. For such problems, there is a need to
Inverse Conduction Heat Transfer regularize the underlying system of equations.
Approximations and Examples If one considers the simple diffusion equation
that has been presented in earlier sections and
Inverse analysis refers to a variety of approaches considers the need to specify the source in the
used to determine model parameters, boundary medium given some number of temperature
www.ebook777.com
measurements in the slab, the process to deter- data are noisy. We see that for each temperature
mine the source is an inverse analysis. An impor- measurement, there is a corresponding quadra-
tant fundamental consideration in inverse ture rule that maps the integral equation to an
analysis is the notion that experimental noise algebraic equation. We see that this process
exists in the measured data and that such noise results in a matrix equation in which the source
corrupts the inversion process. description explicitly shows up as an unknown.
Given the ODE, we can formulate an integral Using typical linear algebra techniques, we find
equation for the temperatures using a delta func- that the results for the source are quite noisy and
tion source [19, 20]. This is simply a Greens’ not representative of the input source that was
function solution for the problem. In a forward used to drive the forward solution. Note that for
analysis, we would simply integrate the source this exercise, synthetic measured data are sam-
distribution in such a way as to generate a tem- pled from the forward solution.
perature at every point given the source distribu- The Greens’ function solution is constructed
tion. The inversion requires an estimate of the using a delta function source:
source, given some measurements of tempera-
ture. Further, it is assumed that the temperature
d2 ϕ 000 Γ d2 G
Γ 2
¼ S ðxÞ which for the Green’s function is 00 ¼ δðx x0 Þ
dx So dx2
Integrating the delta function yields: The final solution for the dependent variable is
G ¼ ax þ b x < x0 ð x0
x þε L 00 0
dG 0 00
ϕðx0 Þ ¼ ð1 x0 =LÞ ð1 þ x=LÞSo S ðxÞdx
Γ ¼ So H s ðx x0 Þ 2Γ L
ðL
dx x0 ε L 00 0
þ ð1 þ x0 =LÞ ð1 x=LÞSo S ðxÞdx
G ¼ Ax þ B x > x0 2Γ x0
For a case in which ϕ is zero at L and L, we get

For the case of a constant source we can show
the Green’s function solutions to be:
that we retrieve the exact solution. For an inverse
So L
00 formulation, we can imagine that several experi-
Gðx; x0 Þ ¼ ð1 x0 =LÞð1 þ x=LÞ x < x0 mental values of ϕ are assumed to be available
2Γ
00 through measurements. The linear algebraic sys-
So L
Gðx; x0 Þ ¼ ð1 þ x0 =LÞð1 x=LÞ x > x0 tem is then:
2Γ
1þðxX
i þLÞ=Δx X
! L 000 L N 000
ϕ ð xi Þ ¼ ð1 xi =LÞ 1 þ x j =L S x j Δx þ ð1 þ xi =LÞ 1 x j =L S x j Δx
2Γ j¼1
2Γ j¼1þðxi þLÞ=Δx
2L
x j ¼ L þ ð j 1Þ
N1
! !
ϕ i ¼ A i, j S j
50 O.A. Ezekoye
We see that the mn matrix operator Ai,j is details of these methods. When applied to the
not square when there are a limited number example problem, TSVD regularization allows
!
us to easily invert for the source profile.
(m) of measurements ϕ i that require many more
For the case in which Γ ¼ 0:25,
evaluations (n) of the source term. To overcome
L < x < L, and S ¼ 4, we assume that
the sensitivity to the data noise and the limited
measurements are made at x ¼ (.67 L,
number of available measurements, regulariza-
0.33 L, 0.33 L, 0.67 L). We will use N ¼ 7
tion methods are used to generate smooth
grid points to evaluate the integral in the Greens
approximations to inversely formulated
function solution. This results in a grid spacing of
problems. Examples of regularization methods
Δx ¼ 0:33L. For this problem we add random
include truncated singular value decomposition !
(TSVD), Tikhonov regularization, and conjugate noise to the measured ϕ i values to simulate
gradient based approaches. More details can be measurement error.
found in Hansen [19] and Press et al. [18], on The matrix operations are then
2 3
2 3 S1 2 3
0 0:37 0:296 0:222 0:148 0:074 0 6 7 4:94
60 S
6 0:296 0:593 0:444 296 0:148 0 76 27 6 7
76 ⋮ 7 ¼ 6 7:13 7
40 0:148 0:296 0:444 0:593 56
0:296 0 4 5 7 4 7:61 5
S6
0 0:074 0:148 0:222 0:296 0:37 0 4:47
S7
The singular value decomposition of the matrix singular value to the smallest singular value,
A results in A ¼ UΣV T . The matrix Σ is a diago- called the condition number of the matrix,
nal matrix with the singular values ordered from reflects how singular the matrix is and how
largest to smallest as the diagonal elements. The much amplification of error might be associated
number of singular values provides an indication with small perturbations propagated through the
of the number of independent equations in the inversion. For this problem, the singular value
system of equations. Also the ratio of the largest decomposition is:
0 1
0 0 0 0
B C
0 10 1 B 0:317 0:5 0:632 0:5 C
0:357 0:5 0:61 0:5 1:358 0 0 0 B C
B 0:5 C
B 0:61 C B C B 0:516 0:5 0:258 C
B 0:5 0:357 0:5 C B 0 0:444 0 0 CB C
B CB C B 0:516 0 0:258 0 C
@ 0:61 0:5 0:357 0:5 A @ 0 0 0:178 0 AB C
B 0:516 0:5 0:258 0:5 C
0:357 0:5 0:61 0:5 0 0 0 0:148 BB
C
C
@ 0:317 0:5 0:632 0:5 A
0 0 0 0
The system of equations is inverted using the SVD (TSVD) pseudo-inverse. In the TSVD inver-
numerically accurate inversion (SVD pseudo- sion, the smallest singular value (0.148) is not used
inverse) and also using a so-called truncated in the inversion. The inversion is given by
www.ebook777.com
7 measurement is made using a modification of the

6 Duhamel integral. The Duhamel integral repre-
5 sentation looks very much like the Greens func-
Input Source tion formulation that was detailed. In both cases,
4
S'''
3 SVD inverse one is solving a Fredholm integral equation for

2 TSVD Inverse some underlying system forcing.
1
0
−1 0 1 References
x/L
1. Carslaw, H. S. and Jaeger, J. C. Conduction of Heat in
Fig. 2.25 Comparison of SVD inverse and TSVD Solids. Oxford University, Oxford, UK (1959).
inverse of noisy input data 2. Bergman, T. L., Lavine, A. S., Incropera, F. P., &
DeWitt, D. P. Fundamentals of heat and mass trans-
fer. Wiley. (2011).
!
3. Mills, A. F. Basic heat and mass transfer (Vol. 2, pp.
Xk !
! ui ϕ ! 745–833). Upper Saddler River, NJ: Prentice hall.
S¼ vi (1999).
i¼1
σi 4. Moran, M. J., Shapiro, H. N., Boettner, D. D., &
Bailey, M. Fundamentals of engineering thermody-
! !
In this expression the u i and v i are the column namics. Wiley. (2010).
5. Schmidt, P. S., Ezekoye, O. A., Howell, J. R., &
vectors of the U and V matrices in the SVD
Baker, D. K. Thermodynamics: an integrated learning
decomposition. The summation is over the total system. Wiley. (2006).
number of retained singular values. Figure 2.25 6. Heskestad, G., & Bill, R. G. Quantification of thermal
shows a comparison of the SVD and TSVD responsiveness of automatic sprinklers including con-
duction effects. Fire Safety Journal, 14(1), 113–125.
inversions of the noisy measurements. We see
(1988).
that there is considerably more variation and devi- 7. Kaviany, M. Principles of heat transfer New York,
ation from the exact source values for the SVD USA, John Wiley Publishers (2002).
inversion relative to the TSVD inversion. The 8. Kaviany, M. Heat transfer physics Cambridge, UK:
Cambridge University Press. (2008).
process of truncating the representation of the A
9. Gebhart, Benjamin. Heat conduction and mass diffu-
matrix acts to smooth or regularize the solution. sion. New York: McGraw-Hill, (1993).
This problem is only one example of the many 10. Wald, F., Simões da Silva, L., Moore, D. B., Lennon,
types of techniques that can be applied in inverse T., Chladná, M., Santiago, A., and Borges,
L. Experimental behaviour of a steel structure under
heat transfer analysis. Many other types of inver-
natural fire. Fire Safety Journal, 41(7), 509–522.
sion processes might be used for fire problems. (2006).
There are many examples of time dependent 11. Bejan, A, Heat Transfer, New York, John Wiley
inversion, as might be done to determine a sur- Publishers, (1993).
12. Arpaci, V.S., Conduction Heat Transfer, Addison
face heat flux when measuring internal
Wesley, (1966).
temperatures that present many different solution 13. Ozisik, M.N., Heat Conduction, New York, John
approaches for inversion [21, 22]. An interesting Wiley Publishers, (1993).
analogy exists between the steady example using 14. Olenick, Stephen M., and Douglas J. Carpenter. “An
updated international survey of computer models for
TSVD based inversion and some types of tempo-
fire and smoke.” Journal of Fire Protection Engineer-
rally evolving inverse problems. For linear time ing 13.2 87–110. (2003).
dependent problems, analytical solution forms 15. Shih, T.M., Numerical Heat Transfer, Hemisphere
exist for time dependent heating using the publishing Corp, (1984).
16. Ferziger, J.H. Numerical Methods for Engineering
Duhamel integral formulation. A simple example
Applications, New York, John Wiley, (1981).
from Beck et al. [21] shows the process for deter- 17. Fletcher, C.A.J., Computational Galerkin Methods,
mining the heat flux when a single temperature Springer (1984).
52 O.A. Ezekoye
18. Press, William H., Brian P. Flannery, Saul 21. Beck, J. V., St Clair, C. R., & Blackwell, B. Inverse
A. Teukolsky, and William T. Vetterling. Numerical heat conduction, John Wiley, (1985).
Recipes in FORTRAN 77: Volume 1, Volume 1 of 22. Ozisik, M. N. Inverse heat transfer: fundamentals and
Fortran Numerical Recipes: The Art of Scientific Com- applications, Taylor & Francis, (2000).
puting. Vol. 1. Cambridge university press, (1992).
19. Hansen, Per Christian. Rank-deficient and discrete
ill-posed problems: numerical aspects of linear inver- Ofodike A. Ezekoye is a professor in the Department of
sion. Vol. 4. Society for Industrial Mathematics, (1987). Mechanical Engineering at the University of Texas at
20. Mathews, J., and Walker, R.L., Mathematical Austin.
Methods of Physics, Addison-Wesley, (1970).
www.ebook777.com
Convection Heat Transfer

3
Arvind Atreya
This chapter is respectfully dedicated to the author’s father, Dr. Dharam Dev Atreya.
Introduction heat transfer. A further subdivision based on

whether the flow occurs inside (e.g., in a pipe)
There are only two fundamental physical modes or outside the body under consideration is also
of energy transfer, conduction and radiation. In often made. For application of convective heat
conduction, energy slowly diffuses through a transfer to fire science, natural convection around
medium from a point of higher temperature to objects is clearly far more important than forced
a point of lower temperature, whereas in radia- convection inside a pipe. Thus greater attention
tion, energy is transmitted with the speed of light is devoted here to natural convective heat trans-
by electromagnetic waves (or photons), and a fer and external flows.
transmitting medium is not required. Thus The objective of this chapter is to provide a
from a conceptual viewpoint, convection is not firm understanding of the physical mechanisms
a basic mode of heat transfer. Instead, it occurs that underlie convective heat transfer, as well as
by a combined effect of conduction (and/or to develop the means to perform convection heat
radiation) and the motion of the transmitting transfer calculations. In the first section of the
medium. chapter, basic concepts and relations are devel-
Nevertheless, convection plays a very impor- oped; calculation methods are illustrated with the
tant role in fires. It transports the enormous help of examples in the second section. Tables of
amount of chemical energy released during a empirical and theoretical results (including their
fire to the surrounding environment by the range of applicability) are also provided for
motion of hot gases. This motion may be induced quick reference.
naturally by the fire itself (hot gases rise and cold
air rushes to replace them) or by a source external
to the fire, such as a prevailing wind. Based on Concepts and Basic Relations
this distinction, the subject of convective heat
transfer is usually subdivided into natural (free) A simple, everyday problem of drying a wet
and forced convection. Obviously, both natural body in a stream of warm dry air is shown in
and forced convection may occur simulta- Fig. 3.1. From our knowledge of fluid mechanics,
neously, resulting in a mixed mode of convective we expect the flow of air to slow down next to the
surface of the wet body, thus transferring some of
its momentum to the body. Conversely, the body
will experience a drag force if it moves through
stationary air. In addition to this exchange of
A. Atreya (*) momentum, the body also losses some of its
University of Michigan moisture; that is, transfer of mass takes place

54 A. Atreya
from the wet surface to the warm air. Further- Basic Laws of Molecular Transfer
more, for moisture to evaporate at the wet surface
the necessary heat must also be transferred from Newton’s Law of Viscosity An isothermal sys-
the warm air to the wet body. Hence, the body tem is shown in Fig. 3.2. It consists of a fluid
experiences a simultaneous transfer of momen- trapped between two impervious flat plates
tum, mass, and energy. that are infinite in extent and separated by a
To obtain a quantitative description of the distance, δ. Experiments show that if the lower
above process, note first that all the quantities plate is fixed and the upper plate is moved at a
(mass, momentum, and energy) being transported constant velocity, u1, then the velocity of the
in the example are conserved, so conservation fluid between the two plates varies from u1
laws govern their rate of transfer. near the top plate to zero at the fixed plate.
These conservation laws need to be Under steady laminar conditions, a linear veloc-
supplemented by basic or constitutive relations ity profile is established as shown in Fig. 3.2.
that relate the rate of transfer to the driving forces The fluid thus exerts a shear force on the
and fluid properties. These basic laws are: stationary plate. Experiments also show that τ,
(1) Newton’s law of viscosity, which relates the the shear force exerted on the bottom plate per
rate of change of momentum to velocity unit area, is directly proportional to u1 and
gradients; (2) Fourier’s law of heat conduction, inversely proportional to the separation distance,
which relates the rate of heat transfer to tempera- δ; that is,
ture gradients; and (3) Fick’s law of mass diffu-
sion, which relates the rate of mass transfer to u1 u1
τα or τ ¼ μ
concentration gradients. With this framework of δ δ
conservation and basic laws a majority of lami-
nar convective heat transfer problems can be The constant of proportionality, μ, is called the
analyzed, at least in principle. For the turbulent dynamic viscosity of the fluid (units: Ns/m2)
case this framework provides guidance for devel- and the force per unit area, τ, is called the shear
oping useful empirical correlations. stress. (In some texts a negative sign is
This section covers the basic laws in the con- introduced to emphasize the direction of net
text of laminar flows and their relationships to momentum transfer, i.e., from the fluid at higher
the more familiar heat transfer coefficients. velocity to the fluid at lower velocity.) Note that
Later, we see how these heat transfer coefficients the shear stress τ exerted by the fluid on the fixed
can be determined by the application of conser- plate is in the positive x-direction and that
vation laws. The effect of turbulence is also exerted by the fixed plate on the fluid is equal
discussed and empirical correlations presented. and opposite in direction.) In differential form
Moving plate u∞
Warm air Direction

of momentum
Surface of transfer δ
a wet body
y
τ
x
Fixed plate
τ
Fig. 3.2 Steady-state velocity distribution in a Newto-

Fig. 3.1 Drying of a wet body in a stream of warm air nian fluid
www.ebook777.com
3 Convection Heat Transfer 55
00
this relationship expresses the shear stress at any time through the fluid ( q_ J/m2s) is found to be
location, y, in the fluid as directly proportional to the temperature differ-
ence, T1 Ts, and inversely proportional to
du the separation distance, δt; that is,
τ¼μ ð3:1Þ
dy
This is Newton’s law of viscosity. Equation 3.1 00 T1 Ts T1 Ts
q_ α ¼ k
states that the shear stress experienced by a fluid δt δt
layer is directly proportional to the velocity gra- The constant of proportionality, k, is called
dient inside the fluid at that location. the thermal conductivity (units: J/mKs). The
Fluids that behave according to Equation 3.1 minus sign is a consequence of the second law
(i.e., τ is linearly related to the velocity gradient, of thermodynamics, which requires the heat to
du/dy, and μ is not a function of the velocity flow in the direction of decreasing temperature.
gradient) are called Newtonian fluids. Fortu- In differential form the heat across any fluid layer
nately, all gases and most simple liquids such as is given by
water obey this simple law. For gases, μ roughly
increases as the square root of temperature as 00 dT
q_ ¼ k ð3:2Þ
predicted by the kinetic theory of dilute gases. dy
Liquids, on the other hand, become “thinner”
(less viscous), that is, μ decreases with increase This is known as Fourier’s law of heat conduc-
in temperature. For non-Newtonian fluids (e.g., tion, which states that the heat flux is directly
pastes, slurries, blood, etc.) the dynamic viscos- proportional to the temperature gradient and the
ity also depends on the velocity gradient or the heat flux vector is oriented in the direction of
rate of shear. decreasing temperature. The thermal conductiv-
ity, k, like viscosity, μ, is a physical property of
Fourier’s Law of Heat Conduction Two sta- the fluid. The thermal conductivity of gases at
tionary parallel plates separated by a distance, δt, low densities increases with increasing tempera-
pffiffiffi
are shown in Fig. 3.3. Let the temperature of the ture (roughly as T according to the kinetic
upper plate be T1 and that of the lower plate be theory of dilute gases) whereas the thermal con-
Ts. Under steady conditions and for temperature ductivity of most liquids decreases with increas-
independent properties of the trapped fluid, a ing temperature.
linear temperature distribution as shown in
Fig. 3.3 is obtained. Thus, as expected, heat is Fick’s Law of Mass Diffusion Once again,
transferred by the stationary fluid from the hot to consider the parallel plate example. This time,
the cold plate. The heat flow per unit area per unit there are no temperature gradients and no
directed motion of the plates. Instead, the top
plate is maintained at a higher concentration of
Hot plate T∞ species A (CA1 , kg of species A/m3), assuming
that it is wet, and the bottom plate is maintained
at a fixed but lower concentration of species
Direction A (CAW, kg of A/m3). Then, under steady
of heat conditions a concentration profile, as shown in
transfer δt
Fig. 3.4, is established. This nonuniform concen-
y tration field is the driving force for species A to
diffuse from the top to the bottom plate. The
x mass flux of species A, ṁ00 (units: kg/m2s), leav-
Cold plate Ts ing the top plate and arriving at the bottom plate
through the fluid, B, is found to be directly pro-
Fig. 3.3 Steady-state temperature distribution portional to the concentration difference and
56 A. Atreya
CA∞ predicted by the kinetic theory of gases. In fact,

the physical mechanism of all three transport
processes is easily understood by considering
the random motion of molecules in an ideal gas.
Direction
of mass Molecules of a gas, even in the absence of
transfer δd bulk fluid motion, move around randomly at
high speeds and bump into each other. Thus, a
y
given molecule may be found anywhere between
the two parallel plates. The problem, however, is
x
how to distinguish one molecule from another.
CAW When the upper plate is moving relative to the
bottom plate, the molecules adjacent to the upper
Fig. 3.4 Steady-state concentration distribution
plate attain a directed velocity over and above
their random motion. Consequently, as the
inversely proportional to the separation distance molecules near the upper plate find themselves
δd, that is, in lower fluid layers (and vice versa) they
exchange directed motion (or momentum) by
00 CA1 CAW CA1 CAW
m_ A α ¼ DAB bumping into each other. Similarly, the gas
δd δd molecules near the hot upper plate are distin-
The constant of proportionality, DAB, is called guished from those adjacent to the cold lower
diffusivity of species A through species B and has plate because they possess a higher kinetic
units of m2/s. The negative sign once again energy. Once again, by virtue of random motion,
indicates that the net mass transfer of species these higher kinetic energy molecules find them-
A occurs in the direction of decreasing concen- selves near the cold plate and collide with low
tration. In the differential form, the mass flux of kinetic energy molecules (or vice versa), thus
species A across any layer of fluid B is given by transporting energy. In the case of mass diffu-
sion, the molecules are chemically labeled and
00 dCA their random motion results in mass transfer.
m_ ¼ DAB ð3:3Þ
dy Since increasing the gas temperature increases
This is known as Fick’s law of mass diffusion, the random molecular motion, the transport pro-
which states that the mass flux across a fluid layer cesses become more efficient at higher gas
is directly proportional to the local concentration temperatures. For gases, the macroscopic physi-
gradient. For binary gas mixtures, DAB increases cal properties (viscosity, thermal conductivity,
with increasing temperature roughly as three- and diffusivity) that characterize momentum,
halves power, as predicted by the kinetic theory heat, and mass transport must also increase with
of dilute gases. temperature.
Assuming constant properties, Equations 3.1
Discussion Equations 3.1, 3.2, and 3.3 look very and 3.2 can be rewritten in the following forms:
similar—they all relate the flux of the transported
quantity to their respective local gradients. Actu- μ d ðρuÞ
τ¼ ð3:1aÞ
ally, these equations may be considered ρ dy
definitions of the three macroscopic physical
properties, μ, k, and DAB of the fluid. In general, 00 k d ρc p T
q_ ¼ ð3:2aÞ
these properties are functions of temperature, ρc p dy
pressure, and composition. As noted earlier, for Here, μ/ρ is known as the kinematic viscosity, ν,
low pressure binary gas mixtures (<10 atm) the and has units of m2/s. The product, ρu, has units
pressure, temperature, and composition depen- of (kg m/s)/m3; that is, momentum per unit vol-
dence of these properties can be approximately ume. The quantity, k/ρcp, is known as the thermal
www.ebook777.com
diffusivity, α, and it too has units of m2/s. The If the surface conditions are such that Ts 6¼ T1,
product, ρcpT, then becomes enthalpy per unit we know that convection heat transfer will occur.
volume and has units of J/m3. Comparing these The convection heat transfer problem is to
00
with Equation 3.3, where the mass diffusivity, relate the local heat flux, q_ , to its driving force,
DAB, also has units of m2/s and CA is expressed Ts T1. By expressing the heat flux as
in kg of species A/m3, we find that the fluxes are
00
related to their corresponding gradients of volu- q_ ¼ hðT s T 1 Þ ð3:4Þ
metric concentration. Furthermore, the ratios of
various physical constants yield the familiar non- the problem is reduced to determining h, which is
dimensional numbers. These are called the heat transfer coefficient. From Equa-
tion 3.2 it is clear that the heat transfer coefficient
Prandtl number, Pr ¼ ν/α may be expressed as
Schmidt number, Sc ¼ ν/DAB
Lewis number, Le ¼ α/DAB
kð∂T=∂yÞ y¼0 J
h¼ ð3:5Þ
Here, the Prandtl number compares the rela- ðTs T1Þ m2 s K
tive magnitude of momentum transfer to heat
transfer, the Schmidt number compares momen- In Equation 3.5 the partial derivative is used
tum transfer to mass transfer, and the Lewis because, in general, temperature is a function of
number compares heat transfer to mass transfer. x, y, z, and time. Thus, if the thermal conductivity
The significance of these nondimensional num- of the fluid is known, the problem of determining
bers will become obvious when we discuss the local heat transfer coefficient is reduced to
boundary layer transfer processes. The balance that of determining the local temperature gradi-
of this section will discuss primarily heat trans- ent in the fluid adjacent to the surface. This local
fer. The treatment of mass transfer is similar and temperature gradient can be experimentally
will not be discussed. However, since fluid measured or obtained theoretically from the solu-
motion is central to convective heat transfer, it tion of the conservation laws. Obviously, the
is necessary to understand momentum transport temperature gradient will vary from point to
to solve convective heat transfer problems. point along the surface of the body. Often, such
detail is not required and it may only be neces-
sary to determine an average heat transfer coeffi-
Relationship of Basic Laws to Transfer cient, h. This is obtained by integrating over the
Coefficients entire surface area, As. The total rate of heat
transfer from the body to the fluid is given by
A flow condition is shown in Fig. 3.5. A fluid of ð
00
velocity, u1, and temperature, T1, flows over an q_ ðJ=sÞ ¼ q_ J=m2 s dAs m2 ð3:6Þ
arbitrarily shaped stationary surface of area, As. As
Defining an average heat transfer coefficient, h, as

q″
As Ts
q_ ¼ hAs ðT s T 1 Þ ð3:7Þ
y τs The following is obtained from the use of

x Equations 3.4, 3.6, and 3.7
u∞
T∞ dAs ð
1
h¼ hðT s T 1 ÞdAs ð3:8Þ
As ð T s T 1 Þ A s
Fig. 3.5 Example of the convection heat transfer

If the surface temperature, Ts, is held constant,
problem then
58 A. Atreya
ð
1 J gradients in the fluid adjacent to the surface is
h¼ hdAs ð3:9Þ required. To obtain these gradients it is necessary
As As m2 s K
to be more specific about the surface geometry
Similarly, the local shear stress at the surface, and the flow conditions. The governing conser-
τs, can be related to its cause, the fluid velocity, vation equations for an arbitrary surface geome-
u1. This relation is derived by defining a try will be first derived and then applied to a flat
local nondimensional friction coefficient, Cf, plate to illustrate the methodology.
according to the equation

1 2 Conservation Equations for Convection
τs C f ρu ð3:10Þ
2 1 Heat Transfer
Once again, the problem is reduced to the deter- It has been shown that to determine the heat
mination of Cf. Using Equation 3.1 we obtain transfer and friction coefficients, the temperature
and velocity distributions in the flow are needed.
μð∂u=∂yÞ y¼0 In principle, these may be obtained from the
Cf ¼ ð3:11Þ
ð1=2Þρu21 solution of the conservation equations with
appropriate boundary conditions. Although in
Thus, the local friction coefficient can be practice it is difficult, the very knowledge of
evaluated from the knowledge of the local veloc- conservation equations and their solutions for
ity gradient in the fluid adjacent to the surface. simple cases (such as a flat plate) provide
The average friction coefficient can easily be considerable insight about the parameters
obtained by integrating the local shear stress, τs, influencing the heat transfer and friction
over the entire surface area, As. The total drag coefficients. Thus, the necessary equations will
force, D, experienced by the body is given by the be first developed and then applied to a flat plate.
product of average shear stress, τs , and the sur- Consider the flow over the surface shown in
face area, As. In other words, Fig. 3.6. To simplify the development assume
ð two-dimensional flow conditions, for which x is
D τ s As ¼ τs dAs ð3:12Þ the direction along the surface and y is normal to
As the surface. Extension of this to three-dimensional
flows is available in the literature [1–5].
Assuming u1 to be the same at all locations
ð
1=2ρu21
τs ¼ C f As ð3:13Þ
As As dy
Defining the average friction coefficient, C f , as dx

q″ u∞
τs T∞
Cf ð3:14Þ
1=2ρu21 As Ts
δ
we get y x
u∞
ð T∞ Ts
1
Cf ¼ C f dAs ð3:15Þ
As As
To obtain h, h, Cf, and C f a knowledge of Fig. 3.6 Velocity and temperature distributions inside
fluid properties and temperature and velocity the fluid for flow over a hot surface
www.ebook777.com
Conservation of Mass (Continuity Equation) Similarly,

The first conservation law that is pertinent to the

problem is that matter is neither created nor ∂ðρvÞ
ρvðx, y þ dyÞ ¼ ρvðx; yÞ þ dy
destroyed. When applied to the differential ∂y x, y
control volume shown in Figs. 3.6 and 3.7 it states
that the net rate of mass flow entering the elemen- Thus, mass leaving the right face per unit time,
tal control volume in the x-direction, plus the net M_ xþdx , is given by
rate of mass flow entering the elemental control

volume in the y-direction equals the net rate of
_ ∂ð puÞ
M xþdx ¼ ρu þ dx dy 1 ð3:17Þ
increase of mass stored in the control volume. ∂x
Mass enters and leaves the control volume exclu-
sively through gross fluid motions. Such a trans- and mass leaving the top face per unit time,
port is often referred to as convective transport. M_ yþdy , is
For a control volume of unit depth in the z-

direction, mass entering the left face per unit ∂ðpvÞ
M_ yþdy ¼ ρv þ dy dx 1 ð3:18Þ
time, Ṁx, is given by ∂y

M_ x ¼ ρ kg=m3 uðm= sec Þ dyðmÞ 1ðmÞ Finally, the rate of increase (or decrease) of
mass stored in the control volume, Ṁs, is of the
Similarly, mass entering the bottom face per unit form
time, Ṁy, is given by
∂ ∂p
M_ s ¼ ðρ dx dy 1Þ ¼ dxdy ð3:19Þ
M_ y ¼ ρ v dx 1 ∂t ∂t
where v is the fluid velocity in the y-direction. Thus, conservation of mass requirement may
Since ρu and ρv are continuous functions of x, now be expressed as
y, and time, in general they will be different at
different locations. To determine ρu at (x + dx), ∂ðρuÞ
ðρuÞdy þ ðρvÞdx ρu þ dx dy
expand ρu about the point x in the Taylor ∂x

series as ∂ðρvÞ ∂p
ρv þ dy dx ¼ dxdy
∂y ∂t

∂ðρuÞ ð3:20Þ
ρuðx þ dx, yÞ ¼ ρuðx; yÞ þ dx
∂x x, y
After canceling terms and dividing by dx dy
ð3:16Þ
∂ρ ∂ðρuÞ ∂ðρvÞ
þ þ ¼0 ð3:21Þ
∂t ∂x ∂y
∂
ρv + ⎯ (ρv) dy
∂y
This is the continuity equation, which is an
y expression of the overall mass conservation
requirement and must be satisfied at every point
∂ in the flow. This equation applies for a single
x ρu dy ρu + ⎯ (ρu) dx
∂x
species fluid, as well as for mixtures in which
dx
species diffusion and chemical reactions may be
(x,y) occurring.
ρv
Fig. 3.7 Differential control volume for mass conserva- Conservation of Momentum The second con-
tion in a two-dimensional flow servation law pertinent to the convection heat
60 A. Atreya
transfer problem is Newton’s second law of face, corresponds to the viscous shear force per
motion. For a differential control volume in a unit area in the negative y-direction on a face
flow field, this requirement states that the sum whose normal is in the negative x-direction—
of all forces acting on the control volume must resulting in a positive shear stress. All the vis-
equal the rate of increase of the fluid momentum cous stresses shown in Fig. 3.8 are positive
within the control volume, plus the net rate at according to the adopted convention. It should
which momentum leaves the control volume be noted that these forces act on the fluid inside
(outflow-inflow). the control volume and are caused by its interac-
The forces acting on the fluid may be tion with the surrounding fluid. Thus, these vis-
categorized into body forces that are proportional cous stresses will vanish if the fluid velocity, or
to the volume, and surface forces, which are more specifically the velocity gradient in the
proportional to the area. Gravitational, centrifu- fluid, becomes zero. The normal viscous stresses
gal, magnetic, and electric fields are familiar shown in Fig. 3.8 must not be confused with
examples of body forces. Of these, gravitational static pressure, p, which does not vanish for
body force is the most important from the fire zero velocity. Since these stresses are continuous
science point of view. The x- and y-components functions of x, y, and time, the customary
of this body force per unit volume of the fluid Taylor’s expansion is used to express the stresses
will be designated as FBx and FBy, respectively. on the top and right faces of the control volume
The surface forces, Fs, acting on the fluid are shown in Fig. 3.8. Thus, the net surface force in
called stresses (force/area). These are due to fluid the x- and y-directions may be expressed as
static pressure, p, and viscous stresses. Since

pressure is always normal to the surface, the ∂σ xx ∂ p ∂τ yx
Fsx ¼ þ dxdy ð3:22Þ
viscous stresses are also resolved into normal ∂x ∂x ∂y
stresses, σii, which act normal to the surface,
and shear stresses, τij, which act along or parallel
∂σ yy ∂ p ∂τxy
to the surface. Figure 3.8 shows the various vis- Fsy ¼ þ dxdy ð3:23Þ
∂y ∂y ∂x
cous stresses acting on the surface of a differen-
tial control volume. A double subscript notation To use Newton’s second law, the time rate of
is used to specify the stress components. The first change of momentum and the momentum influx
subscript indicates the direction of the outward and outflux must also be evaluated. To focus on
normal to the surface, and the second subscript the x-direction, the relevant momentum fluxes
indicates the direction of the force component. are shown in Fig. 3.9. The mass flux through
Accordingly, the stress τxy, acting on the left the left face is ρu and hence the corresponding
Fig. 3.8 Static pressure, p, ∂

and normal and shear σyy (σyy) dy
∂y
viscous stresses acting on a ∂p
differential control volume p dy
τyx ∂ (τ ) dy
∂y yx
in a two-dimensional flow ∂y
y ∂ (τ ) dx
p τxy xy
∂x
∂
x σxx dy σxx (σxx) dx
∂x
τxy ∂p
z dx p dx
∂x
(x,y ) τ
yx
σyy p
www.ebook777.com
Fig. 3.9 Influx and outflux ∂

of x-momentum in the (ρv)u + ⎯ [(ρv)u] dy
y,v ∂y
control volume
x,u dy
∂
(ρu)u (ρu)u + ⎯ [(ρu)u] dx
∂x
z dx
x,y
(ρv)u

momentum flux is (ρu)u. Similarly, the ∂v ∂v ∂v
ρ þu þv
x-momentum flux due to mass flow in the ∂t ∂x ∂y
y-direction through the bottom face is (ρv)u.
∂ ∂τxy
Thus the net rate at which momentum leaves ¼ σ yy p þ þ FBy ð3:26Þ
∂y ∂x
the control volume is given by (x-momentum
outflow inflow)
In Equations 3.25 and 3.26, the first term on
the left side represents the increase in momentum
∂ ∂½ðρvÞu
½ðρuÞudx dy þ dy dx of the fluid inside the control volume, and the
∂x ∂y remaining terms represent the net rate of momen-
tum efflux from the control volume. The terms on
In addition, the time rate of change of x-momentum the right side of the equations account for the
of the fluid within the control volume is net viscous, pressure, and body forces acting
given by on the control volume. These equations must be
satisfied at every point in the fluid. A solution
∂ of Equations 3.21, 3.25, and 3.26 along with
ðρuÞdx dy
∂t appropriate boundary conditions yields the
velocity field needed to determine the friction
Equating the total rate of change in the coefficient.
x-direction to the sum of forces in the x-direction, Before a solution to the above equations
we obtain can be obtained, it is necessary to relate the
viscous stresses to the velocity gradients. For a
∂ðρuÞ ∂½ðρuÞu ∂½ðρvÞu one-dimensional flow of a Newtonian fluid,
þ þ
∂t ∂x ∂y Equation 3.1 relates the shear stress to the veloc-
∂σ xx ∂ p ∂τ yx ity gradient in the fluid. For a two-dimensional
¼ þ þ FBx ð3:24Þ flow of Newtonian fluid [2], the required stress-
∂x ∂x ∂y
velocity gradient expressions are
By using Equation 3.21, Equation 3.24 may
be expressed in a more convenient form as ∂u 2 ∂u ∂v
σ xx ¼ 2μ μ þ ð3:27Þ
∂x 3 ∂x ∂y
∂u ∂u ∂u
ρ þu þv
∂t ∂x ∂y ∂v 2 ∂u ∂v
σ yy ¼ 2μ μ þ ð3:28Þ
∂ ∂τ yx ∂y 3 ∂x ∂y
¼ ðσ xx pÞ þ þ FBx ð3:25Þ
∂x ∂y

∂u ∂v
A similar expression is obtained for the τxy ¼ τ yx ¼μ þ ð3:29Þ
y-direction. This is ∂y ∂x
62 A. Atreya
On substituting Equations 3.27, 3.28, and 3.29 the x- and y-momentum equations is obtained.
into Equations 3.25 and 3.26, the desired form of These are
x-momentum equation
2
∂u ∂u ∂u ∂p ∂ u ∂2 u 1 ∂ ∂u ∂v
ρ þu þv ¼ þμ þ þ μ þ þ FBx ð3:30Þ
∂t ∂x ∂y ∂x ∂x2 ∂y2 3 ∂x ∂x ∂y
y-momentum equation
2
∂v ∂v ∂v ∂p ∂ v ∂2 v 1 ∂ ∂u ∂v
ρ þu þv ¼ þμ þ þ μ þ þ FBy ð3:31Þ
∂t ∂x ∂y ∂x ∂x2 ∂y2 3 ∂y ∂x ∂y
For an isothermal system, Equations 3.21, 3.30, 3. Energy generated per unit volume per unit
and 3.31 along with the equation of state ( p ¼ time inside the control volume (Q watts/m3).
ρRT for an ideal gas) provide a complete set for This may be due to chemical reactions (endo-
determining the four dependent variables (u, v, p, thermic or exothermic) or may be caused by
and ρ) as a function of the three independent the radiative loss of heat. Although the spe-
variables x, y, and t. However, for a cific form of Q will depend on the nature of
nonisothermal system such as a fire, energy bal- the physical process, here we will treat it only
ance must also be considered. as a rate of heat loss or gain per unit volume.
4. The rate of work done by surface or body
Conservation of Energy The temperature field forces. For surface forces, FS, it is given by:
inside the fluid, T(x, y, t), needed to determine [FS (velocity vector)] (surface area); and for
the heat transfer coefficient is obtained by apply- the body forces, FB, it is given by: [FB
ing the first law of thermodynamics to the (velocity vector)] (volume). Both these
differential control volume shown in Fig. 3.6. expressions have units of watts or work done
Before writing the energy balance for this per unit time.
control volume, it is necessary to identify the With these definitions, consider the control
items that must be included in the energy budget. volume shown in Figs. 3.6 and 3.10. The conser-
These are: vation of energy for this control volume can be
1. The stored energy. This includes the specific simply stated as
internal or thermal energy, e; J/kg, and the
kinetic energy of the fluid per unit mass, V2/2
¼ (u2 + v2)/2. Potential energy is neglected
4 9
because for most problems in convective dy
heat transfer it is substantially smaller than y
thermal and kinetic energy. Hence, the total
energy content per unit volume is given by: 1 3
ρ(e + V2/2). x 5 10
2. Conduction of thermal energy across the 6 8
surfaces of the control volume. Here, the rate z
of energy transported per unit area per
unit time across the control surface is given x,y dx
00
by Equation 3.2 as q_ x ¼ k∂T=∂x for the 2 7
00
x-direction, and q_ y ¼ k∂T=∂y for the Fig. 3.10 A control volume showing the energy
y-direction. conducted and convected through its control surfaces
www.ebook777.com
0 1 0 1 0 1
Rate of increase Net rate of energy flow Flow of heat through
B C B C B C
@ of energy inside A ¼ @ into the control volume A þ @ the control surface A
the control volume by bulk and fluid motion by conduction ð3:32Þ
! !
Rate of work done by body Energy generated inside
þ þ
and surface forces the control volume
In Equation 3.32 the rate of increase of energy obtained by subtracting the energy going out
inside the control volume is given by from that coming in, to yield

∂ V2 ∂ V2 ∂ V2
ρ eþ dxdy ρu e þ dxdy ρv e þ dxdy
∂t 2 ∂x 2 ∂y 2
The net rate at which the energy enters the con- Similarly, the heat flowing into the control
trol volume by convection or bulk fluid motion is volume by conduction is given by
00 00 !
∂q_ x ∂q_ y ∂ ∂T ∂ ∂T
þ dxdy ¼ k þ k dxdy
∂x ∂y ∂x ∂x ∂y ∂y
Finally, the net rate at which work is done on

the fluid inside the control volume (Fig. 3.11) is
given by the expression

∂ð puÞ ∂ðpvÞ ∂ ∂ ∂ ∂
uFBx þ vFBy dxdy þ dxdy þ ðuσ xx Þ þ vσ yy þ uτ yx þ vτxy dxdy
∂x ∂y ∂x ∂y ∂y ∂x
On substituting these expressions into Equa-

tion 3.32, simplifying and using Equations 3.27
through 3.31, we obtain

∂T ∂T ∂T ∂ ∂T ∂ ∂T ∂p ∂p ∂p
ρC p þ ρuC p þ ρvC p ¼ k þ k þ þu þv þ μΦ
∂t ∂x ∂y ∂x ∂x ∂y ∂y ∂t ∂x ∂y
ð3:33Þ
" 2 2 # 2
In Equation 3.33, the thermodynamic definition ∂u∂v ∂u ∂v
μΦ ¼ 2μ þ þ þ
of enthalpy (i ¼ e + p/ρ and di ¼ CpdT ) has ∂x∂y ∂y ∂x
been used. Also, the term μΦ is called the viscous 2
2 ∂u ∂v
dissipation and is given by þ
3 ∂x ∂y
ð3:34Þ
64 A. Atreya
9 7 (i.e., FBx ¼ ρgx and FBy ¼ ρgy),

Equations 3.21, 3.30, 3.31, and 3.33 become
8 Continuity
y
10
6 ∂u ∂v
dy þ ¼ 0 ð3:35Þ
x 5 11 ∂x ∂y
12
4
z dx x-momentum
x,y 1 2
∂u ∂u 1∂p ∂ u ∂2 u
u þv ¼ þv þ gx
2 3 ∂x ∂y ρ ∂x ∂x2 ∂y2
ð3:36Þ
Fig. 3.11 Control volume showing the rate of work done
by various surface forces. All units are in watts
y-momentum
2
∂v ∂v 1∂p ∂ v ∂2 v
Equations 3.21, 3.30, 3.31, and 3.33, along u þv ¼ þv þ gy
with the equation of state ( p ¼ ρRT, for an ∂x ∂y ρ ∂y ∂x2 ∂y2
ideal gas), provide a complete set for determin- ð3:37Þ
ing the temperature and velocity field [T(x, y, t), u
(x, y, t), v(x, y, t)] inside the fluid. However, it is Energy equation
not possible to solve the above set of coupled 2
nonlinear partial differential equations. There- ∂T ∂T ∂ T ∂2 T
u þv ¼α þ
fore, several simplifying approximations are ∂x ∂y ∂x2 ∂y2

made. These are discussed below. 1 ∂p ∂p
þ u þv ð3:38Þ
ρC p ∂x ∂y
Simplifications
1. Low velocity. For most problems encountered Often, the energy equation is further simplified
in convective heat transfer, the flow velocity by assuming that the terms (u ∂p/∂x) and (v ∂p/
is low enough (Mach number < 1/3) to ignore ∂y) are negligible. This assumption is justified
the contribution of viscous work in the energy since most processes of interest are nearly iso-
equation. This allows the term μϕ in Equa- baric. Thus the energy equation becomes
tion 3.33 to be dropped.
2
2. Incompressible flow. Fluid density is assumed ∂T ∂T ∂ T ∂2 T
to be constant except in the buoyancy terms u þv ¼α þ ð3:39Þ
∂x ∂y ∂x2 ∂y2
(FBx, FBy) of Equations 3.30 and 3.31. This is
called the Boussinesq approximation and will Equations 3.35, 3.36, 3.37, and 3.39, along
be discussed later in greater detail. with the equation of state ( p ¼ ρRT, for an
3. Steady flow. This approximation allows all the ideal gas), provide a complete set for determin-
time derivative terms in the above equations ing u(x, y), v(x, y), T(x, y), ρ(x, y), and p(x, y).
to be dropped. Once these dependent variables are known, the
4. Constant properties. Specific heat, thermal desired heat transfer coefficient and friction
conductivity, and viscosity are all assumed factor are obtained from Equations 3.5 and
to be constant; that is, independent of temper- 3.11, respectively. However, the above equations
ature and pressure. are still too difficult to solve and a further simpli-
With these simplifications and assuming fication, known as the boundary layer approxi-
that the body force is only due to gravity mation, is often made.
www.ebook777.com
The Boundary Layer Concept viscosity and decreases with increasing free-
stream velocity. By defining the Reynolds num-
In 1904, Prandtl proposed that all the viscous ber, Re, as Re ¼ u1 L/ν, where L is the charac-
effects are concentrated in a thin layer near the teristic length of the plate, the boundary layer
boundary and that outside this layer the fluid thickness decreases with increasing Re. For
behaves as though it is inviscid. Thus, the flow most flows of practical interest, the Reynolds
over a body, such as the one shown in Fig. 3.6, number is large enough such that δ is much less
can be divided into two zones: (1) a thin viscous than the characteristic length, L (δ L ).
layer near the surface, called the boundary layer, Just as a velocity boundary layer develops
and (2) inviscid external flow, which can be because of viscous effects near the surface, a
closely approximated by the potential flow the- thermal boundary layer develops due to heat
ory. As will be seen later, the fact that the bound- transfer between the free stream and the surface
ary layer is thin compared to the characteristic if their temperatures are different. The fluid
dimensions of the object is exploited to simplify particles that come into contact with the plate
the governing equations and obtain a useful solu- surface achieve thermal equilibrium at the plate’s
tion. This boundary layer approximation plays an surface temperature. In turn, these particles
important role in convective heat transfer, since exchange energy with those in the adjoining
the gradients of velocity and temperature at the fluid layer, and temperature gradients develop
surface of the body are required to determine the in the fluid. As shown in Fig. 3.12, the region of
heat transfer coefficient and the friction factor. the fluid in which these temperature gradients
To illustrate these ideas, consider fluid flow exist is the thermal boundary layer, and its thick-
over a flat plate as shown in Fig. 3.12. The fluid ness, δt, is defined as the value of y for which the
particles in contact with the plate surface must ratio [(T – TS)/(T1 TS)] ¼ 0.99. The thermal
assume zero velocity because of no slip at the boundary layer thickness increases with the ther-
wall, whereas the fluid particles far away from mal diffusivity, α, of the fluid and decreases with
the wall continue to move at the free stream increasing free stream velocity. In other words,
velocity, u1. The transition of fluid velocity δt is inversely proportional to the product of
from zero to u1 takes place in a small distance, the Reynolds number and Prandtl number
δ, which is known as the boundary layer thick- (Re Pr ¼ (u1L/ν)(ν/α) ¼ u1L/α). For air,
ness and is defined as the value of y for which Pr 0.7 and the Reynolds number is sufficiently
u ¼ 0.99u1. As is intuitively obvious, the thick- large for flows of practical interest, consequently
ness of the boundary layer increases with fluid δt L.
T∞
u∞
δt
y T
δ u
x Ts
Fig. 3.12 Velocity and thermal boundary layers on a flat plate

66 A. Atreya
Boundary Layer Approximation The Thus all terms in Equation 3.42 are of order of
governing Equations 3.35 through 3.37 and 3.39 magnitude unity except the term [(ν/Lu1)∂2u*/
can be further simplified for the case when ∂x*2], which is much less than 1 and can be
the Reynolds number is reasonably large ignored. Thus, Equation 3.36 is simplified to
(Re (L/δ)2; that is, Re is of the order (L/δ)2)
such that δ L. To compare the various terms ∂u ∂u 1dp ∂2 u
u þv ¼ gx þ ν 2 ð3:43Þ
in the governing equations, first normalize all the ∂x ∂y ρ dx ∂y
variables so that they are of the order of magni-
tude unity. By defining Similarly, Equations 3.37 and 3.39 reduce to
x y u ∂p
x* ¼ y* ¼ u* ¼ 0 ð3:44Þ
L δ u1 ∂y
and and
T Ts
T* ¼ ð3:40Þ ∂T ∂T ∂2 T
T1 Ts u þv ¼α 2 ð3:45Þ
∂x ∂y ∂y
variables that change from 0 to 1 inside the
Equation 3.44 simply implies that p ¼ p(x),
boundary layer are obtained. Substituting these
that is, the pressure at any plane where x ¼ con-
into Equation 3.35 we find that
stant does not vary with y inside the boundary
layer and hence is equal to the free stream
∂u* L ∂v
¼ pressure. To summarize, the boundary layer
∂x* δu1 ∂y* approximation yields a simpler set of governing
equations that are valid inside the boundary
This suggests that
layer. These equations for steady flow of an
incompressible fluid with constant properties are
Lv
v* ¼ Continuity
δu1
∂u ∂v
so that þ ¼0 ð3:35Þ
∂x ∂y
∂u* ∂v*
þ ¼0 ð3:41Þ x-momentum
∂x* ∂y*
∂u ∂u 1∂p ∂2 u
Substituting x*, y*, u*, and v* into Equation 3.36 u þv ¼ gx þ ν 2 ð3:43Þ
and simplifying ∂x ∂y ρ ∂x ∂y
Energy
∂u* ∂u* ∂ p* ν
u* þ v* ¼ g*x þ
∂x* ∂y* ∂x* Lu1
∂T ∂T ∂2 T
" 2 2 # u þv ¼α 2 ð3:45Þ
∂2 u* L ∂ u* ∂x ∂y ∂y
þ
∂x*2 δ ∂y*2
To illustrate the use of these equations in deter-
ð3:42Þ mining the heat transfer coefficient, consider two
classical examples: (1) laminar forced convection
where p* p=ρu21 and g*x gx L=u21 : over a flat surface, and (2) laminar free convection
In Equation 3.42, the quantity ν/Lu1 is on a vertical flat surface. Forced convection is
recognized as 1/Re which is of the order (δ/L)2. chosen as a precursor to free convection because
www.ebook777.com
it is simpler and also allows us to illustrate the Nondimensionalizing Equations 3.35, 3.45,
difference between them. A flat geometry is also 3.47, and 3.48 according to Equation 3.401
chosen in both cases for simplicity. we obtain
Laminar Forced Convection Over a Flat ∂u* ∂v*

þ ¼0 ð3:49Þ
Surface A schematic of this problem is ∂x* ∂y*
presented in Fig. 3.12, and the objective here is
to obtain the gradients of temperature and veloc- ∂u* ∂u* 1 ∂2 u*
ity profile at y ¼ 0. By applying the Bernoulli u* þ v* ¼ ð3:50Þ
∂x* ∂y* ReL ∂y*2
Equation in the potential flow region outside the
boundary layer we obtain ∂T* ∂T* 1 ∂2 T*
u* þ v* ¼ ð3:51Þ
∂x* ∂y* ReL Pr ∂y*2
u21 p
þ þ gh ¼ constant ð3:46Þ
2 ρ along with the boundary conditions
Since the free stream velocity, u1, is constant,
u* ¼ v* ¼ T* ¼ 0 at y* ¼ 0
for a given height y ¼ h above the flat surface we
obtain that p ¼ constant, that is, p 6¼ p(x) out- and
side the boundary layer in the potential flow
region. From Equation 3.44 note that p 6¼ p( y) u* ¼ T* ¼ 1 at y* ! 1 ð3:52Þ
inside the boundary layer. Hence, p ¼ constant
both inside and outside the boundary layer over a where
flat surface. This implies that the term ∂p/∂x
equals zero in Equation 3.43. Also, since the u1 L
ReL
flow is forced (i.e., generated by an external ν
agent such as a fan, rather than by buoyancy)
the gravitational force, gx, in Equation 3.43 is the Reynolds number based on length, L, and
does not contribute to the increase in momentum Pr ν/α is the Prandtl number. Note that
represented by the left side of the equation, and Equations 3.49 and 3.50 are sufficient for deter-
gx ¼ 0. Thus Equation 3.43 becomes mining u* (x*, y*) and v* (x*, y*) and that once
these are known, Equation 3.51 can be indepen-
∂u ∂u ∂2 u dently solved for T* (x*, y*) Also note that for
u þv ¼ν 2 ð3:47Þ Pr ¼ 1, Equations 3.50 and 3.51 as well as their
∂x ∂y ∂y
corresponding boundary conditions are identical.
Equations 3.35, 3.45, and 3.47 govern the Thus for Pr ¼ 1 only Equations 3.49 and 3.50
temperature and velocity distributions inside need to be solved.
the boundary layer shown in Fig. 3.12. The A similarity solution of Equations 3.49 and
associated boundary conditions areno-slip 3.50 along with the boundary conditions (Equa-
tion 3.52) was obtained by Blasius [2]. Blasius
u¼v¼0 at y¼0 observed that since the system under consider-
ation has no preferred length, it is reasonable to
and suppose that the velocity profiles at different
T ¼ Ts at y¼0 ð3:48Þ
also
1
A more convenient definition of y* ¼ y/L and v* ¼ v/
u1 has been used since we are no longer interested in
quantities of order of magnitude unity; instead we are
u ¼ u1 and T ¼ T 1 as y ! 1 simply interested in eliminating units.
68 A. Atreya
u∞
u∞
u∞ u∞
δ (x )
y δ (x )
δ (x )
Fig. 3.13 Observed velocity profiles at different values of x
ð
values of x have similar shapes; that is, if u and
ψ ¼ ϕðηÞdy* þ f 1 ðx*Þ
y are suitably scaled then the velocity profile may
be expressed by a single function for all values of pffiffiffiffiffi ð
x*
x (Fig. 3.13). An obvious choice is ¼ pffiffiffiffiffiffiffiffi ϕðηÞdη þ f 1 ðx*Þ
ReL

u y
¼ϕ ϕ ð3:53Þ Since v* ¼ 0 at y* ¼ 0, f1(x*) is at best an arbi-
u1 δðxÞ trary constant which is takenð as zero. Also, defin-
This choice, as it stands, is not very useful ing a new function f ðηÞ ϕðηÞdη, we obtain
because δ(x) is not known. However, in accor-
dance with the boundary layer approximation, pffiffiffiffiffi f ðηÞ
Rex u1x/ν ~ (x/δ)2. Therefore, ψ¼ x* pffiffiffiffiffiffiffiffi ð3:56Þ
ReL
rffiffiffiffiffiffi
νx therefore,
δ
u1

∂ψ ∂ψ ∂η
can be expected. Substituting into Equation 3.53 u* ¼ ¼
∂y* x* ∂η x* ∂y* x*
we obtain
0 df
h ypffiffiffiffiffiffiffiffii ¼ f ðη Þ ¼ ð3:57Þ
u dη
u* ¼ ¼ϕ Rex ¼ ϕðηÞ ð3:54Þ
u1 x
and
where
∂ψ ∂ψ ∂ψ ∂η
v* ¼ ¼ þ
ypffiffiffiffiffiffiffiffi y* pffiffiffiffiffiffiffiffi ∂x* y* ∂x* η ∂η x* ∂x* y*
η Rex ¼ pffiffiffiffiffi ReL 1

0 pffiffiffiffiffiffiffiffiffiffiffiffi
x x* ¼ f ηf x*ReL
2
is the similarity variable. By introducing a stream ð3:58Þ
function, ψ, such that
On substituting u*, v* into Equation 3.50 and
∂ψ ∂ψ simplifying we obtain
u* ¼ and v* ¼ ð3:55Þ
∂y* ∂x* 00
0 00
2f þ f f ¼ 0 ð3:59Þ
Equation 3.49 is identically satisfied.
Substituting Equation 3.54 into Equation 3.55 where primes represent differentiation with
and integrating, we get respect to η. Equation 3.59 is a third-order
www.ebook777.com
nonlinear ordinary differential equation. Recall Therefore,

that η was a combination of two independent
variables, x* and y*, and it was assumed that rffiffiffiffiffiffiffiffiffi
y u1 x
u* ¼ ϕ(η). If this similarity assumption was η¼ ¼5 for y ¼ δ ¼ δt
x ν
incorrect, then the partial differential Equa-
tion 3.50 would not have reduced to an ordinary or for Pr ¼ 1,
differential Equation 3.59—that is, x* would not
have completely disappeared from the governing 5x
equation. Note also that even though Equa- δ ¼ δt ¼ pffiffiffiffiffiffiffiffi ð3:61Þ
Rex
tion 3.59 is nonlinear and has to be solved numer-
ically, there are no parameters and therefore From Equation 3.61 it is clear that δ and
it needs to be solved only once. Boundary δt increase with x but decrease with increasing
conditions corresponding to Equation 3.59 u1 (the larger the free stream velocity, the thin-
become ner the boundary layer). Now, to determine the
heat transfer coefficient and the friction factor we

f ¼ f 0 ¼ 0 at η ¼ 0, and f 0 need ∂T=∂y y¼0 and ∂u=∂y y¼0 . From Fig. 3.14,
¼ 1 as η ! 1 ð3:60Þ we have
A numerical solution of Equation 3.59 along
∂T* ∂u*
with the boundary conditions, Equation 3.60 is ¼ ¼ 0:332
shown in Fig. 3.14. Note that for Pr ¼ 1, the ∂η η¼0 ∂η η¼0
solution for T* is the same as that for u*. Also,
once T* (x*, y*) and u* (x*, y*) are known the Thus,
heat transfer coefficient and friction factor can
easily be obtained from Equations 3.5 and 3.11. ∂u pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
τs ¼ μ ¼ 0:332u1 ρμu1 =x ð3:62Þ
Furthermore, from the definition of thermal and ∂y y¼0
velocity boundary layer thickness (T* ¼ u*
¼ 0.99), we find that η ¼ 5. and
1.0 1.0
0.8604
0.8 0.8
2
.33
=0
T * = u * = f ′ (η)
0.6 0.6
v * √x * √ReL
pe
Slo
0.4 0.4
0.2 0.2
0
0 1 2 3 4 5 6
η = (y */√x * ) √ReL
Fig. 3.14 Nondimensional velocity profiles in laminar boundary layer over a flat plate
70 A. Atreya

00 ∂T pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Cf , h, δ
q_ s ¼ k ¼ 0:332ðT s T 1 Þk ρu1 =μx u∞, T∞
∂y y¼0 h (x ) ~ x –1/2
ð3:63Þ Cf (x ) ~ x –1/2
δ(x ) ~ x1/2
Hence the local friction and heat transfer
coefficients are
0:664
C f ¼ pffiffiffiffiffiffiffiffi ð3:64Þ
Rex
and x
kpffiffiffiffiffiffiffiffi Fig. 3.15 Variation of Cf, h, and δ with x for flow over a
h ¼ 0:332 Rex ð3:65Þ flat plate
x
Equation 3.65 is often rewritten in terms of a The results for the friction factor, Cf, and the
nondimensional heat transfer coefficient called Nusselt number, Nu, given by Equations 3.64 and
the Nusselt number, Nu, as 3.68 are for local values; that is, Cf and Nu change
with x. This variation is shown in Fig. 3.15. At
hx pffiffiffiffiffiffiffiffi x ¼ 0, both Cf and h tend to infinity. This is
Nu ¼ ¼ 0:332 Rex ð3:66Þ physically incorrect and happens because near
k
x ¼ 0 the boundary layer approximation breaks
All the above results are for the case when down since δ is no longer much less than x.
Pr ¼ 1. When Pr 6¼ 1, Equation 3.51 must also For many applications, only average values of
be solved with the help of the solution just the heat transfer coefficient, h, and friction factor,
obtained for Equations 3.49 and 3.50. Equa- C f , are required. These are obtained by using
tion 3.51 becomes Equations 3.9 and 3.15. In these equations dAs ¼
dx · (the unit width of the flat plate), and the
d2 T* Pr dT* average can be obtained from x ¼ 0 to any
þ f ¼0
dη2 2 dη length, L (which may be the total length of the
plate). Simple integration leads to the following
with T*(0) ¼ 0 and T*(1) ¼ 1. This solution results:
does not change the expressions for δ and Cf
given by Equations 3.61 and 3.64. However CfL ¼ 1:328 ReL 1=2 ¼ 2C f ðEvaluated at x ¼ LÞ
δt and Nu become [2]
ð3:69Þ
5x and
δt ¼ pffiffiffiffiffiffiffiffi Pr1=3 ¼ δPr1=3 ð3:67Þ
Rex
hL L
and NuL ¼ 0:664 ReL 1=2 Pr1=3
k
pffiffiffiffiffiffiffiffi ¼ 2 NuðEvaluated at x ¼ LÞ ð3:70Þ
Nu ¼ 0:332 Rex Pr1=3 ðPr≳0:6Þ ð3:68Þ
It is interesting to note that Cf and Nu are closely
Note that for Pr < 1 (usually true for gases), related. For example, from Equations 3.69 and
δt is greater than δ; that is, the thermal boundary 3.70 one can easily obtain
layer is thicker than the momentum boundary
layer. This is to be expected since Pr < 1 implies CfL ReL Pr1=3
NuL ¼
that ν is less than α. 2
www.ebook777.com
or form of a plume or a buoyant jet. A buoyant

plume above a fire is a familiar example. How-
NuL CfL 2=3 ever, here we will focus on free convection flows
St ¼ Pr ð3:71Þ that are bounded by a surface. A classical exam-
ReL Pr 2
ple of boundary layer development on a heated
where St is known as the Stanton number. vertical flat plate is discussed below.
This analogy between heat and momentum
transfer is called the Reynolds analogy which is Heated, Vertical Flat Plate Consider the flat
significant because the heat transfer coefficient plate shown in Fig. 3.16. The plate is immersed
can be determined from the knowledge of the in an extensive, quiescent fluid, with Ts > T1.
friction factor. This analogy is especially useful The density of the fluid close to the plate is less
for cases where mathematical solutions are not than that of the fluid that is farther from the plate.
available. Buoyancy forces therefore induce a free convec-
tion boundary layer in which the heated fluid
Laminar Free Convection In contrast with rises vertically, entraining fluid from the quies-
forced convection, where the fluid motion is cent region.
externally imposed, for free convection the fluid Under steady-state laminar flow conditions,
motion is caused by the buoyancy forces. Buoy- Equations 3.35, 3.43, and 3.45 describe the
ancy is due to the combined effect of density mass, momentum, and energy balances for
gradients within the fluid and a body force that the two-dimensional boundary layer shown in
is proportional to the fluid density. In practice the Fig. 3.16. Assume that the temperature
relevant body force is usually gravitational,
although it may be centrifugal, magnetic, or elec-
tric. Of the several ways in which a density Ts > T∞
gradient may arise in a fluid, the two most com-
mon situations are due to (1) the presence of u (y )
temperature gradients, and (2) the presence of
concentration gradients in a multicomponent sys-
tem such as a fire. Here, the focus will be on free
convection problems in which the density gradi- δ(x ) u∞ = 0
ent is due to temperature and the body force is T∞, ρ ∞
gravitational. Note, however, that the presence of
density gradients in a gravitational field does not
ensure the existence of free convection currents. Ts
For example, the high temperature, lighter fluid
T (y )
may be on top of a low temperature, denser fluid, T∞
resulting in a stable situation. It is only when the g
condition is unstable that convection currents are δt (x)
generated. An example of an unstable situation
would be a denser fluid on top of a lighter fluid. x, u
In a stable situation there is no fluid motion and,
therefore, heat transfer occurs purely by conduc-
tion. Here we will only consider the unstable
situation that results in convection currents.
Free convection flow may be further classified
y, v
according to whether or not the flow is bounded
by a surface. In the absence of an adjoining Fig. 3.16 Boundary layer development on a heated ver-
surface, free boundary flows may occur in the tical plate
72 A. Atreya
differences are moderate, such that the fluid may u¼v¼0 T ¼ T1 as y ! 1 ð3:76Þ
be treated as having constant properties. Also,
with the exception of the buoyancy force term Nondimensionalizing Equations 3.35, 3.45, 3.75,
(gx in Equation 3.43), the fluid can be assumed to and 3.76 with x* ¼ x/L, y* y/L, u ¼ u/u0,
be incompressible. Outside the boundary layer, v* v/u0, and T* (T T1)/(Ts T1), we
Equation 3.36 is valid, and since u ¼ v ¼ 0 obtain
outside the boundary layer we obtain: dp/dx ¼
ρ1 gx. Since ∂p/∂x ¼ 0 because of the bound- ∂u* ∂v*
þ ¼0 ð3:77Þ
ary layer approximation (i.e., p 6¼ p( y) inside ∂x* ∂y*
the boundary layer; Equation 3.44), dp/dx
inside the boundary layer must be equal to its ∂u* ∂u* gβT*ðT s T 1 ÞL
corresponding value outside, that is, dp/dx ¼ u* þ v* ¼
∂x* ∂y* u20
ρ1 gx.
Substituting this into Equation 3.43 1 ∂2 u*
þ ð3:78Þ
ReL ∂y*2
∂u ∂u ð ρ ρÞ ∂2 u
u þv ¼ gx 1 þν 2 ð3:72Þ and
∂x ∂y ρ ∂y
The first term on the right side of Equa- ∂T* ∂T* 1 ∂2 T*

u* þ v* ¼ ð3:79Þ
tion 3.72 is the buoyancy force, and the flow ∂x* ∂y* ReL Pr ∂y*2
originates because the density ρ is variable. By
introducing the coefficient of volumetric thermal Note that u0 in Equation 3.78 is an unknown
expansion, β, reference velocity and not the free stream veloc-
ity as in the case of forced convection. Also, the

1 ∂p 1 ðρ1 ρÞ dimensionless parameter
β¼ ð3:73Þ
ρ ∂T ρ ρ ðT 1 T Þ
gβðT s T 1 ÞL
it follows that u20
ð ρ1 ρÞ is a direct result of buoyancy forces. To eliminate

¼ βðT T 1 Þ ð3:74Þ the unknown reference velocity, u0 from the
ρ
dimensionless parameter, we define
Substituting Equation 3.74 into Equation 3.72 a

useful form of the x-momentum is obtained as gβðT s T 1 ÞL Lu0 2
Grashof number, GrL
u20 ν
∂u ∂u ∂2 u gβðT s T 1 ÞL3
u þv ¼ gx β ð T T 1 Þ þ ν 2 ð3:75Þ ¼
∂x ∂y ∂y ν2
From Equation 3.75 it is now apparent how Thus, the first term on the right side of Equa-
buoyancy force is related to temperature differ- tion 3.78 becomes GrL/(ReL)2. The Grashof num-
ence. Note that the appearance of the buoyancy ber plays the same role in free convection as the
term in the momentum equation mathematically Reynolds number does in forced convection.
complicates the situation. The decoupling Gr is the ratio of buoyancy and viscous forces.
between the hydrodynamic and the thermal The governing equations now contain three
problems achieved in forced convection is no parameters—the Grashof number, Reynolds
longer possible, since T appears in both number, and Prandtl number. For the forced
Equations 3.45 and 3.75. The boundary convection case it is seen (Equation 3.68) that
conditions associated with the governing Nu ¼ Nu (Re, Pr); thus for the free convection
equations, Equations 3.35, 3.45, and 3.75, are case, we expect Nu ¼ Nu (Re, Gr, Pr). If the
www.ebook777.com
buoyancy term in Equation 3.79 is Gr/(Re)2 convective case we found that δ ~ x1/2;
1, then we primarily have free convection; Fig. 3.15). Following a reasoning similar to the
that is, Nu ¼ Nu(Gr, Pr). For Gr/(Re)2 1, forced convective case, a similarity variable
the forced convective case exists, where as has ξ y/δ(x) or ξ ¼ Ay/x1/4 may be found, where
already been seen, Nu ¼ Nu (Re, Pr). However, A is an arbitrary constant. Also, motivated by
when Gr/(Re)2 ~ 1 a mixed (free and forced) Equation 3.57 for forced convection, it is hoped
convection case is obtained. For the present prob- that u ¼ Bx1/2f0 (ξ) where B is an arbitrary con-
lem we will assume that Gr
(Re)2, thus, Nu stant. Expressing these in nondimensional
must be a function of only Gr and Pr. variables, we get
Since Gr
Re2, it follows that buoyancy
forces are much larger than inertia forces; in ζ ¼ Ay* =x*1=4 ð3:80Þ
other words, the primary balance is between the
buoyancy and viscous forces. Since the left and
side of Equation 3.78 represents the inertia
0
forces, the primary balance is between the two u* ¼ Bx*1=2 f ðξÞ
terms on the right side, that is,
where f 0 (ξ) ¼ df/dξ. Note that the definitions

gβT*ðT s T 1 ÞL ν ∂ u* 2 of the arbitrary constants A and B have been
changed during nondimensionalization. By
u20 u0 L ∂y*2
introducing a stream function, ψ, as in Equa-
Crudely approximating the various terms, we tion 3.55, Equation 3.77 is identically satisfied.
have in dimensional variables Thus,
ð
u 0
gβðT 1 T Þ ν 2 ð3:aÞ ψ ¼ Bx*1=2 f ðξÞdy* þ f 1 ðx*Þ
δ
ð
Similarly approximating Equations 3.77 and 3.79 B 3=4 0
¼ x* f ðξÞdξ þ f 1 ðx*Þ ð3:81Þ
and expressing the result in dimensional form A
(it is more convenient to use Equations 3.35 and B
3.45), we get from Equation 3.35 or 3.77 ¼ x*3=4 f ðξÞ þ f 1 ðx*Þ
A
u v δu Since v* ¼ 0 at y* ¼ 0 (or ξ ¼ 0), f1(x*) is at
or v ð3:bÞ
x δ x best an arbitrary constant which is taken to
be zero without any loss of generality. From
and from Equation 3.79 or 3.45 along with Equations 3.55 and 3.81 we get
relation (3.b)
B h 0
i
ðT 1 T Þ ðT 1 T Þ αx v* ¼ 3 f ð ξ Þ ξ f ð ξ Þ ð3:82Þ
u α or u ð3:cÞ 4Ax*1=4
x δ2 δ2
By using Equations 3.80 and 3.82, Equations 3.78
Combining (3.a) and (3.c) we obtain an expres- and 3.79 can be reduced to
sion for the boundary layer thickness, δ,
00
0 2
1=4 0 00
f þ 3f f 2 f þ T* ¼ 0 ð3:83Þ
ναx
δ
gβðT 1 T Þ
and
Thus, we expect δ to scale with x and u to 1/4
00 0
scale with x1/2. (Note that in the forced T* þ 3Pr f T* ¼ 0 ð3:84Þ
74 A. Atreya
0
where the following definitions of the arbitrary f ¼ f ¼ 0 and T * ¼ 1 when ξ ¼ 0
constants A and B have been used:
and
1=2
4gβðT s T 1 ÞL
B¼ 0
u20 f ¼ 0 T * ¼ 0 at ξ ¼ 1 ð3:87Þ
1=4 ð3:85Þ
gβðT s T 1 ÞL3
A¼ A numerical solution of Equations 3.83 and
4ν2
3.84 along with the boundary conditions given
Note that in Equation 3.84 it has been assumed by Equation 3.87 are shown in Fig. 3.17. Note
that T* is a function of ξ only. From Equa- that the nondimensional x-velocity component,
tion 3.85 it follows that u, may be readily obtained from Fig. 3.17 part
(a) through the use of Equations 3.80 and 3.85.
1=4 Note also that, through the definition of the
gβðT s T 1 ÞL3
y*
ξ¼ similarity variable, ξ, Fig. 3.17 may be used to
x*1=4 4ν2 obtain values of u* and T* for any value of x*

y* GrL 1=4 and y*. Once u*(x*, y*) and T *(x*, y*) are
¼ 1=4 ð3:86Þ known, the heat transfer coefficient can easily
x* 4
be obtained from Equation 3.5. Thus, the temper-
The associated boundary conditions given by ature gradient at y ¼ 0 after using Equation 3.86,
Equation 3.76 become becomes
a b
0.3 0.6
0.2 0.4
Pr = 0.01 T* Pr = 0.01
0.1 0.2
0 0
6 8 10 12 14 16 18 20 22 24 6 8 10 12 14 16 18 20 22 24
0.7 ξ ξ
0.6
1.0
0.5
f ′(ξ) = ⎯ Grx– 1/2
Pr = 0.01 0.8 Pr = 0.01

T * = (T − T∞) / (Ts − T∞)
0.4
2ν
ux
0.6
0.3
0.72 0.72
1 0.4 1
0.2
2 2
10 10
0.1 100 0.2 100
1000 1000
0 0
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6
y Grx 1/4 y Grx 1/4
ξ = ⎯ (⎯⎯) ξ = ⎯ (⎯⎯)
x 4 x 4
Fig. 3.17 Laminar free convection boundary layer on an isothermal, vertical surface
www.ebook777.com
Table 3.1 Dimensionless temperature gradient for free Thus,

convection on a vertical flat plate

Pr 0.01 0.72 1 2 10 100 1000 hL L 4 GrL 1=4
g(Pr) 0.081 0.505 0.567 0.716 1.169 2.191 3.966 NuL ¼ ¼ gðPrÞ ð3:91Þ
k 3 4

∂T ðT s T 1 Þ ∂T* or from Equation 3.89, with x ¼ L we get
¼
∂y y¼0 L ∂y* y*¼0
!
Evaluated
ðT s T 1 Þ GrL 1=4 dT* 4
NuL ¼ Nu ð3:92Þ
¼
Lx*1=4 4 dξ ξ¼0 3 at x ¼ L
The local heat transfer coefficient is It should be noted that the foregoing results
apply irrespective of whether Ts > T1 or Ts <
1=4
k GrL dT* T1. If Ts < T1, the conditions are inverted from
h¼ ð3:88Þ
Lx*1=4 4 dξ ξ¼0 those shown in Fig. 3.16. The loading edge is on
the top of the plate, and positive x is defined in
or the direction of the gravity force.
1=4
hx GrL dT*
Nu ¼ ¼ x*3=4
k 4 dξ ξ¼0 Integral Solution Methods
1=4
Grx
¼ gðPrÞ ð3:89Þ For many problems an exact similarity solution
4 is not possible. An alternate approach is to use
approximate integral methods originally pro-
As is evident from Fig. 3.17, the dimensionless
posed by von Kárman in 1921 [6]. Intelligent
temperature gradient at ξ ¼ 0 is a function of the
use of integral methods gives good result and it
Prandtl number. In Equation 3.89 this function is
can be very useful for solving difficult fire
expressed as –g(Pr). Values of g(Pr) obtained
problems. This method is also very useful for
from the numerical solution are listed in Table 3.1.
conduction problems.
From Equation 3.88 for the local heat transfer
coefficient, the average heat transfer coefficient
for a surface of length L is obtained by using
Equation 3.9 as follows: Forced Flow Laminar Boundary Layer
on a Flat Plate
ð ðL
1 L k GrL 1=4 dx
hL ¼ hðdx 1Þ ¼ 7=4 gðPrÞ 1=4 Referring to Fig. 3.12, one can essentially inte-
L 0 L 4 0 x
grate the equations in the y direction across the
4 k GrL 1=4 boundary layer. This eliminates one variable—
¼ gðPrÞ
3L 4 resulting in ODEs. Integrating the continuity
ð3:90Þ equation:
ðδ ðδ ðδ
∂u ∂V ∂u
dy þ dy ¼ 0; Since, V ¼ 0 at y ¼ 0, V ð y ¼ δÞ ¼ dy
∂x ∂y ∂x
0 0 0
76 A. Atreya
Integrating the momentum equation: thicknesses are determined and the assumed
functional forms are then completely
ðδ ðδ specified.
∂u δ ∂V ∂u ∂u
u dy þ uVj0 u dy ¼ v
∂yδ ∂y0
4. While this method is approximate, it fre-
∂x ∂y
0 0 quently leads to accurate results for the
surface parameters that are usually of interest
Eliminating V by using continuity and in fire problems.
rearranging, the integral form of the boundary Now consider the hydrodynamic boundary
layer momentum equation results. It basically layer, for which the appropriate boundary
states that the decrease in kinetic energy of the conditions are: uð y ¼ 0Þ ¼ 0; uð y ¼ δÞ ¼
fluid in the B.L. occurs because of the wall fric- u1 ; ð∂u=∂yÞ y¼δ ¼ 0 Since u ¼ v ¼ 0 ; at y ¼ 0,
tion at y ¼ 0.
∂2 u=∂y2 y¼0
¼ 0:
2δ 3
ð With these conditions, we could approximate
d4 ∂u
ðu1 uÞu dy5 ¼ v the velocity profile as a third degree polynomial
dx ∂y y¼0 of the form:
0
In a similar fashion, the integral form of the u

y
y2
y3
¼ a1 þ a2 þ a3 þ a4
boundary layer energy equations may be u1 δ δ δ
obtained as:
Applying the boundary conditions to determine
2 δt 3
ð the coefficients a1 to a4. It is easily verified that
d4 ∂T
ðT 1 T Þu dy5 ¼ α a1 ¼ a3 ¼ 0, a2 ¼ 3=2 and a4 ¼ 1=2, in which
dx ∂y
0
y¼0 case:
These equations satisfy the x-momentum and u 3

y 1
y3
¼
energy conservation requirements in an integral u1 2 δ 2 δ
(or average) fashion over the entire boundary
layer. In contrast, the original conservation This velocity profile is specified in terms of the
equations satisfy the conservation requirements unknown boundary layer thickness δ which is
locally at each point in the boundary layer. determined by substituting into the momentum
The solution procedure involves: equation and integrating over y to obtain:
1. Assuming reasonable functional forms for the
unknowns u and T in terms of the 13 u1 dδ 1 140 ν δ2
¼ ; or δdδ ¼ dx i:e:
corresponding (unknown) boundary layer 140 ν dx δ 13 u1 2
thicknesses. 140 νx
2. These functional forms must satisfy appropri- ¼ þ const:
13 u1
ate boundary conditions.
3. Substituting these forms into the integral Since δ ¼ 0 at the leading edge ( x ¼ 0 ), the
equations, expressions for the boundary layer integration constant ¼ 0. Thus:
rffiffiffiffiffiffi
νx
δ ¼ 4:641 ¼ 4:641 x Re1=2
x ; This completely specifies the above velocity profile:
u1

∂u τs 0:646
Thus τs ¼ μ ; differenciating the velocity profile, we get : C f , x 2 ¼ :
∂y y¼0 ρu1 =2 Re1=2 x
The exact solution gave us 0.664 instead of 0.646

as the constant—a minor difference.
www.ebook777.com
Integral Solution for the Thermal Integral Solution for a Heated Vertical
Laminar Boundary Layer Plate (Laminar Free Convection)
Similar to the momentum boundary layer, the The problem shown in Fig. 3.16, for which the
temperature profile of the following form is similarity solution was obtained above, can now
assumed: be solved using integral methods.
2 3
T Ts y y y
T ¼ *
¼ b1 þ b2 þ b3 þ b4 Integral Conservation Equations
T1 Ts δt δt δt
Boundary conditions are:
The integral form of the momentum conservation
*
equation with buoyancy force is derived by using
T ð y ¼ 0Þ ¼ 0; T ð y ¼ δt Þ ¼ 1; ∂T =∂y y¼δt
* *
the Archimedes’ principle. The buoyancy force

¼ 0; & ∂2 T * =∂y2 y¼0 ¼ 0: per unit volume acting on an element of the

3 warmer fluid in the boundary layer is gðρ1 ρÞ
Applying these conditions: T* ¼ 32 δyt 12 δyt . directed vertically upward, where g is the gravi-
Substituting both velocity and temperature tational acceleration, ρ1 is the constant fluid
profiles into the integral energy equation and density outside the boundary layer, and ρ is the
manipulating for Pr 1: density of the fluid element that varies across the
boundary layer. In Fig. 3.18, Newton’s second
δ Pr1=3
¼ ; This result is in good agreement law of motion is applied to the elemental control
δt 1:026
volume of unit depth Δx and extending to y ¼ Y,
with the exact solution of 1.0 instead of 1.026.
where Y is greater than the boundary layer thick-
Thus, the heat transfer coefficient may be then
ness. The net momentum outflow from the vol-
computed from:
ume is equal to the buoyancy force minus the
viscous drag force exerted by the wall:
k f ∂T=∂y y¼0 3 kf
h ¼ ; ðY ðY
Ts T1 2 δt
ρu2 dyxþΔx ρu2 dyx
hx
Nux ¼ 0:332Re1=2 Pr1=3 0 0
kf ðY
∂u
Same as before within 3 decimal. This analysis ¼ gðρ1 ρÞdyΔx μ Δx
0 ∂y y¼0
proves that good solutions may be obtained using
approximate integral methods. These could be Notice that there is no momentum flow across the
very useful in fire problems. boundary at y ¼ Y since uðY Þ ¼ 0.
Fig. 3.18 Force balances Y

g(r∞ − r) dyΔx
on an elemental control
Y
∫
∫
volume in the boundary
layer of a heated vertical ru2dy 0
g
Layer
plate 0
x + Δx
x + Δx
∂u Δx
μ
∂y 0
Boundary
x
y =Y
Y
∫ 0
ru2dy
x
78 A. Atreya
Dividing by Δx and letting Δx ! 0 Assuming that the hydrodynamic and thermal

boundary layers have the same thickness,
ð ðY
d Y 2 ∂u i:e: δ ¼ δt , which seems reasonable because the
ρu dy ¼ gðρ1 ρÞdy μ
dx 0 0 ∂y y¼0 momentum is driven by thermal differences.
Also assuming the following forms for the veloc-
The Boussinesq approximation is introduced by ity and temperature profiles that satisfy the
taking the density to be constant except in the boundary conditions:
buoyancy term; dividing by ρ then gives:
u y
y 2 ðT T 1 Þ
y2
ðY ðY ¼ 1 , and ¼ 1
d ρ ρ ∂u u0 δ δ TS T1 δ
u2 dy ¼ g 1 dy ν
dx 0 0 ρ ∂y y¼0
Here u0 is a scaling velocity, which is a function
Since ρ ¼ ρðP, T Þ a Taylor expansion can be of x only and is yet to be determined. Figure 3.19
written as: ðρ1 ρÞ ¼ ð∂ρ=∂T Þ p ðT 1 T Þ þ shows these profiles; the maximum velocity is
ð∂ρ=∂T ÞT ðP1 PÞ þ higher order terms: The 0.148u0 at y ¼ δ /3. In the forced-convection
pressure variation across the boundary layer is problem, the known free-stream velocity was
negligible; thus, P ¼ P1 to obtain: used to scale the velocity profile; here the scaling
velocity u0, as well as the boundary layer -
thickness δ, are unknowns. The chosen
ρ1 ρ 1 ∂ρ
¼ ðT T 1 Þ ¼ βðT T 1 Þ, boundary layer profiles give ∂y ∂u
¼ 0 and ∂T ¼0
ρ ρ ∂T P ∂y
at y ¼ δ (i.e., they are smooth at the edge of the
where β is the volumetric coefficient of thermal boundary layer). However, the velocity profile
expansion, as before. For an ideal gas, β ¼ 1=T: ∂2 u
does not have the correct limiting value of ∂y 2 at
Substituting and letting Y ! 1, since there is no
contribution to the integrals for y > Y, the wall; also, the assumption of δ ¼ δt is not
valid for high-Prandtl-number fluids. But the use
ð1 ð1
d ∂u of these profiles gives surprisingly good results.
u dy ¼
2
gβðT T 1 Þdy ν Substituting the profiles and performing the
dx 0 0 ∂y y¼0
indicated operations gives:
ð3:93Þ

The integral form of the energy conservation d u20 δ νu0 gβðT S T 1 Þδ
¼ þ
equation is identical to that for forced flow: dx 105 δ 3

21 3 d u0 δ 2α
ð ¼
d4 ∂T dx 30 δ
ðT 1 T Þu dy5 ¼ α ð3:94Þ
dx ∂y
y¼0
0
Simultaneous Solution of the Equations 1

(T − T∞)
TS − T∞
The two ordinary differential equations above,

T − T∞
Equations (3.93) and (3.94) are coupled since
or
the variable T appears in both; hence, they must TS − T∞

u0
u
be solved simultaneously. Boundary conditions

for the velocity and temperature profiles are:
u
y ¼ 0 : u ¼ 0, T ¼ T s , u0
a constant for an isothermal wall, and as

Fig. 3.19 Shape of chosen velocity and temperature
y ! 1 ðor δÞ : u ¼ 0, T ¼ T 1 profiles
www.ebook777.com
These are two ordinary differential equations Substituting for D1 and rearranging gives:
for two unknowns: u0 and δ. Assuming
power law variations for δ and U of the form: 1=4
Pr
δ ¼ D1 xm ; u0 ¼ D2 xn , which give δ ¼ 0 and u0 Nux ¼ 0:508 Rax1=4 ; where Rax
0:952 þ Pr
¼ 0 at x ¼ 0. Substituting in the above equation
gives: β ðT S T 1 Þg x3
¼ is the Rayleigh number
να
d D22 D1 x2nþm νD2 xnm
¼ 2
dx 105 D1 This result agrees very well with the exact
gβðT S T 1 ÞD1 xm numerical solutions of the differential conserva-
þ tion equations and has been widely used.
3
The x dependence cancels if 2n + m 1 ¼ n Example: Laminar Natural-Convection Bound-
m ¼ m, which requires m ¼ 1/4, and n ¼ 1/2. ary Layer on a Vertical Flat Plate A vertical
The differential equations then reduce to two plate at 320 K is immersed in water at 300 K.
algebraic equations for D2 and D1: At a location 10 cm from the bottom of the plate,
determine δ, u0, Nux, hx, and qs . Also plot the
5 D22 D1 νD2 gβðT S T 1 ÞD1 velocity [u(y)] and temperature [T(y)] profiles.
¼ 2þ
4 105 D1 3
2 Solution Properties are evaluated at the mean
3 D2 D1 2α
¼ film temperature of 310 K; k ¼ 0.628 W/m K,
4 30 D1
ρ ¼ 993 kg/m3, ν ¼ 0. 70 x 10-6 m2/s, Pr ¼ 4.6.
Solving gives: Also, α ¼ ν/Pr ¼ 1.52 10-7 m2/s, β ¼ 3.62
10-4 K-1. The Rayleigh number is checked to
80α see if the flow is laminar:
D2 ¼
D21
1=4
ð20=21Þα2 þ να βðT S T 1 Þgx3
D1 ¼ 3:94 Rax ¼ ¼ 6:68 108
gβðT S T 1 Þ να
< 109 ðlaminarÞ
The wall heat flux is obtained from the tempera-
ture profile as:
The boundary layer thickness is δ ¼ Dx1=4 ,

∂T 2k where D is given by the equation above:
qS ¼ k ¼ ðT S T 1 Þ ¼ hx ðT S T 1 Þ
∂y y¼0 δ
hx 2 2 hx x 2x3=4
¼ ¼ ; Thus, Nux ¼ ¼
k δ D1 x1=4 k D1
1=4
ð20=21Þα2 þ να 3
3 1
δ ¼ D1 x ; and D1 ¼ 3:94
1=4
or, δ ¼ 4:57 10 m4 ð0:1 mÞ4
gβðT S T 1 Þ
¼ 2:57 103 m ¼ 2:57 mm
80 A. Atreya
320 0.05
T (K) u (m/s)
0.04
315
0.03
u (m/s)
T (K)
310
0.02
305
0.01
300 0
0 0.001 0.002
y (m)
The scaling velocity is: u0 ¼ D2 x1=2 , where hx ¼ ðk=xÞNux ¼ ð0:628=0:1Þð77 :9Þ

80α ¼ 489 W=m2 K
D2 ¼ 2 ¼ 0:582 m1=2 =s;
D1 qs ¼ hx ð T S T 1 Þ
Thus, u0 ¼ ð0:582Þð0:1Þ1=2 ¼ 0:184 m=s: ¼ 489ð320 300Þ
The local Nusselt number:
1=4 ¼ 9790 W=m2
Pr
Nux ¼ 0:508 Ra1=4 ¼ 77:9 ;
0:952 þ Pr x
At x ¼ 0.1 m. Hence, the local heat transfer The velocity and temperature profiles are
coefficient and heat flux are: obtained from:
u y
y2 ðT T 1 Þ
y2
¼ 1 , and ¼ 1 ; Substituting : δ ¼ D1 x1=4 ; u0 ¼ D2 x1=2
u0 δ δ TS T1 δ
2 2
y x1=4 D2 y y
u¼ 1 , and T ¼ T 1 þ ðT S T 1 Þ 1
D1 D1 x1=4 D1 x1=4
From above : D1 ¼ 4:57 10‐3 m3=4 , D2 ¼ 0:582m1=2 =s, and x ¼ 0:1m
The plot is shown above:
Complications in Practical Problems geometry, onset of turbulence, changes in fluid

properties with temperature, and because of
In the previous section, two relatively simple simultaneous mass transfer from the surface as
problems of laminar forced and free convection illustrated in Fig. 3.17. For such cases, empirical
on a flat surface were solved. These solutions correlations are obtained. These correlations are
illustrate the methodology for determining the discussed in the next section and the various
heat transfer coefficient and provide the neces- complications are individually discussed below.
sary insight regarding the relationship between
the various dimensionless parameters. Most Effect of Turbulence In both forced and free
practical situations are often more complex, and convective flows, small disturbances may be
mathematical solutions, such as those presented amplified downstream, leading to transition
in the previous section, are not always possible. from laminar to turbulent flow conditions.
Complexities arise due to more complex These disturbances may originate from the free
www.ebook777.com
stream or be induced by surface roughness. fluctuations. These fluctuations enhance the

Whether these disturbances are amplified or momentum and energy transfers and hence
attenuated depends on the ratio of inertia to vis- increase the surface friction and convection heat
cous forces for forced flows (the Reynolds num- transfer rate. Also, due to the mixing of fluid
ber) and the ratio of buoyancy to viscous forces resulting from the turbulent fluctuations, the tur-
for free convective flows (the Grashof number). bulent boundary layer is thicker and the bound-
Note that in both Reynolds and Grashof numbers, ary layer profiles (of velocity, temperature, and
viscosity appears in the denominator. Thus for concentration) are flatter than in laminar flow.
relatively large viscous forces or small Reynolds In a fully turbulent flow, the primary mecha-
and Grashof numbers, the naturally occurring nism of momentum and heat transfer involves
disturbances are dissipated, and the flow remains macroscopic lumps of fluid randomly moving
laminar. However, for sufficiently large about in the flow. Turbulent flow contrasts with
Reynolds and Grashof numbers (Re > 5 105 the random molecular motion resulting in molec-
and Gr > 4 108, for flow over a flat plate) ular properties discussed at the beginning of this
disturbances are amplified, and a transition to chapter. In the turbulent region, eddy viscosity
turbulence occurs. and eddy thermal conductivity are important.
The onset of turbulence is associated with the These eddy properties may be ten times as large
existence of random fluctuations in the fluid, and as their molecular counterparts.
on a small scale the flow is unsteady. As shown If one measures the variation of an arbitrary
in Fig. 3.20, there are sharp differences between flow variable, P, as a function of time at some
laminar and turbulent flows. In the laminar location in a turbulent boundary layer, then the
boundary layer, fluid motion is highly ordered typical behavior observed is shown in Fig. 3.21.
and it is possible to identify streamlines along The variable P, which may be a velocity compo-
which fluid particles move. In contrast, fluid nent, fluid temperature, pressure, or species
motion in the turbulent boundary layer is highly concentration, can be represented as the sum
irregular and is characterized by velocity of a time-mean value, P, and a fluctuating
Turbulent
Ts >T∞
a
v
Streamline
y, v u u∞ g
Transition
u
x, u
u∞ = 0
u∞ u∞ Turbulent region
xc
Buffer layer
Laminar
Laminar sublayer
xc x
x Entrainment
Laminar Transition Turbulent y
Fig. 3.20 (a) Velocity boundary layer development on a flat plate for forced flow; (b) velocity boundary layer
development on a vertical flat plate for free convective flow
82 A. Atreya
Fig. 3.21 Variation in the P

variable P with time at
some point in a turbulent
boundary layer
P′
P
component, P0 . The average is taken over a time Equations 3.96 through 3.98 are similar to
interval that is large compared with the period of the laminar boundary layer equations expressed
a typical fluctuation, and if P is time independent in mean flow variables, except for the presence
0 0 0 0
then the mean flow is steady. Thus, the instanta- of additional terms ρu v and ρC p v T . Physical
neous values of each of the velocity components, arguments [2] show that these terms result
pressure, and temperature are given by from the motion of macroscopic fluid lumps
0 0 0
and account for the effect of the turbulent
u¼uþu, v¼vþv, p¼ pþ p fluctuations on momentum and energy
0 0
T ¼TþT, and ρ¼ρþρ transport.
ð3:95Þ On the basis of the foregoing result it is cus-
tomary to speak of total shear stress and total heat
Substituting these expressions for each of the flow flux, which are defined as
variables into the boundary layer equations

(Equations 3.35, 3.43, and 3.45) and assuming ∂u 0 0
the mean flow to be steady, incompressible τtot μ ρu v

∂y
(ρ ¼ constant) with constant properties, and using
the well-established time averaging procedures and
[1–4], the following governing equations are
obtained:
∂T 0 0
Continuity q_ tot k ρC p v T ð3:99Þ
∂y
∂u ∂v
þ ¼0 ð3:96Þ 0
∂x ∂y 0 0 0
The terms ρu v and ρC p v T are always
x-momentum negative and so result in a positive contribution
0 0
to total shear stress and heat flux. The term ρu v

∂u ∂u ∂ ∂u 0 0 ∂p represents the transport of momentum flux due
ρ u þv ¼ μ ρu v ρgx to turbulent fluctuations (or eddies), and it is
∂z ∂y ∂y ∂y ∂x
ð3:97Þ known as the Reynolds stress. The notion of
transport of heat and momentum by turbulent
Energy
eddies has prompted the introduction of transport
coefficients, which are defined as the eddy
∂T ∂T ∂T ∂T 0 0
diffusivity for momentum transfer, εM, and
ρC p u þv ¼ k ρC p v T
∂x ∂y ∂y ∂y eddy diffusivity for heat transfer, εH, and have
ð3:98Þ the form
www.ebook777.com
Fig. 3.22 Boundary layer

u∞(x)
formation and separation
on a circular cylinder in
cross flow x θ
V + Wake
Separation point
Forward
stagnation point Boundary layer
∂u 0 0 Complex Geometry In a previous section on

εM u v
∂y the boundary layer concept, analysis was limited
ð3:100Þ to the simplest possible geometry, that is, a flat
∂T 0 0
εH v T plate. This provided considerable simplification
∂y because dp/dx ¼ 0 in Equation 3.43 for the
forced flow case. However, the situation is not
Thus Equation 3.99 becomes
as simple for fluid flow over bodies with a finite
radius of curvature.
∂u
τtot ρðv þ εM Þ Consider a common example of flow across a
∂y circular cylinder shown in Fig. 3.22. Boundary
and layer formation is initiated at the forward stagna-
tion point, where the fluid is brought to rest with
00 ∂T an accompanying rise in pressure. The pressure is
q_ tot ρC p ðα þ εH Þ ð3:101Þ a maximum at this point and decreases with
∂y
increasing x, the streamline coordinate, and θ,
As noted earlier, eddy diffusivities are much the angular coordinate. (Note: In the boundary
larger than molecular diffusivities, therefore the layer approximation, the pressure is the same
heat and momentum transfer rates are much inside and outside the boundary layer. This can
larger for turbulent flow than for laminar flow. be seen from Equation 3.44.) The boundary layer
A fundamental problem in performing turbulent then develops under the influence of a favorable
boundary layer analysis involves determining the pressure gradient (dp/dx < 0). At the top of the
eddy diffusivities as a function of the mean cylinder (i.e., at θ ¼ 90 ) the pressure eventually
properties of the flow. Unlike the molecular reaches a minimum and then begins to increase
diffusivities, which are strictly fluid properties, toward the rear of the cylinder. Thus, for 90 <
the eddy diffusivities depend strongly on the θ < 180 , the boundary layer development
nature of the flow. They vary across the boundary occurs in the presence of an adverse pressure
layer and the variation can only be determined gradient (dp/dx > 0).
from experimental data. This is an important Unlike parallel flow over a flat plate, for
point, because all analyses of turbulent flow curved surfaces the free stream velocity, u1,
must eventually rely on experimental data. To varies with x. (Note that in Fig. 3.22 a distinc-
date, there is no adequate theory for predicting tion has been made between the fluid velocity
turbulent flow behavior. upstream of the cylinder, V, and the velocity
84 A. Atreya
outside the boundary layer, u1(x).) At the stag- is delayed and the separation point moves far-
nation point, θ ¼ 0 , u1 ¼ 0. As the pressure ther downstream. This happens because the tur-
decreases for θ > 0 , u1 increases according to bulent boundary layer has more momentum than
the Bernoulli equation, Equation 3.46, and the laminar boundary layer to overcome the
becomes maximum at θ ¼ 90 . For θ > 90 , adverse pressure gradient.
the adverse pressure gradient decelerates the The foregoing processes strongly influence
fluid, and conversion of kinetic energy to pres- both the rate of heat transfer from the cylinder
sure occurs in accordance with Equation 3.46, surface and the drag force acting on the cylinder.
which applies only to the inviscid flow outside Because of the complexities associated with flow
the boundary layer. The fluid inside the bound- over a cylinder, experimental methods are used
ary layer has considerably slowed down because to determine the heat transfer coefficient. Such
of viscous friction and does not have enough experimental results for the variation of the local
momentum to overcome the adverse pressure Nusselt number with θ are shown in Fig. 3.24 for
gradient, eventually leading to boundary layer a cylinder in a cross flow of air. Consider the
separation, which is illustrated more clearly in results for ReD 105 (note: ReD is defined as
Fig. 3.23. At some location in the fluid, the VD/ν). Starting at the stagnation point, Nuθ
velocity gradient at the surface becomes zero decreases with increasing θ due to the develop-
and the boundary layer detaches or separates ment of the laminar boundary layer. However, a
from the surface. Farther downstream of the minimum is reached at θ 80 . At this point
separation point, flow reversal occurs and a separation occurs, and Nuθ increases with θ due
wake is formed behind the solid. Flow in this to the mixing associated with vortex formation in
region is characterized by vortex formation and the wake. For ReD 105, the variation of Nuθ
is highly irregular. The separation point is with θ is characterized by two minima. The
defined as the location at which (∂u/∂y)y ¼ decline in Nuθ from the value at the stagnation
0 ¼ 0. If the boundary layer transition to turbu- point is again due to laminar boundary layer
lence occurs prior to separation, the separation development, but the sharp increase that occurs
Favorable Adverse
pressure pressure
gradient gradient
∂p ∂p
0 0
∂x ∂x
Separation point
u∞(x) ( )
∂u
∂y
0
y=0
Wake
Vortices
Fig. 3.23 Velocity profiles associated with separation on a circular cylinder in cross flow
www.ebook777.com
between 80 and 100 is now due to boundary Changes in Fluid Properties In the analysis
layer transition to turbulence. With further devel- and discussion presented thus far, fluid properties
opment of the turbulent boundary layer, Nuθ were assumed to be constant. However, fluid
must again begin to decline. However, separation properties vary with temperature across the
eventually occurs (θ 140 ), and Nuθ increases boundary layer and this variation will have a
due to considerable mixing associated with the significant impact on the heat transfer rate. In
wake region. the empirical heat transfer correlations this influ-
The foregoing example clearly illustrates the ence is accounted for in one of two ways: (1) in
complications introduced by nonplanar geome- correlating the experimental data, all properties
try. Heat transfer correlations for these cases are are evaluated at the mean boundary layer temper-
often based on experimental data. Fortunately, ature, Tf ¼ (Ts + T1)/2, called the film tempera-
for most engineering calculations the local varia- ture, and (2) alternatively, all properties are
tion in the heat transfer coefficient such as that evaluated at T1 and an additional parameter is
presented in Fig. 3.24 is not required; only the used to account for the property variation. This
overall average conditions are needed. Empirical parameter is commonly of the form (Pr1/Prs)r or
correlations for average heat transfer coefficients (μ1/μs)r, where the subscripts 1 and s designate
will be presented in the next section. evaluation of properties at the free stream and
surface temperatures, respectively, and r is an
empirically determined constant.
800 It is important to note that in the empirical
correlations to be presented in the next section,
the same method that is employed in deriving the
700 correlation should be used when applying the
correlation.
600
Effect of Mass Transfer Special attention
ReD = 2.19 × 105 needs to be given to the effect that species mass
transfer from the surface of the solid has on the
500 1.86 × 10
5
velocity and thermal boundary layers. Recall that

1.70 × 105
the velocity boundary layer development is gen-
1.40 × 105 erally characterized by the existence of zero fluid
Nuθ 400
velocity at the surface. This condition applies to
1.01 × 105
the velocity component v normal to the surface,
300 as well as to the velocity component u parallel to
0.71 × 105 the surface. However, if there is simultaneous
mass transfer to or from the surface, it is evident
200 that v can no longer be zero at the surface. Nev-
ertheless, for the problems discussed in this chap-
ter, mass transfer is assumed to have a negligible
100 effect, that is, v 0. This assumption is reason-
able for problems involving some evaporation
from gas-liquid or sublimation from gas-solid
0 interfaces. For larger surface mass transfer rates
0 40 80 120 160
a correction factor (often called the blowing cor-
Angular coordinate, θ
rection) is utilized. This correction factor is sim-
Fig. 3.24 Local Nusselt number for airflow normal to a ply stated here, and discussed in greater detail by
circular cylinder Bird et al. [1]. The correction factor is defined as
86 A. Atreya
E(ϕ) h*/h, where h* is the corrected heat discussed; in addition the empirical formulas
transfer coefficient and h is the heat transfer that will be used in the “Applications” section
coefficient in the absence of mass transfer. of this chapter will be summarized.
According to film theory, E(ϕ) is given by
Functional Form of Solutions The nondimen-
ϕ sional Equations 3.49, 3.50, 3.51, and 3.78 are
Eð ϕ Þ ¼ ϕ ð3:102Þ
ðe 1Þ extremely useful from the standpoint of
suggesting how important boundary layer results
where can be generalized. For example, the momentum
equation, Equation 3.50, suggests that although
00
m_ Cpg conditions in the velocity boundary layer depend
ϕ¼
h on the fluid properties, ρ and μ, the velocity, μ1,
00
and the length scale, L, this dependence may be
m_ ¼ ρs vs is the mass flux coming out of the simplified by grouping these variables in a non-
surface and Cpg is the specific heat of the gas. dimensional form called the Reynolds number.
We therefore anticipate that the solution of Equa-
tion 3.50 will be of the form
Empirical Relations of Convection
Heat Transfer
d p*
u* ¼ f 1 x*, y*, ReL , ð3:103Þ
dx*
The analysis and discussion presented in the sec-
tion on the boundary layer concept have shown Note that the pressure distribution, p*(x*),
that for simple cases the convection heat transfer depends on the surface geometry and may be
coefficient may be determined directly from the obtained independently by considering flow
conservation equations. In the previous section it conditions outside the boundary layer in the
was noted that the complications inherent to most free stream. Hence, as discussed in the section
practical problems do not always permit analyti- on complex geometry, the appearance of dp*/dx*
cal solutions, and that it is necessary to resort to in Equation 3.103 represents the influence of
experimental methods. Experimental results are geometry on the velocity distribution. Note also
usually expressed in the form of either empirical that in Equation 3.50 the term dp*/dx* did not
formulas or graphical charts so that they appear because it was equal to zero for a flat
may be utilized with maximum generality. plate.
Difficulties are encountered in the process of Similarly we anticipate that the solution of
trying to generalize the experimental results in Equation 3.78 will be of the form
the form of empirical correlations. The availabil-
ity of an analytical solution for a simpler but u* ¼ f 2 ðx*, y*, GrL , PrÞ ð3:104Þ
similar problem greatly assists in guessing the
functional form of the results. Experimental Here, the Prandtl number is included because of
data is then used to obtain values of constants the coupling between Equations 3.78 and 3.79. If
or exponents for certain significant parameters, the flow is mixed, that is, buoyant as well as
such as the Reynolds or Prandtl numbers. If an forced, then the Reynolds number must also be
analytical solution for a similar problem is not included in the functional relationship expressed
available, it is necessary to rely on the physical by Equation 3.104.
understanding of the problem and on dimen- From Equation 3.1, the shear stress at the
sional or order-of-magnitude analysis. In this surface, y* ¼ 0, may be expressed as
section the experimental methods, the dimen-
∂u
μu ∂u*
sionless groups, and the functional form of the 1
τs ¼ μ ¼
relationships expected between them will be ∂y y¼0 L ∂y* y*¼0
www.ebook777.com
and from Equation 3.10 it follows that the fric- Thus

tion coefficient is
hL ∂T*
Nu ¼
τs 2 ∂u* k ∂y*
y*¼0
Cf ¼ ¼ y*¼0 ð3:105Þ
1=2ρu1 ReL ∂y*
2
Note that the Nusselt number, Nu, is equal to the
From Equation 3.103 it is clear that dimensionless temperature gradient at the sur-
face. From Equation 3.108 or Equation 3.107 it

∂u* d p* follows that for a prescribed geometry, i.e.,
¼ f x*, Re , ð3:106Þ
∂y* y*¼0 3 L known dp*/dx*
dx*
Hence, for a prescribed geometry (i.e., dp*/dx* Nu ¼ f 6 ðx*, ReL , PrÞ ð3:111Þ
is known from the free stream conditions) we
have for forced flow, and
Nu ¼ f 7 ðx*, GrL , PrÞ ð3:112Þ

2
Cf ¼ f ðx*, ReL Þ ð3:107Þ
ReL 3 for free convective flow. The Nusselt number is
to the thermal boundary layer what the friction
Equation 3.107 is very significant because it
factor is to the velocity boundary layer.
states that the friction coefficient may be
Equations 3.111 and 3.112 imply that for a
expressed exclusively in terms of a dimension-
given geometry, the Nusselt number must be
less space coordinate and the Reynolds number.
some universal function of x*, ReL, and Pr. If
For a prescribed geometry, the function that
this function were known, it could be used to
relates Cf to x* and ReL can be expected to be
compute the value of Nu for different fluids and
universally applicable. That is, it can be expected
different values of u1, T1, and L. Furthermore,
to apply to different fluids and over a wide range
since the average heat transfer coefficient is
of values for u1 and L.
obtained by integrating over the surface of the
Similar results may be obtained for the heat
body, it must be independent of the spatial vari-
transfer coefficient. Equation 3.51 suggests that
able, x*. Hence, the functional dependence of the
the solution may be expressed in the form
average Nusselt number is

d p*
T* ¼ f 4 x*, y*, ReL , ð3:108Þ hL
dx* Nu ¼ ¼ f 8 ðReL ; PrÞ ð3:113Þ
k
for forced flow, and for forced flow, and
T* ¼ f 5 ðx*, y*, GrL , PrÞ ð3:109Þ Nu ¼ f 9 ðGrL ; PrÞ ð3:114Þ
for free convective flow. Here ReL, GrL, and dp*/ for free convective flows.
dx* originate from the influence of fluid motion Although it is very helpful to know the func-
(u* and v*) on Equation 3.51. tional dependence of Nu, the task is far from
From the definition of the convection heat complete, because the function may be any of
transfer coefficient, Equation 3.5, and Equa- millions of possibilities. It may be a sine, expo-
tion 3.40 with y* ¼ y/L we obtain nential, or a logarithmic function. The exact form
of this function can only be determined by an
kð∂T=∂yÞ y¼0 k ∂T* analytical solution of the governing equations,
h¼ ¼þ ð3:110Þ
ðT s T 1 Þ L ∂y* y*¼0 such as Equations 3.70 and 3.91.
88 A. Atreya
Experimental Determination of Heat Transfer numbers—can be easily computed from their

Coefficient The manner in which a convection definitions.
heat transfer correlation may be obtained experi- The foregoing procedure is repeated for a vari-
mentally is illustrated in Fig. 3.25. If a prescribed ety of test conditions. We could vary the velocity,
geometry, such as the flat plate in parallel flow, is u1, the plate length, L, and the temperature dif-
heated electrically to maintain Ts > T1, convec- ference (Ts – T1), as well as the fluid properties,
tion heat transfer occurs from the surface to the using, for example, fluids such as air, water, and
fluid. It would be a simple matter to measure Ts engine oil, which have substantially different
and T1 as well as the electrical power, E · I, Prandtl numbers. Many different values of the
which is equal to the total heat transfer rate, q. _ Nusselt number would result, corresponding to a
The average convection coefficient, hL , can now wide range of Reynolds and Prandtl numbers. At
easily be computed from Equation 3.7. Also, this stage, an analytical solution to a similar but
from the knowledge of the characteristic length, simpler problem proves very useful in guiding
L, and the fluid properties, the values of the how the various nondimensional numbers should
various nondimensional numbers—such as the be correlated. For laminar flow over a flat plate it
Nusselt, Reynolds, Grashof, and Prandtl has been seen that in Equation 3.70 the relation-
ship is of the form
u∞, T∞ I ⋅ E = q = hLAs (Ts –T∞) Nu ¼ CReLm Prn
Ts, As Thus, we plot the results on a log-log graph as

shown in Fig. 3.26 and determine the values of C,
L m, and n. Because such a relationship is inferred
from experimental measurements, it is called an
empirical correlation. Along with this empirical
correlation it is specified how the temperature-
dependent properties were determined for calcu-
lating the various nondimensional numbers.
I E Insulation When such a correlation is used, it is important
that the properties must be calculated in exactly
Fig. 3.25 Experiment for measuring the average convec- the manner specified. If they are not specified,
tion heat transfer coefficient, hL
a b
Pr3
m
NuL = C ReLPrn NuL m
⎯⎯ = C ReL
Pr2 n
Pr
Pr1
Log NuL
Log ( )
NuL
Prn
Log ReL Log ReL
Fig. 3.26 Dimensionless representation of convection heat transfer measurements
www.ebook777.com
then the mean boundary layer temperature, Tf, range of conditions specified, but for most engi-
called the film temperature, must be used. neering calculations one should not expect the
accuracy to be much better than 20%.
Ts þ T1 For proper use of the foregoing correlations it
Tf ð3:115Þ
2 is important to note that the flow may not be
laminar or turbulent over the entire length of
A Summary of Empirical and Practical the plate under consideration. Instead, transition
Formulas In this section, selected dimension- to turbulence may occur at a distance xc (xc < L,
less groups (Table 3.2) and a variety of convec- where L is the plate length) from the leading
tion correlations (Tables 3.3 and 3.4) for external edge of the plate. In this mixed boundary layer
flow conditions are tabulated. Correlations for situation, the average convection heat transfer
both forced and free convection are presented coefficient for the entire plate is obtained by
along with their range of applicability. The integrating first over the laminar region (0 x
contents of this section are more or less a collec- xc) and then over the turbulent region (xc < x
tion of “recipes.” Proper use of these recipes is L ) as follows:
essential to solving practical problems. The ð xc ðL
reader should not view these correlations as sac- 1
hL ¼ hlam dx þ hturb dx ð3:116Þ
rosanct; each correlation is reasonable over the L 0 xc
Table 3.2 Selected dimensionless groups

Group Definition Interpretation
Friction coefficient Local τ2 Dimensionless surface shear stress
Cf ¼ 2
ρu 1=2
Average τ2
Cf ¼ 2
ρu 1=2
Reynolds number Location x Re ¼ u1 x Ratio of inertia and viscous forces
x
ν
Length L u1 L
ReL ¼
ν
Diameter D u1 D
ReD ¼
ν
Prandtl number C pμ ν Ratio or molecular momentum and thermal
Pr ¼ ¼
k α diffusivities
Grashof number Location x gβðT s T 1 Þx3 Ratio of buoyancy to viscous forces
Grx ¼
ν2
Length L gβðT s T 1 ÞL3
GrL ¼
ν2
Diameter D gβðT s T 1 ÞD3
GrD ¼
ν2
Rayleigh number Location x gβðT s T 1 Þx3 Product of Grashof and Prandtl numbers
Rax ¼ Grx Pr ¼
να
Replace x by Land D
to get RaL and RaD
Nusselt number Location x hx Ratio of convection heat transfer
Nux ¼
k to conduction in a fluid slab of thickness x
Replace x by L and D
to get NuL and NuD
00
Modified Grashof number Location x gβq_ s x4 Product of Grashof and Nusselt numbers
Gr*x ¼ Grx Nux ¼
k ν2
Stanton number h Nu Dimensionless heat transfer coefficient
St ¼ ¼
ρu1 C p RePr
90
Table 3.3 Summary of forced convection correlations for external flow geometries
Geometry/flow Type Equation Restrictions Comments
Flat plate/laminar Local Nux ¼ 0.332 Rex1/2 Pr1/3 Rex < 5 105 Properties evaluated at
(Ts ¼ constant) Tt ¼ (Ts + T1)/2
Average 1=2 0.6 Pr 50
NuL ¼ 0:664 ReL Pr1=3
Boundary layer δ
x ¼ 5Rex1=2
thickness
Flat plate/laminar Local Nux ¼ 0.453 Rex1/2 Pr1/3 Rex < 5 105 Properties evaluated at Tf.
00
(q_ s ¼ constant) However, Ts is not known.
00
Instead, qs is known. Thus, T f

¼ T 1 þ T s T 1 =2 where
qs L=K
Ts T1 ¼ 1=2 1=3
0:6795ReL Pr
0.6 Pr 50
Flat plate/turbulent (Ts Local Nux ¼ 0.0296 Rex4/5 Pr1/3 Rex < 108 Properties evaluated at Tf
¼ constant)
Boundary layer δ 0.6 Pr 60
x ¼ 0:37Rex1=5
thickness

www.ebook777.com
Mixed average 4=5 Transition to
NuL ¼ 0:037 ReL 871 Pr1=3
(laminar turbulent) turbulence at
Recrit ¼ 5 105
Flow across cylinders Average NuD ¼ C ReDm Pr1=3 0.4 < ReD < Properties evaluated at Tf
circular cylinder 4 105
ReD C m
0.4–4 0.989 0.330
4–40 0.911 0.385
40–4000 0.683 0.466
0.193 0.618
4 103–4 104
A. Atreya
3
Other geometries 4 104–4 105 0.027 0.805

square
V D
5 103–105 0.246 0.588
V D 5 103–105 0.102 0.675

Hexagon
V D
V D
5 103–1.95 104 0.160 0.638
1.95 104–105 0.0385 0.782

Vertical plate
V D 5 103–105 0.153 0.638
3 4
4 10 –1.5 10 0.228 0.731

Flow across spheres Average 1=2 2=3 3.5 < ReD < Properties
NuD ¼ 2 þ 0:4ReD þ 0:06ReD
1=4
7.6 104 evaluated
Pr0:4ðμ1 =μs Þ at T1
0.71 < Pr < 380

1:0 < μμ1 < 3:2
s
h i
Falling drop Average 1=2 :07 where x is the Properties
NuD ¼ 2 þ 0:6ReD Pr1=3 25 Dx
falling distance evaluated
measured from at T1
rest
91
92
Table 3.4 Summary of free convection correlations for external flow geometries
Vertical plates Local: (Ts ¼ const) 1=4 Grx 4 108 (laminar) Properties evaluated at
Grx
Nuxg(Pr) from Table 3.1 ¼ gðPrÞ (Tf ¼ Ts + T1)/2
4
Average: (Ts ¼ const) Grx 4 108 (laminar) Properties at Tf
4 GrL 1=4
NuL ¼ gðPrÞ
3 4
8 92
Average: (Ts ¼ const) > > None Properties at Tf
< 1=6 =
0:387RaL This correlation may be applied to
NuL ¼ 0:825 þ h
>
:
i8=27 >
; vertical
cylinders if
1 þ ð0:492=PrÞ9=16 D

1=4
35=GrL
L
or

00 1=5
Local: q_ s ¼ const Nux ¼ 0:6 Gr*x Pr 105 < Grx* < 1011 (laminar) Properties at Tf

00 1=4 2 1013 < Grx*Pr < 1016 Properties at Tf
www.ebook777.com
Local: q_ s ¼ const Nux ¼ 0:17 Gr*x Pr

00 1=5
Average: q_ s ¼ const NuL ¼ 0:75 Gr *L Pr 105 < Grx* < 1011 (laminar) Properties at Tf
Horizontal plates (hot Average: (Ts ¼ const) 1=4 105 ≲ RaL ≲ 107 Properties at Tf characteristic length
NuL ¼ 0:54 RaL
surface up or cold surface 1=3 L is defined as L ¼ As /P
NuL ¼ 0:15 RaL 107 ≲ RaL ≲ 1010
down)
00 where
ð1Þ 1=3 RaL 2 108 As ¼ plate surface area
Average: q_ s ¼ const NuL ¼ 0:16 RaL
F ¼ perimeter of the plate (1) All
1=4
properties except β are evaluated at
Horizontal plates (cold Average: (Ts ¼ const) NuL ¼ 0:27 RaL 105 RaL 1010
T e ¼ T s 14ðT s T 1 Þ β is evaluated
surface up or hot surface
00 ð1Þ 1=3 2 108 RaL 1011 at Tf
down) Average: q_ s ¼ const NuL ¼ 0:16 RaL

A. Atreya
3

00
Inclined plates Average: q_ s ¼ const NuL ¼ 0:56 ðRaL cos θÞ1=4 (hot surface θ < 8800 Properties evaluated at Te ¼ Ts – 1/4
facing down) 105 < RaL cos θ < 1011 (Ts – T1)
+θ −θ
For hot surface –15* > θ > –75* Grashof number
h facing up i 105 < RaL cos θ < 1011
NuL ¼ 0:14 ðGrL PrÞ1=3 ðGrc PrÞ1=3
Hot surface
þ 0:56ðRaL cos θÞ1=4
θ ¼ –15*; Grc ¼ 5 109
–30*; 2 109
–60*; 108
–75*; 105
8 92
Horizontal cylinders Average: (Ts ¼ const) > > 10–5 < RaD < 1012 Properties evaluated at Tf
< 1=6 =
0:387 RaD
NuD ¼ 0:6 þ h i 8=27
>
: >
;
1 þ ð0:559=PrÞ9=16
Spheres Average: (Ts ¼ const) 1=4 1 < RaD < 105 Properties evaluated at Tf
NuD ¼ 2 þ 0:43 RaD
1=4 Pr 1
NuD ¼ 2 þ 0:5 RaD
3 105 < Ra < 8 108
93
94 A. Atreya
where xc may be obtained from the critical correlations that require property evaluation
Reynolds or Grashof numbers. at the free stream temperature and include a
Also, several correlations given in Tables 3.3 property ratio to account for the nonconstant
and 3.4 are for the constant heat flux property effect.
00
q_ s ¼ constant boundary condition. Thus, the 3. Determine whether the flow is laminar or tur-
bulent. This determination is made by calcu-
surface temperature of the object is unknown and lating the Reynolds number and comparing the
yet the fluid properties are to be determined at value with the appropriate transition criterion.
Tf ¼ (Ts + T1)/2. For such cases an iterative For example, if a problem involves parallel
procedure is employed and the average surface flow over a flat plate for which the Reynolds
temperature can be determined as follows: number is ReL ¼ 106 and the transition crite-
rion is Recrit ¼ 5 105, it is obvious that a
00 NuL mixed boundary layer condition exists.
q_ s ðKnownÞ ¼ hðT s T 1 Þ ¼ ðT s T 1 Þ
ðL=kÞ 4. Decide whether a local or surface average
coefficient is required. Recall that the local
thus coefficient is used to determine the flux at a
00
particular point on the surface, whereas the
q_ s ðL=kÞ average coefficient determines the transfer
T s ðAverageÞ ¼ T 1 þ ð3:117Þ
NuL rate for the entire surface.
Having complied with the foregoing rules,
The use of correlations given in Tables 3.3 sufficient information will be available to select
and 3.4 is illustrated via examples in the next the appropriate correlation for the problem.
section.
Example 1 Electrical strip heaters are assembled
to construct a flat radiant heater 1 m wide for
Applications conducting fire experiments in a wind tunnel.
The heater strips are 5 cm wide and are indepen-
This section briefly summarizes the methodology dently controlled to maintain the surface temper-
for convection calculations and then presents ature at 500 C. Construction details are shown in
examples to illustrate the use of various Fig. 3.27. If air at 25 C and 60 m/s flows over the
correlations. plate, at which strip is the electrical input maxi-
mum? What is the value of this input? The radia-
Methodology for Convection tive heat loss is ignored.
Calculations The application of a convection
correlation for any flow situation is facilitated Solution Assumptions
by following a few simple rules: Steady-state conditions, neglect radiation losses,
1. Become immediately cognizant of the flow and no heat loss through the bottom surface.
geometry. Does the problem involve flow
over a flat plate, a sphere, a cylinder, and so Properties
forth? The specific form of the convection Tf ¼ 535 K: ρ ¼ 1 atm. From air property
correlation depends, of course, on the Table 3.5, k ¼ 42.9 10–3 W/mK; ν ¼ 43.5
geometry. 10–6 m2/s; Pr ¼ 0.683.
2. Specify the appropriate reference temperature
and then evaluate the pertinent fluid properties Analysis
at that temperature. For moderate boundary The strip heater requiring the maximum
layer temperature differences, it has been electrical power is that for which the average
found that the film temperature may be used convection coefficient is the largest. From the
for this purpose. However, there are knowledge of variation of the local convection
www.ebook777.com
T∞ = 25°C
u∞ = 60 m/s
Air 1 m wide
Ts = 500°C
5 cm
x Insulation
L2
Strip 5
L3 Typical
heater
Fig. 3.27 Construction details for wind tunnel experiments
Table 3.5 Thermophysical properties of air at atmospheric pressure

TK ρ kg/m3 cp kJ/kgK μ 107 N s/m2 ν 106 m2/s k 103 W/m K α 106 m2/s Pr
100 3.5562 1.032 71.1 2.00 9.34 2.54 0.786
150 2.3364 1.012 103.4 4.426 13.8 5.84 0.758
200 1.7458 1.007 132.5 7.590 18.1 10.3 0.737
250 1.3947 1.006 159.6 11.44 22.3 15.9 0.720
300 1.1614 1.007 184.6 15.89 26.3 22.5 0.707
350 0.9950 1.009 208.2 20.92 30.0 29.9 0.700
400 0.8711 1.014 230.1 26.41 33.8 38.3 0.690
450 0.7740 1.021 250.7 32.39 37.3 47.2 0.686
500 0.6964 1.030 270.1 38.79 40.7 56.7 0.684
550 0.6329 1.040 288.4 45.57 43.9 66.7 0.683
600 0.5804 1.051 305.8 52.69 46.9 76.9 0.685
650 0.5356 1.063 322.5 60.21 49.7 87.3 0.690
700 0.4975 1.075 338.8 68.10 52.4 98.0 0.695
750 0.4643 1.087 354.6 76.37 54.9 109 0.702
800 0.4354 1.099 369.8 84.93 57.3 120 0.709
850 0.4097 1.110 384.3 93.80 59.6 131 0.716
900 0.3868 1.121 398.1 102.9 62.0 143 0.720
950 0.3666 1.131 411.3 112.2 64.3 155 0.723
1000 0.3482 1.141 424.4 121.9 66.7 168 0.726
1100 0.3166 1.159 449.0 141.8 71.5 195 0.728
1200 0.2902 1.175 473.0 162.9 76.3 224 0.728
1300 0.2679 1.189 496.0 185.1 82 238 0.719
1400 0.2488 1.207 530 213 91 303 0.703
1500 0.2322 1.230 557 240 100 350 0.685
1600 0.2177 1.248 584 268 106 390 0.688
1700 0.2049 1.267 611 298 113 435 0.685
1800 0.1935 1.286 637 329 120 482 0.683
1900 0.1833 1.307 663 362 128 534 0.677
2000 0.1741 1.337 689 396 137 589 0.672
2100 0.1658 1.372 715 431 147 646 0.667
2200 0.1582 1.417 740 468 160 714 0.655
2300 0.1513 1.478 766 506 175 783 0.647
2400 0.1448 1.558 792 547 196 869 0.630
2500 0.1389 1.665 818 589 222 960 0.613
3000 0.1135 2.726 955 841 486 1570 0.536
96 A. Atreya
coefficient with distance from the leading edge, associated with the first seven heaters from that
the local maximum can be found. Figure 3.15 associated with the first eight heaters. Thus
shows that a possible location is the leading
edge on the first plate. A second likely location qconv, 8 ¼ h18 L8 W ðT s T 1 Þ
is where the flow becomes turbulent. To deter-
mine the point of boundary layer transition to h17 L7 W ðT s T 1 Þ
turbulence assume that the critical Reynolds
number is 5 105. It follows that transition The value of h17 is obtained from the equation
will occur at xc, where applicable to laminar conditions (Table 3.3).
Thus
νRecrit 43:5 106 5 105
xc ¼ ¼ m
u1 60 Nu17 ¼ 0:664Re7 1=2 Pr1=3
¼ 0:36 m or on the eighth strip
60 7 0:05 1=2
¼ 0:664 ð0:683Þ1=3
43:5 106
Thus there are three possibilities: ¼ 406:3
1. Heater strip 1, since it corresponds to the
largest local, laminar convection coefficient Nu17 k 406:3 42:9 103
h17 ¼ ¼
2. Heater strip 8, since it corresponds to the L7 7 0:05
largest local turbulent convection coefficient
¼ 49:8 W=m2 K
3. Heater strip 9, since turbulent conditions exist
over the entire heater
By contrast, the eighth heater is characterized by
For the first heater strip
mixed boundary layer conditions. Thus use the
formula (Table 3.3).
qconv, 1 ¼ h1 L1 W ðT s T 1 Þ

4=5
where h1 is determined from the equation below Nu18 ¼ 0:037Re8 871 Pr1=3
(see also Table 3.3).
Re8 ¼ 8 Re1 ¼ 5:52 105
Nu1 ¼ 0:664Re1 1=2 Pr1=3 Nu18 ¼ 510:5

Nu18 k
60 0:05 1=2 h18 ¼ ¼ 54:7W=m2 K
¼ 0:664 ð0:683Þ1=3 L8
43:5 106
¼ 153:6 The rate of heat transfer from the eighth strip is
then
hence,
qconv, 8 ¼ ð54:7 8 0:05 49:8 7 0:05Þ
Nu1 k 153:6 42:9 103
h1 ¼ ¼ ð500 25Þ ¼ 2113:8 W
L1 0:05
¼ 131:8W=m2 K The power requirement for the ninth heater

hence, strip may be obtained by either subtracting the
total heat loss associated with the first eight from
qconv, 1 ¼ ð131:8Þð0:05Þð1 mÞð500 25Þ that associated with the first nine, or by
¼ 3129 W integrating over the local turbulent expression,
since the flow is completely turbulent over the
The power requirement for the eighth strip may entire width of the strip. The latter approach
be obtained by subtracting the total heat loss produces
www.ebook777.com
Height, L = 0.71 m k ¼ 33.8 10–3 W/mK; ν ¼ 26.41 106

Width, W = 1.02 m 2
Glass panel m /s;
α ¼ 38.3 10–6 m2/s; Pr ¼ 0.69; β ¼ 1/
Tf ¼ 0.0025 K–1
Fire qconv = ? T∞ = 23°C
Analysis
The rate of heat transfer by free convection from
Ts = 232°C
the panel to the room is given by
Fig. 3.28 Glass panel fire screen q ¼ hAs ðT s T 1 Þ
where h is obtained from the following equation

u 4=5 ð L9
k 1 dx from Table 3.4.
h9 ¼ 0:0296 Pr1=3
L9 L8 v L8 x 1=5
8 92
>
< 1=6 >
=
4=5 0:387ReL
42:9 103 60 NuL ¼ 0:825 þ h i8=27 >
h9 ¼ 0:0296 >
: ;
0:05 43:5 106 1 þ ð0:492=PrÞ9=16
ð L9
dx
ð0:683Þ1=3 1=5 here
L8 x
h i
¼ 1825:22 ð0:45Þ0:8 ð0:4Þ0:8 ¼ 86:7 gβðT s T 1 ÞL3
RaL ¼
αν
qconv , 9 ¼ 86:7 0:05 1 ð500 25Þ 9:8 0:0025 ð232 23Þ ð0:71Þ3
¼
¼ 2059 W 38:3 106 26:4 106
¼ 1:813 109
hence
Since RaL > 109, transition to turbulence will
qconv, 1 > qconv, 8 > qconv, 9 occur on the glass panel and the appropriate
correlation from Table 3.4 has been chosen
and the first heater strip has the largest power
8 92
requirement. >
<
9 1=6 >
=
0:387 1:813 10
NuL ¼ 0:825 þ h i8=27 >
Example 2 A glass-door fire screen, shown in >
: ;
1 þ ð0:492=0:69Þ9=16
Fig. 3.28, is used to reduce exfiltration of room
air through a chimney. It has a height of 0.71 m, a ¼ 147
width of 1.02 m, and reaches a temperature of
232 C. If the room temperature is 23 C, esti- Hence
mate the convection heat transfer rate from the
fireplace to the room. NuL k 147 33:8 103
h¼ ¼
L 0:71
Solution Assumptions ¼ 7 W=m K
2
The screen is at a uniform temperature, Ts, and

and
room air is quiescent.
W
Properties q¼7 ð1:02m 0:71mÞ ð232 23Þ C
m2 K
Tf ¼ 400 K, P ¼ 1 atm. From air property table
(Table 3.5): ¼ 1060 W
98 A. Atreya
Note: in this case radiation heat transfer Solution We expect the argon near the heated
calculations would show that radiant heat trans- wall to rise, and that near the cold wall to
fer is greater than free convection heat transfer. descend. Also, for all practical purposes, the
two parallel steel plates are infinite in extent
Example 3 A fire door is constructed as shown in compared with the gap 2b. Since the gap is
Fig. 3.29. Two thin (but rigid) stainless steel walls sealed, under steady-state conditions any mass
are separated by an argon-filled gap of thickness going up must be equal to the mass coming
2b. The door height, approximately equal to the down to conserve the total mass of the gas inside
door width, is much greater than 2b. All other the gap.
sides of the door with one dimension 2b seal the The velocity vector has three components: VX,
argon-filled gap and are assumed to be insulated. VY, and VZ. Of these, only VZ is nonzero. Also,
In the event of a fire, the two vertical walls will be based on the assumption made, VZ varies only
at different temperatures, designated as “hot” and with x. The continuity and momentum equations
“cold.” Assuming that the wall temperatures T1 thus become
and T2 remain constant and using the Boussinesq Continuity
approximation, derive expressions for velocity
and temperature profiles in argon under steady- ∂V Z
state conditions. Also determine the heat flux ¼0
∂Z
through argon. Does it provide adequate insula-
tion? Note that the gas motion in the y direction is The Z-component of the momentum equation
expected to be small; thus it may be neglected becomes
along with any edge effects. Viscous dissipation
may also be neglected.
d2 V Z
μ þ ρm gβðT m T Þ ¼ 0
dx2
Fluid flow
Here Tm is the mean temperature, yet to be

defined. The constant value of density ρm is also
taken at this mean temperature. Any variation
Cold wall @ T2
above or below this mean temperature will pro-

Hot wall @ T1
vide the buoyancy force to drive the fluid. Other

than in the buoyancy term, throughout the
equations ρm is taken as constant—this is
the Boussinesq approximation. Note that with
the increase in the mean temperature, the pres-
sure will rise given the rigid walls of the door.
Thus, the mean density will remain constant
regardless of the temperature. However, the den-
sity will vary with temperature.
2b The boundary conditions are
Gravity
z BC1 : x ¼ b, V Z ¼ 0
BC2 : x ¼ 0, V Z ¼ by symmetry
x
Temperature can depend on at most three space

y
variables and time. For this problem T ¼ T(x)
because T1 and T2 are constant and conditions
Fig. 3.29 Example 3 are steady. Thus, the energy equation reduces to
www.ebook777.com
d2 T Fig. 3.30 Example

k ¼0 3 solution
dx2
(T1 – T2)
qx" = k
With the boundary conditions 2b
T1
BC1 : x ¼ b, T ¼ T 1
BC2 : x ¼ b, T ¼ T 2
The energy equation can be integrated directly

T2
to give
T1 þ T2 x
T¼ ðT 1 T 2 Þ VZ
2 2b
Thus, Tm can be conveniently defined as
T1 þ T2
Tm ¼
2
z
In terms of the mean temperature, the tempera-
ture profile becomes
x
T Tm x
¼
T1 T2 2b
Thus,
As shown in Fig. 3.30, despite the fluid
motion inside the door, the temperature profile
00 dT T1 T2
qx ¼ k ¼k is linear and effectively heat is conducted
dx 2b through low thermal conductivity argon gas
With mean temperature defined, we can now providing excellent insulation. For this condition
rewrite the momentum equation as to hold, door height must be much greater
than 2b.
d2 V Z ρm gβ ρ gβxðT 2 T 1 Þ
¼ ðT T m Þ ¼ m
dx 2 μ 2bμ
Nomenclature
This can also be directly integrated as
A Area (m2)
ρm gβx3 As Surface area (m2)
VZ ¼ ðT 2 T 1 Þ þ C3 x þ C4 Bi Biot number
12bμ
C Molar concentration (kmol/m3)
Applying the boundary conditions, we obtain the Cf Friction coefficient
velocity distribution as cp Specific heat at constant pressure
(J/kgK)
ρm gβx3 ρ gβbx cv Specific heat at constant volume
VZ ¼ ðT 2 T 1 Þ m ðT 2 T 1 Þ (J/kgK)
12bμ 12μ
D Diameter (m)
ρm gβb2 x x3 DAB Binary mass diffusion coefficient
¼ ðT 2 T 1 Þ 3
12μ b b (m2/s)
100 A. Atreya
Dh Hydraulic diameter (m) u, v, w Mass average fluid velocity

E Specific internal or thermal components (m/s)
(sensible) energy (J/kg) x, y, z Rectangular coordinates (m)
FBX, FBY, Components of the body force per xfd,h Hydrodynamic entry length (m)
FBZ unit volume (N/m3) xrd,t Thermal entry length (m)
FSX, FSY, Components of the surface force
FSZ Greek Letters
F Friction factor
Gr Grashof number α Thermal diffusivity (m2/s)
G Gravitational acceleration (m/s2) β Volumetric thermal expansion coefficient
H Convection heat transfer coefficient (K–1)
(W/m2 K) δ Hydrodynamic boundary layer thickness
h Average convection heat transfer (m)
coefficient (W/m2 K) δt Thermal boundary layer thickness (m)
hm Convection mass transfer coeffi- δd Mass transfer boundary layer thickness (m)
cient (m/s) η Similarity variable
hrad Radiation heat transfer coefficient θ Zenith angle (rad)
(W/m2 K) ϕ Azimuthal angle (rad)
K Thermal conductivity (W/mK) μ Viscosity (kg/sm)
L Characteristic length (m) ν Kinematic viscosity (m2/s)
Le Lewis number ρ Mass density (kg/m3)
M Mass (kg) σij Components of the stress tensor (N/m2)
Ṁ Mass flow rate (kg/s) ψ Stream function (m2/s)
ṁ00 Mass flux (kg/m2 s) τ Shear stress (N/m2)
00
ṁi Mass flux of species i (kg/m2 s)
Nu Nusselt number Subscripts
P Perimeter (m)
Pe Peclet number (RePr) A,B Species in a binary mixture
Pr Prandtl number Conv Convection
P Pressure (N/m2) D Diameter; drag
Q Energy generation rate per unit vol- F Fluid properties
ume (W/m3) Fd Fully developed conditions
0
q_ Heat transfer rate per unit length H Heat transfer conditions
(W/m) H Hydrodynamic; hot fluid
00
q_ Heat flux (W/m2) L Based on characteristic length
R Universal gas constant Max Maximum fluid velocity
Ra Rayleigh number S Surface conditions
Re Reynolds number Sur Surroundings
r, ϕ, z Cylindrical coordinates T Thermal
r, θ, ϕ Spherical coordinates X Local conditions on a surface
Sc Schmidt number 1 Free stream conditions
Sh Sherwood number
St Stanton number
T Temperature (K) References
T Time (s)
U Overall heat transfer coefficient 1. R.B. Bird, W.E. Stewart, and E.N. Lightfoot, Transport
(W/m2 K) Phenomena, Wiley, New York (1966).
www.ebook777.com
2. H. Schlichting, Boundary Layer Theory, McGraw-Hill, 6. T. von Karman, Uber laminare und turbulente reibung,
New York (1979). ZAMM,1(4), pp. 233–235, 1921
3. V.S. Arpaci and P.S. Larsen, Convection Heat Trans-
fer, Prentice Hall, Englewood Cliffs, NJ (1984).
4. E.R.G. Eckert and R.M. Drake, Analysis of Heat and Arvind Atreya is a professor of mechanical engineering
Mass Transfer, McGraw-Hill, New York (1973). at the University of Michigan. He has been actively
5. F.P. Incropera and D.P. Dewitt, Fundamentals of Heat involved in fire research since 1979.
Transfer, Wiley, New York (1981).
Radiation Heat Transfer

4
Revised by C. Lautenberger
Original chapter authored by C.L. Tien, K.Y. Lee,
and A.J. Stretton
Introduction vacuum, we have all enjoyed its radiative heat

transfer on cool days, and cursed it on hot ones.
Thermal radiation is the dominant mode of heat All objects with a finite temperature emit
transfer in flames with characteristic lengths thermal radiation through a physical mechanism
exceeding approximately 0.2 m. It is for this related to electron oscillations and transitions. As
reason that quantitative analysis of fire dynamics an object’s absolute temperature increases, these
requires a working knowledge of thermal electron oscillations and transitions become
radiation. This chapter will introduce the more rapid, resulting in increased radiant emis-
fundamentals of thermal radiation and offer sev- sion. Since all objects emit radiation, all objects
eral methods for calculating radiant heat transfer also have a certain amount of thermal radiation
in fires. Basic thermal radiation concepts are impinging upon them (originating from other
presented with an emphasis on application to emitting objects). It is the net difference between
fire phenomena; the reader is referred the litera- incoming and outgoing thermal radiation that
ture for specialized topics [1–4]. leads to a net rate of radiant heat transfer between
objects at different temperatures, and quantifica-
tion of this rate is usually the ultimate goal of a
Basic Concepts radiation heat transfer analysis.
The nature of thermal radiation transport can
The Nature of Thermal Radiation be explained on the basis of quantum mechanics
or electromagnetic wave theory. In the general
Whereas conduction and convection require direct quantum mechanical consideration, electromag-
contact for objects at different temperatures to netic radiation is viewed as the propagation of
exchange heat, thermal radiation is a distinct an ensemble of particles (usually called photons
mechanism of heat transfer that allows spatially or quanta). These particles, being generated by
separated objects at different temperatures to sub-molecular processes that are fed by an
transfer heat. Although the Earth is separated object’s internal energy, carry different energies.
from its Sun by 1.5 1011 m of near perfect The energy of a photon (e, J) is proportional to its
frequency (ν, s1):
Revised by C. Lautenberger (*) e ¼ hν ð4:1Þ

Reax Engineering Inc., 1921 University Ave., Berkeley,
CA 94704, USA The constant of proportionality in Equation 4.1 is
Original chapter authored by C.L. Tien, K.Y. Lee, and Planck’s constant, h ¼ 6.6256 1034 J s. It
A.J. Stretton is seen that the higher the frequency, the higher

www.ebook777.com
4 Radiation Heat Transfer 103
the photon energy. Thus, a radiation field is fully For comparison, visible light has wavelengths
described when the flux of photons (or energy) between 0.4 and 0.7 μm. Thermal radiation
is known for all points in the field for all with wavelengths between 0.7 and 100 μm is
directions and for all frequencies. infrared thermal radiation, whereas ultraviolet
Due to its wave-particle duality, electromag- thermal radiation has wavelengths between 0.1
netic radiation exhibits properties of both and 0.4 μm.
particles and waves. Therefore, thermal radiation
can also be explained as the propagation of Example 1 Calculate the energy of photons (in a
electromagnetic waves. In this context, wave- vacuum) with a wavelength of 101 μm (ultra-
length (λ) is related to frequency (ν) and the violet limit of thermal radiation) and 102 μm
speed of light (c) as: (infrared limit of thermal radiation).
c Solution Equation 4.3 can be used to calculate

λ¼ ð4:2Þ
ν photon energies. For the photon at the ultraviolet
The speed of light in a particular medium is limit:
denoted c, and in a vacuum it is denoted c0
where c0 ¼ 2.998 108 m/s. Wavelength has e¼
hc0
units of length, and for convenience it is usually λ

given in microns (μm, or micrometers) where 6:6256 1034 J s 2:998 108 m=s
1 μm ¼ 106 m. Substituting Equation 4.2 into ¼
101 μm 106 m=μm
Equation 4.1 shows that a photon’s energy
increases as wavelength decreases: ¼ 1:99 1018 J
hc And for the photon at the infrared limit:

e¼ ð4:3Þ
λ
hc0
Electromagnetic waves of practical signifi- e¼
cance have wavelengths ranging from 105 to λ

104 μm. Figure 4.1 shows the electromagnetic 6:6256 1034 J s 2:998 108 m=s
¼
spectrum spanning this range. Thermal radia- 102 μm 106 m=μm
tion usually refers to electromagnetic waves
¼ 1:99 1021 J
with wavelengths between 101 and 102 μm.
Visible
yellow
green
violet
blue
red
X rays Infrared
Ultraviolet Microwave
Thermal radiation
Gamma rays
10-5 10-4 10-3 10-2 10-1 100 101 102 103 104
λ , μm
Fig. 4.1 Electromagnetic spectrum. Adapted from Incropera and DeWitt [4]
104 Revised by C. Lautenberger
Spectral Distribution of Radiation from (or the second radiation constant). Note that k is
a Perfect Emitter the Boltzmann constant (k ¼ 1.3805 1023 J/
K). In Equation 4.4 and throughout this chapter, a
A diffuse surface is an idealized surface that subscript “b” indicates “blackbody” and a sub-
emits thermal radiation equally in all directions, script λ indicates “wavelength”, e.g. Eλ,b is the
i.e. its emission exhibits no directional depen- blackbody emissive power at a particular wave-
dency. A perfect emitter is an idealized surface length λ.
that emits the maximum possible thermal radia- Spectral emissive power is plotted in Fig. 4.2
tion at every wavelength. A blackbody is a dif- as a function of wavelength for several different
fuse perfect emitter that also absorbs all incident blackbody temperatures. Also shown in Fig. 4.2
radiation. is a line labeled λmax ¼ C3/T (where the third
The energy spectrum of radiation emitted by a radiation constant is C3 ¼ 2897.8 μm K) that
diffuse perfect emitter, or blackbody, can be relates the wavelength corresponding to the peak
calculated from Planck’s quantum theory. In spectral emissive power (λmax) to the blackbody
particular, the spectral (or monochromatic) temperature T. This is Wien’s displacement law,
blackbody emissive power (Eλ,b, W/m2 μm) is which is obtained by differentiating Equation 4.4
given by the Planck distribution: with respect to T, setting that result equal to zero,
and solving for λT. Wien’s displacement law
C1 shows that the maximum monochromatic emis-
Eλ, b ðλ; T Þ ¼ ð4:4Þ
λ ðexpðC2 =λT Þ 1Þ
5
sive power of a blackbody shifts to shorter
wavelengths as its temperature increases. From
where C1 ¼ 2πhc20 ¼ 3:742 108 W μm4 =m2 Equation 4.3, it is also seen that the photons
is Planck’s first constant (often called the emitted from blackbodies high temperature
first radiation constant) and C2 ¼ hc0 =k ¼ 1:439 (shorter wavelengths) carry more energy than
104 μm K is Planck’s second constant photons emitted from blackbodies at lower
Fig. 4.2 The Planck distribution: blackbody emissive power and Wien’s displacement law
www.ebook777.com
temperatures (longer wavelengths), a result that Table 4.1 Blackbody radiation fractions
one intuitively expects. λT λT
The Stefan-Boltzman Law is obtained by (μm K) F0!λ () (μm K) F0!λ ()
integrating the spectral blackbody emissive 200 0.000000 6200 0.754140
power over all wavelengths: 400 0.000000 6400 0.769234
600 0.000000 6600 0.783199
ð1 800 0.000016 6800 0.796129
Eb ¼ Eλ, b ðλÞdλ 1000 0.000321 7000 0.808109
0
ð1 1200 0.002134 7200 0.819217
C1
¼ dλ ¼ σT 4 ð4:5Þ 1400 0.007790 7400 0.829527
0 λ5 ðexpðC2 =λT Þ 1Þ 1600 0.019718 7600 0.839102
1800 0.039341 7800 0.848005
where σ is the Stefan-Boltzman constant (σ 2000 0.066728 8000 0.856288
¼ 5.67 108 W/m2 K4) and Eb is the total 2200 0.100888 8500 0.874608
(integrated over all wavelengths) blackbody 2400 0.140256 9000 0.890029
emissive power. It is seen that the oft-cited 2600 0.183120 9500 0.903085
“fourth power dependency” of thermal radiation 2800 0.227897 10,000 0.914199
on temperature is a direct consequence of 3000 0.273232 10,500 0.923710
integrating the Planck spectral distribution over 3200 0.318102 11,000 0.931890
all wavelengths. 3400 0.361735 11,500 0.939959
Equations 4.4 and 4.5 can be used to calculate 3600 0.403607 12,000 0.945098
the fraction of a blackbody’s radiant emission 3800 0.443382 13,000 0.955139
in a particular wavelength band, i.e. between 4000 0.480877 14,000 0.962898
4200 0.516014 15,000 0.969981
two wavelengths. Denote Fλ1 !λ2 as the fraction
4400 0.548796 16,000 0.973814
of a blackbody’s radiant emission between
4600 0.579820 18,000 0.980860
wavelengths λ1 and λ2. It is then calculated as:
4800 0.607559 20,000 0.985602
ð λ2 ð λ1 5000 0.633747 25,000 0.992215
Eλ, b dλ Eλ, b dλ 5200 0.658970 30,000 0.995340
5400 0.680360 40,000 0.997967
Fλ1 !λ2 ¼ 0 0
Eb 5600 0.701046 50,000 0.998953
¼ F0!λ2 F0!λ1 ð4:6Þ 5800 0.720158 75,000 0.999713
6000 0.737818 100,000 0.999905
Thus, Fλ1 !λ2 can be calculated from two values of
F0!λ , which is readily tabulated from Equa-
tion 4.4 as a function of λT (Table 4.1). This fraction of thermal radiation does this surface
then makes it possible to calculate the fraction emit in the visible range at temperatures of
of emission between two wavelengths. 800 and 1429 K?
Example 2 Consider an electrically heated sur- Solution The visible range is from 0.4 to
face used as a heater in a flammability test. What 0.7 μm. Thus for the 800 K emitter:
ðλT Þ1 ¼ 0:4 μm 800 K ¼ 320 μm K and F0!λ1 ¼ 0:000000
Fλ1 !λ2 ¼ F0!λ2 F0!λ1 ¼ 0:000000 0:000000 ¼ 0:000000

Since the fraction of radiant between 0.4 For the 1429 K emitter:
and 0.7 μm is 0 (to six decimal places), the heater
would not appear to be “glowing” at 800 K.
Fλ1 !λ2 ¼ F0!λ2 F0!λ1 ¼ 0:000321 0:000000 ¼ 0:000321
Since the emission fraction in the visible range is of a sphere having radius r is 4πr2 and a unit
nonzero (albeit very small) the heater would sphere, i.e. a sphere with a radius of 1, has a
appear to be glowing (if it didn’t melt first). surface area of 4π. The solid angle subtended
by a surface is the area of a unit sphere covered
by the surface’s projection onto that unit sphere.
Radiant Intensity and Heat Flux For example, a hemisphere subtends a solid angle
of 2π steradians, and all space subtends a solid
When analyzing fire phenomena, we usually speak angle of 4π steradians. Referring to the spherical
in terms of heat fluxes. For example, a radiant heat coordinate system in Fig. 4.3, the differential
00
flux (q_ r) of 20 kW/m2 to the floor is often quoted as solid angle dΩ is defined as:
a rule of thumb for determining the onset of flash-
over in a compartment. Consider a target located dA
dΩ ¼ 2 ¼ sin ðθÞdθdϕ ð4:7Þ
on the floor of a compartment as it approaches !
R
flashover: the radiant heat flux “felt” by this target
is the sum of all thermal radiation incident on this
target, regardless of where the radiation originated.
Some of the radiation incident to the target may
have been emitted by flames, another part
may have been emitted by the ceiling or walls,
and another part may have been emitted by soot
particles located in the hot gas layer. Thus, incident
radiation comes in from all directions and the
radiation felt by the target passes through an imag-
inary hemisphere surrounding the target. The radi-
ant intensity passing through different parts of
this hemisphere will, in general, vary spatially.
The radiant intensity passing through part of the
hemisphere facing the flames is likely greater than
the radiant intensity passing through part of the
hemisphere facing away from the flames. It is seen
that in order properly analyze a radiant heat trans-
fer problem, it is necessary to take into account the
directional nature of radiation. The concept of
radiant intensity is introduced as a tool to analyze
the directional nature of thermal radiation.
Radiant intensity is defined on a per unit solid
angle (Ω, sr or steradians) basis. The surface area Fig. 4.3 Coordinate system for radiation intensity
www.ebook777.com
where ϕ is azimuthal angle (radians), θ is polar intensity is a directional (and possibly spectral)
angle (radians), and dA is the differential area quantity. Radiant heat flux is obtained by sum-
normal to the θ and ϕ directions. In simple terms, ming (or integrating) individual contributions
ϕ can be thought of as degrees longitude and θ over all directions (Equation 4.9) and usually
can be thought of as degrees latitude on a globe. wavelengths (Equation 4.10).
Spectral radiant intensity (W/m2 sr μm) is an
inherently directional quantity defined as radiant
power per unit area normal to the emitting surface Emission, Irradiation, and Radiosity
per unit solid angle per unit wavelength:
Emission
dq_
I λ ðλ; θ; ϕÞ ¼ ð4:8Þ
dA dΩ dλ Now define Iλ,e(λ, θ, ϕ) as the spectral intensity
The radiant heat flux at a single wavelength of radiation emitted by a surface (subscript “e”
across a surface of an arbitrary orientation is the means emission or emitted). The emissive power
spectral radiant heat flux [5, 6]: of that surface at wavelength λ is defined in an
00
analogous manner to Equation 4.9, but with q_ λ
ð 4π replaced by Eλ and Iλ replaced by Iλ,e:
00
q_ λ ðλÞ ¼ I λ ðλ; θ; ϕÞ cos ðθÞdΩ
0 ð 4π
ð 2π ð π=2 Eλ ðλÞ ¼ I λ, e ðλ; θ; ϕÞ cos ðθÞdΩ
¼ I λ ðλ; θ; ϕÞ cos ðθÞ sin ðθÞdθdϕ 0
ð 2π ð π=2
0 0
ð4:9Þ ¼ I λ, e ðλ; θ; ϕÞ cos ðθÞ sin ðθÞdθdϕ
0 0
where Iλ is the radiation intensity at wavelength λ ð4:11Þ
per unit solid angle (Fig. 4.3). Intensity is a useful
measure for thermal radiation because the inten- It is seen that for a surface at a given temperature,
sity of a radiant beam remains constant if it is the intensity of emitted radiation depends on wave-
traveling through a nonparticipating medium. length and direction. For engineering applications,
The total radiant heat flux is obtained by this directional dependency of surface emission is
integrating Equation 4.9 over all wavelengths: usually neglected and we instead (implicitly or
explicitly) work with hemispherical radiation emis-
ð1
00 00 sion. Hemispherical emission can be envisioned by
q_ r ¼ q_ λ ðλÞdλ ð4:10Þ picturing a small (differential) element located at
0
the center of an imaginary hemispherical enclosure
The salient point here is that radiant intensity (see Fig. 4.4). Due to the directional dependency of
is not the same as radiant heat flux. Radiant surface emission, the intensity of radiation emitted
Fig. 4.4 Directional

emission from a differential
surface. Length of arrows
represents spectral
directional intensity of
emitted thermal radiation
ð1
by this surface that impinges on the imaginary
E¼ Eλ ðλÞdλ ¼ πI e ð4:13Þ
hemispherical enclosure may vary with location. 0
However, for engineering purposes, it is usually
adequate to neglect the potentially directional char-
An important result that is obtained from
acter of surface emission and consider only hemi-
Equation 4.4 after performing the integration in
spherical emission, meaning the radiant emission
Equation 4.13 is the Stefan-Boltzman Law,
that impinges on some part of the imaginary hemi-
already presented as Equation 4.5.
spherical enclosure. In practice, this is equivalent to
For a diffuse surface, the fraction of radiation
assuming that a surface is a diffuse emitter, mean-
emitted in angle range ϕ1 ϕ ϕ2 and θ1 θ
ing the intensity of emitted radiation is independent
θ2 can be calculated as:
of direction.
For the reasons described above, in most engi-
neering applications surfaces are approximated ð ϕ2 ð θ 2
1
as diffuse, which means that Iλ,e is a constant that Fðθ1 ; θ2 ; ϕ1 ; ϕ2 Þ ¼ cos ðθÞ sin ðθÞdθdϕ
π ϕ1 θ1
does not vary with θ and ϕ, even though emission
ð4:14Þ
of thermal radiation from all real surfaces
exhibits some directional dependency. Under
this diffuse approximation Iλ,e can be removed Example 3 What is the emissive power of a
from the integrand in Equation 4.11: blackbody at 1000 K? What is its emissive
power between 1 and 5 μm for all emission
ð 2π ð π=2 angles? What is its emissive power between
E λ ðλ Þ ¼ Iλ, e ðλ; θ; ϕÞ cos ðθÞ sin ðθÞdθdϕ 1 and 5 μm for 0 ϕ 2π and 0 θ π=4?
0 0
ð 2π ð π=2
¼ I λ, e ðλÞ cos ðθÞ sin ðθÞdθdϕ Solution Blackbody emissive power is calcu-
0 0 lated from Equation 4.5:
¼ πI λ, e ðλÞ
ð4:12Þ Eb ¼ σT 4 ¼ 5:67 108 10004
and it is seen that the emissive power of a diffuse ¼ 56:7 kW=m2

surface equals its spectral intensity multiplied by π
steradians (for diffuse surfaces E ¼ πI and for this The fraction of this radiation emitted at
reason Equation 4.4 can also be written as Eλ, b wavelengths between 1 and 5 μm can be
ðλ; T Þ ¼ πI λ, b ðλ; T Þ). The total emissive power is calculated with Equation 4.6 and Table 4.1 as
obtained by integrating over all wavelengths: follows:
ðλT Þ1 ¼ 1 μm 1000 K ¼ 1000 μm K and F0!λ1 ¼ 0:000321
ðλT Þ2 ¼ 5 μm 1000 K ¼ 5000 μm K and F0!λ2 ¼ 0:633747
Fλ1 !λ2 ¼ F0!λ2 F0!λ1 ¼ 0:633747 0:000321 ¼ 0:633426
The emissive power between 1 and 5 μm, for The fraction of radiation emitted in the direc-
all emission angles, is 56.7 kW/m2 0.633 ¼ tion 0 ϕ 2π and 0 θ π=4 can be calcu-
35.9 kW/m2. lated from Equation 4.14 as:
www.ebook777.com
ð 2π ð π=4 Solution Total irradiation can be calculated

1
Fðθ1 ; θ2 ; ϕ1 ; ϕ2 Þ ¼ cos ðθÞ sin ðθÞdθdϕ from Equation 4.16 as:
π 0 0
ð 2π θ¼π=4 ! ð1 ð8 ð 20
1 1
Fðθ1 ; θ2 ; ϕ1 ; ϕ2 Þ ¼ cos ðθÞ
2
dϕ G¼ Gλ ðλÞdλ ¼ 5dλ þ 10dλ
π 0 2 θ¼0 0 2 8
ð 2π ¼ 5 ð8 2Þ þ 10 ð20 8Þ
1 2π
¼ dϕ ¼ ¼ 0:5
4π 0 4π ¼ 150 kW=m2
The emissive power between 1 and 5 μm for this
angle range is 35.9 kW/m2 0.5 ¼ 18 kW/m2.
Radiosity
Irradiation As will be discussed later, a certain fraction of

radiation impinging on a surface may be reflected
Spectral irradiation Gλ(λ) (W/m2 μm) is the by that surface. Thus, the total amount of radiation
radiant heat flux at wavelength λ to a surface leaving a surface is the sum of the radiation emit-
incident from all directions. It is obtained by ted by that surface plus the radiation reflected by
integrating the incident spectral radiation inten- that surface. The total radiation leaving a surface,
sity over all angles in a manner directly analo- whether emitted or reflected, is called radiosity, J.
gous to the way that spectral emissive power was Spectral radiosity Jλ is:
defined in Equation 4.11:
ð 2π ð π=2
ð 2π ð π=2 J λ ð λÞ ¼ I λ, e, r ðλ; θ; ϕÞ cos ðθÞ sin ðθÞdθdϕ
Gλ ðλÞ ¼ I λ, i ðλ; θ; ϕÞ cos ðθÞ sin ðθÞdθdϕ 0 0
0 0 ð4:17Þ
ð4:15Þ
Total radiosity (J, W/m2) is obtained by
2
Total irradiation (G, W/m ) is obtained by integrating Jλ over all wavelengths:
integrating Gλ over all wavelengths:
ð1
ð1 J¼ J λ ðλÞdλ ð4:18Þ
G¼ Gλ ðλÞdλ ð4:16Þ 0
0
Note that Gλ has no subscript i (for incident)

because by definition irradiation is incident on a Surface Properties
surface so the “i” would be redundant. G is the
total radiant heat flux incident to a target. Thermal radiation may be absorbed at, reflected
by, or transmitted through a surface. Imprecisely,
Example 4 A surface is uniformly irradiated with absorptivity (α) is the fraction absorbed at the
a source having the following characteristics: surface, reflectivity (ρ) is the fraction reflected by
the surface, and transmissivity (τ) is the fraction
Gλ ¼ 0 kW=m2 μm for λ 2 μm transmitted through the surface. It follows from a
Gλ ¼ 5 kW=m2 μm for 2 < λ 8 μm radiation balance:
Gλ ¼ 10 kW=m2 μm for 8 < λ 20 μm αþρþτ ¼1 ð4:19Þ

Gλ ¼ 0 kW=m2 μm for λ > 20 μm
where each property in Equation 4.29 may
What is its total irradiation? exhibit spectral and directional characteristics
(but such dependency is not explicitly shown). Eλ ðλ; T Þ

ελ ðλ; T Þ ¼ ð4:21Þ
Additionally, emissivity (ε) is the ratio of the Eλ, b ðλ; T Þ
actual amount of radiation emitted by a surface
to the maximum possible amount of radiation Spectral normal emissivity (very close to
that could be emitted by that surface if it was a hemispherical emissivity) is shown for several
blackbody. These properties are defined more materials in Fig. 4.5 [4].
precisely in the sections that follow. Total emissivity is obtained by integrating
Equation 4.21 over all wavelengths:
Emissivity Eð T Þ
εð T Þ ¼ ð4:22Þ
Eb ð T Þ
Since no surface can emit more thermal radiation
than a blackbody, a logical tool for normalizing By definition, the emissivity of a blackbody
thermal emission from real surfaces is the black- (whether ελ(λ, θ, ϕ, T ), ελ(λ, T), or ε(T )) is
body. Spectral surface emissivity is defined as unity. Total normal emissivity (very close to
the ratio of the actual spectral intensity of radia- hemispherical emissivity) is shown graphically
tion emitted by a surface to the blackbody spec- in Fig. 4.6 for several materials [4]. Representa-
tral intensity: tive values of total hemispherical emissivity are
tabulated for several materials in Table 4.2 [7]
I λ, e ðλ; θ; ϕ; T Þ (metals) and Table 4.3 [7] (non-metals).
ελ ðλ; θ; ϕ; T Þ ¼ ð4:20Þ
I λ, b ð T Þ
As described earlier, hemispherical radiation Absorptivity

quantities are usually applied in engineering
applications. The spectral hemispherical emis- In a fire, one of the most important radiative
sivity is defined in terms of the blackbody emis- characteristics of a material or surface is its
sive power at wavelength λ and is obtained by absorptivity, defined loosely as the fraction of
integrating Equation 4.20 over all directions with the incident radiation that is absorbed by
the result: the material. The absorptivity is strongly
0.8
Spectral normal emissivity (-)
Silicon carbide
(1000 K)
0.6
Aluminum oxide
0.4 (1400 K)
Stainless steel (1200 K)
heavily oxidized
0.2
Stainless steel (800 K)
Tungsten (2800 K) lightly oxidized
Tungsten (1600 K)
0
0.1 1 10 100
Wavelength (mm)
Fig. 4.5 Spectral normal emissivity of several materials [4]
www.ebook777.com
Silicon carbide
0.8
Stainless steel
Total normal emissivity (-)
heavily oxidized
0.6
0.4
Aluminum oxide
Stainless steel
lightly oxidized
0.2 Tungsten
0
300 600 900 1200 1500 1800 2100
Temperature (K)
Fig. 4.6 Total normal emissivity of several materials [4].
Table 4.2 Representative total hemispherical emissivity of several metals [7]

Material Description Emissivity
Aluminum Crude 0.07–0.08 (0–200 C)
Foil, bright 0.01 (9 C), 0.04 (1 C), 0.087 (200 C)
Highly polished 0.04–0.05 (1 C)
Ordinarily rolled 0.035 (100 C), 0.05 (500 C)
Oxidized 0.11 (200 C), 0.19 (600 C)
Roughed 0.044–0.066 (40 C)
Unoxidized 0.022 (25 C), 0.06 (500 C)
Bismuth Unoxidized 0.048 (25 C), 0.061 (100 C)
Brass After rolling 0.06 (30 C)
Browned 0.5 (20–300 C)
Polished 0.03 (300 C)
Chromium Polished 0.07 (150 C)
Unoxidized 0.08 (100 C)
Cobalt Unoxidized 0.13 (500 C), 0.23 (1000 C)
Copper Black oxidized 0.78 (40 C)
Highly polished 0.03 (1 C)
Molten 0.15
Matte 0.22 (40 C)
New 0.07 (40–100 C)
Oxidized 0.56 (40–200 C), 0.61 (200 C), 0.88 (540 C)
Polished 0.04 (40 C), 0.05 (260 C), 0.17 (1100 C)
Rolled 0.64 (40 C)
Gold Polished 0.02 (40 C), 0.03 (1100 C)
Electroytically deposited 0.02 (40 C), 0.03 (1100 C)
(continued)
Table 4.2 (continued)

Material Description Emissivity
Inconel Sandblasted 0.79 (800 C), 0.91 (1150 C)
Stably oxidized 0.69 (300 C), 0.82 (1000 C)
Untreated 0.3 (40–260 C)
Rolled 0.69 (300 C), 0.88 (1150 C)
Inconel X Stably oxidized 0.89 (300 C), 0.93 (1100 C)
Iron Cast 0.21 (40 C)
Cast, freshly turned 0.44 (40 C), 0.7 (1100 C)
Galvanized 0.22–0.28 (0–200 C)
Molten 0.02–0.05 (1100 C)
Plate, rusted red 0.61 (40 C)
Pure polished 0.06 (40 C), 0.13 (540 C)
Red iron oxide 0.96 (40 C), 0.67 (540 C)
Rough ingot 0.95 (1100 C)
Smooth sheet 0.6 (1100 C)
Wrought, polished 0.28 (40–260 C)
Lead Oxidized 0.28 (00–200 C)
Unoxidized 0.05 (100 C)
Magnesium 0.13 (260 C), 0.18 (310 C)
Mercury 0.09 (0 C), 0.12 (100 C)
Molybdenum Oxidized 0.78–0.81 (300–540 C)
Monel Oxidized 0.43 (20 C)
Polished 0.09 (20 C)
Nichrome Rolled 0.36 (800 C), 0.8 (1150 C)
Sandblasted 0.81 (800 C), 0.87 (1150 C)
Nickel Electrolytic 0.04 (40 C), 0.1 (540 C)
Oxidized 0.31–0.39 (40 C), 0.67 (540 C)
Wire 0.1 (260 C), 0.19 (1100 C)
Platinum Oxidized 0.07 (260 C), 0.11 (540 C)
Unoxidized 0.04 (25 C), 0.05 (100 C), 0.15 (1000 C)
Silver Polished 0.01 (40 C), 0.02 (260 C), 0.03 (540 C)
Steel Calorized 0.5–0.56 (40–540 C)
Cold rolled 0.08 (100 C)
Ground sheet 0.61 (1100 C)
Oxidized 0.79 (260–540 C)
Plate, rough 0.94–0.97 (40–540 C)
Polished 0.07 (40 C), 0.1 (260 C), 0.14 (540 C), 0.23 (1100 C)
Rolled sheet 0.66 (40 C)
Type 347, oxidized 0.87–0.91 (300–1100 C)
Type AISI 303, oxidized 0.74–0.87 (300–1100 C)
Type 310, oxidized & rolled 0.56 (800 C), 0.81 (1150 C)
Sandblasted 0.82 (800 C), 0.93 (1150 C)
Stellite 0.18 (20 C)
Tantalum 0.19 (1300 C)
Tin Unoxidized 0.04–0.05 (25–100 C)
Tungsten Filament 0.18 (40 C), 0.11 (540 C), 0.39 (2800 C)
Zinc Oxidized 0.11 (260 C)
Polished 0.02 (40 C), 0.03 (260 C)
www.ebook777.com
Table 4.3 Representative total hemispherical emissivity Material Description Emissivity

of several non-metals [7]
Glass Convex D 0.8–0.76
Material Description Emissivity (100–500 C)
Bricks Chrome refractory 0.94 (540 C), 0.98 Fused quartz 0.75–0.8
(1100 C) (100–500 C)
Fire clay 0.75 (1400 C) Nonex 0.82–0.78
Light buff 0.8 (540 C) (100–500 C)
Magnesite refractory 0.38 (1000 C) Pyrex 0.8–0.9 (40 C)
Sand lime red 0.59 (1400 C) Smooth 0.92–0.95
Silica 0.84 (1400 C) (0–200 C)
Various refractories 0.71–0.88 (1100 C) Waterglass 0.96 (20 C)
White refractory 0.89 (260 C), 0.68 Ice Smooth 0.92 (0 C)
(540 C) Oxides Al2O3 0.35–0.54
Building Asbestos, board 0.96 (40 C) (850–1300 C)
materials Asphalt pavement 0.85–0.93 (40 C) C2O 0.27 (850–1300 C)
Clay 0.39 (20 C) Cr2O3 0.73–0.95
(850–1300 C)
Concrete, rough 0.94 (0–100 C)
Fe2O3 0.57–0.78
Granite 0.44 (40 C)
(850–1300 C)
Gravel 0.28 (40 C)
MgO 0.29–0.5
Gypsum 0.9 (40 C) (850–1300 C)
Marble, polished 0.93 (40 C) NiO 0.52–0.86
Mica 0.75 (40 C) (500–1200 C)
Plaster 0.89 (40 C), 0.48 ZnO 0.3–0.65
(540 C) (850–1300 C)
Quartz 0.76 (40 C) Paints Aluminum 0.27–0.7 (1–100 C)
Sand 0.83 (40 C) Enamel, snow white 0.91 (40 C)
Sandstone 0.83 (40 C) Lacquer 0.85–0.93 (40 C)
Slate 0.67 (40–260 C) Lampblack 0.94–0.97 (40 C)
Carbon Baked 0.52–0.79 Oil 0.89–0.97
(1000–2400 C) (0–200 C)
Filament 0.95 (260 C) White 0.89–0.97 (40 C)
Graphitized 0.76–0.71 Paper White 0.95 (40 C), 0.82
(100–500 C) (540 C)
Rough 0.77 (100–320 C) Roofing Aluminum surfaces 0.22 (40 C)
Soot (candle) 0.95 (120 C) materials Asbestos cement 0.65 (1400 C)
Soot (coal) 0.95 (20 C) Bituminous felt 0.89 (1400 C)
Unoxidized 0.8 (25–500 C) Enameled steel, 0.65 (1400 C)
Ceramics Alumina coating 0.65 (430 C), 0.45 white
on inconel (1100 C) Galvanized iron, 0.90 (1400 C)
Zirconia coating 0.62 (430 C), 0.45 dirty
on inconel (1100 C) Galvanized iron, new 0.42 (1400 C)
Earthenware, glazed 0.9 (1 C) Roofing sheet, brown 0.8 (1400 C)
Earthenware, matte 0.93 (1 C) Roofing sheet, green 0.87 (1400 C)
Procelain 0.92 (40 C) Tiles, uncolored 0.63 (1400 C)
Refractory, black 0.94 (100 C) Tiles, brown 0.87 (1400 C)
Refractory, light buff 0.92 (100 C) Tiles, black 0.94 (1400 C)
Refractory, white 0.9 (100 C) Tiles, asbestos 0.66 (1400 C)
Al2O3 cement
Cloth Cotton 0.77 (20 C) Weathered asphalt 0.88 (1400 C)
Silk 0.78 (20 C) (continued)
Table 4.3 (continued) Spectral hemispherical absorptivity, a

Material Description Emissivity directionally-averaged property that is obtained
Rubber Hard, black, glossy 0.95 (40 C) by integrating over all incident angles, is the
Soft, gray 0.86 (40 C) ratio of the spectral irradiation absorbed by
Snow Fine 0.82 (10 C) the surface (Gλ,abs) to the spectral irradiation of
Frost 0.98 (0 C) the surface Gλ:
Granular 0.89 (10 C)
Soils Black loam 0.66 (20 C) ð 2π ð π=2
Plowed field 0.38 (20 C) Iλ, i, abs ðλ; θ; ϕÞ cos ðθÞ sin ðθÞdθdϕ
Water 0.92–0.96 (0–40 C) αλ ðλÞ ¼ 0ð 2π 0ð π=2
Wood Beech 0.91 (70 C) Iλ, i ðλ; θ; ϕÞ cos ðθÞ sin ðθÞdθdϕ
0 0
Oak, planed 0.91 (40 C)
Sawdust 0.75 (40 C) Gλ, abs ðλÞ
¼
Spruce, sanded 0.82 (100 C) Gλ ðλÞ
ð4:24Þ
wavelength–dependent. For example, at
Finally, total hemispherical absorptivity is
wavelengths below 1 μm the absorptivity of clear
obtained by integrating spectral hemispherical
Polymethylmethacrylate is close to zero, but at
absorptivity over all wavelengths:
wavelengths above 3 μm it approaches unity.
A blackbody absorbs all incident radiation ð1 ð1
with no spectral or directional dependency. As Gλ, abs ðλÞdλ αλ ðλÞGλ ðλÞdλ
with emissivity, the idealized blackbody behav- α ¼ 0ð 1 ¼ 0 ð1
ior is used as a normalization tool to quantify Gλ ðλÞdλ Gλ ðλÞdλ
the amount of radiation absorbed by a surface 0 0
relative to the maximum possible amount the Gabs

¼
surface may absorb (i.e., if it was a blackbody). G
Spectral radiant intensity incident on a surface ð4:25Þ
is denoted Iλ,i; it is, in general, a function of λ, θ,
and ϕ. Spectral, directional absorptivity is the Example 5 A particular diffuse material is
ratio if the spectral directional radiant intensity idealized as having a spectral absorptivity of
absorbed by a surface Iλ,i,abs(λ, θ, ϕ) to the zero for wavelengths less than 3 μm and unity
spectral directional radiant intensity incident for wavelengths greater than 3 μm. Calculate its
on that surface Iλ,i(λ, θ, ϕ) (because the latter is total hemispherical absorptivity for a blackbody
the maximum possible radiation that could at 800, 1200, and 2000 K.
be absorbed by that surface, i.e. if it was a
blackbody): Solution Assume Gλ(λ) ¼ Eλ,b(λ,Τe) where Te is
the temperature of the emitter (800, 1200, and
I λ, i, abs ðλ; θ; ϕÞ 2000 K) and use Equation 4.25:
αλ ðλ; θ; ϕÞ ¼ ð4:23Þ
I λ, i ðλ; θ; ϕÞ
ð1 ð3 ð1 ð1
αλ ðλÞEλ, b ðλ; T e Þdλ 0 Eλ, b ðλ; T e Þdλ þ 1 Eλ, b ðλ; T e Þdλ Eλ, b ðλ; T e Þdλ
αðT e Þ ¼ 0 ð1 ¼ 0 3
¼ 3
σT 4e σT 4e
Eλ, b ðλ; T e Þdλ
0
www.ebook777.com
It is possible to put this in a form that allows use several solids and then determined the integrated
of the radiation fraction tabulated in Table 4.1: surface absorptivity of different solids irradiated
ð1 ð3 by hexane flames, blackbodies between 1000 and
3500 K, and solar energy. His absorptivity data
Eλ, b ðλ; T e Þdλ Eλ, b ðλ; T e Þdλ
are reproduced in Table 4.4. Note that the total
α¼ 0 0
σT 4e hemispherical absorptivity of some materials is
ð3 relatively insensitive to the temperature of the
σT 4e Eλ, b ðλ; T e Þdλ radiation source (black PMMA) but others are
¼ 0 quite sensitive. For example, the absorptivity of
σT 4e clear PMMA decreases from 0.85 for a 1000 K
ð3
blackbody to 0.25 for a 3500 K blackbody.
Eλ, b ðλ; T e Þdλ Similar measurements were made by Wesson
¼1 0
¼ 1 F0!3 ð3T e Þ et al. [11] for undegraded wood. Their results are
σT 4e
reproduced in Table 4.5. More recently, Försth
For Te ¼ 800 K, λTe ¼ 2400 μm K and from and Roos [12] conducted similar measurements
Table 4.1 F0!3 ¼ 0:14 so α ¼ 0.86. for wood products (Table 4.6), carpet (Table 4.7),
For Te ¼ 1200 K, λTe ¼ 3600 μm K and from painted plywood (Table 4.8), and plastics
Table 4.1 F0!3 ¼ 0:40 so α ¼ 0.60. (Table 4.9).
For Te ¼ 2000 K, λTe ¼ 6000 μm K and from During a fire, a material’s radiative
Table 4.1 F0!3 ¼ 0:76 so α ¼ 0.24. characteristics may change. Although the
It is seen that, for this idealized material, the integrated absorptivities from Wesson et al. [11]
effective absorptivity is a strong function of (reproduced in Table 4.5) are relatively low, the
emitter temperature. absorptivity of charred wood is generally not the
In a fire we are usually interested in the total same as that of virgin wood. Janssens [13]
hemispherical absorptivity defined in Equa- suggested that blackening causes the absorptivity
tion 4.25. However, as demonstrated above, the of wood to increase from ~0.76 (based on Refer-
total hemispherical absorptivity depends on the ence [11]) to approximately unity as the surface
spectral energy distribution of the radiation temperature approaches the ignition temperature.
source. Therefore, a material technically cannot He therefore used an average value of 0.88 in his
be assigned a single absorptivity value because ignition analyses, and recommends using an
the spectral distribution of the incoming radia- integrated absorptivity of 1.0 during flaming
tion depends on the temperature of the emitter. combustion [14]. Interestingly, Försth and Roos
Due to Wien’s displacement law and the Planck [12] noted the opposite trend, i.e. a reduction in
distribution, this is true even if the emitter effective absorptivity as wood darkens. More
behaves as a blackbody. In fires, the temperature research is needed in this area.
of radiation sources ranges from approximately Wood is not the only class of materials that
~600 K (smoke layer, hot surfaces) to ~2000 K exhibits a change in radiative characteristics dur-
(flames). Additionally, certain bench–scale fire ing a fire. Under nonflaming conditions, low
tests use tungsten–filament heaters that operate density polyethylene has been observed to
at temperatures near 3000 K. Thus, the effect of change from visually opaque to transparent,
source temperature on the integrated eventually followed by a darkening of the surface
(or effective) absorptivity has relevance for [15]. This indicates that a change in the
both real fires and bench-scale fire testing. material’s radiative characteristics occurred
Hallman’s 1971 Ph.D. dissertation [8] and (at least in the visible range). Modak and Croce
subsequent publications [9, 10] remain some of [16] reported that for clear PMMA, 39 % of
the most comprehensive sources of information flame radiation is transmitted through the sur-
on the change of polymers’ total hemispherical face, but for “charred” PMMA (previously
absorptivity with the temperature of the emitter. exposed to a fire environment and then cooled)
Hallman measured the spectral absorptivity of no radiation penetrates in depth. Bubbling
Table 4.4 Integrated surface absorptivities for polymers from Hallman [9]
Blackbody emitter temperature (K)

Generic name Trade name 1000 1500 2000 2500 3000 3500 Flame
Acrylonitrile butadiene styrene Cycolac® 0.91 0.86 0.77 0.71 0.65 0.61 0.92
Cellulose acetate butyrate Uvex® 0.84 0.71 0.56 0.43 0.34 0.27 0.88
Cork 0.64 0.56 0.49 0.46 0.44 0.44 0.60
Melamine/formaldehyde Formica® 0.91 0.88 0.85 0.82 0.80 0.79 0.91
Nylon 6/6 0.93 0.90 0.86 0.82 0.75 0.71 0.93
Phenolic Bakelite 0.90 0.86 0.81 0.77 0.75 0.75 0.91
Polycarbonate (rough surface) Lexan® 0.87 0.83 0.78 0.75 0.72 0.71 0.88
Polyethylene (low density) 0.92 0.88 0.82 0.77 0.72 0.68 0.93
Polymethylmethacrylate (black) Plexiglas® 0.94 0.94 0.95 0.95 0.95 0.95 0.94
Polymethylmethacrylate (clear) Plexiglas® 0.85 0.69 0.54 0.41 0.31 0.25 0.89
Polymethylmethacrylate (white) Plexiglas® 0.91 0.86 0.78 0.70 0.62 0.56 0.92
Polyoxymethylene Delrin® 0.92 0.86 0.78 0.71 0.64 0.59 0.93
Polyphenylene oxide 0.86 0.78 0.70 0.63 0.57 0.53 0.88
Polypropylene 0.87 0.83 0.78 0.74 0.70 0.68 0.86
Polystyrene (clear) Styrolux® 0.75 0.60 0.46 0.35 0.28 0.22 0.78
Polystyrene (white) 0.86 0.75 0.63 0.53 0.45 0.40 0.88
Polyurethane thermoplastic Texin® 0.92 0.89 0.83 0.77 0.72 0.68 0.93
Polyvinyl chloride (clear) 0.81 0.65 0.49 0.38 0.30 0.24 0.85
Polyvinyl chloride (gray) 0.90 0.90 0.89 0.89 0.89 0.89 0.91
PVC/acrylic (gray, rolled) Kydex® 0.88 0.87 0.86 0.85 0.84 0.83 0.88
PVC/acrylic (red cast) Kydex® 0.91 0.90 0.89 0.88 0.87 0.86 0.92
Rubber (Buna–N) 0.92 0.93 0.93 0.93 0.93 0.93 0.92
Rubber (Butyl IIR) 0.92 0.93 0.94 0.94 0.95 0.95 0.92
Rubber (natural, gum) 0.88 0.82 0.76 0.72 0.69 0.68 0.89
Rubber (neoprene) 0.91 0.92 0.93 0.93 0.93 0.93 0.91
Rubber (silicone) 0.79 0.66 0.58 0.54 0.52 0.53 0.79
occurring near the surface of polymers can

Table 4.5 Integrated surface absorptivity for wood from
different emitters (From Wesson et al. [11]) change their radiative characteristics, but this
effect is has not yet been reliably quantified. In
Flame Tungsten Solar
Wood radiation lamp radiation radiation a real fire, materials may become coated in soot
Alaskan cedar 0.76 0.44 0.36 from flames or a smoke layer, causing their
Ash 0.76 0.46 0.36 absorptivities to approach unity.
Balsa 0.75 0.41 0.35
Birch 0.77 0.47 0.39
Cottonwood 0.76 0.48 0.40 Reflectivity
Mahogany 0.76 0.49 0.52
Mansonia 0.76 0.47 0.51 A fraction of radiation incident on a surface may
Maple 0.76 0.49 0.44 be reflected. One complicating factor is that
Oak 0.77 0.56 0.49 reflection may be diffuse, specular, or (most
Redgum 0.77 0.52 0.56 likely) some combination of these two
Redwood 0.77 0.51 0.55 idealizations. A diffuse reflector is a surface for
Spruce 0.76 0.45 0.35 which, analogous to a diffuse emitter, the inten-
White pine 0.76 0.49 0.43 sity of reflected radiation is equal in all directions
Masonite 0.75 0.52 0.61
and does not depend on the angle of incoming
www.ebook777.com
Table 4.6 Effective absorptivities for different grey body temperatures for various wood products (From Försth and
Roos [12])
Grey body emitter T (K) 674 852 1025 1153 1300 5777
Cone calorimeter irradiation (kWm2) 10 25 50 75 100 Sun
Product αeff
Plywood 0.86 0.84 0.81 0.79 0.76 0.40
Dark heat-treated lacquered ash tree floor 0.89 0.88 0.85 0.83 0.80 0.63
Dark heat-treated non-lacquered ash tree floor 0.83 0.81 0.79 0.77 0.74 0.62
Light lacquered ash tree flooring 0.90 0.88 0.86 0.84 0.82 0.40
Light non-lacquered oak flooring 0.86 0.84 0.81 0.80 0.77 0.37
Medium dark lacquered oak flooring 0.91 0.89 0.87 0.85 0.83 0.56
Medium dark non-lacquered oak flooring 0.86 0.84 0.82 0.80 0.77 0.50
Table 4.7 Effective absorptivities for different grey body temperatures for various carpets (From Försth and Roos
[12])
Product αeff
Beige PVC carpet 0.92 0.91 0.90 0.89 0.87 0.60
Pink PVC carpet 0.90 0.88 0.86 0.84 0.81 0.39
Red PVC carpet 0.92 0.92 0.91 0.90 0.89 0.80
Blue PVC carpet 0.89 0.87 0.85 0.83 0.80 0.43
Grey PVC carpet 0.90 0.88 0.86 0.84 0.82 0.43
Black PVC carpet 0.93 0.93 0.93 0.93 0.93 0.92
Grey rubber mat 0.91 0.91 0.91 0.91 0.91 0.82
Black rubber mat 0.90 0.90 0.91 0.91 0.91 0.95
White vinyl carpet 0.88 0.86 0.83 0.81 0.79 0.44
Beige vinyl carpet 0.91 0.90 0.89 0.87 0.85 0.51
Brown vinyl carpet 0.90 0.89 0.88 0.88 0.87 0.77
Grey vinyl carpet 0.92 0.91 0.89 0.88 0.87 0.57
Black vinyl carpet 0.93 0.93 0.93 0.93 0.93 0.94
Beige linoleum carpet 0.92 0.91 0.89 0.88 0.86 0.55
radiation. This contrasts to a specular emitter spectral directional reflectivity is defined as the
which is an idealized surface where the angle of ratio of the reflected spectral radiant intensity to
reflected radiation is equal to the angle of inci- the incident spectral radiant intensity:
dent radiation, like a billiard ball bouncing off
the rail. Rough surfaces approximate diffuse I λ, i, re f ðλ; θ; ϕÞ
ρλ ðλ; θ; ϕÞ ¼ ð4:26Þ
emitters, and polished surfaces are close to spec- I λ, i ðλ; θ; ϕÞ
ular surfaces. It is seen that, in its most general
form, surface reflection is a bidirectional process note that in Equation 4.26, θ and ϕ refer to the
meaning the intensity of reflected radiation direction of the incident radiation, not the reflected
depends not only on the angle of incident radiation (since, for simplification, no consider-
radiation, but also on the angle of reflected radiation is given to the direction of reflected radia-
ation. As a simplification, we look only at tion). Spectral hemispherical reflectivity is
hemispherically-integrated reflection. Then the obtained by integrating over all incident angles:
Table 4.8 Effective absorptivities for different grey body temperatures for various paints painted on plywood (From
Försth and Roos [12])
Product αeff
White ceiling water paint 0.86 0.83 0.81 0.78 0.75 0.30
White floor water paint 0.86 0.84 0.81 0.78 0.75 0.24
Mid gray floor water paint 0.90 0.89 0.89 0.89 0.88 0.76
White priming water paint 0.86 0.83 0.81 0.78 0.75 0.27
Red priming water paint 0.90 0.89 0.87 0.86 0.83 0.71
Red priming water paint 0.89 0.87 0.85 0.83 0.80 0.70
White top water paint 0.87 0.84 0.81 0.78 0.75 0.25
Yellow top water paint 0.90 0.88 0.86 0.84 0.82 0.44
Red top water paint 0.89 0.88 0.86 0.84 0.81 0.55
Blue top water paint 0.89 0.87 0.85 0.83 0.81 0.73
White wall water paint 0.84 0.81 0.77 0.74 0.71 0.23
Black wall water paint 0.93 0.93 0.93 0.93 0.93 0.95
Blue wall water paint 0.92 0.91 0.91 0.90 0.90 0.75
White lacquer paint 0.86 0.84 0.81 0.79 0.76 0.26
Blue lacquer paint 0.90 0.89 0.88 0.88 0.87 0.74
Black lacquer paint 0.92 0.92 0.92 0.92 0.93 0.95
Red ceiling lacquer paint 0.87 0.84 0.82 0.79 0.77 0.70
Black ceiling lacquer paint 0.92 0.93 0.93 0.93 0.93 0.95
Table 4.9 Effective absorptivities for different grey body temperatures for various plastics and other materials (From
Försth and Roos [12])
Product αeff
White ABS 0.91 0.90 0.88 0.86 0.84 0.31
Black ABS 0.92 0.92 0.92 0.92 0.92 0.94
Nature acetal 0.93 0.93 0.93 0.92 0.91 0.46
Nature PA-6 0.93 0.93 0.94 0.94 0.94 0.74
Clear PC 0.92 0.91 0.89 0.87 0.85 0.24
Clear PC Ultra UV 0.92 0.91 0.90 0.88 0.85 0.46
Brown PC 0.93 0.93 0.93 0.93 0.93 0.93
Nature PE 0.93 0.93 0.93 0.93 0.93 0.62
Yellow PE 0.93 0.92 0.92 0.91 0.90 0.53
Black PE 0.93 0.93 0.93 0.93 0.93 0.95
Clear PMMA G 0.93 0.92 0.91 0.89 0.87 0.23
Yellow PMMA G 0.94 0.93 0.93 0.93 0.92 0.62
Brown PMMA G 0.94 0.94 0.94 0.94 0.94 0.94
Clear PMMA XT 0.93 0.92 0.91 0.89 0.87 0.21
Grey PP 0.92 0.92 0.91 0.91 0.90 0.66
Nature PTFE 0.84 0.78 0.73 0.70 0.66 0.10
Clear PVC 0.91 0.90 0.88 0.86 0.84 0.27
White PVC 0.91 0.89 0.87 0.85 0.82 0.31
White PVC expostandard (foamed) 0.82 0.80 0.78 0.76 0.73 0.31
Grey PVC 0.91 0.90 0.90 0.90 0.90 0.88
Black PVC 0.93 0.93 0.93 0.93 0.93 0.95
Nature PVDF 0.94 0.94 0.94 0.94 0.94 0.78
www.ebook777.com
ð 2π ð π=2
Kirchhoff’s Law: Relation Between
Iλ, i, re f ðλ; θ; ϕÞ cos ðθÞ sin ðθÞdθdϕ
Emissivity and Absorptivity
ρλ ðλÞ ¼ 0ð 2π 0ð π=2
Iλ, i ðλ; θ; ϕÞ cos ðθÞ sin ðθÞdθdϕ
0 0
Kirchoff’s law is used extensively in radiation
Gλ, re f ðλÞ heat transfer calculations. In its most general
¼
Gλ ð λ Þ form, Kirchoff’s law states that in order to main-
ð4:27Þ tain thermal equilibrium, the spectral directional
absorptivity must be equal to the spectral direc-
Final, total hemispherical reflectivity is obtained tional emissivity:
by integrating over all wavelengths:
ð1 αλ ðλ; θ; ϕÞ ¼ ελ ðλ; θ; ϕ; T Þ ð4:32Þ
Gλ, re f ðλÞ
Gre f
ρ ¼ 0ð 1 ¼ ð4:28Þ
G Using the relations presented earlier in the chap-
Gλ ðλÞ
0
ter, it can be shown that if the irradiation is
diffuse or the surface is diffuse, then Kirchoff’s
law has no directional dependency, i.e.:
Transmissivity
αλ ðλÞ ¼ ελ ðλ; T Þ ð4:33Þ
Directional spectral transmissivity is defined in
an analogous manner to the other radiation
If Equation 4.33 applies (i.e., the irradiation is
properties discussed here:
diffuse or the surface is diffuse), then if the
surface is also gray (meaning αλ and ελ are
I λ, i, trans ðλ; θ; ϕÞ
τλ ðλ; θ; ϕÞ ¼ ð4:29Þ invariant with λ) or the surface is irradiated
I λ, i ðλ; θ; ϕÞ
only by radiation emitted from a blackbody at
the same temperature as the surface, its total
Hemispherical spectral transmissivity is:
absorptivity is equal to its total emissivity:
ð 2π ð π=2
I λ, i, trans ðλ; θ; ϕÞ cos ðθÞ sin ðθÞdθdϕ α¼ε ð4:34Þ
τ λ ðλ Þ ¼ 0 0
ð 2π ð π=2
Iλ, i ðλ; θ; ϕÞ cos ðθÞ sin ðθÞdθdϕ For engineering calculations, Equation 4.34 is
0 0
Gλ, trans ðλÞ most commonly applied for the special case of
¼
G λ ðλ Þ diffuse and gray surfaces. Fortunately, this is a
ð4:30Þ reasonable approximation for many radiation
heat transfer engineering models for
And total transmissivity is then obtained by participating media in fire applications.
integrating over all wavelengths: Although real surfaces may exhibit an emis-
sivity that varies with wavelength (see Fig. 4.7),
ð1
an effective emissivity can be selected so that
Gλ, trans ðλÞ the integrated emissive power of the gray surface
0ð Gtrans
τ¼ 1 ¼ ð4:31Þ matches the integrated emissive power of the real
G
G λ ðλÞ surface at a particular temperature.
0
4500
Blackbody
4000
Spectral emissive power (Ebλ, W/m2•μm)
3500
Gray body, ε = 0.8
3000
Gray body, ε = 0.6
2500
2000
1500
Real surface
1000
500
0
0 5 10 15
Wavelength (l, mm)
Fig. 4.7 Monochromatic emissive power for black body, two gray bodies, and real surface
Radiant Heat Transfer A2

in Nonparticipating Media dA2
In this section, cases are examined where the

surfaces are separated by a medium that does not
b2
emit, absorb, or scatter radiation. A vacuum meets
this requirement exactly, and common diatomic n2
gases of symmetric molecular structure such as R

N2, O2, and H2 are very nearly nonparticipating
media within the thermal radiation spectrum. The
radiative energy transfer between the surfaces
depends on the geometry, orientation, tempera- n1 b1
ture, and radiation properties of the surfaces. In
practice, surfaces are usually idealized as
isothermal, diffuse, and gray to make engineering
A1
calculations tractable. The geometry and orienta-
dA1
tion of each surface is commonly accounted for in
calculations by one or more configuration factors,
also known as view factors, shape factors, angle Fig. 4.8 Coordinate system for shape factors
factors, and geometric factors [1–7, 17–19].
defined as the fraction of radiation leaving one

View Factors surface which is intercepted by the other surface.
Consider the two arbitrarily oriented surfaces
A view factor, or configuration factor, is a purely A1 and A2 in Fig. 4.8. Assuming that the radiosity
geometrical relation between two surfaces. It is from differential area dA1 is diffuse, the
www.ebook777.com
X
configuration factor from dA1 to the finite area Fi j ¼ 1 ð4:38Þ
A2, Fd1 2, is given by j
ð where Fij relate to surfaces that subtend a

cos ðβ1 Þ cos ðβ2 Þ
Fd12 ¼ !2 dA2 ð4:35Þ closed system. It is possible for a concave surface

A2
πR to “see” itself, which can make Fii important in
certain situations.
where the separation distance between the two In many cases, it is advantageous to define a
!
single surface ( j) as a composite surface
surfaces is R , β is the angle between the line of
!
consisting of multiple (real or imaginary)
!
sight R and the surface normal n , and A2 is the surfaces (k), i.e.:
area of surface 2. If the radiosity from all of
X
surface A1 (not just differential are dA1) is dif- Aj ¼ Ak ð4:39Þ
fuse, then the configuration factor for the finite
area A1 to A2, F1 2, is calculated as: For a composite surface j, made up of multiple
surfaces k, since view factors are additive:
ð ð
1 cos ðβ1 Þ cos ðβ2 Þ
F12 ¼ !2 dA1 dA2 X
A1 Fi j ¼ Aik ð4:40Þ
A1 A2
πR
ð4:36Þ Example 6 For the geometry shown below,
use shape factor algebra to develop an expres-
All configuration factors can be derived using the sion for the view factor between surface
multiple integration of Equations 4.35 and 4.36, 1 and surface 4 that could be evaluated from
but this is generally very tedious except for the shape factor relations provided in Appen-
simple geometries. Several cases have been dix D.
tabulated with the numerical results or algebraic
formulas available in various references [1–7, 17,
18]. Several configuration factors are provided in
Appendix D.
4
The configuration factors in Appendix D can
be extended to other geometries by using config- 3
uration factor algebra and the method of surface
2 1
decomposition. In surface decomposition,
unknown factors can be determined from
known factors for convenient areas or for imagi-
nary surfaces which can extend real surfaces or Solution The desired view factor is F14. For
form an enclosure [1, 6]. simplicity of nomenclature, denote surface A as
When the radiant fluxes from both surfaces a composite surface made up of surfaces 1 and
are uniformly and diffusely distributed 2. Similarly, denote surface B as a composite
(a common engineering assumption), a reciproc- surface made up of surfaces 3 and 4. Then,
ity relation for any given pair of configuration from the additive property of shape factors:
factors in a group of exchanging surfaces is:
AA FAB ¼ A1 F1B þ A2 F2B
Ai Fi j ¼ A j F ji ð4:37Þ
Both FAB and F2B can be calculated from the
The summation rule is another useful relation appropriate shape factor in Appendix D. Note
for calculating unknown configuration factors that
F1B ¼ F13 þ F14 A convenient method to analyze radiative

energy exchange in a diffuse gray enclosure
F13 ¼ FA3 F23 relies on the concepts of radiosity and irradiation
introduced earlier. The irradiation of surface
Combining these two equations gives an expres-
i (Gi) is the radiative flux reaching the ith surface
sion for F1B:
regardless of its origin:
F1B ¼ FA3 F23 þ F14 X
Gi ¼ Fi j J j ð4:41Þ
j
Substituting this expression into the first equation
above:
where Jj is the surface radiosity, defined as the
total radiative flux leaving the jth surface includ-
AA FAB ¼ A1 ðFA3 F23 þ F14 Þ þ A2 F2B ing both emitted and reflected radiation:
¼ A1 FA3 A1 F23 þ A1 F14 þ A2 F2B
J i ¼ Ei þ ρi Gi ¼ εi Ebi þ ρi Gi ð4:42Þ
Solving for F14:
The net rate at which radiation leaves surface i is
1
F14 ¼ ðAA FAB A2 F2B A1 FA3 þ A1 F23 Þ given by
A1
1 Qi ¼ Ai ðJ i Gi Þ ¼ Ai ðEi þ ρi Gi Gi Þ
¼ ðAA FAB A2 F2B Þ þ F23 FA3
A1 ¼ Ai ðEi Gi ð1 ρi ÞÞ
¼ Ai ðEi αi Gi Þ ð4:43Þ
Note that all of the view factors in the above
example can be evaluated from the shape factor since, for a diffuse gray opaque surface ρi ¼ 1
relations provided in Appendix D. αi. It must be emphasized that the radiosity-
irradiation formulation is based on the assump-
tion that each surface has uniform radiosity and
Gray Diffuse Surfaces irradiation (or equivalently, uniform temperature
and uniform heat flux). Physically unrealistic
For engineering applications, thermal emission calculations can result if each surface does not
from most surfaces is treated as having diffuse approximately satisfy this condition. Larger
directional characteristics independent of wave- surfaces should be subdivided into smaller
length and temperature. Real surfaces exhibit surfaces if necessary.
radiation properties that are so complex that The radiosity-irradiation formulation allows a
information about these property measurements more physical and graphic interpretation using
for many common materials is not available. The the resistance network analogy. Eliminating the
gray diffuse surface is a useful model that irradiation Gi from Equations 4.41, 4.42 and
alleviates many of the complexities associated 4.43, and substituting ρi ¼ 1 εi gives
with a detailed radiation analysis, while
providing reasonably accurate results in many Ebi J i X Ji J j
practical situations. The advantage of diffuse Qi ¼ ¼ 1 ð4:44Þ
ð1 εi Þ=ðεi Ai Þ j Ai Fi j
surface analysis is that radiation leaving the sur-
face is independent of the direction of the incom- Note that the second equality in Equation 4.44
ing radiation, which greatly reduces the amount can be written as:
of computation required to solve the governing
equations. Discussions for specularly reflecting X X
Qi ¼ Ai Fi j J i J j ¼ Qi j ð4:45Þ
surfaces and nongray surfaces can be found in the j j
literature [1, 6].
www.ebook777.com
Fig. 4.9 Network analogy Eb2

for radiative exchange 1
———
A2F23
1– ∈2
Refractory
wall ∈2 A2 J2 Eb3
1 1
——— ———
A1F12 A1F13
3
2 J1
1 – ∈1
1 ∈1 A1
Eb1
The denominator in the rightmost term of Equa- and scattering within the medium. The intensity,
tion 4.44 corresponds to resistance in electric Iλ(S), is coupled with the spatial distribution of
circuits. This electrical resistance analogy was the extinction coefficient and with temperature
first proposed by Oppenheim [20]. As illustrated through conservation of energy in the medium.
in Fig. 4.9, the diffuse-gray surface has a radia- The contributions of intensity passing through an
tion potential difference (Ebi Ji) and a resis- area must be integrated over all directions to
tance (1 εi)/εiAi. This example also illustrates calculate a net radiative energy flux. The integral
that an adiabatic surface, such as a reradiating or nature of radiation makes analysis difficult
refractory wall, exhibits a surface temperature and simplifications necessary for engineering
that is independent of the surface emissivity or practice.
reflectivity.
Spectral Emissivity and Absorptivity

Thermal Radiation in Participating
Media From a microscopic viewpoint, emission and
absorption of radiation is attributed to changes
The Equation of Transfer in energy levels of atoms and molecules caused
by interactions with photons. Tien [21] discusses
The equation of transfer describes the variation these effects in gases from an engineering
in intensity of a radiant beam at any position perspective.
along its path in an absorbing-emitting-scattering Consider a monochromatic beam of radiation
medium. This equation is the foundation upon passing through a radiating medium of thickness
which detailed radiation analyses are based, and L. For the special case where the temperature and
the source of approximate solutions when properties of the medium are uniform along this
simplifying assumptions are made. For a given path, the intensity of radiant beam at point x is
direction line in the medium, the equation of obtained by integrating Equation 4.46:
transfer is
I λ ðxÞ ¼ I λ ð0Þexpðκλ xÞ
1 dI λ ðSÞ þ I λ, b ð1 expðκ λ xÞÞ ð4:47Þ
þ I λ ðSÞ ¼ I λ, b ðT Þ ð4:46Þ
κλ ðT; SÞ dS
which accounts for the loss of intensity by
where S represents the physical pathlength and κλ absorption and the gain by emission, and where
represents the spectral extinction coefficient, κλ denotes the extinction coefficient. The extinc-
which includes the effects of both absorption tion coefficient is generally the sum of two parts:
the absorption coefficient and the scattering terms of mean (gray-gas) absorption coefficients
coefficient. In many engineering applications, representing average properties over the whole
the effects of scattering are negligible and the spectrum of wavelengths. The appropriate mean
extinction coefficient represents only absorption. absorption coefficients are the Planck mean, κP,
The spectral emissivity for pathlength S in a for optically thin media, and the Rosseland mean,
uniform gas volume can be readily expressed κR, for optically thick media [5, 6, 21].
by considering the case of no incident radiation The Planck mean absorption coefficient is
(or Iλ(0) ¼ 0): defined as
ð1
Iλ
ελ ¼ ¼ 1 expðκ λ SÞ ð4:48Þ I λ, b κ λ dλ ð
I λ, b π 1
ð
κP 1
0
¼ 4 I bλ κ λ dλ ð4:50Þ
σT 0
which compares the fraction of energy emitted to I λ, b dλ
0
the maximum (blackbody) emission at the same
temperature for the pathlength S through the This form of the absorption coefficient is a
material. function of temperature alone and is independent
The term κ λS in Equation 4.48 is called optical of pressure. The effect of the beam source temper-
pathlength or opacity and is denoted τλ (not to be ature (e.g., a hot or cold wall) in the gas absorptiv-
confused with transmissivity). It can be defined ity is approximated by a ratio correction [21, 22]
more generally for nonhomogeneous media as:
ðS Ts
κm ¼ κP ðT s Þ ð4:51Þ
τλ ¼ κλ ðxÞdx ð4:49Þ Tg
0
where Ts is the source temperature and Tg is the
If τλ << 1, the medium is optically thin at wave- gas temperature. When the Planck mean absorp-
length λ and the properties of the participating tion coefficient is used to estimate the emissivity
medium can generally approximated as ελ τλ . of a gas, the source temperature is set equal to the
The medium is considered optically thick when gas temperature.
τλ >> 1, which implies that the mean penetra- The formulation of radiative transfer is
tion distance is much less than the characteristic simplified when the medium is optically thick.
length of the medium. In optically thick media, In this case, the radiative transfer can be regarded
as will be described below, the local radiant as a diffusion process (the Rosseland or diffusion
intensity results only from local emission and approximation), and the governing equation is
the equation of transfer can be approximated by approximated by:
a diffusion equation.

00 4 1 ∂Eλ, b 4 1 ∂ σT 4
qr ¼
3 κ R ∂x 3 κ R ∂x
Planck and Rosseland Mean Absorption
Coefficients 16σ 3 1 ∂T
¼ T ð4:52Þ
3 κ R ∂x
The mean absorption coefficient is often useful
when radiative energy transport theory must Evaluation of the total heat flux in an optically
be used to describe the local state of a gas at thick medium is simplified by defining the
various locations. The mathematical complexity Rosseland mean absorption coefficient which is
involved in the calculations often dictates a solu- independent of wavelength:
tion based on the gray-gas assumption, where all ð1
radiation parameters are considered to be wave- 1 1 ∂Eλ, b
dλ ð4:53Þ
length independent. Thus solutions are given in κR 0 κλ ∂Eb
www.ebook777.com
In Equation 4.53, ∂Eλ, b =∂Eb is evaluated from the which in turn varies with solid angle Ω according
Planck distribution after setting T ¼ ðEb =σ Þ . 1=4 to the gas body geometry. In practical situations,
00
The Rosseland mean absorption coefficient is not the calculation of q_ r is more convenient in terms
well defined for gases under ordinary conditions of total emissivity, which is often available in
because astronomically long pathlengths are chart form. From the definition of total emissiv-
required to make the windows between the ity, Equation 4.54 can be expressed as:
bands optically thick. However, the Rosseland
ð
limit is useful when dealing with gases in the 00 σT 4
q_ r ¼ εðXÞ cos ðθÞdΩ σT 4 εðLÞ ð4:56Þ
presence of soot particles, which are characterized π Ω
by a continuous spectrum. The source temperature
effect is accounted for by using Equation 4.51 which gives the definition of mean beam length,
in the same manner as for the Planck mean absorp- L, for a gas body, where ε(L ) has the same
tion coefficient. functional form as ε(X). Physically, the mean
The radiating gas in many actual fire systems beam length represents the equivalent radius of
is neither optically thin nor optically thick, so a hemispherical gas body such that it radiates a
it may be necessary to use band theory to flux to the center of its base equal to the average
rigorously calculate a mean absorption coeffi- flux radiated to the boundary surface by the
cient, κ m. However, with a reasonable estimate actual volume of gas. The determination of the
of the mean absorption coefficient, radiative mean beam length is simplified when the gas is
transport calculations are much more convenient. optically thin and only the geometry of the gas
body enters the calculation. In the optically thin
limit, it is convenient to define
Mean Beam Length for Homogeneous
Gas Bodies ð
1
L ¼ L0 X cos ðθÞdΩ ð4:57Þ
π Ω
The concept of mean beam length is a powerful
and convenient tool to calculate the energy flux where L0 is called the geometric mean beam
from a radiating homogeneous gas volume to its length. In the optically thick limit, a correction
boundary surface. It may also be used to approx- factor (C) can be used to obtain reasonable radi-
imate radiative energy flux for a nonhomoge- ative heat flux estimates:
neous gas, especially when more elaborate
calculations are not feasible. Consider the coor- L CL0 ð4:58Þ
dinate system given in Fig. 4.3, where dA is a
differential area on the boundary surface of the In Table 4.10, L0 and C have been provided
gas body. The radiative heat flux from the gas for a variety of gas body shapes. For an arbi-
body to dA is trarily shaped gas volume, the geometric beam
ð1 ð length from the gas volume to the entire bound-
00
q_ r ¼ ελ ðXÞI λ, b cos ðθÞdλdΩ ð4:54Þ ary surface can be estimated by:
0 Ω
4V
where the spectral emissivity, ελ, is a function of L0 ¼ ð4:59Þ
A
pressure pathlength:
ðS where V and A are the volume and the area of the
boundary surface of the gas body, respectively.
X Pa xðξÞdξ ð4:55Þ
0 The correction factor C is approximately 0.9.
Table 4.10 Mean beam lengths for various gas body shapes
Geometric mean Correction
Geometry of gas body Radiating to beam length L0 factor C
Sphere Entire surface 0.66 D 0.97
Cylinder Plane and surface 0.48 D 0.90

H ¼ 0.5 D Concave surface 0.52 D 0.88
Entire surface 0.50 D 0.90
Cylinder Center of base 0.77 D 0.92
H¼D Entire surface 0.66 D 0.90
Cylinder Plane end surface 0.73 D 0.82

H¼2D Concave surface 0.82 D 0.93
Semi-infinite Center of base 1.00 D 0.90

cylinder Entire base 0.81 D 0.80
H!1
Infinite slab Surface element 2.00 D 0.90

Both bounding planes 2.00 D 0.90
Cube 1 4 face 0.90 D 0.91

D D 4D 1 1 face 0.86 D 0.83
heat flux is a concern, gas properties may be

Thermal Radiation Properties calculated using the exponential wide-band
of Combustion Products model [23]. The uncertainties involved in
estimating parameters to calculate radiative heat
Radiation Properties of Gases flux make average properties such as total emis-
sivity a useful tool. The first comprehensive total
The emissivity of any gas is a strong function of emissivity charts were formulated by Hottel and
wavelength, varying by as much as several orders coworkers to summarize work performed up to
of magnitude over small changes in wavelength. about 1945. Modern formulations for the emissiv-
However, the level of accuracy required in engi- ity of gases have been summarized by
neering calculations, where many of the Edwards [22].
parameters are difficult to measure or estimate, Total emissivity charts for water vapor and
seldom requires high resolution emissivity spec- carbon dioxide [22] are provided in Figs. 4.10
tra. Where wavelength dependence of radiative and 4.11, respectively. Gas emissivity can be
www.ebook777.com
1.0 1.0
PaL (atm – m)
PaL (atm – m)
4.0
2.0
1.0
0.1 0.1 4.0

0.4
2.0
1.0
Gas emissivity, ∈t
Gas emissivity, ∈t
0.2 0.4
0.2
0.1 0.1
0.04
0.04
0.01 0.02
0.01
0.01
0.02
0.004
H2O, Pe = 1 CO2, Pe = 1
0.01
0.002
0.004
0.001 0.002 0.001
0.001 0.001
100 200 500 1000 2000 5000
100 200 500 1000 2000 5000
Gas temperature, T (K) Gas temperature, T (K)
Fig. 4.10 Total emissivity of water vapor Fig. 4.11 Total emissivity of carbon dioxide
read off these charts from the partial pressure and in Figs. 4.10 and 4.11, and the use of wide-band
temperature of each gas and the mean beam models is advised to estimate the band overlap
length for the gas volume geometry. Correction correction instead of using the correction charts
factors for the chart emissivities are available in at these lower temperatures [26]. For crucial
the literature for the pressure effect on water engineering decisions, wide-band model block
vapor emissivity [24], the pressure effect on calculations as detailed by Edwards [22] are
carbon dioxide emissivity [5, 6], and the band recommended over the graphical chart method
overlap for mixtures of the two gases [25]. For to determine total emissivity.
most fire protection engineering applications, the Other gases such as sulfur dioxide, ammonia,
pressure correction factors are 1.0 and the band hydrogen chloride, nitric oxide, and methane
overlap correction is approximately Δε 12εCO2 have been summarized in chart form [5]. The
for medium to large fires. Assuming the carrier carbon monoxide chart by Hottel is not
gas is transparent (e.g., air), the emissivity is: recommended for use [27] due to uncertainties
most likely introduced by traces of carbon diox-
1
εg ¼ CH2 O εH2 O þ CCO2 εCO2 Δε εH2 O þ εCO2 ide in the original experiments. Spectral and total
2
ð4:60Þ properties have been published for some of the
important hydrocarbon gases, e.g., methane,
At temperatures below 400 K, the older charts by acetylene, and propylene [28–30]. Mixtures of
Hottel [5, 6] may be more reliable than the charts several hydrocarbon gases are subject to band
overlapping, and appropriate corrections must be interparticle spacing so large that the radiation for
made to avoid overestimating total emissivity of each particle can be treated independently.
a mixture of fuels. Soot particles are produced as a result of incom-
The total emissivity for a gas in the optically plete combustion and are usually observed to be in
thin limit can be calculated from the Planck mean the form of spheres, agglomerated chunks, and
absorption coefficient. Graphs of the Planck mean long chains. They are generally very small
absorption coefficient for various gases that are (50–1000 Å where 1 Å ¼ 1010 m ¼ 104 μm)
important in fires are shown in Fig. 4.12, which compared to infrared wavelengths, so that
can be used with Equation 4.48 to estimate the the Rayleigh limit is applicable to the calculation
total emissivity (by assuming that total properties of radiation properties [31, 32]. Soot particles
represent a spectral average value). are normally characterized by their optical
properties, size, shape, and chemical composition
(hydrogen-carbon ratio). From a heat transfer
Radiation Properties of Soot viewpoint, radiation from a soot cloud is
predominantly affected by the particle size
In a nonhomogeneous (e.g., with soot) medium, distribution and can be considered independent
scattering becomes an important radiative mech- of the chemical composition [31]. Soot optical
anism in addition to absorption and emission. properties are relatively insensitive to temperature
The absorption and scattering behavior of a changes at elevated temperatures, but as
single particle can be described by solving shown in Fig. 4.13, room temperature values rep-
the electromagnetic field equations; however, resentative of soot in smoke do show appreciable
many physical idealizations and mathematical deviations.
approximations are necessary. The most common By choosing appropriate values of optical
assumptions include perfectly spherical particles, constants for soot, the solution for the electro-
uniformly or randomly distributed particles, and magnetic field equations gives [33]
Fig. 4.12 Planck mean 100

absorption coefficient H2O
for various gases 50
CO2
Planck mean absorption coefficient (atm·m)–1
20
C2H2
10 CH4
2
CO
1
0.5
0.2
0.1
250 500 1000 1500 2000
Temperature (K)
www.ebook777.com
Fig. 4.13 Optical 6

constants for soot
Refractive index
k
1
0.8
0.6
1600 K
1000 K
0.4 300 K
0.2
0.2 0.4 0.6 0.8 1 2 4 6 8 10 20
Wavelength, λ (μ)
C0 A mean coefficient that may be used for the

kλ ¼ f ð4:61Þ
λ v entire range of optical thickness is suggested as
where fv is the soot volume fraction (generally C0
about 106 in flames) and C0, a constant between κ R ¼ 3:72 f T ð4:65Þ
C2 v
2 and 6 dependent on the complex index of
refraction m ¼ n ik, is given by to be used in Equation 4.66 for the soot radiation
calculations. Typical temperatures, volume
36πnk
C0 ¼ 2 ð4:62Þ fractions, and mean absorption coefficients for
n2 k þ 2 þ 4n2 k2
2
soot particles in the luminous flames of various
fuels are tabulated in Table 4.11.
Equations 4.61 and 4.62 can be used to evalu-
ate the Planck mean absorption coefficient in the
optically thin limit [34], giving: Radiation Properties of Gas-Soot
C0
Mixtures
κP ¼ 3:83 f T ð4:63Þ
C2 v
The calculation of the total emissivity of a gas-soot
where C2 is Planck’s second constant mixture requires information on basic flame
(1.4388 102 m-K). The Rosseland mean parameters such as soot volume fraction, soot
absorption coefficient in the optically thick limit is absorption coefficient, temperature and geometric
length of the flame, and partial pressure of
C0 the participating gas components [35]. These
κR ¼ 3:6 f T ð4:64Þ
C2 v parameters can be estimated for various types of
Table 4.11 Radiative properties for soot particles

Fuel, composition κS (m1) fv 106 Ts(K )
Gas fuels Methane, CH4 6.45 4.49 1289
Ethane, C2H6 6.39 3.30 1590
Propane, C2H8 13.32 7.09 1561
Isobutane, (CH3)3CH 16.81 9.17 1554
Ethylene, C2H4 11.92 5.55 1722
Propylene, C3H6 24.07 13.6 1490
n-butane, (CH3)(CH2)2(CH3) 12.59 6.41 1612
Isobutylene, (CH3)2CCH2 30.72 18.7 1409
1,3-butadiene, CH2CHCHCH2 45.42 29.5 1348
Solid Fuels Wood, (CH2O)n 0.8 0.362 1732
Plexiglas, (C5H8O2)n 0.5 0.272 1538
Polystyrene, (C8H8)n 1.2 0.674 1486
Fig. 4.14 Mean 8

absorption coefficients for
luminous flames and smoke PCO2 PH2O
X
C0fv C0fv
Mean absorption coefficients κ/(C0fv × 105) (m–1)
104 atm
6
105
103
4
104
103
2
κR
κP
0
500 1000 1500 2000
Temperature (K)
fuel when actual measurements are unavailable for where S is the physical pathlength, εg is the total
a particular situation. The following equation is a emissivity of the gas alone, and κ s is the effective
good approximation [36] for total emissivity of absorption coefficient of the soot. The Planck
homogeneous gas-soot mixtures: mean absorption coefficients for gas-soot
mixtures in luminous flames and smoke are
εt ¼ ð1 expðκSÞÞ þ εg expðκs SÞ ð4:66Þ shown in Fig. 4.14.
www.ebook777.com
ð
I λ ! !
00
dq_ r
Application to Flame and Fire ¼ ! n R dΩ ð4:68Þ
dλ Ω R
Heat Flux Calculation from a Flame

!
Where n is a unit vector normal to the target
Prediction of the radiative heat flux from a flame !
is important in determining ignition and fire element dA and R is the line-of-sight vector
spread hazard, and in the development of fire extending between dA and the far side of the
detection devices. The shape of flames under flame cylinder. Evaluation of Equation 4.68 is
actual conditions is transient, which makes quite lengthy, but under the condition of
detailed radiation analysis cumbersome. In most L/r 3, it can be simplified to [37]
calculations, flames are idealized as simple geo-
00
metric shapes such as plane layers or axisymmet- dq_ r
ric cylinders and cones. A cylindrical geometry, ¼ πI λ, b ελ ðF1 þ F2 þ F3 Þ ð4:69Þ
dλ
shown in Fig. 4.15, will be analyzed here and
used in a sample calculation. where the shape factor constants and emittance
Assuming κ λ is independent of pathlength, are defined as
integration of the transport equation (Equa-
tion 4.48) yields [37] u r 2
F1 ¼ ðπ 2θ0 þ sin ð2θ0 ÞÞ ð4:70aÞ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 4π L
2κλ
I λ ¼ I bλ 1 exp
sin ðθÞ
r 2 L2 cos 2 ðϕÞ v r
F2 ¼ ðπ 2θ0 þ sin ð2θ0 ÞÞ ð4:70bÞ
4π L
ð4:67Þ
w r
where θ, ϕ. r, and L are geometric variables F3 ¼ cos 2 ðθ0 Þ ð4:70cÞ
π L
defined in Fig. 4.15. The monochromatic radia-
tive heat flux on the target element is given by ελ ¼ 1 expð0:7μλ Þ ð4:71Þ
Fig. 4.15 Schematic of

a cylindrical flame
H
S
q0
z q
n Y
dA
r
n y
dA
x f
L
Detector 1.2 m
Flames
y
1.8 m x
z
~ 0.5 m φ
1m
3.6 m
1.4 m
2.4 m
Fig. 4.16 Example calculation for flux to target element from flame
The parameters in the definitions are given by Solution First, the condition of L/r 3 should
be checked to verify that the previous analysis is
θ0 ¼ tan 1 ðL=H Þ ð4:72aÞ applicable.
κλ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
μλ ¼ 2r θ π
ð4:72bÞ L 1:22 þ 1:82
sin 2
0
þ 4 ¼ 8:65 > 3 ð4:74Þ
r 0:25
! ! ! !
n ¼ u i þ v j þ wk ð4:72cÞ The unit normal vector to the detector is given
! !
If the flame is considered to be homogeneous and by n ¼ k , the polar angle θ0 ¼ tan1(1.818)
Equation 4.69 is integrated over all wavelengths, ¼ 1.068 is determined from Equation 4.72a,
the total heat flux is: and the shape factors are evaluated from
Equations 4.70a, 4.70b, and 4.70c:
00
X
3
q_ r ¼ εm Eb Fj ð4:73Þ F1 ¼ 0 ð4:75aÞ
j¼1
Example 7 As shown in Fig. 4.16, a fire detector F2 ¼ 0 ð4:75bÞ

is located at the center of the ceiling in a room
(2.4 3.6 2.4 m) with wood wall linings. The 1 0:25
F3 ¼ cos 2 ð1:068Þ ¼ 0:0102
sprinkler system is capable of extinguishing fires π 1:818
smaller than 0.5 m in diameter 1.0 m high. For ð4:75cÞ
this example, determine the appropriate heat flux
setting for the detector, using a worst case sce- From Equation 4.73, the radiant heat flux can
nario of ignition in one of the upper ceiling be calculated as:
corners.
www.ebook777.com
00
q_ r ¼ ð1 expðκ m SÞÞσT 4f F3
¼ ð1 expð0:8 0:5ÞÞ 5:67 108 17324 0:0102 ð4:76Þ
¼ 1:7 kW=m2
where wood flame properties were taken from Hot upper gas layers are composed of strongly
Table 4.11. If the geometry of the example had participating media such as carbon dioxide,
been L/r < 3, it would have been necessary to water vapor, and soot particles. Heat flux from
interpolate between the L/r ¼ 3 case and the the smoke layer is directly related to ignition of
L/r ¼ 0 case, which has been obtained accu- remote surface locations such as furniture or
rately [6, 37]. If the detector is pointed directly floor carpets. The schematic in Fig. 4.17 will be
! considered in a radiative transport analysis and
at the burning corner in this example (i.e., n ¼
! ! example calculation. The calculation is based on
0:55 i þ 0:83 j ), the calculated heat flux jumps a considerably simplified formulation which
to 9.0 kW/m2, showing the strong influence of provides reasonable results with only a small
direction in calculations of radiation heat penalty in accuracy.
transfer. Integration of Equation 4.46 over the
pathlength S through the smoke layer yields
Heat Flux Calculation from a Smoke 4 !

Layer σT 4 Tw
I ð SÞ ¼ 1 expðκSÞ ð4:77Þ
π T
Consider the situation shown below in Fig. 4.17
involving radiative heat transfer in a compart- The monochromatic radiative heat flux on a dif-
ment fire with a hot gas layer located below the ferential target element is again given by Equa-
ceiling. tion 4.68. However, for the present geometry of
Fig. 4.17 Example SMOKE LAYER

calculation for flux to target
element from smoke layer
0.5 m
1.9 m q
3.6 m
y
z θ
n R sin
f
x
2.4 m
DIFFERENTIAL
TARGET AREA
the ceiling layer and enclosure surface, integra- factors, the remaining shape factors are deter-
tion of Equation 4.68 is quite time-consuming mined by shape factor algebra:
since the upper and lower bounds of the integral
vary with the angle of the pathlength. The calcu- F12 ¼ 0:3242
lation can be simplified by assuming as a first
order approximation that the lower face of the F31 ¼ 0:1831
smoke layer is an isothermal surface. Using this
A3
assumption, the problem can be handled using F13 ¼ F31 ¼ 0:0002
the simple relations of radiative exchange in a A1
nonparticipating medium between gray surfaces
F14 ¼ 1 F12 F13 ¼ 0:6756
(the absorption of the clear air below the smoke
layer is negligible). The radiosity and irradiation Continuing in a similar fashion, the other shape
of each surface in the enclosure is: factors are obtained as:
J i ¼ εi σT 4i þ ð1 εi ÞGi ð4:78aÞ F21 ¼ 0:3242 F31 ¼ 0:1831 F41 ¼ 0:2560

F22 ¼ 0:0000 F32 ¼ 0:0000 F42 ¼ 0:2561
X
Gi ¼ Fi j J j ð4:78bÞ F23 ¼ 0:0000 F33 ¼ 0:0000 F43 ¼ 0:0003
j F24 ¼ 0:6758 F34 ¼ 0:8169 F44 ¼ 0:4876
After solving the simultaneous equations for all The emissivity for the smoke layer can be
Ji and Gi, the net heat flux on any of the surfaces estimated from the mean absorption coefficient
can be calculated from for a wood flame (Table 4.11) as:
00
q_ r, i ¼ J i Gi ð4:79Þ ε1 ¼ 1 expðκ m SÞ ¼ 1 expð0:8 0:5Þ
¼ 0:33
This situation is considered in Example 8 below.
The blackbody emissive power of each surface is
Example 8 A smoke layer 0.5 m thick is floating calculated as σT4, for example:
near the ceiling of a room with dimensions of 4
3.6 2.4 2.4 m. (See Fig. 4.17.) The floor is σT 1 ¼ 5:6696 108 14004
made from wood (emissivity ¼ 0.9), and the ¼ 217:8 kW=m2
four side walls are painted concrete (emissivity
¼ 0.94). The calculation will determine the heat From Equations 4.78a and 4.78b, the radiative
flux in a bottom corner of the room, assuming fluxes to and from each surface are determined
that each surface in the enclosure is kept at con- by solving the eight simultaneous equations:
stant temperature: the smoke layer at 1400 K, the
side walls at 800 K, and the floor at 300 K. J1 ¼ 88:7 kW=m2 G1 ¼ 17:7 kW=m2
Assume there is a differential target area J2 ¼ 4:7 kW=m2 G2 ¼ 43:3 kW=m2
0.01 m2 in one of the corners of the floor, and J3 ¼ 3:9 kW=m2 G3 ¼ 34:8 kW=m2
also at the floor temperature of 300 K. J4 ¼ 23:9 kW=m2 G4 ¼ 34:3 kW=m2
Solution The bottom of the smoke layer will be The net radiative heat flux on the target element
designated surface 1, the floor will be surface from Equation 4.79 is
2, and the differential target area in the bottom
00
corner will be surface 3. Only four surfaces are q_ r, 3 ¼ J 3 G3 ¼ 30:9 kW=m2
required since the four side walls can be treated
as a single surface 4. Shape factors F12 and F31 where the negative sign indicates that heat must
can be found in Appendix D, and from these two be removed from the target element so it remains
www.ebook777.com
Fig. 4.18 Equivalent Eb1

resistance network for an
enclosure
1 – e1
e1 A1
J1
1 1
——— ———
A1F12 A1F14
1
J2 ——— J4
A1F13
Eb 2 Eb 4
1 – e2 1 1 – e4
———
e2 A 2 A2F24 e4 A 4
1
———
A2F23
J3
1 – e3
e3 A 3
Eb 3
in equilibrium. This example also could have I! ! ! Radiation intensity (W/m2)

been solved by the resistance network method i , j, k Cartesian coordinate direction vectors
shown in Fig. 4.18) J Radiosity or radiative heat flux leav-
ing surface (W/m2)
k Boltzmann constant (1.3806
Nomenclature 1023 J/K), or infrared optical con-
stant of soot (imaginary component),
A Area (m2) or thermal conductivity (W/m K)
C Correction factor for mean beam L Mean beam length or distance (m)
length L0 Geometrical mean beam length (m)
C0 Soot concentration parameter n Index of refraction (c0/c) orinfrared opti-
C2 Planck’s second constant (1.4388 cal constant of soot (real component)
!
102 m K) n Unit normal vector
c Speed of light in the medium (m/s) Pa Partial pressure of absorbing gas (Pa)
c0 Speed of light in a vacuum Pe Effective pressure (Pa)
(2.998 108 m/s) Q Energy rate (W)
E Radiative emissive power (W/m2) q˙00 Heat flux (W/m2)
Fij Configuration factor from surface i to !
surface j R Line of sight vector
fv Soot volume fraction r Radius of cylinder (m)
G Irradiation or radiative heat flux S Pathlength (m)
received by surface (W/m2) T Temperature (K)
H Height (m) t Time (s)
!
h Planck’s constant (6.6256 u, v, w Cartesian components of unit vector n
1034 J s) V Volume (m3)
ðs
X
Pressure pathlength, Pa xðξÞdðξÞ References
0
(atm-m) 1. J. deRis, 17th Symposium (International) on Combus-
x Spatial coordinate (m) tion, 1003, Combustion Institute, Pittsburgh, PA
(1979).
Greek Symbols 2. S.C. Lee and C.L. Tien, Progress in Energy and
Combustion Science, 8, 41 (1982).
3. G.M. Faeth, S.M. Jeng, and J. Gore, in Heat Transfer
α Absorptivity or thermal diffusivity k/pcp in Fire and Combustion Systems, American Society of
(m2/s) Mechanical Engineers, New York (1985).
β Angle from normal (radians) 4. Incropera, F.P. and DeWitt, D.P., Fundamentals of
Heat and Mass Transfer, John Wiley & Sons,
ε Emissivity New York, 2002.
θ Polar angle (radians) 5. H.C. Hottel and A.F. Sarofim, Radiative Heat Trans-
κ Extinction coefficient or absorption coeffi- fer, McGraw-Hill, New York (1967).
cient (m l) 6. R. Siegel and H.R. Howell, Thermal Radiation Heat
Transfer, McGraw-Hill, New York (1981).
λ Wavelength (m) 7. Bejan, A., Heat Transfer, John Wiley & Sons,
μ Micron (106 m) New York, 1993.
μλ Defined parameter, Equation 4.73 8. Hallman, J.R., “Ignition characteristics of plastics and
ν Frequency (st) rubber,” Ph.D. Dissertation, University of Oklahoma,
1971.
ξ Integration dummy variable 9. Hallman, J.R., Welker, J.R., and Sliepcevich, C.M.,
ρ Reflectivity or density (kg/m3) “Polymer surface reflectance–absorptance
Ω Solid angle (steradians) characteristics,” Polymer Engineering and Science
σ Stefan-Boltzmann constant (5.6696 14: 717–723 (1974).
10. Hallman, J.R., Sliepcevich, C.M., and Walker, J.R.,
10B W/m2K4) “Radiation absorption for polymers: The radiant panel
τ Transmissivity or optical pathlength and carbon arcs as radiant heat sources,” Journal of
ϕ Azimuthal angle (radians) Fire & Flammability 9: 353–366 (1978).
χ Fractional measure 11. Wesson, H.R., Welker, J.R., and Sliepcevich, C.M.,
“The piloted ignition of wood by thermal radiation,”
Combustion and Flame 16: 303–310 (1971).
Subscripts 12. Försth, M. and Roos, A., “Absorptivity and its Depen-
dence on Heat Source Temperature and Degree of
a Actual Thermal Breakdown,” Fire and Materials 35:
285–301 (2011).
b Blackbody or base 13. Janssens, M., “Piloted ignition of wood: a review,”
e External Fire and Materials 15: 151–167 (1991).
f Flame 14. Janssens, M. and Douglas, B., “Wood and wood
g Gas products,” in Handbook of Building Materials for
Fire Protection, Ed. Harper, C.A., pp. 7.1–7.58,
i Initial or ith surface McGraw–Hill, New York, 2004.
j Summation variable or jth surface 15. Kashiwagi, T. and Ohlemiller, T.J., “A study of oxy-
m Mean value gen effects on nonflaming transient gasification of
0 Original PMMA and PE during thermal irradiation,”
Proceedings of the Combustion Institute 19:
P Planck mean 815–823 (1982).
R Rosseland mean 16. Modak, A.T. and Croce, P.A., “Plastic pool fires,”
s Surface or soot Combustion and Flame 30: 251–265 (1977).
t Total 17. E.M. Sparrow and R.D. Cess, Radiation Heat Trans-
fer, McGraw-Hill, New York (1978).
w Wall 18. J.R. Howell, A Catalog of Radiation Configuration
λ Spectral wavelength Factors, McGraw-Hill, New York (1982).
ν Spectral frequency 19. C.L. Tien, in Handbook of Heat Transfer
1 Ambient Fundamentals, McGraw-Hill, pp 14.36, New York
(1985).
www.ebook777.com
20. Oppenheim, A.K, Trans. ASME, 65, 725, 1956. 31. S.C. Lee and C.L. Tien, 18th Symposium (Interna-
21. C.L. Tien, Advances in Heat Transfer, 5, 253 (1968). tional) on Combustion, Combustion Institute, 1159,
22. D.K. Edwards, in Handbook of Heat Transfer Pittsburgh (1981).
Fundamentals, McGraw-Hill, pp 14.53, New York 32. C.L. Tien, in Handbook of Heat Transfer
(1985). Fundamentals, McGraw-Hill, pp 14.83, New York
23. D.K. Edwards, Advances in Heat Transfer, 12, 115 (1985).
(1976). 33. G.L. Hubbard and C.L. Tien, Journal of Heat Trans-
24. G.B. Ludwig, W. Malkmus, J.E. Reardon, and fer, 100, 235 (1978).
J.A.L. Thompson, Handbook of Radiation from 34. J.D. Felske and C.L. Tien, Journal of Heat Transfer,
Combustion Gases, NASA SP- 3080, Washington 99, 458 (1977).
(1973). 35. J.D. Felske and C.L. Tien, Combustion Science and
25. T.F. Smith, Z.F. Shen, and J.N. Friedman, Journal of Technology, 7, 25 (1977).
Heat Transfer, 104, 602 (1982). 36. W.W. Yuen and C.L. Tien, 16th Symposium (Interna-
26. J.D. Felske and C.L. Tien, Combustion Science and tional) on Combustion, Combustion Institute, 1481,
Technology, 11, 111 (1975). Pittsburgh (1977).
27. M.M. Abu-Romia and C.L. Tien,, Journal of Quanti- 37. A. Dayan and C.L. Tien, Combustion Science and
tative Spectroscopy and Radiative Transfer, 107, 143 Technology, 9, 41 (1974).
(1966).
28. M.A. Brosmer and C.L. Tien, Journal of Quantita-
tive Spectroscopy and Radiative Transfer, 33, 521 Chris Lautenberger is a fire protection engineer
(1985).
at Reax Engineering Inc. in Berkeley, CA. He is
29. M.A. Brosmer and C.L. Tien, Journal of Heat Trans-
fer, 107, 943 (1985). also an Instructor in Cal Poly’s Fire Protection
30. M.A. Brosmer and C.L. Tien, Combustion Science Engineering program where he co-teaches
and Technology, 48, 163 (1986). courses on Fire Modeling and Fire Dynamics.
Thermochemistry
5
D.D. Drysdale
Introduction this is not unreasonable, as flaming combustion

takes place in the gas phase. It may also be
Thermochemistry is the branch of physical assumed that the ideal gas law applies, that is,
chemistry that is concerned with the amounts of
PV ¼ n RT ð5:1Þ
energy released or absorbed when a chemical
change (reaction) takes place [1–3]. Inasmuch where P and V are the pressure and volume of
as fire is fundamentally a manifestation of a n moles of gas at a temperature, T (in degrees
particular type of chemical reaction, viz., com- Kelvin); values of the universal gas constant
bustion, thermochemistry provides methods by (R) in various sets of units are summarized in
which the energy released during fire processes Table 5.1. At ambient temperatures, deviations
can be calculated from data available in the sci- from “ideal behavior” can be detected with most
entific and technical literature. gases and vapors, while at elevated temperatures
To place it in context, thermochemistry is a such deviations become less significant. In this
major derivative of the first law of thermodynam- context, it should be noted that whereas the terms
ics, which is a statement of the principle of conser- gas and vapor are sometimes used interchange-
vation of energy. However, while concerned with ably, it is best to make a distinction. Both are in
chemical change, thermodynamics does not indi- the gaseous state, but as a general rule a vapor at
cate anything about the rate at which such a change normal temperatures can be made to liquefy if
takes place or about the mechanism of conversion; the pressure of the vapor is increased sufficiently.
this falls within the topic of chemical kinetics Thus, propane can be stored as a liquid under a
[4]. Consequently, the information it provides is relatively low pressure (approximately 8.4 bar at
normally used in association with other data, for 293 Kelvin) whereas the permanent gases (par-
example, to enable the rate of heat release to be ticularly hydrogen, helium, oxygen, and nitro-
calculated from the rate of burning. gen) can only be stored as gases at ambient
temperatures, typically in pressure cylinders at
2000 psi (c. 138 bar). Again, as a general rule,
The First Law of Thermodynamics vapors tend to deviate more strongly from ideal
gas behavior than do the permanent gases.
It is convenient to limit the present discussion to
chemical and physical changes involving gases; Internal Energy
D.D. Drysdale (*)
BRE Centre for Fire Safety Engineering, University As a statement of the principle of conservation of
of Edinburgh, Scotland, UK energy, the first law of thermodynamics deals

www.ebook777.com
5 Thermochemistry 139
Table 5.1 Values of the ideal gas constant, R dx ¼ Distance through which the piston is
Units of Units of moved; the increment in volume is therefore
pressure volume Units of R Value of R dV ¼ A · dx
Pa (N/m2) m3 J/Kmol 8.31431 The total work done is obtained by integrating
atm cm3 cm3atm/ 82.0575 Equation 5.5 from the initial to the final state;
Kmol that is,
atm l latm/ 0.0820575
Kmol ð final
atm m3 m3atm/ 8.20575 10–5 w ¼ P dV ð5:6Þ
Kmol initial
Combining Equations 5.3 and 5.5, the differ-

ential change in internal energy can be written
with the relationship between work and heat.
dE ¼ dq P dV ð5:7Þ
Confining our attention to a “closed system”—
for which there is no exchange of matter with the This shows that if the volume remains con-
surroundings—it is known that there will be a stant, as P · dV ¼ 0, then dE ¼ dq; if this is
change if heat is added or taken away, or if work integrated, we obtain
is done on or by “the system” (e.g., by compres-
sion). This change is usually accompanied by an ΔE ¼ qv ð5:8Þ
increase or decrease in temperature and can be
where qv is the heat transferred to the constant
quantified if we first define a function of state
volume system; that is, the change in internal
known as the internal energy of the system, E.
energy is equal to the heat absorbed (or lost) at
Any change in the internal energy of the system
constant volume.
(ΔE) is then given by
ΔE ¼ q w ð5:2Þ
Enthalpy
where q is the heat transferred to the system, and
w is the work done by the system. This can be With the exception of explosions in closed
expressed in differential form vessels, fires occur under conditions of constant
pressure. Consequently, the work done as a result
dE ¼ dq dw ð5:3Þ
of expansion of the fire gases must be taken into
Being a function of state, E varies with tempera- account. At constant pressure, Equation 5.5 may
ture and pressure, that is, E ¼ E(T, P). be integrated to give
According to the standard definition, work, w,
w ¼ P ðV 2 V 1 Þ ð5:9Þ
is done when a force, F, moves its point of
application through a distance, x, thus, in the where V1 and V2 are the initial and final volumes,
limit respectively. Equation 5.2 then becomes
dw ¼ F dx ð5:4Þ ΔE ¼ E2 E1 ¼ q p þ PV 1 PV 2 ð5:10Þ
The work done during the expansion of a gas or, rearranging,
can be derived by considering a cylinder/piston
assembly (Fig. 5.1); thus q p ¼ ðE2 þ PV 2 Þ ðE1 þ PV 1 Þ
dw ¼ P A dx ¼ P dV ð5:5Þ ¼ H2 H1 ð5:11Þ
where where qp is the heat transferred at constant pres-

P ¼ Pressure of the gas sure, and H is known as the enthalpy (H E +
A ¼ Area of the piston PV). The change in enthalpy is therefore the heat
140 D.D. Drysdale
Fig. 5.1 Cylinder/piston Gas pressure P

assembly
dx
Force F
Area A
absorbed (or lost) at constant pressure (provided 0.028 kg nitrogen (N2)

that only P V work is done), and consequently 0.032 kg oxygen (O2)
it is the change in enthalpy that must be consid- 0.044 kg carbon dioxide (CO2)
ered in fire-related problems. 0.028 kg carbon monoxide (CO)
0.016 kg methane (CH4)
0.044 kg propane (C3H8)
The concept of specific heat is normally
Specific Heat associated with solids and liquids, but it is
equally applicable to gases. Such specific heats
Specific heat, or heat capacity, of a body or are required for calculating flame temperatures,
“system” is defined as the amount of heat as described below. Values for a number of
required to raise the temperature of unit mass important gases at constant pressure and a range
by 1 C; the units are J/kg K, although for most of temperatures are given in Table 5.2.
thermochemical problems the units J/mol K are It is important to note that there are two dis-
more convenient. The formal definition of the tinct heat capacities; at constant pressure, Cp, and
“mole” is the amount of a substance (solid, liq- at constant volume, Cv. Thus, at constant
uid, or gas) that contains as many elementary pressure
units (atoms or molecules) as there are carbon
atoms in exactly 0.012 kg of carbon-12 (C12). dq p ¼ dH ¼ C p dT ð5:12Þ
This number—known as Avogadro’s number—
is actually 6.023 1023; in its original form, while at constant volume
Avogadro’s hypothesis was applied to gases and dqv ¼ dE ¼ Cv dT ð5:13Þ
stated that equal numbers of molecules of differ-
ent gases at the same temperature and pressure For an ideal gas, C p ¼ Cv þ R:
occupy the same volume. Thus, the quantity of a
substance that corresponds to a mole is simply
the gram-molecular weight, but expressed in Heats of Combustion
kilograms to conform with SI units. For example,
the following quantities of the gases N2, O2, CO2, Chemical Reactions and Stoichiometry
and CO represent 1 mole of the respective gas
and, according to Avogadro’s hypothesis, will When chemical reactions occur, they are nor-
each occupy 0.022414 m3 at 273 K and mally accompanied by the release or absorption
760 mmHg (101.1 kPa): of heat. Thermochemistry deals with the
www.ebook777.com
Table 5.2 Heat capacities of selected gases at constant pressure (101.1 kN/m2) [5]
Cp (J/mol K)
Temperature (K) 298 500 1000 1500 2000
Species
CO 29.14 29.79 33.18 35.22 36.25
CO2 37.129 44.626 54.308 58.379 60.350
H2O(g) 33.577 35.208 41.217 46.999 51.103
N2 29.125 29.577 32.698 34.852 35.987
O2 29.372 31.091 34.878 36.560 37.777
He 20.786 20.786 20.786 20.786 20.786
CH4 35.639 46.342 71.797 86.559 94.399
quantification of the associated energy changes. into account, although it does not participate to
This requires a definition of the initial and final any significant extent in the chemical change. As
states, normally expressed in terms of an appro- the ratio of oxygen to nitrogen in air is approxi-
priate chemical equation, for example, mately 21:79 (or 1:3.76), Reaction 5.R1 can be
rewritten
C3 H8 þ 5O2 ! 3CO2 þ 4H2 O ð5:R1Þ
C3 H8 þ 5O2 þ 18:8N2 ! 3CO2 þ 4H2 O þ 18:8N2
in which the reactants (propane and oxygen) and
ð5:R3Þ
products (carbon dioxide and water) are
specified. This balanced chemical equation (where 18.8 ¼ 5 3.76), showing that 44 g pro-
defines the stoichiometry of the reaction, that is, pane requires (160 + 18.8 28), or 686.4 g of
the exact proportions of the two reactants (pro- “air” for complete combustion, that is, 15.6 g
pane and oxygen) for complete conversion to air/g propane. Calculations of this type are valu-
products (no reactants remaining). Note that the able in assessing the air requirements of fires.
physical states of the reactants and products Thus, on the assumption that wood has the
should also be specified. In most cases, the initial empirical formula [5] CH1.5O0.75, it can be
conditions correspond to ambient (i.e., 25 C and shown that its stoichiometric air requirement is
atmospheric pressure) so that there should be no 5.38 g air for each gram of fuel, assuming com-
doubt about the state of the reactants. In this case plete combustion of wood to CO2 and H2O. The
both are gaseous, but it is more common in fires relevant stoichiometric equation is
for the “fuel” to be in a condensed state, either
liquid or solid. As an example, the oxidation of n- CH1:5 O0:75 þ O2 þ 3:76N2
hexane can be written ¼ CO2 þ 0:75H2 O þ 3:76N2 ð5:R4Þ
C6 H14 þ 9:5O2 ! 6CO2 þ 7H2 O ð5:R2Þ In this calculation no distinction is made of the
fact that flaming combustion of wood involves
but the fuel may be in either the liquid or the oxidation of the volatile gases and vapors pro-
vapor state. The consequences of this will be duced by the pyrolysis of wood, while the resid-
discussed below. ual char burns much more slowly by surface
Reaction 5.R1 may be used to calculate the oxidation.
mass of oxygen or air required for the complete
oxidation of a given mass of propane. Thus, we
deduce that 1 mole of propane (44 g) reacts Measurement of Heats of Combustion
completely with 5 moles of oxygen
(5 32 ¼ 160 g); that is, 1 g propane requires The heat of combustion of a fuel is defined as the
3.64 g oxygen. If the propane is burning in air, amount of heat released when unit quantity is
then the presence of nitrogen needs to be taken oxidized completely to yield stable end products.
142 D.D. Drysdale
In the present context, the relevant combustion that there is a heat of gasification associated with
processes occur at constant pressure so that we any condensed fuel (liquid or solid); a correction
are concerned with an enthalpy change, ΔHc. By must be made for this if the heat of combustion of
convention, for exothermic reactions such as oxi- the fuel vapor is required.
dation, values of ΔHc are negative; that is, the Table 5.3 contains the heats of combustion
reaction produces energy that can then be lost (ΔHc) of a number of combustible gases, liquids,
from the system. (By contrast, an endothermic and solids, expressed in various ways, viz.,
reaction such as the conversion of water to kJ/mole (fuel), kJ/g (fuel), kJ/g (oxygen), and
hydrogen and oxygen will take place only if kJ/g (air). The first of these is the form normally
energy is provided in a suitable form.) encountered in chemistry texts and reference
Heats of combustion are measured in the books, whereas the second is more commonly
combustion bomb calorimeter in which a precise found in sources relating to chemical engineering
amount of fuel is burned in pure oxygen inside a and fuel technology and is more useful to the fire
pressure vessel whose temperature is strictly protection engineer. However, the third and, par-
monitored. The apparatus is designed to reduce ticularly, the fourth have very specific uses in
heat losses from the calorimeter to a minimum relation to fire problems. It is immediately appar-
so that the amount of heat released can be calcu- ent from Table 5.3 that ΔHc (O2) and ΔHc (air)
lated from the rise in temperature and the total are approximately constant for most of the fuels
thermal capacity of the system; corrections can listed, having average values of 13.1 kJ/g and
be made for any heat loss. In the past, combus- 3 kJ/g, respectively. (See the section on “Rate
tion bomb calorimetry received a great deal of of Heat Release in Fires”.)
attention within physical chemistry [1, 6] as the The data quoted in Table 5.3 refer to heats of
technique was able to provide a wealth of data of combustion measured at ambient temperature,
relevance to thermochemistry. However, the normally 25 C. These data will be satisfactory
experiment gives the heat released at constant for virtually all relevant fire problems, but occa-
volume; that is, the change in internal energy, sionally it may be necessary to consider the heat
ΔE (Equation 5.8). The change in enthalpy is released when combustion takes place at higher
given by temperatures. This requires a simple application
of the first law of thermodynamics. If the reaction
ΔH ¼ ΔE þ ΔðPV Þ ð5:14Þ
involves reactants at temperature T0 reacting to
where Δ(PV) is calculated using the ideal gas law give products at the final temperature TF, the
process can be regarded in two ways:
ΔðPV Þ ¼ ΔðnRT Þ ð5:15Þ 1. The products are formed at T0, absorb the heat
of combustion, and are heated to the final
The method gives the gross heat of combus-
temperature TF.
tion—that is, in which the reactants and products
2. The heat of combustion is imagined first to
are in their standard states. The net heat of com-
heat the reactants to TF, then the reaction
bustion, on the other hand, refers to the situation
proceeds to completion, with no further
in which the products are in the sate in which
temperature rise.
they are formed. For Reaction 5.R1, for example,
By the first law, we can write
water is formed in the gaseous phase so that the
amount of energy released is less than the gross
ðΔH c ÞT 0 þ CPrp ðT F T 0 Þ
heat of combustion by an amount equivalent to
the latent heat of evaporation of water (2.26 kJ/ ¼ ðΔHc ÞT F þ C Rp ðT F T 0 Þ ð5:16Þ
g). The net heat of combustion is the value that
should be used in fire calculations. This is where CpPr and CpR are the total heat capacities of
illustrated in the next section: see Reactions the products and reactants, respectively. This
5.R5a and 5.R5b. It should also be remembered may be rearranged to give
www.ebook777.com
Table 5.3 Heats of combustion of selected fuels at 25 C (298 K) [7]

Fuel ΔHc (kJ/mol) ΔHc (kJ/g) ΔHcb (kJ/g[O2]) ΔHc (kJ/g[air])
Carbon monoxide (CO) 283 10.10 17.69 4.10
Methane (CH4) 800 50.00 12.54 2.91
Ethane (C2H6) 1423 47.45 11.21 2.96
Ethene (C2H4) 1411 50.53 14.74 3.42
Ethyne (C2H2) 1253 48.20 15.73 3.65
Propane (C3H8) 2044 46.45 12.80 2.97
n-Butane (n-C4H10) 2650 45.69 12.80 2.97
n-Pentane (n-C5H12) 3259 45.27 12.80 2.97
n-Hexane 3861 44.90
c-Hexane (c-C6H12) 3680 43.81 12.80 2.97
n-Octane (n-C8H18) 5104 44.77 12.80 2.97
Benzene (C6H6) 3120 40.00 13.06 3.03
Methanol (CH3OH) 635 19.83 13.22 3.07
Ethanol (C2H5OH) 1232 26.78 12.88 2.99
Acetone (CH3COCH3) 1786 30.79 14.00 3.25
D-glucose (C6H12O6) 2772 15.40 13.27 3.08
Cellulosec — 16.09 13.59 3.15
Polyethylene — 43.28 12.65 2.93
Polypropylene — 43.31 12.66 2.94
Polystyrene — 39.85 12.97 3.01
Polyvinylchloride — 16.43 12.84 2.98
Polymethylmethacrylate — 24.89 12.98 3.01
Polyacrylonitrile — 30.80 13.61 3.16
Polyoxymethylene — 15.46 14.50 3.36
Polyethyleneterephthalate — 22.00 13.21 3.06
Polycarbonate — 29.72 13.12 3.04
Nylon 6,6 — 29.58 12.67 2.94
Polyester — 23.8 — —
Wool — 20.5 — —
Wood (European beech) — 19.5 — —
Wood volatiles (European beech) — 16.6 — —
Wood char (European beech) — 34.3 — —
Wood (Ponderosa pine) — 19.4 — —
a
Apart from the solids (D-glucose, et seq.), the initial state of the fuel and of all the products is taken to be gaseous
b
ΔHc(O2) ¼ 13.1 kJ/g is used in the oxygen consumption method for calculating rate of heat release
c
Cotton and rayon are virtually pure cellulose and can be assumed to have the same heat of combustion
ðΔHc ÞT F ðΔH c ÞT 0 where ΔC p ¼ CPrp -C Rp . This may be used in

¼ ΔC p ð5:17Þ
TF T0 integrated form to calculate the heat of combus-
tion at temperature T2 if ΔHc is known at tem-
or, in differential form, we have Kirchoff’s perature T1 and information is available on the
equation heat capacities of the reactants and products, thus
dðΔH c Þ ð T2
¼ ΔC p ð5:18Þ
dT ðΔH c ÞT 2 ¼ ðΔH c ÞT 1 þ ΔC p dT ð5:19Þ
T1
144 D.D. Drysdale
Where H2 ðgÞ þ 0:5O2 ðgÞ ! H2 OðgÞΔH f ½H2 OðgÞ

X X ¼ 241:84 kJ=mol
ΔC p ¼ C p ðproductsÞ C p ðreactantsÞ
ð5:20Þ ð5:R5bÞ
and Cp is a function of temperature, which can By definition, the heats of formation of all the
normally be expressed as a power series in T, for elements are set arbitrarily to zero at all
example, temperatures. This then allows the heats of reac-
tion to be calculated from the heats of formation
C p ¼ a þ bT þ cT 2 þ . . . ð5:21Þ of the reactants and products, thus
Information on heat capacities of a number of ΔH ¼ ΔH f ðproductsÞ ΔH f ðreactantsÞ

species and their variation with temperature may ð5:22Þ
be found in Stull and Prophet [7] and Strehlow
[8]. Some data are summarized in Table 5.2. where ΔH is the heat (enthalpy) of the relevant
reaction. However, most heats of formation can-
not be obtained as easily as heats of combustion.
The example given in Reaction (5.R5a) is
Heats of Formation
unusual in that the heat of formation of water
also happens to be the heat of combustion of
The first law of thermodynamics implies that the
hydrogen. Similarly, the heat of combustion of
change in internal energy (or enthalpy) of a sys-
carbon in its most stable form under ambient
tem depends only on the initial and final states of
conditions (graphite) is the heat of formation of
the system and is thus independent of the inter-
carbon dioxide.
mediate stages. This is embodied in thermo-
Combustion calorimetry can be used indi-
chemistry as Hess’s law, which applies directly
rectly to calculate heats of formation. The heat
to chemical reactions. From this, we can develop
of formation of ethyne (acetylene), which is the
the concept of heat of formation, which provides
enthalpy change of the reaction
a means of comparing the relative stabilities of
different chemical compounds and may be used 2CðgraphiteÞ þ H2 ! C2 H2 ð5:R6Þ
to calculate heats of chemical reactions that can-
not be measured directly. can be deduced in the following way: the heat of
The heat of formation of a compound is combustion of ethyne has been determined by
defined as the enthalpy change when 1 mole of bomb calorimetry as 1255.5 kJ/mol at 25 C
that compound is formed from its constituent (298 K). This is the heat of the reaction
elements in their standard state (at 1 atm pressure
C2 H2 þ 2:5O2 ! 2CO2 þ H2 O ð5:R7Þ
and 298 K). Thus, the heat of formation of liquid
water is the enthalpy change of the reaction which, by Hess’s law (see Equation 5.22), can be
(at 298 K) equated to
H 2 ðgÞ þ 0:5O2 ðgÞ ! H2 OðlÞΔH f 298
ð5:R5aÞ ðΔHc Þ298 ðC2 H2 Þ ¼ 2 ΔH f ðCO2 Þ
¼ 285:8kJ=mol 298 298
þ ΔH f ðH2 OÞ ΔH f ðC 2 H 2 Þ
298
so that ΔHf (H2O) (l) ¼ 285.8 kJ/mole at 2:5 ΔH f ðO2 Þ
25 C. This differs from the heat released by
ð5:23Þ
the reaction if the product is water vapor rather
than liquid (“The heat of formation of water We know that
vapor” kJ/mol [ΔHf {H2O(g)} ¼ –241.84 kJ/
mol]) by the latent heat of evaporation of water (ΔHc)298 (C2H2) ¼ 1255.5 kJ/mol
at 25 C (43.96 kJ/mol). Thus (ΔHf)298 (CO2) ¼ 393.5 kJ/mol
www.ebook777.com
(ΔHf)298 (H2O) ¼ 241.8 kJ/mol Table 5.4 Heats of formation at 25 C (298 K)
(ΔHf)298 (O2) ¼ 0.0 kJ/mol (by definition) Compound (ΔHf)298 (kJ/mol)

Hydrogen (atomic) +218.00
so that by rearrangement, Equation 5.23 Oxygen (atomic) +249.17
yields Hydroxyl (OH) +38.99
(ΔHf)298 (C2H2) ¼ +226.7 kJ/mol Chlorine (atomic) +121.29
Carbon monoxide –110.53
This compound has a positive heat of forma- Carbon dioxide –393.52
tion, unlike CO2 and H2O. This indicates that it is Water (liquid) –285.8
an endothermic compound and is therefore less Water (vapor) –241.83
stable than the parent elements. Under appropri- Hydrogen chloride –92.31
ate conditions, ethyne can decompose violently Hydrogen cyanide (gas) +135.14
to give more stable species. Nitric oxide +90.29
The heats of formation of a number of Nitrogen dioxide +33.85
compounds are given in Table 5.4. The most Ammonia –45.90
stable compounds (CO2 and H2O) have the larg- Methane –74.87
est negative values, while positive values tend to Ethane –84.5
indicate an instability with respect to the parent Ethene +52.6
elements. This can indicate a high chemical reac- Ethyne (acetylene) +226.9
tivity, and indeed heats of formation have been Propane –103.6
used in preliminary hazard assessment to provide n-Butane –124.3
Iso-butanea –131.2
an indication of the risks associated with new
Methanol –242.1
processes in the chemical industry. It should be a
noted that the heats of combustion of endother- Heats of formation of other hydrocarbons are tabulated in
Weast [9]
mic compounds do not give any indication of any
associated reactivity (compare ethane, ethene,
and ethyne in Tables 5.3 and 5.4).
natural fires, which involve diffusion flames
rather than premixed flames. Air and fuel have
to mix by the process of diffusion (laminar or
Rate of Heat Release in Fires turbulent, depending on the size of the fire)
before combustion can occur. The mixing pro-
Although thermochemistry can give information cess is relatively inefficient, and despite the fact
relating to the total amount of energy that can be that excess air is drawn (or entrained) into the
released when a fuel is burned to completion, it is flame, the products of combustion will contain
rarely (if ever) possible to use heats of combus- some species that are only partially oxidized,
tion directly to calculate the heat released in such as carbon monoxide, aldehydes, ketones,
“real” fires. Indeed, it can be argued that the and particulate matter in the form of soot or
rate of heat release is more important than the smoke. Their presence indicates that not all the
total available [10]. When a single item is burn- available combustion energy has been released.
ing in isolation, the rate of burning and the rate of The “combustion efficiency” is likely to vary
heat release in the flame are coupled. It has been from around 0.3–0.4 for heavily fire-retarded
common to express the rate of heat release as the materials to 0.9 or higher in the case of oxygen-
product of the burning rate (i.e., the rate of mass containing products (e.g., polyoxymethylene)
loss ṁ [kg/s]) and the net heat of combustion of [10, 11]. This is discussed in detail by
the fuel (ΔHc kJ/kg). Tewarson [12].
Q_ c ¼ m_ ΔH c ð5:24Þ Fires burning in compartments present a
completely different problem. In the first place,
However, this assumes that combustion is com- there is likely to be a range of different fuels
plete, although it is known that this is never so in present, each with a different stoichiometric air
146 D.D. Drysdale
requirement. These will burn at different rates, Burning gases (i.e., flames) are carried outside
dictated not just by the nature of the fuel but also the compartment, indicating that not all of the
by the levels of radiant heat existing within the heat of combustion is released within the com-
compartment during the fire. The rate of heat partment. For a fully developed (postflashover)
release during the fully developed stage of a fire, it is perhaps more likely that the ratio ṁair/
compartment fire is required for calculating ṁfuel (the “equivalence ratio”) is less than the
postflashover temperature-time histories for stoichiometric ratio r—that is, insufficient air is
estimating fire exposure of elements of structure, entering the compartment to burn all the fuel
as in the method developed by Pettersson vapors. Under these circumstances, excess fuel
et al. [13]. To calculate the rate of heat release vapors will escape from the compartment and
within the compartment, it is assumed that the burn outside as they mix with external air. The
fire is ventilation controlled and that all combus- external flame length will depend inter alia on
tion takes place within the compartment. The rate the equivalence ratio [15].
of heat release ( Q_ c ) can be obtained from the Regardless of whether the equivalence ratio is
expression equal to or greater than the stoichiometric ratio,
fuel vapor will burn outside the compartment and
Q_ c ¼ m_ air ΔH c ðairÞ ð5:25Þ temperatures based on Equation 5.25 will be
high. The method will also overestimate the
where ΔHc (air) is the heat of combustion per
temperatures achieved if the equivalence ratio is
unit mass of air consumed (3 kJ/g; see Table 5.3),
much greater than the stoichiometric ratio. Under
and ṁair is the mass flow rate of air into the
these conditions, excess air is drawn into the
compartment, given approximately by the
compartment and will act as a diluent and reduce
expression
the average temperatures—if the ventilation is
m_ air ¼ 0:52A0 H 0
1=2
ð5:26Þ high enough, the rate of heat release will be
controlled by the area of the burning surface
where A0 is the effective area of ventilation (m2) [3, 16].
and H0 is the height of the ventilation opening Note that the concept of equivalence ratio is
(m) [14]. The compartment temperature (as a used elsewhere in this handbook, specifically by
function of time) is then obtained from heat bal- Tewarson [12] and by Gottuk and Lattimer [17]
ance calculations, as described in Drysdale [3], in discussing the yields of products generated in
Tewarson [12], and Walton and Thomas [14]. the upper layer during the preflashover fire.
The assumption behind Equation 5.25 is that Much useful data on the fire behavior of com-
the burning process is stoichiometric and that all bustible materials can be obtained by using the
the fuel vapors are burned within the compart- technique of “oxygen consumption calorimetry.”
ment—air is supplied at exactly the rate required This is the basis of the “cone calorimeter,” in
to consume the fuel vapors, that is, which the rate of heat release from a small sam-
ple of material burning under an imposed radiant
Rate of supply of air
¼r heat flux is determined by measuring the rate of
Rate of supply of fuel oxygen consumption [18]. The latter can be
where r is the stoichiometric air-fuel ratio and the converted into a rate of heat release using the
maximum possible temperatures will be conversion factor 13.1 kJ/g of oxygen consumed.
achieved. However, it is worth noting that this (A small correction is required for incomplete
does not take into account the fact that the rate of combustion, based on the yield of CO.) This
heat release is not instantaneous. Although technique can be used on a larger scale to mea-
(in principle) the ideal stoichiometric mixture is sure the rate of heat release from items of furni-
created within the compartment, burning gases ture, wall lining materials, and so on [19, 20] and
will emerge from the opening(s) simply because is now used routinely in both fire research and
the reaction takes time to reach completion. fire testing facilities.
www.ebook777.com
Table 5.5 Thermal capacity of the products of combus-

Calculation of Adiabatic Flame tion of a stoichiometric propane/air mixture
Temperatures Thermal capacity at 1000 K
No. of
moles (J/molK) (J/K)
In the previous sections, no consideration has CO2 3 54.3 162.9
been given to the fate of the energy released by H2O 4 41.2 164.8
the combustion reactions. Initially it will be N2 18.8 32.7 614.8
absorbed within the reaction system itself by Total thermal capacity ¼ 942.5 J/K
(1) unreacted reactants, (2) combustion products, (per mole of propane)
and (3) diluents, although it will ultimately be
lost from the system by various heat transfer
processes. This is particularly true for natural
fires in enclosed spaces. However, if we consider 2044000
ΔT ¼ ¼ 2169 K
a premixed reaction system, such as a flammable 942:5
vapor-air mixture, and assume it to be adiabatic, giving the final (adiabatic) temperature as
that is, there is no transfer of heat (or mass) to or 2169 + 298 ¼ 2467 K. In fact, this figure is
from the system, then we can calculate the maxi- approximate for the following reasons:
mum theoretical temperature, the adiabatic flame 1. Thermal capacities change with temperature,
temperature. and average values over the range of
Consider a flame propagating through a temperatures appropriate to the problem
stoichiometric propane-air mixture of infinite have been used.
extent (i.e., there are no surfaces to which 2. The system cannot be adiabatic as there will
heat may be transferred) and that is initially at be heat loss by radiation from the hot gases
25 C. The appropriate equation is given by (CO2 and H2O).
Reaction 5.R8: 3. At high temperatures, dissociation of the
C3 H8 þ 5O2 þ 18:8N2 ! products will occur; as these are endothermic
ð5:R8Þ processes, there will be a reduction in the final
3CO2 þ 4H2 O þ 18:8N2 temperature.
Of these, (2) and (3) determine that the actual
This reaction releases 2044 kJ for every mole
flame temperature will be much lower than
of propane consumed. This quantity of energy
predicted. These effects can be taken into
goes toward heating the reaction products, that
account. Thus, with propane burning in air, the
is, 3 moles of carbon dioxide, 4 moles of water
final temperature may not exceed 2000 K.
(vapor), and 18.8 moles of nitrogen for every
If the propane were burning as a stoichiomet-
mole of propane burned. The thermal capacity
ric mixture in pure oxygen, then in the absence of
of this mixture can be calculated from the ther-
nitrogen as a “heat sink,” much higher
mal capacities of the individual gases, which
temperatures would be achieved. The total ther-
are available in the literature (e.g., JANAF)
mal capacity would be (942.5 614.8)
[7]. The procedure is straightforward, provided
¼ 327.7 J/K. However, the amount of heat
that an average value of Cp is taken for each gas
released remains unchanged (2044 kJ) so that
in the temperature range involved, giving
the maximum temperature rise would be
942.5 kJ/K as the total thermal capacity of the
products per mole of propane consumed (see
2044000
Table 5.5). ΔT ¼ ¼ 6238K
327:7
As 2044 kJ are released at the same time as
these species are formed, the maximum temper- predicting a final temperature of 6263 C.
ature rise will be Because dissociation will be a dominant factor,
148 D.D. Drysdale
this cannot be achieved and the temperature of Table 5.6 Adiabatic flame temperature of lower-
the flame will not exceed approximately 3500 K. limiting hydrocarbon/air mixtures
The occurrence of dissociation at Adiabatic flame temperature
temperatures in the region of 2000 K and Gas at lower flammability limit (K)
above makes it necessary to take dissociation Methane 1446
into account. Dissociation is discussed in Ethane 1502
Propane 1554
Chap. 6. However, the simple calculation
n-Butane 1612
outlined above can be used to estimate the
n-Pentane 1564
temperatures of near-limit flames, when the tem-
n-Heptane 1692
perature is significantly lower and dissociation
n-Octane 1632
can be neglected.
It is known that the lower flammability limit
of propane is 2.2 %. The oxidation reaction tak- atmosphere, and it is decided to keep the atmo-
ing place in this mixture can be described by the sphere inert by the addition of nitrogen. Calcu-
following equation: late the percentage of nitrogen necessary to
prevent ignition of a mixture in which the pro-
0:022C3 H8 þ 0:978ð0:21O2 þ 0:79N2 Þ ! products
pane and air are in stoichiometric proportions.
Dividing through by 0.022 allows this to be
written Solution The stoichiometric reaction for pro-
pane in air is
C3 H8 þ 9:34O2 þ 35:12N2 ! 3CO2 þ 4H2 O
C3 H8 þ 5O2 þ 18:8N2 !
þ 4:34O2 þ 35:12N2 ð5:R10Þ
3CO2 þ 4H2 O þ 18:8N2
ð5:R9Þ
and the heat of combustion of propane is
showing that the heat released by the oxidation of
2044 kJ/mole. It is assumed that the heat of
1 mole of propane is now absorbed by excess
combustion is absorbed by the products 3CO2 +
oxygen (4.34 moles) and an increased amount of
4H2O + 18.8 N2. It was shown above that the
nitrogen as well as the combustion products.
adiabatic flame temperature (i.e., the temperature
Carrying out the same calculation as before, it
of the product gases, assuming no heat losses)
can be shown that the total thermal capacity of
will be 2169 K. If the flame temperature can be
the products per mole of propane consumed is
held below 1600 K (or 1554 K, according to
1627.6 kJ/K, which gives the adiabatic flame
Table 5.6), then flame propagation will not be
temperature for this limiting mixture is 1281 C
possible and the introduction of an ignition
(1554 K). If the same calculation is carried out
source will not lead to an explosion. Suppose
for the other hydrocarbon gases, it is found that
that the extra quantity of nitrogen required
the adiabatic limiting flame temperature lies in a
to form an “inert atmosphere” corresponds to
fairly narrow band, 1600 100 K (Table 5.6).
X moles per mole of propane. Then
This can be interpreted by assuming that the limit
exists because heat losses (by radiation from the C3 H8 þ 5O2 þ 18:8N2 þ XN2
flame) exceed the rate of heat production (within ! 3CO2 þ 4H2 O þ ð18:8 þ XÞN2 ð5:R11Þ
the flame). As a consequence, flame cannot sus-
tain itself. This concept can be applied to certain Following the procedure illustrated in Table 5.5,
practical problems relating to the lower flamma- the thermal capacity of the product gases—
bility limit. 3CO2 + 4H2O + (18.8 + X)N2—will be 3
54.3 + 4 41.2 + (18.8 + X) 32.7 ¼ ΣCp. If
Example 1 It is recognized that the leak of pro- sufficient nitrogen has been added to reduce the
pane into a test cell could lead to a flammable adiabatic flame temperature to 1554 K, then the
www.ebook777.com
thermal capacity of the product gases will be If we consider that 1 mole of fuel passes
given by through each of the six cylinders, but of the
6 moles only three are burned, we have overall
X 2044000
Cp ¼
1554 298 6CH4 þ 12O2 þ 45:12N2 ! 3CH4 þ 3CO2
¼ 1627:4 kJ=mole of propane þ 6H2 O þ 6O2 þ 45:12N2
ð5:R13Þ
Thus
Dividing through by 3 gives
3 54:3 þ 4 41:2 þ ð18:8 þ XÞ 32:7
¼ 1627:4 2CH4 þ 4O2 þ 15:04N2 ! CH4 þ 2O2 þ CO2
X ¼ 20:9 þ 2H2 O þ 15:04N2
ð5:R14Þ
Consequently, the amount of nitrogen added
to the air in the test cell to render the atmosphere The mixture discharged into the exhaust man-
“inert” with respect to a leak of propane ifold has the composition given by the right-hand
corresponds to 20.9 moles of N2 for every side of Reaction 5.R14. If this “burns” at 700 K,
(5 + 18.8) ¼ 23.8 moles of air, that is, the mix- the final abiabatic flame temperature may be
ture in the cell must be 47 % nitrogen, the bal- calculated on the basis of the reaction
ance being air. (Experimentally, a significantly
lower figure is obtained—c. 40 %. It should be CH4 þ 2O2 þ CO2 þ 2H2 O þ 15:04N2 ! 2CO2
remembered that in the above calculation it is þ 4H2 O þ 15:04N2
assumed that the adiabatic temperature assump- ð5:R15Þ
tion is valid and that the reaction will go to
completion.) The total thermal capacity of the product
gases (2CO2 + 4H2O + 15.04 N2) (at 1000 K)
Example 2 A mechanical engineering research can be shown to be 765.3 J per mole of methane
laboratory contains a six-cylinder internal com- burned. Using Kirchoff’s equation (Equa-
bustion engine that is being used for research into tion 5.19), ΔHc(CH4) at 700 K is calculated as
the performance of spark plugs. The fuel being 802.8 kJ/mol, giving ΔT ¼ 802800/765.3 ¼
used is methane, CH4, and the fuel-air mixture 1049 K. This gives a final temperature of
can be adjusted at will. The combustion products 1749 K, which is significantly higher than the
are extracted from the exhaust manifold through limiting flame temperature (1600 K) discussed
a 30 cm square duct, 20 m long. It is found that above. This indicates that there is a risk of
the engine will continue to operate with a stoi- explosion, and measures should be applied to
chiometric mixture when only three of the prevent this mixture being discharged into
cylinders are firing. If under these conditions the duct.
the average temperature of the gases entering It should be noted that at 700 K there will be a
the duct from the manifold is 700 K, is there a “slow” reaction between methane and the oxygen
risk of an explosion in the duct? present, which could invalidate the tacit assump-
tion that the duct becomes completely filled with
Solution The stoichiometric reaction for meth- the mixture described by the right-hand side of
ane in air is Reaction 5.R13. However, slow oxidation of the
methane will tend to make the mixture less flam-
CH4 þ 2O2 þ 7:52N2 ! CO2 þ 2H2 O þ 7:52N2 mable, and so the calculation gives a conserva-
ð5:R12Þ tive answer.
150 D.D. Drysdale
ed. (P.J. DiNenno et al., eds.), National Fire Protection

Nomenclature Association, Quincy, MA, pp. 1-220–1-230 (2008).
5. A.F. Roberts, Combustion and Flame, 8, p. 245
(1964).
A Area (Equation 5.5) 6. G M Barrow, “Physical Chemistry” 4th Edition,
McGraw-Hill Book Co. (New York, 1961)
Aw Area of ventilation opening 7. NIST-JANAF Thermochemical Tables: see http://
Cp Specific heat kinetics.nist.gov/janaf/
E Internal energy 8. R.A. Strehlow, Combustion Fundamentals, McGraw-
F Force (Equation 5.4) Hill, New York (1984).
9. R.C. Weast, Handbook of Chemistry and Physics,
H Height of ventilation opening Chemical Rubber Co., Cleveland, OH (1973).
H Enthalpy 10. V. Babrauskas and R. Peacock, “Heat Release Rate:
ΔHc Heat of combustion The Single Most Important Variable in Fire Hazard,”
ΔHf Heat of formation in Fire Safety Journal, 18, pp. 255–272 (1992).
11. A. Tewarson, in Flame Retardant Polymeric
ṁ Mass rate of burning Materials (M. Lewin, ed.), Plenum, New York (1982).
ṁair Mass flow rate of air 12. M. Khan, A. Tewarson, and M. Chaos, “Combustion
n Number of moles Characteristics of Materials and Generation of Fire
p Pressure Products,” in SFPE Handbook of Fire Protection
Engineering, 5th ed. Springer, New York, 2015.
q Energy 13. O. Pettersson, S.E. Magnusson, and J. Thor, Fire
Qc Rate of heat release Engineering Design of Structures, Swedish Institute
R Universal gas constant of Steel Construction, Publication, 50 (1976).
T Temperature 14. W.D. Walton and P.H. Thomas, “Estimating
Temperatures in Compartment Fires,” in SFPE Hand-
V Volume book of Fire Protection Engineering, 4th
w Work ed. (P.J. DiNenno et al., eds.), Society of Fire Protec-
tion Engineers, Bethesda, MD, pp. 3-204–3-221
Subscripts (2008).
15. M.L. Bullen and P.H. Thomas, Seventeenth Sympo-
sium (International) on Combustion, Combustion
c Combustion Institute, Pittsburgh, PA (1979).
F Final 16. P.H. Thomas and A.J.M. Heselden, “Fully Developed
f Formation Fires in Compartments,” CIB Report No. 20; Fire
Research Note No. 923, Conseil International du
o Initial Batiment, France (1972).
p Constant pressure 17. D.T. Gottuk and B.Y. Lattimer, “Effect of Combus-
v Constant volume tion Conditions on Species Production,” in SFPE
Handbook of Fire Protection Engineering, 4th
ed. (P.J. DiNenno et al., eds.), National Fire Protection
Superscripts Association, Quincy, MA, pp. 2-67–2-95 (2008).
18. V. Babrauskas, “The Cone Calorimeter,” in SFPE
Pr Products Handbook of Fire Protection Engineering, 4th
R Reactants ed. (P.J. DiNenno et al., eds.), National Fire Protection
Association, Quincy, MA, pp. 3-90–3-108 (2008).
19. V. Babrauskas and S.J. Grayson (eds.), Heat
Release in Fires, Elsevier Applied Science, London
(1992).
References 20. M.L. Janssens, “Calorimetry,” in SFPE Handbook of
Fire Protection Engineering, 5th ed. Springer, 2015.
1. W.J. Moore, Physical Chemistry, 5th ed., Longman,
London (1974).
2. P Atkins and J de Paula, “Atkins’ Physical Chemistry”
9th Edition (Oxford University Press, 2009) D.D. Drysdale is professor emeritus in the BRE Centre
3. D.D. Drysdale, Introduction to Fire Dynamics, 3rd ed., for Fire Safety Engineering, School of Engineering, at the
John Wiley and Sons, Chichester, UK (2011). University of Edinburgh, Scotland. His research interests
4. J.F. Griffiths, “Combustion Kinetics,” in SFPE Hand- lie in fire science, fire dynamics, and the fire behavior of
book of Fire Protection Engineering, 4th combustible materials.
www.ebook777.com
Chemical Equilibrium
6
Raymond Friedman
Introduction The chemical equilibrium calculation yields

not only the temperature but the equilibrium
The temperature of a flame must be known in composition of the products. Thus, the genera-
order to calculate convective and radiative heat tion rate of certain toxic or corrosive products
transfer rates, which control pool-fire burning such as carbon monoxide, nitric oxide, or hydro-
rates, flame spread rates, remote ignitions, dam- gen chloride may be calculated, insofar as the
age to exposed items (e.g., structural steel, assumption of equilibrium is valid.
wiring), and response of thermal fire detectors For a fire in a closed volume, the final pressure
or automatic sprinklers. as well as the temperature will depend on the
Chapter 5 provides a simple technique for dissociations and therefore require a calculation
calculating flame temperature, based on ignoring taking chemical equilibrium into account.
the dissociations that occur at high temperature. From a fire research viewpoint, there is interest
Although the error is small for near-limit flames, in correlating flammability limits, extinguish-
this technique gives answers that are too high. ment, soot formation, toxicity, flame radiation,
For example, if propane (C3H8) burns in stoi- or other phenomena; and chemical equilibrium
chiometric proportions with air at 300 K, and it calculations in some cases will be a useful tool
is assumed that the only products are CO2, H2O, in such correlations.
and N2, then the simple thermochemical calcula- In a later part of this chapter, departure of
tion yields a flame temperature of 2394 K. On the actual fires from chemical equilibrium will be
other hand, if chemical equilibrium is consid- discussed.
ered, so that the species CO, O2, H2, OH, H, O,
and NO are assumed present in the products, then
the flame temperature, calculated by methods The Chemical Equilibrium Constant
described in this section, comes out to be
2268 K. Flame temperature measurements in Consider a chemical transformation, such as
laminar premixed propane-air flames agree with
the latter value. (The discrepancy in flame tem- 2CO þ O2 ! 2CO2 ð6:1Þ
perature caused by neglecting dissociation would
If this process can occur, presumably the reverse
be even greater for fires in oxygen-enriched
process can also occur (principle of microscopic
atmospheres.)
reversibility, or principle of detailed balancing):
2CO2 ! 2CO þ O2 ð6:2Þ

R. Friedman (*)
Retired from FM Global

152 R. Friedman
If both processes occur at finite rates in a closed (where M is any molecule) followed by
system, then, after a sufficient time, a condition of
chemical equilibrium will be reached, after which O þ CO þ M ! CO2 þ M ð6:7Þ
no further change occurs as long as the tempera-
ture and pressure remain constant and no addi- Now, observe how Equation 6.3 can be obtained
tional reactants are introduced. This condition of from this reaction sequence.
equilibrium can be expressed as a mathematical The reverse of O2 + M ! 2O + M, namely
constraint on the system, which, for the gaseous 2O + M ! O2 + M, can also occur, and the
reaction 2CO þ O2 ⇄ 2CO2 , can be written equilibrium constant for this pair of reactions,
which actually do occur, is K6 ¼
p2 CO2 pO 2 pM = pO2 pM ¼ pO 2 = pO2 . (The pM term
K3 ¼ ð6:3Þ
p2 CO pO2 cancels.)
Similarly the reverse reaction CO2 + M !
where the pi are partial pressures (atm)1 and K3 is O + CO + M can occur, and the equilibrium
the equilibrium constant. This expression can be constant is K7 ¼ pCO2 = pCO pO . If we now multi-
rationalized by the following argument.
ply K72 by K6, we obtain
According to the chemical “law of mass
action,” first stated a century ago, the rate of the 2
pCO2 pO 2 p2
forward reaction (Equation 6.1) at a given tem- K 27 K 6 ¼ ¼ 2 CO2
pCO pO pO2 p CO pO2
perature is given by k f p2CO pO2 while the rate of
the reverse reaction (Equation 6.2) is given by kr ¼ K3 ð6:8Þ
p2CO2 : At equilibrium, the forward rate must be
Thus, Equation 6.3 is perfectly valid, even if the
equal to the reverse rate:
“law of mass action” does not correctly describe
k f p2CO pO2 ¼ kr p2CO2 ð6:4Þ the reaction process involving CO and O2.
To get a further understanding of the validity
which may be rearranged to of the equilibrium constant concept, consider the
following facts: CO will not react with O2—even
p2 CO2 kf by the above mechanism involving O atoms—
2
¼ ¼ K3 ð6:5Þ
p CO pO2 kr unless first heated to quite high temperatures.
However, at least a trace of moisture is usually
Although this appears to be a satisfactory explana-
present, and in such cases the reaction occurs by
tion, research over the past hundred years has
the following process, which can occur at lower
shown that chemical reactions in fact rarely pro-
temperatures. First, H and OH are formed
ceed as suggested by the stoichiometric equation.
by dissociation of H2O. Then, the CO is
(This is discussed more fully in Chap. 13.) For
converted by
example, the three-body collision of two CO
molecules and an O2 molecule, resulting in the pCO2 pH
formation of two CO2 molecules, simply does not CO þ OH ⇄ CO2 þ H K 9 ¼ ð6:9Þ
pCO pOH
happen. Rather, the reaction would occur as
follows: while the O2 reacts with H:
O2 þ M ! 2O þ M ð6:6Þ pOH pO
O2 þ H ⇄ OH þ O K 10 ¼ ð6:10Þ
pO 2 pH
If the quantity K92K10 is now calculated,

1
In place of partial pressures, the concentrations of the p2 CO2 pH pO
species in moles/liter can be used in these formulae K 29 K 10 ¼ ð6:11Þ
instead (see Chap. 13). p2CO pO2 pOH
www.ebook777.com
6 Chemical Equilibrium 153
But, the reaction H + O + M ! OH + M Now the system is completely defined by the

can occur, as well as its reverse, OH + M ! simultaneous solution of these three equations.
H + O + M. It does not matter if these reactions The equilibrium constant varies with temperature
are actually important in the rate of oxidation of but is independent of pressure (except at rather
CO in the presence of H2O. As long as these high pressures). It is also independent of the
reactions can occur, then at equilibrium presence of other reactive chemical species.
k f pH pO pM ¼ kr pOH pM
and Generalized Definition of Equilibrium

Constant
kf p
¼ K 12 ¼ OH ð6:12Þ
kr pH pO For a generalized reaction
Substituting this into Equation 6.11 aX1 þ bX2 ⇄ cY1 þ dY2
p2 CO2 K would be given by

¼ K 29 K 10 K 12 ¼ K 3 ð6:13Þ
p2 CO pO2
ð pY1 Þc ð pY2 Þd
K¼
Thus, the ratio p2CO2 = p2CO pO2 is a constant at ð pX1 Þa ð pX2 Þb
equilibrium (at a given temperature) regardless
of the reaction mechanism, even if other Attention should be paid to the manner in which
(hydrogen-containing) species are involved, a chemical reaction is written. For example,
because by the principle of microscopic revers- instead of writing 2CO þ O2 ⇄ 2CO2 one could
ibility, these other species (catalysts) affect the equally well have written CO þ 1=2O2 ⇄ CO2 .
reverse reaction as well as the forward reaction. The equilibrium constant for the latter formula-
Let us now consider the mathematical specifi- tion is
cation of the CO–CO2–O2 system at equilibrium. pCO2
K 16 ¼ ð6:16Þ
The system, at a given temperature and pressure, 1=2
pCO pO2
may be described by three variables, namely the
partial pressures of the three species: pCO, pO2 , By comparison of Equation 6.16 with Equa-
pffiffiffiffiffiffi
and pCO2 . There are already two well-known tion 6.3, it is clear that K 16 ¼ K 3 : If the reaction
constraints on the system: (1) The sum of the was written as 2CO2 ⇄ 2CO þ O2 the equilib-
partial pressures must equal the total pressure, p rium constant would be equal to 1/K3.
pCO þ pO2 þ pCO2 ¼ p ð6:14Þ
and (2) the ratio of carbon atoms to oxygen atoms Simultaneous Equilibria
in the system must remain at the original, pre-
sumably known, value of C/O: In most real chemical systems, one must deal
with a number of simultaneous chemical
C pCO þ pCO2 equilibria. For example, air at 2500 K will con-
¼ ð6:15Þ
O pCO þ 2 pO2 þ 2 pCO2 tain the species N2, O2, NO, and O. The follow-
ing simultaneous equilibria may be considered
A third constraint, that of chemical equilibrium,
provides a third equation involving pCO, pO2 , pO 2
O2 ¼ 2O K 17 ¼ ð6:17Þ
and pCO2 : pO 2
p2 CO2 p2 NO
¼ K3 ð6:3Þ N2 þ O2 ¼ 2NO K 18 ¼ ð6:18Þ
p2 CO pO2 pN2 pO2
154 R. Friedman
p2 NO species one includes, the more difficult is the

N2 þ 2O ¼ 2NO K 19 ¼ ð6:19Þ calculation. There is no need to include any spe-
pN 2 p O 2
cies that will be present in very small quantity at
It is easily seen from the above relations that equilibrium. Some guidelines can be provided.
K19 ¼ K18/K17. Hence, Equations 6.17, 6.18, and For combustion of a C–H–O compound in air,
6.19 are not three independent equations, and any it is usually sufficient to include the species CO2,
two of these equations may be used to describe H2O, N2, O2, CO, H2, OH, H, O, and NO. These
the equilibrium condition; the third would be species are adequate if the air-fuel ratio is suffi-
redundant. To determine the four unknowns, ciently large so that the O/C atomic ratio is
pN2 , pO2 , pNO, and pO, one would solve the greater than one. If the O/C atomic ratio is less
selected two equilibrium relations plus the fol- than one, then solid carbon must be considered,
lowing two relations: as well as many additional gaseous species.
If chlorine is present, then HCl, Cl2, and Cl
pNO þ pN 2 þ pO2 þ pO ¼ p ð6:20Þ must be added. If sulfur is present, then SO2
and SO3 are the primary species, unless there is
And a deficiency of oxygen.
pNO þ 2 pN2
¼ 3:76 ð6:21Þ
pNO þ 2 pO2 þ pO The Quantification of Equilibrium
Constants
where 3.76 is the ratio of nitrogen atoms to
oxygen atoms in air. While a chemist might establish the numerical
If one knows the temperature, the equilibrium value of an equilibrium constant for A ⇄ B by
constants may be calculated from the thermody- direct measurement of the partial pressures of
namic properties of the reactants and products, as A and B in a system at equilibrium, this is rarely
discussed in the next section. However, since the done because it is difficult to make such
various equilibrium reactions release or absorb measurements in a high-temperature system,
energy, and accordingly raise or lower the tem- and it takes a long time to establish equilibrium
perature of an adiabatic system respectively, the in a low-temperature system. Instead, the equi-
determination of equilibrium composition of an librium constant is generally determined from
adiabatic system must proceed simultaneously the thermodynamic relation first deduced by
with the calculation of its temperature; that is, van’t Hoff in 1886 [1]
an energy balance must be satisfied as well as the
equilibrium equations, the atom-ratio equations, ΔFo ¼ RT ln K ð6:22Þ
and the p ¼ ∑ pi equation.
If this equation is applied to A ⇄ 2B at absolute
As a general rule, a gaseous chemical system
temperature T, then K ¼ p2B = pA ; and ΔFo is the
at a given temperature, containing s kinds of
chemical species involving e chemical elements, free energy of two moles (mol) of B at 1 atm and
requires s-e equilibrium relations, e-1 atom-ratio temperature T, minus the free energy of 1 mol of
relations, and a p ¼ ∑ pi equation, in order to A at 1 atm and temperature T. (The superscript o
specify it. If the temperature is unknown, an designates that each substance is in its “standard
energy balance equation is also needed. (If the state,” that is, an ideal gas at 1 atm.) By definition
pressure is unknown but the volume is known, ΔFo ¼ ΔH o TΔSo
then the equation of state must be used in the ¼ ΔEo þ ð pV o Þ TΔSo ð6:23Þ
pressure equation.)
In order to solve an actual problem, one must Accordingly, if ΔSo, the entropy difference, and
select the species to be considered. The more either ΔHo, the enthalpy difference, or ΔEo, the
www.ebook777.com
energy difference, are known for the substances large negative number and can dominate the
involved in an equilibrium at temperature T, then other terms to cause K to be small, that is, the
the equilibrium constant, K, may be calculated. It reaction prefers to go in the reverse, or exother-
happens that ΔSo, ΔHo, and ΔEo are well known mic, direction and reduces the energy of the
for almost all substances expected to be present system. (Most spontaneous reactions are exother-
at equilibrium in combustion gases at any tem- mic.) The Δn term is generally small compared
perature up to 4000 K, so the calculation of with the other terms and represents the work
equilibrium constants is straightforward. done by the expanding system on the surround-
The variation of the equilibrium constant with ings, or the work done on the contracting system
temperature was shown by van’t Hoff [1] to be by the surroundings.
given by In summary, Equation 6.25 represents the bal-
ance of these various tendencies and determines
d ln K ΔH o ΔH the relative proportions of reactants and products
¼ ¼ for ideal gases
dT RT 2 RT 2 at equilibrium. Notice that the term ΔEo/RT
ð6:24Þ becomes small at sufficiently high temperature,
and the entropy term then dominates. In other
Thus, for an exothermic reaction occurring at words, all molecules break down into atoms at
temperature T, ΔH is negative and K decreases sufficiently high temperature, to maximize
as T increases. The converse is true for endother- entropy. The important conclusion from this dis-
mic reactions. cussion is that there is no need to consider rates of
It is appropriate to inquire about the underly- forward and reverse processes to determine
ing physical reason for the value of K to be equilibrium.
governed by ΔFo (actually ΔHo and ΔSo). An Table 6.1 provides values of equilibrium
explanation is as follows: any chemical system constants for 13 reactions involving most species
being held at constant temperature will seek to found in fire products at equilibrium, over a
reduce its energy, E, and to increase its entropy, temperature range from 600 K to 4000 K. Equi-
S. The reduction of energy is analogous to a ball librium constants for other reactions involving
rolling downhill. The increase of entropy is anal- the same species may be obtained by combining
ogous to shuffling a sequentially arranged deck these constants, as in Equation 6.13, or as
of cards, yielding a random arrangement. These illustrated in the examples below.
two tendencies will often affect the equilibrium Table 6.1 does not include the ½ N2 ¼ N
constant in opposite directions. equilibrium, because fire temperatures are gener-
Consider the equation ally not high enough for significant N to form.
Tables 6.2 and 6.3 present information on the
ΔSo ΔEo degree to which various gases are dissociated at
ln K ¼ Δn ð6:25Þ
R RT various temperatures.
In performing calculations, remember that even
where Δn is the increase in the number of moles if a relatively small fraction of dissociation occurs,
of product relative to reactant. Equation 6.25 is a rather large amount of energy may be absorbed in
obtained by combining Equations 6.22 and 6.23 the dissociation, with a corresponding large
with the ideal gas law at constant temperature increase in the energy of the system. For example,
Δ(pVo) ¼ ΔnRT. Inspection of Equation 6.25 if water vapor initially at 2800 K is allowed to
shows that, if ΔSo is a large positive quantity dissociate adiabatically at 1 atm, only 5.7 % of
and ΔSo/R dominates the other terms, K will be the H2O molecules will dissociate, but the temper-
large, that is, the reaction is driven by the “urge” ature will drop from 2800 K to 2491 K; that is, the
to increase entropy. Again, if the reaction is temperature relative to a 300 K baseline is lower
highly endothermic, then –ΔEo/RT will be a by 12.4 %.
156 R. Friedman
Table 6.1 Values of log10 K for selected reactions

KA KB KC KD KE KF KG KH KI KJ KK KL KM
600 –18.574 –16.336 18.633 –2.568 34.405 14.318 –7.210 –3.814 –7.710 –5.641 24.077 8.530 5.036
700 –15.449 –13.599 15.583 –2.085 29.506 12.946 –6.086 –2.810 –6.182 –4.431 20.677 7.368 4.374
800 –13.101 –11.539 13.289 –1.724 25.830 11.914 –5.243 –2.053 –5.031 –3.522 18.125 6.494 3.876
900 –11.272 –9.934 11.498 –1.444 22.970 11.108 –4.587 –1.462 –4.133 –2.814 16.137 5.812 3.486
1000 –9.807 –8.646 10.062 –1.222 20.680 10.459 –4.062 –0.988 –3.413 –2.245 14.544 5.265 3.173
1100 –8.606 –7.589 8.883 –1.041 18.806 9.926 –3.633 –0.599 –2.822 –1.799 13.240 4.816 2.917
1200 –7.604 –6.707 7.899 –0.890 17.243 9.479 –3.275 –0.273 –2.328 –1.389 12.152 4.442 2.702
1300 –6.755 –5.958 7.064 –0.764 15.920 9.099 –2.972 0.003 –1.909 –1.059 11.230 4.124 2.520
1400 –6.027 –5.315 6.347 –0.656 14.785 8.771 –2.712 0.240 –1.549 –0.775 10.438 3.852 2.364
1500 –5.395 –4.756 5.725 –0.563 13.801 8.485 –2.487 0.447 –1.236 –0.527 9.752 3.615 2.229
1600 –4.842 –4.266 5.180 –0.482 12.940 8.234 –2.290 0.627 –0.962 –0.311 9.191 3.408 2.110
1700 –4.353 –3.833 4.699 –0.410 12.180 8.011 –2.116 0.788 –0.720 –0.119 8.420 3.225 2.006
1800 –3.918 –3.448 4.270 –0.347 11.504 7.811 –1.962 0.930 –0.504 0.053 8.147 3.062 1.913
1900 –3.529 –3.102 3.886 –0.291 10.898 7.631 –1.823 1.058 –0.310 0.207 7.724 2.916 1.829
2000 –3.178 –2.790 3.540 –0.240 10.353 7.469 –1.699 1.173 –0.136 0.346 7.343 2.785 1.754
2100 –2.860 –2.508 3.227 –0.195 9.860 7.321 –1.586 1.277 0.022 0.472 6.998 2.666 1.686
2200 –2.571 –2.251 2.942 –0.153 9.411 7.185 –1.484 1.372 0.166 0.587 6.684 2.558 1.625
2300 –2.307 –2.016 2.682 –0.116 9.001 7.061 –1.391 1.459 0.298 0.692 6.396 2.459 1.568
2400 –2.065 –1.800 2.443 –0.082 8.625 6.946 –1.305 1.539 0.419 0.789 6.134 2.368 1.517
2500 –1.842 –1.601 2.224 –0.050 8.280 6.840 –1.227 1.613 0.530 0.879 5.892 2.285 1.469
2600 –1.636 –1.417 2.021 –0.021 7.960 6.741 –1.154 1.681 0.633 0.962 5.668 2.208 1.425
2700 –1.446 –1.247 1.833 0.005 7.664 6.649 –1.087 1.744 0.729 1.039 5.460 2.136 1.384
2800 –1.268 –1.089 1.658 0.030 7.388 6.563 –1.025 1.802 0.818 1.110 5.268 2.070 1.347
2900 –1.103 –0.941 1.495 0.053 7.132 6.483 –0.967 1.857 0.900 1.178 5.088 2.008 1.311
3000 –0.949 –0.803 1.343 0.074 6.892 6.407 –0.913 1.908 0.978 1.240 4.920 1.950 1.278
3100 –0.805 –0.674 1.201 0.094 6.668 6.336 –0.863 1.956 1.050 1.299 4.763 1.896 1.248
3200 –0.670 –0.553 1.067 0.112 6.458 6.269 –0.815 2.001 1.118 1.355 4.616 1.845 1.219
3300 –0.543 –0.439 0.942 0.129 6.260 6.206 –0.771 2.043 1.182 1.407 4.478 1.798 1.192
3400 –0.423 –0.332 0.824 0.145 6.074 6.145 –0.729 2.082 1.242 1.459 4.347 1.753 1.166
3500 –0.310 –0.231 0.712 0.160 5.898 6.088 –0.690 2.120 1.299 1.503 4.224 1.710 1.142
3600 –0.204 –0.135 0.607 0.174 5.732 6.034 –0.653 2.155 1.353 1.547 4.108 1.670 1.119
3700 –0.103 –0.044 0.507 0.188 5.574 5.982 –0.618 2.189 1.404 1.589 3.998 1.632 1.098
3800 –0.007 0.042 0.413 0.200 5.425 5.933 –0.585 2.220 1.452 1.629 3.894 1.596 1.077
3900 0.084 0.123 0.323 0.212 5.283 5.886 –0.554 2.251 1.498 1.666 3.795 1.562 1.058
4000 0.170 0.201 0.238 0.223 5.149 5.841 –0.524 2.280 1.541 1.703 3.700 1.529 1.039
Partial pressures of all gases are expressed in atmospheres (Pascals/101,325). Graphite, C(S), is assigned a value of
unity in the equilibrium expressions for KE and KF
Table 6.2 Temperature (K) at Which a given fraction of

a pure gas at 1 atm is dissociated
Fraction CO2 H2O H2 O2 N2 Table 6.3 Temperature at which air at equilibrium
0.001 1600 1700 2050 2200 4000 contains a given fraction of nitric oxide, at 1 atm
0.004 1800 1900 2300 2400 — Fraction Temperature (K)
0.01 1950 2100 2450 2600 — 0.001 1450
0.04 2200 2400 2700 2900 — 0.004 1750
0.1 2450 2700 2900 3200 — 0.01 2100
0.4 2950 3200 3350 3700 — 0.04 2800
www.ebook777.com
2
5:1 103 ¼ 2:6 105 atm: Since, by pro-
Carbon Formation in Oxygen-
Deficient Systems cess α, 2 mol of CO must decompose for each
mole of CO2 formed, we conclude that
Solid carbon (soot) may be expected to form in 2 2.6 10–5 or 5.2 10–5 mol of CO will
oxygen-deficient combustion products, under decompose to C(s) plus CO2, per mole of CO
some conditions. Since solid carbon does not originally present, after which we will have
melt or boil until extremely high temperatures reached an equilibrium state. In other words,
(~4000 K), we only need concern ourselves with about 1/20,000 of the CO will decompose.
solid carbon C(s), not liquid C(l) or gaseous car- If the original mixture had consisted of CO at
bon C(g). 1 atm plus CO2 at any pressure greater than
Consider pure carbon monoxide at 2000 K. 2.6 10–5 atm, at 2000 K, then we could con-
There are three conceivable ways in which it clude that no carbon whatsoever would form.
might form solid carbon: It can also be shown that addition of a trace of
O2 or H2O to CO at 2000 K would completely
1=2 suppress carbon formation. As a general state-
1 1 pCO2
α : CO ⇄ CðsÞ þ CO2 Kα ¼ ment, for a chemical system containing fewer
2 2 p carbon atoms than oxygen atoms, the equilibrium
1=2CO
1 pO2 condition will favor CO formation rather than
β : CO ⇄ CðsÞ þ O2 Kβ ¼
2 pCO that of solid carbon.
p
γ : CO ⇄ CðsÞ þ O Kγ ¼ O For a carbon-containing system with little or
pCO no oxygen, carbon may or may not form,
depending on the hydrogen partial pressure. For
Note that solid carbon does not appear in any example, carbon may form according to
of the equilibrium expressions. (By convention, a C2 H2 ⇄ CðsÞ þ H2 . The equilibrium expression
solid in equilibrium with gases is assigned a for this reaction is written
value of unity.)
From Table 6.1, we see that, at 2000 K, pH2
¼ K ð¼ 13:9 at 3000KÞ
p C2 H 2
1=2
KE 10:353
Kα ¼ ¼ antilog10 7:469 Again, note that solid carbon does not appear in
KF 2 the expression. If we rewrite the expression in the
¼ 5:1 103 form pH2 > 13:9 pC2 H2 , it becomes the criterion
for suppression of carbon formation at 3000 K. In
1
Kβ ¼ ¼ antilog10 ½0 7:469 ¼ 3:4 108 other words, as long as pH2 is more than 13.9
KF
times as large as pC2 H2 , no carbon will form at
KA 3000 K and any carbon present will be converted
Kγ ¼ ¼ antilog10 ½3:178 7:469 to C2H2. On the other hand, pure C2H2 will
KF
¼ 2:2 1011 decompose to C(s) plus H2 until the H2/C2H2
ratio reaches 13.9, after which no further decom-
position will occur at 3000 K.
We see that Kα, Kβ, and Kγ are all small
Another way to view this is to say that H2,
compared with unity, so very little of the CO
C2H2, and solid carbon at 3000 K will be in a
would decompose by any of these modes. How-
state of equilibrium if and only if the ratio
ever, Kα is much larger than either Kβ or Kγ, so it
pH2 = pC2 H2 ¼ 13:9, and this is true regardless of
is the dominant mode for whatever decomposi-
tion may occur. the quantity of solid carbon present, and also
regardless of the presence of other gases.
Thus, from the expression pCO2 ¼ ðKa pCO Þ2 ,
For a C–H–O–N system, the threshold
and taking pCO as 1 atm, we calculate pCO2 ¼ conditions for equilibrium carbon formation are
158 R. Friedman
Table 6.4 Threshold atomic C/O ratios for carbon for- equilibrium condition is only a limiting case that
mation (equilibrium at 1 atm, N/O ¼ 3.76) real flames may approach. The products of a
Temperature (K) nonluminous laminar flame more than a few
Atomic H/C ratio 1600 2000 2400 2800 millimeters from any cold surface will always
0 1.00 1.00 1.00 1.00 be very nearly in equilibrium.
2 1.00 1.02 1.09 1.30
4 1.00 1.05 1.16 1.56
Sample Problems
somewhat more complicated, but the trends are
illustrated by the calculated values shown in Example 1 Given a mixture of an equal number
Table 6.4 for carbon formation thresholds in of moles of steam and carbon monoxide, what
carbon-hydrogen-air systems at 1 atm. will the equilibrium composition be at 1700 K
It must be noted that carbon forms more read- and 1 atm?
ily in actual flames than Table 6.4 indicates,
because of nonequilibrium effects. In premixed Solution We would expect the species CO, H2O,
laminar flames, incipient carbon formation CO2, and H2 to be present. From Table 6.2, we
occurs at a C/O ratio roughly 60 % of the values see that the equilibria H2 ⇄ 2H, O2 ⇄ 2O, and
shown in Table 6.4. See the next section for H2 O ⇄ 1=2H2 þ OH can all be neglected at
further comments on nonequilibrium. 1700 K, so the species H, O, and OH will not
be present in significant quantities.
Since we have four species involving three
Departure from Equilibrium chemical elements, we will require 4—3 or
1, equilibrium relationship, for the equilibrium
This procedure of specifying chemical systems H2 O þ CO ⇄ H2 þ CO2 .
by equilibrium equations will only yield correct The relationship is
results if the system is truly in equilibrium. If one pH2 pCO2
¼K ð6:26Þ
prepares a mixture of H2 and O2 at room temper- pH2 O pCO
ature and then ages the mixture for a year, it will
be found that essentially nothing has happened In addition, we need 3—1, or 2, atom-ratio
and the system will still be very far from equilib- relations, which are
rium. On the other hand, such a system at a high
H 2 pH 2 þ 2 pH 2 O
temperature characteristic of combustion will : ¼2 ð6:27Þ
C pCO þ pCO2
reach equilibrium in a small fraction of a second.
For example, a hydrogen atom, H, in the pres- (because the original mixture of H2O + CO
ence of O2 at partial pressure 0.1 atm will react so contains two H atoms per C atom) and
fast at 1400 K that its half-life is only about 2 μs.
(At room temperature, the half-life of this reac- O pH2 O þ pCO þ 2 pCO2
: ¼2 ð6:28Þ
tion is about 300 days.) C pCO þ pCO2
Since peak flame temperatures are almost
always above 1400 K, and sometimes as high as (because the original mixture of H2O + CO
2400 K, it would appear that equilibrium would contains two O atoms per C atom). Finally, the
always be reached in flames. However, luminous sum of the partial pressures equals 1 atm:
(yellow) flames rapidly lose heat by radiation, pH2 O þ pCO þ pH2 þ pCO2 ¼ 1 ð6:29Þ
turbulent flames may be partially quenched by
the action of steep velocity gradients, and flames We now have a well-set problem, four equations
burning very close to a cold wall may be partially and four unknowns, which may be solved as soon
quenched by heat transfer to the wall. Thus, the as K is quantified.
www.ebook777.com
We do not find the equilibrium H2 O þ CO ⇄ From Table 6.1, log10KB ¼ –1.601 at 2500 K,
H2 þ CO2 in Table 6.1. However, if we calculate and therefore KB ¼ 0.0251. Upon substitution
(KE/KFKC) from Table 6.1, we see that into Equation 6.32 and elimination of pH2
1=2 1=2 between Equations 6.31 and 6.32, one obtains
KE pCO2 ð1Þ pO2 p H p O2
¼ 2
K F K C ð1Þ pO2 pCO p H2 O pH 2 þ 0:000316 pH 0:00258 ¼ 0 ð6:33Þ
pCO2 pH2
¼ ¼K
pCO pH2 O This equation yields a positive and a negative
root. The negative root has no physical signifi-
From Table 6.1, log10 (KEI/KFKC) at 1700 K cance. The positive root is pH ¼ 0:0506 atm.
¼ 12.180 – 8.011 – 4.699 ¼ –0.51, and K ¼ Then, Equation 6.32 yields pH2 ¼ 4:08 atm, and
antilog10 (–0.51) ¼ 0.309. the total final pressure is 4.08 + 0.0506 ¼
Upon substituting K ¼ 0.309 into Equa- 4.13 atm. The degree of dissociation, α, comes
tion 6.26, and then simultaneously solving out to be 0.0062. (This is less dissociation than
Equations 6.26, 6.27, 6.28, and 6.29, we obtain indicated by Table 6.2 because the pressure is
pCO2 ¼ pH2 ¼ 0:179 atm well above 1 atm.)
and Example 3 Propane is burned adiabatically at

1 atm with a stoichiometric proportion of air.
pH2 O ¼ pCO ¼ 0:321 atm Calculate the final temperature and composition.
Example 2 One mole of hydrogen is introduced The initial temperature is 300 K.
into a 50-L vessel that is maintained at 2500 K.
Solution The problem must be solved by a series
How much dissociation will occur, and what will
the pressure be? of iterations. The first step is to assume a final
temperature, either based on experience or by
selecting a temperature substantially below the
Solution Let α be the degree of dissociation of
the hydrogen defined by ( α ¼ ð pH =2Þ= value calculated by assuming that CO2 and H2O

are the only products of combustion. The second
pH2 þ ð pH =2Þ ). Thus, α ranges from zero to
step is to solve the set of equations that specify
one. One mole of H2 partially dissociates to pro-
the equilibrium composition at the assumed final
duce 2 α mol of H, leaving 1—α mol of H2. The
temperature. The third step is to consult an over-
total number of moles is then 2 α + 1 – α, or
all enthalpy balance equation, which will show
α + 1. In view of the definition of α, the total
that the assumed final temperature was either too
number of moles present is ( pH þ pH2 =

high or too low. The fourth step is to assume an
pH2 þ ð pH =2Þ ). appropriate new final temperature. The fifth and
By the ideal gas law, PV ¼ nRT. sixth steps are repeats of the second and third
pH þ pH2 steps. If the correct final temperature is now
pH þ pH2 ð50Þ ¼ ð0:08206Þð2500Þ found to be bracketed between these two
pH2 þ ð pH =2Þ
assumed temperatures, then an interpolation
ð6:30Þ should give a fairly accurate value of the true
which reduces to final temperature. Additional iterations may be
made to improve the accuracy of the results to the
pH
pH2 þ ¼ 4:103 ð6:31Þ degree desired. As a guess, the final temperature
2 is assumed to be 2300 K.
The equilibrium equation is Now the equilibrium equations at 2300 K are
set up. The species to be considered are three
pH
1=2 ¼ K B ð6:32Þ principal species: CO2, H2O, and N2, and seven
pH 2 minor species: H2, O2, OH, H, O, CO, and NO.
160 R. Friedman
(Based on chemical experience, the following H 8 p þ pOH þ 2 pH2 O þ pH2

: ¼ H ð6:41Þ
possible species may be neglected at 2300 K N 37:6 2 pN2 þ pNO
when stoichiometric oxygen is present: N, C(g),
NH, CN, CH, C2, HO2, HCN, O3, C3, NO2, HNO, O 10
: ¼
C2H, CH2, C2O, CHO, and NH2.) Thus, we con- C 3
sider ten species involving four elements, so pO þ pOH þ pNO þ pCO þ pH2 O þ 2 pO2 þ 2 pCO2
10—4 or 6, equilibrium equations are needed. pCO2 þ pCO
Any six independent equilibria may be selected. ð6:42Þ
We can assure independence by requiring that
each successive equilibrium expression we Finally,
write will introduce at least one new chemical pCO2 þ pH2 O þ pN2 þ pH2 þ pO2 þ pOH þ pH
species. Observe that this requirement is met in
þ pO þ pCO þ pNO ¼ 1
the following list:
ð6:43Þ
1 pCO2 KE
CO þ O2 ¼ CO2 1=2 ¼ K ¼ From Table 6.1 at 2300 K:
2 pCO pO2 KF
ð6:34Þ log10 x x
KE 9.001 –
1 pO KF 7.061 –
O2 ¼ O 1=2 ¼ K A ð6:35Þ KE/KF 9.001–7.061 87.1
2 pO 2
KA –2.307 0.00493
KG –1.391 0.0406
1 1 pNO KC 2.682 481
O2 þ N2 ¼ NO 1=2 ¼ K G
2 2 pO2 pN2 KD –0.116 0.766
KB –2.016 0.00964
ð6:36Þ
We insert these K values into Equations 6.34,
1 pH 2 O 6.35, 6.36, 6.37, 6.38, and 6.39, and then solve
H2 þ O2 ¼ H2 O 1=2 ¼ K C ð6:37Þ
2 pH2 pO2 the set of ten equations, Equations 6.34, 6.35,
6.36, 6.37, 6.38, 6.39, 6.40, 6.41, 6.42, and
1 pOH 6.43, for the equilibrium values of the ten partial
H2 þ O2 ¼ H2 O 1=2 ¼ K D ð6:38Þ pressures at 2300 K. This solution may be
2 pH 2 pO 2 obtained by a tedious set of successive
approximations. The first approximation is
1 pH obtained by solving for the three principal spe-
H2 ¼ H 1=2 ¼ K B ð6:39Þ
2 pH 2 cies N2, CO2, and H2O, assuming the partial
pressures of the remaining species are zero.
Four additional equations are needed to deter- Then, using this trial value of pCO2 , solve for
mine the ten unknown partial pressures. These pCO and pO2 , using Equation 6.34 and assuming
are three atom-ratio equations and a summation that pCO ¼ 2 pO2 . Next, using pH2 O and pO2 as
of the partial pressures to equal the total pressure. determined, use Equation 6.37 to determine a
To obtain the atom ratios, we take air to consist trial value of pH2 . Then, using all the foregoing
of 3.76 parts of N2 (by volume) per part of O2, partial pressures, determine pO from Equa-
neglecting argon, and other species. Then, from tion 6.35, pNO from Equation 6.36, pOH from
stoichiometry, C3H8 + 5O2 + (5 3.76)N2 ! Equation 6.38, and pH from Equation 6.39.
3CO2 + 4H2O + 18.8 N2. Thus, ten trial values of the partial pressures are
H 8 pH þ pOH þ 2 pH2 O þ 2 pH2 found. However, upon substitution into
: ¼ ð6:40Þ Equations 6.40, 6.41, 6.42, and 6.43, none of
C 3 pCO2 þ pCO
these equations will be quite satisfied. The partial
www.ebook777.com
pressures of the principal species must then be reactant temperature Tr, and nj and Hj are the
adjusted so as to satisfy Equations 6.40, 6.41, number of moles and the enthalpy per mol of
6.42, and 6.43, and then a second iteration with each product species at product temperature Tp.
the equilibrium equations must be carried out to The enthalpy of each reactant or product spe-
establish new values for the minor species. After cies x at temperature T is given by
four or five such iterations, the results should
converge to a set of partial pressures satisfying H x, T ¼ ΔH of þ H o H 298
o
ð6:45Þ
298:15
all equations.
A faster method is to use a computer program where (ΔHfo, 298.15)x
is the enthalpy of formation
to solve the equations. (See the following of a mol of species x from its constituent
section.) elements in their standard states at 298 K (see
The equilibrium partial pressures at 2300 K Chap. 5). These constituent elements are H2, O2,
will come out to be: N2, and C(s), so ΔHfo, 298.15 for each of these four
species is zero, by definition.
PN2 0.7195 atm Values of (ΔHfo)298.15 and Ho H298 o 2
for
PH2 O 0.1474 atm
various species are contained in Table 6.5 on
PCO2 0.1006 atm
page 1–109. Substitution of numerical values
PCO 0.0143 atm
into Equation 6.44 yields:
PO2 0.0066 atm
PH2 0.0038 atm o
H 2300
POH 0.0037 atm o
Reactant (ΔHf )298.15 o
H 298 o
H2300 ni o
niHi.2300
PNO 0.0028 atm species (kJ/mol) (kJ/mol) (kJ/mol) (mol) (kJ)
PH 0.0006 atm C3H8 –103.85 0.16 –103.69 0.0383 –3.971
PO 0.0004 atm O2 0 0.05 0.05 0.1915 +0.010
N2 0 0.05 0.05 0.7200 +0.036
Now, we must determine if 2300 K was too Total ¼ 3.925
high or too low a guess, by writing the enthalpy
balance equation (see Chap. 5). and
As a basis for the enthalpy balance, we o
H 2300
assume that we have exactly 1 mol of products, Product o
(Hfo)298.15 H 298 o
H2300 ni o
niHi.2300
at 1 atm. Then, if PCO2 ¼ 0:1006 atm (see species (kJ/mol) (kJ/mol) (kJ/mol) (mol) (kJ)
above), we must have 0.1006 mol of CO2. Simi- N2 0 66.99 66.99 0.7195 +48.199
larly, we have 0.0143 mol of CO. Since these are H2O –241.83 88.29 –153.54 0.1474 –22.632
the only two carbon compounds in the products, CO2 –393.52 109.67 –283.85 0.1006 –28.555
and since 3 mol of CO2 + CO must form from CO –110.53 67.68 –42.85 0.0143 –0.613
each mole of C3H8 burned, it follows that O2 0 70.60 70.60 0.0066 +0.466
(0.1006 + 0.0143)/3 ¼ 0.0383 mol of C3H8 H2 0 63.39 63.39 0.0038 +0.241
must have burned. Since the original C3H8-air OH 38.99 64.28 103.27 0.0037 +0.382
mixture was stoichiometric, it follows that the NO 90.29 68.91 159.20 0.0028 +0.446
reactants also consisted of 5 0.0383 ¼ 0.1915 H 218.00 41.61 259.61 0.0006 +0.156
O 249.17 41.96 291.13 0.0004 0.116
mol of O2 and 3.76 0.1915 ¼ 0.7200 mol of
Total ¼ 1.794
N2. (Thus, a total of 0.9498 mol of reactant form
1 mol of product, if the product is indeed at The enthalpy of the products (–1.794 kJ) is
equilibrium at 2300 K.) seen to be 2.131 kJ larger than the enthalpy of
The enthalpy balance equation is the reactants (–3.925 kJ). To put this 2.131 kJ
X X
ni H i , T r ¼ n jH j, T p ð6:44Þ
2
If H o H 298
o
is not available from a table, it may be
Z T
where ni and Hi are the number of moles and the evaluated from the equation H o H 298
o
¼ C p dT:
298
enthalpy per mol of each reactant species at For C3H8, Cp ¼ 0.09 kJ/molK at 298 K.
Table 6.5 Enthalpies of selected combustion productsa
162
Species N2 O2 O NO H2 H H2O (g) OH CO2 CO

(ΔHfo)298.15 0.00 kJ/mol 0.00 kJ/mol 249.17 kJ/mol 90.29 kJ/mol 0.00 kJ/mol 218.00 kJ/mol –241.83 kJ/mol 38.99 kJ/mol –393.52 kJ/mol –110.53 kJ/mol
o o o o o o o o o o
H o H 298 ; H o H 298 ; H o H 298 ; H o H 298 ; H o H 298 ; H o H 298 ; H o H 298 ; H o H 298 ; H o H 298 ; H o H 298 ;
Temp(K) kJ/mol kJ/mol kJ/mol kJ/mol kJ/mol kJ/mol kJ/mol kJ/mol kJ/mol kJ/mol
100 –5.77 –5.78 –4.52 –6.07 –5.47 –4.12 –6.61 –6.14 –6.46 –5.77
200 –2.86 –2.87 –2.19 –2.95 –2.77 –2.04 –3.28 –2.97 –3.41 –2.87
298 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
300 0.05 0.05 0.04 0.05 0.05 0.04 0.06 0.05 0.07 0.05
400 2.97 3.03 2.21 3.04 2.96 2.12 3.45 3.03 4.01 2.97
500 5.91 6.08 4.34 6.06 5.88 4.20 6.92 5.99 8.31 5.93
600 8.90 9.24 6.46 9.15 8.81 6.28 10.50 8.94 12.92 8.94
700 11.94 12.50 8.57 12.31 11.75 8.35 14.18 11.90 17.76 12.02
800 15.05 15.84 10.67 15.55 14.70 10.43 17.99 14.88 22.82 15.18
900 18.22 19.24 12.77 18.86 17.68 12.51 21.92 17.89 28.04 18.40
1000 21.46 22.70 14.86 22.23 20.68 14.59 25.98 20.94 33.41 21.69
1100 24.76 26.21 16.95 25.65 23.72 16.67 30.17 24.02 38.89 25.03
1200 28.11 29.76 19.04 29.12 26.80 18.74 34.48 27.16 44.48 28.43
1300 31.50 33.34 21.13 32.63 29.92 20.82 38.90 30.34 50.16 31.87
1400 34.94 36.96 23.21 36.17 33.08 22.90 43.45 33.57 55.91 35.34
1500 38.40 40.60 25.30 39.73 36.29 24.98 48.10 36.84 61.71 38.85
1600 41.90 44.27 27.38 43.32 39.54 27.06 52.84 40.15 67.58 42.38
www.ebook777.com
1700 45.43 47.96 29.46 46.93 42.84 29.14 57.68 43.50 73.49 45.94
1800 48.98 51.67 31.55 50.56 46.17 31.22 62.61 46.89 79.44 49.52
1900 52.55 55.41 33.63 54.20 49.54 33.30 67.61 50.31 85.43 53.12
2000 56.14 59.17 35.71 57.86 52.95 35.38 72.69 53.76 91.45 56.74
2100 59.74 62.96 37.79 61.53 56.40 37.46 77.83 57.25 97.50 60.38
2200 63.36 66.77 39.88 65.22 59.88 39.53 83.04 60.75 103.57 64.02
2300 66.99 70.60 41.96 68.91 63.39 41.61 88.29 64.28 109.67 67.68
2400 70.64 74.45 44.04 72.61 66.93 43.69 93.60 67.84 115.79 71.35
2500 74.30 78.33 46.13 76.32 70.50 45.77 98.96 71.42 121.93 75.02
2600 77.96 82.22 48.22 80.04 74.09 47.85 104.37 75.01 128.08 78.71
R. Friedman
6
2700 81.64 86.14 50.30 83.76 77.72 49.92 109.81 78.63 134.26 82.41
2800 85.32 90.08 52.39 87.49 81.37 52.00 115.29 82.27 140.44 86.12
2900 89.01 94.04 54.48 91.23 85.04 54.08 120.81 85.92 146.65 89.83
3000 92.71 98.01 56.58 94.98 88.74 56.16 126.36 89.58 152.86 93.54
3100 96.42 102.01 58.67 98.73 92.46 58.24 131.94 93.27 159.09 97.27
3200 100.14 106.02 60.77 102.48 96.20 60.32 137.55 96.96 165.33 101.00
Chemical Equilibrium
3300 103.85 110.05 62.87 106.24 99.96 62.40 143.19 100.67 171.59 104.73
3400 107.57 114.10 64.97 110.00 103.75 64.48 148.85 104.39 177.85 108.48
3500 111.31 118.16 67.08 113.77 107.55 66.55 154.54 108.12 184.12 112.22
C(s) F2 F HF Cl2 Cl HCl Br2 Br HBr
0.00 kJ/mol 0.00 kJ/mol 78.91 kJ/mol –272.55 kJ/mol 0.00 kJ/mol 121.29 kJ/mol –92.31 kJ/mol 0.00 kJ/mol 111.86 kJ/mol –36.44 kJ/mol
o o o o o o o o o
H o H 298 ; H o H 298 ; H o H 298 ; H o H 298 ; H o H 298 ; H o H 298 ; H o H 298 ; H o H 298 ; H o H 298 ;
o
kJ/mol kJ/mol kJ/mol kJ/mol kJ/mol H o H 298 ; kJ/mol kJ/mol kJ/mol kJ/mol kJ/mol
–0.99 –5.92 –4.43 –5.77 –6.27 –4.19 –5.77 –21.72 –4.12 –5.77
–0.67 –2.99 –2.23 –2.86 –3.23 –2.10 –2.86 –16.82 –2.04 –2.86
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
0.02 0.06 0.04 0.05 0.06 0.04 0.05 0.14 0.04 0.05
1.04 3.28 2.30 2.97 3.54 2.26 2.97 34.61 2.12 2.97
2.36 6.64 4.53 5.88 7.10 4.52 5.89 38.31 4.20 5.90
3.94 10.11 6.72 8.80 10.74 6.80 8.84 42.02 6.28 8.87
5.72 13.66 8.90 11.73 14.41 9.08 11.81 45.76 8.36 11.88
7.64 17.27 11.05 14.68 18.12 11.34 14.84 49.51 10.46 14.96
9.67 20.91 13.19 17.64 21.84 13.59 17.91 53.27 12.57 18.10
11.79 24.59 15.33 20.64 25.59 15.82 21.05 57.03 14.70 21.30
13.99 28.30 17.45 23.68 29.34 18.03 24.24 60.81 16.84 24.56
16.24 32.03 19.56 26.76 33.10 20.23 27.48 64.58 19.01 27.87
18.54 35.77 21.67 29.87 36.88 22.41 30.78 68.37 21.20 31.24
20.88 39.54 23.78 33.04 40.66 24.60 34.12 72.16 23.40 34.65
23.25 43.32 25.89 36.24 44.45 26.77 37.51 75.96 25.61 38.10
25.66 47.11 27.99 39.48 48.25 28.93 40.93 79.76 27.85 41.59
28.09 50.91 30.09 42.76 52.05 31.09 44.39 83.57 30.09 45.11
(continued)
163
164
C(s) F2 F HF Cl2 Cl HCl Br2 Br HBr

0.00 kJ/mol 0.00 kJ/mol 78.91 kJ/mol –272.55 kJ/mol 0.00 kJ/mol 121.29 kJ/mol –92.31 kJ/mol 0.00 kJ/mol 111.86 kJ/mol –36.44 kJ/mol
o o o o o o o o o
H o H298 ; H o H 298 ; H o H 298 ; H o H 298 ; H o H 298 ; H o H 298 ; Ho H 298 ; H o H 298 ; H o H 298 ;
o o
kJ/mol kJ/mol kJ/mol kJ/mol kJ/mol H H 298 ; kJ/mol kJ/mol kJ/mol kJ/mol kJ/mol
30.55 54.72 32.18 46.09 55.86 33.23 47.89 87.38 32.35 48.66
33.02 58.54 34.28 49.44 59.68 35.38 51.41 91.20 34.61 52.24
35.53 62.38 36.37 52.83 63.51 37.51 54.96 95.02 36.88 55.84
38.05 66.22 38.46 56.25 67.34 39.64 58.53 98.85 39.15 59.46
40.58 70.07 40.55 59.69 71.18 41.77 62.12 102.68 41.43 63.10
43.13 73.93 42.64 63.17 75.02 43.89 65.73 106.52 43.70 66.76
45.71 77.80 44.73 66.66 78.88 46.02 69.37 110.36 45.98 70.44
48.29 81.67 46.82 70.18 82.74 48.13 73.01 114.20 48.26 74.13
50.89 85.55 48.91 73.73 86.61 50.25 76.68 118.05 50.54 77.83
53.50 89.45 50.99 77.29 90.50 52.36 80.36 121.91 52.81 81.55
56.13 93.35 53.08 80.87 94.39 54.48 84.06 125.77 55.09 85.28
58.77 97.25 55.17 84.47 98.29 56.58 87.76 129.63 57.36 89.02
61.43 101.16 57.25 88.09 102.21 58.69 91.48 133.49 59.63 92.77
64.09 105.08 59.34 91.72 106.14 60.79 95.21 137.37 61.89 96.53
66.78 109.01 61.42 95.37 110.08 62.90 98.95 141.24 64.15 100.31
69.47 112.94 63.50 99.03 114.03 65.00 102.70 145.13 66.41 104.09
72.17 116.88 65.59 102.71 118.00 67.10 106.46 149.01 68.67 107.88
www.ebook777.com
74.89 120.83 67.67 106.39 121.98 69.20 110.23 152.90 70.92 111.68
a
These data are taken from the JANNAF thermochemical tables [2]
R. Friedman
difference in perspective, note that the heat of temperature and composition in a combustion
combustion of 0.0383 mol of propane at 298 K, process, a computer program for executing
to form 3 mol of CO2 and 4 mol of H2O per mole these calculations would be desirable. Fortu-
of propane, is 0.0383 (3 393.52 + 4 241.83 nately, such programs have been developed.
– 103.85) ¼ 78.29 kJ. Thus, the 2.131 kJ dis- However, the user of a computer program
crepancy when compared with 78.29 kJ is rather should be warned that thorough understanding
small, showing that the 2300 K “first guess” was of the material in this chapter is needed to avoid
very close. Since the products, at 2300 K, are misinterpreting the computer output. Further,
seen to have a slightly higher enthalpy than the given such understanding, simple manual
reactants, the correct temperature must be calculations can be performed to obtain indepen-
slightly less than 2300 K. dent checks of the computer output.
To continue the calculation, the next step is to One program, entitled GASEQ, can be used
assume that the final temperature is 2200 K instead with any computer using Windows. It can be
of 2300 K. The details will not be presented, but downloaded from http://www.gaseq.co.uk.
this will yield a new and slightly different set of Alternatively, a program may be obtained from
values of the ten partial pressures of the products. Reaction Design, 6440 Lusk Blvd, Suite D209,
Thus, a new enthalpy balance may be attempted, in San Diego, CA 92121. Their e-mail address is
the same manner as before. When this is done, the <chemkin@ReactionDesign.com>.
result will be that this time the enthalpy of the These programs will calculate the final equi-
reactants will come out to be slightly higher than librium conditions for adiabatic combustion at
the enthalpy of the products, showing that the either constant pressure or constant volume,
correct temperature is above 2200 K. given the initial conditions. For the constant-
An interpolation may be made between the pressure calculations, one specifies the initial
2200 K enthalpy discrepancy and the 2300 K temperature, the pressure, and the identities and
enthalpy discrepancy, which will show that the relative proportions of the reactants. The com-
correct final temperature is 2268 K. Furthermore, puter programs contain the properties of selected
the partial pressures of each product species may reactants including: air, oxygen, nitrogen, hydro-
be obtained by interpolating between the 2200 K gen, graphite, methane, acetylene, ethylene, eth-
partial pressures and the 2300 K partial ane, propane, butane, 1-butene, heptane, octane,
pressures, with results as follows: benzene, toluene, JP-4, JP-5, methanol, ethanol,
and polyethylene. If the fire only involves
T ¼ 2268 K PN2 0.7207 atm reactants from this list, no further input is neces-
PH2 O 0.1484 atm sary. If the fire involves a reactant not on this list,
PCO2 0.1026 atm the input data must include the elemental com-
PCO 0.0125 atm position and the enthalpy of formation of the
PO2 0.0059 atm reactant at 298 K, as well as enthalpy versus
PH2 0.0034 atm temperature data for the reactant over the tem-
POH 0.0032 atm perature range from 298 K to the initial tempera-
PNO 0.0025 atm ture. (If the initial temperature is 298 K, the last
PH 0.0005 atm item is not needed.)
PO 0.0003 atm The computer programs can handle reactants
containing any of the following elements: A,
Al, B, Br, C, Cl, F, Fe, H, He, K, Li, Mg, N,
Computer Programs for Chemical Na, Ne, O, P, S, Si, and Xe. Data are included in
Equilibrium Calculations the program on all known compounds, including
liquids and solids, that can form at elevated
In view of the extremely tedious calculations temperatures from combinations of these
needed for determination of the equilibrium elements. It is not necessary for the user to
166 R. Friedman
specify which product species to consider. The K Equilibrium constant (based on partial
program can consider them all, and will print out pressures expressed in atmospheres)
all equilibrium species present with mole K Degrees Kelvin
fractions greater than 5 10–6, unless instructed n Number of moles (e.g., a mole of oxygen
to print out trace values down to some lower is 32 g)
specified level. pi Partial pressure of ith species (atm)
The program can calculate Chapman-Jouguet p Total pressure (atm)
detonation products as well as constant-pressure R Gas constant (kJ/mol K)
or constant-volume combustion products, if ΔSo Entropy of products relative to entropy of
desired. reactants, all at temperature T and 1 atm
An addition to the program permits calcula- (kJ/mol)
tion of viscosity and thermal conductivity of T Absolute temperature (K)
gaseous mixtures, selected from 154 gaseous
species, at temperatures from 300 K to 5000 K. References
1. J. van’t Hoff, cf. G. Lewis, M. Randall, K. Pitzer, and
Nomenclature L. Brewer, Thermodynamics, McGraw-Hill, New York
(1961).
2. D.R. Stull and H. Prophet, JANNAF Thermochemical
Cp Heat capacity at constant pressure Tables, 2nd ed., NDRS-NBS 37, National Bureau of
(kJ/molK) Standards, Washington, DC (1971).
ΔEo Energy of products relative to energy of
reactants, all at temperature T and 1 atm
Raymond Friedman was with Factory Mutual Research
(kJ/mol) from 1969 through 1993. During most of this time he was
ΔFo Free energy of products relative to free vice president and manager of their Research Division.
energy of reactants, all at temperature Currently he is an independent consultant. He has past
experience at Westinghouse Research Laboratories and
T and 1 atm (kJ/mol)
Atlantic Research Corporation. He is a past president of
ΔHo Enthalpy of products relative to enthalpy The Combustion Institute, past vice chairman and past
of reactants, all at temperature T and secretary of the International Association for Fire Safety
1 atm (kJ/mol) Science, and an expert in fire research and combustion.
www.ebook777.com
Thermal Decomposition of Polymeric

Materials 7
Artur Witkowski, Anna A. Stec, and T. Richard Hull
Introduction homopolymer. If the material is composed of

more than one type of repeat unit it is known as
Polymers are composed of large numbers of a copolymer.
repeat units forming very long chain molecules. Thermal decomposition is “a process of exten-
Most polymers are based on carbon, and hence sive chemical species change caused by heat”.
known as organic polymers. The long-chain Thermal degradation is “a process whereby
structure means that polymers can exist in solid the action of heat or elevated temperature on a
or liquid form, but are too large to be volatile. material, product, or assembly causes a loss of
Polymers fuel the vast majority of unwanted physical, mechanical, or electrical properties”
fires, as wood, paper, fabrics, foams and plastics. [1]. Used correctly, thermal degradation may
Flaming combustion is a gas phase process, and describe processes occurring before around 1 %
it is necessary to understand the stages in the of the mass is lost, while thermal decomposition
conversion of long molecular chains into volatile includes the entire mass loss process.
fragments. This is often referred to as “pyrolysis”
or “gasification”, but these terms encompass a
complex set of chemical and physical processes, Polymer Classification
leading to the production of volatile flammable
molecules. Polymers represent the largest class of combusti-
ble materials fuelling unwanted fires. The differ-
ent ways they can be subdivided provides a
Polymeric Materials useful introduction to this wide and important
class of materials.
A polymer is a large molecule constructed from
many smaller structural units called monomers, Natural, Synthetic, Semi-natural
covalently bonded together in any conceivable and Biobased
pattern (but often, and most simply in long Before the widespread use of synthetic polymers
chains). in the second half of the twentieth century almost
If the material is composed of only one type of all unwanted fires were fuelled by natural
repeating structural unit, it is known as a polymers such as wood, paper, cotton, wool etc.
Synthetic polymers are generally derived from
oil or coal, and share the flammability of those
A. Witkowski (*) • A.A. Stec • T.R. Hull
raw materials. The ease of manufacture and
Centre for Fire and Hazards Science, University of
Central Lancashire (UCLan), Preston, Lancashire PR1 processing of synthetic polymers has driven the
2HE, UK increase in their use. As most of the common

168 A. Witkowski et al.
synthetic polymers are more flammable then (X is OH). Other “polymers” such as
their natural counterparts, this has also increased polyamides, polyesters, epoxy resins and
the number and severity of unwanted fires, polyurethanes are actually classes of materials
despite significant advances in fire prevention, with similar bonding between the repeat units,
detection and control. but different structures within each repeat unit.
Semi-natural polymers are typically purified,
and chemically modified natural polymers, such Chemistry of Polymerisation
as cellulose acetate, cellophane or rayon. Drivers Synthetic polymers are made by chemical reac-
towards greater environmental sustainability tion of monomers to form long polymer chains.
have led to increased developments in biobased Since the reverse process, the decomposition of
polymers, which includes the semi-natural polymers into smaller volatile fuel molecules, is
polymers and a new polymeric materials, using often related to their synthesis, this can also
raw materials derived from living, rather than the provide insight into their decomposition
fossilised carbon sources. It seems likely that behaviour. Broadly the process occurs either by
these materials, such as polylactic acid (PLA) step-growth, or chain-growth polymerisation.
and styrene-soya oil-divinyl benzene (SSD), Step-growth polymerisation occurs by chemical
will increase their market share in the next reaction of two functional groups to form the
decade. linkage, with the release of a small molecule
(such as water). This is known as step-growth
Chemical Composition of the Repeat Units polymerization as it takes place one molecule at a
Most natural polymeric materials involved in time. Figure 7.2 shows the familiar esterification
fires are cellulose based. Cellulose is a polymer reaction, in this case used to produce a polyester.
with alternating repeat units of glucose (Fig. 7.1). As water is often the molecule released, this
Several of the common synthetic polymers, process is also known as condensation polymeri-
the so-called vinyl polymers made from vinyl zation. The synthetic process is slow, and the
monomers (CH2 ¼ CHX), have a repeat lack of water or other secondary product in the
units of (CH2CHX), such as polyethylene polymer prevents decomposition being simply
(X is H); polypropylene (X is CH3); polyvinyl the reverse of polymerisation. The second pro-
chloride (X is Cl); polystyrene (X is C6H5); cess, chain-growth polymerisation, involves
polyacrylonitrile (X is C N); polyvinyl opening of double bonds to form consecutive
acetate (X is OCOCH3); polyvinyl alcohol links in the polymer chain. For example, styrene
Fig. 7.1 Part of a cellulose polymer chain, and a single glucose unit (monomer)
www.ebook777.com
7 Thermal Decomposition of Polymeric Materials 169
H2O
HO
O
OH
HOOC COOH + HO H
O O
O n
Fig. 7.2 Step-growth polymerisation of polyethylene terephthalate (PET)
CH2 CH2 CH2 CH2 CH2 CH2

HC HC HC HC HC HC
+ + + + +
H2 H2 H2 H2 H2
C C C C C
H H H H H H
C C C C C C
Fig. 7.3 Chain-growth polymerisation of poly(styrene) (PS)
monomers combine to form polystyrene, without the monomer. The initiator splits to form two free
release of a secondary product (Fig. 7.3). radicals, which attach themselves to a carbon
The polymerisation reaction is known as atom of the monomer. When this occurs, the
chain-growth or addition polymerisation. Typi- reactive site is transferred to another carbon
cally, once the monomer becomes activated, the atom in the monomer, and the chain begins to
chain grows extremely quickly (less than 1 s), so, propagate (Fig. 7.4). Finally the reactivity of the
during the polymerisation process, only long propagating chain end is lost by combination of
polymer chains and volatile monomer units will two reactive sites, or rearrangement, collectively
be present. It is highly likely that some monomer known as termination. In order to maximise the
will remain at the end of the polymerisation degree of polymerization, the remaining
process, since the final material will have very monomers must diffuse towards the reactive
long chains and a high viscosity. During the chain end, before termination reactions occur.
decomposition of polystyrene, the reverse pro- Addition polymerisation can involve free rad-
cess occurs, resulting in monomer, dimer and ical, cationic, anionic, catalytic, or ring opening
trimer predominating in the vapour phase. processes. The means of polymerisation will
Addition polymerization involves three dis- affect the degree of branching, and the molecular
tinct stages—initiation, propagation and termina- weight and dispersity of the polymer (section
tion. To start the process, an initiator is added to “Molecular Mass or Polymer Chain Length”),
I + I + I + etc.
X X X X X X
Fig. 7.4 Free radical polymerisation of an alkene (CH2 ¼ CHX)
which will in turn affect its decomposition molecular weight less than 500,000 (around
(section “Decomposition Mechanisms”). 5000 repeat units) has sufficiently low viscosity
to be melt-processable, whereas PMMA with
Thermoplastics and Thermosets molecular weight of 5,000,000 does not soften
Synthetic polymers (often referred to as plastics, sufficiently to allow its shape to be changed on
because of their mouldability) can be subdivided heating, hence it is known as cast PMMA.
into those which can be repeatedly deformed The number of repeating units will usually vary
by heating (thermoplastics), so they are melt- as a statistical distribution of chain lengths. This
processable; and those which once the may be classified as a number-average molecular

polymerisation process is complete, cannot be mass Mn , or a weight-average molecular mass

melted or have their shape changed by heating, Mw . Within most synthetic polymers, the most
and decompose directly from solid to vapour common unit is actually the monomer, although
(thermosets or thermosetting polymers). Most this only represents a tiny proportion by mass.
thermosets have additional covalent bonds forming Natural polymers often have narrower molecular
crosslinkages between the polymer chains. mass ranges, or even identical molar masses.
Common thermoplastics include polyethylene
The ratio Mn =Mw is known as the dispersity
(PE), polypropylene (PP), polystyrene (PS),
(or heterogeneity index), providing a simple
polymethylmethacrylate (PMMA), polyvinyl
index of the range of chain lengths present. An
chloride (PVC), polyamides (PA) and some
enzyme, where all the molecules have the same
polyesters and polyurethanes.
molecular mass is described as monodisperse, a
Thermosets are generally stronger, but more
polymer produced by anionic polymerisation
brittle than thermoplastics, have higher thermal
will typically have a dispersity around 1.1, a
stability, higher dimensional stability, higher
step growth polymer will have a dispersity
rigidity, and resistance to creep and deformation
around 2, while one produced by free radical
under load. Epoxies, vulcanized rubbers,
polymerisation will usually lie between 1.5 and 10.
phenolics, unsaturated polyester resins, rigid
polyurethanes, urea-formaldehyde and melamine-
formaldehyde are examples of thermosets. Consequences of High Molar Mass
Increasing the interaction between polymer
Molecular Mass or Polymer Chain Length molecules leads to increasing cohesive energy
The number of repeating units in a synthetic per molecule. In polymeric materials this gives
polymer exerts a significant influence on its phys- rise to certain properties characteristically
ical properties. The high molecular weights of associated with high molar mass, regardless of
polymers increase their viscosity when molten their chemical structure:
and in solution, decrease their solubility, and of (i) High crystal melting point (if the polymer is
course prevent their volatilisation. Chain length crystalline)
can also be a controlling factor in determining the (ii) High viscosity in the melt and in solution
solubility, elasticity, fibre-forming capacity, tear (iii) High mechanical strength
strength, and impact strength in many polymers. (iv) High flexibility and ductility (unless highly
The chain length of commercial polymers is cross-linked)
optimised during synthesis for the intended appli- (v) High resistance to dissolution (especially
cation. For example, PMMA, with an average crystalline polymers)
www.ebook777.com
The average degree of polymerization, or DP,

is usually defined as the number of monomeric
units in a polymer [2]. For a homopolymer, there
is only one type of monomeric unit and the
number-average degree of polymerization is
given by
Number average molecular mass of polymer
DPn ¼
molecular weight of the monomer unit
Mn
¼
M
ð7:1Þ
For most industrial purposes, degrees of poly-

merization in the thousands or tens of thousands
are desired. Fig. 7.5 Typical stress—strain plots for a fibre, a flexible
Low molecular weight polymeric materials plastic, and an elastomer
(having short chains) are generally weaker.
Although sections may be crystalline, only
weak Van der Waals forces hold the crystallites plastics, fibres, and elastomers. Although there
together. This allows the crystalline layers to is no firm dividing line between the groups,
slip past one another causing a break in the some distinction between these categories can
material. Amorphous polymers, with high DP be obtained from a typical stress—strain plot
(such as cast PMMA) have greater strength (Fig. 7.5). Rigid plastics and fibres are resistant
because the molecules become tangled between to deformation and are characterized by a high
layers. elastic modulus and low percentage elongation.
Elastomers readily undergo deformation and
Physical Properties exhibit large reversible elongations under small
Polymers can be readily processed by forming or applied stresses, i.e., they exhibit elasticity. The
moulding into shapes, either once (thermosets) or flexible plastics are intermediate in behaviour,
repeatedly (thermoplastics). Many are important during the elastic stage of their deformation,
engineering materials with a wide range of and then yield inelastically with typical plastic
properties, some of which are unattainable from deformation.
any other material types, and they are generally During the final stages of processing, fibres
low in cost. The desirable properties of polymer are stretched to three or more times their original
products include light weight, availability in a length (“drawing”) when in a semi-crystalline
wide range of colours, low thermal and electrical state, to produce increased chain alignment, crys-
conductivity, toughness, and good resistance to tallinity and strength.
acids, bases and moisture. For amorphous materials, the temperature at
The applications of a polymer depend most which the molecules move relative to one
strongly on its mechanical behaviour. The another, known as the glass transition tempera-
“deformability” of a polymer can be expressed ture (Tg—described in more detail in section
as the ratio of the deformation (strain) resulting “Glass Transition Temperature”) exerts a pro-
from a constant applied stress. The ratio of stress found influence on the physical properties. In
to strain is the elastic modulus. general, elastomers have values of Tg well
A large number of synthetic polymers now below room temperature, while rigid, structural
exist covering a wide range of properties. polymers have Tg values above room
These can be grouped into three major classes: temperature.
Structural Physical and Decomposition chain releases groups from its ends most easily,
Property Data known as end-chain scission or unzipping. In
In order to better understand the range of com- others, the chain breaks at random points along
mon polymers, their names, abbreviations, and its length, known as random chain scission. A
molecular structures have been listed in third process, where groups attached to the back-
Table 7.1. Where available, measurements of bone as side chains can leave as stable molecules,
their crystallinity, glass transition temperature is known as chain-stripping. If the polymer, or
(Tg) and melting point have also been provided, the chain resulting from chain stripping, does
in order to give the reader an instant guide to the not undergo chain scission to form volatiles or
main properties of interest. More detailed lose further substituents, it may undergo
tabulations providing citable data have been carbonisation, resulting in char formation. Thus
published [2]. In addition, Lyon et al. [3] have the conversion of an organic polymer to volatile
tabulated the decomposition temperatures and organic molecules, and/or a char, may follow one
Arrhenius parameters Ea and A (section “Kinet- or more of the four general mechanisms. While
ics of Polymer Decomposition”) for a number of some polymers fall exclusively into one cate-
common polymers from literature and the data is gory, others exhibit mixed behaviour, often
reproduced here. Individual polymeric materials dependent on the decomposition conditions.
will differ according to their method of prepara- The temperature of a material is a measure of
tion, thermal history and the definition and the kinetic energy of its molecules. At very low
method of measurement of the decomposition temperatures (close to 0 K), molecules are almost
properties. Therefore individual literature values stationary, but at all normal temperatures in the
have been provided to give an indication of the solid phase, molecules are in a state of constant
uncertainties in the reported data. vibration. As the temperature increases, the
vibrations become stronger, while the strength
of the chemical bonds remains constant. For a
Polymers and Fire particular polymer, a critical temperature is
reached where there is sufficient kinetic energy
The flammable components of our built and nat- to rupture one of the bonds holding the repeat
ural environments are almost all based on units, or the side chains, of the polymer together.
organic polymers, and the vast majority of If the resultant molecules are small enough to be
unwanted fires are fuelled by these polymers. volatile, they may escape from the surface of the
Smouldering combustion typically occurs by polymer, or in the case of a thermoplastic, form
reaction of atmospheric oxygen with a porous, bubbles within it. When sufficient fuel is present
combustible solid matrix, with a reaction zone in the vapour phase mixed with air, it can react
moving through the solid, releasing gaseous with oxygen, releasing heat and increasing the
products. Flaming combustion requires the fuel free radical concentration. In the presence of a
to be present in molecular form in the vapour pilot flame or spark, additional free radicals
phase, where it can undergo much more rapid accelerate the ignition process. Ignition and
reaction with atmospheric oxygen. Since flaming combustion occurs when there is suffi-
polymers are much too large to exist in the cient heat from the flame to replace the gas phase
vapour phase (because the bonding forces fuel by further pyrolysis. As molecules are
holding them in the condensed phase is propor- released from the decomposing polymer, partic-
tional to their large surface area) they must first ularly by chain stripping, this leaves active sites
break down into volatile fragments. The pyroly- for further reaction. In many cases, such as cellu-
sis of a polymer, turning molecular chains losic materials, this can result in cross-linking
of 10,000–100,000 carbon atoms into species reactions to other polymer chains, leading to
small enough to be volatilised, often involves char formation. In bulk cellulosic materials
breaking the polymer chain. In some cases, the such as wood, this can result in the build-up of
www.ebook777.com
Table 7.1 Common properties and some of their physical/decomposition properties
Common or Melting Decomposition Kinetic parameters [3]

Chemical name Structure trade name Abbreviation Crystallinity Tg/K point/K temperature/K Ea/J mol1 A/s1
Polyethylene Polythene PE 40–80 % 175–260 473–508 687 3.00 105 6.5 1020
CH2 CH2
n 677 2.64 105 2.3 1018
679 2.77 105 2.0 1019
644 3.01 105 2.6 1022
649 2.64 105 1.8 1019
Polypropylene PP 65 % 253 443 660 2.43 105 1.7 1017
CH CH2
n 624 2.43 105 2.2 1018
CH3
Polyisobutylene CH3 PIB – 200 229 621 2.05 105 1.8 1015
C CH2
589 2.05 105 1.5 1016
n
CH3
Polyisoprene Natural Low 203 303 596 2.50 105 8.2 1019
C CH CH2 CH2
n Rubber 574 2.49 105 2.3 1019
CH3
Polybutadiene Butyl rubber 680 2.60 105 9.4 1017
CH CH CH2 CH2
n 2.60 105 3.5 1018
Polystyrene PS Low 363–378 503 637 2.30 105 7.3 1016
CH CH2
n 549 2.01 105 1.3 1017
606 2.30 105 6.7 1017
Polyacrylonitrile “Acrylic” PAN Low 413 590 512 1.30 105 1.8 1011
CH CH2
n (wool or
CN textile)
Poly(vinyl alcohol) PVAL or 343–373 503–533 na – –
CH CH2
n PVA
OH
Poly(vinyl chloride) PVC Low 5 % 353–358 348–378 563–733 1.34 105 7.8 1020
CH CH2
n (plasticised) (485) 1.38 105 3.0 1012
Cl Med 15 %
unplasticised
Poly(vinyl acetate) Acetate PVAc 526 2.24 105 1.8 1020
CH CH2
n
OOCCH3
Poly(vinylidene F Kynar, Hylar PVDF High 313 443–448 738–758 2.01 105 4.9 1013
fluoride)
C CH2
n
F
(continued)

Poly(methyl
CH CH2
acrylate) n
COOCH3
Poly(methyl CH3 Perspex, PMMA Low 388 (synd); 363–378 633–663 2.18 105 4.7 1016
methacrylate) Plexiglass 378 (atact); 591 2.18 105 1.9 1017
C CH2
n 318 (isotac)
COOCH3 577 2.18 105 5.4 1017
514 1.26 105 6.4 1010
Polyformaldehyde, Acetal POM High 364–383 448–454 501 1.09 105 2.3 109
OCH2
polyoxymethylene n
Poly(ethylene oxide) PEG and 618 1.93 105 2.1 1014
OCH2 CH2
or Polyethylene n PEO 601 1.92 105 4.9 1014
glycol
Poly(tetrafluoro F F Teflon PTFE High 398 600 782 3.39 105 4.4 1020
ethylene) C C 773 3.14 105 1.7 1019
n
F F 742 3.37 105 5.3 1021
Engineering
thermoplastics
Poly(ethylene Polyester PET High 343–353 523–533 698–713 1.59 105 2.6 1011
terephthalate) OCH2 CH2 OOC CO
n
Poly(butylene O O Polyester PBT High 318–333 493–503 681 – –

terephthalate)
C C O (CH 2) 4 O
n
Poly(hexamethylene Nylon 6.6, PA 6.6 Medium 343–363 498–538 703–746 – –
www.ebook777.com
NH (CH2) 6 NHCO (CH2) 4 CO
adipamide) n Polyamide
6.6
Polycaprolactam Nylon PA 6 Medium 323–353 498–508 708 – –
NHCO (CH2)5
n 6, Polyamide
6
Polycarbonate CH3 O Lexan PC Low 418–423 488–503 753–758 – –
O C O C
n
CH3
High Thermal
Stability Polymers
Poly(phenylene With PS as PPO – –
oxide) O Noryl
n
Poly(ether ether O PEEK High 32 % 416 607 843 – –
ketone) O C O
n
Polyether sulphone O PES Low 853 – –

S O
n
O
Polyphenylene PPS 777
Sulphide S n
Polyether imide O PEI Low 490 800 – –

O
O H3C N
C
CH3 O
N
n
O
O
Thermosets
Epoxy resin CH3 CH3 H2 EP Cross-linked 273–453 673–723 – –
H2 H2
C C C
OH
O C O O C O
C C C
)
C C C
( H2 H2 H2 CH3 H2 H H2 CH3
Urea Formaldehyde H2 H2 UF Cross-linked >473

resin C C
N O N )
Melamine H H Melaware MF Cross-linked 293–333 – –
formaldehyde resin N N
H
O N N O
H
HN
NH
O
Phenol OH OH Bakelite PF Cross-linked 353–393 723–753 – –

formaldehyde resin H2 H2 H2
C C C
( )
Polyurethane H H PUR and Cross-linked 283–493 – –

N N O O PUF
1 2
C R C R )
(
O O
(continued)

Copolymers
Poly(ethylene EVA Medium 233–293 303–383 753 3.45 105 1.27 1022
co-vinyl acetate) ... CH2 CH2 ... CH2 CH ...
n m
573 1.66 105 6.11 1011
O
C
H3C O
Acrylonitrile- ABS Low 188–378 383–398 693–701 – –

butadiene-styrene ... CH2 CH ... CH2 CH CH CH2 ... CH CH2 ...
n m o
copolymer CN
Styrene acrylonitrile SAN Low 373– 393 393 693 – –

copolymer ... CH2 CH ... CH CH2 ...
n m
CN
Natural and
biopolymers
HO
Cellulose HO H High Decomposes – –
H H OH H
H
H O H OH
HO H O H OH
O HO
O O
www.ebook777.com
H O
H H
OH
H O H OH H O
OH OH H
HO OH H H OH
Keratin O O Wool H-bonded – –

H 2
H 4
C N R C N R
1 3
R C N R C N
H H
O O
Polylactic acid (OCH(CH3)CO-)n PLA 318–338 423–433 623–643 – –

the protective layer, shielding the wood from crystalline materials. As such, they all tend to
external radiation, slowing the rate of further have highly ordered and regular structures.
fuel pyrolysis, and hence the rate of burning. Amorphous materials, by contrast, have their
The burning process can be viewed on either molecules arranged randomly and in long chains
the molecular scale or the macro scale. which wrap around each other, without any long
The molecular scale

Stage I Heating An external source supplies heat causing the temperature of the substance to increase.
The extent of temperature change depends on the specific heat of the material. Physical,
mechanical, and thermal properties change in the case of polymers. This may include
softening, melting and volatilsation
Stage II Decomposition At higher temperatures the majority of the bonds reach failure point, causing the release of
gaseous molecules which differ depending on the material burning. This can be accelerated
by attack of oxygen on the surface of the polymer, producing carbon dioxide and carbon
monoxide
Stage III Oxidation In the presence of oxygen at high temperatures, oxidation of the gaseous fragments
proceeds rapidly, releasing heat, and combustion products (mostly carbon dioxide and
water)
The macro scale

Stage I Heating Heat causes a temperature rise which will depend on the thermal inertia (kρC) of the material
Stage II Pyrolysis Heat causes decomposition of the fuel, followed by pyrolysis of fuel to the gas phase
Stage III Ignition Fuel accumulates above the surface, and reacts with oxygen. Once the critical concentration
of free radicals is reached, flashing will occur. When the total heat flux to the surface from
fuel oxidation is sufficient to pyrolyse enough fuel to replace it, ignition will occur, the rate of
reaction will increase and produce carbon dioxide and water
Stage IV Flame spread As the radiant heat flux increases it will pyrolyse adjacent materials, leading to a repeated
series of ignitions, resulting in fire growth
Stage V Fire As the flame gets larger, it will no longer be able to been entrain sufficient oxygen, and
development products of incomplete combustion such as carbon monoxide and soot will be produced,
increasing the radiative component of heat transfer
Polymer Crystallinity range order. Both crystalline and amorphous

phases exist in polymers.
Although most polymers are solids at room tem- Polymers can form crystallites either by
perature they have more properties in common straightening out the molecules and packing as
with glass than with crystalline solids, such as rods (extended-chain crystals), or each chain
sugar or salt. Glass has an amorphous morphol- folds back and forth, so that crystallisation occurs
ogy with properties different to crystalline solids. by short segments of the same chain packing
When heated, it gradually changes, from a brittle together ( folded-chain crystals) (Fig. 7.6).
solid-like material, softening, and eventually Chain-folding is kinetically favoured, as it is a
becoming a viscous liquid. In contrast, the appli- unimolecular process, but orientation of a poly-
cation of heat to a crystalline solid turns it mer (such as the “drawing” of a fibre) leads to
sharply to a low viscosity liquid at a particular chain-extension crystal structures. A complete
temperature. The difference lies mainly in the polymer chain is likely pass through many small
structure of each phase. Crystalline materials crystals, tying them together into a strong, coher-
have their molecules arranged in repeating ent mass, surrounded by amorphous sections of
patterns. Salt, sugar, ice and most metals are polymer. If individual polymer molecules were
CHAIN-FOLDED CRYSTAL
CHAIN-EXTENDED
CRYSTAL
POLYMER CRYSTALLITE
Fig. 7.6 Schematic of chain-folded crystal, chain extended crystal, and polymer crystallite
confined to individual crystals, they would be different decomposition and burning behaviour if
extremely brittle. The mixtures of small crystals they were formed into plaques under different
and amorphous material in polymers cause them thermal conditions. This can be compensated
to melt over a range of temperatures without a for by subsequent annealing (heating and holding
sharp melting point. In most polymers, the com- each specimen at an appropriate temperature
bination of crystalline and amorphous structures below the crystalline melting point, followed by
forms a material with advantageous properties of controlled cooling). If the polymer is cooled
toughness and rigidity. slowly, this will produce a significant increase
Some polymers, such as polystyrene and in crystallinity, and relieve internal stresses.
PMMA, are completely amorphous, others have The size and shape of the side chains of mono-
a combination of disordered regions and small mer also influence the polymer morphology. If
crystallites. An amorphous polymer results from the monomers are large or irregular, relative to
polymer chains that lack regular order. If parts of the polymer backbone, as in polystyrene, it is
two chains do not pack well together, crystallites difficult for the polymer chains to arrange
do not form. Shorter chains organize themselves themselves in an ordered manner, resulting in a
into crystalline structures more readily than lon- more amorphous material. Likewise, smaller
ger molecules, as those with a high degree of monomers, such as polypropylene, and polymers
polymerization (DP) tend to become tangled. that have a very regular structure, such as the
The DP is an important factor in determining rod-like structure of PTFE, will form highly
the degree of crystallinity of a polymer. The crystalline polymers.
cooling rate also influences the degree of crystal-
linity (section “Differential Thermal Analysis
and Differential Scanning Calorimetry”). Slow Thermal Response Characteristics
cooling provides time for crystallization to of Polymers
occur. Fast cooling yields highly amorphous
materials. When characterising the flammability Physical Transitions
of a material, in order, for example to predict its
large scale fire behaviour, it is essential to ensure The physical processes occurring during thermal
that all the material has the same thermal history. decomposition depend on the material.
For example, cone calorimeter plaques prepared Thermoplastics can be softened and melted by
from the same semicrystalline polymer may have heating; once polymerisation is complete,
www.ebook777.com
thermosetting polymers are infusible and phase (section “Differential Thermal Analysis and Dif-
changes such as melting cannot occur. The melt- ferential Scanning Calorimetry”).
ing/softening behaviour of thermoplastics on
heating depends on the degree of crystallinity. Melting
For crystalline materials the intermolecular The backbone of a typical (CH2CHX)n poly-
forces are usually identical, so melting occurs at mer is composed of a chain of tetrahedrally
a well-defined temperature; for amorphous bonded carbon atoms covalently bonded to each
materials a range of intermolecular forces hold other so that the molecule can be represented as
the polymer chains to each other, so the polymer an extended zigzag chain. The repeating units in
will soften over a wider temperature range. a polymer chain are often free to rotate relative to
However, many materials cannot undergo the one another. If they are all in a particular posi-
transition to a viscous state without undergoing tion, the polymer molecule will have a linear
thermal decomposition. Neither thermosets nor zigzag shape. If they rotate, this will result in
cellulosic materials have a fluid state, so they bends in the polymer molecule. While the linear
neither melt or soften. In thermosets, the zigzags can stack together easily, each bend will
3-dimensional network of cross-linking covalent be an obstacle to crystallinity. For polyethylene
bonds prevents the polymer chains from moving (–CH2CHX)n, a typical value of the molecular
relative to each other. In cellulosic polymers, weight is 1.6 105 g mol1, so the chain
the extensive hydrogen bonding between the contains 10,000 carbon atoms; thus in the
hydroxyl groups and oxygen atoms keeps the extended zigzag state, the chain would be about
polymer chains in place. 1260 nm long and 0.3 nm diameter. As every
group of four atoms in the chain can have three
Glass Transition Temperature possible stable rotational positions (linear, zigzag
As the temperature of a polymer rises above a or kinking to left or right), a total of 310,000
certain critical point, its glass transition tempera- shapes (or additional degrees of freedom) are
ture, Tg, it becomes more rubber-like. Con- available to this particular chain, only one of
versely, as the temperature drops below Tg, it which is the fully extended zigzag. Even though
behaves in an increasingly brittle manner. The this has the lowest energy, the most probable
glass-transition temperature is the point at which conformation will be some kind of randomly
the polymer chains in a non-crystalline (amor- angled amorphous shape (Fig. 7.7).
phous) material acquire sufficient thermal energy The number of different possible states which
to undergo significant translational motion, char- can exist when leaving the crystalline state
acteristic of the liquid-like or rubbery state. Below (or extended zigzag conformation), known as
Tg the chains are frozen into a glassy state, where the entropy change, ΔS, exerts the greatest influ-
only very localised atomic movement, such as ence on the melting temperature of a polymer. A
vibration, is possible. If a molten polymer is large part of this entropy is due to the additional
cooled so quickly that the Tg is reached before freedom that allows the chain conformational
the polymer can fully crystallise, then the polymer changes to occur in the melt; i.e., the restrictions
will remain frozen in its glassy (amorphous) of the crystalline lattice no longer apply.
state until its temperature is raised above Tg. In any phase transition, the free energy
The Tg is a transition which is characteristic of change involved, ΔG, is zero, since the two
non-crystalline phases; it is an important parame- phases must be in equilibrium. The free energy
ter in the selection of materials for particular is the sum of the chemical bonding forces, or
applications, such as whether rigid or elastomeric enthalpy ΔH, and the disorder, or number of
properties are required. Above Tg, but well below possible ways the molecule can be arranged.
its melting point, an incompletely crystallised This is the product of the temperature and the
polymer can undergo further crystallisation entropy change, TΔS.
Crystalline portion
Break in crystalline region
amorphous region
Fig. 7.7 Crystalline and amorphous regions of a polymer
ΔG ¼ ΔH TΔS ¼ 0 ð7:2Þ Table 7.2 Structures and melting temperatures of four

polymers
so that during the polymer melting process; the
melting temperature Tm is given by Polymer Tm(°C)
T m ¼ ΔH fusion =ΔSfusion ð7:3Þ
135
The enthalpy of fusion (ΔHfusion) reflects the
CH3 CH3 CH3 CH3 CH3
strength of attractive forces between the
165
molecules. For chemically similar materials
(e.g. hydrocarbon polymers) these will not vary
greatly. Thus ΔSfusion, the entropy change on 380
melting, is the only significant variable.
For polymers, entropy is related to the number
of degrees of freedom each polymer molecule > 600
has. A molecule in the liquid state has more
potential degrees of freedom (bond rotations,
bond angle flexions and inversions, translational translational and rotational degrees of freedom
and rotational motions) than if it is frozen in a are acquired on melting, leading to extremely
crystal. Clearly, the more flexible the polymer high values of Tm, often well above the decom-
chain, the more degrees of freedom it acquires position temperature of the polymer.
on melting, and the greater is ΔSfusion. A high Table 7.2 shows the variation in melting point
ΔSfusion leads to a low Tm. for four polymers. In PE every bond is free to
Conversely, the stiffer the polymer chain, the rotate; PP is constrained to a small extent by the
fewer degrees of freedom it acquires on melting, coiling of the chain to accommodate the methyl
the lower is ΔSfusion and the higher is Tm. groups in a regular manner. Aromatic rings do
For completely rigid polymer chains, only not allow any rotation so poly-1,4-phenylene
www.ebook777.com
ethylene (PhCH2CH2Ph)n can only rotate Influence of Oxygen

between the two –CH2–CH2– carbon atoms. The thermal decomposition of polymers may pro-
In poly-1,4-phenylene (Ph–Ph–)n no rotation is ceed by heat alone, or by the combined action of
possible (as the conjugated aromatic rings must heat and oxygen. In many polymers, the thermal
all lie in the same plane, so the polymer molecule decomposition processes are accelerated by oxy-
exists as a rigid rod), there is no “entropy advan- gen, lowering the minimum decomposition
tage” to the polymer on melting. Poly-1,4- temperatures. Prior to ignition, thermo-oxidative
phenylene decomposes before melting, so it is decomposition results in pyrolysis of fuel and
not melt-processable, so there is no way of forming other species. After ignition, during steady
it into shapes, and so its practical applications flaming, even in well-ventilated conditions, pyrol-
are limited to heat resistant fabrics, electrical insu- ysis of the condensed phase (pyrolysis zone) is
lation etc [4]. essentially anaerobic, with all the oxidation taking
place in the gas phase (flame zone) [5]. Thus, the
Bubble Formation mass loss, resulting from pyrolysis, the residue
As a consequence of the chemical processes formation etc., of a flaming sample corresponds
of polymer decomposition, leading to volatile to a decomposition of the material under an inert
formation, volatile molecules will start to atmosphere [6]. Unfortunately, there are several
accumulate within the decomposing polymer. published studies of the development of fire
If the polymer is molten when decomposition retarded materials which appear oblivious to this
commences, bubbles will form and migrate fundamental principle of fire science. Only for
upwards, eventually erupting from the surface. ignition, non-intense flaming, samples near and
This causes physical swelling, reducing the ther- after extinction, and non-igniting samples will
mal inertia of the material, accelerating the rate thermo-oxidative decomposition be relevant to
of surface heating and the onset of ignition. the behaviour in a fire. Indeed the observation of
bubbles of volatile fuel in decomposing polymers,
around the time to ignition (which have been
Chemical Transformations characterised by immersing the test specimen
in liquid nitrogen), showed that for many
In the case of thermosets and cellulosic thermoplastics, even prior to ignition, most vola-
materials, the polymer molecule starts to decom- tile formation comes from the bulk of the polymer,
pose before the chains have acquired sufficient not its surface, and hence the critical decomposi-
energy to overcome the forces holding them in tion condition remains anaerobic [7].
place. These materials tend to produce carbona- The thermal decomposition of polymers has
ceous chars on thermal decomposition. The been broken into for general chemical
physical structure of these chars will profoundly mechanisms. The first three essentially describe
affect the heat transfer, volatile release, and the conversion of an involatile polymer molecule
access of oxygen, all of which will impact on into fragments small enough to be volatile. In
the thermal decomposition processes. The char many cases the decomposition follows more
can undergo glowing combustion in the presence than one of the mechanisms.
of oxygen. However, it is unlikely that both (i) Random-chain scission, in which chain
glowing combustion of the char and significant scissions occur at apparently random
flaming can occur simultaneously in the same locations in the polymer chain.
zone above the surface, since the flame will con- (ii) End-chain scission, in which individual
sume all the available oxygen, and the flow of monomer units are successively removed
volatiles through the char will tend to drive oxy- at the chain end.
gen away from the char surface. Therefore, in (iii) Chain-stripping, in which atoms or groups
general, char oxidation will only occur after not part of the polymer chain (or backbone)
flaming has subsided. are cleaved.
(iv) Cross-linking, in which bonds are created Chain Branching

between polymer chains. With two chain branches on every other carbon
These are discussed further in section atom in the chain, polyisobutylene (CH2C
“Decomposition Mechanisms”. (CH3)2)n has the lowest thermal stability,
followed by polypropylene (CH2CH(CH3))n
with one branching point on every other carbon
Influence of Chemical Structure atom. Commercial polyethylene is not composed
on Thermal Stability only of straight polymer chains. It actually
contains a number of branches of its linear
The combustion behaviour of a polymeric mate- chains, either of small groups such as CH3– or
rial can be interpreted in terms of the properties longer side chains, which occur randomly during
of the volatiles, particularly their composition, the polymerisation process. These are the most
reactivity and rate of formation. Thermal stabil- reactive parts of otherwise unreactive structures.
ity can be quantified from the temperature depen- Polymethylene (CH2)n is the name given to
dence of decomposition. the special, unbranched form of polyethylene.
Detailed studies by Madorsky [8] in the 1960s The number of branching points in normal
of the effects of chemical structure on the ther- polyethylene also affects its crystallinity.
mal stability of polymeric materials underpin Low density polyethylene (LDPE) has around
our understanding of the factors controlling 60 branching points per 1000 carbon atoms. An
the thermal decomposition of polymers. These intermediate density, linear low density polyeth-
experiments investigated the thermal stability ylene (LLDPE) is actually a copolymer of ethane
by determining the temperature, Th, at which and an alkene such as oct-1-ene, so the regularity
50 % of a small polymer sample will volatilise of the polymer chain is deliberately disrupted
in 30 min in an inert atmosphere. Table 7.3 by the presence of side chains, 6 carbon atoms
summarises the effects of chemical structure on in length. High density polyethylene (HDPE) is
the thermal stability of polymers, and provides closer to the idealised polymethylene with
examples of that behaviour. The individual around 7 branching points per 1000 carbon
effects are discussed below. atoms. During thermal decomposition, the
Table 7.3 Factors affecting the thermal stability of polymers (From Madorsky [8])
Factor Effect on thermal stability Examples Th/K
Chain branching Weakens Polymethylene 688
Polyethylene 679
Polypropylene 660
Polyisobutylene 621
Double bonds in polymer backbone Weakens Polypropylene 660
Polyisoprene 596
Aromatic ring in polymer backbone Strengthens Poly-1,4-phenylene methylene 703
Polystyrene 637
High molecular weight Strengthens PMMA B (MW ¼ 5.1 l06) 600
PMMA A (MW ¼ 1.5 l05) 556
Cross-linking Strengthens Polydivinyl benzene 672
Polystyrene 637
Oxygen in the polymer backbone Weakens Polymethylene 688
Polyethylene oxide 618
Polyoxymethylene <473
www.ebook777.com
branching points are the first to break, initiating Interaction of Chemical and Physical
chain decomposition reactions. Processes
Double Bonds The chemical composition of the gas phase fuel,

In a similar way, the double bonds in the polymer and its production rate, depend both on the chem-
chain of polyisoprene (CH2CH2CH ¼ C ical structure of the polymer as it breaks down to
(CH3))n also provide reactive sites for the initia- release fuel, and the physical properties of the
tion of chain decomposition reactions, reducing its material. In particular, the transfer of heat
thermal stability compared to polypropylene. through the solid/liquid and its rheological
properties (affecting both heat and gas transport),
Aromatics in Backbone will influence the decomposition and burning
Conversely, the presence of an aromatic ring, in behaviour, under a defined set of external
the polymer backbone, such as poly-1,4-phenylene conditions. On the microscale (such as that in
methylene (CH2C6H4)n, compared to poly- TGA, section “Thermogravimetric Analysis”, or
styrene (CH2CH(C6H4))n, where the aro- MCC, section “Microscale Combustion Calorim-
matic ring is a side chain, increases the rigidity of etry”), a single piece of sample of mass 5 mg,
the chain, lowering the entropy of the liquid state, heated at 10 K per minute may be in thermal
thus raising the melting point. The higher melting equilibrium to within a few degrees K, but an
point reduces the rate of pyrolysis, since most of escaping monomer would still have to pass over a
the material is protected by the surface layers. million repeat units, even if following the
shortest straight line trajectory from the centre
Molecular Weight of the sample to the edge, in order to escape. In
In end-chain scission the reaction starts at the end actuality, a much more tortuous path around the
of a polymer chain, since the end molecule is only polymer crystallites would need to be followed in
held by one bond. PMMA A with 1500 repeat order for fuel to escape. Along this path, the
units decomposes 55 K lower than PMMA B with monomer or other volatile fragment could
50,000 repeat units. The rate of the decomposi- recombine with the polymer; only when bubbles
tion reaction will then depend on the number of form is the reverse process of repolymerization
end groups available. The higher the molecular effectively prevented.
weight, the smaller that number will be. Thermoplastics can melt without chemical
reaction to form a viscous state (polymer
Cross-Linking melt), but they may decompose thermally by
Cross-linking will also prevent melting, random chain scission, reducing their molecular
inhibiting the transport of molecules to the sam- mass, increasing the proportion of short chain
ple surface for pyrolysis as gas phase fuel. polymer molecules in the liquid phase, before
Polydivinylbenzene has two reactive groups per melting. In the absence of a flame the fluidity of
monomer unit, and therefore each monomer can the polymer will accelerate the access of oxy-
be attached to two chains. gen, and its release of volatile decomposition
products.
Oxygen in Backbone
Oxygen in a polymer chain, such as in
polyoxymethylene (CH2O)n, will also pro- Thermal Analysis: Methods
vide a reactive site for polymer decomposition to for Quantifying the Thermal Response
start, lowering the thermal stability. of Polymers
In general, the Th data indicates the ease of
conversion to volatiles, which are likely to act Thermal analysis describes a family of
predominantly as fuel. However, the fire techniques which measure changes in physical
behaviour will depend on combustibility of the and chemical properties during controlled
volatiles as they are produced. heating. The programme may take many forms:
(i) The sample may be subjected to a constant wide ones, and platinum crucibles catalyse
heating (or cooling) rate (dT/dt ¼ β), for some reactions more than alumina ones. The
example 10 K min1. type of holder or clamping used for thermome-
(ii) The sample may be held isothermally chanical methods is equally important. The
(β ¼ 0). results are also unlikely to be entirely indepen-
(iii) A “modulated temperature programme” dent of the make and type of instrument used.
may be used where a sinusoidal or other The Rate of heating: this has most important
alteration is superimposed onto the under- effects. A very slow heating rate will allow
lying heating rate (section “Differential the reactions to come closer to equilibrium
Thermal Analysis and Differential Scan- and there will be less thermal lag in the appa-
ning Calorimetry”). ratus. Conversely, high heating rates will give
(iv) To simulate special industrial or other pro- a faster experiment, may be more representa-
cesses, a stepwise or complex programme tive of the heating rates in fires (section
may be used. For example, the sample might “Choice of Atmosphere and Heating Rate in
be heated at l0 K min1 to 373 K, and held Thermal Analysis”), but deviate more from
for 10 min, to drive off any absorbed water, equilibrium and result in greater thermal
then heated at l0 K min1 to 673 K, then lag. The parameters of special heating
held there for 30 min. In addition, the atmo- programmes, such as modulated temperature
sphere can be changed during a particular or sample control, will also affect the results.
part of the heating regime, such as switching The Atmosphere: both the transfer of heat, the
from nitrogen to oxygen above 1000 K, to supply and removal of gaseous reactants, and
quantify the presence, by oxidation, of car- the nature of the reactions which occur, or are
bonaceous residues. prevented, depend on the chemical composi-
(v) The heating may be controlled by the tion of the atmosphere and its flow.
response of the sample itself (e.g. high res- Oxidations will occur quickly in oxygen,
olution TGA). more slowly in air, and not at all in nitrogen
In order to understand, and alter, the or other inert gas; product removal by a fairly
behaviour of a material in a fire, it is necessary rapid gas flow may prevent reverse reactions
to know as much as possible about the processes occurring.
of decomposition. These processes are often The Mass of the sample: a large mass of sample
highly dependent on the conditions, particularly will require more energy, and heat transfer
the heating rate and atmosphere. Moreover, sen- will be determined by sample mass and
sitivity to these parameters can give the vital dimensions. These include the volume, pack-
clues needed to interpret the fire behaviour. Ther- ing, and particle size of the sample. Fine
mal analysis tends to be far more sensitive to powders react rapidly, lumps more slowly.
instrumental parameters than other branches of Large samples may allow the detection of
chemical analysis. The following summarises the small effects. Comparison of runs should pref-
key parameters needed to ensure the validity and erably be made using similar sample masses,
reproducibility of results obtained by thermal sizes and shapes.
analysis, conveniently summarised by the acro-
nym SCRAM.
The Sample: the chemical composition, the Thermogravimetric Analysis
source and pre-treatments, together with the
history of the sample, impurities and dilution “Thermogravimetry is a technique in which the
with inert material can all affect results. mass of a test specimen is measured as a function
The Crucible: the material and shape of the cru- of temperature or time, while the test specimen is
cible or sample holder is important. Deep subjected to a controlled temperature program.”
crucibles may restrict gas flow more than flat, [9] Thermogravimetric analysis (TGA) is the
www.ebook777.com
most commonly used method for investigating A sample is suspended on a highly sensitive
the complex thermal decomposition processes balance over a precisely controlled furnace. Usu-
of polymers. In TGA experiments, at heating ally heating rates of 5–20 K min1 are used to
rates around 10 K min1, a sample size around look for broad decomposition stages, while
5–10 mg is small enough to ensure that it is in slower heating rates, around 1 K min1 are better
thermal equilibrium with the apparatus. for isolating individual events. Sample sizes are
In isothermal TGA, the sample is brought usually kept as small as possible, within the
quickly up to the desired temperature, usually limits of sensitivity of the apparatus this is usu-
by rapid insertion into a preheated furnace, and ally around 5 mg per run. This reduces bulk
the weight of the sample is monitored during the effects, and at higher heating rates, avoids ther-
course of thermal decomposition. In practice, the mal gradients being set up within the sample.
sample does not heat instantaneously, so errors A typical microbalance has a rotating pivot,
arise where the selected temperature is high and is controlled electronically using a zero detec-
enough for significant decomposition to occur tion device, as used in a galvanometer, usually a
within the first few minutes. light and photocell and a magnet and moving coil
In dynamic TGA, the sample is subjected to system to restore balance. The control system
temperature programmed heating. This varies varies the current passed through the coil to keep
from a fixed rate, such as 10 K min1, to a variable the beam of the balance in the zero position. This
rate, designed to highlight particular features is known as a null deflection system and has the
(such as 10 K min1 up to 650 K followed by advantage that it keeps the sample in the same
1 K min1 up to 660 K, in order to focus on a position in the furnace throughout the run.
particular step. A further modification is high res- The results of TGA experiments may be
olution TGA, in which the changes in sample mass presented as mass losses, usually as a percentage
are used to slow the heating programme, to get of total mass as a function of temperature,
better resolution of particular thermal events. As (or time, for isothermal TGA), or may be
isothermal TGA is falling into disuse, it is com- presented as the differential, showing the peaks
mon to refer to “dynamic TGA” simply as TGA. of mass loss (DTG). Both formats have their
The cylindrical furnace is designed to have a advantages—the mass vs temperature plot gives
long constant temperature zone in its centre. direct information about sample composition, the
TGA experiments rarely use static air, because percentage of non-volatiles etc., while DTG is
of the uncertainty of its composition during a run, much easier to see small differences in peak
and the possibility of reverse reactions occurring decomposition temperatures, and allows easy
with the vapour phase effluent. A flowing purge quantification of the maximum rate of mass loss
gas is almost always used, although operation (or fuel production in fire science). Figure 7.8
under a vacuum is sometimes possible, and use- shows the TGA and DTG curves for polyacrylo-
ful for techniques involving evolved gas analysis nitrile (PAN) in nitrogen at a heating rate of
(although the typically boiling points and gasifi- 10 K min1, showing three distinct decomposi-
cation temperatures are lowered by 150–250 K). tion stages at 550, 700 and 1100 K.
The best crucibles are made of platinum, which is There are three common designs of TGA instru-
inert, does not melt below 2042 K and may be ment, each having advantages and disadvantages
cleaned in strong acid. To reduce their thermal (Fig. 7.9). The hangdown thermobalance is
inertia, they are thin, and hence delicate. They probably the simplest and most robust design,
can chemisorb hydrogen, giving rise to a spuri- and dominated in the early instruments. The
ous weight gain, although this is unlikely in top-loading thermobalance has the advantages
routine polymer decomposition studies. The that, for evolved gas analysis the volatiles may be
alternative, ceramic crucibles, can suffer from sampled more directly, and the surrounding atmo-
fusion with molten samples and be very difficult sphere moves more naturally around the crucible
to clean. (though this will give different results, for example
Fig. 7.8 TGA and DTG curves for polyacrylonitrile decomposition in nitrogen at 10 K min1
Fig. 7.9 Three common TGA instrument designs
in studies of char oxidation, where the rate is (v) Quantifying thermal stability
dependent on collisions of oxygen gas with the (vi) Determination of the content of low
sample surface). The horizontal thermobalance volatiles (plasticizer, solvents)
minimises the effect of gas flow on the recorded (vii) Quantitative determination of single poly-
mass. In the other two instrument designs this is mer components
compensated for by subtracting a blank run from (viii) Filler content (e.g. carbon black, chalk,
the baseline, with an empty crucible with the same glass fibre)
gas and flow rate over the entire temperature range. (ix) Oxidative stability
Data from thermogravimetric analysis
(TGA) is commonly used in the following
determinations. Differential Thermal Analysis
(i) Identification of polymers/materials pres- and Differential Scanning Calorimetry
ent in composite materials
(ii) Studies of polymer decomposition These two techniques allow quantification of the
(iii) Generation of evolved products energy required, or released, during temperature
(iv) Determination of kinetic data programmed heating. Originally, differential
www.ebook777.com
scanning calorimetry (DSC), invented by Perkin- Advanced Material Analysis

Elmer, was deemed by the International Confer- Decomposition and oxidative susceptibility
ence on Thermal Analysis (ICTA) to describe Reaction kinetics
only the technique where a quantifiable amount Purity determinations
of energy, as an electric current, was supplied to The basis for obtaining kinetic parameters
the sample (or reference) to maintain the sample from DSC is to identify the rate of reaction with
and reference temperatures equal, now known as the DSC signal, and the extent of reaction with
power compensation DSC. the fractional area of the peak plotted against
In differential thermal analysis (DTA), the time. It is possible to obtain the three variables,
temperature of the sample is compared to the rate of reaction, extent of reaction and tempera-
temperature of a reference (usually a matching ture by carrying out a series of isothermal
empty crucible) during programmed heating experiments at different temperatures in much
(or cooling). Raw DTA data is a recorded as the same way as in classic kinetic investigation.
temperature difference (between the sample and The set-up of the experimental procedure is not
reference pans) as a function of temperature. without difficulty, but the interpretation of the
Prior calibration using materials of known heat result is less contentious than with the alternative
capacity, such as sapphire discs, together with dynamic procedures [10].
software converts the temperature difference Standard protocols have been published cov-
into energy units, to make the results comparable ering most of the common DSC measurements
to power compensation DSC. Thus, calibrated [11]. Some of these, such as chemical kinetic
DTA became known as heat flux DSC. Now it parameters, and the enthalpies of these
is recognized that both techniques provide the transitions, are of direct value in prediction of
same information and are both classed as DSC thermal decomposition, where others, such as the
[10]. Data is reported as a differential, showing heat capacity and thermal conductivity, are
endothermic (heat to sample) or exothermic (heat required when modelling the heat transfer
from sample) processes. Unfortunately some through the condensed phase (such as a 6 100
manufacturers have the exo-peaks going upward, 100 mm cone calorimeter sample) from
while others show them going down. Thus there microscale thermal analysis data.
is an established convention to label all DSC Excellent thermal contact is essential between
plots with an arrow labelled “exo” to show how the sample and reference pans and the instrument.
the results have been reported. DSC provides Such contact is vulnerable to attack from acids
useful information about polymeric materials such as hydrogen chloride and carbon rich vapours
and their decomposition. As the most commonly from the decomposition products of polymers. For
used thermal analysis technique, DSC also these reasons, and the importance of the DSC for
provides the following quantitative characterising the properties of polymers and other
measurements of physical processes and materials, DSC is used with some reluctance in
characteristics. polymer decomposition studies.
Thermophysical Properties Prior to thermal decomposition, the enthalpy
Specific heat capacity change of processes occurring in the condensed
Product Identification and Characterization phase may be quantified using DSC or DTA. The
Melting temperatures glass transition, Tg is manifested by a step rather
Transition enthalpies than a peak in the trace, resulting from an
Phase transformations, phase diagrams increase in specific heat capacity of the sample.
Crystallization temperatures On a molecular level, below the Tg, the polymer
Degree of crystallinity chains are held in place relative to one another.
Glass transition temperatures Above the Tg, movement is possible, as there are
more ways for the polymer to absorb energy, maintain pressures up to three atmospheres.
hence the increase in specific heat. For For measurements in oxidising atmospheres,
non-crystalline polymers Tg is in fact the only such as air, of boiling temperatures, or heats of
observable thermal transition prior to decomposi- vaporisation, a small hole may be pierced in the
tion. For a crystalline polymer, a defined melting lid. Polymer samples which do not decompose,
point will also be observable. If the polymer has may be pre-melted and pressed onto the base of
been cooled more quickly than the time required the pan to optimise thermal conductivity, and
for crystallization, then an exothermic peak enhance the signal. For higher temperatures, plat-
known as the cold-crystallization temperature inum, silver, gold, quartz, alumina, or graphite
may be observed, when the polymer chains are may be suitable.
free to move and become more closely aligned.
In some polymers, crystallization is almost Modulated Temperature DSC (MTDSC)
instantaneous, for others it can take hours or A minor revolution in thermal analysis occurred
days. Crystallization can only occur above Tg. with the development by Reading [12] of
The DSC heating-trace for a crystallisable MTDSC. This allows the two components
polymer, which has been cooled so rapidly from of enthalpy, such as the heat capacity and a
the melt that the glassy state was reached before process occurring at a particular temperature,
crystallisation occurred, is shown on in Fig. 7.10. such as Tg or melting, to be separated.
Therefore, when detailed studies are
dq
performed on polymeric materials, it is essential ¼ Cs β þ f ðt; T Þ ð7:4Þ
to ensure that all samples have the same thermal dt
history (e.g. the same degree of crystallinity) if Equation 7.4 above separates the two
consistent results are to be obtained. components of the DSC signal, Csβ, the heat
capacity component, and f(t,T) which describes
Crucibles for DSC and DTA thermal events such as an endothermic transition.
Most commonly, single-use aluminium pans are The heat capacity of the product of the thermal
used for measurements from 120 to 870 K. event will usually differ from that of the reactant,
Although aluminium does not melt until 933 K, but will be masked by the thermal event. Using a
there is a risk of irreversible and expensive sinusoidally varying heating rate, the reversing
alloying of the pan and temperature sensors (heat capacity), and non-reversing (thermal
above 870 K. The use of a lid on the pan typically event) components can be separated. Ideally
gives an order of magnitude improvement in 4–6 oscillations should occur over the duration
sensitivity. Crimp-on lids can be used which of the thermal event.
Glass transition Simultaneous Thermal Analysis

Cold crystallisation
Simultaneous Thermal Analysis (STA) combines
the benefits of TGA and DTA/DSC into a single
EXOTHERMIC
experiment. The combination has two distinct

advantages. It allows unambiguous separation of,
Crystal melting or unification of, particular thermal events, such as
whether an endothermic process preceded the
mass loss stage, or whether the mass loss itself
TEMPERATURE was the endothermic process. The subtle
differences in instrumental design of the separate
Fig. 7.10 Typical DSC curve obtained during heating
of a crystalline polymer that has been subject to fairly TGA and DSC instruments can make such differ-
rapid cooling entiation problematic. More importantly, during
www.ebook777.com
thermal decomposition mass is often lost. In order about by chemical reactions. For this type of
to obtain a mass-specific data for a thermal transi- measurement the specimen is subjected to an
tion (or, for example, the heat capacity of a poly- oscillating stress, usually following a sinusoidal
mer char), it is necessary to know what mass of waveform. The applied stress produces a
material is giving rise to the signal. This can be corresponding deformation or strain.
obtained directly from STA data, since the mass is Although TMA and DMA are rarely used for
known across the temperature range. studying the thermal decomposition of polymers,
DMA is an essential tool for the development of
fire-safe composite materials, such as the carbon-
Thermomechanical Analysis fibre construction of the Airbus A380 or Boeing
and Dynamic Mechanical Analysis 787, where maintenance of structural integrity in
a fire is at least as important as the suppression of
In thermomechanical analysis (TMA) a probe flammability.
applies a force to a sample during heating, and
the movement of the probe is recorded as a func-
tion of temperature. It involves measurement of a Rheology
specimen’s dimensions (length or volume) as a
function of temperature whilst it is subjected to a A decomposing thermoplastic changes from a
constant mechanical stress. In this way thermal fluid-like or gel-like solid, to a gel-like or viscous
expansion coefficients can be determined and liquid often with no distinct transition from one
changes in this property with temperature phase to the other. TMA and DMA are designed
(and/or time) can be monitored. Many materials for measurement of the fluid or gel-like
will deform under the applied stress at a particu- behaviour of a solid, where rheology is designed
lar temperature which is often connected with the to quantify the gel-like or viscous behaviour of a
material melting or undergoing a glass transition. polymer in the liquid state. There is considerable
Alternatively, the specimen may possess residual overlap, and many material states can be
stresses which have been “frozen-in” during investigated by both techniques. The term rheol-
preparation. On heating, dimensional changes ogy (from rheo meaning “to flow”) refers to both
will occur as a consequence of the relaxation of physical deformation and flow of material under
these stresses. Usually the force is static, though an applied force. Thus, the rheological behaviour
modern instruments can offer the facility of an of polymers encompasses a wide range of mac-
oscillating force. Useful physical data, such as roscopic phenomena including the flow of vis-
compressive and tensile strength, softening, cous liquids, the mechanical properties of elastic
shrinking, thermal expansion, glass transition, solids, and viscoelasticity (the time dependent
and melting can all be obtained using TMA. mechanical properties of a polymer).
In dynamic mechanical analysis (DMA) a The mechanical properties of elastic solids
solid sample is held between two moveable show a region where an applied stress is propor-
clamps and subject to oscillatory forces, and the tional to the resultant strain (Hooke’s law). For
response measured during temperature liquids, Newton’s law describes a region where
programmed heating (or cooling). This allows the applied stress is proportional to the rate of
the measurement of the mechanical properties, strain. Both are valid only for small strains or
such as mechanical modulus or stiffness and rates of strain. In many cases, a polymer will
damping of a specimen as a function of tempera- show the characteristics of both a liquid and a
ture. DMA is a sensitive probe of molecular solid, and neither of these limiting laws will
mobility within materials and is most commonly adequately describe its behaviour. The system is
used to measure the glass transition temperature then said to be in a viscoelastic state, where it
and other transitions in macromolecules, or to will respond like an elastic solid to a sudden
follow changes in mechanical properties brought force, but like a viscous liquid to a slow force.
0 00
η* ¼ η iη ð7:6Þ
The release of fuel from a decomposing polymer,

as volatiles or bubbles migrating through a vis-
coelastic medium is a function of its rheological
properties.
The Five Regions of Viscoelastic

Behaviour
The physical properties of an amorphous polymer
are related to its molecular motion, which is
governed by the flexibility of the polymer chains
and their temperature. For a linear, amorphous
polymer, five distinguishable states can be observed
if the elastic modulus is measured over a range of
temperature. The relaxation modulus Er (the time
dependent change in stress under a constant strain),
for polystyrene, plotted as log Er against tempera-
ture, shows these five distinct regions, Fig. 7.12.
(i) The glassy state: Below 360 K, even short
range motion of one polymer molecule rel-
Fig. 7.11 Schematic diagram of a parallel plate rheome- ative to another is frozen, giving the mate-
ter with environmental chamber (oven) rial high rigidity, but with corresponding
brittleness.
(ii) Retarded elastic state: The glass transition
Rheometry provides an essential tool kit for temperature Tg is in this region, and the
adjusting the conditions in polymer processing. sharp decrease in modulus with temperature
The use of rheometry for studying aspects of the reflects the increase in molecular motion as
thermal decomposition of polymers is less com- the temperature rises from Tg to 30 K above
mon. Probably the parallel plate rheometer, with Tg. Just above Tg the movement of the chain
environmental chamber (Fig. 7.11) is most suited segments is still rather slow, giving leathery
to investigating the changing properties of a properties to the material.
decomposing polymer. The polymer sample is (iii) The rubbery state: At approximately 30 K
mounted between two parallel circular metal above Tg the modulus curve begins to flat-
plates. One plate is driven mechanically with an ten out into the plateau region, giving a
oscillating motion of fixed angle or torque, and softer elastic material, since the polymer
the resistance to the motion is recorded. Various chains can move relative to each other
parameters including the elastic and relaxation over a short range, but liquid flow (or long
modulus and the viscosity can be determined. range movement) is still inhibited.
The complex dynamic modulus G* can be (iv) Rubbery flow: After the rubbery plateau,
separated into two components, the storage mod- the modulus again decreases as liquid-like
ulus G’ (representing the solid or gel like flow becomes possible.
behaviour) and the loss modulus G” (representing (v) Viscous state: Above 450 K, there is little
the viscous or liquid like behaviour). evidence of any elastic recovery in the poly-
0
G* ¼ G þ iG
00
ð7:5Þ mer, and all the characteristics of a viscous
liquid become evident. Here, there is a
These are related to the corresponding dynamic steady decrease of the modulus as the tem-
viscosities η*. perature increases.
www.ebook777.com
Fig. 7.12 The five regions

of viscoelasticity for an
amorphous polymer
(i) glassy state; (ii) retarded
elastic state; (iii) rubbery
state; (iv) rubbery flow
state; and (v) viscous state
In the glassy, elastic and rubbery states move- were released. The most widely used fire
ment of bubbles through the polymer matrix will retardant, aluminium hydroxide releases water
be inhibited. Sufficient fuel release to support vapour as it decomposes; surface oxidation can
ignition is unlikely in the rubbery flow state. As result in carbon dioxide (CO2) formation,
the polymer moves through the viscous state fuel which tends to delay ignition; evolution of
release will become increasingly easy, allowing a hydrocarbons is indicative of the fuel release
critical concentration to accumulate above the step leading to ignition.
sample to support ignition. In the simplest form of evolved gas analysis
(EGA) a non-dispersive infrared analyser (NDIR)
may be attached to the exhaust line of a TGA, DSC
Techniques Involving Chemical Analysis or STA, in order to provide a time resolved profile
of Decomposition Products of CO2, carbon monoxide (CO), hydrocarbons and
oxygen concentrations as a function of time, dur-
Evolved Gas Analysis (EGA) ing the TGA run [13]. This technique allows the
Thermogravimetric Analysis (TGA) provides the major decomposition products to be distinguished
crucial data with which to interpret the during thermal decomposition, particularly in
complexities of burning behaviour. For example, relation to fuel gases and non-combustible
ignition can only occur when there is sufficient products. The evolution of CO2 formed on the
fuel being released into the gas phase. The tem- surface of the decomposing polymer is indicative
perature at which the first sharp mass loss occurs of the consumption of oxygen and fuel, but
normally corresponds to the critical surface tem- accompanied by self-heating. CO2 is not formed
perature for ignition, and if, for example, the in the gas phase below 500 C.
addition of fire retardants delays this onset, they
may also delay ignition. However, if this mass Thermal Analysis with Fourier Transform
loss is non-combustible (such as the loss of HCl Infrared Analysis
from PVC), it will clearly not result in ignition; it Fourier Transform Infrared (FTIR) has
is, therefore, important to know the chemical revolutionised infrared analysis, producing spec-
nature of the products released during mass loss tra of higher resolution in a much shorter time.
in TGA, as well as the temperature at which they For polymer decomposition studies this allows the
STA
Computer Balance mechanism
EVA+Nanocomposite in Air
0.14
0.12
0.10
0.08
Dry air/nitrogen
Abs
0.06
0.04
0.02
−0.00
4000 3000
cm-1
2000 1000
Furnace with thermocouple
Balance /Thermocouple beam
Sample/reference pan
Thermocouple assembly
FT-IR Spectrometer Gases flow
Gas cell
Laser
Heated Line
Gases out Window Mirror
Fig. 7.13 Schematic of a typical STA-FTIR instrument set-up
spectra of gas phase products to be collected con- Pyrolysis-Gas Chromatography Mass

tinuously as a function of temperature. Gas phase Spectrometry
infrared spectra are much sharper than condensed Pyrolysis-gas chromatography–mass spectrome-
phase spectra, and generally identification of indi- try (py-GCMS) is a three stage process
vidual molecules is possible in the gas phase, consisting of pyrolysing about 1 mg of sample,
whereas only the presence of functional groups separating the mixture of volatile products by
may be obtained from condensed phase spectra. gas chromatography, and analysing each of the
However, when a large number of similar components by mass spectrometry. From its eso-
molecules are present, each with their distinctive teric origins, gas chromatography–mass spec-
spectra superimposed, deconvolution software trometry instruments are now standard, low
may be necessary in order to identify and quantify cost and do not depend on specialist technical
the individual components. Figure 7.13 shows a support (Fig. 7.15).
schematic of a typical STA FTIR set up [14]. Given the large numbers of different products
TGA, DSC or STA, combined with FTIR that can result from a single stage in the decom-
analysis of the vapour phase provides a very position of a polymer, separation of the products
rich profile of polymer decomposition through is often required, before they can be identified.
the spectra of the evolved products. This is Controlled temperature and heating rate pyroly-
shown in Fig. 7.14 for ethylene-vinyl acetate sis may be followed first by sample collection on
copolymer in nitrogen, and can be deconvoluted a sorption tube, or for rapid pyrolysis, direct
to provide profiles with respect to temperature of injection into a gas chromatography column
the evolution of individual species. The tech- (py-GC). In its simplest form, a gas chromato-
nique is gaining popularity amongst the fire retar- graph consists of several metres of capillary tube
dant community, but the deconvolution of with a well-controlled flow of a carrier gas
spectra, particularly where compounds are pres- through it. The tube or “column” is lined with
ent for which no calibration data has been an absorbent solid coated with an involatile liq-
recorded, limits the applications of the technique. uid that will adsorb and desorb constituents in the
www.ebook777.com
0.10 Linked spectrum at T=327C
CH3COOH
CO2
Abs
0.05
0.00
Linked spectrum at T=412C
1797
2895 C2H4
Abs
0.10
0.00
Linked spectrum at T=493C
CO H2O CO2
0.01
Abs
0.00
4000 3000 2000 1000
cm-1
Fig. 7.14 FTIR spectra recorded during decomposition of ethylene vinyl acetate at a heating rate of 10 K min1
Fig. 7.15 Typical pyrolysis GC-MS apparatus

sample, depending on their affinity for the liquid, the temperature evolution profile of a species
and their volatility. The column is located in a may be monitored.
temperature programmable oven. Molecules with
different adsorption properties relative to the col- Thermal Volatilization Analysis
umn packing will reach the end of the column at Another useful method for quantitatively
different times, and hundreds of compounds may collecting and separating polymer decomposition
be separated in a single run. A detector placed at products is thermal volatilization analysis (TVA)
the exit of the gas chromatograph will respond to based upon the apparatus and techniques
the concentration of molecules, other than the described by McNeill et al. [16]. The apparatus
carrier gas, giving a series of peaks. For a single consists of a sample chamber (heated by a pro-
peak, the time from injection is characteristic of grammable tube furnace) connected in series to a
the molecule, and the area under the peak is primary liquid nitrogen-cooled sub-ambient trap
proportional to its concentration. Column lining, and a set of four secondary liquid nitrogen-
column temperature programming, carrier gas cooled cold traps. The whole system is continu-
flow rate, sample size, and detector type can be ously pumped to a vacuum of 1 107 atm.
adjusted to achieve optimal separation of the Volatile condensable products from a ~25 mg
decomposition products. sample are initially trapped at two stages: the
Simple detectors include the flame ionization ‘cold-ring’ (T 285 K) immediately above the
detector (FID) which gives a signal roughly pro- heated area of sample tube (for high boiling point
portional to the carbon content of the molecule, materials) and the primary liquid nitrogen cooled
and hence its calorific potential. They have a sub-ambient trap (T 77 K) for lower boiling
dynamic range spanning several orders of mag- point species. The pressure increase is propor-
nitude. However the reduction in cost of the mass tional to the evolution of both condensable and
spectral analysers, which provide a chemical fin- non-condensable volatiles from the sample
gerprint for each separated molecule, which can [17]. These separated fractions may be subse-
then be compared to databases such as the NIST quently analysed by gas-phase FTIR (section
mass spectral library, provide the rapid and “Thermal Analysis with Fourier Transform Infra-
powerful tool for investigating the true complex- red Analysis”) and GC–MS (section “Pyrolysis-
ity of polymer decomposition. Figure 7.16 shows Gas Chromatography Mass Spectrometry”).
py-GC/MS chromatograph of the decomposition
products of soy-styrene-divinyl benzene (SSD) Microscale Combustion Calorimetry
polymer [15] collected from 323–873 K at A relatively recent innovation in fire calorimetry
10 K min1. has been the development of the microscale com-
Pyrolysis GC-MS complements TGA-FTIR bustion calorimeter (MCC) [18]. The apparatus
as it provides snapshot identifying all evolved was developed by Lyon and Walters at the
species over a particular temperature range. U.S. Federal Aviation Administration (FAA)
TGA FTIR shows the variation of the spectrum [19]. A schematic of the apparatus is shown in
as a function of temperature, once identified, Fig. 7.17. It thermally decomposes 1–3 mg of
SSD_50to500_2 Scan EI+

5.71 TIC
100
3.97e7
16.53
7.08
8.75 8.88
%
3.80
2.35 2.98 5.12 7.27 8.26 10.24 16.84
6.75 10.00
2.08 10.46 11.85 13.06 14.47 15.02
0
Fig. 7.16 Py-GC/MS chromatographs of the decomposition products of soy-styrene-divinyl benzene (SSD) polymer
collected from 323 K to 873 K at 10 K min1
www.ebook777.com
Fig. 7.17 Schematic diagram of MCC
sample in an alumina crucible at a heating rate of its popularity has led, for example, to a number of
1 K s1 in an atmosphere of nitrogen (or synthetic fire retardant development laboratories publishing
air). The pyrolysis gases are then passed into the articles describing their latest fire retardant
combustor at 1173 K and mixed with oxygen. formulations, based solely on MCC data, paying
The oxygen depletion of the dry gas is measured, less regard to what that data actually means.
and a computer algorithm converts the oxygen Figure 7.18 shows the heat release curves for
depletion and mass flow rate data into specific five common polymers measured using the
heat release rate (W/g), as a function of tempera- microscale combustion calorimeter. In particu-
ture. For polymers that decompose to release lar, the pyrolysis temperature and the peak of
only fuel, the information obtained is similar to specific heat release rate of the most useful
that for DTG in nitrogen. For materials showing indicators of fire performance.
gas phase inhibition, the severe oxidising condi-
tion of 1173 K will usually suppress any reduc-
tion in oxidation. For materials producing char in Choice of Atmosphere and Heating
nitrogen, the mass of residue will increase, Rate in Thermal Analysis
though if the experiment is repeated using syn-
thetic air as the pyrolysis gas, the char yield can Thermal analysis is generally conducted in nitro-
be separated from the formation of an inorganic gen or in dry air. Studies of thermal stability (for
residue. example to ensure that a material does not
The apparatus has grown in popularity due to degrade at its normal operating temperature)
its excellent design, relatively low cost and ease of would normally be carried out in air, or for
use. It is particularly popular with the aircraft accelerated ageing, in elevated oxygen
manufacturer, Boeing, who uses it for both atmospheres. The decomposition of a polymer
materials screening and quality control. However, is known to be dependent on the atmosphere in
Fig. 7.18 MCC heat

release rate curves for
common polymers
which the decomposition occurs. For some with which TGA can be run in flowing air, and
polymers the decomposition occurs 200 K the potential for added insight (such as the for-
lower in air than in nitrogen, while for a few mation of a more stable char in an oxidising
others it makes little difference. In other cases, atmosphere) it is often also worthwhile obtaining
certain decomposition stages are delayed by the the TGA in air.
presence of oxygen, perhaps through the forma- The choice of heating rate during thermal
tion of a more thermally stable char. The decom- decomposition of polymers has been the subject
position of polypropylene has been widely of much controversy. Low heating rates (around
studied (section “Polypropylene”) and it has 1 K min1) produce better quality data and better
been shown that the process is accelerated by defined and separated chemical events. High
oxygen, and that the oxygen permeates the heating rates (around 100 K min1) are more
decomposing polymer, not only affecting the representative of those found in a fire. However,
surface layers. when analysing the products during the thermal
In contrast, in order to understand the decomposition of a polymer, for example in
reaction-to-fire behaviour of a material, it should TGA-FTIR, it is often necessary to use larger
be recognized that amounts of sample (~15 mg) in order to obtain
(i) Heating rates in fires are probably closer to representative spectra. With such larger samples,
100 K min-1 than the 10 K min1 typical of it is more difficult to maintain thermal equilib-
thermal analysis, causing faster fuel pro- rium across the sample bulk, and therefore lower
duction flows and preventing the diffusion heating rates (around 10 K min1) are preferable.
of oxygen into the polymer. Conversely, a faster heating rate will produce a
(ii) The atmosphere under a flame is practically higher concentration of volatiles in a shorter
free of oxygen. time, although these may become mixed in the
(iii) Fuel is generally released from a sampling line before analysis.
decomposing polymer as bubbles of vapour,
forcing their way out through a molten or
porous condensed phase, and released from Char and Residue Analysis
the polymer under anaerobic conditions.
These factors all indicate that fire behaviour, The complexities of the thermal decomposition of
both before and after ignition, is better polymeric materials can be investigated by a
represented by decomposition in nitrogen, than combination of analysing the volatile components
by decomposition in air. However, given the ease as described above, and by analysing the
www.ebook777.com
chemical composition of the char or residue. In scattered wavelengths of lower energy infrared
terms of improved fire retardancy, enhancement radiation emitted from the sample can be used
of the char layer normally provides the best route very effectively to quantify the presence of
for improving the fire safety. Chars may show the carbon-carbon single bonds, carbon-carbon dou-
chemistry of the solid phase processes prior to ble bonds and carbon-carbon aromatic bonds in a
carbonisation. For example, polyacrylonitrile carbonaceous char layer, although care must be
fibre decomposes to produce materials which ulti- taken to avoid using a laser wavelength that
mately become carbon fibre (a modern, heterocy- causes fluorescence. More sophisticated analysis
clic material composed of graphite layers with has been undertaken using nuclear magnetic res-
nitrogen heterocycles along their perimeters). onance (nmr) with a solids probe (magic angle
Unfortunately, chars are predominantly car- spinning) to examine the chemical environment
bon, and can be difficult to analyse. In addition of any atomic nuclei with uneven numbers of
to the chemical composition of chars, the three nucleons, such as C-13, F-19, P-31 etc. [20].
dimensional and microscopic structure is also
important. If the char provides a coherent, uninter-
rupted layer which seals the underlying polymer Decomposition of Polymers
from the attack of radiant heat and oxygen, while
preventing the escape of fuel, it will be much more Different polymers decompose in different ways.
effective in reducing the attack of fire. Scanning When a polymer is heated its chains will start to
electron microscopy (SEM) is the most effective break down, which eventually results in the for-
tool surveying the permeability of the char or mation of volatile fuel molecules. The pyrolysis
residue layer. With the addition of elemental anal- of a polymer, which turns polymer chains of
ysis, more information about the chemistry of the 10,000–500,000 carbon atoms into species
formation of the protective layer can be obtained. small enough to be volatilized, often involves
For example, some inorganic materials migrate to breaking the polymer chain. In some cases, the
the surface of the decomposing polymer and form chain releases groups from its ends most easily,
low melting inorganic glasses. known as end-chain scission or unzipping. In
Analysis of the chemical composition of char others, the chain breaks at random points along
and residue layers is complicated by the high its length, known as random chain scission. A
absorptivity of carbonaceous carbon. If spectro- third process, in which groups attached to the
scopic techniques such as Fourier transform backbone as side chains are released, is known
infrared (FTIR) are used for the analysis of the as chain stripping. If the resulting chain may be
carbonaceous layer, very low signals are often prevented from undergoing chain scission to
found because the infrared radiation is generally form volatiles or lose further substituents, it
absorbed by the black solid, rather than specifi- may instead undergo carbonization that results
cally by the organic functional groups of interest. in char formation. Thus, the conversion of
In addition, a high proportion of carbon-carbon organic polymer into volatile organic molecules
bonds will be non-polar, and so will have no may follow four general mechanisms. While
dipole moment, and therefore no ability to absorb some polymers fall exclusively into one cate-
infrared radiation, so that relatively little useful gory, others exhibit mixed behaviour
information can be obtained. Surface analysis (Table 7.4).
techniques such as diamond-attenuated total
reflectance FTIR (d-ATR FTIR) can identify
the presence of functional groups within the sur- Decomposition Mechanisms
face layers of the char. Sample preparation is
trivial and results can be obtained rapidly. In Random chain scission usually takes place when
addition Raman spectroscopy, which uses visible the bonding energies are similar along the chain.
frequency laser light of high intensity and detects The distinction between random chain and end
Table 7.4 Generalised mechanisms of polymer decomposition

Mechanism Examples of polymer Typical products
Random chain scission Polyethylene Alkanes, alkenes, very little monomer
Polypropylene Alkanes, alkenes, very little monomer
Polystyrene Styrene monomer, dimer and trimer
. . .more generally Monomers and oligomers
End chain scission Polymethylmethacrylate 90–100 % monomer
Polytetrafluoroethylene 90–100 % monomer
. . ..more generally monomer
Chain stripping Polyvinyl chloride Hydrogen chloride, aromatic hydrocarbons and char
Polyvinyl alcohol Water and char
Generally Small molecules and char
Cross-linking Polyacrylonitrile Hydrogen cyanide and char
Polyetheretherketone Volatile aromatics and char
Generally Char and volatile products
chain scission is generally based on the extent of kinetic factors may also play a role in deter-
monomer evolution, and on the measurement of mining the abundance of a specific compound.
the polymer’s average molecular mass when The formation of small radicals from the end
heating the polymer below its typical pyrolysis of a polymeric chain can be kinetically
temperature. The slow evolution of monomer favoured, and, as a result, formation of small
and rapid decrease of the molecular mass is an radicals in the initiation step is more common
indication of random chain scission. than predicted by thermodynamic criteria. In
Both end and random chain scission are polystyrene, end chain scission produces sty-
dominated by free radical processes. The decom- rene monomer in competition with random
position of a polymer by a free radical mecha- chain scission, which ultimately produces the
nism can be divided into three processes. dimer, trimer, tetramer etc.
Initiation—the time taken and for the first bond Propagation—once a polymer chain has a free
to rupture, generating two free radicals. In radical, the reactivity will propagate along the
random chain scission each radical will be at polymer macromolecule very quickly, releasing
the end of a polymer chain, in end chain monomers or other small volatile molecules.
scission the H · or · CH3 radical released Termination—at some point the rapid propaga-
will attach itself to another polymer chain. tion step must end. This may be at the end of
This is likely to occur at the weakest link in the polymer chain, at the branching point, or
the polymer chain, such as a C–C single some other structural irregularity where two
bond one bond away from a double bond, radicals can combine.
triple bond or an aromatic ring
γ
C Cβ Cα ¼ C , the β-carbon
shown here. These bonds are generally Thermodynamics of Polymer
weaker than other C–C or C–H bonds, and Decomposition
significantly weaker than aromatic, double,
or triple bonds. Bond breaking can take At every propagation step in the depolymerisation
place as a random or end chain scission. of a vinyl polymer –(CH2CHX)n– a C–C σ-bond
Since larger free radicals are more stable is broken, and a C ¼ C π-bond is formed
than those on small molecules, the fragmenta- (Fig. 7.19).
tion in the middle of the polymeric chain is The enthalpy on breaking a σ-bond is around
favoured thermodynamically compared to the 320 kJ mol1, and that of forming the π-bond
formation of smaller molecules. However, in the volatile fragment is around 260 kJ mol1.
www.ebook777.com
H 2C
C C CH
X X X X X X X X
σ−bond broken π−bond formed
Fig. 7.19 Generalised mechanism for free radical depolymerisation of a vinyl polymer
Thus, in purely enthalpic terms, depolymerisation phase and the gas phase and, for sustained burn-
requires about 60 kJ per mole of volatile alkene ing, in the gas phase. The applied heat and exo-
formed, i.e. ΔH ¼ +60 kJ mol1. However, a thermic release of energy by chemical reactions
reaction can only proceed when the free energy such as combustion must exceed the energy
change ΔG is negative, and since: required for endothermic process such as heating,
melting, depolymerisation and volatilisation.
ΔG ¼ ΔH TΔS ð7:7Þ
the entropy change involved must also be consid-

Kinetics of Polymer Decomposition
ered. The entropy (or disorder) of a decomposing
polymer is obviously increasing, since the num-
Chemical kinetics is the study of reaction rates.
ber gas molecules is increasing. Thus ΔS is posi-
Theoretically, the rate of a chemical reaction, as
tive, and is typically around +120 J mol1 K1.
a function of state variables such as temperature,
At room temperature, 300 K, the free energy
pressure, and concentration, can be predicted for
change on depolymerisation is thus:
any set of conditions. However, since many
ΔG ¼ þ60 000 ðþ300 120Þ J mol1 chemical reactions are the sum of a number of
competing processes, such predictions do not
ΔG ¼ þ24 000 J mol1 ¼ þ24 kJ mol1
often agree with the experimental data. The
This demonstrates the obvious fact that at room most common reason for carrying out kinetic
temperature depolymerisation is unfavourable, studies is to investigate reaction mechanisms,
and hence polymers are thermodynamically sta- although empirical estimates of gasification
ble. However, at a much higher temperature of rates provide essential data for fire models that
570 K, the TΔS term is greatly increased: include condensed phase fuel. Kinetic data
provides insight into the mechanism, although
ΔG ¼ þ60 000 ðþ570 120Þ J mol1 this generally needs to be corroborated with
ΔG ¼ þ8 400 J mol1 ¼ 8:4 kJ mol1 chemical evidence. The simplest media for
investigating the kinetics of chemical reactions
From this it can be concluded that, above a cer- are gas and solution phases, which are described
tain temperature, sometimes known as the ceiling briefly below.
temperature, depolymerisation will prevail, as
polymer chains (formed at lower temperatures) Gas and Solution Phase Kinetics
“unzip” to regenerate monomers. The ceiling For a reaction of A ! Products, simple gas or
temperature for polymethylmethacrylate is solution phase kinetics classifies reactions in
about 500 K, and for polystyrene 580 K. terms of order (Table 7.5).
In addition to this simplified view of In gas and solution phases, most reaction rates
depolymerisation, a burning polymer is a complex lie between zero and second order—genuine
situation, where chemical reactions can take place third order reactions, where three molecules col-
in the condensed phase, between the condensed lide simultaneously, and possess sufficient
Table 7.5 Orders of chemical reactions, and their corresponding rate equations
Reaction Differential rate
order Description equation Integrated rate equation Examples
Zero Reactant d½A k t ¼ ½A e.g. vaporisation, where the
concentration does ¼k reactants have to pass
dt
not affect rate through a narrow orifice, or
when heat is supplied at a
fixed rate.

First Rate of reaction d½A ½A0 Typical of a simple
depends on amount dt ¼ k ½A ln
½A
¼ kt decomposition reaction, or
of reactant radioactive decay.
Second Rate of reaction d½A 1 1 Common for gas phase
2
¼ kt
depends on amount dt ¼ k½A ½A ½A0 processes such as
present of two 2A ! Products. Kinetics
reactants (which more complex for
may both be A) A + B ! Products
!
nth Rate is not a simple d½A n 1 1 1 May be a simple fraction
function of reactant Rate ¼ dt ¼ k ½A kt ¼ (based on kinetic analysis)
n 1 ½An1 ½A0n1
concentration or an experimental value
(for a complex processes)
energy to react, are very unlikely. Where com- dependence of the reaction, Ea (the proportion
plex reaction pathways occur, particularly with of molecules with sufficient energy to overcome
multiple competing processes, precise fractional the barriers to reaction, Ea) to be quantified.
order processes are also observed. Unfortunately, even in the gas and solution
phases, all theoretical attempts to make general
Solid Phase Kinetics predictions of the value of the pre-exponential
In the solid phase the reacting molecules do not factor have been unsuccessful.
move freely and collide at a rate controlled by the The thermal decomposition of a polymer
thermal energy of the system in contrast to gases involves many competing chemical processes.
and liquids. The reacting species cannot move The goal of chemical kinetics is to parameterise
through the solid and so the rate will not depend all the competing processes, so they switch in or
on concentration within the solid. If a diffusion out under different conditions. The reality is that
process controls the mechanism, the rate will be such universal mechanisms have so far only been
dependent on concentration gradients within the applied to simple gas phase systems such as the
solid. Alternatively, a reaction interface may combustion methane (which itself involves over
move through the bulk with pure reactant ahead 50 reactions) [21].
of it and pure product behind. Here concentration Generally, solid-phase pyrolysis is assumed
is meaningless and is replaced by fraction to follow the Arrhenius equation with three
reacted, α. parameters defining the reaction kinetics. These
Experimentally, the rate constant (k) has been kinetic parameters are reaction order, activation
found to vary as a function of two variables, energy (Ea) and pre-exponential factor (A).
A and Ea, defined by the Arrhenius equation. Figure 7.20 shows how the pairs of values of
Ea A and Ea affect the mass loss rate, for a particular
k ¼ A e RT : ð7:8Þ peak pyrolysis temperature (Tp), based on the
This is a very flexible mathematical function, analysis of Lyon [22]. Rearranging the Arrhenius
allowing the influence of the underlying rate to equation gives
be described by the pre-exponential factor A,
(including the number of collisions, the molecu- Ea
Tp ¼ ð7:9Þ
lar orientation, and factors intrinsic to the Rln kAp
reaction mechanism) and the temperature
www.ebook777.com
Ea/kJmol−1: A /s−1
0.005
250 : 1019 292 : 1019
0.004 338 : 1019
Mass loss rate/s-1
210 : 1013
180 : 1013
0.003 242 : 1013
130 : 107
0.002 110 : 107 150 : 107
0.001
450 550 650 750 850 950
Temperature/K
Fig. 7.20 Relationship of Arrhenius parameters A and Ea, shown as Ea in kJmol1: A in s1 by the peak (Tp) of each
mass loss curve, to the pyrolysis temperature and mass loss rate
Polymer decompositions are frequently free remaining polymer chains is the same as the
radical process, where first a bond is broken initial average molecular weight.
resulting in two free radicals (initiation), then For end-chain initiation, the rate of mass loss,
the reactive free radical causes a rapid sequence dm
dt , is described by
of similar reactions (propagation), until a process
occurs which removes the activity of the free dm
¼ D p kei 2n ð7:11Þ
radical (termination). dt
The initiation mechanism in decomposition
n the number of the polymer chains, so 2n is the
kinetics can be considered as random- or
number of chain ends, and kei is the rate constant
end-chain initiation with propagation by unzipping
for end-chain initiation. The number of polymer
or random chain scission, without termination
chains is simply the mass of the sample divided
reactions other than exhaustion of the polymer
by the molecular weight of each chain, or
chain by unzipping. The rate of mass loss, dmdt , for
random-chain initiation can be considered as m
n¼ ð7:12Þ
D pWm
dm
¼ D p kri m ð7:10Þ
dt where Wm is the molecular weight of the
Where Dp is the degree of polymerisation, monomer, which allows the expression to be
(or number of monomer units per polymer rearranged, to give
chain); and kri is the rate constant for the dm 2kei m
random-chain initiation reaction. Here, the rate ¼ ð7:13Þ
dt Wm
constant of the propagation reaction is ignored in
the expression because it is very rapid relative to It is apparent that for random initiation the
the initiation step. The degree of polymerisation rate is dependent on the chain length of the poly-
is included in the equation because, for each mer, or “dispersity” of the sample; whereas for
initiation, Dp monomer units will be released; end-chain initiation the rate is dependent on the
the remaining mass, m, corresponds to the reac- number of chain ends, or initiation sites. How-
tant concentration; and the process is assumed to ever, in both cases, the rate is assumed first order
follow first order kinetics. The polymer unzips with respect to the mass of the sample and it is
completely so the molecular weight of all assumed that the average value used for Dp
adequately represents the effect of chain length without the aid of computers. Only in the last
on the kinetic process. 25 years has thermal analysis data been readily
In the random-chain initiation expression, the available in digital form. A great deal of pains-
longer chains are more likely to decompose first taking experimental work and analysis relied on
because they have more initiation sites, and these manual techniques, and the parameters
therefore the longer chains would decompose obtained are still valid, and in use today. A com-
more quickly than the shorter chains. If the initial pilation of kinetic parameters for several com-
sample had a range of molecular weights, the mon polymers, prepared by Lyon et al. [3], is
distribution of the molecular weight would included in Table 7.1. In order to set the different
change with time. In this case, the rate equation methods into context the brief description of the
would be, approaches is provided below.
Kinetics describing condensed phase decom-
dm 2kei mn position processes often involves physical and
¼ ð7:14Þ
dt Wm chemical mechanisms. In pyrolysis modelling,
the overall kinetic expressions are highly
with 1 < n < 2. simplified, if considered at all. The most com-
Depending on the molecular weight distribu- mon assumption is that the reaction rate is
tion the reaction order normally lies between one described by first order or sometimes zero order
and two for random-chain initiation leading to kinetics.
complete unzipping of a polydisperse system [23]. Since the kinetic parameters cannot be
This comparison shows how the decomposi- measured directly, they have to be extracted by
tion process can be influenced by the estimation methods based on microscale data.
polymerisation process, and its influence on the Generally there are two approaches to property
molecular weight distribution. The same polymer estimation: direct, analytical methods [32,
may be polymerised using free radical, cationic, 24–26], and curve-fitting sometimes with the
anionic or catalytic reactions, which result in use of evolutionary algorithms [27–32]. Recently,
different molecular weight distribution curves methods based on optimisation algorithms and
for polymers of the same average molecular curve fitting have gained popularity. These
weight. For a particular polymer sample, both methods use specific software for performing
end and random chain initiation reactions may the simulation, and can analyse complicated,
be important. The activation energies for the overlapping reactions and noisy data but still
different initiation steps may be quite different, may require significant computational time for
leading to large variations in the relative rates the iteration process to converge. Due to the
with temperature. For instance, in PMMA, at low kinetic compensation effect (where a change in
temperatures (around 570 K), end-chain initia- the activation energy (Ea) can be partially or
tion predominates. At higher temperatures completely compensated for by a change in the
(around 770 K), the random-chain initiation frequency factor (A) [33], described in section
step dominates. “Solid Phase Kinetics” and Fig. 7.20), and other
factors, the solution is not unique. Furthermore,
the process is stochastic, so if it is repeated, it
Modelling Polymer Decomposition may not produce exactly the same parameters.
and Pyrolysis Much simpler and faster are analytic methods,
which give unique solutions based on reference
Historically (from 1950–1990), determination of points (such as the peak reaction rate) without
kinetic parameters for polymer decomposition requiring lengthy iterations, but often have
involved analysis of paper traces from thermal more restrictions and limited accuracy. They
analysis output and numerical calculation, only operate well on simple, non-noisy and
www.ebook777.com
non-overlapping, well-separated reactions, and “Techniques Involving Chemical Analysis of

are often limited to a specific reaction path. Decomposition Products”). The most detailed
If the reaction chemistry is known it can be detection methods for volatiles produced by ther-
used to identify the reaction path. For an engineer- mal analysis techniques are Fourier transform
ing solution of polymer pyrolysis rates, this is infrared (FTIR) spectroscopy, and mass-
usually too complicated and ambitious. In poly- spectrometry (MS), particularly when gas chro-
mer pyrolysis, several competing reactions matography (GC) is used for prior separation of
occur simultaneously, which may overlap in the individual components.
time and temperature. This leaves pyrolysis and The kinetics of thermal decomposition reactions
fire modellers with the choice of describing were traditionally determined under isothermal
each reaction, with the resultant complex conditions. Dynamic, linear heating rates are now
computations, or to mathematically mimic the more common as computers are able to tackle
mass loss rate data in its simplest form, and estab- the greater mathematical complexity of the data
lish model-specific kinetics that can reproduce the analyses. Various approaches to kinetic data analy-
experimental curve. It has been established that sis aim to provide consistent, predictive methods of
the model parameters can compensate for other reaction rate parameterisation, and contribute to a
shortcomings and simplifications [34, 35]. As better understanding of the physical and chemical
approximations are inherent in a simplified reac- characteristics of solid state decomposition.
tion mechanism, and the “components” included
in the mechanism may have no fundamental phys- Mathematical Models of Polymer
ical significance, choosing their values freely Decomposition Kinetics
provides the best possible approximation to fit The pyrolysis process of many polymers may be
the experimental data. Within the two broad described by a simple reaction scheme:
categories of curve-fitting and analytical methods,
a number of different approaches have been devel- Asolid ! Bsolid þ Cvolatile
oped for estimating the kinetic parameters for the
In the kinetic analysis of depolymerisation
thermal decomposition of polymers.
(section “Kinetics of Polymer Decomposition”)
it was assumed all the products were volatile, so
the sample mass represented the amount of
Experimental Determination of Kinetic
reactants. Here, both reactant and product con-
Parameters
tribute to the mass, so the rate of reaction must be
expressed in terms of the degree of conversion.
Solid state kinetic data are of practical interest for
Kinetic models are usually based on a single-step
both thermal decomposition and combustion of
rate Equation 7.15.
polymeric materials [36]. The most popular
experimental techniques used to study kinetics dα
of thermally stimulated reactions are thermogra- ¼ kðT Þ f ðαÞ ð7:15Þ
dt
vimetric analysis (TGA) and differential scanning
calorimetry (DSC) (sections “Thermogravimetric Where t is time, T is temperature, α is the degree
Analysis and “Differential Thermal Analysis and of conversion of reactants to products, and f(α) is
Differential Scanning Calorimetry”). These the reaction model. The rate constant, k, has the
techniques are capable of measuring the global usual Arrhenius form. The kinetic parameters
kinetics of many types of thermally activated can be decoupled by specifying f(α). A reaction
reactions but are not themselves able to give any model that is widely used in thermal analysis
evidence about the nature of these reactions. [37–44], but less in pyrolysis models for fire
Techniques that can provide additional informa- simulation [44, 45], is the nth order reaction
tion on the mechanisms of solid state reactions model in which the residual conversion 1α is
are based on evolved gas analysis (section raised to the power n, so f(α) ¼ (1α)n.
Solid Phase Mechanisms solids such as calcium oxalate and calcium car-
bonate, involving transport processes of less
Historically, several physical mechanisms of complexity than those of a decomposing
solid phase thermal decomposition kinetics polymer.
were proposed [46], typically for simple inor- The first difficulty in studying solid-state
ganic systems, in order to explain experimentally kinetics is in deciding which of such a large
derived thermal analysis data. The mechanisms number of equations is being obeyed. Several
have a common starting point of nucleation, equations may give very similar curves and thus
where initial bond breaking leads to rearrange- appear to give equally good fits to some experi-
ment, followed by the formation of gaseous and mental data. The range of tools and approaches
solid products. This may be repeated randomly, for kinetic parameter determination of solid state
putting strain on the bonds around the product decomposition reactions are shown in Fig. 7.21.
nucleus, and may be geometrically spherical, For many years the different methods of
cylindrical or linear, often dependent upon the obtaining the kinetic parameters for pyrolysis
structure. Alternatively, a two-dimensional inter- reactions were surrounded by controversy, as to
face sweeps through the reactant from one end to which method was the most reliable and effi-
the other. A large number of mechanisms based cient. Each method had its limitations, frequently
on diffusion of product have been proposed, ignoring many factors in order to simplify the
which assume the rate depends on movement of model. For example, the limitations of
gaseous reactants or products through the solid. nonisothermal methods include the need to
The most commonly used are shown in Table 7.6 choose a particular reaction model, the error
[47, 48, 59, 64, 65], but there is no consensus as associated with predicting rates outside the
to which ones are valid. Most of these models range of experimentally determined heating
were devised in an attempt to interpret the sim- rates, and if the reaction order is used as an
pler solid-state decompositions of inorganic adjustable parameter the inability to derive
Table 7.6 Reaction models used to describe the diffusional processes of solid state thermal decomposition
Reaction model f(α) g(α)
2 =2
1 3=
1 Power law
3α α 2
1= 1=
2 Power law 2α 2
α 2
2= 1=
3 Power law 3α 3
α 3
3= 1=
4 Power law 4α 4 α4
5 One-dimensional diffusion 1 1
2α
α2
6 Exponential law α ln(α)
7 Mampel (first order) α1 lnð1 αÞ
8 Second order ð1 αÞ2 ð1 αÞ1 1
9 Third-order ð1 αÞ 3 1h i
ð1 αÞ2 1
2
10 Avrami-Erofe’eva 4ð1 αÞ½lnð1 αÞ3=4 ½lnð1 αÞ1=4
a
11 Avrami-Erofe’ev 3ð1 αÞ½lnð1 αÞ2=3 ½lnð1 αÞ1=3
a 1=2
12 Avrami-Erofe’ev 2ð1 αÞ½lnð1 αÞ ½lnð1 αÞ1=2
1 h i2
13 Three-dimensional diffusion
2ð1 αÞ2=3 1 ð1 αÞ1=3 1 ð1 αÞ1=3
14 Two-dimensional diffusion ½lnð1 αÞ1 ½ð1 αÞlnð1 αÞ þ α
2=3
15 Contracting sphere 3ð1 αÞ 1 ð1 αÞ1=3
16 Contracting cylinder 2ð1 αÞ1=2 1 ð1 αÞ1=2
a
Representing 3, 2 and 1 dimensional growth of nuclei respectively
www.ebook777.com
Experimental Methods
TGA, DTA, DSC, MCC, EGA
Analytical Methods Curve-fitting Optimisation Methods
Model-fitting Model-free Optimisation Pyrolysis

Algorithms Models
- Genetic - ThermaKin
Algorithm - GPyro
Isothermal Nonisothermal Isothermal Nonisothermal
- Hybrid Genetic - Condense
Algorithm Phase CFD
Conventional - Direct Differential - Standard - Kissinger - Shuffled - Pyrolysis
- Freeman-Caroll - Friedman - Ozawa-Flynn-Wall Complex Model of FDS
- Coats-Redfern - Vyazovkin - Vyazovkin Evolution with - Pyrolysis
- McGrattan-Lyon Competitive Submodel of
- Direct Method Complex FireFOAM
- Generalised Direct Evolution
Method
Fig. 7.21 Grouping of methods developed to study solid-phase chemical kinetics
mechanistic information from it. This could be complicated methods yield more accurate results
described as curve-fitting where the physical for clean, simple data.
significance of the parameter is (conveniently)
ignored. The kinetic parameters that are required
to calculate the reaction rates can be obtained Isothermal Model-Fitting Method
from nonisothermal analyses by direct (inspec- (Conventional Method)
tion) and indirect (numerical regression)
methods. Direct methods are fast, simple and This method is identical to that in the kinetics of
sufficiently accurate for certain applications. a simple gas or solution phase reaction. It
Indirect methods are more flexible but require involves determination of the rate constant (k),
sophisticated software, and the resulting kinetic for isothermal decomposition at a number of
parameters may not be unique. different temperatures.
Very often the accuracy of a particular kinetic
method and reaction model is not perfect; how- gðαÞ ¼ kt ð7:16Þ
ever such results are sufficient for predicting the From which specific kinetic parameters such
time to ignition and surface temperature of a as the activation energy (Ea) and frequency fac-
burning solid in fire simulations using a numeri- tor (A) of the Arrhenius Equation 7.17 can be
cal pyrolysis model. Analytical methods can determined.
substitute the computationally expensive optimi-

zation methods when estimating the parameters Ea
k ¼ A exp ð7:17Þ
of pyrolysis kinetics. The choice of method RT
depends on the accuracy required, and the com-
plexity of the experimental data. Methods with First g(α) is calculated using models, such as
fewer reference values work better in the case of zero or first order, or the diffusion models in
noisy data and overlapping reactions, while more Table 7.6. Once the best fit model is found, the
slope of g(α) vs time provides k for each temper- Coats-Redfern Method

ature. Ln k may be plotted against 1/T to obtain This method [53, 54] uses the integral form of the
Ea (slope) and A (intercept). non-isothermal rate law
ðT
A Ea
gð α Þ ¼ exp dT ð7:24Þ
Non-isothermal Model Fitting Analytical β 0 RT
Methods
Coats and Redfern utilized the asymptotic
Direct Differential Method series expansion for approximating the tempera-
This method [49, 50] uses the differential form of ture integral, producing:
the non-isothermal rate law
gðαÞ AR 2RT exp
ln 2 ¼ ln 1
dα A Ea T βEa Ea
¼ exp f ðαÞ ð7:18Þ
dT β RT Ea
ð7:25Þ
RT
Taking the logarithm of the non-isothermal rate
law, gives: Where, Texp is the mean experimental
temperature.
dα=dT A Ea Plotting the left-hand side of Equation 7.25
ln ¼ ln ð7:19Þ
f ðαÞ β RT versus 1/T gives the activation energy (Ea) and
frequency factor (A) from the slope and inter-
Plotting the left-hand side versus 1/T gives the cept, respectively. The model with the best linear
activation energy Ea and frequency factor A from fit is chosen as the final solution. The Coats-
the slope and intercept, respectively. The model Redfern equation was originally used with a
that gives the best linear fit is usually chosen. first-order model g(α) ¼ ln(1-α) but later has
been generalized to other reaction models
Freeman–Carroll Method as well.
The Freeman and Carroll method [51, 52] is a
differential method originally developed with a Kissinger Method
reaction model f(α) ¼ (1α)n. Taking the natu- Kissinger [41, 42] proposed a kinetic analysis
ral logarithm of the differential form of the method for reaction-order models ( f(α) ¼
non-isothermal rate law gives, (1α)n) based on taking the derivative of
dα A Ea Equation 7.18 giving d2α/dT2. Accordingly, the
ln ¼ ln þ ln f ðαÞ ð7:20Þ maximum reaction rate occurs when the second
dT β RT
derivative is zero, from which the following
dα Ea 1 equation can be obtained:
Δln ¼ Δln f ðαÞ Δ ð7:21Þ
dT R T
βEa E
n1 a
¼ A n ð 1 αmax Þ exp
Δln dT
dα
Δln f ðαÞ Ea RT 2max RT max
¼ ð7:22Þ
ΔT1 ΔT1 R ð7:26Þ
Δln dT
dα
Ea ΔT1 Where, Tmax is the temperature of the maximum
¼ ð7:23Þ rate, β is the heating rate and αmax is the conver-
Δln f ðαÞ R Δln f ðαÞ
sion value at that maximum rate. The maximum
The activation energy can be obtained from inter- reaction rate represents the peak DTG curve.
cept –Ea/R of Equation 7.22, or from the –Ea/R Taking the natural logarithm of equation above
slope of Equation 7.23. and rearranging gives,
www.ebook777.com
0 1
gðαÞ ¼ AeRT t
Ea
β AR n ð 1 α Þn1
ð7:28Þ
A Ea
max
ln 2 ¼ ln@
T max Ea RT max to give:
ð7:27Þ Ea
lngðαÞ ¼ lnA þ lnt ð7:29Þ
The activation energy (Ea) is obtained by RT
plotting the left-hand side of Equation 7.27
This can be rearranged to give:
versus 1/Tmax for a series of runs at different

heating rates. Equation above has been A Ea
lnt ¼ ln ð7:30Þ
generalized to any reaction model f(α) [55]. gð α Þ RT
As the Kissinger method is a model-free
method it does not require any modelistic A plot of –ln t versus 1/T for each degree of
assumptions to calculate Ea. However, it is not conversion, α, gives Ea from the slope for that α
an isoconversional method because it does not regardless of the model.
calculate Ea values at progressive values of α but
rather assumes a constant Ea [38]. Ozawa, Flynn and Wall (OFW) Method
Ozawa [61], and Flynn and Wall [62] indepen-
dently developed an isoconversional calculation
Model-Free/Isoconversional Methods method for nonisothermal data which is com-
monly referred to as the OFW method. Taking
Isoconversional methods do not rely on mathe- the natural logarithm of the nonisothermal rate
matical models such as those in Table 7.6, law (Equation 7.24) and using Doyle’s approxi-
instead they evaluate kinetic parameters at pro- mation [63] for the temperature integral gives the
gressive conversion values of α [56]. These following,
methods require several kinetic curves to per- AEa Ea
form the analysis and have therefore been lnβ ¼ ln 5:331 1:052 ð7:31Þ
RgðαÞ RT
called, “multi-curve” methods [57, 58]. The
term isoconversional derives from the analysis For a fixed degree of conversion, α, the plot of
of several curves at different heating rates at the ln β vs. 1/T, obtained from thermograms
same degree of conversion, α. They determine recorded at several heating rates, should be a
the activation energy for each conversion point straight line whose slope can be used to evaluate
(Ea, α), drawing an isoconversional plot (Ea the activation energy.
vs. α). The terms, “isoconversional” and
“model-free” are sometimes used interchange- Vyazovkin’s Methods
ably, but not all model-free methods are The non-isothermal rate law (Equation 7.24) can
isoconversional. be transformed by substituting x ¼ Ea/RT to give
Model-free methods usually only calculate ð
activation energies (Ea) from the slope of a linear AEa 1 ex
gðαÞ ¼ dx ð7:32Þ
equation while terms such as the frequency factor βR x x2
(A) and model are grouped into the intercept,
and cannot be determined without assuming a or more simply gðαÞ ¼ AE βR pðxÞ where pðxÞ ¼
a
ð 1 x
particular model. Isoconversional methodologies e
dx. The temperature integral p(x) is a
can be used to analyse both isothermal and x x2
non-isothermal data, as described below. function of Ea and temperature, so can also be
represented as pðxÞ ¼ I ðEa ; T Þ.
Standard Isoconversional Method The first Vyazovkin method [64] is based on
This method [59, 60] can be derived by taking the assumption that the reaction model is inde-
the logarithm of the isothermal rate law pendent of heating rate. Therefore
AEa AEa Later, Vyazovkin [68] modified his

gð α Þ ¼ I ð Ea ; T 1 Þ ¼ I ð Ea ; T 2 Þ
β1 R β2 R isoconversional method to analyse kinetics from
AEa isothermal and nonlinear heating rate
¼ ¼ I ð Ea ; T n Þ ð7:33Þ experiments. In this version, the heating rate β
βn R
in the nonisothermal rate law (Equation 7.24)
where β1, β2,. . ., βn, T1, T2, . . ., Tn are different represents the heating function with respect to
heating rates and temperatures respectively, and time. Therefore, the temperature integral in
together with Ea and A they correspond to the Equation 7.24 becomes a time integral, as
same degree of conversion α. shown below:
Equation 7.33 can be reduced to: ðt
Ea
gðαÞ ¼ A exp dt ð7:39Þ
I ðEa ; T 1 Þ I ðEa ; T 2 Þ I ð Ea ; T n Þ 0 RT ðtÞ
¼ ¼ ¼
β1 β2 βn where T(t) is the heating program.
¼σ ð7:34Þ This method allows for use of linear and non-
linear heating rates and also is suitable for
where, is a constant. isothermal analysis.
Therefore: Before analogical procedures should be
employed as used to get to Equation 7.38
β 1 I ð Ea ; T 2 Þ σ the temperature integral I(Ea, T ) needs to be
¼ ¼1 ð7:35Þ
β 2 I ð Ea ; T 1 Þ σ replaced by time integral J(Ea, T(t)).
Vyazovkin [69] presented further modifica-
For n heating rates, the summation of pairs of tion as the advanced isoconversional method
such inverse ratios can be generalised to: (AIC). The principle enhancement in the AIC is
the integration over smaller time intervals, to
X X β i I Ea ; T j
n n
j6¼i ¼ nðn 1Þ ð7:36Þ better account for variations in Ea.
i¼1
β j I ð Ea ; T i Þ Therefore, Equation 7.39 was modified to
give:
or
ðt
! Ea
X X β i I Ea ; T j gðαÞ ¼ A exp dt ð7:40Þ
n n
nð n 1Þ ¼ 0 tΔα RT ðtÞ
j6¼i
i¼1
β j I ð Ea ; T i Þ
where Δα ¼ 1=m and m is the number of
ð7:37Þ
segments into which the integration is divided,
However for experimental data such a differ- typically m ¼ 10–50.
ence might not converge to zero, so separate As in the previous Vyazovkin methods, the
values of Ea could be found for each degree of activation energy Ea at each α is the value that
conversion, α, to find the minimum value of the minimises Ω, as shown in Equation 7.38.
following equation: Some other analytical methods for obtaining

the kinetic parameters are listed in Table 7.7.

X X β I E ; T
These methods are fast and easy to use and they

i a j

n n
j6¼i
¼Ω ð7:38Þ

i¼1
β j I ð Ea ; T i Þ
provide relatively unique and efficient solutions
for most engineering applications. The direct
where Ω is non-zero constant. method (DM) and generalized direct method
It is known that exponential temperature (GDM) proposed by Matala et al. [70] consider
integral of the non-isothermal rate law, Equa- a multiple step of nth order reaction. They both
tion 7.24 has no analytical solution [65, 66] but describe reactions being well separated, where
its approximation can be found in mathematical the DM, generally, is based on the same approach
tables [67]. as Friedman’s [26] single-step reaction model
www.ebook777.com
Table 7.7 Analytical methods for obtaining kinetic parameters

Method Kinetic parameters
Direct method Ea r pk
Ek ¼ N k R T2
αk α pk pk

A ðαk αk1 ÞNk1 Ek
Ak ¼ r pk β Nk exp
αk α pk RT pk
pffiffiffiffiffiffiffiffiffiffiffi
Generalized direct method Ea
E ¼ b 2ab 4ac
2
Where
2
αk αpp
a¼
R2 T 4pp
2
2r pp N k αk αpp 2 αk αpp
b¼
RT 2pp RT 3pp

c ¼ r pp ðN k 1ÞN k r T pp N k αk αpp
2

A ðαk αk1 ÞNk1 Ek
Ak ¼ r p pk β Nk exp
αk α p pk RT pp

Friedman Ea r 2k αk α1k
ln þ N k ln
r 1k αk α2k
Ek ¼ R
1 1

T 2k T 1k
A r ðα α ÞNk1
Ak ¼ β 1k k k1
exp RTE1k ðαk αk1 ÞNk
McGrattan et al. and Lyon et al. Ea eRT 2pk r pk
Ek ¼
αk αk1

A eβr pk Ek
Ak ¼ exp
αk αk1 RT pk
Adapted from Ref. [70]
Note:
Reference points from first derivative of mass data (MLR curve) are: r p , r 1T , r 2T , T p , T pk , T 1k , T 2k , α p , αk , αk1 , α1k , α2k
Values to specify from second derivative of mass data are: r pp , r p pk , r T pp , T pp , αpp , α p pk
and the GDM relies on a higher-order derivative performance based design. A broad spectrum of
of the mass data. McGrattan et al. [24] and Lyon options for simulation is currently available,
et al. [25] presented equations with slightly from simple programmes to various high-level
different derivations but essentially based on the packages implementing advanced methods.
same idea. They both analysed first-order reac- Despite the continuous development of tools for
tion kinetics. Lyon et al. worked on a single-step fire modelling, most of the development has
reaction path but additionally offered an alterna- focused on gas phase processes, particularly
tive method for extracting the reaction zone and CFD models. The current state of the
parameters using the heat release rate from micro- art is still not capable of reliable prediction of the
scale combustion calorimeter (MCC) data [71]. time to ignition of a solid polymer exposed to a
constant applied heat flux. This leaves
predictions of flammability, flame spread, fire
Computational Models of Polymer growth rate, and fire suppression even further
Decomposition behind. Condensed phase fuel production,
which plays a pivotal role in fire growth, has
Computer simulation has become an integral part become an increasingly large research area in
of fire safety engineering and this tendency is the last decade, resulting in the development of
expected to increase with the evolution of numerical models of pyrolysis incorporating
complex arrays of physical and chemical (i.e. char and/or virgin solid) may also be temper-
processes. ature dependent. Pyrolysis gases are assumed to
Generally, one of the first modelling tasks is to be in thermal equilibrium with the solid.
select the appropriate mechanism to represent the Very often it is difficult to establish and quantify
physical phenomena. A computer simulation the kinetic mechanism of solid phase decomposi-
environment attempts to translate real-world tion with certainty, especially for materials with
physical laws into their virtual form. How much complex kinetics (multistep decomposition,
simplification takes place in the translation pro- sequential and parallel reactions etc.) such as
cess will determine the accuracy of the resulting most polymers. Usually, condensed and gas phase
model. However, this process is often subjective, species in the reactions are not characterised exper-
and detailed justifications of the inclusion or imentally. However, Bustamante [88] showed that
exclusion of the different mechanisms are often numerical approaches can be used to derive a
lacking. multiple hypothetical decomposition mechanism
First, a new multistep decomposition method for polyurethane foam. Additional parameters,
was proposed by Ohlemiller [72] where the such as chemical analysis of gas and condensed
kinetic parameters are calculated by numerical phases support the assignment of the actual decom-
iteration from TGA mass loss (ML) and mass position mechanisms better than merely obtaining
loss rate (MLR) measurements. This method was agreement with the MLR curves. The kinetic
used by Rein et al. [28] to simulate the decompo- parameters in multiple reaction mechanisms, due
sition of polyurethane foam and by Matala [73], to their complexity, must be often determined by
Lautenberger [74], Lyon and Stoliarov [75, 76] several curve-fitting algorithms. Evolutionary
to simulate numerically the decomposition of algorithms are most commonly used, including
various materials. A substantial number of genetic algorithms (GA) [27–31, 89] or shuffled
studies [29, 32, 77–80] have demonstrated that complex evolution (SCE) [90, 91] algorithms.
numerical pyrolysis modelling (using e.g. GPyro They operate on a “survival of the fittest” principle.
[30, 81–83], the general pyrolysis model of The algorithm uses a mathematical formulation of
FDS [84], pyrolysis submodel of FireFOAM the experimental pyrolysis data, and the
CFD code [85] or ThermaKin [86, 87]) can optimisation process starts from a random set of
describe degradation of solid materials exposed automated trial solutions and tests their fitness
to external heat flux, and determine the against the experimental (TGA) curve. The suit-
relationships between the fundamental physical ability of the stochastic processes is assessed by its
and chemical properties of polymeric materials fitness value; the better the fitness value, the greater
and their gasification behaviour. Most of the the probability of the trial solution surviving to the
models calculate the ML and MLR of a sample next iteration round. The method has a stochastic
exposed to a heat source in one-dimension. component in the mutation (one or more
Models with a controlled volume approach com- parameters are replaced by a random number),
bine the transfer of thermal energy with Arrhenius crossover (creating new trial solution by uniting
kinetics for the decomposition of the polymer. two older solutions) and selection operations to
They can predict the overall behaviour of a ensure wide exploration and to avoid becoming
pyrolysing polymer by solving a set of mass and trapped in local minima or maxima. These
energy conservation equations using a fully algorithms are very efficient for high-dimensional
implicit scheme. In many cases only three species problems with many parameters, resistant to
are considered, virgin solid, char (where applica- becoming trapped in local optima caused by
ble), and pyrolysis products, and it is assumed that overlapping reactions or even noisy data. However,
the virgin solid decomposes to char and/or gas they are a bit more complicated to use than the
phase products through a single heterogeneous analytical methods; they are heuristic in nature;
nth order Arrhenius-type reaction. Material inefficient for small problems, and they require
properties for the condensed phase species significant amount of computer resources.
www.ebook777.com
Recently, modelling tools have created a each element being characterised by component
more flexible simulation environment with the mass and temperature. Additionally, the model
possibility of adding any physical effect to the describes the transport of gaseous products
model. All the relevant physical aspects of through the condensed phase and follows changes
particular design may be included. Adaptability in the volume of the bulk material.
of modelling platform allows customised For thermally thick solids (typically,
solutions to be developed, applicable to unique thicknesses above 15 mm [97]) the thermal inertia,
circumstances. This is very often required for the kρc, the product of thermal conductivity (k), den-
complicated process of polymer decomposition. sity (ρ), and specific heat (c), of a material governs
Using tools like parameterized geometry, interac- its ignition and flame spread properties. This
tive meshing, and custom solver sequences, determines the rate of rise in surface temperature
they can quickly adapt to different, complex and consequently, the time to ignition [98]. The
requirements. With this kind of all-inclusive time to ignition (tig) of a thermally thick solid
modelling environment, increasing the prospect exposed to a constant net heat flux QR ¼ Qext –
of building models with real-world precision. CHF, where Qext is the external heat flux from fire
The growth of complexity in the models has been or radiant heater and CHF is the critical heat flux
justified by the implicit assumption that models for ignition, has been expressed in Equation 7.41.
with a higher number of mechanisms should be 2
more accurate. However, as direct consequence, π T ig T 0
tig ¼ kρc ð7:41Þ
Q_ R_
4 2
the number of parameters required to perform a
simulation has also increased significantly. It is
important to control the accumulation of uncer- where Tig and T0 are the ignition and ambient
tainty in the input parameters in order to prevent temperatures, respectively. The time to ignition
the model’s output to extend beyond the limits of of a thermally thin solid exposed to a constant net
reliable prediction [92]. The global error induced heat flux has also been expressed in Equation 7.42.
by the uncertainty of a large number of parameters
may exceed the improvements obtained with the T ig T 0
tig ¼ ρcτ ð7:42Þ
incorporation of complex mechanisms; thus there Q_ _
R
remains an optimum level of model complexity.
Where τ refers to material thickness.
Equations 7.41 and 7.42 follow from the concept
Computational Modelling of Polymer of a constant ignition temperature Tign and
Combustion temperature-independent thermal inertia. Once
ignition has occurred and a flame is established
The same pyrolysis models may also be used to on the surface, the net heat flux becomes QR ¼
calculate the mass loss rate of a one-dimensional Qext + Qflame – CHFb, where Qflame is the addi-
sample of solid fuel exposed to a uniform heat tional heat flux supplied by the flame and CHFb
flux. ThermaKin is an example of such a model,
σT 4b is the critical heat flux for burning in terms
which has been effectively utilised as a practical of the surface burning temperature Tb and the
tool for the prediction and/or extrapolation of Boltzmann radiation consant σ. It has been
the results of fire calorimetry experiments shown that Tb
Tp where Tp is the pyrolysis
[3, 93–96]. The model, which combines the temperature measured in laboratory thermal anal-
absorption and transfer of thermal energy with ysis experiments using small samples and con-
Arrhenius kinetics for the decomposition of the stant heating rates [22]. Thus, polymers with high
polymer, predicts the overall behaviour of a pyrolysis temperatures reradiate more of the inci-
pyrolysing object through mass and energy dent heat flux from the heater and flame back to
conservation equations. These equations are the surroundings, and the net heat flux that drives
formulated in terms of rectangular finite elements, the burning process is reduced accordingly.
Re
f le
ct ion
io iss
n Em
Infrared Radiation – Absorption,
Absorption reflection, emission
Thermal
Thermal Inertia
Inertia kρc
kρc Heat transfer through solid – thermal inertia kρC
Heat losses
Endothermic thermal decomposition and gasification -

First Order Arrhenius kinetics
Gas transport leading to critical mass flux
radiation
convection
Ignition
Convective and Radiative Heat Transfer
Steady Burning – Rate of fuel pyrolysis controlled by

radiation from flame
Fig. 7.22 Schematic of processes occurring in the cone calorimeter, as modelled by ThermaKin
The processes modelled by ThermaKin have applications and thermal decomposition is

been summarised in Fig. 7.22. For this study, provided below.
radiant heat from above the sample is absorbed,
emitted or reflected, and the condensed phase
heat transfer process is modelled through the Thermoplastics
solid. The resulting temperature increases drives
endothermic decomposition processes, leading to Polyethylene
the gasification of volatile fuel components. Three types of polyethylene are generally avail-
When a critical mass flux for ignition is reached, able, high density polyethylene (HDPE), linear
ignition will occur, and the incident radiant flux low density polyethylene (LLDPE), and low den-
is augmented by radiation from the flame. There- sity polyethylene (LDPE). HDPE is composed of
after, quasi-steady state conditions pertain, until structurally regular chains with very few branch
the sample is so thin that it has no more capacity points (less than 1 % of carbon atoms) which pack
to absorb heat, and the rate of pyrolysis increases. efficiently, resulting in a highly crystalline mate-
rial with a correspondingly higher density. The
polymer is used to manufacture bottles, crates and
Behaviour of Individual Polymers pipes. LDPE is approximately 6 % branched with
a much lower crystallinity and lower density,
In general, each polymer decomposes in an indi- with good film forming properties, so that its
vidual way, and most generalisations are of lim- largest application is as film for packaging, and
ited value. A brief description of the distinctive cable coverings. LLDPE fills the gap between the
features of selected individual polymers, their two materials and is generally prepared as a
www.ebook777.com
copolymer of ethylene with 8–10 % of an α-olefin The kinetics of polyethylene pyrolysis has
such as but-1-ene, or oct-1-ene. This produces a been studied frequently [99]. It has been shown
chain with a controlled number of similar short that random chain scission decomposition of
chain branches, with densities intermediate polyethylene does not rigorously follow first
between HDPE and LDPE. Initiation of decom- order kinetics. The reaction order, in both
position of each of the different types of polyeth- dynamic and isothermal decomposition of a
ylene occurs at the branching points, as observed high density polyethylene, was determined
by Madorsky (section “Chain Branching”). Thus [100] to be 0.55. The thermal decomposition of
LDPE, having a greater number of branching polyethylene with different characteristics, such
points and other abnormalities in its structure, as density or melt flow index, showed no signifi-
decomposes at a slightly lower temperature than cant differences, except in the presence of an
HDPE, although the general mechanism is the added catalyst [101].
same for each type of PE. The mechanism of The thermal decomposition of polyethylene
thermal decomposition of polyethylene follows comprises several steps, presented schematically
random chain scission, resulting in a mixture of in Fig. 7.23. The temperature increase causes
alkanes, alkenes, and dienes. The elimination of random scission of the polymer backbone
ethylene (the monomer) is not favoured energeti- resulting in the formation and release of a large
cally and pyrolysis of polyethylene does not amount of small molecules and radicals.
occur by end chain scission. Formation of cyclic hydrocarbons and aromatic
130
H3C CH3
Temperature (°C)
Polymer
Condensed phase
chain 180
230
Biradicals
280
330
Interface
+ O2
supplied
Oligoradicals 380
Heat
H2C CH2 CH3 CH3 CH2 CH2

CH3
H C CH
3 3
H2C CH CH C CH C CH2HC
330
H C
3 CH2 CH2 O CH
CH CH2 H C
3
CH2 CH2 CH2 CH2 CH
3
C CH
CH2
Gas phase
H3 C CH
280
3
Volatile
species 230
CH2 CH2 H2C CH2

CH CH
H3 C C CH C
180
CH2 H2C C CH CH2
HC CH O
CH CH2 CH3 CH3
CH CH H2C CH2
130
Fig. 7.23 LDPE decomposition processes [103]

rings occurs at higher temperature by Diels-Alder carbon. The CO also comes mainly from the C
cyclisation reactions of alkenes. The aromatic (1) [methylene] carbon (~80 %). This is in con-
content of the pyrolysate increases with tempera- trast to the solid-phase oxidation products, which
ture. The thermal decomposition in air (shown have been found (by C-13 NMR on these same
on the right hand side of Fig. 7.23) also labelled PP materials) to originate predominantly
produces partially oxygenated products such (80–85 %) from oxidation at the C(2) [tertiary]
as aldehydes, ketones and major combustion carbon. Thus oxygen preferentially attacks the
products such as carbon monoxide, carbon main carbon backbone, resulting in release of
dioxide, water etc.) [102]. volatile fuel at lower temperatures than from
anaerobic decomposition.
Polypropylene A further investigation [105] of the volatile
Three types of polypropylene (PP) can be products provided a detailed description of the
synthesised; isotactic, syndiotactic and atactic, organic carbon compounds found in the vapour
although only the isotactic form is used commer- phase above the decomposing polypropylene.
cially (the difference lies in the stereochemical Approximately half of the volatile products
positioning of the methyl group relative to the identified have a methyl ketone at one end of
main chain). Isotactic PP is highly crystalline, the molecule. In all of the methyl ketones, the
with a sharp melting point at 443 K. The thermal carbonyl carbon originates from the C(2) [ter-
decomposition of polypropylene in an inert tiary] position within the PP molecule, and the
atmosphere also proceeds via random chain scis- terminal methyl group originates from the C
sion, to generate between 10 % and 25 % (3) [methyl] position. This demonstrates the sus-
propene, in addition to other fragments, typically ceptibility to thermal decomposition of the C
containing multiples of three carbon atoms (2) branching point (c.f. LDPE and HDPE).
(e.g. 6, 9, 12 up to 18 carbon atoms). The thermal
decomposition of PP is particularly sensitive to Polystyrene PS
the presence of even small (<1 %) quantities of Polystyrene and has good physical properties
oxygen, causing earlier fuel release, and this allowing for easy processing by injection mould-
aspect has been studied extensively. In a detailed ing, transparency, good electrical insulating
study [104] using a set of three isotactic polypro- characteristics etc. It is typically found in atactic
pylene samples that had been individually and amorphous forms, for example in a CD or
labelled with carbon-13 at each of the three DVD “jewel case”, as well as in expanded form
positions in the monomer unit (Fig. 7.24), as EPS packaging, or EPS and extruded PS
experiments were conducted to determine the (XPS), for thermal insulation. It is entirely amor-
original position of the carbon forming carbon phous, with no defined melting point, above Tg it
monoxide and carbon dioxide during thermal shows the classic five regions of viscoelastic
decomposition. By GC-MS analysis, it was behaviour. The rate of decomposition is
found that 2/3 of the CO2 was derived from the influenced by the method of synthesis of polysty-
C(1) [methylene] carbon, while the remaining rene—anionically initiated polystyrene is more
1/3 comes from the C(2) [tertiary] carbon, with thermally stable than that produced by thermal
none coming from the C(3) [methyl group] initiation or with free radical initiators [106],
presumably because it has less structural
Fig. 7.24 Polypropylene H2 irregularities in its backbone.
repeat unit showing C 1C Thermal analysis shows that polystyrene
(1) methylene, C(2) tertiary
degrades thermally in a single step, and that
and C(3) methyl group 2C
carbon positions monomeric styrene (~40 %) is the principal vola-
H
tile, together with a much smaller amount of ben-
3CH zene and toluene. However there is also a
3
significant amount of dimer, trimer, tetramer and
www.ebook777.com
pentamer. These oligomers are formed in intramo- In both cases, the decomposition follows a
lecular transfer reactions, in direct competition free radical mechanism, stabilised by the four
with the monomer-producing depolymerisation substituents on the α-carbon atom. This allows
process. A significant feature of the thermal deg- the unpaired electron to reside on the α-carbon
radation of polystyrene is the rapid initial decrease atom long enough for the double bond to reform,
in molecular weight, which is followed by a more releasing methyl methacrylate monomer. The
gradual fall as volatilisation exceeds 10 % [107]. monomer the yield has been quantified as
Oxygen plays a very important role in the 59–95 % depending on the decomposition
degradation of PS. The degradation mechanism conditions [4]. During the decomposition of
involves depropagation, thermo-oxidative high molecular weight PMMA, the monomer is
products include benzaldehyde, benzoic acid, evolved from the decomposing solid, avoiding
phenol, and benzyl alcohol. the complexity of sample dripping. The combi-
nation of a single gas phase fuel molecule, and
Polymethylmethacrylate (PMMA) avoiding dripping, are possible reasons for this
PMMA is probably the polymer with most widely polymer’s popularity in fire science. Its atypical
studied the thermal decomposition, and the clear decomposition may be one of the reasons fire
favourite amongst fire scientists as a model fuel. scientists have not made the necessary progress
Like polystyrene, PMMA is an amorphous on condensed phase decomposition and pyrolysis
polymer with no defined crystalline melting of polymers.
point. Low molecular mass “extrudable” PMMA
softens on heating to form a viscous liquid, while Polycarbonates
higher molecular mass “cast” PMMA Polycarbonates are a class of polymers with good
decomposes prior to softening. Almost uniquely mechanical properties, containing the repeat
amongst the common polymers, lower molecular group (OC(O)O)n in their backbone.
weight PMMA decomposes by end chain scission They are crystal clear, highly impact resistant,
(Fig. 7.25), while higher molecular weight amorphous engineering plastics. In addition to
PMMA decomposes by a combination of end CD and DVD discs, they are used in the construc-
and random chain scission (Fig. 7.26) [108]. tion industry and car headlights for tough plastic
CH3
C
CH3 CH2
C
CH3 CH3 CH3 CH3 CH3 CH3 CH3 CH3 CH3 CH3 CH3 CH3 C O
CH2 O
C C C C C C C C C C C C C
C C C C C C C C C C C O CH3
H2 H2 H2 H2 H2 H2 H2 H2 H2 H2 H2 O
C C C C C C C C C C C C
O O O O O O O O O O O O
O O O O O O O O O O O O CH3
CH3 CH3 CH3 CH3 CH3 CH3 CH3 CH3 CH3 CH3 CH3 CH3
Fig. 7.25 End chain scission of PMMA
CH3 CH3 CH3 CH3 CH3 CH3 CH3

CH3 CH3 CH3 CH3 CH3
C C C C C C C
C C C C C
C C C C C C
C C C C H2 C H2 H2 H2 H2 H2 H2
H2 H2 H2 H2 C C C C C C C
C C C C C
O O O O O O O
O O O O O
O O O O O O O
O O O O O
CH3 CH3 CH3 CH3 CH3 CH3 CH3

CH3 CH3 CH3 CH3 CH3
Fig. 7.26 Random chain scission of PMMA

glazing. Polycarbonates are relatively resistant to Aliphatic Polyamides

thermal decomposition, which may take place Polyamides are the most widely used engineering
either by an anionic ester interchange route, or by plastics. In addition to fabrics and fibres,
free radical chain scission [109]. In either case, polyamides are widely used in motor vehicles.
carbon dioxide is the main decomposition product They are semicrystalline polymers with reason-
(~35 %) and the resultant organic material forms a able thermal stability but high water absorption,
char. Below 675 K the decomposition products frequently used with glass fibre reinforcement for
also include cyclic dimers and cross-linked spe- increased stiffness at elevated temperatures. The
cies. The decomposition mechanism depends on C—N bonds are the weakest in any polyamide,
the temperature and heating rate. Above 775 K and these tend to break first, followed by the C
most functionality of the original polymer is lost (O)—CH2 bonds. In an inert atmosphere, the
with the formation of aromatic compounds, lead- decomposition products of polyamide 6 (NH
ing to extensive char formation above 800 K. The (CH2)5CO)n at 875 K is predominantly the
maximum rate of decomposition of polycarbonate cyclic monomer (caprolactam) 73 %, with small
in nitrogen is around 745 K. The formation of amounts of CO2, water and other minor decom-
carbon dioxide in the thermal decomposition of position products. The decomposition products
polycarbonate [110] can be explained by reactions from polyamides made from a diamine and a
of the type shown in Fig. 7.27. dicarboxylic acid, such as polyamide 6.6, the
The ether typically continues the decomposi- most widely used polyamide, include a mixture
tion process and for this reason there are of hydrocarbons, nitriles, amides and acids
similarities between the pyrolysis products of [111]. Decomposition at 875 K in an inert atmo-
aromatic ethers and aromatic carbonates (see sphere produced cyclopentane, 39 %, carbon
section “Polyetheretherketone (PEEK)”) dioxide, 26 %, water, 6 %, and surprisingly cap-
rolactam, 6 % (believed to be formed by the
O radical bite-back mechanism [112]) and hexane
dinitrile 2 %.
O C O
Polyesters, Polyethylene Terephthalate

and Polybutylene Terephthalate
heat Polyesters are commonly used both as
thermoplastics in various high volume automo-
HO tive, electrical, and other engineering
C O
applications and as textile fibres. They are widely
used as structural composites in glass reinforced
polyesters (GRP) or “fibreglass”. Commercially
O the most important are polyethylene terephthal-
ate (PET) and polybutylene terephthalate (PBT),
although recent demands for sustainability have
favoured the development of polylactic acid
heat - CO2
(PLA), derived from biologically sourced
lactic acid.
Thermal decomposition of polyesters usually
starts with the scission of the polymer chain
through a six-membered ring transition stage
O
[113] (Fig. 7.28). It is believed that this scission
is mostly heterolytic and not a free-radical pro-
Fig. 7.27 Loss of CO2 and ether formation in polycar- cess, although this is still under discussion in the
bonate decomposition literature. Secondary reactions mostly involve
www.ebook777.com
O O O O O O
C H2 C C H2 C C H2 C
C C C
O O O O O O
C C C
H2 H2 H2
O H O O O O O
H
C C C C H2 C C C
H2
C C
O O O O O
O C C C
H2 H2 H2
O H O O O O O
H
HC C C C C C C
H2 H2
C C
O O O O O
O C C C
H2 H2 H2
OH O O O O O
H
C C C C H2 C C C
H2
O C C
H2C O O O O O
C C
H2 H2
Fig. 7.28 Ring formation in the decomposition of PET
the formation of vinyl ester chain ends (CH2 ¼ such as wool or cotton. It is also used
CH-O-C(O)-Ar-) which lead to polyene extensively as a copolymer, for example in
structures, and crosslinking, but also evolution acrylonitrile-butadiene-styrene (ABS)
of light volatile products. In air, polyesters formulations. Depending on conditions, polyac-
undergo crosslinking, a relatively minor process rylonitrile either decomposes by cyclisation, or
in an inert atmosphere. chain scission resulting in volatile formation, or
Pyrolysis of PET showed evolution of benzoic chain stripping with the release of hydrogen
acid (43 %), acetaldehyde (16 %), CO2 (10 %), cyanide [115]. The resulting structure may
and the vinyl and divinyl esters of benzoic acid either decompose further producing volatiles,
(7 and 4 % respectively). Rapid volatilization such as ethyne by random chain scission, and
of light fragments makes polyesters easily ignit- rearrangements such as cyclisation, resulting in
able polymers. Despite the presence of benzene ethyne, ethane, benzene and higher aromatics,
groups in the main polymer chain, thermoplastic or, under different (usually slower) conditions
polyesters show very limited tendency to undergo rearrangements such that the polyene
char, but instead, aromatic-containing polymer may cross-link, ultimately resulting in the for-
fragments volatilize and feed the flame [114]. mation of a protective char. This may shield the
remaining polymer from radiative heating,
Polyacrylonitrile (PAN) while acting as a barrier to fuel and oxygen.
Polacrylonitrile (PAN) is a significant bulk Indeed, under controlled conditions of slow
polymer, extensively employed in the textile heating this process is used to manufacture car-
industry in fibre form, which is significantly bon fibre from polyacrylonitrile fibre, and has
more flammable than its natural counterparts, been studied extensively [116].
Horrocks [117] identified three pairs of

competing volatilisation and cyclisation/
A
carbonisation decomposition reactions in pure C C C C
PAN which depended on atmosphere and heating N N N N
X
rate. At the slowest heating rates in air,
cyclisation (~625 K) was followed by the first
carbonisation processes, around 725 K, followed
by volatilisation at around 825 K. At higher
heating rates, chain scission leading to B
volatilisation, at around 625 K competes with X N N N N
cyclisation or carbonisation processes. The resi-
due then undergoes the first carbonisation stage,
and the final volatilisation stage. It is this
volatilisation stage which is responsible for the O O
high flammability of acrylonitrile polymers in
general.
In the production of carbon fibres from PAN, C
the initial stabilisation stage is conducted under
carefully controlled conditions in the presence of X N N N NH2
H H
oxygen, where a strongly exothermic process
occurs around 473 K. In DSC studies, a sharp, Fig. 7.29 Stages in the decomposition of PAN
narrow exothermic peak is observed under inert
conditions in contrast to a broad exotherm, Halogenated Polymers
some 50 K higher, associated with oxidising
atmospheres [118, 119]. The atmosphere is then Polyvinyl Chloride (PVC)
changed for nitrogen in the subsequent thermal PVC is ranked third behind polyethylene and
decomposition stages, in order not to oxidise the polypropylene in terms of worldwide polymer
carbon fibres. consumption because it has a good chemical
During the stabilisation stage, the following resistance and can be used in many different
processes, shown in Fig. 7.29 are believed to applications, such as cables, pipes, furniture,
occur [120]. Polyacrylonitrile (A), particularly etc. As the pure polymer is rather brittle it does
as a result of nucleophilic attack (X), undergoes not enjoy widespread use. The popularity of PVC
cyclisation of nitrile groups leading to the lies in its unmatched ability to form a stable, dry,
formation of hydronaphthiridine rings [121] flexible and easily processed material when
(B). The hydronaphthiridine rings may undergo plasticized. However, this involves adding a sig-
oxidative dehydrogenation leading to acridone nificant amount, typically 30 %, of flammable
and other structures [122] (C) increasing the organic liquid such as di-iso-octyl phthalate as a
aromaticity [123]. plasticiser, to an otherwise low flammability
The result is a ladder polymer containing a polymer.
mixture of acridone, pyridine, hydronaphiridine At elevated temperatures PVC undergoes a
and other structures. In the early stages of dehydrochlorination reaction to release hydrogen
decomposition this ladder polymer alternates chloride by β-elimination to form a conjugated
with unchanged PAN [124]. Figure 7.30 shows polyene [125]. β-elimination is catalysed by
typical ratios of these different components acids, such as HCl, and hence the decomposition
obtained under certain conditions, which started is auto catalytic and very rapid. As a conse-
to form the ladder polymer at 180 C, as shown quence PVC materials usually contain stabilisers
by FTIR [121]. which absorb hydrogen chloride to allow the
www.ebook777.com
Fig. 7.30 Typical ladder O O OH

polymer structure [116]
* ** ** ** co
*
C
* N n
** N n
** N n
** N N
N
H
acridone pyridine hydronaphthiridine others
40% 30% 20% 10%
CI CI CI CI CI CI CI CI CI CI CI
H H H
C C C C C C C C C C C CH3
H3C H C H C H C C H C H C C H C H C C H C
H2 H2 H2 H2 H2 H2 H2 H2 H2 H2 H2
473 K
H CI H CI H CI H CI H CI H CI H CI H CI H CI H CI H CI
H H H H H H H H H H H
C C C C C C C C C C C CH3
H3C C C C C C C C C C C C
H H H H H H H H H H H
rapid decomposition slow decomposition

673 K H H H2 H
C C C C
HC CH HC C CH
H
+
CH
HC
+ 823 K
HC
CH
HC CH
C C C C
H H H H
Formation of char precursors

Volatilisation of unsaturated and aromtic molecules
Fig. 7.31 PVC decomposition
extrusion and processing without decomposition. 925 K, 75–90 % of chlorine has been recovered
At higher temperatures the polyene undergoes as HCl. As HCl is a reasonably efficient free
further rearrangements and product elimination radical trap, it removes high energy H·, ·O · and ·
to produce a complex pattern of hydrocarbons OH radicals from the flame zone, which can
with aromatic materials predominating quench gas phase flaming. In some formulations,
(Fig. 7.31). At temperatures between 475 and some chlorine may remain in the residue. For
575 K, 80–95 % of the chlorine content of rigid example calcium carbonate, a common filler in
PVC is released as HCl, 70 % within one minute PVC, will react with HCl to produce non-volatile
at 575 K [126]. Under combustion conditions at calcium chloride, while releasing non-fuel CO2
from the carbonate. The chorine content of the ¼ CF2) at 800 K, although at 873–973 K larger
residue depends on the decomposition fragments predominate (C2F4, 16 %; C3F6, 26 %;
conditions. When a plasticised PVC containing C4H8 58 %) [132]. In fires PTFE can evolve
calcium carbonate was decomposed under carbonyl fluoride (COF2), which hydrolyses to
non-flaming conditions 58 % of chlorine was produce toxic carbon monoxide and hydrogen
released as HCl. Under combustion conditions fluoride. Under certain conditions, PTFE was
at 925 K, the recovery decreased to 34 %, but at found to produce supertoxic particulates on ther-
1275 K, calcium carbonate decomposes to cal- mal decomposition [133].
cium oxide and gas phase chloride recovery
increased to 43 % [127].
A small proportion of the chlorine is released Elastomers
as other chorine containing gas or vapour species
and a number of chlorine containing species have Silicone Polymers
been identified from large-scale PVC fires, includ- Silicone polymers are typically heat resistant
ing mono- and dichlorobenzenes and other chloro- elastomers, used as sealants, adhesives,
aromatic and chloro-aliphatic hydrocarbons lubricants, and for electrical insulation. Somewhat
[128]. Evidence exists to show that, depending surprisingly, in North America and in aircraft,
on the fire situation, as much as 20 % of the silicone foams have also been used, with some
chlorine may exist in an organic form [129]. controversy, as firestops [134]. Polysiloxanes are
PVC has a low heat of combustion, and burns inorganic–organic polymers that have a backbone
with a low heat release rate, because the halogen formed from alternating oxygen and silicon atoms
atoms in the structure release HCl, almost 60 % of the form (OSi(R2))n. Various substituents
of its mass, which then inhibits the conversion of (as R) are added to the silicon atom. Polydimethyl
CO to CO2. This is the major heat release step in siloxane, for example, has two methyl groups
polymer combustion. When hydrogen chloride attached to the silicon atom. In addition to the
gas (HCl) comes off on heating, this causes a linear backbone, polysiloxanes can undergo
double bond to form between alternate carbons, crosslinking of the macromolecular chains
strengthening the chain. As this residue gets hot- through oxygen bridges, giving them rubber-like
ter, the chain either cross-links, as double bonds properties. Low molecular weight polysiloxanes
open and attachments to neighbouring chains are sold as silicone fluids with many practical
occur, eventually leading to char formation, or applications. Higher molecular weight siloxanes
they break down and cyclise to form volatile are cross-linked to make silicone rubbers.
aromatic hydrocarbons. If the aromatic The pyrolysate of polydimethyl siloxane
hydrocarbons only contain a few rings, they consists of various cyclic siloxanes, such as
may form into soot particles, if they are large or hexamethylcyclotrisiloxane, octamethylcyclotri-
cross-linked the residue may form a stable pro- siloxane, decamethylcyclotrisiloxane etc. The
tective char layer [130]. formation of these compounds is exemplified
below in Fig. 7.32 for the formation of
Polytetrafluoroethylene (PTFE) hexamethylcyclotrisiloxane [4].
PTFE has many applications due to its excellent
resistant to thermal decomposition, electrical Polyisoprenes and Other Rubbers
insulation and mechanical toughness. Despite An elastomer is a polymer with elastic properties
being polymerised by a free radical process, the (section “Physical Properties”) whereas a rubber
polymer has very little chain branching, and is generally refers to a cross-linked compound
highly crystalline. The thermal decomposition, comprising elastomer and additives (Fig. 7.33).
by end chain scission, starts at 713 K, reaching Polyisoprene was the world’s first commer-
a maximum at 813 K [131]. The predominant cially produced elastomer, originally derived
species in the vapour phase is the monomer (CF2 from the latex of the rubber tree Hevea
www.ebook777.com
Fig. 7.32 Example CH3 CH3 CH3 CH3

decomposition of
polydimethyl siloxane to
hexamethylcyclotrisil- Si Si Si Si
oxane O O O OH
CH3 CH3 CH3 CH3
heat
CH3 H3C CH3 CH3

O
Si Si Si
CH3
O
CH3 HO O
Si
H3C CH3
heat
CH3 H3C CH3

O
Si Si
Si + H3C CH3
OH
O O
CH3
Si
H3C CH3
CH3 alternative polymers have higher damping and

dynamic stiffness [135], typically leading to
CH2 C = CH CH2
n higher noise and greater transmission of vibration.
The thermal decomposition of polyisoprenes
Fig. 7.33 Repeat unit polyisoprene has been reviewed elsewhere [136]. NR and IR
raw polymer have identical TGA curves with a
Brasilienis, although a synthetic equivalent was single large maximum mass loss peak [137] at
developed on an industrial scale during World 646 K in N2 with less than 0.5 % residue at
War 2. The main disadvantages of natural 773 K. When compounded with carbon black
(NR) and synthetic isoprene (IR) rubbers are and sulphur into a typical commercial formula-
their limited resistance to high temperatures and tion a difference is noted; the NR mix retains a
oils, which has led to the use of other elastomers single peak temperature for mass loss, but the IR
such as chloroprene (CR), acrylonitrile butadiene compound gains a second mass loss peak [138] at
rubber (NBR) and ethylene propylene diene around 700 K. It is believed that IR polymer may
modified rubber (EPDM). Nearly all of these undergo cyclisation, catalysed by small
quantities of residual titanium tetrachloride used Thermosetting Polymers

in polymerisation.
Polyisoprene decomposes predominantly by Epoxy Resins
random chain scission, initiated at a ‘weak link’ Epoxy resins are tough, thermally stable and
due to a chain defect. This defect may be due to an important class of thermosetting polymeric
residues from polymerisation such as initiator, or material. They are widely used in electronic
some head to head linkages. Common chemical applications, binding carbon fibres together in
species detected from decomposition include iso- metal replacement composites enjoying wide
prene (2-methylbutadiene) and dipentene use in aerospace, automotive and even bicycle
(1-methyl-4-(1-methylethenyl)cyclohexene), for engineering applications, and find use as excel-
which the yield quantities vary considerably, lent adhesives. Typically synthesised from epi-
depending on the temperature and conditions chlorohydrin and bis-phenol A, the pre-polymer
of thermal decomposition. Typically, at contains reactive epoxide groups as part of a
temperatures below 573 K, the yields of isoprene polymeric diglycidylether (Fig. 7.35), terminated
and dipentene may be in the region of 5 and 15 % with epoxy groups. On the addition of an initiator
[139]. At higher temperatures, however, the iso- or hardener, the epoxy groups react by cross-
prene yield may be significantly higher. A linking, resulting in a three dimensional network
simplified process of depolymerisation at lower of polymer chains.
temperatures has been summarised by Cataldo The thermal stability of epoxy resins, as well
[140], although at higher temperatures more spe- as their flammability, depends on the structure of
cies are generated, including 2,4-dimethyl-2- the monomer, the structure of the curing agent
ethenylcyclohexene (Fig. 7.34). and the crosslink density [141]. In general, the
Overall, polyisoprene decomposes dramati- thermal stabilities of aromatic epoxy resins are
cally to low molecular weight fragments which higher than those of aliphatic ones, even though
vaporise into fuel for combustion, with almost no the crosslink densities of the aromatic networks
char to slow down decomposition. This is consis- may be lower. Usually, thermal decomposition of
tent with the observation that polyisoprenes burn any epoxy resin starts from the dehydration of the
readily with high heat release and with high secondary alcohol (R1C(R2)–OH), leading to the
yields of small aromatic species that will create formation of vinylene ethers (the first step in
high levels of smoke. Fig. 7.36). Upon further decomposition, aliphatic
+ II
O O
[O2]
+
I
+ +
III
Fig. 7.34 Decomposition scheme for polyisoprene (I—polyisoprene; II—dipentene; III—isoprene) (After Ref. [142])
www.ebook777.com
CH3
O CI
H2C C C + HO C OH
H H2
CH3
CH3 CH3 H2 H
C C CH2
O OH
O C O O C O
O
H2C C C C C C
H H2 CH3 H2 H H2 CH3
Fig. 7.35 Synthesis of an epoxy resin, showing basic repeat unit, before reaction of terminal epoxy groups
OH
O O
C C C
H2 H H2
− H 2O
O O
C C C
H2 H H
+ O2 (g) O
OH H 3C CH3
+
OH
C O
CH
HC H 2C
+
CH3
Hydrocarbons
Fig. 7.36 Decomposition of epoxy resins, showing loss of water, to form a vinylene ether, followed by Claisen
rearrangement to form a tri-substituted benzene char precursor
chain ends produce light combustible gases, allyl Fig. 7.36) which changes the paraphenylene
alcohol, acetone and various hydrocarbons. group to a 1,2,4-trisubstituted benzene with
Alternatively, the allylic ethers or amides, increased thermal stability [142]. This structure
formed after losing water, can undergo the is partially responsible for further crosslinking
Claisen rearrangement (the second step in and charring of epoxy resins.
Polyurethanes Variation of R1, R2 and the number of functional

Polyurethanes (PU) represent an important class groups in the polyol and isocyante leads to a large
of polymers and elastomers with a wide variety variety of PU polymers. In flexible PU foams, the
of structures and properties. As solids, they are polymer decomposes to form a pool, isocyanate
used for elastomeric wheels (revolutionising tends to be released first from the pool, followed
skateboards and rollerblades in the 1980s), sus- by the polyol. In rigid PU foams the greater cross-
pension components, electrical potting linking can lead to charring of the foam.
compounds, hoses, high performance adhesives, During combustion, particularly in under
surface coatings and sealants. As flexible foams ventilated conditions, polyurethanes produce
(PUF) they are very widely used in upholstered toxicologically significant quantities of hydrogen
furnishing and carpet underlay. As rigid foams cyanide [143]. This is believed to be a major
(PUR), they are widely used in thermal insulation contributor to fire deaths [144]. The low thermal
products in buildings etc. Although included in inertia of PU foams, their resultant flammability
the section on thermosets, a class of thermoplas- and their widespread use, means that that they
tic polyurethanes (TPU) also exist. make a particularly important contribution to fire
Polyurethanes are formed by reaction of an risk and hazard.
isocyanate (O ¼ C ¼ NR1N ¼ C ¼ O) with
a polyol (HOR2OH) to form a polyurethane Phenolic Resins
(C(O)NHR1NHC(O)R2O)n. How- Some of the oldest synthetic polymers were phe-
ever use of isocyantes or polyols with more than nolic resins, originally sold as Bakelite. Today
two functional groups, such as R1(N ¼ C ¼ O)3 their applications include low flammability rigid
or R2(OH)3 will lead to a complex cross-linked components of railway seats, snooker balls, cir-
structure. For example, in such formulations, the cuit boards, adhesives (particularly for plywood
presence of traces of water is one way to promote and other wood-based composites) and as foam
cross-linking between two unreacted isocyanate thermal insulation of lower flammability than
groups. XPS, EPS or PUR. Phenolic resins are formed
by reaction of a phenol with formaldehyde. They
2R N ¼ C ¼ O þ H2 O contain both aromatic rings and aliphatic carbon
! R NH CðOÞ NH R þ CO2 in the polymer backbone, and are fully cross-
linked thermosetting polymers.
The reaction can continue between further Ten percent of mass loss is observed between
isocyanate groups, and -NH groups by the forma- 373 and 423 K, and a further 30 % between
tion of biuret linkages (Fig. 7.37). 673 K and 1173 K in nitrogen (Fig. 7.38),
The decomposition of polyurethanes is depen- through loss of water. The main volatile products
dent on the chemistry of both the polyol and the of the thermal decomposition of phenolic resins
isocyanate and the polymerisation process. found in a study using thermogravimetric
Fig. 7.37 Formation O O

of a biuret linkage
in polyurethanes 1 1 1 1
R C R R C R
N N N N
H H H
C O
+
HN
2
R N C O
2
R
www.ebook777.com
CH2
CH
HO
OH
HC
+ H2C + 4 H2O
CH2
CH
HO
OH
HC
H2C
Fig. 7.38 Cross linking of phenolic resins between 373 K and 423 K accounting for 10 % mass loss
OH
OH
H3C CH3
H 3C
CH3 cresol
CH3OH OH OH OH OH
HOH2C CH2 CH2 CH2
CH2 OH
OH
H3C OH phenol
H3C CH3
CH3
xylenol
Fig. 7.39 Formation of phenol and its methyl derivatives
analysis and Fourier transform infrared spectros- Polymers with High Thermal Stability
copy (TGA-FTIR) at 30 K per minute include
tars, various phenols and cresols, water, carbon Polyetheretherketone (PEEK)
dioxide, methane and carbon monoxide [145] Poly(oxy-1,4-phenyleneoxy-1,4-phenylenecarbo-
(Fig. 7.39). The char residue was measured as nyl-1,4-phenylene), (PEEK), is a semi-
63 % at 975 K and 60 % at 1275 K. crystalline polymer with excellent mechanical,
* O O O
C C
C n
* N R
O N n
O C C
Fig. 7.40 Repeat unit of polyetheretherketone O O
Fig. 7.41 Structure of a repeat unit of polyimide (PI)

chemical and thermal properties which permits
for its use in a variety of industries and partic- glass and metal products. PI contains the
ularly as an ideal metal replacement with sequence CO – NR – CO as part of a ring struc-
exceptionally high thermal stability [146]. ture along the backbone, shown in Fig. 7.41, the
The structure of PEEK is shown in Fig. 7.40. presence of which gives the polymer its high
The stable aromatic backbone, which makes up temperature properties [152].
the bulk of the monomer unit, is attributed to PI has high oxidative stability and therefore
giving the polymer its excellent thermal can withstand temperatures of up to 533 K
properties [147, 148]. PEEK has a continuous [154]. They are resistant to weak acids and
use temperature of 533 K and a melting point of organic solvents but not bases. PI also resists
616 K [149]. The onset of thermal decomposi- ionising radiation and has good electrical
tion occurs between 848 and 853 K. The thermal properties, however, exposure to water or steam
decomposition of PEEK is different in both oxy- above 100 C may cause parts to crack rendering
gen and nitrogen environments, however, both poor hydrolysis resistance [154]. Common uses
show a two step decomposition process. In the for PI are appliance bearings, seals and electronic
first step around 50 % of the mass is lost, components and products.
(~873 K in nitrogen and ~823 K in air), random
chain scission of the ketone and to a lesser Polyetherimide (PEI)
extent, the ether bonds is believed to be the Polyetherimide (PEI) is an amorphous thermo-
main mechanism [150]. The second step is plastic with high temperature resistance, rigidity
slower, with over 35 % of the residue remaining and impact strength. PEI has similar properties to
at 1273 K in nitrogen; in air mass loss is com- PI and although it will burn, both have self-
plete at 1000 K. extinguishing properties [154]. PEI also has aro-
The main decomposition products contain matic rings in its backbone (Fig. 7.42)
lower molecular weight volatiles as well as ben- PEI has low smoke emission and is resistant to
zene and methylbenzene. As the randomness of alcohols, acids and hydrocarbon solvents how-
the main chain scission increases, other volatile ever, dissolves in partially halogenated solvents.
products are formed such as diphenylether and PEI is sold under the trade name of Ultem™ and
CO and CO2 at 923 K and dibenzofuran, biphe- similar to PI, the polymer is used widely thor-
nyl and naphthalene at and above 1023 K [151]. ough out the electronics industry. PEI is also
The high char yield suggests that random used in the automotive industry for temperature
chain scission is accompanied by carbonisation sensors and lamp sockets and due to its dimen-
and pyGC/MS data indicates that carbonisation sional stability is used for large, flat parts such as
might be the dominant pyrolysis pathway at computer hard disks [154].
temperatures above 1023 K [153].
Polyphenylene Sulphide (PPS)
Polyimide (PI) Polyphenylene Sulphide (PPS) is a high temper-
Polyimide (PI) is a high-temperature, linear ther- ature, high strength and highly chemically
moplastic. Similar to PEEK, PI is used in high resistant thermoplastic. This is indicated by
performance applications as a replacement for the presence of an aromatic benzene ring on the
www.ebook777.com
Fig. 7.42 Structure of a O

repeat unit of polyether O
imide (PEI)
n
N
N CH3
O
O O
O CH3
Fig. 7.43 Structure of a higher temperatures irreversible water loss is also

repeat unit of S observed, through chemical dehydration leading
n
Polyphenylene Sulphide
to crosslinking and char formation (Fig. 7.44).
(PPS)
Wood based products are primarily composed
of three components; hemicellulose, cellulose
Table 7.8 Onset decomposition (Td), peak mass loss and lignin. Breakdown of these components,
temperature (Tp), and ignition (Tig) temperatures of however, is not entirely simultaneous. The hemi-
PEEK, PEI and PPS [155]
cellulose, particularly its pentosans, decompose
Polymer Td/K Tp/K Tig/K first, largely between 473 and 533 K, followed by
PEEK 843 873 843 cellulose at 513 K to 623 K, and finally by the
PEI 800 828 801 lignin at 653–773 K.
PPS 777 818 848
Cellulose evolves water in the first stage of
thermal decomposition, before any other signifi-
backbone of the structure linked with an electro- cant changes are observable. Early in the pyrolysis
negative sulphur atom [154] shown in Fig. 7.43. of cellulose some of the carbon-oxygen bonds
Due to its highly crystalline nature, PPS is in the links between the glucose units may be
brittle and is often reinforced with glass fibre. expected to undergo random chain scission
PPS has similar mechanical properties to other (Fig. 7.44). During pyrolysis the water present
engineering thermoplastics and its intrinsic from the first decomposition stage, and if hemi-
flame resistance allows for applications in the cellulose is also present, the acids resulting from
electronic industry and instances where the its decomposition will also be present; both
material’s mechanical properties need to remain serve to promote hydrolysis. The decomposition
constant at elevated temperatures. As such, PPS continues until molecules are small enough to be
is used for impellers and pump housings [154]. volatile are produced. These include formalde-
The onset of decomposition (Td) in PEEK hyde, acetone, glyoxal, glycolic aldehyde,
occurs at a higher temperature (843 K) than in glycolic acid, lactic and dilactic acid, formic and
PEI and PPS as is shown in Table 7.8 [153]. In acetic acid, as well as water, carbon monoxide and
both PEEK and PEI, the peak mass loss tempera- carbon dioxide.
ture (Tp) occurs after ignition, in PPS this occurs
before. PPS also ignites at a higher temperature Proteins
than PEI and PEEK. Proteins (poly α-amino acids) (NHCHRC
(O))n are linked in the same way as polyamides,
but as the R substituent is different for each of over
Natural Polymers 20 amino acids, these affect their decomposition.
Thermal decomposition of proteins has probably
Polysaccharides been more extensively studied than the sum of all
Cellulosic materials contain both bound and other polymers, and is more usually described as
unbound water. Unbound water is lost below its cooking. The most important protein polymers
normal boiling point of 373 K. Bound water is from a fire safety perspective are silk and wool
released above 373 K in a reversible process. At and others associated with textiles. When heated,
HO
HO H
H H OH H H
H O H OH
HO H O H OH
O HO
O O
H O
H H
OH H O H OH H O
OH OH H
HO OH H H
OH
Cleavage of cellulose chain
Fig. 7.44 Part of a cellulose chain showing cleavage between two glucose units
silk starts to decompose above 250 C and forms a

char. This charring characteristic is probably
largely influenced by the dehydrating and cross-
linking tendency of the hydroxyl group within the
serine-CH2OH substituent. Wool fibres and fabrics
have significantly greater commercial applications
in products such as protective clothing and con-
tract upholstery, where high levels of fire-resistant
performance are required. Wool comprises
18 α-amino acids, including nearly 10 % by mass
cysteine, (NHCH(CH2SH)C(O)) which
dimerises forming cystine linkages (two R groups Fig. 7.45 Structure of cellulose nitrate
from different parts of the chain form linkages
CH2SSCH2) providing cross-links number of hydroxyl groups on the sugar mole-
between adjacent chains. The high sulphur content cule, which leave the polymer as water molecules
(3–4 wt%) coupled with the high nitrogen content during decomposition, result in char formation.
(15–16 wt%) contributes to the inherently low Numerous derivates of cellulose have been
flammability of wool. The fibre also contains prepared, including cellulose acetate (“acetate”)
about 15 wt% of adsorbed moisture under normal and cellulose nitrate (once used as a medium for
atmospheric conditions. When wool is heated, it photographic films, now presenting a problem for
starts to give off its adsorbed moisture at 100 C safe storage of archive film), which is also used as
and above, and then starts to thermally degrade a paint. Cellulose nitrate (or nitrocellulose)
rapidly above 200 C, with the notable evolution deserves special mention as it contains enough
of hydrogen sulphide (H2S) from cleavage of cys- oxygen to fuel its own decomposition, and there-
teine linkages above 230 C alongside char forma- fore can burn under water, and its heat release is
tion. The low flammability of the volatiles, not measurable by oxygen depletion calorimetry.
coupled with char formation, from cross-linking Fortunately, its only other common use is for
and dehydration of many α-substituents gives a table-tennis balls (Fig. 7.45).
relatively high ignition temperature of 840–870 K.
Polyesters Polylactic Acid (PLA),
Polyhydroxybutyrate (PHB)
Biopolymers These materials have an aliphatic backbone, and
are therefore more flammable than their aromatic
Cellulose Based counterparts. As the monomer may be derived
Cellulose derived from wood pulp has an average from living sources, rather than fossilised mate-
of 3000 repeat units, while cellulose derived from rial, they are favoured as being part of a cycle
cotton has around 15,000 repeat units. The large which does not contribute to the global increase
www.ebook777.com
in carbon dioxide. They are, however, flammable chloride and borax which formed a glassy layer
and this is one of the limitations on their wider on heating [156]. However, the main driver for
use. They are used for packaging, films etc. development came with the growth of the plastics
Polylactic acid decomposes to produce water, industries and the resultant widespread use of
carbon dioxide, acetaldehyde, acetone, acrylic synthetic polymers. During the 1960s and 1970s
acid, acetic acid and cyclic oligomers [154]. fires became more common and more severe.
Polyhydroxybutyrate decomposes to give dimer Anecdotally, fire fighters reported a change from
(41 %) crotonic acid (35 %) trimer (12 %) and fires with limited visibility, to those with almost
tetramer (3 %) below 611 K, with crotonic acid no visibility due to dense smoke, primarily
predominating at higher temperatures [155]. resulting from newly available low-cost polyure-
thane foam furniture. The increasing severity of
the fire problem led to the development from
Fire Retardants empirical tests for flammability to engineering
models capable of providing data on burning
In general, fire retardants are more expensive behaviour. This was accompanied by a shift in
than their host polymers, and are only added in emphasis from ignitability to peak heat release
order to meet regulatory requirements. If a par- rate. The cone calorimeter provided a reproduc-
ticular polymer is used in a high risk situation ible means of quantifying penetrative burning into
(mass transport, electrical and electronic, a sample, but not surface spread of flame. The
upholstered furniture, or certain construction main drivers in fire retardant development have
applications) and it is too flammable, fire been summarised in Table 7.9.
retardants may be incorporated to ensure that it According to a survey carried out by SRI Con-
meets regulatory criteria. Understandably, sulting [157], the total market for flame retardants
manufacturers will optimise their materials in in the United States, Europe and Asia in 2007
order to pass regulatory tests, and it is incumbent amounted to about 1.8 million metric tons
on the regulators to ensure that their tests con- (Fig. 7.46). This market was expected to grow at
tinue to be appropriate to the types of fire retar- an average annual rate of about 3.7 % per year on
dant technologies being deployed. a volume basis over the period 2007–2012. It is
split roughly equally between Europe, America
and Asia, with halogenated flame retardants and
Drivers in Fire Retardant Development antimony oxide comprising around 37.5 %.
The history of fire retardants goes back to Egyp-

tian times when solutions of alum (hydrated potas- Fire Retardant Strategies
sium aluminium sulphate (KAl(SO4)2 .12H2O))
were used to treat timber. Gay-Lussac protected The wealth of different fire retardant strategies is
theatre fabrics from fire by treatment with a testament to the ingenuity of the chemists who
mixtures of ammonium phosphate, ammonium developed them. Fire retardant agents have been
Table 7.9 The main drivers in fire retardant development

Decade Event Demand
1960s Widespread availability of cheap polymer Reduced ignitability
products—more serious fires
1970s Smoke much worse Reduced smoke
1980–1990s Development of Cone Calorimeter (and emphasis Reduced Peak Heat Release
on peak heat release rate, rather than ignitability)
Increase in deaths from smoke inhalation Reduced Fire Toxicity
2000s Halogen FRs found across the ecosystem Halogen-free FRs
2010s Climate change and other environmental Sustainable FRs
concerns become mainstream
Fig. 7.46 Volume

Aluminium
consumption of flame
hydroxide
retardants in the USA,
Europe and Asia [159]
Brominated
Organophosphates
Other
Europe
Asia
Antimony Oxide
America
Chlorinated
0 100 200 300 400 500 600 700 800

Volume /x1000 tonnes
Fig. 7.47 Classification of fire retardant strategies
classified in many ways: place of action—gas or condensed phase [158]. It is useful to distinguish
condensed phase; mode of action—physical or within the broadest category: fire retardants
chemical; chemical nature of agent—halogen, include any material which reduces the flamma-
phosphorus, metal hydroxide or carbonate, etc.; bility of the polymer; flame retardants inhibit the
means of incorporation of agent—additive or gas phase free radical reactions responsible for
reactive (i.e. bound onto the polymer chain etc.). flaming combustion. Figure 7.47 illustrates how
Unfortunately, these classifications cannot be the major fire retardant strategies could be
unambiguously applied to particular fire subdivided. Fire retardants have evolved over
retardants—for example, the most widely used the last four decades in order to meet the demands
fire retardant, aluminium hydroxide releases of industry and regulators, from halogen based
water, so acting in the gas phase, but in doing so flame inhibitors to cleaner, char promoters,
absorbs heat and leaves a protective residue in the resulting in less smoke and toxic gas emissions.
www.ebook777.com
While halogenated flame retardants continue to occur. The exothermic processes are thus
be used in a wide range of existing products, very stopped, the thermal feedback to the condensed
little new work has been published on them. phase is reduced, so the supply of flammable
Instead, research has been heavily focussed on volatiles is reduced.
finding suitable halogen-free replacement fire • Reaction in the solid phase: Here two types
retardants, indicating that the industry has indeed of reaction can take place. Firstly, breakdown
recognised the need for change, in the face of of the polymer can be accelerated by the
increasing pressure, predominantly resulting fire retardant, causing pronounced flow of
from environmental concerns. the polymer allowing it to recede from the
A fire retardant should inhibit ignition, and flame, usually by dripping. Secondly, the fire
may also suppress the combustion process. retardant can promote the formation of a layer
Most fire retardants interfere with a particular of carbonaceous char on the polymer surface.
stage of combustion, e.g. during heating, decom- This can occur, for example, through the
position, pyrolysis, ignition or sustained burning. dehydrating action of the fire retardant
generating double bonds in the polymer. Ulti-
mately, these form the carbonaceous layer by
Physical Action cyclizing and cross-linking.
There are several ways in which the combustion

process can be retarded by physical action: Additive vs Reactive Fire Retardants
• By cooling: Endothermic processes such as
decomposition of additives, cool the substrate Additive fire retardants are incorporated in the
to a temperature below that required for sus- plastic either prior to, during, or, more fre-
taining the combustion process. quently, following polymerisation. They are
• By formation of a protective layer: The com- used especially in thermoplastics. If they are
bustible components of the condensed phase compatible with the plastic they may act as
can be shielded from the gaseous phase with a plasticisers, otherwise they are considered as
protective solid layer. The condensed phase is fillers. They are sometimes volatile or tend to
thus protected from radiation, and oxidative leach out of the polymer, so fire retardance may
attack, while pyrolysis gas evolved is inhibited. be gradually lost. The development of high
• By dilution: The incorporation of inert molecular weight products (oligomeric and poly-
substances (e.g. fillers) and additives which meric fire retardants) can eliminate this problem.
evolve inert gases on decomposition dilutes
the fuel in the condensed and gaseous phases Reactive fire retardants are built into the polymer
so that the lower ignition limit of the gas molecule, for example by attaching a fire retarding
mixture is not exceeded, and heat to the group to a monomer group. This prevents them
condensed phase is dissipated. from bleeding out of the polymer and volatilising
and their fire retardance is thus retained. In addi-
tion, they have no plasticising effect and do not
Chemical Action affect the thermal stability of the polymer. They are
used mainly in thermosets (especially polyesters,
Chemical reactions in the condensed and gas epoxy resins and polyurethanes) in which they can
phases may interfere with the combustion process. be easily incorporated. They are, however, nor-
• Reaction in the gas phase: The radical mech- mally a more expensive solution than additive fire
anism of the combustion process which takes retardants. In crystalline polymers, where the struc-
place in the gas phase is interrupted by the tural integrity, strength and other physical
flame retardant. The radical concentration falls properties depends on the microcrystalline struc-
below a critical value, and flaming cannot ture, the presence of foreign groups on the polymer
chain is likely to have a detrimental effect on the This occurs in reaction 1 and 2 where each
physical properties. “·”represents an unpaired electron.
Combinations of additive or reactive fire
retardants with further additives can produce an H þO2 ! OH þ O ð7:43Þ
additive, synergistic or antagonistic effect.
While the additive effect is the sum of the indi- O þH2 ! OH þH ð7:44Þ
vidual actions, synergism is a greater than addi- Halogen-containing flame retardants act by
tive effect, and antagonism a less than additive interfering with the radical chain mechanism taking
effect. When used alone, some additives show no place in the gas phase. The high-energy OH · and
or only negligible effectiveness. The synergistic H · radicals formed by chain branching are
effect occurs when they are used together with removed by the halogen-containing flame retardant.
other specific fire retardants. Such synergists At first the flame retardant breaks down to
have achieved great importance in practical use
because they are often more effective or less RX ! R þX ð7:45Þ
expensive than the single fire retardants.
where X · is either Cl · or Br · .
The halogen radical reacts to form the hydro-
gen halide:
Halogenated Flame Retardants
X þRH ! R þHX ð7:46Þ
There are more than 40 brominated flame
retardants in current use and an additional num- which in turn interferes with the radical chain
ber of chlorinated FRs [159]. These are used by mechanism:
simple melt blending with the polymer (additive
HX þ H ! H2 þ X ð7:47Þ
FRs) or by incorporation into the polymer chain
during polymerisation, (reactive FRs). Some
additive FRs are small molecules while others The removal of H · is key to elimination of
are oligomers or polymers. From a health and the main chain branching step (when 1 unpaired
environmental perspective, small molecules will electron becomes 3).
be easiest to release from the polymer matrix
HX þ OH ! H2 O þ X ð7:48Þ
(by evaporation, leaching, end-of-life processing
etc.) Oligomeric or polymeric additives may be The removal of OH · blocks the main heat
released during end-of-life processes, while release step of hydrocarbon combustion, the con-
reactive FRs, which have been successfully version of CO to CO2, by replacement with less
incorporated into the polymer are only likely to reactive halogen free radicals in the gas phase
be problematic during burning or incineration or [160]. The H and OH radicals are essential for
other end-of-life processes. many flame reactions and are involved in the
main heat release in reaction 7.49.
Mode of Action of Halogenated Flame
Retardants CO þ OH ! CO2 þ H ð7:49Þ
Halogenated flame retardants act by releasing
hydrogen halides (HX) during thermal decompo- Loss of H and OH will increase the yield of
sition. If the hydrogen halide release coincides toxic carbon monoxide and other products of
with fuel gasification, then HX can interfere with incomplete combustion (hydrogen cyanide
the gas phase combustion processes. Flaming (HCN), organoirritants and soot).
combustion involves a very small number of The high-energy H · and OH · radicals are
highly reactive free radicals to propagate the removed by reaction with HX and replaced with
gas phase oxidation processes. For ignition to lower-energy X · radicals. The actual flame retar-
occur, the number of radicals must increase. dant effect is thus produced by HX. Kinetic reaction
www.ebook777.com
schemes predict that HBr is recycled around seven achieved with a filler material such as a metal
times in order to account for the observed flame hydroxide. For example, aluminium hydroxide
inhibition [161], while the HCl is not recycled. not only evolves water vapour, but also
In the condensed phase, the resulting unsatu- absorbs a vast amount of heat as it is
rated polyenes may act as char precursors, dehydrated.
forming products with a tendency to cyclize and
condense to yield carbonaceous products, which 180200 C
2 AlðOHÞ3 ðsÞ ! Al2 O3 ðsÞ þ 3 H2 OðgÞ
protect the condensed phase below the flame
zone against attack by oxygen and radiant heat. ΔH ¼ þ1:3 kJ g1
In PVC, after loss of 60 % mass as hydrogen
chloride from the surface layers, char formation This is about the same amount of energy that
is a significant fire retardant mechanism would heat 1.6 g of polythene to 673 K (its
protecting the underlying polymer. decomposition temperature).
The halogen content in the polymer compound, In addition, aluminium hydroxide is a good
and its chemical binding, will dictate the flame conductor of heat, reducing the local hot spots,
retardant behaviour. In the presence of antimony which are responsible for starting fires.
oxide (Sb2O3), the efficiency of halogenated Aluminium hydroxide (incorrectly referred to
flame retardants is improved, although antimony in industry as alumina trihydrate or ATH),
has no flame retardant effect on its own. This is accounts for half of all the fire retardant
believed to result from the formation of volatile additives used by weight. However, this is
SbX3 and other species which are more effective partly due to its low price and the requirement
halogen carriers than HX. for it to be present in a polymer at a 50–70 %
loading to be effective.
Alternatives to aluminium hydroxide have
Metal Hydroxide and Carbonate Fillers also been investigated (Table 7.10).
The absolute contributions of these fillers to
These involve incorporation of a material that fire retardancy have been estimated [160] as
evolves an inert gas on heating. This is often shown in Table 7.11 and Fig. 7.48.
Table 7.10 Mineral filler fire retardants

Filler Formula Tdecomp/ C ΔHdecomp/kJ g1
Aluminium hydroxide Al(OH)3 180–200 1300
Magnesium hydroxide Mg(OH)2 300–320 1450
Calcium hydroxide Ca(OH)2 430–450 1150
Nesquehonite MgCO3.3H2O 70–100 1750
Hydromagnesite Mg5(CO3)4(OH)2.4H2O 220–240 1300
Huntite Mg3Ca(CO3)4 400 980
Ultracarb Hydromagnesite/Huntite 220–400 1172
Boehmite AlO(OH) 340–350 560
Table 7.11 Assumptions for quantification of the magnitude of physical fire retardant effects of mineral fillers
Effect How quantified
Diluting polymer in condensed phase Heat capacity of the filler prior to decomposition
Endothermic decomposition of filler Heat of decomposition
Presence of inert residue Heat capacity of the residue after decomposition
Presence of diluent gases Heat capacity of the diluent gases
Fig. 7.48 Absolute estimation of the contributions of individual effects to the overall fire retardancy of mineral fillers
[160]
– To meet a regulatory requirement; and

Burning Behaviour of Polymeric – To demonstrate that the material being tested
Materials will perform adequately in a specific fire
scenario [166].
Quantifying Fire Behaviour A wide range of industry standard tests are used
to demonstrate a product’s suitability for a partic-
Polymer flammability has no specific meaning; the ular application, and ultimately all products with
flammability of a polymer is often defined by the specified fire performance must meet these
method used to measure it [162]; flammability has criteria. However, the specification of materials,
been described as the ease at which a substance will or the design of fire retardant materials requires
ignite [163] but is also used to indicate the rate of simple quantifications of a material’s fire
fire growth after ignition: flammability is a function behaviour. A relatively small number of fire tests
of both gas and solid phase chemistry [164]. Fire are employed in order to quantify the fire
tests are crucial to the development, screening and behaviour of a material. The most common of
evaluation of materials with improved fire safety. these are the limiting oxygen index (LOI) and the
Techniques employed to measure the ignition and Underwriters’ Laboratory UL 94 test, both ease of
burning behaviour of a polymer are numerous. extinction tests, and the cone calorimeter, which
Some examples of fire tests are shown in measures time to ignition and rate of heat release.
Table 7.12 together with the individual fire
properties they can quantify [165]. Ignitability
It can be seen from Table 7.12 that no one test Ignition is a very important parameter
covers all the parameters describing fire safety controlling flame spread and fire growth. How-
behaviour, but many address more than one ever, the source of ignition necessarily impacts
parameter. When selecting a test method, it is on the result, and therefore ignition temperature
necessary to determine the end use of the prod- is found to depend on the design of the test used
uct, and the likely fire scenarios. Aside from to measure it. The characteristics of several com-
materials’ development, in fire testing, there are mon ignition sources have been reported [167]
generally two end purposes: and are shown in Table 7.13. Tests which use
www.ebook777.com
Table 7.12 Common fire tests and parameters assessed

Glow Small- Laterally Limiting Smoke Steady
Radiant wire test flame UL-94 induced oxygen density state tube
ignition (IEC ignitability (IEC Cone flame test index chamber furnace
test (ISO 60695-2- test (ISO 60695- calorimeter (ISO (ISO (ISO (ISO
5657) 10–13) 11925) 11-10) (ISO 5660) 5658) 4589-2) 5659) 19700)
Ignitability x x x x x x
Rate of x x x
flame
spread
Rate of x
heat
release
Ease of x
extinction
Smoke x x
production
Effluent x
toxicity
Table 7.13 Characteristics of some common ignition sources

Source of ignition Duration of source/s Total heat/kJ Maximum heat flux/kW m2
Match flames 2–35 6 18–20
Cigarette lighter 30 24 16–24
Diffusion flame, small 30 8 18–32
Diffusion flame, large 30 15 6–37
Premixed flame, small 30 50 58
Premixed flame, large 30 120
Electric spark <100 mJ
Electric arc 1 0.4
Electric arc 5 15
Electric bulb, 60 W 30 3
Electric bulb, 100 W 30 8
Electric hot plate, 1 kW 30 30
Electric radiator 30 90 20–25
Crumpled paper
1/2 sheet 85 175 7–10
1 sheet 152 340 7–22
2 sheets 223 680 7–21
3 sheets 333 1020 5–22
4 sheets 335 1600 6–23
Folded paper
5 sheets 380 1680 14
10 sheets 420 3500 15
Wastepaper basket 360 3400 10–40
1600 5000 10–40
Small stuffed toy 330 9500 20–39
Scatter cushion 513 11,000 17–28
Bedding 1200 130,000 26
sparks, electric arcs, hot surfaces or open flames behaviour of vertically or horizontally oriented
for less than 30 s will not deliver more than specimens made from plastic and other
100 kJ, and represent low severity tests. Crum- non-metallic materials, exposed to a small-
pled or folded paper can deliver between 200 and flame ignition source of nominal 50 W
4000 kJ in 1–8 min, representing a medium power [178].
severity exposure, while burning bedding can The method determines the linear burning rate
deliver 130,000 kJ in 20 min, representing a and the afterflame/afterglow times, as well as the
high severity exposure. Many standard tests use damaged length of specimens, and is applicable
a gas or liquid fuel ignition source, of specified to solid and cellular materials with density of at
energy or power, corresponding to one of the least 0.25 g cm-3, provided they do not shrink
“unwanted fire” ignition sources. away from the applied flame without igniting.
Ignition is a complex subject, which has been The test method described provides a classifica-
addressed by several authors [168–172] and com- tion system, which may be used for quality assur-
prehensively described and summarized else- ance, or the pre-selection of component materials
where [173]. Ignition occurs when the oxidising of products, provided that the test sample thick-
volatiles feed enough heat back to the polymer to ness is the thinnest to be used in the application.
volatilise a similar concentration under the The Underwriter’s Laboratory designed this stan-
conditions of the test. Thus, the fraction of the dard to indicate a plastic’s flammability for use
heat of combustion passed back to the polymer as part of an electrical appliance, rather than the
for a given mass of fuel must be greater than its hazards of a material under actual fire conditions.
heat of gasification. This critical condition can be UL 94 flammability testing is the first step toward
described by the mass loss rate at ignition. For the obtaining a plastic recognition and subsequent
cone calorimeter the critical mass loss rate is listing in the “Plastics Recognized Component
around 1–6 g s1 m2, and the resulting heat Directory” (formerly known as “Yellow Cards”).
release rate at ignition (HRRig) is around The 94 V test describes the Vertical Burn test
20–100 kW m2 [174–176]. Ignition does not (Fig. 7.10), which is a more stringent test than the
directly or necessarily correspond to “flammabil- Horizontal Burn method 94HB. The set up uses a
ity” measured by LOI or UL 94, since both of very small Bunsen flame with a manometer and
these are ease of extinction tests, and correspond needle valve to control the gas flow. The criteria
better to the minimum mass loss rate needed for for each classification are shown below
sustained burning or fire propagation [177]. Ignit- (Table 7.3). While the test is crude, it is a realistic
ability can be determined by measurement of the ignition scenario, and lets the user see what is
ignition delay time, an important fire parameter, happening during the test. It is easy to set up a
for example using a cone calorimeter. small test burner with a 15 mm blue flame in
order to provide an instant simulation of the test
Ease of Extinction Tests (Table 7.14).
UL-94 ‘Bunsen Burner’ Test IEC 60695-11-10 For the UL-94 vertical burning test, the
This is a small-scale laboratory screening conditions and measures are depicted in
procedure for comparing the relative burning Fig. 7.49.
Table 7.14 UL94 classifications

V-0 Burning stops within 10 s after two applications of ten seconds each of a flame to a test bar. NO
Vertical burn flaming drips are allowed
V-1 Burning stops within 60 s after two applications of ten seconds each of a flame to a test bar. NO
Vertical burn flaming drips are allowed
V-2 Burning stops within 60 s after two applications of ten seconds each of a flame to a test bar. Flaming
Vertical burn drips ARE allowed
H-B Slow horizontal burning on a 3 mm thick specimen, with a burning rate less than 3“ per minute or
Horizontal burn stops burning before the 5” mark. H-B rated materials are considered “self-extinguishing”
This is the lowest (least flame retardant) UL94 rating
www.ebook777.com
Test strip
10 mm 57 kW m−2
5 mm 66 kW m−2
0 mm
−10 mm
25 kW m−2
10 mm
10 mm
Test flame
and burner
Fig. 7.50 The time-average incident burner flux in the
300 mm UL-94 V test showing the heat flux at the top of the flame,
5 mm below the top, and at the base of the sample strip
Cotton wool [189]
Fig. 7.49 The UL-94 vertical test arrangement

Pilot flame, removed after ignition
Attempts at explaining the phenomena occur-
ring during the test have been made, however
understanding the relationship between ignition
and small flames has been described as ‘poor or
lacking’ [179]. Several studies have endeavoured Sample
to correlate the results of the UL-94 test to other
flammability measures, most commonly the cone
calorimeter (ISO 5660 [180–185] although cor-
relation to data, where little consensus exists on
its interpretation, is ambiguous [96, 186]. The
heat flux generated by the Bunsen burner, due
to its partially premixed, blue diffusion flame is Gas inlet – O2/N2 mixture
responsible for the ignition of the polymer and
as a result, research has been conducted on Fig. 7.51 Limiting oxygen index test
determining the values for this parameter
[187]. The time-average incident heat flux at 3 min, or propagates down the length of the
height of 0.5 cm and 1 cm for the burner are sample, the test is repeated at lower oxygen
shown in Fig. 7.50. concentrations. If it self-extinguishes the test is
repeated at higher oxygen concentration. The
Limiting Oxygen Index oxygen concentration is adjusted in this manner
This test relates to the minimum concentration of until the specimen just supports combustion.
oxygen that will just support flaming combustion The oxygen concentration reported is the volume
in a flowing mixture of oxygen and nitrogen percent, with repeatability often as good as
(Fig. 7.51) [188]. A specimen is positioned verti- 0.1 % O2.
cally in a transparent borosilicate glass test col- Downward flame spread may be regarded as a
umn and a mixture of oxygen and nitrogen is best case scenario, and while a material with
forced upwards through the column. The speci- limiting oxygen index (LOI) < 21 % should be
men is ignited at the top. If the flame remains for considered to support downward flame spread,
materials with LOI > > 21 % should still be Bench-Scale Measurement of Heat
considered flammable. Particular problems arise Release
with materials with a high dripping propensity,
since ignition will only occur under extreme The Cone Calorimeter
circumstances. Very thin materials often have The Cone Calorimeter [180, 189] (Fig. 7.52) was
insufficient heat release per unit area to support developed specifically to determine the rate of
combustion, while thicker materials conduct too heat release and effective heat of combustion of
much heat away from the flame zone. Thus there building materials (ISO 5660–1). It was subse-
is a “most flammable thickness” for many quently modified to determine smoke generation
materials around 1.6 mm. For non-charring (ISO 5660–2) and later applied to furniture.
materials, the criteria for ignition (heat transfer A horizontal specimen, 100 mm square, typi-
from the flame > heat of gasification per unit cally 3–6 mm, but up to 50 mm, thick is mounted
mass) are replicated in the criteria for extinction. under a steel frame, such that only the surfaces,
However, while ignition requires a source but not the edges are exposed to a conical radia-
(whose energy input will affect the result), tor pre-set to between 10 and 100 kW m2
extinction has no such dependence. The dilution mounted beneath an instrumented hood and
of the flame by nitrogen causes the flame to duct. A spark ignition is used and the specimen
swell, reducing the amount of heat fed back to is mounted on a load cell. Heat release is
the sample below the flame. As a rule of thumb, quantified by oxygen depletion calorimetry.
there is generally some correlation between the Measurement of heat release from real fires by
time to ignition in the cone calorimeter and the oxygen depletion calorimetry is well established,
LOI, but none between LOI and heat release rate. and gives sensible values which relate to the
Fig. 7.52 Diagram

of cone calorimeter
Exhaust duct, leading to gas sampling
Conical Radiant Heater
Sample holder
Spark igniter
Load cell
www.ebook777.com
extent of burning. Provided the effluent flow pHRR), fire growth rate index (FIGRA), THR
through the exhaust is carefully controlled, the etc., for regulatory purposes.
heat release will be proportional to the oxygen These applications of the cone calorimeter
depletion. A sample of the effluent is cooled to define different techniques and data evaluation.
remove water and analysed using a paramagnetic For regulatory purposes, its strengths are its well-
analyser and non-dispersive infrared CO and defined conditions, reproducibility and unambig-
CO2 analysers. It will not take into account the uous data evaluation of one or two characteristic
reduction in heat release due to the endothermic values. The use of defined, and in some way
decomposition of metal hydroxide fire retardants, ideal, burning behaviour is suitable for develop-
such as aluminium hydroxide (ATH), although ing pyrolysis and burning models and for
this can be compensated for separately. (For obtaining reasonable input values for the simula-
PMMA containing 60 % ATH this would result tion of fires. However, as a fire scenario, it is not
in an overestimation of total heat release by representative of most real fires. Small fires are
~8 %). A detailed description of the use and not usually initiated with radiation from
interpretation of data from the cone calorimeter above, piloted by a spark ignition source, and
for fire retardant materials development has been surrounded by a frame which acts as a large
published [193]. The cone calorimeter monitors a heat sink, producing an unusual gas flow field
comprehensive set of fire properties in a well around the flame zone, and where the effects of
defined fire scenario. The results can be used to sample dripping are negligible.
evaluate material specific properties, setting it
apart from many of the established fire tests Heat Release Curves from Cone Calorimetry
which are designed to monitor the fire response The heat release rate (HRR) during the cone
of a certain specimen. calorimeter experiment gives rise to a character-
The cone calorimeter covers ignition followed istic heat release rate curves versus the time
by essentially penetrative flaming combustion, (Fig. 7.53) [174, 190].
where the flame front moves through the bulk
of the sample. The ignition parameter measured Microscale Measurement of Heat Release
in the cone calorimeter is the time to ignition, Microscale Combustion Calorimetry [192]
which depends on the thermal inertia, critical (MCC) (section “Microscale Combustion Calo-
heat flux and critical mass loss for ignition, or rimetry”) evaluates the combustibility of milli-
alternatively the critical surface temperature for gram samples by separately reproducing the
ignition. Fire response parameters measured in solid state and gas phase processes of flaming
the cone calorimeter include mass loss, heat combustion by controlled pyrolysis of the sam-
release rate (HRR), total heat release (THR), ple in an inert gas stream, followed by high
smoke production and CO production. Fire temperature oxidation of the volatile pyrolysis
response properties more typical of fully devel- products. Oxygen consumption calorimetry is
oped or post flashover fire scenarios are not used to measure the heat of combustion of the
replicated in the cone calorimeter. pyrolysis products. The maximum amount of
There are three distinct uses of cone calorim- heat released per unit mass per degree of tem-
eter data: perature (J g1 K1) is a material property that
– To compare the fire response of materials: to appears to be a good predictor of “flammability”.
assess their fire performance; to perform The heat release capacity (HRC) and total heat
screening for materials development; or to release (THR), obtained by MCC, are related to
develop pyrolysis and burning models. the char yield and the heat of complete combus-
– To determine data for input to simulations or tion of the volatiles. It takes no account of phys-
predictions of full-scale fire behaviour. ical effects, such as dripping, wicking, and
– To determine characteristic parameters such sample thickness; or chemical effects such as
as the maximum HRR (peak heat release rate, flame inhibition, because the conditions force
HRR Thermally thick non-charring (or non-residue forming)

Thick Sample samples show a strong initial increase after ignition up to a
(Non Charring) steady HRR191. This plateau remains until near the end of
the test, when an additional pHRR occurs. This peak is
caused by reduction in heat transfer by conductivity through
the increasingly thin sample, since the glass wool supporting
the sample prevents heat transfer to the sample holder as
the pyrolysis zone approaches it.
time
HRR
For intermediate thermally-thick, non-charring samples the
plateau vanishes. The averaged or steady HRR is only
marked by a shoulder. The pHRR increases in comparison
Intermediate with thermally thick non-charring samples since its origin is
Thickness half way between the thermally-thick non-charring and
(Non Charring) thermally-thin behaviour.
time
Thermally-thick charring (residue forming) samples show an

HRR
initial increase in HRR until an effective char layer is formed.
As the char layer thickens, this results in a decrease in HRR.
Thick Sample The maximum reached at the beginning is the steady HRR
(Charring) and the pHRR.
Some thermally thick charring materials, such as wood,
show a pHRR at the beginning, typical for charring, and a
second pHRR at the end. The second peak may be caused by
cracking char or increase in effective pyrolysis temperature,
time as observed with the thick non-charring materials.
HRR Thermally-thin samples are characterized by a sharp peak in

HRR, since the whole sample is pyrolysed at the same time.
In this case, the pHRR is determined by their total fire load.
Thin Sample
(Charring/
Non Charring)
time
Fig. 7.53 Types of heat release curves from cone calorimetry [191]
pyrolysis and combustion to completion [193]. Influence of Physical Properties

However, varying the combustion temperature on Flammability
or oxygen concentration results in incomplete
combustion as occurring in real fires. The The following factors affect polymer combustion
THR results have been correlated to LOI; in real fires, and should therefore influence the
HRC and char residue to LOI; and HRC and outcome of a suitably designed test.
THR with peak heat release rate (pHRR) in Fuel production—when the fuel in the gas
the cone calorimeter. It has been used as a phase reaches a critical concentration, ignition
screening test for efficacy of flame retardant and flaming can occur. While the fuel produc-
additives [194]. tion rate during heating is essentially a
www.ebook777.com
material property, the air flow around the Smoke Formation—Smoke can act as both the
sample may profoundly alter the ignition source of radiation (a sooty yellow diffusion
temperature. flame radiates much more than a blue
Presence of inhibitors or diluents—Cl · or premixed flame) or block radiation from the
Br · or PO · are stable radicals which will flame back to the polymer.
reduce the critical concentration of active
radicals such as H · and OH·, in the flame
zone. The effect is most pronounced at igni- Char Formation
tion, and least evident under developed fire
conditions. Classifying polymers by the structural units
Rheology of decomposing polymer—Some they contain has been used to calculate various
polymers depolymerise during decomposition flammability parameters and predict burning
reducing their viscosity. This allows better behaviour [195, 196]. The char-forming
dispersion of heat, and material flow away tendency (CFT) of polymers may be estimated
from the source of heat. This can result in from the contributions from each structural
harmless dripping away from the flame zone, group, referred to as “molar group
or flaming drips allowing flaming to spread contributions”. Van Krevelen has taken the
downwards. Some additives (e.g. high surface char-forming tendency of the individual struc-
area fillers, such as nanofillers) will increase tural units of polymers as an additive quantity,
the viscosity reducing dripping resulting in a and based on this, the following relationship has
more rapid increase in the surface tempera- been created:
ture. This will reduce the time to ignition. In X
some cases free radical initiators are added ðCFT Þi
CR
1200 ð7:50Þ
purely to promote dripping to remove the M
fuel from the source of heat.
Char formation—the formation of a char on Where
the surface of the polymer will reduce the CR ¼ Char Residue ð%Þ
flow of heat to and fuel from the sample. CFT ¼ Char Forming Tendency ðno unitsÞ
Intumescent chars bubble up and provide a M ¼ Molecular Weight of repeat unit g mol1
more effective barrier. However, in a typical
fire test, the direction of swelling is often Each structural group is assigned a value. This
towards the heat source, increasing the radiant is known as the char-forming tendency (CFT)
flux to the sample. shown in Table 7.15. Aliphatic groups are gener-
Orientation of sample—as flames rise, flame ally assigned a value of zero, although if they are
spread is easiest from below (going upwards) connected to aromatic nuclei they can have neg-
and hardest from above (going downwards). ative values (Table 7.16). Char forming tendency
Because of flow of molten material and ulti- cannot be calculated for polymers which contain
mately dripping, it is very difficult to correlate halogenated species as their soot-forming
vertical burning behaviour with horizontal tendencies would significantly affect the char
burning behaviour. formation.
Absorption of radiation—radiation from flames The char-forming tendency is a statistical con-
or a radiant panel must be absorbed by the cept. For example the phenyl group has a CFT
polymer. The presence of absorbing centres value of 1 C equivalent, which means that on
(conjugated double bonds, or black pigments) average only 1 in 6 phenyl groups in the polymer
can increase the localisation of the heating. forms a char, where the other 5 contribute to tar
Conversely, a highly reflective surface can and gas formation. If the benzene ring contains
significantly lengthen the time to ignition in 4 non-hydrogen, non-aliphatic substituents, all
certain tests. the rings will contribute to the char.
Table 7.15 Molar group contributions for char formation
Contribution to residue Contribution to residue

per structural in per structural in
Group unit CFT C-equiv. Group unit CFT C-equiv.
“All” aliphatic groupsa 0 0 N N 12 1
C C
O
-CHOH-(exception) 4 1/3 36 3.5
C C
N NH
CH
12 1 CH S 42 3.5
C C
N
24 2 CH CH 42 3.5
C N
N
36 3 H 84 7
N
N
48 4 O 84 7
N
60 5 N 108 9
72 6 O N 132 11
60 5 H H 120 10
N N
N N
96 8 O O 144 12
N N
O O
(continued)
www.ebook777.com
Contribution to residue Contribution to residue

per structural in per structural in
Group unit CFT C-equiv. Group unit CFT C-equiv.
72 6 N N 120 10
N N
O O
120 10 180 15
N N
N N
168 14
a
Without halogen groups
Table 7.16 Molar group contributions for char forma- 144 g per structural unit of PEEK. The molecular
tion of aliphatic groups, connected to aromatic nuclei weight of the PEEK monomer unit is
supplying hydrogen for the disproportionation reaction 288.3 g mol1. These can be used to estimate
(H shift)
the mass of the char residue (CR) to give:
Contribution to residue
12
Group per structural unit CFT in C-equivalent CR
1200 ¼ 50%
>CH2 and 12 1 288
>CH-CH2- The calculated char residue (CR) is 50 %.
-CH3 18 3/2 This is slightly greater than char yields deter-
-C(CH3)2 36 3 mined by experimental methods which give
-CH(CH3)2 48 4
values ranging from 41 % [198] to 47 %
[199, 200].
Example: Calculation of Char Forming

Tendency (CFT) Calculating Polymer Flammability from
Figure 7.54 shows the values assigned to the Molar Group Contributions
structural groups which make up the monomer
unit in polyetheretherketone (PEEK) [197]: Recently, another useful method has been devel-
These values are summarised in Table 7.17. oped to calculate the heat release capacity (HRC)
The CFT of PEEK has been determined as from additive molar group contributions. As a
12 and therefore the char residue will amount to material flammability parameter [201–203] the
O Example: Calculating Heat Release

Capacity (HRC)
C O n The calculation of heat release capacity is
4 4 4 illustrated by example of molar group
O
contributions for a diglycidylether of bisphenol-
Fig. 7.54 PEEK assigned with structural group A (BPA epoxy) cured by anionic ring opening
contributions for char polymerization. The chemical structure of the
repeat unit of the polymer is shown in Fig. 7.55.
Table 7.17 Summary of structural group contributions
The polymer repeat unit is comprised of six
and their char-forming tendencies basic chemical groups, and the heat release
Chemical group Value N CFT
capacity is calculated from the associated Ni,
4 3 12
Mi, and ψi for these groups, which are listed in
Table 7.19.
The molar heat release capacity (ηc) is
0 2 0
obtained by summing the group contributions
O according to their mole fraction in the repeat
O 0 1 0
unit, then dividing by the molar mass of the
C repeat unit to give the heat release capacity on a
Total 12 mass basis in units of J g1 K1.
X X
Ψ i ni Ψ i iNiΨ i
ηc ¼ ¼ X ¼X
HRC has been recognized as a tool for analysing M i ni M i i N i Mi
fire response and flammability of polymers. A
quantitative laboratory pyrolysis-combustion 204:5 kJ=mole K
¼ ¼ 601 J g1 K 1
method for directly measuring the heat release 340g=mole
capacity has been established [204–206], and
experimental data are presented which indicate The predicted value of 601 J g1 K1
that HRC can be used to correlate polymer struc- compares favourably with the measured value
ture with fire behaviour. of 657 J g1 K1 for this polymer.
The contribution of over 40 M groups has
been correlated to HRC [207, 208] as shown in
Table 7.18. The measured and estimated heat Conclusions
release capacities for over 80 polymers agree to
within 15 %, demonstrating a capability for Polymeric materials fuel nearly all unwanted
prediction of polymer flammability from chemi- fires. All polymeric materials contain large
cal structure. molecular chains, giving them greater strength
Specific heat release rate is a molecular-level and resilience, than either small molecules or
fire response parameter of a burning polymer. metallic structures. However, as almost all con-
Lyon et al. determined the specific heat release tain carbon and hydrogen they are easily oxidised
rate using the MCC (section “Microscale Com- and burn readily.
bustion Calorimetry”). Dividing the specific heat The diversity in the range of polymeric
release rate (W g-1) by the rate of temperature materials is huge and polymers may be classified
rise (K s1) gives a material fire parameter with in several ways: natural, biobased or synthetic;
the units (J g1 K1) representing the HRC. They means of polymerization; thermoplastics or
argue that the HRC is a true material property thermosets; molecular mass distribution; or phys-
that is rooted in the chemical structure of the ical properties. Each has impacts on their burning
polymer, and is calculable from additive molar behaviour. Some polymers, such as polyamide
group contributions [199]. 6, polyvinyl chloride and polyacrylonitrile differ
www.ebook777.com
Table 7.18 Structural groups and their molar contribution to the heat release capacity (molar group contributions
derived from a single polymer are marked*)
Contribution/ Contribution/ Contribution/
Structural group kJ mol1 K1 Structural group kJ mol1 K1 Structural group kJ mol1 K1
118* H 8.1 HO 19.8
O 77.0 7.6 Br 22.0

CH3
P NH
69.5 CH2 O 4.18 O 22.0
C
30.6 1.8 23.2*
CF2
C O
H3C 29.5 Cl Cl 0.1 O 25.5

C
N
H3C C
O
28.8 H 8.8 Cl 34.7
N
N
28.3 S 10.9* O O 36.4*
C N N
O O
26.6 O 11.6 O Pendant: 39.5

Backbone: 13.7
CH
C O
H3C 22.5 13.8 43.0*

O
P
N
N
O
19.0 H2N 13.9* O 49.0
O C O
18.7 F3C 14.8 53.5*
N
Si
16.7 C N 17.6 N 66.7

CH2
N N
15.1 N N 18.9* O 74.5

N
O O
O N
N
O
9.7 O 19.2 O 76.7

C C S O P O
O O
O CH2 H3C
CH CH2 O O CH2 HC
* *
H3C CH2 O
Fig. 7.55 Repeat unit of diglycidylether bisphenol A
Table 7.19 Group contributions used in the calculation of the heat release capacity of bisphenol-A epoxy
Molar mass Molar heat release capacity

Chemical group, i Number of moles N Mi/g mol1 Ψ/kJ mol1 K1 NiMi/g mol1 Ni ψ/kJ mol1 K1
1 12 28.3 12 28.3
C
2 13 26.6 26 53.2
CH
H3C 2 15 22.5 30 45.0

2 76 28.8 152 57.6
4 14 16.7 56 66.8
CH2
O 4 16 11.6 64 46.4
Total: 340 204.5
only in their average molecular mass, its distribu- most polymers exert a degree of crystallinity,
tion and any impurities arising from manufacture. increasing with the duration of cooling, giving a
Others such as polyethylene, differ as a result of sharper transition between solid and liquid
the polymerization process, so free radical phases. The chemical composition of the molec-
polymerised low-density polyethylene (LDPE) ular chains exerts a profound influence on the
has the most branching points (or starting points thermal decomposition of the polymers, with
for decomposition), catalytically polymerised chain branching, double bonds, or oxygen in the
linear low-density polyethylene (LLDPE) has a polymer backbone reducing the thermal stability,
smaller number of identical branching points, and and aromatic rings and crosslinking of the poly-
high density polyethylene (HDPE) has the least of mer backbone increasing the thermal stability.
branching points and hence the highest decompo- Polymer decomposition can best be studied on
sition temperature. Many polymers, such as a microscale by thermogravimetric analysis,
polyamides, polyesters, polyurethanes and which provides fundamental information
epoxies represent diverse classes, having only about gaseous fuel production rates, quantified
the chemistry of the linkages (e.g. esters or by the Arrhenius parameters A and Ea. Other
urethanes) between repeat units in common. techniques, such as DSC, DTA, DMTA, and
Both the physics and chemistry of polymers rheology provide additional information on the
affect their thermal decomposition and burning physical transformations occurring, while
behaviour. Depending on their thermal history, evolved gas analysis, using FTIR or GCMS, or
www.ebook777.com
oxygen depletion calorimetry, and char analysis, 4.07, American Society for Testing Materials, West
illuminate the chemical processes. Conshohocken.
2. J.M.G. Cowie and V. Arrighi, Polymers: Chemistry
The thermal decomposition of polymers and Physics of Modern Materials, 3rd Edition, CRC
is thermodynamically driven, as higher Press, Boca Raton (2008).
temperatures favour the formation of gaseous 3. S.I. Stoliarov, N. Safronava, and R.E. Lyon, “The
molecules, and is controlled by a frequently com- effect of variation in polymer properties on the rate
of burning”, Fire and Materials, 33, pp. 257–271
plex array of competing kinetic processes. (2009).
Attempts to identify individual reactions have 4. S.C. Moldoveanu, Analytical Pyrolysis of Synthetic
generally failed, and there is a broad consensus Organic Polymers, Techniques and Instrumentation
that predicting the rate of fuel gasification will in Analytical Chemistry, Volume 25, 1st Edition,
Elsevier B.V (2005).
suffice as input to pyrolysis models, and fire 5. R.E. Lyon, “Plastics and Rubber”, In Handbook of
models that include condensed phase processes. Building Materials for Fire Protection, Harper CA
The breakdown of individual polymers can (ed), McGraw-Hill, Chap 3:3.1–3.51 (2004).
follow up to four competing pathways: end 6. B. Schartel and T.R. Hull, “Development of fire-
retarded materials - Interpretation of cone calorime-
chain scission (PMMA, PTFE, PS); random ter data”, Fire and Materials, 31 (5), pp. 327–354
chain scission (PE, PP, PS, PA, polyisoprene (2007).
etc.); chain-stripping (PVC, PAN, PVA, cellu- 7. A. Fina and G. Camino, Ignition mechanisms in
lose etc.); and char formation (PAN, PEEK, cel- polymers and polymer nanocomposites, Polym.
Adv. Technol., 22, 1147–1155, (2011)
lulose etc.), with a significant variation from 8. S. L. Madorsky, Thermal Degradation of Organic
individual polymers. Polymers, Interscience, John Wiley, New York
Fire retardants are added to our around a third (1964).
of plastic materials in order to meet regulatory 9. “Plastics – Thermogravimetry (TG) of polymers –
General Principles”, ISO 11358 (1997).
requirements. In general, these apply to high risk 10. P.G. Laye, Differential Thermal Analysis and
applications, such as construction products, Differential Scanning Calorimetry, in Principles of
upholstered furnishings, electrical and electronic Thermal Analysis and Calorimetry, Edited by
goods, and materials for mass transport P.J. Haines, Royal Society of Chemistry, Cambridge,
UK (2002).
applications. There is considerable diversity in 11. ISO 11357–1 to 6:2005–2011 “Plastics - Differential
the different fire retardant mechanisms, from the scanning calorimetry (DSC) - Parts 1–6.
gas phase flame inhibitors, using halogens or 12. M. Reading, A. Luget, and R. Wilson, “Modulated
organophosphorus compounds, to condensed differential scanning calorimetry”, Thermochimica
Acta, 238, pp. 295–307 (1994).
phase processes ranging from intumescents and 13. S. Zhang, T.R. Hull, A.R. Horrocks, G. Smart,
char formation, to endothermic dehydration and B.K. Kandola, J. Ebdon, P. Joseph and B. Hunt:
formation of a refractory shield. Thermal degradation analysis and XRD
Fire behaviour may be quantified on a bench characterisation of fibre-forming synthetic polypro-
pylene containing nanoclay: Polym.Degrad.Stab.,
scale using ease of ignition tests such as the UL 92, 727–732, (2007).
94, or the much criticised limiting oxygen index, 14. A. Witkowski, A.A. Stec and T.R. Hull, The influ-
or using more sophisticated apparatus such as ence of metal hydroxide fire retardants and nanoclay
the cone calorimeter or microscale combustion on the thermal decomposition of EVA, Polym.
Degrad. Stab., 97, 2231–2240, (2012).
calorimeter. The physical properties of polymers 15. M. Sacristan, T.R. Hull, A.A. Stec, J.C. Ronda,
exert an influence on this process, with perhaps M. Galia, and V. Cadiz, “Cone calorimetry studies
the greatest benefit being conferred by char of fire retardant soybean-oil-based copolymers
formation, reducing the rate of thermal attack. containing silicon or boron: Comparison of additive
and reactive approaches,” Polymer Degradation and
Stability, 95, pp. 1269–1274 (2010).
16. I.C. McNeill, L. Ackerman, S.N. Gupta,
M. Zulfiquar, and S. Zulfiquar, “Part A: Polymer
References Chemistry”, Journal of Polymer Science,
15, p. 2381 (1977).
1. ASTM E176, “Standard Terminology of Fire 17. J.P. Lewicki, K. Pielichowski, P.T. De La Croix,
Standards”, in Annual Book of ASTM Standards, B. Janowski, D. Todd, and J.J. Liggat, “Thermal
degradation studies of polyurethane/POSS
nanohybrid elastomers”, Polymer Degradation and determination of solid fuel properties for CFD fire
Stability, 95, pp. 1099–105 (2010). models using bench-scale pyrolysis tests”,
18. ASTM D7309-11, “Standard test method for deter- Proceedings of the Combustion Institute, 33(2),
mining flammability characteristics of plastics and pp. 2599–2606 (2011).
other solid materials using microscale combustion 33. A.K. Galwey and M.E. Brown, “Arrhenius
calorimetry” (2011). parameters and compensation behaviour in solid-
19. R.E. Lyon and R.N. Walters, “Pyrolysis combustion state decompositions”, Thermochimica Acta,
flow calorimetry”, Journal of Analytical and Applied 300, pp. 107–115 (1997).
Pyrolysis, 71, pp. 27–46 (2004). 34. A.V. Nikolaev, V.A. Logvinenko, and
20. S. Bourbigot, M.L. Bras, F. Dabrowski, J.W. Gilman V.M. Gorbatchev, “Special features of the compen-
and T. Kashiwagi, “PA-6 clay nanocomposite hybrid sation effect in nonisothermal kinetics of solid-phase
as char forming agent in intumescent formulations”, reactions”, Journal of Thermal Analysis, 6, pp.
Fire Mater., 24, 201–208, (2000). 473–577 (1974).
21. M.V. Petrova, F.A. Williams, A small detailed 35. A. Matala and S. Hostikka, “Pyrolysis modeling of
chemical-kinetic mechanism for hydrocarbon com- PVC cable materials”, Fire Safety Science, 10, pp.
bustion, Combustion and Flame, 144, 526–544, 917–930 (2011).
(2006). 36. J.H. Flynn, in Encyclopedia of Polymer Science and
22. R. E. Lyon, N. Safronava, and S. I. Stoliarov, The Engineering, ed. H.F. Mark, N.M. Bikales,
Role of Thermal Decomposition Kinetics in the C.G. Overberger, and G. Menges, pp. 690–723
Burning of Polymers. Proceedings of the 12th Inter- (Suppl.), New York, Wiley (1989).
national Conference on Fire Science and Engineer- 37. S. Vyazovkin and C.A. Wight, “Kinetics in Solids”,
ing (INTERFLAM), 2010. Annual Review of Physical Chemistry, 48,125-149
23. L. Reich and S.S. Stivala, Elements of Polymer Deg- (1997).
radation, McGraw-Hill, New York (1971). 38. S. Vyazovkin, and C.A. Wight, “Model-free and
24. K. McGrattan, R. McDermott, and W. Mell, et al., Model-Fitting Approaches to Kinetic Analysis of
“Modeling the burning of complicated objects using Isothermal and Nonisothermal Data”,
Lagrangian particles”, in Conference proceedings of Thermochimica Acta, 340/341, pp. 53–68 (1999).
the twelfth international interflame conference, 39. J.H. Flynn and L.A. Wall, “General Treatment of the
pp. 743–753 (2010). Thermogravimetry of Polymers”, Journal of
25. R.E. Lyon, N. Safronava, and E. Oztekin, “A simple Research of the National Bureau of Standards-A,
method for determining kinetic parameters for Physics and Chemistry, 70A(6), pp. 487–523 (1966).
materials in fire models”, Fire Safety Science, 40. A.K. Galwey and M.E. Brown, “Kinetic Background
10, pp. 765–777 (2011). to Thermal Analysis and Calorimetry”, in Handbook
26. H. L. Friedman, “New methods for evaluating kinetic of Thermal Analysis and Calorimetry, Volume
parameters from thermal analysis data”, Journal of 1, Principles and Practice (M.E. Brown, ed.),
Polymer Science, Polymer Letter 7(1), pp. 41–46 Elsevier, New York, pp. 147–224 (1998).
(1969). 41. H.E. Kissinger, “Variation of Peak Temperature with
27. C. Lautenberger, G. Rein, and C. Fernandez-Pello, Heating Rate in Differential Thermal Analysis”,
“The application of a genetic algorithm to estimate Journal of Research of the National Bureau of
material properties for fire modeling from bench- Standards, 57(4), pp. 217–221 (1956).
scale fire test data”, Fire Safety Journal, 41, pp. 42. H.E. Kissinger, “Reaction Kinetics in Differential
204–214 (2006). Thermal Analysis”, Analytical Chemistry, 29(11),
28. G. Rein, C. Lautenberger, and C. Fernandez-Pello, pp. 1702–1706 (1957).
“Application of genetic algorithms and 43. J.H. Flynn, “Temperature Dependence of the Rate of
thermogravimetry to determine the kinetics of poly- Reaction in Thermal Analysis”, Journal of Thermal
urethane foam in smoldering combustion”, Combus- Analysis, 36, pp. 1579–1573 (1990).
tion and Flame, 146, pp. 95–108 (2006). 44. T. Ozawa, “Kinetic Analysis of Derivative Curves in
29. A. Matala, S. Hostikka and J. Mangs, “Estimation of Thermal Analysis”, Journal of Thermal Analysis,
pyrolysis model parameters for solid materials using 2, pp. 301–324 (1970).
thermogravimetric data”, Fire Safety Science, 9, pp. 45. K.B. McGrattan, B. Klein, S. Hostikka, and J. Floyd,
1213–1223 (2009). “Fire dynamics simulator (version 5) user’s guide”,
30. C. Lautenberger and C. Fernandez-Pello, NIST Special Publication 1019–5, National Institute
“Generalized pyrolysis model for combustible of Standards and Technology, Gaithersburg, MD
solids”, Fire Safety Journal, 44, pp. 819–839 (2009). (2007).
31. C. Lautenberger and C. Fernandez-Pello, “Optimiza- 46. G.R. Heal, “Thermogravimetry and Derivative
tion algorithms for material pyrolysis property estima- Thermogravimetry in Principles of Thermal
tion”, Fire Safety Science, 10, pp. 751–764 (2011). Analysis and Calorimetry”, (P. J. Haines,
32. M. Chaos, M. M. Khan, and N. Krishnamoorthy, ed.), Royal Society of Chemistry, Cambridge, UK
et al., “Evaluation of optimization schemes and (2002).
www.ebook777.com
47. J. Criado, M. Gonzalez, A. Ortega and C. Real, 63. C. Doyle, “Kinetic analysis of thermogravimetric
“Some considerations regarding the determination data”, Journal of Applied Polymer Science, Vol.
of the activation energy of solid-state reactions 5, No. 15, pp. 285–292 (1961).
from a series of isothermal data”, Journal of Thermal 64. S. Vyazovkin and D. Dollimore, “Linear and Non-
Analysis, 29, pp. 243–250 (1984). linear Procedures in Isoconversional Computations
48. F. Rogers and T. Ohlemiller, “Pyrolysis kinetics of a of the Activation Energy of Nonisothermal
polyurethane foam by thermogravimetry; a general Reactions in Solids,” Journal of Chemical Informa-
kinetic method”, Journal of Macromolecular Sci- tion and Computer Sciences, 36, pp. 42–45 (1996).
ence, 1, pp. 169–185 (1981). 65. M.E. Brown, Introduction to thermal analysis:
49. J.H. Sharp and S.A. Wentworth, ”Kinetic analysis of Techniques and applications, Chapter 10, 2nd ed,
thermogravimetric data,” Analytical Chemistry, Kluwer, Amsterdam (2001).
41, pp. 2060–2062 (1969). 66. A.K. Galwey and M.E. Brown, “Thermal decompo-
50. B.N.N. Achar, G.W. Brindley and J.H. Sharp, sition of ionic solids: Chemical properties and
“Kinetics and mechanism of dehydroxylation reactivities of ionic crystalline phases”, Elsevier,
processes,” Proceedings of the International Clay Amsterdam, pp. 139–171 (1999).
Conference, p. 67, Jerusalem (1966). 67. W. Gautschi and W.F. Cahill, “Exponential integral
51. E.S. Freeman and B. Carroll, “The Application of and related functions”, in Handbook of mathematical
Thermoanalytical Decomposition of Calcium Oxa- functions with formulas, graphs, and mathematical
late Monohydrate,” Journal of Physical Chemistry, tables (M. Abramowitz and I. Stegun, eds.), National
62, pp. 394–397 (1958). Bureau of Standards, Washington, DC, pp. 227–237
52. E.S. Freeman and B. Carroll, “Interpretation of the (1964).
kinetics of thermogravimetric analysis,” Journal of 68. S. Vyazovkin, “Evaluation of activation energy of
Physical Chemistry, 73, pp. 751–752 (1969). thermally stimulated solid-state reactions under arbi-
53. A.W. Coats and J.P. Redfern, “Kinetic Parameters trary variation of temperature,” Journal of Computa-
from Thermogravimetric Data,” Nature, 201, pp. tional Chemistry, 18, pp. 393–402 (1997).
68–69 (1964). 69. S. Vyazovkin, “Modification of the integral
54. A.W. Coats and J.P. Redfern, “Kinetic parameters isoconversional method to account for variation in
from thermogravimetric data. II.,” Journal of Poly- the activation energy,” Journal of Computational
mer Science, Part B: Polymer Letter 3, pp. 917–920 Chemistry, 22, pp. 178–183 (2001).
(1965). 70. A. Matala, C. Lautenberger and S. Hostikka,
55. J.P. Elder, “The general applicability of the “Generalized direct method for pyrolysis kinetic
Kissinger equation in thermal analysis,” Journal of parameter estimation and comparison to existing
Thermal Analysis, 30, pp. 657–669 (1985). methods”, Journal of Fire Sciences, 30(4),
56. P. Simon, “Isoconversional methods: Fundamentals, pp. 339–356 (2012).
meaning and application,” Journal of Thermal Anal- 71. R.E. Lyon and R.N. Walters, “Pyrolysis combustion
ysis and Calorimetry, 76, pp. 123–132 (2004). flow calorimetry”, Journal of Analytical and Applied
57. J. Zsako, “Kinetic Analysis of Thermogravimetric Pyrolysis, 71, pp. 27–46 (2004).
Data, VI, Some Problems of Deriving Kinetic 72. T. Ohlemiller, “Modeling of smoldering combustion
Parameters from TG Curves,” Journal of Thermal propagation”, Progress in Energy and Combustion
Analysis, 5, pp. 239–251 (1973). Science, 11, 277–310 (1985).
58. J. Zsako, “Kinetic analysis of thermogravimetric 73. A. Matala, “Estimation of Solid Phase Reaction
data XXIX. Remarks on the ‘many curves’ method,” Parameters for Fire Simulation”, Master’s thesis,
Journal of Thermal Analysis, 46, pp. 1845–1864 Helsinki University of Technology, Finland (2008).
(1996). 74. C. Lautenberger, “A Generalized Pyrolysis Model
59. S. Vyazovkin, “Computational aspects of kinetic for Combustible Solids”, Users’ guide (2009).
analysis.: Part C. The ICTAC Kinetics Project — 75. S.I. Stoliarov, S. Crowley, R.E. Lyon, and
the light at the end of the tunnel?,” Thermochimica G.T. Linteris, “Prediction of the burning rates of
Acta, 355, pp. 155–163 (2000). non-charring polymers”, Combustion and Flame,
60. A. Khawam and D.R. Flanagan, “Role of 156, pp. 1068–1083 (2009).
isoconversional methods in varying activation 76. S.I. Stoliarov, N. Safronava, R.E. Lyon, “The effect
energies of solid-state kinetics. I. Isothermal kinetic of variation in polymer properties on the rate of
studies,” Thermochimica Acta, 429, pp. 93–102 burning,” Fire and Materials, 33, pp. 257–271
(2005). (2009).
61. T. Ozawa, “A new method of analyzing thermogra- 77. C. Lautenberger, E. Kim, N. Dembsey, and C. -
vimetric data,” Bulletin of the Chemical Society of Fernandez-Pello, “The Role of Decomposition
Japan, 38, pp. 1881–1886 (1965). Kinetics in Pyrolysis Modeling – Application to a
62. J.H. Flynn and L.A. Wall, “A quick, direct method Fire Retardant Polyester Composite”, Fire Safety
for the determination of activation energy from Science, 9, pp. 1201–1212 (2008).
thermogravimetric data,” Journal of Polymer Sci- 78. J.E.J. Staggs, “A theory for quasi-steady single-step
ence, Part B: Polymer Letter, 4, pp. 323–328 (1966). thermal degradation of polymers”, Fire and
Materials, 22, 1998, pp. 109–118 (1998).
79. J. Zhang, M.A. Delichatsios, and S. Bourbigot, 94. S.I. Stoliarov, S. Crowley, R.E. Lyon, and
“Experimental and numerical study of the effects of G.T. Linteris, “Prediction of the Burning Rates of
nanoparticles on pyrolysis of polyamide 6 (PA6) Non-Charring Polymers”, Combustion and Flame,
nanocomposite in the cone calorimeter”, Combus- 156, pp. 1068–1083 (2009).
tion and Flame, 156, pp. 2056–2062 (2009). 95. S.S. Rahatekar, M. Zammarano, S. Matko,
80. F. Jia, E.R. Galea, and M.K. Patel, “The numerical K.K. Koziol, M.H. Windle, T. Kashiwagi, and
simulation of the non-charring pyrolysis process and J.W. Gilman, “Effect of Carbon Nanotubes and
fire development within a compartment”, Applied Montmorillonite on the Flammability of Epoxy
Mathematical Modelling, 23, pp. 587–607 (1999). Nanocomposites”, Polymer Degradation and
81. C. Lautenberger, Ph.D. Thesis, University of Stability, 98, pp. 870–879 (2010).
California, Berkeley, CA, USA (2007) available at 96. P. Patel, T.R. Hull, A.A. Stec, and R. E. Lyon, “Influ-
<http://repositories.cdlib.org/cpl/fs/> ence of physical properties on polymer flammability in
82. C. Lautenberger, “Gpyro – A Generalized Pyrolysis the cone calorimeter,” Polymers for Advanced
Model for Combustible Solids” – Users’ Guide Ver- Technologies, 22, pp. 1100–1107 (2011).
sion 0.700 (February 19, 2009). 97. J.G. Quintiere, Principles of Fire Behaviour,
83. GPyro, available at <http://code.google.com/p/ Delmar, Albany, NY (1997).
gpyro/> 98. D. Drysdale, An Introduction to Fire Dynamics, 2nd
84. K.B. Mc Grattan, S. Hostikka, J.E. Floyd, H.R. Baum, Edition, John Wiley & Sons, Chichester (1999).
and R.G. Rehm, “Fire Dynamics Simulator (Version 99. T. Faravelli, G. Bozzano, M. Colombo, E. Ranzi,
5). Technical Reference Guide”, Volume 1: Mathe- M. Dente, Kinetic modeling of the thermal degrada-
matical Model, NIST Special Publication 1018–5, tion of polyethylene and polystyrene mixtures, Jour-
Gaithersburg, MD, (October 2007). nal of Analytical and Applied Pyrolysis,
85. FireFOAM Code, available at <http://code.google. 70, 761–777, 2003.
com/p/firefoam-dev/>. 100. Z. Gao, I. Amasaki, and M. Nakada, ”A thermogra-
86. S.I. Stoliarov and R.E. Lyon, “Thermo-Kinetic vimetric study on thermal degradation of polyethyl-
Model of Burning”, Federal Aviation Administration ene,” Journal of Analytical and Applied Pyrolysis,
Technical Note, DOT/FAA/AR-TN08/17 (2008), 67 (1), pp. 1–9 (2003).
<www.fire.tc.faa.gov/reports/reports.asp> 101. A. Marcilla, A. Gomez, A.N. Garcia, and
87. S.I. Stoliarov and R.E. Lyon, “Thermo-kinetic model M.M. Olaya, “Kinetic study of the catalytic decom-
of burning for pyrolyzing materials”, in Proceedings position of different commercial polyethylenes over
of the Ninth International Symposium on Fire Safety an MCM-41 catalyst”, Journal of Analytical and
Science, pp. 1141–1152 (2009). Applied Pyrolysis, 64, pp. 85-101(2002).
88. L. Bustamante Valencia, Experimental and 102. C.F Cullis and M.M Hirschler, The combustion of
Numerical Investigation of the Thermal Decomposi- organic polymers, New York, NY: Oxford Univer-
tion of Materials at Three Scales: Application to sity Press (1981).
Polyether Polyurethane Foam used in Upholstered 103. S.L. Madorsky, "Thermal degradation of organic
Furniture, Ph.D. Thesis, ENSMA, Poitiers, France polymers”, Interscience Publishers, A Division of
(2009). John Wiley & Sons Inc. (1964).
89. C.R. Houck, J.A. Joines, and M.G. Kay, “GAOT: A 104. S.M. Thornberg, R. Bernstein, D.K. Derzon,
Genetic Algorithm for Function Optimization: A A.N. Irwin, S.B. Klamo, and R.L. Clough, “The
Matlab Implementation”, Report NCSU-IE TR genesis of CO2 and CO in the thermooxidative deg-
95–09 (1995), available at <http://www.ise.ncsu. radation of polypropylene”, Polymer Degradation
edu/kay/> and Stability, 92, pp. 94–102 (2007).
90. Q. Duan, V.K. Gupta and S. Sorooshian, “A shuffled 105. R. Bernstein, S.M. Thornberg, R.A. Assink,
complex evolution approach for effective and A.N. Irwin, J.M. Hochrein, J.R. Brown,
efficient global minimization,” Journal of Optimiza- D.K. Derzon, S.B. Klamo, and R.L. Clough, “The
tion Theory and Applications, 76, pp. 501–521 origins of volatile oxidation products in the thermal
(1993). degradation of polypropylene, identified by selective
91. Q. Duan, S. Sorooshian and V.K. Gupta, “Optimal isotopic labelling,” Polymer Degradation and Sta-
Use of the SCEUA Global Optimization Method for bility, 92, pp. 2076–2094 (2007).
Calibrating Watershed Models,” Journal of Hydrol- 106. G.G. Cameron, W.A.J. Bryce, I.T. McWalter, “Ther-
ogy, 158, pp. 265–284 (1994). mal degradation of polystyrene-5. Effects of initiator
92. N. Bal, Uncertainty and complexity in pyrolysis residues,” European Polymer Journal, 20, pp.
modelling, PhD Thesis, University of Edinburgh, 563–569 (1984).
UK (2012), available at http://www.era.lib.ed.ac. 107. N. Grassie and G. Scott, Polymer Degradation and
uk/handle/1842/6511 Stabilisation, Cambridge University Press,
93. S.I. Stoliarov and R.E. Lyon, “Thermo-Kinetic Cambridge, UK (1985).
Model of Burning”, Federal Aviation Administration 108. W.R. Zeng, S.F Li, and W.K. Chow, “PMMA
Technical Note DOT/FAA/AR-TN-08/17 (2008). Review on Chemical Reactions of Burning Poly
www.ebook777.com
(methylmethacrylate)”, Journal of Fire Sciences, 125. W.D. Woolley, “Decomposition Products of PVC
20, p. 401 (2002). for Studies of Fires”, British Polymer Journal, 3(4),
109. I.C. McNeill and A. Rincon, “Thermal degradation pp. 186–193 (1971).
of polycarbonates: Reaction conditions and reaction 126. . W.D. Wolley, “Studies of the dehydrochlorination
mechanisms,” Polymer Degradation and Stability, of PVC in nitrogen and air”, Building Research
39, pp. 13–19 (1993). Establishment, Current Paper CP 9/74 (1974).
110. . A. Davis and J.H. Golden, J. Macromol. Scie. Rev. 127. Purser, D.A., Fardell, P.J., Rowley, J., Vollam,
Macromol. Chem. C, 3, p. 49 (1969). S. and Bridgeman, B. An improved tube furnace
111. S.C. Moldoveanu, “Analytical Pyrolysis of Synthetic method for the generation and measurement of
Organic Polymers”, Techniques and Instrumentation toxic combustion products under a wide range of
in Analytical Chemistry, Volume 25, 1st Edition, fire conditions. Proceedings of the 6th International
Elsevier (2005). Conference Flame Retardants ‘94, London, UK
112. S. Smith, “The re-equilibration of polycaproamide,” (26–27 Jan 1994). Interscience Communications.
Journal of Polymer Science, 30, pp. 459–478 (1958). 128. K.T. Paul, “Feasibility Study to Demonstrate the
113. L.H. Buxbaum, “The degradation of poly(ethylene Potential of Smoke Hoods in Simulated Aircraft
terephthalate),” Angewandte Chemie International Fire Atmospheres: Development of the fire model”,
Edition, 7, pp. 182–190 (1968). Fire and Materials, 14, pp. 43–58, (1989).
114. S.V. Levchik and E.D. Weil, “A review on thermal 129. K. Lebek, T.R. Hull, and D. Price, “Products of
decomposition and combustion of thermoplastic burning rigid PVC burning under different fire
polyesters”, Polymers for Advanced Technologies, conditions Fire and Polymers”, Materials and
15, pp. 691–700 (2004). Concepts for Hazard Prevention, ACS Symposium
115. T.R. Hull, A.A. Stec, and S. Nazare, “TGA-FTIR Series No. 922, Oxford University Press, p. 334–347
Investigation of The Fire Retardant Mechanism of (2005).
Acrylonitrile Copolymers Containing Nanofillers,” 130. T.R. Hull, A.A. Stec, and K.T. Paul, Proceedings of
in 235th American Chemical Society National the 9th International Symposium on Fire Safety Sci-
Meeting, APR 06–10, New Orleans, LA (2008). ence, 665–676 (2008).
116. Z. Bashir, “A critical review of the stabilisation of 131. H.F. Mark, N. Bikales, C.G. Overberger, and
polyacrylonitrile,” Carbon, 29, pp. 1081–1090 (1991). J.I. Kroschwitz, eds., Encyclopedia of Polymer Sci-
117. A.R. Horrocks, J. Zhang and M.E. Hall, “Flammabil- ence and Engineering, Wiley Europe, vol 1–4
ity of polyacrylonitrile and its copolymers II. Thermal (1989).
behaviour and mechanism of degradation,” Polymer 132. E.E. Lewis and M.A. Naylor, “Pyrolysis of
International, 33, pp. 303–314 (1994). Polytetrafluoroethylene”. Journal of the American
118. N. Grassie, Developments in polymer degradation, Chemical Society, 69, p. 1968–70 (1947).
Applied Science, Vol. 1, p. 137, London (1977). 133. A. Stec and R. Hull, Fire Toxicity, Woodhead Pub-
119. E. Fitzer and D. Muller, “The influence of oxygen on lishing, Cambridge, 2010.
the chemical reactions during stabilization of PAN 134. E. Ackerman, Firestopping Through-Penetrations,
as carbon fiber precursor,” Carbon, 13, p. 63–69 in Science and Technology of Building Seals,
(1975). Sealants, Glazing, and Waterproofing: Seventh Vol-
120. L.T. Memetea, N.C. Billingham, and E.T.H. Then, ume (J.M. Klosowski, ed.), ASTM STP 1334, Amer-
“Hydroperoxides in polyacrylonitrile and their role ican Society for Testing and Materials, West
in carbon-fibre formation,” Polymer Degradation Conshohocken, PA (1998).
and Stability, 47, pp. 189–201 (1995). 135. J. Harris, A. Stevenson, “On the role of nonlinearity
121. N. Grassie, J.N. Hay and I.C. McNeill, “Coloration in the dynamic behavior of rubber components”,
in acrylonitrile and methacrylonitrile polymers,” Rubber Chemistry and Technology, 59 (5),
Journal of Polymer Science, 31, p. 205 (1958). pp. 740-764 (2011).
122. J. Brandrup and L.H. Peebles, “On the chromophore 136. D.J. Kind and T.R. Hull, “A review of candidate fire
of polyacrylonitrile. IV. Thermal oxidation of retardants for polyisoprene,” Polymer Degradation
polyacrylonitrile and other nitrile-containing and Stability, 97, pp. 201–213 (2012).
compounds”, Macromolecules, 1, 64–72, (1968). 137. D.W. Brazier and G.H. Nickel, “Thermoanalytical
123. M.A. Geiderikh, B.E. Davydov, B.A. Krentsel, methods in vulcanizate analysis. Derivative
I.M. Kustanovich, L.S. Polak, A.V. Topchiev, and thermogravimetric analysis”, Rubber Chemistry
R.M. Voitenko, “Preparation of polymeric materials and Technology, 48 (4), pp. 661–677 (1975).
with semiconductor properties,” Journal of Polymer 138. A.K. Sircar, “Identification of natural and synthetic
Science, 54, pp. 621–626 (1961). polyisoprene vulcanizates by thermal analysis”,
124. S.C. Martin, J.J. Liggat and C.E. Snape, “In situ Rubber Chemistry and Technology., 50 (1),
NMR investigation into the thermal degradation pp. 71–82 (1977).
and stabilisation of PAN,” Polymer Degradation 139. S. Straus and S.L. Madorsky, “Thermal Degradation
and Stability, 74, pp. 407–412 (2001). of Unvulcanized and Vulcanized Rubber in a
Vacuum”, Industrial and engineering chemistry, 48 154. F. D. Kopinke, M. Remmler, K. Mackenzie, Thermal
(7), pp. 1212–1219 (1956). decomposition of biodegradable polyesters-I: Poly
140. F. Cataldo, “Thermal depolymerization and pyroly- (hydroxybutyric acid). Polym. Degrad. Stab.,
sis of cis-1,4-polyisoprene: preparation of liquid 52, 25–38, 1996.
polyisoprene and terpene resin”, Journal of Analyti- 155. H. Morikawa, R.H. Marchessault, Pyrolysis of bac-
cal and Applied Pyrolysis, 44(2), pp. 121–130 terial polyalkanoates, Canadian Journal of Chemis-
(1998). try 59, 2306,1981
141. S.V. Levchik and E.D. Weil, “Thermal decomposi- 156. J.L. Gay-Lussac, Ann. Chim. Phys., 18, p. 211 (1821).
tion, combustion and flame-retardancy of epoxy 157. SRI Consulting, Report on Flame Retardants,
resins: a review of the recent literature,” Polymer Published December 2008
International, 53, pp. 1901–1929 (2004). 158. T.R. Hull, A. Witkowski, L.A. Hollingbery, “Fire
142. S.C. Lin, B.J. Bulkin and E.M. Pearce, “Thermal retardant action of mineral fillers”, Polymer Degra-
Degradation Study Of Phenolphthalein Polycarbon- dation and Stability, 96, pp. 1462–1469 (2011).
ate”, Journal of polymer science, Part A-1, Polymer 159. A. Bergman, A. Ryden, R.J. Law, J. de Boer,
chemistry, 19, 2773–2797, (1981). A. Covaci, M. Alaee, L. Birnbaum, M. Petreas,
143. B.C. Levin, M. Paabo, J.L. Gurman and S.E. Harris, M. Rose, S. Sakai, N. Van den Eede and I. van der
“Effects of exposure to single or multiple Veen, “A novel abbreviation standard for organo-
combinations of the predominant toxic gases and bromine, organochlorine and organophosphorus
low oxygen atmospheres produced in fires” Toxico- flame retardants and some characteristics of the che-
logical Sciences, 9, 236–250 (1987). micals” Environment International, 49, 57–82, (2012).
144. D.A. Purser, Asphyxiant components of the fire 160. A. Schnipper, L. Smith-Hansen, and S.E. Thomsen,
effluents, in Fire Toxicity, (A.A Stec and T.R. Hull, “Reduced Combustion Efficiency of Chlorinated
eds.), Woodhead Publishing, Cambridge (2010). Compounds Resulting In Higher Yields of Carbon
145. J. Wang, H. Jiang and N. Jiang, Study on the Monoxide”, Fire and Materials, 19, pp. 61–64, (1995).
pyrolysis of phenol-formaldehyde (PF) resin and 161. V. Babushok, W. Tsang, G.T. Linteris, and
modified PF resin. Thermochimica Acta, 2009, 496, D. Reinelt, “Chemical Limits to Flame Inhibition”,
136–142 Combustion and Flame, 115, pp. 551–560 (1998).
146. A. Murari and A. Barzon, “Comparison of New 162. M.I. Nelson and J. Brindley, “Polymer combustion:
PEEK Seals with Traditional Helicoflex for Ultra Effects of flame emissivity” Philosophical
High Vacuum Applications”, Vacuum, Volume Transactions of the Royal Society A: Mathematical,
72, Issue 3, pp. 327–334 (2003). Physical and Engineering Sciences, 358, 3655–3673
147. S.K. Yesodha, C.K.S. Pillai, and N. Tsutsuni, (2000).
“Stable Polymeric Materials for Non-Linear Optics: 163. J.G. Quintiere, Principles of Fire Behaviour, Delmar
A Review Based on Azobenzene Systems”, Progress Publishers, New York, USA, (1997).
in Polymer Science, Volume 29, Issue 1, pp. 45–74 164. . H. Zhang, Fire-Safe Polymers and Polymer
(2004). Composites, US Department of Transport. Report
148. M.P. Stevens, Polymer Chemistry: An Introduction, Number: DOT/FAA/AR-04/11 (2004).
Third Edition. Oxford University Press, New York, 165. P. Patel, T.R Hull, and Colin Moffatt, “PEEK poly-
USA (1999). mer flammability and the inadequacy of the UL‐94
149. M.C. Kuo, C.M. Tsai, J.C. Huang, and M. Chen, classification,” Fire and Materials, 36, pp. 185–201
“PEEK Composites Reinforced by Nano-Sized (2012).
SiO2 and Al2O3 Particulates”, Materials Chemistry 166. V. Babrauskas, “Fire Test Methods for Evaluation of
and Physics, Volume 90, pp. 185–195 (2005). Fire-Retardant Efficacy in Polymeric Materials”,
150. L.H. Perng, C.J. Tsai, and Y.C. Ling, “Mechanism Chapter 3, in Fire Retardancy of Polymeric
and Kinetic Modelling of PEEK Pyrolysis by Materials (A.F. Grand and C.A. Wilkie, eds.), CRC
TG/MS”, Polymer, Volume 40, pp. 731–732 (1999). Press, New York, USA (2000).
151. P. Patel, T. R. Hull, R. W. McCabe, D. Flath, 167. K.T. Paul and S.D. Christian, “Standard flaming
J. Grasmeder, and M. Percy, Mechanism of ignition sources for upholstered composites, furni-
thermal decomposition of poly(ether ether ketone) ture and bed assembly,” Journal of Fire Sciences, 5
(PEEK) from a review of decomposition studies, (3), pp. 178–211 (1987).
Polymer Degradation and Stability, 95, pp. 168. D. Hopkins Jr and J.G. Quintiere, “Materials fire
709–718 (2010). properties and predictions for thermoplastics”, Fire
152. A.-M.M. Baker and J. Mead, Thermoplastics, Safety Journal, 26, pp. 241–268 (1996).
Chapter 1, In C.A. Harper, Modern Plastics Hand- 169. D.J. Rasbash, D.D. Drysdale, and D. Deepak, “Criti-
book, McGraw-Hill Professional Publishing, Ohio, cal heat and mass transfer at pilot ignition and
USA (2000). extinction of a material”, Fire Safety Journal,
153. R.E. Lyon and M.L. Janssens, Polymer Flammabil- 10, pp. 1–10 (1986).
ity, US Department of Transport, Report Number: 170. H.E. Thomson, D.D. Drysdale, and C.L. Beyler, “An
DOT/FAA/AR-05/14 (2005). experimental evaluation of critical surface
www.ebook777.com
temperatures as a criterion for piloted ignition of 186. B. Schartel and T.R. Hull, “Application of Cone
solid fuels”, Fire Safety Journal, 13, pp. 185–196 Calorimetry to the Development of Materials with
(1988). Improved Fire Performance”, Fire and Materials,
171. E. Mikkola and I.S. Wichman, “On the thermal igni- 31, pp. 327–354 (2007).
tion of combustible materials”, Fire and Materials, 187. J.G. Quintiere, B.P. Downey, and R.E. Lyon,
14, pp. 87–96 (1989). “An Investigation of the Vertical Bunsen Burner
172. T. Kashiwagi, “Radiative ignition mechanism of Test for Flammability of Plastics”, US Department
solid fuels”, Fire Safety Journal, 3, pp. 185–200 of Transport, Report Number: DOT/FAA/AR-TN
(1981). (2010).
173. V. Babrauskas, Ignition Handbook, Fire Science 188. ISO 4589–2 “Plastics – Determination of burning
Publishers, Issaquah WA, USA and SFPE, USA behaviour by oxygen index – Part-2: Ambient tem-
(2003). perature test”, (1996).
174. R.E. Lyon, “Plastics and Rubber”, in Handbook of 189. ISO 5660–2 “Reaction-to-fire tests – Heat release,
Building Materials for Fire Protection, (C.A. Harper, smoke production and mass loss rate – Part 2:
ed), McGraw-Hill, chap 3, 3.1-3.51 (2004). Smoke production rate (dynamic measurement)”,
175. R.E. Lyon, R.N. Walters, and S.I. Stoliarov, “Ther- (2002).
mal Analysis of Polymer Flammability”, Presented 190. B. Schartel and T.R. Hull, “Application of cone
at 228th ACS Meeting Philadelphia (2004). calorimetry to the development of materials with
176. A. Tewarson, “Generation of Heat and Chemical improved fire performance”, Fire and Materials,
Compounds in Fires”, in The SFPE Handbook of 31, pp. 327–354 (2007).
Fire Protection Engineering, 3rd edition 191. R.E. Lyon, in Recent Advances in Flame Retardancy
(P.J. DiNenno, D.D. Drysdale, C.L. Beyler, of Polymers, vol. 13, (M. Lewin, ed.), BCC, Inc.,
W.D. Walton, R.L.P Custer, J.R. Hall Jr and pp. 14-25 (2002)
J.M. Watts Jr, eds), National Fire Protection Associ- 192. R.E. Lyon and R.N. Walters, “Pyrolysis combustion
ation, Inc., chap 3.4,3-82-3-161 (2002). flow calorimetry”, Journal of Analytical and Applied
177. M. Sibulkin and M.W. Little, “Propagation and Pyrolysis, 71, pp. 27–46 (2004).
extinction of downward burning fires”, Combustion 193. B. Schartel, K.H. Pawlowski, and R.E. Lyon, “Pyrol-
Flame, 31, pp. 197–208 (1978). ysis combustion flow calorimeter: A tool to assess
178. IEC 60695-11-10 “Fire hazard testing - Part 11–10: flame retarded PC/ABS materials?”, Thermochimica
Test flames - 50 W horizontal and vertical flame test Acta, 462, pp. 1–14 (2007).
methods,” (1999). 194. R.E. Lyon, R.N. Walters, M. Beach, and F.P. Schall,
179. V. Babrauskas, “Ignition: A Century of Research and “Flammability Screening of Plastics Containing Flame
an Assessment of our Current Status”, Journal of Retardant Additives”, ADDITIVES 2007, 16th Inter-
Fire Protection Engineering, 17(3), pp. 165–183 national Conference, San Antonio, TX (2007).
(2007). 195. D.W. Van Krevelen, Properties of Polymers.
180. ISO 5660–1 “Fire tests – Reaction to fire – Part 1: Chapter 21 – Thermal Decomposition, 4th
Rate of heat release from building products (cone Edition, Elsevier Science Publishers, Amsterdam
calorimeter method)”, (1993). (2009).
181. A.B. Morgan and M. Bundy, “Cone Calorimeter 196. R. Walters and R.E. Lyon, Calculating Polymer
Analysis of UL-94 V-Rated Plastics”, Fire and Flammability from Molar Group Contributions,
Materials, 31, pp. 257–283 (2007). DOT/FAA/AR-01/31 (2001).
182. Y. Wang, F. Zhang, X. Chen, Y. Jin, and J. Zhang, 197. P. Patel, Doctoral Thesis, University of Central Lan-
“Burning and Dripping Behaviours of Polymers cashire, UK (2011).
under the UL-94 Vertical Burn Test Conditions”, 198. H. Zhang, Fire-Safe Polymers, and Polymer
Fire and Materials, 34, pp. 203–215 (2009). Composites, US Department Of Transport, Report
183. M. Bundy and T. Ohlemiller, “Bench-Scale Flam- Number: DOT/FAA/AR-04/11, Federal Aviation
mability Measures for Electronic Equipment”, Administration (2004).
National Institute of Standards and Technology, 199. R.E. Lyon and M.L. Janssens, Polymer Flammabil-
NISTIR 7031 (2003). ity, US Department of Transport, Report Number:
184. S. Hong, J. Yang, S. Ahn, Y. Mun, and G. Lee, DOT/FAA/AR-05/14 (2005).
“Flame Retardant Performance of Various 200. P. Patel, T.R. Hull, R.E. Lyon, S.I. Stoliarov,
UL-94 Classified Materials Exposed to External R.N. Walters, S. Crowley, and N. Safronava, “Inves-
Ignition Sources”, Fire and Materials, 28, pp. tigation of the Thermal Decomposition and Flamma-
25–31 (2004). bility of PEEK and its Carbon and Glass-Fibre
185. B. Schartel and U. Braun, “Comprehensive Fire Composites”, Polymer Degradation and Stability,
Behaviour Assessment of Polymeric Materials In Press (2011).
Based on Cone Calorimeter Investigations”, 201. R.E. Lyon, “Solid-State Thermochemistry of
e-Polymers, Article 13, pp. 1–14 (2003). Flaming Combustion,” in Fire Retardancy of
Polymeric Materials (C.A. Wilkie and A.F Grand, 208. R.N. Walters and R.E. Lyon, “Calculating Polymer
eds.), Marcel Dekker, Inc., NY (2000). Flammability from Molar Group Contributions,”
202. R.E. Lyon, “Heat Release Capacity,” Proceedings Proceedings of the BCC Conference on Flame
of the 7th International Conference on Fire and Retardancy of Polymeric Materials, Stamford, CT
Materials, San Francisco, CA, pp. 285–300 (2001). (May 22–24, 2000).
203. R.E. Lyon, “Heat Release Kinetics,” Fire and
Materials, 24, pp. 179–186 (2000).
204. R.N. Walters and R.E. Lyon, “A Microscale Dr Witkowski is a Lecturer at the University of Central
Combustion Calorimeter for Determining Flamma- Lancashire (UCLan), UK. His work focuses on thermal
bility Parameters of Materials,” Proceedings 42nd decomposition of solids and pyrolysis mechanisms, and
International SAMPE Symposium and Exhibition, numerical modelling.
42(2), pp. 1335–1344 (1997).
205. R.N. Walters and R.E. Lyon, “A Microscale Com- Dr Stec is an Associate Professor in Fire Chemistry and
bustion Calorimeter for Determining Flammability Toxicity at UCLan. Her work focuses on quantification
Parameters of Materials,” NISTIR 5904 (K. Beall, of combustion products and the factors affecting fire
ed.), pp. 89–90 (1996). toxicity from bench- and largescale tests.
206. R.E. Lyon and R.N. Walters, U.S. Patent
5981290, Microscale Combustion Calorimeter, Prof Hull is a Professor of Chemistry and Fire Science at
11/09/1999. the University of Central Lancashire (UCLan). He
207. R.N. Walters and R.E. Lyon, “Molar Group obtained his PhD in 1987 in Fire Retardant Mechanisms
Contributions to Polymer Flammability,” PMSE from the University of Salford, UK. His current research
Preprints, 83, 86, ACS National Meeting, interests include the development of fire retardant
Washington, D.C. (August 2000). materials and the assessment of fire toxicity.
www.ebook777.com
Structural Mechanics
8
Luke A. Bisby
Introduction (ambient) temperature conditions takes little

(if any) explicit account of the possible effects
Structural mechanics, sometimes called ‘solid of fire; it typically consists of the following steps:
mechanics’ or ‘mechanics of materials’ is 1. The architect and the engineer establish the
concerned with describing the behavior of struc- aesthetic, general structural layout, and func-
tural members under loading, as occurs in all tional requirements for the building.
buildings and other structures due to the effects of 2. The overall structural framing system and
gravity and other forces (e.g. wind, earthquake, building materials are selected and a structural
etc.). A detailed understanding of structural concept is proposed, based on the competing
mechanics is essential for anyone seeking to per- interests of architectural, functional, eco-
form structural fire engineering analysis or design. nomic, and sustainability considerations. At
It is not possible within this brief chapter to this stage only approximate sizes of structural
provide a complete treatment of the topic; how- elements are known.
ever when reviewed in conjunction with Chap. 9 3. The likely loads which will act on the struc-
of this handbook, the current chapter provides a ture are estimated in accordance with struc-
basic description of structural mechanics as is tural engineering principles (discussed
required for an initial understanding of the below), with due consideration given to all
means by which structural stability, and to a cer- possible loads and their likelihood of acting
tain extent integrity, against fire spread and insu- (on their own or in combination with other
lation during fire, are engineered through careful loads).
selection and design of building materials. 4. A structural analysis is performed to deter-
mine the load path by which all loads are
transferred through the structure from the
Philosophy of Structural Design location where they act and into the
structure’s foundations. The internal forces
Structural design is both a creative art and a and stresses acting within the various struc-
science, and structural engineers may use consid- tural elements are subsequently determined.
erable creativity in determining the load bearing 5. The likely stresses and forces acting in each
system for a particular building. In general, the structural element (the design loads) are
process of structural design under normal checked against the capacities of the respec-
tive structural elements (the design
L.A. Bisby (*)
resistances) to ensure that the structural
School of Engineering, University of Edinburgh, UK,
The King’s Buildings, Mayfield Road, Edinburgh, elements have sufficient strength and stiffness
UK EH93JL to (a) resist the applied loads without collapse

256 L.A. Bisby
(called ultimate design considerations), and characterized worst case credible loads to be
(b) provide a suitable level of in-service per- assumed in the particular jurisdiction in which
formance of the structure (called serviceabil- the design code is in force. The most common
ity considerations; for instance deflections, loads for which typical structures are designed
vibrations, durability, etc.). are given below.
6. Steps 3–5 are repeated as necessary, until the 1. Dead Loads: These are loads which are
structural elements satisfy both ultimate and always present, and include the self weight
serviceability design requirements and the of the structure as well as loads arising from
building is deemed to satisfactorily achieve permanent fixtures and equipment. Dead
the functional requirements set out in the loads may include the weight of floor
early stages of the design process. coverings, walls, doors, suspended ceilings,
It is clear from the above steps that the struc- etc., and are usually estimated based on the
tural design process fundamentally involves a dimensions and construction materials of the
comparison between the loads acting on and trial structure under analysis.
within the structure, and the resistances or 2. Live Loads: Sometimes called Imposed Loads,
capacities of the structural elements from which these are typically specified by design codes
the structure is made. This chapter focuses on on the basis of data obtained from surveys of
how a structural element’s resistance to load can real buildings and account for the weight of
be determined using basic structural mechanics; people and moveable fixtures and equipment.
however a description of the means by which the It is important to recognize that live loads are
loads acting on a structure are typically estimated likely to be variable throughout the life of a
by structural engineers is instructive and is briefly structure, and thus the specified values of live
treated first. loads given in codes may be considerably
higher than those which are actually experi-
enced on a day-to-day basis.
Structural Design at Ambient 3. Snow Loads: As the name implies, these are
Temperature loads due to the weight of snow and ice which
can accumulate on structures in cold climates.
To understand the goals of structural fire engi- Snow loads are estimated based on geo-
neering and the means by which these goals are graphic and climatic data which has been col-
met during design it is first necessary to under- lected and calibrated over many decades. In
stand the general framework through which some design codes loads due to snow may be
structures are designed to resist the full suite of treated as live loads.
potential loads to which they might be subjected 4. Wind Loads: Wind loads may cause lateral
during their lifetime, as well as the probabilistic forces which act on the vertical surfaces of a
basis of this framework which is intended to building, but may also cause uplift on hori-
provide a suitably low probability of failure. zontal surfaces such as roofs and slabs. Wind
loads are highly variable and again are treated
in design codes using empirical correlations
Loads and Load Combinations based on geography, topography, and the
form of the building. Wind loads are particu-
Throughout a structure’s lifetime it will be larly important during the design of tall
subjected to a wide variety of loads. During the buildings, for which lateral load resistance
structural design process it is essential that all often governs the design of the overall struc-
credible loading scenarios be considered and tural system. Modern tall building design typ-
addressed. Loads to be used in structural design ically includes complex wind tunnel tests to
are typically specified in design codes, which determine the possible distributions of wind
provide empirically determined and statistically pressures over the building’s surface given its
www.ebook777.com
8 Structural Mechanics 257
geometry and the surrounding climate and only those loads noted previously might be act-
topography. ing, include:
5. Seismic Loads: In many parts of the world,
1:4D
loads arising form both vertical and horizontal
ground acceleration during earthquake must 1:2D þ 1:6L þ 0:5S
be considered during design. Design earth- 1:2D þ 1:6S þ ðL or 0:8W Þ
quake loads are given in design codes and ð8:1Þ
1:2D þ 1:6W þ L þ 0:5S
account for geography, probability and mag-
nitude of possible earthquakes, soil 1:2D þ E þ L þ 0:2S
conditions, etc. It is worth noting that the
statistical likelihood of a fire and earthquake
occurring simultaneously is very low. and so on, where:
It is clear that there is uncertainty associated D ¼ dead load;
with the likely magnitude of each of the different L ¼ live load;
loads, and also with the likelihood that each of S ¼ snow load;
the loads might be acting at its full (or some W ¼ wind load; and
lesser) value at any given point in time during E ¼ earthquake load.
the life of a structure. For instance, the self The various load combinations are based on
weight of a structure, once designed, is the philosophy that, under any given set of
reasonably well known and can be assumed to circumstances, the worst case loads on a structure
always be acting, whereas the weight of the peo- will be described by one of these combinations of
ple in a structure has large variability and may loads—with a selected level of statistical confi-
not ever be known with any degree of certainty dence as described in the following sections.
either spatially or temporally. Furthermore, it is
highly unlikely that all of the noted loads will be
acting at their full value at any given time (i.e. the Working Stress Design
chances are low that a building will be
completely full of people, in the middle of win- Once the worst case load acting on a structural
ter, with the wind blowing a gale, and during an element at any given instant is determined using
earthquake). the procedures described above, the structural
Most modern building codes deal with the engineer must assess whether or not the perfor-
uncertainty around loading using a series of load mance of the candidate structural element design
combinations which seek to statistically account is satisfactory under that load. There are a num-
for the variability in magnitude and occurrence of ber of means by which this can be accomplished
the respective loads when acting in combination. so as to ensure a reasonable level of confidence
These load combinations help engineers to decide that the design will not fail. Most modern struc-
which combinations of loads they must consider tural design codes use a procedure which is
in designing a structure; any given structure may called Load and Resistance Factor Design
need to be checked under a variety of potential (LRFD), or in some codes Limit States Design
load combinations to determine the worst possi- (described in detail in the next section). How-
ble case which must be used in designing the ever, some older codes still use an approach
individual structural elements. called Working Stress or Allowable Stress
As an illustrative example, some of the Design.
loading combinations required by The Ameri- It is important to recognize that all building
can Society of Civil Engineers’ Minimum materials have their own distinct response under
Design Loads for Buildings and Other loading (refer to Chap. 9), and that this specific
Structures (ASCE-7-05) [1], assuming that response under loading profoundly influences
258 L.A. Bisby
their ability to resist deformation and eventual Limit States Design

failure. The key parameters of interest are a
material’s strength (usually described in terms Limit States Design is now the preferred method
of its failure stress) and its resistance to deforma- of design in most national building codes, largely
tion (usually referred to as its stiffness). Strength because it (1) removes some of the unnecessary
and stiffness are described in the following conservatism which is inherent in Working
sections, but it should be noted that there are Stress Design and (2) attempts to rationally
differing amounts of uncertainty associated with assess and account for the statistical variability
both of these properties—for all building of both the loads acting on a structure and the
materials. For instance, timber, which is a natu- resistance of the structural elements, including
rally occurring material with associated defects, variability associated with material response.
has a relatively high level uncertainty associated Limit States Design accomplishes this by apply-
with its mechanical properties, since these ing reliability concepts to both loads and
depend on numerous factors including the spe- resistances such that a consistent level of safety
cies of tree, the climate in which the tree was or safety index is achieved for all designs.
grown, the grade of lumber, the in-service As the name implies, Limit States Design uses
humidity condition, etc., whereas structural a variety of so-called ‘limit states’ which repre-
steel has relatively little uncertainty since it is sent the functional requirements for a structure.
manufactured under well controlled factory Ultimate Limit States (ULS) are those associated
conditions and is relatively insensitive to humid- with structural failure or collapse, and are
ity, etc. LRFD and Working Stress design pro- addressed by checking the capacity of the struc-
vide alternative means by which this variability tural element, with material and member
and uncertainty in material response to loading strengths artificially (statistically) reduced to
can be accounted for during design. account for known variability in material
In Working Stress Design, the loads expected properties, errors and uncertainties in construc-
to be acting on the structure during service are tion, etc., against the credible worst case loads
compared against the permissible stress levels which might act on the structure. The most likely
which are considered safe for the structural (mean) loads are artificially increased to account
elements under long term loads. The loads to be for their spatial and temporal variability. Service-
considered in Working Stress design are deter- ability Limit States (SLS) are those associated
mined based on guidance given in building with the in-service performance of the structure
codes, and are intended to represent a conserva- and are checked against the loads assumed to be
tive estimate of the most likely in-service loading acting in service. Since serviceability limit states
on the structure. The analysis of the structure is are not associated with life safety, the service
subsequently performed under these loads, and loads need not be unduly increased during
the stresses in the structural elements are calcu- design.
lated (using principles presented later). The The design checks which are made in Limit
resulting stresses are compared against the allow- States Design can be expressed in general as:
able stresses for the materials in question; these
are also specified in building codes. αE ϕR ð8:2Þ
Working stress design loads and allowable
stresses have been calibrated over time to pro- where:
vide safe designs by implicitly building a rela- E ¼ the specified effect of loads acting on the
tively large safety factor into the allowable stress structure;
values specified in codes. These methods have α ¼ load factors applied to the specified loads
now fallen out of favour in most jurisdictions, which take into account the variability of
and all modern building codes are moving the load and load patterns and, to some
towards a Limit States design approach. extent, inaccuracy in the structural analysis;
www.ebook777.com
R ¼ the calculated resistance of a member based build from steel, the resistance factor is typically
on specified material properties and cross- reduced somewhat to ensure that member failure
sectional dimensions; and occurs before connection failure; this is preferred
ϕ ¼ the resistance factor applied to the calculated because connection failures can occur with little
resistance or to specified properties and warning whereas member failures generally give
dimensions, workmanship, type of failure some warning prior to failure. Concrete tends to
(e.g. brittle versus ductile) and uncertainty have lower specified material resistance factors
in the prediction of resistance. in the range of 0.60–0.65 to reflect its higher
variability; designers and codes are statistically
less confident of its nominal strength.
Serviceability Limit States
For the serviceability limit states, the factored Safety Index

resistance on the right side of Equation 8.1 is
replaced with a serviceability criterion such as In Limit States Design, the values of both the
an allowable deflection, acceleration, etc. The load factors, α, and the material or member resis-
load factors, α, are assigned different values, as tance factors, ϕ, in a given building code have
described for example in ASCE-7-05 (ASCE been calibrated to provide the desired level of
2005); typically close to or less than 1.0. safety (or rather to give an acceptable probability
of failure). This is generally accomplished using
the concept of a safety index, β.
Material or Member Resistance Factors In reality, both E and R in Equation 8.1 are
random statistical variables with an associated
As indicated in the definitions of ϕ and R, for probability distribution about a mean value.
ultimate strength design the resistance, R, is This is shown schematically in Fig. 8.1, where
scaled by a resistance factor, ϕ, which is typi- the probability that either the load, E, or the
cally less than 1.0 to reflect the probability that resistance, R, take on given values is plotted.
the full theoretical value of R may not be Clearly, the mean resistance must be greater
achieved at all times (i.e., in some cases the than the mean load effect to prevent failure;
structural members and materials may not be as however, because both load and resistance are
strong as we calculate them to be based on nomi- probabilistic in nature there is always a small
nal material properties, dimensions, tolerances, chance of failure (occurring when then resistance
and construction qualities). This results from a is less than the applied load). The probability of
statistical consideration of the likely ability of a failure is represented by the shaded overlapping
structural member to resist load. In some area in Fig. 8.1.
jurisdictions these resistance factors are applied
to structural elements of different types based on
Frequency
the member type (e.g. beam, column, wall, etc.)
and materials of construction (e.g. steel, con-
Resistance
crete, timber, etc.). Values may vary between
about 0.60 and about 0.95. In other jurisdictions Load
resistance factors may be applied to materials Failure
rather than to structural elements. Values of
resistance factors for structural materials also
vary depending on the particular building code
E R Design Variable
and jurisdiction. For structural steel, the value of
ϕ is typically in the range of 0.85–0.95. Interest- Fig. 8.1 Probabilistic nature of load effect, E, and struc-
ingly, for connections (i.e., bolts, welds, etc.) tural resistance, R
260 L.A. Bisby
It is clear that the probability of failure can be data are not very well known—particularly
reduced by artificially increasing the true (nomi- during fire.
nal) resistance, R, of a structure by imposing
smaller values of ϕ (i.e. by shifting the R-curve
to the right in Fig. 8.1) or by artificially decreas- Structural Design Under Fire
ing the true loads by imposing larger values of α Conditions
(i.e. by shifting the E-curve to the left). The
target safety index, β, allows building code Structural design for fire conditions generally
developers to determine what these values should follows the same approach as for structural
be based on known (or approximated) probability design under ambient conditions, however
distributions for both the loads and the material because a severe fire in most buildings is a statis-
or member resistances. tically ‘rare’ event, the load and resistance
Since failure will occur if R < E, the proba- factors specified in building codes for the fire
bility of failure can be represented by the distri- limit state change to reflect this fact.
bution of Z ¼ RE. This is shown schematically
in Fig. 8.2, where failure is again represented by
the shaded region. This failure probability distri- Philosophy and Goals
bution also has a mean value and an associated
standard deviation. The goal of Limit States The design equation during fire is similar to
Design is to ensure that the mean value is suffi- Equation 8.2, and can be expressed in general as:
ciently above zero.
As shown in Fig. 8.2, the Safety Index is simply αθ Eθ ϕθ Rθ ð8:4Þ
the number of standard deviations that the mean
value of the Z curve is greater than zero. In most where the subscript θ is added to denote the
international codes this value is set between 2 and effects of elevated temperature. Elevated temper-
3. The safety Index can be quantified, provided ature may have an effect on each of the terms in
that the mean and standard deviations of the load Equation 8.4. For instance:
and resistance distributions are known, using: Eθ ¼ the specified effect of loads acting on the
structure at elevated temperature. It should
Z RE be noted that thermal expansion of structural
β¼ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi ð8:3Þ
σz σR 2 þ σE 2 elements may introduce new loads into the
structure due to restraint to thermal expan-
Clearly, to be able to quantify the safety
sion, and these should be considered;
index, it is necessary to have good statistical
αθ ¼ load factors applied to the specified loads
data for both the applied loads on a structural
for the elevated temperature condition.
member and its resistance; in many cases these
These are typically reduced as compared
with the ambient temperature values to
Frequency
reflect the most likely load condition at
the time of a fire (service loading condi-
tion). Typical load combinations for fire
are given later in this chapter;
Rθ ¼ the calculated resistance of a member at
Failure elevated temperature, based on material
b × sZ properties (and in some cases reduced
cross-sectional dimensions) which have
been reduced due to the damaging effects
Z Z=R-E
of heating. The resistance of a structural
Fig. 8.2 Probabilistic nature of Z ¼ R–E, and definition element will reduce during the course of a
of the Safety Index, β fire as it heats up; and
www.ebook777.com
ϕθ ¼ the resistance factor applied to the calculated most important outcome of assuming these
resistance or to specified properties and reduced loads during fire is the realization that,
dimensions, workmanship, type of failure, und er day-to-day conditions which are typically
and uncertainty in the prediction of resis- used to assess structural performance in fire,
tance at high temperature. These factors are most structures are subjected to loads of 50 %
typically set to 1.0 such that the nominal or less of their ultimate design capacities [2].
member or material strength at elevated tem- It should be noted that the value of the load
perature is used in calculations. or load effect resulting from the extraordinary
It is also important to recognize that structural event (fire) should be included and is denoted
design for fire typically considers three distinct by Ak in Equation 8.5.
modes of failure which must be prevented when
satisfying Equation 8.4. Recognizing that ensur-
ing that fire does not spread beyond the compart- Structural Mechanics
ment of origin for the requisite period of time is a
fundamental goal of fire safety engineering, these Thus far, this chapter has concerned itself with
failure modes are: the method that structural engineers use to quan-
1. loss of load bearing capacity (i.e. structural tify the likelihood of failure of, and hence design,
collapse); structural members under the influence of the
2. passage of flame or hot gas through a building various combinations of loads to which they
element (e.g. wall or floor), which would rep- might be subjected. Structural mechanics is the
resent a breach of fire compartmentation; and branch of physics which allows structural
3. excessive temperature rise at the exposed face engineers to determine the strength, or load bear-
of the structural element, which may also rep- ing capacity, and deformation of structural
resent a breach of fire compartmentation. elements of various types (e.g. beams, columns,
slabs) under load. To provide a basic overview of
the procedures used, the following sections give
a brief summary of the necessary concepts; the
Structural Fire Design Loads and Load steps in any analysis typically include:
Combinations 1. Calculation of external reaction (support)
forces;
Load combinations for use in Limit States Design 2. Determination of internal forces (axial, bend-
for ultimate capacity at ambient conditions were ing, shear, and torsion); and
given previously. In the case of structural fire anal- 3. Prediction of failure modes depending on the
ysis, the load combinations are altered to reflect the materials of construction, the geometry, the
statistical unlikelihood of a severe fire occurring in support conditions, and the loads.
the first place, as well as the fact that the actual
likely loads acting on a structure on a day-to-day
basis are typically much less than those used for Statics
ultimate strength design. Various countries apply
slightly different load combinations for fire. As one With the previous issues in mind, we now move
example, ASCE-7-05 [1] suggests the following to a discussion of the physics which are used to
load combination for fire: evaluate the capacities of various types of struc-
tural elements. The first of these topics is statics.
1:2D þ Ak þ ð0:5L or 0:25Þ ð8:5Þ
Statics provides the means by which both the
It should be noted that other codes may also external reactions and internal forces within a
include the effects of snow and wind loads during structural element can be determined. If a struc-
fire, however again at reduced levels as com- ture is in equilibrium (i.e. it is not moving but
pared with ambient temperature design. The ‘static’) then the algebraic sum of all of the forces
262 L.A. Bisby
X
and moments acting on that structure is equal to F y ¼ 0 : 6 þ RBy þ RCy 1ð24Þ
zero. Otherwise the structure would be 2ð18Þ=2 ¼ 0
accelerating.
X
Mz ¼ 0 : 6ð6Þ þ 1ð24Þð6Þ
Static Equilibrium and Reaction Forces RCy ð18Þ þ 2ð18Þð9Þ=2 ¼ 0
The first step in the analysis of structural element
under load is the determination of its support
∴RCy ¼ 18; RBy ¼ 30
reactions. For simple two-dimensional
(i.e. planar) structures, static equilibrium can be Note that if this structure had one or more
applied using an orthogonal coordinate system, additional supports there would not have been
such that the algebraic sum of the forces in the sufficient equilibrium equations to solve for all of
x (horizontal) and y (vertical) directions must be the unknown reactions, and the structure would
zero, and also the sum of the moments acting on have been referred to as statically indeterminate.
the structure must be zero. These equations can Unknown support reactions for statically indeter-
be expressed as follows: minate structures can only be obtained by consid-
X X X ering compatibility of their deformation under
Fx ¼ 0, F y ¼ 0, Mz ¼ 0 ð8:6Þ
load in addition to equilibrium; such methods are
In reality structures are three-dimensional, beyond the scope of this introductory discussion.
and in this case there would be six equations
governing static equilibrium; three translations Internal Forces
along x, y and z-axes and moments about all Once the support reactions for a statically deter-
three axes. As is typical in elementary structural minate structure are known the internal forces
mechanics this chapter considers only planar can be determined at any desired location.
structures for simplicity. Again, for a two-dimensional planar structure
To illustrate the use of these equations con- there are three internal forces which must be
sider the planar beam structure shown in Fig. 8.3; considered: Axial force, N(x), shear force, V(x),
this is similar to the example structure used in the and moment, M(x). The internal forces are found
previous edition of this handbook [3, 4]. The by taking a section through the structure, which
structure is a solid beam of constant cross- leads to the development of three unknown inter-
sectional area and materials of construction nal forces. Again the three equations of equilib-
along its length. The beam is supported on an rium can be applied to solve for the unknowns.
idealized pin support at Location B and an As an example, for the two-dimensional pla-
idealized roller support at Location C. It is loaded nar structure shown in Fig. 8.3c, the internal
by a diagonal tension force at Location A, as well forces at any location between B and C can be
as a distributed load of varying magnitude along determined from the following equilibrium
its length. equations (again with units of kN and m):
Figure 8.3b shows a free body diagram of the X
beam, where the supports have been replaced by Fx ¼ 0 : 8 þ 8 þ N ¼ 0 ! N ¼ 0
the unknown reaction forces which they would X
generate (in the directions in which they prevent F y ¼ 0 : 6 1ð6 þ xÞ þ 30
motion). Note that there are three unknown reac- ð2x=18ÞðxÞ=2 þ V ¼ 0
tion forces in this case, and since three equations ∴V ðxÞ ¼ x2 =18 x þ 18
of equilibrium are available in two dimensions,
X
the unknowns can be determined. Such structures Mz ¼ 0 : M þ 6ð6 þ xÞ þ 1ð6 þ xÞð6 þ xÞ=2
are referred to as statically determinate. The
30ðxÞ þ ð2x=18ÞðxÞðx=3Þ=2 ¼ 0
equations are applied as follows (with units of
kN and m): ∴MðxÞ ¼ x3 =3 x2 =2 þ 18x 54
www.ebook777.com
a
w = 1 kN/m w = 3 kN/m
A
C
B
53.13°
10 kN 6m 18 m
w = 2 kN/m
b
w = 1 kN/m
A
C
8 kN RBx B
RBy RCy
6 kN
c w = 2x/18 kN/m
w = 1 kN/m
M(x)
A
N(x)
8 kN 6 kN B
30 kN V(x)
6 kN
d
8 8
Axial Force
0 0
Diagram (kN)
(tension +ive)
18
e Shear Force
0 0 Diagram (kN)
−6
−12
−18
58.9
f
Bending Moment
0 0
Diagram (kN·m)
−54
Fig. 8.3 (a) Example structure, (b) free body diagram, (c) section to the left of location x, (d) axial force diagram, (e)
shear force diagram, and (f) bending moment diagram
By taking sections at successive locations forces along the length of the structure; these are
along the length of the beam and calculating the called the axial force, shear force, and bending
internal forces it is possible to develop diagrams moment diagrams, respectively, and are shown in
which plot the variation of the respective internal Fig. 8.3d–f for the structure in question.
264 L.A. Bisby
Strength of Simple Structural Elements A
Once the internal forces within a structural ele- P P

ment are known using the principles briefly
presented above (or using more advanced Fig. 8.4 Tensile loading of cylindrical bar of cross-
sectional area, A
techniques for statically indeterminate structures)
the effects of these forces must be determined so
as to check the capacity of the structural element P
P σ=
in question against the loading demand. To do A
this the likely failure mode for the structural
element must be determined. Fig. 8.5 Determination of average axial stress, σ, of
cylindrical bar of cross-sectional area, A, under tensile
There are a variety of failure modes which load, P
must be considered, depending on the type of
structural element (i.e. beam, column, etc.) and
the material from which it is constructed (some s
materials are more prone to certain types of
failures than other materials). The most impor- sy
tant failure modes in buildings are typically ten- eultimate > 5%
sion failure, compression failure, and bending
failure. For some structures or structural
elements under certain conditions shear failure
E = 200GPa
σy ≈ 250-450MPa
may also be important; however shear is not E
treated in this introductory discussion.
1
Tension Members (Cables and Ties) ey e
Tension members are much less common in
Fig. 8.6 Idealized axial stress, σ, versus axial strain, ε,
buildings that compression elements (columns)
for typical structural steel
or flexural (bending) elements (beams and slabs),
however tension members are the simplest struc-
This shows that stress in the bar is propor-
tural element because their failure mode can be
tional to the internal force and is given in units
described in relatively simple terms. Tension
of force per unit area (in this case N/mm2 or
elements in real buildings include diagonal brac-
MPa). Using Hooke’s Law the strain, ε, and
ing, cable-supports and ties, and hangar bars.
hence elongation, δ, of the bar can be determined
Consider a cylindrical steel bar of a given
from the following expressions (within the lin-
material and length L and cross-sectional area
ear-elastic range of material response) from:
A which is subjected to a tensile axial load, P,
in the direction of its longitudinal axis (Fig. 8.4). σ ¼ εE
Using the principles of the preceding section, if δ ð8:8Þ
the bar is sectioned at any internal location, a ε¼
L
tension force, P, will be acting internally
(Fig. 8.5). In the above expressions, E is the modulus of
The force is tensile as it acts to elongate the elasticity of the material from which the bar is
bar. The average tensile stress in the bar, which is made (a material characteristic, see below and
a measure of the intensity of force in a material, Fig. 8.6), and the strain, ε, represents the inten-
can be determined from: sity of deformation.
To determine if the bar will fail under this
P load (and hence stress) the stress versus strain
σ¼ ð8:7Þ
A response of the material from which the bar is
www.ebook777.com
made must be known. As an example, for struc- s

tural steel a typical idealized plot of stress versus fc’ ≈ 25-60MPa
strain is given in Fig. 8.6. Note that in practice fc’
this curve would be determined experimentally.
Figure 8.6 shows that the loading response of
steel is characterized by a linear increase in stress
with increasing strain, with slope E providing the
definition of Hooke’s Law used previously, up to
a given stress, σ y, which is called the yield stress.
Beyond the yield stress the strain increases line- e
arly with no further increase in stress. Hence, the ≈ 0.002 ecc ≈ 0.003
steel can be assumed to fail in tension when it
Fig. 8.7 Idealized axial stress, σ, versus axial strain, ε,
reaches its yield stress, and its load carrying for typical ‘normal’ strength concrete
capacity (resistance), R, can be easily determined
under this condition from:
a
R ¼ σ yA ð8:9Þ P
Calculation of tensile strength at elevated b P

temperature can be performed in a similar man- σ =
A
ner provided that the temperature of the struc-
tural element and the effect of this temperature
on the stress versus strain response of the mate-
rial are known. Chapter 9 provides information
on the probable reductions in mechanical P P
properties of various structural materials at ele-
Fig. 8.8 (a) Compressive loading of short (stocky) cylin-
vated temperature. drical bar of cross-sectional area, A, and (b) determination
For the purposes of illustration, if it is of average axial stress, σ
assumed that the yield strength of steel is reduced
from σ y at ambient temperature to σ yθ at elevated
Compression Members (Columns
temperature, then the tensile strength at elevated
and Struts)
temperature, Rθ, can be determined from:
Compression elements are common in buildings
Rθ ¼ σ yθ Aθ ð8:10Þ and include columns and struts. The strength of
these elements can be determined in a similar
It should be noted that the cross-sectional area manner as for tension elements, with the excep-
at elevated temperature, Aθ, may also be reduced tion that compression elements may also be
in applying Equation 8.10 to reflect a reduction in susceptible to buckling failure.
cross-sectional area due to heating, as would Consider the short, stocky compression
occur for example due to charring of wood for element shown in Fig. 8.8. As for the tension
timber elements in fire. element discussed previously, the column has
It must be recognized that all structural height L, cross-sectional area A, and is subjected
materials display distinct differences in mechani- to an axial load P (in this case compressive).
cal response to loading and widely varying Provided that the element is not prone to
strength and stiffness. For example, a typical buckling failure, its compressive (crushing)
stress versus strain response for ‘normal’ strength strength at ambient or elevated temperature can
concrete is shown in Fig. 8.7, where drastic be approximated using Equations 8.9 and 8.10,
differences in both the qualitative and quantita- respectively. However, in reality all compression
tive aspects as compared with steel are obvious. elements in buildings are susceptible to buckling
266 L.A. Bisby
Fig. 8.9 (a) Compressive a

loading of long (slender) P
c
bar of cross-sectional area, P
A, (b) determination of
average axial stress, σ, of
the bar under compressive δ
load, P, (c) buckling failure
δ
of the bar under load, P, ε= d
and (d) equilibrium of half L y
the bar under buckling
failure b P
P
σ=
L A y
L
2
P P P P
X
failure and to the combined effect of axial load Mz ¼ 0 : M Py ¼ 0 ! M
and the inevitable bending which also occurs;
failure by buckling must therefore also be ¼ Py ð8:11Þ
checked.
The elastic deformation of an element in
To illustrate the alternative compressive fail-
bending can be described by the following
ure mode by buckling, Fig. 8.9 shows a vertical
equation:
column of height L, cross-sectional area A, and
applied load P. The column is assumed to be M d2 y
pin-supported at both ends, such that there is no ¼ ð8:12Þ
EI dx2
rotational restraint of its extremities (note that
this is not the case for most real columns in where y is the displacement perpendicular to the
buildings). When the load is applied to the col- axis of the element, x is the distance along
umn (Fig. 8.9a) the column experiences an axial the element, and M is the internal moment
compressive stress of σ (Fig. 8.9b), and the which is acting at the section. I is the moment
column’s length will reduce by an amount δ of inertia of the element and is a measure of its
(Fig. 8.9a). However, because in reality it is resistance to flexural deformation. The moment
impossible to apply a perfectly concentric com- of inertia it is a function of the element’s cross-
pressive load, and because all structural elements sectional geometry and can be considered as a
contain small imperfections and irregularities, flexural analogue of area, A, for calculating ten-
the column will inevitably also experience a sile strength.
small amount of bending (Fig. 8.9c); the result Substituting Equation 8.12 into Equation 8.11
is a lateral deflection, y. and rearranging gives:
Figure 8.9d shows the bottom half of the col-
d2 y P
umn when it is sectioned at its mid-height. If we y¼0 ð8:13Þ
dx2 EI
consider the section’s equilibrium under this con-
dition, taking moments at the section the follow- This is a second order differential equation
ing moment equilibrium equation is obtained: which has the following solution:
www.ebook777.com
rffiffiffiffiffi ! rffiffiffiffiffi !
P P R
v ¼ C1 sin x þ C2 cos x p 2EI/L2
EI EI
ð8:14Þ
syA
If the ends of the column are pinned (as in this
case), then Equation 8.14 is only satisfied when:
rffiffiffiffiffi ! rffiffiffiffiffi
P P
sin L ¼ 0 or L ¼ nπ ð8:15Þ Real Column
EI EI L
Lcritical
Rearranging for P, the critical buckling load,
Fig. 8.10 Column compressive strength versus buckling
Pcritical, is obtained; this is the theoretical load length
which will cause global buckling failure:
design procedures which have been calibrated
n2 π 2 EI to take account of the necessary factors.
R ¼ Pcritical ¼ where : n ¼ 1, 2, 3 . . .
L2 Calculation of compressive strength at ele-
ð8:16Þ vated temperature is similar to at ambient tem-
perature, however several additional important
It is clear that the lowest value of Pcritical will considerations are required; these are that:
govern and that this occurs for n ¼ 1. Whilst this 1. the strength of materials is reduced at elevated
equation is only valid for linear elastic materials, temperature, such that the crushing strength
as it depends on the assumptions of elastic beam will be less than at ambient temperature (see
theory (i.e. Equation 8.12), it is instructive for Chap. 9), according to Equation 8.10;
studying the propensity of elements to buckling 2. the stiffness (i.e. elastic modulus, E) is
failure under compressive loads. reduced at elevated temperature (possibly
An interesting feature of Equation 8.16 is that more or less severely than the material’s
it shows buckling strength to be proportional to strength), thus reducing the critical buckling
the inverse of the square of the buckling length. load, Pcritical to:
Thus, if the length of a column is doubled then its
buckling strength decreases by a factor of four, n 2 π 2 Eθ I θ
Rθ ¼ Pcritical, θ ¼
and so on. This can be important for the response Lθ 2 ð8:17Þ
of columns in fire in cases where lateral support where : n ¼ 1, 2, 3;
from beams and slabs is removed by heating of
the floorplate. 3. the effective size of the column’s cross sec-
The preceding section has given two means of tion may be reduced, thus reducing the
calculating the strength of a structural element moment of inertia of the section from I to Iθ;
subjected to compressive axial load. Figure 8.10 4. local increases in temperature may result in
plots these two methods versus column length. additional loads and moments due to
For very short columns, the crushing strength interactions with the rest of the structure dur-
given by Equation 8.9 will govern, whereas for ing fire; for instance thermal restraint to
slender columns buckling (Equation 8.16) will expansion of columns by the cool surrounding
govern. In reality, the transition between the structure can increase the compressive loads
two failure modes is more gradual due to column on a column by 20–30 % in some cases; and
imperfections and inadvertent load eccentricities, 5. thermal expansion of the floorplate may result in
and for intermediate column lengths a combined lateral forces and displacements being imposed
buckling crushing failure mode will be observed on columns, resulting in unexpected shear forces
in reality. Building codes contain structural and so-called second-order bending moments.
268 L.A. Bisby
o
M
Compressive
Strain and stress
r
Tensile Strain
and stress
M M
Fig. 8.12 Segment of a beam in bending

Fig. 8.11 Segment of a beam in bending
Flexural Elements (Joists, Beams stretched; this location is called the neutral
and Girders) plane, and occurs at the mid-height for sections
Flexural elements in buildings are those which which are symmetric about a horizontal axis of
resist the applied loads primarily by bending; bending (such as the I-shaped cross section
these include joists, beams, girders, and slabs. shown in Fig. 8.13).
The variation of internal moment in a structural If it is assumed that the beam is homogenous
element under a set of loads, E, can be deter- and fabricated from a linear elastic material
mined using the techniques discussed previously (a helpful simplification for illustrative
with reference to Fig. 8.3. The resistance of an purposes), the distribution of strains over the
element to bending, R, must be determined using cross section is linear as shown in Fig. 8.13.
structural mechanics. Applying Hooke’s Law the stress distribution
When a structural element is subjected to over the cross section is therefore also linear,
bending it experiences curvature. This is shown with maximum compressive stress at the top
in Fig. 8.11, where a short segment of beam is fibre of the cross section and maximum tensile
subjected to a moment couple (i.e. an internal stress at the bottom fibre. These assumptions can
bending moment) which causes the segment to be expressed as:
bend in a concave-up direction. Lines which
εtop εbottom
were previously vertical drawn on the side of εy ¼ y¼ y ð8:18Þ
ytop ybottom
the beam would now both point towards a distant
origin called the center of curvature (denoted by
σ top εtop E
O in Fig. 8.11). The distance to the center of σy ¼ y¼ y ð8:19Þ
curvature is called the radius of curvature, r. ytop ytop
This concave up condition is typically
referred to as a positive or sagging moment. If the beam is in equilibrium, then at any section
When an internal moment causes bending in a to the total compressive forces must be equal to
concave down direction it is referred to as a the total tensile forces. For any small area, dA,
negative or hogging moment. located anywhere on the cross section, the resul-
When the segment is subjected to a sagging tant force is determined as the stress multiplied
moment, material at the top of the beam’s cross by its area:
section is compressed whereas material at the bot- dF ¼ σdA ð8:20Þ
tom of the beam is elongated, as shown in Fig. 8.12.
At one specific location on the beam’s cross Applying equilibrium in the horizontal direction
section it is neither being compressed nor and integrating over the cross section:
www.ebook777.com
etop stop
ytop ey sy
y
Neutral Axis
ybottom
ebottom sbottom
Beam Cross-section Strain distribution Stress Distribution
Fig. 8.13 Beam cross-section, strain distribution, and stress distribution for a homogenous, linear elastic beam in
bending
X ð ð
simple techniques or tables given in solid
Fx ¼ 0 : 0 ¼ dF ¼ σdA
mechanics textbooks. Rearranging and
ðA A ð incorporating Equation 8.19, an equation for the
y σ top
¼ σ top dA ¼ ydA stress at any location in a cross section is
ytop ytop
A A obtained:
ð8:21Þ
My
σy ¼ ð8:24Þ
Since the term in front of the integrand is I
non-zero, then for equilibrium it is required that:
This is referred to as The Flexure Formula,
ð
and can be used (for linear elastic materials) to
0 ¼ ydA ð8:22Þ check that the stress in a cross section is less than
A the failure stress for the material from which a
This is equivalent to stating that the axis of zero beam is made; and hence to define the resistance
strain, called the neutral axis, must pass through of an element, R, with respect to bending failure.
the centroid of the cross section. This approach works for most statically determi-
Stresses in the beam can also be determined nate structures, because for these types of
using equilibrium, which requires that the inter- structures the formation of a single flexural fail-
nal moment, M, be equal to the moment pro- ure point (called a plastic hinge) is sufficient to
duced by integrating the moment contributions cause failure. Statically indeterminate structures
of the individual areas over the cross section. are more complicated as they require the forma-
This can be expressed as: tion of more than one location of flexural failure,
ð ð as discussed below.
X The above approach to flexural analysis can
M ¼M : M ¼ ydF ¼ yðσdAÞ
also be used for elevated temperature analysis
ð
A ! A
ð ð8:23Þ provided that the likely reductions in mechanical
y σ top 2
¼ y σ top dA ¼ y dA properties of a structural member’s constituent
ytop ytop materials are known, and also provided that
A A
reductions in cross sectional area are
The term inside the integrand represents the accounted for.
moment of inertia (or second moment of area), I, Similar mechanics can be used to develop
of the cross section, and can be found using equations for stress in sections made from
270 L.A. Bisby
inelastic or non-linear materials, however these scope of the current chapter, however the
are considerably more involved and are not implications of static indeterminacy for real
discussed here. For beams made from steel, structural response, both at ambient and at ele-
which is often considered to be elastic-plastic, vated temperature, are worthy of brief
as previously shown in Fig. 8.6, or for reinforced discussion.
concrete which is built up from a combination of
steel and concrete (refer to Fig. 8.7), specialist Continuous Beams
texts should be consulted for information in Continuous beams are statically indeterminate
this area. due to their being, as the name implies, continu-
ous over multiple supports. An example of a
Lateral Instability of Beams relatively simple continuous beam is given in
The above equations assume failure of a beam by Fig. 8.14. This figure shows that unlike simply-
excessive bending stresses. This is analogous to supported beams, which require only a single
crushing failure of a column in compression. location of flexural failure for collapse to occur,
However, in certain cases beams may fail due continuous beams require multiple failure
to instability failures which are akin to buckling locations before a failure mechanism can form;
failures in compressively loaded columns. Lat- three failure locations in the case of the beam
eral instability failures result when a beam’s shown. Continuous beams are therefore redun-
compression fibre has insufficient lateral support, dant structures, and they can benefit from benefi-
allowing it to buckle in a direction perpendicular cial structural actions such as moment
to its longitudinal axis. Obviously, this is more of redistribution, both at ambient temperature and
a problem for slender beams or beams built up during fire. Techniques to account for moment
from thin plates. In design the propensity of an redistribution during fire are described in detail
element to lateral instability is accounted for by elsewhere (e.g. [2]).
limiting the maximum stresses which are permit-
ted in the cross section; this is particularly an Frames
issue for structural steel beams (and columns). The structural response of building frames, for
In structural fire design it is important to rec- instance the idealized two-dimensional portal
ognize that any members which provide lateral frame and moment resisting frame shown in
bracing to beams must have sufficient fire resis- Figs. 8.15 and 8.16, respectively, under gravity
tance to be able to continue to provide this brac- loads is complex, and for most real structures
ing in the event of a fire. This is an important requires the use of detailed computer analysis
consideration for both beams and, as already techniques to determine the internal forces and
noted, for columns. Specialist texts should be moments. In fire the response of frames is even
consulted for additional information on lateral further complicated; for instance by the effects of
instability of beams in bending. thermal expansion and thermal restraint at ele-
vated temperature.
To illustrate some of the important behaviors
Continuity and Full Structure Response which may occur in statically indeterminate
frame structures during fire, Fig. 8.15 shows a
As already noted, most real structures are too possible sequence of deformation that is likely to
complicated to use equilibrium alone to deter- occur in a simple planar portal frame structure
mine all of the external and internal forces which subjected to a fire in its interior.
may be acting. In these case more advanced The initial geometry of the structure is
methods of analysis must be used which gener- shown in Fig. 8.15a, where the well-known
ally account also for the deformations of shape of an industrial building such as a ware-
structures under loading (in addition to static house is evident. Under ambient conditions the
equilibrium). Such methods are beyond the stability of the structure is assured by moment
www.ebook777.com
A B C D E
RAx
RAy RBy RCy RDy REy
Bending Moment
Diagram
Simply-supported Equivalent beam

beam with built-in ends
Single-hinge Three-hinge
failure mechanism failure mechanism
Fig. 8.14 Bending moment diagram and failure mechanism for a continuous beam as compared with a simply-
supported beam under a uniformly distributed load
resisting connections at locations B, C, and D The roof elements (B-C and C-D) would initially
(i.e. these connections resist relative rotational be heated by the fire, and two important actions
displacements between the structural elements would occur:
framing into the joints). Without moment 1. The beams would experience an overall lon-
resisting connections at these locations the gitudinal thermal expansion, which would
structure would be a mechanism and would tend to increase the length of the beams
immediately collapse. The connections to the under heating by amount ΔT, as a function of
foundations at A and E are not strictly required the constituent materials’ coefficient of ther-
to be moment resisting for stability under ambi- mal expansion, αT. This thermal expansion, if
ent conditions. Under ambient conditions this occurring without any axial restraint, would
structure resists loads by a combination of be described by the following expressions:
bending and axial compression in its structural
εT ¼ αT ΔT or ΔLT ¼ αT ΔTL ð8:25Þ
elements.
Figure 8.15b shows how this portal frame The result of this thermal expansion would be
might react during the early stages of a fire. to push the columns outwards, and hence to
272 L.A. Bisby
a a
C
B D
A E
b
b
c Fig. 8.16 Schematic showing complexities of structural

response of a statically indeterminate multi-floor moment
resisting frame in fire (Reproduced after Buchanan [2])
construction, the roofing system, etc. The

result of this is that the roof beams would
bow toward the fire and begin to sag purely
as a consequence of the thermal gradient.
As a consequence of the thermal elongation
Fig. 8.15 Schematic showing complexities of structural and thermal bowing, combined with reductions
response and possible failure mode of a statically indeter-
minate portal frame in fire
in the mechanical properties of the beam on
further heating, the peak in the roof would grad-
ually displace downwards under the influence of
increase the moments in the columns and in gravity loads. At some point the peak of the roof,
the connections to the foundation. In reality Point C, may displace below the height of points
there would be some restraint to axial thermal B and D, and the roof would snap-through
expansion, which would also tend to increase forming a catenary.
the mechanical compressive loads in the Figure 8.15c shows the structure once the roof
beams on heating. has snapped through. Under this condition the
2. The beams, which when heated from below roof acts in tension (rather than bending and
would experience greater heating at their bot- compression) to support the load under a
tom fibre than at their top, would undergo severely deformed geometry. In addition, the
thermal curvature and thermal bowing, to var- moments on the column bases would be reversed
ious degrees depending on the materials of and stability of the structure would be assured
www.ebook777.com
only by the moment capacity of the columns and

their base connections to the foundations. Even-
tual failure of the structure would occur with the
columns pulled inward leading to collapse.
This relatively simple example clearly
illustrates that thermal expansion, thermal bow-
ing, large deflections, and alternative load carry-
ing mechanisms can all be expected to play
pivotal roles in the structural mechanics of a
real, however simplified, structure during fire.
A more complicated, yet still highly idealized
structure is shown in Fig. 8.16; this is a
two-dimensional moment resisting frame.
Again, Fig. 8.16a shows the frame supporting
gravity loads under ambient conditions. Again,
the structure resists both vertical (e.g. gravity)
and possibly lateral (e.g. wind) loads through a
combination of bending and compressive
loading.
Figure 8.16b shows a highly exaggerated ide-
alization of the possible deformation of this Fig. 8.17 Schematic representation of compressive and
moment resisting frame under exposure to a fire tensile membrane actions in a reinforced concrete slab
which is localized to a single internal bay at the
ground floor level. This suggests that the Membrane action, which can be either compres-
response of the structure is far more complex sive or tensile, manifests itself in planar
than the response of a single isolated beam structures such as reinforced concrete floor
presented earlier, and that the effects of continu- slabs or steel-concrete composite deck slabs.
ity, axial restraint, thermal elongations and Figure 8.17 shows two-dimensional
rotations, and reductions in mechanical idealizations of both tensile and compressive
properties of the constituent materials will all membrane actions, where compressive mem-
profoundly affect the forces, stresses, brane action can be though of as arching action,
deformations, and ultimately the failure mode and tensile membrane action can be thought of as
of the structure during fire. As a result of the catenary action. In reality these actions normally
heating during a fire, elements are subjected to manifest themselves in three dimensions; think
loads which may never have been considered of a dome (compressive membrane) or a net
during ambient design (for example shearing or (tensile membrane).
unexpected bending of the perimeter columns Compressive membrane action is particularly
due to being pushed laterally by the expanding important in continuous reinforced concrete
floor plate). Such factors must be considered structures during the early stages of a fire,
during design in order for structures to be ratio- where restrained thermal expansion of concrete
nally engineered to resist the effects of fire. Such slabs can lead to the development significant
analysis is extremely complex and requires the lateral restraint forces resulting in arching action
use of specialized computer analysis software. during fire (provided of course that the thrust
forces have a line of action below the neutral
Slabs and Shells (Membrane Actions) axis of bending of the slabs).
An additional structural action which often plays Tensile membrane action is particularly
an important role in the response of real important during the late stages of fires
structures during fire is membrane action. for structures with relatively thin concrete or
274 L.A. Bisby
a axial restraint as in Fig. 8.18b) then the bar is

L prevented from expanding and as a result it
experiences zero thermal strain (i.e. zero elonga-
tion) but an increase of thermal stress. By invok-
ing Hooke’s Law (Equation 8.8 presented
previously) the thermal stress, σ T (or thermal
b force FT), in the bar will be:
σ T ¼ EεT ¼ EαT ΔT
ð8:26Þ
FT ¼ σ T A ¼ EAαT ΔT
DL = aTDTL
c If it is assumed that the bar is made from

FT = EAaTDT structural steel, with a typical yield strength, σ y,
of 350 MPa and modulus of elasticity E of
FT = EAaTDT
200,000 MPa (refer to Fig. 8.6), and if it is further
DL = aTDTL assumed that the coefficient of thermal expansion
of steel, αT, is 13 106/K then the increase of
Fig. 8.18 Schematic representation of the possible temperature required to cause the bar to yield can
effects of restraint to thermal expansion of a cylindrical
steel bar
be determined as follows:
ε ¼ εT ¼ αT ΔT
steel-concrete composite slabs. In these cases σ y ¼ EεT ¼ EαT ΔT
tensile membrane action can prevent structural σy 350
∴ΔT ¼ ¼ ¼ 134 C
collapse (however under large vertical EαT 200;000 13 106
deflections) for durations of fire exposure much ð8:27Þ
greater than would be expected on the basis of
single element analysis. In other words, a change in temperature of the
Membrane actions are discussed in consider- perfectly restrained steel bar of +134 C can
able detail in specialized structural fire engineer- cause the bar to reach its failure stress. This
ing references (e.g. [2, 5, 6]). simple example, whilst clearly not representative
of a real structure, shows that a relatively mild
increase of temperature can have a profound
Thermally-Induced Loading influence on the forces and deformations within
a structure during heating. Only recently has the
As discussed previously, thermal expansion (and significance of thermal interaction within a
more importantly restraint to thermal expansion) structure been widely acknowledged within the
may significantly affect the deformation and structural fire engineering community in terms of
eventual failure of real structures in real fires. its possible influences on the overall stability,
To illustrate the possible importance of thermal and likely failure modes, of a building during
expansion, consider a simple cylindrical steel bar fire (e.g. [6]).
of length L which is uniformly heated an amount
ΔT (Fig. 8.18).
If the bar is free to expand as in Fig. 8.18a Other Considerations
then the change in length during heating is
described by Equation 8.25 given previously. If There are a host of other important structural
however, the bar is rigidly supported between mechanics issues which must be considered in
two immovable walls (i.e. a case of perfect the analysis and design of structures to resist the
www.ebook777.com
damaging effects of fire and elevated tempera- the scope of this introductory discussion; additional
ture; it is not possible to discuss all of these here. guidance is available elsewhere (e.g. [7]).
There are many failure modes and structural
interactions which could occur in any given
structure, and hence structural fire analysis and Summary
design should only be undertaken by individuals
with specialist knowledge in this area. This chapter has provided a brief, introductory
summary of basic structural mechanics, as is
relevant to a surface level understanding of the
Connections response of structural elements and structures to
fire. In conjunction with Chap. 9 of this hand-
One area which has not traditionally been explic- book, it provides a basic understanding of the
itly considered within mainstream structural means by which structural stability, and to a
design for fire is the specific performance of certain extent integrity and insulation during
connections at elevated temperature. Under fire, can be engineered through careful selection
ambient conditions structures are generally and design of building materials.
designed under the assumption that structural
members will fail before their connections. Acknowledgements The overall structure and format of
Indeed, as already noted more onerous strength this chapter of the SFPE Handbook has been based on the
previous (4/e) version, which was authored by Robert
reduction factors are typically applied during
W. Fitzgerald. The significant contribution of Prof.
connection design to ensure that this is the case. Fitzgerald to the development of this handbook must
The performance in fire of connections of various therefore be gratefully acknowledged.
types is currently a topic of considerable research
interest; however a detailed discussion of this
topic is avoided here. It is sufficient for the reader Nomenclature
to be aware that connection performance in fire
should be considered during the fire safe struc- A Area (nm2)
tural design of a building.
Ak Load or load effect resulting from an
extraordinary event (e.g. fire)
Disproportionate Collapse C1 Integration constant
C2 Integration constant
Disproportionate collapse refers to a situation c Distance from the extreme compression
where localized failure of a single structural ele- fiber to the neutral axis of bending (mm)
ment can lead to major or global collapse of a D Dead load
large, or disproportionate, portion of a structure E Earthquake load
[2]. There have been numerous real examples of E Load effect, or modulus of elasticity
disproportionate structural failure due to extreme (Young’s modulus) (GPa)
loading events such as blast, earthquake, and F Force (kN)
even fire. The most notable case of a dispropor- f c’ Compressive strength of concrete (MPa)
tionate structural collapse resulting from fire is I Moment of inertia (mm4)
probably the collapse of Building 7 of the World L Live load
Trade Center complex in New York on L Length (mm)
September 11th, 2001. M Moment (kN∙m)
Design to avoid disproportionate collapse, n 0, 1, 2. . .
termed design for redundancy or resilient design, O Center of curvature
requires the provision of structural redundancy and P Load (kN)
alternative load paths. Again, this topic is beyond R Member resistance, or reaction force (kN)
276 L.A. Bisby
r Radius of gyration (mm) 2. A.H. Buchanan, Structural Design for Fire Safety,
S Snow load Wiley, New York, NY (2001).
3. R.W. Fitzgerald, Mechanics of Materials, Addison-
V Shear force (kN) Wesley, Reading, MA (1982).
W Wind load 4. Fitzgerald, R. “Structural Mechanics,” SFPE Hand-
x Coordinate parallel to the axis of the struc- book of Fire Protection of Engineering, National Fire
tural element (mm) Protection Association, Quincy, MA (2008).
5. J.A. Purkiss, Fire Safety Engineering Design of
y Coordinate normal to the axis of the struc- Structures, Butterworth-Heinemann, New York, NY
tural element (mm), or lateral deflection (2007).
(mm) 6. Y. Wang, I. Burgess, F. Wald, M. Gillie, Performance-
Z Difference between resistance and load Based Fire Engineering of Structures, Spon Press
(2012).
demand 7. Scott et al., 2002 Prevention of Progressive Col-
α Load factor lapse, Multihazard Mitigation Council Of the
αT Coefficient of thermal expansion (K1) National Institute of Building Sciences,
β Safety index Washington, D.C., July.
ε Strain (no units)
εcc Compressive failure strain of concrete Professor Luke Bisby is Arup Chair of Fire and
(no units) Structures, and Royal Academy of Engineering Research
εy Yield strain (no units) Chair, within the BRE Centre for Fire Safety Engineering
at the University of Edinburgh. Educated in Canada as a
ΔL Change in length (mm) structural engineer, his research and teaching are focused
ΔT Change in temperature (K) predominantly in the area of Fire Safety Engineering. His
δ Deformation (mm) reasearch is broadly in the areas of structural engineering
σ Stress (MPa) and the provision of fire safety across the built environ-
ment, with an emphasis on the mechanical response of
σy Yield stress (MPa) buildings and construction materials during fire. He has
ϕ Resistance factor published widely in areas related to the fire behaviour of
π Pi reinforced concrete structures, and on fibre-reinforced
θ Subscript denoting elevated temperature polymers in structural engineering applications. His
current reasearch is interested in the response of novel
ω Uniformly distributed loading (kN/m) structural materials to heating, including high perfor-
mance and high strength concretes, polymers and polymer
References composites for construction, fire protection materials
(including reactive fire protection coatings), and structural
cross-laminated timber. He is also involved in research
1. ASCE, Minimum Design Loads for Buildings and projects seeking to better understand and unpick sociolog-
Other Structures (ASCE-7-05), American Society of ical and psychological issues in fire safety engineering
Civil Engineers (2005). and regulation.
www.ebook777.com
Properties of Building Materials

9
V.K.R. Kodur and T.Z. Harmathy
Introduction the properties of the materials the building com-

ponent is composed of. The availability of mate-
Building components are to be designed to satisfy rial properties at high temperature and
the requirements of serviceability and safety limit temperature distributions permits a mathematical
states. One of the major safety requirements in approach to predicting the performance of
building design is the provision of appropriate building components exposed to fire. When
fire resistance to various building components. a structural member is subjected to a defined
The basis for this requirement can be attributed to temperature-time exposure during a fire, this
the fact that, when other measures of containing exposure will cause a predictable temperature dis-
the fire fail, structural integrity is the last line of tribution in the member. Increased temperatures
defense. In this chapter, the term structural mem- cause deformations and property changes in the
ber is used to refer to both load-bearing (e.g., materials. With knowledge of the deformations
columns, beams, slabs) and non-load-bearing and property changes, the usual methods of struc-
(e.g., partition walls, floors) building components. tural mechanics can be applied to predict fire
Fire resistance is the duration during which resistance performance.
a structural member exhibits resistance with In recent years, significant effort has been
respect to structural integrity, stability, and undertaken to develop material properties of
temperature transmission. Typical fire resistance various construction materials at elevated
rating requirements for different building temperatures. In this chapter, the characteristics
components are specified in building codes. of materials are outlined. The various properties
In the past, the fire resistance of structural that influence fire resistance performance,
members could be determined only by testing. In together with the methods used to develop these
recent years however, the use of numerical properties, is discussed. The trends on the varia-
methods for the calculation of the fire resistance tion of thermal, mechanical, and other material-
of various structural members is gaining accep- specific properties with temperature of commonly
tance because these calculation methods are far used construction materials are presented.
less costly and time consuming. The fire perfor-
mance of a structural member depends, in part, on
Material Characteristics
V.K.R. Kodur (*)
Civil and Environmental Engineering, Michigan State Classification
University, East Lansing, Michigan, USA
T.Z. Harmathy Materials, based on composition, can be classified
as either a homogeneous or heterogeneous type.

278 V.K.R. Kodur and T.Z. Harmathy
Homogeneous materials have the same composi- by hydration. The cement paste in a concrete has
tion and properties throughout their volume and a highly complex microstructure, interspersed
are rarely found in nature. Heterogeneous materials with very fine, elaborate pores.
have different composition and properties. Most Most building materials can be treated as iso-
construction materials are heterogeneous, yet tropic materials, that is, as though they possessed
their heterogeneity is often glossed over when the same properties in all directions. An excep-
dealing with practical problems. tion to this is some of the advanced composite
The heterogeneity of concrete is easily notice- materials, such as fiber-reinforced polymers
able. Other heterogeneities related to the micro- (FRP), which might possess varying properties
structure of materials, that is, their grain and pore in different directions and are classified as aniso-
structures, are rarely detectable by the naked eye. tropic materials.
The microstructure depends greatly on the way Among the material properties, those that are
the materials are formed. In general, materials unambiguously defined by the current composi-
formed by solidification from a melt show the tion and phase are referred to as structure-insen-
highest degree of homogeneity. The result of the sitive. Some others depend on the microstructure
solidification is normally a polycrystalline mate- of the solid or on its previous history. These
rial, comprising polyhedral grains of crystals, properties are structure-sensitive.
which, in general, are equiaxial and randomly
oriented. Severe cold working in metals may
produce an elongated grain structure and crystals Porosity and Moisture Sorption
with preferred orientations.
Noncrystalline solids are called amorphous The fire performance of a material is dependent
materials. Gels and glasses are amorphous on the chemical composition and molecular
materials. Gels are formed by the coagulation structure of the material. The presence of water
of a colloidal solution. Glasses (vitreous in the material composition influences the
materials) are solids with a liquid-like, grainless properties of materials at elevated temperatures.
submicroscopic structure with low crystalline The two commonly associated terms to describe
order. On heating, they will go through a series the composition and the extent of water present
of phases of decreasing viscosity. in a material are porosity and moisture sorption.
Synthetic polymers (plastics) are made up of What is commonly referred to as a solid object
long macromolecules created by polymerization is actually all the material within its visible
from smaller repeating units (monomers). In the boundaries. Clearly, if the solid is porous—and
case of thermoplastic materials, the mobility of most building materials are—the so-called solid
the molecular chains increases on heating. Such consists of at least two phases: (1) a solid-phase
materials soften, much like glass. In some other matrix and (2) a gaseous phase (namely, air) in
types of plastics, called thermosetting materials, the pores within the matrix. Usually, however,
polymerization also produces cross-bonds there is also a liquid or liquid-like phase present:
between the molecular chains. These cross- moisture either absorbed from the atmosphere to
bonds prevent the loosening of the molecular the pore surfaces or held in the pores by capillary
structure and the transition of the material into a condensation. This third phase is always present
liquid-like state. if the pore structure is continuous; discontinuous
Some building materials (e.g., gypsum, brick) pores (like the pores of some foamed plastics) are
are formed from a wet, plastic mass or from not readily accessible to atmospheric moisture.
compacted powders by firing. The resulting The pore structure of materials is
product is a polycrystalline solid with a well- characterized by two properties: porosity, P
developed pore structure. Two important (m3.m–3), the volume fraction of pores within
building materials, concrete and gypsum, are the visible boundaries of the solid; and specific
formed by mixing finely ground powders (and surface, S (m2.m–3), the surface area of the pores
aggregates) with water. The mixture solidifies per unit volume of the material. For a solid with
www.ebook777.com
9 Properties of Building Materials 279
continuous pore structure, the porosity is a mea- m (dimensionless) ¼ Constant that has a value
sure of the maximum amount of water the solid between 1 and +1
can hold when saturated. The specific surface and Hamilton and Crosser recommended the
(to a lesser degree) porosity together determine following rather versatile formula for two-phase
the moisture content the solid holds in equilib- solids [4]:
rium with given atmospheric conditions.
v1 π1 þ γv2 π2
The sorption isotherm shows the relationship π¼ ð9:2Þ
at constant temperature between the equilibrium v1 þ γv2
moisture content of a porous material and the
where
relative humidity of the atmosphere. A sorption
isotherm usually has two branches: (1) an adsorp- nπ1
γ¼ ð9:3Þ
tion branch, obtained by monotonically increas- ðn 1Þπ1 þ π2
ing the relative humidity of the atmosphere from
0 to 100 % through very small equilibrium Here phase 1 must always be the principal
steps; and (2) a desorption branch, obtained by continuous phase. n (dimensionless) is a function
monotonically lowering the relative humidity of the geometry of phase distribution. With
from 100 to 0 %. Derived experimentally, the n ! 1 and n ¼ 1, Equations 9.2 and 9.3 con-
sorption isotherms offer some insight into the vert into Equation 9.1 with m ¼ 1 and m ¼ 1,
nature of the material’s pore structure [1, 2]. respectively. With n ¼ 3, a relation is obtained
For heterogeneous materials consisting of for a two-phase system where the discontinuous
solids of different sorption characteristics (e.g., phase consists of spherical inclusions [5].
concrete, consisting of cement paste and By repeated application, Equations 9.2
aggregates), the sorption isotherms can be and 9.3 can be extended to a three-phase system
estimated using the simple mixture rule (with [6], for example, to a moist, porous solid that
m ¼ 1; see Equation 9.1). consists of three essentially continuous phases
Building materials, such as concrete (or more (the solid matrix, with moisture and air in its
accurately, the cement paste in the concrete) and pores).
wood, because of their large specific surfaces,
can hold water in amounts substantial enough to
be taken into consideration in fire performance Survey of Building Materials
assessments.
There are burnable (combustible) and nonburn-
able (noncombustible) building materials. The
Mixture Rules reason for preferring the use of the words burn-
able and nonburnable has been discussed by
Some properties of materials of mixed composi- Harmathy [2]. To a designer concerned with the
tion or mixed phase can be calculated by simple structural performance of a building during a fire,
rules if the material properties for the constituents the mechanical and thermal properties of these
are known. The simplest mixture rule is [3] materials are of principal interest. Yet burnable
X building materials may become ignited, and
πm ¼ vi π im ð9:1Þ
thereby the positive role assigned to these
i
materials by design (i.e., functioning as structural
where elements of the building) may change into a
π ¼ Material property for the composite negative role—that is, becoming fuel and adding
πi ¼ Material property for the composite’s ith to the severity of fire. Those properties of burn-
constituent able building materials that are related to the
vi (m3.m–3) ¼ Volume fraction of the ith latter role are discussed in other chapters of this
constituent handbook.
From the point of view of their performance in (or Group I/F and L/F)—wood and FRP; and
fire, building materials can be divided into the Group I—gypsum and insulation.
following groups:
1. Group L (load-bearing) materials. Materials
capable of carrying high stresses, usually in
tension or compression. With these materials, Material Properties at Elevated
the mechanical properties related to behavior Temperatures
in tension and/or compression are of principal
interest. The behavior of a structural member exposed to
2. Group L/I (load-bearing/insulating) materials. fire is dependent, in part, on the thermal and
Materials capable of carrying moderate stresses mechanical properties of the material of which
and, in fire, providing thermal protection to the member is composed. While calculation
Group L materials. With Group L/I materials, techniques for predicting the process of deterio-
the mechanical properties (related mainly to ration of building components in fire have devel-
behavior in compression) and the thermal oped rapidly in recent years, research related to
properties are of equal interest. supplying input information into these
3. Group I (insulating) materials. Materials not calculations has not kept pace. The designer of
designed to carry load. Their role in fire is to the fire safety features of buildings will find that
resist the transmission of heat through building information on the properties of building
elements and/or to provide insulation to Group L materials in the temperature range of interest,
or Group L/I materials. With Group I materials, 20–800 C is not easy to come by. Most building
only the thermal properties are of interest. materials are not stable throughout this tempera-
4. Group L/I/F (load-bearing/insulating/fuel) ture range. On heating, they undergo physico-
materials. Group L/I materials that may chemical changes (“reactions” in a generalized
become fuel in fire. sense), accompanied by transformations in their
5. Group I/F (insulating/fuel) materials. Group I microstructure and changes in their properties.
materials that may become fuel in fire. For example, concrete at 500 C is completely
The number of building materials has been different from the material at room temperature.
increasing dramatically during the past few The thermophysical and mechanical
decades. In the last decade or so, a number of properties of most materials change substantially
high-performing materials, such as FRP and within the temperature range associated with
high-strength concrete (HSC), have been devel- building fires. In the field of fire science, applied
oped to achieve cost-effectiveness in construc- materials research faces numerous difficulties. At
tion. Although many of these high-performing elevated temperatures, many building materials
materials possess superior properties at ambient undergo physicochemical changes. Most of the
temperatures, the same cannot be said of properties are temperature dependent and sensi-
their performance at elevated temperatures. tive to testing method parameters such as heating
In materials such as HSC, additional comple- rate, strain rate, temperature gradient, and so
xities such as spalling arise, which may severely on. Harmathy [7] cited the lack of adequate
impact the fire performance of a structural knowledge of the behavior of building materials
member. at elevated temperatures as the most disturbing
By necessity, only a few of those materials trend in fire safety engineering. There has been a
that are commonly used will be discussed in this tendency to use “notional” (also called “typical,”
chapter in some detail. These materials are as “proprietary,” “empirical,” etc.) values for mate-
follows: in Group L—structural steel, light- rial properties in numerical computations—in
gauge steel, and reinforcing/prestressing steel; other words, values that ensure agreement
Group L/I—concrete and brick (including between experimental and analytical results.
fiber-reinforced concrete); Group L/I/F Harmathy warned that this practice might lead
www.ebook777.com
to a proliferation of theories that lack general Stress-Strain Relationships

validity.
Clearly, the generic information available on The mechanical properties of solids are usually
the properties of building materials at room tem- derived from conventional tensile or compres-
perature is seldom applicable in fire safety sive tests. The strength properties are usually
design. It is imperative, therefore, that the fire expressed in stress-strain relations, which are
safety practitioner knows how to extend, based often used as input data in mathematical models
on a priori considerations, the utility of the scanty calculating the fire resistance. Figure 9.1 shows,
data that can be gathered from the technical for a metallic material, the variation of stress, σ
literature. Also, knowledge of unique material- (Pa), with increasing strain (deformation), ε
specific characteristics at elevated temperatures, (mm–1), while the material is strained
such as spalling in concrete or charring in wood, (deformed) in a tensile test at a more or less
is critical to determine the fire performance of a constant rate (i.e., constant crosshead speed),
structural member. These properties are usually of the order of 1 mmmin–1. Generally,
discussed in the following sections. because of a decrease in the strength and ductility
of the material, the slope of the stress-strain
curve decreases with increasing temperature.
Reference Condition
Most building materials are porous and therefore Modulus of Elasticity, Yield Strength,
capable of holding moisture, the amount of Ultimate Strength
which depends on the atmospheric conditions.
Because the presence of moisture may have a The modulus of elasticity is a measure of the
significant and often unpredictable effect on the ability of the material to resist deformation and
properties of materials at any temperature below is expressed as the ratio of the deforming stress to
100 C, it is imperative to conduct all property the strain in the material. Generally, the modulus
tests on specimens brought into a moistureless of elasticity of a material decreases gradually
“reference condition” by some drying technique with increasing temperature.
prior to the test. The reference condition is nor- The tensile or compressive strength of the
mally interpreted as that attained by heating the material is generally expressed by means of
test specimen in an oven at 105 C until its yield strength and ultimate strength. Often the
weight shows no change. A few building
materials however, among them all gypsum
σu u
products, may undergo irreversible physico-
chemical changes when held at that temperature r
for an extended period. To bring them to a refer- σy
ence condition, specimens of these materials y
Stress, σ
should be heated in a vacuum oven at some e

lower temperature level (e.g., at 40 C in the
case of gypsum products).
Mechanical Properties
0
The mechanical properties that determine the fire Strain, ε
0.2%
performance of structural members are strength,
modulus of elasticity, and creep of the compo- Fig. 9.1 Stress-strain curve (strain rate is roughly
nent materials at elevated temperatures. constant)
Fig. 9.2 Variation of 100

strength with temperature
for different materials 90
80
Strength (% of initial)
70
60
50
40 Concrete
30
FRP Wood
20
Structural
10
steel
0
0 100 200 300 400 500 600 700 800
Temperature (°C)
strength at elevated temperature is expressed as a For metals of similar metallurgical

percentage of the compressive (tensile) strength characteristics, the stress-strain curve can be
at room temperature. Figure 9.2 shows the varia- reproduced at room temperature at a reasonable
tion of strength with temperature (ratio of tolerance, and the shape of the curve does not
strength at elevated temperature to that at room depend significantly on the crosshead speed. At
temperature) for concrete, steel, wood, and FRP. sufficiently high temperatures, however, the
For all four materials, the strength decreases with material undergoes plastic deformation even at
increasing temperature; however, the rate of constant stress, and the e-r section of the stress-
strength loss is different. strain curve will depend markedly on the cross-
For materials such as concrete, compressive head speed.
strength is of main interest because it has very
limited tensile strength at higher temperatures.
However, for materials such as steel, both Creep
compressive and tensile strengths are of equal
interest. Creep, often referred to as creep strain, is defined
Section 0-e of the curve in Fig. 9.1 represents as the time-dependent plastic deformation of the
the elastic deformation of the material, which is material and is denoted by εt (mm–1). At normal
instantaneous and reversible. The modulus of stresses and ambient temperatures, the deforma-
elasticity, E (Pa), is the slope of that section. tion due to creep is not significant. At higher
Between points e and u the deformation is plas- stress levels and at elevated temperatures, how-
tic, nonrecoverable, and quasi-instantaneous. ever, the rate of deformation caused by creep can
The plastic behavior of the material is be substantial [8]. Hence, the main factors that
characterized by the yield strength at 0.2 % off- influence creep are the temperatures, the stress
set, σy (Pa), and the ultimate strength, σu (Pa). level, and their duration.
After some localized necking (i.e., reduction of In a creep test the variation of εt is recorded
cross-sectional area), the test specimen ruptures against time, t (h), at constant stress (more accu-
at point r. The modulus of elasticity is more or rately, at constant load) and at constant (ele-
less a structure-insensitive property. vated) temperature T (K). A typical strain-time
www.ebook777.com
a
b
Creep stress, εt
Strain, ε
s2
tan–1z
tan–1εts
•
s1
εt
e
s
E
0 Time, t 0 Temperature-compensated time, θ
Fig. 9.3 (a) Creep strain vs. time curve (T ¼ constant; σ constant); (b) creep strain vs. temperature-compensated
time curve (σ constant)
curve is shown in Fig. 9.3a. The total strain, these two regimes can be satisfactorily described
ε (mm–1), is by the following equation [10]
σ εt0
ε¼ þ εt ð9:4Þ εt ¼ cosh1 2Zθ=εt0 ðσ ’ constantÞ ð9:6Þ
E ln 2
The 0-e section of the strain-time curve or approximated by the simple formula [11]
represents the instantaneous elastic (and revers-
εt εt0 þ Zθ ðσ ’ constantÞ ð9:7Þ
ible) part of the curve; the rest is creep, which
is essentially nonrecoverable. The creep is fast where Z (h–1) is the Zener-Hollomon parameter,
at first (primary creep, section e-s1 in Fig. 9.3a, and εt0 (mm–1) is another creep parameter, the
then proceeds for a long time at an approxi- meaning of which is explained in Fig. 9.3b.
mately constant rate (secondary creep, section The Zener-Hollomon parameter is defined as [12]
s1-s2), and finally accelerates until rupture
occurs (tertiary creep, section s2-r). The curve Z ¼ ε_ ts eΔH=RT ð9:8Þ
becomes steeper if the test is conducted
either at a higher load (stress) or at a higher where ε_ ts (mm–1h–1) is the rate of secondary
temperature. creep at a temperature, T. The two creep
Dorn’s concept is particularly suitable for parameters, Z and εt0, are functions of the applied
dealing with deformation processes developing stress only (i.e., they are independent of the
at varying temperatures [9]. Dorn eliminated the temperature).
temperature as a separate variable by the intro- For most materials, creep becomes noticeable
duction of a new variable: the “temperature- only if the temperature is higher than about
compensated time,” θ (h), defined as one-third of the melting temperature (on the
absolute scale).
ðt
The creep of concrete is due to the presence
θ¼ eΔHc =RT dt ð9:5Þ of water in its microstructure [13]. There is
0
no satisfactory explanation for the creep of
where ΔHc (Jkmol–1) is the activation energy of concrete at elevated temperatures. Anderberg
creep, and R (Jkmol1K–1) is the gas constant. and Thelandersson [14], and Schneider [15]
From a practical point of view, only the pri- suggested techniques for the calculation of the
mary and the secondary creeps are of impor- deformation of concrete under conditions char-
tance. It has been shown that the creep strain in acteristic of fire exposure.
reactions to shift to higher temperatures and to

Thermal Properties develop faster.
The coefficient of linear thermal expansion,
The material properties that influence the tem- β (mm–1K–1), is defined as
perature rise and distribution in a member are its
thermal conductivity, thermal expansion, specific 1 d‘
β¼ ð9:9Þ
heat, thermal diffusivity, and mass loss. These ‘ dT
properties depend on the composition and
Since ‘ ¼ ‘0 the coefficient of linear thermal
characteristics of the constituent materials.
expansion is, for all intents, the tangent to the
dilatometric curve. For solids that are isotropic in
a macroscopic sense, the coefficient of volume
Thermal Expansion
expansion is approximately equal to 3β.
The thermal expansion is measured with a
The thermal expansion characterizes the expan-
dilatometric apparatus, capable of producing
sion (or shrinkage) of a material caused by
curves that show the expansion of the materials
heating and is defined as the expansion (shrink-
with temperature in the range from 20 to
age) of unit length of a material when it is raised
1000 C. Harmathy [7, 16], using a horizontal
1 in temperature. The expansion is considered to
dilatometric apparatus, recorded dilametric
be positive when the material elongates and is
curves for various types of concrete and brick,
considered negative when it shortens. In general,
some of which are presented in later sections.
the thermal expansion of a material is dependent
The sample was 76.2 mm long and about 13 by
on the temperature. The dilatometric curve is a
13 mm in cross section. It was subjected to a
record of the fractional change of a linear dimen-
small spring load that varied during the test.
sion of a solid at a steadily increasing or decreas-
Unfortunately, even this small load caused
ing temperature. With mathematical symbolism,
creep shrinkage with those materials that tended
the dilatometric curve is a plot of
to soften at higher temperatures. Furthermore,
Δ‘ because the apparatus did not provide a means
against T for placing the sample in a nitrogen atmosphere,
‘0
in certain cases oxidation may also have had
where Δ‘ ¼ ‘ – ‘0 and ‘0 (m) and ‘ (m) are the some effect on the shape of the curves.
changed and original dimensions of the solid,
respectively, the latter usually taken at room
temperature. Δ‘ reflects not only the linear Mass Loss
expansion or shrinkage of the material, but also
the dimensional effects brought on by possible The mass loss is often used to express the loss of
physicochemical changes (i.e., “reactions”). mass at elevated temperatures. The thermogra-
The heating of the solid usually takes place at vimetric curve is a record of the fractional varia-
a predetermined rate, 5 Cmin–1 as a rule. tion of the mass of a solid at steadily increasing
Because the physicochemical changes proceed or decreasing temperature. Again, with mathe-
at a finite rate and some of them are irreversible, matical symbolism, a thermogravimetric curve
a dilatometric curve obtained by heating rarely is a plot of
coincides with that obtained during the cooling M
cycle. Sluggish reactions may bring about a against T
M0
steady rise or decline in the slope of the dilato-
metric curve. Discontinuities in the slope indi- where M and M0 (kg) are the changed and
cate very fast reactions. Heating the material at a original masses of the solid, respectively, the
rate higher than 5 Cmin–1 usually causes the latter usually taken at room temperature.
www.ebook777.com
Fig. 9.4 Mass loss of 105

various concrete types
as a function of
100
temperature [18]
95
90
Mass (%)
85
80 Fiber-reinforced
siliceous concrete
75 Fiber-reinforced
carbonate concrete
Plain carbonate
70
concrete
65
0 100 200 300 400 500 600 700 800 900 1000
Temperature (°C)
Generally a heating rate of 5 Cmin–1 is used in with respect to its dilatometric behavior, its den-
the measurements. sity at any temperature can be calculated from
A thermogravimetric curve reflects reactions the thermogravimetric and dilatometric curves.
accompanied by loss or gain of mass but, natu-
ðM=M0 ÞT
rally, it does not reflect changes in the materials’ ρ ¼ ρ0 ð9:10Þ
microstructure or crystalline order. M/M0 ¼ 1 is 1 þ ððΔ‘Þ=ð‘0 ÞÞT
the thermogravimetric curve for a chemically
inert material. Again, an increase in the rate of where ρ0 (kgm–3) is the density of the solid at the
heating usually causes those features of the curve reference temperature (usually room tempera-
that are related to chemical reactions to shift to ture), and the T subscript indicates values
higher temperatures and to develop faster. pertaining to temperature T in the thermogra-
The thermogravimetric curves to be shown vimetric and dilatometric records.
were obtained by a DuPont 951 thermogra- The density of composite solids at room tem-
vimetric analyzer [17], using specimens of perature can be calculated by means of the mix-
10–30 mg in mass, placed in a nitrogen atmo- ture rule in its simplest form (Equation 9.1 with
sphere [7]. The rate of temperature rise was m ¼ 1).
5 Cmin–1. Figure 9.4 shows the variation of X
p¼ v i pi ð9:11Þ
mass loss for concrete in the temperature range i
from 20 to 1000 C.
where the i subscript relates to information on the
ith component. At elevated temperatures, the
Density, Porosity expansion of the components is subject to
constraints, and therefore the mixture rule can
The density, ρ (kgm–3), in an oven-dry condi- yield only a crude approximation.
tion, is the mass of a unit volume of the material, If, as usual, the composition is given in mass
comprising the solid itself and the air-filled fractions rather than in volume fractions, the
pores. Assuming that the material is isotropic volume fractions can be obtained as
w i = pi dξ
vi ¼ P ð9:12Þ c p ¼ c p þ Δh ð9:16Þ
i wi = pi dT
where wi is the mass fraction of the ith compo- where cp (Jkg–1K–1) is the specific heat for that
nent (kgkg–1). mixture of reactants and (solid) products that the
True density, ρt (kgm–3), is the density of the material consists of at a given stage of the con-
solid in a poreless condition. Such a condition is version (as characterized by ξ), and Δ hp (Jkg–1)
nonexistent for many building materials and, is the latent heat associated with the physico-
therefore, may be a theoretical value derived on chemical change.
crystallographic considerations, or determined As Equation 9.16 and Fig. 9.5 show, in
by some standard technique, for example, temperature intervals of physicochemical insta-
ASTM C135 [19]. The relationship between the bility, the apparent specific heat consists of sen-
porosity and density is sible heat and latent heat contributions. The latter
contribution will result in extremities in the
ρt ρ calorimetric curve: a maximum if the reaction
P¼ ð9:13Þ
ρt is endothermic, a minimum if it is exothermic.
In heat flow studies, it is usually the ρcp
The overall porosity of a composite material
product (Jm–3K–1) rather than cp that is needed
consisting of porous components is
as input information. This product is referred to
X as volume specific heat.
P¼ v i Pi ð9:14Þ
i Until the 1980s, adiabatic calorimetry was the
principal method to study the shape of the cp
where, again, the i subscript relates to the ith versus T relationship. Since the 1980s, differen-
component of the material. tial scanning calorimetry (DSC) has been the
most commonly used technique for mapping the
curve in a single temperature sweep at a desired
Specific Heat rate of heating. Unfortunately, the accuracy of the
DSC technique in determining the sensible heat
The specific heat of a material is the characteris- contribution to the apparent specific heat may not
tic that describes the amount of heat required to be particularly good (sometimes it may be as low
raise a unit mass of the material at unit tempera- as 20 %). The rate of temperature rise was
ture. A calorimetric curve describes the variation usually 5 Cmin–1. At higher heating rates, the
with temperature of the apparent specific heat of peaks in the DSC curves tend to shift to higher
a material at constant pressure, cp (Jkg–1K–1).
The apparent specific heat is defined as
δh
cp ¼ ð9:15Þ
δT p
Apparent specific heat, cp
dξ
Δhp ⎯
where h is enthalpy (Jkg–1), and the p subscripts dT
indicate the constancy of pressure. If the heating
of the solid is accompanied by physicochemical
changes (i.e., “reactions”), the enthalpy becomes
a function of the reaction progress variable, ξ cp
(dimensionless), that is, the degree of conversion
at a particular temperature from reactant(s) into
product(s). For any temperature interval where 0 Temperature, T
physicochemical change takes place [2, 6, 20],
0 ξ 1, and Fig. 9.5 The apparent specific heat
www.ebook777.com
temperatures and become sharper. For conductivity of porous materials is, in a strict
temperatures above 600 C, a high-temperature sense, merely a convenient empirical factor that
differential thermal analyzer (DTA) is also used. makes it possible to describe the heat transmis-
Harmathy, with the aid of a DuPont 910 differen- sion process with the aid of the Fourier law.
tial scanning calorimeter, developed calorimetric That empirical factor will depend not only on
curves for a number of materials by placing the the conductivity of the solid matrix but also on
samples, 10–30 mg in mass, in a nitrogen atmo- the porosity of the solid and the size and shape of
sphere [7, 21]. the pores. At elevated temperatures, because of
Materials that undergo exothermic reactions the increasing importance of radiant heat trans-
may yield negative values in the calorimetric mission through the pores, conductivity becomes
curve. A negative value for cp indicates that, at sensitive to the temperature gradient.
the applied (and enforced) rate of heating, the rate Because measured values of the thermal
of evolution of reaction heat exceeds the rate of conductivity depend to some extent on the tem-
absorption of sensible heat by the material. In natu- perature gradient employed in the test, great
ral processes, the apparent specific heat can never discrepancies may be found in thermal conduc-
be negative, because the heat evolving from the tivity data reported by various laboratories.
reaction is either scattered to the surroundings or, A thermal conductivity value yielded by a
if absorbed by the material, causes a very fast tem- particular technique is, in a strict sense, applica-
perature rise. If the heat of reaction is not very high, ble only to heat flow patterns similar to that
obtaining nonnegative values for cp can be characteristic of the technique employed.
achieved by suitably raising the scanning rate. For Experimental data indicate that porosity is not
this reason, some materials undergoing exothermic a greatly complicating factor as long as it is not
reactions must be tested at rates of heating higher larger than about 0.1. With insulating materials,
than 5 Cmin–1, often as high as 50 Cmin–1. however, the porosity may be 0.8 or higher. Con-
If experimental information is not available, duction through the solid matrix may be an insig-
the cp versus T relationship can be calculated nificant part of the overall heat transmission
from data on heat capacity and heat of formation process; therefore, using the Fourier law of heat
for all the components of the material (including conduction in analyzing heat transmission may
reactants and products), tabulated in a number of lead to deceptive conclusions. If the solid is not
handbooks [22, 23]. Examples of calculations are oven-dry, a temperature gradient will induce
presented in Harmathy [2, 6], where information migration of moisture, mainly by an evaporation
is developed for the apparent specific heat versus condensation mechanism [24]. The migration of
temperature relation for a cement paste and four moisture is usually, but not necessarily, in the
kinds of concrete. direction of heat flow and manifests itself as an
increase in the apparent thermal conductivity of
the solid. Furthermore, even oven-dry solids may
Thermal Conductivity undergo decomposition (mainly dehydration)
reactions at elevated temperatures. The sensible
The temperature rise in a member, as a result of heat carried by the gaseous decomposition
heat flow, is a function of the thermal conductiv- products as they move in the pores adds to the
ity of the material. Heat transmission solely by complexity of the heat flow process. At present
conduction can occur only in poreless, there is no way of satisfactorily accounting for
nontransparent solids. In porous solids (most the effect of simultaneous mass transfer on heat
building materials), the mechanism of heat trans- flow processes occurring under fire conditions.
mission is a combination of conduction, radia- The thermal conductivity of layered, multi-
tion, and convection. (If pore size is less than that phase solid mixtures depends on whether the
about 5 mm, the contribution of pores to convec- phases lie in the direction of, or normal to,
tive heat transmission is negligible.) The thermal the direction of heat flow and is determined
using the simple mixture rule [4, 25]. At higher figures to identify thermal conductivity values
temperatures, because of radiative heat transfer obtained by stepwise increasing and stepwise
through the pores, the contribution of the pores to decreasing the temperature of the sample, respec-
the thermal conductivity of the solid must not be tively. Also, often the thermal conductivity of a
disregarded [26]. material is taken as invariant with respect to the
The thermal conductivity of solids is a direction of heat flow.
structure-sensitive property. For crystalline
solids, the thermal conductivity is relatively
high at room temperature and gradually Thermal Diffusivity
decreases as the temperature rises. For predomi-
nantly amorphous solids, on the other hand, the The thermal diffusivity of a material is defined as
conductivity is low at room temperature and the ratio of thermal conductivity to the volumet-
increases slightly with the rise of temperature. ric specific heat of the material. It measures the
The conductivity of porous crystalline materials rate of heat transfer from an exposed surface of a
may also increase at very high temperatures material to the inside. The larger the diffusivity,
because of the radiant conductivity of the pores. the faster the temperature rise at a certain depth
The thermal conductivity of materials such as in the material. Similar to thermal conductivity
concrete or brick can be measured, in the tem- and specific heat, thermal diffusivity varies with
perature range between 20 and 800 C, using a temperature rise in the material. Thermal diffu-
non-steady-state hot wire method [27, 28]. The sivity, α, can be calculated using the relation
thermal conductivity values at discrete tempera-
ture levels can be plotted to obtain a curve. k
α¼ ð9:17Þ
Unfortunately, no scanning technique exists for ρc p
acquiring a continuous thermal conductivity ver-
sus temperature curve from a single temperature where
sweep. Special problems arise with the estima- k ¼ Thermal conductivity
tion of the thermal conductivity for temperature ρ ¼ Density
intervals of physicochemical instability. Both the cp ¼ Specific heat of the material
steady-state and variable-state techniques of
measuring thermal conductivity require the sta-
bilization of a pattern of temperature distribution Special (Material-Specific) Properties
(and thereby a certain microstructural pattern) in
the test sample prior to the test. The test results In addition to thermal and mechanical properties,
can be viewed as points on a continuous thermal certain other properties, such as spalling in con-
conductivity versus temperature curve obtained crete and charring in wood, influence the perfor-
by an imaginary scanning technique performed at mance of a material at elevated temperature.
an extremely slow scanning rate. Because each These properties are unique to specific materials
point pertains to a more or less stabilized and are critical for predicting the fire perfor-
microstructural pattern, there is no way of know- mance of a structural member.
ing how the thermal conductivity would vary in
the course of a physicochemical process devel-
oping at a finite rate and varying microstructure. Critical Temperature
On account of the nonreversible
microstructural changes brought about by In building materials, such as steel and FRP, the
heating, the thermal conductivity of building determination of failure in a structural member
materials (and perhaps most other materials) is exposed to fire is simplified to the calculation of
usually different in the heating and cooling critical temperature. The critical temperature is
cycles. Open and solid circles are used in the defined as the temperature at which the material
www.ebook777.com
loses much of its strength and can no longer Spalling

support the applied load. When this temperature
is reached, the safety factor against failure Spalling is defined as the breaking of layers
becomes less than 1. (pieces) of concrete from the surface of the
North American standards (ASTM E119) concrete elements when the concrete elements
assume a critical or failure temperature of are exposed to high and rapidly rising
538 C (1000 F) for structural steel. It is a typical temperatures, such as those experienced in fires.
failure temperature for columns under full design Spalling can occur soon after exposure to heat
load. This temperature is also regarded as the and can be accompanied by violent explosions,
failure temperature in the calculation of fire resis- or it may happen when concrete has become so
tance of steel members. If a load is applied to weak after heating that, when cracking develops,
the member, the test is continued until the mem- pieces fall off the surface. The consequences
ber actually fails, which, depending on the load may be limited as long as the extent of the
intensity, may occur at a higher or lower steel damage is small, but extensive spalling may
temperature. lead to early loss of stability and integrity due
This concept of critical temperature is also to exposed reinforcement and penetration of
used for reinforced and prestressed steel in partitions.
concrete structural members for evaluating the Although spalling might occur in all
fire resistance ratings. These ratings are gener- concretes, high-strength concrete (HSC) is
ally obtained through the provision of minimum believed to be more susceptible than normal-
member dimensions and minimum thickness strength concrete (NSC) because of its low
of concrete cover. The minimum concrete permeability and low water-cement ratio. In a
cover thickness requirements are intended to number of test observations on HSC specimens,
ensure that the temperature in the reinforcement it has been found that spalling is often of an
does not reach its critical temperature for the explosive nature [30, 31]. Hence, spalling is one
required duration. For reinforcing steel, the of the major concerns in the use of HSC and
critical temperature is 593 C, whereas for should be properly accounted for in evaluating
prestressing steel the critical temperature is fire performance. Spalling in NSC and HSC
426 C [29]. columns is compared in Fig. 9.6 using the data
Fig. 9.6 Spalling in NSC

and HSC columns after
exposure to fire [32]:
(a) normal-strength
concrete column
and (b) high-strength
concrete column
obtained from full-scale fire tests on loaded primary parameters that influence spalling in
columns [32]. It can be seen that the spalling is HSC. The variation of porosity with temperature
quite significant in the HSC column. is an important property needed for predicting
Spalling is believed to be caused by the spalling performance of HSC. Noumowe
buildup of pore pressure during heating. The et al. carried out porosity measurements on
extremely high water vapor pressure, generated NSC and HSC specimens, using a mercury
during exposure to fire, cannot escape due to the porosimeter, at various temperatures [36].
high density (and low permeability) of HSC, and
this pressure buildup often reaches the saturation
vapor pressure. At 300 C, the pressure reaches Charring
approximately 8 MPa; such internal pressures are
often too high to be resisted by the HSC mix Charring is the process of formation of a layer of
having a tensile strength of approximately char at the exposed surface of wood members
5 MPa [33]. The drained conditions at the heated during exposure to fire. The charring process also
surface, and the low permeability of concrete, occurs in other members, such as FRP and some
lead to strong pressure gradients close to the types of plastics. When exposed to heat, wood
surface in the form of the so-called “moisture undergoes thermal degradation (pyrolysis), the
clog.” [2, 34] When the vapor pressure exceeds conversion of wood to char and gas, resulting in
the tensile strength of concrete, chunks of con- a reduction of the density of the wood. Studies
crete fall off from the structural member. The have shown that the charring temperature for
pore pressure is considered to drive progressive wood lies in the range of 280–300 C [29].
failure; that is, the lower the permeability of The charred layer is considered to have prac-
concrete, the greater the spalling. This falling tically no strength. The fire resistance of the
off can often be explosive in nature, depending member depends on the extent of charring and
on the fire and concrete characteristics. the remaining strength of the uncharred portion.
However, other researchers explain the occur- The charring rate, a critical parameter in
rence of spalling on the basis of fracture mechan- determining the fire resistance of a structural
ics and state that the spalling results from wood member, is defined as the rate at which
restrained thermal dilatation close to the heated wood is converted to char. In the standard fire
surface [35]. This leads to compressive stresses resistance test, it has been noted that the average
parallel to the heated surface, which are released rate of charring transverse to the grain is approx-
by brittle fractures of concrete, in other words, imately 0.6 mm/min [29]. The charring rate par-
spalling. allel to the grain of wood is approximately twice
Spalling, which often results in the rapid loss of the rate when it is transverse to the grain.
concrete during a fire, exposes deeper layers of Detailed studies on the charring rates for several
concrete to fire temperatures, thereby increasing specimen and timber types are reported by
the rate of transmission of heat to the inner layers various researchers [37–39] and are summarized
of the member, including the reinforcement. in a report [40]. These charring rates were
When the reinforcement is directly exposed to constant (in each study) and ranged from 0.137
fire, the temperatures in the reinforcement rise at to 0.85 mm/min. The assumption of a constant
a very high rate, leading to a faster decrease in rate of charring is reasonable for thick wood
strength of the structural member. The loss of members.
strength in the reinforcement, added to the loss Charring is influenced by a number of
of concrete due to spalling, significantly decreases parameters, the most important ones being den-
the fire resistance of a structural member. sity, moisture content, and contraction of wood.
In addition to strength and porosity of con- The influence of the moisture content and density
crete mix, density, load intensity, fire intensity, of the wood on the charring rate is illustrated in
aggregate type, and relative humidity are the Fig. 9.7 for Douglas fir exposed to the standard
www.ebook777.com
0.9 Harmathy [2] and the ASCE manual on structural

Moisture content (by weight)
5%
fire protection [29] present a wealth of informa-
0.8 tion on concrete, steel, wood, brick, gypsum, and
10%
15% various plastics. The thermal properties of
0.7 20% 31 building materials are surveyed in an NRCC
Rate of charring (mm/min)
0.6 report [7]. The mechanical and thermal properties

of concrete are discussed in an ACI guide [43],
0.5 and in reports by Bennetts [44] and Schneider
[45]. Those of steel are surveyed in the ACI
0.4
guide, in Bennetts’s report, and in a report by
0.3
Anderberg [46]. Information on the thermal con-
ductivity of more than 50 rocks (potential
0.2 concrete aggregates) is presented in a paper by
Birch and Clark [47]. The relationships for ther-
0.1 mal and mechanical properties, at elevated
temperatures, for some building materials are
0
300 400 500 600 listed in the ASCE structural fire protection
Density (kg/m3 )
manual [29]. In most cases these properties are
expressed, in the temperature range of 0–1000 C,
Fig. 9.7 Rate of charring in Douglas fir as a function of as a function of temperature and other properties
its density (dry condition) for various moisture contents at ambient temperature. These values can be
when exposed to ASTM standard fire [29]
used as input data in mathematical models for
predicting cross sectional temperatures and fire
fire [29]. It can be seen that the charring rate performance of structural members.
decreases with increasing density of the wood
and also with increasing moisture content.
It is important to recognize that the charring Steel
rate in real fires depends on the severity of fire to
which the wood is exposed. It should be noted Steel is a Group L material. The steels most often
that the charring rate is a function of the imposed used in the building industry are either hot-rolled
radiant heat flux. This depends on the fuel load or cold-drawn. The structural steels and concrete
and the ventilation factor of the compartment (for reinforcing bars are hot-rolled, low-carbon,
full details see Chap. 30, in this book). Detailed ferrite-pearlite steels. They have a randomly ori-
information on the charring of untreated wood— ented grain structure, and their strength depends
with expressions for charring rate in terms of the mainly on their carbon content. The prestressing
influencing factors of density, moisture content, steel wires and strands for concrete are usually
external heat flux, and oxygen concentration— made from cold-drawn, high-carbon, pearlitic
when exposed to real fires is given by Hadvig steels with an elongated grain structure, oriented
[41] and Mikkola [42]. in the direction of the cold work. In addition,
light-gauge steel, made from cold-formed steel,
finds wide applications in lightweight framing,
Sources of Information such as walls and floors.
Information on the mechanical properties of
Information on the properties of building two typical steels (a structural steel [ASTM A36]
materials at elevated temperatures is scattered and a prestressing wire [ASTM A421]) is
throughout the literature. There are a few presented in Figs. 9.8, 9.9, and 9.10 and in
publications, however, that may be particularly Table 9.1 [48]. Figures 9.8 and 9.9 are stress-
valuable for fire safety practitioners. A book by strain curves at room temperature (24 C and
Fig. 9.8 Stress-strain

curves for a structural steel 1 = 24°C 7 = 368°C
(ASTM A36) at room 700 2 = 99°C 8 = 427°C
temperature and elevated 3 = 149°C 9 = 482°C
temperatures [48] 4 = 204°C 10 = 536°C
600
5 = 260°C 11 = 593°C
6 = 316°C 12 = 649°C
500 54
67
3
σ (MPa)
400 21
8
1
300 9
2
10
200
11
100 12
0
0 0.02 0.04 0.06 0.08 0.10 0.12
ε (m·m–1)
Fig. 9.9 Stress-strain

curves for prestressing steel 1 = 21°C 7 = 377°C
(ASTM A421) at room 2 = 93°C 8 = 432°C
temperature and elevated 3 = 149°C 9 = 488°C
temperatures [48] 2000 4 = 204°C 10 = 538°C
5 = 257°C 11 = 593°C
6 = 310°C 12 = 649°C
4
1500 5
σ (MPa)
2 1 3
6
1000
7
8
9
500
10
11
12
0
0 0.02 0.04 0.06 0.08 0.10 0.12
ε (m·m–1)
21 C, respectively) and at a number of elevated structure-sensitive property, the creep

temperature levels. Figure 9.10 shows the effect parameters may show a substantial spread, even
of temperature on the yield and ultimate for steels with similar characteristics at room
strengths of the two steels. temperature. The application of the creep
Table 9.1 presents information on the effect of parameters to the calculation of the time of struc-
stress on the two creep parameters, Z and εt0 tural failure in fire is discussed in Hamilton and
(see Equation 9.7). Because creep is a very Crosser [4, 8].
www.ebook777.com
The modulus of elasticity (E) is about (phase change of steel) of the ferrite-pearlite
210 103 MPa for a variety of common steels structure into austenite.
at room temperature. Figure 9.11 shows its Being a structure-sensitive property, the ther-
variation with temperature for structural steels mal conductivity of steel is not easy to define.
[50] and steel reinforcing bars [49]. (E0 in For carbon steels it usually varies within the
Fig. 9.11 is the modulus of elasticity at room range of 46–65 Wm–1K–1.
temperature.) Equations for various properties of steel, as
The density (ρ) of steel is about 7850 kgm–3. functions of temperature, are available in the
Its coefficient of thermal expansion (β) is a ASCE structural fire protection manual [29] and
structure-insensitive property. For an average in Eurocode 3 [51, 52]. In the ASCE manual, the
carbon steel, β is 11.4 106 mm–1K–1 at same set of relationships is applicable for thermal
room temperature. The dilatometric curve properties of both structural and reinforcing steel.
shown in Fig. 9.12 is applicable to most of the However, separate relationships for stress-strain
common steels. The curve reveals substantial and elasticity are given for the two steels with
contraction of the material at about 700 C, slightly conservative values for structural steel.
which is associated with the transformation Recently, Poh proposed a general stress-strain
equation that expresses stress explicitly in terms
1750 of strain in a single continuous curve [53, 54].
The critical temperature of steel is often used as
a benchmark for determining the failure of struc-
1500
tural members exposed to fire. This ensures that
the yield strength is not reduced to less than that of
1250 σu 50 % of ambient value. The critical temperature
σy
for various types of steels is given in Table 9.2.
σy or σu (MPa)
1000 The above discussed high temperature

ASTM properties are generally applicable to conven-
750
A421 tional carbon (mild) steel whose chemical
composition consist of iron, carbon, manganese,
σu sulfur and phosphorous. In recent years, a num-
500
ber of new steels are available and these steel are
ASTM made by adding alloys, such as nickel, titanium,
A36
250 boron and chromium. These alloys influence
σy
durability characteristics, as well mechanical
0 properties of steel. For example, molybdenum,
0 100 200 300 400 500 600 700
chromium and niobium can increase the fire
Temperature (°C) resistance property of steel, while chrome and
Fig. 9.10 The ultimate and yield strengths for a struc- nickel can enhance the corrosion resistance of
tural steel (ASTM A36) and a prestressing steel (ASTM steel [56]. Current design rules on fire resistance
A421) at elevated temperatures [48] of steel structures (EC3 2005b [51], BS:5950
Table 9.1 Creep parameters for a structural steel and a prestressing steel [48]
Steel ΔHc/R (k) εt0(σ) (m m–1) Z(σ) (h–1)
ASTM A36 38,890 3.258 10–17σ1.75 2.365 10–20σ4.7 if σ 103.4 106
1.23 1016 exp (4.35 10–8σ) if 103.4 106 σ 310 106
ASTM A421 30,560 8.845 10–9σ0.67 1.952 10–10σ3 if σ 172.4 106
8.21 1013 exp (1.45 10–8σ) if 172.4 106 σ 690 106
σ is measured in Pa
Fig. 9.11 The effect of 1.0

temperature on the
modulus of elasticity
of (1) structural steels 1
0.8
and (2) steel reinforcing 2
bars [49]
0.6
E/E 0
0.4
0.2
0
0 100 200 300 400 500 600 700
Temperature (°C)
Fig. 9.12 Dilatometric 0.014

curve for steel
0.012
0.010
0.008
0
Δ /
0.006
0.004
0.002
0
0 100 200 300 400 500 600 700 800 900 1000
Temperature (°C)
Table 9.2 Critical temperature for various types of steel slower loss of strength and modulus throughout
Temperature 20–800 C temperature range as compared to
Steel Standard/reference ( C) mild steel. This is mainly due to the presence of
Structural steel ASTM 538 chromium and niobium, which improves fire
Reinforcing steel ASTM 593 resistance properties of steel.
Prestressing steel ASTM 426 Furthermore, tests by Kodur et al. [59]
Light-gauge EC 3 [51] 350 have shown that type of heat treatment has sig-
steel nificant influence on strength properties of steel
Gerlich et al. [55] 400
e.g. annealing and normalizing produces normal
strength steel, whereas quenching and tempering
2003 [57]) are mainly based on experimental produces high strength steel. High strength steel,
data on mild steel and do not account for specific produced using quenching and tempering pro-
property variations in new types of alloy steels. cess, and that is used in bolts (A490 bolts)
Recent research by Wang et al. [58] clearly possesses slightly lower thermal conductivity
show that high strength (Q460) steel exhibits than that of conventional mild steel.
www.ebook777.com
Fig. 9.13 Reduction of 1.0

the yield strength of cold- BS 5950: Part 856
formed light-gauge steel 0.9
2.0% strain
at elevated temperatures
[57–61] 0.8 1.5% strain
Strength retension factor, FyT /Fy

0.7 0.5% strain
0.6
0.5
Gerlich54
0.4
0.3
0.2
Makelainen and Miller 55
0.1
0.0
0 100 200 300 400 500 600 700
Temperature (°C)
Fig. 9.14 Modulus of 1.0

elasticity of cold-formed Makelainen and Miller55
light-gauge steel at 0.9 Steady-state tests
elevated temperatures
Normalized modulus elasticity, ET /E
[57–61] 0.8
0.7
0.6
0.5 Gerlich54
0.4
0.3
0.2
Makelainen and Miller55
0.1 Transient tests
0.0
0 100 200 300 400 500 600 700
Temperature (°C)
The properties of cold-formed light-gauge elevated temperatures, corresponding to 0.5 %,

steel are slightly different from those of 1.5 %, and 2 % strains based on the proposed
hot-rolled structural steel. Gerlich [60] and relationships and on the relationship in BS 5950
Makelainen and Miller [61], based on steady- [57]. The BS 5950 curves represent a conserva-
state and transient tests on cold-formed steel tive 95 % confidence limit (i.e., a 5 % chance that
tension coupons (cut from studs) and galvanized strength would fall below the curve), whereas the
sheets, proposed relationships for yield strength other two curves are representative of mean test
and modulus of elasticity. Figure 9.13 shows the data. Figure 9.14 shows the variation of modulus
variation of yield strength of light-gauge steel at of elasticity of light-gauge steel at elevated
temperatures. The modulus ET represents the principal aggregate. Also, a small amount of
tangent modulus at low stress levels (or initial discontinuous fibers (steel or polypropylene
tangent modulus), because steel stress-strain fibers) is often added to the concrete mix to
relationships become increasingly nonlinear at achieve superior performance; this concrete is
elevated temperatures. The effect of zinc coating referred to as fiber-reinforced concrete (FRC).
on the mechanical properties of steel is of little In this section, the properties of concrete are
significance. discussed under three groups: namely, NSC,
The light-gauge steel has somewhat lower FRC, and HSC.
thermal expansion when compared to similar
expressions for other steels [61]. The other ther-
mal properties of steel, such as specific heat and Normal-Strength Concrete
thermal conductivity, are of little importance for
the thermal modeling of light-gauge steel because A great deal of information is available in
steel framing plays a minor role in the heat trans- the literature on the mechanical properties of
fer mechanism. A review of some of these various types of normal-strength concrete.
properties is presented in a review paper [62]. This information is summarized in reports by
The critical temperature of light-gauge steel Bennetts [44] and Schneider [45], the ACI
is much lower than for other types of steels. guide [43], the ASCE fire protection manual
Although Eurocode 3 limits this to a conservative [29], and in Harmathy’s book [2]. Figure 9.15
value of 350 C, in other cases a critical temper- shows the stress-strain curves for a lightweight
ature of 400 C is used (see Table 9.2). concrete with expanded shale aggregate at room
temperature (24 C) and a few elevated tempera-
ture levels [63]. The shape of the curves may
Concrete depend on the time of holding the test specimen
at the target temperature level before the com-
Concrete is a Group L/I material. The word con- pression test.
crete covers a large number of different The modulus of elasticity (E) of various
materials, with the single common feature that concretes at room temperature may fall within a
they are formed by the hydration of cement. very wide range, 5.0 103–35.0 103 MPa,
Because the hydrated cement paste amounts to
only 24–43 volume percent of the materials pres- 16
ent, the properties of concrete may vary widely 24°C
with the aggregates used.
Traditionally, the compressive strength of 14 260°C
concrete used to be around 20–50 MPa, which

is referred to as normal-strength concrete (NSC). 12 538°C
In recent years, concrete with a compressive
σ (MPa)
strength in the range 50–100 MPa has become

760°C
widely used and is referred to as high-strength 8
concrete (HSC). Depending on the density,

concretes are usually subdivided into two major
4
groups: (1) normal-weight concretes with
densities in the 2150- to 2450-kgm–3 range and
(2) lightweight concretes with densities between 0
0 0.004 0.008 0.012
1350 and 1850 kgm–3. Fire safety practitioners –1)
ε (m • m
again subdivide the normal-weight concretes into
silicate (siliceous) and carbonate aggregate con- Fig. 9.15 Stress-strain curves for a lightweight masonry
crete, according to the composition of the concrete at room and elevated temperatures [63]
www.ebook777.com
dependent mainly on the water-cement ratio in of normal-weight concretes decreases faster with
the mixture, the age of concrete, the method of the rise of temperature than that of lightweight
conditioning, and the amount and nature of the concretes.
aggregates. Cruz found that the modulus of elas- The compressive strength (σu) of NSC may
ticity decreases rapidly with the rise of tempera- also vary within a wide range. Compressive
ture, and the fractional decline does not depend strength is influenced by the same factors as the
significantly on the type of aggregate [64] modulus of elasticity. For conventionally pro-
(in Fig. 9.16, E0 is the modulus of elasticity at duced normal-weight concretes, the strength at
room temperature). From other surveys [2, 44], it room temperature is usually between 20 and
appears, however, that the modulus of elasticity 50 MPa. For lightweight concretes, the strength
is usually between 20 and 40 MPa.
Information on the variation of the compres-
1.0
sive strength with temperature is presented in
Fig. 9.17 (for a silicate aggregate concrete),
Carbonate
0.8 (E0 = 34 000 MPa) Fig. 9.18 (for a carbonate aggregate concrete),
and Fig. 9.19 (for two lightweight aggregate
concretes, one made with the addition of natural
0.6 sand) [65]. ([σu]0 in the figures stands for the
compressive strengths of concrete at room tem-
E/E0
perature.) In some experiments, the specimens

0.4 Lightweight were heated to the test temperature without load
(E0 = 19 000 MPa)
(see curves labeled “unstressed”). In others
they were heated under a load amounting to
0.2 Sulicate
(E0 = 38 000 MPa)
40 % of the ultimate strength (see curves labeled
“stressed”). Again, in others they were heated to
0 the target temperature without load, then cooled
0 100 200 300 400 500 600
to room temperature and stored at 75 % relative
Temperature (°C)
humidity for six days, and finally tested at room
Fig. 9.16 The effect of temperature on the modulus of temperature (see curves labeled “unstressed
elasticity of concretes with various aggregates [61] residual”).
Fig. 9.17 The effect 1.0

of temperature on the
compressive strength Stressed
of a normal-weight 0.8
concrete with silicate
aggregate [65] Unstressed
0.6
σu /(σu )0
Unstressed
residual
0.4
Avg. initial
σu = 26.9 MPa
0.2
0
0 200 400 600 800
Temperature (°C)

temperature on the Stressed
compressive strength
of a normal-weight Unstressed
0.8
concrete with carbonate
aggregate [65]
0.6 Unstressed
σu /(σu)0
residual
0.4
Avg. initial
0.2 σ = 26.9 MPa
0
0 200 400 600 800
Temperature (°C)
Fig 9.19 The effect of 1.0

temperature on the Stressed
compressive strength of (sanded)
two lightweight concretes 0.8
(one with natural sand) [65]
Unstressed
residual
0.6 (sanded)
σu /(σu)0
Unstressed
0.4 (unsanded)
Unstressed
(sanded)
0.2
Avg. initial σu of "unsanded" concrete = 17.9 MPa
Avg. initial σu of "sanded" concrete = 26.9 MPa
0
0 200 400 600 800
Temperature (°C)
Some information on the creep of concrete at dilatometric curves bear the characteristic
elevated temperatures is available from the work features of the curve for the paste.
of Cruz [66], MareÂchal [67], Gross [68], and The dilatometric curves of two normal-weight
Schneider et al. [69] The creep curves shown in concretes (with silicate and carbonate
Fig. 9.20 are those recorded by Cruz for a normal- aggregates) and two lightweight concretes (with
weight concrete with carbonate aggregates. expanded shale and pumice aggregates) are
Because the aggregates amount to 60–75 % of shown in Fig. 9.21 [20]. These curves were
the volume of concrete, the dilatometric curve obtained in the course of a comprehensive study
usually resembles that of the principal aggregate. performed on 16 concretes.
However, some lightweight aggregates, for The results of dilatometric and thermogra-
example, pearlite and vermiculite, are unable to vimetric tests were combined to calculate the
resist the almost continuous shrinkage of the volumetric heat capacity (ρcp) versus tempera-
cement paste on heating, and therefore their ture relation for these four concretes, as shown in
www.ebook777.com
Fig. 9.22. The partial decomposition of the weight and lightweight concretes are shown in
aggregate is responsible for a substantial drop Fig. 9.23. This information, derived by combining
(above 700 C) in the density of concretes made thermodynamic data with thermogravimetric
with carbonate aggregate. observations [2, 6], has since been confirmed by
The aggregate type and moisture content have differential scanning calorimetry [7]. Experimen-
significant influence on the specific heat of con- tal data are also available on a few concretes and
crete. The usual ranges of variation of the volume- some of their constituents [2, 7].
specific heat (i.e., the product ρcp) for normal- The thermal conductivity (k) of concrete
depends mainly on the nature of its aggregates.
0.004 In general, concretes made with dense, crystal-
649°C
line aggregates show higher conductivities than
those made with amorphous or porous
0.002
aggregates. Among common aggregates, quartz
has the highest conductivity; therefore, concretes
0 made with siliceous aggregates are on the whole
more conductive than those made with other
0.002
482°C silicate and carbonate aggregates.
∈t, mm–1
Derived from theoretical considerations [6],

0 the solid curves in Fig. 9.24 describe the varia-
0.001 tion with temperature of the thermal conductivity
316°C
of four concretes. In deriving these curves, two
0
concretes (see curves 1 and 2) were visualized to
0.001
149°C represent limiting cases among normal-weight
0 concretes, and the other two (see curves 3 and
0.001 4), limiting cases among lightweight concretes.
24°C The points in Fig. 9.24 stand for experimental
0
0 1 2 3 4 5 data. They reveal that the upper limiting case is
Time, t probably never reached with aggregates in com-
Fig. 9.20 Creep of a carbonate aggregate concrete at
mon use and that the thermal conductivity of
various temperature levels (applied stress: 12.4 MPa; lightweight concretes may be somewhat higher
compressive strength of the material at room temperature: than predicted on theoretical considerations.
27.6 MPa) [66]

curves for two normal-
weight and two lightweight 2
0.010 1
concretes [20]. (1) normal-
weight concrete with
silicate aggregate, 0.005
(2) normal-weight concrete
0
with carbonate aggregate,

Δ /
(3) lightweight concrete 0

with expanded shale
aggregate, (4) lightweight –0.005 3
concrete with pumice
aggregate
–0.010 4
–0.015
0 100 200 300 400 500 600 700 800 900 1000 1100 1200
Temperature (°C)
Fig. 9.22 Volumetric heat 2100

capacity of two normal-
2000
weight and two lightweight
concretes [20]. (1) normal- 1900
weight concrete with
silicate aggregate, 1800
ρcp (MJ·m–3·K–1)
(2) normal-weight concrete 1700 1
with carbonate aggregate,
(3) lightweight concrete 1600
with expanded shale 1500
aggregate, (4) lightweight
2
concrete with pumice 1400
aggregate
1300
4 3
1200
1100
0 100 200 300 400 500 600 700 800 900 1000 1100 1200
Temperature (°C)
Fig. 9.23 Usual ranges of 7

variation for the volume-
specific heat of normal- 6
weight and lightweight
concretes [6] 5
–1
ρcp , MJ·m ·K
–3
4
Normal weight
3
1
Lightweight
0
0 200 400 600 800
Temperature (°C)
Fig. 9.24 Thermal

conductivity of four
“limiting” concretes and 2.5
some experimental thermal
conductivity data. 6,19 2.0 1
Symbols: ▼, ∇—various
k (W·m ·K )
–1
gravel concretes; ●—
–1
expanded slag concretes; 1.5

■, □—expanded shale 2
concretes; ○—pumice 1.0
concrete
3
0.5
4
0
0 200 400 600 800
Temperature (°C)
www.ebook777.com
Further experimental information on the ther- Studies have shown that polypropylene fibers in
mal conductivity of some normal-weight and a concrete mix are quite effective in minimizing
many lightweight concretes is available from spalling in concrete under fire conditions [71,
the literature [6, 7, 20]. 72]. The polypropylene fibers melt at a relatively
In reinforced concrete structures, the bond low temperature of about 170 C and create
between rebars and concrete (at elevated channels for the steam pressure in concrete to
temperatures) plays a major role in determining escape. This prevents the small explosions that
the fire endurance of structural members. cause the spalling of the concrete. Based on
Diederichs and Schneider investigated the variation these studies, the amount of polypropylene fibers
of bond strength between deformed and plain rebars needed to minimize spalling is about 0.1–0.25 %
and concrete as a function of temperature [70]. They (by volume). The polypropylene fibers were
found that the bond strength reduction follows the found to be most effective for HSC made with
same pattern as compressive strength for deformed normal-weight aggregate.
and rusted plain bars. However, higher reduction in The addition of fibers improves certain
bond strength was observed for new plain bars. mechanical properties, such as tensile strength,
They also found that the bond strength at elevated ductility, and ultimate strain, at room tempera-
temperature increases with decreasing coefficient ture. However, there is very little information on
of thermal expansion of concrete, which is signifi- the high-temperature properties of this type of
cantly influenced by the type of aggregate. concrete [73].
Diederichs and Schneider also concluded that the Steel fiber–reinforced concrete (SFRC)
water-cement ratio and the bar diameter have a exhibits, at elevated temperatures, mechanical
minor effect on the bond strength between steel properties that are more beneficial to fire resis-
and concrete [70]. Figure 9.25 illustrates the varia- tance than those of plain concrete. There is some
tion of bond strength as a function of temperature information available on SFRC’s material
for reinforced and prestressed concrete. properties at elevated temperatures. The effect
of temperature on the compressive strength for
two types of SFRC is shown in Fig. 9.26.
Fiber-Reinforced Concrete The strength of both types of SFRC exceeds the
initial strength of the concretes up to about
Steel and polypropylene discontinuous fibers are 400 C. This is in contrast to the strength of
the two most common fibers used in the concrete plain concrete, which decreases slightly with
mix to improve structural properties of concrete. temperatures up to 400 C. Above approximately
Fig. 9.25 Variation of

1.0
bond strength as a function
of temperature for
Strength (proportion of ambient)
reinforced and prestressed

0.8
concrete [70] Reinforcing steel
0.6
Prestressing steel
0.4
0.2
0
0 200 400 600 800
Temperature (°C)
Fig. 9.26 Effect of 140

temperature on
compressive strength of 120
steel fiber–reinforced Fiber-reinforced
concrete
concrete
Compressive strength
100
(% of initial strength)
80
60
40 Plain concrete
20
0
0 200 400 600 800
Temperature (°C)

temperature on tensile
strength of steel
fiber–reinforced concrete 100 Fiber-reinforced
concrete
(% of initial strength)
80
Tensile strength
60
40 Plain concrete
20
0
0 200 400 600 800
Temperature (°C)
400 C, the strength of SFRC decreases at an The type of aggregate has a significant
accelerated rate [74]. influence on the tensile strength of steel
The effect of temperature on the tensile fiber–reinforced concrete. The decrease in tensile
strength of steel fiber–reinforced carbonate strength for carbonate aggregate concrete is higher
concretes is compared to that of plain concrete than that for siliceous aggregate concrete [75].
in Fig. 9.27 [75]. The strength of SFRC decreases The thermal properties of SFRC, at elevated
at a lower rate than that of plain concrete temperatures, are similar to those of plain concrete.
throughout the temperature range, with the Kodur and Lie [27, 73] have carried out detailed
strength being significantly higher than that of experimental studies and developed dilatometric
plain concrete up to about 350 C. The increased and thermogravimetric curves for various types of
tensile strength delays the propagation of cracks SFRC. Based on these studies, they have also
in fiber-reinforced concrete structural members developed expressions for thermal and mechanical
and is highly beneficial when the member is properties of steel fiber–reinforced concrete in the
subjected to bending stresses. temperature range 0–1000 C [18, 76].
www.ebook777.com
High-Strength Concrete Kodur and Sultan have presented detailed

experimental data on the thermal properties of
The strength of concrete has significant influence HSC (for both plain and steel fiber–reinforced
on the properties of HSC. The material properties concrete types) [80]. The type of aggregate has
of HSC vary differently with temperature than significant influence on the thermal properties
those of NSC. This variation is more pronounced of HSC at elevated temperatures. Figure 9.29
for mechanical properties, which are affected by shows the thermal conductivity and specific
these factors: compressive strength, moisture heat of HSC, with siliceous and carbonate
content, density, heating rate, percentage of silica aggregates, as a function of temperature. Based
fume, and porosity [77]. The available informa- on the test data, Kodur and Sultan have proposed
tion on the mechanical properties of HSC at relationships for thermal conductivity, specific
elevated temperatures is presented in a review heat, thermal expansion, and mass loss of HSC
report by Phan [30]. as a function of temperature [81].
The loss in compressive strength with temper- The variation of thermal expansion with
ature is higher for HSC than that for NSC up to concrete temperature for siliceous and carbonate
about 450 C. Figure 9.28 shows the comparison aggregate HSC is similar to that of NSC, with
of strengths for NSC and HSC types, together the aggregate having a strong influence. Overall,
with CEB and European design curves for NSC. the thermal properties of HSC, at elevated
The difference between compressive strength temperatures, are similar to those of NSC [82].
versus temperature relationships of normal- HSC, due to low porosity, is more susceptible
weight and lightweight aggregate concrete is to spalling than NSC, and explosive spalling
not significant. However, HSC mixture with sil- may occur when HSC is exposed to severe fire
ica fume have higher compressive strength loss conditions. Hence, one of the major concerns for
with increasing temperature than HSC mixture the use of HSC is regarding its behavior in fire,
without silica fume. Based on a series of high- in particular, the occurrence of spalling at
temperature material property tests, Kodur elevated temperatures. For predicting spalling
et al. have proposed a set of stress-strain performance, knowledge of the variation of
relationships for HSC as a function of tempera- porosity with temperature is essential. Fig-
ture [78, 79]. The variation, with temperature, of ure 9.30 shows the variation of porosity with
modulus of elasticity and tensile strength of HSC temperature for NSC and HSC. The data in
is similar to that of NSC. this figure are taken from the measurements
Fig. 9.28 Comparison 1.2

of design compressive
strength and results
of unstressed tests 1
NSC
of lightweight aggregate
concrete [30]
0.8
Fc /Fc (20°C)
0.6
0.4 HSC
CEB design curve

0.2
Eurocode design curve
0
0 200 400 600 800 1000
Temperature (°C)
a
2.5 b 10
Thermal conductivity (W/m°C)
Siliceous
2.0 8 aggregate HSC
Specific heat (KJ/kg°C)

Carbonate
1.5 6 aggregate HSC
1.0 Siliceous 4
aggregate HSC
0.5 Carbonate 2
aggregate HSC
0 0
0 200 400 600 800 0 200 400 600 800 1000
Temperature (°C) Temperature (°C)
Fig. 9.29 Thermal conductivity and specific heat capacity of HSC as a function of temperature: [80] (a) thermal
conductivity of high-strength concrete and (b) specific heat of high-strength concrete
Fig. 9.30 Porosity of HSC 10

and NSC as a function of
NSC
temperature [36]
8
Porosity (%)
6 HSC
0
0 100 200 300 400 500 600
Temperature (°C)
of porosity after exposure to different a b

temperatures [36].
The spalling in HSC can be minimized by
creating pores through which water vapor can
be relieved before vapor pressure reaches critical
values. This is usually done by adding polypro-
pylene fibers to the HSC [71, 72, 83]. Also,
Kodur et al. have reported that spalling in HSC
columns can be minimized to a significant extent
by providing bent ties as lateral confinement Conventional Modified
[77, 84]. Figure 9.31 illustrates conventional tie configuration tie configuration
and improved tie configuration for minimizing
spalling in HSC columns [84]. Fig. 9.31 Tie configuration for achieving higher fire
resistance in concrete structures [79]
www.ebook777.com
of magnitude greater than the stresses allowed in

Brick the design of grouted brickwork. Because brick is
rarely considered for important load-bearing
Building brick belongs in the L/I group of
roles in buildings, there has been little interest
materials. The density (ρ) of bricks ranges from
in the mechanical properties of bricks at elevated
1660 to 2270 kgm–3, depending on the raw
temperatures.
materials used in the manufacture, and on the At room temperature, the coefficient of
molding and firing technique. The true density thermal expansion (α) for clay bricks is about
of the material (ρt) is somewhere between 2600 5.5 106 mm–1 K–1. The dilatometric
and 2800 kgm–3. and thermogravimetric curves for a clay brick
The modulus of elasticity of brick (E) is usu- of 2180 kgm–3 density are shown in Fig. 9.32
ally between 10 103 and 20 103 MPa. Its [7]. The variation with temperature of the spe-
compressive strength (σu) varies in a very wide cific heat and the thermal conductivity of this
range, from 9 to 110 MPa—50 MPa may be brick is shown in Figs 9.33 and 9.34,
regarded as average [85]. This value is an order respectively [7].

and thermogravimetric
curves for a clay brick [7]
0.010
0
Δ /
0.005
1.05
1.00
0.95
0.90
M/M 0
0.85
0.80
0.75
0.70
0 100 200 300 400 500 600 700 800 900 1000
Temperature (°C)
Fig. 9.33 Apparent 3500

specific heat of a clay
brick [7] 3000
2500
Cp (J·kg–1·K–1)
2000
1500
1000
500
0
0 100 200 300 400 500 600 700 800 900 1000
Temperature (°C)
Fig. 9.34 Thermal 1.75

conductivity of a clay
brick. Symbols: 1.50
○—heating cycle,
●—after cooling [7]
1.25
k, W·m–1·K–1
1.00
0.75
0.50
0.25
0
0 100 200 300 400 500 600 700 800 900 1000
Temperature (°C)
locust). The density of Douglas firs varies from

Wood 430 to 480 kgm–3 and that of southern pines
from 510 to 580 kgm–3. The true density of the
Wood is a Group L/I/F or I/F material. As struc- solid material that forms the walls of wood cells
tural members, wood is widely used in residential (αt) is about 1500 kgm–3 for all kinds of wood.
and low-rise constructions. Although about The density of wood decreases with temperature;
180 wood species are commercially grown in the density ratio (ratio of density at elevated
the United States, only about 25 species have temperature to that at room temperature) drops
been assigned working stresses. The two groups to about 0.9 at 200 C and then declines sharply
most extensively used as structural lumber are to about 0.2 at about 350 C [40].
the Douglas firs and the southern pines. Wood is an orthotropic material, so the
The oven-dry density (ρ) of commercially strength and stiffness in longitudinal and trans-
important woods ranges from 300 kgm–3 verse directions are influenced by grain orienta-
(white cedar) to 700 kgm–3 (hickory, black tion. The mechanical properties of wood are
www.ebook777.com

temperature on the
modulus of elasticity
and compressive strength 0.8
of wood [87–89] E/E0
E/E0, or σu /(σu)0
0.6
σu/(σu)0
0.4
0.2
0
0 50 100 150 200 250 300
Temperature (°C)
affected by temperature and are influenced by The coefficient of linear thermal expansion
moisture content, rate of charring, and grain ori- (β) ranges from 3.2 10–6 to 4.6
entation. The modulus of elasticity (E) of air-dry, 10–6 mm–1K–1 along the grain and from
clear wood along the grain varies from 5.5 103 21.6 10–6 to 39.4 10–6 mm–1K–1 across
to 15.0 103 MPa, and its crushing strength (σu) the grain [91]. Wood shrinks at temperatures
varies from 13 to 70 MPa. These properties are above 100 C, because of the reduction in mois-
related and roughly proportional to the density, ture content. Lie [29] reported that the amount of
regardless of the species [86]. shrinkage can be estimated as 8 % in the radial
Figure 9.35 shows the variation of the direction, 12 % in the tangential direction, and an
modulus of elasticity and compressive strength average of 0.1–0.2 % in the longitudinal direc-
of oven-dry, clear wood with temperature tion. The dilatometric and thermogravimetric
[87–89]. (E0 and [σu]0 in the figure are modulus curves of a pine with a 400 kgm–3 oven-dry
of elasticity and compressive strength at room density are shown in Fig. 9.36 [7].
temperature, respectively.) The modulus of elas- The thermal conductivity (k) across the grain
ticity decreases slowly with temperature up to of this pine was measured as 0.86–1.07
about 200 C, when it reaches about 80 %, and Wm–1K–1 between room temperature and
then the decline is more rapid. The compressive 140 C [14]. The thermal conductivity increases
strength also drops linearly to about 80 % at initially up to a temperature range of 150–200 C,
about 200 C, and then the drop is more rapid— then decreases linearly up to 350 C, and
to about 20 % around 280 C. finally increases again beyond 350 C.
The tensile strength exhibits behavior similar Figure 9.37 shows the apparent specific heat
to that of compressive strength, but the decline in for the same pine, as a function of temperature
tensile strength with temperature is less rapid. [7]. The accuracy of the curve (developed by
The moisture content plays a significant role differential scanning calorimeter [DSC]) is some-
in determining the strength and stiffness, with what questionable. However, it provides useful
increased moisture content leading to higher information on the nature of decomposition
reduction. There is very little information on reactions that take place between 150 and 370 C.
stress-strain relationships for wood. The formulas Charring is one of the main high-temperature
for reduced stiffness and design strength can properties associated with wood and should be
be found in Eurocode 5 [90] (Part 1.2). considered in predicting performance under fire

curves for a pine of 0.01
400 kgm–3 density [7]
0
–0.01
Δ / 0
–0.02
–0.03
–0.04
–0.05
1.2
1.0
0.8
M/M 0
0.6
0.4
0.2
0
0 100 200 300 400 500 600 700 800 900 1000
Temperature (°C)
Fig. 9.37 Apparent 3500

specific heat for a pine
of 400 C density [7] 3000
2500
Cp (J·kg–1·K–1)
2000
1500
1000
500
0
0 100 200 300 400 500 600 700 800 900 1000
Temperature (°C)
www.ebook777.com
conditions. The rate of charring is influenced by presence of compressed plies [92]. Comparison
the radiant heat flux or, alternatively, the fire of charring rates indicate that engineered lumber
severity. Generally, a constant transverse-to- has higher rate of charring rate as compared to
grain char rate of 0.6 mm/min can be used for conventional wood [92].
woods subjected to standard fire exposure [29].
The charring rate parallel to the grain of wood is
approximately twice the rate when it is transverse Fiber-Reinforced Polymers
to the grain. These charring rates should be used
only when attempting to model the performance In recent years, there has been a growing interest in
of wood sections in the fire resistance furnace. the use of fiber-reinforced polymers (FRPs) in civil
Charring is influenced by a number of engineering applications due to the advantages,
parameters, the most important ones being den- such as high strength and durability (resistance to
sity, moisture content, and contraction of wood. corrosion), that FRP offers over traditional
It is reasonable to modify the 0.6 mm/min to materials. FRP composites consist of two key
approximately 0.4 mm/min for moist dense elements, namely the fibers (glass, carbon, or
wood or to 0.8 mm/min for dry and light wood. aramid) and a thermosetting polymer matrix such
The fire retardants often used to reduce flame as epoxy, vinyl ester, phenolic, or polyester
spread in wood may only slightly increase the resin. The commonly used types of FRP composite
time until ignition of wood. materials are glass fiber–reinforced plastic
Specific charring rates for different types of (GFRP), carbon fiber–reinforced plastic (CFRP),
wood can be found in “Structural Fire Protec- and aramid fiber–reinforced plastic (AFRP)
tion” [29] and Bénichou and Sultan [40]. composites. FRPs are similar to wood in that they
Eurocode [90] gives an expression for charring will burn when exposed to fire and can be classi-
depth in a wood member exposed to standard fied as an L/I/F type material.
fire. The dependence of charring rate on the FRP is used as an internal reinforcement
radiant heat flux is discussed in Wood (reinforcing bars as an alternative to traditional
Handbook [87]. steel reinforcement) and as external reinforce-
In recent years different types of engineered ment in forms, such as wrapping and sheeting
wood is widely used in residential construction. for the rehabilitation and strengthening of con-
These engineered wood products (ex: joists and crete members. One of the main impediments to
studs) capitalize on the strength of wood and the using FRPs in buildings is the lack of knowledge
efficiency of the sectional shapes (ex: I-shaped about the fire resistance of FRP [93, 94].
joists) to enhance load bearing capacity at There are some major differences associated
ambient conditions, while at the same time with FRP as a material. The properties depend
reducing the mass and cost of the structural on the type and composition of FRP, and the
member. However, there is very limited data on availability of various types of FRP makes it
high temperature thermo-mechanical properties difficult to establish the properties at elevated
of engineered lumber and fire resistance of temperatures. The material properties are con-
engineered joists and studs. Limited research trolled by the fibers in the longitudinal direction
has clearly shown that fire resistance of and by the matrix in the transverse direction. In
engineered joists to be significantly lower than addition to thermal and mechanical properties,
that of conventional wood joists [92]. This was factors such as burning, charring, evolution of
mainly attributed to poor thermal, mechanical smoke, and toxicity in fire also play a significant
and charring properties of engineered lumber as role in determining the fire performance. A sum-
compared to conventional wood products. Typi- mary of typical mechanical properties for various
cally, room temperature thermal conductivity types of FRPs, in comparison to other commonly
and modulus of elasticity of engineered lumber used construction materials, at room tempera-
is higher than other types of wood due to the ture, is presented in Table 9.3.
Table 9.3 Properties of various FRP composites and other materials
Modulus Modulus Tensile Comp. Shear Shear Poisson’s Tensile Comp.

of elasticity of elasticity strength strength modulus strength ratio strength strength
σt1 σc1 G S σc2
Material E1 (MPa) E2 (MPa) (MPa) (MPa) (MPa) (MPa) ν σ2 (MPa) (MPa)
GFRP (glass/ 55,000 18,000 1050 1050 9000 42 0.25 28 140
epoxy)
GFRP (glass/ 42,000 12,000 700 — 5000 72 0.30 30 —
epoxy)
unidirectional
CFRP (carbon/ 180,000 10,000 1500 — 7000 68 0.28 40 —
epoxy)
unidirectional
CFRP (graphite/ 207,000 5200 1050 700 2600 70 0.25 40 120
epoxy)
Boron/epoxy 207,000 21,000 1400 2800 7000 126 0.30 84 280
ARP (aramid/ 76,000 8000 1400 — 3000 34 0.34 12 —
epoxy)
unidirectional
Mild steel 200,000 — 550 240 — 380 — — —
Concrete (normal 31,000 — 4 40 — 7 0.15–0.20 — —
strength)
Wood (Douglas 9800 — 69 — — — — — —
fir)
E1 ¼ modulus of elasticity in longitudinal direction
E2 ¼ modulus of elasticity in transverse direction
There is very little information on the material about 75 % at a temperature of 200 C. The strain
properties of FRPs at elevated temperatures [93]. level, for a given stress, is also higher with the
The impact of high temperatures on the behavior increase in temperature. Recently, Wang and
of FRP composites is severe degradation of their Kodur reported high temperature strength and
properties: reduction of strength and stiffness, stiffness properties of glass and carbon FRP
and increase in deformability, thermal expan- rebars; full details of the tests are reported in
sion, and creep. Above 100 C temperature, the Wang and Kodur [96].
degradation can be quite rapid as the glass tran- The variation of strength with temperature
sition temperature of the matrix is reached. (ratio of strength at elevated temperature to that
The glass transition temperature, which is at room temperature) for FRP along with that of
often considered the upper use temperature, other traditional construction materials is shown
varies with the type of resin used and was in Fig. 9.2. The curve showing the strength deg-
found to be as low as 100 C in some resins and radation of FRP is based on the limited informa-
as high as 220 C in others. From the limited tion reported in the literature [93, 95]. The rate of
studies, it appears that as much as 75 % of the strength loss is much greater for FRP than for
GFRP strength and stiffness is lost by the time concrete and steel, resulting in a 50 % strength
the temperature reaches 250 C [93, 95]. loss by about 200 C.
The stress-strain relationships, from the stud- The bond between FRPs and concrete
ies conducted by Gates [95], for a CFRP com- (or between FRP layers or lap splices in multiply
posite (IM7/5260) are shown in Fig. 9.38 for layup applications) is essential to transfer loads.
various temperatures. It can be seen that the This load transfer occurs through the polymer
tensile strength of IM7/5260 composite reduces resin matrix and thus relies heavily on the
to approximately 50 % at about 125 C and to mechanical properties of the polymer.
www.ebook777.com
Fig. 9.38 Tensile stress- 200

strain curves for CFRP at
various temperatures [95] IM7/5260
23°C
IM7/8320
− (MPa)
150
Tension
70°C
Effective stress σ
100
125°C
150°C
175°C 23°C–70°C
50
200°C 125°C–150°C
175°C
200°C
0
0 200 400 600
Effective plastic strain ε–p
Deterioration of the mechanical properties of the Thus, bond degradation at elevated temperature
matrix material at temperatures above the spe- is a critical factor to be considered in the design
cific polymer’s glass transition temperature, Tg, of FRP-reinforced or -strengthened concrete
have the potential to cause loss of bond at only members. This was observed in full-scale fire
modestly increased temperatures, resulting in tests on FRP-strengthened reinforced concrete
loss of interaction between FRP and concrete. columns [102].
The glass transition temperature of commonly The critical temperature of FRP is much lower
used polymer matrix materials is typically in than that for steel and depends on the composition
the range of 65–140 C. of fibers and matrix. Kodur and Baingo
No specific research has yet been reported on have assumed a critical temperature of 250 C in
the bond between concrete and externally bonded modeling the behavior of FRP-reinforced concrete
FRP strengthening systems at high temperature, slabs [93]. Recently, Wang and Kodur have devel-
although limited data on the high-temperature oped critical temperature information for glass
residual performance of the FRP concrete bond and carbon FRP reinforcing rebars [103, 104].
has recently been presented [97]. They carried out a series of tensile strength
Research on the bond properties of FRP bars tests at high temperatures on two types of com-
for concrete reinforcement applications (internal mercially available FRP rebars. This included
reinforcement) at elevated temperature has been both carbon FRP and glass FRP bars of different
reported in the literature [98–101]. This work has diameters. Conventional steel rebars were also
indicated that dramatic decreases in bond strength tested for comparison. The data were used to
can be expected, to values of about 10 % of room determine the variation of average failure strength
temperature strength, at temperatures between and elastic modulus for each type of reinforcement
100 and 200 C (i.e., at temperatures close to or with increasing temperature. Full details of exper-
above Tg). The observed bond strength reductions imental studies, including specimen preparation,
have been attributed to changes in the properties of test setup, test procedure, and observations as well
the polymer matrix at the surface of the FRP bars. as test data, are described elsewhere [96, 104].
It seems clear that temperature effects on the A summary of the results of these studies are
FRP–FRP and FRP–concrete bond are critical, shown in Fig. 9.39. For the GFRP and CFRP
both in FRP internal reinforcement and in exter- bars, observed failure strengths were used,
nally bonded FRP applications, and a great deal whereas for the steel bars, the 0.2 % proof stress
of additional research is required in this area. was used. The elastic modulus was taken as the
Fig. 9.39 Effect of

strength degradation as 9.5 mm CFRP rebar
a function of temperature 9.5 mm GFRP rebar
for FRP [96] 1.2
10 mm steel rebar
12.7 mm GFRP rebar
1.0
15 mm steel rebar
Normalized strength
0.8
0.6
0.4
0.2
250 325 550

0
0 100 200 300 400 500 600 700 800
Temperature (°C)
Table 9.4 Thermal properties of various FRPs and other materials at room temperature
Coefficient of thermal expansion Thermal conductivity k
(unidirectional) (β: 10–6 m m–1 C) (W m–1 C–1)
Material Longitudinal αL Transverse αΤ Longitudinal kL Transverse kT
Glass/epoxy (S-glass) 6.3 19.8 3.46 0.35
Glass/epoxy (E-glass: 63 % fiber) 7.13 — — —
Carbon/epoxy (high modulus) –0.9 27 48.4–60.6 0.865
Carbon/epoxy (ultra-high modulus) –1.44 30.6 121.1–129.8 1.04
Boron/epoxy 4.5 14.4 1.73 1.04
Aramid/epoxy (Kevlar 49) –3.6 54 1.73 0.73
Concrete 6.16 1.36–1.90
Steel 10.8–18 15.6–46.7
Epoxy — 54–90 — 0.346
slope of a straight line fitted to the initial linear result that agrees well with published data avail-
portion of the recorded stress-strain relationship able in the literature.
for each specimen. The critical temperature for The variation of elastic moduli of FRP with
the FRP reinforcement was derived based on a temperature is different in each direction. Typical
50 % tensile strength reduction, as is the case for values for various types of FRP are given in
steel reinforcement. This resulted in critical Table 9.3 [93]. The three values represent the
temperatures of about 325 C and 250 C for longitudinal, transverse, and shear moduli, respec-
GFRP and CFRP reinforcing bars, respectively. tively, of different unidirectional FRPs. At high
These critical temperatures are significantly less temperature, the elastic moduli of FRPs decreases
than 593 C, the critical temperature for steel at a faster rate than that for concrete or steel.
reinforcement, thus highlighting the presumed Similar to mechanical properties, the thermal
susceptibility of FRP reinforcement to fire. properties of FRP are also dependent on direc-
Figure 9.39 also shows that the steel reinforcing tion, fiber type, fiber orientation, fiber volume
bars in these tests lost about 50 % of their room- fraction, and laminate configuration. Table 9.4
temperature yield strength at about 550 C, a shows thermal properties for various types of
www.ebook777.com
FRP at room temperature. In the longitudinal is likely due to the fact that such information
direction, the thermal expansion of FRPs is is proprietary to the composite materials’
lower than that of steel. However, in the trans- manufacturers. Also, there is not much informa-
verse direction, it is much higher than that of tion on evolution of smoke and toxins in FRP
steel. Some of the information available in the composites exposed to fire.
literature can be found in a review report by Thermal expansion of FRP reinforcement
Kodur and Baingo [93]. At room temperatures, varies in longitudinal and transverse directions,
FRPs in general have low thermal conductivity, and the coefficient of thermal expansion highly
which makes them useful as insulation materials. depends on type of fiber, resin, and volume frac-
With the exception of carbon fibers, FRPs have a tion of fiber. The longitudinal coefficient of
low thermal conductivity. thermal expansion is dominated by properties of
Information on the thermal properties of FRP the fiber, while the transverse coefficient is
at elevated temperatures is very scarce, which dominated by properties of the resin. Figure 9.40
a
25
GFRP100-T1 GFRP100-T2
20
CFRP200-T1 CFRP200-T2
Thermal strain (10−3)
15
10
0
0 50 100 150 200 250 300 350
−5
Temperature (°C)
b 1
0.5
0
0 50 100 150 200 250 300 350
−0.5
Thermal strain (10−3)
−1
−1.5
−2
GFRP100-L1 GFRP100-L2
−2.5
−3
CFRP200-L1 CFRP200-L2
−3.5
−4
Temperature (°C)
Fig. 9.40 Variation of thermal strain in GFRP and CFRP in (a) longitudinal and (b) transverse directions as a function
of temperature
(a and b) shows longitudinal and transverse regular gypsum. When the core of the gypsum
coefficients of thermal expansion for typical board is modified with special core additives or
GFRP and CFRP bars. It can be noted that usu- with enhanced additional properties, to improve
ally there is a change in expansion rate at around the natural fire resistance from regular gypsum
glass transition temperature (Tg), indicating FRP board, it is classified as type X or improved type
reinforcement experiences different coefficients X gypsum board. There might be significant var-
of thermal expansion before and after phase iation in fire performance of the gypsum board
change (Tg). In transverse direction, the dimen- based on the type and the formulation of the core,
sion of GFRP and CFRP rebars increase with which varies from one manufacturer to another.
temperature, and GFRP undergoes higher ther- Gypsum is an ideal fire protection material.
mal expansion than that of CFRP. However, in The water inside the gypsum plays a major role
longitudinal direction, GFRP rebar slightly in defining its thermal properties and response to
expands with temperature, but CFRP rebar fire. On heating, it will lose the two H2O
contracts with increase in temperature. The molecules at temperatures between 125 and
coefficients of thermal expansion in transverse 200 C. The heat of complete dehydration is
direction for GFRP and CFRP rebars can be 0.61 106 J/kg gypsum. Due to the substantial
taken to be 64.5 and 7.79 106/ C, respec- absorption of energy in the dehydration process,
tively, while the corresponding coefficients of a gypsum layer applied to the surface of a build-
thermal expansion in longitudinal direction are ing element is capable of markedly delaying the
2.48 and 7.6 106/ C, respectively [105] penetration of heat into the underlying load-
bearing construction.
The thermal properties of the gypsum board
Gypsum vary depending on the composition of the core.
The variation with temperature of the volume
Gypsum (calcium sulfate dihydrate: CaSO4 specific heat (ρcp) of pure gypsum has been
2H2O) is a Group I material. Gypsum board is illustrated in Harmathy [106], based on informa-
produced by mixing water with plaster of paris tion reported in the literature [107, 108]. The
(calcium sulfate hemihydrate: CaSO41/2H2O) or thermal conductivity of gypsum products is
with Keene’s cement (calcium sulfate anhydrite: difficult to assess, owing to large variations in
CaSO4). The interlocking crystals of CaSO4 their porosities and the nature of the aggregates.
2H2O are responsible for the hardening of the A typical value for plaster boards of about
material. 700 kgm–3 density is 0.25 Wm–1K–1.
Gypsum products are used extensively in the Figures 9.41 and 9.42 illustrate the typical varia-
building industry in the form of boards, including tion of the thermal conductivity and the specific
wallboard, formboard, and sheathing. The core heat, respectively, of the gypsum board core with
of the boards is fabricated with plaster of paris, temperature. The plots reflect the expressions
into which weight- and set-controlling additives proposed recently by Sultan [109], based on
are mixed. Furthermore, plaster of paris, with the tests conducted on type X gypsum board
addition of aggregates (such as sand, pearlite, specimens. The specific heat measurements
vermiculite, or wood fiber) is used in wall plaster were carried out at a heating rate of 2 C/min.
as base coat, and Keene’s cement (neat or mixed The dehydration of gypsum resulted in the two
with lime putty) is used as finishing coat. peaks that appear in the specific heat curve at
Gypsum board, based on composition and temperatures around 100 C and 650 C. The
performance, is classified into various types, peak values are slightly variant to those reported
such as regular gypsum board, type X gypsum earlier by Harmathy [16]; this may be due to the
board, and improved type X gypsum board. A differences in gypsum composition.
gypsum board with naturally occurring fire resis- The coefficient of thermal expansion (β) of
tance from the gypsum in the core is defined as gypsum products may vary between
www.ebook777.com
Fig. 9.41 Thermal 0.6

conductivity of type X
gypsum board core as a
0.5
Thermal conductivity [W/(m°C)]

function of temperature
[109]
0.4
0.3
0.2 Sultan [109]
0.1
0
0 100 200 300 400 500 600 700 800 900 1000
Temperature (°C)
Fig. 9.42 Specific heat of 20

type X gypsum board core
as a function of
temperature [109]
15
Specific heat [kJ/(kg°C)]
10
5
Sultan (1996)
Heating rate: 2°C/min
0
0 100 200 300 400 500 600 700 800 900 1000
Temperature (°C)
11.0 10–6 and 17 10–6 mm–1K–1 at room North American gypsum board products. The
temperature, depending on the nature and attachment details (screw spacing, orientation
amount of aggregates used. The dilatometric of gypsum board joints, stud spacing, etc.) may
and thermogravimetric curves of a so-called have a noticeable effect on the fire performance
fire-resistant gypsum board of 678 kgm–3 den- of the gypsum board.
sity are shown in Fig. 9.43.
There is not much information about the
mechanical properties of the gypsum board at Insulation
elevated temperatures because these properties
are difficult to obtain experimentally. The Insulation is a Group I material and is often used
strength of gypsum board at an elevated temper- as a fire protection material both for heavy struc-
ature is very small and can be neglected. The tural members such as columns and beams and
Gypsum Association [110] lists typical mechani- for lightweight framing assemblies such as floors
cal properties, at room temperature, for some and walls. The insulation helps delay the

curves for a gypsum board 0.01
of 678 kgm–3 density [7]
0
–0.01
Δ / 0
–0.02
–0.03
–0.04
–0.05
1.05
1.00
0.95
0.90
M/M 0
0.85
0.80
0.75
0.70
0 100 200 300 400 500 600 700 800 900 1000
Temperature (°C)
temperature rise of structural members, thereby temperatures are mainly due to variation in the
enhancing fire resistance. There are a number of chemical composition of fiber.
insulation materials available in the market. Min- Full-scale fire resistance tests on walls and
eral wool and glass fiber are the two most widely floors have shown that the mineral fiber insula-
used insulation materials in walls and floors. tion performs better than glass fiber insulation.
Other insulation materials used for fire protection This is mainly because glass fiber melts in the
include intumescent paints, spray mineral fibers, temperature range of 700–800 C and cannot
insulation boards, and compressed fiber board. withstand direct fire exposure. The melting
The thermal properties of insulation play an point for mineral fiber insulation is higher.
important role in determining the fire resistance. The density of glass fiber is about 10 kg/m3 and
However, there is not much information avail- is much lower than that of rock fiber, which is
able on the thermal properties of various types about 33 kg/m3.
of insulation. Figure 9.44 shows the variation of The mineral wool insulation, when installed
thermal conductivity with temperature for glass tightly between the studs, can be beneficial for
and rock fiber insulation types. The differences the fire resistance of non-load-bearing steel stud
in thermal conductivity values at higher walls because it acts as an additional fire barrier
www.ebook777.com
Fig. 9.44 Thermal 2

conductivity of insulation
as a function of Rock fiber
temperature [40] Glass fiber
Thermal conductivity (W/m°C)

1.6
Rock fiber
1.2
0.8
0.4
0
0 200 400 600 800 1000
Temperature (°C)
Table 9.5 Properties of some commonly used insulation materials [105]
Equilibrium
Density Thermal conductivity Specific heat moisture content
Material ρ (kg/m3) k (W/mK) c (J/kgK) %
Spray
Sprayed mineral fibers 300 0.12 1200 1
Perlite or vermiculite plaster 350 0.12 1200 15
High-density perlite or vermiculite plaster 550 0.12 1200 15
Boards
Fiber silicate or fiber calcium silicate 600 0.15 1200 3
Gypsum plaster 800 0.2 1700 20
Compressed fiber boards
Mineral wool, fiber silicate 150 0.2 1200 2
after the fire-exposed gypsum board falls off gypsum, cementitious and mineral fiber and
[111]. On the other hand, cavity insulation other additives such as vermiculite.
slows down the flow of heat through the wall The thermal properties of some of the com-
assembly and can cause an accelerated tempera- monly used insulation systems are given in
ture rise in the fire-exposed gypsum board. Table 9.5 [112]. It should be noted that these
Another common form of fire insulation values are average property values and can vary
applied on steel structural members to achieve depending on the manufacturer and on the
required fire resistance is spray applied fire resis- proportions of different constituent materials.
tive materials (SFRM), which work by delaying Also the moisture content of the insulation mate-
temperature rise in steel. SFRM, available under rial has an effect on the thermal properties.
different trade names, offers several advantages The above listed thermal properties for fire
over other types of fire insulation such as cost insulation are at room temperature and they can
effectiveness, ease of application, and light vary significantly with temperature and also with
weight, and therefore is widely used as fire insulation composition, which can vary for dif-
proofing material for steel structures. SFRM is ferent trade names (from different commercial
mainly composed of base materials such as manufactures) among the same type of insulation
(ex: SFRM). However, in practice fire resistance conductivity among three types of SFRM well
of insulated structural (steel) members is pertains to the variation in their densities and also
evaluated by considering only room temperature to composition of ingredients in each type. The
thermal properties of fire insulation [113]. This is trends in the figure further indicate that tempera-
mainly due to lack of reliable data on the effect of ture has significant effect on thermal conductiv-
temperature on thermal properties of fire insula- ity of SFRM. This variation in thermal
tion. Further, there is no data on relative thermal conductivity at higher temperatures is primarily
performance of similar fire insulation products governed by changes in moisture content and
(ex: SFRM) produced from different commercial density of different SFRM types.
manufactures. Insulation materials such as SFRM experience
Figure 9.45a shows variation of thermal con- shrinkage at higher temperatures, as opposed to
ductivity with temperature for three types of expansion phenomenon in materials such as
commercially available SFRM (A, B, and C) steel, concrete and wood. The variation of ther-
generated in a recent research study [114]. The mal strain for three types of SFRM is plotted as a
thermal conductivity of three SFRM types at function of temperature in Fig. 9.45b [114]. This
room temperature is in the range of 0.07 and variation of thermal strain with temperature is
0.2 W/m.K. This variation of thermal also linked to changes in moisture content.
Fig. 9.45 Effect of a

temperature on (a) thermal 0.3
conductivity and (b) SFRM A
thermal contraction, 0.25 SFRM B
of different SFRMs
Thermal Conductivity (W/mK)
SFRM C
0.2
0.15
0.1
0.05
0
0 100 200 300 400 500 600 700 800
Temperature (°C)
Temperature (°C)
b
0 200 400 600 800 1000 1200
5
0
−5
−10
Strain(ΔL/L)x10-3
−15
−20
−25 SFRM A
−30 SFRM B
−35 SFRM C
−40
−45
−50
www.ebook777.com
Table 9.6 Density of SFRM at room temperature and main inputs needed in these models is the mate-
after exposure to 700 C rial properties at elevated temperatures. The ther-
Density (Kg/m3) mal and mechanical properties of most materials
Room temp. Decrease change substantially within the temperature
Insulation type (20 C) 700 C in density (%) range associated with building fires.
SFRM A 298 241.3 19.0 Even to date, there is lack of adequate knowl-
SFRM B 423.2 349.8 17.3 edge of the behavior of many building materials
SFRM C 451.8 381.2 15.6 at elevated temperatures. Although there is suffi-
cient information available for some materials,
However, the loss of moisture content only such as normal-strength concrete and steel, there
account for the shrinkage phenomenon that is a complete lack of information on certain
occurs in 100–400 C range. The intermediate properties for widely used materials, such as
expansion resulting in increase in thermal strains wood, insulation, and so on. Often, traditional
in 400–800 C range is dictated by the expansion materials are being modified (e.g., high-strength
of intumescent material, such as, vermulite, concrete) to enhance their properties at room
which is added to SFRM to counteract shrinkage temperatures without giving due consideration
and the percentage of Vermiculite in SFRM has to elevated temperatures. In many cases, these
major influence on the level of contraction. modifications will cause the properties to deteri-
The change in density for three types of orate at elevated temperatures and introduce
SFRM at ambient conditions and after exposure additional complexities, such as spalling in HSC.
to 700 C is presented in Table 9.6 [114]. There In the field of fire science, applied materials
is a decrease in density in all three types of research faces numerous difficulties. At elevated
SFRM at 700 C, which is predominantly due temperatures, many building materials undergo
to the loss of moisture. This decrease in density physicochemical changes. Most of the properties
in SFRM is comparable to that in gypsum, and are temperature dependent and sensitive to test-
attributed to dehydration reactions, which takes ing method parameters such as heating rate,
place with increase in temperature [115]. strain rate, temperature gradient, and so on. One
positive note is that in the last two decades, there
has been significant progress in developing mea-
Other Miscellaneous Materials surement techniques and commercial
instruments for measuring the properties. This
Further information is available from the litera- will likely lead to further research in establishing
ture on the dilatometric and thermogravimetric material properties.
behavior, apparent specific heat, and thermal The review on material properties provided in
conductivity of a number of materials in this chapter is a broad outline of the available
Group I, including asbestos cement board, information. Additional details related to specific
expanded plastic insulating boards, mineral conditions on which these properties are devel-
fiber fireproofing, arborite, and glass-reinforced oped can be found in cited references. Also,
cement board [7]. The properties of plastics and when using the material properties presented in
their behavior in fire are discussed in other this chapter, due consideration should be given to
chapters of this handbook and in Harmathy [2]. the material composition and other
characteristics, such as fire and loading, because
the properties at elevated temperatures depend
Summary on a number of factors.
The use of numerical methods for the calculation Disclaimer Certain commercial products are identified
of the fire resistance of various structural in this paper in order to adequately specify the experimen-
members is gaining acceptance. One of the tal procedure. In no case does such identification imply
recommendations or endorsement by the authors, nor does ε_ ts Rate of secondary creep (m.m–1h–1)
it imply that the product or material identified is the best θ Temperature-compensated time (h)
available for the purpose.
ξ Reaction progress variable, dimensionless
π Material property (any)
ρ Density (kgm–3)
Nomenclature
σ Stress; strength (Pa)
σ Stefan-Boltzmann constant
a Material constant, dimensionless
b (5.67 10–8 Wm–2K–4)
Constant, characteristic of pore geome-
try, dimensionless
c Specific heat (Jkg–1K–1)
Subscripts
c Specific heat for a mixture of reactants
g Glass transient (temperature)
and solid products (Jkg–1K–1)
a Of air
E Modulus of elasticity (Pa)
I Of the ith constituent
h Enthalpy (Jkg–1)
p At constant pressure
Δh Latent heat associated with a “reaction”
s Of the solid matrix
(Jkg–1)
t True
ΔHc Activation energy for creep (Jkmol–1)
t Time-dependent (creep)
k Thermal conductivity (Wm–1K–1)
T At temperature T
Lv Heat of gasification of wood
u Ultimate
‘ Dimension (m)
y Yield
Δ‘ ‘ – ‘0
0 Original value, at reference temperature
m Exponent, dimensionless
M Mass (kg)
n Material constant, dimensionless
P Porosity (m3m–3) References
qn Net heat flux to char front
1. T.Z. Harmathy, Technical Paper No. 242, National
R Gas constant (8315 Jkmol–1K–1) Research Council of Canada, Ottawa (1967).
S Specific surface area (m2.m–3) 2. T.Z. Harmathy, Fire Safety Design and Concrete,
t Time (h) Longman Scientific and Technical, Harlow, UK
T Temperature (K or C) (1993).
3. D.A.G. Bruggeman, Physik. Zeitschr., 37, p. 906
v Volume fraction (m–3.m3) (1936).
w Mass fraction (kgkg–1) 4. R.L. Hamilton and O.K. Crosser, Industrial & Engi-
Z Zener-Hollomon parameter (h–1) neering Chemistry Fundamentals, 7, p. 187 (1962).
5. J.C. Maxwell, A Treatise on Electricity and Magne-
tism, 3rd ed., 1, Clarendon Press, Oxford, UK
Greek Letters (1904).
6. T.Z. Harmathy, Journal of Materials, 5, p. 47 (1970).
7. T.Z. Harmathy, DBR Paper No. 1080, NRCC 20956,
α Thermal diffusivity National Research Council of Canada, Ottawa
β Coefficient of linear thermal expansion (1983).
(mm–1) 8. V.K.R. Kodur and M.M.S. Dwaikat, “Effect of high
γ temperature creep on fire response of restrained steel
Expression defined by Equation 9.3,
beams”, J. of Materials and Structures, 43, 10,
dimensionless pp. 1327–1341 (2010)
β0 Charring rate (mm/min) 9. J.E. Dorn, Journal of the Mechanics and Physics of
δ Characteristic pore size (m) Solids, 3, p. 85 (1954).
ε 10. T.Z. Harmathy, in ASTM STP422, American Society
Emissivity of pores, dimensionless
for Testing and Materials, Philadelphia (1967).
ε Strain (deformation) (mm–1) 11. T.Z. Harmathy, “Trans. Am. Soc. Mech. Eng.,”
εt0 Creep parameter (mm–1) Journal of Basic Engineering, 89, p. 496 (1967).
www.ebook777.com
12. C. Zener and J.H. Hollomon, Journal of Applied 33. U. Diederichs, U.M. Jumppanen, and U. Schneider,
Physics, 15, p. 22 (1944). “High Temperature Properties and Spalling
13. F.H. Wittmann (ed.), Fundamental Research on Behaviour of High Strength Concrete,” in
Creep and Shrinkage of Concrete, Martinus Nijhoff, Proceedings of Fourth Weimar Workshop on High
The Hague, Netherlands (1982). Performance Concrete, HAB, Weimar, Germany
14. Y. Anderberg and S. Thelandersson, Bulletin 54, (1995).
Lund Institute of Technology, Lund, Sweden (1976). 34. Y. Anderberg, “Spalling Phenomenon of HPC and
15. U. Schneider, Fire and Materials, 1, p. 103 (1976). OC,” in International Workshop on Fire Perfor-
16. T.Z. Harmathy, Journal of the American Concrete mance of High Strength Concrete, NIST SP 919,
Institute, 65, 959 (1968). NIST, Gaithersburg, MD (1997).
17. 951 Thermogravimetric Analyzer (TGA), DuPont 35. Z.P. Bazant, “Analysis of Pore Pressure, Thermal
Instruments, Wilmington, DE (1977). Stress and Fracture in Rapidly Heated Concrete,” in
18. T.T. Lie and V.K.R. Kodur, “Thermal and Mechani- International Workshop on Fire Performance of
cal Properties of Steel Fibre-Reinforced Concrete at High Strength Concrete, NIST SP 919, NIST,
Elevated Temperatures,” Canadian Journal of Civil Gaithersburg, MD (1997).
Engineering, 23, p. 4 (1996). 36. A.N. Noumowe, P. Clastres, G. Debicki, and J.-L.
19. ASTM Test Method C135 86, 2007 Annual Book Costaz, “Thermal Stresses and Water Vapor Pressure
of ASTM Standards, 15.01, American Society for of High Performance Concrete at High Tempera-
Testing and Materials, Philadelphia (2007). ture,” Proceedings, Fourth International Symposium
20. T.Z. Harmathy and L.W. Allen, Journal of the Amer- on Utilization of High-Strength/High-Performance
ican Concrete Institute, 70, p. 132 (1973). Concrete, Paris, France (1996).
21. 910 Differential Scanning Calorimeter (DSC), 37. J.A. Purkiss, Fire Safety Engineering Design of
DuPont Instruments, Wilmington, DE (1977). Structures, Butterworth Heinemann, Bodmin,
22. J.H. Perry (ed.), Chemical Engineers’ Handbook, Cornwall, UK (1996).
3rd ed., McGraw-Hill, New York (1950). 38. E.L. Schaffer, “Charring Rate of Selected Woods—
23. W. Eitel, Thermochemical Methods in Silicate Inves- Transverse to Grain,” FPL 69, U.S. Department of
tigation, Rutgers University, New Brunswick, Agriculture, Forest Service, Forest Products Labora-
Canada (1952). tory, Madison, WI (1967).
24. T.Z. Harmathy, Industrial & Engineering Chemistry 39. B.F.W. Rogowski, “Charring of Timber in Fire
Fundamentals, 8, p. 92 (1969). Tests,” in Symposium No. 3 Fire and Structural
25. D.A. DeVries, in Problems Relating to Thermal Use of Timber in Buildings, HMSO, London (1969).
Conductivity, Bulletin de l’Institut International du 40. N. Bénichou and M.A. Sultan, “Fire Resistance of
Froid, Annexe 1952–1, Louvain, Belgique, p. 115 Lightweight Wood Frame Assemblies: State-of-the-
(1952). Art Report,” IR 776, IRC, National Research Coun-
26. W.D. Kingery, Introduction to Ceramics, John cil of Canada, Ottawa (1999).
Wiley and Sons, New York (1960). 41. S. Hadvig, Charring of Wood in Building Fires—
27. T.T. Lie and V.K.R. Kodur, “Thermal Properties of Practice, Theory, Instrumentation, Measurements,
Fibre-Reinforced Concrete at Elevated Laboratory of Heating and Air-Conditioning, Tech-
Temperatures,” IR 683, IRC, National Research nical University of Denmark, Lyngby, Denmark
Council of Canada, Ottawa (1995). (1981).
28. Thermal Conductivity Meter (TC-31), Instruction 42. E. Mikkola, “Charring of Wood,” Report 689, Fire
Manual, Kyoto Electronics Manufacturing Co. Ltd., Technology Laboratory, Technical Research Centre
Tokyo, Japan (1993). of Finland, Espoo (1990).
29. ASCE, “Structural Fire Protection: Manual of Prac- 43. Guide for Determining the Fire Endurance of Con-
tice,” No. 78, American Society of Civil Engineers, crete Elements, ACI-216–89, American Concrete
New York (1993). Institute, Detroit, MI (1989).
30. L.T. Phan, “Fire Performance of High-Strength Con- 44. I.D. Bennetts, Report No. MRL/PS23/81/001, BHP
crete: A Report of the State-of-the-Art,” National Melbourne Research Laboratories, Clayton,
Institute of Standards and Technology, Gaithersburg, Australia (1981).
MD (1996). 45. U. Schneider (ed.), Properties of Materials at High
31. U. Danielsen, “Marine Concrete Structures Exposed Temperatures—Concrete, Kassel University, Kassel,
to Hydrocarbon Fires,” Report, SINTEF—The Nor- Germany (1985).
wegian Fire Research Institute, Trondheim, Norway 46. Y. Anderberg (ed.), Properties of Materials at High
(1997). Temperatures—Steel, Lund University, Lund,
32. V.K.R. Kodur and M.A. Sultan, “Structural Sweden (1983).
Behaviour of High Strength Concrete Columns 47. F. Birch and H. Clark, American Journal of Science,
Exposed to Fire,” Proceedings, International Sym- 238, p. 542 (1940).
posium on High Performance and Reactive Powder 48. T.Z. Harmathy and W.W. Stanzak, in ASTM
Concrete, Concrete Canada, Sherbrooke, Canada STP464, American Society for Testing and
(1998). Materials, Philadelphia (1970).
49. Y. Anderberg, “Mechanical Properties of 66. C.R. Cruz, Journal, PCA Research and Development
Reinforcing Steel at Elevated Temperatures,” Laboratories, 10, p. 36 (1968).
Tekniska Meddelande, 36, Sweden (1978). 67. J.C. MareÂchal, in ACI SP 34, American Concrete
50. “European Recommendations for the Fire Safety of Institute, Detroit, MI (1972).
Steel Structures,” European Convention for Con- 68. H. Gross, Nuclear Engineering and Design,
struction Steelwork, Tech. Comm. 3, Elsevier, 32, p. 129 (1975).
New York (1983). 69. U. Schneider, U. Diedrichs, W. Rosenberger, and
51. Eurocode 3, Design of steel structures, Part 1-2: R. Weiss, Sonderforschungsbereich 148,
General rules-structural fire design, Document Arbeitsbericht 1978–1980, Teil II, B 3, Technical
CEN, European Committee for Standardization, University of Braunschweig, Germany (1980).
UK (2005). 70. U. Diederichs and U. Schneider, “Bond Strength at
52. T. Twilt, “Stress-Strain Relationships of Reinforcing High Temperatures,” Magazine of Concrete
Structural Steel at Elevated Temperatures, Analysis Research, 33, 115, pp. 75–84 (1981).
of Various Options and European Proposal,” TNO- 71. V.K.R. Kodur, “Fibre-Reinforced Concrete for
Rep. BI-91-015, TNO Build. and Constr. Res., Delft, Enhancing the Structural Fire Resistance of
Netherlands (1991). Columns,” ACI-SP (2000).
53. K.W. Poh, “General Stress-Strain Equation,” ASCE 72. A. Bilodeau, V.M. Malhotra, and G.C. Hoff,
Journal of Materials in Civil Engineering, Dec. “Hydrocarbon Fire Resistance of High Strength Nor-
(1997). mal Weight and Light Weight Concrete
54. K.W. Poh, “Stress-Strain-Temperature Relationship Incorporating Polypropylene Fibres,” in
for Structural Steel,” ASCE Journal of Materials in Proceedings, International Symposium on High Per-
Civil Engineering, Oct. (2001). formance and Reactive Powder Concrete,
55. J.T. Gerlich, P.C.R. Collier, and A.H. Buchanan, Sherbrooke, Canada (1998).
“Design of Light Steel-Framed Walls for Fire Resis- 73. V.K.R. Kodur and T.T. Lie, “Fire Resistance of
tance,” Fire and Materials, 20, 2 (1996). Fibre-Reinforced Concrete,” in Fibre Reinforced
56. G.Q. Li, S.C. Jiang, and Y.Z. Yin, “Experimental Concrete: Present and the Future, Canadian Society
studies on the properties of constructional steel at of Civil Engineers, Montreal (1997).
elevated temperatures.” J. Struct. Eng., 129, 12, pp. 74. U.-M. Jumppanen, U. Diederichs, and
1717–1721 (2003). K. Heinrichsmeyer, “Materials Properties of
57. BS 5950, “Structural Use of Steelwork in Building,” F-Concrete at High Temperatures,” VTT Research
Part 8, in Code of Practice for Fire Resistant Design, Report No. 452, Technical Research Centre of
British Standards Institution, London (2003). Finland, Espoo (1986).
58. W. Wang, L. Bing and V.K.R. Kodur, “Effect of 75. J.A. Purkiss, “Steel Fibre-Reinforced Concrete at Ele-
temperature on strength and elastic modulus of vated Temperatures,” International Journal of Cement
high strength steel”, in Press: ASCE Journal of Composites and Light Weight Concrete, 6, 3 (1984).
Materials in Civil Engineering, pp. 1–24 (2012). 76. T.T. Lie and V.K.R. Kodur, “Effect of Temperature
59. V.K.R. Kodur and W. Khaliq, “Effect of temperature on Thermal and Mechanical Properties of Steel
on thermal and mechanical properties of steel bolts”, Fibre-Reinforced Concrete,” IR 695, IRC, National
ASCE Journal of Materials in Civil Engineering, Research Council of Canada, Ottawa (1995).
24, 6, pp. 765–774 (2012). 77. V.K.R. Kodur and R. McGrath, “Effect of Silica
60. J.T. Gerlich, “Design of Loadbearing Light Steel Fume and Confinement on Fire Performance of
Frame Walls for Fire Resistance,” Fire Engineering High Strength Concrete Columns,” Canadian Jour-
Research Report 95/3, University of Canterbury, nal of Civil Engineering, p. 24 (2006).
New Zealand (1995). 78. F.P. Cheng, V.K.R. Kodur, and T.C. Wang, “Stress-
61. P. Makelainen and K. Miller, Mechanical Properties Strain Curves for High Strength Concrete at Elevated
of Cold-Formed Galvanized Sheet Steel Z32 at Ele- Temperatures,” ASCE Journal of Materials Engi-
vated Temperatures, Helsinki University of Technol- neering, 16, 1, pp. 84–90 (2004).
ogy, Finland (1983). 79. V.K.R. Kodur, T.C. Wang, and F.P. Cheng,
62. F. Alfawakhiri, M.A. Sultan, and D.H. MacKinnon, “Predicting the Fire Resistance Behaviour of High
“Fire Resistance of Loadbearing Steel-Stud Walls Strength Concrete Columns,” Cements and Concrete
Protected with Gypsum Board: A Review,” Fire Composites Journal, 26, 2, pp. 141–153 (2003).
Technology, 35, 4 (1999). 80. V.K.R. Kodur and M.A. Sultan, “Thermal Properties
63. T.Z. Harmathy and J.E. Berndt, Journal of the Amer- of High Strength Concrete at Elevated
ican Concrete Institute, 63, p. 93 (1966). Temperatures,” CANMET-ACI-JCI International
64. C.R. Cruz, Journal, PCA Research and Development Conference, ACI SP-170, Tokushima, Japan, Ameri-
Laboratories, 8, p. 37 (1966). can Concrete Institute, Detroit, MI (1998).
65. M.S. Abrams, in ACI SP 25, American Concrete 81. V.K.R. Kodur and M.A. Sultan, “Effect of Tempera-
Institute, Detroit, MI (1971). ture on Thermal Properties of High Strength
www.ebook777.com
Concrete,” ASCE Journal of Materials in Civil Engi- Concrete,” MSc Thesis, Kingston, Canada, Depart-
neering, 15, 8, pp. 101–108 (2003). ment of Civil Engineering, Queen’s University
82. V.K.R. Kodur and W. Khaliq, “Effect of temperature (2006).
on thermal properties of different types of high 98. A. Katz and N. Berman, “Modeling the Effect of
strength concrete”, ASCE Journal of Materials in High Temperature on the Bond of FRP Reinforcing
Civil Engineering, 23, 6, pp. 793–801 (2011). Bars to Concrete,” Cement and Concrete Composites
83. V.K.R. Kodur, “Spalling in High Strength Concrete Journal, 22, pp. 433–443 (2000).
Exposed to Fire—Concerns, Causes, Critical 99. A. Katz, N. Berman, and L.C. Bank, “Effect of High
Parameters and Cures,” in Proceedings: ASCE Temperature on the Bond Strength of FRP Rebars,”
Structures Congress, Philadelphia (2000). Journal of Composites for Construction, 3, 2,
84. V.K.R. Kodur, “Guidelines for Fire Resistance pp. 73–81 (1999).
Design of High Strength Concrete Columns,” Jour- 100. A. Sumida, T. Fujisaki, K. Watanabe, and T. Kato,
nal of Fire Protection Engineering, 15, 2, pp. “Heat Resistance of Continuous Fiber Reinforced
93–106 (2005). Plastic Rods,” Proceedings, Fifth Annual Sympo-
85. J.W. McBurney and C.E. Lovewell, ASTM— sium on Fibre-Reinforced-Plastic Reinforcement
Proceedings of the Thirty-Sixth Annual Meeting, for Concrete Structures, Thomas Telford, London,
Vol. 33 (II), American Society for Testing and pp. 557–565 (2001).
Materials, Detroit, MI, p. 636 (1933). 101. N. Galati, B. Vollintine, A. Nanni, L.R. Dharani, and
86. Wood Handbook: Wood as an Engineering Material, M.A. Aiello, “Thermal Effects on Bond Between
Agriculture Handbook No. 72, Forest Products Lab- FRP Rebars and Concrete,” Proceedings, Advanced
oratory, U.S. Government Printing Office, Polymer Composites for Structural Applications in
Washington, DC (1974). Construction, Woodhead Publishing Ltd.,
87. C.C. Gerhards, Wood & Fiber, 14, p. 4 (1981). Cambridge, UK, pp. 501–508 (2004).
88. E.L. Schaffer, Wood & Fiber, 9, p. 145 (1977). 102. V.R. Kodur, L.A. Bisby, and M.F. Green, “Experi-
89. E.L. Schaffer, Research Paper FPL 450, mental Evaluation of the Fire Behavior of Fibre-
U.S. Department of Agriculture, Forest Products Reinforced-Polymer-Strengthened Reinforced Con-
Lab., Madison, WI (1984). crete Columns,” Fire Safety Journal, 41, 7, pp.
90. “Structural Fire Design,” Part 1.2, in Eurocode 5, 547–557 (2005).
CEN, Brussels, Belgium (1995). 103. V.R. Kodur and L.A. Bisby, “Evaluation of Fire
91. F.F. Wangaard, Section 29, in Engineering Endurance of Concrete Slabs Reinforced with FRP
Materials Handbook (C.L. Mantell, ed.), McGraw- Bars,” ASCE Journal of Structural Engineering,
Hill, New York (1958). 131, 1, pp. 34–43 (2005).
92. V.K.R. Kodur, J. Fike, R. Fike, and M. Tabaddoor, 104. Y.C. Wang, P.M.H. Wong, and V.K.R. Kodur, “An
“Factors governing fire resistance of engineered Experimental Study of Mechanical Properties of
wood I-joists”, Proceedings of the Seventh Interna- FRP and Steel Reinforcing Bars at Elevated
tional Conference on Structures in Fire, Zurich, Temperatures,” Composite Structures, 80, 1, pp.
Switzerland, pp. 417–426 (2012). 131–140 (2007).
93. V.K.R. Kodur and D. Baingo, “Fire Resistance of FRP 105. B. Yu, and V.K.R. Kodur, “Effect of Temperature on
Reinforced Concrete Slabs,” IR 758, IRC, National Strength and Stiffness Properties of Near-Surface
Research Council of Canada, Ottawa (1998). Mounted FRP Reinforcement,” Journal of
94. V.K.R. Kodur, “Fire Resistance Requirements for Composites, Part B: Engineering, 58, pp. 510–517
FRP Structural Members,” Proceedings—Vol I, (2014).
1999 CSCE Annual Conference, Canadian Society 106. T.Z. Harmathy, in ASTM STP301, American Society
of Civil Engineers, Regina, Saskatchewan (1999). for Testing and Materials, Philadelphia (1961).
95. T.S. Gates, “Effects of Elevated Temperature on the 107. R.R. West and W.J. Sutton, Journal of the American
Viscoelastic Modeling of Graphite/Polymeric Ceramic Society, 37, p. 221 (1954).
Composites,” NASA Technical Memorandum 108. P. Ljunggren, Journal of the American Ceramic
104160, NASA, Langley Research Center, Society, 43, p. 227 (1960).
Hampton, VA (1991). 109. M.A. Sultan, “A Model for Predicting Heat Transfer
96. Y.C. Wang and V.K.R. Kodur, “Variation of Through Noninsulated Unloaded Steel-Stud Gypsum
Strength and Stiffness of Fibre Reinforced Polymer Board Wall Assemblies Exposed to Fire,” Fire Tech-
Reinforcing Bars with Temperature,” Cement and nology, 32, 3 (1996).
Concrete Composites, 27, pp. 864–874 (2005). 110. “Gypsum Board: Typical Mechanical and Physical
97. SK. Foster, “High Temperature Residual Perfor- Properties,” GA-235–98, Gypsum Association,
mance of Externally-Bonded FRP Systems for Washington, DC (1998).
111. M.A. Sultan, “Effect of Insulation in the Wall Cavity V.K.R. Kodur is a Professor in the department of Civil and
on the Fire Resistance Rating of Full-Scale Asym- Environmental Engineering and also serves as Director of
metrical (1 2) Gypsum Board Protected Wall the Center on Structural Fire Safety and Diagnostics at the
Assemblies,” in Proceedings of the International Michigan State University (MSU). Dr. Kodur’s research
Conference on Fire Research and Engineering, has focused on the evaluation of fire resistance of structural
Orlando, FL, SFPE, Boston (1995). systems through large scale fire experiments and numerical
112. A.H. Buchanan, Structural Design for Fire Safety, modelling; characterization of materials under high temper-
John Wiley & Sons Ltd., Chichester, UK (2002). ature; and non-linear design and analysis of structural
113. V.K.R. Kodur, M. Dwaikat and R. Fike, “High-tem- systems. He is a Fellow of the Canadian Academy of
perature properties of steel for fire resistance Engineering, a Foreign Fellow of Indian National Academy
modeling of structures,” Journal of Materials in of Engineering and Fellow of ASCE, ACI and SEI. He is an
Civil Engineering, 22, 5, pp. 423–434 (2010). Associate Editor of Journal of Structural Engineering,
114. V.K.R. Kodur and A. Shakya, “Effect of temperature Chairman of ACI Fire Protection Committee, and Chairman
on thermal properties of fire insulation”, Fire Safety of ASCE-29 (Fire) Standards Committee.
Journal, 61, pp. 314–323 (2013).
115. S. Park, S.L. Manzello, D.P. Bentz, and T.Z. Harmathy was head of the Fire Research Section,
T. Mizukami, “Determining thermal properties of Institute of Research in Construction, National Research
gypsum board at elevated temperatures”, Fire and Council of Canada, until his retirement in 1988. His
Materials (2009). research centered on materials science and the spread
potential of compartment fires.
www.ebook777.com
Chemical Kinetics and Fire

10
Gregory T. Linteris and John F. Griffiths
Introduction that of pyrolyzing materials. The specialized

fields of propellants, explosives, and material
The purpose of this chapter is to set out the synthesis also require solid-phase chemical
principles of chemical kinetics as they apply to kinetic descriptions, but these are beyond the
combustion in flames and fires. Chemical equilib- scope of the present chapter. Nonetheless, many
rium, which was discussed in a previous chapter, of the fundamental principles of chemical kinet-
deals with the final preferred state of a given set of ics discussed here are relevant regardless of the
reactants after an infinite time has passed. In con- phase of the reacting system.
trast, chemical kinetics deals with the rate at which Gas-phase chemical kinetics is of interest in
the system proceeds to the equilibrium state, i.e., fires for many reasons. The heat release in a fire
the specific participating chemical reactions and typically occurs in the gas phase, and is responsi-
their rates. Chemical equilibrium and chemical ble for the gas-phase temperature field, and hence
kinetics are related in that the thermodynamic, the heat flux to the burning materials (a feedback
equilibrium state provides the driving force for loop which controls the fuel supply rate in the fire,
chemical reaction. The material in this chapter is and hence the geometric growth in fire size with
covered briefly; more detailed descriptions can be time). Some fundamental fire phenomena, such as
found in chemistry [1] and combustion [2–4] text ignition and extinction, are clearly controlled by
books, upon which much of the material is based. the gas-phase chemical kinetics. Fire suppression
The foundations of chemical kinetics have is controlled by the rates of chemical reaction,
validity in gas, liquid or solid phases, but for both for the relatively inert agents (e.g., CO2,
fires, the gas phase has the greatest relevance water) which reduce the temperature (and hence
because the main heat release normally occurs overall reaction rate) to the point of extinction,
during flaming combustion. The role of solid- and for chemically acting agents (e.g., CF3Br and
and liquid-phase chemical kinetics in fires is hydrofluorocarbons) which interfere with the nor-
discussed in Chap. 7. Similarly, smoldering com- mal chemistry of the fuels with air. Similarly, the
bustion is a surface combustion process and the action of the most commonly used fire retardants
chemical kinetic description is closely related to in polymers is controlled by their gas-phase
chemical behavior. In general, chemical reaction
rates must be fast enough to match the local
G.T. Linteris (*)
residence time for transport (either convective
National Institute of Standards and Technology,
100 Bureau Dr. Stop 8665; Gaithersburg, MD, 20899 or diffusive); if not, the flame will extinguish.
The formation of soot, the major radiating
J.F. Griffiths
University of Leeds, School of Chemistry, Leeds, species from fire plumes, is controlled by
LS2 9JT UK gas-phase chemical kinetics, as is the formation

326 G.T. Linteris and J.F. Griffiths
of CO, which is the major toxic compound CH4 þ 2O2 ! CO2 þ H2 O ð10:1Þ
responsible for fire deaths. In fires, the formation
of other toxic compounds, for example of HCN, which is an example of a global (or overall) reac-
as well as environmental pollutants (polycyclic tion. While reaction (10.1) shows the reactants and
aromatic hydrocarbons, dioxins, etc.) is con- products, it does not represent the detailed chemi-
trolled by the chemical kinetics of reactions cal interactions which actually occur. Rather,
occurring in the gas-phase. Clearly, understand- the conversion of CH4 to CO2 and H2O is a
ing chemical kinetics is central to controlling multi-step process involving many species and
unwanted fires and their deleterious effects. It is many reactions. A more complete representation
of great value if the Fire Scientist can answer the might include some intermediate species along the
question: “Is the process at hand controlled by path:
the rate of chemical reactions or by some other
CH4 ! CH3 ! CH2 O ! CHO
physical process?” The goal of the present chap- ! CO ! CO2 ; ð10:2Þ
ter is to provide some fundament materials for
approaching such a question. but again, the details are missing.
The reaction of a fuel (for example methane) Figure 10.1 shows reaction pathways for a
with air to products can be represented by an premixed methane-air flame (initial pressure P0
expression such as: and temperature T0 of 1 bar and 298 K). The
CH4 C2H6
•H •H,O,OH •CH3 •H,O,OH
•H •O •H,O,OH •M
CH3 C2H5 CH3CHO CH3CO CH3
•CH3
•O •CH3 •H •M,O2
•O,OH
CH2O C2H4 CH3,CH2O,CHO
•H,O,OH •H
CHO C2H3
•M,O2,H •H •M,H,O2 •O
•OH •H
CO C2H2 CH2CO CH3
•O •OH
•H
CH CH2 CH2O,CHO
•O,O2 •O,O2
CO CO,CO2
Fig. 10.1 Reaction pathway analysis for premixed stoichiometric methane-air flame (From [5])
www.ebook777.com
10 Chemical Kinetics and Fire 327
arrows connect the initial, intermediate, and the decomposition of methane proceeds mostly
product species for the major reaction steps through the sequence shown in Equation 10.2,
involved in the consumption of methane. For with the molecule oxidized to smaller and
each species in the figure, the major reactants smaller fragments. In contrast, Fig. 10.2 shows
and products are at the ends of the arrows, that under rich conditions, reaction of CH3 pro-
while participating reaction partners are listed ceeds largely through its reaction with other CH3
next to the arrow. The thickness of the arrows molecules, in pyrolysis reactions which tend to
indicates the fraction of the total reaction flux form larger molecules, and finally, acetylene
which proceeds through that particular reaction C2H2, which is a precursor for soot.
path (normalized, in this case, by the total reac- Pictures such as those in Figs. 10.1 and 10.2
tion rate of CH4.) The purpose of such a figure is are very useful tools for understanding the role of
to provide not only detailed knowledge of the chemistry in the physical behavior of combustion
important steps in the consumption of a reactant, systems. Indeed, the widespread availability of
but also to provide a heuristic understanding of numerical codes for performing the simulations,
the general features of the chemical system. For chemical databases for the mechanisms, and fast
example, Fig. 10.2 shows the chemical reaction computers have made such simulations integral
pathways for methane in the same configuration design tools in many chemical and engineering
(a premixed laminar flame), but at a different fields. These include the design of propulsion
fuel-air ratio, in this case, fuel rich. In Fig. 10.1, devices (gas turbines, diesel and spark ignition
CH6 C2H6
O
•H •H,CH •CH3 •H,O,OH
•CH3
•O •H,O,OH •M
CH3 C2H5 CH3CHO CH3CO CH3
•H
•O •CH3 •H •M,O2
•O,OH
CH2O C2H6 CH3,CHO
•H,O,OH •H2OH
CHO C2H3
•M,O2,K •H •M,H,O2 •O
•OH •H
CO C2H2 CH2CO CH3
•OH
CH2O,CHO
Fig. 10.2 Reaction pathway analysis for premixed fuel-rich methane-air flame (From [5])
engines, etc.) and new power plants and the C-H bond. In this case, the fragments, ∙H and
incinerators, particularly in regard to understand- ∙CH3, are radicals. The unpaired electron is typi-
ing the efficiency and pollutant formation. In fire cally shown by “∙”. Radicals tend to be highly
safety, detailed chemical kinetic descriptions are reactive, and are responsible for promoting the
primarily used in research. As computer models, chemistry occurring in combustion systems. The
kinetic mechanisms, and computer speed all unpaired electron in a radical attacks bonds in
improve, however, the contributions of chemical stable molecules, leading to their decomposition.
kinetics to the understanding of fire safety will The energy barrier in radical reactions tends to be
increase, as it has in other fields. very low, which is why their reactions are so fast.
A reaction mechanism (such as that used to In high-temperature gaseous combustion
produce the reaction pathway analyses in systems, the equilibrium concentration of radicals
Figs. 10.1 and 10.2) starts with a list of species increases with temperature, and peak radical
believed to be participating in chemical reactions concentrations can be quite high, with a volume
for the physical system and conditions of interest. fraction on the order of 1 % in the primary reac-
Thermodynamic properties of the species are tion zone of premixed flames. While this number
required, as are the rates of the reaction of each may appear low, recall that radicals are highly
species with all others in the list, and the temper- reactive: their concentrations do not build up
ature and pressure dependence on the rates of higher because they are consumed so fast.
reaction. The development of such mechanism, Typical radicals in combustion are: ∙H, ∙O∙,
and validation of the mechanisms, is a time- ∙OH, ∙CH, ∙C2, ∙CH2, ∙CHO, ∙CH3, ∙R, etc. Here,
consuming and arduous task. Fortunately, there ∙R denotes any hydrocarbon molecule with an
are many combustion and chemical kinetics unpaired electron at one site. For example, ∙R
researchers worldwide working in this area can be ∙CH3, ∙C2H5, ∙C3H7, for the methyl,
[6–9]. The databases are constantly in develop- ethyl, and propyl radicals, formed by the abstrac-
ment, and versions are freely available, as tion (removal) of H atom from CH4, C2H6, and
discussed below. The remainder of this chapter C3H8. In these cases, the stable molecules can
describes the fundamental concepts used in also be represented generically as RH; for exam-
developing chemical mechanisms, and some ple: C2H6 + OH ! C2H5∙ + H2O, which can be
examples of their application. written as RH + OH ! R∙ + H2O .
In the context of combustion kinetics, explo-
sive behavior corresponds to extremely rapid
Fundamentals reaction. There are two types of explosive behav-
ior: thermal explosions, and chain branched
Radical Reactions explosions. The former is due to temperature
rise, while the latter is due to an exponential
In a combustion system, the consumption of a build-up in radical concentrations. In combus-
fuel molecule (and its decomposition products) is tion, the radical pool refers to the chain-carrying
driven largely by attack from radicals. A radical radicals which are involved in the branching
(or free radical) is an atom, molecule, or ion with reactions, for example H, OH, O, and HO2.
one or more unpaired electron or an open shell It often takes time for this pool to develop,
configuration. A radical can be formed by break- which leads to an induction time for ignition
ing the bond of a stable molecule, for example (also called the ignition delay). The buildup,
due to a high-energy collision: CH4 + M ¼> maintenance, and decay of the radical pool in
∙CH3 + ∙H + M. In this reaction, M (discussed combustion is determined by the relative rates
further below) represents any other molecule in of the production and destruction of radicals,
the system which can act as a collision partner and relies upon certain key branching steps in
with CH4, and thereby supply the energy to break the reactions scheme, as described below.
www.ebook777.com
The typical steps of radical chain reactions, too much energy stay together (since it is a rela-
initiation, propagation, branching, and termina- tively small molecule and cannot absorb the
tion, are described here in the context of ignition energy in vibrational or rotational modes of
of H2/O2 systems (following Ref. [2]). The initial energy storage). Another example of such a
radicals required to start the process come from three-body termination reaction is: H + H + M
the breaking of a bond of a stable molecule ! H2 + M, which is important in flames. Reac-
(either thermally through collisional energy tion 10.8 represents the destruction of a radical
transfer, or through photolytic interactions, for without another interacting gas molecule (for
example in the presence of UV light). Since the example through radical quenching at a wall),
bond strengths of most stable molecules relevant which can be very important in many situations
to combustion systems tend to be high, the pro- where solid surfaces are available to the gas-phase
cess is slow (photolytic bond breaking is typi- reactants. Although it looks as if reaction 10.5 is a
cally unimportant in combustion). propagation step (because there is no increase in
Initiation steps, for example reaction 10.3 in the number of unpaired electrons) this is classified
Table 10.1, are those in which two reactive as a chain branching reaction since the number
radicals are formed from stable species. Propa- of active reaction chains has been multiplied.
gation steps (reaction 10.4) involve radicals, There is another class of overall termination
changing the type, but not the total number of steps: the gas-phase catalytic cycles involving
radicals. These reactions tend to have very low flame inhibitors such as FeO, HBr, and HOPO.
activation energies, and hence are very fast. These catalytic cycles serve to reduce radical
Propagation steps are mostly responsible for the concentrations in flames, and are discussed in
consumption of the fuel and its decomposition more detail below.
products in combustion system. Chain branching
steps (e.g., reaction 10.6) increase the number of
radicals, and hence are responsible for the explo- Law of Mass Action
sive growth in the radical concentration, which
leads to rapid reaction in the system as a whole. The rate of disappearance of a reactant is generally
Termination steps reduce the number of radicals, proportional to the concentrations of participating
and thereby shut down the overall combustion. reactants. For an arbitrary chemical reaction:
Reaction 10.7, while it technically is a propaga-
tion step, is usually thought of as a termination A þ B ! products; ð10:9Þ
step because at low temperatures the radical
the proportionality is represented by the Law of
HO2• is relatively unreactive, and its fate is
Mass Action:
often to be destroyed at walls (as in the
reconciling of the explosion limits of the H2–O2 d ½ A
system [2–4]). In reaction 10.7, a non-reacting ¼ ω_ A ¼ k ½A1 ½B1 ð10:10Þ
dt
third body (M) takes away energy from the
radical-radical combination. In this example, Here, the brackets denote concentration (for
after the reaction of H• + O2, HO2• would have example with units of mol cm3), d[A]/dt is the
rate of change in the concentration of A with time
Table 10.1 Radical reactions important in H2/O2 (t), and k is a proportionality constant. This
ignition expression was phenomenologically developed,
Initiation H2 þ O2 ! 2OH (10.3) based on empirical results, but there is a theoret-
Propagation OH þH2 ! H2 O þ H (10.4) ical basis for it. Molecules of A and B must
Branching O þH2 ! OH þH (10.5) collide to react. Their collision rate depends
H þO2 ! OH þO (10.6) upon their concentrations (which depend upon
Termination H þO2 þ M ! HO2 þ M (10.7) the total concentration via the ideal gas law and
H ! ½H2 (10.8) their volume fraction). Today, k is known as the
specific reaction rate constant (or often just the obtained experimentally, and are valid only for
rate constant). The magnitude of k is usually a that experiment and the range of conditions from
function only of temperature, and that depen- which it was developed. The expressions are
dence is often significant. usually complex; the empirical reactions orders
Note that the concentration C of species i, Ci, are typically not integers, can be negative, and
or [i] can be expressed as Ci ¼ Xi ∙ CT, where the depend upon time and the reaction conditions.
total concentration is given by the ideal gas law Extrapolation to other experimental conditions
can be unreliable or incorrect.
NT P
CT ¼ ¼ ð10:11Þ Note that reaction 10.13 is not believed to
V RT actually occur; detailed experiments have
in which NT is the total number of moles in the shown that on the molecular level, two molecules
volume, V, at the given pressure P and tempera- of hydrogen do not collide with one molecule of
ture T, and R is the universal gas constant oxygen to form two molecules of water. In con-
(8.314 J mol1 K1) The mole fraction Xi trast, elementary reactions, such as
(known also as the volume fraction), is
OH þ H2 ! H2 O þ H ð10:15Þ
Ni Ni
Xi ¼ ¼ X ð10:12Þ are believed to represent an actual interaction
NT Ni
i between molecules: an OH molecule collides
with an H2 molecule, and (if there is sufficient
in which Ni is the number of moles of species i, energy involved in the collision) they react to
and the summation is over all species in the form one H2O molecule and an H atom. The
system. Law of Mass Action for this elementary reaction
would be:
Global vs. Elementary Rates d ½H 2

¼ k15 ½OH 1 ½H 2 1 ð10:16Þ
dt
The Law of Mass Action can be written for either
global or elementary reactions. An example of a and the specific reaction rate constant k15
global reaction (also called a net reaction or describes the probability of reaction. The depen-
overall reaction) is dence (i.e. reaction order) on the concentrations
of H and OH (1 in this example) are the
2H2 þ O2 ! 2H2 O ð10:13Þ molecularity with respect to that reactant, and
the overall molecularity is the sum of individual
in which 2 moles of hydrogen reacts with 1 mole
molecularities of the reactants. Generally, for
of oxygen to form 2 moles of water; the Law of
elementary reactions, the determination of reac-
Mass Action would be:
tion orders is easy: they correspond to the
d ½O 2 1
d ½H 2 d ½H 2 O molecularities.
¼ 2 ¼ An overall reaction such as reaction 10.13 is
dt dt dt ð10:14Þ
¼ kG ½H2 n ½O2 m the result of a large number of elementary
reactions. For example the reaction of hydrogen
where kG is the global rate coefficient, n is the and oxygen, in a simple but satisfactory form,
reaction order with respect to H2, m is the order can be described through a sequence of reactions
with respect to O2, and m + n is the overall order. involving about 8 species and 40 reactions; that
Note that a distinction is made on the left hand for methane 30 species and 400 reactions, and
side of the equation between reactant removal hexane, 450 species 1500 reactions.
and product formation. Reaction 10.13 with Comprehensive representations of mechanisms
10.14 describe what happens to the reactants for the combustion of hydrocarbon fuels run into
globally. Relations such as these are typically many hundreds of species involved in many
www.ebook777.com
thousands of reactions, the complexity increasing d ½A

with their size or the number of compounds in the ¼ kuni ½A1 ð10:21Þ
dt
fuel mixture [6]. Describing a chemical system in
terms of elementary reactions is a difficult and Some bimolecular reactions behave as though
time consuming task, but has many advantages. they were unimolecular. For example, in the
The reaction order of elementary reactions is con- bimolecular reaction 10.17 above, A + B !
stant (does not change with system or conditions, products, if the concentration of B is present in
as the global orders might), and the specific reac- large excess as compared to A, its concentration
tion rate constant varies only with temperature; will not change appreciably, so the rate expres-
hence, the rate expressions should be valid over a sion Equation 10.19 would be:
wider range of conditions than those of one-step d ½ A
global reactions. ¼ keff ½A1 ð10:22Þ
dt
in which the effective rate constant is determined
Type of Reactions by keff ¼ k[B]1. Hence, this second-order reac-
tion is pseudo-first order in A.
There are three types of elementary reactions A special example of a pseudo first-order
actually observed in gaseous combustion reaction is the decomposition of the butyl radical
systems: unimolecular, bimolecular, and ∙C4H9 (formed after abstraction of a hydrogen
termolecular (depending upon the number of atom from C4H10) after a collision with another
species involved). molecule M:
C4 H9 þ M ! C3 H6 þ CH3 þ M ð10:23Þ

Bimolecular Reactions
Bimolecular reactions are those given by The collision partner M, represents any other
molecule in the system. It is called a chaperone
AþB!C ð10:17Þ molecule or a third body (as in reaction 10.7).
or Conceptually, its role is to provide the energy
needed (via a collision) to break the necessary
AþB!CþD ð10:18Þ bonds in reactions 10.23, but it does not other-
wise participate in the reaction. In reaction 10.7,
with
it serves the opposite role: to carry away the
d ½ A excess energy resulting from the joining of two
¼ ω_ A ¼ k ½A1 ½B1 ð10:19Þ free radicals.
dt
The role of M can be recognized in the follow-
These are the most common types of reactions; the ing way. Reaction 10.23 constitutes the breaking
molecularity is one for each reactant and two over- of a C–C bond. In order for this to happen it is
all (a so-called second-order reaction). Common necessary for sufficient energy to be accumulated
examples would be H + H2O ! OH + H2, CO + at the appropriate part of the C4H9∙ radical. The
OH ! CO2 + H, or C4H10 + ∙OH ! ∙C4H9 + energy required is approximately equal to the
H2O. This last reaction is called an abstraction bond dissociation energy (see Chap. 5). C4H9∙ is
reaction because the ∙OH radical abstracts a able to gain the necessary thermal energy via
hydrogen atom from the butane molecule. kinetic energy transfer from another species dur-
ing a collision. So M can be any molecule in the
Unimolecular Reactions system and the concentration [M] represents the
Unimolecular reactions are given by total concentration of species in the system.
A ! B or A ! B þ C ð10:20Þ As a result of this special function of “M”
the order of reaction 10.23 can vary, with keff
with showing a complex experimental dependence
on concentration. The reaction is second order and H + H + M ! H2 + M. (Note that for the
overall at low concentrations of species in the latter of these reactions, a factor of two would
system (signifying low pressures), which arises have to be added to Equation 10.27 for the change
from a first order dependence with respect to both in [H] with time, since both A and B (i.e. H) are
the reactant, C4H9∙ and M, e.g. the same. The rates of these third-order reactions
are pressure dependent (each concentration in
d ½C4 H 9
¼ k0 ½C4 H 9 ½M ð10:24Þ Equation 10.27 is proportional to P via Equa-
dt tion 10.11). Also, via Le Châtelier’s principle,
where the subscript o signifies the rate constant the equilibrium is affect by pressure since there
at the low pressure limit, and represents the sec- is a change in the number of moles in reac-
ond order case. However, as the concentration tion 10.26. Finally, the efficiency of different
(or pressure) of the system is raised the reaction species as third bodies can vary substantially.
changes to eventually achieve a first order depen- For example, in the reaction OH + H + M !
dence on the principal reactant and with no H2O + M, the efficiency of N2, CO, H2, CO2,
dependence on M, e.g. and H2O, as a third body is enhanced by a factor
or 1, 1.8, 2, 3.6, and 6.3 relative to N2. These
d ½C 4 H 9 different third-body efficiencies are usually
¼ k1 ½C4 H 9 ; ð10:25Þ
dt accounted for in detailed reaction mechanisms.
where the subscript 1 signifies the rate constant at
the high (infinite) pressure limit, and is strictly a Units of Reaction Rate Constant
first order rate constant. There is a complex depen- Since the derivative of the concentration with
dence of the rate constant on concentration between time d[A]/dt always has the units (concentra-
the two limits, and the overall reaction order tion/time), but since the number of terms on the
varies between 2 and 1. There are sophisticated right hand side of a rate equation varies with the
theories applied in combustion chemistry to inter- reaction order n (c.f. Equations 10.19, 10.21,
pret these data, but the two limiting conditions can 10.27), the units of the specific reaction rate
be derived on the basis of simple algebraic analysis constant k must change accordingly. The units
as is found in chemical kinetic texts (e.g. [1]). of k are: (concentration)(n1)∙s1 (or, for exam-
ple, (mol cm3)(n1) s1) which yields the units
Termolecular Reactions shown in Table 10.2 for reaction orders of 0 to 3.
Termolecular reactions are described by
AþBþM!CþM ð10:26Þ Arrhenius Rate Expression
with Specific reaction rate constants can be a strong
d ½ A function of temperature (and show no pressure
¼ k ter ½A ½B1 ½M1 ð10:27Þ dependence, except in the situations described
dt
above). In the late 1800s, Svante Arrhenius dis-
Examples of such reactions are the radical recom- covered empirically that the rate of chemical
bination reactions OH + H + M ! H2O + M reactions is described well by an exponential
Table 10.2 Units of the rate constant for various reaction orders
Units of k
Reaction order Rate expression, single component, A (in cgs)
0 d½A=dt ¼ ko conc time1 mol cm3 s1
1 d½A=dt ¼ k1 ½A time1 s1
2 d½A=dt ¼ k2 ½A2 conc1 time1 cm3 mol1 s1
3 d½A=dt ¼ k3 ½A2 ½M conc2 time1 cm6 mol2 s1
www.ebook777.com
function of temperature. The equation which Ea,r. For this exothermic reaction, the products
bears his name is have less chemical energy than the reactants (and
for an adiabatic system, this difference is typi-
kðT Þ ¼ A0 eðEa =RT Þ ð10:28Þ cally manifested as an increase in temperature of
the products). The reverse reaction has a higher
in which A0 is the pre-exponential factor and Ea is
activation energy by an amount corresponding to
the activation energy. Later, the pre-exponential
the exothermicity of the forward reaction. Note
factor A0 was found, for many reactions, to have
that potential energy curves are shown for
some temperature dependence, and today, a
un-catalyzed (solid line) and catalyzed (dashed
modified version of the Arrhenius equation is
line) reactions. A catalyst does not change the
more commonly used:
energy states of the products or reactants, but
kðT Þ ¼ AT b eðEa =RT Þ ð10:29Þ lowers the effective activation energy (for both
forward and reverse reaction), so the system can
In combustion, Ea is typically given in units of more rapidly achieve the equilibrium state.
kcal mol1 or kJ mol1; alternatively, Ea is The collision term ATb in the modified
divided by R to give an activation temperature Arrhenius expression represents the frequency
Ta (¼ Ea/R), with units of K. of collisions, times the probability of colli-
The kinetic theory of gases has provided a sion—the so-called steric factor, with a typical
theoretical basis for the modified Arrhenius upper limit of 1013–1014 cm3 mol1 s1. (Note
equation. For a bimolecular reaction (10.18, that the units of ATb are the same as those of k).
10.19), the reaction rate is proportional to the Nonetheless, steric factors are often quite low,
concentrations of reactants present. The rate con- representing the need for the molecules to have
stant k represents the probability of reaction, the correct orientation, and have the energy in the
which in turn depends upon the rate of molecular molecule (vibrational, rotational, translational)
collisions (embodied in the pre-exponential term to be distributed optimally for reaction to occur.
ATb). Not all collisions will have sufficient For unimolecular reactions, A represents the
energy for reaction, however, and the exponen- vibrational energy in the molecule which leads
tial term describes the fraction of molecules in to its decomposition. Termolecular reactions are
the gas with sufficient energy to overcome a actually two bimolecular reactions in rapid
barrier to reaction. Figure 10.3 shows the chemi- sequence; grouping them together can lead to
cal potential energy diagram for a hypothetical unusual values for the A and Ea.
reaction. For the forward reaction, the activation
energy is Ea,f, while for the reverse reaction, it is Effect of Temperature on Reaction Rates
The parameters for the reaction rate expression
are mostly determined through, or validated
Potential Energy
Un-catalyzed in, experimental measurements in devices such

as shock-tubes, static reactors, flow reactors,
Ea,f Ea,r premixed and diffusion flames, rapid combustion
Reactants machines, and combustion bombs. In determin-
ing the Arrhenius parameters experimentally, one
approach, adopted very early, is to measure the
reaction progress for a given set of initial reactant
Catalyzed concentrations. By performing experiments over
Products a range of temperatures, it is possible to construct
the plot of as ln(k) vs. 1/T as in Fig. 10.4. The
Reaction Progress
activation energy Ea is obtained from the negative
slope of the line, while the pre-exponential (as ln
Fig. 10.3 Energy diagram for a chemical reaction (A0 )) is obtained from the intercept. The results
Fig. 10.4 Arrhenius plot 28 2000 1000 600 T / K

from the experimentally
determined rate constant
for the reaction H + O2
26
ln (k / mol−1 cm3 s−1)

24
22
20
0.5 1.0 1.5
103 K/T
Fig. 10.5 Dependence of log (k / cm3mol-1s-1)

reaction rate constant k on
14
temperature for various
reactions important in
H + HBr = H2 + Br
combustion 13
C3H8 + OH = iC3H7 + H2O

12
CH4+ OH = CH3 + H2O
CO + OH = CO2+H
11
10
9
H + O2 = OH + O
7
H + H2O = H2 + OH
1000 K/T
6
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
presented in the figure were obtained over a Figure 10.5 shows the rate constant for vari-
very wide temperature range by many different ous important reactions in combustion with data
research groups using a number of different from Ref. [10]. As indicated, the reactions CO +
experimental techniques. OH ! CO2 + H and H + HBr ¼ H2 + Br have
www.ebook777.com
very low activation energies (E/R ¼120 K and

290 K respectively), and show very little temper- Important Concepts in Hydrocarbon
ature sensitivity, typical of a radical propagation Combustion Kinetics
reactions. The chain-branching reaction H + O2
¼ OH + O has a somewhat higher temperature Applications of combustion kinetics to fire will
sensitivity (Ea/R ¼ 9,860 K) as does the propa- be illustrated in examples below.
gation reaction H + H2O ! H2 + OH (for
which the activation energy, Ea ¼ 82 kJ
mol1), both of which are endothermic. The Ignition
lines in Fig. 10.5 that exhibit curvature require
the rate constant to be interpreted over the full Ignition is defined as the initiation of combustion.
temperature range using the three parameter Generally, this involves bringing the gas-phase
representation (Equation 10.29). Otherwise, a reactants to the point of rapid, exothermic reaction
different activation energy would have to be (i.e., explosive behavior). As described above, this
applied within more restricted sections of the high rate of reaction can be induced via thermal or
temperature range, as is reflected in the varying radical chain branching mechanisms. And these, in
gradient. turn, can be induced by a spark, a local hot spot
(due to a hot wire or surface), or a pilot flame.
Effect of Pressure on Reaction Rates Alternatively, spontaneous (or auto-) ignition
Pressure manifests itself primarily through occurs when a fuel and air mixture, for example
its effect on concentration via the Law of Mass in a uniformly heated chamber or next to a hot
Action (Equation 10.10) with the ideal gas law surface, by itself reaches the explosive reaction
(Equation. 10.11), and in three-body reactions conditions. Gas-phase reactions lead to increasing
(Equations. 10.26 and 10.27) which are very radical concentrations (either due to thermal or
pressure dependent. In the context of fire, these chain branching mechanisms), which eventually
considerations are most important with regard are high enough for sustained gas-phase flame
to the laboratory experiments used to: measure propagation. A special category of spontaneous
elementary rates, understand a phenomenon, or ignition takes place when a uniformly mixed gas
validate a chemical mechanism. For example, mixture, at constant initial temperature, is raised
many reaction rates for three-body reactions are instantaneously to a uniformly high temperature
in the fall-off regime at ambient pressure. Hence, after passage of a shock wave, as in a shock tube
in specifying the relevant rate, pressure plays experiment. This configuration, while not experi-
a role. enced often in fire research, is of significance since
Nonetheless, in actual fires, while small pres- most of the experimental and calculated data on
sure differences have a major effect on the flow ignition delay are obtained using this technique.
of gases, the magnitude of pressure changes typ- In fire research however, the term “ignition” is
ically has little effect on combustion kinetics. often used in the context of the initiation of
Some effect of pressure may be relevant at high flaming combustion over a solid or liquid mate-
altitude, or in aircraft fire safety considerations. rial. This process, while also dependent upon the
Of course, pressure rise in explosions can influ- behavior of the gas-phase reactants, largely
ence the kinetics. For typical atmospheric involves the thermodynamic behavior of the
pressure fires, however, the limited changes in solid material as it heats and decomposes (typi-
pressure have little effect on the combustion cally when exposed to an external infrared heat
kinetics. flux, for example from the hot upper layer in a
room fire). Of course, gas-phase reaction and 1000

solid material ignition are related in that the φ = 1.0, P = 1 bar
solid material ignites when: (1) it is producing

gas-phase reactants fast enough, (2) the fuel 100
molecules mix with the oxidizer gas in
proportions which are within the flammability
τign / ms
limits, and (3) the radical concentrations build 10
up sufficiently, due to heat release, chain-
branching, etc. Nonetheless, this last part of the
process (ignition, by an externally imposed 1
source, of the gas-phase reactants which are
within the flammability limits) is distinctly dif-
ferent from the spontaneous ignition of gas-phase
0.1
reactants described in the previous paragraph. 900 950 1000 1050
For the purposes of the discussion below, when Temperature K
discussing “ignition”, it is gas-phase spontaneous
Fig. 10.6 Ignition delay of a stoichiometric H2/air mix-
ignition (not material ignition) with which we are ture as a function of initial gas temperature
concerned.
Spontaneous ignition is a gas-phase chemical
process well described by the detailed kinetic temperature and pressure, and the presence of
models described above (and below)—provided trace species. Figure 10.7 shows the H-atom
that the reaction mechanism embraces processes volume fraction and temperature as a function
that are relevant to an appropriately wide range of time for an H2.air system at an initial temper-
of temperatures. For example, if a reactive gas ature of 1000 K. As indicated, the ignition delay
mixture is exposed to a hot surface (or container), is about 1.6 104 s. The build-up in H radical
and (for the purposes of discussion) uniformly concentration is exponential in time, and
heated, it will reach an explosive condition (i.e., increases drastically for times greater than
rapid reaction) after some length of exposure about 10-5 s. Hence, the explosive behavior is
time to the heated condition. The parameter of predominantly due to a chain-branching mecha-
interest is the ignition time or ignition delay τign, nism rather than thermal initiation. Some fuels
which is the time it takes for a mixture at a given decompose forming H-atoms more readily, and
temperature, to reach the state of rapid reaction. these promote a short ignition delay. Additives
To define the ignition delay, a criterion for the which consume H-atoms retard ignition, while
condition of rapid reaction is required. Often, a those that create H-atoms accelerate ignition.
characteristic temperature rise, the maximum
rate of temperature (or pressure) rise, or some
minimum radical concentration is used (e.g., XOH Competition Between Reactions
> ¼ 104) is used. Figure 10.6 shows the calcu-
lated ignition delay of a stoichiometric H2/air The macroscopic behavior in combustion
mixture. As indicated, the ignition delay is a systems is often determined by which of two
strong function of the initial temperature: it (or more) possible reaction paths dominate.
drops by about three orders of magnitude as the Reactions proceed in parallel, and which reaction
initial temperature rises from 925 to 1025 K. dominates depends upon the temperature and the
Given a comprehensive kinetic mechanism, concentrations of reactants (which can change
ignition delay is readily calculated. The chemis- with time and location in the system), and
try of homogeneous ignition is often different the pressure. Some examples of combustion
from that in a flame. Ignition delays are a strong reactions in which the competition between alter-
function of the reactants, stoichiometry, initial nate reactions are given below.
www.ebook777.com
Fig. 10.7 H-atom and temperature increase with time for stoichiometric H2/air mixture at 1000 K
Initiation Reactions: Thermal and the ratio of the rate constants for these two
Decomposition vs. Oxidation competing reactions. As indicated, reaction
The homogeneous ignition above requires the (10.31) is 4 and 9 orders of magnitude higher at
buildup in radical concentrations. To get this 2000 K and 1000 K, respectively. Hence, for any
process started, there must be radical initiation reasonable concentration of O2, the rate of the
reactions. Two possible initiation routes are the molecular reaction (10.31) (i.e., the rate constant
dissociation reaction of H2, reaction (10.30) and times the relevant concentrations as in Equa-
the molecular reaction between H2 and O2. reaction 10.19) will be much higher, making it the
tion (10.31). Figure 10.8 shows the rate constants favored initiation route.
H2 þ M ! H þH þM k ¼ 6:7 1014 eð51000=TÞ s1 ð10:30Þ
H2 þ O2 ! HO2 þH k ¼ 1:5 1014 eð28, 500=TÞ s1 ð10:31Þ
Relative Rates of Oxidation and where R. represents an alkyl radical generated

Degradation of the Primary Fuel Radical from an alkane (e.g. an ethyl radical (C2H5)
As described above, the propagation reactions from ethane (C2H6)). Whether or not the alkyl
are the principal means of reactant consumption radical then decomposes or oxidizes depends
through, for example, upon the temperature and the concentration of
oxygen. Consider normal undecane (n-C11H24),
a component of kerosene. The undecyl radical
RH þ OH ! R þH2 O ð10:32Þ
formed from it in a reaction such as (10.32) may
undergo the competitive reactions
T/K
2000 1500 1000 500
10 20
8 kH2 + O2 18
kH2 + O2/kH2 + M
6 16
4 kH2 + M 14
log (k / cm3mol-1s-1)
log (kH2 + O2 / kH2 + M)

2 12
0 10
−2 8
−4 6
−6 4
−8 2
−10 0
0.0 0.5 1.0 1.5 2.0
1000 K/T
Fig. 10.8 Rate constant (left scale) and ratio of the rate constants (right scale) for H2 + O2 abstraction reaction and
H2 + M dissociation reaction, as a function of temperature (top scale) and inverse temperature (bottom scale)
n C11 H23 ! n C6 H12 þ n C5 H11 k ¼ 2:5 1013 eð14433=TÞ s1

ð10:33Þ
alkyl radical alkene lower alkyl radical
n C11 H23 þ O2 ! n C11 H22 þ HO2 k ¼ 1:0 1012 eð1000=TÞ mol1 cm3 s1
hydroperoxy ð10:34Þ
alkyl radical alkene
radical
Examining the activation energy for the two decomposition reaction to the consumption of
competing reactions reveals that the first, the n-C11H23∙ increases at higher temperature by
thermal decomposition route (Ea/R ¼ 14,443 K), many orders of magnitude. To compare the
is very temperature sensitive, whereas the second rates of disappearance of the fuel radical
route (Ea/R ¼ 1,000 K), H-atom abstraction by d[n-C11H23∙]/dt by the two routes, it is necessary
O2 (a radical propagation reaction), is less tem- to use Equation 10.21 and Equation 10.19 for
perature sensitive. Figure 10.9 shows the ratio of unimolecular and bimolecular reactions, so that
the rate constant for the two reactions. As the ratio of rates is ωdec/ω+O2 ¼ kdec/(k+O2[O2])
indicated, the relative importance of the in which the concentration of [O2] is determined
www.ebook777.com
T/K
2000 1500 1000 500
0 5
ωdec / ω+O2
-1 4
-2 kdec / k+O2 3
log (kdec/k+O2/ cm3mol-1s-1)
-3 2
log (ωdec/ω+O2)
-4 1
-5 0
−6 −1
−7 −2
−8 −3
−9 −4
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
1000 K/T
Fig. 10.9 Ratio of the rate constant (left scale) and reaction rate (right scale) for decomposition (dec) vs. H abstraction
by O2 (+O2) reactions for n-C11H23 , as a function of temperature (top scale) and inverse temperature (bottom scale)
at each temperature via Equations 10.11 and combustion behavior of different hydrocarbons
10.12, with XO2 ¼ 0.21 in air. As shown in the can be very similar in flames.
figure, the reaction rates are about equal at
850 K; whereas the decomposition rate is 1000 Relative Rates of Reaction of OH with CO
times slower at 600 K, and 1000 times faster at vs. Hydrocarbon
1500 K, then the abstraction route. The reaction of carbon monoxide, an intermedi-
The competition between radical abstraction ate species in hydrocarbon-air flames, controls
and thermal decomposition reactions has several two of the most important features of fires.
consequences. The relative rate of reaction is (1) Most of the heat release in a flame occurs via
dependent upon the oxygen concentration, so in conversion of CO to CO2, and (2) residual CO, the
fuel rich conditions, thermal decomposition will most important toxic by-product of flames (and
be favored, leading to the build-up of small fuel the species responsible for most fire deaths) is
radicals, and ultimately acetylene, which is a often controlled by its reaction rate. As described
precursor for soot. Also, at temperatures below below, the consumption of CO in flames occurs
1000 K, the combustion chemistry tends to be almost entirely by its reaction with OH. Hence, it
specific to the primary fuel structure; whereas at is of interest to compare the rate of OH reaction
higher temperatures the large fuels will tend to with CO to that of OH reaction with other hydro-
decompose to smaller hydrocarbon fragments, so carbon species (for example, the fuel itself).
that ultimately, the important reactions for heat Figure 10.10 shows the reaction rate constant
release all involve combustion of the same, much for OH reaction with n-C4H10 or CO. As the
smaller (e.g., one- and two-carbon) hydrocarbon figure indicates, the reaction of OH with
species. This explains why the macroscopic n-C4H10 is on the order of 140 times faster than
T/K
2000 1500 1000 500
14 200
180
kC4H10+OH
160
140
kC4H10+OH / kCO+OH
120
kC4H10 + OH / kCO+OH
12 100
80
kCO+OH
60
40
20
10 0
0.0 0.5 1.0 1.5 2.0
1000 K/T
Fig. 10.10 Rate constant k (left scale) and ratio of rate constants (right scale) for reaction of OH with CO or n-C4H10,
as a function of temperature (top scale) and inverse temperature (bottom scale)
its reaction with CO at 1800 K. This illustrates decomposition products. Hence, it is of interest
that in many flames, the burnout of CO will be to compare the rates of these two processes.
kinetically limited until the hydrocarbon content Figure 10.11 shows the rate constant (left
is much lower than the CO concentration; i.e., for scale) for the H-atom abstraction reaction
n-C4H10 (and many other hydrocarbons) n-C4H10 + H ! C4H9 + H2 and for the reaction
concentrations of 1/140 that of CO, the rate of the H + O2 ! OH + O (using the rate
consumption will be about equal. The burnout of expressions: k ¼ 3.1 1014 exp(4320/T) / mol1-
the CO is typically the last stage of reaction cm3 s1; and k ¼ 1.99 1014 exp(8460/T)/
sequence and occurs after the hydrocarbon is mol1 cm3 s1, respectively). The ratio of the
essentially gone. rate constants is also shown (right scale). As
indicated, both reactions have a comparatively
H-Atom Reaction with O2 vs. Reaction high temperature sensitivity, so they become
with Fuel increasingly important at high temperatures. Fur-
The reaction of H + O2 ! OH + O (reaction thermore, the abstraction reaction is roughly
10.6) is the most important chain branching reac- 20 times faster at typical flame temperatures
tion in combustion, and greatly increases radical (1200–2000 K) and in the range of 50–1000
concentrations in the flame. Nonetheless, as times faster at temperatures between 650 and
described above, reactions of hydrocarbon spe- 1000 K. Hence, the chain branching reaction is
cies with radicals (chain propagation processes most influential after the hydrocarbon is con-
such as reaction 10.32) are largely responsible sumed. This property of the kinetics has a large
for the consumption of the fuel species and its influence on flame structure. The regime of the
www.ebook777.com
T/K
2000 1500 1000 500
14 1000
kH+C4H10
800
kH+C4H10 / kH+O2
kH+O2
12
kH + C4H10/ kH + O2
600
400
10
200
8 0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
1000 K/T
Fig. 10.11 Rate constant k (left scale) and ratio of rate constants (right scale) for reaction of H with O2 or n-C4H10, as a
function of temperature (top scale) and inverse temperature (bottom scale)
chain-branching reactions is separated, either and time.) For example, in a co-flow laminar
physically or temporally, from the regime of diffusion flame burner, the high-temperature reac-
fuel consumption, and one cannot have a large tion zone (i.e., the flame sheet) forms a conical
build-up in radical concentrations until the fuel shape around the fuel stream above the exit jet of
species are significantly depleted. Nonetheless, in the burner. Figure 10.12 shows of the volume
the context of spatial structure of a flame, the fraction profiles for some of the species as a func-
effect is mitigated somewhat by the diffusion of tion of radial position across the high-temperature
radicals (especially H atoms) which are reaction zone in a cup-burner flame. The peak
transported to regions where they are needed for temperature occurs at r ¼ 8.5 mm, and the
propagation reactions. concentration of O2 and CH4 decrease near that
location. On the fuel side of the reaction zone
(i.e., r < 8.5 mm), the species consist mostly of
Flame Configuration/Structure Effects the fuel (CH4) and its decomposition products
on Chemistry (e.g., CH3); while on the oxidizer side, the major
species present are O2 and the products of com-
As described above, there often exists competi- bustion (e.g., H2O, CO2) which are diffusing radi-
tion between possible reaction paths in combus- ally out from the flame. Due to diffusion, there is
tion systems, controlled largely by the some mixing of species from each side. One nota-
temperature and chemical environment. The ble feature is that the peak of the radical pool
chemical environment can vary due to different occurs on the oxidant side of the reaction zone,
initial conditions, changes with time, or the struc- nearly coincident with the peak temperature.
ture of the flame itself. (“Flame structure” These results can be explained as follows.
constitutes the profile of concentrations of spe- First, the major chain branching reaction H + O2
cies, temperature, and flow velocities over space ! OH + O (reaction 10.6) has a relatively high
Fig. 10.12 Calculated species volume fraction, temperature, and heat release rate at a height of 11 mm above a
cup-burner flame of CH4 and air
activation energy (E ~ 70 kJ mol-1), so its occur- radicals required in both reactions react faster
rence requires high temperature; but more impor- with CH4). The consumption flux of CH4 creates
tantly, it requires O2, so that reaction must occur a dilemma: it needs both CH4 and H, but they
primarily on the air side of the flame. Secondly, cannot both co-exist at high concentration
the rate of reaction of H with CH4 is much faster because the chain-branching reactions producing
than with O2, so reaction 10.6 cannot dominate H will not occur until hydrocarbons are consid-
on the fuel side. Similarly, the reaction CO + OH erably depleted. This is solved by species trans-
! CO2 + H, which is responsible for CO con- port: H atoms are produced at a high rate near the
sumption (as discussed above), must also be on peak temperature, but diffuse rapidly upstream
the oxidant (air) side of the reaction zone: it where they are consumed by reaction with CH4.
needs OH, which would be consumed preferen-
tially by the hydrocarbons on the fuel side. Since
the CO + OH reaction, forming CO2, is also Super Equilibrium
responsible for a large fraction of the heat
release, its location also dominates the location Chemical equilibrium is an idealized state, which
of the heat release and peak temperature. is sometimes achieved for select conditions in
The structure of a premixed flame (Fig. 10.13) fires, but often is not realized in practice. In
also results from the hydrocarbon kinetics premixed and diffusion flames of hydrocarbons
described above. As illustrated, the CO con- with air, for example, the radical pool species H,
sumption and the peak of the chain-branching OH, and O, can achieve concentrations several
reaction 10.6 are both retarded until the CH4 is orders of magnitude higher than those calculated
nearly gone (since, as described above, the at thermodynamic equilibrium. For example,
www.ebook777.com
Fig. 10.13 Major species

profiles and reaction fluxes
for laminar premixed
methane-air flame
(From [3])
Fig. 10.14 shows the ratio of peak to equilibrium forming the opposing flows in the burner. As
H-atom volume fraction in a methane-air coun- indicated, super-equilibrium ratios range from
terflow diffusion flame; the results are plotted as 1 at zero strain, to 400 at a strain rate of 190 s-1.
a function of strain rate, which is proportional to The super-equilibrium concentrations occur
the jet velocities of the methane and air streams because the reactions which form the radicals
Fig. 10.14 Ratio of peak to final equilibrium H-atom volume fraction as a function of strain rate in a methane-air
counterflow diffusion flame
(typically, the chain-branching reactions such as Role of Trace Species

H + O2 and O + H2) are much faster than the
reactions which recombine radicals (typically, There are several classic examples in combustion
the three-body recombination reactions such as in which very low concentrations of reactants
H + OH + M). The species do not reside long completely change both the route for fuel decom-
enough in the main reaction zone for the recom- position, as well as the overall rate of reaction.
bination reaction to establish thermodynamic While they are sometime esoteric, they demon-
equilibrium at the prevailing temperature. The strate the principles which are important for more
result is that radical pool species are present in practical situations.
very high concentration, leading to the rapid
attack on the fuel species, and fast overall reac- Moisture in CO Oxidation
tion. If the radicals are controlled to be closer to In early research, it was found that the oxidation
equilibrium levels (even at the elevated rate of CO to CO2 is highly dependent on
temperatures of the flames), the overall reaction the presence of trace quantities of water in the
rate is greatly reduced. system. In fact, many hydrogen containing
Premixed flames also have large super- compounds (e.g., H2, or a hydrocarbon) can
equilibrium of radicals. For example, in a supply the trace hydrogen atom necessary for
stoichiometric methane-air flame, the peak vol- the faster reaction. In the bone-dry system, the
ume fraction of H, O, and OH (which occur near reaction of CO proceeds as follows. The initia-
the location of the peak of the H + O2 reaction tion step
flux) are 17, 14, and 2.6 times the final equilib-
rium value, respectively. Note that it is gener- CO þ O2 ! CO2 þ O ð10:35Þ
ally accepted that the chemical action of fire
suppressants is to decrease the peak radical is slow. It is believed to be followed by
concentration towards the final equilibrium
O þO2 þ M ! O3 þ M
level.
www.ebook777.com
O3 þ CO ! CO2 þ 2 O ð10:36Þ Fe(CO)5
CO þ O þM ! CO2 þ M ð10:37Þ +O2 +O

Fe FeO2 FeO
which are all slow at combustion temperatures.
In contrast, in the presence of water, the reac- +H
+H2O
tion proceeds via: +H
CO þ O2 ! CO2 þ O ð10:38Þ FeOH Fe(OH)2

+H
O þH2 O ! 2OH Fig. 10.15 Schematic diagram of reaction pathways of

Fe(CO)5 in premixed methane-air flames
CO þ OH ! CO2 þ H ð10:39Þ

This is followed by the formation of the
inhibiting species FeO, Fe(OH)2, and FeOH via
H þO2 ! OH þ O ð10:40Þ the sequence shown in Fig. 10.15:
with the subsequent build-up of the radical pool, Once formed, the three reactions (forming
followed by fast reaction in the system. A similar a triangle on the right side of Fig. 10.15) enter
mechanism exists if H2 is present rather than into a gas-phase catalytic cycle:
H2O. At an H2 content of 0.1 %, the moist reac-
tion route accounts for about 70 % of the CO FeO þ H2 O ! FeðOHÞ2 (10.42)
FeðOHÞ2 þ H ! FeOH þ H2 O (10.43)
reaction. Indeed, for a premixed CO-air flame, FeOH þ H ! FeO þ H2 (10.44)
the presence of only 2000 μL/L (ppmv) of H2 can —————————————
increase the burning velocity by factor of two. Net : H þH ! H2
Flame Inhibition by Iron-Containing which effectively recombines H atoms into H2,

Compounds thereby greatly reducing the radical
In early work with premixed CO-air flames, it concentrations and essentially suppressing the
was also found that trace contamination of the flame propagation chemistry. (Recall that the
CO with iron pentacarbonyl strongly affected radicals are typically present at super-equilibrium
the reaction mechanism. It was found that iron quantities, so the effect of reactions 10.42 to
in the high-pressure steel bottle containing the 10.44 is to drive the system to equilibrium
CO reacted with the CO to form Fe(CO)5, which more rapidly.) In the rate constant expression
had high enough vapor pressure to enter the Equation 10.28, the activation energy (Ea/R)
reactant stream (at trace quantities) with the for these reactions is low (0 K, 302 K, and
CO. Later work showed that the iron enters into 604) K, and the pre-exponential factor A’ is high
catalytic gas-phase reactions which have a very (1013.2 s1, 1014.3 s1, and 1012.1 s1), so these
strong inhibiting effect on the build-up in the reactions proceed at nearly gas-kinetic rates: that
radical pool. The inhibiting effect acts similarly is, as fast as the molecules can collide, they react.
for all hydrocarbon flames. Indeed, it has been In fact, further analysis has shown that these
found that iron compound concentrations as low reactions of iron in flames approach the ideal
as 1 μL/L can reduce the burning velocity of gas-phase catalytic radical recombination
premixed hydrocarbon-air flames by 1 %. The rate [11].
mechanism is as follows. The Fe(CO)5 readily
decomposes, via a sequence of steps, to form Flame Inhibition by Bromine-Containing
Fe and 5 CO: Compounds
Similar catalytic cycles are present for bromine
FeðCOÞ5 ! Fe þ 5CO: ð10:41Þ added to hydrocarbon flames. For example, the
catalytic cycle of HBr in a methane-air flame can be different. Of course, since the reactions are
be represented by: elementary, they should apply to both situations,
and the two mechanisms could be combined into
H þHBr ¼ H2 þ Br (10.45) a larger, comprehensive unit. Nonetheless, the
H þBr2 ¼ HBr þ Br (10.46) run time of a simulation depends on the size of
Br þBr þM ¼ Br2 þ M (10.47)
————————————— the chemical mechanism, so it is usually desir-
Net : H þH ! H2 able to minimize the number of species (i.e., the
number of variables) and reactions in a model.
The net effect of adding HBr to a flame typi- Selecting the right chemical model, and its
cally is to recombine H atoms into H2 (as in the complexity, are an essential first step for under-
iron mechanism above), and thereby reduce the standing the chemical behavior of a system.
overall reaction rate. It should be noted that in Often, one will eliminate chemistry that is not
actual methane-flames, several regeneration important for the present problem. For example,
steps for the inhibiting molecule HBr are also while the chemistry for the formation of NOx is
important, such as CH2O + Br∙ ¼ ∙HCO + HBr, always occurring in high-temperature flames of
and CH4 + Br∙ ¼ ∙CH3 + HBr. These reactions hydrocarbons with air, the NOx species are rela-
contribute to cycles like reactions 10.45 to 10.47; tively trace compounds, and have a minor effect
however, they are not so easy to outline in a on the development and overall energetics of
concise list because there are many reactions flames. Hence, those reactions can be eliminated
involved in the consumption of other radicals, if the goal of a study is to understand a feature
such as ∙CH3 and ∙HCO. This basic mechanism such as the rate of heat release, rather than pol-
is believed to occur also during the action of lutant (i.e., NOx) formation.
other bromine-containing additives such as In combustion, detailed chemical kinetic
CF3Br (halon 1301). models are hierarchical: a mechanism for a larger
reactant will contain the entire kinetic model of
each of the smaller species which are present in
Chemical Kinetic Models the larger model. The simplest kinetic model is
that of the H2/O2 system, and this is contained in
Overview all larger mechanisms; the H2/O2 kinetic model is
followed in complexity by those for CO, CH2O,
A chemical kinetic model starts with a list of CH4, and C2H6. As a result, a mechanism for
species and the reactions between them. The cetane (C16H34) can have 1200 species and more
thermodynamic data for the species are required, than 7000 reactions [6]. Usually, each of the
as are the parameters for the specific reaction rate smaller, simpler mechanisms is experimentally
constants of all reactions. Either the forward or validated prior to development of the next most
reverse reaction data are required, with the equi- complex model, so the reaction sets are internally
librium relationship usually then providing the consistent. Because all the mechanisms are
rate parameters for the counterpart reaction. based on elementary rates, the mechanisms
The relevant reactions necessary for inclusion should be widely applicable. Unlike the example
depend upon the environment, and the desired of NOx chemistry, the chemistry of smaller
information. For example, a mixture of hydrocarbon species cannot be eliminated
hydrocarbons and air would have a different from kinetic models of larger species; rather,
chemical mechanism for understanding the they are essential.
atmospheric chemistry at 298 K as opposed to Even for a kinetic model based on elementary
flame chemistry at 2000 K. In the former, only rates, one still needs to exercise caution in its
reactions prevalent at low temperature are impor- application. The species and reaction set of a
tant, whereas in the latter, high-temperature pro- mechanism are developed for a specific range
cess are considered; the participating species can of conditions (e.g., reactant composition and
www.ebook777.com
stoichiometry, flame type, temperature range,

etc.). It is incumbent upon the user to understand Role of Gas-Phase Kinetics in Some
the conditions for which the mechanism was Fire Problems
developed, and interpret if those conditions are
close enough to the condition of interest for the Understanding Standard Fire
mechanism to be useful. An example of this Tests Through Kinetic Modeling:
is the NIST chemical kinetic mechanism for The Cup Burner Test
hydrofluorocarbon (HFC) inhibition of hydrocar-
bon flames. The mechanism was developed The cup burner method is a standard test [16]
based on the assumption that the flame inhibitors used for specifying the amount of fire suppres-
(for example C2HF5) would be added in small sant to be used in full-scale applications. In this
amounts (a few percent) to hydrocarbon-air test, a small cup (28 mm dia) holds the liquid
flames. Recently, it has become of interest to fuel, which is supplied to the burner so as to
understand the flammability of pure refrigerants maintain the level of the fuel to the top edge of
with air, as well as the reaction of flame the cup. For gaseous fuels, a fine screen
inhibitors at very high loading. For these distributes the low velocity gas evenly over the
conditions, additional species are likely to partic- exit nozzle of the burner. A chimney (10 cm
ipate, and the mechanism will need to be diameter) holds the gaseous oxidizer, which is
extended. usually air mixed with fire suppressant. By
The important reaction pathways can some- increasing the volume fraction of the fire sup-
times be very similar for different fuels. As pressant, its extinguishing concentration is deter-
outlined by Warnatz [12], in combustion at high mined as the volume fraction in air at which the
temperatures (i.e., near 2000 K), larger fuel flame is extinguished.
molecules quickly break down to C1 and C2 As with any standard test method, there is
species, and once these form, their decomposi- always the question as to how the results
tion pathways are similar for all hydrocarbon obtained in small laboratory experiments
flames. Similarly, as outlined by Babushok and corresponds to full-scale fires. Insight into this
Tsang [13], the burning velocity of hydrocarbons question can be gained by numerically modeling
tends to be controlled by a small set of reactions the test method, interpreting the controlling
involving small decomposition products of the physics, and then comparing those with the
larger fuel molecules (and other common species parameters believed to be controlling at full
such as O2). Also, the mechanism of flame inhi- scale. Since flame extinction is intimately
bition of halogenated compounds also tends to be connected with flame chemistry (and overall
very similar for different hydrocarbon fuels reaction rate) a good chemical kinetic description
[13]. The reactions which form and destroy the of the flame is necessary. It is possible to model
radical pool are similar, so the effect of the flame cup-burner flames using direct numerical simu-
inhibitors (which act to reduce radical lation, in which the momentum, energy, and spe-
concentrations) are also related. cies conservation equations are solved directly.
Using available chemical kinetic mechanisms,
the numerical calculation (2-D axi-symmetric)
Databases produces detailed information on the flame struc-
ture: i.e., the species concentrations, gas velocity
There are many sources, both in publications and and temperature, at all locations in the calcula-
online, for thermodynamic data, elementary tion domain, as a function of time. Using post-
reaction rates, and comprehensive kinetic processing techniques on the data to generate the
models. Some thermodynamic data can be reactive and diffusive flux of species, and heat
found in [14] and [15], and reaction rate (and transfer properties of the flame, great insight can
other) data in [6–8, 10]. be gained (for example, Ref. [17]).
Fig. 10.16 Calculated

species profiles at a height
of 16 mm above a
cup-burner flame of CH4
and air, with C2HF5 added
to the air stream at a
volume fraction of 9.3 %
(From [17])
Figure 10.16 show the calculated flame struc- premixed character and a higher reactivity. At
ture for a cup-burner flame of methane, with the extinguishment agent volume fraction, fur-
C2HF5 added to the air stream at a volume frac- ther lift-off does not sufficiently increase the
tion just below extinguishment (0.093). Results reactivity to overcome the loss of reactivity
such as these have been used to answer such caused by agent addition, and the premixed
questions as: flame base cannot stabilize in the flow field.
1. Why do cup-burner flames always extinguish 2. Simulations have shown that essentially inert
due to flame lift-off, rather than extinction agents (such as CO2, N2, Ar, He, etc.) act by
elsewhere? lowering the temperature at the flame base,
2. Does a particular fires suppressant work by reduce reaction rates there, and cause lift off
thermal or chemical kinetic mechanisms? (as described above). For other agents (such
3. Does a fire suppressant in the air stream add to as CF3Br, Br2, CF3H, C2HF5, etc.) the peak
the heat release in cup-burner flame? temperature (as well as the temperature in
4. Why are cup-burner flames in low gravity the stabilization region) is actually increased
harder to extinguish than those in normal with agent addition; hence, the mechanism
gravity? cannot be thermal in origin. Further analysis
Analysis of cup-burner flames has shown that: has shown that chemical reactions at the flame
1. The flame is stabilized at the base, by a base reduce radical concentrations more there
partially-premixed region which has a much than further up in the flame, causing flame
higher reactivity than the rest of the flame. extinguishment preferentially at the base.
Addition of the fire suppressant to the air stream 3. By integrating the heat of reaction for all
lowers the reactivity (either thermally or kinet- reactions over the entire calculation domain,
ically) at the flame base. The base can find a it has been demonstrated that, for agent volume
new stabilization point if it lifts off further: fractions just above that for extinguishment,
higher lift-off allows more time for mixing of CF3Br reduces the total heat release, whereas
the fuel and air streams, which leads to more C2HF5 increases (i.e. nearly doubles) it.
www.ebook777.com
4. Fire safety in spacecraft requires an understand- 8. H. Wang, Available Reaction Models, University of
ing of the behavior of flames in zero gravity Southern California, Los Angeles, CA; http://ignis.
usc.edu/Mechanisms/Model%20release.html, 2012.
(0-gn). Experiments and simulations have 9. F.L. Dryer, Kinetic Models, Princeton University,
shown that despite 0-gn flames being much Princeton, NJ; http://www.princeton.edu/mae/people/
weaker, they require more fire suppressant for faculty/dryer/homepage/kinetic_models/, 2012.
extinguishment. Simulations have explained 10. J.A. Manion, R.E.L.R.D. Huie, D.R. Burgess,
V.L. Orkin, Tsang, W.S. McGivern, V.S. Hudgens,
this unexpected behavior. The normal-gravity V.D. Knyazev, D.B. Atkinson, E. Chai, A.M. Tereza,
(1-gn) flames have buoyancy-induced flicker, C.-Y. Lin, T.C. Allison, W.G. Mallard, F. Westley,
which causes the flame base to oscillate J.T. Herron, R.F. Hampson, D.H. Frizzell, NIST
strongly. This creates a more difficult flow Chemical Kinetics Database, NIST Standard Refer-
ence Database 17, Version 7.0 (Web Version),
field for flame base stabilization, and causes Release 1.4.3, Data Version 2008.12, National Insti-
the flame to blow off earlier than it would with- tute of Standards and Technology, 2012.
out the flame flicker. Hence, the better-stabilized 11. V.I. Babushok, W. Tsang, G.T. Linteris, D. Reinelt,
0-gn flames require more suppressant to lower Combust. Flame 115 (1998) 551–560.
12. J. Warnatz, Pure and Applied Chemistry 72 (2000)
the reaction rate at the base and cause blow-off. 2101–2110.
The significance of the work to understand the 13. V.I. Babushok, W. Tsang, Combust. Flame
chemical kinetic behavior of laboratory flames 123 (2000) 488–506.
goes beyond the goal of understanding standard 14. A. Burcat, B. Ruscic, Third Millennium Ideal Gas and
Condensed Phase Thermochemical Database for
test methods. Twenty years ago, such 2-D time- Combustion With Updates From Active Thermo-
dependent simulations with full chemistry were not chemical Tables, ANL-05/20 and TAE 960
possible, even for cup-burner flames; but today Technion-IIT; ftp://ftp.technion.ac.il/pub/supported/
they are common. As computer speed, numerical aetdd/thermodynamics/BURCAT.THR, Argonne
National Laboratory, 2012.
methods, and chemical kinetic models all improve, 15. in: P.J. Linstrom, W.G. Mallard (Eds.), NIST Chem-
it will not be long before such simulations with istry WebBook, NIST Standard Reference Database
detailed chemistry are possible for the larger Number 69, National Institute of Standards and Tech-
domains typical of fires. When that happens, the nology, Gaithersburg MD, 20899 (http://webbook.
nist.gov), 2001.
full potential of chemical kinetics for understand- 16. NFPA 2001 Standard on Clean Agent Fire
ing fire problems will start to be realized—as Extinguishing Systems 2008 Edition, NFPA, 2007.
has occurred in other areas of combustion. 17. F. Takahashi, G.T. Linteris, V.R. Katta, O. Meier,
Proc. Combust. Inst. 34 (2012) 2707–2717.
References Gregory T. Linteris is a mechanical engineer in the

Flammability Reduction Group of the Fire Research Divi-
1. S. Benson, The Foundations of Chemical Kinetics, sion of the Engineering Laboratory at the National Insti-
McGraw-Hill, New York, 1960. tute of Standards and Technology. Dr. Linteris is a project
2. J. Warnatz, U. Maas, R.W. Dibble, Combustion, leader for research on material flammability in the Flam-
Springer-Verlag, Berlin, 2010. mability Reduction Group. He also conducts research to
3. S.R. Turns, An Introduction to Combustion, McGraw- understand the detailed mechanisms of chemically acting
Hill, Boston, 2000. fire suppressants. In 1997, Dr. Linteris served as a payload
4. I. Glassman, Combustion, Academic Press, San specialist astronaut on two NASA space shuttle missions,
Diego, CA, 1996. conducting microgravity combustion, fluid mechanics
5. J. Warnatz, in: W.C. Gardiner (Ed.), Combustion and material science experiments while in earth orbit for
Chemistry, Springer-Verlag, New York, 1984. 20 days.
6. C.K. Westbrook, Pitts, LLNL Chemical Kinetic
Mechanisms for Combustion, https://www-pls.llnl. John F. Griffiths is emeritus professor of combustion
gov/?url ¼ science_and_technology-chemistry-com chemistry at the University of Leeds in the United
bustion-mechanisms, Lawrence Livermore National Kingdom. His main research interest is the gas-phase
Laboratory, 2012. combustion of hydrocarbons, which is related both to
7. C.K. Law, Law Combustion Group: Chemical Kinetic combustion hazards which may arise from spontaneous
Database, Princeton University, Princeton, NJ, http:// ignition processes in the chemical industry and to the
www.princeton.edu/~cklaw/kinetics/, 2012. efficiency of reciprocating engines.
Diffusion Flames
11
Ali S. Rangwala
Fires involve reactants, usually fuel and air, not The Diffusion Coefficient
intimately mixed at a molecular level before
combustion. Usually, the fuel is in the solid or Diffusion is the phenomena of migration of mass.
liquid state so transfer of material across a phase The mass can be in the form of atoms, molecules,
boundary (phase change) must also occur. The ions, or other particles because of spatial gradient
vaporized fuel must combine with oxygen from of some quantity (concentration, temperature,
air to form a flammable mixture, which when pressure etc.). Similar to conduction heat transfer
ignited forms the flame zone. In most fire (Fourier’s law) and momentum transfer (Newton’s
problems, this mixing of fuel vapor and oxygen law), mass transfer is governed by a law called as
takes place mostly by diffusion and takes orders Fick’s law of diffusion. In a simplified context,
of magnitude longer time compared with that of Fick’s law of diffusion describes the movement
a chemical reaction. Therefore, diffusion of spe- of one chemical species A through a binary
cies is the primary controlling process during mixture of A and B because of concentration gra-
such burning behavior. A fundamental under- dient of A. In most fire problems A is usually fuel
standing of diffusion flames then involves vapor, oxidizer or products of combustion, while B
exploring the mechanisms associated with the represents air. To explain this further, let us con-
transport of the reactants and the resulting flame sider the example of a candle flame shown in
structure. Fig. 11.1. The paraffin of the candle melts because
In Fire Protection Engineering, diffusion flame of heat from the flame; it travels by capillary forces
theory is used in calculating flame-length, flame- through the wick where it then evaporates to
location, and rates of burning. The flame-length, is become paraffin vapor, a gaseous fuel.
used for hazard analysis as it provides information Let F represent fuel vapor, O represent oxy-
to estimate the heat transfer to surrounding gen and P represent products. Fuel vapor will
surfaces. Knowledge of flame location is neces- issue out of the wick because of the heat received
sary for suppression, and finally, the rate of burn- from the flame. If one traverses along the path
ing provides an estimate of the “size” of the fire X-X’ (shown by dashed red line in Fig. 11.1) the
and in combination with the heat of combustion is concentration of fuel vapor is highest at the wick
used to calculate the heat release rate. and reduces until it reaches a concentration most
suitable for chemical reaction with oxygen at X’.
The concentrations of the various species
involved, can be given by the mass fractions,
A.S. Rangwala (*)
Associate Professor, Department of Fire Protection YF, YO and YP, where YF denotes the mass of
Engineering, Worcester Polytechnic Institute, fuel vapor divided by the total mass of
100 Institute Road, Worcester, MA 01609-2280

www.ebook777.com
11 Diffusion Flames 351
Fig. 11.1 The candle flame
gas-mixture in a given volume and the subscripts XX’, represented as x-direction. This is logical
O and P denote oxidizer and products, respec- because the fuel vapor originates at the wick
tively. Initially the mass fraction of F at the wick (location X). Equation 11.2 is also called the
is 1 at the wick and zero just outside the wick, Ficks law of diffusion and forms the starting
and this gradient drives the fuel vapor out of the point of our discussion on diffusion flames.
wick. On approaching the flame zone, the mass Note that similar relationships can be written
fractions of the fuel and oxidizer should be such for oxygen (diffusing towards the flame) and
that a stoichiometric mixture should be formed. products of combustion (diffusing on either side
Now, Fick’s law states that the mass transport of of the flame).
fuel vapor along XX’ due to mass diffusion can Table 11.1 gives the binary diffusion
be described by: coefficients D for many common gases. Values
refer to atmospheric pressure. The first part of the
Y F, X Y F, X0 table gives data for several gases in the presence
J F ¼ ρDFA : ð11:1Þ
XX0 of a large excess of air. The mass diffusivity D, is
an important transport property such as thermal
JF is the mass flow rate of fuel vapor per unit diffusivity α and momentum diffusivity (kine-
area (or mass flux) and is proportional to the mass matic viscosity) υ, and all three have dimensions
fraction difference divided by the distance from of (length)2/time. The ratios of these quantities,
the wick to the flame (XX’). ρ is the density of the taken as a pair, form three important nondimen-
gas-mixture system and DFA is a proportionality sional numbers that play a prominent role in
factor called as the binary mass diffusivity of fuel analyzing most fire problems. They are Prandtl
vapor with respect to air. In differential form, number, Pr ¼ υ/α, Schmidt number, Sc ¼ υ/D
Equation 11.1 can be written as: and Lewis number, Le ¼ α/D. Equation 11.2
assumes that the driving force for diffusion is
dY F concentration gradient. However, it has been
J F ¼ ρDFA : ð11:2Þ observed, as well as predicted by the kinetic
dx
theory of gases, that mass diffusion can occur in
The negative sign denotes that the mass fraction the presence of a temperature gradient and pres-
of fuel vapor will decrease as one moves along sure gradient. Further, the value of the binary
352 A.S. Rangwala
Table 11.1 Diffusion coefficients of common gases at shown by the dashed lines. Products of combus-
0 C, 760 mmHg. Assuming ideal gas behavior, D can be tion and heat diffuse outwards from the flame
calculated for other pressures and temperatures using the
zone to both the sides. One of the characteristics
relation D / P1 T 3=2
of typical hydrocarbon diffusion flames is their
Gas-pair D (cm2/s) yellowness, especially when the fuel can emit
O2—air 0.178 soot, and it’s appearance can be explained using
CO2—air 0.138
Fig. 11.2 [3].
H2—air 0.611
The reaction zone usually has a blue emission,
H2O—air 0.220
especially when the fuel and oxidizer have been
Methane (CH4)—air 0.196
mixed in proper proportions. This is mainly due
Ethane (C2H6)—air 0.108
Propane (C4H10)—air 0.0878
to radiation due to excited CH radicals. The
Butane (C3H8)—air 0.0750 reddish glow arises from radiation from CO2
Pentane (C5H12)—air 0.0671 and water vapor radiation. Most importantly,
n-Octane (C8H18)—air 0.0505 the intense yellow radiation which is a character-
Benzene (C6H6)—air 0.077 istic of most fires is due to the presence of carbon
Toluene (C7H8)—air 0.051 particles or soot. Figure 11.2a illustrates a
Napthalene (C10H8)—air 0.0513 simplified illustrative sketch of three prominent
Anthracen (C14H10)—air 0.0421 zones in a diffusion flame. Note that emphasis is
Methyl alcohol (CH3OH)—air 0.1325 given to the fuel side. The cracking zone is a
Ethyl alcohol (C2H5OH)—air 0.102 region on the fuel side of the reaction zone before
the soot formation zone. This where the
molecules crack and polymerize forming lighter
mass diffusivity itself depends on temperature
fuel molecules which chemically react with the
and pressure (D P1T3/2). The diffusion coef-
oxygen in the reaction zone and rest convert to
ficient of common gas-mixtures at various
carbonaceous and tarry substances in a soot for-
temperatures and pressures can be obtained
mation zone that exists just before the reaction
using this equation in combination with values
zone. Soot generally forms as particles with
listed in Table 11.1. However, this effect is usu-
diameters of the order of several nanometers by
ally neglected in fire problems. Further details on
a process called as inception or nucleation. These
the subject can be found in Ref. [1].
particles then undergo surface growth. One
mechanisms attributed to the surface growth is
called as the Hydrogen Abstraction by C2H2
Structure of Diffusion Flame Addition (HACA) mechanism [6] where
H-atoms impacting on the soot surface activate
The zoomed inset of the part of the flame zone in acetylene addition thereby increasing the mass of
Fig. 11.1 shows an illustrative sketch of the soot particles. The process of nucleation occurs
structure of a diffusion flame. It consists of a concurrently with coagulation, where small
flame separating a fuel-rich zone and an particles coalesce to form larger primary
oxidizer-rich zone. The flame or reaction zone particles, and agglomerations where multiple pri-
incorporates the location of the maximum tem- mary particles line up end-to-end to form larger
perature. For example, for a methane-air flame, structures resembling a string of pearls as shown
this temperature is experimentally observed to be in Fig. 11.2b [5]. When the soot particles pass
around 1950 K [2]. As shown in Fig. 11.1, fuel through the flame front they oxidize whereby
and oxidizer both almost disappear in the flame the mass of soot is decreased by heterogeneous
zone, although there is some fuel and oxidizer surface reactions between soot particles and
leakage outside the designated flame zone as oxidizing species. Incandescent soot results in the
www.ebook777.com
Fig. 11.2 (a) Zones in a diffusion flame [4] (b) soot particle observed under a scanning electron microscope—resemblance
to string of pearls [5]
Fig. 11.3 Upward propagating flame on a PMMA surface
characteristic yellow-color of most hydrocarbon growth, the fuel composition and temperature of
diffusion flames. The processes leading to crack- the flame zone. Since time increases with linear
ing and soot formation are dependent on fuel type dimensions longer flames are usually yellower,
and ambient conditions and form an important smokier than shorter flames, indicating the escap-
research topic in fire safety [7–9]. This is mainly ing of the soot particles out of the flame zone.
because most of the flame radiation from fires Large carbon particles will radiate more heat as
originates from soot particles. The luminous radi- shown in Fig. 11.3 [10]. A laminar natural convec-
ation from soot often makes it difficult to observe tion flame is blue at the base, yellow in the center
the blue emission from the reaction zone. The blue and red/orange at the tip. The latter condition
radiation is observable for small flames (less than means that so much heat has been lost by radiation
15 cm) as shown in Fig. 11.3 where the initial from the carbon particles, that when it does cross
stages of a wall fire (PMMA, 1.2 cm thick, 5 cm the reaction zone into a region where oxygen is
wide, 50 cm long) are shown. The size of the soot available for combustion, it is too cold to burn.
particles depends on the time allowed for their The result is emission of smoke.
354 A.S. Rangwala
A useful material-property called smoke- the diffusion flame is best approached by

point of a fuel can be defined based on this considering a candle flame once again. Focus
radiative heat loss mechanism by soot. A fuels on a control volume in the gas phase as shown
smoke point is the maximum height of its lami- by the dashed boxed. The fuel vapor and oxidizer
nar flame (or fuel mass flow rate) burning in air at diffuse from opposite directions and approach
which soot is just released from the flame tip. the flame in a normal direction (Fig. 11.1,
Another definition (and more applicable to fire LHS). The concentrations of fuel and oxidizer
safety) is the heat release rate at which smoke at the flame are in stoichiometric proportion. In
just begins to be released from the flame tip. other words, the diffusion flame surface is
Smoke point is a simplified ranking scheme for defined as the locus of all points in space where
soot production and was first introduced by Kent the fuel and oxygen meet at stoichiometric
and Wagner [11]. Smoke-point can be easily proportions. A one-step chemical reaction
determined for gases and vapors by adjusting given by Equation 11.3 can be used to represent
the flow rate of the fuel from a simple burner. the overall chemical process.
For liquid fuels a wick-fed lamp (ASTM D1322)
is used. Determining the smoke point for solid 1 g½FUEL þ s g½OXIDIZER
fuels is difficult, although some progress has ! ð1 þ sÞ g½PRODUCTS þ Heat ð11:3Þ
been made in this direction by de Ris and
Cheng [12]. It has been shown [12–14] that The assumption made here is that the net disap-
smoke point can provide a convenient measure pearance rate of the reactants (fuel and oxidizer)
of the flame radiant fraction. A comprehensive is infinitely fast. This is represented by the
review on the development of an engineering zoomed inset in Fig. 11.1, where solid lines are
model capable of predicting the release of soot used to indicate the profiles of temperature and
and radiation given the smoke point of the fuel, mass fractions of the reactants. The flame zone is
stoichiometric mass ratio of the reactants and the infinitesimally thin, and both fuel and oxygen are
adiabatic stoichiometric flame temperature is consumed at this “zero-thickness” flame sheet.
discussed by Lautenberger [15]. However, in the actual scenario, the assumption
of infinitely fast reaction is not true as indicated
by the profiles in dashed lines in the flame struc-
Diffusion Flame Theory ture inset in Fig. 11.1, where the flame zone has
finite thickness and both the oxygen and the fuel
The theory of diffusion flames consists of an leak through this flame zone. The details of flame
analysis of factors controlling the mixing of broadening are beyond the scope of this chapter.
fuel and oxidizer. Main factors controlling the However, an interested reader may refer to a
mixing are mass diffusivity (D), gradient of spe- book related to the topic [16]. For our purposes,
cies mass fraction (dY/dn) normal to the the infinite rate chemistry assumption is suffi-
condensed fuel surface and the flow field. Unlike cient in predicting parameters such as the mean
premixed flame analysis, the rates of the reaction flame zone location and mass burning rate.
mechanisms do not dominate the burning behav- The one-step and infinitely fast reaction
ior in diffusion flames. As discussed earlier, in 000
000
ω_ O
000
ω_ P
diffusion flames, fuel and oxidizer come together assumptions also imply that ω_ F ¼ s ¼ 1þs ,
000
in a reaction zone through diffusion. This diffu- where ω_ denotes the nonlinear rate term
sion can be just molecular transport (candle representing the rate of formation or destruction
flame, laboratory flames) or be enhanced several of a species per unit volume. Subscripts F, O
times by convection, which may be even and P, denote fuel, oxidizer and products, respec-
turbulent (most large-scale fires such as pool, tively. The conservation equations for the control
building, forest etc.) The theoretical solution of volume are given by:
www.ebook777.com
∂Y F ∂Y F ∂ ∂Y F 000
Species Conservation ρ þ ρui ¼ ρD ω_ F ; ð11:9Þ
∂t ∂xi ∂xi ∂xi
Assumption: binary diffusion coefficients are
∂½Y P =ð1 þ sÞ ∂ ½ Y P =ð 1 þ s Þ
equal for all species. ρ þ ρui
∂t ∂xi
000
∂Y O ∂Y O ∂ ∂Y O 000 ∂ ∂½Y P =ð1 þ sÞ ω_
ρ þ ρui ¼ ρD ω_ O ; ð11:4Þ ¼ ρD þ P : ð11:10Þ
∂t ∂xi ∂xi ∂xi ∂xi ∂xi 1þs
∂Y F ∂Y F ∂ ∂Y F 000
ρ þ ρui ¼ ρD ω_ F ; ð11:5Þ
∂t ∂xi ∂xi ∂xi Modified Energy Conservation
Equation
∂Y P ∂Y P ∂ ∂Y P 000
ρ þ ρui ¼ ρD þ ω_ P : ð11:6Þ
∂ Tcp =ΔH c ∂ Tcp =ΔH c
ρ þ ρui
∂t ∂xi

Energy Conservation ∂ λ ∂ Tcp =ΔH c 000
¼ þ ω_ F : ð11:11Þ
∂xi cp ∂xi
∂T ∂T ∂ ∂T
ρcp þ ρui cp ¼ λ Equations 11.8, 11.9, 11.10, and 11.11 can be
∂t ∂xi ∂xi ∂xi combined into a single equation given by
" 000 000
#
000 ω_ O ΔH c ω_ P ΔHc
þ ω_ F ΔH c or or : ð11:7Þ LðβÞ ¼ 0; ð11:12Þ
s 1þs
In the above equations, ρ represents the gas where β can take several values as shown in
phase density, cp represents the specific heat and Table 11.2 and the operator L is expressed as,
λ equals the thermal conductivity. ΔHc represents
the heat of combustion of the fuel and D equals ∂β ∂β ∂ ∂β
LðβÞ ρ þ ρui ρD : ð11:13Þ
the diffusion coefficient which is assumed to be the ∂t ∂xi ∂xi ∂xi
same for oxygen—air, fuel—air and product—air
000 In the operator L, the first term represents the
The nonlinear rate terms ( ω_ ) can be eliminated
from the equations by suitable subtractions and accumulation of thermal energy or chemical spe-
assuming that the Lewis number is unity cies, the second term represents the convection
efflux thorough the control surfaces and the third
Le ¼ ρcλp D ¼ 1 . Multiply Equation 11.4 by 1s , represents the diffusion efflux. The non-linear
1 c
Equation 11.6 by 1þs and Equation 11.7 by ΔHp c to volumetric reactive effects are eliminated using
get the modified conservation equations.
Table 11.2 Different forms of the coupling function
b introduced in Equation 11.12
Modified Species Conservation β Value
Equations βFO Y F YsO
βFP Y F þ 1þs
YP
βFT Tc
Y F þ ΔHpc
∂ðY O =sÞ ∂ðY O =sÞ
ρ þ ρui βOP
∂t ∂xi Y O þ 1þs
YP
000 βOT Tc
Y O þ ΔHpc
∂ ∂ðY O =sÞ ω_ O
¼ ρD ; ð11:8Þ βPT Tc
Y P þ ΔHpc
∂xi ∂xi s
356 A.S. Rangwala
Fig. 11.4 The

onedimensional diffusion
flame mathematical model
the coupling function β, which can take 6 forms apparent, then, that each accidental flame is
as shown in Table 11.2. This methodology is dependent on the details of the environment in
referred to as the Shvab-Zeldovich transforma- which it burns. Thus the possible variations in
tion after two classical papers by Shvab [17] and behavior are endless. The one-dimensional flame
Zeldovich [18] that first used the coupling func- structure discussed earlier will be used once
tion. Note that although Shvab-Zeldovich pro- again to describe the problem. A further set of
posed a general solution, the original idea was assumptions will be imposed to simplify the
first proposed by Burke-Schumann [19] in 1928. math and facilitate physical understanding.
Equation 11.12 can be solved with knowledge
of initial and boundary conditions. However, this
is not an easy task! For example, the convective Assumptions
term is also non-linear unless the velocity is
constant. Further, many added assumptions 1. The oxidizer enters the system at x ¼ 0 with
such as constant, ρD, steady state ( ∂β ∂t
¼ 0), a temperature Ti, a concentration of YO,i and a
one-dimensional system, constant pressure and velocity of u ¼ U which is a constant.
low speed flow are required before analytical 2. The fuel enters the system at x ¼ 1 with a
solutions to some problems can be obtained. temperature Ti, a concentration of YF,i and a
Nevertheless the coupling function β is a power- velocity of u ¼ U Fig. 11.4.
ful tool that is used extensively in diffusion flame 3. The reaction occurs at x ¼ xf in a zone with
problems. One example is considered below. thickness, ε ! 0:
4. All reactants are consumed at the reaction
x < x f YF ¼ 0
zone so for:
x > x f YO ¼ 0
Diffusion Flame Location
The diffusion flame surface is defined as the Conservation Equations

locus of all points in space where the fuel and
oxygen meet at stoichiometric proportions. The Since the flow field is assumed to be known
position of the flame front of a diffusion flame is and constant (u ¼ U), there is no need to solve
dependent on the surrounding geometry and the the overall mass and momentum conservation
flow rates of the various gas streams. It is equations. Thus,
www.ebook777.com
Species Conservation As shown above, the coupling function or the

Shvab-Zedovich transformation eliminates the
000 2
d ðY O =sÞ ω_ F d ðY O =sÞ nonlinear reaction rate terms from the conserva-
U ¼ þD ð11:14Þ
dx sρ dx2 tion equations. We are still left with 6 boundary
000 2 conditions for the three linear equations that can
dY F ω_ F d YF be obtained with βOF, βOT and βFT. We can
U ¼ þD ð11:15Þ
dx ρ dx2 further decrease the difficulty in the solution
procedure without altering the nature of the solu-
Energy conservation tion by replacing β with a new normalized vari-
able, Z ¼ ββðð1xÞβ ð0Þ
Þβð0Þ ; which is called as a mixture
!
d T= ΔH c =c p ω_
000
d2 T= ΔH c =cp fraction. Note that β (0) and β (1) are constants
u ¼ F þα ; obtained from the boundary conditions. Since
dx ρ dx2
the boundary conditions for the variable Z at
ð11:16Þ
the boundaries x ¼ 0 and x ¼ 1 are Z ¼ 0 and
where α ¼ ρcλ p represents the thermal diffusivity. Z ¼ 1, respectively, irrespective of the
β-variable that defines Z, the solution is unique
and we need to obtain the solution to one mixture
Boundary Conditions
fraction equation only. This simple algebraic
x¼0 T ¼ Ti Y O ¼ Y O, i YF ¼ 0 manipulation allows us to obtain a single differ-
x¼1 T ¼ Ti YO ¼ 0 Y F ¼ Y F, i ential equation with normalized boundary
ð11:17Þ conditions given by:
The conservation equations are coupled using 2

dZ d Z x¼0!Z¼0
Equation 11.12. Note that with the steady, 1-D U ¼α where
dx dx2 x¼1!Z¼1
system discussed in the example, the L operator
d β 2 ð11:20Þ
reduces to LðβÞ ρU dβ
dx ρD dx2 : This gives,
2 The solution to this differential equation is
dβ d β
U ¼α : ð11:18Þ
dx dx2
ex=δ 1 α
Z ¼ 1=δ where δ¼
Three β-variables will be selected such that ð e 1Þ U
the values of temperature and mass fractions of ð11:21Þ
fuel and oxygen (three variables obtained by
solving three equations) can be evaluated later.
Going back to the dimensional values (oxygen
The choices are βOF, βOT and βFT. The variables
concentration, fuel concentration and tempera-
and the boundary conditions are listed below:
ture) requires the determination of the flame
Y O, i location. Based on the assumption that the
YO x ¼ 0, βOF ¼
βOF ¼ YF and s flame will place itself where fuel and oxidizer
s x ¼ 1, βOF ¼ Y F, i arrive in stoichiometric proportions, the
Y O, i Tcp flame location can be expressed by:
x ¼ 0, βOT ¼ þ
Y O Tcp s ΔHc β x f ¼ YsO Y F ¼ 0. Substituting the appropri-
βOT ¼ þ and
s ΔH c Tcp
ate expression for Ζ(x) in terms of β and
x ¼ 1, βOT ¼
ΔHc evaluating
Tcp
x ¼ 0, βFT ¼
Tcp ΔHc x=δ
βFT ¼ Y F þ and Y O, i Y O, i e 1
ΔHc Tcp βOF ðxÞ ¼ þ Y F, i ;
x ¼ 1, βFT ¼ Y F, i þ
ΔH c s s e1=δ 1
ð11:19Þ ð11:22Þ
358 A.S. Rangwala
the flame location, xf, is obtained as Zeldovich transformation. This is because

Z couples the transport of heat and species into a
2 1=δ 3
single variable. In some cases, heat and different
Y e 1
x f ¼ δ Ln4 þ 15 :
O, i
ð11:23Þ species may have different diffusivities and
s Y O , i þ Y F, i
s therefore can be transported at different rates.
Consequently, β and Z are no longer conserved.
Finally, assuming there is no oxidizer in the However, by making the equal diffusivity and
fuel zone and no fuel in the oxidizer zone the fuel Le ¼ 1 assumption, we are requiring that these
and oxygen concentrations can be defined as: diffusive fluxes are transported at the same rate
and hence preserving the conserved nature of Z.
x < x f Y O ¼ sβOF ðxÞ and YF ¼ 0 Assumption 4 is used in most fire problems, how-
x > x f YO ¼ 0 and Y F ¼ βOF ðxÞ ever, can be relaxed. Assumptions 5–8 were used
ð11:24Þ to solve our specific example of a one-dimensional
flame. Of these 5 can be most questionable. This is
A similar method can be followed when deter- because the flow-field plays a significant role in
mining the temperature distribution, from f2 for diffusion problems. An exact representation of
x < xf and f3 for x > xf. This is explained in the the flow-field can be obtained by solving the
worked example. overall mass and momentum conservation or the
There are several advantages in introducing Navier–Stokes equations, which require proper
the mixture fraction. The mixture fraction (Z) pressure–velocity coupling since most fires are
should satisfy the balance equation incompressible in nature. Assumption 6 is reason-
able mainly because of the slow regression rates
∂Z ∂Z ∂ ∂Z observed for most condensed fuels. Assumption
LðZÞ ρ þ ρui ρD ¼ 0; 7 facilitated an analytical solution. Assumption
8 is reasonable so long as the properties are chosen
ð11:25Þ
correctly. The correct choice of properties is
based on the definition of the operator L of hugely important in all problems of this nature.
Equation 11.10. The boundary condition equa- This is further discussed below.
tion is Z ¼ 1 in the fuel stream and Z ¼ 0 in the
oxidizer stream. This equation converts all
the Shvab-Zeldovich variables (β) into a single Property Estimation
parameter. Equation 11.25 greatly simplifies the
modeling of diffusion flames. Note that there is Most theoretical and empirical expressions to
no source term in Equation 11.25. solve fire problems usually rely on the assump-
tion of constant thermophysical properties. An
important issue in using these expressions in
practice therefore necessitates a proper method
Comments on the Formulation to evaluate the thermophysical properties such
and Analysis that results obtained through them match with
experimental data. A first step in property esti-
It is important to reiterate the assumptions used mation is obtaining certain average temperature.
in working out the solution discussed earlier. For a diffusion flame, the temperature can
1. One-step and infinitely fast reaction, 2. Lewis vary from a few hundred degrees at the fuel
number ¼ 1, 3. Binary diffusion coefficients side to around 1500–2000 K at the flame zone.
are equal for all species, 4. ρD ¼ constant, Surrounding temperature can also be of the order
5. Velocity (U) is constant, 6. Steady state, 7. of a few tens to a few hundred degrees based on
One-dimensional, and 8. Constant specific heat, the problem. Composition of the gas mixture
thermal diffusivity and density. Assumptions 1, 2 within this range varies from a pure fuel vapor
and 3 are necessary for implementing the Shvab- near the interface to pure air in the far field.
www.ebook777.com
Within this range, there exists many gas species Ethane and 20 % nitrogen as fuel (all percentages
formed because of thermal cracking of the fuel are in volume) do the following: (1) Plot the
vapor as well as combustion products such as mixture fraction as a function of “x”.
CO2, CO and H2O. Therefore the process of (2) Find the flame location (xF). (3) Plot the
arriving at an “average” mixture property based fuel and oxygen concentrations as a function
on specific mixture composition and average of “x”. (4) Plot the temperature as a function of
temperature is a nontrivial issue. “x”. Solve for two situations of U ¼ 1 mm/s and
This problem of an average gas-composition U ¼ 0.1 mm/s and U ¼ 1 mm/s. Comment on
at an average temperature whose properties can what is the meaning of the “characteristic length
be used for correlations with constant property scale δ” and what is the effect of U on δ, the
assumptions has been investigated by several flame location and the flame temperature. Assume
researchers (c.f. Rangwala et al. [20] for a list thermal properties as those of air at 1000 K.
of references related to the topic). The main
method reported in most combustion textbooks
is by Law and Williams [21] and employs flame, Part 1: Mixture Fraction
ambient (or surrounding) and interface tempera-
x
ture to arrive at an average temperature. The A plot of Z ðxÞ ¼ e1δ 1 is shown in Fig. 11.5.
average mixture composition is calculated using eδ 1
some proportions of fuel and air. The disadvan- The value for delta ( δ ¼ Uα ) based on U of
tage of using the Law and Williams [21] scheme in 1 mm/s equals 0.168 m. It is assumed
fire problems is the need to know the fuel-vapor that the thermal diffusivity is that of air at
composition which is difficult to evaluate for com- 1000 K (α ¼ 168 106 m2 =s).
plex materials usually involved. A much simpler
scheme using only properties of air was recently
developed by Rangwala et al. [20]. The scheme Part 2: Flame Location
has been tested in several diffusion controlled
problems involving burning behavior of both gas- The flame is located at the position where βOF ¼
s Y F ¼ 0: This corresponds to a stoichiomet-
YO
eous liquid and solid fuels. The scheme considers
forced convection and variable oxygen con- ric mixture. The flame location can be found
centration. This scheme is simpler to use and using Equation 11.23. Based on the problem
recommended for fire problems. In this scheme, statement, Y O, i ¼ 0:2232þ0:7828
0:2230
¼ 0:244 and Y F, i
the average thermal conductivity is estimated as ¼ 0:830þ0:228 ¼ 0:811; where, the molecular
0:830
the thermal conductivity of air calculated at a
temperature given as one third the sum of ambient weight of ethane is 30 g/mol. s ¼ YYFo is
stoic
and the adiabatic flame temperature. The gas phase
specific heat is estimated as the specific heat of air
at adiabatic flame temperature. Adiabatic flame
temperatures for several fuels are tabulated in
standard fire dynamics textbooks [22, 23].
Solved Example
On a geometrical configuration identical to that

of the one-dimensional non-viscous problem
presented earlier, a mixture of 22 % oxygen and
78 % nitrogen as oxidizer and a mixture 80 % Fig. 11.5 Mixture fraction (Z) vs. x
360 A.S. Rangwala
Fig. 11.6 Profiles of

oxygen, fuel and
temperature
obtained assuming a one step overall reaction of Profiles of fuel, oxidizer and temperature are
ethane reacting with oxygen C2 H6 þ 72O2 ! evaluated in Part 3 and 4 discussed next.
2CO2 þ 3H 2 O and equals s ¼ 112 30 ¼ 3:73.
Substituting these values in Equation 11.23
gives, Part 3 and 4: Profiles of Oxygen,
Fuel and Temperature
0 1 1
e0:168 1
0:244
First assume that there is no oxygen in the fuel
x f ¼ 0:168 ln@ þ 1A ¼ 0:57:
3:73
0:811 þ 0:244
3:73 zone and no fuel in the oxidizer zone. Use f1 to
find the fuel and oxygen mass fractions.
The location of the flame is shown in Fig. 11.6. It
occurs at a location where the concentration of x < x f Y o ¼ ϕ f 1 ðxÞ and Y f ¼ 0
fuel (YF) and oxidizer (YO) are zero. Note that x > x f Y o ¼ 0 and Y f ¼ f 1 ðxÞ
since the chemical reaction is assumed to be infi-
nitely fast, there is no fuel or oxidizer leakage on Since the oxygen and fuel mass fractions
either side of the flame. Further, the flame location are known, the temperature can be plotted using
occurs at the location of maximum temperature. f2 or f3.
x !
eδ 1 Ti Ti Ti
f 3 ðxÞ ¼ ΨðxÞð f 3 ð1Þ f 3 ð0ÞÞ þ f 3 ð0Þ ¼ 1 Y f,i þ þ
eδ 1 ΔH c =cp ΔHc =cp ΔH c =cp
! !
exδ 1 Ti Ti Ti
T ðxÞ ¼ ΔH c =cp Y f,i þ þ Y f ðxÞ
e 1 ΔH =c ΔH =c ΔH c =cp
1
δ c p c p
The specific heat, cp ¼ 1:14gJ @ 1000K and the The characteristic length scale δ represents
heat of combustion is equal to ΔH c ¼ 47500 gJ . the ratio of the mass diffusion to the velocity
of the gas stream. This characteristic distance
The temperature profile is shown in Fig. 11.6.
www.ebook777.com
represents the extent that heat mass is diffusing

against the gas flow into the gas stream. Mass Burning Rate
00
The mass burning rate per unit area (m_ ) or mass
burning flux is an important parameter to quantify
Final Note the hazard associated with fire. As mentioned
earlier, this mass flux varies with B-number [3],
The example discussed uses constant surface which accounts for the thermodynamic effects,
boundary conditions at fuel and oxidizer side. For and the Nusselt number to account for the
the case of condensed fuels at the interface between gas-phase convective effects. A general relation-
fuel and air, specifying a constant mass fraction for ship will be derived in this section to account for a
the species as the boundary condition at the inter- variety of physical flow conditions. In fire
face provides great simplicity in formulating an problems, the fuel is usually a solid or liquid
analytical solution. However, determining an burning in free convective conditions. The fuel
approximate value of YF,x ¼ surface, the fuel mass vaporizes first, then diffuses toward the oxygen
fraction at the interface, is complicated as it from the ambient. To solve such problems one
involves solving a heat and mass balance at the needs to have information on the velocity, temper-
fuel surface [24]. Further complication is that typi- ature and mass fraction profiles. The solution can
cally the species concentrations are discontinuous be initiated by writing down a heat balance at the
at the interface between two materials, whereas surface of the burning fuel,
temperature is continuous. To take an example,
00 00
consider a heptane pool fire. If we are interested m_ L ¼ q_ ; ð11:26Þ
in determining the rate at which heptane vapor is
transferred to the gas-phase, we would need to where L represents the latent heat of vaporization
specify the vapor concentration of heptane in the (for a liquid) or the heat of gasification (for a
00
gas-phase side of the heptane-air interface. The solid). q_ is the net heat flux to the surface and
mass fraction of heptane inside the pool is unity depends on the nature of the flow field and
(neglecting the small amount of oxygen or nitrogen boundary conditions (free stream temperature,
dissolved in heptane). However, it would be incor- ambient oxygen concentration etc.). It also
rect to assume YF,x ¼ surface ¼ 1 for the gas-phase 00
depends on m_ , an effect called as the “blocking
mass fraction of heptane vapor at the interface. effect [22].” If the net heat flux to the surface
This value primarily depends on the interface tem- increases, then logically mass flux or mass loss
perature besides the pressure. Interface tempera- rate per unit area will also increase. This causes
ture is determined by the heat balance analysis the boundary layer to thicken causing reduction
between gas-phase to interface. The conditions at in gradients. The burning rate therefore does not
the interface are based on relationships that are increase linearly with external heat flux and to
theory-based or deduced from experiments. An develop a general solution requires generaliza-
additional nondimensional number called as tion of the heat flux to the fuel. It can be assumed
B-number arises during the solution of establishing that the heat flux to the surface is a product of the
the mass fraction of fuel vapor at the interface. The heat transfer coefficient (h) and temperature dif-
B-number is an important derived property, which ference between the flame and the surface. If the
is dependent on the thermo-physical properties of flame temperature is denoted by Tf and surface
the system, that provides an expression for the temperature by Ts the mass loss rate per unit area
mass burning rate eventually. This leads us to the can be represented as:
next section which discusses an important parame-
ter usually used to quantifying the fire-hazard 00
00 q_ h T f Ts h
associated with a given material—its mass burning m_ ¼ ¼ ¼B ; ð11:27Þ
rate as a function of time. L L cp
362 A.S. Rangwala
Table 11.3 B-number values for different fuels [25]

Solids Formula B-number Liquids Formula B-number
Polypropylene C3H6 1.29 Methanol CH3OH 2.53
Polyethylene C2H4 1.16 Ethanol C2H5OH 2.89
Polystyrene C8H8 1.55 Propanol C3H7OH 3.29
Nylon 6/6 C12H22N2O2 1.27 Butanol C4H8OH 3.35
Polycarbonate C6H14O3 1.41 n-pentane C5H12 7.63
PMMA C5H8O2 1.78 n-hexane C6H14 6.67
PVC C2H3Cl 1.15 n-heptane C7H16 5.92
Fir wood C4.8H8O4 1.75 n-octane (gasoline) C8H18 5.42
α-cellulose C6H10O5 6.96 iso octane C8H18 6.59
Polyoxymethylene CH2O 1.47 n-nonane C9H20 4.89
n-decane C10H22 4.61
n-undecane C11H24 4.43
n-dodecane (kerosene) C12H26 4.13
Acetone C3H2O 7.28
cp ðTf Ts Þ blocking effect and then corrected for this effect
where B ¼ L is nondimensional and
through the ratio Nu/Nuo
represents the mass transfer number. L in
Equation 11.27 denotes the latent heat of gasifica-
00 λ Nu
tion, h represents the convective heat transfer m_ ¼ B Nuo ð11:29Þ
xcp Nuo
coefficient, Tf and Ts denote an average flame tem-
perature and temperature of the fuel surface and cp
Nondimensionalizing both sides by ρU 1 ,
denotes the specific heat of the gas. The B-number
where U 1 is a characteristic velocity repre-
(also called the Spalding mass transfer number)
senting the flow-field subjected to or induced by
was first introduced by Spalding [3] in 1950 to
the diffusion flame, the above expression can be
characterize liquid fuel droplet burning and physi-
re-written as:
cally relates the heat release related to combustion
(the numerator) to the losses associated with 00
m_ λ μ Nu
combustion (the denominator). The heat transfer ¼B Nuo ð11:30Þ
coefficient (h) is expressed in terms of a nondimen- ρU 1 cp μ ρU1 x Nuo
sional Nusselt number, defined as Nu ¼ hxλ , where
Nu
λ denotes the thermal conductivity of the gas- ¼ BðPr ReÞ1 Nuo ð11:31Þ
Nuo
mixture at the interface between the condensed
fuel and air and x is a characteristic length scale.
The quantity cpλμ equals 1/Pr where Pr ¼ Prandtl
In case of a flat fuel surface x can be defined as the
Nu
distance from the leading edge and in the case of a number. The ratio Nu o
, the blocking effect can be
sphere or cylinder refers to the diameter. The ear- calculated for several laminar flows and equals
lier expression is now equal to: lnð1þBÞ
B for small B. For turbulent flows it can be
calculated empirically only. Typical values of
00 λ B for many fuels are listed in Table 11.3.
m_ ¼ B Nu ð11:28Þ
xcp Expressions for Nuo for many geometrical
configurations are available in standard heat and
In order to account for the blocking effect, the mass transfer textbooks and shown in Table 11.4.
Nusselt number is first evaluated without the For example, using Table 11.4, an expression for
www.ebook777.com
gX3 ðT f T o Þ
Table 11.4 Nuo values for some standard geometries and flow conditions. Re ¼ U1υ X, and Gr ¼ T o υ2
Flow field type Illustration Nuo

Free convective flow
Vertical plate (laminar, Gr < 109) 0.59 (GrPr)1/4
Vertical plate (turbulent, Gr > 109)
Horizontal plate burning face up (laminar, Gr < 107) 0.54 (GrPr)1/4
Horizontal plate burning face up (turbulent, Gr > 107) 0.14(GrPr)1/3
Horizontal plate burning face down (turbulent) 0.27(GrPr)1/4
Horizontal cylinder 0.525(GrPr)1/4
Sphere 2 + 0.6 Gr1/4Pr1/3
Forced flow
Horizontal flat plate (laminar) 0.332Re1/2 Pr1/3
Horizontal flat plate (turbulent Re > 105) 0.036Re0.8 Pr0.3
Pool fire (laminar and axisymmetric) 0.11 Re1/2 Pr2/3
(continued)
364 A.S. Rangwala

Flow field type Illustration Nuo
Droplet (laminar) 0.37Red0.6
Cylinder (laminar) 0.891Re1/2
Cylinder (turbulent, Re > 40,000) 0.27 (GrPr)1/4
mass loss rate of a vertical plate in a laminar free (condensed phase). Typical regression rates of
convective flow can be calculated as: most fuels (solids/liquids shown in Table 11.3)
vary between 0.02 and 0.4 mm/s. The regression
00 λ rates are very small and are essentially due to the
m_ ¼ Co Bm ðGr PrÞ1=4 ð11:32Þ
cp x diffusion controlled nature of the problem.
A mass burning rate for the same plate in a

turbulent natural convective flow-field is:
Diffusion Flame Height
00 λ When gaseous fuel issues out of a tube into

m_ ¼ Co Bm ðGr PrÞ1=3 ð11:33Þ ambient atmosphere of air and the gas is ignited
cp x
a flame is established as shown in Fig. 11.7. The
question we will try to answer is how high is the
The constant Co can take different values and
resulting diffusion flame? The problem can be
usually lies between 0.5 and 1.5. The value of the
solved using conservation equations with certain
exponent m on B is 0.5. Recently Ali
approximations as was first shown by Burke and
et al. [26] have extended the correlation (laminar
Schuman [19]. However, for practical purposes a
case) to include several orientation angles. The
simple physical reasoning exercise will be
mass burning rate of a plate oriented at an angle θ
adopted here. It will be shown later that the
with respect to the vertical for a plate in a free
relationship obtained is similar to that obtained
convective flow-field is given by [26]:
using the more rigorous approach.
Imagine a fuel molecule, shown in Fig. 11.7
00 lnð1 þ BÞ
m_ ¼ 0:737ðGr x Prx Þ0:25 ; ð11:34Þ by the red circle, initially located at the center of
B0:15
the burner tube. The molecule can traverse two
extreme paths (depicted by o-x and o-y) to meet
where Gr x ¼ geff βΔTx
3
υ2 and geff ¼ g cos θ for 0 <

with oxygen at the flame surface. The time taken
θ < 90 and geff ¼ gðcos θ sin θÞ for 90 <

to traverse horizontally along o-x is given by
θ 0 . Note that 0 represents the vertical d2/4D, where D is the diffusion coefficient
case, 90 denotes a pool fire and 90 denotes between methane and air. The relationship is
a ceiling configuration. The mass loss rate per obtained by a dimensional analysis of the two
00
unit area m_ can also be expressed as a regression parameters: distance over which diffusion must
00
rate m_ /ρs where ρs is the density of the fuel occur (m) and diffusion coefficient (m2/s).
www.ebook777.com
Table 11.5 Flame height correlations (all quantities in

SI m, m3/s etc.)
Geometry Flame height (Hf)
0:67
Circular QF ðT 1 =T F Þ T 1
H f ¼ 4πDln
ð1þ1sÞ T f
0:67
Square QF ðT 1 =T F Þ
Hf ¼ T1
16D½inverf ð1þsÞ Tf
0:5 2
2 0:33
Slot bβ2 QF Tf
H fm ¼ hIDY Fstoic
T1
TF T1
h 4 4 4 i1=3 h i2=9
QF T 1 Tf
H f B ¼ 9β
8D2 ah4 T 4 T1
F
β ¼ 4inverf1 1þ1
ð sÞ
QF volumetric fuel flow rate (m3/s), D diffusivity (m2/s),
T 1 ambient temperature (K), Tf mean flame temperature
(K), TF fuel temperature (K), s molar stoichiometric
oxidizer-fuel ratio, inverf inverse error function, ω ¼
inverf[erf(ω)],
a ¼ mean buoyant acceleration
Tf
0:6g T1 1 (m2/s)
developed parabolic exit velocity profile. The

subscript “m” denotes that the flow-field is
Fig. 11.7 The diffusion jet-flame momentum controlled and subscript “b” denotes
buoyancy controlled. To determine if the flame is
The only way of obtaining a time scale from momentum or buoyancy controlled the flame
these two quantities is to divide the square of Froude number Fr must be calculated as
the distance by the diffusion coefficient. The
length of the flame will correspond to the condi- ðV e IY F, stoic Þ2
Fr f ¼ : ð11:35Þ
tion that a point on the stream axis where com- aH f
bustion is complete, the average depth of
penetration of air into gas must equal to the If Fr f 1 then flame is momentum con-
radius of the burner tube. trolled and if Fr f
1 then the flame is buoyancy
Similarly, the time taken to traverse vertically controlled. If Fr f 1, then the flame lies in a
along O-Y is given by Hf /V, where Hf is the flame transitional zone which is both momentum and
height and V is the velocity of the gas issuing of buoyancy controlled. In this case,
the burner. Equating the two times gives, Hf /V 8 9
d2/4D or Hf d2V/4D. This simple result is 3 <" 3 #2=3 =
4 H fB H fB
reflected in all the correlations (developed by H fT ¼ H fm 1 þ 3:38 1 ;
9 H fm : H fm ;
Roper [27, 28]) related to diffusion flame height
shown in Table 11.5. An important element in the ð11:36Þ
expression developed is the influence of fuel
properties, geometry (of the duct). The influence where H f m and H f B represent the momentum and
of fuel properties is usually incorporated in the buoyancy controlled flame heights and can be
flame temperature and stoichiometric coefficient obtained using the correlation presented in
(s) shown in Table 11.5. Table 11.3. Note that for a circular and square
The inverse error function is generated in the port geometry the equations are applicable for
same way as inverse trigonometric functions and both momentum and buoyancy controlled.
is tabulated in standard textbooks (For example The equations shown in Table 11.5 demonstrate
see Table 9.4 in Turns [29]). The parameter, that some of the primary controlling parameters
I typically takes values between 1 and 1.5. influencing the flame height of a diffusion flame
I ¼ 1 for uniform flow and I ¼ 1.5 for fully are flow rate, geometry, and stoichiometry. The
366 A.S. Rangwala
flow rate and geometry influence can be

characterized by the nondimensional Froude num- Turbulent Diffusion Flames
ber. The molar stoichiometric ratio (s) is mainly a
function of fuel type. For a generic hydrocarbon Most fires can be classified as turbulent diffusion
CxHy burning in air the chemical balance equation flames, and extensive studies related to the basic
assuming complete combustion can be written as: understanding of the problem can be found in
literature [31–38]. Besides the important aspects
of height of a diffusion flame and rate of burning,
Cx H y þ ‵ a ðO2 þ 3:76N 2 Þ
which were discussed in the previous sections, an
y
! xCO2 þ H2 O þ 3:76aN 2 ; ð11:37Þ important fire safety concern with the advent of
2 turbulence is the enhancement in the transport
moles air processes, changes in the flame structure, extent
ðxþ4yÞ
where, a ¼ x þ 4y and s ¼ moles fuel stoic ¼ XO2 . and shape, and the radiant energy flux emitted by
XO2 is the mole fraction of oxygen in air and the flame [39, 40]. In large fires, a solid fuel is
equals 1/4.76. As the carbon content in the fuel normally gasified by radiative heat flux from the
increases (y/x increases) the flame length fire (and hot zones) and the evolved gases mix
increases. For example a propane flame (C3H8) with surrounding air. The mixing process is
is nearly 2.5 times longer than a methane flame determined by buoyancy induced flow rather
(CH4). Fuels which have oxygen in their mole- than by the forced convection as in jets and
cule (e.g. CO) show an even smaller flame length. sprays usually studied in the field of combustion
If the fuel is diluted with an inert gas, the flame in engines and furnaces, for example. The advent
height reduces. The molar stoichiometric ratio in of turbulence, which is a random process,
necessitates the use of statistical methods, thus
this case, is s ¼ xþy=4
; where Xdil is the mole
1
XO2 making the solution probabilistic in nature. Thus
1Xdil
an exact solution is seldom possible. Given
fraction of the inert in the fuel stream.
our current state of understanding a solution
The discussion so far pertains to gaseous jet
that falls in a reasonable range is acceptable.
flames. Roper’s correlations can be extended to
The coupling function described as the mixture
predicting flame height of condensed fuels as
fraction (Equation 11.25) plays a central role in
well. For example, an approximate fuel flow
reducing problems in the analysis of turbulent
velocity in a candle flame can be obtained by
diffusion flames to problems associated with
assuming the candle wick to be a vertical cylin-
non-reacting turbulent flows.
der. Using Roper’s correlation for a circular
The turbulent flow field can be visualized as
burner (Table 11.3) and substituting QF ¼ mρ_ F being comprised of eddies having wide range of
F
and using heat transfer coefficients for a finite length scales, the larger of which extract energy
vertical cylinder (represents candle wick) and a from the mean flow. The large eddies break up
hot surface facing up (which represents the top into smaller eddies, which in turn break up into
face of candle wick) it can be shown that [30] even smaller ones, until the smallest eddies,
called Kolmogorov eddies form, and dissipate
n the energy back to the main flow as viscous
L 1=4
H f 0:5D Ra 2 ; ð11:38Þ dissipation. This concept was first proposed
D
by Kolmogorov is described as the eddy/energy
where, D equals the diameter of the wick, L is equal cascade hypothesis in turbulent literature [33].
to the length of the wick from the base, Ra is the Note that the size of the smallest eddies are
Rayleigh number defined using the wick diameter, several orders of magnitude larger than the
and n is a constant approximately equal to 0.7. mean free path of the molecules of the gas or
Sunderland et al. [30] have shown the above rela- fluid. Due to this reason, the turbulent flow field
tionship to provide a good match with experimen- can be considered as continuum and the conser-
tal data using candle wicks of different thicknesses. vation equations of mass (species), momentum
www.ebook777.com
Fig. 11.8 Illustrative

sketch of laminar flamelet
concept of turbulent
diffusion flames. The
turbulent intensity
increases from (a) to (d).
Local extinction is
observed in (c) due to flame
stretching
and energy are still applicable. The question then diffusive layers, considered in general as laminar,
arises that how does this flow pattern influence a embedded within an otherwise non-reacting tur-
diffusion flame? For this, the general theory pro- bulent flow field. This allows us to use the mixture
posed by Williams [41] and models by Peters fraction to describe turbulent flames as well.
[38] and Kuznetsov [42] will be discussed. The If we assume that the turbulent transport of
general model is named as the flamelet model momentum, species and thermal energy are all
mainly because of the tendency of turbulence to equal (Sc ¼ Pr ¼ Le) we can substitute existing
break the flame into several tiny flames or correlations from the fluid mechanics literature
flamelets. Both strain and shear at the interface related to turbulent momentum diffusivity (eddy
of eddies increase the concentration gradient viscosity) to apply for turbulent mass diffusivity
between the reactants thereby enhancing the and thermal diffusivity as well.
mixing. Going back to Equation 11.1 (Ficks Figure 11.8 shows the influence of increased
∂Y turbulent intensity on a turbulent flame. As we
law) as the gradient ∂x i
, increases, the rate of
mixing should also increase eventually. How- move from Fig. 11.8a to d, eddies are capable of
ever, this increase is not due to a change in the corrugating a flame and eventually breaking the
molecular diffusivity DAB. Thus, in order to corrugations and distributing the flame sheet into
model this increase in the overall mixing process, smaller flamelets. Thus at sufficiently high strain,
an improvised turbulent diffusivity is defined. there is also local extinction achieved. Another
Note that the turbulent diffusivity is an empirical aspect, and most important from an aspect of fire
concept and incorporates both the molecular dif- safety is an understanding of flame radiation due to
fusivity as well as some function which can repre- large turbulent hydrocarbon diffusion flames. For
sent the influence of intensity and scale of example flame radiation is an important parameter
turbulence on the mixing at the flamelet level. in design of safe separation distances. The primary
The flamelets are considered as thin reactive- factors influencing the problem are the
368 A.S. Rangwala
Fig. 11.9 Spectral measurements of thermal radiation from a turbulent (a) ethylene-air flame [43] and (b) methane-air
flame [44]
concentration of soot, CO2, H2O, temperature μm corresponding to the molecular radiation

profiles (at different locations in the flame) and from CO2 and H2O.
geometry of the problem. Incandescent soot Prediction of flame radiation necessitates
formed within the flame is the primary source of knowledge of formation and oxidation of soot
flame radiation in large-scale fires, although gas- in the flame, its concentration profiles, as well
eous products of combustion (CO2, H2O) can also as concentration and temperature of sources of
emit heat. The radiation emitted from CO2 and molecular radiation such as CO2 and H2O,
H2O is usually called molecular radiation. The besides the type of fuel and the residence time
easiest way to understand (and explore) the differ- available. A common approach used in fire
ence between the radiation from soot and gaseous problems, is identifying a radiant fraction which
products (CO2, H2O) is by comparing two diffu- represents the ratio of radiant heat transfer rate
sion flames. from the flame to the surroundings to the total
Figure 11.9a [43] and Fig. 11.9b [44] depict heat released by the flame. The radiant fraction is
spectral radiation intensities for an ethylene-air denoted as χR, and is expressed as:
and methane-air diffusion flame respectively.
The ethylene-air flame produces more soot Q_ rad
when compared with the methane-air flame. χR ¼ ; ð11:39Þ
m_ F ΔH c
Correspondingly, it exhibits intensity peaks at
1.5 μm corresponding to the in-flame soot.
In comparison, the methane-air flame shows where Q_ rad denotes the radiant energy transferred
significant intensity peaks at 2.5–3 μm and 4–5 from the flame in the form of radiation. Based on
www.ebook777.com
our current understanding and knowledge-base, of a wall fire. Tc which is a characteristic tempera-
χR is estimated using empirical correlations. χR ture can be represented as ΔHc/cp. Substituting
depends on both flame size and heat release rate the nondimensional variables in the energy
and the general trend is for χR to increase with equation gives,
flame size. Depending on the fuel type and flow
!
conditions, the radiant fraction can range from a ∂Tb b
∂T AY on j Y fj eE=RT
few percent to more than 50 % [45–47]. ρc p u1 b
u þb
v ¼ ΔHc
∂bx ∂by s
!
ΔHc ∂2 Tb ∂2 Tb
þλ 2 þ :
Flame Extinction L c p ∂b
x 2 ∂by 2
ð11:40Þ
General strategies for extinguishing diffusion
000
flames are cooling, reactant removal, chemical ω_ ΔH
c
The term O
s , representing the energy
inhibition, and flame removal. These strategies
are achieved for example, by adding sufficiently released due to the chemical reaction between
large quantities of a material (such as water) that fuel and oxidizer is expressed using an empirical
AY on j Y j eE=RT
cools the condensed fuel thereby slowing com- Arhenius expression ΔH c s
f
, or
bustion, removing an essential reactant from the specifically, the rate at which fuel is lost per
system, adding a chemical suppressant (such as unit volume is given by AY n j j E=RT
: In
o Y fe
Halon) that inhibits the reactions taking place in
the expression, A is a constant called as
the flame zone, or physically removing the flame
the pre-exponential constant or the frequency
from the reactant mixture by inducing high gas j j
velocities (for example by explosives in large factor. The terms, Y no Y f represent the mass
production-well fires). fraction of oxidizer and fuel (n and j are empiri-
The rate of gas-phase chemical reaction that cal constants) to emphasize the fact that the
000
was denoted by ω_ i in Equations 11.4, 11.5, and rate of loss of fuel is dependent on the initial
11.6 was thus far ignored. However, it plays an concentration of fuel and oxidizer. To simplify
important role in flame extinction (as well as the mathematical analysis and gain physical
ignition). It is therefore necessary to formulate insight into the problem, we will assume a sim-
000
an expression to quantify ω_ i in order to establish pler expression of the form KeE=RT , where K is a
critical extinction criteria. This is usually constant. Note that so far we have been able to
done by employing a one-step Arhenius neglect this term by assuming the infinitely fast
approximations to the rates. The analysis can be reaction assumption and the resulting Shvab-
generalized using a non-dimensional Damkohler Zeldovich transformation. We therefore never
number [48]. To better understand the concept, worried about defining this term. However, for
let us perform an exercise of nondimensio- both ignition and extinction analysis, the nonlin-
nalizing the energy conservation equation ear source term cannot be ignored. Rearranging
(gas-phase) using nondimensional variables: terms gives
_
u ¼ u=uc , b v ¼ v=uc , bx ¼ x=Lc , b y ¼ y=Lc , !
b
∂T b
∂T KeE=RT Lcp
b
T ¼ T=T c , where uc ¼ some characteristic veloc- b
u þb
v
¼ ΔH c
∂b
x ∂b
y s ρcp u1 ΔH c
ity, which we will denote as u1 and Lc ¼ L is some !
characteristic length scale. The definition of u1 ΔH c Lcp b ∂2 T
∂2 T b
and L, depends on the nature of the problem. For þλ 2 þ 2
L cp ΔH c ρcp u1 ∂b
x 2 ∂b y
example, if the flowfield is natural convective
mostly (large open pool fires, wall fire etc.) u1 ð11:41Þ
pffiffiffiffiffiffi
gL where L is for example the diameter of
the pool fire or the length of the wall in the case or
370 A.S. Rangwala
!
b
∂T ∂Tb eE=RT L=u1 laminar flamelets are distorted to an extent such
b
u þbv ¼ that the fuel vapor does not have sufficient time to
∂b
x ∂by s ρ=K
! react with the oxygen. This results in further dis-

L=u1 ∂2 T b ∂2 T b tortion of the corrugated flame illustratively shown
þ 2 þ 2 ð11:42Þ in the transition from Fig. 11.8c, d. τchemical is the
L =α ∂bx 2 ∂by
chemical reaction time, i.e. the time taken for the
chemical reaction of combustion to occur. τchemical
The rearrangement of the terms as shown in can be reduced by influencing the chemical reac-
Equation 11.42 provides some interesting nondi- tion taking place at the flame. For example chemi-
mensional numbers. Firstly, we can define some cal suppressants such as Halon are effective reduce
time scales such that: the rate of gas phase reaction.
If the conduction losses are increased (denomi-
τresidence ¼ uL1 ¼ Time the reactants are present in
nator of Da2) extinction can be achieved
the CV by quenching due to heat losses (in-depth conduc-
τchemical ¼ Kρ ¼ Time it take s for a chemical tion, re-radiation, and radiative feedback from the
reaction to occur. flame). For example, if the fuel supply can be
decreased by say lowering the vaporization rate of
2 ρcp
τconduction ¼ Lα , α¼ λ ¼ Pre-heat time a condensed fuel by adding water, the flame
approaches the surface and thus looses more heat
We can now express, the terms on the right
hand side of Equation 11.42, in square brackets which may ultimately result in quenching due to
as two nondimensional numbers which are increase in heat losses. The quenching due to lack of
fuel supply has not been studied extensively with
denoted as:
condensed fuels and works by Roberts and Quince
τresidence L=u1 [50] and Torero et al. [51] represent the few studies
Da1 ¼ ¼ ; ð11:43Þ that define a minimum fuel supply stability limit
τchemical ρ=K
for diffusion flames established over a
and condensed fuel. Torero et al. [51] show that for
diffusion flames spreading on condensed fuel
τresidence L=u1 surfaces, such as upward spread along walls, or
Da2 ¼ ¼ 2 : ð11:44Þ
τconduction L =α horizontal spread or lateral spread on floors, the
mechanism of gas-phase flame quenching is depen-
“Da” stands for Damkohler number, after dent on blow-out mechanisms (reduction of Da1)
Damkohler [49]. As Da1 (or Da2) decrease, sup- for the leading edge (location of flame anchored to
pression is obtained. This is explained using the the surface) and quenching (reduction of Da2) for
residence and chemical times defined. τresidence is the trailing edge (location of the flame tip).
the residence time which refers to the length of Extinction of diffusion flames can also be
time the fuel vapors remain in the reaction zone. physically explained using a critical flame tem-
The residence time will depend on the fluid perature concept. The “S-curve” first studied by
dynamics of the flame. A diffusion flame can be Fendell [52] and Linan [53] is usually used to
extinguished by the mechanism of blowout, famil- represent the response of temperature to the sys-
iar with the small flames of matches and candles. tem Damkohler number. For most fire problems
The mechanism involves the distortion of the reac- the critical flame temperature at which a diffu-
tion zone, within the flame in such a way as to sion flame ceases to exist is approximately
reduce the time that the fuel vapors have to react. 1300 C [54, 55]. However, this value should
Blowout can occur if sufficient airflow can be be used with caution. A recent study by Tien
achieved to reduce τresidence and consequently Tf, and Endo [56] show that there is no unique criti-
thus reducing the Da1 number below a critical cal extinction temperature for different materials
value. This behavior is also observed in turbulent and the flame temperature at extinction is a
flames where with sufficiently high intensity, the strong function of flow parameters.
www.ebook777.com
non-premixed hydrocarbon flames. Fire Safety Jour-

Final Comments and Conclusions nal, 2005. 40(2): p. 141–176.
9. Vander Wal, R.L. and Tomasek, A.J., Soot oxidation:
Dependence upon initial nanostructure. Combustion
This chapter discussed the problem of diffusion
and flame, 2003. 134(1): p. 1–9.
flames also called as non-premixed flames. The 10. Rangwala, A.S. Flame spread analysis using a vari-
discussion illustrates two important aspects. Anal- able b-number. in Proc. Fire Safety Sci. 2008.
ysis of diffusion flames involves knowledge of Karlsruhe, Germany.
11. Kent, J. and Wagner, H.G., Who do diffusion flames
transport properties such as diffusivity, specific
emit smoke. Combustion Science and Technology,
heat of species involved in the process and a thor- 1984. 41(5–6): p. 245–269.
ough understanding of the flow system. Both are 12. De Ris, J. and Cheng, X.F., The role of smoke point in
limited by the postulates that are made in setting material flammability testing. Proc. Fire Safety Sci.,
1994. 4: p. 301–312.
up the simple examples discussed in this chapter.
13. Markstein, G.H. Correlations for smoke points and
It is therefore absolutely necessary to do radiant emission of laminar hydrocarbon diffusion
experiments in order to establish the flow regime flames. in Symposium (International) on Combustion.
and quantify material and gas-phase properties. 1989. Elsevier.
14. Delichatsios, M., Smoke yields from turbulent buoyant
The inherent non-linear nature of the governing
jet flames. Fire Safety Journal, 1993. 20(4): p. 299–311.
equations makes the subject difficult and solution 15. Lautenberger, C.W., Cfd simulation of soot formation
to simplified geometries is only possible at this and flame radiation, 2002, WORCESTER POLY-
stage. Other aspects such as onset of instabilities TECHNIC INSTITUTE.
16. Zeldovich, I., Barenblatt, G.I., Librovich, V., and
and turbulent transport are not clearly understood.
Makhviladze, G., Mathematical theory of combustion
This leaves the topic with several fascinating areas and explosions. Moscow: Nauka, 1985.
of research that need further work. 17. Shvab, V.A., Goz. Energ. izd. Moscow-Leningrad,
1948.
18. Zel’dovich, Y.B., Zhur. Tekhn. Fiz. English transla-
tion: NACA Tech. Memo No. 1296 (1950), 1949. 19:
References p. 1199.
19. Burke, S.P.a.S., T. E. W.,, Diffusion flames. Industrial
1. Arias-Zugasti, M. and Rosner, D.E., Soret transport, & Engineering Chemistry, 1928. 20(10): p. 998–1004.
unequal diffusivity, and dilution effects on laminar 20. Rangwala, A.S., Raghavan, V., Sipe, J.E., and Okano,
diffusion flame temperatures and positions. Combus- T., A new property evaluation scheme for mass trans-
tion and flame, 2008. 153(1): p. 33–44. fer analysis in fire problems. Fire Safety Journal,
2. Smyth, K.C., Miller, J.H., Dorfman, R.C., Mallard, 2009. 44(4): p. 652–658.
W.G., and Santoro, R.J., Soot inception in a methane/ 21. Law, C.K. and Williams, F.A., Kinetics and convec-
air diffusion flame as characterized by detailed spe- tion in the combustion of alkane droplets. Comb.
cies profiles. Combustion and flame, 1985. 62(2): Flame, 1972. 19: p. 393–405.
p. 157–181. 22. Quintiere, J.G., Fundamentals of fire phenomena.
3. Spalding, D.B., Combustion of liquid fuel in gas 2006: John Wiley & Sons, New York.
stream. Fuel, 1950. 29: p. 2–7. 23. Drysdale, D., An introduction to fire dynamics. 1998:
4. Spalding, D.B., Some fundamentals of combustion. John Wiley & Sons, New York.
Gas turbine series vol. 2. 1955, London: Academic 24. Raghavan, V., Rangwala, A., and Torero, J., Laminar
Press, Butterworths Scientific Publications. flame propagation on a horizontal fuel surface: Veri-
5. Chakrabarty, R.K., Moosmüller, H., Arnott, W.P., fication of classical emmons solution. Combustion
Garro, M.A., Tian, G., Slowik, J.G., Cross, E.S., Theory and Modelling, 2009. 13(1): p. 121–141.
Han, J.-H., Davidovits, P., and Onasch, T.B., Low 25. Annamalai, K. and Sibulkin, M., Flame spread over
fractal dimension cluster-dilute soot aggregates from combustible surfaces for laminar flow systems part i:
a premixed flame. Physical review letters, 2009. Excess fuel and heat flux. Combustion Science and
102(23): p. 235504. Technology, 1979. 19(5): p. 167–183.
6. Frenklach, M. and Wang, H. Detailed modeling of 26. Ali, S.M., Raghavan, V., and Rangwala, A.S., A
soot particle nucleation and growth. in Symposium numerical study of quasi-steady burning
(International) on Combustion. 1991. Elsevier. characteristics of a condensed fuel: Effect of angular
7. Sunderland, P. and Faeth, G., Soot formation in hydro- orientation of fuel surface. Combustion Theory and
carbon/air laminar jet diffusion flames. Combustion Modelling, 2010. 14(4): p. 495–518.
and flame, 1996. 105(1): p. 132–146. 27. Roper, F., The prediction of laminar jet diffusion
8. Lautenberger, C.W., De Ris, J.L., Dembsey, N.A., flame sizes: Part i. Theoretical model. Combustion
Barnett, J.R., and Baum, H.R., A simplified model and flame, 1977. 29: p. 219–226.
for soot formation and oxidation in cfd simulation of
372 A.S. Rangwala
28. Roper, F., Laminar diffusion flame sizes for curved 46. Orloff, L., De Ris, J., and Delichatsios, M., Radiation
slot burners giving fan-shaped flames. Combustion from buoyant turbulent diffusion flames.
and flame, 1978. 31: p. 251–258. Combustion Science and Technology, 1992. 84
29. Turns, S.R., An introduction to combustion, 2000, (1–6): p. 177–186.
McGraw-Hill, New York. 47. Delichatsios, M., De Ris, J., and Orloff, L. An
30. Sunderland, P., Quintiere, J., Tabaka, G., Lian, D., enhanced flame radiation burner. in Symposium
and Chiu, C.-W., Analysis and measurement of candle (International) on Combustion. 1992. Elsevier.
flame shapes. Proceedings of the Combustion Insti- 48. Williams, F., A review of flame extinction. Fire Safety
tute, 2011. 33(2): p. 2489–2496. Journal, 1981. 3(3): p. 163–175.
31. Bilger, R., Turbulent diffusion flames. Annual Review 49. Damköhler, G., The effect of turbulence on the flame
of Fluid Mechanics, 1989. 21(1): p. 101–135. velocity in gas mixtures. 1947.
32. Bilger, R. The structure of turbulent nonpremixed 50. Roberts, A. and Wuince, B., A limiting condition for
flames. in Symposium (International) on Combustion. the burning of flammable liquids. Combustion and
1989. Elsevier. flame, 1973. 20(2): p. 245–251.
33. Peters, N., Turbulent combustion. 2000: Cambridge 51. Torero, J.L., Vietoris, T., Legros, G. And Joulain, P.,,
university press. Estimation of a total mass transfer number from the
34. Libby, P.A. and Williams, F.A., Turbulent reacting standoff distance of a spreading flame. Combustion
flows. Turbulent Reacting Flows, 1980. 1. Science and Technology, 2002. 174(11): p. 187–203.
35. Dahm, W. and Dimotakis, P., Measurements of 52. Fendell, F.E., Ignition and extinction in combustion of
entrainment and mixing in turbulent jets. AIAA jour- initially unmixed reactants. J. Fluid Mech, 1965.
nal, 1987. 25(9): p. 1216–1223. 21(2): p. 281–304.
36. Takeno, T. and Kotani, Y. Transition and structure of 53. Linan, A., The asymptotic structure of counterflow
turbulent jet diffusion flame. in AIAA, Aerospace diffusion flames for large activation energies. Acta
Sciences Meeting. 1977. Astronautica, 1974. 1(7): p. 1007–1039.
37. Pope, S.B., Turbulent flows. 2000: Cambridge univer- 54. Rasbash, D.J., The extinction of fires by water sprays.
sity press. Fire Res. Abst. Rev., 1962. 4(1)(28–53).
38. Peters, N., Laminar diffusion flamelet models in 55. Quintiere, J. and Rangwala, A., A theory for flame
non-premixed turbulent combustion. Progress in extinction based on flame temperature. Fire and
Energy and Combustion Science, 10(3): p. 319–339. Materials, 2004. 28(5): p. 387–402.
39. Tamanini, F. A numerical model for the prediction of 56. T’ien, J.S. and Endo, M., Material flammability:
radiation-controlled turbulent wall fires. in Sympo- A combustion science perspective. Proceedia Eng.,
sium (International) on Combustion. 1979. Elsevier. 2013. 62: p. 120–129.
40. Delichatsios, M. and Orloff, L. Effects of turbulence
on flame radiation from diffusion flames. in Sympo-
sium (International) on Combustion. 1989. Elsevier. Ali S. Rangwala is an associate professor at the depart-
41. Williams, F., Recent advances in theoretical ment of Fire Protection Engineering at Worcester Poly-
descriptions of turbulent diffusion flames. Turbulent technic Institute (WPI) (2006 – present). He has a BS in
mixing in nonreactive and reactive flows, p. 202–208. Electrical Engineering, from the Government College of
42. Kuznetsov, V., The effect of turbulence on the forma- Engineering, Pune, India (2000), an MS in Fire Protection
tion of large superequilibrium concentrations of Engineering from the University of Maryland, College
atoms and free radicals in diffusion flames. Park (2002), and a PhD in Mechanical and Aerospace
Akademiia Nauk SSSR, Izvestiia, Mekhanika Engineering from the University of California, San
Zhidkosti i Gaza, p. 3–9. Diego (2006).
43. Gore, J. and Faeth, G. Structure and spectral radia- Professor Rangwala’s research interests are in the area
tion properties of turbulent ethylene/air diffusion of industrial fire and explosion safety, specifically
flames. in Proc. Combust. Instit. 1988. Elsevier. deflagration of combustible dust clouds, ignition beha-
44. Faeth, G., Gore, J., Chuech, S., and Jeng, S.-M., vior of combustible dust layers, in-situ burning of oil,
Radiation from turbulent diffusion flames. Annual spread of oil slicks, measuring techniques in fire induced
Review of Heat Transfer, 1989. 2(2). flows, and flame propagation and burning rate behavior
45. De Ris, J.a.O., L.,, The role of buoyancy direction and of condensed fuel surfaces. He teaches three graduate
radiation in turbulent diffusion flames on surfaces. courses: Explosion Protection, Industrial Fire Protection,
Fifteenth Symposium (International) on Combustion, and Combustion at Worcester Polytechnic Institute,
1974. 15: p. 175–182. Department of Fire Protection Engineering
www.ebook777.com
Fundamentals of Premixed Flames

12
Grunde Jomaas
Lavoisier, Berthollet and Dalton were pioneers the process. The rate at which the products and
in the understanding of the mixture composition concomitant heat are generated is governed by
needed for a flame existence, and in the early the Arrhenius reaction rate, m_ Y F Y O eE=RT ,
1800s, Sir Humphry Davy created a miner’s which states that the fuel and oxidizer must
lamp with a fine meshed net that improved the have the proper concentrations in order to react
safety for mine workers, as the mesh was finer and the molecules must have sufficient energy
than the quenching distance and hence reduced in order to begin to chemically react. In a
the number of accidental explosions. When premixed flame, the fuel and oxidizer are
Bunsen created the burner associated with his mixed before the heat required to overcome
name in 1855 [1], the premixed flame was con- the activation energy, E, and to initiate the
sidered to be However, as the following will chemical chain reactions is added. The added
show, there was, in the words of Richard heat is often in the form of a spark, increased
Feynman, plenty of room at the bottom [126]. temperature or an open flame. The ratio of the
There are two main types of flames, namely fuel and oxidizer must be within the flammabil-
the non-premixed flame, which is often referred ity limits (see Chap. 17), which have both lower
to as a diffusion flame, and the premixed flame. and upper bounds, and the heat applied must be
The former is the more common type of flame in sufficient to get the chain reactions going, and
fire situations, where the atmospheric oxidizer this is typically referred to as the minimum
and the vaporized (liquids) or pyrolyzed (solids) ignition energy (MIE) (see Chaps. 17 and 69)
fuel and oxidizer are not mixed until they react in if it is not a purely thermal explosion (for
the flame sheet. In the premixed flame, the fuel Auto-Ignition Temperatures, AIT, see
and the oxidizer are mixed before reaching the Chap. 17). Although premixed flames can be
reaction zone, where chemical energy is formed directly or indirectly from solids and
converted to heat and light. This type of flame liquids (e.g. grain elevator explosion), the
is seen in petrol (gasoline) engines, gas turbines, current discourse will focus on gaseous fuels.
some explosions and elsewhere. A premixed flame is a wave phenomenon and
In both flame configurations, chemical it will propagate into the unburned mixture if not
reactions must convert the fuel and oxidizer forced to do otherwise. The wave can either be
into products and they typically release heat in propagated at subsonic speeds by the heat
release from chemical chain reactions (deflagra-
tion) or at supersonic speeds with a leading
G. Jomaas (*)
Department of Civil Engineering, Brovej, DTU Building shock initiating combustion (detonation). This
118, Technical University of Denmark (DTU), DK-2800 contrasts a premixed flame to the non-premixed
Kgs. Lyngby, Denmark

374 G. Jomaas
(diffusion) flame, which will try to situate itself studies of other premixed flame phenomena. In
at the stoichiometric location of the incoming order to computationally simulate a flame phe-
streams of fuel and oxidizer. nomenon such as the laminar flame speed, reliable
It is quite common to refer to a premixed and accurate data on the chemical, thermody-
flame as an explosion because the reactions and namic, and transport properties must be available
the wave-front created by the expansion of the over a wide range of experimental conditions,
gases are occurring very quickly. However, the and as such this presents a major challenge.
word explosion should technically be reserved These modeling obstacles can be circumvented
for the case of a purely thermal explosion, to some degree by optimizing the kinetic
which is a fast rate of reaction in a premixture mechanisms based on experimental observations
without a wave [2–6]. The classic theories by of fundamental properties such as ignition delay
Semenov and Frank-Kamenetskii [7, 8] use dif- times in shock tubes, extinction limits, flame
ferent approaches to show that the chemical gain structures, and laminar flame speeds. The follow-
from exothermic reactions has to exceed the ing will therefore focus on the theoretical, numer-
conduction losses in the system for a thermal ical and experimental work that has enabled the
runaway to occur. current understanding of premixed flames. Most
As the reaction rate is proportional to the combustion textbooks contain chapters on
energy release, and the reaction rate is exponen- Premixed Flames that have discourses ranging
tial with respect to temperature, the result of from more practical applications to advanced
increasing the temperature is typically a thermal mathematical formulations and derivations, and
runaway (explosion). This requires that the num- the reader is encouraged to consult these for fur-
ber of chain branching reactions is larger than the ther information on the topic [10–24].
number of chain terminating reactions. In the
classic example with H2-O2, the detailed chemi-
cal reaction scheme along with the reaction rates Rankine-Hugoniot Relations –
of the reaction steps will reveal the chain Infinitely Thin Flames
initiating reactions (creation of the first free
radicals), the chain branching reactions (more In order to analyze the propagation of a premixed
free radicals in the products than in the flame, the one-dimensional equations for a
reactants), the chain carrying reactions (equal reacting flow have to be discussed. Figure 12.1
number of free radicals in the reactants and the shows the configuration and the relevant
products) and the chain terminating reactions parameters in the unburned (left) and
(fewer free radicals in the products than in the the burned (right) region of the infinitely thin
reactants), and thus reveal the bottle necks of the flame. In this frame of reference, the flame is
explosive process [9]. As such, a detailed study stationary such that the unburned gas velocity
of the chemistry can help us understand how to
prevent such a runaway, for example by
introducing competing radicals that will inhibit
the chain branching, which is the fundamental ρu ρb
operational aspect of several fire extinguishment uu ub
agents, such as halons and their more environ- pu pb
mentally friendly replacements.
Tu Tb
The laminar flame speed is a thermo-chemical
property embodying information about the diffu-
sive and reactive aspects of a combustible mix-
Fig. 12.1 Schematic of the one-dimensional premixed
ture, and is thus a central link to understanding of
flame, with unburned gases (reactants) to the left and
premixed flames. Furthermore, high fidelity lami- burned gases (products) to the right of the infinitely thin
nar flame speed data enables precise and accurate flame
www.ebook777.com
12 Fundamentals of Premixed Flames 375
and density match the mass consumption rate of After some algebraic manipulation of the four
the flame. equations with five unknowns [13, 15, 16, 19,
The one-dimensional governing equations for 24], we can produce the Rankine-Hugoniot [25,
conservation of mass, momentum and energy 26] diagram, see Fig. 12.2, which enables us to
(the energy release, qc, equals zero for a shock plot the solutions of the underdetermined prob-
wave) are: lem. A key point in this plot is the Chapman-
Jouguet point, which is named after Chapman
ρu uu ¼ ρb ub
[27] and Jouguet [28].
pu þ ρu u2u ¼ pb þ ρb u2b The diagram offers numerous solutions, but
1 1 not all the mathematically possible solutions for
hu þ u2u þ q ¼ hb þ u2b
2 2 the burned state are physically possible. The
velocity uu is real and positive in the upper left
The subscripts refer to the unburned (u) and the
region. However, it does not exist (imaginary) in
burned (b) regions, respectively. The total
the upper right region, as it would require a com-
enthalpy, h, equals the sensible plus the chemical
pression wave to move in the negative direction,
enthalpy (see Chap. 5.)
which is impossible. The lower left corner is
h ¼ c p T þ h0 inaccessible because it is impossible to decrease
the pressure and the specific volume simulta-
In addition, we have the equation of state neously. Similarly, the upper right corner is inac-
pu ¼ ρu RT u cessible because it is impossible to increase the
pressure and the specific volume at the same time.
Fig. 12.2 Rankine-Hugoniot plot for the one-dimensional premixed flame. Two lines are shown, one associated with a
shockwave (q ¼ 0) and one for a wave with a given heat release [13]
376 G. Jomaas
The plot is then left with four solution regions, for the underdetermined set of equations, and
where two are for compression waves and two the flame structure provides this information
are for expansion waves [13, 19]. through the introduction of the heat release rate.
Compression waves (high velocity, the
specific volume decreases, the pressure and the
density increase when moving from the initial
Flame Structure and the Flame Speed
state to the final state)
1. Strong detonation, pb > pCJ, supersonic to
The fact that the flame has a structure is essential,
subsonic (pCJ is the pressure at the CJ point)
as this means that it has a finite thickness,
• Rarefaction waves in the post-combustion
which is needed in order to move away from
gases cause this region to be unstable and
the Landau limit (flame sheet, no thickness) in
waves that occur here will always slow
which the flames are intrinsically unstable to all
down and end up with a velocity
wavelengths [29]. The flame thickness therefore
corresponding to that of the upper CJ
introduces stability.
point. Note that there are theoretical
The detailed structure of a premixed flame
solutions for situations with steady,
contains a preheat zone and a reaction zone, see
overdriven detonation waves [127].
Fig. 12.3. The laminar flame speed depends on
2. Weak detonation, pb < pCJ, subsonic to
the chemical kinetics and the thermal and mass
supersonic
diffusion.
• Does not exist (based on structure analysis)
Although many other methods with more
Expansion waves (low velocity, the specific
mathematical rigor and more refined physical
volume increases, the pressure and the density
assumptions [13, 15, 16, 19, 24] exist, the
decrease when moving from the initial state to
considerations of Mallard and Le Chatelier [30]
the final state)
provide a good starting point for the introductory
3. Weak deflagration, pb > pCJ, L (pCJ, L, pres-
discussion herein. Also, the main outcome is the
sure at lower CJ point), subsonic to subsonic
same of all the theories, namely that the laminar
• Deflagration (expansion waves), which
burning velocity and the flame thickness have a
requires the structure to be analyzed
physical dependency on the fuel-air mixture and
4. Strong deflagration, pb < pjb, subsonic to
the ambient conditions. This knowledge can be
supersonic
used for safety design through stabilization and
• Does not exist, as it is impossible to have
control of the flames.
heat addition and proceed past the sonic
The following will introduce more details
condition in a constant area duct (Rayleigh
and results associated with the laminar flame
flow).
speed, which is a steady, self-sustained propa-
Thus, only region 3 and the upper CJ
gation of an ‘exothermic reaction wave’ at a low
point exist.
In summary, a Rankine-Hugoniot plot arises
from the conservation equations for steady, 1-D
flow. The plot shows that combustion waves can Preheat Zone
Reaction Zone
either be deflagration or detonation waves, that uu = su
is, they can, respectively, either be slower or Tb
faster than the speed of sound in the medium Yu

Reaction Rate
they traverse. In order to study the deflagration
solution, a flame structure is needed, which
means that studies of the species concentrations, Tu
the flame thickness, diffusion rates, and the Yb = 0
temperature become relevant. For the deflagra-

tion scenario, one more equation is needed Fig. 12.3 The structure of a premixed flame [13]
www.ebook777.com
subsonic velocity in a direction normal to itself. to the physical problem. The governing equations
For the most part, radiative heat transfer and indicate that convection, diffusion and reaction
second-order (coupled) transport effects, such are the relevant transport and reaction processes
as thermal diffusion, are neglected, and the that should be considered, see Fig. 12.4. The first
focus is placed on heat conduction, diffusion assumption is that there is no reaction in the
and the rapid exothermic chemical reactions. far upstream and the far downstream domains,
Due to the small pressure difference, it can be so the reaction timescales go to infinity there. In
considered isobaric, and the momentum equa- the preheat zone, the reaction timescale is long
tion is not needed. As a result, only the mass due to the low temperature, so convection and
balance (of each species) and the energy balance diffusion dominate, and have to have matching
are considered. timescales for the physical problem to be bal-
As the analysis focuses on conservation of anced. In the reaction zone, the reaction timescale
mass and energy across the flame (1-D, adiabatic, is small due to the high temperature, so the length
steady and planar flame propagation into station- scale has to be small so that the thermal and mass
ary, combustible mixture in a doubly infinite gradients, which govern the rate of diffusion, are
domain), transport and reaction parameters are large. Because convection is a first order differ-
key. The problem therefore calls for two ential and diffusion a second order differential,
parameters, which are the laminar flame velocity we can deduce that diffusion will be dominating
and the flame thickness. over convection for the small length scales pres-
ent in the reaction zone. Consequently, we have
that diffusion and reaction balance each other in
Mallard and Le Chatelier the reaction zone. Finally, at the ignition temper-
ature, all three timescales have to be of approxi-
The basis for the theoretical development by mately the same order.
Mallard and Le Chatelier [30] requires the con- This approach also allows us to obtain the key
cept of an ignition temperature, but an actual parameter for studies on laminar premixed
ignition temperature does not exist in a flame flames, namely the laminar burning velocity,
(and hence cannot be determined). The ignition without getting into discussions on the reaction
temperature in approach is also not to be confused progress variable and details about the reaction
with the adiabatic flame temperature (See rate and the mathematical challenges it causes
Chaps. 5, 17, and 18). Still, the matching of the through its non-linearity.
timescales and the length scales in a preheat The assumptions in this approach are that the
zone and a reaction zone provides useful insight gases are heated by conduction in the preheat
1/τ
Far Upstream:
No reaction due Matching timescales at T=Ti Far Downstream:
to lack of No reaction due to lack
Reaction zone
temperature of fuel/oxidizer
Preheat zone Diffusion +
Convection + Reaction
Diffusion
Reaction Convection
Fig. 12.4 Overview of the laminar premixed flame problem. See also Figs. 12.1 and 12.3
378 G. Jomaas
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi 1=2
zone and reach ignition at the ignition boundary,
λ=c p T b T i ω_ ω_
that the chemical enthalpy is converted into SL ¼ α
ρ Ti Tu ρ ρ
sensible enthalpy in the chemical reaction
zone, and that the temperature curve is linear. The theoretical results for the laminar flame speed
The enthalpy balance gives have been compared with experimental results for
qpreheat ¼ qconduction more than 100 years, with varying success, preci-
sion and accuracy. In the following, a series of
Tb Ti important experimental findings will be presented.
_ p ðT i T u Þ ¼ λA
mc
δ
where ṁ is the mass burning rate, cp is the heat Flame Speed Measurements
capacity, λ is the thermal conductivity (often
estimated to be the average of the unburned There are numerous methods that enable the
and the burned values), A is the surface area, measurement of the laminar flame speed. One
and δ is the flame thickness. The left hand side of the most basic methods is the use of a
represents the amount of energy that is absorbed Bunsen-type burner, see Fig. 12.5, which
as the unburned mixture flows into the entrance establishes the flame speed, SL ¼ uu sin ðαÞ;
of the preheat zone, raising it to the ignition based on geometrical considerations. Here, α is
temperature, Ti. The right hand side represents the half angle at the top of the flame. Another
the heat flux (conduction) to the interface. As classic method is the use of flat flame burner and
such, the heat needed to raise the unburned gas
from Tu to Ti equals the heat conducted from the
reaction zone.
By using the definition of the mass burning
rate, the area cancels out from the equation, and
the laminar burning speed is introduced and can
be isolated for, as follows.
m_ ¼ ρu uu A ¼ ρu SL A
Tb Ti
c p ρu SL ðT i T u Þ ¼ λ
δ

λ=c p T b T i
SL ¼
ρu δ T i T u
We thereby have that the laminar flame speed

depends on the thermal diffusivity of the mix-
λ=c
ture, α ¼ ρ p ; is inversely proportional to the
flame thickness, and is dependent on the reactiv-
ity of the mixture, as represented by the ratio of
the temperature gradients.
If we then estimate the flame thickness as ρu
_
SL ¼ ωδ; where ω_ is the volumetric
reaction rate, the expression for the flame
speed becomes
Fig. 12.5 Burner for measuring flame speed
www.ebook777.com
after Spalding [31] introduced the concept of chamber. As mentioned, this method requires
correction for heat losses, see Fig. 12.6, much correction for the expansion that is caused by
improved results were obtained. This method is the hot combustion products, and if the
still very relevant today, and substantial research experiments are conducted in a fixed volume
efforts have been undertaken to further develop chamber, then there is also a need for correction
the fidelity of the measurements [32, 33]. due to the pressure increase in the chamber as the
A third method that has been used for more flame propagates.
than half a century is the outwardly propagating However, in 1972 Andrews and Bradley [34]
flame method, which requires that the premixture pointed out that something was not well under-
is contained upon ignition by a spark [10]. This stood when it came to the laminar flame speed
containment can be a soap bubble, a balloon or a measurements, despite the fact that more than
100 years had passed since Bunsen’s break-
through experiment for premixed flames
SL and numerous researchers had worked on
establishing the flame speed for half a century
Flame speed
using all the methods described above. They

created a plot of the type shown in Fig. 12.7,
which displays the majority of the reported lam-
inar flame speeds for stoichiometric (equiva-
lence ratio, ϕ, equals 1) methane-air flames at
normal ambient initial conditions (1 atm, 298 K)
Heat loss as a function of the year the result was
published. It can be seen that there was a
Fig. 12.6 The flame speeds obtained on a flat flame major discrepancy between the reported results
burner have to be corrected for the losses to the burner, for what should have yielded one data point. It is
as pointed out by Spalding [31]
important to recognize that moderate extents of
60
50
Laminar flame speed [cm/s]
40
30
20
Uncorrected flame speed regime
Corrected flame speed regime
10
0
1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010
Year
Fig. 12.7 The reported laminar flame speed for stoichiometric methane-air flames at ambient conditions (1 atm,
298 K) versus the year the results were published
380 G. Jomaas
error in the laminar flame speed could result in

noticeable differences in the chemical
mechanisms developed, as the accuracy of the
developed mechanism must necessarily depend
on the corresponding accuracy of its target
points. Such mechanisms, which are developed
and validated against incorrect experimental
results, will yield poor predictive capabilities
and this could lead to dangerous engine designs.
Furthermore, it is to be noted that this was for
conditions perceived to be simple, as compared
to situations with other fuels, fuel mixtures,
leaner or richer conditions, higher pressures,
other inerts, etc., which will be discussed later. Fig. 12.8 The flame speeds obtained in a burner
Wu and Law [35] suggested that the observed should be corrected for the stretch rate, as pointed out
scatter in existing experimental methane flame by Law [35]
speed data in Fig. 12.7 was caused by the lack of
correction for flame stretch (aerodynamic non-
uniformity and/or unsteady effects that lead to atmospheric conditions from various methods
changes in the flame surface area, which results that have utilized the abovementioned correc-
in a modified reaction rate change, and therefore tion methods [36–39]. The good comparison,
also in the propagation rate) and diffusion also with the numerical results (Wang-
effects, and proposed a methodology to system- mechanism [40]), indicates that the problem of
atically subtract the effects of stretch from flame the laminar flame speed is now well-understood.
speed measurements. In particular, they These results have then enabled increased
designed a stagnation flow experiment in accuracy of the numerical results obtained
which flame stretch (aerodynamic using detailed chemical reaction schemes, a
non-uniformity and/or unsteady effects) could result that also validates the mechanism due to
be measured, and then obtained the unstretched, the fact that the laminar flame speed is the
fundamental flame speed through a linear eigenvalue of the initial problem. As such, the
extrapolation to zero stretch, see Fig. 12.8. theoretical (mathematics, physics and chemis-
Utilizing this technique and applying the stretch try), numerical and experimental development
correction to other available experimental data, combined with the utilization of technological
they were able to significantly reduce the developments (computers, cameras, measuring
discrepancies between measured flame speeds devices) have enabled us to understand the chal-
and narrow the discrepancies in the determined lenging physical problem through a structured
fundamental flame speeds of methane and effort. Often, the error bars posed in practical,
hydrogen, respectively. The impact that this real-life fire problems and in fundamental
correction method had is obvious from the research problems in the field of fire safety are
increased uniformity of the results published of the order that the laminar flame speed prob-
thereafter, see ‘Corrected flame speed regime’ lem used to have, and the structured effort
data in Fig. 12.7. undertaken to solve the flame speed problem
Figure 12.9 is a plot of recent results for the should be of inspiration for other fields of fire
laminar flame speed for methane-air flames at and combustion safety [41].
www.ebook777.com
Fig. 12.9 The laminar flame speed as a function of the line represents numerical results from a detailed chemical
equivalence ratio for methane-air flames at atmospheric mechanism [40]
conditions from different experiments [36–39]. The solid
Parameters Affecting the Laminar around stoichiometric conditions, and research

Flame Speed has shown that the peak lies slightly on the rich
side [42].
The laminar flame speed is a function of a range The expectations for the flame speed based
of practical parameters that will be described in on the temperature results are confirmed in
the following. Fig. 12.11, which shows the experimentally
obtained laminar flame speeds of numerous
Fuel Type and Equivalence Ratio hydrocarbons at standard initial ambient
As discussed, the laminar flame speed primarily conditions (1 atm, 298 K) using the constant-
depends on the thermal diffusivity of the mixture pressure spherical flame technique [44, 45].
and the reaction rate, and hence on the flame Results from the outwardly propagating spheri-
temperature. Given the fact that the flame tem- cal flame configuration are chosen, as this con-
perature is dominated by the chemical potential figuration is known to yield high accuracy results
energy stored in the fuel, we therefore expect and has been central in the increased accuracy of
different fuels to have different laminar flame the measurements and results that has taken place
speeds. The enthalpy of different fuels can be over the last three decades, as shown in Fig. 12.7
found in Chap. 5. Figure 12.10 shows the calcu- [38, 46–51]. This type of apparatus also easily
lated adiabatic flame temperatures for a range of allows for variations of the initial pressure, for
hydrocarbon-air mixtures, and the results suggest which results will be presented and discussed
that acetylene should have the highest flame below. Still, current research is carried out in
speed among the hydrocarbons, followed by eth- order to further improve the understanding of
ylene. As expected, the reactivity decreases as the accuracy of various flame speed measure-
the mixture is moved from stoichiometric ment techniques [137].
conditions (apart from for acetylene), both Acetylene (C2H2) vs. Ethylene (C2H4)
towards leaner and richer conditions. As such, vs. Ethane (C2H6) is one interesting comparison,
most of the fuels have a temperature maximum as the results indicate that the molecular structure
382 G. Jomaas
2700
Flame Temperature [Celsius]

2600
2500
Acetylene
2400
Ethylene
2300
2200 Propylene
2100 Propane
2000 Ethane
1900
Methane
1800
1700
0.60 0.80 1.00 1.20 1.40
Equivalence Ratio, f
Fig. 12.10 The adiabatic flame temperature as a function of equivalence ratio for a range of hydrocarbons in air.
The calculations are based on equilibrium considerations [43]
Fig. 12.11 Laminar flame 160

speeds as a function of the
equivalence ratio for 140
Acetylene
various hydrocarbons in air
120 Ethylene
at standard conditions
100 DME
SL [cm/s]
80 Methane
60 Ethane
40 Propane
20 Propylene
0
0.5 1.0 1.5 2.0
has an effect on the laminar burning velocity. flame speed with a second order effect arising
However, as Fig. 12.10 showed that the tempera- from the transport properties of the fuel.
ture also was significantly different for the three Another point worth noting is that the flame
fuels, further studies were needed to confirm that speed for many alkanes is about 40 cm/s for
the molecular structure played a role. Law diluted stoichiometric conditions and this has been
mixtures of the three fuels with nitrogen so that shown to be the case for numerous fuels in the
they all had the same adiabatic flame temperature, entire range from methane to octane [52]. How-
and the results were conclusive, in that the rank- ever, isomers, alcohols and aromatics display a
ing from highest to lowest flame speed still was larger variation in the flame speed at this condi-
acetylene, ethylene and ethane [13]. Therefore, tion. Just as the flame temperature did, the lami-
the flame temperature dominates the laminar nar flame speed decreases as the mixtures move
www.ebook777.com
away from the stoichiometric condition, both

towards leaner and richer conditions. Thus,
as the flames become leaner/richer, the flame
temperature is reduced, the flame speed is
reduced and the flame becomes thicker (inversely
proportional to the flame speed). The plot for the
flame thickness as a function of the equivalence
ratio is actually not very unlike the plot for the
Minimum Ignition Energy in form (for more on
MIE, see Chaps. 17 and 69) and the plot for
quenching distance, which also scales with the
flame thickness (see Chap. 17). Furthermore, the
flammability limits (see Chap. 17) create a sort of
asymptote for the characteristic parabolic shape Fig. 12.12 Laminar flame speeds as a function of the
that the flame speed, the flame temperature and upstream mixture temperature for methane at different
equivalence ratios at atmospheric pressure [24]
the flame thickness have, and estimates on the
needed flame temperature for sustained reaction
are made in Chap. 5.
Tu " ) ρu # ) SL ". Thus, for an increase
Upstream Temperature in upstream temperature, it is the changes in the
Given the importance that the temperature has on density that lead to the only significant increase
the flame speed, it is also expected that increas- in the laminar flame speed.
ing the upstream (unburned) temperature should Experimental and numerical results confirm
have an effect on the flame speed. First, it is that the laminar flame speed increases as the
envisioned that this increase could influence the upstream temperature is increased, as seen in
adiabatic flame temperature. However, as seen in Fig. 12.12. In summary, the reason for this
Chap. 5, the adiabatic flame temperature can be increase is that the laminar flame speed is pro-
estimated as Tad ¼ Tu + Yu qc/cp, and this effect portional to the thermal diffusivity, α, which in
is therefore small, because Tu is small compared turn scales as α T 3=2 p1 . Thus, the laminar
to Yuqc/cp. As such, from this perspective, the burning velocity goes up as the temperature is
main effect of the increase in the upstream tem- increased.
perature is to facilitate the reaction by increasing
the reaction rate, which has a weak dependence Inert Type and Oxygen Percentage
on the temperature for small increases in the As the creation of an inert or partially inert atmo-
upstream temperature. In the more rare cases sphere is a common fire safety strategy, it is of
where the upstream temperature is significantly interest to understand how this influences the
increased, the effect of the enhanced reaction rate flame speed. The type of inert will affect the
must be given greater attention than in the cur- laminar burning velocity in different ways.
rent string of arguments. Second, there is a small Experiments and calculations with detailed
effect due to change in transport properties, chemistry support this, and the most typical com-

because λ=c p T γ ; but this is a mild depen- parison is one between nitrogen, argon and
dence, as γ is smaller than unity. Finally, there helium as inerts for a fuel/oxygen/inert mixture
are changes associated with the density change. with molar ratio of oxygen to inert equal to 0.21/
For a given fixed burning flux, m_ ρu SL ; the 0.79. In the following, the burning flux, m_ ρu
laminar burning speed will increase with an SL ; is considered, as this parameter is the true
increase in the upstream temperature: eigenvalue of the problem, and thus isolates the
384 G. Jomaas
effects of inert addition better than the flame 0.25

speed does.
The specific heat depends on the degrees of
Laminar burning flux, f 0, gm/cm2-s

freedom for the respective inert, with increasing 0.20
cp for increasing freedom. Monatomic gases He-air

have smaller cp than diatomic gases, so
0.15
c p, He ¼ c p, Ar < c p, N 2 . The thermal conductivity
is inversely related to the mass of the inert, thus Ar-air
λAr λN2 < λHe (somewhat larger for N2 than 0.10
for Ar). As the burning flux is a function of the

ratio of these two parameters, m_ λ=c p , and
0.05 N2-air
as a result, we have that m_ He > m_ Ar > m_ N2 ,
which is shown to hold for methane in
Fig. 12.13 and similar results have been
0.00
reported for other fuels [11, 53]. 4 6 8 10 12 14 16 18
Figure 12.14 shows that the laminar flame % CH4 in various airs
velocity is a strong function of the oxygen con-
centration in the premixture, which is consistent Fig. 12.13 Laminar burning flux for methane-oxygen-
with the fact that the flame temperature is a inert at different methane concentrations (ϕ ¼ 1 at 9.5 %)
at atmospheric pressure [13]. The molar ratio of oxygen-
strong function of the oxygen concentration.
inert was 0.21–0.79 for all mixtures
Only a few experimental data points (solid
markers) are presented along with the numerical
results, as the flames are more and more affected
by cellular flame front instabilities (see below) as
the flame becomes thinner (flame velocity
increases). Also, some of the very rich flames
were clearly sooty, and such conditions tend to
slow the flame due to increased radiative heat
transfer and thus decreased flame temperature.
Finally, it is worth noting that the chemical
kinetic scheme used to obtain the numerical
results is not validated for reduced oxygen
concentrations and the results should therefore
be used with caution.
It is important to note that these experiments
with reduced oxygen concentration reveal that
the flames can still propagate in such an environ-
ment. This is of interest, as a commonly used
safety measure is to reduce the oxygen concen-
tration in a room to around 15 % oxygen in air, as
this typically prevents ignition of a non-premixed
flame or extinguishes the same type of flame Fig. 12.14 Numerical and experimental laminar flame
rapidly. However, if a premixture is created in speeds as a function of the oxygen concentration in the
such an environment, ignition and propagation initial environment for acetylene at different equivalence
ratios at an initial pressure of 10 atm
are still possible.
www.ebook777.com
Fig. 12.15 Numerical 45

and experimental laminar
40 Ethane 1 atm
flame speeds as a function
of pressure for ethane-air 35 2 atm
at different equivalence
ratios 30 5 atm
SL [cm/s]
25 10 atm
20
1 atm Numerical
15
2 atm Numerical
10
5 atm Numerical
5
10 atm Numerical
0
0.60 0.80 1.00 1.20 1.40
Pressure
The functionality of the laminar flame speed with
respect to pressure is SL pðn=2Þ1 , where n is
the reaction order (sometimes referred to as the
pressure exponent), which indicates the influence
of pressure on the concentrations of the reactants
and thus on the reaction rate. For elementary
reactions, n can be 1, 2, or 3, and if the reaction
order is 2, then it is predicted that the laminar
flame velocity should be insensitive to pressure
variations [54]. However, the reaction order is
smaller than 2 for most hydrocarbons, so the
laminar flame speed typically decreases as the
Fig. 12.16 The burning flux of ethane-air (and other
pressure is increased. Experiments have
fuel-air mixtures) increases as the pressure increases.
validated this, as shown in Fig. 12.15. The results are obtained by multiplying the density
However, as mentioned, the true eigenvalue obtained from equilibrium calculations with the experi-
of the problem is the mass burning flux, which is mentally obtained flame velocity
the product of the laminar burning velocity and
the density, and this parameter should increase as to meet the desired design or safety criteria, both
the pressure is increased. Figure 12.16 shows that for the propagation velocity and for the stability
the mass burning flux indeed increases as the (see below) [55–57]. This becomes very relevant
pressure increases. High-pressure accidents are for all the new fuel types, such as syngas and
therefore more severe than low pressure biofuels, as one can either change the burners, or
accidents, because the momentum is much create the mixture (through use of inerts or other
higher. fuels) such that it will perform as desired in the
existing design. Results for a range of syngas
Fuel Mixture mixtures are shown in Fig. 12.17, and similar
Based on the knowledge of the properties of the results for a range of pressures are shown in the
gases, custom-made compositions can be made paper by Sun et al. [55].
386 G. Jomaas
250
p =1 atm
CO:H2 = 50:50 to detonation transition (DDT) is possible.
CO:H2 = 75:25
These large scale phenomena are challenging to
CO:H2 = 95:5
200 model, given the vast span of the scales, but there
CO:H2 = 99:1
Faeth et al., 1997 are validated models that can handle the problem
SL [cm/s]
150 McLean et al., 1994

reasonably well [79].
100
50
Flame Structure and Detailed

0
1 2 3 4 5
Chemistry of Premixed Flames
Standard tools exist to calculate the flame propa-

Fig. 12.17 The laminar flame speed of various mixtures
of hydrogen and carbon monoxide (syngas) in air at stan-
gation with detailed chemistry, and one can
dard conditions [48, 55, 58]. The symbols represents extract details about all the relevant parameters
experimental results, whereas the lines represents calcu- from such calculations. It is important that such
lated values calculations are done in an ‘infinite’ domain so
that boundary conditions do not influence the
calculations. This is easily achieved over a
Flame Instabilities and Acceleration small domain, given the fact that premixed
Due to the practical importance of cellular flames flames are very thin. The calculation results
in combustion at elevated pressures, and also for presented herein were obtained using the Sandia
safety considerations at atmospheric conditions, PREMIX code [80], which simulates the
the last decade has seen significant experimental propagation of a planar flame through a
efforts pursuing a more thorough understanding premixture under adiabatic conditions. The
of the fundamentals of the phenomenon. An out- C2–C3 flames were simulated using the detailed
wardly propagating flame will typically not have reaction mechanism of Qin et al. [40], which
instabilities initially, but the flame front will includes 70 species and 463 reactions. Other
eventually become filled with cellular mechanisms exist, either for specific fuels, or
instabilities, which can either be hydrodynamic for ranges of fuels, with the smallest being that
or thermal-diffusive in nature, something which for hydrogen with 9 species and 21 reactions to
has been shown extensively both in experiments larger mechanisms with hundreds of species and
and theory [29, 57, 59–68]. Figure 12.18 shows several thousand reactions [81–89]. As this is a
the evolution of a hydrogen flame from a point- field in rapid development, newer mechanisms
source ignition (e.g. accidental explosion) for and improvements to existing ones are published
three different initial conditions. Research has continuously, and lately there has been a great
shown that although these cellular instabilities focus on obtaining results and understanding for
accelerate the propagation velocity [69–74], biofuels and other alternative fuels [90].
there is no transition from a laminar to a turbulent In addition to providing global results such as
flame in an unconfined environment [75]. Bradley the propagation velocity of the flames, these
has also shown that large scale experiments cor- detailed mechanisms provide insight into the
relate well with small scale experiments [76, 77]. intermediate steps that take place as reactants
Under confinement in tubes with or without are turned into products. For example,
obstacles [78, 112, 128–131], which is common calculations with hydrocarbons have shown
in places where accidental gas leaks are ignited, (in the reaction coordinate, see Fig. 12.19) that
initially slow flames may accelerate spontane- the CO is formed before the CO2 and that tem-
ously up to supersonic speed, and deflagration perature goes up as the hydrocarbon is
www.ebook777.com
Fig. 12.18 Evolution of an outwardly propagating flame front as a result of the decrease in flame thickness
spherical flame from a point-source ignition in a with the increase in pressure (the flame thickness scales
premixture of hydrogen and air at three different pressures inversely with the laminar flame speed). The increased
[13]. The lean flame (left) has a cellular structure due to surface area leads to an increase in the propagation veloc-
thermal-diffusive instabilities, whereas the rich flame at ity, and the flame no longer traverses into the unburned
the same pressure (middle) is smooth. The rich flame to mixture with the laminar flame velocity
the right has hydrodynamic cellular instabilities over the
388 G. Jomaas
0.20
Equilibrium T
2000
T (K) 0.15
Mole Fraction
H2O
1500
ρ (kg/m ) x 1000
3 CH4
Tign 0.10
1000 CO2
U (cm/s) x 5
0.05
500
CO
0
0
0 0.05 0.10 0.15
0 0.05 0.10 0.15
Distance (cm) Distance (cm)
Fig. 12.19 Reaction coordinate view of the physical properties (left) and species (right) for a methane-air flame [24]
consumed. Further, as the concentration of the excellent ones do exist [89, 90]. In addition,
hydrocarbon diminishes, the CO concentration is numerous new methods have been developed to
at its max, before decaying as it is converted into deal with the complexity and large computational
CO2. The intermediate carbon-containing spe- times that are introduced when detailed chemistry
cies of the reaction, such as CH3, CH2O, HCO is included in reacting flow scenarios and these
and CH exist in a very small zone, whereas HO2 methods should be consulted to obtain high fidel-
and H2O2 exist over a somewhat broader zone. ity results at reasonable calculation cost [93–96].
As such, the reaction zone is very narrow,
whereas the equilibrium temperature is reached
relatively slowly. The increase in the reaction Other Aspects Related to Premixed
rate comes with the activation of the reaction, Flames
whereas the decrease is caused by depletion of
the reactants. The fact that the temperature and There are numerous other aspects to the study of
fuel depletion profiles are not completely sym- premixed flames than the laminar flame speed, but
metric, indicates that the thermal and the mass they are all, to some extent, variations or different
diffusion are not equal throughout the flame, phases of the main propagation mode. Some of
which means that the Lewis number (ratio of these aspects are briefly mentioned in the following.
thermal and mass diffusion) is not equal to
unity, as opposed what is typically assumed in a
basic flame analysis. Very detailed information Adiabatic Flame Temperature
and in-depth understanding of flame chemistry for Constant Pressure Versus Constant
and flame structures can be found in numerous Volume Combustion
classic textbooks [18–22].
Another parameter of interest in the study of
premixed flames is the adiabatic flame tempera-
Numerical Simulations of Premixed ture, which is based on thermochemistry (see
Flames Chap. 5). The flame temperature of a premixed
flame can be estimated as
A proper discussion of Direct Numerical
Y F , u qc
Simulations (DNS), Large Eddy Simulations Tb ¼ Tu þ
cp
(LES) and Computational Fluid Dynamics
(CFD) modeling requires a book of its own, and where qc is the chemical heat release.
www.ebook777.com
The adiabatic flame temperature, Tad, is chemical time scale, which typically is so short
estimated from the total enthalpy, which is the that it is very unlikely for any spark not to exist for
sum of the chemical and the sensible energy (see a sufficient time to ignite a premixture within the
Chap. 5 for further details). For a constant pres- flammability limits. As such, a non-ignition sce-
sure process, we have that the enthalpy of the nario from a spark means that the energy density
reactants should equal the enthalpy of the (a larger spark will have to heat a larger mass) was
products, so that not sufficient. Again, several levels of mathemat-
ical complexity can be offered for the analysis,
Hreact ¼ H prod which reveals that ignition parameters are a func-
X X X h i
N i hi ¼ N i hi ¼ N i h0f , i þ c p, i ðT ad T u Þ : tion of the laminar flame speed (and the flame
react prod prod thickness). The critical ignition radius is propor-
tional to the flame thickness (and thus inversely
For a constant volume process, on the other hand, proportional to the laminar burning velocity).
we have the internal energy of the reactants Given the fact that the laminar burning velocity
match the internal energy of the products, so that is decreasing as the pressure is increased, we have
that ignition of most premixtures is facilitated as
Ureact ¼ Uprod
the pressure is increased.
H react ¼ H prod þ Ru N react T u N prod T ad :
Thus, for the same initial conditions, a constant

volume combustion process results in much Turbulence
higher temperatures than a constant pressure
combustion process does. This is a consequence As reaction is a flame surface phenomenon, it is
of the pressure forces not doing any work when predicted that the turbulent burning velocity is
the volume is fixed. higher than the laminar burning velocity due to
the increased area of a turbulent flame and the
increase in the transport at the flame surface,
which will lead to increased reaction rate and
Ignition
consequently larger velocity. The reader can con-
sult several resources to learn more on the effect
In comparison to other premixed phenomena,
of turbulence on premixed flames [67, 97–110].
ignition can be considered a low temperature phe-
nomenon. As for studies of chemical kinetics,
both low and high temperature and homogenous
and non-homogenous phenomena have to be Tube Flames
understood. In the piloted ignition scenario, a
local energy is supplied, which will provide situa- Reacting flow and premixed flames in tubes are
tion where the Damköhler number (the ratio of of great importance for efficient and safe engine
characteristic fluid time and the characteristic and burner design [11, 111–114]. The pressure
chemical time) has to be larger than a critical dependence in a closed tube creates acoustic
Damköhler number for ignition to occur. Typical excitation (spherical configuration has acoustic
ignition sources for fire scenarios are very large decay). The Rayleigh criterion (oscillations)
compared to the MIE. Sparks are used for shows that heat must be added in phase with
establishing the MIE, and a small spark with pressure if energy is to be transferred into
short duration is enough. The only requirement acoustic waves [115]. The classical demonstra-
is that the duration is long enough compared to the tion consisted of a hydrogen flame burning
390 G. Jomaas
inside an open tube. Pressure variations in the propagation. As such, the detonation velocity is
tube cause the flow of gas, and therefore the heat the velocity of the unburned gases moving into
release, volume expansion, and backpressure on the wave, and at the Chapman-Jouguet point, the
the nozzle to vary during the vibration. In detonation velocity equals the velocity of sound
Rijke’s tube, which is open at both ends, energy in the gases behind the detonation wave plus the
from heat is turned into sound from a standing mass velocity of these gases.
wave [116]. If the product of the fluctuating From a practical point, it could be of interest
parts of the heat release and the backpressure, to know if the detonation velocity of fuel-
integrated over a cycle of the vibration, is posi- oxygen-argon is faster than that of fuel-oxygen-
tive, in the absence of damping the vibration helium. As both have a unity Mach number
will be maintained. In other words, if the behind the detonation, and assuming that the
fluctuating heat release is more in phase than flame temperature is not too different (it is dif-
out of phase with the vibration in the resonator, ferent, but not 10 times), the following shows
conditions are right for feeding energy into the that it all comes down to the molecular weight
vibration. dependence of the speed of sound, and that the
Experiments [132, 133] and theory [134] have detonation velocity in helium is significantly
demonstrated a variety of possible outcomes of higher.
flame-acoustic interaction, such as flame stabili- sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
zation by sound waves, a parametric instability uHe ðγRT ÞHe RHe R0 =MW He
¼ ¼
and spontaneous turbulization of the flame front. uAr ðγRT ÞAr RAr R0 =MW Ar
In particular, the spontaneous turbulization may rffiffiffiffiffiffiffiffiffiffiffiffiffi
MW Ar pffiffiffiffiffi
develop as a result of a flame-acoustic resonance ¼ ¼ 10
with strong increase of pressure and burning rate MW He
in the process [135, 136]. As with the other smaller topics in this subchap-
ter, there is significant knowledge to be found in
the literature on detonations [99, 117–125], as
Detonation well as in Chaps. 69 and 70.
The main motion of this chapter has been the

subsonic propagation of premixed flames, but as
shown in Fig. 12.2, the solutions to the equations The Role of Premixed Flames in Fire
also predict a regime that is supersonic, and this Safety Engineering
propagation is referred to as detonation. Whereas
the propagation velocity of the laminar premixed As fire safety engineers, our main objective is to
flames in the deflagration regime moved with a avoid ignition. This can be achieved by creating
velocity of the order of 1 m/s, detonation an environment with no or small energy sources,
velocities are of the order of 1 km/s. As opposed by keeping the atmosphere outside of the flam-
to what is the case for laminar premixed flames, mability limits through design or by use of detec-
heat conduction and radical diffusion do not con- tion systems, through flame arrestors that will
trol the velocity for detonations, and the hot lead to quenching, through creating an overall
gases follow the wave for detonations. The inert atmosphere (remove oxygen, increase nitro-
shock wave structure and the developed super- gen). However, if ignition does occur, we want to
sonic wave raise the temperature and pressure have appropriately scaled and positioned pres-
substantially and cause explosive reaction and sure vents for explosions, or introduce heat losses
the energy release that sustains the wave or energy absorption for flames, which can be
www.ebook777.com
done either chemically or physically. Further 10. Lewis, B. & von Elbe, G. 1951 Combustion, Flames
information can be found in Chaps. 69 and 70. and Explosions of Gases, Academic Press,
New York.
Igniting a premixture in a tube that is open at 11. Lewis, B. & von Elbe, G. 1987 Combustion, Flames
both ends will lead to a combustion wave with and Explosions of Gases, 3rd edn. Academic.
transport processes involving heat conduction 12. Strehlow, R.A. 1979 Fundamentals of Combustion,
and diffusion of radicals, which therefore will Kreiger Publishing Company, New York.
13. Law, C. K. 2006 Combustion Physics, Cambridge
result in a speed that is lower than the speed of University Press.
sound in the unburned mixture (deflagration). If, 14. Zel’dovich, Y. B., Barenblatt, G. I., Librovich, V. B.
on the other hand, the tube filled with the same & Makhviladze, G. M. 1985 The Mathematical The-
premixture is closed at one end, then there is a ory of Combustion and Explosions. Consultants
Bureau.
possibility of a transition from deflagration to 15. Kuo, K. K. 2002 Principles of Combustion, 2nd edn.
detonation (supersonic). This can for example John Wiley.
happen in mine explosions. Thus, it is clear that 16. Williams, F. A. 1985 Combustion Theory. Addison-
the environment plays a significant role for what Wesley.
17. Turns, S. R. 2000 An Introduction to Combustion:
happens with the ‘explosive’ mixture. Concepts and Applications. McGraw-Hill.
The main lesson to be learned for a fire safety 18. Glassman, I. 1996 Combustion. Academic.
engineer from this chapter is that detailed and 19. Gaydon, A. G. & Wolfhard, H.G. 1970 Flames.
accurate knowledge of a complex problem like Chapman and Hall.
20. Fristrom, R. M. & Westenberg, A. A. 1965 Flame
that of premixed burning is achievable, as shown Structure. McGraw-Hill.
in Fig. 12.7. With structured efforts over time, 21. Gardiner, W. C., Jr. 1999 Combustion Chemistry,
the field of fire safety can grow into a science in Springer-Verlag.
its true form. The challenge lies in scaling up 22. Fenimore, C. P. 1964 Chemistry in Premixed
Flames. Pergamon.
fire [41]. 23. Buckmaster, J. D. & Ludford, G. S. S. 1982 Theory
of Laminar Flames. Cambridge.
24. McAllister, S. Chen, J-Y & Fernandez-Pello,
A.C. 2011 Fundamentals of Combustion Processes,
References Springer.
25. Rankine, W. J. M. (1870). “On the thermodynamic
1. Bunsen, R, & Roscoe H., 1857 Photochemische theory of waves of finite longitudinal disturbances”.
Untersuchungen: Maafsbestimmung der Chemische Philosophical Transactions of the Royal Society
Wirkungen des Lichts, Poggendorff’s Annalen der of London 160: 277–288, doi:10.1098/rstl.1870.
Physik 100, 43–88, http://gallica.bnf.fr/ark:/12148/ 0015.
bpt6k15185v/f55.image. 26. Hugoniot, H. (1887). “Mémoire sur la propagation
2. Semenov, N.N. 1927 Z. Phys., 46, 109; Chemical des mouvements dans les corps et spécialement dans
Kinetics and Chain Reactions, Clarendon, London, les gaz parfaits (première partie) [Memoir on the
1935. propagation of movements in bodies, especially per-
3. Hinshelwood, C.N. and Thompson, H.W. 1928 Proc. fect gases (first part)”] (in French). Journal de
R. Soc. London A, 118, 171; l’Ećole Polytechnique 57: 3–97. See also: Hugoniot,
4. Hinshelwood, C.N. 1929 Kinetics of Chemical H. (1889) “Mémoire sur la propagation des
Change in Gaseous Systems, 2nd Edition, mouvements dans les corps et spécialement dans
Clarendon, Oxford. les gaz parfaits (deuxième partie)” [Memoir on the
5. Houston, P.L. 2001 Chemical Kinetics and Reaction propagation of movements in bodies, especially per-
Dynamics, McGraw-Hill fect gases (second part)], Journal de l’Ećole
6. Laidler, K.J. 1987 Chemical Kinetics, Pearson Polytechnique, vol. 58, pages 1–125.
Education. 27. Chapman, D. L. 1899 “On the rate of explosion in
7. Semenov N. N. 1958 Some Problems in Chemical gases”, Philosophical Magazine, Series 5, 47 (284):
Kinetics and Reactivity. Vol. I. Princeton University 90–104.
Press. 28. Jouguet, J.C.E. 1906 “Sur la propagation des
8. Frank-Kameneskii, D. A. 1969 Diffusion and heat réactions chimiques dans les gaz” [On the propaga-
Transfer in Chemical Kinetics, Plenum Press. tion of chemical reactions in gases], Journal des
9. Wang, X. & Law, C.K. 2013 “An analysis of the Mathématiques Pures et Appliquées, Series 6, Vol.
explosion limits of hydrogen-oxygen mixtures,” 1, pp. 347–425 (1905), continued in Vol. 2, pp. 5–85
Journal of Chemical Physics 138, 134305, 1–12. (1906).
392 G. Jomaas
29. Landau, L.D. 1944 On the theory of slow combus- 45. Jomaas, G., Zheng, X.L., Zhu, D.L., & Law,
tion. Acta Physicochim. U.R.S.S. 19, 77–85. C.K. 2005 Experimental determination of counter-
30. Mallard, E. & Le Chatelier, H.L. 1883 Ann. Mines 4, flow ignition temperatures and laminar flame speeds
379. of C2-C3 hydrocarbons at atmospheric and elevated
31. Botha, J. P. & Spalding, D. B. 1954 The laminar pressures. Proc. Combust. Instit. 30, 193–200.
flame speed of propane/air mixtures with heat 46. Dowdy, D.R., Smith, D.B., Taylor, S.C. & Williams,
extraction from the flame. Proc. R. Soc. Lond., A. 1990 The use of expanding spherical flames to
Series A, Mathematical and Physical Sciences determine burning velocities and stretch effects in
225, 71–96. hydrogen/air mixtures. Proc. Combust. Instit. 23,
32. de Goey, L. P. H., van Maaren, A. & Quax, R. M. 325–332.
1993 Stabilization of adiabatic premixed laminar 47. Bradley, D. & Harper, C.M. 1994 The development
flames on a flat flame burner Combust. Sci. Technol. of instabilities in laminar explosion flames. Com-
92, 201–207. bust. Flame 99, 562–572.
33. Bosschaart, K.J. & de Goey, L.P.H. 2004 “Detailed 48. Hassan, M.I., Aung, K.T. & Faeth, G.M. 1997
analysis of the heat flux method for measuring burn- Properties of laminar premixed CO/H2/air flames at
ing velocities,” Combustion and Flame various pressures, Journal of Propulsion and Power
132, 170–180. 13, 239–245.
34. Andrews, G.E. & Bradley, D. 1972 Determination of 49. Hassan, M.I., Aung, K.T. & Faeth, G.M. 1998
burning velocities: A critical review. Combust. Measured and predicted properties of laminar
Flame 18, 133–153. premixed methane/air flames at various pressures.
35. Wu, C.K. & Law, C.K. 1984 On the determination of Combust. Flame 115, 539–550.
laminar flame speeds from stretched flames. Proc. 50. Hassan, M.I. Aung, K.T. Kwon, O.C. & Faeth,
Combust. Instit. 20, 1941–1949. G.M. 1998 Properties of laminar premixed hydrocar-
36. Hassan, M.I., Aung, K.T. & Faeth, G.M. 1998 bon/air flames at various pressures. J. Prop. Power
Measured and predicted properties of laminar 14, 479–488
premixed methane/air flames at various pressures. 51. Tse, S.D., Zhu, D.L. & Law, C.K. 2000 Morphology
Combust. Flame 115, 539–550. and burning rates of expanding spherical flames in
37. Vagelopoulos, C.M. & Egolfopoulos, F.N. 1998 H2/O2/inert mixtures up to 60 atmospheres. Proc.
Direct experimental determination of laminar flame Combust. Instit. 28, 1793–1799.
speeds. Proc. Combust. Instit. 27, 513–519. 52. Davis, S.G. & Law, C.K. 1998 Determination of fuel
38. Rozenchan, G., Zhu, D.L., Law, C.K. & Tse, structure effects on laminar flame speeds of C1 to C8
S.D. 2002 Outward propagation, burning velocities, hydrocarbons. Combust. Sci. Tech. 140, 427–449.
and chemical effects of methane flames up to 60 atm. 53. Kwon, O.C. & Faeth, G.M. 2001 Flame/stretch
Proc. Combust. Instit. 29, 1461–1469. interactions of premixed hydrogen-fueled flames:
39. Gu, X.J., Haq, M.Z., Lawes, M. & Woolley, Measurements and predictions. Combust. Flame
R. 2000 Laminar burning velocity and Markstein 124, 590–610.
lengths of methane-air mixtures. Combust. Flame 54. Law, C.K. 2006 “Propagation, structure, and limit
121, 41–58. phenomena of laminar flames at elevated pressures,”
40. Qin, Z., Lissianski, V., Yang, H., Gardiner, W.C., Combustion Science and Technology 178, 335–360.
Davis, S.G. & Wang H. 2000 Combustion Chemistry 55. Sun, H., Yang, S.I., Jomaas, G. & Law, C.K. 2007
of Propane: A Case Study of Detailed Reaction High-pressure laminar flame speeds and kinetic
Mechanism Optimization. Proc. Combust. Inst. 28, modeling of carbon monoxide/hydrogen combus-
1663–1669. tion. Proc. Combust. Instit. 31, 439–446.
41. Torero, J.L., 2013 Scaling-up Fire, Proc. Combust. 56. C.K. Law, G. Jomaas, J.K. Bechtold, Proceedings of
Instit. 34 99–124. the Combustion Institute 30 (2005) 159–167.
42. Law, C.K., Makino, A. & Lu, T.F. 2006 “On the 57. G. Jomaas, C.K. Law, J.K. Bechtold, Journal of Fluid
Off-Stoichiometric Peaking of Adiabatic Flame Mechanics 583 (2007) 1–26.
Temperature with Equivalence Ratio,” Combust. 58. Mclean, I.C., Smith, D.B. & Taylor, S.C. 1994 The
Flame 145, 808–819. use of carbon monoxide/hydrogen burning velocities
43. Reynolds, W. C. 1986 The element potential for to examine the rate of the CO + OH reaction. Proc.
chemical equilibrium analysis: implementation in Combust. Instit. 25,749–757.
the interactive program STANJAN. Tech. Rept. 59. Darrieus G. 1938 Propagation d’un front de flamme.
A-3391, Dept. of Mechanical Engineering, Stanford La Technique Moderne 30, No. 18.
University. 60. Markstein, G.H. 1951 Experimental and theoretical
44. Tse, S.D., Zhu, D.L. & Law, C.K. 2004 An optically studies of flame-front stability. J. Aero. Sci. 18,
accessible high-pressure combustion apparatus. Rev. 199–209.
Sci. Instrum. 75, 233–239.
www.ebook777.com
61. Groff, E.G. 1982 The cellular nature of confined spher- 79. Skjold, T., Pedersen, H.H., Bernard, L., Ichard, M.,
ical propane-air flames. Combust. Flame 48, 51–62. Middha, P., Narasimhamurthy, V.D., Landvik, T.,
62. Markstein, G. H. 1964 Nonsteady Flame Propaga- Lea, T & Pesch, L. (2013). A matter of life and
tion. Pergamon. death: validating, qualifying and documenting
63. Frankel, M.L. & Sivashinsky, G.I. 1983 On effects models for simulating flow-related accident
due to thermal expansion and Lewis number in scenarios in the process industry. Fourteenth Inter-
spherical flame propagation. Combust. Sci. Tech. national Symposium on Loss Prevention and Safety
31, 131–138. Promotion in the Process Industries, Florence,
64. Sivashinsky, G.I. 1977 Diffusional-thermal theory of 12–15 May 2013, published in Chemical Engineer-
cellular flames. Combust. Sci. Tech. 15, 137–146. ing Transactions, 31: 187–192. ISBN: 978-88-
65. Bechtold, J.K. & Matalon, M. 1987 Hydrodynamic 95608-22-8. ISSN: 1974-9791, http://www.aidic.it/
and diffusion effects on the stability of spherically cet/13/31/032.pdf.
expanding flames. Combust. Flame 67, 77–90. 80. Kee, R.J., Grcar, J.F., Smooke, M.D. & Miller,
66. Joulin, G. & Clavin, P. 1979 Linear stability analysis J.A. 1985 A FORTAN program for modeling steady
of nonadiabatic flames: Diffusional-thermal model. laminar one-dimensional premixed flames. Sandia
Combust. Flame 35, 139–153. Report SAND85-8240.
67. Clavin, P. 1985 Dynamic behavior of premixed 81. Smith, G.P., Golden, D.M., Frenklach, M., Moriarty,
flame fronts in laminar and turbulent flows. Prog. N.W., Eiteneer, B., Goldenberg, M., Bowman, C.T.,
Energy Combust. Sci. 11, 1–59. Hanson, R.K., Song, S., Gardiner, W.C., Lissianski,
68. Istratov, A.G. & Librovich, V.B. 1969 On the stabil- V. & Qin, Z. GRIMech homepage, Gas Research
ity of gasdynamic discontinuities associated with Institute, Chicago, 1999, http://www.me.berkeley.
chemical reactions. The case of a spherical flame. edu/gri-mech/.
Astronautica Acta 14, 453–467. 82. Li, J., Zhao, Z., Kazakov, A. & Dryer, F.L. 2004 An
69. Sivashinsky, G.I. 1979 On self-turbulization of a updated comprehensive kinetic model of hydrogen
laminar flame. Acta Astronautica 6, 569–591. combustion. Int. J. Chem. Kinet. 36, 566–575.
70. Filyand, L., Sivashinsky, G.I. & Frankel, M.L. 1994 83. Davis, S.G., Joshi, A.V., Wang, H. & Egolfopoulos,
On self-acceleration of outward propagating wrin- F.N. 2005 An optimized kinetic model of H2/CO com-
kled flames. Physica D 72, 110–118. bustion. Proc. Combust. Instit. 30, (2005), 1283–1292.
71. Bychkov, V.V. & Liberman, M.A. 1996 84. Curran, H.J., Fischer, S.L. & Dryer F.L. 2000 The
Stability and the fractal structure of a spherical Reaction Kinetics of Dimethylether. II:
flame in a self-similar regime. Phys. Rev. Lett. 76, Low-Temperature Oxidation in Flow Reactors. Int.
2814–2817. J. Chem. Kinet. 32, 741–759.
72. Gostintsev, Y.A., Istratov, A.G. & Shulenin, 85. Curran, H. J., Gaffuri, P., Pitz, W. J. & Westbrook,
Y.V. 1989 Self-similar propagation of a free turbu- C. K. 2002 A comprehensive modeling study of
lent flame in mixed gas mixtures. Combustion, iso-octane oxidation. Combustion and Flame
Explosion and Shock Waves 24, 563–569. 129, 253–280.
73. Gostintsev, Y.A., Istratov, A.G., Kidin, N.I & 86. Farrell, J. T., Cernansky, N. P, Dryer, F. L., Friend,
Fortov, V.E. 1999 Autoturbulization of gas flames: D. G., Hergart, C. A., Law, C. K., McDavid, R. M.,
Theoretical treatment. High Temperature 37, Mueller, C. J., Patel, A. K. & Pitsch, H. 2007 Devel-
603–607. opment of an experimental database and kinetic
74. Sivashinksy, G.I. 1983 Instabilities, pattern forma- models for surrogate diesel fuels. 2007 SAE World
tion, and turbulence in flames. Ann. Rev. Fluid Mech. Congress 2007-01-0201.
15, 179–199. 87. Fisher, E. M., Pitz, W. J., Curran, H. J. & Westbrook,
75. F. Wu, G. Jomaas, and C.K. Law, “An Investigation C. K. 2000 Detailed chemical kinetic mechanisms
on Self-Acceleration of Cellular Spherical Flames,” for combustion of oxygenated fuels. Proceedings of
Proceedings of the Combustion Institute, Vol. the Combustion Institute 28, 1579–1586.
34, 937–945, 2013. 88. Pitz, W. J., Cernansky, N. P., Dryer, F. L.,
76. Bradley, D. 1999 Instabilities and flame speeds in Egolfopoulos, F. N., Farrell, J. T., Friend, D. G. &
large-scale premixed gaseous explosions. Phil. Pitsch, H. 2007 Development of an experimental
Trans. R. Soc. Lond. A 357, 3567–3581. database and chemical kinetic models for surrogate
77. Bradley, D. Cresswell, T.M. & Puttock, J.S. 2001 gasoline fuels. 2007 SAE World Congress 2007-01-
Flame acceleration due to flame-induced instabilities 0175.
in large-scale explosions. Combust. Flame 124, 89. Sirjean, B., Dames, E., Sheen, D. A., You, X.-Q.,
551–559. Sung, C., Holley, A. T., Egolfopoulos, F. N., Wang,
78. Dorofeev, Kuznetsov, Alekseev, Efimenko, and H., Vasu, S. S., Davidson, D. F., Hanson, R. K.,
Breitung “Evaluation of limits for effective flame Pitsch, H., Bowman, C. T., Kelley, A. P., Law,
acceleration in hydrogen mixtures,” Journal of Loss C. K., Tsang, W., Cernansky, N. P., Miller, D. L.,
Prevention in the Process Industries 14, 2001. Violi, A. & Lindstedt, R. P. 2009 A high-temperature
394 G. Jomaas
chemical kinetic model of n-alkane oxidation, 109. Pope, S. B. 1990 Computations of turbulent combus-
JetSurF version 1.0. http://melchior.usc.edu/ tion: progress and challenges. Proc. Combust. Inst.
JetSurF/JetSurF1.0/Index.html. 23, 591–612.
90. Liu W., Sivaramakrishnan R., Davis M.J., Som S., 110. Pope, S. B. 2000 Turbulent Flows, Cambridge.
Longman D.E., Lu T.F., 2013 “Development of a 111. Bradley, D. 2000 “Flame Propagation in a Tube: The
Reduced Biodiesel Surrogate Model for Compres- Legacy of Henri Guénoche,” Combustion Science
sion Ignition Engine Modeling,” Proc. Combust. and Technology 158, 15–33.
Inst. 34, 401–409. 112. V. Bychkov, A. Petchenko, V. Akkerman, L.-E.
91. D. Poinsot, T. & Veynante, D. 2005 Theoretical and Eriksson, 2005 Theory and Modeling of Accelerating
Numerical Combustion. 2nd Edition R.T. Edwards, Flames in Tubes, Phys. Rev. E 72, 046307.
Inc., Flourtown. 113. V.V. Bychkov, S.M. Golberg, M.A. Liberman,
92. Warnatz, J., Maas, U. & Dibble, R. W. 2001 Com- L.E. Eriksson, 1996 Propagation of Curved Flames
bustion: Physical and Chemical Fundamentals, in Tubes, Phys. Rev. E 54, 3713.
Modeling and Simulation, Experiments, Pollutant 114. Candel, S. 2002 Combustion dynamics and control.
Formation. Springer-Verlag. Proc. Combust. Inst. 29, 1–28.
93. Lu T.F., Law C.K., 2006 “On the Applicability of 115. Rayleigh, Nature 1878 #18 p 379
Directed Relation Graph to the Reduction of 116. Rijke, P.L. 1859, “Notiz über eine neue Art, die in
Reaction Mechanisms,” Combust. Flame, einer an beiden Enden offenen Röhre enthaltene Luft
146, 472–483. in Schwingungen zu versetzen” (Note on a new way
94. Lu T. F., Law C.K., 2009 “Toward Accommodating to create oscillation of the air contained in a tube
Realistic Fuel Chemistry in Large-Scale Computa- with both ends open), Annalen der Physik
tion,” Prog. Energy Combust. Sci., 35, 192–215. 183, 339–343.
95. Lu T.F., Ju Y., Law C.K., 2001 “Complex CSP for 117. Lee, J. H. S. 2001 Detonation waves in gaseous
Chemistry Reduction and Analysis,” Combust. explosives. Handbook of Shock Waves, (eds.
Flame, 126, 1445–1455. Ben-Dor, G., Igra, O. & Elperin, T.), Chapter 17,
96. Pope, S. B., 1997 Computationally efficient imple- Volume 3. Academic.
mentation of combustion chemistry using in situ 118. Döering, W. 1943 On detonation processes in gases.
adaptive tabulation. Combust. Theory Modeling Ann. Phys. 43, 421–436.
1, 41–63. 119. Fickett, W. & Davis, W.C. 1979 Detonation, Univer-
97. Bradley, D. 1992 “How fast can we burn?” Proc. sity of California Press.
Combust. Inst. 24, 247–262. 120. Sharpe, G.J. 2000 The structure of planar and curved
98. Peters, N. 2000 Turbulent Combustion. Cambridge. detonation waves with reversible reactions. Phys.
99. Clavin, P. 2000 Dynamics of combustion fronts in Fluids 12, 3007–3020.
premixed gases: From flames to detonations. Proc. 121. Taylor, G. I. 1950 The dynamics of the combustion
Combust. Instit. 28, 569–585. products behind plane and spherical detonation front
100. Palm-Leis, A. & Strehlow, R.A. 1969 On the propa- in explosives. Proc. R. Soc. Lond. A 200, 235–247.
gation of turbulent flames. Combust. Flame 122. Urtiew, P. A. & Oppenheim, A. K. 1966 Experimen-
13, 111–129. tal observations of the transition to detonation in an
101. Bray, K. N. C. 1980 Turbulent flows with premixed explosive gas. Proc. Roy. Soc. A 295, 13–28.
reactants. Turbulent Reacting Flows (ed. Libby, 123. von Neumann, J. 1942 Theory of detonation waves.
P. A. & Williams, F. A.). pp. 115–183. Springer- John von Neumann, Collected Works 6, (ed. Taub,
Verlag. A. J.), Macmillan, 1963.
102. Bray, K. N. C. 1996 The challenge of turbulent 124. Zel’dovich, Y. B. 1940 On the theory of propagation
combustion. Proc. Combust. Inst. 26, 1–26. of detonation in gaseous systems. Zhur. Eksp. Teor.
103. Bray, K. N. C., Libby, P. A., Masuya, G. & Moss, Fiz. 10: 542–568. (English translation: NACA TM
J. B. 1981 Turbulence production in premixed turbu- 1261,1950).
lent flames. Combust. Sci. Technol. 25, 127–140. 125. Zel’dovich, Y. B. & Kompaneets. A.S. 1960 Theory
104. Borghi, R. 1988 Turbulent combustion modeling. of Detonation, Academic.
Prog. Energy Combust. Sci. 14, 245–292. 126. http://metamodern.com/2009/12/29/theres-plenty-of-
105. Libby, P. A. & Williams, F. A. 1980 Turbulent room-at-the-bottom%E2%80%9D-feynman-1959/
Reacting Flows. Springer-Verlag. 127. Sharpe, G.J. 1999 Linear Stability of pathological-
106. Libby, P. A. & Williams, F. A. 1994 Turbulent detonations. J. Fluid Mech. 401, 311–338.
Reacting Flows. Academic. 128. Bychkov and Akkerman, Phys. Rev. E 73, 066305
107. Poinsot, T. 1996 Using direct numerical simulation (2006).
to understand premixed turbulent combustion. Proc. 129. Bychkov, et al Phys. Rev. Lett. 101, 164501 (2008)
Combust. Inst. 26, 219–232. 1417
108. Pope, S. B. 1985 PDF method for turbulent reactive 130. Valiev, et al Phys. Rev. E 80 036317 (2009).
flows. Prog. Energy Combust. Sci. 11, 119–192. 131. Valiev, et al Combust. Flame 157, 1012 (2010).
132. Searby, Combust. Sci. Techn. 81: 221 (1992).
www.ebook777.com
133. Searby and Rochwerger, J. Fluid. Mech. 231: 529 Expanding and Counterflow Laminar Flames Using
(1991). Direct Measurements and Numerical Simulations,”
134. Bychkov Phys. Fluids 11: 3168 (1999). 1423 Proceedings of the Combustion Institute 35, XXX–
135. Petchenko et al., Phys. Rev. Lett. 97: 164501 (2006). XXX.
136. Petchenko et al., Combust. Flame 149: 418 (2007).
137. J. Jayachadran, A. Lefebvre, R. Zhao, F. Halter,
Emilien Varea,2 B. Renou, F. N. Egolfopoulos, Grunde Jomaas is an Associate Professor at the
2015”Study of Propagation of Spherically Technical University of Denmark.
Fire Plumes, Flame Height, and Air

Entrainment 13
Gunnar Heskestad
Introduction fire heating of building structures, smoke fill-

ing rates, fire venting, and so forth. They can
Practically all fires go through an important, ini- also be important in fire suppression system
tial stage in which a coherent, buoyant gas stream design.
rises above a localized volume undergoing com- This chapter deals primarily with axisymmet-
bustion into surrounding space of essentially ric, turbulent fire plumes and reviews a number
uncontaminated air. This stage begins at ignition, of relations for predicting the properties of such
continues through a possible smoldering interval, plumes. It is assumed throughout the chapter that
into a flaming interval, and may be said to end the surrounding air is uncontaminated by fire
prior to flashover. The buoyant gas stream is products and that it is uniform in temperature,
generally turbulent, except when the fire source except where specifically treated as temperature
is very small. The buoyant flow, including any stratified. Release of gas from a pressurized
flames, is referred to as a fire plume. source is assumed to be vertical unless noted
Combustion may be the result of pyrolysis of otherwise (subsection on “Jet Flames in Horizon-
solid materials or evaporation of liquids because tal Discharge”). The relations cease to be valid
of heat feedback from the combustion volume or at elevations where the plume enters a smoke
of pressurized release of flammable gas. Other layer.
combustion situations may involve discharge of Main topics are flame heights, plume
liquid sprays and aerosols, both liquid and solid, temperatures and velocities, virtual origin, air
but these will not be discussed here. In the case of entrainment, and effects of ambient tempera-
high-pressure releases, the momentum of the ture stratifications. At the end of the chapter, a
release may be important. Flames in these few additional aspects of fire plumes are
situations are usually referred to as diffusion discussed, including flame pulsations, rise
flames, being the result of combustible vapor or times of plume fronts propagating upward
gas mixing or diffusing into an ambient oxidant, from suddenly initiated fires, and
usually air, as opposed to being premixed with an nonaxisymmetric plume situations including
oxidant. wall/corner effects, wind effects, and jet diffu-
Properties of fire plumes are important in sion flames from horizontal gas discharge.
dealing with problems related to fire detection, Knowledge of the behavior of fire plumes
presented in this chapter is based mostly on
observations and dimensional analysis,
complemented by analytical models.
G. Heskestad (*)
Retired from FM Global in 2004 as a Consulting
Research Scientist

www.ebook777.com
13 Fire Plumes, Flame Height, and Air Entrainment 397
Earlier methods were based on Reynolds-

Note on Numerical Modeling Averaged Navier Stokes (RANS) models and
are included in a survey of fire models prepared
Methods of numerical analysis are now showing by Olenick and Carpenter [10].
promise in modeling turbulent fires, e.g., the The ultimate approach is to solve the
Fire Dynamics Simulator (FDS). FDS is based governing equations more fundamentally with
on large eddy simulation (LES) of the turbulent Direct Numerical Simulation (DNS), which is
convective motion [1, 2] and requires a combus- becoming gradually more feasible for fire
tion model. Recent FDS contributions for fire problems with the availability of increasing
plumes incorporate subgrid mixture-fraction computing power. At this writing, DNS still
based combustion models and include Ma remains a great challenge and is not likely to
and Quintiere [3], who compared general become a viable tool in fire applications for
predictive capabilities to selected literature some time because of its demand on resources.
data characterized as “the best”, and Xin Numerical modeling may eventually repro-
et al. (two separate studies) who compared duce actual experiments, with a bonus of
detailed predictions to careful measurements in numerous variables of the mean and turbulent
the base regions up to an estimated one fourth flow fields hardly accessible by physical
of the flame height of a 7.1-cm diameter methane experiments.
burner [4] and a 1-m diameter methane fire [5]
(measurements by Tieszen et al. [6]). The chal-
lenge is to reproduce experimental results to Fire Plume Features
a high degree of accuracy, which was achieved
to a considerable extent in the base-region Figure 13.1 shows a schematic representation of
calculations by Xin et al. [4, 5], although the a turbulent fire plume originating at a flaming
temperature calculations required an assumption source, which may be solid or liquid. Volatiles
about radiant heat loss. driven off from the combustible by heat feed-
Another approach is the LES solver, back from the fire mix with the surrounding air
FireFOAM [7], which is based on the open source and form a diffusion flame. Laboratory
code OpenFOAM [8] and also incorporates a simulations often employ controlled release of
subgrid mixture-fraction based combustion flammable gas through a horizontal, porous sur-
model. The code has been applied by Wang face. The mean height of the flame is L.
et al. [9] to model fire plumes on a 0.3-m square Surrounding the flame and extending upward is
burner for comparison with existing data. With a boundary (broken lines) that confines the
an assumption of a constant radiant fraction of entire buoyant flow of combustion products
20 % (although the code allows for a more fun- and entrained air. The air is entrained across
damental treatment of thermal radiation), results this boundary, which instantaneously is very
of axial plume temperatures and velocities are in sharp, highly convoluted, and easily discernible
approximate agreement with the experiments. in smoky fires. The flow profile could be the
Flame heights and air entrainment compare time-averaged temperature rise above the ambi-
favorably with experimental values. Fluctuating ent temperature, the concentration of a gas (such
velocities are comparable to reported experi- as CO2) generated by the fire, or the axial veloc-
mental values, whereas fluctuating temperatures ity in the fire plume.
are excessive. (The long-term goal is stated to be Figure 13.1 suggests qualitatively, based on
predictive capability for large scale industrial experimental observations [12–16], how the
fires and water-based suppression on the temperature rise on the centerline, ΔT0, and
OpenFOAM platform, incorporating pyrolysis the velocity on the centerline, u0, might
and soot models, already available capabilities behave along the plume axis. In this example
in FireFOAM.) of a relatively tall flame, the temperatures are
398 G. Heskestad
Fig. 13.1 Features of a Z Z

turbulent fire plume,
including axial variations
on the centerline of mean
excess temperature, ΔT0,
and mean velocity, u0 [11] ΔT0 u0
Entrained
flow Flow
profile
Flame L
ΔT0;u 0
0
nearly constant in the lower portion of the For large fires, the radiative fraction tends to
flame. Temperatures begin to decay in the decrease with increasing fire size [19] and the
intermittent, upper portion of the flame as the convective fraction tends to increase. This
combustion reactions trail off and air entrained result is especially striking in smoky hydrocar-
from the surroundings cools the flow. The bon fires [20] where radiation is increasingly
centerline velocities, u0, tend to have their absorbed in the mantle of combustion products
maxima slightly below the mean flame height surrounding the flames (dashed lines in
and always decay toward higher elevations. If Fig. 13.1) as the fire increases in size, although
the combustible is porous and supports inter- the smoke itself will emit some radiation.
nal combustion, there may not be as pro- However, even mantles of combustion products
nounced a falloff in the gas velocity toward from sufficiently large fires of a clean burning
the top of the combustible as suggested in fuel, such as liquefied natural gas (LNG), will
Fig. 13.1 [14]. absorb significant and increasing amounts of
The total (or chemical) heat release rate of radiation as the fires become larger because of
a fire source, Q, _ is either convected, Q_ c , or water vapor and carbon dioxide, both major
radiated, Q_ r , away from the combustion region. combustion products and radiation-absorbing
In a fire deep in a porous combustible pile (e.g., a gases. Another reason for decreasing radiative
stack of wood pallets), some of the total heat fraction with increasing size of larger fires is
generated is trapped by and stored in the not yet that the flames are optically thick and the
burning material; the rest escapes from the com- radiation-emitting area (flame area) per unit
bustible array as either convective or radiative fire power (kW) decreases with increasing fire
energy flux. If most of the volatiles released size (based on Equation 13.8 and constant heat
undergo combustion above the fuel array, as in release rate per unit fire area).
pool fires of liquids and other horizontal-surface The total heat release rate, Q, _ is often
fires, and even in well-developed porous pile assumed to be equal to the theoretical heat
fires, then the convective fraction of the total release rate, which is based on complete com-
heat release rate is rarely measured at less than bustion of the burning material. The theoretical
60–70 % of the total heat release rate [17, heat release rate in kW is evaluated as the mass
18]. The convective flux, Q_ c , is carried away by burning rate in kg/s multiplied by the lower heat of
the plume above the flames, while the remainder complete combustion in kJ/kg. The ratio of the
of the total heat liberated, Q_ r , is radiated away in
total heat release rate to the theoretical heat
release rate, which is the combustion efficiency,
all directions.
www.ebook777.com
is indeed close to unity for some fire sources (e.g., Typically, the luminosity of the lower part of
methanol and heptane pools) [17], but may devi- the flaming region appears fairly steady, while
ate significantly from unity for others (e.g., a the upper part appears to be intermittent. Some-
polystyrene fire, for which a combustion effi- times vortex structures, more or less pronounced,
ciency of about 45 % [18] has been measured, can be observed to form near the base of the
and a fully involved stack of wood pallets, for flame and shed upward. [22, 23].
which a combustion efficiency has been measured Figure 13.2 helps to define the mean flame
at 63 % [17]). height, L [23]. It shows schematically the variation
of flame intermittency, I, versus distance above
the fire source, z, where I(z) is defined as the
Calculation Methods fraction of time that at least part of the flame lies
above the elevation, z. The intermittency decreases
This section presents calculation methods for from unity deep in the flame to smaller values in
flame height; plume temperature and velocity; the intermittent flame region, eventually reaching
effects of ambient temperature stratification zero. The mean flame height, L, is the distance
on plume temperature, velocity, and other above the fire source where the intermittency has
plume variables; virtual origin of a fire declined to 0.5. Objective determinations of
plume, a point source from which the plume mean flame height according to intermittency
appears to originate; and entrainment of air measurements are fairly consistent with (although
from the surroundings into a fire plume. The tending to be slightly lower than) flame heights
calculation methods are illustrated by that are averaged by the human eye [23].
examples. The mean flame height is an important quan-
tity that marks the level where the combustion
reactions are essentially complete and the inert
Flame Heights plume can be considered to begin. Several
expressions for mean flame height have been
The visible flames above a fire source contain the proposed. Figure 13.3, taken from McCaffrey
combustion reactions. Tamanini [21] has [24], shows normalized flame heights, L/D, as a
function of a Froude number, Q_ (represented as
investigated the manner in which combustion *
approaches completion with respect to height in

Q_
*2=5
diffusion flames. to compress the horizontal scale), from
Fig. 13.2 Definition by

Zukoski et al. [23] of mean
flame height, L, from 1.0
measurements of
intermittency, I
I 0.5
0 z (arbitrary units)
400 G. Heskestad
Fig. 13.3 Flame height Q*

correlations compiled by
10 –2 100 10 2 104 105
McCaffrey [24]. Capital 1000
letters without subscripts C3H8 +
correspond to various W
K
researchers as follows: +
+ H2
B Becker and Liang [25],
+
C Cox and Chitty [26], B K
H Heskestad [27], 100
K Kalghatgi [28], W
S Steward [29], T Thomas T Jet
[30], W Hawthorne flames
et al. [31], and Z Zukoski
L/D
[32]. Capital letters with 10 S
subscripts represent Pool
chemical formulae fires
1 B
Z
C
0.1
0.1 1 10 100 1000
Q *2/5
data correlations available in the literature. This flow rate and is several hundred times the size
Froude number is defined as of the source diameter.”
_ ¼ Q_ Buoyancy Regime The correlation by Heskestad

Q* pffiffiffiffiffiffi ð13:1Þ
ρ1 c p T 1 gDD2 _
(H) represented in Fig. 13.3 covers the entire Q*
range except the momentum regime and has
where
the following form given by McCaffrey [24]:
Q_ ¼ Total heat release rateðgiven in terms of
L
¼ 1:02 þ 3:7Q_
*2=5
the mass burning rate, m_ f , as m_ f H c Þ ð13:2Þ
ρ1 and T 1 ¼ Ambient density and temperature, D
respectively Actually, this correlation was originally
c p ¼ Specific heat of air at constant pressure presented in the form [27]
g ¼ Acceleration of gravity
D ¼ Diameter of the fire source L
¼ 1:02 þ 15:6N 1=5 ð13:3Þ
D
Quoting McCaffrey with respect to this figure:
“On the left are pool-configured fires with flame As before, D is the diameter of the fire source
heights of the same order of magnitude as the (or effective diameter for noncircular fire
base dimension D. In the middle is the interme- sources such that πD2/4 ¼ area of fire source)
diate regime where all flames are similar and the and N is the nondimensional parameter defined
Q_
*2=5
is seen as a 45 degree line in the figure. by
Finally, in the upper right is the high Froude " # 2
number, high-momentum jet flame regime c pT1 Q_
N¼ 3
ð13:4Þ
where flame height ceases to vary with fuel gρ21 ðH c =r Þ D5
www.ebook777.com
where Hc is the actual lower heat of combustion values calculated according to Equation 13.3
and r is the actual mass stoichiometric ratio of air may be associated with subjective flame-height
to volatiles. averaging by eye or photography (including
It is readily shown that N and Q*_ are related as assigning a value for the instantaneous flame
follows: height when detached flame fragments appear),
departure of fire area from circular, nonuniform
c p T 1 3 _ *2 volatilization rates across a burning surface
N¼ Q ð13:5Þ
Hc =r (as observed in Wood et al. [40]), density of
volatiles, effects of the surrounding topography
This equation, combined with Equation 13.3, and wind, and effects of flame Reynolds number
leads to Equation 13.2 when typical values are (increases with fire size).
substituted for the environmental and fuel Some large scale pool fires of LNG on water as
variables. well as on land, reviewed by Mudan [41], appear to
The parameter N was derived specifically by have produced flame heights generally a little lower
consideration of the flaming region [27], whereas than, but still quite consistent with Equation 13.3
_ was originally derived by Zukoski [33] from
Q* for both water and land sites and from two different
analysis of the nonreacting turbulent plume. Sub- investigations. Flame heights from land
sequently, Heskestad [34] presented results of measurements of a third investigation had great
flame height measurements at widely varying scatter, with a smallest flame height value about
ambient temperatures. The parameter Q* _ did one third as large as calculated from Equation 13.3.
not account correctly for the observed variations Subsequent to its derivation, the Equation 13.3
in flame height (increasing flame height with was found also to represent large, deep storages
increasing ambient temperature), while the when the flames extended above the storage and
parameter N did. For that reason, N is considered flame heights were measured above the base of
the more appropriate scaling parameter. the fire (bottom of storage in the experiments)
Equation 13.3 [27] is anchored in the 50 % [42]. The storages investigated included 4.5-m-
intermittency flame heights measured by high palletized storage of different commodities,
Zukoski et al. [23] for 81 experimental fires of 3- to 6-m-high rack storage of two different
natural gas (essentially methane), burned on cir- commodities, and wood pallets stacked
cular porous beds of three diameters, 0.10, 0.19, 0.3–3.3 m high. In these cases the fire diameter
and 0.50 m. The author extended this body of was calculated as the diameter of a fire area equal
data into the low-N region based on propane fires to the ratio of heat release rate to heat release rate
on a 0.76-m diameter sand bed [35]. Other per unit area.
contributing data [27, 35] were based on time- A convenient form of Equation 13.3 can be
averaged flame heights, mostly by eye, including developed. Let
(but not limited to) 51 and 102 mm- diameter gas
" #1=5
jets of various fuels [36]; 102 and 279 mm- c pT1
diameter natural-gas sandbox burners [37]; pool A ¼ 15:6 ð13:6Þ
gρ21 ðHc =r Þ3
fires of gasoline varying from 0.3 to 23 m in
diameter [38]; pool fires of JP-4 fuel varying Then Equation 13.3 can be written in the dimen-
from 1 1 m to 10 10 m [39] (treated as circu- sional form
lar pools of diameters corresponding to their
L ¼ 1:02D þ AQ_
burn areas); sand filled, 1.52 m-diameter pans 2=5
ð13:7Þ
initially saturated with acetone or methanol
[40]; and pool fires of diameters 1.22, 1.74, and The coefficient, A, varies over a rather narrow
2.44 m of methanol, heptane, a silicone trans- range, associated with the fact that Hc/r, the heat
former fluid and a hydrocarbon transformer liberated per unit mass of air entering the com-
fluid [15]. Deviations of observed values from bustion reactions, does not vary appreciably
402 G. Heskestad
among various combustibles. For a large number Newman and Wieczorek [46] have examined
of gaseous and liquid fuels, Hc/r remains within “chemical flame heights”, defined by the ratio of
the range of 2900–3200 kJ/kg, for which the CO to CO2 yields decaying to the limit for well-
associated range of A under normal atmospheric ventilated fires of the combustible material. They
conditions (293 K, 760 mmHg) is 0.240–0.226 found good agreement with determinations of
(m kW–2/5), with a typical value of A ¼ 0.235. flame height based on flame luminosity. They
Hence, under normal atmospheric conditions claim certain attributes for this method, such as
the results not being subject to visual bias and
L ¼ 1:02D þ 0:235Q_
2=5
ð13:8Þ available in cases where visual observations are
not feasible (presumably as in cases of obscuring
(L and D in m; Q_ in kW).
smoke, where infrared imaging may also work).
Fairly common fuels that deviate significantly
from the cited range 0.240–0.226 for A include
EXAMPLE 1 Consider a 1.5-m-diameter pan
acetylene and hydrogen (0.211). In general, the
fire of methyl alcohol with a heat release inten-
coefficient A ¼ 0.235 in Equation 13.8 may be
sity of 500 kW/m2 of surface area. Normal atmo-
considered adequate unless actual values of Hc
spheric conditions prevail (760 mmHg, 293 K).
and r are known that indicate otherwise and/or
Calculate the mean flame height.
atmospheric conditions deviate significantly
from normal. While values of Hc/r for complete
SOLUTION Available values of the lower heat of
combustion have been used to calculate the
combustion (Hc ¼ 21,100 kJ/kg) and stoichiomet-
coefficients A, Huggett [43] found a number of
ric ratio (r ¼ 6.48) give Hc /r ¼ 3260 kJ/kg. With
years ago that the heat of combustion per unit
mass of oxygen consumed, and hence Hc/r for a this value for Hc/r substituted in Equation 13.6,
standard atmospheric composition, was little together with cp ¼ 1.00 kJ/kg K, T1 ¼ 293 K,
affected by incompleteness of combustion in a g ¼ 9.81 m/s2, and ρ1 ¼ 1.20 kg/m3, the coeffi-
number of test calculations.1 cient A is calculated as 0.223 (m kW–2/5). The total
Referring to any of the flame-height relations heat release rate is Q_ ¼ 500π1:52 =4 ¼ 884 kW.
in Equations 13.3, 13.7, and 13.8, it can be seen Equation 13.7 gives a mean flame height of L ¼
that negative flame heights are calculated for 1.02 · 1.5 + 0.223 · 8842/5 ¼ 1.83 m.
sufficiently small values of the heat release rate.
Of course, this situation is unphysical and the EXAMPLE 2 This example is similar to Exam-
correlation is not valid here. For pool fires, ple 1, except for new atmospheric conditions
there are indications that a single flaming area representative of Denver, Colorado, on a hot
breaks down into several zones when heat release day: 630 mmHg pressure and 310 K temperature.
rates decrease to the point where negative flame
height (L ) is calculated [35]. SOLUTION Using Equation 13.6, the new coef-
Grove and Quintiere [45] have developed new ficient, A, increases from 0.223 to 0.249 [most
correlations for flame height, including linear readily calculated from (310/293)3/5 (760/630)2/5
sources (not considered here). They present 0.223 ¼ 0.249, where the equation of state for a
_ rather than N.
their results in terms of Q* perfect gas has been used]. Using Equation 13.7,
the new flame height is L ¼ 2.23 m, increased
from 1.83 m for normal atmospheric conditions.
1
As a further aid in assessing variations in A, Tewarson
[44], in his Table 3–4.12, lists values of ΔHO for complete Note Assuming that gas radiation is dominant in
combustion of many fuels, the lower heat of combustion heat transfer to the pool surface in the example,
per unit mass of oxygen consumed. From these values, that the absorption coefficient varies linearly
Hc/r (kJ/kg) can be easily calculated, the lower heat of
combustion per unit mass of air (of standard composition) with atmospheric pressure [47] and using the
consumed and, hence, the coefficient A. methanol pool fire data of Burgess et al. [102],
www.ebook777.com
the burning rate associated with the reduced Heskestad’s work [53] was based on an
atmospheric pressure is estimated to decrease extension of the author’s correlation for
by a negligible amount. buoyancy-controlled turbulent diffusion flames.
A momentum parameter is defined, which is the
EXAMPLE 3 One 1.2-m-high stack of wood ratio of gas release momentum to the momentum
pallets (1.07 1.07 m) burns at a total heat generated by a purely buoyant diffusion flame:
release rate of 2600 kW under normal atmo- " #
spheric conditions. Calculate the mean flame T1 c p ΔT L 4=5 ρ1 =ρs 2=5
RM ¼ 1:36 N
height above the base of the pallet stack. TL ðH c =r Þ r2
ð13:9Þ
SOLUTION The square flaming area can Here, TL and ΔTL are the plume centerline tem-
be converted to an equivalent diameter: perature and excess temperature (above ambi-
πD2/4 ¼ 1.072, which gives the equivalent ent), respectively, at the mean flame height of
diameter, D, of 1.21 m. Since the combustion purely buoyant diffusion flames, and ρs is the
efficiency of wood is considerably less than density of the source gas in the discharge stream.
100 %, it is difficult to select reliable and con- A value of 500 K is assigned [53] to ΔTL. Note
sistent values for Hc and r to form the ratio Hc/r. that the first two sets of parentheses are nearly
Instead, it can be assumed that A ¼ 0.235, the constant for normal ambient temperatures and
typical value. Using Equation 13.7, the mean fuels with comparable values of (Hc/r). Under
flame height above the base of the pallet stack these circumstances, the momentum parameter
is calculated as L ¼ 1.02 · 1.21 + 0.235 · is closely linked to the parameter N but is
26002/5 ¼ 4.22 m. affected quite significantly by the source gas
density at the discharge conditions as well as
Momentum Regime In Fig. 13.3 it is seen the mass stoichiometric ratio. If the gas discharge
that at high values of Q* _ the normalized is sonic or choked, the density of the source gas
flame heights begin to level off and eventually can be considerably higher than is the case at
attain constant values, but not at the same atmospheric pressure.
value of Q* _ and not at the same normalized Figure 13.4 presents flame heights of jet dif-
flame height. fusion flames in the form L/LB versus RM, where
Flame heights of vertical turbulent jet flames L is the flame height reported by various
have been studied by a number of investigators investigators and LB is the buoyancy-controlled
reviewed by Blake and McDonald [48, 49], who flame height according to Equation 13.3. The
proposed a new correlation of normalized flame data scatter about a value L/LB ¼ 1.2, approxi-
heights versus a “density-weighted Froude num- mately, for RM < 0.1. At higher values of RM,
ber.” Although an improvement over previous the flame height ratio approaches an asymptotic
work, the correlation still exhibits significant slope of 1/2, indicated by a dashed line. The
scatter. At about the same time, Delichatsios associated values of N are so large that we can
[50] proposed an alternative approach. Previ- take LB/D/N1/5 (see Equation 13.3), which
ously, other authors had proposed flame height together with Equation 13.9 imply that L/D is
relations, including Becker and coworkers [25, constant when the slope is equal to 1/2 (for
51], and Peters and Göttgens [52]. Subsequent to constant source gas and discharge density). Con-
these publications, Heskestad [53] also consid- stant slope and constant L/D (for a given gas and
ered the high-momentum regime, especially with density) appear to be achieved reasonably
respect to defining an unambiguous transition to quickly above RM ¼ 0.1.
momentum control and flame heights in this The fact that the low-RM flame height ratios in
regime. Fig. 13.4 tend to scatter about a level higher than
404 G. Heskestad
Fig. 13.4 Data on flame 10

heights of turbulent jet
diffusion flames in ratio to
the corresponding
buoyancy-controlled flame
heights, plotted versus the
ratio of gas release
momentum to buoyancy
momentum (from
L/LB
Heskestad [53]). Data 1
plotted as + pertain to
choked discharge of +
hydrogen ++
++
++
0.1
1.00E–04 1.00E–03 1.00E–02 1.00E–01 1.00E+00 1.00E+01
RM
unity has been attributed to several possible manner on the mass stoichiometric ratio and the
factors [53]. One of the two most important source gas density at discharge.
may be the working definition of mean flame It should be pointed out that the transition
height employed by some investigators, produc- to the momentum regime, RM ¼ 0.1, and the
ing greater values than the 50 % flame- flame height in the momentum regime,
intermittency height. Another may be retinal Equations 13.10 and 13.11, differ significantly
retention of flame images in visual averaging of from previously proposed relations, as discussed
rapidly pulsating flames (typical of the scales of in Heskestad [53].
the experiments), tending to make an observer
exaggerate the mean flame height. EXAMPLE 4 Calculate the normalized height
Above RM ¼ 0.1, adopting the dashed line in of a hydrogen jet flame from a 5-mm-dia-
Fig. 13.4 as representative of the momentum meter nozzle connected to a reservoir (tank,
regime, the normalized flame height is pipe, etc.) at ambient temperature of 293 K
" and a pressure of either (a) 150 kPa or
1=2 # 1=2
LM TL H c =r 2=5 ρs (b) 300 kPa.
¼ 5:42 r
D T1 c p ΔT L ρ1
SOLUTION
ð13:10Þ (a) The ratio of ambient pressure (101 kPa) to
the reservoir pressure (150 kPa) is 0.673,
where LM is the flame height in the momentum
corresponding to subsonic discharge (sonic
regime. For Hc/r ¼ 3100 kJ/kg (many common
discharge occurs at a pressure ratio of 0.528,
gases), ΔTL ¼ 500 K, and T1 ¼ 293 K, Equa-
as for air). The mass flow of hydrogen from
tion 13.10 becomes
the nozzle is calculated with the aid of a
1=2 compressible flow formula (e.g., Shapiro
LM ρ
¼ 18:5 s r ð13:11Þ [54]) as 1.74 g/s, using a ratio of specific
D ρ1
heats k ¼ 1.4 (as for air). Based on a heat of
In this case the nondimensional flame height in combustion of 120,000 kJ/kg, the heat
the momentum regime depends in a simple release rate is Q_ ¼ 209 kW. The source
www.ebook777.com
gas density in the discharge stream, ρs, is ¼ 6.93 104 N2/5 ¼ 0.67, indicating the
calculated from the source gas density at flame is in the momentum regime as in (a).
ambient temperature and pressure, ρs1, as The normalized flame height is calculated
follows: from Equation 13.10 as LM/D ¼ 239, some-
ρs ¼ ðρs =ρs1 Þρs1 ¼ ðρs =ρs0 Þðρs0 =ρs1 Þρs1 , what higher than for the lower discharge
where ρs0 is the source gas density in the gas pressure in (a).
reservoir. The density ratios can be expressed
in terms of pressure ratios with the result Note If the nozzles of cases (a) and (b) are
1=k 1 sharp-edged holes or openings instead, it is
ρs ¼ ð ps = ps0 Þ ð ps0 = p Þρs1 , where ps is
the pressure in the discharge stream (ambient recommended that the source diameter be
pressure for subsonic discharge) and ps0 is the multiplied by (discharge coefficient)1/2; see,
pressure of the gas reservoir. Finally, we for example, Shapiro [54] for values of the dis-
obtain ρs ¼ (101/150)1/1.4 (150/101) ρs1 ¼ charge coefficient of sharp-edged orifices in
1:12 0:083 ¼ 0:093 kg=m3 , where 0.083 compressible flow (varying from 0.60 near
(kg/m3) is the density of hydrogen at ambient incompressible flow conditions to 0.77 for
temperature and pressure. Now the momen- choked flow).
tum parameter can be calculated from Equa-
tion 13.9, taking ΔTL ¼ 500 K,
Hc ¼ 120,000 kJ/kg, r ¼ 34.3, and ρs Plume Temperatures and Velocities
¼ 0.093 kg/m3, yielding RM ¼ 1.16 103
N2/5. The parameter N is calculated from The first plume theories (Schmidt [55], Rouse
Equation 13.4 with the result N ¼ 6.76 et al. [56], Morton et al. [57]) assumed the
106, which results in RM ¼ 0.62 and following:
places the flame in the momentum regime. 1. Turbulent flow
The normalized flame height is calculated 2. Point source of buoyancy
from Equation 13.10 as LM/D ¼ 185. 3. Variations of density in field of motion small
NOTE: The calculated height may include compared to ambient density
the visual-averaging bias toward somewhat 4. Air entrainment velocity at the edge of the
higher than actual values built into the plume proportional to the local vertical
database. plume velocity
(b) The ratio of ambient pressure (101 kPa) to 5. Profiles of vertical velocity and buoyancy
reservoir pressure (300 kPa) is 0.337, force in horizontal sections of similar form
corresponding to sonic, or choked, dis- at all heights
charge. The mass flow of hydrogen from Morton et al. [57] developed an integral for-
the nozzle is calculated with the aid of an mulation on the further assumption that the
appropriate compressible flow formula for profiles are uniform “top hat” profiles. The
choked discharge (e.g., Shapiro [54]) as mean motion is then governed by the following
3.65 g/s, corresponding to a heat release three conservation equations for continuity,
rate of 436 kW. The source gas density momentum, and buoyancy:
for choked flow is calculated as in (a), d 2
Continuity : b u ¼ 2αbu ð13:12Þ
except the ratio ( ps/ps0) is set equal to the dz
value for a Mach number of unity, 0.528; d 2 2 ρ ρ
that is, ρs ¼ 0:5281=k ðρs0 =ρs1 Þ 0:083 ¼ Momentum : b u ¼ b2 g 1
dz ρ1
0:634ð300=101Þ 0:083 ¼ 0:156 kg=m3 .
ð13:13Þ
The parameter N is calculated from Equa-
tion 13.4 as 294 107 and the momentum d 2 ρ1 ρ
Buoyancy : b ug ¼ 0 ð13:14Þ
parameter from Equation 13.9 as RM dz ρ1
406 G. Heskestad
In these equations, z is the elevation above the 13.19, and 13.20 is replaced by z z0. In addi-
point source of buoyancy; b is the radius to the tion, to accommodate large density deficiencies
edge of the plume; u is the vertical velocity in as are present in fire plumes, Morton’s extension
the plume; α is the entrainment coefficient (the of the weak-plume theory [59] leads to the result
proportionality constant relating the inflow that Δρ/ρ1 in Equation 13.20 should be replaced
velocity due to entrainment at the edge of the by Δρ/ρ (¼ ΔT/T1 using the ideal gas law).
plume to u); ρ is the density in the plume; and ρ1 Also, Equation 13.18 for growth in plume
is the ambient density. Equation 13.14 can be radius should incorporate the additional factor

integrated immediately to ðρ1 =ρÞ1=2 ¼ ðT=T 1 Þ1=2 using the ideal gas
ρ1 ρ law] on the right side of the equation. Relaxing
b2 ug ¼ B ¼ constant ð13:15Þ
ρ1 the assumption that the flow profiles are uniform
renders the numerical coefficients in the resulting
Here, B is the buoyancy flux in the plume which equations in doubt.
remains constant at all heights. The flux can be Measurements in fire plumes above the flames
related to the convective heat in the plume, Q_ c, by have to a large extent supported the theory. The
noting plume radius and centerline values of mean
excess temperature and mean velocity have
Q_ c ¼ ρuπb2 c p ðT T 1 Þ been found [11] to obey the following relations:
¼ πub2 c p ðρ1 ρÞT 1 ð13:16Þ 1=2
T0
bΔT ¼ 0:12 ðz z0 Þ ð13:21Þ
where the ideal gas law has been used. In T1
this equation, T is the plume temperature and !1=3
T1 is the ambient temperature. Combining T1
Q_ c ðz z0 Þ5=3
2=3
ΔT 0 ¼ 9:1
Equations 13.15 and 13.16 gives gc2ρ ρ21
1
B ¼ g πc p T 1 ρ1 Q_ c ð13:17Þ ð13:22Þ
1=3
Solutions to Equations 13.12, 13.13, and g
Q_ c ðz z0 Þ1=3
1=3
u0 ¼ 3:4
13.15 can be determined [57] in the form c p ρ1 T 1
(expressing B in terms of Q_ c using ð13:23Þ
Equation 13.17)
Here, bΔT is the plume radius to the point where
6α the temperature rise has declined to 0.5ΔT0; T0 is
b¼ z ð13:18Þ
5 the centerline temperature, Q_ c is the convective
1=3 heat release rate, z is the elevation above the fire
5 9 1=3 1=3 1=3
u¼ 2
g1=3 c p ρ1 T 1 Q_ c z source, and z0 is the elevation of the virtual origin
6 10πα above the fire source.2 (If z0 is negative, the
ð13:19Þ virtual origin lies below the top of the fire
source).
1=3
Δρ 5 9π2 α4 2=3 2=3 5=3 The virtual origin is the equivalent point
¼ g1=3 c p ρ1 T 1 Q_ c z
ρ1 6 10 source height of a finite area fire. This origin is
ð13:20Þ
Equations 13.18, 13.19, and 13.20 are the 2

For normal atmospheric conditions (T1 ¼ 293 K, g
weak plume (small density deficiency) relations ¼ 9.81 m/s2, cp ¼ 1.00 kJ/kg K, ρ1 ¼ 1.2 kg/m3), the
h i1=3
for point sources. To account for area sources, a factor 9:1 T 1 = ðgc2p p21 has the numerical value
virtual source location or virtual origin, z0, is 25.0 K m5/3 kW–2/3, and the factor 3.4[g/(cpρ1T1)]1/3 has
introduced [57, 58] and z in Equations 13.18, the numerical value 1.03 m4/3 s1 kW–1/3.
www.ebook777.com
usually located near the fuel surface for pool fires 0.86 [56], 1.00 [61], 1.08 and 1.24 [62], 1.31
and may be assumed coincident with the fuel [16], 1.05 [12], and 1.5 [13]. The widely
surface when the plume flow is predicted at differing results can probably be attributed to
high elevations. Near the fire source, however, the difficulty of positioning the measuring probes
it is important to know the location of the virtual accurately with respect to the plume centerline
origin for accurate predictions. Calculation of the and to different, intrinsic errors associated with
virtual origin is discussed in the following sec- the diverse types of anemometers used (pitot
tion for both pool fires and three-dimensional tube, bidirectional flow probe, hot wire, vane
storage arrays. anemometer, cross-correlation techniques, laser
Equations 13.21, 13.22, and 13.23 are known Doppler anemometer).
as the strong plume relations. The numerical Often profiles of temperature rise and velocity
coefficients for the relations have been determined are represented as Gaussian in shape, although
from data sets for which the locations of the virtual there is no theoretical foundation for this
origin, z0, have been established and the convec- distribution.
tive heat release rates, Q_ c , are known [15, 60]. " #
We may compare the experimentally derived R 2
ΔT ¼ ΔT 0 exp ð13:24Þ
numerical coefficients in Equations 13.21, 13.22, σ ΔT
and 13.23 to the theoretical coefficients indicated
" #
in Equations 13.18, 13.19, and 13.20, which are R 2
based on the integral theory of Morton et al. [57] u ¼ u0 exp ð13:25Þ
σu
for weak plumes, point sources, and top hat
profiles. Forcing equality between the predictions
Here, ΔT and u are the local values, at the radius,
of Equations 13.22 and 13.20 (interpreting Δρ/ρ1
R, in the plume of temperature rise and gas
as ΔT/T1), we obtain a value for the entrainment
velocity. The quantities σΔT and σu are measures
coefficient of α ¼ 0.0964. With this value for α,
of the plume width, corresponding to the radii
the theoretical coefficient for centerline velocity
where local values of temperature rise and veloc-
in Equation 13.19 becomes 2.61, compared to the
ity are e 1 ¼ 0.368 multiplied by the centerline
experimental value 3.4 in Equation 13.23. The
values. For Gaussian profiles, the plume radii σΔT
theoretical coefficient for plume radius in
and σu are 1.201 multiplied by the plume radii,
Equation 13.18 becomes 0.116, compared to the
bΔT and bu, discussed previously.
experimental value 0.12 in Equation 13.21. There
Equations 13.21, 13.22, and 13.23 cease to be
is good consistency between the theoretical and
valid at and below the mean flame height.
experimental coefficients. However, the theoreti-
However, it is possible to represent ΔT0 such
cal expression for mass flow rate in a weak plume,
that a general plot of experimental temperature
generated from the product ρ1u(πb2) (using
variations is produced throughout the length
Equations 13.18 and 13.19) and the value for α
of the plume, including the flames. The method
above, produces a numerical coefficient that
is based on the observation that Q_ ðz z0 Þ5=3
2=3
is only 56 % of the coefficient based on c
experiments (see discussion of Equation 13.40 in Equation 13.22 can be written as
h i
2=5 5=3
later in the chapter). ðz z0 Þ=Q_ c . This result suggests plot-
In addition to the temperature radius of a
ting ΔT0 versus ðz z0 Þ=Q_ c . Figure 13.5
2=5
plume, bΔT, a velocity radius, bu, can also be
defined. The velocity radius is the plume radius shows the result in logarithmic coordinates for
to the point where the gas velocity has declined normal atmospheric conditions. For values of the
to 0.5 u0. The most reliable measurements [60] abscissa greater than 0.15–0.20 (m/kW2/5), the
indicate that bu is perhaps 10 % larger than bΔT. centerline temperature rise falls off with the 5=3
Other measurements indicate ratios bu/bΔT of power of the abscissa, in accordance with the
408 G. Heskestad
Fig. 13.5 Temperature 1000

rise on the plume centerline
of pool fires for normal
atmospheric conditions
[11] in a form attributable
to McCaffrey [12], and
Kung and Stavrianidis [15]
100
Slope = –5/3
ΔT0(K)
10
1
0.01 0.1 1 10
·
(Z – Z 0)/Qc2/5 (m·kW –2/5)
plume law for temperature (Equation 13.22). temperature, Equation 13.22, to produce the fol-
Abscissa values in the 0.15–0.20 range corre- lowing useful nondimensional parameter: [14]
spond to the mean flame height (see Equa- " 1=5 #
tion 13.37); an associated temperature rise of 1 c p ρ1
T 2=5 u0
ξ¼ 1=5 ð13:26Þ
about 500 K is indicated in Fig. 13.5. At smaller g2=5 ΔT 0 Q_ c
abscissa values, the experimentally observed
temperature rise increases more slowly, In the plume region where Equations 13.22
approaching a value deep in the flame of approx- and 13.23 are valid, their numerical coefficients
imately ΔT0 ¼ 900 K. When closer to the fuel correspond to a constant value ξ ¼ 2.2. This
surface than represented in Fig. 13.5, the value has been confirmed for a number of test
temperatures on the plume axis tend to decrease fires [14], at heights as low as the mean flame
again [12, 13, 16]. height and even somewhat lower. Equation 13.26
Figure 13.5 is based on the pool fire data of with ξ ¼ 2.2 is a useful relation for determining
Kung and Stavrianidis [15] of methanol, heptane, the maximum velocity in the plume, which
a silicone transformer fluid, and a hydrocarbon occurs slightly below the mean flame height
transformer fluid, for the three pool diameters where the temperature rise may be taken at
1.22, 1.74 and 2.44 m and using 1-mm diameter approximately ΔT0 ¼ 650 K. For normal atmo-
bare bead thermocouples. For the fires that pro- spheric conditions and the value ξ ¼ 2.2, Equa-
duced the highest values of temperature rise in tion 13.26 becomes
the deep-flame region, near 900 K, which include u0
1=5 ¼ 0:54 ð13:27Þ
all the fluids except the silicone fluid, the data ΔT 0 Q_ c
from the various combinations of fluid and pool
diameter were in good agreement. The maximum velocity just under the mean
The plume law for velocity, Equation 13.23, flame height, u0m, is obtained by setting
may be combined with the plume law for ΔT0 ¼ 650 K
www.ebook777.com
u0m ¼ 1:97Q_ c
1=5 u0m being overestimated, although the effect is
ð13:28Þ
probably not very large because of the slow, 1=5 th
power dependence on ΔT0).
Fires with low flame height-to-diameter The turbulence intensities in a fire plume are
ratios have not been investigated extensively
quite high. On the axis, George et al. [60] report
and may require special consideration. For one
an intensity of temperature fluctuations of
particular fire with very low flame height [15] approximately T0 /ΔT0 ¼ 0.38, where T0 is the
in which a proprietary silicone transformer
root mean square (rms) temperature fluctuation.
fluid was burned in a 2.44-m-diameter pool, a
Centerline values of the intensity of axial veloc-
flame height ratio of L/D ¼ 0.14 was measured3 ity fluctuations were measured near u0 /u0 ¼ 0.27
at a convective heat release rate of
by George et al. [60] and near u0 /u0 ¼ 0.33 by
Q_ c ¼ 327 kW. Using the results in the next Gengembre et al. [16], where u0 is the rms veloc-
section, the virtual origin is calculated at z0 ¼ ity fluctuation in the axial direction.
1.5 m, assuming Q_ c =Q_ ¼ 0:7. With respect
to the abscissa in Fig. 13.5, the lowest EXAMPLE 5 Example 1 concerned a 1.5-m-
possible value is z0 =Q_
2=5
corresponding diameter methyl alcohol fire burning under normal
c
to the fuel surface, z ¼ 0. For the present case, atmospheric conditions, generating Q_ ¼ 884 kW
with a calculated mean flame height of 1.83 m.
z0 =Q_
2=5
c ¼ 1:5=3272=5 ¼ 0:15 m kW2=5 . At
For an elevation of 5 m and given a virtual origin
this abscissa value, a centerline temperature rise
z0 ¼ 0.3 m (from Example 8), calculate the
of 580 K is indicated in Fig. 13.5. From the exper-
temperature radius, bΔT, as well as the centerline
iment [15], a near surface ΔT0 of 440 K can be
value of temperature rise, ΔT0, and gas velocity,
determined by slight extrapolation, fairly close to
u0. Also calculate the maximum gas velocity in
the prediction from Fig. 13.5. Fires with very low
the flame.
flame height-to-diameter ratios may generally be
expected to produce lower maximum mean
temperatures than other fires. However, it is not SOLUTION Assume4 Q_ c ¼ 0:8 Q_ and first
yet clear whether the type of prediction attempted calculate the temperature rise, using Equa-
here for a particular low L/D fire is generally valid. tion 13.22 and properties for normal atmospheric
There is also uncertainty associated with conditions (T1 ¼ 293 K, g ¼ 9.81 m/s2, cp
assuming that ξ in Equation 13.26 remains ¼ 1.00 kJ/kg K, ρ1 ¼ 1.20 kg/m3)
completely constant down to the flame level in 1=3
low L/D fires. It may be found that ξ still remains 293
ΔT 0 ¼ 9:1 ð0:8 884Þ2=3
approximately constant down to the height where 9:81 1:002 1:202
the maximum gas velocity occurs, although this ð5 þ 0:3Þ5=3 ¼ 123 K
maximum will probably occur above the flames.
The associated temperatures at this height cannot The temperature radius can now be calculated
as yet be predicted. Consequently, the relation in from Equation 13.21 as
Equation 13.28 becomes somewhat uncertain as

L/D decreases (ΔT0 is overestimated, resulting in 123 þ 293 1=2
bΔT ¼ 0:12 ð5 þ 0:3Þ ¼ 0:76 m
293
3
A ratio L/D ¼ 0.02 can be calculated from Equation 13.7
assuming Hc/r ¼ 3470 kJ/kg, an average for silicone oils
from values reported by Tewarson [63] and assuming a
convective heat fraction Q_ c =Q_ ¼ 0:7. If a value of Hc/r
near the bottom of the reported range [63] is selected,
4
Without specific knowledge, Q_ c =Q_ may usually be
3230 kJ/kg, the observed value L/D ¼ 0.14 is reproduced; assumed at 0.7. However, methyl alcohol produces a fire
slight changes in the assumed convective fraction will of low luminosity and radiation, for which Q_ c =Q_ ¼ 0:8 is
also reproduce the measured value. a good estimate.
410 G. Heskestad
The velocity is calculated from Equation 13.23 as increases in temperature with height, which is
1=3 normal in buildings, and the fire source is weak,
9:81 the temperature difference between the plume
u0 ¼ 3:4 ð0:8 884Þ1=3
1:00 1:20 293 and the ambient, which gives the plume buoy-
ð5 þ 0:3Þ1=3 ¼ 5:3 m=s ancy, may vanish and actually reverse in sign.
Eventually the plume ceases to rise.
Instead of Equation 13.23, the velocity can also The maximum height achieved by plumes
be calculated from Equation 13.26 in this case, in temperature-stratified space has been given
since ΔT0 is already known. Actually, because by Heskestad [64], based on pioneering theo-
normal ambient conditions prevail, Equa- retical and experimental work by Morton
tion 13.27 can be used as follows: et al. [57]
u0 ¼ 0:54ð123 0:8 884Þ1=5 ¼ 5:3 m=s " #1=8 3=8

T a1 _ 1=4 dT a
zm ¼ 3:79 2 Qc
Finally, the maximum velocity in the flame is g ρa1 c p dz
given by Equation 13.28 as ð13:29Þ
1=5
u0m ¼ 1:97ð0:8 884Þ ¼ 7:3 m=s Here, dTa/dz is the ambient temperature gradient,
EXAMPLE 6 Recalculate the quantities called Ta1 and ρa1 are the ambient temperature and
for in Example 5 using ambient conditions repre- density, respectively, at the level of the fire
sentative of Denver, Colorado, on a hot day: source, and the constant 3.79 traces to
630 mmHg pressure and 310 K temperature. experiments using dyed light liquid injected
(See Note on heat release rate for EXAMPLE 2.) into a density stratified salt solution [57]. Other
results are presented in Fig. 13.6, which shows
SOLUTION Changed ambient variables enter- the ratio on the plume centerline of stratified
ing the equations include T1 ¼ 310 K and ρ1 value versus unstratified value for various
¼ 0.78 kg/m3. From Equation 13.22, the new plume variables: temperature rise relative to the
temperature rise is pre-existing value at each level (curve A), axial
velocity (curve B), plume radius (curve C), and
ΔT 0 ¼ 167 Kðversus 123 K in Example 5Þ volume concentration of a combustion species
(curve D). The ratios are plotted against the frac-
The new velocity from Equation 13.23 is tion of maximum elevation achieved by the
u0 ¼ 6:0 m=sðversus 5:3 m=sÞ plume, z/zm. By definition, the stratified velocity
(B) decreases to zero at z/zm ¼ 1. The stratified
For the new ambient conditions, the relation temperature rise (A) becomes negative below the
analogous to Equation 13.28 is calculated as maximum reach. The stratified plume radius
(C) grows rapidly in approach to the maximum
u0m ¼ 2:10Q_ c
1=5
plume reach. However, there is little effect of the
stratification on the centerline variation of con-
from which the new maximum velocity in the
centration of a combustion species.
flame is
Fire experiments in temperature stratified
u0m ¼ 7:8 m=s ðversus 7:3 m=sÞ space [66] have largely supported the validity
of Fig. 13.6 for temperature rise (A) and volume
concentrations (D), except that the experimental
Plumes in Temperature-Stratified values needed an incremental height, roughly
Ambients equal to 25 % of the theoretical plume reach, to
return to zero.
When a buoyant, turbulent plume rises, it cools The maximum plume reach can be interpreted
by entrainment of ambient air. If the ambient air in terms of a critical ambient temperature rise
www.ebook777.com
Fig. 13.6 Theoretical

behavior of centerline C
plume variables in linearly 1.6
temperature-stratified
ambients. From Heskestad D
1.2
[64, 65], traceable to
Morton et al. [57] Curve A:
ratio of temperature rises 0.8
(stratified versus
unstratified). Curve B: ratio
of axial velocities. Curve 0.4
C: ratio of plume radii.
Curve D: ratio of B
volumetric species 0
concentrations 0.2 0.4 0.6 0.8 1.0
–0.4 Z/Zm
–0.8
–1.2 A
from the source level to an elevated observation SOLUTION The temperature rise in unstratified
plane, just strong enough to prevent plume fluid space is required as a reference and is calculated
from penetrating the plane. Experiments [66] from Equation 13.22, taking z0 ¼ 0 for simplic-
show that the critical ambient temperature rise ity since deviations of the virtual origin from the
for a linear profile is 7.4 times the centerline level of the fire can be assumed to be inconse-
temperature rise at the level of the observation quential in this case. We have
plane that results from a fire source in a uniform
ΔT 0 ¼ 25:0Q_ c z5=3
2=3
environment. Furthermore, the critical tempera- ð13:30Þ
ture rise is surprisingly insensitive to the shape of
the stratification profile. For a profile where The temperature rise of the stratification, 5 K, is
one-half of the ambient temperature rise to the 7.4 times the value of ΔT0 for this associated
observation plane occurred higher than 75 % of unstratified-space fire, which will just drive the
the elevation of the observation plane above the plume to the ceiling. Solving Equation 13.30 for
source, the critical ambient temperature rise was Q_ c, setting ΔT0 ¼ 5/7.4 K and z ¼ 20 m, we get
only 12 % greater than that for the linear profile. Q_ c ¼ 7:9 kW. Assuming a ratio of 0.7 for con-
Effects of ambient temperature stratifications vective in ratio to the total heat release rate,
have been considered anew by Watanabe and the latter is Q_ ¼ 11:4 kW. If the ceiling height
Tanaka [67]. They developed their own model is doubled to 40 m, the new result is
and tested it against experiments using a smol- Q_ ¼ 64 kW.
dering fire source.
EXAMPLE 7 Consider a 20-m-high atrium Virtual Origin

where the temperature rise, floor to ceiling, is
5 K. What heat release rate is required of a Pool Fires As pointed out earlier in this chapter,
floor-level fire to drive the plume to the ceiling? knowledge of the virtual origin of fire plumes is
What would be the effect of doubling the ceiling important for predicting the near-source plume
height? behavior. The virtual origin is a point source
412 G. Heskestad
from which the plume above the flames appears mean centerline temperature at the mean flame
to originate. height, TL. It appeared that F could be considered
The virtual origin of a test fire is most conve- a constant for wide variations in ambient temper-
niently determined from temperature data above ature and pressure but might be affected by wide
the flames along the plume axis. According to swings in the fuel variables, Hc/r, and convective
3=5 fraction. The available data did not reflect any
Equation 13.22, a plot of ΔT 0 versus z should
produce a straight line that intercepts the z-axis at sensitivity to fuel identity within their scatter and
z0. Despite this apparent simplicity of obtaining led to the determination F ¼ 0.083 m kW–2/5,
z0, the task is very difficult in practice. Slight with Equation 13.31 becoming
inaccuracies in the determinations of centerline
Q_ Q_ in kW, D in m
2=5
temperatures have large effects on the intercept, z0
¼ 1:02 þ 0:083
z0; such inaccuracies may be associated with D D
off-axis placement of sensors, radiation-induced ð13:32Þ
errors in the temperature signal, or inadequate
Later, Hasemi and Tokunaga [69] analyzed
averaging of the signal.
their temperature measurements in plumes from
Data obtained in this manner on the virtual
gas burners of diameters in the 0.2–0.5 m range
origin for pool fires varying in diameter from
and established alternative correlations for the
0.16 to 2.4 m [12, 14, 15], were examined for
virtual origin. In terms of the nondimensional
consistency with a theoretical model by _ defined in Equation 13.1, their
parameter Q*
Heskestad [68]. The model relied heavily on the
correlations are
flame-height correlation represented by Equa-
tion 13.3 and led to the prediction z0 *2=5
¼ 2:4 Q_ Q_ 1:0
*
1
D
Q_
2=5
z0 *2=3 ð13:33Þ
¼ 1:02 þ F ð13:31Þ z0
¼ 2:4 Q_ Q_
*2=5
Q_ < 1:0
*
D D D
where F is a rather complex dimensional func-
For normal ambient conditions, these
tion of environmental variables cp, T1, ρ1, g;
correlations can be written in terms of the
Hc/r for the combustible, the fraction of the total
variable Q_ =D (cf. Equation 13.32) as
2=5
heat release carried away by convection, and the
Q_ Q_
2=5 2=5
z0
¼ 2:4 þ 0:145 16:5
D D D
!5=3 ð13:34Þ
Q_ Q_ Q_
2=5 2=5 2=5
z0
¼ 0:0224 0:145 < 16:5
D D D D
Cetegen et al. [70] have proposed correlations burners located some distance above the floor
for the virtual origin on the basis of their air of the laboratory. Their correlations for the vir-
entrainment measurements in fire plumes and tual origin are
attempts to apply entrainment theory for a point
z0
¼ c þ 1:09Q_ Q_ > 1
source to the laboratory fires. Their experiments *2=5 *
involved gas burners (natural gas) with diameters D

ð13:35Þ
z0
of 0.10, 0.19, 0.30, and 0.50 m. The experiments ¼ c þ 1:09Q_ Q_ 1
*2=3 *
were performed with and without a floor D

mounted flush with the upper surface of the
www.ebook777.com
Fig. 13.7 Correlations 2 1 3 4

for the virtual origin 2
of pool fires. Curve 1:
Equation 13.32; Curve 2:
Equation 13.34; Curve 3:
Equation 13.36 with floor;
Curve 4: Equation 13.36 1
without floor
Z0 /D
0
–1
0 10 20 30 40
Q 2/5 /D [kW2/5 m–1]
where Q_ has been defined by Equation 13.1, and

*
Using Equation 13.1, Equation 13.35 can be
where c ¼ 0.50 with a flush floor around the written in terms of Q_ =D yielding
2=5
burners and c ¼ 0.80 without a flush floor.
Q_ Q_
2=5 2=5
z0
¼ c þ 0:0659 > 16:5
D D D
!5=3 ð13:36Þ
Q_ Q_
2=5 2=5
z0
¼ c þ 0:01015 16:5
D D D
z0 ¼ L 0:175Q_ c
2=5
where c ¼ 0.50 and c ¼ 0.80 with and with-
out a flush floor, respectively. ð13:37Þ
L and z0 in m; Q_ c in kW
Figure 13.7 is a composite plot of the various
correlations for the virtual origin of pool fires, In addition to representing pool fires, Equa-
plotted as z0/D versus Q_ =D. Despite the
2=5
tion 13.37 has also been verified to represent
diverse approaches, the overall correlations are deep storage fires [42], allowing the location of
surprisingly similar. Precise measurements are the virtual origin to be calculated from knowl-
not yet available to clearly identify an optimal edge of the mean flame height and the convective
correlation. In the meantime, curve 1 in Fig. 13.7 heat release rate. As discussed earlier, mean
(i.e., Equation 13.32) is recommended for its flame heights above the base of a fire in storages
simplicity, clear foundation in theory [68], and can be determined from Equation 13.3 when the
central position among the other correlations. flames extend above the storage, which implies
that values of z0 calculated refer to the distance
Other Fire Types The original derivation of above the base of the fire (usually the base of the
Equation 13.31 for pool fires [68] includes the storage). Equation 13.37 may also be assumed to
following expression: be valid for turbulent jet fires.
414 G. Heskestad
EXAMPLE 8 Example 1 concerned a 1.5-m- ¼ 0.69 m. From Equation 13.8, calculate the
diameter methyl alcohol fire generating flame height as 3.67 m (above base of storage),
Q_ ¼ 884 kW. Calculate the virtual origin. which is 0.67 m above the top of the storage.
The height of the virtual origin above the base of
SOLUTION In this example, D ¼ 1.5 m. Direct the storage is calculated from Equation 13.37,
substitution into Equation 13.32 gives assuming Q_ c ¼ 0:7 Q_ ¼ 1050 kW, yielding z0
¼ 3.67 0.175 10502/5 ¼ 0.84 m.
z0 0:083ð884Þ2=5
¼ 1:02 þ
D 1:5
¼ 1:02 þ 0:83 ¼ 0:19 Entrainment
from which After ignition, the fire plume carries fire products
z0 ¼ 0:19 1:5 ¼ 0:29 m diluted in entrained air to the ceiling. A layer of
diluted fire products, or smoke, forms under the
This is the value for z0 (rounded off) used in ceiling, which thickens and generally becomes
Example 5. hotter with time. The fire environment is inti-
mately tied to the behavior of this layer, which,
EXAMPLE 9 Negative values for z0 are often in turn, depends to a major extent on the mass
calculated for low heat release fires and suffi- flow rate of plume fluid into the layer. Conse-
ciently large fire diameters, as in Example 8. quently, it is important to be able to predict the
Positive virtual origins are often found for high mass flow rate that may occur in a fire plume.
heat release fires. Substituting heptane for methyl The mass flow at a particular elevation in a
alcohol in Example 8 (2500 kW/m2 rather than fire plume is almost completely attributable to air
500 kW/m2 measured for methyl alcohol) [17], entrained by the plume at lower elevations. The
calculate the new virtual origin. mass flow contributed by the fire source itself is
insignificant in comparison.
SOLUTION The new heat release rate is For a weak plume, the mass flow rate at a
cross section can be written
π1:5 2
Q_ ¼ 2500 ¼ 4420 kW
4 m_ ent ¼ Eρ1 u0 b2u ð13:38Þ
From Equation 13.32, where E is a nondimensional constant of
proportionality. With the aid of Equation 13.23
z0 44202=5
¼ 1:02 þ 0:083 ¼ 0:57 and the equivalent of Equation 13.21 written for
D 1:5 bu (setting T0/T1 ¼ 1 because of the weak plume
from which assumption), Equation 13.38 becomes
z0 ¼ 0:57 1:5 ¼ 0:85 m 1=3

gρ21
Q_ c ðz z0 Þ5=3 ð13:39Þ
1=3
m_ ent ¼E
EXAMPLE 10 A 3-m-deep storage is known to c pT1
produce a heat release rate per unit floor area of
4000 kW/m2 when fully involved. At a stage of Early measurements by Yih [71] indicated a
fire development in such a storage, a heat release value E ¼ 0.153.
rate of 1500 kW is reached. What is the location Cetegen et al. [70, 72] concluded from theo-
of the virtual origin? retical analysis that Equation 13.39 also applies
to strongly buoyant plumes. From extensive
SOLUTION First determine the flame height. entrained-flow measurements for natural gas
Evaluate the effective fire diameter from burners of several diameters, these authors pro-
πD2/4 ¼ 1500/4000 ¼ 0.375 m2, which gives D posed a coefficient E ¼ 0.21 based on the total
www.ebook777.com
heat release rate, corresponding to E ¼ 0.24 and taking ΔTL ¼ 500 K. Interestingly, ṁent,L is
based on the convective heat release rate as in independent of the acceleration of gravity, g.
Equation 13.39 (assuming a convective fraction Mass flow rates in fire plumes at levels below
of 0.7). However, the plume flow rates at large the flame tip have been found to increase linearly
heights were somewhat overpredicted and those with height for fire diameters of 0.3 m and
at low heights, approaching the flames, were greater [73], where the flames are substantially
somewhat underpredicted. turbulent, from zero (essentially) at the fire base
Heskestad [73] reconsidered the entrainment to the flame-tip value in Equation 13.41, that is,
problem for strong plumes, assuming self-
preserving density deficiency profiles instead of m_ ent ¼ m_ ent, L z=L ð13:42Þ
self-preserving excess temperature profiles as
Measurements in the flaming region for fire
traditionally assumed. This approach led to the
diameters smaller than 0.30 m do not show a
following extension of Equation 13.39:
linear variation of mass flow rate with height,
1=3 including data by Cetegen et al. [70, 72] (fire
gρ21
Q_ c ðz z0 Þ5=3
1=3
m_ ent ¼ E diameters of 0.10 and 0.19 m) and Beyler [74]
c pT1
" # (0.19 and 0.13 m). (It is important to note, how-
GQ_ c
2=3
ever, that all these data are consistent with an
1þ 2=3 approach to the mass flow rate at the mean flame
g1=2 c p ρ1 T 1 ðz zo Þ5=3
height given by Equation 13.41 [73].) Neither do
ð13:40Þ the mass flow measurements in turbulent jet
Equation 13.40, with E ¼ 0.196 and G ¼ 2.9, flames by Delichatsios and Orloff [75] show a
was found to represent the data of Cetegen linear variation with height (estimated values of
et al. [70, 72] very well over the entire N in range 50–6300 and momentum parameter
nonreacting plume for all their fire diameters, RM in range 0.0015–0.010); in fact, these
ranging from 0.10 to 0.50 m [73]. At large measurements indicate a 5/2 power dependence,
heights, the bracketed term involving within the flames, of mass flow rate on height
G approaches unity, and at levels approaching above the nozzle exit. As a guide to entrainment
the flame tip (Equation 13.3), this term estimates in the flaming region for sources
approaches 1.5, approximately. Equation 13.40, smaller than 0.30 m in diameter, it appears that
with E ¼ 0.196 and G ¼ 2.9 is the recommended a second power variation of mass flow rate with
relation for calculating mass flow rates in plumes, height is quite representative of the fire sources
at and above the mean flame height. referred to above of diameters 0.13 and 0.10 m,
The entrained flow at the mean flame height, in which case Equation 13.42 is replaced by
ṁent,L, follows from setting z z0 ¼ L z0 in z 2
Equation 13.40 (with E ¼ 0.196 and G ¼ 2.9), m_ ent ¼ m_ ent, L ð13:43Þ
L
taking L from Equation 13.3 and z0 from Equa-
tion 13.31 (with substitution of full expression The 0.19-m-diameter sources generated scattered
for F), with the result, results between the linear and second power var-
" # iation [73]. Delichatsios [76] as well as Quintiere
T L 5=6 T 1 Q_ c and Grove [77] have also analyzed mass flow
m_ ent, L ¼ 0:878 þ 0:647
T1 ΔT L c pT1 rates in the flaming region.
ð13:41Þ We digress briefly on the appropriateness of
relating entrainment behavior to the diameter of
The numerical values are linked to the experi- the fire source. The governing parameters for fire
mental calibration coefficient for F (based on plumes from horizontal, circular sources have
F ¼ 0.083 m kW–2/5 at normal atmospheric been considered so far to be the parameter
conditions as indicated under Equation 13.31) N and the momentum parameter RM. However,
416 G. Heskestad
for small fire sources it is common to see a Under the prevailing assumptions, and the fur-
laminar flame sheet preceding transition to tur- ther assumption Q_ c =Q_ ¼ 0:7 and Hc/r 3100
bulence around the rim of the fire source, and the kJ/kg, Equation 13.45 implies that the mass flow
degree to which such laminar regions and effects at the flame tip is 13 times the mass stoichiomet-
exist will depend on the flame Reynolds number. ric requirement of the fuel [73].
A flame Reynolds number can be formulated as Fires with very low flame height-to-diameter
u0mbum/vm, where u0m is the characteristic gas (L/D) ratios have not been investigated exten-
velocity in the flame, proportional to Q_
1=5
sively. It is not clear to what L/D limit the
according to Equation 13.28; bum is the entrained-flow relations presented here apply,
associated characteristic flame radius, propor- but this limit is smaller than 0.9, the lowest L/D
ratio associated with the data of Cetegen
tional to [78] Q_ ; and vm is the kinematic
2=5
et al. [70, 72] For plume mass flows above the
viscosity evaluated at the mean maximum flame
flames, there is no L/D limit for predictions at the
temperature, which can be considered constant.
higher elevations, but predictions of mass flows
Hence, the flame Reynolds number can be con-
at levels just above the flames may begin to
sidered proportional to Q_ . Assuming the dis-
3=5
deteriorate before L/D ¼ 0.14 is reached, as
charge momentum is not important (small RM), seems to be implied in the observations follow-
the flame entrainment behavior should be a func- ing Equation 13.28.
tion of N and Q. _ When the entrainment behavior Further, mention should be made of a plume
is represented on N, Q_ coordinates for the various mass flow formula often used because of its
test fires indicated above, it is found that the fires simplicity, originally developed for the flaming
with linear increase of mass flow rate with height region of large fires by Thomas et al. [79]
in the flame plot above a line Q_ / N 1=2 , with 1=2
some uncertainty about the precise level. With m_ ent ¼ 0:096 gρ1 ρ f ‘ Wf z3=2 ð13:47Þ
the aid of Equation 13.4 it becomes clear that this
relation implies an equivalent limit line D ¼
constant, which justifies relating the entrainment Here ρf‘ is the gas density in the flames and Wf is
behavior to the source diameter. the fire perimeter. This formula has also been
For normal atmospheric conditions tested against mass flow data above the flames
Equations 13.40, 13.41, and 13.42 can be written by Hinkley [80], who claims it is very satisfac-
as follows for the plume mass flow rate at various tory for heights up to 10 times the linear dimen-
heights: sion (or diameter) of a fire, although there is little
Above the mean flame height, L ( Q_ c in kW, theoretical justification for its use above the
z and z0 in m) flames. The following version of Equation 13.47
is often used [80] (based on normal atmospheric
m_ ent ðkg=sÞ ¼ 0:071Q_ c ðz z0 Þ5=3
1=3
conditions and an assumed flame temperature):
h i
1 þ 0:027Q_ c ðz z0 Þ5=3 m_ ent ðkg=sÞ ¼ 0:188Wf ðmÞzðmÞ3=2 ð13:48Þ
2=3
ð13:44Þ It is instructive to compare the predictions of

Equations 13.44 and 13.48 for plume regions
At the mean flame height, L(ΔTL ¼ 500 K)
above the flames. In a number of comparisons
m_ ent, L ðkg=sÞ ¼ 0:0058Q_ c ðkWÞ ð13:45Þ for heat release rates in the range 1000–8000 kW,
heat release rates per unit area in the range
At and below the mean flame height, L, for fire 250–1000 kW/m2, and heights varying from the
source diameters of 0.3 m and greater flame level to 128 m, the predictions of Equa-
z tion 13.48 range from 0.64 to 1.38 times the
m_ ent ðkg=sÞ ¼ 0:0058Q_ c ðkWÞ ð13:46Þ predictions of Equation 13.44.
L
www.ebook777.com
Cetegen et al. [70, 72], whose data have limitations. Units used throughout this section
contributed most to the mass flow are kW for heat release rate, m for length, s for
recommendations in this text, have carefully time, K for temperature, and m/s for velocity.
pointed out that their fire plumes were produced The example can be used of a large building
in as quiet an atmosphere as could be maintained that will allow clear, uncontaminated air to exist
in their laboratory. They report that small ambi- around a particular growing fire for at least
ent disturbances could provide 20–50 % 10 min before smoke begins to recirculate into
increases in the measured plume mass flows. the region. Normal atmospheric conditions pre-
Clearly, there is need for further research. vail. Wood pallets are stored in a large, continu-
Grove and Quintiere [45] have developed new ous array on the floor to a height of 1.2 m. This
correlations for entrainment, including linear array is ignited locally at an interior point, and
sources (not considered here). They state their the fire spreads in a circular pattern at constant
near-field entrainment correlations show some radial speed (as predicted and observed for wood
differences with others. cribs) [81], such that the heat release rate grows
with the second power of time as
EXAMPLE 11 Calculate plume mass flow 2
rates for the methyl alcohol fire of Examples 1, _ t
Q ðkW Þ ¼ 1000 ð13:49Þ
5, and 8. tg
Here, t is time and tg is the so-called growth time.

SOLUTION From Example 1, Q_ ¼ 884 kW
When tg is 60 s, the fire grows through a magni-
and L ¼ 1.83 m; from Example 5, Q_ c ¼ 0:8 tude of 1000 kW in 60 s. When tg is 600 s, the fire
Q_ ¼ 707 kW; from Example 8, z0 ¼ 0.29 m. grows through a magnitude of 1000 kW in 600 s,
At the mean flame height, 1.83 m, the mass flow a much slower growth rate. In this illustration,
rate follows from Equation 13.45: it is assumed that the growth time is tg ¼ 140 s. It
is also assumed that the fully involved pallet
m_ ent, L ¼ 0:0058 707 ¼ 4:1 kg=s
storage generates a total heat release rate of
Mass flow rates in the flaming region are calcu- 2270 kW/m2 of floor area. [17] The objective is
lated using Equation 13.46 as to determine flame height as a function of time,
as well as the variation of plume centerline tem-
z
m_ ent ðkg=sÞ ¼ 4:1 ¼ 2:2 z ðmÞ perature, plume centerline velocity, and plume
1:83
width at an elevation of 5 m above the base of the
Mass flow rates above the flames are obtained fuel array where a structural member may cross
from Equation 13.44; for example, at a height of and be heated by the plume.5
3.66 m (twice the flame height) For the assumed growth time, tg ¼ 140 s, the
variation of total heat release rate with time
m_ ent ¼ 0:071 7071=3 ð3:66 þ 0:29Þ5=3 comes from Equation 13.49 as follows:
h i
1 þ 0:027 7072=3 ð3:66 þ 0:29Þ5=3 Q_ ¼ 5:10 102 t2 ð13:50Þ
¼ 6:24ð1 þ 0:22Þ The convective heat release rate is assumed at
¼ 7:6 kg=s 70 % of the total heat release rate as
Illustration
5
In addition to the previous examples, it is instruc- In addition to convective heating, which depends on gas
temperature and velocity, radiative heating would also be
tive to work through a somewhat larger problem important in such cases and might even dominate over
to illustrate handling of the equations and their convective heating if the structure is immersed in flames.
418 G. Heskestad
Q_ c ¼ 3:57 102 t2 ð13:51Þ The fire diameter, D, is also plotted in Fig. 13.8,
based on Equation 13.53.
The instantaneous fire diameter, D, is deter- The virtual origin, z0, is determined from
mined as follows. Since the heat release rate per Equation 13.32, with substitutions for Q_ from
unit floor area is 2270 kW/m2, Equation 13.50 and for D from Equation 13.53
πD 2
z0 ¼ 5:46 103 t þ 2:52
Q_ ¼ 2270 ð13:52Þ
4
102 t4=5 ð13:55Þ
Upon eliminating Q_ between Equations 13.50
and 13.52, the following can be obtained: The curve labeled z0 in Fig. 13.8 represents the
virtual origin according to Equation 13.55. It is
D ¼ 5:35 103 t ð13:53Þ seen that z0 nearly levels off in the time interval
plotted in the figure; actually, z0 begins to
First, the behavior of flame height may be
decrease again at somewhat larger times.
calculated using Equation 13.8. Substitution of
With this foundation, there is sufficient infor-
Equations 13.50 and 13.53 into Equation 13.8
mation to calculate gas temperatures, velocities,
gives the following relation of flame height as a
and plume widths at the 5 m height above the
function of time:
base of the fuel array.
L ¼ 5:46 103 t þ 7:15 102 t4=5 ð13:54Þ The temperature rise on the plume centerline
at the selected height is determined from
This relation is plotted in Fig. 13.8 for the Equation 13.22 by substituting z ¼ 5 (m); z0
10-min (600-s) fire interval and is labeled L. from Equation 13.55; Q_ c from Equation 13.51;
and values of T1, g, cp, and ρ1 for the normal
atmosphere, yielding
10
2:71t4=3
ΔT 0 ¼ 5=3
5 þ 5:46 103 t 2:52 102 t4=5
ð13:56Þ
8
This relation is valid up to the time that a
L temperature rise associated with the flame tip,
ΔT0 ¼ 500 K, is felt at the selected height,
6
which occurs at t ¼ 303 s. The plot of ΔT0 in
D, L, Z0 (m)
Fig. 13.9 is according to Equation 13.56 up to the

time t ¼ 303 s. At larger times, ΔT0 is deter-
4
mined from Fig. 13.5 in the following manner:
at each selected time, z z0 is calculated using
Equation 13.55; Q_ c is calculated from Equa-
tion 13.51; the quantity (z z0)/ Q_
2=5
D c is deter-
2
mined and ΔT0 is read from Fig. 13.5. The
resulting extension of the ΔT0 curve is seen in
Z0 Fig. 13.9.
0 The centerline gas velocity at the 5 m height
0 200 400 600 above the base of the fuel array can then be
Time (s)
considered. Equation 13.23 can be used up to
Fig. 13.8 Growing fire illustration: fire diameter, D, the moment that the flame tip reaches the 5 m
flame height, L, and virtual origin, z0 height, that is, at t ¼ 303 s. After substitution of
www.ebook777.com
Fig. 13.9 Growing fire 1000 20

illustration: plume width,
bΔT, and centerline values
of temperature rise, ΔT0,
and velocity, u0, at 5 m
ΔT0
above the base of the fuel
800 16
u0
u0 (m/s); bΔT •10 (m)

600 12
ΔT0 (K)
bΔT
400 8
200 4
303 s
0 0
0 200 400 600
Time (s)
z ¼ 5 (m), z0 from Equation 13.55, Q_ c from with a lower limit of ðz z0 Þ=Q_ c ¼ 0:08,
2=5
Equation 13.51, and normal ambient conditions, according to available measurements [12, 16].
Equation 13.23 becomes Since the largest time in Fig. 13.9 corresponds
to ðz z0 Þ=Q_
2=5
0:339t2=3 c ¼ 0:092, Equation 13.28 has
u0 ¼ 1=3 been used to calculate the entire extension of
5 þ 5:46 103 t 2:52 102 t4=5 the u0 curve in Fig. 13.9.
ð13:57Þ The temperature radius of the plume at the
5 m height above the fuel array is calculated
The u0 curve in Fig. 13.9 follows Equation 13.57 from Equation 13.21, which can be written
to the limit, t ¼ 303 s. As stated in conjunction

with Equation 13.27, the maximum velocity (for ΔT 0 1=2
bΔT ¼ 0:12 1 þ ðz z0 Þ ð13:58Þ
a given size fire) occurs just below the mean T1
flame height where ΔT0 ¼ 650 K, which
With substitution of z ¼ 5 (m), ΔT0 from Equa-
corresponds to (z z0)/ Q_ c ¼ 0:135 according
2=5
tion 13.56, and z0 from Equation 13.55,
to Fig. 13.5. Using z ¼ 5 (m), and z0 and Q_ c from Equation 13.58 becomes
Equations 13.55 and 13.51, the 0.135 limit is

1=2
found to correspond to a time of t ¼ 385 s, 1þ9:25103 t4=3
where Equation 13.28 gives the centerline veloc- bΔT ¼ 0:12
ð5þ5:46103 t2:52102 t4=5 Þ
5=3
ity in terms of Q_ c . In fact, it appears that Equa-

tion 13.28 can be used with good accuracy to 5 þ 5:46 103 t 2:52 102 t4=5
even larger times, at least to times associated ð13:59Þ
420 G. Heskestad
This equation is plotted in Fig. 13.9 up to the (5) effects of wind, and (6) behavior of turbulent
time the flames reach the 5 m height at t ¼ 303 s. jet diffusion flames in horizontal discharge.
The temperature radius at the 5 m height is seen These topics are discussed briefly here.
to vary from 0.59 m early in the fire to 0.83 m at
303 s. Plume fluid will reach a minimum of twice
the temperature radius, bΔT; hence, the total Flame Intermittency Length Scale
width of the plume in this example will be at
least four times bΔT, growing from a minimum The flame intermittency has been discussed in
of 2.4 m early in the fire to a minimum of 3.3 m connection with the definition of flame height in
as the flames reach the 5 m height. Fig. 13.2, based on the work of Zukoski
et al. [23]. Zukoski and coworkers [82] also
defined a flame intermittency length scale, LI,
Additional Plume and Flame Topics with the aid of the maximum (negative) slope
of the intermittency (I) curve and the
Other aspects of fire plumes than those already intersections of this slope with I ¼ 0 and I ¼ 1.
discussed have been studied and may be of inter- LI was defined as the difference in elevations
est, including (1) flame intermittency length z between these two intersections.
scale, (2) flame pulsations and (3) the rise of a Figure 13.10 presents results [27] from simul-
plume front or cap after ignition into undisturbed taneous measurements of the flame height ratio,
air. Both of these aspects are essentially axisym- L/D, and the flame intermittency length scale to
metric, whereas three others are not: (4) effects flame height ratio, LI/L, for the three circular
on a fire plume of proximity to a wall or corner, porous bed fires of natural gas investigated by
Fig. 13.10 Flame heights and intermittency scales according to flame-intermittency data of Zukoski et al. [82]
www.ebook777.com
Zukoski et al. [23, 82] The open symbols repre- As a measure of the data scatter it is noted that
sent the nondimensional flame heights L/D, measured frequencies near a given diameter
which compare well with the reference curve agree within a factor of two.
according to Equation 13.3. It is seen that LI/L Flame pulsations are not always easy to detect
generally increases as N and the nondimensional visually. For example, Wood et al. [40] reported
flame height decrease, implying an visual pulsations for a flame on their 1.52-m
accompanying increase in flame fluctuations. diameter sandfilled pan when initially saturated
Wood et al. [40] reported that coherent flaming with methanol, but not when initially saturated
in their pool fires changed to distributed flamelets with acetone. Other than visually, pulsations have
for L/D values smaller than about 0.5 (log10 L/ been detected based on temperature and velocity
D ¼ 0.30), which occurs near log10 N ¼ 5 records in the flames, motion picture records, or
(N ¼ 105) in Fig. 13.10. This behavior seems to pressure fluctuations on the surface of the fire
correlate well with LI/L near unity. source.
Flame Pulsations Rise of Plume Front
Flame pulsations have been studied by a number Measurements by Tanaka et al. [86] of the rise
of investigators, tracing at least as far back as times of plume fronts from suddenly initiated
Rashbash et al. [83] and reviewed in conjunction fires at steady heat release rate (such as pool
with an investigation reported by Cetegen and fires) have been represented by Heskestad [87]
Ahmed [84]. Byram and Nelson [85] describe in the following formula:
the pulsation cycle as starting with the expansion
of the flames near the base of the fire, followed t*R ¼ 0:46 ð13:61Þ
by a sudden collapse of these flames toward the
Here, tR* is the nondimensional rise time to a
center of the fire. A flame bulge then travels
given height, z, defined as
upward to the flame tip in an even, wavelike
1=3 1=3 4=3
motion. Expansion of the lower part of the flames t*R ¼ g= c p T 1 ρ1 Q_ c z tR ð13:62Þ
starts the cycle again.
Cetegen and Ahmed [84] summarize the where tR is the physical rise time to the
published data on pulsation frequency in a single elevation z.
plot for burner or pool diameters ranging from As an example, if Q_ ¼ 100 kW and assuming
0.03 to 20 m and propose the simple curve fit Q_ c ¼ 0:7 100 ¼ 70 kW, the physical rise time
to an elevation of 30 m in a normal atmosphere is
f ðHzÞ ¼ 1:5½DðmÞ1=2 ð13:60Þ calculated as
tR ¼ 0:46 ½9:81=ð1:00 293 1:2Þ1=3 701=3 304=3

tR ¼ 34 s
New measurements by Hu et al. [88] have Wall/Corner Effects

closely reproduced the result in Equation 13.61
(0.43 compared to 0.46). In addition, these McCaffrey [24] has reviewed effects on flame
investigators measured rise times for plumes height of placing fire sources next to a wall or
near a building wall, giving essentially the same in a corner, referring to experiments by Hasemi
result as the case of a free plume, and near a and Tokunaga [89], Back et al. [90], Mizuno and
building corner (interior) where the rise times Kawagoe [91], and Kokkala [92]. The effects are
were about ¾ of the free plume values. generally reported to be small. In more recent
422 G. Heskestad
work, Poreh and Garrad [93] and Lattimer and agreement of measured flame lengths with lab-
Sorathia [94] have made further studies. oratory results in one case (a methane
series) and 10–40 % smaller lengths in
another case (a propane series) [53]. The
Windblown Flames smaller lengths may have been associated with
a slight wind.
The main effect of wind is to bend or deflect the
flames away from the vertical. Another effect,
observed in wind tunnel studies by Welker and
Jet Flames in Horizontal Discharge
Sliepcevich [95], is “flame trailing,” in which the
flames trail off the burner along the floor in the
Becker et al. [100] have studied flame geometries
downwind direction for a significant distance.
resulting from horizontal gas discharge and have
Flame trailing was thought to be associated
provided extensive data and correlations. For
primarily with fuel vapors of greater density
ease of application, their results have been
(higher molecular weight) than air, as was the
reformulated by the author as functions of the
case with all the various liquid fuels used in the
momentum parameter RM of Equation 13.9 (and
experiments.
Fig. 13.4), using supporting data from Becker
Wind tunnel measurements of flame deflec-
and Liang [25].
tion angle, involving fire diameters in the range
The flames were stabilized against blow-off
0.10–0.60 m, and large-scale data for square
with hydrogen or oxygen, when necessary, and
LNG pools in the effective diameter range
photographed from the side in 1-s exposures. The
2–28 m, obtained by Attalah and Raj [96], have
results are presented in Fig. 13.11, where the
been found to correlate well against the ratio of
inset defines the selected measures of flame
wind velocity to the maximum velocity in the
geometry with X and Y being the maximum hori-
flame according to Equation 13.28 [65]. The rela-
zontal and vertical flame excursions, respec-
tionship indicates that a flame deflection angle of
tively. In the figure, maximum excursions
approximately 25 can be expected for a velocity
(derived from faired curves [100]) have been
ratio of 0.10. Effects of wind on flame length
made nondimensional with flame heights in ver-
were minor for velocity ratios up to 0.35 (flame
tical discharge (from faired curves [24]), L, and
deflection angle of approximately 60 ). Data by
plotted as functions of RM. Seven different gases
Huffman et al. [97] indicate that at the consider-
were employed in the experiments, as indicated.
ably higher velocity ratio of 1.0, flame lengths
Note that the vertical flame heights, L, were
are approximately 30 % greater than under qui-
interpreted as the highest points “at which
escent conditions.
flaming gas was seen to dwell with an apprecia-
For turbulent jet diffusion flames, discussed
ble frequency.” [25]
under the subheading “Momentum Regime” of
Calculation of RM has been illustrated in
the section “Flame Heights,” Brzustowski
Example 4. Note in the figure that as RM
et al. [98] found reduced flame length in a
approaches 1, the entire flame practically
range of small wind velocities for a laboratory
projects horizontally and has a length nearly the
hydrogen flame. Related to this finding is an
same as in vertical discharge (X/L near 0.9).
observation reported to the author by a staff
There is some uncertainty in this comparison
member of a major supplier of flare equipment
since the bases of the X and L measurements
that the flames of flares are typically seen to
are not quite identical.
increase in length by some 30 % when winds
The inset sketch in Fig. 13.11 is actually the
calm down from about 0.4 m/s. Finally we call
outline of an ethane flame from a photograph
attention to measurements by Sönju and Hustad
[100] at RM ¼ 0.14.
[99], conducted mostly outdoor, which indicate
www.ebook777.com
1
Fig. 13.11 Data from C3H8 X/L
Becker et al. [100] on C3H8 Y/L
maximum nondimensional 0.9 CH4 X/L
horizontal and vertical X /L
CH4 Y/L
flame excursions of
C2H6 X/L
turbulent jet diffusion 0.8
C2H6 Y/L
flames in horizontal gas
discharge, formulated as C2H4 X/L
functions of the momentum 0.7 C2H4 Y/L
parameter RM of C2H2 X/L
Equation 13.9 and Fig. 13.4 0.6 C2H2 Y/L
X
CO X/L
X/L; Y/L
CO Y/L
0.5
H2 X/L
Y H2 Y/L
0.4
0.3
0.2
0.1
Y/L
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
RM
Fires with low flame height-to-diameter ratios

Research Needs were discussed following Equation 13.28 and
several uncertainties noted. Research on such
All of the areas discussed in this chapter will fires can fill in areas of uncertainty.
benefit from further study. However, certain There may be secondary effects on flame
needs stand out from the discussion. height (and other flame properties) of variables
Temperature measurements in fire with enumerated in the paragraph before Equation 13.6,
bare-bead thermocouples usually involve radia- which should be investigated, including nonuni-
tion errors and undoubtedly such errors were form volatilization rate across the fuel bed, depar-
present in the measurements for Fig. 13.5. ture of fire area from circular, density of volatiles,
Newman and Croce [101] in their study of and effect of flame Reynolds number (which
aspirated thermocouples determined radiation depends on fire diameter and heat release rate).
errors for a centerline location close to the Attention is directed to the end of subsection
flame tip of a 0.230-m diameter acetone pan “Pool Fires” of the section “Virtual Origin,”
fire. For a 1-mm diameter bare-bead thermocou- where more precise measurements of virtual
ple their measurements translate to a thermo- origins for pool fires are recognized as a need.
couple temperature 170 K less than the actual Near the end of subsection “Entrainment,” the
gas temperature as a result of radiation to the observations of Cetegen et al. [68, 70] on the sensi-
environment. (The aspirated thermocouple tivity of air entrainment in a plume to ambient
design of Newman and Croce nearly disturbances were noted. Since air entrainment
eliminated the radiation error.) Careful scenarios (as in room fires) often involve
experiments are needed to establish near error- disturbances, it is important to be able to account
free temperatures. for these.
424 G. Heskestad
LI Intermittency length scale

Data Sources LM Momentum controlled flame height
(m)
NFPA 92B, Standard for Smoke Management ṁent Entrained mass flow rate in plume
Systems in Malls, Atria, and Large Spaces (kg/s)
[103], was referenced in this chapter for tables ṁent,L ṁent at the mean flame height, L(kg/s)
of heat release rate per unit floor area, kW/m2, ṁf Mass burning rate (kg/s)
and growth times, tg, of a number of fuel arrays. N Nondimensional parameter defined in
The same information has been incorporated Equation 13.4
by Alpert and Ward [104], together with ps Pressure in source gas discharge
additional data. stream (Pa)
In Appendix 3 tables are provided to estimate ps0 Pressure in source gas reservoir (Pa)
combustion efficiencies as well as total and con- Q_ ṁfHctotal heat release rate (kW)
vective heat release rates per unit exposed area of
Q_ c Convective heat release rate (kW)
materials under full-scale burning conditions.
Q_
*
Radiative heat release rate (kW)
Q_
*
Nondimensional parameter defined in
Nomenclature 13.1
R Radius (m)
A Defined in Equation 13.6 (mkW ) –2/5 r Actual mass stoichiometric ratio, air
B Buoyancy flux defined in Equa- to fuel volatiles
tion 13.15 (m4s3) RM Momentum parameter defined in
b Plume radius (m) Equation 13.8
bΔT Plume radius to point where ΔT/ΔT0 ¼ T Mean temperature (K)
0.5 (m) T0 Mean centerline temperature in plume
bu Plume radius to point where (K)
u/u0 ¼ 0.5 (m) T1 Ambient temperature (K)
bum bu at level of maximum gas velocity T0 rms temperature fluctuation (K)
near flame tip (m) Ta(z) Ambient temperature at level z (K)
c Adjustable constant, Equation 13.35 Ta1 Ambient temperature at source level
cp Specific heat of air at constant pres- (K)
sure (kJ/kgK) TL T0 at mean flame height (K)
D Diameter (m) ΔT T T1, mean temperature rise above
F Function (cp, T1, ρ1, g); see Equa- ambient (K)
tion 13.31 (mkW–2/5) ΔT0 Value of ΔT on plume centerline (K)
f Frequency (s1) ΔTL TL T1(K)
g Acceleration due to gravity (m/s2) t Time (s)
Hc Actual lower heat of combustion tg Growth time; see Equation 13.49 (s)
(kJ/kg) tR Rise time of plume front (s)
ΔHO Tewarson’s [44] lower heat of com- tR* Nondimensional rise time of plume
bustion per unit mass of oxygen con- front, see 13.62
sumed (kJ/kg) u Mean axial velocity (m/s)
I Intermittency u0 Mean axial velocity on centerline
k Ratio of specific heats, constant- (m/s)
pressure versus constant-volume u0m Maximum value of u0, near flame tip
L Mean flame height above base of fire (m/s)
(m) u0 rms velocity fluctuation in axial direc-
LB Buoyancy controlled flame height (m) tion (m/s)
www.ebook777.com
Wf Fire perimeter (m) Simulations of a One-Meter Diameter Methane

z Height above base of fire (m) Fire,” Combustion and Flame, 153, pp.499–509
(2008).
z0 Height of virtual origin above base of 6. Tieszen, S.R., O’Hern, T.J., Schefer, R.W.,
fire (m) Weckman, E.J. and Blanchat, T.K., “Experimental
zm Maximum vertical penetration of Study of the Flow Field In and Around a One Meter
plume fluid in stratified ambient (m) Diameter Methane Fire,” Combustion and Flame,
129, pp. 378–391 (2002).
α Entrainment coefficient 7. FireFoam, released by FM Global, available from <
ξ Nondimensional parameter defined in http://code.google.com/p/firefoam-dev/>
Equation 13.26 8. OpenFoam, produced by OpenCFD Ltd, available
vm Kinematic viscosity of flame gases at from < http://www.openfoam.com/>
9. Wang, Y, Chatterjee, P. and de Ris, J.L., “Large
maximum flame temperature (m2 s1) Eddy Simulation of Fire Plumes,” Proceedings
ρ Mean density (kg/m3) of the Combustion Institute, 33, pp. 2473–2480
ρa1 Ambient density at source level (2011).
(kg/m3) 10. Olenick, S.M. and Carpenter, D. J., “An Updated
International Survey of Computer Models for Fire
ρfe Mean density in flames (kg/m3) and Smoke,” Journal of Fire Protection Engineering,
ρs Density of source gas discharge 13, pp. 87–110 (2003).
stream (kg/m3) 11. G. Heskestad, “Engineering Relations for Fire
ρs0 Density of source gas in reservoir Plumes,” Fire Safety Journal, 7, pp. 25–32 (1984).
12. B.J. McCaffrey, “Purely Buoyant Diffusion Flames:
(kg/m3) Some Experimental Results,” NBSIR 79–1910,
ρs1 Density of source gas at ambient tem- National Bureau of Standards, Washington, DC
perature and pressure (kg/m3) (1979).
ρ1 Ambient density (kg/m3) 13. G. Cox and R. Chitty, “A Study of the Deterministic
Properties of Unbounded Fire Plumes,” Combustion
Δρ ρ1 ρ, mean density deficiency and Flame, 39, pp. 191–209 (1980).
(kg/m3) 14. G. Heskestad, “Peak Gas Velocities and Flame
σΔT Plume radius to point where ΔT/ΔT0 ¼ Heights of Buoyancy-Controlled Turbulent Diffu-
e1 (m) sion Flames,” 18th Symposium on Combustion,
Combustion Institute, Pittsburgh, PA, pp. 951–960
σu Plume radius to point where u/u0 ¼ (1981).
e1 (m) 15. H.C. Kung and P. Stavrianidis, “Buoyant Plumes of
Large-Scale Pool Fires,” 19th Symposium on Com-
bustion, Combustion Institute, Pittsburgh, PA,
References pp. 905–912 (1983).
16. E. Gengembre, P. Cambray, D. Karmed, and
1. Rehm, R.G, Baum, H.R., “The Equations of Motion J.C. Bellet, “Turbulent Diffusion Flames with
for Thermally Driven, Buoyant Flows,” J. Res. Nat. Large Buoyancy Effects,” Combustion Science and
Bur. Stand. 83, pp. 297–308 (1978). Technology, 41, pp. 55–67 (1984).
2. H.R. Baum, K.B. McGrattan, and R.G. Rehm, 17. G. Heskestad, “A Fire Products Collector for Calo-
“Mathematical Modelling and Computer Simulation rimetry into the MW Range,” Report OC2E1.RA,
of Fire Phenomena,” Fire Safety Science- Factory Mutual Research Corp., Norwood, MA
Proceedings Fourth International Symposium, Inter- (1981).
national Association of Fire Safety Science, London, 18. A. Tewarson, “Physico-Chemical and Combustion/
UK (ed. T. Kashiwagi), pp. 185–193 (1994). Pyrolysis Properties of Polymeric Materials,” NBS-
3. T.G. Ma and J.G. Quintiere, “Numerical Simulation of GGR-80-295, National Bureau of Standards,
Axi-Symmetric Fire Plumes: Accuracy and Washington, DC (1982).
Limitations,” Fire Safety Journal, 38, pp. 467–492 19. C.L. Beyler, “Fire Hazard Calculations for Large,
(2003). Open Hydrocarbon Fires,” The SFPE Handbook of
4. Y. Xin, J.P. Gore, K.B. McGrattan, R.G. Rehm, and Fire Protection Engineering, 3rd ed., Society of Fire
H.R. Baum, “Fire Dynamics Simulation of a Turbu- Protection Engineering and National Fire Protection
lent Buoyant Flame Using Mixture-Fraction-Based Association, Quincy, MA (2002).
Combustion Models,” Combustion and Flame, 20. B. Hägglund and L.E. Persson, “The Heat Radiation
141, pp. 329–335 (2005). from Petroleum Fires,” Försvarets Forskning-
5. Xin, Y., Filatyev, S.A., Biswas, K., Gore, J.P., sanstalt, Stockholm, FDA Report C20126-D6(A3)
Rehm, R.G., and Baum, H.R., “Fire Dynamics (1976).
426 G. Heskestad
21. F. Tamanini, “Direct Measurements of the Longitu- 38. V.I. Blinov and G.N. Khudiakov, “Certain Laws
dinal Variation of Burning Rate and Product Yield in Governing Diffusive Burning of Liquids,” Dokl.
Turbulent Diffusion Flames,” Combustion and Acad. Nauk SSSR, 113, p 1094–1098 (1957).
Flame, 51, pp. 231–243 (1983). 39. B. Hägglund and L.E. Persson, “The Heat Radiation
22. E.E. Zukoski, T. Kubota, and B. Cetegen, “Entrain- from Petroleum Fires,” Försvarets Forskningsanstalt,
ment in Fire Plumes,” Fire Safety Journal, 3, pp. Stockholm, FDA Rep. C20126-D6 (A3), 1976.
107–121 (1980–81). 40. B.D. Wood, P.L. Blackshear, Jr. and E.R.G. Eckert,
23. E.E. Zukoski, B.M. Cetegen, and T. Kubota, “Visi- “Mass Fire Model: An Experimental study of the
ble Structure of Buoyant Diffusion Flames,” 20th Heat Transfer to Liquid Fuel Burning from a Sand-
Symposium on Combustion, Combustion Institute, Filled Pan Burner,” Combust. Sci. Technol., 4, p. 113
Pittsburgh, PA, pp. 361–366 (1985). (1971).
24. B. McCaffrey, “Flame Height,” The SFPE Hand- 41. K.S. Mudan, “Thermal Radiation Hazards from
book of Fire Protection Engineering, 2nd ed., Soci- Hydrocarbon Pool Fires,” Prog. Energy Combust.
ety of Fire Protection Engineers and National Fire Sci., 10, pp. 59–80 (1984).
Protection Association, Quincy, MA, pp. 2-1–2- 42. G. Heskestad, “Flame Heights of Fuel Arrays with
8 (1995). Combustion in Depth,” Fire Safety Science—
25. H.A. Becker and D. Liang, “Visible Length of Verti- Proceedings of the Fifth International Symposium,
cal Free Turbulent Diffusion Flames,” Combustion International Association for Fire Safety Science,
and Flame, 32, pp. 115–137 (1978). pp. 427–438 (1998).
26. G. Cox and R. Chitty, “Some Source-Dependent 43. Huggett, C., “Estimation of Rate of Heat Release by
Effects of Unbounded Fires,” Combustion and Means of Oxygen Consumption Measurements,”
Flame, 60, pp. 219–232 (1985). Fire Mater. 4, pp. 61–65 (1980).
27. G. Heskestad, “Luminous Heights of Turbulent Dif- 44. Tewarson, A.,” Generation of Heat and Gaseous,
fusion Flames,” Fire Safety Journal, 5, pp. 103–108 Liquid, and Solid Products in Fires,” The SFPE
(1983). Handbook of Fire Protection Engineering, 4th ed,
28. G.T. Kalghatgi, “Lift-Off Heights and Visible Society of Fire Protection Engineers and National
Lengths of Vertical Turbulent Jet Diffusion Flames Fire Protection Association, Quincy, MA,
in Still Air,” Combustion Science and Technology, pp. 3-109–3-194 (2008).
41, pp. 17–29 (1984). 45. B.S. Grove and J.G. Quintiere, “Calculating Entrain-
29. F.R. Steward, “Prediction of the Height of Turbulent ment and Flame Height in Fire Plumes of Axisym-
Diffusion Buoyant Flames,” Combustion Science metric and Infinite Line Geometries,” Journal of Fire
and Technology, 2, pp. 203–212 (1970). Protection Engineering, 12, pp. 117–137 (2002).
30. P.H. Thomas, “The Size of Flames from Natural 46. Newman, J.S. and Wieczorek, C.J., “Chemical
Fires,” Ninth Symposium on Combustion, Combus- Flame Heights,” Fire Safety Journal, 39, pp.
tion Institute, Pittsburgh, PA, pp. 844–859 (1963). 375–382 (2004).
31. W.R. Hawthorne, D.S. Weddel, and H.C. Hottel, 47. de Ris, J, Wu, P. and Heskestad, G., “Radiation Fire
“Mixing and Combustion in Turbulent Gas Jets,” Modeling,” Proceedings of the Combustion Institute,
Third Symposium on Combustion, Williams and 118, pp. 51–60 (1999).
Wilkins, Baltimore, pp. 288–300 (1949). 48. T.R. Blake and M. McDonald, “An Examination of
32. E.E. Zukoski, “Fluid Dynamic Aspects of Room Flame Length Data from Vertical Turbulent Diffu-
Fires,” Fire Safety Science—Proceedings of the sion Flames,” Combustion and Flame, 94, pp.
First International Symposium, Hemisphere, 426–432 (1993).
New York, pp. 1–30 (1984). 49. T.R. Blake and M. McDonald, “Similitude and the
33. E.E. Zukoski, “Convective Flows Associated with Interpretation of Turbulent Diffusion Flames,” Com-
Room Fires,” Semi-Annual Progress Report to bustion and Flame, 101, pp. 175–184 (1995).
National Science Foundation, California Institute 50. M.A. Delichatsios, “Transition from Momentum to
of Technology, Pasadena (1975). Buoyancy-Controlled Turbulent Jet Diffusion
34. G. Heskestad, “On Q* and the Dynamics of Turbu- Flames and Flame Height Relationships,” Combus-
lent Diffusion Flames,” Fire Safety Journal, 30, pp. tion and Flame, 33, pp. 349–364 (1993).
215–227 (1998). 51. H.A. Becker and S. Yamazaki, “Entrainment,
35. G. Heskestad, “A Reduced-Scale Mass Fire Experi- Momentum Flux and Temperature in Vertical Free
ment,” Combustion and Flame, 83, pp. 293–301 Turbulent Diffusion Flames,” Combustion and
(1991). Flame, 33, pp. 123–149 (1978).
36. H. Vienneau, “Mixing Controlled Flame Heights 52. N. Peters and J. Göttgens, “Scaling of Buoyant Tur-
from Circular Jets,” BSc Thesis, Dept. Chem. Eng., bulent Jet Diffusion Flames,” Combustion and
Univ. New Brunswick, Fredericton, N.B., 1964. Flame, 85, pp. 206–214 (1991).
37. M.V. D’Souza and J.H. McGuire, “ASTM E-84 and 53. G. Heskestad, “Turbulent Jet Diffusion Flames: Con-
the Flammability of Foamed Thermosetting solidation of Flame Height Data,” Combustion and
Plastics,” Fire Technology, 13, p 85–94 (1977). Flame, 118, pp. 51–60 (1999).
www.ebook777.com
54. A.H. Shapiro, The Dynamics and Thermodynamics Daniel and Florence Guggenheim Jet Propulsion
of Compressible Fluid Flow, Vol. 1, The Ronald Center, Pasadena (1982).
Press Company, New York (1953). 73. G. Heskestad, “Fire Plume Air Entrainment
55. W. Schmidt, “Turbulente Ausbreitung eines Stromes According to Two Competing Assumptions,” 21st
erhitzer Luft,” Zeitschrift für Angewandte Symposium on Combustion, Combustion Institute,
Mathematik und Mechanik, 21, pp. 265–278 (1941). Pittsburgh, PA, pp. 111–120 (1986).
56. H. Rouse, C.S. Yih, and H.W. Humphreys, “Gravi- 74. C.L. Beyler, Development and Burning of a Layer of
tational Convection from a Boundary Source,” Products of Incomplete Combustion Generated by a
Tellus, 4, pp. 201–210 (1952). Buoyant Diffusion Flame, Ph.D. Thesis, Harvard
57. B.R. Morton, G.I. Taylor, and J.S. Turner, “Turbu- University, Cambridge, MA (1983).
lent Gravitational Convection from Maintained and 75. M.A. Delichatsios and L. Orloff, “Entrainment
Instantaneous Sources,” Proceedings of the Royal Measurements in Turbulent Buoyant Jet Flames
Society A, 234, pp. 1–23 (1956). and Implications for Modeling,” 20th Symposium
58. B.R. Morton, “Forced Plumes,” Journal of Fluid on Combustion, Combustion Institute, Pittsburgh,
Mechanics, 5, pp. 151–163 (1959). PA (1985).
59. B.R. Morton, “Modeling of Fire Plumes,” 10th Sym- 76. M.A. Delichatsios, “Air Entrainment into Buoyant
posium on Combustion, Combustion Institute, Jet Flames and Pool Fires,” The SFPE Handbook of
Pittsburgh, PA, pp. 973–982 (1965). Fire Protection Engineering, 2nd ed., Society of Fire
60. W.K. George, R.L. Alpert, and F. Tamanini, “Tur- Protection Engineers and National Fire Protection
bulence Measurements in an Axisymmetric Buoyant Association, Quincy, MA, pp. 2-20–2-31 (1995).
Plume,” International Journal of Heat and Mass 77. J.Q. Quintiere and B.S. Grove, “A Unified Analysis
Transfer, 20, pp. 1145–1154 (1977). for Fire Plumes,” 27th Symposium on Combustion,
61. S. Yokoi, “Study on the Prevention of Fire-Spread Combustion Institute, Pittsburgh, PA,
Caused by Hot Upward Current,” Report No. 34, pp. 2757–2766 (1998).
Building Research Institute, Japan (1960). 78. G. Heskestad and T. Hamada, “Ceiling Jets of Strong
62. G. Heskestad, “Fire Plume Simulator,” Report Fire Plumes,” Fire Safety Journal, 21, pp. 69–82
18792, Factory Mutual Research Corp., Norwood, (1993).
MA (1974). 79. P.H. Thomas, P.L. Hinkley, C.R. Theobald, and
63. A. Tewarson, “Experimental Evaluation of Flamma- D.L. Sims, “Investigation into the Flow of Hot
bility Parameters of Polymeric Materials,” in Flame- Gases in Roof Venting,” Fire Technical Paper
Retardant Polymeric Materials, Plenum, New York, No. 7, H. M. Stationery Office, Joint Fire Research
pp. 97–153 (1982). Organization, London (1963).
64. G. Heskestad, “Note on Maximum Rise of Fire 80. P.L. Hinkley, “Rates of ‘Production’ of Hot Gases in
Plumes in Temperature-Stratified Ambients,” Fire Roof Venting Experiments,” Fire Safety Journal,
Safety Journal, 15, pp. 271–276 (1989). 10, pp. 57–65 (1986).
65. G. Heskestad, “Dynamics of the Fire Plume,” Philo- 81. M.A. Delichatsios, “Fire Growth Rates in Wood
sophical Transactions of the Royal Society of Cribs,” Combustion and Flame, 27, pp. 267–278
London A, 356, pp. 2815–2833 (1998). (1976).
66. G. Heskestad, “Fire Plume Behavior in Temperature 82. Zukoski, E.E., Kubota, T. and Cetegen, B., “Entrain-
Stratified Ambients,” Combustion Science and Tech- ment in the Near Field of Fire Plumes,” California
nology, 106, pp. 207–228 (1995). Institute of Technology, Daniel and Florence
67. J-I. Watanabe and T. Tanaka, “Experimental Inves- Guggenheim Jet Propulsion Center, August 1981
tigation into Penetration of a Weak Fire Plume into a 83. D.J. Rasbash, Z.W. Rogowski, and G.W.V. Stark,
Hot Upper Layer,” Journal of Fire Sciences, 22, pp. “Properties of Fires of Liquids,” Fuel, 35, pp.
405–420 (2004). 94–107 (1956).
68. G. Heskestad, “Virtual Origins of Fire Plumes,” Fire 84. B.M. Cetegen and T.A. Ahmed, “Experiments on the
Safety Journal, 5, pp. 109–114 (1983). Periodic instability of Buoyant Plumes and Pool Fires,”
69. Y. Hasemi and T. Tokunaga, “Flame Geometry Combustion and Flame, 23, pp. 157–184 (1993).
Effects on the Buoyant Plumes from Turbulent Dif- 85. G.M. Byram and R.M. Nelson, Jr., “The Modelling
fusion Flames,” Fire Science and Technology, 4, pp. of Pulsating Fires,” Fire Technology, 6, pp. 102–110
15–26 (1984). (1970).
70. B.M. Cetegen, E.E. Zukoski, and T. Kubota, 86. T. Tanaka, T. Fujita, and J. Yamaguchi, “Investiga-
“Entrainment in the Near and Far Field of Fire tion into Rise Time of Buoyant Fire Plume Fronts,”
Plumes,” Combustion Science and Technology, International Journal of Engineering Performance-
39, pp. 305–331 (1984). Based Fire Codes, 2, pp. 14–25 (2000).
71. C.-S. Yih, “Free Convection Due to a Point Source of 87. G. Heskestad, “Rise of Plume Front from Starting
Heat,” Proceedings of the U.S. National Congress of Fires,” Fire Safety Journal, 36, pp. 201–204 (2001).
Applied Mechanics, New York, pp. 941–947 (1952). 88. L.H. Hu, Y.Z. Li, R. Huo, L. Yi, and C.L. Shi,
72. B.M. Cetegen, E.E. Zukoski, and T. Kubota, “Experimental Studies on the Rise-Time of Buoyant
“Entrainment and Flame Geometry of Fire Plumes,” Fire Plume Fronts Induced by Pool Fires,” Journal of
Report G8-9014, California Institute of Technology, Fire Sciences, 22 pp. 69–84 (2004).
428 G. Heskestad
89. Y. Hasemi and T. Tokunaga, “Some Experimental 98. T.A. Brzustowski, S.R. Gollahalli, and
Aspects of Turbulent Diffusion Flames and Buoyant H.F. Sullivan, “The Turbulent Hydrogen Diffusion
Plumes from Fire Sources Against a Wall and in a Flame in Cross-Wind,” Combustion Science and
Corner of Walls,” Combustion Science and Technol- Technology, 11, pp. 29–33 (1975).
ogy, 40, pp. 1–17 (1984). 99. O.K. Sönju and J. Hustad, “An Experimental Study
90. J. Back, C. Beyler, and P. DiNenno, “Wall Incident of Turbulent Jet Diffusion Flame,” 9th ICODERS,
Heat Flux Distributions Resulting from Adjacent American Institute of Aeronautics and Astronautics,
Flames,” Proceedings of the Fourth International Poitiers, France (1984).
Symposium on Fire Safety Science, International 100. H.A. Becker, D. Liang, and C.I. Downey, “Effect
Association for Fire Safety Science, London, UK, of Burner Orientation and Ambient Airflow on
pp. 241–252 (1994). Geometry of Turbulent Free Diffusion
91. T. Mizuno and K. Kawagoe, “Burning Rate of Flames,” 18th Symposium on Combustion,
Upholstered Chairs in the Center, Alongside a Wall Combustion Institute, Pittsburgh, PA,
and in a Corner of a Compartment,” Fire Safety Sci- pp. 1061–1071 (1981).
ence—Proceedings of the First International Sympo- 101. Newman, Jeffrey S. and Croce, Paul A., “A
sium, Hemisphere, New York, pp. 849–857 (1984). Simple Aspirated Thermocouple for Use in Fires,”
92. M.A. Kokkala, “Characteristics of a Flame in an Journal of Fire and Flammability, 10, pp. 326–336
Open Corner of Walls,” Interflam 1993, Interscience (1979).
Communications Limited, London (1993). 102. Burgess, D.S., Grumer, J., and Wolfhard, H.G.,
93. M. Poreh and G. Garrad, “A Study of Wall and “Burning Rates of Liquid Fuels in Large and Small
Corner Fire Plumes,” Fire Safety Journal, 34, pp. Open Trays,” International Symposium on the Use
81–98 (2000). of Models in Fire Research, Publication 786,
94. B.Y. Lattimer and U. Sorathia, “Thermal National Academy of Sciences - National Research
Characteristics of Fires in a Noncombustible Cor- Council, Washington, DC, 1961, p68.
ner,” Fire Safety Journal, 38 pp. 709–745 (2003). 103. NFPA 92B, Standard for Smoke Management
95. J.R. Welker and C.M. Sliepcevich, “The Effect of Systems in Malls, Atria, and Large Spaces, National
Wind on Flames,” Technical Report No. 2, NBS Fire Protection Association, Quincy, MA (2005).
Contract XST 1142 with University of Oklahoma, 104. R.L. Alpert and E.J. Ward, “Evaluation of
Norman (1965). Unsprinklered Fire Hazards,” Fire Safety Journal,
96. S. Attalah and P.K. Raj, “Radiation from LNG 7, pp. 127–143 (1984).
Fires,” Interim Report on Phase II Work, Project
IS-3.1 LNG Safety Program, American Gas Associ-
ation, Arlington, VA (1974). Gunnar Heskestad retired from FM Global in 2004 as an
97. K.G. Huffman, J.R. Welker, and C.M. Sliepcevich, assistant vice president and consulting research scientist.
“Wind and Interaction Effects on Free-Burning Fires,” He has specialized in fluid mechanics and heat transfer of
Technical Report No. 1441–3, NBS Contract CST 1142 fire with applications to fire protection issues.
with University of Oklahoma, Norman (1967).
www.ebook777.com
Ceiling Jet Flows

14
Ronald L. Alpert
Introduction shows an idealization of an axisymmetric

ceiling jet flow at varying radial positions, r,
Much of the hardware associated with detection beneath an unconfined ceiling. In actual fires
and suppression of fire in commercial, within buildings, the simple conditions pic-
manufacturing, storage, and modern residential tured—a hot, rapidly moving gas layer
buildings is located near the ceiling surfaces. In sandwiched between the ceiling surface and tran-
case of a fire, hot gases in the fire plume rise quil, ambient-temperature air—exist only at the
directly above the burning fuel and impinge on beginning of a fire, when the quantity of combus-
the ceiling. The ceiling surface causes the flow to tion gases produced is not sufficient to accumu-
turn and move horizontally under the ceiling to late into a stagnant, heated gas layer in the upper
other areas of the building remote from the fire portion of the compartment. Venting the ceiling
position. The response of smoke detectors, heat jet flow through openings in the ceiling surface
detectors, and sprinklers installed below the ceil- or edges can retard the accumulation of this
ing so as to be submerged in this hot flow of heated gas layer.
combustion products from a fire provides the As shown in Fig. 14.1, the ceiling jet flow
basis for building fire protection. emerges from the region of plume impingement
Studies quantifying the flow of hot gases under on the ceiling, flowing radially away from the
a ceiling resulting from the impingement of a fire fire. As it does, the layer grows thicker by
plume have been conducted since the 1950s. entraining room air at the lower boundary. This
Studies at the Fire Research Station in Great entrained air cools the gases in the jet and
Britain [1, 2], Factory Mutual Research Corpora- reduces its velocity. As the hot gases move out
tion [3–7], the National Institute of Standards and across the ceiling, heat transfer cools the portion
Technology (NIST) [8, 9], and at other research adjacent to the ceiling surface.
laboratories [10, 11] have sought to quantify the
gas temperatures and velocities in the hottest
portion of the flow produced by steady fires Steady Flow Under Horizontal,
beneath smooth, unconfined horizontal ceilings. Unconfined Ceilings
Ceiling jet refers to the relatively rapid gas
flow in a shallow layer beneath the ceiling sur- Weak Plume-Driven Flow Field
face that is driven by the buoyancy of the hot
combustion products from the plume. Figure 14.1 A generalized theory to predict gas velocities,
gas temperatures, and the thickness (or depth)
R.L. Alpert (*) of a steady fire-driven ceiling jet flow has been
Alpert FireProtection Science developed by Alpert [4] for the case of a weak

430 R.L. Alpert
Fig. 14.1 Ceiling jet flow r

beneath an unconfined
ceiling
•
Q
plume, when flame height is much less than Within the ceiling jet flow, the location of
the height, H, of the ceiling above the burning maximum excess temperature and velocity are
fuel. This work involves the use of several predicted [4] to be highly scale dependent, even
idealizations in the construction of the theoretical after normalization by the ceiling height.
model, but results are likely to provide reason- Measurements of the distance below the ceiling
able estimates over radial distances of one or two at which these maxima occur have been made
ceiling heights from the point of fire plume mainly for 1-m scale experiments [12, 13].
impingement on the ceiling. Results show distances below the ceiling ranging
from about 1 % to 2 % of the ceiling height for r/H
Ceiling Jet Thickness Alpert defined the thick- from less than 1 to 2, with predicted reductions
ness of the ceiling jet, ‘T, as the distance below in the percent of ceiling height at larger scales.
the ceiling where the excess of gas temperature Much of the discussion below deals with
above the ambient value, ΔT, drops to 1/e predictions and correlations of the maximum
(1/2.718 . . .) of the maximum excess tempera- excess temperature and velocity in the ceiling
ture. Based on this definition, measurements jet flow, which occur, as already noted,
obtained with a liquid pool fire 8 m beneath a relatively close to the ceiling surface. Often fire
ceiling show that ‘T/H is about 0.075 at an r/H of detectors or sprinklers are placed at ceiling stand-
0.6, increasing to a value of 0.11 for r/H from off distances that are outside of this region and
about 1 to 2. These results are in good agreement therefore will experience cooler temperatures
with detailed measurements and analysis for the and lower velocities than predicted. In facilities
region r/H < 2 performed by Motevalli and Marks with very high ceilings, the detectors could be
[12] during their small-scale (0.5- and 1.0-m ceil- closer to the ceiling than 1 % of the fire source-
ing heights) experiments. The following correla- to-ceiling separation and will fall in the ceiling
tion for ‘T/H developed by Motevalli and Marks jet thermal and viscous boundary layers. In
from their temperature data confirms the predicted low-ceiling facilities, it is possible for sprinklers
constancy of ceiling jet thickness (at about or detectors to be placed outside of the ceiling jet
10–12 % of H) for r/H from Alpert’s theory: flow entirely if the standoff is greater than 12 %
h of the fire source-to-ceiling height. In this case,
‘T r i
¼ 0:112 1 exp 2:24 response time could be drastically increased.
H H ð14:1Þ
r
for 0:26 2:0 Ceiling Jet Excess Temperature and
H Velocity Alpert [3] has developed easy-to-use
Additional measurements of ceiling jet thick- correlations to quantify the maximum excess
ness, for steady flows induced by strong plumes gas temperature (above the ambient value) and
and for transient flows, are discussed later. velocity at a given position in a ceiling jet flow
www.ebook777.com
14 Ceiling Jet Flows 431
produced by a steady fire. These correlations are cardboard boxes, plastic materials in cardboard
widely used in hazard analysis calculations. boxes, and liquid fuels. Heat release rates for
Evans and Stroup [14] have employed the these fuels range from 600 kW to 98 MW, total
correlations in the development of a generalized ceiling heights range from 4.6 to 18 m, and radial
program to predict heat detector response for positions for most measurements range out to a
the case of a detector totally submerged in the little more than twice the ceiling height. In SI
ceiling jet flow. The correlations are based on units, Alpert’s [3] correlations for maximum
measurements collected during fire tests involv- ceiling jet excess temperatures and velocities
ing fuel arrays of wood and plastic pallets, empty are as follows:
Q_
2=3
T T 1 ¼ 16:9 for r=H 0:18 ð14:2Þ
H5=3
Q_ =H 5=3
2=3
T T 1 ¼ 5:38 for r=H > 0:18 ð14:3Þ
ðr=H Þ2=3
!1=3
Q_
U ¼ 0:947 for r=H 0:15 ð14:4Þ
H
1=3
_
Q=H
U ¼ 0:197 for r=H > 0:15 ð14:5Þ
ðr=H Þ5=6
where temperature, T, is in C; velocity, U, is in varies from about 60 % to 70 % of the actual

m/s; total heat release rate, Q_ is in kW; and radial heat release rate for mixed plastic/cardboard
position and ceiling height (r and H ) are in m. commodities and wood, respectively, with flam-
Data from these fire tests are correlated using mable liquids similar to heptane being in the
the rate at which heat is actually released in the middle of this range. Hence, for general
_ based on measured fuel mass loss rates
fire, Q, commodities, it would be desirable to use ceiling
and the best estimates for actual heat of combus- jet excess temperature and velocity correlations
tion that were available in 1970–1971. Even based on convective heat release rate (see such
though it is the convective component of this correlations in Equations 14.7 and 14.8).
total heat release rate that is directly related to The preceding correlations for both
the buoyancy of the fire, accurate estimates for temperatures and velocities (Equations 14.2,
this convective component were not readily 14.3, 14.4, and 14.5) are broken into two parts.
available for all the fuels tested when the One part applies for the ceiling jet in the area of
correlations were first developed. For the liquid the impingement point where the upward flow
alcohol pool fires that constitute the primary of gas in the buoyant plume turns to flow
basis of the correlation developed by Alpert, the out beneath the ceiling horizontally, with an
convective heat release rate, Q_ c , is now known to
assumed unchanged velocity magnitude. The
impingement point or turning region correlations
be about 74 % of the actual heat release rate.
(Equations 14.2, 14.3 and 14.4) are independent
However, for the remaining solid commodities
of radius and represent plume temperatures
and pallets, the convective heat release rate
432 R.L. Alpert
and velocities calculated at the ceiling height as in Equations 14.3 and 14.5 but based on Q_ c
above the fire source. The other correlations instead of Q_ and zH zv instead of H is shown in
(Equations 14.3 and 14.5) apply outside of this Figs. 14.2 and 14.3. Further details of this reanal-
turning region as the flow moves away from the ysis of ceiling jet data from the early 1970s is
impingement area. It is important to recognize provided in a recent lecture by the author [15].
that these correlations implicitly assume that In Figs. 14.2 and 14.3, the ordinates are the
there is a point buoyancy source for the imping- dimensional quantities
ing plume located at the top surface of the burn-
ing fuel. Hence, there is no dependence on the ðT T 1 ÞðzH zv Þ5=3 U ðzH zv Þ1=3
and
Q_ Q_
horizontal dimension (e.g., effective diameter) of 2=3 1=3
c c
the fire source in the correlations, and the height
of the fuel array is restricted to being a small respectively. Values for these ordinates at the
fraction of the ceiling height. plume axis (see Chap. 13) are shown for compar-
In order to remedy deficiencies discussed ison with the ceiling jet values. Regression fits
above in the existing correlations, much of the based on data only from the ethanol pool fires, for
original data from the 1970s has been reanalyzed which heat release rates are known most accu-
by the author. This new analysis ignores a small rately, are also shown in these figures. Based on
number of measurements where an accurate esti- the data from all of the full-scale fire tests, the
mate of convective heat release rate for the following new correlations are obtained for
corresponding fire sources (plastic pallets and excess gas temperature and velocity in the ceiling
one cardboard box commodity) would be very jet:
difficult to obtain. For the remaining large-scale 0:678
Q_ c
2=3
fire tests, the convective heat release rate is cal- r
T T 1 ¼ 7:22
culated from measurements of fuel mass loss rate ðzH zv Þ5=3 zH zv
and handbook values of the convective heat of ð14:7Þ
combustion. Instead of arbitrarily correlating
measurements with the ceiling elevation above r
for > 0:16
the top fuel surface, H, the new analysis uses the zH zv
ceiling elevation above the location of the virtual
1:017
Q_ c
1=3
plume origin, zH zv. The location, zv, of the r
U ¼ 0:229
virtual plume origin above a reference location ðzH zv Þ1=3 zH zv
is given (see Chap. 13) by a correlation based on
the actual heat release rate, Q,_ and the effective r
diameter, Deff, of the fuel array. When height in for > 0:228 ð14:8Þ
zH zv
the plume, z, is measured from the base of the
r
flame zone instead of the top surface of the burn- where the limits on in Equations 14.7 and
ing fuel, the following expression for virtual zH zv
origin height has been found to be applicable 14.8 are obtained from the respective intersections
even to complex fuel arrays: of the ceiling jet regression fits with the axis values
shown from the plume correlations. Within the
r
turning region (i.e., for zH z the limits shown)
zv ¼ 0:083Q_ 1:02Deff
2=5
ð14:6Þ v
existing correlations for maximum temperature
In Equation 14.6, zv is the distance above the and velocity in the plume can be used.
base of a burning fuel array and Deff is the diame- A further improvement to the ceiling jet
ter of a circle having the same plan area as for the excess temperature and velocity correlations
actual array. The result of correlating excess tem- can be obtained [15] by using just the ethanol
perature and velocity measurements from full- pool and heptane spray fire data, not only
scale tests using the same functional relationships because these are the best documented fire
www.ebook777.com
100
Ceiling Jet Excess Temperature Correlation Based on

Symbol Fire Description Deff [m] - Clg Ht, zH[m]
Ethanol Pool, 0.963x0.963m 1.087 - 8.6

Convective HRR and ZH-ZV [Km5/3/kW2/3] Heptane Spray, 174-434 g/s 3.66 - 4.6 & 7.9
PS Jars in Boxes, 4.11 m high 2.94 - 18
PE Bottles in Boxes, 4.57 m high 2.77 -18
Wood Pallets, 2.44 m high 1.38 - 18
Value at Axis from Heskestad Plume Correlation
Power-Law Fit to Ethanol Pool Fire Data
10
y = 6.5175x-0.655
1
0.1 1 10
r/(zH-zv) [-]
Fig. 14.2 Correlation of maximum ceiling jet excess the commodity with descriptions in a test report; Also
temperature data from full-scale fire tests (Note that “PE note that previously, data for heptane sprays had not
bottles” previously was called “PVC bottles,” an error contained the virtual origin correction applied to data for
discovered by comparing known test numbers involving the other commodities, due to a spreadsheet error)
1 Symbol Fire Description

Ceiling jet velocity correlation based on convective
Ethanol pool fires

Heptane spray fires, 434 & 521 g/s
Wood pallet fire
Value at axis from Heskestad plume correlation
HRR and zH – zv [m4/3/(s kW1/3)]
Power-law fit to ethanol pool fire data
y = 0.241x –1.0865
0.1
0.1 1 10
r/(zH – zv) [–]
Fig. 14.3 Correlation of maximum ceiling velocity data from full-scale fire tests (See legend from Fig. 14.2)
434 R.L. Alpert
sources from the original study, in terms of available at time) in the original ceiling-jet
combustion parameters, but also because these study. As a result, the value of Q_ is 13 % greater
are the only near steady-state fire sources. Fires for ethanol and 8 % less for heptane compared
in piles of solid fuels are inherently transient, to the values used for the correlations in
which makes a data correlation difficult when Equations 14.7 and 14.8. With these new
transient velocity and temperature data are not values, virtual source heights and convective
available. heat release rates can be obtained [15]. For
Modern handbook values [16] for actual and both the heptane spray and ethanol pool data
convective heats of combustion for the two taken together, the resulting regression fit
fuels selected will now be used instead of equations and regression coefficients (R2
what had been assumed (from the knowledge values) are given below:
0:6545
Q_ c
2=3
r r
T T 1 ¼ 6:721 R2 ¼ 0:958 for > 0:134 ð14:7AÞ
ðzH zv Þ 5=3 zH zv zH zv
1:0739
Q_ c
1=3
r r
U ¼ 0:2526 R2 ¼ 0:972 for > 0:246 ð14:8AÞ
ðzH zv Þ 1=3 zH zv zH zv
Certain constraints should be understood burner can be found in the section following on
when applying these correlations in the analysis “Corner Configuration with Strong Plumes”.
of fire flows. The correlations apply only during Experiments have shown that unless great
times after fire ignition when the ceiling flow care is taken to ensure that the fuel perimeter is
may be considered unconfined; that is, no in contact with the wall surfaces, the method of
accumulated warm upper layer is present. Walls reflection used to estimate the effects of the walls
close to the fire affect the temperatures and on ceiling jet temperature will be inaccurate. For
velocity in the ceiling jet independent of any example, Zukoski et al. [17] found that a circular
effect on the fire-burning rate due to radiant burner placed against a wall so that only one
heat received from the walls. The correlations point on the perimeter contacted the wall
were developed from test data to apply in cases behaved almost identically to a fire far from the
where the fire source is at least a distance 1.8 wall with plume entrainment only decreasing by
times the ceiling height from the enclosure walls. 3 %. When using Equations 14.2, 14.3, 14.4, and
For special cases where burning fuel is located 14.5, this fire would be represented by replacing
against a flat wall surface or two wall surfaces Q_ with 1:05Q_ and not 2Q_ as would be predicted
forming a 90 corner, the correlations are by the method of reflections. The value of 2Q_
adjusted based on the method of reflection. This would be appropriate for a semicircular burner
method makes use of symmetry to account for with the entire flat side pushed against the wall
the effects of the walls in blocking entrainment surface.
of air into the fire plume. For the case of a fire Consider the following calculations, which
adjacent to a flat wall, 2Q_ is substituted for Q_ in demonstrate typical uses of the correlations,
the correlations. For a fire in a 90 corner, 4Q_ is Equations 14.2, 14.3, 14.4, and 14.5.
substituted for Q_ in the correlations [3]. More (a) The maximum excess temperature under a
accurate formulas for ceiling jet gas temperature, ceiling 10 m directly above a 1.0-MW heat-
as well as ceiling heat flux, in a 90 corner that release-rate fire is calculated using Equa-
were obtained from experiments with a propane tion 14.2 as
www.ebook777.com
16:9ð1000Þ2=3 based on alcohol pool-fire tests performed at the

T T1 ¼ U.K. Fire Research Station in the 1950s. These
105=3
correlations are cast in the following heat release
16:9ð100Þ
¼ rate, excess temperature, and velocity variables
46:42 that are nondimensional (indicated by the super-
ΔT ¼ 36:4 C script asterisk) and applicable to steady-state
fires under unconfined ceilings (indicated by the
(b) For a fire that is against noncombustible
subscript 0):
walls in a corner of a building and 12 m
below the ceiling, the minimum heat release Q_
Q_ 0 ¼
*
rate needed to raise the temperature of the ð14:9Þ
ρ1 c p T 1 g1=2 H 5=2
gas below the ceiling 50 C at a distance 5 m
from the corner is calculated using Equa-
ΔT=T
tion 14.3 and the symmetry substitution of ΔT *0 ¼ 2=3 ð14:10Þ
4Q_ for Q_ to account for the effects of the Q_
*
0
corner as
2=3 5=3 pffiffiffiffiffiffi
U= gH
4Q_ =H U *0 ¼ 1=3 ð14:11Þ
T T 1 ¼ 5:38
Q_ 0
*
ðr=HÞ2=3
2=3
4Q_ Figure 14.4 shows a plot of the Heskestad
50 ¼ 5:38 2=3
12 5=3
ð5=12Þ correlation for excess temperature and velocity
3=2 data as solid line curves. The correlations
5 50ð12Þ
Q_ ¼ developed by Alpert [3] are plotted as broken
4 5:38 curves, using the same dimensionless para-
Q_ ¼ 1472 kW ¼ 1:472 MW meters with assumed ambient temperature of
293 K (20 C), normal atmospheric pressure,
(c) The maximum velocity at this position is and convective heat release rate equal to the
calculated from Equation 14.5, modified to total heat release rate, Q_ c ¼ Q.
_ Generally, the
account for the effects of the corner as results of Heskestad [7] predict slightly higher
1=3 excess temperatures and substantially greater
_
4Q=H
U ¼ 0:197 gas velocities than Alpert’s [3] results.
ðr=H Þ5=6 Another curve shown in Fig. 14.2 is a fit to
the mean ceiling jet velocity predicted by the
0:197ð5888Þ1=3
¼ generalized theory of Alpert [4], which also
ð5=12Þ5=6 121=3 predicts that the turning-region boundary
U ¼ 3:2 m=s should be at r/H ¼ 0.17. This predicted veloc-
ity is reasonably close to Heskestad’s [7]
Nondimensional Ceiling Jet Relations experimental correlation for velocity. Based
Heskestad [7] developed correlations1 for maxi- on the results shown in Fig. 14.4, the nondi-
mum ceiling jet excess temperature and velocity mensional excess temperature from the
Heskestad [7] correlation and the nondimen-
sional velocity from Alpert’s theory [4, 13] are
1
Originally developed by G. Heskestad and C. Yao in recommended for the prediction of steady ceil-
“A New Approach to Development of Installation ing jet flows beneath unobstructed ceilings.
Standards for Fired Detectors,” Technical Proposal The Heskestad correlation and the Alpert the-
No. 19574, prepared for The Fire Detection Institute, by
Factory Mutual Research Corporation, Norwood, MA ory are adequately fit, respectively, by the
(1971). following expressions:
436 R.L. Alpert
Fig. 14.4 Dimensionless 10

correlations for maximum
ceiling jet temperatures and
Nondimensional ceiling jet velocity

velocities produced by
steady fires. Solid line:
and excess temperature

Heskestad [7]; dotted line:
Alpert [3]
1
U 0*
0.1
0.1 1 10
r/H
Nondimensional excess temperature: Alpert3
Nondimensional excess temperature: Heskestad7
Nondimensional velocity: Alpert3
Nondimensional velocity: Heskestad7
Nondimensional velocity theory: Alpert3
4=3
ΔT *0 ¼ 0:225 þ 0:27Hr for 0:2 r=H < 4:0 ð14:12Þ
ΔT *0 ¼ 6:3 for r=H 0:2 ð14:13Þ
r 0:69
U *0 ¼ 1:06 for 0:17 r=H < 4:0 ð14:14Þ
H
U *0 ¼ 3:61 for r=H 0:17 ð14:15Þ
Heskestad and Delichatsios [18] examined the just steady-state alcohol pool fires having weakly
original data from Heskestad [7] and concluded buoyant plumes. For example, this relationship
that nondimensional velocity and temperature between ceiling jet velocity and excess tempera-
could be related by the following equation: ture is consistent with measurements [18] for
U *0 r 0:63 time-dependent fires having strong plumes.
qffiffiffiffiffiffiffiffi
ffi ¼ 0:68 for r=H 0:3 Other methods used to calculate ceiling jet
ΔT *0 H velocity and maximum possible (when the ceil-
ing is adiabatic) ceiling jet temperatures are
ð14:16Þ
reported by Cooper and Woodhouse [9]. A criti-
The preceding relation has been found appli- cal review of correlation formulas for excess
cable to a much wider range of conditions than temperature and velocity in the ceiling jet under
www.ebook777.com
a variety of conditions has been assembled by burner fires having heat release rates from 12 to
Beyler [19]. To apply these and the preceding 764 kW and beneath ceilings up to 2.5 m in
expressions to realistic burning situations, it is height. This correlation is found to be accurate
recommended that the convective heat release for ratios of free flame height to ceiling height
rate should be used. less than or equal to about 2.0. At greater flame-
height ratios, significant heat released in the ceil-
ing jet itself appears to be the cause for a lack of
Strong Plume-Driven Flow Field agreement with the correlation.
When the flame height of a fire plume is compa- Flame Lengths in the Ceiling Jet It is very
rable to the height of the ceiling above the burn- interesting to note an often-overlooked finding of
ing fuel, the resultant ceiling jet is driven by a Heskestad and Hamada [6]. When there is flame
strong plume. Additional information about this impingement on the ceiling (flame-height ratio
type of flow field is provided in the section on >1), the mean flame radius along the ceiling from
“Sloped Ceilings” (see the special case of zero the plume centerline is observed to be about equal
inclination angle, i.e., a horizontal ceiling). to the difference between the free flame height and
the ceiling height. Hence, Heskestad and Hamada
Ceiling Jet Temperature Heskestad and find that the total average length of flame from the
Hamada [6] measured ceiling jet temperatures burning fuel to the flame tip under the ceiling is
for ratios of free flame height (in the absence of virtually the same as the free flame height.
a ceiling, obtained from existing knowledge of In an earlier study involving small
flame heights) to ceiling height ranging from 0.3 (0.36–8 kW) pool fires beneath ceilings up to
up to 3. A correlation of excess temperatures 0.336 m in height, Yu (You)2 and Faeth [10]
could be achieved by using the plume radius, b, measure the mean flame radius along the ceiling.
at the ceiling as a normalizing length scale, rather Their results yield a flame radius about one-half
than the ceiling height used for the case of a weak the difference between the free flame height and
plume. This correlation takes the form: the ceiling height, or one-half that of Heskestad
r 1 h and Hamada, perhaps due to the smaller scale of
ΔT r i their experiment.
¼ 1:92 exp 1:61 1
ΔT p b b
r
for 1 40 Ceiling Jet Thickness For strong plumes,
b
Atkinson and Drysdale [20] demonstrate that
ð14:17Þ
much of the plume kinetic energy is lost (possi-
where ΔTp is the excess temperature on the bly 75 % of that in the incident plume) during the
plume centerline at the level of the ceiling process of ceiling impingement. As a result of
(obtained from Equations 14.2 or 14.13 or other this kinetic energy loss, the initial ceiling jet
fire-plume relations) and b is the radius where thickness after the turning region may be twice
the velocity of the impinging plume is one-half that expected for the case of weak plumes, about
the centerline value. The expression for this char- 11 % of the ceiling height at r/H ¼ 0.2.
acteristic plume radius is given by Measurements made by Atkinson and Drysdale
and by Yu [5] show that the ceiling jet thickness
h i1=2 T 1=2 Q_ 2=5 may reach a minimum of 8 % of ceiling height at
4=5
b ¼ 0:42 c p ρ1 T 3=5
1 g
2=5 p c r/H ¼ 0.5 and then increase up to 12 % of ceiling
ΔT 3=5
p height at large radial distances, as for weak
ð14:18Þ plumes.
The Heskestad and Hamada [6] correlation is

2
derived from measurements made with propane H. Z. Yu formerly published under the spelling You.
438 R.L. Alpert
applied to actual heat-transfer problems, the ceil-

Convective Heat Transfer ing height should be corrected for the location of
to Horizontal Unconfined Ceilings the virtual point source for the plume.
Note that the heat-flux parameter on the left
Convection is the dominant mode of heat transfer side of Equation 14.19 is proportional to the
for the case of weak plumes impinging on classic heat-transfer Stanton number and that
ceilings. This heat-transfer regime is important the Rayleigh number is proportional to the cube
for the prediction of activation times for detection of the plume Reynolds number, Re (defined in
devices and the prediction of damage for objects, terms of centerline velocity, characteristic plume
such as cables or pipes, suspended below the diameter, 2b, and kinematic viscosity at the
ceiling. However, damage to the ceiling structure plume centerline temperature).
itself will much more likely be the result of strong Equation 14.19 has been established for
plume (flame) impingement, for which heat trans- mainly weak plumes with Rayleigh numbers
fer due to thermal radiation will be just as impor- from 109 to 1015. Kokkala [22] has verified this
tant or more important than convection [21]. The impingement zone heat-transfer correlation,
maximum convective heat flux to a ceiling occurs using up to 10 kW natural gas flames, for flame
when the ceiling surface is at or near ambient heights up to 70 % of the ceiling height. For
temperature, T1, before there has been any greater flame height to ceiling height ratios,
significant heating of the ceiling material. This Kokkala [22] finds that heat-transfer rates are
maximum convective flux is the subject of the many times higher than predicted, partly due to
following discussion. For additional discussion of thermal radiation.
ceiling heat loss, see Chap. 25. Alpert [23] performed small-scale (0.3 m
ceiling height) experiments at elevated air
pressures, which allow Rayleigh numbers greater
Weak Plume Impingement (Turning) than 2 1015 to be achieved while maintaining
Region somewhat better control of ambient disturbances
than in 1-atm experiments. Results of these
Quantification of convective heat transfer from experiments essentially confirm the predictions
weak fire plumes impinging on ceiling surfaces
of the correlation in Equation 14.19, as well as
has been an area of research activity for many
an expression recommended for the plume
years. In the turning region, a widely used corre- impingement region by Cooper [8]. The latter
lation is derived by Yu and Faeth [10] from
expression yields nondimensional ceiling heat
experiments with small pool fires (convective
transfer, in terms of the plume Reynolds number
heat release rates, Q_ c , from 0.05 to 3.46 kW; defined by Alpert [23], as follows:
ceiling heights, H, less than 1 m). This correlation
00
gives convective heat flux to the ceiling, q_ , as 00
q_ H 2
¼ 49 Re1=2
00
q_ H 2 31:2 38:6 Q_ c
¼ 3=5 ¼ 1=6 ð14:19Þ !1=2
_ 1=6 Ra
Qc Pr Ra Q_ c H2=3
1=3
¼ 105 ð14:21Þ
where Pr is the Prandtl number, and the plume v
Rayleigh number, Ra, is given by
gQ_ c H 2 0:027Q_ c H 2 Although Equations 14.19 and 14.21 have

Ra ¼ ¼ ð14:20Þ identical dependence of impingement heat flux
3:5 pv3 v3
on fire heat release rate and ceiling height, heat-
for gases similar to air, having ambient absolute flux values from Equation 14.21 are about 50 %
pressure, p, and kinematic viscosity, ν. It is higher, since this expression is derived from data
recommended that when these expressions are on turbulent jets.
www.ebook777.com
00
Ceiling Jet Region be expressed as follows, since q_ ¼ hΔT, with
ΔT given by Equation 14.3:
Outside of the turning region, the convective flux 00
q_ H2 r 1:36 r
to the ceiling is known to drop off sharply with
¼ 1:323 f for 0:2 < 4:0
increasing radial distance from the plume axis. Q_ c H H
The experiments of Yu and Faeth [10] described ð14:25Þ
in the preceding section were also used to deter-
mine this radial variation in ceiling jet convec- Equations 14.22 and 14.25 are in good agreement
tive flux. Their own data, as well as data from for a friction factor of 0.03, which is comparable
small-scale experiments (ceiling heights of 0.5 to with the value of 0.02 deduced from Alpert’s [4]
0.8 m) by Alpert [13] and by Veldman [11] are theory.
all consistent with the following correlation that
is given by Yu and Faeth [10]3:
Sloped Ceilings
00
q_ H 2 r 1=3 r
¼ 0:04 for 0:2 < 2:0
Q_ c H H There have been very few studies of the ceiling
ð14:22Þ jet flow resulting from plume impingement on an
inclined, flat ceiling, i.e., where the ceiling is
An alternate derivation of Equation 14.22 can inclined at some angle, θ, to the horizontal. One
be obtained by using Alpert’s correlation for such study, by Kung et al. [24], obtained
ceiling jet excess temperature (Equation 14.3) measurements showing pronounced effects in
and Alpert’s theory for average ceiling jet veloc- the velocity variation along the steepest run
ity (Equation 14.14) with the Reynolds/Colburn from the point of impingement of a strong
analogy, as discussed by Yu and Faeth [10] and plume, both in the upward and downward
Veldman [11]. From the Reynolds/Colburn anal- directions. In the upward direction, the rate of
ogy, the heat-transfer coefficient at the ceiling, h, velocity decrease with distance, r, from the inter-
should be related to ceiling jet average velocity section of the plume vertical axis with the ceiling
and density as follows: was reduced significantly as the ceiling slope
increased. In the downward direction, the flow
h f
¼ Pr2=3 ð14:23Þ separated from the ceiling and turned upward at a
ρ1 Ucp 2 location, r, denoted by Kung et al. [24] as the
where Pr is the Prandtl number and f is the ceiling penetration distance. These results were the out-
friction factor. By using Equation 14.14 for aver- come of experiments with 0.15- and 0.228-m-
diameter pan fires located 0.279 to 0.889 m
age ceiling jet velocity, U, the ceiling heat-
transfer coefficient becomes beneath an inclined 2.4-m square ceiling and
were limited to convective heat release rates in
!
Q_ c
1=3
r 0:69 r
the range of 3–13 kW.
h ¼ 0:246 f for 0:17 < 4:0 Following Heskestad and Hamada [6], Kung
H H H
et al. developed correlations by scaling near-
ð14:24Þ maximum excess temperature and velocity, as
well as radial distance along the ceiling, in
With f ¼ 0.03, Equation 14.24 is identical to
terms of the quantities in the undeflected plume
the simplified expression listed in Beyler’s exten-
at the impingement point. These correlations take
sive compilation [19]. The nondimensional heat
the following form:
flux to a ceiling at ambient temperature can then
r 0:7
ΔT
¼ exp ð0:12 sin θ 0:42Þ 1
3
ΔT p b
Note that there is a typographical error in the exponent
of r/H in Equation 14.17 of this reference. ð14:26Þ
440 R.L. Alpert
r 0:6
U perpendicular to the “upslope” directions are
¼ exp ð0:79 sin θ 0:52Þ 1 provided for slope angles from 0 to 60 . In
Vp b
many of these experiments, there is flame
ð14:27Þ
impingement on the ceiling and flame in the
for r/b 1(upward direction from the im- ceiling jet.
pingement point, i.e., r ¼ rup) and θ ¼ 0 30 ; Very recently (about 10 years after the pre-
r ceding studies by Sugawa et al.), new
ΔT
¼ ð0:15 sin θ 0:11Þ correlations have been developed by Y. Oka
ΔTp b and colleagues at Yokohama National University
þ 0:97 0:06 sin θ ð14:28Þ from significant additional measurements under
r ceilings inclined from the horizontal up to a
U maximum angle, θ, of 40 and having a center-
¼ ð0:21 sin θ 0:10Þ
Vp b ð14:29Þ line clearance, H, above the fuel surface of 1 m.
þ 0:99 1:17 sin θ In these studies [26, 27], ceiling jet gas velocity
is measured not only with bi-directional tubes but
for r/b < 0 (downward direction from the also with a Particle Imaging Velocimetry (PIV)
impingement point), valid only for θ ¼ 0 30 , system that makes use of smoke particles natu-
and for ΔT and U 0. rally present in the plume from the 0.285 m
In Equations 14.26, 14.27, 14.28, and 14.29, square heptane pan fire (heat release rate of
the characteristic plume radius is proportional to 43 kW). By having two methods for measuring
that defined in Equation 14.18 but with a slightly velocity, the authors [26] determined that the
different magnitude, namely, bidirectional probe generally provided a bulk
mean velocity whereas the high resolution PIV
h 4=5 i1=2 T 1=2 Q_ 2=5
b ¼ 0:548 c p ρ1 T 3=5 2=5 p c system could provide a true maximum velocity.
1 g
ΔT 3=5
p The ratio of the former to the latter is determined
ð14:30Þ to be 0.828 instead of 0.707, as would be
expected for the classic half-Gaussian velocity
Equation 14.29 shows that the ceiling jet profile. Algebraic expressions for ceiling jet
velocity first becomes zero in the downward velocity have been obtained [26] both for the
direction at values of r/b equal to 5.6, 3.5, case of the flame tip below the inclined ceiling
and 2.0 for ceiling slopes of 10 , 20 , and 30 , and for flame impingement on the inclined ceil-
respectively. ing. In the latter case, gas velocity continues to
About 10 years after the work by Kung et al., increase in the flame zone in the steepest upward
additional measurements of gas temperature direction from the impingement point (as occurs
and velocity under inclined ceilings with in the fire plume itself) before becoming nearly
thermocouples and bidirectional tubes, respec- constant and then decreasing with distance in the
tively, were obtained by Sugawa et al. [25]. upward direction. Oka et al. [27] have now devel-
These experiments involved nearly full-scale oped, from data for 0 < θ < 40 , a single alge-
conditions (ceiling clearance, H, on the fuel cen- braic expression for maximum ceiling jet gas
terline of 1.25–2.5 m) with a 200 mm diameter velocity, Uup, at the steepest upward radial dis-
propane gas burner fire providing 10–100 kW heat tance, rup, from plume impingement, which
release rates. Formulas for ceiling jet excess tem- covers cases both with and without flame
perature and velocity along the “upslope” and impingement on a ceiling, as follows:

U up r up cos θ β * 1=3
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi ¼ α Qc ð1 þ sin θÞ ð14:31Þ
g H þ r up sin θ H þ r up sin θ
www.ebook777.com
where, for

0:037 < r up cos θ=H þ r up sin θ 0:151 α ¼ 6:051, β ¼ 0:458
0:151 < r up cos θ=H þ r up sin θ 0:350 α ¼ 2:540, β¼0
0:350 < r up cos θ= H þ r up sin θ 1:80 α ¼ 0:855, β ¼ 1:040
and Qc* ¼ the usual definition (see Equa- temperature from extensive thermocouple
tion 14.9) with Q_ the convective component of measurements that covers cases both with and
heat release rate without flame impingement on the ceiling is
The corresponding algebraic expression given by the following:
for near-maximum excess ceiling jet gas

ΔT r up cos θ 0:781 * 2=3
¼ 2:778 Qc ð1 þ sin θÞ 0:1 r up cos θ=H 2:4 ð14:32Þ
T1 H
The detailed velocity and excess temperature assumed. This assumption implies that when a
measurements discussed above have allowed change in heat release rate occurs at the fire
Oka et al. [26] also to derive algebraic source, full effects of the change are immediately
expressions for Gaussian ceiling jet velocity felt everywhere in the flow field. In a room-sized
and excess temperature thickness under sloped enclosure, under conditions where the fire is
ceilings. They determined that ceiling jet thick- growing slowly, this assumption is reasonable.
ness was not affected by flame impingement as However, in other cases, the time for the heat
long as data corresponding to any region of con- release rate to change significantly may be com-
tinuous flaming are excluded. The expression for parable to or less than the time, tf ti, for gas to
Gaussian thermal thickness under a sloped ceil- travel from the burning fuel to a detector
ing is given by: submerged in the ceiling jet. The quasi-steady
LT h r i assumption may not be appropriate in this situa-
up
¼ f0:00254θ þ 0:112g 1 exp βT tion, unless the following condition is satisfied,
H H depending on the accuracy desired:
ð14:33Þ
Q_
where βT ¼ 2.91 + 2.20[1 exp(0.0662θ)] > t f ti ð14:34Þ
_
dQ=dt
and valid for 0.4 rup/H 2.4; 0 θ 40 .
where ti is an ignition reference time.
The quasi-steady assumption, together with
Time-Dependent Fires the strong plume-driven ceiling jet analysis of
Heskestad and Hamada [6], has been used by
Quasi-Steady Assumption Kung et al. [28] to correlate ceiling jet velocity
and temperature induced by growing rack-
For time-dependent fires, all estimates from the storage fires. Although gas travel times for
previous section may still be used, but with the these large-scale experiments may amount to
constant heat release rate, Q,_ replaced by an many seconds, Equation 14.34 shows that a suf-
appropriate time-dependent Q_ ðtÞ. In making this ficiently small fire-growth rate allows a quasi-
replacement, a “quasi-steady” flow has been steady analysis to be used.
442 R.L. Alpert
Testing has shown that the heat release rate were validated in an extensive series of tests
during the growth phase of many fires can often conducted by Factory Mutual Research Corpora-
be characterized by simple time-dependent poly- tion [18, 31], where measurements were made of
nomial or exponential functions. The most exten- maximum ceiling jet temperatures and velocities
sive research and analysis have been performed during the growth of fires in three different sizes
with heat release rates that vary with the second of wood crib. Subsequent to this original experi-
power of time. mental study, Heskestad and Delichatsios [32]
_ computed for
corrected the heat release rate, Q,
the crib tests and also generalized their results to
Power-Law Fire Growth other types of fuels by using the more relevant
convective heat release rate, Q_ c . The resulting
The growth phase of many fires can be dimensionless correlations for maximum ceiling
characterized by a heat release rate increasing jet temperatures and velocities are given by
proportionally with a power, p, of time measured
from the ignition reference time, ti, as follows: ΔT *2 ¼ 0 t*2 t*2 f ð14:37Þ
Q_ ¼ αðt ti Þ p ð14:35Þ !4=3

t*2 t*2 f
Figure 14.5 shows one case where the heat ΔT *2 ¼ t*2 > t*2 f
0:126 þ 0:210r=H
release rate for a burning foam sofa during the
growth phase of the fire, more than 80 s (ti) after ð14:38Þ
ignition [29], can be represented by the following r 0:63
equation: U *2
qffiffiffiffiffiffiffiffi
ffi ¼ 0:59 ð14:39Þ
ΔT *2 H
Q_ ¼ 0:1736ðt 80Þ2 ð14:36Þ
Heskestad [30] used the general power-law where

behavior given by Equation 14.35 to propose a t ti
set of theoretical modeling relations for the tran- t*2 ¼ 1=5 ð14:40Þ
Aαc H4
sient ceiling jet flow that would result from such
a time-varying heat release rate. These relations
Fig. 14.5 Heat release Foam Sofa

rate history for a burning 4000
foam sofa [29]
•
Q = 0.1736 (t – 80)2
Rate of heat release (kW)
3000
2000
1000
0
0 100 200 300 400 500
Time (s)
www.ebook777.com
U quasi-steady limit is achieved, as shown by an

U *2 ¼ ð14:41Þ
ðAαc H Þ 1=5 alternative method by Heskestad [18]. The value
of the quasi-steady excess temperature, (ΔT2*)qs,
in this limit of t2* » (t2*)f becomes, from
ðT T 1 Þ=T 1 Equation 14.38
ΔT *2 ¼ ð14:42Þ
ðAαc Þ2=5 g1 H 3=5 4=3
t*2
ΔT *2 qs ¼ ð14:48Þ
g 0:126 þ 0:210r=H
A¼ ð14:43Þ
ρ1 cp T 1
The preceding correlations of ceiling jet
temperatures and velocities are the basis for the
Q_ c calculated values of fire detector spacing found
αc ¼ ð14:44Þ
ðt t i Þ2 in NFPA 72®, National Fire Alarm Code®,
Appendix B, “Engineering Guide for Automatic
* r Fire Detector Spacing” [33]. In NFPA 72, three
t2 f ¼ 0:813 1 þ ð14:45Þ
H or four selected fire heat release rates assumed to
increase proportionally with the square of time
and where dimensionless variables are indicated
are used as the basis for the evaluation. These
with the superscript asterisk.
fire heat release rate histories are chosen to
Notice that in Equation 14.38 the dimension-
be representative of actual fires involving
less time, t2* , has been reduced by the time (t2*)f.
different commodities and geometric storage
This reduction accounts for the gas travel time,
arrangements. The chosen release rate histories
tf ti, between the fire source and the location of
are as follows:
interest along the ceiling at the specified r/H. For
dimensionless times after ignition less than (t2*)f, Slow Q_ ¼ 0:00293t2 ð14:49Þ
the initial heat front has not yet arrived at r/H so
the gas temperature is still at the ambient value,
as shown in Equation 14.37. In dimensional Medium Q_ ¼ 0:01172t2 ð14:50Þ
terms, the gas travel time is given by the follow-
ing, after using the definition of t2* in Fast Q_ ¼ 0:0469t2 ð14:51Þ
Equation 14.45:
Ultrafast Q_ ¼ 0:1876t2 ð14:52Þ
4=5 0:813ð1 þ r=H Þ
t f ti ¼ H ð14:46Þ
ðAαc Þ1=5 where Q_ is in kW and t is in s.
Substitution of Equation 14.35 into Equa- EXAMPLE Sofa fire: Consider how the follow-
tion 14.34 shows that for power-law fire growth, ing calculation demonstrates a use of the correla-
the quasi-steady assumption will always be valid tion (Equations 14.38 and 14.39) for calculating
beginning at a sufficiently long time after igni- the ceiling jet maximum temperature and veloc-
tion. For the specific case of t2 fire growth, sub- ity produced by a t2 fire growth.
stitution of Equation 14.44 and the expression for A foam sofa, of the type analyzed in Fig. 14.5,
the gas travel time, Equation 14.46, into Equa- is burning in a showroom 5 m below a suspended
tion 14.34 results in the following requirement if ceiling. The showroom temperature remote from
a quasi-steady analysis is to be appropriate: the fire remains at 20 C at floor level as the fire
begins to grow. Determine the gas temperature
t ti 0:813ð1 þ r=H Þ and velocity at the position of a ceiling-mounted
> H4=5 ð14:47Þ
2 ðAαc Þ1=5 fire detector submerged in the ceiling jet flow 4 m
away from the fire axis when the convective heat
In the limit of very large values of t ti, release rate (assumed to equal the total heat
Equation 14.47 will always be satisfied and a release rate) first reaches 2.5 MW.
444 R.L. Alpert
Figure 14.5 shows that the heat release rate EXAMPLE Rack storage: Yu and Stavrianidis
from the sofa first reaches 2.5 MW (2500 kW) at [34] were interested in predicting activation
about 200 s after ignition. Using the analytic times of quick-response sprinklers protecting
formula for the time-dependent heat release high rack storage of plastics. Since the sprinklers
rate, Equation 14.36, the time from the virtual are activated typically in less than 1 min by the
ignition of the sofa at 80 s to reach 2500 kW is ceiling jet flow, information on flow temperature
and velocity shortly after ignition is required.
2500 0:1736ðt 80Þ2 The objective was to correlate properties of the
ðt 80Þ ¼ 120 s ceiling jet induced by fires in 2- to 5-tier-high
In this problem, the low-level heat release rate up rack storage, consisting of polystyrene cups
packaged in corrugated paper cartons on pallets.
to 80 s after actual ignition of the sofa is ignored.
Thus, the sofa fire can be treated as having started When this fuel array is ignited at its base,
at t ¼ 80 s and grown to 2.5 MW in the following the initial growth period (tf ti 25 s) can be
characterized as heat release rates increasing by
120 s. Equations 14.40, 14.41, 14.42, 14.43, 14.44,
and 14.45 are used to evaluate parameters of the third power of time, as follows:
the problem, using the dimensionless correlations Q_ c ¼ αc ðt ti Þ3 ð14:53Þ
for ceiling jet temperature and velocity.
For the sofa fire in the showroom example, where αc ¼ 0.0448. Because of upward and lat-
T1 ¼ 293 K, ρ ¼ 1.204 kg/m3, cp ¼ 1 kJ/kgK, eral flame propagation during the transient rack-
g ¼ 9.8 m/s2, αc ¼ 0.1736 kW/s2, A ¼ 0.0278 storage fire, the virtual origin elevation, zo, of the

m4/kJ · s2, r ¼ 4 m, H ¼ 5 m, t*2 f ¼ 1:46, t plume changes during the course of fire growth,
ti ¼ 120 s, and t*2 ¼ 11:40. For the conditions of as follows:

interest, t*2 > t*2 f , so the correlation (Equa-
zo ¼ 2:4 þ 0:095Q_ c
2=5
ð14:54Þ
tion 14.38) is used to evaluate the dimensionless
ceiling jet temperature thereby complicating the effort to correlate ceil-
4=3 ing jet properties. Nevertheless, Yu and
11:40 1:46 Stavrianidis were able to develop correlations
ΔT *2 ¼
0:126 þ 0:210ð4=5Þ based on the following dimensional temperature
ΔT *2 ¼ 109:3 and velocity variables, which are similar to those
first proposed by Heskestad [30] for power-law
Equation 14.39 is used to calculate the dimen- fire growth:
sionless ceiling jet velocity
ΔT m
pffiffiffiffiffiffiffiffiffiffiffi ΔT^ m ¼ α1=3
c ðH zo Þ1=3 ð14:55Þ
U*2 ¼ 0:59ð4=5Þ0:63 109:3 ¼ 7:10 T1
The dimensional excess temperature and velocity ^ m ¼ α1=6 ðH zo Þ1=3 Um

U ð14:56Þ
c
are calculated using Equations 14.42 and 14.41,
respectively, to yield where the maximum ceiling jet excess tempera-
ture, ΔT^ m , and velocity, Ûm, variables depend on
ΔT ¼ 147 K the following heat release rate and radial distance
T ¼ 147 K þ 293 K ¼ 440 K ¼ 167 C parameters, respectively:
U ¼ 3:37 m=s X ¼ α1=6 ðH zo Þ2=3 Q_ c
1=3
ð14:57Þ
c
The corresponding gas temperature calculated r
with the quasi-steady analysis of Equation 14.48 R^ ¼ ð14:58Þ
H zo
instead of the fire analysis is 197 C.
www.ebook777.com
The exact form of the preceding correlations, in depth, hb, must be greater than the quantity (H/10)
terms of detailed formulas, is provided by Yu and (‘b/H)1/3. Downstream of where the ceiling jet
Stavrianidis [34]. flow is parallel to the beams or corridor walls and
In addition to maximum excess gas tempera- in the absence of spillage, Delichatsios [35] deter-
ture and velocity, Yu and Stavrianidis [34] also mined that the average excess ceiling jet tempera-
measured the depth of the ceiling jet, in terms of ture and velocity within the primary channel are
the distance below the ceiling where the velocity given by the following:
and excess temperature are 1/e of the respective 1=3 " #
maximum values. Results show the ceiling jet ΔT H Y ‘b 1=3
¼a exp 6:67St
depth based on velocity to be very similar to ΔT p ‘b H H
that based on excess temperature and both depths ð14:59Þ
to be fairly insensitive to the transient fire growth
process. Typical values for the ratio of ceiling
pffiffiffiffiffiffiffiffiffiffi H 1=6
jet temperature depth to effective ceiling height, U ¼ 1:102 HΔT ð14:60Þ
‘b
‘T/(H zo), for radial positions, r/(H zo) of
0.217, 0.365, 1.75, and 4.33 are about 0.07, 0.1, under the conditions
0.14, and 0.2, respectively.
Y > ‘b
hb =H > 0:1ð‘b =H Þ1=3
Confined Ceilings
‘b =H > 0:2
Channel Configuration
Y ‘b 1=3
0:5 < < 3:0
H H
Previous discussions of ceiling jets in this chapter
have all dealt with unconfined radial spread of the where
gas flow away from a ceiling impingement point. ΔTp ¼ Excess temperature on the plume center-
In practice this flow may be interrupted by ceiling line defined previously in Equation 14.17
beams or corridor walls, creating a long channel Y ¼ Distance along the channel measured from
that partially confines the flow. Knowledge of the the plume impingement point
resultant ceiling jet flows is important in deter- St ¼ Stanton number, whose value is
mining fire detector response times. For the chan- recommended to be 0.03
nel configuration, the flow near the impingement Based on the minimum value of ‘b/H ¼ 0.2,
point will remain radial (i.e., axisymmetric), but the limit on hb/H implies that the beam depth to
after spreading to the walls or beams that bound ceiling height ratio must be at least 0.17 for the
the ceiling, the flow will become generally paral- fire gases to be restricted to the primary channel.
lel with the confining boundary. Delichatsios [35] The constant a in Equation 14.59 is determined
has developed correlations for steady-state ceil- by Delichatsios to be in the range 0.24–0.29. This
ing jet temperature and velocity, which apply to equation is based on the concept that the channel
the channel flow between beams and down flow has undergone a hydraulic jump, which
corridors. In the case of corridors, the correlations results in greatly reduced entrainment of cooler,
apply when the corridor half-width, ‘b, is greater ambient air from below. Reductions in ceiling jet
than 0.2 times the ceiling height, H, above the fire temperature or velocity are then mainly due to
source. Note that this value of ‘b corresponds heat losses to the ceiling and would thus be
approximately to the outer radius of the ceiling dependent on ceiling composition to some
jet turning region. In the case of beams, the flow extent.
must also be contained fully so that only a flow in Additional detailed measurements of the
a primary channel results, without spillage under ceiling jet flow in a primary beamed channel
the beams to the adjoining secondary channels. have been obtained by Koslowski and Motevalli
For the latter condition to be satisfied, the beam [36]. Their data generally validate the
446 R.L. Alpert
Delichatsios beamed ceiling correlation (Equa- corner, is often used as a hazardous environment
tion 14.61) and ceiling jet flow behavior, but in which to test the flammability of wall and
additional measurements for a range of beam ceiling linings. This wall-ceiling-corner configu-
depth to ceiling height ratios has allowed the ration also occurs naturally in many types of
correlation to be generalized. Furthermore, enclosures (see below) where hot gases from
Koslowski and Motevalli recast the correlation the fire source may be partially or completely
in terms of the nondimensional heat release confined by more than just the ceiling and corner
rate defined by Heskestad and Delichatsios walls themselves, resulting in the formation of a
(Equations 14.9 and 14.10), instead of centerline hot gas layer near the ceiling. In this section, the
plume conditions at the ceiling, with the follow- environment of an open corner with inert lining
ing result: surfaces is discussed, where a ceiling jet
1=3 " #
develops due to impingement of a fire plume or
H Y ‘b 1=3 flames from the source fire at the base of the wall
ΔT 0 ¼ C
*
exp 6:67 St
‘b H H corner onto the ceiling covering the wall-corner.
A careful study of this environment based on
ð14:61Þ
full-scale tests was conducted by Lattimer and
Sorathia [38]. These tests used a ceiling clear-
where the Stanton number is recommended to be
ance of 2.25 m above the surface of a
0.04, rather than 0.03, and the constant, C, has
0.17–0.50 m2 or L-shaped line (each leg being
the following dependence on the ratio of beam
0.17–0.50 m) sand burner having propane heat
depth, hb, to ceiling height, H:
release rates from 50 to 300 kW.
2 Thermocouple measurements [38] of excess
hb hb
C ¼ 25:38 þ 13:58 þ 2:01 gas temperature at a radial distance from the
H H
ð14:62Þ corner, r, in the ceiling jet could be correlated
Y (with a regression coefficient of 0.85) by the
for 0:5 1:6
H following formulas:
To derive Equation 14.62, Koslowski and
rþH
Motevalli vary the hb/H ratio from 0.07 up to T T 1 ¼ 950 for 0:55 ð14:63Þ
L f , tip
0.28. In so doing, they note that C increases
steadily with this ratio until leveling off near
hb/H ¼ 0.17, determined by Delichatsios as the
condition for the fire gases to be restricted to r þ H 2 rþH
T T1 ¼C for > 0:55
the primary channel. Between values of hb/H of L f , tip L f , tip
0.07 (or even much less) and 0.17, spillage from ð14:64Þ
the primary channel to adjacent secondary
channels is steadily reduced, thereby increasing The specific value of 950 for the maximum
temperatures in the primary channel. excess of corner fire gas temperature above
Characteristics of the ceiling jet flow in the sec- ambient in Equation 14.63 may vary for fire
ondary channels, as well as the primary channel, sources other than the propane burner or for
have also been studied by Koslowski and corner walls having thermal characteristics dif-
Motevalli [37]. ferent from those used in these specific tests.
However, it is expected that the functional
dependencies for ceiling jet temperature should
Corner Configuration with Strong be preserved. Note that the constant, C, in Equa-
Plumes tion 14.64 is 288 for the square burner of side, D,
and 340 for the L-shaped line burner, each leg of
An open configuration of two walls at a 90 angle which is length, D and that Lf,tip is the flame
to form a corner, covered by a ceiling, with a fire length from the surface of either type burner to
source at the base of and in close contact with the the flame tip, the furthest location where flame
www.ebook777.com
tips are observed visually, as determined from heated layer of gas will accumulate in the upper
the correlation [38], portion of the enclosure. This heated layer can be
qffiffiffiffiffiffi deep enough to totally submerge the ceiling jet
L f , tip
¼ 5:9 Q* ð14:65Þ flow. In this case, temperatures in the ceiling jet
D
can be expected to be greater than if the ceiling
where Q* is based on actual fire heat release rate jet were entraining gas from a cooler, ambient-
and the burner length-scale, D (instead of the temperature layer. It has been shown by Yu and
usual ceiling clearance, H). Faeth [10] that the submerged ceiling jet also
Lattimer and Sorathia [38] also used twenty results roughly in a 35 % increase in the heat
Schmidt-Boelter gauges to measure heat flux to transfer rate to the ceiling.
the bounding surfaces of the corner configuration There are analytical formulas to predict tem-
from the propane sand burner flames. Their perature and velocity in such a two-layer envi-
00
measurements of total heat flux, q_ , to the ceiling ronment, in which the ceiling jet is contained in a
surface from the ceiling jet flames and/or hot heated upper layer and the fire is burning in a
gases could be correlated by the following for lower, cool layer. This type of prediction, which
either the square or L-shaped line burner: has been developed by Evans [40, 41], Cooper
[42], and Zukoski and Kubota [43], can best be
00 rþH
q_ ¼ 120 for 0:58 ð14:66Þ used to check the proper implementation of read-
L f , tip ily available numerical models (e.g., zone or
3:5 field/CFD) of fire-induced flows in enclosures.
00 rþH rþH
q_ ¼ 18 for > 0:58 An example of a zone model to predict activation
L f , tip L f , tip of thermal detectors by a ceiling jet submerged in
ð14:67Þ a heated layer is the algorithm developed by
Davis [44]. This model, which assumes that ther-
where the flame tip total length is given by Equa-
mally activated links are always located below
tion 14.65, above. This same formula is found
the ceiling at the point of maximum ceiling jet
also to predict the maximum heat flux to the top
temperature and velocity, is based partly on a
portion of the wall from the ceiling jet flow,
model and thoroughly documented software
where now the variable, r, represents distance
developed by Cooper [45].
from the corner along the top of the wall.
Formulas to predict the effect of the heated
Again, the specific maximum heat flux of
upper layer in an enclosure are based on the
120 kW/m2 that was measured in the corner
assumption that the ceiling jet results from a
configuration by Lattimer and Sorathia [38]
fire contained in a uniform environment at the
may vary for fuels with thermal radiation
heated upper-layer temperature. This substitute
characteristics much different from those of pro-
fire has a heat release rate, Q_ 2 , and location
pane or for different burner configurations. For
below the ceiling, H2, differing from those of
example, it is well known that peak heat fluxes in
the real fire. Calculation of the substitute
pool and solid fuel fires can exceed 140–160 kW/
m2, as discussed by Coutts [39]. quantities Q_ 2 and H2, depends on the heat release
rate and location of the real fire, as well as the
depths and temperatures of the upper and lower
General Enclosure Configurations layers within the enclosure.
Following the development by Evans [41], the
The analyses in preceding sections for uncon- substitute source heat release rate and distance
fined ceiling jet flows may be sufficient for below the ceiling are calculated from
large industrial or commercial storage facilities. Equations 14.68, 14.69, 14.70, and 14.71. Origi-
In smaller rooms, or for very long times after fire nally developed for the purpose of sprinkler and
ignition in larger industrial facilities, a quiescent, heat detector response time calculations, these
448 R.L. Alpert
equations are applicable during the growth phase Then, analogous to Equations 14.69, 14.70, and
of enclosure fires. 14.71 of the previous method:
0 13=2
2=3 2=5 1 þ σ 1=3
1 þ CT Q_ I, 1
*
1A Z I, 2 ¼ ZI, 1 ξ3=5 m*2 ð14:74Þ
Q_ I, 2 @
*
¼ ð14:68Þ σ
ξCT CT

σm*2
Q_ c, 2 ¼ Q_ c, 1 ð14:75Þ
8 92=5 1þσ
>
> >
>
< ξQ_ I, 1 CT
* =
ZI, 2 ¼
>

2
Z I, 1 H 2 ¼ H 1 ZI, 1 þ ZI, 2 ð14:76Þ
> * 1=3 * 2=3 >>
:Q_ I, 2 ðξ 1Þ β þ 1 þ ξCT Q_ I, 2 ;
The last step is to use the substitute source values
ð14:69Þ of heat release rate and distance below the ceiling,
as well as heated upper-layer properties for ambi-
Q_ c, 2 ¼ Q_ I, 2 ρ1, 2 c p1 T 1, 2 g1=2 Z I, 2
* 5=2 ent conditions, in the correlations developed for
ð14:70Þ
ceiling jet flows in uniform environments.
To demonstrate the use of the techniques, the
H 2 ¼ H 1 ZI, 1 þ ZI, 2 ð14:71Þ
previous example in which a sofa was imagined to
be burning in a showroom may be expanded.
Further explanation of variables is contained in Let all the parameters of the problem remain the
the nomenclature section. same except that at 200 s after ignition (t ti
Cooper [42] has formulated an alternative cal- ¼ 120 s), when the fire heat release rate has
culation of substitute source heat release rate and reached 2.5 MW, a quiescent heated layer of gas
distance below the ceiling that provides for gener- at a temperature of 50 C is assumed to have
alization to situations in which portions of the accumulated under the ceiling to a depth of 2 m.
time-averaged plume flow in the lower layer are For this case, the two-layer analysis is needed to
at temperatures below the upper-layer tempera- determine the ceiling jet maximum temperature at
ture. In these cases, only part of the plume flow the same position as calculated previously (a radial
may penetrate the upper layer sufficiently to distance of 4 m from the plume impingement point
impact on the ceiling. The remaining portion at on the ceiling).
low temperature may not penetrate into the hotter All of the two-layer calculations presented
upper layer. In the extreme, when the maximum assume quasi-steady conditions. From Equa-
temperature in the lower-layer plume flow is less tion 14.47 with the values of parameters in the
than the upper-layer temperature, none of the single-layer calculation, it can be shown that the
plume flow will penetrate significantly into the time after sofa ignition must be at least 31 s for
upper layer. This could be the case during the a quasi-steady analysis to be acceptable. Since
decay phases of an enclosure fire, when the heat the actual time after ignition is 120 s, such an
release rate is small compared to earlier in the fire analysis is appropriate. It will be assumed that
growth history. In this calculation of substitute fire- this finding will carry over to the
source quantities, the first step is to calculate the two-layer case.
fraction of the plume mass flow penetrating the Using Equations 14.68, 14.69, 14.70, and
upper layer, m2*, from Equations 14.72 and 14.73. 14.71 from the work of Evans [41], values of
1:04599σ þ 0:360391σ 2 the heat release rate and position of the substitute
m*2 ¼ ð14:72Þ fire source that compensates for the two-layer
1 þ 1:37748σ þ 0:360391σ 2
effects on the plume flow can be calculated.
where The dimensionless heat release rate of the real
2 3 fire source evaluated at the position of the inter-
1 þ C Q_ * 2=3 face between the upper and lower layers is as
ξ 6 T I , 1 7
σ¼ 4 15 ð14:73Þ follows:
ξ1 ξ
www.ebook777.com
Q_ Using the ceiling height above the substitute

Q_ I, 1 ¼
*
5=2
ð14:77Þ source, this equation yields the result
ρ1 c p1 T 1 g1=2 ZI, 1
4=3
11:40
For an actual heat release rate of 2500 kW, ambi- ΔT *2 ¼
ent temperature of 293 K, and distance between
qs 0:126 þ 0:210ð4=5:161Þ
the fire source and the interface between the lower ¼ 134:4
and upper layers of 3 m, Equation 14.77 becomes
For the given time after ignition of 120 s and the
2500 assumed fire growth, the calculated Q_ 2 value
Q_ I, 1
*
¼
1:204 1 293 9:81=2 35=2 implies that α equals 0.1606, instead of the origi-
¼ 0:1452 nal sofa fire growth factor of 0.1736. Substitution
of this new α in Equation 14.42, along with H2
Using the ratio of upper-layer temperature to and the upper-layer temperature as the new ambi-
lower-layer temperature, ξ ¼ 323/293 ¼ 1.1024, ent value, yields the following dimensional
and the constant, CT ¼ 9.115, the dimensionless excess temperature at the 4-m radial position in
heat release rate for the substitute fire source is the ceiling jet:
Q_ I, 2 ¼ 0:1179
*
134:4 323 ð0:0278 :01606Þ2=5
ΔT ¼
9:8 5:1613=5
Using the value for the constant β2 ¼ 0.913, the
ΔT ¼ 190 K
position of the substitute fire source relative to
the two-layer interface is T ¼ 190 K þ 323 K ¼ 513 K ¼ 240 C
Z I, 2 ¼ 3:161 This is 73 C above the temperature calculated
previously using the quasi-steady analysis and a
Now, from Equations 14.76 and 14.77, the
uniform 20 C ambient, demonstrating the effect
dimensional heat release rate and position rela-
of flow confinement on gas temperature.
tive to the ceiling are found to be
Q_ 2 ¼ 2313 kW H 2 ¼ 5:161 m
Ceiling Jet Development
The analogous calculations for substitute fire-
source heat release rate and position following At the beginning of a fire, the initial buoyant flow
the analysis of Cooper [42], Equations 14.72, from the fire must spread across the ceiling,
14.73, 14.74, 14.75, 14.76, and 14.77, are driven by buoyancy, to penetrate the cooler
ambient air ahead of the flow. Research studies
σ ¼ 23:60
designed to quantify the temperatures and
m*2 ¼ 0:962 velocities of this initial spreading flow have
ZI, 2 ¼ 3:176 been initiated [46]. At a minimum, it is useful
to become aware of the many fluid mechanical
Q_ 2 ¼ 2308 kW
phenomena embodied in a description of the
H 2 ¼ 5:176 m ceiling jet flow in a corridor up to the time
These two results are essentially identical for this when the ceiling jet is totally submerged in a
quiescent, warm upper layer. Borrowing heavily
type of analysis.
from a description of this flow provided by
Since it has been shown that the quasi-steady
analysis is appropriate for this example, the dimen- Zukoski et al. [46], the process is as follows.
A fire starts in a small room with an open door
sionless maximum temperature in the ceiling jet
to a long corridor having a small vent near the
flow, 4 m from the impingement point, can now
be calculated from (ΔT2*)qs in Equation 14.48. floor at the end opposite the door. As the fire
450 R.L. Alpert
Fig. 14.6 Transient a b

ceiling jet flow in a room
and corridor [45]
c Vf
Hydraulic jump
d Vf
e Vw
Vw
starts, smoke and hot gases rise to form a layer the fire room in the same manner that the plume
near the fire room ceiling. The layer is contained within the room flows over the interior ceiling.
in the small room by the door soffit (Fig. 14.6a). The gas flow in this jet is supercritical, analogous
As the fire continues, hot gas from the room to the shooting flow of liquids over a weir. The
begins to spill out under the soffit into the hall- velocity of the gas in this flow is greater than the
way. The fire grows to a relatively constant heat speed of gravity waves on the interface between
release rate. the hot gas and the cooler ambient air. The inter-
The outflowing gas forms a short, buoyant action of the leading edge of this flow with the
plume (Fig. 14.6b) that impinges on the hallway ambient air ahead of it produces a hydraulic,
ceiling, producing a thin jet that flows away from jumplike condition, as shown in Fig. 14.6c.
www.ebook777.com
A substantial amount of ambient air is entrained beams or corridor walls, are very useful for
at this jump. Downstream of the jump, the veloc- verifying that detailed numerical models of
ity of the gas flow is reduced and mass flow is fire phenomena (e.g., Hara and Shinsuke [50])
increased due to the entrainment at the jump. have been implemented properly. The predictive
A head is formed at the leading edge of the techniques are the basis for acceptable design
flow. Mixing between this ceiling-layer flow of fire detection systems, as exemplified by
and the ambient cooler air occurs behind Appendix B of NFPA 72®, National Fire Alarm
this head. Code [33].
The flow that is formed travels along the hall-
way ceiling (Fig. 14.6c, d) with constant velocity
and depth until it impinges on the end wall
(Fig. 14.6e). A group of waves are reflected Nomenclature
back toward the jump near the fire room, traveling
on the interface. Mixing occurs during the wall A g/(ρ1cpT1)(m2/kg)
impingement process (Fig. 14.6f), but no signifi- a Constant in Equation 14.59, equal to
cant entrainment occurs during the travel of the 0.24–0.29
nonbreaking reflected wave. When these waves b Effective plume radius at the intersec-
reach the jump near the fire room door, the jump tion with the ceiling elevation (m)
is submerged in the warm gas layer, eliminating CT Constant [17], related to plume flow,
the entrainment of ambient lower-layer air at this equal to 9.115
position (Fig. 14.6g). cp heat capacity at constant pressure
After several wave reflections up and down (J/kg K)
the corridor along the interface, the wave motion D Burner dimension (m)
dies out, and a ceiling layer uniform in depth is Deff Effective diameter of the base of the
produced. This layer slowly grows deeper as the flame zone or the burning fuel
hot gas continues to flow into the hallway from f Ceiling friction factor
the fire room. g Gravitational acceleration (m/s2)
It is clear from the preceding description that H Ceiling height above fire source; for
quantification of effects during development of a sloped ceiling, on the fire axis (m)
submerged ceiling jet flow is quite complex. h Heat transfer coefficient (kW/m2 K)
Analyses and experiments have been hb Depth of beams in a primary beam
performed to better understand the major features channel (m)
of a developing ceiling jet flow in a corridor Lf,tip Visible flame length from burner to
[47, 48]. One such study [49] contains a descrip- furthest flame tip (m)
tion somewhat different from that already given. ‘b Half-width for corridor or primary
beam channel (m)
‘T Ceiling jet thickness based on 1/e
depth of excess temperature profile
Summary (m)
m 2* Fraction of fire-plume mass flux
Reliable formulas are available to predict maxi- penetrating upper layer
mum gas temperatures and velocities and approx- p Ambient air pressure (Pa); also, as
imate temperature/velocity profiles in fire-driven exponent of time for general power-
ceiling jet flows beneath unobstructed ceilings for law fire growth
both steady and power-law fire growth. These Pr Prandtl number
predictive formulas, which also apply to certain Q_ Total heat release rate (kW)
situations where the ceiling jet flow is confined by Q_ c Convective heat release rate (kW)
452 R.L. Alpert

Q*, Q_ 0 _ ρ1 c p T 1 pffiffigffiH 5=2
*
Q= ν Kinematic viscosity (m2/s)
p ffiffi
ffi θ
Q00c* Q_ c = ρ1 c p T 1 gH 5=2 Angle of inclination of the ceiling with
q_ respect to the horizontal (degrees)
Rate of heat transfer per unit area
ρ Gas density (kg/m3)
(heat flux) to the ceiling surface
σ Parameter defined in Equation 14.73
(kW/m2)
ξ Ratio of temperatures, T1,2/T1,1
R Radial distance to detector (m)
R^ r/(H zo)
Ra Rayleigh number
Subscripts
Re Reynolds number
0 Based on steady-state fire source
r Radial distance from axis of fire
1 Associated with lower layer
plume (m)
2 Associated with upper layer; or parameter
rup Radial distance in steepest upward
associated with t2 fire growth
direction from axis of fire plume (m)
1 Ambient, outside ceiling jet or plume
St Stanton number, h/(ρUcp)
flows
T Ceiling jet gas temperature (K)
c Convective fraction
T1 Ambient air temperature (K)
f Associated with gas travel time delay
Tp Peak gas temperature in plume at the
I Value at the interface position between the
intersection with ceiling elevation
heated upper layer and cool lower layer
(K)
i Reference value at ignition
ΔT Excess gas temperature, T T1
p Associated with plume flow
(K) or ( C)
qs Quasi-steady flow condition
t Time (s)
U Ceiling jet gas velocity (m/s)
Uup Maximum ceiling jet gas velocity in Superscripts
the steepest upward direction (m/s)
Vp Maximum plume velocity at the inter- * Dimensionless quantity
section with ceiling elevation (m/s) ^ Quantity related to transient rack-storage
Y Distance along channel or corridor, fire
measured from plume axis (m)
ZI Distance of layer interface above the References
real or substitute fire source (m)
z Vertical distance above the base of 1. R.W. Pickard, D. Hird, and P. Nash, “The Thermal
Testing of Heat-Sensitive Fire Detectors,” F.R. Note
the flame zone 247, Building Research Establishment,
zH Distance of ceiling above the base of Borehamwood, UK (1957).
the flame zone 2. P.H. Thomas, “The Distribution of Temperature and
zo Virtual origin elevation in a transient Velocity Due to Fires Beneath Ceilings,” F.R. Note
141, Building Research Establishment,
rack storage fire Borehamwood, UK (1955).
zv Distance of virtual plume origin 3. R.L. Alpert, “Calculation of Response Time of
above the base of the flame zone Ceiling-Mounted Fire Detectors,” Fire Technology,
Rate of change of heat release rate 8, p. 181 (1972)
_
dQ=dt 4. R.L. Alpert, “Turbulent Ceiling Jet Induced by Large-
with time (kW/s) Scale Fires,” Combustion Science and Technology,
11, 197 (1975)
Greek Letters 5. H.Z. Yu (You), “An Investigation of Fire-Plume
Impingement on a Horizontal Ceiling:
2-Impingement and Ceiling-Jet Regions,” Fire and
α Growth parameter for t2 fires (kW/s2) Materials, 9, 46 (1985)
β2 Constant [17] related to plume flow, equal 6. G. Heskestad and T. Hamada, “Ceiling Jets of Strong
to 0.913 Fire Plumes,” Fire Safety Journal, 21, 69, (1993)
www.ebook777.com
7. G. Heskestad, “Physical Modeling of Fire,” Journal of Safety Science, Proceedings of the Third International
Fire & Flammability, 6, p. 253 (1975). Symposium (G. Cox and B. Langford, eds.), Elsevier
8. L.Y. Cooper, “Heat Transfer from a Buoyant Plume to Applied Science, New York, p. 261 (1991).
an Unconfined Ceiling,” Journal of Heat Transfer, 23. R.L. Alpert, “Convective Heat Transfer in the
104, p. 446 (1982). Impingement Region of a Buoyant Plume,” ASME
9. L.Y. Cooper and A. Woodhouse, “The Buoyant Plume- Journal of Heat Transfer, 109, p. 120 (1987).
Driven Adiabatic Ceiling Temperature Revisited,” 24. H.C. Kung, R.D. Spaulding, and P. Stavrianidis, “Fire
Journal of Heat Transfer, 108, p. 822 (1986). Induced Flow Under a Sloped Ceiling,” Fire Safety
10. H.Z. Yu (You) and G.M. Faeth, “Ceiling Heat Trans- Science, Proceedings of the Third International Sym-
fer during Fire Plume and Fire Impingement,” Fire posium (G. Cox and B. Langford, eds.), Elsevier
and Materials, 3, 140 (1979) Applied Science, New York, p. 271 (1991).
11. C.C. Veldman, T. Kubota, and E.E. Zukoski, “An 25. O. Sugawa, T. Hosozawa, N. Nakamura, A. Itoh and
Experimental Investigation of the Heat Transfer Y. Matsubara, “Flow Behavior under Sloped Ceiling,”
from a Buoyant Gas Plume to a Horizontal Ceil- in Fifteenth Meeting of UJNR Panel on Fire Research
ing—Part 1: Unobstructed Ceiling,” NBS-GCR- and Safety, NISTIR 6588, National Institute of
77–97, National Bureau of Standards, Washington, Standards and Technology, Gaithersburg, MD USA,
DC (1977). March 2000.
12. V. Motevalli and C.H. Marks, “Characterizing the 26. Y. Oka, M. Ando and K. Kamiya, “Ceiling Jet Flow
Unconfined Ceiling Jet Under Steady-State Properties for Flames Impinging on an Inclined Ceil-
Conditions: A Reassessment,” Fire Safety Science, ing,” Fire Safety Science, Proceedings of the Tenth
Proceedings of the Third International Symposium International Symposium, International Association
(G. Cox and B. Langford, eds.), Elsevier Applied for Fire Safety Science, London, 2012.
Science, New York, p. 301 (1991). 27. Y. Oka and M. Ando, “Temperature and Velocity
13. R.L. Alpert, “Fire Induced Turbulent Ceiling-Jet,” Properties of a Ceiling Jet Impinging on an Uncon-
Technical Report Serial No. 19722–2, Factory Mutual fined Inclined Ceiling,” accepted for publication in
Research Corporation, Norwood, MA, p. 35 (1971). Fire Safety Journal, 2012.
14. D.D. Evans and D.W. Stroup, “Methods to Calculate 28. H.C. Kung, H.Z. Yu (You), and R.D. Spaulding,
the Response Time of Heat and Smoke Detectors “Ceiling Flows of Growing Rack Storage Fires,”
Installed Below Large Unobstructed Ceilings,” Fire 21st Symposium (International) on Combustion, Com-
Technology, 22, 54 (1986). bustion Institute, Pittsburgh, PA, p. 121 (1986).
15. R.L. Alpert, “The Fire Induced Ceiling-Jet Revisited,” 29. R.P. Schifilliti, Use of Fire Plume Theory in the
in The Science of Suppression, Proceedings of Design and Analysis of Fire Detector and Sprinkler
Fireseat 2011 at the National Museum of Scotland, Response, Thesis, Worcester Polytechnic Institute,
9 November 2011, The University of Edinburgh, Worcester, MA (1986).
Edinburgh, Scotland, pp. 1–21. 30. G. Heskestad, “Similarity Relations for the Initial
16. A. Tewarson, “Generation of Heat and Gaseous, Liq- Convective Flow Generated by Fire,” ASME Paper
uid and Solid Products in Fires,” SFPE Handbook of No. 72-WA/HT-17, American Society of Mechanical
Fire Protection Engineering, this volume, (p. 3–142 in Engineers, New York (1972).
4th Edition). 31. G. Heskestad and M.A. Delichatsios, “Environments
17. E.E. Zukoski, T. Kubota, and B. Cetegen, “Entrainment of Fire Detectors,” NBS-GCR-77-86 and NBSGCR-
in Fire Plumes,” Fire Safety Journal, 3, 107 (1981) 77-95, National Bureau of Standards, Washington,
18. G. Heskestad and M.A. Delichatsios, “The Initial DC (1977).
Convective Flow in Fire,” 17th International Sympo- 32. G. Heskestad and M.A. Delichatsios, “Update: The
sium on Combustion, Combustion Institute, Initial Convective Flow in Fire,” Short Communica-
Pittsburgh, PA (1978). tion, Fire Safety Journal, 15, p. 471 (1989).
19. C.L. Beyler, “Fire Plumes and Ceiling Jets,” Fire 33. NFPA 72®, National Fire Alarm Code®, National
Safety Journal, 11, p. 53 (1986). Fire Protection Association, Quincy, MA (1999).
20. G.T. Atkinson and D.D. Drysdale, “Convective Heat 34. H.Z. Yu and P. Stavrianidis, “The Transient Ceiling
Transfer from Fire Gases,” Fire Safety Journal, Flows of Growing Rack Storage Fires,” Fire Safety
19, p. 217 (1992). Science, Proceedings of the Third International Sym-
21. Y. Hasemi, S. Yokobayashi, T. Wakamatsu, and posium (G. Cox and B. Langford, eds.), Elsevier
A. Ptchelintsev, “Fire Safety of Building Components Applied Science, New York, p. 281 (1991).
Exposed to a Localized Fire: Scope and Experiments 35. M.A. Delichatsios, “The Flow of Fire Gases Under a
on Ceiling/Beam System Exposed to a Localized Beamed Ceiling,” Combustion and Flame,
Fire,” AsiaFlam 95—1st International Conference, 43, 1 (1981).
Interscience Communications, Ltd., London, p. 351 36. C. Koslowski and V. Motevalli, “Behavior of a
(1995). 2-Dimensional Ceiling Jet Flow: A Beamed Ceiling
22. M.A. Kokkala, “Experimental Study of Heat Transfer Configuration,” Fire Safety Science, Proceedings of
to Ceiling from an Impinging Diffusion Flame,” Fire the Fourth International Symposium (T. Kashiwagi,
454 R.L. Alpert
ed.), International Association of Fire Safety Science, Ceiling Vents—Theory,” Fire Safety Journal,
Bethesda, MD, p. 469 (1994). 16, pp. 137–163 (1990).
37. C.C. Koslowski and V. Motevalli, “Effects of Beams 46. E.E. Zukoski, T. Kubota, and C.S. Lim, “Experimental
on Ceiling Jet Behavior and Heat Detector Opera- Study of Environment and Heat Transfer in a Room
tion,” Journal of Fire Protection Engineering, 5, 3, Fire,” NBS-GCR-85-493, National Bureau of
p. 97 (1993). Standards, Washington, DC (1985).
38. B. Lattimer and U. Sorathia, “Thermal Characteristics 47. H.W. Emmons, “The Ceiling Jet in Fires,” Fire Safety
of Fires in a Noncombustible Corner,” Fire Safety Science, Proceedings of the Third International Sym-
Journal, 38, p. 709 (2003). posium (G. Cox and B. Langford, eds.), Elsevier
39. D. Coutts, “An Emissive Power Correlation for Solid Applied Science, New York, p. 249 (1991).
Fuel Packages,” Journal of Fire Protection Engineer- 48. W.R. Chan, E.E. Zukowski, and T. Kubota, “Experi-
ing, 21, p. 133, 2011. mental and Numerical Studies on Two-Dimensional
40. D.D. Evans, “Thermal Actuation of Extinguishing Gravity Currents in a Horizontal Channel,” NIST-
Systems,” Combustion Science and Technology, GCR-93-630, National Institute of Standards and
40, p. 79 (1984). Technology, Gaithersburg, MD (1993).
41. D.D. Evans, “Calculating Sprinkler Actuation 49. G. Heskestad, “Propagation of Fire Smoke in a Corri-
Time in Compartments,” Fire Safety Journal, 9, 147 dor,” Proceedings of the 1987 ASME/JSME Thermal
(1985). Engineering Conference, Vol. 1, American Society of
42. L.Y. Cooper, “A Buoyant Source in the Lower of Two Mechanical Engineers, New York (1987).
Homogeneous, Stably Stratified Layers,” 20th Inter- 50. T. Hara and K. Shinsuke, “Numerical Simulation of
national Symposium on Combustion, Combustion Fire Plume-Induced Ceiling Jets Using the Standard
Institute, Pittsburgh, PA (1984). κ – ε Model,” Fire Technology, 42, p. 131 (2006).
43. E.E. Zukoski and T. Kubota, “An Experimental Inves-
tigation of the Heat Transfer from a Buoyant Gas
Plume to a Horizontal Ceiling—Part 2: Effects of Dr. Ronald L. Alpert received his undergraduate and
Ceiling Layer,” NBS-GCR-77-98, National Bureau graduate education at the Massachusetts Institute of Tech-
of Standards, Washington, DC (1977). nology, where he majored in mechanical engineering. For
44. W.D. Davis, “The Zone Fire Model Jet: A Model for nearly 35 years, he was with FM Global in various tech-
the Prediction of Detector Activation and Gas Tem- nical and managerial positions, ending his career there as
perature in the Presence of a Smoke Layer,” NISTIR an assistant vice president and manager of the Flamma-
6324, National Institute of Standards and Technology, bility Technology Research Program. Dr. Alpert was edi-
Gaithersburg, MD (1999). tor in chief of the Journal of Fire Protection Engineering
45. L.Y. Cooper, “Estimating the Environment and the for 10 years and a section editor of the NFPA Fire
Response of Sprinkler Links in Compartment Fires Protection Handbook, 20th edition. He has published
with Draft Curtains and Fusible Link-Actuated numerous papers in refereed journals and technical
reports.
www.ebook777.com
Vent Flows
15
Takeyoshi Tanaka
Introduction When a pressure difference exists across a

vent, fluid (liquid or gas) will be pushed through.
Fire releases a great amount of heat that causes Precise calculation of such flows from the basic
the heated gas to expand. The expansion pro- laws of nature can only be performed today by
duced by a fire in a room drives some of the gas the largest computers. For fire purposes, and all
out of the room. Any opening through which gas engineering purposes, calculations are carried
can flow out of the fire room is called a vent. out with sufficient precision using the methods
The most obvious vents in a fire room are of hydraulics. Since these formulas are only
open doors and open or broken windows. Venti- approximate, they are made sufficiently accurate
lation ducts also provide important routes for gas (often to within a few percent) by a flow coeffi-
release. A room in an average building may have cient. These coefficients are determined by
all of its doors and windows closed and, if venti- experimental measurements.
lation ducts are also closed, the gas will leak
around normally closed doors and windows and
through any holes made for pipes or wires. These Calculation Methods for Nonbuoyant
holes will act as vents. (If a room were hermeti- Flows
cally sealed, a relatively small fire would raise
the pressure in the room and burst the window, If a pressure drop, Δp ¼ p1 p2, exists across a
door, or walls.) vent of area, A, with a fluid density, ρ, the flow
Gas will move only if it is pushed. The only through the vent has (Fig. 15.1) [1]
forces acting on the gas are the gas pressure and rffiffiffiffiffiffiffiffiffi
gravity. Since gravity acts vertically, it might 2Δ p
Velocity u ¼ ð15:1Þ
seem that gas could only flow through a hole in ρ
the floor or ceiling. Gravity, however, can pro-
duce horizontal pressure changes, which will be rffiffiffiffiffiffiffiffiffi
2Δ p
explained in detail below. A gas flow that is Volume flow V_ ¼ CA ð15:2Þ
ρ
caused directly or indirectly by gravity is called
a buoyant flow. and
pffiffiffiffiffiffiffiffiffiffiffiffi
Mass flow m_ ¼ CA 2ρΔ p ð15:3Þ
In these formulas C is flow coefficient and the SI

units are Δ p ¼ ðPaÞ ¼ ðN=m2 Þ, A ¼ ðm2 Þ, ρ ¼
T. Tanaka (*)
Fire Science and Technology, Kyoto University ðkg=m3 Þ, u ¼ ðm=sÞ, V_ ¼ ðm3 =sÞ, m_ ¼ ðkg=sÞ.

456 T. Tanaka
Fig. 15.1 Most fire vents a b

are orifices
Area
A
Area
A
Orifice Nozzle
pg
where A is area of vent and A1 is area of supply
pipe. The factor 6895 converts pressure in lb/in.2
Area A h i
to Pascals while the factor 1 ðA=A1 Þ2
Area A1 corrects Δp for the dynamic effect of the inlet
velocity in the supply hose or pipe.
In the atmosphere, the pressure at the ground,
pa, is just sufficient to support the weight of the
air above. If the air density is ρa, the pressure, p,
pg Area A1 at height, h, is less than pa by the weight of the air
below height, h. Thus the pressure difference is
Δ p ¼ pa p ¼ ρa gh ð15:5Þ
Area A
It is sometimes convenient when considering fire
gases to use h ¼ Δ p=ρa g, the pressure head, in
meters of ambient air, in the velocity and flow
Fig. 15.2 A hose nozzle and a sprinkler nozzle
rate formulas given above.
The previous discussion supposes that the
flowing fluid is of constant density. For liquids
If the flow of water from a fire hose or sprin-
this is true for all practical situations. The density
kler (Fig. 15.2) is to be calculated and the pres-
of air or fire gases will not change significantly
sure, pg, is read on a gauge (in lb/in.2) at the
during the flow through the vent so long as the
entrance to the nozzle where the area is A1, the
pressure change is small, so they can also be
previous formulas provide the velocity, volume
treated as constant density fluids.
flow, and mass flow by using
If the pressure drop is large, the equations
6895 pg become more complicated [2]. If the pressure
Δp ¼ ð15:4Þ and density upstream of the vent are p1 and ρ1
1 ðA=A1 Þ2 while the pressure after the vent is p2, the
equations for velocity and mass flow become
sffiffiffiffiffiffiffiffi( 2=γ " ðγ1Þ=γ #)1=2

2 p1 γ p2 p2
u¼ 1 ð15:6Þ
ρ1 γ 1 p 1 p1
www.ebook777.com
15 Vent Flows 457
( 2=γ " ðγ1Þ=γ #)1=2

pffiffiffiffiffiffiffiffiffiffiffiffi γ p2 p2
m_ ¼ CA 2ρ1 p1 1 ð15:7Þ
γ 1 p1 p1
where γ ¼ c p =cv .
The value of γ depends on the complexity of ρ ¼ Density of the fluid approaching the vent
the molecules of the flowing gas. For fire gases μ ¼ Viscosity of the fluid approaching the vent
(which always contain a large amount of air) the A door or window vent is almost always rect-
value of γ will fall between 1.33 and 1.40. For angular, not circular. The D to be used in the
most fire purposes the diatomic gas value (air) of Reynolds number should be the hydraulic
1.40 is sufficiently accurate. diameter
The mass flow given by the previous equation
4A
has a maximum at D¼ ð15:11Þ
P
γ=ðγ1Þ
p2 2 where
¼ ð15:8Þ
p1 γþ1 A ¼ Area of the vent
P ¼ Perimeter of vent
For γ ¼ 1.40, the maximum flow is reached for a For a rectangular vent, a wide and b high, A
downstream pressure p2 ¼ 0:528 p1 . For all ¼ ab, P ¼ 2ða þ bÞ.
lower back pressures the flow remains constant
at its maximum 2ab
D¼ ð15:12Þ
" ð a þ bÞ
ðγþ1Þ ðγ1Þ #1=2
pffiffiffiffiffiffiffiffiffiffi 2
m_ ¼ CA ρ1 p1 γ The experimental values of the flow
γþ1
coefficients for nozzles and orifices, C, are
ð15:9Þ given in Fig. 15.3 [2]. Flow coefficients for
nozzles are near unity while for orifices are
With these equations, the mathematical descrip- approximately 0.6, as can be seen in Fig. 15.1,
tion of the rate of flow of liquids and gases through where the flow from an orifice separates from the
holes is complete as soon as the appropriate flow edge of the orifice and decreases to a much
coefficients are known. The coefficients, found by smaller area, in fact about 0.6 of the orifice area.
experiment, correct the formulas for the effect of For most fire applications the Reynolds num-
the fluid viscosity, the nonuniformity of the veloc- ber will be about 106. Sprinklers and fire nozzles
ity over the vent, turbulence and heat transfer are small but the velocity is quite high. Con-
effects, the details of nozzle shape, the location of versely, ventilation systems of buildings are
the pressure measurement points, and so forth. The larger but have a lower velocity. Finally, doors
corrections also depend on the properties and and windows in the areas of a building not too
velocity of the fluid. The most important coeffi- near the fire are still larger but the velocity is still
cient correction for any given vent geometry is the smaller. For most purposes the flow coefficient
dimensionless combination of variables called the can be set as C ¼ 0:98 for a nozzle and C ¼ 0:60
Reynolds number, Re, and for an orifice.
uDρ
Re ¼ ð15:10Þ
μ
Buoyant Flows Through Vertical Vents
where
u ¼ Velocity of the fluid given by the previous A fire in a room causes gases to flow out through
equations a vent by two processes. The heating of the air in
D ¼ Diameter of the nozzle or orifice a room causes the air to expand, pushing other air
458 T. Tanaka
Fig. 15.3 Orifice and 1.0

nozzle flow coefficients
Nozzle
0.9
Flow coefficient
P1 P2
0.8
P1 P 2
0.7
Orifice
0.6
0.5
104 105 106 107
Reynolds number (Re)
Fig. 15.4 Pressure a b

gradients: (a) each side of a h
door; (b) superimposed on
a pressure versus height
graph
hv
hn
p
pf pa pf pa
out through all available vents and hence Δ p f ¼ p f pa ð15:13Þ

throughout the entire building. At the same
time, the heated air, with products of combustion where
and smoke, rises in a plume to the ceiling. When pf ¼ Pressure at the floor inside the room in front
the hot layer of gas at the ceiling becomes deep of the vent
enough to fall below the top of a vent, some hot pa ¼ Pressure at the floor level outside of the
gas will flow out through the vent. As the fire room just beyond the vent
grows, the buoyant flow out will exceed the gas The pressure at height y is less than the pres-
expansion by the fire. Thus the pressure in the fire sure at the floor and can be found by the follow-
room at the floor will fall below atmospheric, and ing hydrostatic equations:
outside air will flow in at the bottom. A familiar ðy
sight develops, where smoke and perhaps flames Inside p1 ¼ p f ρ1 gdy ð15:14Þ
0
issue out the top of a window while fresh air
ðy
flows in near the bottom. This buoyant flow
Outside p2 ¼ pa ρ2 gdy ð15:15Þ
mechanism allows a fire to draw in new oxygen 0
so essential for its continuation.
For these buoyantly driven flows to occur, The pressure difference at height, h, is
there must be a pressure difference across the Δ p ¼ p1 p2
vent. Figure 15.4 illustrates how these pressure ðh
differences are produced. The pressure differ- ¼ Δ p f þ ðρ2 ρ1 Þg dy ð15:16Þ
ence at the floor is 0
www.ebook777.com
15 Vent Flows 459
Since the outside density, ρ2, is greater than the Thus the buoyantly driven flow velocities
inside density, ρ1, the integral is positive so that induced by a room fire could be as high as
Δp is often positive (outflow) at the top of the rffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
vent and negative (inflow) at the bottom. The 2Δ p 2 28:84
u¼ ¼ ¼ 7:00 m=s ð23 ft=sÞ
flow properties at the elevation, h, are the same ρ 1:176
as previously given. Since the pressure varies with height and time,
sffiffiffiffiffiffiffiffiffi a series of pressure probes is required and each
2Δ p
u¼ ð15:17Þ should have its own meter or a rapid activation
ρ switch. Although standard pitot tubes are the
most accurate dynamic pressure probes, they
sffiffiffiffiffiffiffiffiffi
are sensitive to flow direction and would have
V_ 2Δ p
¼C ð15:18Þ to be adjusted at each location for the direction of
A ρ
the local flow, especially for outflow and inflow.
The probe orientation would need to be continu-
m_ pffiffiffiffiffiffiffiffiffiffiffiffi ally changed as the fire progressed.
¼ C 2ρΔ p ð15:19Þ
A A single string of fixed-orientation pressure
probes arranged vertically down the center of
Since they are not the same at different heights in
the door increases convenience of the measure-
the vent, the volume and mass flow are given as
ment but forces a decrease in accuracy. The
flow per unit area.
out-in flow problem is avoided by use of bidirec-
tional probes in place of pitot tubes [3]
(Fig. 15.5). These probes give velocities within
Measuring Vent Flows in a Fire
10 % over an angular range of 50 degrees of
Experiment
the probe axis in any direction.
Determination of the local velocity also
Sufficient measurements must be made to evalu-
requires the measurement of the local gas den-
ate ρ and Δp to allow use of Equation 15.19.
sity. The density of fire gases can be determined
There are four different available methods that
from measured gas temperatures with sufficient
differ in simplicity, accuracy, and cost.
accuracy by the ideal gas law
Method 1 The dynamic pressure distribution Mp
ρ¼ ð15:20Þ
can be measured in the plane of the vent. This RT
measurement requires a sensitive pressure meter.
where
The pressure difference is almost always less
than the atmospheric pressure difference M ¼ Average molecular weight of flowing gas
between the floor, pf, and the ceiling, pc. For a J
room 2.5 m in height the atmospheric pressure R ¼ 8314 ¼ Universal gas constant
kg mol K
difference is
As noted previously, the pressure changes only
p f pc ¼ ρa gH ¼ 1:176 9:81 2:5 by a very small percentage throughout a building
so its effect on gas density is negligible.
¼ 28:84 Pascals ð3:0 mm H2 OÞ
Fire gases contain large quantities of nitrogen
from the air and a variety of other compounds.
This is only The average molecular weight of the mixture will
be close to but somewhat larger than that of air.
p f pc 28:84
¼ Incomplete knowledge of the actual composition
pa 101, 325
of fire gas prevents high-accuracy calculations.
¼ 0:00028 fraction of atmospheric pressure For most fire calculations, it is accurate enough
460 T. Tanaka
Fig. 15.5 A bidirectional Support tubes

flow probe (Connect press taps to indicating instrument)
0.572D
0.286D
Press.
taps Barrier
0.857D ( )
D 7"
8
0.143D
2D
to neglect the effect of the change of molecular strongly with the distance from the ceiling.
weight from that of air (Ma ¼ 28.95). Density of Thus a string of thermocouples hanging verti-
gas is determined primarily by its temperature cally on the centerline of the vent is usually
(which may vary by a factor of 4 in a fire). Thus considered to be the best that can be done in a
practical fire test. Special care must be exercised
352:8 kg
ρ¼ ð15:21Þ to keep the test fire some distance away from the
T m3 entrance to the vent. Since a fire near a vent has
where T is temperature in Kelvin (¼ C + 273). effects at present unknown, fire model
A string of thermocouples must be included calculations of real fire vent flows under such
along with the bidirectional probes to measure conditions will be of unknown accuracy. The
vent flows. For higher accuracy, aspirated velocity distribution vertically in the vent is
thermocouples must be used or a correction given by
made for the effect of fire radiation [3]. The sffiffiffiffiffiffiffiffiffi
temperature and, hence, the gas density will 2Δ p
u ¼ 0:93 ð15:22Þ
vary over the entire hot vent outflow. To deter- ρ
mine the temperature distribution so completely
would require an impracticably large number of where ρ follows from Equation 15.21 using the
thermocouples. Fortunately, the temperature in temperature distribution in the vent with a cali-
the vent is a reflection of the temperature distri- bration factor of 0.93 for the bidirectional probes
bution in the hot layer inside the room, which [4]. Using ρ from Equation 15.21 gives the
normally is stratified, and hence varies most directly useful forms

pffiffiffiffiffiffiffiffiffiffi N
u ¼ 0:070 TΔ p Δp 2
m
pffiffiffiffiffiffiffiffiffiffi lb
u ¼ 5:81 TΔ p Δp pressure measured with bidirectional probe ð15:23Þ
in:2
www.ebook777.com
15 Vent Flows 461
where If the bidirectional probe pressures are

measured in psi, the coefficient 16.79 must be
u is in ðm=sÞ
replaced by 1394.
T is in ðKÞ
Method 2 A somewhat simpler but less accurate
Except for very early stages of a room fire, procedure to measure vent flows requires the
there will be flow out at the top (u, Δp > 0) and measurement of the pressure difference at the
flow in at the bottom (u, Δp > 0).1 Thus there is floor (or some other height). One pressure differ-
a position in the vent at which u ¼ 0, which ence measurement together with the vertical tem-
is the vertical location where the pressure inside perature distribution measurement, T1, inside the
is equal to that outside. This elevation, hn, is room (about one vent width in from the vent) and
called the neutral axis. Defining the elevation of T2, outside the vent (well away from the vent
the vent sill as hb (hb ¼ 0 for a door) and the flow), provides the density information required
elevation of the soffit as ht, the flows are given by to find the pressure drop at all elevations
ð ht (Equation 15.16).
Flow out m_ u ¼ C ρub dy ð15:24Þ ðy
hn 1 1
Δ p ¼ Δ p f þ 3461 dy ð15:28Þ
ð hn 0 T2 T1
Flow in _
m d ¼ C ρub dy ð15:25Þ
hb
For most fires, Δpf will be negative; that is,
where the pressure at the floor inside the fire room will
b ¼ Width of the vent be less than the pressure outside. This is only
C ¼ Experimentally determined flow coefficient true for a fire room with a normal size vent
(¼ 0.68) [5] (door, window). For a completely closed room
Using Equations 15.21 and 15.22 and the inside pressure is well above the outside
C ¼ 0.68 into these equations, the most conve- pressure. Since the temperature inside the
nient forms are fire room is higher than that outside, Equa-
tion 15.28 gives a Δp that becomes less nega-
ð ht rffiffiffiffiffiffiffi
Δp tive, passes through zero at the neutral axis, hn,
Flow out m_ u ¼ 16:79 b dy ðkg=sÞ
hn Tv and becomes positive at higher levels in the
ð15:26Þ fire room. The vertical location of the neutral
axis is, therefore, readily found from
ð hn rffiffiffiffiffiffiffi
Δp Equation 15.28.
Flow in m_ d ¼ 16:79 b dy ðkg=sÞ The calculation of the pressure distribution
hb Tv
requires measurement of the temperature
ð15:27Þ
distribution both inside, T1, and outside, T2, of
where the vent. However, calculation of the flow
Δp ¼ Pressure drop in Pascals measured with requires a knowledge of the density distribution
bidirectional probe as a function of y in the vent itself. Thus a third thermocouple
b ¼ Width of the vent in m string is required to measure the temperature
TV ¼ Temperature (K) in the vent measured as a distribution, Tu, in the vent. The desired flow
function of y properties [6] are
Velocity
1
Equation 15.23 should be written u ¼ (sign Δp)K rffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðy
pffiffiffiffiffiffiffiffiffiffiffiffiffi 2Δ p 1 1
TjΔ pj since when Δp < 0 the absolute value must be u¼ ¼ 4:33 T v dy ðm=sÞ
used to avoid the square root of a negative number and the ρ hn T 2 T1
sign of the velocity changes since the flow is in and
ð15:29Þ
not out.
462 T. Tanaka
Fig. 15.6 Buoyant flow

out of the window of a
fire room ρ
y
hV
δ
hn ρa ht
hi d
hb
Room 1 Room 2
Flow out photographically) locating the bottom of the out-

ð ht flow in the vent during the test at the position of
m_ u ¼ C ρbu dy the neutral axis, hn, where Δp ¼ 0. Method 3 is
hn the same as Method 2 except that the neutral axis
ð ht ð y 1=2 location is found directly by experiment rather
1 1 1
¼ 1063 b dy dy than being deduced from the pressures. The dis-
hn T v hn T 2 T 1
tribution of pressure drop across the vent is found
ð15:30Þ by integrating Equation 15.16 above (Δp > 0)
and below (Δp < 0) hn using the density distri-
Flow in
bution inside, ρ1, and outside, ρ2, the room. The
ð hn flow properties are computed as before from
m_ d ¼ C ρbu dy Equations 15.29, 15.30 and 15.31.
hb
ð ht ð y 1=2
1 1 1 Method 4 A simpler but less accurate method
¼ 1063 b dy dy
hn T v hb T 2 T 1 uses the fair assumption that the gas in the fire
ð15:31Þ room soon separates into a nearly uniform hot
layer of density, ρ, with a nearly uniform cold
where layer below of density, ρd. This separation
b ¼ Width of the vent at height y with appropriate notation is shown in Fig. 15.6.
Δp ¼ Calculated from Equation 15.16 using the In this approximation the appropriate flow
temperatures (and thus densities) inside and formulas [5] are
outside of the room

ρ ¼ Density computed from the temperature in ρa ρ 1=2
Outflow uu ¼ 2g y ð15:32Þ
the vent ρ
(Note that for inflow Δp is negative. There-
fore, the equation takes the square root of the where y is distance above the neutral plane
magnitude |Δp| while its sign gives the flow
2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
direction.) m_ u ¼ Cb 2gρðρa ρÞðhv hn Þ3=2 ð15:33Þ
3
Method 3 The use of a sensitive pressure meter The inflow by this two-layer method depends
can be avoided entirely by visually (or better, on δ, which is small and cannot be determined
www.ebook777.com
15 Vent Flows 463
with sufficient accuracy because of the effect of The flow coefficient to be used for buoyant
gas motions in the fire room. flows is 0.68 as determined by specific
The neutral axis may be found in several experiments designed for the purpose. For
ways: nonbuoyant flows (nozzles and orifices), the
1. It may be located visually or photographically flow coefficients are determined to better than
during the test. 1 % and presented as a function of the Reynolds
2. It may be found from the vent temperature number as in Fig. 15.3. This accuracy is possible
distribution by locating [visually on a plot of because the fluid can be collected and measured
TV( y)] the position just below the most rapid (by weight or volume).
temperature rise from bottom to top of For buoyant flows the experiments are much
the vent. more difficult because the hot outflow and cold
The lower-layer temperature, Td, of the inflow cannot be collected and weighed. The
two-layer model is taken as the gas tempera- best fire-gas vent flow coefficient measurements
ture just above the vent sill. The upper-layer to date [6, 7] have 10 % accuracy with occa-
temperature, Tu, is chosen so that the sional values as bad as 100 % (for inflow).
two-layer model has the same total mass (i.e., The most accurate buoyant flow coefficients
the same mean density) in the vent as the real were measured not for fire gases but for two
flow.2 nonmiscible liquids (kerosene and water)
[7]. In this case the two fluids could be separated
ð and measured, and the value 0.68 was found
1 1 hv dy hn hv hn
¼ ¼ þ ð15:34Þ except for the very low flow rates (near the
T hv 0 T hv T d hv T u
beginning of a fire). When buoyant flow
coefficients can be measured within a few per-
The densities ρa and ρ are found using Equa- cent accuracy, they will be a function of the
tion 15.21 from the temperatures Ta and Tu, Reynolds number, Re ¼ uhvρ/μ; the Froude
respectively. number, Fr ¼ u2ρa/ghv(ρ ρa); and the depth
The outflow velocity and mass flow are found parameter, hn/hv.
from Equations 15.32 and 15.33. The best option now available is to use
An estimate of the air inflow rate can be found C ¼ 0.68 and expect 10 % errors in flow
if the test has included the measurement of the calculations.
oxygen concentration in the gases leaving the fire Note that the above four methods require a
room. The gas outflow rate is equal to the inflow knowledge of hn, the dividing line between out-
rate plus the fuel vaporized, except for the effect flow above and inflow below. It would be useful
of transient variations in the hot layer depth. to have a simple formula by which hn could be
Thus calculated without any special measurements.
What determines hn?
1 þ yO2 λ
m_ d ¼ m_ u ð15:35Þ The fire at the start sends a plume of heated
1 þ 0:23λ
gas toward the ceiling and, by gas expansion,
where λ is the effective fuel-air ratio. pushes some gas out of the vent. The hot plume
gases accumulate at the ceiling with little, if
any, flowing out the vent. After a time,
depending on the size of the room, the hot
layer depth becomes so large that its lower sur-
face falls below the top of the vent. Hot gas
2
Sometimes the mean temperatures, T of the two-layer begins to flow out.
model and the real flow are also used and both hn and Tu
are determined (using Td as above). The requirement of When a fire has progressed to a second room,
identical T is arbitrary, sometimes leads to impractical there is a hot layer on each side of a connecting
results, and is not recommended. vent. Thus (with two layers on each side) there
464 T. Tanaka
Fig. 15.7 Some selected a b c d e

two-layer vent pressure
drop distributions. Dotted
line is pressure distribution n n
in room 1; solid line is
pressure distribution in n n
room 2 n n
are as many as four different gas densities: ρd1 is In Fig. 15.7d, the densities (slopes of pressure
greater than ρ1, densities below and above in distribution lines) are somewhat different than
room 1, and ρd2 is greater than ρ2, densities those in Fig. 15.7c (the hot layer in room 2 is
below and above in room 2. There are also four less deep but hotter than that in room 1).
pertinent levels: hb, sill height (0 if the vent is a Consequently, there are two neutral axes
door); ht, soffit height; hi1 , interface height in with a new small inflow layer at the top,
room 1; hi2 , interface height in room 2. There three flow layers in all—two in and one out.
are many different flow situations possible In Fig. 15.7e, the densities and floor level pres-
depending on these eight values. sure difference are such that there are four
The pressure variation from floor to ceiling flow layers, two out and two in, with three
in each room depends on the densities and neutral axes.
layer heights in the room. In addition, the pres- These five cases do not exhaust the possible
sure difference between the two rooms (at the vent flow situations.
floor, for example) may have any value Figure 15.7a, b accounts for all cases early in a
depending on the fire in each room, all the fire and all cases of vents from inside to outside of a
room vents, and especially the vent (or vents) building. They are also the only cases for which
connecting the two rooms. Figure 15.7 shows a experimental data are available. The case
few of the possible pressure distributions. The illustrated in Fig. 15.7c is common inside a build-
pressure distribution in room 1 is shown with a ing after a fire has progressed to the point that hot
dotted line while that in room 2 is shown as a layers exist in the two rooms on each side of a vent.
solid line. The cases illustrated in Fig. 15.7d, e have not been
In Fig. 15.7a, there are no hot layers, the pressure directly observed but probably account for an
in room 1 at every level is higher than that in occasional confused flow pattern. (In fact, the
room 2, and the flow is everywhere out (posi- above discussion assumes two distinct layers in
tive) (room 1 to room 2). each room.) The layers are seldom sharply defined
In Fig. 15.7b, a common situation exists. The and in this case there may be many neutral axes, or
density in room 2 is uniform (perhaps the regions, with a confusing array of in-out flow
outside atmosphere). Room 1 has a hot layer layers. These confused flow situations are probably
and a floor pressure difference such that there not of much importance in a fire since they seldom
is outflow at the top, inflow at the bottom, and occur and when they do they don’t last very long.
a single neutral axis somewhat above the The previous discussion of the possible
hot-cold interface in the room. two-layer flow situation is very important for
In Fig. 15.7c, the flow situation is similar to that the zone modeling of a fire. Fire models to date
in Fig. 15.7b, although there are hot layers in are all two-layer models (a three- or more layer
both rooms (but with a neutral axis above the model will present far more complex vent flows
interface in room 1 and below the interface in than those pictured in Fig. 15.7). In fire compu-
room 2). tation by a zone model, cases such as (d) and
www.ebook777.com
15 Vent Flows 465
Fig. 15.8 Sample fire

room and doorway 1.85
temperature distributions
In doorway
Height above floor (m)

1.5
In room
0.5
300 350 400 450

Temperature (K)
(e) in Fig. 15.7 will be unimportant to fire devel- Thus

opment. However, since these situations can (
arise, they should be handled via fire computa- Density in room 1 at height hþj if α > 0
ρ¼
tion, that is, by computing the flow layer by Density in room 2 at height hþ
i if α < 0
layer. Each layer has a linear pressure variation
from sill, interface, or neutral axis up to the next This flow calculation appears complex but can be
interface, neutral axis, or soffit. coded quite easily for computer use and then
By use of the pressure drop at the floor and the used to calculate all the possible cases.
room densities on each side of the vent in Equa- Although all vent flows can now be calcu-
tion 15.16, the position, hi, of all layers and the lated, the path of each layer of gas flow when it
sill, interfaces, neutral axes, and soffit will be enters a room is still needed for fire modeling. If
known. Thus, for each layer (defined as j) the the two-layer model is to be preserved, each
pressure drop at the bottom, Δpj, and at the top, inflow must mix with the hot layer or the cold
Δpj+1, will be known. Since the room densities layer, or be divided between them. No informa-
are constant in each room for each layer, the tion is yet available as to the best solution to this
vent pressure drop will vary linearly from Δpj problem.
to Δpj+1. The flow in each layer from room 1 to To illustrate these various methods of flow
room 2, found by integration [8], is given by calculation, some test data from a steady burner
fire in a room at the U.S. Bureau of Standards [6]
2 pffiffiffiffiffi are used. Some typical data are shown in
m_ i ¼ ðsign αÞC b h jþ1 hj 2ρ Fig. 15.8. Accurate results, even in a steady-
0 3
q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi 1
Δ pj þ Δ pj Δ pjþ1 þ Δ pjþ1 state fire, are difficult to obtain and questions
B C
@ q ffiffiffiffiffiffiffiffiffiffiffi q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi A about the data in this figure will be noted as
Δ pj þ Δ pjþ1 appropriate. The vent temperatures were
measured by small-diameter bare thermocouples
ð15:36Þ
for which there is some unknown radiation cor-
rection. This unknown correction may account
where
for the top vent temperature being higher than

Δ pj þ Δ pjþ1 that in the fire room.
α¼ whose sign determines The vent was 1.83 m high, 0.737 m wide, and
2
the in‐out direction of the flow the outflow measured with bidirectional probes
(not corrected for flow angle) was 0.588 kg/s
ρ ¼ Density of the gas flowing in the flow layer i
for a fire output of 0.63 kW. The ambient
466 T. Tanaka
temperature was 21.3 C (¼ 294.3 K). This flow In the figure the rapid temperature rise in the vent
was determined by using Method 1. begins at about 1 m. Hence this height is chosen
Method 2 uses the known location of the neu- as the neutral axis. The lowest inlet temperature
tral axis and requires the integration of is Td ¼ 308 K. By computing (1/Tv) the average
Equations 15.30 and 15.31. In this way the data value was found to be (1/Tv) ¼ 2.875 103.
of Fig. 15.8 gives outflow of 0.599 kg/s, 1.8 % Now by Equation 15.34
higher compared to Method 1 and inflow of
0.652 kg/s. A measured (by bidirectional probes) 1:00 0 1:83 1:00
2:875 103 ¼ þ
inflow is not given, but it seems odd that the 1:83 308 1:83 T u
inflow is greater than the outflow since inflow
must be smaller than the outflow by the mass rate Thus Tu ¼ 411.9 K. The corresponding density
of fuel burned at steady state. is ρ ¼ 352.8/411.9 ¼ 0.8565 kg/m3. From the
Data for use of Method 3 are not available. ambient temperature, Ta, we find ρa ¼ 352.8/
Method 4 requires the selection from Fig. 15.8 294.3 ¼ 1.199 kg/m3. Thus the outflow by Equa-
of a neutral axis location and inlet temperature. tion 15.33 is
2
m_ u ¼ 0:68 0:737½2 9:81 0:8565ð1:199 0:8565Þ1=2
3
ð1:83 1Þ3=2 ¼ 0:607 kg=s
This value is 3.2 % higher compared to sloped surfaces. Such buoyant flows through hor-
Method 1. izontal vents can become issue in some situations
associated with fire, particularly in such a config-
uration as exemplified in Fig. 15.9, which may
Buoyant Flows Through Horizontal arise in fires in basements of buildings, holds in
Vents ships and multi-floor building containing rooms
closed to the outdoors.
Unlike flows through vents in a vertical wall or The flow rate through a horizontal opening
non-buoyant flows through orifices, not ample can be treated, as done in vertical vents, by
quantitative works have been done on buoyant using Bernoulli’s equation if there is no temper-
flows through vents in horizontal or slightly ature difference between the connecting spaces
Fig. 15.9 Possible VENT

scenarios of horizontal vent
flow in fire where the
standard Bernoulli’s flow
equation becomes
inadequate
VENT
FIRE
FIRE
www.ebook777.com
15 Vent Flows 467
or if there is a large enough pressure difference pU and pL ¼ Pressures at the vent elevation in
across the vent. Let Δp be the pressure difference the upper and lower spaces, respectively.
across the vent defined by The standard vent flow model using
Bernoulli’s equation would predict the flow rate
Δ p ¼ pU pL ð15:37Þ
through a horizontal vent as follows according to
where the value of Δp, illustrated in Fig. 15.10:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðaÞ when Δ p < 0 : V_ U ¼ CAV 2Δ p=ρL ; V_ L ¼ 0 ð15:38aÞ
ð bÞ when Δ p ¼ 0 : V_ U ¼ 0; V_ L ¼ 0 ð15:38bÞ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðcÞ when Δ p > 0 : V_ U ¼ 0; V_ L ¼ CAV 2jΔ pj=ρU ð15:38cÞ
where room set in a full-scale thermostatic chamber,

V_ U and V_ L ¼ Volume rates of flows to upper and where the air in the test room was heated by an
lower spaces through the vent, respectively electric heating system while the temperature in
AV ¼ Vent area the thermostatic chamber was controlled by an air
C ¼ Vena contracta conditioning system. While circular orifices were
ρL and ρU ¼ Air densities in lower and upper used in the brine-water experiments by Estein
spaces, respectively. and others, square openings were adopted in the
However, this method will fail, over certain Heiselberg’s experiments as the vent. The concep-
ranges of pressure difference, when temperature tual configuration of the vents in their experimen-
difference exists between the spaces due to the tal setup is illustrated in Fig. 15.11.
antagonistic effect of buoyancy to pressure gra-
dient. Consider the situation where the lower
space temperature is higher than upper space a
temperature, for example. There is a flow across
the vent due to the air density difference
Δρð¼ ρU ρL Þ even if the pressure difference,
Δp, is basically zero. In such a case as mentioned
earlier in Fig. 15.9, where a pace has no other air Δp < 0 : VU > 0; VL = 0
passage besides the vent between the connecting b
spaces, a bidirectional flow will occur through
the vent in a certain range around Δp ¼ 0 and the
upward and downward flow rates will become Δp = 0 : VU = 0; VL = 0
equal due to continuity.
The horizontal vent flows under zero pressure c
difference, Δp ¼ 0, was investigated by Epstein
[9], Tan and Jaluria [10] and others, using brine-
water scale models. In these experiments, the
effects of vent configuration on the flow
characteristics were investigated for a large Δp > 0 : VU = 0; VL > 0
range of opening size and aspect ratio. Also,
Heiselberg [11] conducted the experiments for Fig. 15.10 The standard vent flow model for horizontal
the same purpose using a single opening test vents
468 T. Tanaka
Regime II (0.15 < L/D < 0.4): Countercurrent

orifice flow
TU, rU Regime III (0.4 < L/D < 3.25): Combination of
D an orifice flow and turbulent diffusion flow
Regime IV (3.25 < L/D): Turbulent and slow
VENT
L exchange flow, where counterflow within the
tube appears to be mixture of warm and cold
fluids with chaotic motion
where
D ¼ Diameter of the vent
TL, rL
L ¼ Length of the vent
Epstein defined the Froude number, Fr, as
V_ EX
Fig. 15.11 Schematic configuration of vents in the Fr ¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð15:39Þ
existing experiments for horizontal vent flow
gðT L T U ÞL5
TL
The flow through a horizontal vent under
Δp ¼ 0 condition is often unstable. Epstein [9]
identified four regimes of the exchange flow as a and developed the formulas for the exchange
function of aspect ratio of the vent as follows: flow through a horizontal vent according to the
Regime I (L/D < 0.15): Oscillatory exchange flow above regimes based on his brine-water experi-
ment data as follows:
8
>
> L
>
> 0:055 < 0:15
>
> D
>
>
>
> 1=2
>
> L
>
> 0:147 DL 0:15 < < 0:4
>
< D
Fr ¼ ð15:40Þ
>
> q
1
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
L
>
> 0:093 3 0:4 < < 3:25
>
> D
>
> 1 þ 0:084 DL 0:4
>
>
>
>
>
> L 3=2 L
>
: 0:32 D 3:25 < < 10
D
8
where >
>
L
< 0:115
>
> 0:050
V_ EX ¼ Exchange volume flow rate >
> D
>
>
>
< 1=2
TL ¼ Temperature in lower space L L
TU ¼ Temperature in upper space Fr ¼ 0:147 0:115 < < 0:55
>
> D D
g ¼ acceleration due to gravity >
>
>
> 1=2
>
> L L
Heiselberg, adopting the same Fr as Equa- >
: 0:077 0:55 < < 4:455
tion 15.39, developed a similar formula based D D
on his experimental data using the airs at differ- ð15:41Þ
ent temperature as follows:
www.ebook777.com
15 Vent Flows 469
Heiselberg data Epstein formula Epstein data Heiselberg formula

0.15
0.1
Fr
0.05
0
0.01 0.1 1 10 100
L/D
Fig. 15.12 Comparison of Epstein’s brine-water scale experiments and Heiselberg’s full-scale experiments for
horizontal vent flow
Since the vents in Heiselberg’s experiments were Kenton [12], Tan and Jaluria [10] and Heskestad
square, the vent diameter D was converted from and Spaulding [13]. Of these, the pressure differ-
the side length of the square vent, AV, in his ence, Δp, was not directly measured in the
experiments as experiments by Epstein and Kenton so the critical
rffiffiffi condition of transition from bi-directional to
4 uni-directional flow was determined by the flow
D¼ S 1:128S ð15:42Þ
π rate. The critical (flooding) condition is depen-
dent on vent condition and buoyancy strength.
Figure 15.12 plots the experimental data and the
For analyzing the above existing results, Cooper
equations from Epstein and Heiselberg together.
[14] introduced the following non-dimensional
Albeit the difference of the fluids used in their
parameters defined as follows:
experiments, the distributions of the measured
data seem to show a similar trend when plotted
Grashof number
in terms of the Fr.
In such an unstable configuration as where a
2gD3 ρΔρ
high density fluid is above and low density fluid is Gr ¼ ð15:43Þ
μ2
below, a bi-directional flow takes place through
the vent in the region of pressure around Δp ¼ 0. Froude number
However, as the pressure difference, Δp, is
increased the exchange flow in the direction of V_ U
Fr ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð15:44Þ
opposing to the pressure gradient decreases. If ΔT
AV 2gD
Δp exceeds some critical value, ΔpC, then the T
flow becomes uni-directional. The flows in such
ranges of Δp were investigated by Epstein and Non-dimensional pressure difference
470 T. Tanaka
Epstein and Kenton Tan and Jaluria Heskestad and Spaulding

0.5
0.4
0.3
FrFL
0.2
0.1
0
1.E+06 1.E+07 1.E+08
Gr
Fig. 15.13 Plot of FrFL as a function of Gr for small L/D data
Δp be attributed to the relatively small vent size in

Π¼ ð15:45Þ
4gΔρD their experiments. Similar dependence of FrFL
on Gr is found by Heskestad and Spaulding
where [13]. The experiments for Gr < 1.4 107 may
AV ¼ Vent area be in the range of laminar or transitional flow.
T ¼ ðT U þ T L Þ=2 From the viewpoint of practical applications to
realistic building fire configurations, the horizon-
ρ ¼ ðρU þ ρL Þ=2 tal vent flows of interest are probably in the large
Δρ ¼ ρL ρU Gr range, i.e., turbulent flow regime. The critical
Froude number, FrFL, from the existing data in
ΔT ¼ T L T U ; Temperature difference large Gr range are plotted vs non-dimensional
Δ p ¼ pU pL temperature difference, ΔT=T, in Fig. 15.14.
Let the critical volume flow rate and pressure Note here that when ΔT < 0, FrFL denotes the
difference where transition from bi-directional critical Froude number to purge the downward
flow to uni-directional flow just occur be V_ U, FL flow and attain the upward flooding flow. The
and ΔpFL, respectively, and let the corresponding critical (flooding) nondimensional pressure
Fr and Π be FrFL and ΠFL, respectively. Fig- difference, ΠFL, is plotted also vs ΔT=T in
ure 15.13 plots the critical (flooding) Froude Fig. 15.15 although the data are only available
number, FrFL, vs Grashof number, Gr, which from Heskestad and Spaulding [13].
were reduced from the existing experiments. As The effect of buoyancy will not disappear as
seen in this figure, FrFL is relatively insensitive soon as the pressure difference exceeds the criti-
to changes in Gr in the range about 3 107 < cal value and the vent flow has changed from
Gr < 3 108. However, there is a significant bi-directional to uni-directional. Even after
increase in FrFL for Gr < 1.4 107, i.e., for uni-directional flow has been established, signif-
the data of Tan and Jaluria [10], which seems to icant buoyancy effect still remains. For this
www.ebook777.com
15 Vent Flows 471
Fig. 15.14 Plot of FrFL as Epstein and Kenton Heskestad and Spaulding
a function of ΔT=T for 0.3
small L/D data
0.25
0.2
FrFL
0.15
0.1
0.05
0
−0.6 −0.4 −0.2 0 0.2 0.4 0.6
ΔT/T
Heskestad and Spaulding

0.8
0.6
PFL
0.4
0.2
0
−0.6 −0.4 −0.2 0 0.2 0.4 0.6
ΔT/T
Fig. 15.15 Plot of ΠFL as a function of ΔT=T for small L/D data
uni-directional flow regime, Cooper presented a seen that the standard Bernoulli flow equation,
plot of V_ U =V_ U, FL vs Δp/ΔpFL as seen in expressed by dashed line, overestimates the
Fig. 15.16, where V_ U, FL and ΔpFL are critical expected flow rate by a factor in excess of
(flooding) volume flow rate and pressure differ- 3, and that only after Δp/ΔpFL exceeds 3 or
ence, respectively. From this figure, it can be 4 the standard model provides flow estimate
472 T. Tanaka
Non flooding data Flooding data Bernoulli eq. Series4

8
6
VU/VU,FL
0
0 1 2 3 4
Δp/ΔpFL
Fig. 15.16 Plot of V_ U =V_ U, FL as a function of Δp/ΔpFL (¼ Fr=Fr FL as a function of Π/ΠpFL) for uni-directional flow
regime
correct to within a few tens of percent. If the The maximum value of the exchange flow
values of FrFL and ΠFL, which depend on tem- rate, V_ EX, MX , is attained at Δp ¼0. For shallow
perature difference, are determined using circular vents, this is obtained from Epstein [9] as
Figs. 15.14 and 15.15, respectively, V_ U, FL and rffiffiffiffiffiffiffiffiffiffiffiffiffi
ΔpFL can be calculated using Equations 15.44 4 ΔT
V_ EX, MX ¼ 0:055 AV gD ð15:46Þ
and 15.45, respectively. Then it follows that π T
uni-directional flow rate, V_ U , can be obtained
Cooper [14] presented the relationship of
for arbitrary pressure difference, Δp, from
V L =V_ Ex, MX and V_ N =V_ U, FL as a function of
_
Fig. 15.16.
Δp/ΔpFL as shown in Fig. 15.17. Then, it follows
The range of pressure 0 Δ p Δ pFL is the
that using this figure and Equation 15.46 together
mixed flow regime where there will be a
with the critical (flooding) pressure and volume
bi-directional flow through the vent, i.e., a flow
flow rate, the exchange flow rate and the net flow
from the upper to the lower space, V_ U , and a flow
rate can be estimated. Cooper also showed that
from the lower to the upper space, V_ L . The V_ L is his method well correlated the data of the flow
also called exchange flow, V_ EX . At the two through tube-like vents (i.e., moderate-to-large
extremes of the mixed flow regime, V_ U ¼ V_ L at L/D) rather than shallow vents, as shown in
Δ p ¼ 0 and V_ U ¼ V_ U, FL; V_ L ¼ 0 at Δ p ¼ Δ pFL . Fig. 15.18.
Let V_ N be the net volume flow rate from the upper,
or the higher pressure, space to the lower space, Accuracy of Vent Flow Calculations
i.e., V_ U V_ L . Generally, in the intermediate
range 0 < Δ p < Δ pFL , V_ N ¼ V_ U V_ L ¼ For nonbuoyant flows (using nozzles or orifices
_
V U V EX 0. This net volume flow rate, V_ N ,
_ in a straight run of pipe made and calibrated with
can be said pressure-driven part of the flow. a specific geometry over a known Reynolds
www.ebook777.com
15 Vent Flows 473
0.8 VL/VEx,MX VN/VU,FL
0.6
0.4
0.2
0
0.0 0.2 0.4 0.6 0.8 1.0
Δp/ΔpFL
Fig. 15.17 Plots of V_ L =V_ Ex, MX and V_ N =V_ U, FL as a function of Δp/ΔpFL
0.39 < L/D < 5.0 (tubes)

0.8
L/D = 0.019 (vent)
L/D = 0.114 (vent)
0.6
VL/VEx,MX
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
VN/VU,FL
Fig. 15.18 Plots of V_ L =V_ Ex, MX from Epstein and Kenton [12] as a function of V_ N =V_ U, FL
474 T. Tanaka
number range) one easily obtains 2 % accuracy. in the building. All these flows are initially
Thus, Equations 15.1, 15.2, 15.3, 15.4, 15.5, nonbuoyant. The flow through the building is
15.6, 15.7, 15.8, and 15.9 are capable of high simply flow through a complex system of pipes
accuracy. and orifices. As the fire grows larger, hot gas
For vents in vertical walls with limited inter- flows buoyantly out of the place of origin, while
nal room fire circulations, the best methods of cold gas flows in below. Thus, while the net flow
measurement may get 5 % accuracy. However, in (out-in) is just sufficient to accommodate the fire
real fires, induced circulations are often severe gas expansion, the actual volumetric hot gas out-
and unknown. Thus, errors of 10 % or higher flow may be 2.5 times larger than the inflow. A
must be expected. Even if flow instrumentation layer of hot gas moves along the ceiling of
is located in the vent itself, there is never enough connected spaces and at the first opportunity
to really account for variations over the vent proceeds up a stairwell or other ceiling (roof)
surface and time fluctuations originating in the opening into regions above [15]. The
fire phenomena inside the fire room. accumulating hot gas will help spread the fire
For vents in a horizontal surface, the majority while the newly created hot fire gases build a
of the results presented here are from water-brine new hot layer in the adjacent spaces. The flow
experiments in small holes. The experimental and pressure drop across each vent will then
accuracy is 10 %. However, for a real fire, the progress through a succession of situations as
errors are probably much higher. A typical case previously discussed. The flow throughout the
is a hole in the ceiling burned through by the building is, therefore, determined by the vent
flames from below. The hole geometry is very and flow friction drops along all of the available
irregular and is completely unknown. Further- flow paths from the fire to the outside of the
more, a fire directly below the hole supplies hot building.
gas with a considerable vertical velocity. Also, The vent flow calculation procedures
the ceiling jet flow often provides considerable described in this section are sufficiently accurate
cross flow. and general to compute the required flow-
Full-scale experimental results determining pressure drop relations for building flow
the effects of fire circulation, large density ratios, networks (except slow buoyant flows through
and large Reynolds numbers are needed. horizontal vents).
Vents as Part of the Building Flow

Network Room Pressure
A building is an enclosed space generally with General Equation to Control Room

floors and walls that divide the space both verti- Pressure
cally and horizontally into rooms, corridors, and
stairwells. A fire that starts at any place in the The above vent flow equations give flow rates of
building causes gas expansion, which raises the air and smoke through a vent as a function of
local pressure and pushes air throughout the pressure difference between the spaces at both
building through all pathways leading to the out- sides of an arbitrary vent. The next question is
side. If a window is open in the room of fire “How could we know the pressure difference?” It
origin, and there is little or no wind, little flow can be obtained from the conservations of mass
moves through the remainder of the building. If and heat of the rooms concerned and the ideal gas
there is no open window, the flow will move law. That is, the following equations hold for an
toward cracks and leaks wherever they may be arbitrary room, i:
www.ebook777.com
15 Vent Flows 475
Mass conservation Now consider a leaky room condition in which

X the magnitude of pressure buildup induced by the
d
ðρV R Þ ¼ m_ i j þ m_ ji ð15:47Þ heat release of the fire is insignificant compared
dt with the absolute atmospheric pressure, which is
Heat conservation the case in most building fires. Since ρT is a
constant under such conditions, the left-hand
d X
c p ðρV R T Þ ¼ Q_ Q_ h þ c p m_ i j T þ m_ ji T j side of Equation 15.48 becomes 0; hence,
dt X
ð15:48Þ Q_ ¼ Q_ h þ cp m_ i j T þ m_ ji T ¼ 0
Ideal gas law ð15:52Þ
Mp The mass flow rates in Equation 15.52, ṁij and

ρT ¼ ð15:49Þ ṁji, are given as a function of the differences of
R
the pressure between room i and adjacent spaces;
where in other words, this equation can be regarded as
j ¼ Index of adjacent room connected by the condition that the pressure of room i must
vents with the room considered satisfy. Incidentally, the same equation can be
V, VR ¼ Volume of the room obtained by using dρ=dt ¼ ðρ=T ÞdT=dt, from
ṁij ¼ Mass outflow rate from room i to an ρT ¼ constant, and Equation 15.51 into Equa-
adjacent room j tion 15.47, the mass conservation.
ṁji ¼ Mass inflow rate to room i from an In case a room contains stratified upper and
adjacent room j lower layers, the heat conservation equations for
_
Q ¼ Heat release rate of fire source the layers are written as
Q_ h ¼ Heat loss from the room gas due to heat
d
transfer cp ðρu V u T u Þ ¼ Q_ ðuÞ Q_ hðuÞ
∑ ¼ Summation with respect to all the vents dt
X
between all the adjacent rooms þ cp m_ i j T þ m_ ji Tj
In these equations subscript i is omitted for ρ, T, u
_ Q_ h , and VR of the room considered for simplicity

Q, ð15:53aÞ
of equation form. Normally, Q_ ¼ 0 for rooms other d
than the fire room. Strictly speaking, the terms for cp ðρl V l T l Þ ¼ Q_ ðlÞ Q_ hðlÞ
dt
mass fuel input and heat input due to temperature X
þ cp m_ i j T þ m_ ji Tj
rise of the fuel gasification should be added to
l
Equations 15.47 and 15.48, respectively, if the
ð15:53bÞ
room being considered is a fire room. Here these
terms are disregarded for clarity of the equations where (u) and (l ) denote that only the quantities
and for their relatively small significance. associated with the upper and lower layers are
Since the room volume, VR, is constant, the left- considered, respectively. It is meant in the above
hand side of Equation 15.48 can be extended as equations that heat transfer by radiation and mass
d dT d transport through a fire plume and/or entrainment
cp ðρV R T Þ ¼ cp ρV R þ cp T ðρV R Þ of opening jet that can occur between the upper
dt dt dt
and lower layers are included as well as those
ð15:50Þ
between adjacent rooms.
Hence, using Equation 15.47 into Equation 15.48 Now, adding up Equations 15.53a and 15.53b,
yields the equation for temperature as follows: noting that ρuTu ¼ ρlTl ¼ constant and Vu + Vl
¼ VR (constant) and that the heat and mass trans-
dT X
cp ρV R ¼ Q Q_ h þ cp m_ ji Tji T port between the upper and lower layers in the
dt same room offset each other, yields the same
ð15:51Þ equation as Equation 15.52; that is,
476 T. Tanaka
Equation 15.52 holds for the two-layer configu- The vent flow rate can be obtained by consid-
ration also. Note also that Equation 15.52 holds ering the heat conservations for the layers as
universally for transient as well as steady-state follows:
conditions.
d ðρu V u T u Þ
cp ¼ Q_ Q_ h þ c p T a m_ p ð15:56aÞ
EXAMPLE 1 For a clearer understanding of the dt
relationship between the conservations of mass d ð ρa V l T a Þ
and heat and the room pressure, consider the cp ¼ c p T a m_ p þ c p T a m_ d
dt
simple example shown in Fig. 15.19. An upper ð15:56bÞ
hot layer is developing under the ceiling due to
the fire plume flow above a fire source in a room. Noting that ρu T u ¼ ρa T a and V u þ V l ¼ V R
The room has only one vent at the bottom. The (constant) so that
mass conservations in this configuration can be
d d ðV u þ V l Þ
written for the upper and lower air layers, respec- c p ð ρu T u V u þ ρ a T a V l Þ ¼ c p ρa T a
tively, as dt dt
dV R
d ð ρu V u Þ ¼ c p ρa T a ¼0
¼ m_ p ð15:54aÞ dt
dt
adding up Equations 15.56a and 15.56b yields
d ð ρa V l Þ
¼ m_ p m_ d ð15:54bÞ Q_ Q_ h c p m_ d T a ¼ 0 ð15:57Þ
dt
where md is the rate of air flow through the bottom Therefore, the air flow rate through the vent at
vent. Adding up these equations yields the mass such a stage of fire is given by
conservation of the room as a whole as follows:
Q_ Q_ h
d m_ d ¼ ð15:58Þ
ðρ V u þ ρa V l Þ ¼ m_ d ð15:55Þ c pTa
dt u
Invoking Equation 15.3, the vent flow equa-
This equation implies that, in the smoke- tion, in Equation 15.58 or 15.57, the pressure
filling stage, the upper-layer volume with smaller difference between the room and the outside
density is increasing while the lower layer with can be obtained as [16]
greater density is decreasing; thereby the total
!2
mass in the room is decreasing just at the rate 1 Q_ Q_ h
of the vent flow. Δp ¼ ð15:59Þ
2ρa ðCAÞ2 c p T a
That is, the room pressure in this configuration

depends on the net heat input (i.e., heat release –
heat loss to boundary walls). At the very first
Qh
stage the upper layer is thin in thickness and low
in temperature, so the heat transfer rate, Q_ h , is
small but as the upper layer develops the Q_ h
Q increases. Therefore, if the heat release rate, Q,_
is constant, the room pressure will rise very
mp
rapidly at first and then will go down gradually.
l md Although it is not necessary to know the room
a a
pressure to calculate the vent flow rate in this
particular case, it is required to know the pressure
Fig. 15.19 Fire room at the stage of smoke layer filling in general cases such as when there is a pressure
www.ebook777.com
15 Vent Flows 477
difference distribution over the height of a vent Δp

hn ¼ ð15:61Þ
and when multiple vents are involved. ðρa ρÞg
Since Δp and hn are in one-to-one correspon-

Equation to Control Pressure
dence, solving the pressure and solving the neutral
at Steady State
plane height means the same. Incidentally, Δp in
Equation 15.61 is negative in this particular case.
Strictly speaking, there is no steady state in
The mass inflow and outflow rates through the
building fire phenomena but there are stages at
vent, md and mu, are given as a function of hn as
which fire behavior can be regarded as nearly
follows:
steady. At steady state dT=dt ¼ 0 in Equa-
tion 15.51; hence, Equation 15.52 becomes pffiffiffi
2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 3=2
X m_ d ¼ Cb 2gρa ðρa ρÞhn ð15:62aÞ
Q_ Q_ h þ cp m_ ij T þ m_ ji Tj 3
h X i
¼ Q_ Q_ h þ c p m_ ji Tj T pffiffiffi
X
þ cp T m_ þ m_ ji 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X ij m_ u ¼ Cb 2gρa ðρa ρÞðhv hn Þ3=2
¼ 0 þ cp T m_ i j þ m_ ji ¼ 0 3
ð15:62bÞ
Hence, Equation 15.52 is reduced to
X The neutral plane height, hn, can be solved by
m_ i j þ m_ ji ¼ 0 ð15:60Þ invoking Equation 15.60, which is simply m_ d ¼
m_ u in this case, as follows:
that is, the mass conservation equation. Of
course, this relation can also be obtained by hv hv
letting dρ/dt ¼ 0 in Equation 15.47. hn ¼
1=3 ¼
1=3 ð15:63Þ
ρa
1þ ρ 1þ T
Ta
EXAMPLE 2 Consider a room with only one
vent that is at steady state and uniform tempera- Using Equation 15.63 into Equations 15.62a
ture as shown in Fig. 15.20. Letting Δp be the and 15.62b, the vent mass flow rates can be
room pressure difference relative to the outdoor concretely obtained. Incidentally, Equation 15.63
space at the height of the sill of the vent, the is identical to the following pressure difference at
neutral plane height, hn, is the height of the sill of the vent:
Fig. 15.20 Single room at

steady state with uniform
temperature
Fire room
mu
hv – h n
hv
Neutral plane md
hn
a a
478 T. Tanaka
Fig. 15.21 Effective vent

area of a series of vents
m1 m2 mk mN
A1 A2 Ak AN
Δp1 Δp2 Δpk ΔpN
Δp
ðρa ρÞhv where Ae is the equivalent vent area, which is

Δ p ¼ ðρa ρÞhn ¼
1=3 ð15:64Þ given by
1 þ TTa
1=2
1 1 1 1
EXAMPLE 3 Equivalent vent area of a series Ae ¼ 2
þ 2 þ
þ 2 þ
þ 2
A 1 A2 Ak AN
of vents may be a typical example in which
ð15:69Þ
steady-state mass conservation law is invoked.
Figure 15.21 illustrates a system of spaces Incidentally, since the equation
connected by N vents with different areas pffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi
through which vent airflows are induced due to m_ ¼ CAk 2ρΔ pk ¼ CAe 2ρΔ p ð15:70Þ
a given pressure difference between both ends,
Δp. In this case, from the mass conservation holds for any of the vents, the pressure difference
for the N – 1 rooms involved in the system of across an arbitrary vent, Δpk, can be calculated
spaces, all the vent flow rates have to be the simply by
same; that is, 2
Ae
Δ pk ¼ Δp ð15:71Þ
m_ 1 ¼ m_ 2

¼ m_ k

m_ N ð m_ Þ ð15:65Þ Ak
The pressure difference across an arbitrary EXAMPLE 4 Natural smoke venting is a smoke
vent, Δpk, is unknown but the following equation control method that removes smoke from fire in a
holds building space to the outdoor space through a
smoke vent arranged in a roof or upper part of a
Δ p1 þ Δ p2 þ

þ Δ pk þ

þ Δ pN ¼ Δ p wall of a room. Figure 15.22a illustrates a typical
ð15:66Þ configuration of natural smoke venting at an
early stage of fire in a room having a smoke
and the pressure difference across an arbitrary vent at height, he, and an inlet above the floor,
vent, Δpk, is expressed as where we assume a constant heat release rate of
_ and steady-state smoke layer
m_ 2 fire source, Q,
Δ pk ¼ ð15:67Þ properties such as temperature, T, and layer inter-
2ρðCAk Þ2
face height, y. Under such a condition, the pres-
where ρ is the flowing air density. sure difference as shown by Fig. 15.22b develops
Substituting Equation 15.67 into Equa- with height due to the temperature difference
tion 15.66 reveals that the mass flow rate, ṁ, between the indoor and the outdoor space. As a
can be calculated using the given pressure differ- result, the pressure difference across the smoke
ence, Δp, as vent, Δp(he), which is the driving force of the
pffiffiffiffiffiffiffiffiffiffiffiffi smoke discharge, is induced. If the indoor
m_ ¼ CAe 2ρΔ p ð15:68Þ pressure at the floor level is Δp(0), relative to
www.ebook777.com
15 Vent Flows 479
a Ae
b Δp(he)
c Δp⬘(he)
me
he
mp
y a
y
md
Ad
Δp(0) Δp⬘(0)
Fig. 15.22 Configuration and pressure profile of natural smoke venting. (a) Natural smoke venting configuration.
(b) Pressure difference profile for large Ad. (c) Pressure difference profile for small Ad
the outdoor space, the pressure difference at Furthermore, an additional equation to obtain
height, he, is smoke layer density is necessary. It can be
provided from some appropriate energy conser-
Δ pðhe Þ ¼ Δ pð0Þ vation equation for the smoke layer; for example,
þ ð ρa ρÞ g ð h e y Þ ð15:72Þ
Q_ ¼ c p m_ ðT T a Þ þ αk AW ðT T a Þ ð15:77Þ
so that the rate of smoke exhaustion through the
vent, me, is where AW is the boundary wall area exposed to
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi heat transfer from the smoke layer and αK is the
m_ e ¼ CAe pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2ρΔ pðhe Þ ffi effective heat transfer coefficient [17–19].
¼ CAe 2ρ½Δ pð0Þ þ ðρa ρÞgðhe yÞ Using the above equations and the equation of
ð15:73Þ gas state, we can obtain the smoke layer tempera-
ture and interface height, the vent flow rates, the
In Equation 15.73, Δp(0), ρ, and y are plume flow rate at the interface height, and the
unknown so we need to seek additional pressure differences, although some trial-and-
relationships. error method must be invoked because of the non-
The air inflow rate through the inlet above the linearity of Equation 15.75.
floor, ṁd, is expressed as On the other hand, it is more practical to know
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi the necessary smoke vent area to assure a certain
m_ d ¼ CAd 2ρa Δ pð0Þ ð15:74Þ
level of clear height for evacuation or fire fight-
The plume flow rate at the layer interface ing than to predict the behaviors of the smoke
height, mp, can be estimated using some appro- layer and vent flows under arbitrary conditions.
priate plume formula such as If the critical layer interface height, yc, is
specified, the required smoke vent area, Ae, can
m_ p ¼ 0:08Q_
1=3 5=3
y ð15:75Þ be obtained by simply following the procedure
below one by one, without invoking any trial-
At steady state, mass conservation holds for and-error method [18].
the smoke and the air layers so that 1. Calculate the plume flow rate at height, yc:
m_ ¼ 0:08Q_
1=3 5=3
m_ e ¼ m_ p ¼ m_ d ð m_ Þ ð15:76Þ yc ð15:78aÞ
480 T. Tanaka
2. Calculate the smoke layer temperature: numerical methods if multiple vents, mechanical
ventilation, transient condition, and fuel mass
Q_ input are involved.
T ¼ Ta þ ð15:78bÞ
c p m þ αK AW When considering the pressures in building
rooms, it is convenient to take them relative to
3. Calculate the smoke layer density:
the outdoor space rather than to use absolute
352:8 atmospheric pressure since vent flows in fire are
ρ¼ ð15:78cÞ induced by only trivial fractions of the atmo-
T
spheric pressure. In the case of a single room, if
4. Calculate the pressure at floor level: the room pressure, p, is taken as such a relative
m_ 2 pressure to the outdoor space at the level of the
Δ pð 0Þ ¼ ð15:78dÞ room floor, any vent flow rate can be given as a
2ρa ðCAd Þ2
function of the only pressure. Therefore, Equa-
5. Calculate the pressure at the smoke vent: tion 15.52 can be expressed as
Δ pðhe Þ ¼ Δ pð0Þ f ð pÞ Q_ Q_ h
þ ðρa ρÞgðhe yc Þ ð15:78eÞ X
þ cp ðm_ u T þ m_ d T a Þ
6. Calculate the smoke vent area: ¼0 ð15:79Þ
m_
Ae ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð15:78fÞ Because the vent flow terms in Equation 15.79
C 2ρΔ pðhe Þ are a function of p, the problem is reduced to the
solution of f( p) ¼ 0 for p.
Incidentally, Equation 15.73 might suggest that
If the room is at uniform temperature, for
the smoke exhaust rate would increase in propor-
example, the neutral plane height is given as a
tion to the smoke vent area, but it is not necessarily
function of p as
the case. It is true that the larger the smoke vent
area the larger the smoke exhaust rate, but the p
hn ¼ ð15:80Þ
smoke exhaust rate is affected by the size of the g ð ρa ρÞ
air inlet at the bottom. This is because the mass
Pressure profile across a vent varies depending
conservation expressed by Equation 15.76 requires
on the height of the vent relative to the neutral
the smoke venting rate, ṁe, be the same as the air
plane as shown in Fig. 15.23. The vent flow rates
inflow rate through the inlet, ṁd. Even though the
are given as a function of the neutral plane height
smoke vent area, Ae, is very large, the indoor
as follows:
pressure at floor height, Δp(0), becomes low
1. when hn < hb
when the inlet area is small, which eventually
results in a small pressure difference across the 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
smoke vent, Δp(he), as understood from Equa- m_ u ¼ Cb 2ρgðρa ρÞ
3h i
tion 15.72 and illustrated by Fig. 15.22c. ðht hn Þ3=2 ðhb hn Þ3=2
m_ d ¼ 0
ð15:81aÞ
Numerical Computation of Room
Pressure 2. when hb < hn < ht
Except in particularly simple cases such as the 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

above examples, solving room pressure problems m_ u ¼ Cb 2ρgðρa ρÞðht hn Þ3=2
3
requires using numerical computation. Even in 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
m_ d ¼ Cb 2ρgðρa ρÞðhn hb Þ3=2
the case of single fire rooms, it is too difficult to 3
solve room pressure problems without using ð15:81bÞ
www.ebook777.com
15 Vent Flows 481
Fig. 15.23 Pressure

profile across vents for
nonstratified room (a) T, ρ Tα, ρα
condition
Neutral plane
ht
hb ht
(b)
hn
(c)
ht hb
hb
p (<0) p (<0)
Fig. 15.24 Pressure

profile across vents for
stratified room condition (a) T, ρ Tα, ρα
Neutral plane
ht
hb ht
(b)
hn hi
(c)
ht hb
hb
p (<0) p (<0)
3. when ht < hn p
hn ¼ þ hi ð15:82Þ
g ð ρa ρÞ
m_ u ¼ 0
2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi The neutral plane height, hn, can be lower
m_ d ¼ Cb 2ρa gðρa ρÞ ð15:81cÞ than the interface height, hi, under some
3h i
ðhn hb Þ3=2 ðhn ht Þ3=2 conditions, such as rapid transient expansion
of the upper layer and mechanical air supply
From Equations 15.80 and 15.81, it is easily to the room, although most often hi is lower
understood that Equation 15.79 is a nonlinear than hn. In case hi is lower than hn, the
function with respect to p. vent flow rates are calculated as follows:
If the room is stratified into two uniform [15, 20, 21]
layers, a little more complicated situation can 1. when hn < hb
emerge; that is, the pressure profile across a
vent is affected by the layer interface as well as 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
m_ u ¼ Cb 2ρgðρa ρÞ
the neutral plane as exemplified in Fig. 15.24. 3h i
In this particular example in which the lower ðht hn Þ3=2 ðhb hn Þ3=2
layer is at the same temperature as the outdoor m_ d ¼ 0
space, the neutral plane height is given by ð15:83aÞ
482 T. Tanaka
2. when hb < hi < hn < ht 1

df
pðkþ1Þ ¼ pðkÞ f pð k Þ ð15:84Þ
2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dp
m_ u ¼ Cb 2ρgðρa ρÞðht hn Þ3=2
3 where p(k) is the kth estimate of the pressure.
2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
m_ d ¼ Cb 2ρgðρa ρÞ ðht hn Þ3=2 In the case of a structure having multiple
3 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
rooms, every room has to satisfy Equation 15.52
þCb 2ρa ð pÞðhi hb Þ and the flow through the vents in the boundary of
ð15:83bÞ the room is a function of the pressure difference
with the adjacent rooms. If a structure consists of
3. when ht < hi
N rooms connected arbitrarily to each other,
Equation 15.52 for an arbitrary room i may be
m_ u ¼ 0 expressed in the most general form as [20, 21]
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð15:83cÞ
m_ d ¼ Cb 2ρa ð pÞðht hb Þ f i ð p1 ; . . . ; pi ; . . . ; p N Þ
X
Needless to say, Equation 15.79 becomes a non- Q_ Q_ h þ c p m_ i j T þ m_ ji T j ¼ 0
linear function with respect to p in this case also. ð15:85Þ
If the lower temperature is different from the
outdoor space, the pressure profiles across the Since every room in the structure must satisfy
vents are generally more complex. Nevertheless, the identical equation, the problem is reduced
the vent flow rates can be given as a function of p. to the solution of the coupling equations as
Since Equation 15.79 is nonlinear, it is impos- follows:
sible to obtain an analytical solution except in
9
very limited cases, so usually a numerical f 1 ð p1 , . . . , pi . . . , pN Þ ¼ 0 >
>
method is employed. The idea of the ...........................> >
=
Newton–Raphson method, which is a typical f i ð p1 , . . . , p i . . . , pN Þ ¼ 0 ð15:86Þ
>
example of iterative numerical solution methods, ...........................> >
>
;
is illustrated in Fig. 15.25. In this method the f N ð p1 , . . . , pi . . . , pN Þ ¼ 0
pressure is estimated successively starting from
the first estimate until the pressure acceptably A multidimensional Newton–Raphson method
satisfies f( p) ¼ 0 according to the procedure as may be used for the solution of Equation 15.86.
follows: The iteration procedure is expressed by
Fig. 15.25 Iterative f (p)

numerical solution of room
pressure by Newton–
Raphson method f (p (1))
f (p (3))
p (2)
p
p (1)
p (3)
f (p (2))
www.ebook777.com
15 Vent Flows 483
49
0.38
0.41 21 27 41
1.63 1.74 1.27 1.38

1.99
20 0.76 20 0.79 0.61
20 25 39
1.46 0.90
1.10 0.69
1.37
20 0.77 20 0.75 0.55
25
20
23 35
1.35 0.04 1.17
0.34
0.70 0.54 0.71
0.30
20 20 0.24 0.34
0.41
21 4.57
102 391 207

0.65
0.11 2.21
1.29 1.08 1.14
0.18 38 0.96 226 0.31 0.97 0.93
38
0.69
0.69 0.69 0.69
3.29
20 20 20 21
20
Fig. 15.26 An example of smoke movement prediction by a two-layer zone model (Number: temperature [ C],
number !: vent flow rate or plume flow rate [kg/s])
h i1 h i
pðkþ1Þ ¼ pðkÞ JðkÞ f pðkÞ ð15:87Þ pðkþ1Þ ¼ pðkÞ þ Δ p ð15:90Þ
Figure 15.26 shows an example of smoke

where p ¼ ( p1,. . ., pi,. . ., pN) and f ¼ ( f1,. . ., movement in a building predicted using the
f2,. . ., fN) and [J] is the Jacobian matrix defined by two-layer zone model BRI2, in which vent
2 3 smoke and air flows were calculated using the
∂ f 1 =∂ p1 ... ∂ f 1 =∂ p j ... ∂ f 1 =∂ pN
6 ... ... ... ... ... 7 above multidimensional Newton–Raphson
6 7
½J ¼ 6
6 ∂ f i =∂ p1 ... ∂ f i =∂ p j ... ∂ f j =∂ pN 7
7 method.
4 ... ... ... ... ... 5
∂ f N =∂ p1 ... ∂ f N =∂ p j ... ∂ f N =∂ pN
ð15:88Þ
Summary
In actual calculation, the inverse matrix of the
Jacobian matrix, [J(k)]1, is not calculated but the Most of the equations for vent flows in this chap-
coupling linear equations ter are basically derived from Bernoulli’s equa-
h i h i tion for steady flow of ideal and incompressible
J ðkÞ Δ p ¼ f pðkÞ ð15:89Þ fluid. The mass and volume vent flows are given
as a function of the flowing gas density, area of
are directly solved using an appropriate method the vent, and the pressure difference across the
such as Gaussian elimination for correction pres- vent with a coefficient called flow coefficient or
sure increments Δp and the new estimate is cal- opening coefficient. The value of the flow coeffi-
culated as cient, C, varies depending on the size and shape
484 T. Tanaka
of a vent. For usual doorway openings and P Perimeter (m)

windows in a building, the value of the C is p Pressure (Pa)
known to be from 0.6 to 0.7 from many experi- Q_ Heat release rate of fire source (kW)
mental measurements. Q_ h Heat loss by heat transfer (kW)
In a building fire, the heat release and the R Gas constant (J/kg mol K)
flows induced by the fire cause temperature Re Reynolds number ()
distributions in spaces in the building, which T Temperature (K)
then cause vertical pressure difference distribu- u Velocity (m/s)
tion across a vent. When such a vertical temper- V Volume (m3)
ature distribution exists across a vent, it is often V_ Volume flow rate (m3/s)
convenient for calculation of the vent flow rates VR Room volume (m3)
to first obtain the neutral plane height, which is y Vertical coordinate (m)
given as a function of the pressure difference at a αK Effective heat transfer coefficient
reference height and gas density difference in the (kW/m2K)
spaces at both sides of the vent. Δ Increment of
The equation to control the space (room) δ Depth (see Fig. 15.6) (m)
pressures at a reference height in building spaces γ ¼ cp/cv Isentropic exponent ()
can be obtained by considering the mass and heat Π Non-dimensional pressure ()
conservations in the spaces and the equation of ρ Density (kg/m3)
gas state. In general, the equation is a function of μ Viscosity (Ns/m2)
the vent flow rates, the temperatures of the spaces,
and the heat addition and loss, but at steady state, Subscripts
it is reduced to be the mass conservations of the
spaces. When multiple spaces are involved in a a Atmosphere
calculation of the flows in fire, analytical b Sill of vent
solutions are only possible for very limited c Ceiling of room
conditions, so generally some numerical calcula- d Lower
tion method must be invoked to solve the cou- f Floor
pling nonlinear equations for the space pressures. g Gauge
i Hot-cold interface
ij From room (layer) i to room (layer) j
Nomenclature j Index of layer
A Area (m2) L, l Lower
a Length (m) n Neutral axis
b Width (m) O2 Oxygen
C Flow coefficient () t Soffit of vent
cp Specific heat at constant pressure u, U Upper
(kJ/kg K) v, V Vent, in the vent
cv Specific heat at constant volume 0 Reference height
(kJ/kg K) 1 Upstream of orifice
D Orifice diameter (m) 2 Downstream of orifice
Fr Froude number ()
g Gravity constant (m/s2)
Gr Grashof number () References
h Height (m)
1. H. Rouse, Fluid Mechanics for Hydraulic Engineers,
[J] Jacobian matrix McGraw-Hill, New York (1938).
L Orifice length (m) 2. Mark’s Mechanical Engineers Handbook, McGraw-
M Molecular weight (kg/kg mol) Hill, New York (1958).
ṁ Mass flow rate (kg/s)
www.ebook777.com
15 Vent Flows 485
3. J.S. Newman and P.A. Croce, Serial No. 21011.4, Fac- 14. L. Y. Cooper, “Combined buoyancy- and pressure-
tory Mutual Research Corp., Norwood, MA (1985). driven flow through a shallow, horizontal, circular
4. D.J. McCaffrey and G. Heskestad, “Robust Bidirec- vent”, HTD-Vol. 299, Heat Transfer With Combined
tional Low-Velocity Probe for Flame and Fire Appli- Modes, ASME, Chicago (1994).
cation—Brief Communications,” Combustion and 15. T. Tanaka, “A Model of Multiroom Fire Spread,” Fire
Flame, 26, pp. 125–127 (1976). Science and Technology, 3, p. 105 (1983).
5. J. Prahl and H.W. Emmons, “Fire Induced Flow 16. S. Yamada and T. Tanaka, “Reduced Scale
Through an Opening,” Combustion and Flame, Experiments for Convective Heat Transfer in the
25, pp. 369–385 (1975). Early Stage of Fires,” International Journal on Engi-
6. K.D. Steckler, H.R. Baum, and J. Quintiere, 20th neering Performance-Based Codes, 1, 3, pp. 196–203
Symposium on Combustion, Pittsburgh, PA (1984). (1999).
7. J. Quintiere and K. DenBraven, NBSIR 78–1512, 17. T. Tanaka and T. Yamana, “Smoke Control in Large
National Bureau of Standards, Washington, DC Scale Spaces (Part 1, Analytic theories for simple
(1978). smoke control problems),” Fire Science and Technol-
8. H.E. Mitler and H.W. Emmons, NBS-GCR-81-344, ogy, 5, 1, pp. 31–40 (1985).
National Bureau of Standards, Washington, DC 18. T. Tanaka, “Performance-Based Fire Safety Design
(1981). Standards and FSE Tools for Compliance Verifica-
9. M. Epstein, “Buoyancy-driven exchange flow through tion,” International Journal on Engineering
small openings in horizontal partition, with special Performance-Based Codes, 1, 3, pp. 104–117 (1999).
reference to flows in multicompartment enclosures”, 19. B.J. McCaffrey, J.G. Quintiere, and
Journal of Heat Transfer, 110, pp.885–893 (1988) M.F. Herkeleroad, “Estimating Room Temperature
10. Q. Tan and Y. Jaluria, NIST-G&R-92-607, National and Likelihood of Flashover Using Fire Test Data
Institute of Standards and Technology, Gaithersburg, Corrections,” Fire Technology, 17, 2, pp. 98–119
MD (1992). (1981).
11. Heiselberg, P. and Li, Z., (2007), "Experimental study 20. T. Tanaka and K. Nakamura, “A Model for Predicting
of buoyancy driven natural ventilation through hori- Smoke Transport in Buildings,” Report of the Build-
zontal openings", Proceedings of Roomvent 2007 : ing Research Institute, No. 123, Ministry of Construc-
Helsinki 13–15 June 2007.. tion, Tsukuba, Japan (1989).
12. M. Epstein and M.A. Kenton, “Combined Natural 21. T. Tanaka and S. Yamada, “BRI2002: Two Layer
Convection and Forced Flow Through Small Zone Smoke Transport Model,” Fire Science and
Openings in a Horizontal Partition, with Special Ref- Technology, 23, Special Issue (2004).
erence to Flows in Multicompartment Enclosures,”
Journal of Heat Transfer, 111, pp. 980–987 (1989).
13. G. Heskestad and R. D. Spaulding, “Inflow of air Takeyoshi Tanaka is a professor emeritus at Kyoto Uni-
required at wall and ceiling apertures to prevent versity. His performance–based areas of expertise are fire
escape of fire smoke”, Proceeding of the 3rd Interna- modeling, smoke control, and fire safety design. His pro-
tional Symposium on Fire Safety Science, fessional experience includes research for the Building
pp.919–928 (1991) Research Institute of Japan’s Ministry of Construction.
Effect of Combustion Conditions

on Species Production 16
Daniel T. Gottuk and Brian Y. Lattimer
Introduction Species product formation is affected by the

compartment geometry, ventilation, fluid dynam-
A complete compartment fire hazard assessment ics, thermal environment, chemistry, and mode of
requires a knowledge of toxic chemical species burning. The mode of burning and ventilation are
production. Although combustion products two of the key conditions that dictate product
include a vast number of chemical species, in formation. These conditions can be used to classify
practical circumstances the bulk of the product fires into three general categories: (1) smoldering,
gas mixture can be characterized by less than (2) free- (or open-) burning fires, and (3) ventila-
10 species. Of these, carbon monoxide tion-limited fires. Smoldering is a slow combus-
(CO) represents the most common fire toxicant tion process characterized by low gas temperatures
(see Chap. 63). Over half of all fire fatalities and no flaming. Under these conditions, high
have been attributed to CO inhalation levels of CO can be generated. Chapters 63, 19,
[1, 2]. Concentrations as low as 4000 ppm (0.4 % and 36 discuss this mode of burning in detail; thus,
by volume) can be fatal in less than an hour, and it will not be discussed further here. Free-burning
carbon monoxide levels of several percent have fires are flaming fires that have an excess supply of
been observed in full-scale compartment fires. A air. These well-ventilated fires (discussed in
complete toxicity assessment should not only Chap. 36) are generally of little concern in terms
include the toxicity of CO but also include the of generating toxic species. This chapter focuses
synergistic effects of other combustion products, primarily on the third category, ventilation-limited
such as elevated CO2 and deficient O2 levels. flaming fires. These fires consist of burning
The transport of combustion products away materials inside an enclosure, such as a room, in
from the room of the fire’s origin is of the utmost which airflow to the fire is restricted due to limited
importance, because nearly 75 % of the fatalities ventilation openings in the space. As a fire grows,
due to smoke inhalation occur in these remote conditions in the space will transition from
locations [3]. However, conditions close to the overventilated to underventilated (fuel rich). It is
compartment of origin will govern the levels that normally during underventilated conditions that
are transported to remote locations. The research formation of high levels of combustion products,
in this area has focused on characterizing species including CO, creates a major fire hazard.
levels produced under a variety of conditions, both This chapter discusses the production of species
inside and nearby the compartment of fire origin. within a compartment fire and the transport of
these gases out of the fire compartment to adjacent
areas. Engineering correlations are presented
D.T. Gottuk (*) • B.Y. Lattimer
Jensen Hughes, 3610 Commerce Drive, Suite 817, along with brief reviews of pertinent work on
Baltimore, MD 21227, USA

www.ebook777.com
16 Effect of Combustion Conditions on Species Production 487
species production in compartment fires. These It is generally under these conditions that
sections provide the background and basis for products of incomplete combustion are created.
understanding the available engineering Typically, the fire plume extends into the upper
correlations and the range of applicability and layer, such that layer gases recirculate through
limitations. An engineering methodology is the upper part of the plume.
presented to utilize the information given in this Depending on both the size of the room and
chapter. the size of the fire, it is possible to have a fire
This chapter is organized according to the plume that cannot be contained within the room,
following outline: resulting in flame extension out of windows or
Basic Concepts doors. Flame extension can occur when the fire
Species Production Within Fire Compartments plume impinges on the ceiling and the ceiling jets
Hood Experiments are longer than the distance from the plume to
Compartment Fire Experiments outside vent openings (Fig. 16.2). Flame exten-
Chemical Kinetics sion is different from a second burning phenom-
Fire Plume Effects enon outside of the fire compartment, called
Transient Conditions external burning, which is discussed below. The
Species Transport to Adjacent Spaces main point to understand is that flame extension
Engineering Methodology outside of the fire compartment is a result of a fire
that is too large to be contained in the room.
Flame extension can occur during both over-
Basic Concepts and underventilated burning conditions. To esti-
mate when flame extension may occur, the max-
In a typical compartment fire, a two-layer system imum heat release rate that can be supported by
is created. The upper layer consists of hot the compartment ventilation needs to be deter-
products of combustion that collect below the mined. Flame length correlations can then be
ceiling, and the lower layer consists of primarily used to determine whether flames will extend
ambient air that is entrained into the base of the outside of the compartment.
fire (Fig. 16.1). Initially, the fire plume is totally As a fire progresses and the upper layer
in the lower layer, and the fire burns in an descends, the layer will spill below the top of
overventilated mode similar to open burning. doorways or other openings into adjacent areas.
Due to excess air and near-complete combustion, The hot, vitiated, fuel-rich gases flowing into
little CO formation is expected in this mode. (See adjacent areas can mix with air that has high O2
Chap. 36, for yields.) As the fire grows, ventila- concentrations to create a secondary burning
tion paths in the room restrict airflow, creating zone outside of the compartment (Fig. 16.3).
underventilated (fuel-rich) burning conditions. This is referred to as external burning. External
Upper layer
Flame
extension
Lower layer
Upper
Air layer
Fig. 16.1 An overventilated compartment fire with the Fig. 16.2 A fire compartment with flame extension out
fire plume below the layer interface of the doorway
488 D.T. Gottuk and B.Y. Lattimer
nprod ¼ Number of moles of products of

complete combustion per mole of reactants
(stoichiometric mixture of fuel and oxidant
External streams)
burning
Cp ¼ Heat capacity of products of complete
Air
Layer combustion (kJ/gmol K)
burning The use of the ignition index is discussed in
Air detail in Chap. 17, of this book. An ignition
index greater than 1.0 indicates that ignition is
expected if the mixture contains sufficient fuel.
Fig. 16.3 An underventilated compartment fire with
external burning of fuel-rich upper layer gases
Species Yields
burning can also be accompanied by layer burn-
ing. Layer burning is the ignition of fuel-rich The generation of fire products in compartment
upper-layer gases at the interface between the fires can be quantified in terms of species yields,
upper and lower layers. External burning and Yi, defined as the mass of species i produced per
layer burning occur due to the buildup of suffi- mass of fuel burned (g/g):
cient fuel in an atmosphere that is able to mix
with available oxygen. These phenomena can mi
Yi ¼ ð16:2Þ
only occur during underventilated burning mf
conditions. In some circumstances, external
Similarly, oxygen is expressed as the depletion of
burning can decrease human fire hazard through
O2 (i.e., DO2), which is the grams of O2 consumed
the oxidation of CO and smoke leaving the fire
per gram of fuel burned:
compartment (see the section in this chapter,
“Species Transport to Adjacent Spaces”). mO2
DO2 ¼ ð16:3Þ
The occurrence of external burning has been mf
predicted using a compartment layer ignition
model developed by Beyler [4] (see Chap. 17). The normalized yield, fi, is the yield divided by
Beyler derived a relationship called the ignition the theoretical maximum yield of species i for the
index to predict the ignition of gases at the inter- given fuel, ki. For the case of oxygen, f O2 is the
face of the upper and lower layers inside a com- normalized depletion rate, where ki is the theoret-
partment. The ignition index, I, is defined as ical maximum depletion of oxygen for the given
amount of fuel. As a matter of convenience, the
X C j =100 ΔH c, j use of the term yield throughout this chapter will
I¼ ð T SL, j 1:0 ð16:1Þ
j
also include the concept of oxygen depletion.
nprod C p dT As in Chap. 36, the normalized yield is also
To
aptly referred to as the “generation efficiency”
where of compound i. By definition, the normalized
j ¼ Fuel species of interest yields range from 0 to 1, and are thus good
Cj ¼ Volume concentration of fuel j when fuel indicators of the completeness of combustion.
stream is stoichiometrically mixed with For example, under complete combustion
oxidant stream conditions the normalized yields of CO2, H2O,
ΔHc,j ¼ Heat of combustion of the species and O2 are 1. As a fire burns more inefficiently,
j (kJ/gmol) these yields decrease. The use of normalized
TSL,j ¼ Adiabatic flame temperature at the stoi- yields is also useful for establishing mass
chiometric limit for fuel species j (K) balances. The conservation of carbon requires
To ¼ Temperature of the gas mixture prior to that
reaction (K)
www.ebook777.com
f CO þ f CO2 þ FTHC þ f resid:C ¼ 1 ð16:4Þ depending on conditions, is typically 10–20 %

by volume.
where fTHC is the normalized yield of gas-phase
total hydrocarbons and fresid.C is the normalized Xiwet ¼ ð1 XH2 Owet ÞXidry ð16:8Þ
yield of residual carbon, such as soot in smoke or
Reliable water concentration measurements are
high molecular weight hydrocarbons that con-
difficult to obtain. Therefore, investigators have
dense out of the gas sample.
calculated wet species concentrations using
For two-layer systems the yield of all species
the above relationship with the assumption that
except oxygen can be calculated as follows:
the molar ratio, C, of H2O to CO2 at any equiva-

Xiwet m_ f þ m_ a Mi lence ratio is equal to the calculated molar ratio
Yi ¼ ð16:5Þ at stoichiometric conditions [5, 6]. Based on this
m_ f Mmix
assumption, Equation 16.9 can be used to calcu-
where late wet species concentrations from dry concen-
tration measurements.
Xiwet ¼ The wet mole fraction of speciesi
m_ a ¼ The mass air entrainment rate Xidry
Xiwet ¼ ð16:9Þ
1 þ CXCO2dry
into the upper layer
m_ f ¼ The mass loss rate of fuel
Mi ¼ The molecular weight of speciesi Equivalence Ratio
Mmix ¼ The molecular weight of the mixture
ðtypically assumed to be that of airÞ The concept of a global equivalence ratio (GER)
can be used to express the overall ventilation of a
The depletion rate of oxygen is calculated as control volume, such as a fire compartment.
However, due to the complex interaction
0:21m_ a MO2 =Ma XO2wet m_ f þ m_ a MO2 =Mmix
DO2 ¼ between the plume and the upper and lower
m_ f
layers, as well as the potential extension of the
ð16:6Þ fire beyond the initial compartment, a unique
definition for the GER does not exist. Therefore,
The normalized yield, fi, is simply calculated
if one uses the term GER, it must be associated
by dividing the yield by the maximum theoretical
with a defined control volume.
yield
The first efforts in developing the GER
Yi concept were based on hood experiments
fi ¼ ð16:7Þ
ki [7–11] (e.g., as in Fig. 16.4) in which the idea
of a plume equivalence ratio was introduced. The
Typical operation of common gas analyzers plume equivalence ratio, ϕp, is the ratio of the
requires that water be removed from the gas mass of fuel burning, mf, to the mass of oxygen
sample before entering the instrument. Conse- entrained, ma, into the fire plume (below the
quently, the measured gas concentration is con- upper layer) normalized by the stoichiometric
sidered dry and will be higher than the actual fuel-to-oxygen ratio, r O2 .
wet concentration. Equation 16.8 can be used to
calculate the wet mole fraction of species Xiwet , m f =mO2
from the measured dry mole fraction, Xidry . As ϕp ¼ ð16:10aÞ
r O2
can be seen from Equation 16.8, the percent
difference between Xidry and Xiwet is on the order Since oxygen is typically entrained into a fire
of the actual H2O concentration which, plume via air, ϕp is commonly defined as

the two-layer system
created in the hood
experiments of Beyler [8, 9]
Layer
interface
Exhaust and
gas sampling Air
Burner
m f =ma can be defined as the time required for a unit

ϕp ¼ ð16:10bÞ volume of air to move through the upper-layer
r
where ma is the mass of air entrained into volume, and can be characterized according to
the plume (in the lower layer) and r is the stoi- Equation 16.11.
chiometric mass fuel-to-air ratio. As discussed in
V ul ρul
the section on species production within fire tR ¼ ð16:11Þ
compartments, this simple characterization of m_ exhaust
the equivalence ratio well represented the where
global conditions that existed in the first m_ exhaust ¼ Mass flow rate of gases out of the layer ðkg=sÞ
hood and compartment fire experimental ρul ¼ Density of the upper‐layer gases ðkg=m3 Þ
configurations. V ul ¼ Volume of the upper layer ðm3 Þ
In order to more accurately describe the For example, consider a compartment fire
time integrated conditions within the upper burning with a plume equivalence ratio of 0.5
layer, a second equivalence ratio was defined with upper layer gases that have a residence
for this control volume [7, 10, 11]. The upper- time of 20 s. If the fire grows quickly such that ϕp
layer equivalence ratio, ϕul, is the ratio of the increases to a value of 1.5 in about 5 s, ϕul would
mass of the upper layer that originates from fuel now lag behind (less than 1.5). The fuel rich
sources, to the mass of the upper layer that (ϕp ¼ 1.5) gas mixture from the plume is effec-
originates from any source of air into the upper tively diluted by the upper-layer gases since there
layer, divided by the stoichiometric fuel-to-air has not been sufficient time (greater than 20 s) for
ratio. the layer gases to change over. The result is that
The two equivalence ratios (ϕp and ϕul) are ϕul will have a value between 0.5 and 1.5. Another
not necessarily the same. As a fire grows, the instance when ϕul can differ from ϕp is when
upper-layer composition represents a collective additional fuel or air enters the upper layer
time history of products. In an ideal two-layer directly. An example of this would be the burning
fire, where all air enters the upper layer through of wood paneling in the upper layer.
the plume, ϕul is the same as ϕp only during The calculation of ϕul can be a complex task.
steady burning conditions. If the burning rate of Either a fairly complete knowledge of the gas
the fire changes quickly compared to the resi- composition is needed [7] or time histories of
dence time of the gases in the upper layer, the ventilation flows and layer residence times are
upper-layer equivalence ratio lags behind the needed to be able to calculate ϕul. Toner [7] and
plume equivalence ratio. The residence time, tR, Morehart [12] present detailed methodologies
www.ebook777.com
for calculating ϕul from gas composition m_ f m_ f r o

ϕ¼ ra ¼ ð16:13bÞ
measurements. Equation 16.12 can be used to m_ a m_ a Y O2, air
calculate ϕul if the mass flow rates can be
expressed as a function of time. where
ðt r a ¼ Mass air‐to‐fuel ratio
1 m_ f ðt0 Þdt0 r o ¼ Oxygen‐to‐fuel mass ratio
ϕul ¼ ð tttR
ð16:12Þ Y O2, air ¼ Mass fraction of oxygen in air ð0:23Þ
0 0
r m_ a ðt Þdt The formulation of Equation 16.13b allows
ttR
direct use of values tabulated for various fuels
Although termed the upper-layer equivalence in Appendix 3, Table 3.2, of this handbook.
ratio, ϕul actually represents the temporal aspect Another useful expression for ϕ can be derived
of the equivalence ratio no matter what the con- from Equation 16.13b by multiplying the numer-
trol volume. For instance, the control volume may ator and denominator by the fuel heat of combus-
be the whole compartment, as shown in Fig. 16.1. tion, Δhc, and recognizing that the heat release
In this case, the compartment equivalence ratio, per mass of oxygen consumed, E, is equal to Δhc
ϕc, is defined as the ratio of the mass, mf, of any over ro, yielding
fuel entering or burning in the compartment to the
mass, ma, of air entering the compartment Q_ 1 Q_ 1
ϕ¼ ¼ ð16:13cÞ
normalized by the stoichiometric fuel-to-air ratio. m_ a EY O2, air m_ a 3030
In a compartment fire, air is typically drawn
where
into the space through a door or window style
vent. If all of the air drawn into the compartment Q_ ¼ Ideal heat release rate of the fire ðkWÞ
is entrained into the lower layer portion of the m_ a ¼ Air flow rate ðkg=sÞ
plume, then the plume equivalence ratio can be E 13,100 kJ/kg (Drysdale [13])
an adequate representation of the fire environ- Note that Q is the ideal heat release rate,
ment. However, if layer burning occurs, or mul- which is determined by multiplying the mass
tiple vents cause air to enter the upper layer loss rate by the heat of combustion, and is not
directly, the use of a compartment equivalence limited by the amount of air flowing into the
ratio is more appropriate. As a practical note, for compartment or control volume. To date, Equa-
fires within a single compartment, the equiva- tion 16.13c has not been utilized in the literature
lence ratio is calculated (and experimentally and therefore has not been well established.
measured) based on the instantaneous fuel mass However, it offers a convenient means to calcu-
loss rate, ṁf, and air flow rate, ṁa, into the late the equivalence ratio without the need to
compartment (Equation 16.13a). know the fuel chemistry.
The equivalence ratio is an indicator of two
m_ f m_ a distinct burning regimes, overventilated (fuel
ϕ¼ ð16:13aÞ
r lean) and underventilated (fuel rich).
As noted previously, r is defined as the stoichio- Overventilated conditions are represented by
metric fuel-to-air ratio. Unfortunately, the ratio r is equivalence ratios less than one, while
sometimes defined as the air-to-fuel ratio, ra. underventilated conditions are represented by
Therefore, consideration must be given to values equivalence ratios greater than one. An equivalence
obtained from tabulated data. Keeping with the ratio of unity signifies stoichiometric burning,
nomenclature of this chapter, the equivalence which, in an ideal process, represents complete
ratio can also be expressed as combustion of the fuel to CO2 and H2O with no
excess oxygen. During underventilated conditions CH1.74O0.323 N0.07. Calculate the stoichiometric
there is insufficient oxygen to completely burn fuel-to-air ratio, the maximum yields of CO, CO2,
the fuel; therefore, products of combustion will and H2O, and the maximum depletion of O2.
also include excess fuel (hydrocarbons), carbon
monoxide, and hydrogen. It follows that the Solution For complete combustion of the fuel to
highest levels of CO production in flaming fires CO2 and H2O, the following chemical equation
is expected when underventilated conditions can be written
occur in the compartment on fire. This basic
chemistry also suggests that species production CH1:74 O0:323 N0:07 þ 1:2735ðO2 þ 3:76 N2 Þ
can be correlated with respect to the equivalence ! 1:0CO2 þ 0:87H2 O þ 4:823N2
ratio. Although the not-so-ideal behavior of
The molecular weight of the fuel, M f , ¼ 1ð12Þ
actual fires prevents accurate theoretical predic-
þ1:74ð1Þ þ 0:323ð16Þ þ 0:07ð14Þ ¼ 19:888.
tion of products of combustion, experimental
The stoichiometric fuel-to-air ratio is
correlations have been established.
A simple model for the most complete com-
ð1 mole fuelÞ M f 19:888
bustion of a fuel can be represented by the fol- r¼ ¼
ðmoles of airÞ ðMa Þ 1:2735 ð4:76Þ ð28:8Þ
lowing expressions: [8]
r ¼ 0:1139
f CO2 ¼ f O2 ¼ f H2 O ¼ 1 for ϕ < 1 The stoichiometric air-to-fuel ratio is
ð16:14aÞ
1
¼ 8:78
r
f CO2 ¼ f O2 ¼ f H2 O ¼ 1=ϕ for ϕ > 1
The maximum yield of CO (i.e., kCO), is
ð16:14bÞ
calculated by assuming that all carbon in the
fuel is converted to CO. Therefore, the number
f CO ¼ f H2 ¼ 0 for all ϕ ð16:14cÞ of moles of CO formed, nCO, equals the number
of moles of carbon in one mole of fuel. For the
f THC ¼ 0 for ϕ < 1 ð16:14dÞ polyurethane foam, nCO ¼ 1.
nCO ðMCO Þ ð1Þ ð28Þ

kCO ¼ ¼ ¼ 1:41
f THC ¼ 1 1=ϕ for ϕ < 1 ð16:14eÞ n f ðMfuel Þ ð1Þ ð19:888Þ
These expressions assume that for ϕ greater Similarly, kCO2 and kH2 O are calculated as
than 1, all excess fuel can be characterized as
ð1Þ ð44Þ
unburned hydrocarbons. Since compartment fire kCO2 ¼ ¼ 2:21
experiments have shown that significant levels 19:888
ð0:87Þ ð18Þ
of both CO and H2 are produced at higher equiv- k H2 O ¼ ¼ 0:787
19:888
alence ratios, Expression 16.14c is not always
representative, and reveals a shortcoming of The maximum depletion of oxygen, kO2 , refers
assuming this simple ideal behavior. However, to the mass of oxygen needed to completely
for the products of complete combustion (CO2, combust one mole of fuel to CO2 and
O2, and H2O), this model serves as a good bench- H2O. This is the same as the stoichiometric
mark for comparison of experimental results. requirement of oxygen.
nO2 ðMO2 Þ ð1:2735Þ ð32Þ
Example 1 Consider a piece of cushioned kO2 ¼ ¼ 2:05
furniture to be primarily polyurethane foam. The nfMf ð1Þ19:888
nominal chemical formula of the foam is
www.ebook777.com

Example 2 XCOwet m_ f þ ma MCO
The fuel specified in Example 1 is burning at a Y CO ¼
m_ f Ma
rate of 9 g/s and entraining air at a rate of 56 g/s.
ð0:033Þ ð9 þ 56Þ ð28Þ
Measurements of the upper layer gas composi- ¼ ¼ 0:23
tion reveal dry concentrations of 3.7 % by vol- 9ð28:8Þ
ume CO, 14.3 % CO2, and 0.49 % O2. Correct the Y CO 0:23
f CO ¼ ¼ ¼ 0:16
concentrations for the water removed during the kCO 1:41
gas analysis process (i.e., calculate the wet ð0:127Þ ð9 þ 56Þ ð44Þ
Y CO2 ¼ ¼ 1:40
concentrations). 9ð28:8Þ
1:40
Solution In order to use Equation 16.9 to calcu- f CO2 ¼ ¼ 0:63
2:21
late the wet mole fractions, the stoichiometric ð0:11Þ ð9 þ 56Þ ð18Þ
molar ratio of H2O to CO2 for C needs to be Y H2 O ¼ ¼ 0:50
9ð28:8Þ
calculated. This ratio is simply obtained
0:50
from the stoichiometric chemical equation in f H2 O ¼ ¼ 0:63
Example 1. 0:787
The depletion of oxygen is calculated using
nH2 O 0:87
C¼ ¼ ¼ 0:87 Equation 16.6, assuming the molecular weight
nCO2 1 of the gas mixture, Mmix, to be approximately
that of air.
Once C is obtained, the wet mole fractions can

be calculated as 0:21m_ a MO2 =Ma XO2wet m_ f þ m_ a MO2 =Mmix
DO2 ¼
m_ f
0:037
XCOwet ¼ ¼ 0:033 0:21ð56Þ32=28:8 0:0044ð9 þ 56Þ32=28:8
1 þ 0:87ð0:143Þ DO2 ¼
9
0:143 DO2 ¼ 1:42
XCO2wet ¼ ¼ 0:127
1 þ 0:87ð0:143Þ
The normalized yield is calculated as
0:0049
XO2wet ¼ ¼ 0:0044
1 þ 0:87ð0:143Þ DO2 1:42
f O2 ¼ ¼ ¼ 0:69
k O2 2:05
The estimated mole fraction of water is
XH2 O ¼ CXCO2wet ¼ 0:87ð0:127Þ ¼ 0:11

Species Production Within Fire
Compartments
Therefore, the corrected gas concentrations on
a percent volume basis are 3.3 % CO, 12.7 %
Hood Experiments
CO2, and 0.44 % O2.
Example 3 Continuing from Example 2, calculate Beyler [8, 9] was the first to publish major
the yields and normalized yields for each species species production rates in a small-scale two-
measured. The wet mole fractions are layer environment. The experiments perfor-
XCOwet ¼ 0:033, XCO2wet ¼ 0:127, and XO2wet ¼ med consisted of situating a burner under a
0:0044. 1-m-diameter, insulated collection hood. The
result was the formation of a layer of combustion
Solution Using Equations 16.5 and 16.7, the products in the hood similar to that found in a
yield and normalized yield of CO, CO2, and two-layer compartment fire (see Fig. 16.4). By
H2O can be calculated. The maximum yields varying the fuel supply rates and the distance
calculated in Example 1 are kCO ¼ 1:41, between the burner and layer interface, and, con-
kCO2 ¼ 2:21, kH2 O ¼ 0:787, and kO2 ¼ 2:05. sequently, the air entrainment rate, a range of
equivalence ratios was obtained. Layer gases For comparison, the expressions for ideal
were exhausted at a constant, metered flow rate complete combustion (Equations 16.14a, 16.14b,
from the periphery of the hood at a depth of 16.14c, 16.14d, and 16.14e) are shown on each
15 cm below the ceiling. The general procedure plot in Fig. 16.5. The CO2 yield departs from
was to allow steady-state burning conditions to Equation 16.14b as CO production increases at
develop, so the layer maintained a constant depth higher equivalence ratios. This departure, which
below the exhaust flow location. Tests revealed a is fairly independent of ϕ for ϕ > 1, has been
reasonably well-mixed uniform layer both in described by the yield coefficient [5]. The ratios
temperature and chemical composition during of the normalized yield of CO2, H2O, or normalized
the steady-state conditions. Gas analysis was depletion of O2 to the theoretical maximums
performed on samples taken from the exhaust expressed by Equations 16.14a, 16.14b, 16.14c,
stream. Table 16.1 shows the physicochemical 16.14d, and 16.14e are defined as the yield
properties of the fuels tested. coefficients, BCO2 , BH2 O , and BO2 respectively [5].
Beyler’s results show that species yields cor-
relate very well with the plume equivalence ratio. f CO2
BCO2 ¼ for ϕ < 1 ð16:15aÞ
Figure 16.5 shows normalized yields of major 1
species for propane fires plotted against the
f CO2
plume equivalence ratio. The trends seen in BCO2 ¼ for ϕ > 1 ð16:15bÞ
these plots for propane are fairly representative 1=ϕ
of the other fuels tested. For overventilated f H2 O
conditions, the yield of CO2 and H2O and deple- BH 2 O ¼ for ϕ < 1 ð16:16aÞ
1
tion of O2 are at a maximum, and there is virtu-
ally no production of CO, H2, or unburned f H2 O
BH 2 O ¼ for ϕ > 1 ð16:16bÞ
hydrocarbons (THC). As underventilated burning 1=ϕ
conditions ðϕ 1Þ are approached, products of
f O2
incomplete combustion (CO, H2, and THC) are BO2 ¼ for ϕ < 1 ð16:17aÞ
generated. 1
Table 16.1 Physicochemical data for selected fuels
Empirical chemical Empirical Maximum theoretical yields

Fuel formula of volatiles molecular weight kCO kCO2 kO2 kH2 O 1/rc
Acetone C3H6O 58 1.45 2.28 2.21 0.93 9.45
Ethanol C2H5OH 46 1.22 1.91 2.09 1.17 8.94
Hexane C6H14 86 1.95 3.07 3.53 1.47 15.1
Isopropanol C3H7OH 60 1.40 2.20 2.40 1.20 10.3
Methane CH4 16 1.75 2.75 4.00 2.25 17.2
Methanol CH3OH 32 0.88 1.38 1.50 1.13 6.43
Propane C3H8 44 1.91 3.00 3.64 1.64 15.6
Propene C3H6 42 2.00 3.14 3.43 1.29 14.7
Polyurethane foam CH1.74O0.323 N0.0698 20 1.41 2.21 2.05 0.79 8.78
Polymethylmethacrylate C5H8O2 100 1.40 2.20 1.92 0.72 8.23
Toluene C7H8 92 2.13 3.35 3.13 0.78 13.4
Wood (ponderosa pinea) C0.95H2.4O 30 0.89 1.40 1.13 0.73 4.83
Wood (spruceb) CH3.584O1.55 40 0.69 1.09 0.89 0.80 3.87
a
From Beyler [9] chemical formula estimated from ϕ < 1 yield data
b
Gottuk et al. [5]
c
r ¼ stoichiometric fuel-to-air ratio
www.ebook777.com
16
a b c
0.38
0.34 XX
XX X XX X X
XXXXX X
1.0
XXXX X
1.0 X
XXX
XXX
X
XX XX X 0.30 XXX XXX XX
X
X X XX X
X XXX
0.26 XX
XX
XXXX
XX 0.22 X
XXX
XX
XXX
0.6 X 0.6
XXXXX
X 0.18 X
XXX
O2 Yield
CO Yield
CO2 Yield
X 0.14
X X
X X
XX X
X
X X
XXX
0.10 XXX X X
0.2 0.06 X 0.2
X
X
0.02 X XX
X
X
XXXX X
XXXXXX
0.2 0.6 1.0 1.4 1.8 0.2 0.6 1.0 1.4 1.8 0.2 0.6 1.0 1.4 1.8
Equivalence ratio Equivalence ratio Equivalence ratio
d e f
X
X XXX X
X
X XXXX X XXX X
XX XXX 1.8 0.44 X
1.0
XX
XXXXXXX
X X
X X XXX
XXX 0.36 X
XX
1.4 XX
X
Effect of Combustion Conditions on Species Production
XXXXXX XX
X
X
XX X
0.28 X X
0.6 X 1.0 X
X
H2 Yield
0.20
H2O Yield
THC Yield
XX
0.6 X X
XX
XXXXX X 0.12 X X
X X
0.2 XX X
XX
0.2 X 0.04 X
XXX X X
XXX
X
XX
XXX XX X X X
X XX XX
0.2 0.6 1.0 1.4 1.8 0.2 0.6 1.0 1.4 1.8 0.2 0.6 1.0 1.4 1.8
Equivalence ratio Equivalence ratio Equivalence ratio
Fig. 16.5 Normalized yields of measured chemical species as a function of the equivalence ratio for propane experiments using a 13 cm (o) or 19 cm (x) burner with supply
rates corresponding to 8 to 32 kW theoretical heat release rate [8]
495
f O2 yields from toluene fires remain fairly constant

BO 2 ¼ for ϕ > 1 ð16:17bÞ at about 0.09 for both overventilated and
1=ϕ
underventilated burning conditions.
These terms are useful in discussing It should be noted that Beyler originally
characteristics of the combustion efficiency. For presented all correlations with normalized yields,
example, an O2 yield of 1 indicates complete fCO. However, better agreement is found between
utilization of available O2. In the case of CO2 unnormalized CO yield-equivalence ratio
and H2O, deviation from the model (as indicated correlations for different fuels, YCO (shown in
by BCO2 or BH2 O < 1) is a measure of the degree Fig. 16.6), rather than normalized yields. One
of incomplete combustion. It can be seen from point of interest, though, is that when CO
Fig. 16.5 that the production of CO is primarily at production is correlated as normalized yield, a
the expense of CO2 (i.e., BO2 and BH2 O remain more distinct separation of the data occurs for
nearly 1, while BCO2 is about 0.8). Table 16.2 ϕp greater than 1. The degree of carbon monox-
shows average yield coefficients for ide production (represented as fCO) during
underventilated fires. underventilated conditions can be ranked by
Figure 16.6 shows unnormalized CO yields chemical structure according to oxygenated
plotted against the plume equivalence ratio for hydrocarbons greater than hydrocarbons greater
fuels tested by Beyler [8, 9]. The correlations than aromatics. This ranking is not observed for
agree quite well for all fuels. Below an equiva- unnormalized yield correlations.
lence ratio of 0.6, minimal CO production is Toner et al. [7] and Zukoski et al. [10, 11]
observed. Above ϕp equal to 0.6, carbon monox- performed similar hood experiments with a differ-
ide yield increases with ϕp and, for most fuels, ent experimental setup. The hood used was a 1.2 m
tends to level out at equivalence ratios greater cube, insulated on the inside with ceramic fiber
than 1.2. Toluene, which creates large amounts insulation board. The layer in the hood formed to
of soot, is anomalous compared to the other fuels the lower edges where layer gases were allowed to
studied. As can be seen in Fig. 16.6, the CO spill out. Gas sampling was done using an
Table 16.2 Average yield coefficients and upper-layer temperatures for underventilated firesa (values in parentheses
are standard deviations)
Fuel BCO2 BO2 BH2 O Temperature (K) Reference
Acetone 0.78 (0.03) 0.92 (0.04) 0.99 (0.04) 529 (76) Beyler [8]
Ethanol 0.79 (0.01) 0.97 (0.01) 1.00 (0.04) 523 (72) Beyler [8]
Hexane 0.61 (0.10) 0.82 (0.02) 0.87 (0.03) 529 (25) Beyler [8]
Hexane 0.83 (0.05) 0.96 (0.06) NA 1038 (62) Gottuk et al. [5]
Isopropanol 0.75 (0.01) 0.89 (0.01) 0.96 (0.01) 513 (33) Beyler [8]
Methane 0.80 (0.05) 1.00 (0.04) 1.01 (0.03) 713 (101) Toner et al. [7]
Methane 0.69 (0.03) 0.87 (0.07) 0.86 (0.06) 547 (12) Morehart et al. [12]
Methanol 0.79 (0.03) 0.99 (0.00) 0.94 (0.02) 566 (53) Beyler [8]
Propane 0.78 (0.05) 0.97 (0.03) 1.05 (0.04) 557 (62) Beyler [8]
Propene 0.77 (0.08) 0.92 (0.08) 1.02 (0.10) 629 (51) Beyler [8]
Polyurethane foam 0.87 (0.04) 0.97 (0.02) NA 910 (122) Gottuk et al. [5]
Polymethylmethacrylate 0.77 (0.06) 0.92 (0.19) 0.72 (0.04) 525 (37) Beyler [9]
Polymethylmethacrylate 0.93 (0.04) 0.98 (0.04) NA 1165 (126) Gottuk et al. [5]
Toluene 0.57 (0.04) 0.62 (0.05) 0.78 (0.03) 509 (23) Beyler [8]
Wood (ponderosa pine) 0.85 (0.05) 0.89 (0.03) 0.79 (0.10) 537 (37) Beyler [9]
Wood (spruce) 0.90 (0.00) 0.95 (0.00) NA 890 (0) Gottuk et al. [5]
a
Values have been calculated from data found in the cited references. Values for Toner et al. [7], Beyler [8], Beyler [9],
and Morehart et al. [12] are from hood experiments, and values for Gottuk et al. [5] are for a reduced-scale enclosure
www.ebook777.com
Fig. 16.6 Unnormalized 0.4

carbon monoxide yields as
Notes:
a function of the plume
equivalence ratio for = Methanol
= Ethanol
Unnormalized CO yield
various fuels studied by 0.3 = Isopropanol
Beyler in a hood apparatus X = Propane
X
[8, 9] X X XX
X
= Acetone X X X
X X XX
X
X XXXX
XX XXXXXXX
XX X XXXX
0.2 X X XX XX
X
X X
X X XX X
XXX X
X
X
XXX
X
X
0.1 XX
X
XX
X
XXX
X
XX
X
X
X X
X
XX
X
X X XXXX X
X X XXX X
XX X
X XX X
XXX
0 XXX X
0 0.5 1.0 1.5
0.4
Notes:
X = Prop ane
= Propene
0.3 = Hexane
+ = Toluene X
X X XX
X
• = PMMA X X X
X X XX
X XX
= Equation 14 XX
X
X
X XXX
X XXX
XX
0.2 X X
•
X X XX X• X
•• X • X
X
XXX X
• •
X
XXX
•• X
0.1 XX •
+ X+
X
++ XX + +
X
+ + +
• XXX
X
X
X
XXX
XX
X• X
XX X
XX X XXX XX
X
0 X X • XX ••XXXX• X
0 0.5 1.0 1.5

Plume equivalence ratio
uncooled stainless-steel probe inserted into the conditions, with mean upper-layer residence
layer. Detailed gas species measurements were times of about 25–180 s, it can be concluded
made using a gas chromatograph system. The that ϕp and ϕul were equal.
upper-layer equivalence ratio was determined The data of Toner et al. [7] have been used to
from conservation of atoms using the chemical plot CO and CH4 yields versus upper-layer
species measurements, the measured composition equivalence ratio in Figs. 16.7 and 16.8, respec-
of the fuel, and the metered fuel flow rate. Natural tively. The correlations are qualitatively similar
gas flames with heat release rates of 20–200 kW on to the correlations obtained by Beyler for differ-
a 19-cm-diameter burner were studied. The layer ent fuels. An analysis of these test results also
in the hood was allowed to form and reach a showed that normalized CO2 and O2 yield versus
steady-state condition before gas sampling was equivalence ratio data is represented reasonably
performed. well by Equations 16.15, 16.16, and 16.17. Simi-
It was concluded that species concentrations lar to Beyler’s propane results, the average BO2
were well correlated to the upper-layer equiva- value is about 1 and BCO2 is 0.8 for
lence ratio, ϕul, and insensitive to temperatures underventilated burning conditions (the use of
for the range studied (490–870 K). Since these yield coefficients is discussed further in the sec-
experiments were conducted during steady-state tion on engineering correlations).

carbon monoxide yields as
Notes:
a function of equivalence
= Toner
ratio for methane fires + = Morehart
0.3
studied by Toner et al. [7]
and Morehart et al. [12] in
hood experiments
Equation 22
0.2
Equation 21
0.1
0
0 0.5 1.0 1.5 2.0 2.5 3.0
Equivalence ratio
Fig. 16.8 Normalized 1.0

yields as a function of
Notes:
equivalence ratio for + = Toner
methane fires studied by 0.8 = Morehart
Toner et al. [7] and
Normalized CH4 yield
Morehart et al. [12] in hood 1 – 1/φ

experiments 0.6
0.4 +
+
+ + +
++
+ +
0.2 + +++
+
+
+ +++
0 ++ ++ +
++ + + ++
0 0.5 1.0 1.5 2.0 2.5 3.0
Equivalence ratio
Toner compared the measured species (490–870 K), Toner et al. concluded that, if a
concentrations to the calculated equilibrium freeze-out temperature existed, it must be higher
composition of the reactants at constant temper- than 900 K. Work by Pitts [14] and by Gottuk
ature and pressure. The layer composition was et al. [15], discussed later, shows that a freeze-
modeled quite well by the chemical equilibrium out temperature does exist in the range of
composition for very overventilated conditions 800–900 K, depending on other factors.
but not for underventilated conditions. His obser- Zukoski, Morehart, et al. [11] performed a sec-
vance of CO production for near-stoichiometric ond series of tests similar to that described above
and underventilated fires, at the expense of CO2 for Zukoski et al. [10] and Toner et al. Much of the
production, led them to suggest that the oxidation same apparatus was used except for a different
of CO was “frozen out” before completion. collection hood. The hood, 1.8 m square by 1.2 m
(At low temperatures, there is insufficient energy high, was larger than that used by Toner et al. and
for CO to chemically react to CO2.) [7] Since the was uninsulated.
results showed that species production was inde- Morehart et al. [12] experiments consisted of
pendent of temperature for the range studied establishing steady-state burning conditions such
www.ebook777.com
that the burner-to-layer interface height was was reduced by 25 % [11, 12]. Excluding one
constant. A constant ϕp was maintained based outlier data point, CO concentrations increased
on this constant interface height in conjunction by 25 %. This is an important result. Although
with the fact that the mass burning rate of fuel the gas temperatures were well below 800 K, an
was metered at a constant rate. Additional air was increase in the layer temperature resulted in more
then injected into the upper layer at a known flow fuel being combusted to products of complete
rate until a new steady-state condition was combustion and additional CO (see the section
achieved. This procedure established a ϕul that “Chemical Kinetics” later in this chapter).
was lower than the ϕp, since ϕp was based on the
ratio of the mass burning rate to the mass of air
entrained into the plume from room air below the Compartment Fire Experiments
layer interface. By increasing the air supply rate
to the upper layer, a range of ϕul was established The hood experiments performed by Beyler and
while maintaining a constant ϕp. Zukoski et al. differ from actual compartment
Although similar, the correlations obtained by fires in several ways. The hood setup allowed
Morehart et al. deviated from those obtained by considerable radiation to the lab space below.
Toner et al. Figs. 16.7 and 16.8 compare the CO Conversely, a real compartment would contain
and CH4 yields calculated from the data of most of the radiation, thus resulting in higher
Morehart et al. with the yields calculated from wall and upper-layer temperatures. Conse-
the data of Toner et al. For overventilated quently, higher fuel mass loss rates for pool fires
conditions, Morehart et al. observed higher CO would be expected for an actual compartment
and CH4 yields, signifying that the fires fire. Also, the hood setup results in a lower layer
conducted by Morehart et al. burned less that has an infinite supply of air which is neither
completely. For underventilated methane fires, vitiated nor heated. In a real compartment fire, the
Morehart et al. observed lower CO, CO2, and air supply is limited by ventilation openings
H2O and higher CH4 and O2 concentrations (doors, windows, etc.) and the depth of the
than Toner et al. The only apparent differences upper layer. The air that is entrained into the
between experiments was that Morehart found lower layer of an actual compartment fire can be
layer temperatures were 120–200 K lower for convectively heated by hot compartment surfaces
fires with the same equivalence ratio as those prior to fire plume entrainment. The hood
observed by Toner, that is, they ranged from experiments did not include any significant ceil-
488 to 675 K. Due to the similarity in experimen- ing and wall flame jets. These dynamic flame
tal apparatus, except for the hood, Morehart structures enhance mixing of the upper layer in
concluded that the temperature difference actual compartment fires and extend the flame
resulted from having a larger uninsulated hood. zone beyond the plume. Lastly, the hood experi-
Morehart studied the effect of increasing tem- ment correlations were developed from sustained
perature on layer composition by adding differ- steady-state burning conditions. Actual fires of
ent levels of insulation to the hood. Except for interest are usually in a continual growth stage,
the insulation, the test conditions (e.g., ϕ of 1.45 and, thus, are more transient in nature.
and layer interface height) were held constant. Tewarson reported that CO and CO2 yields
Residence times of layer gases in the hood were and O2 depletion were correlated well by the
in the range of 200–300 s. For the range of air-to-fuel stoichiometric fraction (i.e., the recip-
temperatures studied (500–675 K), substantial rocal of the equivalence ratio) for wood crib
increases in products of complete combustion enclosure fires [16]. Enclosure fire data was
(i.e., CO2 and H2O) and decreases in fuel and taken from previous work in the literature for
oxygen occurred with increasing layer tempera- cellulosic-base fiberboard and pine wood cribs
ture. Upper-layer oxygen mass fraction was burned in various compartment geometries,
reduced by approximately 70 % and methane 0.21–21.8 m3 in volume, with single and dual
horizontal and vertical openings centered on the exhaust vent. This location for the probe was
end walls. Additional data were obtained for pine chosen after species concentration and tempera-
wood cribs burned in a small-scale flammability ture measurements, taken at several locations in
apparatus that exposed the samples to variable the upper layer, showed a well-mixed, uniform
external radiant heat fluxes with either natural or layer.
forced airflow from below. Table 16.1 shows the physicochemical
The characteristics of the correlations properties used for the four fuels. It should be
presented by Tewarson are similar to the noted that in determining properties of a fuel,
correlations developed by Beyler. The CO2 such as maximum yields or the stoichiometric
yield and O2 depletion are relatively constant fuel-to-air ratio, the chemical formula must char-
for low equivalence ratios and decrease sharply acterize the volatiles, not necessarily the base
as the equivalence ratio increases for fuel. For liquid fuels or simple polymers, such
underventilated conditions. The CO yield as PMMA, the composition of the volatiles is the
correlates with the equivalence ratio but with a same as the base fuel. However, more complex
fair amount of scatter in the data. fuels can char or contain nonvolatile fillers, as
Due to the lack of measurements, the air found in polyurethane foams. As a result, the
entrainment rate used to calculate the mass air- composition of the volatiles differs from that of
to-fuel ratio was estimated from the ventilation the base material. As an example, the composi-
parameter, Ah1/2, where A is the cross-sectional tion of the wood volatiles used in this study was
area and h is the height of the vent. Although the obtained by adjusting the analyzed wood compo-
general shape of the correlations are valid, the sition for an observed average of 25 % char [5].
use of the ventilation parameter assumption The results of these compartment tests
causes the equivalence ratio data to be suspect. showed similarities to Beyler’s hood
In addition, the elemental composition of the fuel experiments. However, some significant quanti-
volatiles for the wood was not corrected for char tative differences exist. Figure 16.9 compares the
yield. A correction of this sort would tend to CO yield correlations from Beyler’s hood study
decrease the calculated equivalence ratio and and that of these compartment tests for hexane
increase the CO and CO2 yields. fires. This plot illustrates the primary difference
Gottuk et al. [5, 17] conducted reduced-scale observed between the hood and compartment
compartment fire tests specifically designed to hexane and PMMA fire test results. An offset
determine the yield-equivalence ratio exists between the rise in CO yield for the two
correlations that exist for various fuels burning studies. For the hood experiments, higher CO
in a compartment fire. A 2.2 m3 (1.2 m 1.5 m production was observed for overventilated (ϕp
1.2 m high) test compartment was used to < 1) and slightly underventilated burning
investigate the burning of hexane, PMMA, conditions. For the compartment fire
spruce, and flexible polyurethane foam. The test experiments, negligible CO was produced until
compartment was specially designed with a underventilated conditions were reached. Con-
two-ventilation path system that allowed the sistent with the increased CO production and
direct measurement of the air entrainment rate the conservation of carbon, CO2 yields were
and the fuel volatilization rate. The setup created lower for the hood experiments compared to the
a two-layer system by establishing a buoyancy- compartment fires. The spruce and polyurethane
driven flow of air from inlet vents along the floor, compartment fires produced similar CO yield-
up through the plume, and exhausting through a equivalence ratio correlations to those observed
window-style exhaust vent in the upper layer. by Beyler in hood experiments (i.e., high CO
There was no inflow of air through the exhaust yields were observed for overventilated fires).
vent. The upper-layer gas mixture was sampled The differences in CO formation can be
using an uncooled stainless steel probe placed explained in terms of temperature effects. For
into the compartment through the center of the the region of discrepancy between equivalence
www.ebook777.com
Fig. 16.9 Comparison of 0.4

unnormalized CO yield
Notes:
correlations for hexane
= Gottuk et al.
fires in a compartment and + = Beyler
0.3 Equation 22
under a hood apparatus.
(Figure taken from Gottuk
et al. [5])
0.2 +
++ +++
+ + + +
+
+
Equation 21
0.1 +
+
+
0
0 0.5 1.0 1.5 2.0 2.5 3.0
Equivalence ratio
ratios of 0.5 and 1.5, upper-layer temperatures in The National Institute of Standards and Tech-
Beyler’s hood experiments and the spruce and nology (NIST), Building and Fire Research Lab-
polyurethane compartment fire experiments were oratory, has also performed reduced-scale
typically below 850 K, whereas temperatures for compartment fire experiments using a natural
the hexane and PMMA fires were above 920 K gas burner for the heat source [6]. The compart-
(temperatures typically associated with ment (0.98 m 1.46 m 0.98 m high) had a
postflashover fires) [18]. single ventilation opening consisting of a 0.48 m
As is detailed in the section on chemical kinet- wide by 0.81 m high doorway. A large number of
ics, the temperature range between 800 and tests were conducted covering a range of heat
900 K is a transition range over which the oxida- release rates from 7 to 650 kW. Fires greater
tion of CO to CO2 changes from a very slow to a than 150 kW resulted in upper-layer
fast reaction. That is, for upper-layer temperatures greater than 870 K and flames
temperatures below 800 K, the conversion of 0.5–1.5 m out of the door. This single ventilation
CO to CO2 does not occur at an appreciable opening and the large fires (up to 650 kW) pro-
rate to affect CO yields. Since the oxidation of duced nonuniform upper-layer conditions. For
a fuel first results in the production of CO, which fires with heat release rates greater than about
then further reacts to form CO2, the low 250 kW (ϕ > 1.5), carbon monoxide concen-
temperatures (<800 K) prevent CO from trations in the front of the compartment were
oxidizing. This results in high CO yields. For approximately 30–60 % higher than in the rear.
temperatures greater than 900 K, the reactions Temperature gradients of 200–300 C were
that convert CO to CO2 occur faster as tempera- observed from the back to the front of the com-
ture increases. Therefore, for the overventilated partment. Due to the nonuniform air entrainment
conditions discussed above, the high at the base of the fire and possible mixing of
temperatures associated with the hexane and additional air near the front, it is difficult to
PMMA compartment fires resulted in virtually determine the local equivalence ratio for each
all CO being oxidized to CO2 for ϕp < 1. region. The concentration gradient from front to
Overall, the compartment fire test results rear of the compartment may have been due to
revealed that the production of CO is primarily differences in the local equivalence ratios. None-
dependent on the compartment flow dynamics theless, plots of concentration measurements in
(i.e., the equivalence ratio) and the upper-layer the rear of the compartment versus equivalence
temperature. ratio are quite similar to the data of Zukoski
et al. and Toner et al. Yield data for these results the upper layer of a reduced-scale enclosure fire.
have not been reported. The enclosure was the same as used by Gottuk
A second set of experiments was performed et al. [5], measuring 1.5 m wide, 1.2 m high, and
by NIST to investigate the generation of CO in 1.2 m deep. Two primary sets of tests were
wood-lined compartments [19]. Douglas fir ply- conducted for cases with and without Douglas fir
wood (6.4 mm thick) was lined on the ceiling and plywood suspended below the ceiling, with (1) a
on the top 36 cm of the walls of the compartment 0.12 m2 window vent opening and (2) a 0.375 m2
described above. Natural gas fires ranging from doorway opening, both leading to a hallway.
40 to 600 kW were burned in the compartment. In the compartment with a window opening
The results showed that, for tests in which wood and wood burning in the upper layer, Lattimer
pyrolysis occurred, increased levels of CO were et al. measured CO concentrations of 10 % on
observed compared to burning the natural gas average, which is nearly three times greater than
alone. Carbon monoxide concentrations (dry) the levels measured without the wood. Peak
reached levels of 7 % in the front and 14 % in concentrations were as high as 14 %, the same
the rear of the compartment. These are extremely as measured by Pitts et al. [19] CO
high concentrations compared to the peak levels concentrations were similarly high when the
of 2–4 % observed in the unlined compartment doorway opening was used. In this case, the
fire tests with the methane burner only. Typical quasi-steady state average CO concentrations
peak CO concentrations observed for a range of were 8 % with peaks greater than 10.6 % with
fuels (including wood) in hood experiments wood compared to approximately 5.7 % average
[8–11] and the compartment fire experiments of levels with a doorway vent and no wood. Regard-
Gottuk et al. [5] also ranged from 2 % to 4 %. less of the vent opening, these tests showed that
However, concentrations greater than 5 % have wood in the upper layer resulted in CO
also been reported for cellulosic fuels burning in concentrations increasing dramatically (10.1 %,
enclosures [16, 20]. vs. 3.2 % without wood) with only small
Since wood is an oxygenated fuel, it does not increases in the CO2 concentrations (11.6 %,
require additional oxygen from entrained air to vs. 10.4 % without wood). These trends are
form CO. This enhances the ability of the wood summarized in Table 16.3, which presents the
to generate CO in a vitiated atmosphere. There-
fore, there are two reasons that high CO Table 16.3 Summary of quasi-steady state average spe-
concentrations can result in fires with oxygenated cies concentrations (percent volume dry) for
fuels in the upper layer. First, the fuel-bound underventilated reduced-scale compartment fire tests
with and without wood in the upper layer [21, 22]
oxygen allows the fuel to generate CO during
pyrolysis. Second, due to preferential oxidation Window vent No wood in Wood in
tests [21] upper layer upper layer
of hydrocarbons over CO, the limited oxygen in
CO 3.2 10.1
the upper layer reacts with the pyrolyzing wood
CO2 10.3 11.6
to form additional CO. Aspects of this chemistry
O2 0.2 0.04
are discussed in the next section. Doorway vent test [21]
These initial test results for fires with wood on CO 5.7 8.0
the walls and ceiling emphasized the importance CO2 8.7 9.6
of adding additional fuel to the upper layer. The O2 0.2 0.1
practical implications are significant, as many NIST results [22]
structures have cellulosic-based wall coverings Doorway vent
and other combustible interior finishes. Because CO 2.6/1.8a 5.5/11.5a
of the initial studies by NIST, Lattimer et al. [21] CO2 6.5/7.5a 10/15.5a
conducted a series of tests to evaluate the effect on O2 0.1/0a 0/0.5a
species production from the addition of wood in a
Front and back, respectively
www.ebook777.com
average upper-layer species concentrations for included in these plots are the data from the com-
tests with and without wood for both window partment fires of Gottuk et al. [5] The results show
and doorway vent conditions. For comparison, that the global equivalence ratio concept is capable
the data from the NIST research has also been of predicting the CO2, O2, and CO yields, although
included. somewhat fortuitously, in a compartment with
The compartment equivalence ratio was calcu- wood pyrolyzing in the hot, vitiated upper layer.
lated for both the tests with and without wood in These tests also indicate that the correlations hold
the upper layer when the window vent was used. to fairly high equivalence ratios of about 5.5, as
Figures 16.10, 16.11, and 16.12 show the observed for the tests with wood. More work is
corresponding calculated yields of CO2, O2, and needed to determine whether the global equiva-
CO plotted as a function of equivalence ratio. Also lence ratio concept can predict species levels when
Fig. 16.10 The 1.2

normalized CO2 yield data
1.1
of Gottuk et al. [5] (○),
data from Lattimer 1.0
[(kg of CO2/kg of fuel)/YCO2, max]
et al. [21] with no wood in 0.9

the compartment upper
Normalized CO2 yield
layer (●), and data from 0.8

Lattimer et al. [21] with 0.7
wood in the upper layer
(~). Also shown in this plot 0.6
is the normalized CO2 yield 0.5
estimated using the
0.4
complete combustion
model of Equation 16.14 0.3
(—)
0.2
0.1
0.0
0 1 2 3 4 5 6
Compartment equivalence ratio, φc
Fig. 16.11 The 1.2

normalized O2 yield data of
1.1
Gottuk et al. [5] (○), data
from Lattimer et al. [21] 1.0
with no wood in the
[(kg of O2/kg of fuel)/YO2, max]
0.9
compartment upper layer
Normalized O2 yield
(●), and data from 0.8

Lattimer et al. [21] with 0.7
wood in the upper layer
(~). Also shown in this plot 0.6
is the normalized O2 yield 0.5
estimated using the
0.4
complete combustion
model of Equation 16.14 0.3
(—) 0.2
0.1
0.0
0 1 2 3 4 5 6
Fig. 16.12 The 0.30

unnormalized CO yield
data of Gottuk et al. [5]
(○), data from Lattimer 0.25
CO yield (kg of CO/kg of fuel)

et al. [21] with no wood in
the compartment upper
layer (●), and data from 0.20
Lattimer et al. [21] with
wood in the upper layer (~)
0.15
0.10
0.05
0.00
0 1 2 3 4 5 6
nonoxygenated fuels are in the upper layer. It is enclosure (as defined by ISO 9705) to compare
also unclear whether other oxygenated fuels will the results from the NIST reduced-scale enclo-
follow the correlations as well as the available sure tests to fires conducted in a full-scale enclo-
wood fire data. sure [23–26]. The enclosure measured 2.44 m
The data in Table 16.3 should provide an wide, 3.67 m deep, and 2.44 m tall, with a door
assessment of the effect of the ventilation open- (0.76 m by 2 m) centered at one end of the
ing on species generation. However, it is uncer- compartment. The fires consisted of a 35-cm-
tain whether the differences are due more to diameter natural gas burner centered in the enclo-
differences in sampling locations relative to the sure. The burner was scaled to provide the same
flame regions. In the tests with a doorway vent, exit gas velocities as in the reduced-scale enclo-
the larger opening resulted in larger air flow rates sure tests. Twelve tests were conducted, with
and, thus, larger fires in the compartment fires ranging in size from 0.5 to 3.4 MW. In one
(approximately 500 kW vs. 220 kW with the test, the ceiling and upper portions of the walls
window vent). The larger fires increased the were lined with 12.7 mm thick plywood.
flame zone within the compartment. Conse- In the full-scale enclosure, fires greater than
quently, the sampling probe was probably within 1250 kW created underventilated conditions. The
the flame zone at times, which would yield NIST researchers concluded that although the
higher CO and lower CO2 concentrations than reduced-scale and full-scale enclosures were
measurements from the window vent tests in geometrically similar, with good agreement
which the sampling probe was not sampling between predicted mass flows, the differences
from a flame zone. With the window vent, the in measured gas concentrations indicated that
fires were small enough such that there were no the generation of combustion products is not
ceiling jets at the level of the sampling probe. entirely controlled by the ventilation within the
The research discussed thus far has compartment. CO concentrations (upwards of
concentrated on reduced-scale enclosures. Lim- 6 % by volume) were as much as two times
ited full-scale studies have been reported in the higher in the full-scale enclosure than in the
literature to date. One study by NIST systemati- reduced-scale tests. These results also coincided
cally examined the production of species in light with higher upper-layer temperatures,
of the global equivalence ratio concept. NIST approaching 1400–1500 K. The variation in CO
conducted a set of tests using a standard concentrations from front to back in the
www.ebook777.com
enclosure was reversed in the full-scale enclosure different hood experiments and between hood
compared to the reduced-scale enclosure. In the and compartment fire experiments have indicated
full-scale enclosure, higher CO concentrations that upper-layer temperatures have an effect on
were observed in the back of the compartment. CO production. The results of these chemical
In the reduced-scale enclosure, higher concen- kinetics studies provide insights into CO genera-
trations were measured in the front. Pitts primar- tion in compartment fires, which also serve to
ily associates the higher CO concentrations with explain the differences in CO yields between
the high layer temperatures that are in the range experiments with respect to temperature effects.
that strongly favor the formation of CO toward These studies primarily focused on the question
equilibrium concentrations (values can approach “What would the resulting composition be if the
16 % at ϕ of 3) [26]. upper-layer gases in the hood experiments
One full-scale enclosure test was conducted existed at different isothermal conditions (con-
with wood in the upper layer. This test resulted in stant temperature)?” A particular focus was to
high CO concentrations of 8 % in the front and examine the resulting compositions for cases
12 % in the rear for a 2 MW fire. The modeled under the high temperatures character-
temperatures were lower than those observed in istic of compartment fires. Chemical kinetics
the full-scale tests without wood. These results models calculate the change in species
are similar to those observed in the NIST concentrations with respect to time. Calculations
reduced-scale enclosure. are dependent on the reaction mechanism (i.e.,
the set of elementary reactions and associated
kinetic data) and the thermodynamic data
Chemical Kinetics base used.
Thermodynamic data are fairly well known
The field of chemical kinetics can be used to and introduce little uncertainty into the
describe the changes in gas composition with modeling. However, reaction mechanisms do
time that result from chemical reactions. The vary. Pitts presents a comparison of the use of
kinetics of actual combusting flows are depen- various mechanisms in the literature [14]. The
dent on the initial species present, temperature, comparison indicates that reaction kinetics for
pressure, and the fluid dynamics of the gases. high temperatures (greater than 1100 K) are
Due to the inability to adequately characterize fairly well understood. However, the elementary
the complex mixing processes and the significant reactions for the range of 800–1000 K are not as
temperature gradients in turbulent flames, the use certain; therefore, quantitative modeling results
of kinetic models is restricted to simplified in this range may be suspect. Nevertheless, the
combusting flow processes. Consequently, the general trends presented below are valid despite
fire plume in a compartment fire is beyond cur- any uncertainty associated with the
rent chemical kinetics models. However, the mechanisms used.
reactivity of the upper-layer gas composition Chemical kinetics modeling shows that signif-
can be reasonably modeled if one assumes that icantly different trends occur for overventilated
the layer can be characterized as a perfectly and underventilated burning conditions. This can
stirred reactor, or that the layer gases flow away be seen in Figs. 16.13 and 16.14, which present
from the fire plume in a plug-flow-type process major species concentrations with respect to time
[14, 15]. Pitts has shown that no significant for an overventilated and underventilated condi-
differences between results exist for either tion, respectively. Figure 16.13 shows a modeled
modeling approach when applied to these upper case for ϕ equal to 0.91 and a temperature of
layers [14]. 900 K. The initial composition is taken from
Several kinetics studies have been performed Beyler’s data for a fire with a layer temperature
to examine aspects of the reactivity of upper- of 587 K. The temperature of 900 K corresponds
layer gases [12, 14, 15]. Comparisons between to the temperature observed by Gottuk et al. for a
Fig. 16.13 Chemical 10

kinetics model calculated
species concentrations
Upper-layer concentrations (% vol)

versus time for an 8 H2 O/2
overventilated (ϕ ¼ 0.91)
burning condition with an
upper-layer temperature of 6
900 K [15] CO2 /2
O2
2
C2 H4 H CO
2
0
0 10 20 30 40
Time (s)
Fig. 16.14 Chemical 1300 K

kinetics model calculated 0.05
species concentrations
versus time for an H2O/4
underventilated (ϕ ¼ 2.17)
burning condition with an 0.04 CH4/2
upper-layer temperature of
1300 K [14]
CO2/2
H2
Mole fraction
0.03
CO
0.02
0.01
O2
0 0.4 0.8 1.2 1.6 2

Time (s)
hexane compartment fire at the same global given sufficient time. This is representative of
equivalence ratio. For overventilated conditions, the case of the overventilated hexane and
increased temperatures cause CO concentrations PMMA compartment fires studied by Gottuk
to initially increase. As can be seen in Fig. 16.13, et al., in which the higher compartment
this is due to the incomplete oxidation of temperatures, compared to the hood tests of
hydrocarbons (modeled as C2H4). Once the Beyler, resulted in near-zero CO yields for ϕ
hydrocarbons are consumed, available O2 is less than 1.
used in the oxidation of CO to CO2. Since Figure 16.14 shows an underventilated case for
overventilated conditions indicate excess oxy- ϕ equal to 2.17 and a temperature of 1300 K. The
gen, CO concentrations are reduced to zero initial composition is taken from Morehart et al.
www.ebook777.com
Fig. 16.15 Carbon

monoxide concentrations 1300 K
as a function of time for a 0.026
range of isothermal
conditions. Initial 1200 K 1100 K
concentrations from a
methane hood fire at ϕ ¼
CO mole fraction
2.17 [14]
0.022
K
00
10
0.018 800 K
900 K
700 K
0.014
0 4 8 12 16 20
Time (s)
for a methane hood experiment [12]. Similar to the an underventilated fire (ϕ ¼ 2.17). Pitts noted that
overventilated conditions, CO increases due to the there are three distinct temperature regimes. At
oxidation of hydrocarbons (CH4). However, the temperatures under 800 K, the gas mixture is
available oxygen is depleted before the unreactive and the CO to CO2 reactions are said
hydrocarbons are fully oxidized. The resulting to be “frozen out.” As the temperature increases in
composition consists of higher levels of CO and the range of 800–1000 K, the mixture becomes
H2 and decreased levels of unburned fuel. Carbon more reactive and CO is formed at faster rates,
dioxide levels remain virtually unchanged. The due to the oxidation of unburned hydrocarbons.
much higher temperature studied in this case For the time period shown, it is interesting to
results in much quicker reaction rates, as is note that the ultimate concentration is approxi-
reflected in the 2 s time scale for Fig. 16.14 com- mately constant1 for each case in this temperature
pared to 30 s for Fig. 16.13. range. The third regime of high temperatures
It is clear from Figs. 16.13 and 16.14 that hydro- above 1100 K is characterized by fast reaction
carbon oxidation to CO and H2 is much faster than rates and much higher CO production for the 20 s
CO and H2 oxidation to CO2 and H2O, respec- reaction time shown. With sufficient time, the
tively. This is a result of the preferential combina- ultimate CO concentration for the 800–1000 K
tion of free radicals, such as OH, with conditions would approach the same value as that
hydrocarbons over CO. Carbon monoxide is seen for the higher temperatures.
oxidized almost exclusively by OH to CO2 Results of Zukoski et al. [10] and Gottuk
[27]. Therefore, it is not until the hydrocarbons et al. [15] indicated that layer temperatures of
are consumed that free radicals are able to oxidize 850–900 K or higher are needed for the layer
CO to CO2. gases to be reactive. Considering that the
The formation and consumption of CO in a
reactive gas mixture is dependent on both the
1
temperature of the mixture and the amount of Note that although the ultimate CO concentration is
time over which the mixture reacts. This point is roughly constant, the value of 2.1 % for this illustration
is not to be taken as a universal limit for this temperature
illustrated in Fig. 16.15, which shows the resulting
range. In general, the resulting CO concentration will
CO concentrations at different isothermal depend on the initial gas composition and the time over
conditions from an initial gas mixture taken from which the mixture is allowed to react.
minimum (freeze out) temperature above which The temperature of the fire plume has a
a gas mixture is reactive is dependent on the time significant effect on species production from the
scale evaluated. These values are consistent with fire plume. It is reasonable to assume that
the results shown in Fig. 16.12. In terms of com- differences in upper-layer temperature are also
partment fires, the time over which the gases reflective of a similar trend in the average tem-
react can be taken as the residence time of the perature of the fire plume gases. An increase in
gases in the upper layer, which is calculated the upper-layer temperature can increase the fire
according to Equation 16.11. In many practical plume temperature in two ways. For plumes that
cases of high-temperature compartment fires, it extend into the upper layer, entrainment of hotter
would be reasonable to assume that the residence upper-layer gases will result in increased plume
time of layer gases would be longer than the time temperatures compared to plumes in layers with
needed for the gas mixture to react fully. lower temperature gases. Secondly, an increase
in the surrounding temperature (both gases and
compartment surfaces) reduces the radiant heat
Fire Plume Effects loss from the plume, thus resulting in a higher
plume temperature.
Although a fire plume is too complex to ade- The effect of temperature on species genera-
quately model the chemistry, the hood tion in a fire plume can be found in the methane
experiments discussed earlier provide significant hood experiments of Morehart et al. [12] and
insights with respect to the fire plume and species Zukoski et al. [11] Morehart studied the effect
production in compartment fires. Results of of increasing temperature on layer composition
Beyler’s hood experiments suggest that the pro- by adding different levels of insulation to his
duction of upper-layer gases is independent of hood. Except for the insulation, the test
the structure and fluid dynamics of the flame. conditions (e.g., ϕ of 1.45 and layer interface
Beyler modified a 19 cm propane burner by height) were held constant. For the range of
including a 2.8 cm lip to enhance turbulence and temperatures studied (500–675 K), substantial
the large-scale structure of the flame [8]. Com- increases in products of complete combustion
pared to the no-lip burner, the flame was mark- and decreases in fuel and oxygen occurred with
edly changed, and air entrainment was increased increasing layer temperature. Upper-layer oxy-
by 30 %. Yet, the upper-layer species-equiva- gen mass fraction was reduced by approximately
lence ratio correlations were the same for both 70 % and methane was reduced by 25 %. Exclud-
burners. Additionally, as shown in Fig. 16.5, ing one outlier data point, CO concentrations
correlations for different size burners are also increased by 25 %. The temperatures of the
identical. Morehart et al. upper layer were well below
The insensitivity of species yields to the 700 K. Therefore, based on kinetics modeling,
details of the flame structure is also suggested these layers were unreactive at these low
by the compartment fire hexane results of Gottuk temperatures. It follows that the change in layer
et al. [5] The correlations include data from fires composition must have been due to changes in
utilizing various size burn pans and with a wide the plume chemistry. The more complete com-
range of air entrainment rates. In several cases, bustion can be attributed to an extension of
nearly equal steady-state equivalence ratio fires the flammability limits (or reaction zone) in the
were obtained with quite different burning rates plume due to raising the flame temperature. The
and air entrainment rates. Although the above discussion demonstrates that increasing
conditions varied significantly, the positive cor- the plume temperature substantially increases
relation between yields and equivalence ratio the consumption of O2 and fuel, and primarily
suggests that the yields are not sensitive to the increases the levels of products of complete
details of the flame structure. combustion.
www.ebook777.com
The effect of changing temperature on a com- a ϕul that is less than ϕp. Conversely, a fire that is
partment fire upper-layer composition is twofold: dying down quickly, such that ϕp is decreasing
(1) the generation of species in the fire plume is rapidly, will have a ϕul that is higher than ϕp.
changed, and (2) oxidation of post-flame gases in These trends result due to the upper layer being a
the layer is affected. Elevated compartment temporary collection reservoir for the gases from
temperatures correlate with increased fire plume the fire plume.
temperatures and more complete oxidation of the In an effort to characterize transient
fuel to CO2 and H2O within the fire plume. The conditions, Gottuk et al. defined a steady-state
layer temperature dictates post-flame oxidation time ratio, τSS, as the ratio of the residence time,
in the upper layer. tR, to a characteristic growth time of the fire.
Upper-layer temperatures below about 800 K Since fire growth is directly related to the fuel
indicate chemically unreactive layers. As such, volatilization rate, a representative growth time
combustion within the fire plume controls the of the fire was defined as the ratio of the fuel mass
final CO levels that would be measured in the loss rate, ṁf, to the derivative of the fuel volatili-
upper layer. At these low temperatures significant zation rate, m€ f . An increase in τSS is indicative of
levels of CO can be generated even for some more transient conditions.
overventilated conditions (0.5 < ϕ < 1). The
yield of CO is inversely proportional to tempera- tR
τSS ¼ ð16:18Þ
ture for overventilated conditions and directly pro- m_ f =m€f
portional to temperature for underventilated
conditions. An analysis of the transient nature of the com-
Upper-layer temperatures of about 900 K and partment fires studied by Gottuk et al. showed
higher indicate chemically reactive layer gases. As that values well below 1 indicated near steady-
such, reactions in the layer dictate final CO pro- state conditions, such that the plume and upper-
duction. Temperatures above 900 K allow nearly layer equivalence ratios could be considered
complete oxidation of CO to CO2 for overven- equal. Investigation of individual fires showed
tilated conditions. For underventilated fires, chem- that the steady-state time ratio decreased below
ical kinetics modeling indicates that higher 1.0 at very early times in the fire. Typically, the
temperature environments may result in slightly ratio was 0.1 or less for the quasi-steady-state
higher CO yields due to preferentially accele- periods over which data was averaged. For
rated hydrocarbon oxidation compared to CO some fires, during the highly transient transition
oxidation. from overventilated to underventilated condi-
During underventilated conditions, two tions, the τSS increased quickly, approaching a
mechanisms affecting net CO formation compete value of 1.
(i.e., CO and hydrocarbon oxidation). Increasing The correlations presented in the engineering
gas temperatures above 900 K depletes CO by methodology represent data that have been aver-
accelerating the CO to CO2 conversion. How- aged over steady-state (hood experiments) or
ever, incomplete oxidation of unburned quasi-steady-state (compartment fires) periods.
hydrocarbons increases the CO production. For the purpose of modeling fires with respect
Since hydrocarbon oxidation is much faster to time it is of interest to know how the species
than CO oxidation, net CO levels increase until yields correlate with the equivalence ratio during
all available oxygen is consumed. transient conditions (i.e., as the fire is growing).
Determining this correlation was accomplished
by plotting the yield to equivalence ratio data for
Transient Conditions individual fires from the time of ignition to the
steady-state period. These transient correlations
Transient conditions cause the upper-layer were compared to the steady-state correlations
equivalence ratio to differ from the plume equiv- obtained from steady-state averaged data from
alence ratio. A fast-growing fire will tend to have all tests (e.g., the CO yield correlation shown in
Fig. 16.16 Comparison 0.4 1.0

•
between a transient, •
•• Notes:
unnormalized CO yield = Steady-state yield
correlation for a hexane fire 0.8
+ = Transient yield
Steady-state time ratio

0.3 •
with an average steady- • = τ
ss
state ϕp of 3 and the steady-
+ ++ + 0.6
state correlation for all + +
++ ++ +
+
0.2 + +
hexane fires studied by +
Gottuk et al. The steady- + + 0.4
state time ratio, τSS, data •
++ +
+
are shown as solid dots [17] 0.1 + ++
•••
• •• + ++ ++ + 0.2
• • •••• •• • •
•• • • • •
• ••• • • • • • •• • • •• • •• • • • ••
•• •• ••••• • • ••••• • • •
+
++ +
• ••• • •
•
0 +++++++++++++ + + + • •• • 0
0 0.5 1.0 1.5 2.0 2.5 3.0
Equivalence ratio
Fig. 16.9). An example of one comparison is having relatively large fires compared to the
shown in Fig. 16.16. Figure 16.16 shows the compartment volume. Until flashover conditions
steady-state hexane CO yield correlation along are approached, a full-scale compartment fire
with the transient yield vs. equivalence ratio data will most likely have smaller fires compared to
for a hexane compartment fire that obtained a the volume of the space. As a result, the resi-
steady-state average ϕp of 3. The solid dots in dence time of gases in the upper layer of a full-
Fig. 16.16 represent the steady-state time ratio scale fire may be much longer. Times on the
data, τSS. For this example, τSS remained fairly order of 5–10 min may not be unrealistic in
constant at about 0.1 for the entire fire. And as some cases. Therefore, in the case of a fast-
can be seen, the agreement between the transient growing full-scale fire, values of ϕp could
and steady-state correlations is quite good, even increase relative to ϕul. The application point is
for the transition to underventilated conditions. that the control volume used for the equivalence
Good agreement between transient and steady- ratio must be considered with respect to the resi-
state data was also observed for CO2 and O2 yield dence time of gases in the upper layer.
correlations.
Although more transient in nature than the hood
experiments, the compartment fires are charac- Species Transport to Adjacent Spaces
terized as primarily quasi-steady and, therefore,
do not differ significantly from Beyler’s hood The species levels transported from a compart-
experiments in this respect. This analysis also ment depend on a variety of conditions produced
shows that the species yield correlations developed during the course of the fire. As compartment fire
for steady-state conditions are representative of gases exit the compartment, they entrain the
the transient growth periods of these fires. gases present in the adjacent space (Fig. 16.17).
In terms of full-scale application, these results If a fuel-rich mixture is produced in the compart-
suggest that ϕp and ϕul are approximately equal ment, the gases flowing out may ignite, causing
for compartment fires characterized by relatively burning in the adjacent space. This burning is an
slow growth compared to the upper-layer resi- indication that oxidation reactions are taking
dence time (i.e., τSS 1). However, the low τSS place, which ultimately affects the species levels
values observed in the reduced-scale compart- transported to remote areas. As the gases con-
ment fires may not always be representative of tinue to flow through the adjacent space, they are
full-scale fires. The reduced-scale compartment cooled by mixing with surrounding gases and
fires had residence times typically between 4 and heat losses to the boundaries. Eventually, gases
20 s. These short residence times were a result of are cooled to a temperature below which
www.ebook777.com
Fig. 16.17 Phenomena Lf,tip

controlling species
transport to remote Oxidation
Transport
reactions •
locations Qloss to remote
locations
Reaction
mair is frozen
•
mair
•
mfuel
•
oxidation reactions do not readily occur. At this “Chemical Kinetics” in this chapter). In the
point the reactions are said to be frozen. The presence of oxygen, hydrocarbons begin to react
amount of combustion products that exist at this efficiently when temperatures are above 700 K
point will continue to flow throughout the rest [14]. Perhaps more importantly, the oxidation of
of the structure. As a result, combustion CO to CO2 does not readily occur until temper-
product levels in the overall structure will accu- atures rise above 800 K. The occurrence of burning
mulate as the fire inside the compartment persists outside the compartment, either from flame exten-
and/or additional items in the structure begin sion or the onset of flashover, will result in local
to burn. temperatures in excess of 1300 K [4]. At these
Conditions inside the fire compartment and in temperatures, oxidation reactions for both
the directly adjacent space will govern the spe- hydrocarbons and CO occur in the presence of
cies levels transported to remote locations. The available oxygen.
primary consideration is the conditions that The occurrence of burning outside the com-
develop inside the compartment. If burning out- partment has been shown, in most situations, to
side of the fire compartment occurs due to either reduce the incomplete combustion products
flame extension or external burning, gases will (including smoke and CO levels) transported to
continue to react outside of the compartment remote locations [5, 28–30]. The burning in
influencing the species levels transported to unconfined adjacent areas (e.g., open surround-
remote areas. The degree to which gases react ings) has been measured to decrease incomplete
outside the compartment depends on the mixing combustion products more efficiently than burn-
of oxygen with the fuel-rich gases flowing out of ing in confined adjacent areas (e.g., a corridor).
the compartment, and the addition of fuel to the In addition, the consumption of incomplete com-
gases flowing along the adjacent space. bustion products during burning in confined
areas was found to be sensitive to the air entrain-
ment into the plume/ceiling jet flow. This
General Effects of Burning Outside entrainment is a function of the mass flow from
the Compartment the compartment, the geometry of the opening
between the compartment of fire origin and the
Species levels transported to remote locations will adjacent space, and the geometry of the adjacent
be equivalent to those formed inside the compart- space itself. Smoke layers that develop in con-
ment unless burning occurs outside. Chemical fined adjacent spaces can cause lower oxygen
kinetics indicate that oxidation reactions cannot levels to be entrained into the plume/ceiling
occur efficiently outside the compartment unless jet flow in the adjacent space, increasing the
gas temperatures are near those produced at the incomplete combustion product levels
onset of flashover (775–875 K) (see the section transported to remote areas.
Burning in Unconfined Adjacent Areas Burning in Confined Adjacent Areas
Unconfined adjacent areas are those areas where Burning in a confined adjacent area, such as a
the flame extending from the compartment of corridor or room, causes the external flame to
origin is not redirected by the boundaries of the impinge on a ceiling, and possibly on walls.
adjacent area, and the gases are allowed to burn Compared with flow in an unconfined area, the
as a buoyant plume. Examples of unconfined ceiling and walls in the confined area will reduce
adjacent areas include outdoor surroundings, the amount of air entrainment into the gas jet
atriums, and corridors with high ceilings relative exiting the fire compartment. The effects of
to the door height of the burning compartment. burning in confined areas on species transport
Gottuk et al. [15, 31] investigated the impact of have been investigated by Ewens et al. [28, 32]
external burning on combustion products down- and Lattimer et al. [29, 30] using the same fire
stream of an unconfined jet. The compartment compartment design in the unconfined external
was connected to its surroundings through a win- burning study (Figs. 16.19 and 16.20)
dow opening. Tests were performed with com- [17, 31]. The transport of species to remote
partment fires with and without external burning. locations is geometry dependent and can be
In these tests, a compartment equivalence ratio of affected by smoke layers that develop in the
1.6 was the lowest ϕc where external burning was confined area. However, species levels transported
noted to occur. The effects of external burning on to remote areas can be predicted by defining an
the CO levels downstream of the fire are shown equivalence ratio for a control volume involving
in Fig. 16.18. With a compartment equivalence both the fire compartment and the burning in the
ratio greater than 1.6, CO levels were measured adjacent space.
decreasing below the fire compartment levels. Species transport has been evaluated for two
The CO yield is shown to decrease to a minimum common compartment-hallway configurations.
of 0.02 at ϕ greater than 2.0. The decrease in CO Ewens [28] and Lattimer et al. [30] evaluated
represents a 75–90% reduction of the CO species transport in a configuration with the fire
generated in the fire compartment. compartment on the end of a hallway (see
Fig. 16.18 Effect of 0.40

external burning on CO
levels downstream of an 0.35
unconfined adjacent area.
CO yield versus 0.30

compartment equivalence
ratio for hexane 0.25
compartment fires with an
0.20
exhaust jet to the open
atmosphere through a 0.15
window opening; (○)
compartment levels, (☉) 0.10
downstream levels [17, 31]
0.05
0.00
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
www.ebook777.com
1.52 m Inlet soffit

Air Compartment
Fire
location Hallway
Exhaust
1.22 m duct
Adjustable
size opening
1.22 m
Air plenum
1.67 m
3.66 m
1.22 m
Fig. 16.19 Compartment on the end of a hallway [28, 30, 32]
Compartment
1.22 m 1.52 m
5.18 m
Exhaust
1.22 m Hallway duct
Fire
location 1.67 m
Inlet soffit
Air plenum
Adjustable
Air inlet duct
size opening 1.22 m
Exit soffit
Air
Fig. 16.20 Compartment on the side of a hallway [29, 30]
Fig. 16.19). Lattimer et al. [29, 30] performed a Using the apparatus shown in Fig. 16.19,
study with the fire compartment on the side of a Ewens [28] evaluated the effects of different
hallway (see Fig. 16.20). In both studies, most geometric variables and compartment stoichiom-
tests were performed with a window connecting etry on CO levels (in addition to other species)
the compartment and hallway. downstream of the external burning where

external burning on CO
levels downstream of a 0.35
postflashover compartment
fire. Adjacent hallway with 0.30
a (~) 0 m inlet soffit and (∎)
0.20 m inlet soffit. (☉) 0.25
Unconfined area and (○)
0.20
inside the compartment
[17, 28, 31, 32] 0.15
0.10
0.05
0.00
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Fig. 16.22 The mass flow 1e + 0

rates in a space adjacent to
a postflashover fire with O2
external burning plotted 1e – 1 CO2
Mass flow rate (kg/s)
versus the ratio of the CO

distance down the hall, x, to THC
the length of the flame to its 1e – 2
Total
tip, Lf,tip. Opening of
0.12 m2, no inlet soffit
above the opening, and an 1e – 3
average compartment ϕ ¼
2.0 [28]
1e – 4
1e – 5
0.0 0.5 1.0 1.5 2.0
x/Lf,tip
reactions were considered frozen. As shown in vs. no soffit) or reduce the layer depth in the
Fig. 16.21, Ewens [28] demonstrated that burning adjacent space will enhance the oxidation
in confined areas outside the compartment reactions and result in lower levels of incomplete
decreases incomplete combustion product levels, combustion products, such as CO.
but not always to the same extent as when the fire Fire gases flowing into and through the adja-
gases spill into an unconfined space. The degree cent space entrain surrounding gases as they flow
of oxidation was found to be dependent on both away from the fire compartment. These gases can
air entrainment into the gases exiting the com- undergo chemical reactions, particularly as they
partment and the fuel flowing out of the compart- are within the flaming region. A sample plot of
ment. The data of Ewens show that both fire size the species variation along the flame length in the
and geometric parameters will affect species pro- hallway is shown in Fig. 16.22 [28]. Results
duction. The two important results to understand shown in Fig. 16.22 are from postflashover hex-
are that geometries which increase air entrain- ane fire tests with an average compartment
ment in the adjacent space (e.g., an inlet soffit equivalence ratio of ϕ ¼ 2.0, a 0.12 m2 opening
www.ebook777.com
(0.50 m wide 0.24 m high) connecting the where burning occurred. In this set of data, the
compartment and adjacent space, and no inlet mixing in the adjacent space (hallway) was
soffit above the opening in the hallway. These varied by using windows with different areas
data represent time-averaged conditions 0.025 m and aspect ratios to connect the compartment to
below the ceiling along the hallway during the the adjacent space, and by adding a 0.20 m soffit
quasi-steady-state period of the fire, when exter- above the window. The equivalence ratio in this
nal burning was occurring. analysis was calculated using a control volume
Gases entered the adjacent space as a ceiling that extended to the sampling point located in the
jet, but were allowed to expand horizontally until adjacent space. Using this control volume, the
intersecting the walls of the hallway. For this mass flow rate of air for the ϕcv calculation in
geometry, the largest increase in the total mass Equation 16.13 was the air flow rate into the
flow rate was during the first half of the flame compartment plus the air entrainment into the
length, which was 2.7 m (on average) from the plume/ceiling jet flow in the adjacent space, up
compartment. By x/Lf,tip ¼ 0.75, the total mass to the sampling location. In these experiments,
flow rate had reached a maximum. This indicated gas sampling was always performed in or just
that all of the entrainment into the ceiling jet flow downstream of the flame within the hallway.
occurred by x/Lf,tip ¼ 0.75. The majority of the In Figs. 16.23, 16.24, 16.25, and 16.26 the
oxidation reactions had also occurred by x/Lf,tip species yields are plotted vs. the control volume
¼ 0.75. The mass flow rates of CO, CO2, and O2 equivalence ratio, ϕcv. Note that CO yields are
were essentially constant downstream of x/Lf,tip not normalized because for various fuels
¼ 0.75. This indicates that the oxidation of CO to unnormalized CO yields were found to correlate
CO2 was frozen by an x/Lf,tip ¼ 0.75. Small best with ϕ. The lines in the normalized O2
amounts of total hydrocarbons (THC) continued depletion and CO2 and THC yield plots represent
to react from 0.75 < x/Lf,tip < 1.0; however, this the results from the complete combustion model
was not measured to significantly increase CO presented in Equation 16.14. Due to limited data
levels. near the compartment, there are few data points
Analysis of data in Fig. 16.22, as well as other at high ϕcv.
data by Ewens et al. [28, 32] and Lattimer The trends in the species data were similar to
et al. [29, 30], indicates that by the flame tip all those observed in the hood experiments by
of the oxidation reactions have occurred. As a Beyler [8, 9] and in the compartment
result, the mass flow rates of the major combus- experiments by Gottuk et al. [5] The normalized
tion products beyond the flame tip will be O2 depletion is approximately unity at ϕcv less
transported to remote locations. The mass flow than 1.0 and decays at the rate prescribed by the
rate levels will be influenced by the oxygen complete combustion model at higher ϕcv. CO2
availability in the flaming region. Based on test normalized yields are near unity up to a ϕcv of
results from Ewens et al. [28, 32] and Lattimer approximately 0.8. At ϕcv greater than 0.8, the
et al. [29, 30], mixing near the compartment has CO2 levels begin to decay and are consistently
been shown to have the most significant influ- less than the level predicted by the complete
ence on the combustion products transported to combustion model. This behavior is consistent
remote areas. with the rise in incomplete combustion products,
such as CO and THC, at ϕcv ranging from 0.6 to
0.8. These results indicate that species levels in
Predicting Species Levels adjacent spaces can be adequately correlated by
the same global equivalence ratio correlations
Lattimer [33] performed additional analysis on obtained for species production in fire
Ewens’ data to develop a correlation between compartments as long as ϕcv is calculated using
species transported and ϕ for a control volume the appropriately defined control volume. ϕcv
consisting of part of the area in the adjacent space accounts for the effects of external burning on

O2 depletion of gases in a
space adjacent to a 1.4
postflashover compartment
1.2
Normalized O2 yield
fire. Control volume
includes fire compartment 1.0
and a portion of the
adjacent space where 0.8
burning occurs
0.6
0.4
0.2
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Control volume equivalence ratio, φ cv

CO2 yields of gases in a
postflashover
1.2
compartment fire
1.0
0.8
0.6
0.4
0.2
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Control volume equivalence ratio, φ cv

CO yields of gases in a
postflashover
0.30
compartment fire.
0.25
0.20
0.15
0.10
0.05
0.00
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Control volume equivalence ratio, φcv
www.ebook777.com

THC yields of gases in a 0.45
space adjacent to a
postflashover 0.40
Normalized THC yield

compartment fire 0.35
0.30
0.25
0.20
0.15
0.10
0.05
0.00
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Control volume equivalence ratio, φcv
species levels and can be used with Equations ignition for tests with different layer depths.
16.19, 16.20, 16.21, 16.22, 16.23, 16.24, and Layer depth is represented as a dimensionless
16.25 to estimate species transported to remote depth, γ ¼ δ/z, which relates the distance
areas. between the ceiling and the bottom of the visible
smoke layer, δ, to the distance between the ceil-
ing and the bottom of the gases flowing out of the
Effects of Oxygen-Deficient Smoke compartment, z. (For a window configuration,
Layers in Adjacent Spaces z is measured to the bottom of the window.) As
shown in Fig. 16.27, the smoke layer did not
The development of hot, oxygen-deficient smoke affect the time to ignition until the visible
layers in the adjacent space affects both the smoke layer was nearly deep enough to prevent
entrainment into the plume/ceiling jet and the ignition altogether (indicated by the infinite time
amount of oxygen mixing with the fuel-rich to ignition). At layer depths greater than γ ¼ 1.7,
gases flowing from the fire compartment. external burning did not occur since the exiting
Ewens et al. [28, 32] demonstrated that layers fire gases were not able to entrain sufficient fresh
as thin as 0.20 m may have an impact on incom- air to provide the necessary oxygen for combus-
plete combustion products being transported to tion. Rather, the gases exiting the fire compart-
adjacent areas. Lattimer et al. [29, 30] performed ment entrained primarily vitiated gases in the
a series of tests with different oxygen-deficient upper layer of the adjacent space.
layer depths in the space directly adjacent to a The CO, CO2, and THC yields measured at a
postflashover fire (see Fig. 16.20). Tests were remote location (in the exhaust duct) for the
performed with three different opening sizes different window opening sizes are shown plot-
connecting the compartment to the adjacent ted in Figs. 16.28, 16.29, and 16.30 with respect
space, but the compartment stoichiometry was to the smoke layer depth. Each data point is the
approximately the same in all tests. In each test, time-averaged yield during the quasi-steady-state
the layer depth was kept at a constant level by the part of the fire. For this geometry, combustion
use of an exit soffit. In order to change the layer product levels were not significantly affected by
depth from test to test, the height of the exit soffit the smoke layer until it fell below the bottom of
was adjusted for each test. the opening (γ > 1.0). As the layer depth
Except for cases with a deep smoke layer in increased from γ ¼ 1.0–1.8, the burning outside
the adjacent space, external burning occurred in the compartment became increasingly less effi-
all tests. Figure 16.27 contains a plot of time to cient. This resulted in an increase in CO and
Fig. 16.27 Time for

external burning in tests
with a range of layer depths 500
Time to external burning (s)

in the adjacent space with a 450
6.32 m2 floor plan area [30] 400
350
300
250
200
150 A vent = 0.12 m2
100 A vent = 0.08 m2
A vent = 0.04 m2
50
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Dimensionless smoke layer depth, δ/z
Fig. 16.28 The effect of 0.5 3.0

an oxygen-deficient upper
layer on downstream CO
2.5
Normalized THC yield and
THC
unnormalized CO yields
species yields from a 0.4

CO2
postflashover fire extending 2.0
into a hallway. Opening of 0.3
0.12 m2, 0.20 m soffit
1.5
above the opening, and an
average ϕc ¼ 3.1. Open 0.2
symbols are tests with no 1.0
external burning. γ > 1 0.1
indicates layer is below the 0.5
bottom of the vent [28, 29]
0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Dimensionless layer depth, γ = δ/z

an oxygen-deficient upper-
Normalized THC yields and
0.4 THC 0.8

species yields from a

CO2
postflashover fire extending
into a hallway. Opening of 0.3 0.6
0.08 m2, 0.20 m soffit
average ϕc ¼ 2.8. Open 0.2 0.4
symbols are tests with no
external burning [30]
0.1 0.2
0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
www.ebook777.com

an oxygen-deficient upper
Normalized THC yields and

THC
species from a 0.4 0.8

CO2
postflashover fire extending
into a hallway. Opening of 0.3 0.6
0.04 m2, 0.20 m soffit
average ϕc ¼ 2.8. Open 0.2 0.4
symbols are tests with no
external burning [29, 30] 0.1 0.2
0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
THC yields and a decrease in the CO2 yield. The explored. These variables include the effects of
increase in CO and THC yields was attributed to air addition through forced ventilation, addi-
reducing the oxygen available to oxidize the tional fuel decomposition in the adjacent space,
combustion products. When the smoke layer and heat losses to the ceiling and walls.
was increased to a dimensionless layer depth of The addition of air to the system through
γ ¼ 1.7–1.8, external burning was not observed forced ventilation may have an influence on spe-
and downstream species yields were consistent cies levels, depending on where the air is added
with levels inside the fire compartment. relative to the external burning. Forced ventila-
External burning in some tests with deep oxy- tion in the region where external burning is
gen deficient upper layers actually caused addi- occurring will introduce additional oxygen into
tional CO formation in the adjacent space. As the flow, and possibly induce additional mixing.
shown in Fig. 16.28, in tests with the largest This may result in better oxidation of incomplete
opening, 0.12 m2, and a dimensionless layer combustion products, such as CO. Addition of air
depth of γ ¼ 1.3–1.5, CO yields increased to to the system downstream of the external burning
approximately 0.27 kg/kg, which is approxi- will dilute the gases, but will not reduce the
mately 0.05 kg/kg higher than compartment amount (in terms of mass) of incomplete com-
levels. In addition, corresponding normalized bustion products being transported to remote
THC yields were on average 0.06 kg/kg lower locations.
than compartment levels. These results indicate Decomposition of fuel in the adjacent space
that available oxygen is being used to preferen- may affect species levels transported to remote
tially oxidize THC instead of CO. This oxidation locations. This effect may be sensitive to the
of THC forms additional CO, causing an increase location of the decomposing fuel, the type of
in CO levels transported to remote locations. The fuel, whether it is flaming or smoldering, and
exact conditions in the adjacent space necessary the conditions surrounding the fuel.
to produce these results have not been fully Heat losses to the ceiling and walls can cause
established. gas temperatures in the adjacent space to
decrease more readily for some materials, com-
pared to well-insulated boundaries. An example
Other Considerations of this may be steel decks and bulkheads on
ships. A decrease in gas temperature may cause
There are other variables that may influence the temperatures to reach levels where no reactions
combustion product levels being transported to can occur sooner than those observed in well-
remote locations that have not been fully insulated cases. As a result, higher heat loss to
the boundaries may result in higher incomplete 1. Determine the compartment equivalence
combustion products, including CO, to be ratio, ϕc.
transported to remote areas. 2. If ϕc is less than 1, estimate species levels
Species concentrations may not always be using the global equivalence ratio-yield
transported away from the compartment in a correlations presented in Equations 16.19,
uniform manner. In experiments performed by 16.20, 16.21, 16.22, 16.23, 16.24, and 16.25
Lattimer et al. [29, 30] with the compartment with the ϕc.
located on the side of a hallway, the bulk flow 3. If ϕc is greater than 1, determine whether
from the compartment was measured to flow external burning will occur outside of the
across the hallway and down the side of the compartment. External burning can be
hallway opposite the fire compartment. This assumed to occur at a ϕc (plume or compart-
resulted in higher CO, CO2, and THC levels ment) of 1.6, or by calculating the ignition
(and lower O2) flowing along the side of the index using Equation 16.1.
hallway opposite the fire compartment. For 4. If there is no external burning, use the ϕc and
example, CO levels were measured to be as Equations 16.19, 16.20, 16.21, 16.22, 16.23,
high as 1.9 % along the side of the hallway 16.24, and 16.25 to calculate the species
opposite the fire compartment, while on the side transported.
of the hallway with the fire compartment the 5. If external burning is occurring, determine the
maximum CO level was measured to be 1.0 %. effect of the smoke layer using the dimension-
As these gases flow farther from the compart- less smoke depth, γ ¼ δ/z, where δ is the
ment, they are expected to become more uniform depth of the layer below the ceiling and z is
across the hallway. However, the distance away the lowest elevation of gases exiting the
from the compartment where mostly uniform compartment.
flow occurs was not quantified. 6. If γ is greater than 1.0 and external burning is
predicted, the smoke layer can be assumed to
inhibit oxidation in the adjacent space. CO
Engineering Methodology and other incomplete-combustion products
are not reduced. The ϕc and Equations 16.19,
In light of the experimental work and chemical 16.20, 16.21, 16.22, 16.23, 16.24, and 16.25
kinetics considerations discussed previously, sev- can be used to estimate the species transported
eral correlations can be used as guidelines for fire to remote locations.
protection engineering. The production of chem- 7. If γ is less than 1.0 and external burning is
ical species in compartment fires has been shown predicted, the smoke layer does not inhibit the
to be correlated with the control volume equiva- oxidation in the adjacent space and
lence ratio, ϕcv. For most purposes, the equiva- incomplete-combustion products, such as
lence ratio can be calculated using instantaneous CO, will be reduced. The species transported
fuel burning rates and air mass flow rates assum- can be estimated using Equations 16.19,
ing quasi-steady-state conditions. The following 16.20, 16.21, and 16.22 and the equivalence
methodology presents a guide to determining ratio for a control volume, ϕcv, that
bounds on species production as well as incorporates the compartment and the adja-
comments on the limits of this approach. cent space out to the flame tip (see Fig. 16.17).
The methodology for estimating species It should be noted that this methodology may
transported to remote locations is provided in not provide the maximum levels of incomplete
Fig. 16.31. This approach considers the occur- combustion products that can be produced in a
rence of external burning outside the compart- fire. Equations presented in this methodology for
ment. In general, the primary steps in the species yields as a function of equivalence ratio
analysis are: have been shown to provide good correlations
www.ebook777.com
Fig. 16.31 General

methodology for predicting Calculate φc
species levels transported
to remote locations from a
fire compartment
No
φc > 1.0?
Yes
External No
burning?
Yes
Calculate smoke layer
depth, δ, relative to lowest
elevation of gases exiting
compartment, z, (γ = δ/z)
No External burning
γ = δ/z < 1.0?
has no effect
Yes
Calculate φcv with control

volume out to flame tip
in adjacent space
Use φcv to determine Use φc to determine

species yields species yields
even for wood as a secondary fuel pyrolyzing in CO. Equations 16.19a and 16.19b represents a
the hot upper layer. However, it is not clear “zeroth order” correlation between CO yield and
whether these correlations will hold for equivalence ratio. For overventilated burning
nonoxygenated fuels in the upper layer or how conditions, there is no CO production and for
well they will represent other oxygenated fuels. underventilated conditions CO is produced at a
Several empirical correlations have been yield of 0.2 g per gram of fuel burned. This
developed to predict species levels at a range of correlation applies best to fires with average
equivalence ratios. Different correlations are upper-layer temperatures greater than 900 K.
given in the following paragraphs to accommo-
f CO ¼ 0 for ϕ < 1 ð16:19aÞ
date analyses of various levels of complexity.
Due to its toxicity, CO production is of pri- f CO ¼ 0:2 for ϕ > 1 ð16:19bÞ
mary importance. Four correlations (see
Equations 16.19, 16.20, 16.21, and 16.22) are Equations 16.20a, 16.20b, and 16.20c
presented, representing varying degrees of com- accounts for some of the temperature effect by
plexity. In each case, the correlations basically including a linear rise in CO yield over the tran-
represent a lower bound for the yield of sition region from ϕ of 0.5 to 1.5.
f CO ¼ 0 for ϕ < 0:5 ð16:20aÞ The simple model presented as

Equations 16.14a, 16.14b, 16.14c, 16.14d, and
f CO ¼ 0:2ϕ 0:1 for 0:5 < ϕ < 1:5 16.14e with the inclusion of the empirically deter-
mined yield coefficients, is fairly adequate for
ð16:20bÞ
predicting CO2, O2, and H2O normalized yields
(see Equations 16.23, 16.24, and 16.25).
f CO ¼ 0:2 for ϕ > 1:5 ð16:20cÞ Suggested average yield coefficients for compart-
ment fires of elevated temperatures (T > 900 K)
The temperature effect on CO production is are 0.88 for BCO2 and 0.97 for BO2 [5]. Suggested
best represented in the following two values for low upper-layer temperatures
correlations. Equation 16.21, which represents a (T < 800 K) are 0.77 for BCO2 , 0.92 for BO2 , and
fit to the hexane data of Beyler’s hood 0.95 for BH2 O . Average yield coefficients for
experiments, is suggested for compartment fires underventilated fires are shown in Table 16.2.
with average upper-layer temperatures below
800 K. Equation 16.22 is used for fires with f CO2 ¼ 1 for ϕ < 1 ð16:23aÞ
upper-layer temperatures above 900 K. Equation
f CO2 ¼ BCO2 =ϕ for ϕ > 1 ð16:23bÞ
16.15 is an approximate fit to the compartment
fire hexane data of Gottuk et al. For the most part, f O2 ¼ 1 for ϕ < 1 ð16:24aÞ
CO yields from hexane fires represent lower
limits observed for the fuels studied to date [5, f O2 ¼ BO2 =ϕ for ϕ > 1 ð16:24bÞ
8]. Therefore, these equations provide a mini-
mum CO production that can be used for hazard f H2 O ¼ 1 for ϕ < 1 ð16:25aÞ
analysis.
f H2 O ¼ BH2 O =ϕ for ϕ > 1 ð16:25bÞ
1
Y CO ¼ð0:19=180Þ tan ðXÞ
ð16:21Þ As presented in Equations 16.23, 16.24, and
þ 0:095 for T < 800 K
16.25, normalized chemical species yields, f, can
whereX ¼ 10ðϕ 0:8Þ and tan1 (X) is in degrees be correlated quite well by the global equiva-
lence ratio. This is true for a wide range of fuel
Y CO ¼ð0:22=180Þ tan 1 ðXÞ types. However, it is worthwhile to point out that
ð16:22Þ for different fuels, the CO2, O2, and H2O yields
þ 0:11 for T > 900 K
to equivalence ratio correlations only collapse
where X ¼ 10ðϕ 1:25Þ and tan1(X) is in down to a single curve when the yields are
degrees normalized by the maximum possible yield for
The figures presented earlier of CO yield a given fuel (i.e., presented as f rather than Y ).
versus equivalence ratio also show plots of Although complete combustion does not occur,
Equations 16.21 and 16.22. Figure 16.7 shows combustion efficiencies with respect to equiva-
the CO yield data for methane hood experiment lence ratio are similar enough between fuels that
fires in which upper-layer temperatures ranged the stoichiometry of a particular fuel will dictate
from 490 to 870 K. The CO yield data of the generation of CO2 and the depletion of O2.
Zukoski et al. and Toner et al. lie between the Therefore, the species associated with complete
curves of Equations 16.21 and 16.22, particu- combustion (CO2, O2, and H2O) are not expected
larly for slightly overventilated and stoichio- to have equal yields for different fuels, since
metric conditions. This is consistent with the varying fuel compositions will dictate different
fact that these fires had temperatures that were limits of CO2 and H2O that can be generated and
higher than those represented by Equation O2 that can be consumed for a gram of fuel
16.14 and some were within the transition burned. By normalizing the yields, the variability
range of 800–900 K. of fuel composition is removed.
www.ebook777.com
On the other hand, carbon monoxide produc- concentrations in Fig. 16.32) represent a reason-
tion is best correlated by the equivalence ratio able lower bound for a range of typical fuels.
when represented as a simple yield, YCO, rather Higher concentrations of CO can be created,
than a normalized yield, fCO. This is one indicator particularly when additional fuel is added to a
that CO production is not strongly dependent on vitiated upper layer. Corresponding to Fig. 16.32,
fuel type, as is production of CO2 and O2. The CO2 and O2 concentrations for hexane compart-
reason for this is believed to be due to the fact ment fires are shown in Figs. 16.33 and 16.34,
that CO is effectively an intermediate product respectively. Even though the peak
that depends more on the elementary chemistry concentrations of CO2 will be dependent on the
than on fuel composition, which determines fuel type, the oxygen concentration as a function
products of complete combustion. of will be similar for most hydrocarbons [5].
Once yields are determined using the above The ratio of CO to CO2 concentrations can be
correlations, species gas concentrations can be used as an indicator of the combustion mode.
calculated. Equation 16.26 can be used to calcu- Higher combustion efficiency is obtained as
late the concentration of species i for all species more fuel is burned completely to CO2 and H2O
except oxygen. Oxygen concentrations can be and is indicated by a ratio of CO to CO2 near
calculated from the depletion of oxygen using zero. Since CO is a product of incomplete com-
Equation 16.27. bustion, the ratio of CO to CO2 concentrations
will increase as fires burn less efficiently. The
Y i m_ f Mmix
Xiwet ¼ ð16:26Þ ratio increases with equivalence ratio even for
m_ f þ m_ a Mi underventilated conditions, as evidenced by
experimental data (e.g., Gottuk et al. [5]) and
0:21m_ a MO2 =Ma DO2 m_ f
XO2wet ¼ ð16:27Þ the engineering correlations presented above.
m_ f þ m_ a MO2 =Ma
The yield-equivalence ratio correlations Example 4 Consider that the piece of furniture
shown in Fig. 16.9, which are also represented described in Example 1 is burning in a room such
by Equations 16.21 and 16.15, have been that a two-layer system develops. The only ven-
replotted as CO concentration vs. equivalence tilation to the room is an open doorway through
ratio in Fig. 16.32. As indicated previously, the which 217 g/s of air is being entrained. The
yield correlations in Fig. 16.9 (and thus, the material is burning at a rate of 37 g/s, and the
Fig. 16.32 Carbon 5

monoxide concentrations
Notes:
as a function of
= Gottuk et al.
CO concentration (% vol wet)
equivalence ratio for 4 + = Beyler

hexane fires in a
compartment (o) and under
a hood (+). Data represent 3
the same tests shown in
Fig. 16.9 as unnormalized
yields 2
+++ +
++
1 ++
++
++
+
0 +
0 0.5 1.0 1.5 2.0 2.5 3.0
Equivalence ratio
Fig. 16.33 CO2 20

concentrations as a
CO2 concentration (% vol wet)

function of equivalence
ratio for hexane 15
compartment fires [5]
10
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Fig. 16.34 O2 20
concentrations as a
O2 concentration (% vol wet)
function of equivalence
ratio for hexane 15
compartment fires [5]
10
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0

average temperature of the upper layer is 700 C. X ¼ 10 ϕ p 1:25 ¼ 10ð1:5 1:25Þ ¼ 2:5

Calculate the plume equivalence ratio and deter- 0:22
mine the yield of CO and depletion of O2. Y CO ¼ tan 1 ðXÞ þ 0:11
180
0:22
Solution The plume equivalence ratio is calcu- Y CO ¼ tan 1 ð2:5Þ þ 0:11
180
lated using Equation 16.10b. The stoichiometric Y CO ¼ 0:19
fuel-to-air ratio, r, has already been calculated in
Example 1. Therefore, 0.19 g of CO are produced for every
gram of polyurethane foam that burns. The pro-
m_ f =m_ a 37=217
ϕp ¼ ¼ ¼ 1:5 duction rate of CO is equal to that yield, YCO,
r 0:1139
multiplied by the fuel burning rate (0.19 37 g/s
Because the average upper-layer temperature ¼ 7.0 g/s).
ð700 C þ 273 ¼ 973 KÞ is above 900 K, Equa- The normalized yield of oxygen is determined
tion 16.22 is used to calculate the yield of using Equation 16.24, and the recommended
CO. The argument, X, of the inverse tangent is yield coefficient, BO2 , of 0.97.
www.ebook777.com
BO2 0:97 The mass flow rate of air, ṁa, into the room is
f O2 ¼ ¼ ¼ 0:65
ϕ 1:5 estimated using the ventilation parameter [33] as
follows:
From Example 1, we obtain the maximum theo- pffiffiffi pffiffiffiffiffiffiffiffiffi
retical depletion of oxygen, kO2, and calculate the m_ a ¼ 0:5A h ¼ 0:5 1:56 2:05 ¼ 1:12 kg=s
depletion of oxygen as
Substituting ṁa into the equation above for the
DO2 ¼ f O2 kO2 ¼ 0:65 ð2:05Þ compartment equivalence ratio yields a ϕc of 1.3.
¼ 1:33 g of O2 per gram of fuel burned Since ϕc is greater than 1, the occurrence of
external burning must be considered. However,
The depletion rate of oxygen is 49.2 g/
using the criteria that ϕc is less than 1.6, it is
s (1.33 37 g/s).
assumed that no external burning will occur.
Example 5 For the piece of furniture burning in Species levels inside the room are calculated by
Example 4, calculate the CO and O2 Equations 16.19, 16.20, 16.21, 16.22, 16.23,
concentrations in the upper layer. 16.24, and 16.25 using ϕc. The yield of CO is
calculated using Equation 16.21 or 16.22,
Solution Gas concentrations can be calculated depending on the temperature of the upper
from the yields determined in Example 4 using layer. The upper-layer temperature can be
Equation 16.26 for CO and Equation 16.27 for O2. estimated using the McCaffrey, Quintiere, and
Harkleroad (MQH) correlation [34] that is
Y CO m_ f Mmix 0:19ð37Þ ð28:8Þ
XCOwet ¼ ¼ ¼ 0:028 presented in Chap. 30 of this book. According
m_ f þ m_ a MCO ð37 þ 217Þ ð28Þ
to the MQH correlation, the upper-layer gas
0:21m_ a MO2 =Ma DO2 m_ f temperatures exceed 900 K for fires above
X O2 ¼
m_ a þ m_ f MO2 =Ma 1100 kW. For temperatures above 900 K, Equa-
0:21ð217Þ 32=28:8 1:33ð37Þ tion 16.22 is used to calculate the CO yield as
¼
ð32 þ 217Þ 32=28:8
X O2 ¼ 0:005 Y CO ¼ ð0:22=180Þ tan 1 ½10ðϕ 1:25Þ
þ 0:11 ¼ 0:14
The resulting concentrations of CO and O2 are 2.8
and 0.5 % by volume, respectively.
Since there is no external burning, the CO
generated in the compartment (0.14 kg of CO
Example 6 A fire is burning in a room that has
per kg of fuel burned) will flow to other parts of
one door open and no other ventilation. The room
the building. Before dilution occurs away from
is 7 m wide and 4 m deep with a 2.43 m high
the fire compartment, the initial concentration of
ceiling. The door measures 0.76 m wide and
CO in the gases from the fire compartment can be
2.05 m high (area ¼ 1.56 m2). The peak heat
calculated using Equation 16.26:
release rate of the fire has been estimated to be
4.5 MW. Determine how much CO can be
Y CO m_ f Mmix 0:14 m_ f 28:8
transported to other rooms in the building. XCOwet ¼ ¼
m_ f þ m_ a MCO m_ f þ 1:12 28
Solution The first step is to calculate the com-
partment equivalence ratio, ϕc. Since details of Since there is no information on the contents
the fire are not provided, the mass burning rate of burning in the room, an accurate assessment of
the fuel is not known. Therefore, ϕc is estimated the fuel mass burning rate, ṁf, cannot be
via Equation 16.13c using the heat release rate, obtained. An estimate of ṁf can be made using
Q, of 4.5 MW. Equation 16.13a, with an assumed value of r, the
stoichiometric fuel-to-air ratio. Values of r are
Q 1 4, 500 kW presented in Table 16.1 for various fuels as 1/r.
ϕc ¼ ¼
m_ a 3030 m_ a 3030 In order to bound the possible CO
concentrations, values of 1/r of 4–15 are chosen reactants (stoichiometric mixture of

to represent a reasonable range of hydrocarbon fuel and oxidant streams)
fuels that may be burning in the room. The fol- Q ideal heat release rate
lowing shows an example of the ṁf calculation r stoichiometric fuel-to-air ratio
using Equation 16.13a and 1/r ¼ 4: ra stoichiometric air-to-fuel ratio
rO2 stoichiometric fuel-to-oxygen ratio
ϕ m_ a 1:3ð1:12 kg=sÞ
m_ f ¼ ¼ ¼ 0:36 kg=s T temperature
1=r 4 TSL,j adiabatic flame temperature at the
The corresponding calculation for 1/r of 15 yields stoichiometric limit for fuel species
an ṁf of 0.097 kg/s. Substituting the values for ṁf j (K)
into the above equation for XCOwet results in CO To temperature of the gas mixture prior
concentrations of 3.5 and 1.1 %, respectively. to reaction (K)
t time
tr residence time of gases in the upper
Nomenclature layer
τSS steady-state time ratio
Bi yield coefficients of species i Vul volume of the upper layer
C stoichiometric molar ratio of water to X mole fraction
carbon dioxide Xidry dry mole fraction of species i (H2O
Cj volume concentration of fuel j when removed from sample)
fuel stream is stoichiometrically Xiwet wet mole fraction of species i
mixed with oxidant stream Y yield (g/g) also refers to DO2
Cp heat capacity of products of complete YO2 , air mass fraction of oxygen in air
combustion, (kJ/g mol K) z distance between the bottom of the
DO2 mass depletion of oxygen per gram of compartment outflow and the ceiling
fuel burned (g/g) in the adjacent space
E energy released per kg of oxygen γ dimensionless layer depth in adjacent
consumed space ðγ ¼ δ=zÞ
F normalized yield or generation δ layer depth in the adjacent space
efficiency ϕ equivalence ratio
ΔHc,j heat of combustion of the species j, ϕc compartment equivalence ratio
(kJ/g mol) ϕcv equivalence ratio defined per a
j fuel species of interest specified control volume
k maximum theoretical yield ϕp plume equivalence ratio
Lf,tip length of flame tip for flame ϕul upper-layer equivalence ratio
extending down a corridor ceiling ρul density of the upper layer
M molecular weight
ma mass of air Subscripts
ṁa mass flow rate of air
mf mass of fuel A air
ṁf mass loss rate of fuel f fuel
€f
m derivative of the fuel mass loss rate CO carbon monoxide
ṁexhaust mass flow rate out of the layer O2 oxygen
n molar quantity CO2 carbon dioxide
nprod number of moles of products of com- H 2O water
plete combustion per mole of H2 hydrogen
www.ebook777.com
THC total unburned hydrocarbons 13. D. Drysdale, An Introduction to Fire Dynamics, 2nd
resid,C residual carbon ed., John Wiley and Sons, Chichester, UK (1999).
14. W.M. Pitts, 24th Symposium (International) on Com-
Xiwet wet gas concentration with water in bustion, Combustion Institute, Pittsburgh, PA (1992).
the mixture 15. D.T. Gottuk, R.J. Roby, and C.L. Beyler, “The Role of
Xidry dry gas concentration with no water in Temperature on Carbon Monoxide Production in
the mixture Compartment Fires,” Fire Safety Journal, 24, pp.
315–331 (1995).
16. A. Tewarson, “Fully Enveloped Enclosure Fires of
Wood Cribs,” 20th Symposium (International) on
Combustion, Combustion Institute, Pittsburgh, PA,
References p. 1555 (1984).
17. D.T. Gottuk, “The Generation of Carbon Monoxide in
1. R.A. Anderson, A.A. Watson, and W.A. Harland, Compartment Fires,” NBS-GCR-92-619, National
“Fire Deaths in the Glasgow Area: II The Role of Institute of Standards and Technology, Gaithersburg,
Carbon Monoxide,” Medicine, Science, & the Law, MD (1992).
21, pp. 289–294 (1981). 18. W.D. Walton and P.H. Thomas, “Estimating
2. B. Harwood and J.R. Hall, “What Kills in Fires: Temperatures in Compartment Fires,” The SPFE
Smoke Inhalation or Burns?” Fire Journal, 83, pp. Handbook of Fire Protection Engineering, National
29–34 (1989). Fire Protection Association, Quincy, MA, Ch. 2-
3. R.J. Gann, V. Babrauskas, and R.D. Peacock, “Fire 2 (1988).
Conditions for Smoke Toxicity Measurements,” Fire 19. W.M. Pitts, E.L. Johnsson, and N.P. Bryner, “Carbon
and Materials, 18, 3, pp. 193–199 (1994). Monoxide Formation in Fires by High-Temperature
4. C.L. Beyler, “Ignition and Burning of a Layer of Anaerobic Wood Pyrolysis,” presented at the 25th
Incomplete Combustion Products,” Combustion Sci- Symposium (International) on Combustion, Combus-
ence and Technology, 39, pp. 287–303 (1984). tion Institute, Pittsburgh, PA (1994).
5. D.T. Gottuk, R.J. Roby, M.J. Peatross, and 20. D. Gross and A.F. Robertson, 10th Symposium (Inter-
C.L. Beyler, “Carbon Monoxide Production in Com- national) on Combustion, Combustion Institute,
partment Fires,” Journal of Fire Protection Engineer- Pittsburgh, PA, pp. 931–942 (1965).
ing, 4, pp. 133–150 (1992). 21. B.Y. Lattimer, U. Vandsburger, and R.J. Roby, “Car-
6. N.P. Bryner, E.L. Johnsson, and W.M. Pitts, “Carbon bon Monoxide Levels in Structure Fires: Effects of
Monoxide Production in Compartment Fires— Wood in the Upper Layer of a Post-Flashover Com-
Reduced-Scale Enclosure Test Facility,” NISTIR partment Fire,” Fire Technology, 34, 4 (1998).
5568, National Institute of Standards and Technology, 22. W.M. Pitts, “The Global Equivalence Ratio Concept
Gaithersburg, MD (1995). and the Prediction of Carbon Monoxide Formation in
7. S.J. Toner, E.E. Zukoski, and T. Kubota, “Entrain- Enclosure Fires,” NIST Monograph 179, National
ment, Chemistry, and Structure of Fire Plumes,” NBS- Institute of Standards and Technology, Gaithersburg,
GCR-87-528, National Institute of Standards and MD (1994).
Technology, Gaithersburg, MD (1987). 23. N.P. Bryner, E.L. Johnsson, and W.M. Pitts, “Carbon
8. C.L. Beyler, “Major Species Production by Diffusion Monoxide Production in Compartment Fires: Full-
Flames in a Two-Layer Compartment Fire Environ- Scale Enclosure Burns,” in Proceedings of the Annual
ment,” Fire Safety Journal, 10, pp. 47–56 (1986). Conference on Fire Research, NISTIR 5499, National
9. C.L. Beyler, Fire Safety Science—Proceedings of Institute of Standards and Technology, Gaithersburg,
First International Symposium, Hemisphere, MD (1994).
Washington, DC, pp. 430–431 (1986). 24. W.M. Pitts, N.P. Bryner, and E.L. Johnsson, “Com-
10. E.E. Zukoski, S.J. Toner, J.H. Morehart, and bustion Product Formation in Under and
T. Kubota, Fire Safety Science—Proceedings of the Overventilated Full-Scale Enclosure Fires,” in
Second International Symposium, Hemisphere, Proceedings of Combustion Fundamentals and
Washington, DC, pp. 295–304 (1989). Applications, Joint Technical Meeting, San Antonio,
11. E.E. Zukoski, J.H. Morehart, T. Kubota, and TX (1995).
S.J. Toner, “Species Production and Heat Release 25. N.P. Bryner, E. L. Johnsson, and W.M. Pitts, “Scaling
Rates in Two-Layered Natural Gas Fires,” Combus- Compartment Fires—Reduced- and Full-Scale Enclo-
tion and Flame, 83, pp. 324–332 (1991). sure Burns,” in Proceedings, International Confer-
12. J.H. Morehart, E.E. Zukoski, and T. Kubota, “Species ence on Fire Research and Engineering (D.P. Lund
Produced in Fires Burning in Two-Layered and and E.A. Angell, eds.), Society of Fire Engineers,
Homogeneous Vitiated Environments,” NBS-GCR- Boston (1995).
90-585, National Institute of Standards and Technol- 26. W.M. Pitts, “An Algorithm for Estimating Carbon
ogy, Gaithersburg, MD (1990). Monoxide Formation in Enclosure Fires,” Fire Safety
Science—Proceedings of the Fifth International Sym- Compartment Fires with External Burning,” 24th
posium, International Association of Fire Safety Sci- Symposium (International) on Combustion, Combus-
ence,” pp. 535–546 (1997). tion Institute, Pittsburgh, PA, pp. 1729–1735 (1992).
27. J. Warnatz, “Rate Coefficients in the C/H/O System,” 32. B.Y. Lattimer, D.S. Ewens, U. Vandsburger, and
in Combustion Chemistry, (W.C. Gardiner, ed.), R.J. Roby, “Transport and Oxidation of Compartment
Springer-Verlag, New York, pp. 224–232 (1984). Fire Exhaust Gases in Adjacent Corridor,” Journal of
28. D.S. Ewens, “The Transport and Remote Oxidation of Fire Protection Engineering, 6, 4 (1994).
Compartment Fire Exhaust Gases,” M.S. Thesis, 33. B.Y. Lattimer, unpublished data (2000).
Virginia Polytechnic Institute and State University, 34. B.J. McCaffery, J.G. Quintiere, and M.F. Harkleroad,
Department of Mechanical Engineering, Blacksburg, “Estimating Room Fire Temperatures and the Likeli-
VA (1994). hood of Flashover Using Fire Test Data Correlations,”
29. B.Y. Lattimer, U. Vandsburger, and R.J. Roby, “The Fire Technology, 17, 2, pp. 98–119 (1981).
Transport of Carbon Monoxide from a Burning Com-
partment Located on the Side of a Hallway,” 26th
Symposium (International) on Combustion, Combus- Dr. Daniel T. Gottuk is VP of Specialty Services and the
tion Institute, Naples, Italy, pp. 1541–1547 (1996). Technical Director of Jensen Hughes. He is actively involved
30. B.Y. Lattimer, U. Vandsburger, and R.J. Roby, “The in fire hazard analyses, fire research and testing, and forensic
Transport of High Concentrations of Carbon Monox- engineering relative to fire dynamics and fire detection.
ide to Locations Remote from the Burning Compart-
ment,” NIST-GCR-97-713, U.S. Department of Brian Y. Lattimer is a Professor in Mechanical
Commerce (1997). Engineering at Virginia Tech. His research areas include
31. D.T. Gottuk, R.J. Roby, and C.L. Beyler, “A Study of fire dynamics, heat transfer from fires, and material
Carbon Monoxide and Smoke Yields from response to fires.
www.ebook777.com
Flammability Limits of Premixed

and Diffusion Flames 17
Craig Beyler
Introduction propagate are known as the upper and lower

flammable limits and are expressed as
It is well known that not all fuel/oxidant/diluent concentrations of the fuel in a specified oxidant/
mixtures can propagate flame. Normal flame- diluent mixture at a specified temperature and
type combustion cannot be sustained outside pressure. For instance, the lower flammable
certain limits definable in terms of fuel/oxidant/ limit (LFL) of methane in air at normal tempera-
diluent composition. Definition of these limits ture and pressure is 5 % by volume, and the upper
has received a great deal of attention in premixed flammable limit (UFL) is 15 % by volume.
combustion conditions, that is, in systems where As such, only methane/air mixtures with meth-
the fuel and oxidant are mixed prior to combus- ane concentrations between 5 % and 15 %
tion. Despite scientific interest in the subject methane will support propagation of flame. For
dating back to the nineteenth century, the mech- most simple hydrocarbons, the lower and upper
anism responsible for flammable limits is not yet flammable limits in air correspond to an
understood. Nonetheless, a great deal has been equivalence ratio of approximately 0.5 and
learned that has practical application. 3, respectively. The lower flammable limit
Much less investigation into the nature and concentrations for these fuels is approximately
cause of limits in diffusion flames has been 48 g/m3 (Fig. 17.1) [1].
undertaken. Empirically, clear parallels exist The most widely used method of measuring
between diffusion and premixed limits, and flammable limits was developed by the
these will be explored in the latter portion of U.S. Bureau of Mines [2]. The apparatus consists
this chapter. of a 1.5-m-long, 0.05-m-diameter vertical tube
which is filled with the fuel/oxidant/diluent mix-
ture to be tested. The top of the tube is closed,
Premixed Combustion and the base of the tube can be closed until the
start of the test to prevent diffusion of the mixture
Premixed flame fronts can only propagate within from the tube. With the base of the tube open, the
a range of compositions of fuel and oxidant. The mixture is ignited by a spark or small pilot flame
composition limits within which a flame can at the base of the tube, and the travel of the flame
front up the tube is observed. The mixture is
deemed to be within the flammable limits if the
flame can propagate halfway up the 1.5 m tube.
The test is designed to identify the range of
C. Beyler (*)
Jensen Hughes, 3610 Commerce Drive, Suite 817, mixture compositions capable of flame propaga-
Baltimore, MD 21227, USA tion remote from the ignition source.

530 C. Beyler

molecular weight on lower
Lower limit of flammability (vol %)

limits of flammability of X
5
alkanes at 25 C [1]
4 50 mg/L
3 X
2 X
X
X
X
1 X
45 mg/L X X X X
0
0 20 40 60 80 100 120 140 160 180 200
Molecular weight (g/mol)
designed for use in special conditions, and others

simply reflect national differences. Although
15 each method gives substantially similar results,
some variations in results do exist (see, for exam-
X X
ple, Smedt et al. [8] and Goethals et al. [9]).
Methane in air (vol %)
X
Mixtures are capable of combustion outside
10 the flammable limits, but external energy must be
provided throughout the mixture volume in order
to allow propagation of a flame [10]. An example
X X X
of this behavior is shown in Fig. 17.3. A small
hydrogen diffusion flame is used as a pilot source
5
in a lean methane/air mixture. At methane
concentrations less than 5 %, combustion occurs
only in the wake of the pilot flame. Above 5 %,
0
the flame can propagate away from the pilot
0 4 8 flame, regardless of the orientation of the pilot
Tube diameter (cm) flame.
Fig. 17.2 Upper and lower flammable limits of methane Flammable limits are a function of the oxygen
as determined in a vertical tube apparatus for upward and inert concentrations, as well as the mixture
propagation (circles), and for downward propagation temperature and pressure. As the concentration
(crosses) [3] of inerts is reduced and the oxygen concentration
is increased, the upper flammable limit is
The apparatus can be used with ignition at the increased, whereas the lower limit is relatively
top of the tube, but the flammable limits deter- unchanged. This result can be understood by
mined for downward propagation are narrower observing that at the lower flammable limit
than for upward propagation. The 0.05 m dia- there is always more than enough oxygen present
meter of the tube was chosen as the smallest for complete combustion, but at the upper limit
diameter at which the heat losses from the less than the stoichiometrically required oxygen
flame to the tube wall had minimal effect on the is present. Hence, at the upper limit the addi-
flammable limits determined (Fig. 17.2) [3]. Sev- tional oxygen participates in the combustion pro-
eral other methods for determining flammable cess, whereas at the lower limit the additional
limits are available [4–7]. Some methods are oxygen simply replaces inert gas.
www.ebook777.com
17 Flammability Limits of Premixed and Diffusion Flames 531
Methane
flame
H2
Air Air + 3.1% CH4 Air + 3.7% CH4 Air + 4.3% CH4 Air + 5.23% CH4
H2
Methane
flame
Air Air + 3.1% CH4 Air + 3.7% CH4 Air + 4.95% CH4 Air + 5.64% CH4
Initial flame Light blue Strong blue
Fig. 17.3 A small jet diffusion flame in a coflowing (a) The stream velocity is 0.222 m/s, and the hydrogen jet
and contraflowing (b) stream as the concentration of the diameter is 1.52 mm [10]
fuel in the stream is gradually increased up to ignition.
The lower flammable limit is also insensitive If the liquid is not enclosed fully, there will be
to the pressure, except at pressures well below a location above the surface of the liquid where
atmospheric. The upper limit shares this insensi- the fuel/air mixture will be diluted below the
tivity at subatmospheric pressures, but the upper upper flammable limit and will ignite if an igni-
limit increases with increasing pressure above tion source is present.
atmospheric (Fig. 17.4) [1].
The flammable limits widen with increases in
mixture temperature as illustrated in Fig. 17.5; Predicting Lower Flammable Limits
[1] this aspect will be discussed further later in of Mixtures of Flammable Gases
this chapter. Figure 17.5 also relates flammable (Le Chatelier’s Rule)
limits with the saturation vapor curve and the
autoignition temperature (AIT). The flashpoint Based on an empirical rule developed by Le
of a liquid is given in the figure as TL. At that Chatelier in the late nineteenth century, the
temperature, the vapor pressure at the liquid sur- lower flammable limit of mixtures of multiple
face is at the lower flammable limit. The flammable gases in air can be determined. A
corresponding upper limit temperature is given generalization of Le Chatelier’s rule was given
as TU. If a liquid is contained within a closed by Coward et al. [11]
vessel and the vapors are allowed to come into Xn
Ci
equilibrium at temperatures above the upper 1 ð17:1Þ
i¼1
LFL i
limit temperature, the vapors in the vessel will
be above the upper flammable limit, for example, where Ci is the volume percent of fuel gas, i, in
as typically occurs in an automobile gas tank. the fuel/air mixture, and LFLi is the volume
532 C. Beyler

pressure on the limits of
% of air = 100% – % natural gas
flammability of natural gas
in air at 28 C [1] 50
Natural gas (vol %)

40
30
Flammable mixtures
20
10
0
0 100 200 300 400 500 600 700 800
Initial pressure (atm)
Fig. 17.5 Effect of

temperature on limits of Saturated Upper
flammability of a vapor-air limit
combustible vapor in air at mixtures
constant initial pressure [1]
Combustible concentration
Auto-
Flammable ignition
Mist mixtures
B
A Lower
limit
TL TU AIT
Temperature
percent of fuel gas, i, at its lower flammable limit where C f i is the volume percent of fuel gas i in
in air alone. If the indicated sum is greater than the fuel gas mixture.
unity, the mixture is above the lower flammable
limit. This relationship can be restated in terms of Example 1 A mixture of 50 % methane, 25 %
the lower flammable limit concentration of the carbon monoxide, and 25 % hydrogen is mixed
fuel mixture, LFLm , as follows: with air. Calculate the lower flammmable limit of
this fuel gas mixture.
100
LFLm ¼ X
n ð17:2Þ
Solution Referring to Table 17.1, LFLs of
C f i =LFLi
i¼1 methane, carbon monoxide, and hydrogen are
www.ebook777.com
Table 17.1 Summary of limits of flammability, Lower Temperature Limits (TL), and Minimum Autoignition
Temperatures (AIT) of individual gases and vapors in air at atmospheric pressure [1]
Limits of flammability (vol %) Limits of flammability (vol %)
Combustible LFLa UFLa TL ( C) AIT ( C) Combustible LFLa UFLa TL ( C) AIT ( C)
Acetal 1.6 10 37 230 Cumene 0.88b 6.5b – 425
Acetaldehyde 4.0 60 – 175 Cyanogen 6.6 – – –
Acetic acid 5.4b – 40 465 Cycloheptane 1.1 6.7 – –
Acetic anhydride 2.7c 10d 47 390 Cyclohexane 1.3 7.8 – 245
Acetanilide 1.0e – – 545 Cyclohexanol 1.2e – – 300
Acetone 2.6 13 – 465 Cyclohexene 1.2b – – –
Acetophenone 1.1e – – 570 Cyclohexyl acetate 1.0e – – 335
Acetylacetone 1.7e – – 340 Cyclopropane 2.4 10.4 – 500
Acetyl chloride 5.0e – – 390 Cymene 0.85b 6.5b – 435
Acetylene 2.5 100 – 305 Decaborane 0.2 – – –
Acrolein 2.8 31 – 235 Decalin 0.74b 4.9b 57 250
Acrylonitrile 3.0 – 6 – n-Decane 0.75f 5.6g 46 210
Acetone- 2.2 12 – – Deuterium 4.9 75 – –
cyanohydrin
Adipic acid 1.6e – – 420 Diborane 0.8 88 – –
Aldol 2.0e – – 250 Diesel fuel (60 cetane) – – – 225
Allyl alcohol 2.5 18 22 – Diethyl amine 1.8 10 – –
Allyl amine 2.2 22 – 375 Diethyl analine 0.8e – 80 630
Allyl bromide 2.7e – – 295 1,4-Diethyl benzene 0.8b – – 430
Allyl chloride 2.9 – 32 485 Diethyl cyclohexene 0.75 – – 240
o-Aminodiphenyl 0.66 4.1 – 450 Diethyl ether 1.9 36 – 160
Ammonia 15.0 28 – – 3,3-Diethyl pentane 0.7b – – 290
n-Amyl acetate 1.0b 7.1b 25 360 Diethyl ketone 1.6 – – 450
n-Amyl alcohol 1.4b 10b 38 300 Diisobutyl carbinol 0.82b 6.1h – –
tert-Amyl alcohol 1.4e – – 435 Diisobutyl ketone 0.79b 6.2b – –
n-Amyl chloride 1.6i 8.6b – 260 2-4,Diisocyanate – – 120 –
tert-Amyl chloride 1.5 j
– 12 345 Diisopropyl ether 1.4 7.9 – –
n-Amyl ether 0.7e – – 170 Dimethyl amine 2.8 – – 400
Amyl nitrite 1.0e – – 210 2,2-Dimethyl butane 1.2 7.0 – –
n-Amyl propionate 1.0e – – 380 2,3-Dimethyl butane 1.2 7.0 – –
Amylene 1.4 8.7 – 275 Dimethyl decalin 0.69b 5.3k – 235
Aniline 1.2l 8.3l – 615 Dimethyl 3.4 – – –
dichlorosilane
Anthracene 0.65e – – 540 Dimethyl ether 3.4 27 – 350
n-Amyl nitrate 1.1 – – 195 n,n-Dimethyl 1.8b 14b 57 435
formamide
Benzene 1.3b 7.9b – 560 2,3-Dimethyl pentane 1.1 6.8 – 335
Benzyl benzoate 0.7e – – 480 2,2-Dimethyl propane 1.4 7.5 – 450
Benzyl chloride 1.2e – – 585 Dimethyl sulfide 2.2 20 – 205
Bicyclohexyl 0.65b 5.1m 74 245 Dimethyl sulfoxide – – 84 –
Biphenyl 0.70k – 110 540 Dioxane 2.0 22 – 265
2-Biphenylamine 0.8e – – 450 Dipentene 0.75m 6.1m 45 237
Bromobenzene 1.6e – – 565 Diphenylamine 0.7e – – 635
(continued)
534 C. Beyler

Butadiene (1,3) 2.0 12 – 420 Diphenyl ether 0.8e – – 620
n-Butane 1.8 8.4 72 405 Diphenyl methane 0.7e – – 485
1,3-Butandiol 1.9e – – 395 Divinyl ether 1.7 27 – –
Butene-1 1.6 10 – 385 n-Dodecane 0.60e – 74 205
Butene-2 1.7 9.7 – 325 Ethane 3.0 12.4 130 515
n-Butyl acetate 1.4i 8.0b – 425 Ethyl acetate 2.2 11 – –
n-Butyl alcohol 1.7b 12b – – Ethyl alcohol 3.3 19n – 365
sec-Butyl alcohol 1.7b 9.8b 21 405 Ethyl amine 3.5 – – 385
tert-Butyl alcohol 1.9b 9.0b 11 480 Ethyl benzene 1.0b 6.7b – 430
tert-Butyl amine 1.7b 8.9b – 380 Ethyl chloride 3.8 – – –
n-Butyl benzene 0.82b 5.8b – 410 Ethyl cyclobutane 1.2 7.7 – 210
sec-Butyl benzene 0.77b 5.8b – 420 Ethyl cyclohexane 2.0o 6.6o – 260
tert-Butyl benzene 0.77b 5.8b – 450 Ethyl cyclopentane 1.1 6.7 – 260
n-Butyl bromide 2.5b – – 265 Ethyl formate 2.8 16 – 455
Butyl cellosolve 1.1m 11h – 245 Ethyl lactate 1.5 – – 400
n-Butyl chloride 1.8 10b – – Ethyl mercaptan 2.8 18 – 300
n-Butyl formate 1.7 8.2 – – Ethyl nitrate 4.0 – – –
n-Butyl stearate 0.3e – – 355 Ethyl nitrite 3.0 50 – –
Butyric acid 2.1e – – 450 Ethyl propionate 1.8 11 – 440
α-Butryolactone 2.0m – – – Ethyl propyl ether 1.7 9 – –
Carbon disulfide 1.3 50 – 90 Ethylene 2.7 36 – 490
Carbon monoxide 12.5 74 – – Ethyleneimine 3.6 46 – 320
Chlorobenzene 1.4 – 21 640 Ethylene glycol 3.5e – – 400
m-Cresol 1.1m – – – Ethylene oxide 3.6 100 – –
Crotonaldehyde 2.1 16n – – Furfural alcohol 1.8p 16q 72 390
Gasoline 2-Monoisopropyl
100/130 1.3 7.1 – 440 biphenyl 0.53h 3.2r 141 435
115/145 1.2 7.1 – 470 Monomethylhydrazine 4 – – –
Glycerine – – – 370 Naphthalene 0.88s 5.9t – 526
n-Heptane 1.05 6.7 4 215 Nicotine 0.75b – – –
n-Hexadecane 0.43e – 126 205 Nitroethane 3.4 – 30 –
n-Hexane 1.2 7.4 26 225 Nitromethane 7.3 – 33 –
n-Hexyl alcohol 1.2b – – – 1-Nitropropane 2.2 – 34 –
n-Hexyl ether 0.6e – – 185 2-Nitropropane 2.5 – 27 –
Hydrazine 4.7 100 – – n-Nonane 0.85u – 31 205
Hydrogen 4.0 75 – 400 n-Octane 0.95 – 13 220
Hydrogen cyanide 5.6 40 – – Paraldehyde 1.3 – – –
Hydrogen sulfide 4.0 44 – – Pentaborane 0.42 – – –
Isoamyl acetate 1.1 7.0b 25 360 n-Pentane 1.4 7.8 48 260
Isoamyl alcohol 1.4 9.0b – 350 Pentamethylene glycol – – – 335
Isobutane 1.8 8.4 81 460 Phthalic anhydride 1.2l 9.2v 140 570
Isobutyl alcohol 1.7b 11b – – 3-Picoline 1.4e – – 500
Isobutyl benzene 0.82b 6.0h – 430 Pinane 0.74w 7.2w – –
Isobutyl formate 2.0 8.9 – – Propadiene 2.16 – – –
(continued)
www.ebook777.com

Isobutylene 1.8 9.6 – 465 Propane 2.1 9.5 102 450
Isopentane 1.4 – – – 1,2-Propandiol 2.5e – – 410
Isophorone 0.84 – – 460 b-Propiolactone 2.9d – – –
Isopropylacetate 1.7e – – – Propionaldehyde 2.9 17 – –
Isopropyl alcohol 2.2 – – – n-Propyl acetate 1.8 8 – –
Isopropyl biphenyl 0.6e – – 440 n-Propyl alcohol 2.2n 14b – 440
Jet fuel Propyl amine 2.0 –– – –
JP-4 1.3 8 – 240 Propyl chloride 2.4e – – –
JP-6 – – – 230 n-Propyl nitrate 1.8x 100x 21 175
Kerosene – – – 210 Propylene 2.4 11 – 460
Methane 5.0 15.0 187 540 Propylene dichloride 3.1e – – –
Methyl acetate 3.2 16 – – Propylene glycol 2.6y – – –
Methyl acetylene 1.7 – – – Propylene oxide 2.8 37 – –
Methyl alcohol 6.7 36n – 385 Pyridine 1.8n 12a – –
e
Methyl amine 4.2 – – 430 Propargyl alcohol 2.4i – – –
Methyl bromide 10 15 – – Quinoline 1.0e – – –
3-Methyl butene-1 1.5 9.1 – – Styrene 1.1z – – –
Methyl butyl ketone S51.2 8.0b – – Sulfur 2.0aa – 247 –
Methyl cellosolve 2.5x 20l – 380 p-Terphenyl 0.96e – – 535
Methyl cellosolve n-Tetradecane 0.5e – – 200
acetate 1.7m – 46 – Tetrahydrofurane 2.0 – – –
Methyl ethyl ether 2.2e – – – Tetralin 0.84b 5.0m 71 385
Methyl chloride 7e – – – 2,2,3,3-Tetramethyl
Methyl cyclohexane 1.1 6.7 – 250 pentane 0.8 – – 430
Methyl Tetramethylene glycol – – – 390
cyclopentadiene 1.3b 7.6b 49 445 Toluene 1.2b 7.1b – 480
Methyl ethyl ketone 1.9 10 – – Trichloroethane – – – 500
Methyl ethyl ketone Trichloroethylene 12bb 40a 30 420
peroxide – – 40 390 Triethyl amine 1.2 8.0 – –
Methyl formate 5.0 23 – 465 Triethylene glycol 0.9l 9.2bb – –
Methyl 1.0e – – 295 2,2,3-Trimethyl butane 1.0 – – 420
cyclohexanol
Methyl isobutyl Trimethyl amine 2.0 12 – –
carbinol 1.3e – 40 – 2,2,4-Trimethyl
Methyl isopropenyl pentane 0.95 – – 415
ketone 1.8i 9.0e – – Trimethylene glycol 1.7e – – 400
Methyl lactate 2.2b – – – Trioxane 3.2e – – –
α-Methyl 0.8e – – 530 Turpentine 0.7b – – –
naphthalene
2,Methyl pentane 1.2e – – – Unsymmetrical
Methyl propionate 2.4 13 – – dimethylhydrazine 2.0 95 – –
Methyl propyl 1.6 8.2 – – Vinyl acetate 2.6 – – –
ketone
Methyl styrene 1.0e – 49 495 Vinyl chloride 3.6 33 – –
(continued)
536 C. Beyler

Methyl vinyl ether 2.6 39 – – m-Xylene 1.1b 6.4b – 530
Methylene chloride – – – 615 o-Xylene 1.1b 6.4b – 465
Monoisopropyl p-Xylene 1.1b 6.6b – 530
bicyclohexyl 0.52 4.1r 124 230
a
T ¼ 70 C
b
T ¼ 100 C
c
T ¼ 75 C
d
T ¼ 75 C
e
Calculated
f
T ¼ 53 C
g
T ¼ 86 C
h
T ¼ 175 C
i
T ¼ 50 C
j
T ¼ 85 C
k
T ¼ 110 C
l
T ¼ 140 C
m
T ¼ 150 C
n
T ¼ 60 C
o
T ¼ 130 C
p
T ¼ 72 C
q
T ¼ 117 C
r
T ¼ 200 C
s
T ¼ 78 C
t
T ¼ 122 C
u
T ¼ 43 C
v
T ¼ 195 C
w
T ¼ 160 C
x
T ¼ 125 C
y
T ¼ 96 C
z
T ¼ 29 C
aa
T ¼ 247 C
bb
T ¼ 30 C
5.0 %, 12.5 %, and 4.0 % by volume, respec- rate per unit volume of normal alkane/air lower
tively. Using Equation 17.2 we find flammable mixtures at room temperature. Since
the heat capacity of the products of complete
100
LFLm ¼ ¼ 5:48% combustion are nearly the same for all
50=5 þ 25=12:5 þ 25=4
hydrocarbons, their observation also implies
The composition of the lower flammable limit that the adiabatic flame temperature (AFT) at
fuel/air mixture is 2.74 % methane, 1.37 % car- the lower flammable limit is a constant. Exami-
bon monoxide, 1.37 % hydrogen, and 94.5 % air. nation of a wide range of C,H,O-containing fuels
indicates that the adiabatic flame temperature at
the LFL is approximately 1600 K (150 K) for
Critical Adiabatic Flame Temperature at most C,H,O-containing fuels, with the following
the Lower Flammable Limit notable exceptions: hydrogen, 980 K; carbon
monoxide, 1300 K; and acetylene, 1280 K. This
As early as 1911, Burgess and Wheeler [12] result indicates that the adiabatic flame tempera-
noted the constancy of the potential heat release ture at the lower flammable limit is an indication
www.ebook777.com
of the reactivity of the fuel. The lower the adia- utilized to predict the effect of variable mixture
batic flame temperature, the more reactive temperature and diluents on the flammable limits
the fuel. of a mixture. Coward and Jones [2] have exam-
The utility of the concept of a critical adia- ined variable oxygen/diluent ratios, using nitro-
batic flame temperature at the lower flammable gen, carbon dioxide, water, argon, and helium as
limit goes beyond that outlined above. It has been diluents. Their work shows that the limit temper-
demonstrated that the adiabatic flame tempera- ature is insensitive to the oxygen/diluent ratio.
ture at the lower flammable limit is relatively Figure 17.6, adapted from Macek [16], illustrates
insensitive (100 K) to the diluent used and to the change in adiabatic flame temperature at the
the initial temperature of the mixture [13–15]. lower flammable limit as additional nitrogen is
The adiabatic flame temperature at the limit is added to decrease the oxygen/nitrogen ratio. The
insensitive to initial temperature only so long as figure shows an increase in the adiabatic flame
significant preflame combustion reactions do not temperature at the lower flammable limit from
occur. As such, for a mixture near or above its 1550 K to over 1700 K as we move from normal
autoignition temperature (AIT) for a significant air to the stoichiometric limit. Beyond the stoi-
length of time, the adiabatic flame temperature at chiometric limit, no fuel-lean mixture can burn.
the limit is not expected to be constant. Weinberg The region beyond the stoichiometric limit can
[15] has shown that a mixture of 1 % methane be best understood in the context of flammability
(LFL ¼ 5 % at 293 K) in air can burn if it is diagrams and upper flammable limits. We will
preheated to 1270 K, even though the flame only examine these later in the chapter.
increases that temperature by about 250 K, in The insensitivity of the limit temperature to
accordance with the expected adiabatic flame the chemical structure of C,H,O-containing fuels
temperature. This result was achieved by mixing contributes significantly to the utility of the con-
the methane and air just before the flame so that cept of a critical adiabatic flame temperature at
preflame reactions were not allowed to proceed the lower flammable limit. No systematic evalua-
significantly. tion of the limit temperature concept for fuels
Due to the constancy of the adiabatic flame containing sulfur, nitrogen, or halogens has
temperature at the lower limit, the concept can be been undertaken. Existing data indicate that
Fig. 17.6 Computed 0.19 0.18 0.17 0.16 0.14 0.13 0.12 0.11
adiabatic flame 1800
temperature along the Mol fraction O2 in mixture
lower branch of the 1700
flammability limits of
Adiabatic flame temperature (K)
propane (Adapted from 10 SL

Macek) [16]. SL and NP 1600
are defined in Fig. 17.9
1500
LL
1400
NP
1300
1200
1100
10 20 30 40
Added nitrogen (vol %)

538 C. Beyler
halogen-containing fuels have limit temperatures This equation uses concepts developed in
several hundred degrees higher than C,H,O fuels. Chap. 5. For the present purposes, it is suitable
Since halogens are combustion inhibitors, this con- to use an average value of the heat capacity. This
clusion is consistent with the idea that the adiabatic adjustment reduces Equation 17.3 to
flame temperature at the lower flammable limit is
indicative of the reactivity of the fuel. Thus, possi- LFL
ΔH c ¼ nC p T f , LFL T 0 ð17:4Þ
ble exceptions to the generalization that the adia- 100
batic flame temperature at the lower flammable
limit is approximately 1600 K may be identifiable We know that Tf ,LFL ¼ 1600 K, and for T0 ¼
by considering the reactivity of the fuel gas. 20 C, we also know that LFL ¼ 2.1 %.
Egerton and Powling [17] have shown that the Rearranging Equation 17.4 yields
limit temperatures at the upper flammable limit
ΔH c T f , LFL T 0
for hydrogen and carbon monoxide are equal to ¼
their limit temperatures at the lower flammable nC p LFL=100
1600 K 293 K
limit. Stull [18] has reported the same result for ¼
methane. However, it is not generally possible to 2:1=100
¼ 6:22 104 K
calculate the adiabatic flame temperature for
other fuels, since the products of combustion
Both the heat of combustion and the heat capac-
under fuel-rich conditions include a mixture of
ity are weak functions of temperature, and these
products of combustion and pyrolysis, which
effects will be ignored. As such we can use the
cannot be predicted by assuming chemical equi-
above expression to predict the lower flammable
librium is achieved or by detailed chemical
limit for an initial temperature of 200 C.
kinetics calculations. Equilibrium calculations
indicate that the only carbon-containing species T f , LFL T 0 1600 K 473 K
¼
that should be produced are CO, CO2, CH4, and LFL=100 LFL=100
solid carbon. This conclusion is not generally a ¼ 6:22 104 K
good approximation under fuel-rich conditions. LFL ¼ 1:8 percent
Example 2 The lower flammable limit of pro-

pane at 20 C is 2.1 % by volume. Find the lower Flammability Diagrams
flammable limit at 200 C.
Whereas the flammable limits of a fuel in air
Solution For adiabatic combustion, all the heat can be characterized by the lower and upper
released is absorbed by the products of flammable limits, it is necessary to represent
combustion: flammable limits of more general fuel/oxidant/
ð T f , LFL inert mixtures by using flammability diagrams.
LFL Examples of flammability diagrams for
ΔH c ¼ nC p dT ð17:3Þ
100 T0 methane/oxygen/nitrogen mixtures are shown in
Figs. 17.7 and 17.8. Based on an extensive series
where of tests with a range of mixture compositions, a
ΔHc ¼ Heat of combustion of the fuel flammability diagram can be constructed
LFL/100 ¼ Mole fraction of fuel indicating the regions of mixture compositions
n ¼ Number of moles of products of combustion within the flammable limits.
per mol of fuel/air mixture Two types of flammability diagrams are often
Cp ¼ Heat capacity of the products of used. The first type uses three axes in which each
combustion of the three constituent gases is explicitly
T0 ¼ Initial temperature of the fuel/air mixture represented, and the second diagram utilizes
Tf, LFL ¼ Adiabatic flame temperature of a lower only two axes in which the third gas concentration
flammable limit mixture is determined by the difference between the sum
www.ebook777.com
Fig. 17.7 Three-axis C

flammability diagram for 100 0
the system methane/
oxygen/nitrogen at 90 10
atmospheric pressure and
80 20
26 C [1]
70 30
Nit
60 +CH4 40
l%
rog
–N2 –O2
(vo
en
50 MI 50
ne
+O2
(vo
+N2
tha
l%
Me
40 –CH4 60
)
30 70
Flammable Limit
20 mixtures Min O2
80
line
10 Critical C/N 90
0 100
O 100 90 80 70 60 50 40 30 20 10 0 H
Oxygen (vol %)
Fig. 17.8 Two-axis C

flammability diagram for 100
the system methane/
oxygen/nitrogen at %O2 = 100% – % CH4 – %N2
atmospheric pressure and
80
26 C [1]
Methane (vol %)
60 +CH4
–N2
MI
+O2 +N2
Air
40 –CH4 line
Flammable
mixtures
20 Limit line
Critical C/N Min O2

N
A
0 20 40 60 80 100
Nitrogen (vol %)
of the other two gases and 100 %. Both types give mixture with an oxygen-to-nitrogen ratio less
the same information. than that of the limit line will not support flame
Shown in Figs. 17.7 and 17.8 are the air and propagation when mixed with any amount of
limit lines. Anywhere along the air line the ratio methane. This condition is known as the limiting
of oxygen to nitrogen is the same as in air. The oxygen concentration (LOC).
limit line represents a range of mixtures with a The LOC is an important concept in inerting.
fixed oxygen-to-nitrogen ratio which is tangent If the oxygen concentration can be maintained
to the flammable region. Any oxygen/nitrogen below the LOC, then premixed burning can be
540 C. Beyler
prevented. The LOC is a function of the temper- and pressure. Like flammable limits, the dynam-
ature, pressure, fuel, and inert gas. Table 17.2 ics of the LOC can generally be understood using
shows the LOC [19–23] of a wide range of the AFT concepts.
fuels with nitrogen and carbon dioxide as the As can be seen in Table 17.2 with nitrogen
inert diluents. The tabulated values apply to diluent, the LOC is generally in the 10–12 %
diluted air/fuel mixtures at normal temperature range. Fuels like carbon monoxide and hydrogen
Table 17.2 Limiting oxygen concentrations at normal temperature and pressure

Limiting oxidant concentration Limiting oxidant concentration
N2/air (volume % O2 above which CO2/air (volume % O2 above
Gas or vapor deflagration can take place) which deflagration can take place) Reference
Ethane 11 13.5 Coward and Jones [19]
Propane 11.5 14.5 Coward and Jones [19]
n-Butane 12 14.5 Coward and Jones [19]
Isobutane 12 15 Coward and Jones [19]
n-Pentane 12 14.5 Coward and Jones [19]
Isopentane 12 14.5 Jones et al. [20]
n-Hexane 12 14.5 Coward and Jones [19]
n-Heptane 11.5 14.5 Jones et al. [20]
Ethylene 10 11.5 Coward and Jones [19]
Propylene 11.5 14 Coward and Jones [19]
1-Butene 11.5 14 Coward and Jones [19]
Isobutylene 12 15 Jones et al. [20]
Butadiene 10.5 13 Coward and Jones [19]
3-Methyl-1- 11.5 14 Zabetakis [22]
butene
Benzene 11.4 14 Coward and Jones [19]
Gasoline
(73/100) 12 15 Jones et al. [20]
(100/130) 12 15 Jones et al. [20]
(115/145) 12 14.5 Jones et al. [20]
Kerosene 10 (150 C) 13 (150 C) Zabetakis and Rosen [23]
JP-1 fuel 10.5 (150 C) 14 (150 C) Jones et al. [20]
JP-3 fuel 12 14.5 Jones et al. [20]
JP-4 fuel 11.5 14.5 Jones et al. [20]
Natural gas
(Pittsburgh) 12 14.5 Coward and Jones [19]
n-Butyl chloride 14 – Kuchta et al. [21]
12 (100 C) – Kuchta et al. [21]
Methylene 19 (30 C) – Kuchta et al. [21]
chloride 17 (100 C) – Kuchta et al. [21]
Ethylene 13 – Kuchta et al. [21]
dichloride 11.5 (100 C) – Kuchta et al. [21]
1,1,1- 14 – Kuchta et al. [21]
Trichloroethane
Trichloroethylene 9 (100 C) – Kuchta et al. [21]
Acetone 11.5 14 Zabetakis [22]
n-Butanol – 16.5 (150 C) Zabetakis [22]
Carbon disulfide 5 7.5 Zabetakis [22]
(continued)
www.ebook777.com

Limiting oxidant concentration Limiting oxidant concentration
N2/air (volume % O2 above which CO2/air (volume % O2 above
Gas or vapor deflagration can take place) which deflagration can take place) Reference
Carbon monoxide 5.5 5.5 Zabetakis [22]
Ethanol 10.5 13 Zabetakis [22]
2-Ethyl butanol 9.5 (150 C) – Zabetakis [22]
Ethyl ether 10.5 13 Zabetakis [22]
Hydrogen 5 5.2 Zabetakis [22]
Hydrogen sulfide 7.5 11.5 Zabetakis [22]
Isobutyl formate 12.5 15 Zabetakis [22]
Methanol 10 12 Zabetakis [22]
Methyl acetate 11 13.5 Zabetakis [22]
Methyl ether 10.5 13 Zabetakis [22]
Methyl formate 10 12.5 Zabetakis [22]
Methyl ethyl 11 13.5 Zabetakis [22]
ketone
Data were determined by laboratory experiment conducted at atmospheric temperature and pressure. Vapor-air-inert
gas samples were placed in explosion tubes and ignited by electric spark or pilot flame
Source: Adapted from Table C-1, NFPA 69, Standard on Explosion Prevention Systems
have lower LOCs, and chlorinated fuels have 8

higher values. These trends are expected based % air = 100% – % n-hexane – % inert
on AFT concepts at the LFL. For carbon dioxide 7
as a diluent, the LOCs are generally 2–3 %
higher than for nitrogen diluent. Again this is
6
expected based on AFT concepts due to the
higher molar heat capacity of carbon dioxide.
n-Hexane vapor (vol %)
Figure 17.9 is yet another representation of 5

the flammable limits of fuel/oxidant/inert
mixtures. The dilution of a fuel/air mixture is 4
given by the percent of inert gas in excess of
the nitrogen present in air. Figure 17.9 includes N2
3
only mixtures that lie to the right of the air line,
Cst
and as such is a magnification of a portion of the
CO2
region included in Figs. 17.7 and 17.8. Also 2
NP
shown in Fig. 17.9 are several lines and points
of specific interest. The highest concentration of 1
SL
nitrogen that will allow propagation of a flame is
known as the nitrogen point (NP). Of course, this
0
concept can be generalized to any inert (IP). If 0 10 20 30 40 50
the concentration of the inert is greater than that
Added inert (vol %)
at the inert point, no mixture of fuel and oxidant
will propagate a flame remote from the ignition Fig. 17.9 Limits of flammability of various n-hexane/
source. inert gas/air mixtures at 25 C and atmospheric pressure [1]
542 C. Beyler
16 Flammability diagrams are useful not only in

% air = 100% – % methane – % inert determining the flammability of a given mix-
ture, but also in developing strategies for
14
avoiding flammable mixtures while diluting
fuel-rich mixtures. In order to make use of the
12 diagrams in this fashion, we must examine the
change in position on the diagram when fuel,
10 CO2 oxygen, or inert gas is added to the mixture.
H2O
Methane (vol %)
Consider a mixture given by point MI in the

three-axis diagram (Fig. 17.7). The arrows indi-
8 He
Cst cate the change in the mixture composition with
Flammable the addition or removal of each gas species. In
6 mixtures the three-axis diagram, moving toward the ver-
N2
tex corresponding to 100 % of any one of the
gases corresponds to the addition of that gas,
4
since adding an infinite amount of a single gas
will reduce the concentrations of the other gases
2 to zero. Adding air corresponds to moving
toward the point on the air line at which there
is no fuel. Clearly, following these examples,
0
0 10 20 30 40 50 the effect of adding any gas or gas mixture can
Added inert (vol %) be plotted in the three-axis diagram. In the
two-axis diagram in Fig. 17.8, moving toward
Fig. 17.10 Limits of flammability of various methane/ the vertex with 0 % inert, 0 % fuel corresponds
inert gas/air mixtures at 25 C and atmospheric pressure [1] to the addition of oxygen. In Fig. 17.9 moving
toward the 0 % inert, 0 % fuel vertex
As shown in Fig. 17.9, the stoichiometric line corresponds to adding air.
passes through the flammable region. The point Figure 17.10 shows the effect of various inert
at which the stoichiometric line intersects the diluents on the flammable region. As indicated
boundary of the flammable region is known as by the critical adiabatic flame temperature con-
the stoichiometric limit (SL). The SL is the most cept, the lower flammable limit is increased in
dilute stoichiometric mixture that will propagate proportion to the heat capacity of the diluent (see
a flame remote from the ignition source. In the Chap. 5).
case of methane, the peak of the flammable
region occurs near the stoichiometric limit Example 3 A methane leak fills a 200 m3 room
(Fig. 17.10). For longer chain alkanes, the peak until the methane concentration is 30 % by vol-
shifts to the rich side of the stoichiometric line ume. Calculate how much nitrogen must be
(Fig. 17.9). For C5 and higher hydrocarbons, the added to the room before air can be allowed in
peak of the flammable region is bisected by the the space.
stoichiometric line defined by combustion to CO
rather than to products of complete combustion. Solution The initial mixture in the room is given
This shift has been attributed to incomplete com- by the point B in Fig. 17.11. Adding nitrogen
bustion [16] and to preferential diffusion of moves along the line toward pure nitrogen (the N
reactants [24]. A similar shift of the maximum point). Drawing the line from the air point, A,
burning velocity to the rich side of stoichiometry tangent to the flammable region defines the mix-
is also observed. In this case, preferential diffu- ture C: the mixture with the least nitrogen added
sion of reactants has been shown to be the that, on mixing with air, will not form a flamma-
responsible factor. ble mixture. Referring to Fig. 17.11 we see that
www.ebook777.com
Fig. 17.11 Graphic C

solution of Example 3 100 0
(Adapted from
90 10
Zubetakis) [1]
80 20
70 30
+CH4
–N2 –O2
Nit
60 40
l%
rog
M1
(vo
en
50 +O2 +N2 50
ne
(vo
tha
–CH4
l%
Me
40 Air 60
)
30 B 70
Flammable
20 mixtures 80
D
10 C 90
0 100
O 100 90 80 70 60 50 40 30 A L 0 N
Oxygen (vol %)
point C corresponds to a methane concentration

of 13 %. In order to reduce the methane concen- Rearranging this equation we find
tration from 30 % to 13 %, an as yet unknown
amount of nitrogen must be added. If we could C 13
V N ¼ V ln ¼ 200 m ln
3
remove only the initial mixture and replace it C0 30
with nitrogen, the amount of nitrogen would ¼ 167 m3
simply be
Of course, the flow of gases out of the room
30 13 contains methane and may burn on mixing with
200 m3 ¼ 113 m3
30 air. Mixing air and the initial gases in the room
results in mixtures along the line AB (see
However, there is generally no way to prevent
Fig. 17.11), some of which are clearly flamma-
mixing of the initial mixture to be exhausted
ble. As such, ignition sources must be excluded
and the nitrogen being introduced to replace
near the room exhaust, or the exhaust also needs
it. As such, inerting nitrogen is also lost. We
to be inerted.
can model this occurrence by assuming that the
room is well mixed during nitrogen injection so
Example 4 A 1 kg/s flow of methane is being
that the concentrations are uniform everywhere.
dumped into the atmosphere. How much nitro-
Under these conditions the methane concentra-
gen must be mixed with methane to avoid a
tion, C, is given by
flammable mixture in the open?

V N
C ¼ C0 exp Solution In order to make the methane nonflam-
V
mable, it needs to be diluted with enough nitro-
where gen so that on further addition of air the
C0 ¼ Initial methane concentration flammable region is missed. Such a mixture of
VN ¼ Volume of nitrogen added methane and nitrogen is given by extrapolating
V ¼ Volume of the room the line AC back to zero oxygen; that is, point D
544 C. Beyler
on Fig. 17.11, where the mixture is 82 % nitro- of the flame to the outside. We have already
gen, 18 % methane. The ratio of the flow rates of discussed the first of these. Some low-power
nitrogen to methane must equal the ratio of the electrical equipment can be designed such that
concentrations of nitrogen and methane. Since the worst fault condition cannot produce the
concentrations expressed as volume percent are minimum ignition energy for a specified gas.
directly related to mole fractions, the flow rates Such equipment is termed “intrinsically safe”
of nitrogen and methane must be expressed as and may be used where there is a risk of a
molar flow rates, ṅ, flammable atmosphere being formed.
Where this method is not feasible, the electrical
n_ N2 CN2
¼ equipment may be housed in an “explosionproof”
n_ CH4 CCH4 enclosure that will not allow propagation of the
The molar flow rate of methane is given by flame out of the enclosure, which is accomplished
by making the size of the openings small enough
m_ CH4 that sufficient heat is lost by the flame as it passes
n_ CH4 ¼
MW CH4 through the opening that it is quenched. The
quenching distance is most often determined by
where MW is the molecular weight and ṁ is the placing a pair of flanged electrodes in a gas mix-
mass flow rate. ture and attempting to ignite the gases. The flanges
are parallel plates, and if the mixture can be
m_ CH4 CN2
n_ N2 ¼ ignited in the presence of the plates, the separation
MW CH4 CCH4
of the plates is greater than the quenching dis-
100 g=s 82% tance. The quenching distance with parallel
¼
16 g=mol 18% plates, d||, is 65 % of the quenching diameter in
¼ 285 mol=s circular tubes. Figure 17.12 [25] shows the rela-
tion of the quenching distance to the minimum
m_ N2 ¼ n_ N2 MW N2 ignition energy for a number of hydrocarbon/air
¼ ð285 mol=sÞð28 g=molÞ mixtures. The relation can be expressed as Emin ¼
0.06d||2, where Emin is the minimum ignition
¼ 7970 g=s or 7:97 kg=s
energy in air given in mJ and d|| is the quenching
distance in air given in mm.
Ignition Energies and Quenching Because the hot quenched flame gases in an
Diameters enclosure will expand through the opening, they
may autoignite outside the enclosure. It has been
The energy required to ignite flammable found that the minimum experimental safe gap
mixtures is generally quite low, on the order of (MESG) for most hydrocarbons is approximately
a few tenths of a millijoule (mJ) for near- half the quenching distance [25].
stoichiometric mixtures in air and as low as a
few thousandths of a millijoule in oxygen. Here
again, preferential diffusion causes the minimum Dusts and Mists
to occur for rich mixtures for fuels with molecu-
lar weights greater than that of air [24]. As the The lower flammable limit of dusts and mists
flammable limits are approached, the ignition would be expected to be higher than their gas-
energy increases sharply. eous counterparts due to the need to volatilize the
Several methods exist for preventing the initi- dust or mist. For very small particles with high
ation of an explosion. These include avoiding surface-area-to-volume ratios, the lower flamma-
flammable mixtures, excluding ignition sources ble limit is independent of particle diameter, and
whose energy is greater than the minimum igni- the limit concentrations are approximately the
tion energy, and enclosing any ignition sources in same as the analogous gaseous fuel for fuels that
an enclosure that will not allow the propagation volatilize completely. Hertzberg et al. [26] have
www.ebook777.com
20
to the particle, both the ability of heat to penetrate
the particle and the rate of heating are affected.
10 It is well known that the fraction of the material
8
6 volatilized increases with the rate of heating. It is
4 not expected that the lower flammable limit can
be reduced below 50 g/m3, even at 100 % oxygen.
2 As the particle size increases, it would be
1
expected that the lower flammable limit would
0.8 also increase due to the difficulty of getting the
Minimum ignition energy (mJ)
0.6
fuel into the gas phase where combustion will
0.4
take place. This result does in fact occur, but
0.2 depending on the geometry of the test, the appar-
ent lower flammable limit of mists can actually
0.1
0.08 decrease with increasing particle diameter due to
0.06 the effects of gravity [27]. If the ignition source
0.04 is at the bottom of the container and the aerosol is
not kept well mixed, the particles can begin to
0.02
settle out, causing the local concentration in the
0.01 lower portions of the apparatus to be higher. This
0.008
0.006 laboratory effect can also be expected to operate
0.004 under actual conditions, depending on the degree
of mixing of the aerosol.
0.002
Although it is in principle possible for flame
0.001 propagation to occur as a result of heterogeneous
0.1 0.2 0.4 1.0 2.0 4.0 10.0 20.0 combustion of particles, this appears not to be an
Quenching distance (mm)
important mechanism for organic materials.
Fig. 17.12 The relation between flat-plate quenching Lower flammable limits of anthracite coal dusts
and spark minimum ignition energies for a number of with only a 20 % volatile yield can be explained
hydrocarbon-air mixtures [25] solely on the basis of gas-phase combustion
[28]. On the other hand, flame propagation by
shown that bituminous coal dusts with particle heterogeneous combustion is important for metal
diameters of 50 mm or less and polyethylene and graphite dusts.
dusts with particle diameters of 100 μm or less
have lower flammable limits in air that are inde-
pendent of particle diameter. Figure 17.13 shows
the measured lower flammable limit concentra- Diffusion Flame Limits
tion for Pittsburgh bituminous coal as a function
of average particle diameter and oxygen concen- The limits of flammability for diffusion flames
tration. Notice that the lower flammable limit in were first examined by Simmons and Wolfhard
the small-particle limit is a function of the oxygen [29]. In their experiments, they determined the
concentration, unlike gaseous fuels. Also note minimum level of dilution of the oxidant stream
that the lower flammable limit concentration is necessary to prevent the stabilization of a diffu-
much higher than the 48 g/m3 typical of gaseous sion flame for a variety of gas and liquid fuels.
hydrocarbons. These effects are due to the fact The oxygen mole fraction, XO2 , of the oxidant
that not all the coal dust is volatilized. The frac- stream at the flammability limit is known as the
tion of dust that is volatilized is a function of the limiting oxygen index (LOI), or simply the oxy-
particle diameter and the oxygen concentration. gen index (OI). Simmons and Wolfhard’s results
As the oxygen concentration affects the maxi- are included in Table 17.3. They observed that
mum flame temperature and, hence, the heat flux the oxygen index of their diffusion flames
546 C. Beyler
Fig. 17.13 Lean Pittsburg bituminous

flammability limit data for 700
Pittsburgh bituminous coal
as a function of particle size 8-Liter chamber,
for three oxygen 600
chemical match ignition
concentrations [26]
Lean flammable limit (g/m3)

500 Key
15.5% O2
400 21% O2
X 50% O2
300
X
200
X
100 X X
X
X
0
1 2 5 10 20 50 100 200 500
Mean diameter, Ds (µm)
Table 17.3 Thermodynamic equilibrium properties at extinction (Adapted from Macek [16] and Strehlow [25])
Fuel LFL (vol %) T(LFL) (K) X(SL)a T(SL) (K) X(NP)a T(NP) (K) OIa X(OI)a T(OI) (K)
CH4 5.0 1480 0.123 1720 0.117 1610 0.139 0.130 1780
C2H2 2.7 0.085 1540
C2H4 2.7 0.105 1610
C2H6 3.0 1530 0.114 1620 0.111 1540 0.118 0.114 1630
C3H8 2.1 1540 0.125 1730 0.114 1470 0.127 0.124 1720
n-C4H10 1.8 1640 0.134 1830 0.121 1490 – – –
n-C5H12 1.4 1590 0.135 1810 0.115 1410 0.1325 0.130 1760
n-C6H14 1.2 1610 0.135 1800 0.117 1420 0.1335 0.132 1770
n-C7H16 1.05 1620 0.134 1770 0.118 1430 – – –
n-C8H18 0.90 1650 0.134 1770 0.118 1440 0.134 0.133 1780
n-C10H22 0.1345 0.133 1780
CH3COCH3 2.6 0.1285 1730
CH3OH 6.7 1550 0.112 1690 0.085 1430 0.111 0.103 1530
C2H5OH 3.3 1490 0.118 1700 0.106 1430 0.126 0.121 1670
n-C3H7OH 2.2 1490 0.128 0.124 1700
n-C4H9OH 1.7 1510 0.129 0.126 1710
n-C5H11OH 1.4 1550 0.130 0.128 1730
n-C6H13OH 1.2 1490 0.1315 0.130 1740
n-C8H17OH 0.1315 0.130 1750
C6H6 1.3 0.133 1810
C6H12 1.2 0.134 1770
H2 4.0 0.054 1080
CO 12.5 0.076 1450
a
Expressed as mole fraction of oxygen
www.ebook777.com
Fig. 17.14 Computed X

adiabatic flame X
temperatures at 1800 X
flammability limits for
n-alkanes (Adapted from X X
1750
Macek) [16] X
X
Adiabatic flame temperature (K)

1700
1650
X
1600
Stoichiometric
X
limit (premixed)
1550 Oxygen index
(diffusion)
1500 Lean limit
(premixed)
1450
0 1 2 3 4 5 6 7 8 9 10
C Atoms in alkane
equaled the ratio, XO2 =ðXO2 þ Xdiluent Þ, found in a flame is evaluated by examining the flammability
premixed stoichiometric-limit mixture involving of a premixed stoichiometric mixture of the fuel
the same fuel. This result implies that the adia- and oxidant.
batic flame temperature for the limit diffusion To do this, we assume that Le Chatelier’s rule
flame, calculated on the basis of stoichiometric holds at the stoichiometric limit; that is,
combustion of the fuel and oxidant streams, is X Ci
equal to the adiabatic flame temperature at the n
i¼1 1 ð17:5Þ
stoichiometric limit of a premixed system SLi
involving the same fuel, oxidant, and diluent. and that the adiabatic flame temperature at the
Figure 17.14 graphically illustrates the rela- stoichiometric limit for each fuel is a constant,
tionship of the adiabatic flame temperatures at leading to the expression
the lean, premixed limit in air, at the stoichiomet-
ric limit (premixed), and at the oxygen index X ðCi =100ÞΔH c, i
n
i¼1 ð T f , SL, i 1 ð17:6Þ
(premixed). As the figure shows, the adiabatic
flame temperature at the stoichiometric limit n p C p dT
T0
and the oxygen index are essentially equal, and
the adiabatic flame temperature at the lower where
flammable limit in air is approximately 150 K Ci ¼ Volume percent of fuel species, i, when the
less. Ishizuka and Tsuji [30] verified Simmons fuel stream is mixed stoichiometrically with
and Wolfhard’s results for methane and hydro- the oxidant stream
gen, and showed that the adiabatic flame temper- Tf,SL,I ¼ Adiabatic flame temperature of the stoi-
ature at the limit is the same whether dilution is chiometric limit mixture for fuel species i
of the fuel or oxidizer stream. ¼ 1700 K for most hydrocarbons
The information in Fig. 17.14 forms the basis ¼ 1450 K for carbon monoxide
of a method for the evaluation of diffusion flame ¼ 1080 K for hydrogen
limits for fuel mixtures. In essence, the ability of T0 ¼ Temperature of the stoichiometric mixture
a fuel and oxidant pair to react in a diffusion prior to reaction
548 C. Beyler
ΔHc,I ¼ Heat of combustion of fuel species All of the above results are for relatively qui-
¼620 kJ/mol for hydrocarbons (per car- escent conditions. It is well known that at higher
bon, assuming H/C ¼ 2) strain rates, the oxygen concentration at extinc-
¼283 kJ/mol for carbon monoxide tion increases. This phenomenon can most easily
¼242 kJ/mol for hydrogen be seen in counterflow diffusion flame extinction
np ¼ Number of moles of products of combus- experiments such as Hamins et al. [35]
tion per mole of reactants (stoichiometric
mixture of the fuel and oxidant streams) Example 5 As part of a hazard analysis of a
Cp ¼ Heat capacity of the products of particular room fire, the composition of the hot
combustion layer during fire development has been
This approach has been successfully used to estimated. The results of the analysis indicate
predict the flammability of the hot gas layer that the following composition represents the
formed in enclosure fires [31]. highest concentration of fuel gases expected:
Although the hot gas layer formed in enclosure Hot layer—700 K, 10 % total hydrocarbons
fires can become flammable, under some (THC), in the form of CH2, 2 % CO, 1 % H2,
conditions the oxygen concentration in the hot 15 % CO2, 2 % O2, 70 % N2
layer can cause extinction of flames fully immersed Cold layer—300 K, 21 % O2, 79 % N2
in the hot layer. Based on the analogies between Will the hot layer burn?
premixed and diffusion flames, one would expect
the oxygen concentration in the layer at extinction Solution The working equation is Equation 17.6.
to be approximately equal to the premixed LOC. In The first step is to write a balanced chemical
fact, comparing the nitrogen diluent in Table 17.2 equation for stoichiometric burning:
with Table 17.3, one can see a very close corre-
0:1 CH2 þ 0:02 CO þ 0:01 H2 þ 0:02 O2 þ 0:7 N2
spondence between the LOC and the LOI.
Morehart, Zukoski, and Kubota [32] exam- þ0:15 CO2 þ xðO2 þ 3:78 N2 Þ ! 0:27 CO2
ined the oxygen concentration at extinction of þ0:11 H2 O þ ð0:7 þ 3:78XÞN2
flames by dilution of air with combustion
products. They found that flames were We can find x by requiring that both sides of
extinguished at oxygen concentrations of this equation have the same amount of oxygen:
12.4–14.3 %, with the lower value occurring for CO O2 CO2 air CO2 H2 O
a 50-cm-diameter pool burner and the higher 0:02 0:11
value occurring for a 9 cm pool burner. These þ 0:02 þ 0:15 þ x ¼ 0:27 þ ! x ¼ 0:145
2 2
results are consistent with diesel pan fire tests
(0.62 m and 0.84 m diameters) conducted by The concentrations in the stoichiometric mixture
Peatross and Beyler [33] in which oxygen can be determined from the balanced chemical
concentrations below 14 % could not be achieved equation:
during pool burning in a compartment. It is also
ni
consistent with the results of Back et al. [34], Ci ¼ 100%
nT
who measured oxygen concentrations at extinc-
nT ¼ 0:1 þ 0:02 þ 0:01 þ 0:02 þ 0:7 þ 0:15
tion in water mist extinguishment tests in
obstructed machinery space fires. They found þ 0:145 þ 0:145ð3:78Þ ¼ 1:693

an average oxygen concentration of 14.5 % for 0:1
heptane spray fires and 13.5 % for pool fires at CTHC ¼ 100% ¼ 5:9%
1:693
extinction. Since the molar heat capacity of water
0:02
vapor is midway between nitrogen and carbon CCO ¼ 100% ¼ 1:2%
1:693
dioxide, one would expect water mist and com-
bustion product extinction limits to be between 0:01
CH 2 ¼ 100% ¼ 0:6%
nitrogen and carbon dioxide. 1:693
www.ebook777.com
Similarly, the number of moles of products per X ðCi =100ÞΔH c, i ð0:059Þ ð620Þ103
mole of reactants can be determined from the
n ¼
i¼1
C p T f , SL, i T 0 35:3ð1700 537Þ
chemical equation
ð0:012Þ ð283Þ103
½0:27 þ 0:11 þ 0:7 þ 0:145ð3:78Þ þ
np ¼ 35:3ð1450 537Þ
1:693
ð0:006Þ ð242Þ103
¼ 0:962 þ ¼ 1:07
35:3ð1080 537Þ
This result is lower than typical values of 1–1.1, ð17:7Þ
because the unknown hydrocarbon mixture is
taken as CH2 . This choice is not an error, since Since the result is greater than one, the hot layer
CH2 has been consistently used for the heat will ignite and burn.
release and heat capacity as well. For conve- Although the approach to the onset of layer
nience, we will use constant average specific burning used in Example 5 has a great deal of
heats taken from Drysdale: [3] generality, it requires a very detailed characteri-
zation of the upper and lower layers. It has been
Cp (J/mol · K) C (%)a shown by Beyler [31] that a much simpler
CO2 54.3 16.2 method can be used to evaluate the conditions
H2O 41.2 6.6 required for layer burning.
N2 32.7 77.2 The method [31] is based on the very simple
chemical model
Calculated by the same method as the fuel gas
Fuel þ Oxidizer
concentrations (
Products þ Excess oxidizer for ϕ < 1
X Ci !
n pC p ¼ n p C p, i Products þ Excess fuel for ϕ > 1
100
ð17:8Þ
¼ 0:96½ð0:162Þð54:3Þ þ ð0:066Þ
where the equivalence ratio, ϕ, is given by
ð41:2Þ þ ð0:772Þ ð32:7Þ
¼ 35:3 J=mol K m_ f
ϕ¼
m_ air r
ð17:9Þ
Notice that the average specific heat is near mf
that of nitrogen, since it is the major constituent r ¼
mair Stoichiometric
of the mixture. In calculating T0 , the initial
temperature of the mixture, we will ignore According to this model, the fuel mass frac-
variations in Cp between the hot and cold tion in the upper layer is
layers.
Yf ¼ 0 for ϕ < 1
nh Th þ nc Tc 1 1=ϕ ð17:10Þ
T0 ¼ nh þ nc ¼ nT Yf ¼ for ϕ > 1
nh þ nc 1 þ 1=ϕr
ð1Þ ð700 K Þ þ ð0:69Þ ð300 KÞ
¼ ¼ 537 K Equation 17.6 can be expressed on a mass basis
1:69 for this application as
where nh and nc are the number of moles Y f ΔH c
originating in the hot and cold layers, respec- 1 ð17:11Þ
m p C p ðT SL T 0 Þ
tively. Substituting into Equation 17.5,
550 C. Beyler
where ΔHc is the heat of combustion of the fuel, T0 is the precombustion temperature resulting
and mp is the mass of products resulting from from stoichiometric mixing of the air and fuel
burning a unit mass of upper layer gases. streams. Here, the upper layer contains the fuel
Substituting the ϕ > 1 relationship for Yf into and the lower layer contains the air. T0 can be
Equation 17.6, expressing the heat release in expressed as
terms of oxygen consumed using
T u þ Y f =r T l
ΔH O2 Y O2 T0 ¼ ð17:16Þ
ΔH c ¼ ð17:12Þ 1 þ Y f =r
r
Using Equations 17.15 and 17.16, a relationship
and recognizing that
between the critical ignition equivalence ratio and
Yf the layer temperatures can be developed. Using
mp ¼ 1 þ ð17:13Þ normal values for the semi-universal constants,
r
ΔHO2 ¼ 13:4 MJ=kg, Cp ¼ 1.1 kJ/kg K, TSL ¼
Yields 1700 K. Using air properties for the lower layer,
Y O2 ¼ 0:233 and Tl ¼ 300 K. Using a typical
1 1=ϕ ΔHO2 Y O2
1 ð17:14Þ r ¼ 0.07 yields the relationship between ϕig and
1þr C p ðT SL T 0 Þ
Tu shown in Fig. 17.15. The results shown in
Equation 17.14 can be solved for the equality Fig. 17.15 are consistent with the measurements
condition to give the equivalence ratio at which of Beyler [31], where ϕig was found to be 1.7 for Tu
layer burning begins, ϕig, of 500–600 K. Gottuk [36] found that external
burning was first observed in flashes at ϕ ¼ 1.4
k 0.4, and sustained external burning was first
ϕig ¼ ð17:15Þ
kr1 observed at f ¼ 1.9 0.3 when Tu was in the
where range 900–1100 K. While in Gottuk’s [36]
experiments it was difficult to observe burning at
ΔH O2 Y O2 the layer interface due to soot deposits on the
k¼
C p ðT SL T 0 Þ viewing ports, layer interface burning was
2.2
2.0
Ignition equivalence ratio
1.8
1.6
1.4
1.2
1.0
300 500 700 900 1100 1300
Upper layer temperature (K)
Fig. 17.15 Equivalence ratio required for upper layer constants, ΔH O2 ¼ 13:4 MJ=kg, Cp ¼ 1.1 kJ/kg K, TSL
ignition as a function of the upper layer temperature ¼ 1700 K. Using air properties for the lower layer,
determined using Equations 17.15 and 17.16 with typical Y O2 ¼ 0:233, Tl ¼ 300 K. Using a typical r ¼ 0.07 yields
properties. Using normal values for the semi-universal the relationship between ϕig and Tu
www.ebook777.com
generally observed shortly after the initiation of LFL Lower flammable limit (volume percent)
flashes in the exhaust. Because the exhaust flow M Mass (kg)
was isolated from the inflow in the experiment, n Moles
there is some issue of the availability of a pilot NP Nitrogen point
flame which does not arise in normal OI Oxygen index
two-directional vents found in most fires. Thus, r Stoichiometric fuel/air ratio
Gottuk’s work is generally consistent with SL Stoichiometric limit (volume percent)
Fig. 17.15. T Temperature (C or K)
V Volume (m3)
X Mole fraction
Oxygen Index Test Method Y Mass fraction
ΔHc Heat of combustion (J/kg)
The original oxygen index test method, used to ϕ Equivalence ratio
determine the oxygen index of liquid and gas
fuels, utilizes a counterflow diffusion flame Subscripts
formed at the stagnation region of a porous cyl-
inder or sphere through which fuel vapors are C Combustion
fed. A low-velocity oxidant stream passes over i Species
the porous body. This arrangement yields the ig Ignition
most favorable aerodynamic conditions for f Flame or fuel
flame stabilization. As such, fuel and oxidant l Lower layer
streams that can burn in the low-velocity coun- L Liquid or lower limit
terflow system may not burn under less favorable m Mixture
aerodynamic conditions characterized by higher N Nitrogen
velocities and shear. O Initial or ambient
It is also important to point out the difference p Products of combustion
between the oxygen index as measured for gas u Upper layer
and liquid fuels and the oxygen index of solids as U Upper limit
measured using a candle-type test [37, 38]. The
oxygen indexes of the gas and liquid fuels as
tested by Simmons and Wolfhard [29] were References
governed by gas-phase effects. In the American
Society for Testing and Materials test [33] for 1. M.G. Zabetakis, Bulletin No. 627, U.S. Bureau of
solids, the extinction can be caused by gas- and Mines, Washington, DC (1965).
2. H.F. Coward and G.W. Jones, Bulletin No. 503,
solid-phase effects. As such, the oxygen index of U.S. Bureau of Mines, Washington, DC (1952).
a solid fuel is not directly relevant to gas-phase 3. D.D. Drysdale, An Introduction to Fire Dynamics,
diffusion flame limits and should not be used to John Wiley and Sons, New York (1999).
calculate adiabatic flame temperature at the limit 4. ASTM E681-94, Standard Test Method for Concen-
tration Limits of Flammability of Chemicals, Ameri-
for use in the expressions presented here. can Society for Testing and Materials, Philadelphia
(1994).
5. ASTM E918-83, Standard Test Method for Concen-
tration Limits of Flammability of Chemicals, Ameri-
Nomenclature can Society for Testing and Materials, Philadelphia
(1993).
AIT Autoignition temperature (C or K) 6. DIN 51 649 Teil 1, Bestimmung der
C Concentration (volume percent) Explosionsgrenzen von Gasen and Gasgemischen in
Cp Heat capacity (J/kg K)
552 C. Beyler
Luft, Deutsches Institute fur Normung, Berlin, Flammability Characteristics of Aircraft Fuels,” Tech-
Germany (1986). nical Report 52–35, Supplement I, Wright Air Devel-
7. VDI 2263 Part 1, Test Methods for the Determination opment Center, Wright-Patterson AFB, OH (1954).
of the Safety Characteristics of Dusts, Verein 21. J.M. Kuchta, A.L. Furno, A. Bartkowiak, and
Deutscher Ingenieure (1990). G.H. Martindill, “Effect of Pressure and Temperature
8. G. Smedt, F. Corte, R. Notele, and J. Berghmans, on Flammability Limits of Chlorinated Combustibles
“Comparison of Two Standard Test Methods for in Oxygen-Nitrogen and Nitrogen Tetroxide-Nitrogen
Determining Explosion Limits of Gases at Atmo- Atmospheres,” Journal of Chemical and Engineering
spheric Conditions,” Journal of Hazardous Materials, Data, 13, 3, p. 421 (1968).
A70, pp. 105–113 (1999). 22. M.G. Zabetakis, “Flammability Characteristics of
9. M. Goethals, B. Vanderstraeten, J. Berghmans, Combustible Gases and Vapors,” Bulletin 627,
G. Smedt, S. Vliegen, and E. Van’t Oost, “Experimen- U.S. Bureau of Mines, Washington, DC (1965).
tal Study of the Flammability Limits of Toluene–Air 23. M.G. Zabetakis and B.H. Rosen, “Considerations
Mixtures at Elevated Pressure and Temperature,” Involved in Handling Kerosene,” Proceedings AP1,
Journal of Hazardous Materials, A70, pp. 99–104 37, p. 296 (1957).
(1999). 24. B. Lewis and G. Von Elbe, Combustion, Flame, and
10. G.A. Karim, I. Wierzba, M. Metwally, and K. Mohon, Explosions of Gases, Academic, New York (1961).
“Combustion of a Fuel Jet in a Stream of Lean Gas- 25. R.A. Strehlow, Combustion Fundamentals, McGraw-
eous Fuel-Air Mixtures,” in 18th Symposium (Inter- Hill, New York (1984).
national) on Combustion, Combustion Institute, 26. M. Hertzberg, K. Cashdollar, and R. Conti, “Domains
Pittsburgh, PA (1981). of Flammability and Thermal Ignitability for
11. H.F. Coward, C.W. Carpenter, and W. Payman, “The Pulverized Coals and Other Dusts: Particle Size
Dilution Limits of Inflammability of Gaseous Dependences and Microscopic Residue Analyses,”
Mixtures. Part III. The Lower Limits of Some Mixed in the 19th Symposium (International) on Combus-
Inflammable Gases with Air. Part IV. The Upper tion, Combustion Institute, Pittsburgh, PA,
Limits of Some Gases, Singly and Mixed, in Air,” pp. 717–729 (1981).
Journal of the Chemical Society, 115, pp. 27–36 27. J.H. Burgoyne and L. Cohen, “The Effect of Droplet
(1919). Size on Flame Propagation of Liquid Aerosols,”
12. M.J. Burgess and R.V. Wheeler, “The Lower Limit of Proceedings of the Royal Society A, 225, 375,
Inflammation of Mixtures of the Paraffin pp. 375–392 (1954).
Hydrocarbons with Air,” Journal of the Chemical 28. M. Hertzberg, K. Cashdollar, and C. Lazzara, “The
Society, 99, pp. 2013–2030 (1911). Limits of Flammability of Coals and Other Dusts,”
13. A.G. White, “Limits for the Propagation of Flame in 18th Symposium (International) on Combustion, Com-
Inflammable Gas–Air Mixtures. Part III. The Effects bustion Institute, Pittsburgh, PA, pp. 717–730 (1981).
of Temperature on the Limits,” Journal of the Chemi- 29. R.F. Simmons and H.G. Wolfhard, “Some Limiting
cal Society, 127, pp. 672–684 (1925). Oxygen Concentrations for Diffusion Flames in Air
14. M.G. Zabetakis, S. Lambiris, and G.S. Scott, “Flame Diluted with Nitrogen,” Combustion and Flame,
Temperatures of Limit Mixtures,” in 7th Symposium 1, pp. 155–161 (1957).
(International) on Combustion, Combustion Institute, 30. S. Ishizuka and H. Tsuji, “An Experimental Study of
Pittsburgh, PA. Effect of Inert Gas on Extinction of Laminar Diffu-
15. F.J. Weinberg, “Combustion Temperatures: The sion Flames,” 18th Symposium (International) on
Future?” Nature, 283, 239 (1971). Combustion, Combustion Institute, Pittsburgh, PA,
16. A. Macek, “Flammability Limits: A Re-Examina- pp. 695–703 (1981).
tion,” Combustion Science and Technology, 21, pp. 31. C.L. Beyler, “Ignition and Burning of a Layer of
43–52 (1979). Incomplete Combustion Products,” Combustion Sci-
17. A. Egerton and J. Powling, “The Limits of Flame ence and Technology, 39, pp. 287–303 (1984).
Propagation at Atmospheric Pressure. II. The Influ- 32. J. Morehart, E. Zukoski, and T. Kubota,
ence of Changes in the Physical Properties,” “Characteristics of Large Diffusion Flames Burning
Proceedings of the Royal Society A, 193, London, in a Vitiated Atmosphere,” in Third International
UK, pp. 190–209 (1948). Symposium on Fire Safety Science, Elsevier Science
18. D.R. Stull, Fire Research Abstracts and Reviews, Publishers, UK, pp. 575–583 (1991).
13, 161 (1971). 33. M. Peatross and C. Beyler, “Ventilation Effects on
19. H.F. Coward and G.W. Jones, “Limits of Flammabil- Compartment Fire Characterization,” in Fifth Interna-
ity of Gases and Vapors,” Bulletin 503, U.S. Bureau of tional Symposium on Fire Safety Science, Elsevier
Mines, Washington, DC (1952). Science Publishers, UK, pp. 403–414 (1997).
20. G.W. Jones, M.G. Zabetakis, J.K. Richmond, 34. G. Back, C. Beyler, R. Hansen, “A Quasi Steady-State
G.S. Scott, and A.L. Furno, “Research on the Model for Predicting Fire Suppression in Spaces
www.ebook777.com
Protected by Water Mist Systems,” Fire Safety Jour- 38. ASTM D2863-97, Standard Test Method for Measur-
nal, 35, pp. 327–362. ing the Minimum Oxygen Concentration to Support
35. A. Hamins, D. Trees, K. Seshadri, H. Chelliah, Candle-Like Combustion of Plastics (Oxygen Index),
“Extinction of Nonpremixed Flames with American Society for Testing and Materials,
Halogenated Fire Suppressants,” Combustion and Philadelphia (1997).
Flame, 99, pp. 221–230 (1994).
36. D.T. Gottuk, “The Generation of Carbon Monoxide in
Compartment Fires,” PhD Dissertation, Virginia Poly- Dr. Craig Beyler earned a Ph.D. in engineering science
technic and State University, Blacksburg, VA (1992). at Harvard University under the direction of Professor
[Also in NIST-GCR-92-619, National Institute of Howard Emmons and served on the faculty of Worcester
Standards and Technology, Gaithersburg, MD (1992).] Polytechnic Institute’s Center for Firesafety Studies.
37. C.P. Fenimore and F.J. Martin, “Flammability of Dr. Beyler is currently technical director of Hughes
Polymers,” Combustion and Flame, 10, 135 (1966). Associates, Inc., Fire Science and Engineering.
Ignition of Liquids
18
D.D. Drysdale
Introduction thermoplastics (e.g., polypropylene and polysty-

rene), first soften and liquefy before producing
The purpose of this chapter is to discuss the molecular fragments that are small enough to
ignition characteristics of combustible liquids vaporize. Others such as wood do not liquefy
that are in widespread use as fuels and solvents but release gases and vapors directly leaving
and are encountered as process fluids in the behind an involatile carbonaceous char that, if
chemical and process industries. Ignition leads permitted to do so, will undergo surface oxida-
to flaming combustion in which the fuel tion (smoldering) at a much slower rate. As a
undergoes a change of state and is converted general rule, fires involving combustible liquids
from liquid to vapor. are associated only with flaming combustion, but
Unlike the flaming combustion of solid fuels, there are exceptions that will be discussed later.
this conversion does not involve any chemical The underlying physics of the vaporization
change to the fuel molecules that simply evapo- process for liquids provides a relatively simple
rate from the exposed surface.1 The flammable key to understanding the conditions under which
vapors mix with air to burn as a diffusion flame. liquids can be ignited. The vapors from combus-
When combustible solids exhibit flaming com- tible liquids are flammable and exhibit exactly
bustion, the change of state from solid to vapor the same properties and behavior as the common
involves chemical decomposition (see Chap. 7). flammable gases such as methane and propane
Unlike liquids for which the process of evapora- (see Chap. 17). Thus, we can identify flammabil-
tion is reversible (the evolved vapors can be ity limits, autoignition temperatures, minimum
converted back to the original liquid by cooling ignition energies, quenching distances, and so
or by compression), the conversion is irrevers- on. Of these, the most important are the flamma-
ible, breaking down the large polymeric bility limits. If the concentration of vapor above a
molecules of which the solid is composed into liquid surface is below the lower flammability
fragments that are small enough to vaporize and limit, then the vapors cannot be ignited, flame
enter the gas phase. Some solids, such as the will not propagate through the vapor-air mixture,
and the liquid will not “burn.” The limiting con-
dition of the liquid at which the vapors are at the
1
There are exceptions to this generalization. High molec- lower flammability limit is known as the
ular weight liquids with high flashpoints (e.g., cooking oil,
flashpoint. Experimentally, this can be measured
flashpoint 321 C) will be undergoing some chemical
decomposition at temperatures associated with vapor in a closed cup apparatus in which the vapor-air
formation. mixture in the closed volume above the surface
D.D. Drysdale (*) (the “headspace”) is at equilibrium with the liq-
uid—the vapor will be at a pressure (the

www.ebook777.com
18 Ignition of Liquids 555
saturation vapor pressure) that is defined by the many solvents (e.g., acetone, diethyl ether, etc.),
temperature of the liquid. This “closed cup some paints and varnishes, and so on. Most are
flashpoint” provides us with a relatively simple blends, but for convenience and clarity in the
method of ranking flammable liquids according following discussion, a one-component system
to the hazard they present in everyday use. In (such as pure n-hexane) will be considered.
principle, the concept of flashpoint can also be The classic phase diagram for a
applied to combustible solids, but because the one-component system is shown schematically
phase change (solid to vapor) is irreversible in Fig. 18.1. The variables are pressure and tem-
there is no corresponding simple method for perature and the so-called “phase space” is
classifying solids according to their ignition haz- divided into three areas corresponding, respec-
ard. The ignition of solids depends on a large tively, to solid, liquid, and gas (vapor). For a pure
number of factors including the physical form compound at constant pressure (illustrated by the
of solid and the mode and intensity of the heat horizontal dashed line) we can identify the melt-
transfer process. Such issues are discussed in ing point (TM) and the boiling point (TB), which
Chap. 21. are uniquely defined at any given pressure. The
For combustible liquids, the flashpoint is values quoted in the literature refer to normal
closely linked to the flammability limits of the atmospheric pressure. The upper pair of lines
vapor. If the liquid is in an unconfined cup or that intersect at the point T in Fig. 18.1 represent
present as a pool, the minimum liquid tempera- equilibrium states between solid and liquid and
ture at which the vapors can be ignited and burn between liquid and vapor, respectively.2
is found to be higher than the “closed cup The line (TC) defines how the vapor pressure
flashpoint” as defined above and is called the of the liquid varies with temperature. Thus, for
“open cup flashpoint.” The reason for this is n-hexane at normal atmospheric pressure, TB ¼
simply that the vapors will diffuse away from 69 C, which corresponds to the temperature at
the liquid surface and for successful ignition which the (saturated) vapor pressure is
from a “pilot” (a small flame or a spark) the 101.3 kPa. The variation of boiling point with
pilot must be located in a region where the mix- pressure is best illustrated using pure water as the
ture is flammable. In general, a higher liquid example. At sea level (101.3 kPa) it has a boiling
temperature is required to ensure that the pilot point of 100 C, but, as with all other liquids, this
is in a flammable zone. However, the “flash” of point decreases with elevation. In Banff,
flame that occurs as flame propagates through the Scotland (at sea level), water boils at 100 C
flammable mixture is not necessarily followed by but in Banff, Alberta (elevation 1463 m), it
sustained burning of the liquid. A criticality must boils at about 95 C. On the summit of Mount
be exceeded before this will occur. It is only then Everest (8848 m) where the pressure is approxi-
that the liquid can properly be said to have been mately 33 kPa, or one-third of the value at sea
ignited to flaming combustion. This is known as level, it boils at about 72 C. As will be seen, the
the firepoint, which will be discussed in a later flashpoints of combustible liquids also change
section entitled “Measurement of Flashpoint and with a change of atmospheric pressure but for a
Firepoint.” subtly different reason, as will be discussed later.
On the phase diagram, temperature and pres-
sure may be varied independently provided that
Vaporization of Liquids only one phase is present: there are two degrees
The liquids of general interest to the fire protec-

tion engineer are those that are stable at normal 2
The third line in Fig. 18.1, below the intersection at T,
atmospheric temperatures and pressures (say,
represents the equilibrium states between solid and vapor.
10–30 C and 101.3 kPa). These include com- Solid converts directly to vapor by the process of subli-
mon liquid fuels (such as gasoline and kerosene), mation. It will not be considered further here.
556 D.D. Drysdale
Fig. 18.1 Typical phase Pressure

Solid
diagram for a
one-component system.
Points on the curve TC
correspond to the
equilibrium (“saturated”)
vapor pressure of the
liquid, as given in Liquid C
Equation 18.3. T is the
“triple point” and C defines Vaporization
the critical temperature and Condensation
pressure (Table 18.1)
M B
Melting
Freezing
Sublimation
Gas
Deposition
TM TB Temperature
of freedom (i.e., independent variables), which increase the temperature, the resulting vapor
in this case are temperature and pressure. Thus, pressure is defined by the line BC and (unless
a gas can be compressed and heated at the atmospheric pressure is increased in step) com-
same time and still remain a gas (no change in plete conversion of liquid to vapor will occur,
state). The ideal gas law encapsulates this in the and the number of phases present is reduced from
equation two to one as expressed in Gibbs’s phase rule:
PV ¼ nRT ð18:1Þ F¼c pþ2 ð18:2Þ
where
P ¼ Pressure where f is the number of degrees of freedom
V ¼ Volume (independent variables), c is the number of
T ¼ Temperature (K) components, and p is the number of phases pres-
n ¼ Number of moles of gas present (mass ent (e.g., see Moore [2] and Atkins and de Paula
divided by the molecular weight) [3]). For the one-component system (e.g., pure
R ¼ Ideal gas constant3 n-hexane), c ¼ 1, so that when p ¼ 2 (liquid and
However, when two phases are present and in vapor present) the number of degrees of freedom
equilibrium, corresponding to a point on one of f ¼ 1. That is, we can change either the tempera-
the lines on the phase diagram, then P and ture or the pressure, but we cannot change them
T cannot be varied independently without chang- independently without changing the number of
ing the number of phases present. For example, at phases present. (Note that the intersection of the
point B, liquid and vapor are in equilibrium, with three lines on the phase diagram marked T is
the saturated vapor pressure of n-hexane equal to known as the triple point, where the three phases
101.3 kPa (760 mmHg, or 1 bar) at 69 C. If we are in equilibrium; that is, p ¼ 3. The number of
degrees of freedom is, therefore, zero so that this
point is uniquely defined.)
3
The numerical value of R depends on the units used for P In summary, the lines that divide the phases in
and V (see Chap. 5). Fig. 18.1 represent equilibrium states: the line
www.ebook777.com
Table 18.1 Critical temperatures and pressures [1]

Normal boiling point ( C) Critical temperature ( C) Critical pressure (bar)
Hydrogen (H2) 252.9 240 13
Nitrogen (N2) 195.8 146.9 34
Oxygen (O2) 183.0 118.5 50.5
Methane (CH4) 164 82.3 46.5
Ethane (C2H6) 88.6 32.2 48.3
Propane (C3H8) 42.1 96.6 42.5
n-Butane (n-C4H10) 0.5 152.3 37
n-Hexane (n-C6H14) 69 234.5 29.9
that separates the liquid and gaseous phases condensation of vapor back to the liquid state
defines how the saturated vapor pressure of the (see Fig. 18.1). For this reason, if the liquid is
liquid varies with temperature. However, this unconfined (e.g., forming a pool in the open), the
line does not continue indefinitely but ceases at liquid will eventually undergo complete evapo-
the critical point marked C. At temperatures ration as vapor continuously diffuses away from
and pressures above the critical point, only one the surface of the liquid. Consequently, the vapor
phase exists—the distinction between the pressure at the surface will be less than the
liquid and gas disappears. Some values of critical saturated vapor pressure and equilibrium cannot
temperatures and pressures are given in be achieved. The rate of mass loss by evaporation
Table 18.1. A gas such as propane (boiling will be determined by the temperature of the
point 42 C) is below its critical temperature liquid, the exposed area of the pool, and any air
at ambient temperatures and can be liquefied by movement over the liquid surface (see, for exam-
pressurization. However, the so-called “perma- ple, Wade [4] and Clancy [5]). Boiling occurs
nent gases,” which include oxygen and nitrogen, when the vapor pressure is equal to atmospheric
are above their respective critical temperatures pressure, as discussed above.
and cannot exist as liquids at ambient tempera- However, if the liquid is in a closed (sealed)
ture (e.g., 25 C) regardless of the pressure. They container capable of withstanding high internal
are stored under pressure as gases in cylinders, pressures, the two phases (liquid and vapor) will
typically at 140 bar. A single phase then exists remain in equilibrium at temperatures well above
within the cylinder. The only way that a perma- the atmospheric boiling point. Thus, propane and
nent gas can be stored as a pressurized liquid is to butane (which have normal boiling points of
cool it below its critical temperature. Large 42.1 C and 0.5 C, respectively) can be
quantities of natural gas (mainly methane) can stored as liquids at 25 C at 9.6 bar and 2.3 bar
be stored economically as a refrigerated (cryo- (957 and 231 kPa), respectively, in appropriate
genic) liquid: its critical temperature is 82.3 C pressure vessels. These pressures correspond to
and its normal boiling point is 164 C as given the saturated vapor pressures of these two
in Table 18.1. hydrocarbons at 25 C. The reduction in volume
If a liquid is in an enclosed space, such as a associated with condensation is very large,
can, tank, or bottle, the vapor will be contained making liquefaction a particularly effective
within the headspace and quickly reach equilib- means of storing these and similar gases. They
rium (i.e., the saturated vapor pressure will be can be liquefied simply by compression,
reached). This value is predicted in the phase although this is not possible with the so-called
diagram and is a function of temperature (see permanent gases, as discussed above.
Equations 18.3, 18.4, and 18.5). It represents a Clearly, methane, propane, and butane cannot
dynamic state in which vaporization continues exist as stable liquids at normal temperatures and
but at a rate that is balanced exactly by pressures. If liquefied methane (at 163 C) is
558 D.D. Drysdale
released from a refrigerated tank and spilled on dðln p∘ Þ Lv

the ground, it will form a pool and boil ¼ ð18:3Þ
dT RT 2
vigorously until the surface of the ground has
cooled to about 163 C. Thereafter, it will where Lv is the latent heat of evaporation of the
behave as a stable liquid, evaporating at a rate liquid (kJ/kg) and R is the ideal gas constant. The
dictated by the rate of heat transfer from the derivation of this expression requires a number
ground (see Thyer [6]). of approximations and may be found in most
Although methane is much lighter than air at texts on physical chemistry [2, 3]. Integration of
ambient temperature, the vapor that evolves from the equation gives the vapor pressure as a func-
the pool will be initially at 163 C and much tion of temperature; thus,
denser than the surrounding air. Consequently, it po ¼ Cexp½Lv =RT ð18:4Þ
will spread horizontally until it gains sufficient
heat from the surroundings to regain its buoy- or
ancy. Although propane is sometimes stored as a Lv
cryogenic liquid, propane and butane are more ln p∘ ¼ lnC ð18:5Þ
RT
commonly stored in pressure vessels. Cata-
strophic release due to vessel failure gives rise A plot of ln p versus 1/T will be a line of slope
to a BLEVE (boiling liquid expanding vapor –Lv/R, although it is not strictly linear over an
explosion), a term originally coined for a pres- extended temperature range. However, it may be
sure burst of a boiler containing superheated assumed to be linear within the range of
water (see Chap. 66). (It is defined by the Centre temperatures with which we are concerned (i.e.,
for Chemical Process Safety as “an explosion we can assume that Lv is constant). Values of Lv
resulting from the failure of a vessel containing for a range of liquids are given in Table 18.2.
a liquid at a temperature significantly above its The expression for vapor pressure is normally
boiling point at normal atmospheric pressure.” given in the form shown in Equation 18.5. The
[7]) The liquid boils throughout its volume once 53rd edition of the CRC Handbook of Chemistry
the pressure is released and a substantial quantity and Physics [13] (and perhaps some later
will convert to vapor. The heat of vaporization is editions) gives an extensive table of data on p
taken from the remaining liquid so that the (T), but in a modified form as follows:
BLEVE produces a vapor cloud containing a
A
significant proportion of the original mass as log10 po ¼ 0:2185 þB ð18:6Þ
T
liquid droplets. These may fall to the ground,
although if ignition occurs (as it will if the pres- where p is given in mmHg. Values of A and B for
sure burst has been the result of exposure of the some typical liquid fuels are given in Table 18.3
vessel to fire), there will be a fireball that will (converting the data from log10 to loge [i.e., ln] and
burn out rapidly (see, for example, Abbassi and from mmHg to kPa is a hazardous process that has
Abbassi [8]). not been attempted here). Vapor pressures may
also be calculated from data in Yaws [12].
Calculation of Vapor Pressure Example 1 Using the data in Table 18.3, calcu-
late the pressure in a cylinder containing liquid
If the space above the liquid is enclosed (as in a isobutane at 25 C. How can you determine how
bottle or other container), evaporation will take much fuel remains in the cylinder after drawing
place until the vapor pressure reaches its satura- gas from it for a period of time?
tion value. This equilibrium is described by a
form of the Clapeyron-Clausius equation, which Solution For isobutane, A ¼ 5416.2 K and
gives the saturated vapor pressure ( p ) as a func- B ¼ 7.349085. T ¼ 25 + 273 K ¼ 298 K.
tion of the temperature of liquid (T K). Substituting these values in Equation 18.4 gives
www.ebook777.com
Table 18.2 Selected ignition properties of some fuels in aira
Flammability
limitsb (%
Molecular Boiling Lv H Flashpoint ( C) by volume) AIT

Fuel Formula weight point ( C) (kJ/kg) (MJ/kg) Closed Open Lower Upper ( C)
Alkanes
Methane CH4 16 162 509 50.2 – – 5.0 15.0 600
Ethane C2H6 30 89 489 47.6 – 135 3.0 12.5 515
Propane C3H8 44 42 426 46.4 – 104 2.1 9.5 450
n-Butane C4H10 58 0 386 45.9 – 60 1.8 8.5 405
i-Butane – – 10 366 – 117 – 1.8 8.4 460
n-Pentane C5H12 72 36 365 45.5 – 49 1.4 7.8 260
i-Pentane – – 13 371 – – 51 1.4 7.6 420
n-Hexane C6H14 86 69 365 45.2 22 – 1.2 7.5 234
i-Hexane – – – – – 29 – 1.2 7.0 –
n-Heptane C7H16 100 98 365 45.0 4 – 1.0 7.0 223
i-Heptane – – – – 18 – 1.0 6.0 –
n-Octane C8H18 114 125 298 44.9 13 – 0.8 6.5 220
i-Octane – – – – – 12 – 1.0 5.6 –
n-Nonane C8H20 128 151 288 44.8 31 – 0.7 5.6 206
n-Decane C10H22 142 174 360 44.7 44 – 0.75 5.4 208
n-Undecane C11H24 156 196 308 44.6 – 65 0.7 4.8 202
n-Dodecane C12H26 170 216 293 44.6 72 – 0.60 4.7 204
Kerosene{ ~C14H30 ~198 ~232 ~291 ~44.0a ~49 – (~0.6) (~5.6) ~260
Alkenes
Ethylene C2H4 29 104 516 47.3 121 – 2.7 36 450
Propene C3H6 42 48 437 45.9 108 – 2.0 11.0 457
1-Butene C4H8 56 6 398a 45.4 80 – 1.6 9.3 384
1-Pentene C5H10 70 30 314 46.9 – 1.8 1.5 8.7 273
Hexelene C6H12 84 67 388 47.5 – – – – 253
Cycloparaffins
Cyclopropane C3H6 42 34 588 46.3 95 – 2.4 10.4 498
Cyclobutane C4H8 56 13 483 44.8 65 – 1.8 11.1 427
Cyclopentane C5H10 70 49 443 44.3 37 – 1.4 9.4 361
Cyclohexane C6H12 84 81 358 43.9 20 – 1.3 8.0 260
Cycloheptane C7H14 98 119 376 43.7 9a – 1.1 7.1 –
Dimethyl C8H16 112 119 300 46.3a 11 – 0.9 6.5 304
cyclohexane
Aromatics
Benzene C6H6 78 80 432 40.7 11 – 1.4 7.1 562
Toluene C7H8 92 110 362 41.0 4 7 1.2 7.1 536
m-Xylene C8H10 106 139 343 41.3 25 – 1.1 7.0 528
o-Xylene – – 141 347 41.3 17 24 1.0 6.0 464
p-Xylene – – 137 339 41.3 25 – 1.1 7.0 529
Styrene C8H8 104 145 – 40.5 32 – 1.1 6.1 490
bi-Phenyl C12H10 154 254 – 40.6 113 124 0.8 6.7 540
Naphthalene C10H8 128 218 316c 40.3 79 88 0.9 5.9 587
Anthracene C13H10 166 340 310c 40.0c 121 196 0.7 – 540
Ethyl benzene C8H10 106 136 320c 43.1 15 24 1.0 6.7 432
Butyl benzene C10H14 134 173 277c 43.7 49 63 0.8 5.8 412
(continued)
560 D.D. Drysdale
Flammability
limitsb (%

Alcohols
Methanol CH3OH 32 64 1101 20.8 12 16 7.3 36.0 469
Ethanol C2H5OH 46 78 837 27.8 13 22 4.3 19.0 423
n-Propanol C3H7OH 60 97 686 31.3 15 29 2.0 12.0 371
i-Propanol – – 82 667 33.1 12 – 2.0 12.6 399
Allyl alcohol C3H6O 58 95 684 31.9 21 24 2.5 18.0 378
n-Butanol C4H9OH 74 117 621 36.1 29 43 1.4 11.2 343
i-Butanol – – 107 578 36.1 28 – 1.7 9.8 406
2-Pentanol C5H11OH 88 119 575c – – 41 1.5 9.7 343
i-Amyl alcohol C5H11OH 88 130 501 35.3 43 46 1.2 9.0 350
3-Pentanol 118 575c – 34 39 1.2 9.0 435
n-Hexanol C6H13OH 102 159 458 36.4 45 74 1.2a 8.2 285
Cyclohexanol – – 161 460c 36.6 68 – 1.2 9.3 300
n-Heptanol C7H15OH 116 176 439 39.8 – 71 1.0 7.2 –
1n-Octanol C8H17OH 130 196 408 40.6 81 – 0.9 6.4 282
2n-Octanol 180 419 – 74 82 0.8 6.5 –
Nonanol C9H19OH 144 214 403 40.3 – – 0.8 6.1 –
i-Decanol C10H21OH 158 235 373 – – – 0.7 5.5 –
Carbonyls
Formaldehyde CH2O 30 97 826 18.7 93 – 7.0 73.0 430
37 % in H2O – – 97 826c – 54 93 (7.0c) – 424
Acetaldehyde C2H4O 44 21 570 25.1 38 – 1.6 10.4 185
i-Butyraldehyde C4H8O 72 61 444c 33.8 40 24 1.6 10.6 254
Crotonaldehyde C4H6O 70 102 490c 34.8 13 – 2.1 15.5 232
Diethyl C4H12O 76 118 500c – 294 – – – –
acetaldehyde
Ethyl C8H16O 128 163 325c 39.4 – 52 – – –
hexaldehyde
Paraldehyde C6H12O3 132 124 328 – 17 36 1.3 16.2 238
Salicyl aldehyde C7H6O2 122 196 396 – 78 – 1.4 8.4 –
Benzaldehyde C7H6O 106 179 362 – 64 74 1.4 7.8 192
Ketones
Acetone C3H6O 58 56 521 29.1 18 9 2.6 128 538
2-Butanone C4H8O 72 80 443 33.8 2 1 1.8 10.0 516
Diethyl ketone C5H10O 86 101 380 33.7 – 13 1.5 8.0 452
Methyl i-butyl C6H12O 100 116 345c 35.2 23 24 1.4 7.5 454
ketone
Dipropyl ketone C7H14O 114 144 317 38.6 – – 1.1 7.0 –
Methyl n-propyl C5H10O 86 375 376c 33.7 7 16 1.5 8.2 452
ketone
Methyl vinyl C4H6O 70 81 440c – 7c – – – –
ketone
Acids
Formic acid CH2O2 46 101 502 5.7 69 – 18.0 57.0 601
Acetic acid C2H4O2 60 118 405 14.6 40 57 5.4 16.0 427
Benzoic acid C7H6O2 122 250 s 270c 24.4 121 – 1.4 8.0 574
(continued)
www.ebook777.com
Flammability
limitsb (%

Miscellany
Camphor C10H16O 152 204 s 265c 38.8 66 93 0.6 3.5 466
Carbon disulfide CS2 76 47 – 13.6 30 – 1.3 50.0 90
m-Creosol C7H8O 108 203 – 34.6 86 – 1.1 7.6 559
o-Creosol – – 191 – 34.1 81 – 1.4 7.6 599
p-Creosol – – 202 – 34.1 86 – 1.1 7.6 559
Furan C4H4O 68 31 399 – 35 – 2.3 14.3 –
Pyridine C5H5N 79 114 449 35.0 20 – 1.8 12.4 482
Aniline C6H7N 93 183 434 36.5 76 91 1.3 11.0 617
Acetal C6H14O2 118 103 277 31.8 21 – 1.6 10.4 230
p-Cymene C10H14 134 176 283 43.9 47 63 0.7 5.6 436
o-Dichloro C6H4Cl2 146 180 – 19.3 66 74 2.2 9.2 647
benzene
1,1-Dichloro C2H2Cl2 96 37 – – – 10 7.3 16.0 582
ethylene
1,2-Dichloro – – 61 – – 6 – 5.6 12.8 460
ethylene
Monochloro C6H5Cl 112 132 – – 32 38 1.3 7.1 638
benzene
Resorcinol C6H6O2 110 276 – 26.0 127 – 1.6 9.8 567
Ethyl formate C3H6O2 74 54 – 22.5 20 12 2.7 13.5 455
Ethyl acetate C4H8O2 88 77 – 25.9 4 1 2.2 11.4 427
Methyl C4H8O3 104 80 – 22.2 2 – (2.4) (13.0) 469
propionate
Acrolein C3H4O 56 53 – 29.1 – 26 2.8 31.0 234
Acrylonitrile C3H3N 53 77 – 24.5 – 0 2.4 17.3 481
n-Amyl acetate C7H14O2 130 149 – 33.5 24 27 1.1 6.8 357
1-Amyl acetate – – 153 – – 25 38 1.0 7.5 360
1, 3-Butadiene C4H6 54 4 – – 76 – 2.0 11.5 429
n-Butyl acetate C6H12O2 116 127 – 30.0 22 32 1.7 7.6 421
n-Butyl ether C8H18O 130 141 – 39.7 25 38 1.5 7.6 –
Dimethyl ether C2H6O 46 24 – 31.6 41 – 3.4 18.0 350
Divinyl ether C4H4O 70 39 – – 30 – (1.7) (27) 360
Diethyl ether C4H10O 74 35 – 37.4 45 – 1.9 48 180
Gasoline{ – – ~33 – ~44.1 ~45 – (~1.4) (~6.8) ~371
Naptha{ – – ~177 – – ~41 – (~0.8) (~5.0) ~246
Petroleum ether{ – – ~78 – – ~ 18 – (~1.4) (~5.9) ~288
Note: s indicates sublimes at normal pressures; Lv is latent heat of evaporation; H is heat of combustion; ~ indicates
approximate values; – indicates not available; { indicates liquid blend
a
Adapted from Kanury [9]. The data were originally from references International Critical Tables of Numerical Data
[10] and Handbook of Industrial Loss Prevention [11] but the flammability limits and autoignition temperatures have
been taken from Yaws [12]. It is not clear whether the flashpoint measurements quoted in Yaws [12] refer to the closed
cup or the open cup tests so these have not been adopted
b
The figures in brackets are taken from Kanury [9]
c
Estimated value
562 D.D. Drysdale
Table 18.3 Calculation of saturated vapor pressures 760 mmHg, or 5.7 mmHg. The calculation is
A (K) B () exactly the same as for the previous example,
Methane (CH4) 2128.8 7.027729 except that log10( p ) ¼ log10(5.7) ¼ 0.756. Thus
Propane (C3H8) 4811.8 7.392262
10, 912:0
n-Butane (n-C4H10) 5801.2 7.492753 0:756 ¼ 0:2185 þ 8:24809
i-Butane (i-C4H10) 5416.2 7.349085
T
n-Pentane (C5H12) 6595.1 7.489673 7:4922T ¼ 2384:272
n-Hexane (C6H14) 7627.2 7.717119
n-Heptane (C7H16) 8928.8 8.258500
T ¼ 318:2 K ¼ 45:2∘ C
i-Octane (C8H18) 9086.6 8.113870
The closed cup flashpoint of n-decane is given
n-Decane (C10H22) 10,912.0 8.248089
n-Dodecane (C12H26) 11,857.7 8.150997
in Table 18.2 as 317 K, or 44 C. Sources of the
difference between the calculated flashpoint
Data for Equation 18.5 [13]
and that measured in a standard test will be
discussed below.
log10( po) ¼ 3.378, or po ¼ 2386.7 mmHg, or
3.14 bar. If the temperature remains 25 C, this
pressure will remain unchanged for as long as
Vapor Pressure of Liquid Blends
there is any liquid isobutane left in the container.
Pressure is no guide to the amount of isobutane
Most commonly encountered fuels are in fact
remaining. The only way to determine how much
blends of different compounds. Gasoline, for
liquid is left is to weigh the container, assuming
example, contains several hundred individual
that you know the tare.
hydrocarbons including a significant proportion
of aromatics. The total vapor pressure is the sum
Example 2 Calculate the normal boiling point of
of the partial vapor pressures of the individual
n-hexane from the data in Table 18.3, assuming
components, which in turn depend on the con-
the atmospheric pressure is 760 mmHg.
centration of the individual components in the
blend. To illustrate how the vapor pressures of
Solution Take po ¼ 760 mmHg, so that
the components may be calculated, consider a
log10(760) ¼ 2.881. For n-hexane, A ¼ 7627.2 K
mixture of two hydrocarbon liquids, A and B.
and B ¼ 7.717119, so by substitution and
At a given temperature, the partial vapor
rearrangement:
pressures of components A and B are given by
7627:2 Raoult’s Law [2, 3]:
2:881 ¼ 0:2185 þ 7:717119
T
pA ¼ xA pAo pB ¼ xB pBo ð18:7Þ
4:836T ¼ 1666:543
where xA and xB are the mole fractions of A and
T ¼ 344 K ¼ 71:6∘ C
B, respectively, given by
which is about 2 K higher than the measured nA nB
value quoted in Table 18.2. xA ¼ xB ¼ ð18:8Þ
nA þ nB nA þ nB
Example 3 Calculate the temperature at which and nA and nB are the number of moles of A and
the vapor pressure of n-decane corresponds to the B present (i.e., the mass of the component present
lower flammability limit for n-decane vapor. divided by its molecular weight).
Assume that this vapor pressure is 0.75 % by Suppose that A and B are n-hexane and
volume (Table 18.3) and that the atmospheric n-decane, respectively, and the mixture is at a
pressure is 760 mmHg. temperature of 25 C. These hydrocarbons form
an “ideal mixture” in that the molecules of
Solution The vapor pressure corresponding to the A and B are so similar that they do not interact
lower flammability limit of n-hexane is 0.75 % of with one another, either physically or chemically
www.ebook777.com
(i.e., the interactions between A and B are no where the molecular weights are MWA ¼ 86
different from the interactions between A and A, and MWB ¼ 142. According to Equation 18.6
or B and B). and Table 18.3, the partial pressures of
As an example, consider a mixture containing n-hexane and n-decane are 10.66 mmHg and
5 % hexane (by mass) in n-decane. Would its 1.65 mmHg, respectively. Using Equation 18.8
flashpoint be above or below 25 C? (The with Ln-hexane ¼ 1.2 % and Ln-decane ¼ 0.75 %
flashpoints of n-hexane and n-decane are (see Table 18.2),
29 C and 44 C, respectively.) This can be
ascertained by calculating the partial vapor 10:66=760 1:65=760
þ ¼ 1:46 ð18:11Þ
pressures of the two components at 25 C and 0:012 0:0075
using Le Chatelier’s Principle (see Chap. 17) to
discover if the total vapor pressure ( ptotal ¼ indicating that the mixture is above the lower
pn-hexane pn-decane) is above or below the flammability limit at 25 C (i.e., the flashpoint
lower flammability limit. The principle states of this mixture is below 25 C).
that a mixture of flammable vapors in air will
be at the lower flammability limit if Example 4 Determine by calculation whether n-
X li decane containing 1 % n-pentane (by volume)
¼1 ð18:9Þ would be classified as a Class 1C or a Class II
i
Li
flammable liquid according to the NFPA
where li is the percentage composition (molar Standard. [14] (This is equivalent to posing the
proportion) of component i in the vapor-air mix- question, “Is the flashpoint above or below
ture and Li is the corresponding value for the 37.8 C?”)
lower flammability limit of component i. To cal-
culate the equilibrium partial vapor pressures of Solution This calculation is identical to that
n-hexane and n-decane above a 5 % hexane/95 % discussed previously, but the densities of the
n-decane mixture (by mass), the respective mole two liquids must be taken into account and the
fractions must be calculated; thus, calculation carried out at 37.8 C. The mixture
can be taken as 0.01 626 kg of n-pentane +
0:05=MW A 0.99 730 kg of n-decane (where the densities
xnhexane ¼
0:05=MW A þ 0:95=MW B of n-pentane and n-decane are 626 kg/m3
ð18:10Þ
0:95=MW B and 730 kg/m3, respectively). The mole
xndecane ¼
0:05=MW A þ 0:95=MW B fractions are
ð0:01 626Þ=MW A
xnpentane ¼
ð0:01 626Þ=MW A þ ð0:99 730Þ=MW B
ð18:12Þ
ð0:99 730Þ=MW B
xndecane ¼
ð0:01 626Þ=MW A þ ð0:99 730Þ=MW B
where now MWA ¼ 72 (the molecular weight of 10.05 mmHg and 3.72 mmHg. Applying the Le
pentane) and MWB ¼ 142. These give xn-pentane Chatelier Principle (with the lower flammability
¼ 0.0141 and xn-decane ¼ 0.986. From Equa- limit of n-pentane vapor as 1.4 %),
tion 18.6 and Table 18.3, the saturated vapor X li 10:05=760 3:72=760
pressures of n-pentane and n-decane at 37.8 C ¼ þ ¼ 1:6
L i 0:014 0:0075
are 713.11 mmHg and 3.773 mmHg, respec- i
tively, so that the partial pressures are ð18:13Þ

564 D.D. Drysdale
This is above the lower flammability limit and, Table 18.4 Examples of data for the Van Laar equation
consequently, the mixture has a flashpoint below for binary (two-component) systems [15]
37.8 C and is definitely not a Class II liquid. Component A Component B CA CB
(Further calculation could be carried out to ascer- Ethanol Water 0.67 0.42
tain if the mixture is Class IB or IC; see below.) Methanol Water 0.25 0.20
This calculation reveals that the partial vapor Acetone Water 0.89 0.65
pressure of the more volatile component can be n-Heptane CCl4 0.2164 0.0618
disproportionately high and for this reason it will
evaporate from the mixture much more rapidly γ ¼ 1). For a two-component mixture of A and
than the less volatile component. Consequently, B, the activity coefficients are given by the Van
care must be taken when determining the Laar equations:
flashpoints of such mixtures. The liquid to be
tested should be kept in a closed container and CA
log10 γA ¼
a sample transferred to the flashpoint apparatus ½1 þ ðCA xA =CB xB Þ2
CB ð18:16Þ
as quickly as possible to minimize evaporative log10 γB ¼
loss. In some circumstances, it might be wise to ½1 þ ðCB xB =CA xA Þ2
refrigerate the liquid and chill the apparatus. The
author has experience of assessing the flashpoint Essentially the same set of calculations
of a sample of crude oil that (without refrigera- can be carried out to establish the flammability
tion) gave a flashpoint of 28 C, but a flashpoint properties of nonideal mixtures, but the activity
of 15 C occurred if the liquid (and the apparatus) coefficients (Equations 18.15 and 18.16) must be
was cooled to 0 C before opening the sample calculated from Equations 18.16 using data such
container. (The problem of evaporative loss is as those contained in Table 18.4.
also encountered in the more extreme example Another more general data set than that given
of trying to identify traces of gasoline or other in Table 18.4 is given by Babrauskas [16].
flammable liquids that may have been used in an
arson attack.)
In the examples discussed above, the vapor Effect of Atmospheric Pressure
pressure of liquid mixtures was calculated using on Flashpoint
Raoult’s Law (Equation 18.7), which applies
only to ideal mixtures such as blends of The calculations that are provided above all refer
hydrocarbons. It is important to note that many to the standard atmosphere at sea level where the
other liquid mixtures, such as alcohol and water, pressure is 101.3 kPa (760 mmHg), convention-
are not ideal as there is some interaction between ally normalized as 1 bar. If the atmospheric pres-
the molecules of the different components (A and sure changes, this change has no significant
B). Instead of Equation 18.7, it is necessary to use effect on the vapor pressure, which is a function
Equation 18.14: of the temperature of the liquid. At a constant
temperature but a reduced pressure, the vapor-air
pA ¼ αA pAo pB ¼ αB pBo ð18:14Þ ratio in the headspace will be increased (i.e., it
will become richer in fuel). This has significant
where αA is known as the activity of component consequences for liquid fuels because it will
A in the mixture, and pAo is the saturation vapor reduce the flashpoint.
pressure of pure A, and so on. The activity coeffi- Consider the following argument. In Example
cient αA is the product of the mole fraction of 3, the temperature at which the saturated vapor
A (Equation 18.15) and the activity coefficient γA: pressure of n-decane corresponds to the lower
αA ¼ γA nA αB ¼ γB nB ð18:15Þ flammability limit was shown by calculation to
be 45.2 C, which compares well with the
where γA is the activity coefficient of component measured value of the closed cup flashpoint
A in the mixture (note that for a pure liquid, (44 C). At 45.2 C, the vapor pressure was
www.ebook777.com
assumed to be 5.7 mmHg, which is 0.75 % of 35–63 C [19], as measured at sea level. As an
normal atmospheric pressure (760 mmHg). If aircraft gains altitude after takeoff, the air pres-
the temperature remains the same (45.2 C) but sure in the headspace will fall relatively rapidly,
the pressure is reduced—say to the value appro- while the fuel will cool rather slowly. There is
priate to Denver, Colorado (at 1 mile high, the potential for the vapor-air mixture in the
631 mmHg)—then the volumetric concentration headspace to become flammable. On long-haul
of n-decane vapor in air becomes 5.7/631 ¼ flights, of course, the hazard will be relatively
0.009, or 0.9 %. It has been shown that the short-lived as the fuel loses heat and cools to
lower flammability limit is remarkably insensi- below the local flashpoint, relevant to the
tive to a reduction in pressure until it falls below pressure at cruising altitude. This phenomenon
200–300 mmHg (27–40 kPa) [17, 18]. Clearly, at is discussed in NFPA’s Fire Protection Hand-
45 C the saturated vapor pressure of n-decane is book [19].
above the lower flammability limit. The effect on
the flashpoint can be shown in the following
example. Measurement of Flashpoint
and Firepoint
EXAMPLE 5 Calculate the flashpoint of
n-decane if measured in Denver, Colorado, There are a number of standard tests available for
where the atmospheric pressure is 631 mmHg. measuring the closed cup [20, 21] and open cup
Assume that the lower flammability limit of [22, 23] flashpoints (Fig. 18.2). The former mea-
n-decane vapor is 0.75 %. surement is directly related to the lower flamma-
bility limit of the fuel vapor and is used to
SOLUTION The vapor pressure corresponding classify liquids according to their ignition hazard
to the lower flammability limit of n-hexane is [18]. Its relationship to equilibrium vapor pres-
0.75 % of 631 mmHg, or 4.73 mmHg. The sure of the liquid is discussed in an earlier
calculation is exactly the same as in Example 3, section.
except that log10( p ) ¼ log10(4.73) ¼ 0.675.
Thus,
Closed Cup Flashpoints
10, 912:0
0:075 ¼ 0:2185 þ 8:24809
T
In the closed cup test, such as the Pensky-
7:573T ¼ 2384:272 Martens apparatus [20] and the Tag tester [21],
T ¼ 314:8 K ¼ 41:8∘ C the flammability of the saturated (equilibrium)
vapor-air mixture in the space above the liquid
The value obtained in Example 3 at normal surface (i.e., the headspace) is tested by
atmospheric pressure was 45.2 C. The differ- introducing a small pilot flame (see Fig. 18.2).
ence is not insignificant and could be very impor- The apparatus is designed to allow the miniature
tant for liquids close to the boundary between explosion within the headspace to vent through
two classifications (see later discussion). The an aperture that is opened to admit the pilot
issue becomes more significant at higher ignition source, which also allows the “flash” of
altitudes such as Mexico City (2240 m) and flame to be observed. The procedure involves
Lhasa in Tibet (3650 m). In these cities, the raising the temperature of the liquid slowly
flashpoint of n-decane would be approximately from approximately 10–20 K below the
39.4 C and 35.9 C, respectively. anticipated flashpoint at a rate of 5–6 K/min,
An interesting consequence of this relates to introducing the ignition source at intervals
the headspace in the fuel tanks of aircraft. The corresponding to about a 1 C (1 K) temperature
kerosene grades of commercial aviation fuel rise. The slow rate of heating is intended to allow
have closed cup flashpoints in the range of enough time for equilibrium conditions to be
566 D.D. Drysdale
Fig. 18.2 Four of the

commonly used Flame tip
apparatuses for
determining flashpoints of Test flame
Sample
flammable or combustible applicator device
liquids [19]
Sample
Bath Bath
Tag closed cup Tag open cup

ASTM D56 ASTM D1310
Stirrer
Test flame
applicator device
Test flame
applicator
device
Test cup
Test cup
Cleveland open cup Pensky-Martens closed cup

ASTM D92 ASTM D93
reached within the headspace (see below). The Corrected flashpoint ¼ T 0:033 ð760 PÞ
lowest temperature at which a flash of flame is
ð18:17Þ
observed is recorded as the closed cup flashpoint.
It is expected that it can be determined to an where T is the measured flashpoint ( C) and P is
accuracy of better than 1 C for liquids with the ambient atmospheric (barometric) pressure
flashpoints below 100 C. Values of the closed (mmHg). This is intended for relatively small
cup flashpoint for a range of liquids are given in excursions that are commonly experienced on a
Table 18.2. All refer to standard atmospheric day-to-day basis. No guidance is given that is
pressure (101.3 kPa). If the closed cup flashpoint relevant to high-altitude locations.
is measured when the atmospheric pressure In general, there is reasonable but not exact
differs from 760 mmHg, the value may be agreement between measured values and those
corrected using Equation 18.17: calculated on the basis that the vapor pressure
www.ebook777.com
must correspond to the lower flammability limit. obtained some results in the Setaflash apparatus
The reason for this may be that the lower flam- [16]. It is useful in identifying the temperature
mability limit is based on the ability of a flame to range within which the vapor-air mixture in the
propagate approximately 75 cm inside a vertical headspace is flammable. For example, at ambient
tube, 5 cm in diameter [24], whereas the temperatures, the vapor-air mixture in the head-
flashpoint is observed as a localized ignition in space of a gasoline tank is well above the upper
the vicinity of the ignition source. Similar flammability limit and cannot be ignited. How-
localized ignition occurs in the flammability ever, the upper flashpoint of the lower alcohols
limit apparatus but at a concentration of fuel in (in particular methanol and ethanol) appears to
air that sustains only limited flame propagation. be in the mid-20s, only 10–15 K or so above the
If this explanation is accurate, the “calculated” lower flashpoint. This means that at ambient
flashpoint would be expected to be greater than temperatures (say, 15–20 C) a partially full can
the measured one—as indeed the calculation of alcohol contains a flammable vapor-air mix-
above shows (Example 3). ture that can be easily ignited. This is a signifi-
Care should be taken when testing liquids of cant hazard that can give rise to serious
reduced flammability, such as certain chlorinated consequences. For example, if an attempt is
hydrocarbons. James and Tyler [25] investigated made to top-up a conventional flambé lamp
reports of fire and explosions that involved directly from the fuel container before the flame
a commercial cleaning fluid, of which the has extinguished and if the temperature of the
principal component was methyl chloroform alcohol in the container is between the lower and
(1,1,1 trichloroethane, CCl3CH3). This com- upper flashpoints, flame will propagate into the
pound does not give a flashpoint in the standard container, perhaps causing it to burst or other-
test, but a flashpoint of 12 C was recorded in wise expel burning liquid. Such occurrences
vessels of diameter greater than 12.4 cm [26]. have led to a number of serious accidents in
Babrauskas [16] draws attention to a problem restaurants [28]. Provision of a flame arrester in
with blends containing halogenated components. the opening of the container would prevent such
If these are of high volatility, the blend may give an occurrence.
a high flashpoint as a consequence of the
inhibiting effect of the halogenated component.
However, if this halogenated component is lost Open Cup Flashpoints and Firepoints
as a result of preferential evaporation over a
period of time, the effective flashpoint can Open cup flashpoints are not routinely available
decrease, which is the reverse of the effect of in the literature although they are clearly relevant
the preferential loss of lighter hydrocarbons to the ignition of open pools of liquid. They
from fuel blends as discussed above. are determined using an open cup, the most
The closed cup flashpoint is sometimes common of which is the Cleveland apparatus
referred to as the “lower flashpoint.” Although [22] as shown in Fig. 18.2. Instead of the vapor
not widely used, this term does emphasize the accumulating immediately above the liquid sur-
link to the lower flammability limit of the vapor face, it is lost to the atmosphere by diffusion.
and allows the concept of the “upper flashpoint” Consequently, the concentration of vapor in air
to be introduced. This term corresponds to the deceases with height above the liquid surface. In
temperature at which the vapor concentration in the standard test, the ignition source (a small dif-
the headspace is at the upper flammability limit, fusion flame at the end of a swivel arm) is moved
signifying that the mixture will not ignite when across the top of the cup, no more than 2 mm
an ignition source is introduced, although a weak above its rim, in a trajectory that carries the flame
diffusion flame may exist briefly at the open over the center. The process of heating the fuel is
aperture. Upper flashpoint is seldom measured, essentially the same as for the closed cup test, but
although Hasegawa and Takishi [27] have in this case the result is more strongly apparatus
568 D.D. Drysdale
a b
Decane Decane
64 64
Temperature (°C)
Temperature (°C)
60 n -Decane 60
56 56
Firepoint Firepoint
Flashpoint
52 52
0 2 4 6 8 10 12 0 2 4 6 8 10 12
Height of ignition source (mm) Height of ignition source (mm)
Fig. 18.3 Open cup flashpoint (o) and firepoint (•) of (61.5 C); (b) Firepoint as a function of the height
n-decane as a function of the height of the ignition source of the ignition source, showing that it is relatively
above the liquid surface. (a) Flashpoint, revealing insensitive to heights less than about 9 mm. The
how the onset of sustained burning occurs when the arrow shows the height of the ignition source in the
temperature of the liquid is above the firepoint standard test [29]
dependent. A flash of flame is observed when the firepoint. Glassman and Dryer [29] found the
ignition source first encounters a mixture at the firepoint to be much less sensitive to the height
lower flammability limit. For this reason, the of the ignition source, as shown in Fig. 18.3b.
measured open cup flashpoint is very sensitive to In general, firepoints are not routinely
the height of the ignition source above the surface. measured and there is not a good database. Some
This was demonstrated very clearly by Glassman values quoted by Babrauskas [16] are given in
and Dryer [29], as shown in Fig. 18.3a. Clearly, Table 18.5. A different selection is given by
this measurement is apparatus-specific and cannot Kanury [9], but these are all blends that are poorly
provide information about the ignitability of the defined. Typically the firepoint is 10–20 K above
liquid that can properly be generalized—the the closed cup flashpoint, but one cannot rely on
flashpoint of an open pool of liquid will depend this generalization. The difference appears to be
on the distance the vapor has to travel before erratic and can be much greater and more uncertain
meeting a suitable ignition source. Indeed, instead for high-flashpoint liquids (see Babrauskas [16]).
of observing a flashpoint, the liquid may catch fire The lower alcohols seem to behave in a very dif-
and continue to burn (i.e., its temperature is above ferent manner. Glassman and Dryer [29] found the
the firepoint [see Fig. 18.3a]). open cup flashpoints and the firepoints of methanol
This burning will occur when fuel vapors are and ethanol were equal and—even more surpris-
being released at a high enough rate to support a ingly—considerably less than the closed cup
diffusion flame. At the flashpoint (closed cup and flashpoint. This anomaly disappeared if a spark
open cup), the mixture is fuel lean and all the fuel ignition source was used instead of a flame in the
vapor is consumed in the premixed flame. How- open cup measurement: the open cup flashpoint
ever, if the temperature of the liquid is high and the firepoint remained equal but were now
enough to produce a fuel-rich vapor-air mixture, higher than the closed cup flashpoint (see
a self-sustained diffusion flame becomes possi- Table 18.5). This observation has still to be
ble, as illustrated clearly in Fig. 18.3a, which explained satisfactorily, but clearly the behavior
reveals that a minimum fuel temperature must of the alcohols is not typical.
be achieved for this result to occur. This mini- Several attempts have been made to define the
mum temperature at which a self-sustaining dif- firepoint of liquids (and indeed solids) in terms
fusion flame becomes possible is known as the of the heat and mass transfer processes involved
www.ebook777.com
Table 18.5 Some values of closed cup flashpoint, open Roberts and Quince [30], Rasbash [31], and
cup flashpoint, and firepoint temperatures Beyler [32]. In particular, they have used
Closed cup Open cup Firepoint Spalding’s B-number, first used to describe the
FP ( C) FP ( C) ( C) rate of burning of fuel droplets [33], to develop
n-Hexane 22 a
NA the concept of ignition [30, 31] and extinction
n-Heptane 4 1 2 [31, 32] criticalities. It is a dimensionless transfer
Methanolb 12 1.0, 13.5b 1.0, 13.5b
number that can be used to express the conserva-
n-Octane 12 17 18
tion of heat (BH) or mass (BM), the values of
Ethanolb 13 6, 18.0b 6, 18.0b
which can be used to define the rates of heat
s-Butanol 24 NA 29
and mass transfer, respectively. They can be
m-Xylene 25 NA 44
p-Xylene 25 31 44
expressed as follows:
n-Butanol 29 36 36, 38, 50
mog H=r c T g T ls
n-Nonane 31 37 42 BH ¼ ð18:18Þ
Q
o-Xylene 32 36 42
JP-6 NA 38 43 and
n-Decane 44 52 61.5, 66
Decalin NA 57 63 mfs mog =r
BM ¼ ð18:19Þ
Tetraline NA 71 74 1 mfs
Bicyclohexyl NA 74 79
n-Dodecane 74 NA 103 where mog is the mass fraction of oxygen in the
Fuel oil no. 2 124 NA 129 atmosphere, mfs is the mass fraction of fuel
Fuel oil no. 6 146 NA 177 vapor immediately above the liquid surface,
Glycerol 160 176 207 H is the heat of combustion of the fuel vapor,
Motor oil 216 NA 224 r is the stoichiometric ratio (mass of O2
Unless otherwise stated, these data come from the Factory required to burn unit mass of fuel), c is the
Mutual Handbook, as quoted by Babrauskas [16]
specific heat, Tg is the ambient air temperature,
NA Not available
a
The open cup flashpoint of n-hexane is quoted as 26 C and Tls is the temperature of the surface of the
in the original Factory Mutual Handbook and repeated in liquid. BH and BM are assumed equal when the
Babrauskas [16]. This is incorrect diffusivities of heat and mass are equal (the
b
Data from Glassman and Dryer [29]. The lower values
Lewis number is unity). However, this assump-
were obtained with ignition by a pilot flame. The upper
values refer to spark ignition tion carries with it the hidden assumption that
radiative heat transfer can be ignored and only
convection need be considered. For small
in the combustion of the fuel vapors close to the flames—particularly those associated with the
fuel surface. For a diffusion flame to become burning of small droplets for which this
established at the surface of the liquid, the rate approach was developed—this approximation
of evolution of flammable vapor must be greater is reasonable.
than a certain critical value. It has been argued The rate of burning can be expressed as a mass
that it is determined by the need to establish a flux (ṁ00 , the rate of mass transfer per unit surface
self-sustaining process whereby the energy area) in terms of the B-number using the follow-
required to maintain (and promote) the evolution ing equation:
of vapors comes from the flame by convective
and radiative heat transfer. However, if the flow 00 h
m_ ¼ lnð1 þ BÞ ð18:20Þ
rate of vapors is too small, the flame will be too c
close to the surface and self-extinguish as a con- where h is the (convective) heat transfer coeffi-
sequence of heat losses to the surface. cient. Following the argument developed by
Valuable contributions to the definition of Roberts and Quince [9], which invokes the con-
firepoint as a criticality have been made by cept that there is a critical temperature below
570 D.D. Drysdale
which a flame will extinguish (see Chap. 5), a enough from the surface so that the quenching
critical B-number can be formulated as process does not occur. The critical flow rate of
vapors at the firepoint will, therefore, be given by
mog T f , max T ls
Bcrit ¼ ð18:21Þ 00 h
r T f , max T f , crit m_ crit ¼ lnð1 þ Bcrit Þ ð18:22Þ
c
00
where Tf,max is the theoretical flame temperature ṁcrit has not been determined for any liquid fuels
assuming no heat losses to the surface of the but values have been reported for a range of
liquid, Tf,crit is the critical flame temperature solids (see Chap. 36).
below which the flame will extinguish, and Tls
is the surface temperature of the liquid—the
firepoint temperature. Bcrit can be calculated Classification of Liquid Fuels
from Equation 18.19 for BM, substituting for mfs
the mass concentration of fuel vapor above the Although this chapter is entitled “Ignition of
liquid surface at the firepoint (calculated from the Liquids,” most of the emphasis has been on
saturation vapor pressure derived from data sim- understanding the flashpoint, the minimum liquid
ilar to that contained in Table 18.3), allowing the temperature at which the vapor can be ignited. It
critical temperature hypothesis to be tested. The is clear that it is the firepoint that determines
theoretical temperature Tf,max can be deduced whether or not sustained flaming of the liquid
from a heat balance at the surface, assuming will occur, yet combustible liquids are classi-
that the flame loses no heat to the surface (i.e., fied—quite properly—in terms of their
it is adiabatic). For a range of fuels (identified in flashpoints. Measurement of the closed cup
Table 18.2), Tf,crit was found to have a mean flashpoint provides a method of classifying flam-
value of 1350 C (albeit 100 K), which is not mable liquids according to the hazard they
inconsistent with measured and predicted values represent. Systems of classification have been
for premixed flames close to the lower flamma- developed in several countries, but they have as
bility limit (about 1300 C) (see Chap. 5). the common basis the need to identify and make
Observations of the firepoint temperatures of provision for those liquids that can be easily
a number of fuels reveal that the saturated vapor ignited at ambient temperatures. Thus, in the
pressure at the firepoint is above stoichiometric. United Kingdom under the Highly Flammable
Roberts and Quince [9] reported values from Liquids and Liquefied Petroleum Gases
1.33 to 1.92 stoichiometric. Clearly, the mix- (HFL/LP-gas) Regulation 1972 [34], liquids
ture immediately above the surface is rich by a with closed cup flashpoints less than 32 C were
significant margin but is still within the flamma- classified as “highly flammable liquids.” NFPA
bility range. (Zabetakis [24] has shown that the 30, Flammable and Combustible Liquids Code
upper flammability limit is between 2.5 and 4 [14], assigns liquids with flashpoints less than
the stoichiometric concentration.) The firepoint 37.8 C (100 F) to a similar category, known as
represents a criticality, the rate of evolution of Class I. Figure 18.4 compares the U.K. and
vapors being just sufficient to allow the estab- U.S. systems and shows how the Class I liquids
lishment of a diffusion flame at the surface. It is are subdivided into three subclasses A, B, and
closely linked to the “quenching distance,” a C. The boundary between I (A and B) and IC is
characteristic of premixed flames that are set at 22.8 C (73 F), whereas Class IA liquids
quenched (extinguished) within 1 or 2 mm of are distinguished from Class IB in having normal
the surface due to heat losses and (probably) the boiling points less than 37.8 C (100 F).
loss of free radicals (see Chap. 12). The flow rate “Flammable liquids” (1972 Regulations,
of vapors at the firepoint must be sufficient to U.K.) and Class II liquids (U.S.) have a common
allow a nascent diffusion flame to form far upper bound of 60 C. These are liquids that must
www.ebook777.com
Fig. 18.4 A comparison U.K. U.S.A. GHS

of the U.K. and
U.S. classifications of
flammable and combustible Class IIIB
liquids with the UN
100°C Combustible 212°F 100°C Combustible
Globally Harmonized liquids
93.4°C
200°F liquids
System
Class IIIA
60°C 60°C
140°F
Class II
50°C Flammable 50°C Flammable
liquids 37.8°C liquids*
100°F
32°C
Class IC
23°C
73°F
Highly Class Highly
flammable IA + IB flammable
0°C liquids 32°F 0°C liquids*
GHS—UN “Globally Harmonized System”
*Liquids with flashpoint <0°C and boiling points <35°C are classified
as “extremely” flammable.”
be heated significantly above ambient tempera- Table 18.6 Classification system used in DSEAR
ture before the evolved vapor can be ignited, 2002 [35]
although it is clear that some caution is required Boiling
if ambient temperatures above 32–37.8 C are Classification Flashpoint point
encountered. For example, strict application of Extremely <0 C <35 C
the U.K. Highly Flammable Liquid Regulations flammable
or NFPA 30 would be highly inappropriate in Highly flammable <21 C >35 C
Dubai where the average maximum temperature Flammable 21 C and <55 C
during the summer months is about 40 C.
Liquids with flashpoints above 60 C are consid- EU system and as a consequence the 1972
ered to be relatively “safe” in that significant U.K. Regulations have effectively been
temperature increases are required before they replaced by the Dangerous Substances and
can be ignited. In the United Kingdom these are Explosive Atmospheres Regulations (DSEAR
called “combustible liquids,” whereas in the 2002) in which the classification system shown
United States they belong to Class III (which is in Table 18.6 has been adopted.
further divided into Class IIIA and Class IIIB, as This is not the place to attempt to describe the
shown in Fig. 18.4). finer points of these classification systems
Within the European Union (EU), new classi- (of which there are many), but it is relevant to
fication systems have evolved following the
introduction of new regulations arising from EU
4
Directives. These relate, rather confusingly, to Storage: “The Highly Flammable Liquids and Liquefied
the three main “endpoints” (i.e., storage, supply, Petroleum Gases Regulations (HFL/LP-gas)”; Supply:
“The Chemical (Hazard Information and Packaging for
or transport, each slightly different4). Over
Supply) Regulations (CHIPS)”; and Transport: “The Car-
the years, the United Kingdom has modified the riage of Dangerous Goods and Use of Transportable Pres-
HFL/LP-gas regulations to conform to the sure Equipment Regulations (CDG).”
572 D.D. Drysdale
Table 18.7 Classification developed by the UN for piloted ignition: spontaneous ignition, or
transport of hazardous chemicals, known as the GHS sys- “autoignition,” is discussed below.
tem [36]
A pool of a high-firepoint liquid at ambient
Category Criteria temperature is difficult to ignite by means of a
1 Flashpoint <23 C and initial boiling point locally applied ignition source, such as a flame. If
35 C
a flame is applied to the surface (e.g., using a
2 Flashpoint <23 C and initial boiling point
>35 C blowtorch), convection currents are established
3 Flashpoint 23 C and 60 C that remove hot liquid from the point of flame
4 Flashpoint >60 C and 93 C application, replacing it with cooler (cold) liquid
from below. This process occurs because of a
surface tension–driven flow at the surface, first
draw the reader’s attention to the classification identified by Sirignano and Glassman [37] in
that has recently been adopted for transportation their study of flame spread over liquid surfaces.
by the United Nations (UN) to facilitate interna- It is a consequence of the fact that surface tension
tional trade. It defines four categories of liquid, decreases with temperature so that there is a net
as shown in Table 18.7. This classification is force at the surface that draws the hotter liquid on
the UN Globally Harmonized System (GHS) the surface at the point of flame impingement
and is compared in Fig. 18.4 with the original toward the cooler regions. The movement at the
U.K. (1972 Regulations) and U.S. systems of surface created by this force effectively disperses
classification. the heat transferred to the surface and delays
ignition. However, if the liquid is absorbed onto
a porous substrate—a “wick”—then application
Sustained Ignition of Liquids of an ignition source in the form of a small flame
will quickly raise the temperature of the liquid to
Provided that the temperature of a liquid is above its firepoint and burning will be established.
its firepoint, ignition of the vapors above a pool There are two factors that contribute to this
will be followed by the establishment of a flame behavior: first, the liquid is held as a thin film
at the surface. If the temperature is only slightly on the material of the wick and surface
above the firepoint, the initial diffusion flame tension–driven flows are suppressed; and, sec-
will be weak but will strengthen as it transfers ond, most wick materials have low thermal
heat to the surface, causing the temperature to conductivities and insulate the liquid from heat
rise thus generating an increased rate of supply of loss through the wick (by conduction). The best-
fuel vapors. Eventually, a steady-state rate of known example of the effectiveness of a wick in
burning will be achieved, controlled by the heat promoting ease of ignition is with the common
and mass balance of the surface of the fuel at a candle. The wax melts at a relatively low tem-
temperature close to the normal boiling point. perature (about 100 C), but the wick holds the
[18] The standard open cup tests are designed molten wax in place, permitting easy ignition by
to heat the liquid uniformly (bulk heating), but in a match flame. The amount of energy required
practice such bulk heating is difficult to achieve has never been measured, but in principle it could
unless the liquid is used in some process that be calculated assuming the liquid film acts as a
requires an elevated temperature. An obvious thermally thin “solid.” However, the heat transfer
example is the deep fat fryer and similar devices characteristics of such ignition sources are diffi-
used in cooking, but oils and other high- cult to quantify and the calculation would be
flashpoint liquids may be used in heat exchangers rather academic.
and other devices at temperatures above their Burgoyne and Roberts [38] studied the igni-
firepoints. If they are released, then in principle tion of pools of high-flashpoint liquids from
a small ignition source may ignite the vapors and flames established on wicks that dipped into the
cause a fire. This form of ignition is known as liquid at one end of a 0.4-m-long tank. They
www.ebook777.com
Fig. 18.5 Spread of flame 400

over the surface of a liquid
Hexanol
from a flame established on
Isopentanol
Length of induction period (s)

a wick at one end of a pool
Butanol
contained in a long trough 300
[38]. Effect of liquid depth
on the duration of the
induction period (the time
taken for flame to begin to 200
spread over the surface).
The open cup flashpoints of
hexanol, isopentanol (aka
isoamyl alcohol) and 100
butanol, are 74 C, 46 C
and 43 C, respectively.
(The firepoints are not
available)
0 5 10 15 20
Depth of liquid (mm)
found that ignition of the pool would occur after a

delay (induction period), the length of which was Autoignition
determined by the depth of the pool. In this case,
ignition of the pool comprised gradual heating of If the temperature of a flammable vapor-air
the surface layers by heat transfer from the flame mixture is increased sufficiently, it can ignite
on the wick until the firepoint temperature was spontaneously without the introduction of a
achieved in the vicinity of the wick, after which source of ignition—a “pilot”—such as a flame
the flame would spread over the surface of the or spark. The concentration of fuel in air at
pool. Minimum induction periods were observed which the autoignition temperature is a minimum
for layers less than 5 mm, increasing by as much as corresponds to the stoichiometric vapor-air
an order of magnitude as the depth was increased mixture, which is known to be the most reactive
to 8–10 mm, depending on the fuel. As an illustra- [24]. Values of the (minimum) autoignition
tion, Fig. 18.5 shows the results obtained by temperatures (AIT) of a large number of liquid
Burgoyne and Roberts for three alcohols fuels (gases and liquids) are given in Table 18.2.
[38]. There appears to have been no further sys- These have been obtained in a standard test using
tematic studies of this mode of ignition, but it is the Setchkin apparatus [40, 41]. The measurement
reasonable to assume that the time to full ignition is made (for liquids) by dropping a small quantity
of a pool of liquid fuel will depend both on the of the liquid (0.1 ml) into a 500 ml spherical flask
firepoint of the liquid and the depth of the pool. that has been heated to a carefully controlled
It is appropriate to note at this point that high- temperature [18, 41]. If ignition is not observed
flashpoint liquids (both “flammable” and “com- within 10 min, the experiment is repeated at
bustible” according to the U.K. classification higher temperatures. The minimum autoignition
system as shown in Fig. 18.4) can be ignited temperature (AIT) is determined by repeated
very easily if dispersed in air as a spray or mist. experiments, “bracketing” the final value. The
This behavior is analogous to a dust explosion, values of AIT quoted in Table 18.2 were obtained
which involves dispersion of fine solid particles in this way, but in fact these measurements are
in air (see Chap. 70). The concept of flammability highly apparatus dependent. Lower values of AIT
limits applies and in all other respects a mist of are found if the test is carried out in a larger vessel,
combustible liquid exhibits the same properties as specifically with a lower surface-to-volume ratio
a flammable vapor-air mixture. Spray ignition is (Table 18.8). This is entirely consistent with the
discussed in the review by Aggarwal [39]. theory of spontaneous combustion (see Chap. 20).
574 D.D. Drysdale
Table 18.8 Comparison of the AIT ( C) of combustible less than 100 C) boils at >370 C, but the AIT
liquids in spherical vessels of different sizes [41] is quoted as 245 C. In principle, if the conditions
Volume of vessel (m3 106) are right, the hot vapor produced when paraffin
8 35 200 1000 1200 wax is dropped onto a hot surface (>370 C) may
Diethylether 212 197 180 170 160 autoignite if a flammable vapor/air mixture is
Kerosene 283 248 233 227 210 formed near the surface at a sufficiently high
Benzene 668 619 579 559 – temperature (certainly >245 C). However, the
Methanol 498 473 441 428 386 temperature of the surface at which autoignition
n-Pentane 295 273 – 258 – will occur cannot be defined uniquely. Colwell
n-Heptane 255 248 – 223 – and Resa [42] studied the autoignition of a range
of liquid fuels and engine lubricants when single
drops fell on to a flat, heated plate. It was
However, the Setchkin test [40] provides data necessary for them to present their results in
on autoignition under highly idealized conditions terms of the probability of ignition, carrying out
in which the liquid vaporizes inside an enclosed a large number of tests over a range of plate
volume at a uniform temperature. A relatively temperatures. Jet A fuel (similar to kerosene)
uniform fuel/air mixture will be formed that did not autoignite at temperatures below
will undergo autoignition if the temperature is 510 C, despite the fact that the AIT (of kerosene)
high enough. However, this scenario is rarely as measured in the Setchkin Test is quoted as
encountered in practice: more commonly, a liq- 233 C (see Table 18.8). If the plate temperature
uid fuel may come into contact with a hot sur- was 583 C, the ignition probability was 50 %,
face, such as a bearing assembly or an exhaust increasing to 100 % above 650 C. The probabil-
manifold, which is exposed to the open atmo- ity distribution for kerosene and other fuels and
sphere. The liquid evaporates from the hot sur- lubricants, as determined by Colwell and Resa, is
face and the vapor mixes with air at ambient shown in Fig. 18.6. Similar studies, with similar
temperature, or at a temperature that is signifi- conclusions, have been carried out by Shaw and
cantly lower than that of the hot surface. Under Weckman [43] for diesel fuels and Davis
these circumstances, the minimum AIT derived et al. [44] for high performance fuels used in
from the Setchkin test is not relevant as neither motorsports. Note that these results are specific
the temperature nor the concentration of vapor in to a flat surface, 0.579 m by 0.108 m which is
air will be uniform. Logically one would expect horizontal and unconfined: the distribution will
the autoignition process to take place where shift to lower temperatures if the surface is con-
the fuel concentration is at or close to stoichio- fined, or significantly profiled, or to higher
metric (the most reactive mixture) and the local temperatures if the surface is vertical or signifi-
temperature is sufficiently high. To achieve this, cantly smaller in area.
the surface temperature will have to be much If a high-boiling point liquid (such as cooking
higher than the quoted literature value of the oil) is overheated, autoignition can occur, with
corresponding AIT. flame appearing spontaneously in the plume of
If the boiling point of the liquid is lower than hot vapor rising from the surface. As the liquid is
the AIT (e.g., for n-decane, these temperatures already very hot (perhaps close to its boiling
are 174 C and 201 C, respectively), then spill- point), the fire that follows will immediately be
ing the liquid onto a hot, open surface cannot burning at or close to its maximum rate. This
give rise to autoignition because the vapor cannot type of behavior can be demonstrated using
be at a temperature greater than the boiling point corn oil in a Cleveland open cup and is likely to
of the liquid. The situation would be different for be the cause of flaming fires in deep fat fryers.
a combustible liquid with a boiling point much This cannot occur with low-boiling liquid fuels
higher than the AIT. Paraffin wax (a solid at such as gasoline (as explained above), which
room temperature but melting at temperatures accounts for the fact that if gasoline is spilled
www.ebook777.com
Fig. 18.6 Ignition probability as a function of the surface temperature of a flat plate for aviation fluids [42]. With kind
permission from Springer Science+Business Media B.V.
on to a hot exhaust, autoignition will not occur, process in which the liquid reacts heteroge-
although ignition will occur if there is a pilot neously with oxygen from the air. If there is
source (e.g. a spark or flame) nearby. On the sufficient insulation, the rate of heat released by
other hand, heavier fuels and lubricating oils this process will be greater than the rate at which
(with high boiling points) may autoignite under it can be removed and the temperature will rise.
these circumstances. There has been much inter- The phenomenon of self-heating in solids is
est in this issue [16]. discussed at length in Chap. 20.
The best-known example of self-heating (and
spontaneous combustion) of liquids involves the
Ignition of Liquids in Porous Materials so-called “drying oils,” such as linseed oil and
tung oil. To quote Bowes [45], “The risk of self-
This chapter has so far dealt with the ignition of heating and ignition in textile fibres impregnated
liquids, leading either to a transient premixed with readily oxidisable oils has been recognised
flame or to sustained, flaming combustion. How- for a long time,” citing a paper by Hapke that
ever, under the right conditions, liquid fuels with refers to such incidents dating back to the eigh-
high boiling points can undergo a form of smol- teenth century. Self-heating of some of the vege-
dering combustion, although this form of com- table oils that were used in wool blending in the
bustion is normally associated with combustible nineteenth century were known to be the cause of
solids that yield a solid char on heating (see fires and for this reason a test was devised in the
Chap. 19). Clearly, a bulk liquid cannot undergo United Kingdom to identify the oils that were
charring, but if it is dispersed as a thin film within particularly hazardous (the Mackey Test [45, 46]).
a porous substrate (such as an insulating mate- The most hazardous are the unsaturated vegetable
rial) and provided it is of low volatility, it may oils containing a high proportion of carbon–carbon
undergo a slow exothermic reaction that leaves a double bonds, which are prone to oxidation.
charlike residue on the fibers of the substrate. The proportion of these unsaturated linkages can
The process starts insidiously, by a self-heating be quantified in terms of the “iodine number,”
576 D.D. Drysdale
which is effectively the percentage of iodine Another situation in which liquids can initiate
absorption due to the reaction of iodine at the self-heating, leading to spontaneous fires, is in
carbon–carbon double bonds. The higher the iodine the case of oil-soaked lagging [45, 48]. This situ-
number, the greater the propensity of the oil to ation is encountered in industrial plants where
exhibit self-heating when dispersed onto rags, and thermal insulation is provided around vessels
so on. Typical ranges of values are shown in and pipework to conserve heat and maintain pro-
Table 18.9 (extracted from Bowes [45]). cess fluids at high temperatures. If the fluid leaks
The most hazardous oils can self-heat when into the insulation, it will be dispersed through
dispersed onto quite small amounts of mate- the open structure of the insulant. If the leak is
rial—as low as 25 g has been reported. This too large, then the material will become logged
means that rags used to apply linseed oil or with the fluid and self-heating will not occur.
tung oil–based finishes to wood surfaces can However, if the leak is small, then the fluid will
self-heat if they are not disposed of properly. become dispersed onto the fibers of the insulation
The less reactive oils will present problems if material, creating a large area of fuel exposed to
dispersed on much larger quantities of porous air within the interstices of the material—an
material, such as bales of wool. Qualitative ideal situation to promote self-heating, particu-
information on a wide range of liquids may be larly as both the fluid and the insulation are
found in NFPA’s Fire Protection Handbook® already hot. The volatility of the fluid is impor-
[47], which provides a list of materials that are tant here: if at the relevant temperature its vapor
subject to spontaneous heating. This information pressure is too high, then the oxygen concentra-
has been extracted and is presented in tion in the pores of the insulation may be too low
Table 18.10. to allow the self-heating process to develop
Under the right conditions, self-heating of sufficiently.
these dispersed liquids will lead to the onset of There have been studies of the propensity of
smoldering combustion. The self-heating process different fluids to undergo self-heating when dis-
is a slow oxidation involving the carbon–carbon persed on porous insulation materials and an
double bonds, resulting in changes to the struc- empirical expression has been derived to assess
ture of the absorbed liquid. The chemical mech- the hazard. This is reported by Lindner and
anism has not been studied in detail, but the Seibring [49] and developed by Britton [50].
consequence appears to be that char is produced Self-heating may be an issue if
on the fibers, which can then oxidize more AIT
vigorously, producing a smolder that has the > 1:55 ð18:23Þ
AIT FP
potential to undergo transition to flaming. The
smolder will involve the substrate if it is combus- where AIT is the autoignition temperature and
tible (e.g., cotton rags, wool, etc.), but it is not FP is the (closed cup) flashpoint. Although it may
necessary for the substrate to be combustible, at seem illogical to incorporate AIT (which refers
least for the more reactive oils. to autoignition in the gas phase) into a correlation
that deals with a heterogeneous process, the sig-
nificance of AIT is that it is a measure of the
reactivity of the fuel. FP is also associated with a
Table 18.9 Ranges of iodine values [45] gas-phase process, but in this expression, it is a
Oil Iodine number Hazard surrogate for a measure of the volatility of the
Tung, linseed 160–185 Most hazardous fuel as it correlates with the boiling point (see,
Sunflower seed 127–136 for example, Babrauskas [16]). A “reactive fuel”
Soya bean 124–133 (low AIT) of low volatility (high FP) will, there-
Cottonseed 103–111 fore, give a high value of AIT/(AIT – FP)—
Olive oil 80–85 consistent with the above inequality. However,
Coconut oil 8–10 Least hazardous this is at least two steps removed from
www.ebook777.com
Table 18.10 Liquids capable of self-heating when dispersed on fibrous materials [47]
Tendency to Usual shipping
spontaneous container or
Name heating storage method Precautions Remarks
a
Castor oil Very slight Metal barrels, Possible heating of saturated fabrics in badly
metal cans in ventilated piles
wooden boxes
a
Coconut oil Very slight Drums, cans, glass Only dangerous if fabrics, etc., are impregnated
a
Cod liver oil High Drums, cans, glass Impregnated organic materials are extremely
dangerous
a
Corn oil Moderate Barrels, tank cars Dangerous heating of meals, etc., unlikely unless
stored in large piles while hot
a
Cottonseed Moderate Barrels, tank cars May cause heating of saturated material in badly
oil ventilated piles
a
Fish oil High Barrels, drums Impregnated porous or fibrous materials are
extremely dangerous. Tendency of various fish oils
to heat varies with origin
a
Lanolin Negligible Glass, cans, metal Heating possible in contaminated fibrous matter
drums, barrels
a
Lard oil Slight Wooden barrels Dangerous on fibrous combustible substances
a
Linseed oil High Tank cars, drums, Rags or fabrics impregnated with oil are extremely
cans, glass dangerous. Avoid piles etc.
Store in closed containers, preferably metal
a
Menhaden Moderate to Barrels, drums, Dangerous on fibrous product
oil high tank cars
a
Mustard oil, No Barrels Avoid contamination of fibrous combustible
black materials
a
Olive oil Moderate to Tank cars, drums, Impregnated fibrous materials may heat unless
low cans, glass ventilated. Tendency varies with origin of oil
a
Paint Moderate Drums, cans, glass Fabrics, rags, etc. impregnated with paints that
containing contain drying oils and dryers and extremely
drying oilb dangerous. Store in closed containers, preferably
metal
a
Palm oil Low Wooden barrels Impregnated fibrous materials may heat unless
ventilated. Tendency varies with origin of oil
a
Peanut oil Low Wooden barrels, tin Impregnated fibrous materials may heat unless
cans ventilated. Tendency varies with origin of oil
a
Perilla oil Moderate to Tin cans, barrels Impregnated fibrous materials may heat unless
high ventilated. Tendency varies with origin of oil
a
Pine oil Moderate Glass, drums Impregnated fibrous materials may heat unless
ventilated. Tendency varies with origin of oil
a
Red oil Moderate Glass bottles, Impregnated porous or fibrous materials may heat
wooden barrels unless ventilated. Tendency varies with origin of oil
a
Soybean oil Moderate Tin cans, barrels, Impregnated fibrous materials may heat unless well
tank cars ventilated
a
Tung oil Moderate Tin cans, barrels, Impregnated fibrous materials may heat unless well
tank cars ventilated. Tendency varies with origin of oil
a
Whale oil Moderate Barrels and tank Impregnated fibrous materials may heat unless well
cars ventilated. Tendency varies with origin of oil
a
In every case, the recommended precaution to prevent spontaneous heating is to “avoid contact of leakage from
containers with rags, cotton, or other fibrous combustible materials”
b
This is a well-known hazard in the trade. Impregnated rags are laid out flat to dry. Self-heating will only occur if the
rags are left in a crumpled state, providing insulation for the interior
578 D.D. Drysdale
understanding the process involved. The theory liquids in practical situations is considered (e.g.,
of self-heating is described in Chap. 20, but it is spilling gasoline onto a hot exhaust).
only recently that the parameters relevant to If a combustible liquid is absorbed into a
understanding lagging fires have been examined porous medium, such as a rag or the insulation
in detail [51]. around a hot pipe, then it may undergo a self-
heating process leading initially to a smoldering-
type reaction. Reactive drying oils such as lin-
Summary seed will undergo this process at ambient
temperatures and can lead to flaming combustion
Flammable and combustible liquids present a under suitable conditions. Similarly, hot process
range of fire hazards in our everyday lives, fluid contaminating the insulation around
whether it be in industry, commerce, or the pipework in industrial plants can lead to “lagging
home. In this chapter, an attempt has been made fires,” which are difficult to control. The hazard
to outline the fundamental parameters that deter- associated with liquids under these conditions is
mine the flammability of liquids, bearing in mind poorly understood, which makes awareness of
that the hazard may present itself in a number of the problem even more important.
ways: as an unconfined pool of liquid, as a dis-
persion of droplets, or as liquid absorbed on a
porous substrate. The simplest way of classifying References
the flammability of a liquid is in terms of
1. D.R. Lide (ed.), Handbook of Chemistry and Physics,
its flashpoint, the lowest temperature at which
80th ed., CRC, Boca Raton, FL (2001/2002).
the saturated vapor pressure of the liquid 2. W.J. Moore, Physical Chemistry, 5th ed., Longman,
corresponds to the lower flammability limit of London, UK (1972).
the vapor in air at normal atmospheric pressure. 3. P.W. Atkins and de Paula, Physical Chemistry, 8th
ed., Oxford University Press, Oxford, UK (2006).
The classification schemes used in the United
4. S.H. Wade, “Evaporation of Liquids in Currents of
States and the United Kingdom are based simply Air,” in Proceedings of the Institution of Chemical
on the flashpoint, but it is important to recognize Engineers, Institution of Chemical Engineers, Rugby,
that the flashpoint temperature decreases with UK (Jan. 1942).
5. V.J. Clancey, “The Evaporation and Dispersion of
atmospheric pressure, while the actual flashpoint
Flammable Liquid Spillages,” in Proceedings of the
is not relevant if the liquid is presented to an Symposium on Chemical Process Hazards, University
ignition source as a mist or spray. Indeed, liquids of Manchester, Institution of Chemical Engineers,
with flashpoints above 300 C may be ignited by Rugby, UK (1974).
6. A.M. Thyer, “A Review of Data on Spreading and
a small flame or a spark if they are dispersed in
Vaporization of Cryogenic Liquids Spills,” Journal of
droplet form. Hazardous Materials, 99, pp. 31–40 (2003).
A distinction must be made between 7. Center for Chemical Process Safety, “Guidelines for
flashpoint, as defined above, and firepoint. The Evaluating the Characteristics of Vapor Cloud
Explosions, Flash Fires and BLEVEs,” AIChE,
latter refers to the lowest temperature at which a
New York (1994).
liquid will continue to burn following ignition of 8. T. Abbassi and S.A. Abbassi, “The Boiling Liquid
the vapors by a pilot ignition source. It is higher Expanding Vapor Explosion (BLEVE): Mechanism,
than the flashpoint, generally by 10–20 K. Values Consequence, Assessment, Management,” Journal of
Hazardous Materials, 141, pp. 489–519 (2007).
of autoignition temperature (AIT) are quoted
9. A.M. Kanury, “Ignition of Liquid Fuels,” in SFPE
in the literature, but these are not related to either Handbook, 3rd ed., National Fire Protection Associa-
the flashpoint or the firepoint. The AIT represents tion, Quincy, MA, pp. 2-188–2-199 (2002).
the lowest temperature at which flame develops 10. E.W. Washburn (ed.), International Critical Tables of
Numerical Data: Physics, Chemistry and Technology,
spontaneously in a uniformly heated closed
McGraw-Hill, New York (1927).
vessel into which a small sample of liquid is 11. Factory Mutual System, Handbook of Industrial Loss
introduced. This idealized configuration must Prevention, Factory Mutual Research Corporation,
be taken into account when “autoignition” of Norwood, MA (1968).
www.ebook777.com
12. C.L. Yaws (ed.), Chemical Properties Handbook: 29. I. Glassman and F.L. Dryer, “Flame Spreading Across
Physical, Thermodynamic, Environmental, Transport, Liquid Fuels,” Fire Safety Journal, 3, pp. 123–138
Safety and Health Related Properties for Organic and (1980/1981).
Inorganic Chemicals. McGraw Hill, New York 30. A.F. Roberts and B.W. Quince, “A Limiting Condi-
(1999). tion for the Burning of Flammable Liquids,” Combus-
13. R.C. Weast (ed.), Handbook of Chemistry and Phys- tion and Flame, 20, pp. 245–251 (1973).
ics, 53rd ed., CRC, Cleveland, OH (1972/1973). 31. D.J. Rasbash, “Relevance of Firepoint Theory to the
14. NFPA 30, Flammable and Combustible Liquids Code, Assessment of Fire Behavior of Combustible
National Fire Protection Association, Quincy, MA Materials,” International Symposium on Combustible
(2008). Materials, Edinburgh University, Edinburgh,
15. P.F. Thorne, Fire Research Note 1022, Fire Research Scotland (1974).
Station, Borehamwood, UK; P.F. Thorne, 32. C.L. Beyler, “A Unified Model of Fire Suppression,”
“Flashpoints of Mixtures of Flammable and Journal of Fire Protection Engineering, 4, pp. 5–16
Non-Flammable Liquids,” Fire and Materials, 1, pp. (1992).
134–140 (1976). 33. D.B. Spalding, “Some Fundamentals of Combustion,”
16. V. Babrauskas, Ignition Handbook, Fire Science Butterworths, London, UK (1957).
Publishers, Issaquah, WA (2003). 34. Highly Flammable Liquids and Liquefied Petroleum
17. B.P. Mullins and S.S. Penner, Explosions, Gases Regulations, Her Majesty’s Stationery Office,
Detonations, Flammability and Ignition, Pergamon London, UK (1972).
Press, London, UK (1959). 35. Chemicals (Hazard Information and Packaging for
18. D. Drysdale, Introduction to Fire Dynamics, 3rd ed., Supply): Regulations 2002, HSE Booklet L131,
John Wiley and Sons, New York (2011). Health and Safety Executive, London, UK (2002).
19. NFPA, Fire Protection Handbook, 20th ed., National 36. Globally Harmonised System of Classification and
Fire Protection Association, Quincy, MA, Labelling of Chemicals (GHS), 1st rev. ed., UN Pub-
pp. 21-135–21-149 (2008). lication ST/SG/AC. 10/30/Rev.1, United Nations,
20. American Society for Testing and Materials, “Stan- Geneva, Switzerland (2005).
dard Test Method for Flashpoint by the Pensky- 37. W.A. Sirignano and I. Glassman, “Flame Spreading
Martens Closed Tester,” ASTM D-93-94, ASTM, Above Liquid Fuels: Surface Temperature Driven
West Conshohocken, PA (1994). Flows,” Combustion Science and Technology, 1,
21. American Society for Testing and Materials, “Stan- pp. 307–312 (1970).
dard Test Method for Flashpoint by the Tag Closed 38. J.H. Burgoyne and A.F. Roberts, “Spread of Flame
Tester,” ASTM D-56-87, ASTM, West Across a Liquid Surface Part 2,” Proceedings of the
Conshohocken, PA (1987). Royal Society, London, A308, pp. 55–68 (1968).
22. American Society for Testing and Materials, “Stan- 39. S.K. Aggarwal, “A Review of Spray Ignition Phenom-
dard Test Method for Flashpoint and Firepoint by the ena: Present Status and Future Research,” Progress in
Cleveland Open Cup,” ASTM D-92-90, ASTM, West Energy and Combustion Science, 24, pp. 565–600
Conshohocken, PA (1990). (1998).
23. American Society for Testing and Materials, “Stan- 40. American Society for Testing and Materials, “Stan-
dard Test Method for Flashpoint and Firepoint of dard Test Method for Autoignition Temperature of
Liquids by the Tag Open Cup Apparatus,” ASTM Liquid Chemicals,” ASTME, 659–78 (Reapproved
D-1310-86, ASTM, West Conshohocken, PA (1986). 2000) (1978).
24. M.G. Zabetakis, “Flammability Characteristics of 41. N.P. Setchkin, “Self-Ignition Temperatures of Com-
Combustible Gases and Vapours,” Bulletin 627, bustible Liquids,” Journal of Research National
U.S. Bureau of Mines, Washington, DC (1961). Bureau of Standards, 53, pp. 49–66 (1954).
25. J.J. James, “A Method for Flammability Testing of 42. J.D. Colwell and A. Reza,. “Hot surface ignition
Low Flammability Liquids,” MSc Thesis, University of automotive and aviation fluids” Fire Technology
of Manchester, Manchester, UK (1991). 41 105–123. (2005).
26. B.J. Tyler, personal communication. 43. A Shaw and E Weckman, “Evaluation of the ignition
27. K. Hasegawa and K. Kashuki, “A Method for Mea- of diesel fuels on hot surfaces” Fire Technology
suring the Upper Flashpoint—Practical Method Using 46 407– (2010).
the Setaflash Closed Cup Apparatus,” Report of the 44. S. Davis, S Kelly and V Somandepalli, “Hot surface
Fire Research Institute No. 71, Fire Research Institute, ignition of performance fuels” Fire Technology
Tokyo, Japan (1991). 46 363–374 (2010).
28. H. Mundwiler, “Brand im Hotel International Zürich, 45. P.C. Bowes, “Self-Heating: Evaluating and
16. Februar 1988”/“Incendie à l’Hôtel International, Controlling the Hazards,” HMSO, London (1984).
Zurich, 16 février 1988”; Schweizerische 46. W. McD Mackey, Journal of the Society of Chemical
Feuerwehrzeitung/Journal des Sapeurs-Pompiers Industry, London, 14, p. 940 (1895); 15, p. 90 (1896).
Suisses/Giornale dei Pompieri Svizzeri, 116, 3, pp. 47. National Fire Protection Association, Fire Protection
144–162 (Mar. 1990). Handbook, 20th ed., Table 6.17.11 “Materials Subject
580 D.D. Drysdale
to Spontaneous Heating,” National Fire Protection 51. A.C. McIntosh, M. Bains, W. Crocombe, and
Association, Quincy, MA (2008). J.F. Griffiths, “Autoignition of Combustible Fluids in
48. P.C. Bowes, “Fires in Oil-Soaked Lagging,” BRE Porous Insulating Materials,” Combustion and Flame,
Current Paper, CP 35/74, Fire Research Station, 99, pp. 541–550 (1994).
Borehamwood, UK (1974).
49. H. Lindner and H. Seibring, “Self-Ignition of Organic
Substances in Lagging Material,” Chemie Ingenieur D.D. Drysdale is professor emeritus in the BRE Centre
Technik, 39, p. 667 (1967). for Fire Safety Engineering, School of Engineering, at the
50. L.G. Britton, “Spontaneous Insulation Fires,” AIChE University of Edinburgh, Scotland. His research interests
Loss Prevention Symposium, San Diego, CA (August lie in fire science, fire dynamics, and the fire behavior of
19–22, 1990). combustible materials.
www.ebook777.com
Smoldering Combustion
19
Guillermo Rein
Introduction charcoal embers are familiar examples of smol-

dering combustion. The latter is illustrated in
Smoldering combustion is the slow, low temper- Fig. 19.1 (left). A less familiar case but more
ature, flameless burning of porous fuels and the common fire threat is the burning of foam in
most persistent type of combustion phenomena. upholstery as seen in Fig. 19.1 (right). The
The heat is released when oxygen directly attacks smouldering combustion of polymer foams and
the surface of a solid fuel [1]. It is especially other building insulation materials is gaining
common in porous fuels which form a char on importance because of their extensive use by
heating [2], like cellulosic insulation, polyure- the construction industry which is seeking
thane foam or peat. Smoldering combustion is higher energy efficiency.
among the leading causes of residential fires Despite its broad implications to safety and
[3, 4], and it is a source of safety concerns in the environment, our current understanding of
industrial premises as well as in commercial and smoldering combustion is limited, and consider-
space flights. Smoldering is also the dominant ably less advanced than flaming combustion.
combustion phenomena in wildfires of natural This is due to its complexity and coupling of
deposits of peat and coal which are the largest transport and thermochemical processes inside a
and longest burning fires on Earth [5]. These reactive porous media, The most comprehensive
fires contribute considerably to global green- reviews of smouldering combustion in the litera-
house gas emissions, and result in widespread ture include Ohlemiller [1], T’ien et al. [6],
destruction of ecosystems and the waste of Ohlemiller [7], Babrauskas [8], Rein [9] and
natural resources. Drysdale [2].
Smoldering constitutes a hazard in the built
environment, as emphasized in this chapter, for
two main reasons. First, it typically yields a Smoldering vs. Flaming Combustion
higher conversion of the fuel to toxic
compounds than flaming, though this occurs The core of any combustion process is a global
more slowly. Second, smoldering provides a exothermic reaction that results in the release
path to a flaming fire via heat sources too weak of heat, and both gaseous and solid products.
to directly ignite a flame. A burning cigarette or Whether smoldering or flaming will be the dom-
inant mode is dictated by which chemical spe-
cies is oxidized. If the oxidation takes place in
the solid phase, smouldering is dominant; if the
G. Rein (*)
Department of Mechanical Engineering, Imperial College oxidation takes place in the gas phase then
London, London, SW72AZ, UK flaming dominates.

582 G. Rein
In general, the combustion of a solid fuel approximated by two lumped chemical pathways:
involves countless elementary chemical reactions, pyrolysis (Equation 19.1) followed by oxidation
but the global reaction, in its simplest form, can be (Equations 19.2a or 19.2b).
Pyrolysis:
Fuel ðsolidÞ þ Heat ! Pyrolyzate ðgasÞ þ Char ðsolidÞ þ Ash ðsolidÞ ð19:1Þ
Heterogeneous oxidation:
Char ðsolidÞ þ O2 ðgasÞ ! Heat þ CO2 þ H2 O þ other gases þ Ash ðsolidÞ ð19:2aÞ
Gas-phase oxidation:
Pyrolyzate ðgasÞ þ O2 ðgasÞ ! Heat þ CO2 þ H2 O þ other gases ð19:2bÞ
Pyrolysis1 results in gaseous and solid General Characteristics of Smoldering

products. Of these, both the pyrolysate and the Combustion
char are susceptible to subsequent oxidation.2
Char oxidation (Equation 19.2a) is a heteroge- The characteristic temperature, spread rate and
neous reaction (involving solid and gaseous spe- power of smoldering combustion are low com-
cies) and takes place on the char produced by the pared to flaming combustion. Typical peak
pyrolysis reaction and leads to smoldering com- temperatures for smoldering are in the range
bustion. Oxidation of the pyrolysate (Equa- from 450 C to 700 C, although very energetic
tion 19.2b) is airborne and takes place in the and dense fuels such as coal can reach peaks at
gas phase, which is a homogenous reaction around 1000 C. The effective heat of combus-
(involving only gaseous species) and leads to tion taking into account the combustion effi-
flaming combustion. ciency is in the range from 6 to 12 kJ/g. These
Char is a carbon-rich porous material, also are much lower compared to typical values of
called charcoal or black carbon, with a high around 1500 C and 16–30 kJ/g respectively for
surface-to-volume ratio and a high heat of reac- flaming combustion. The heat release rate per
tion. The attack of oxygen is facilitated by metal unit area of burning front is low and ranges
and mineral impurities which can catalyze the from 10 to 30 kW/m2 [10]. Because of these
oxidation process. The other important solid spe- characteristics and despite the considerable vari-
cies is ash, a mineral-rich residue of negligible ation in the chemical nature of smoldering fuels,
reactivity that is left after the fire. smoldering spreads in a creeping fashion, typi-
cally around 1 mm/min, which is two orders of
magnitude slower than flame spread.
As we will see in the following sections, smol-
dering combustion can be initiated with weaker
ignition sources and is more difficult to suppress
than flaming combustion. This makes smoldering
1
Pyrolysis in this context is the chemical decomposition the most persistent combustion mode. In the
of a solid material solely by heating. It does not involve presence of a large quantity of fuel, a smoldering
oxidation reactions and it is endothermic. It involves the
fire can burn for very long periods (days, weeks,
irreversible and simultaneous change of chemical compo-
sition and physical phase. months) despite fire-fighting attempts, extensive
2
Oxidation in this context is the reaction of a species with rain or weather changes. Indeed, the longest-
the oxygen in the air. It is an exothermic reaction. burning urban fire in modern history occurred
www.ebook777.com
19 Smoldering Combustion 583
Fig. 19.1 (Left) Smoldering embers and ash residue 125 mm in diameter smoldered in microgravity
(Photo by J. B. Nielsen, Public domain, via Wikimedia conditions (Photo by NASA)
Commons). (Right) Cross-section of a polyurethane slab
after the collapse of the Word Trade Center For most smoldering fuels under typical
towers on Sept. 11, 2001. The immense piles of conditions, the two mechanisms controlling the
debris smoldered for 3–5 months in the middle of rate of spread are the oxygen supply and the heat
one of the most heavily populated cities of the transfer [1]. Both mechanisms are dictated by
world, resisting all suppression attempts until the heat and mass transfer in porous media. At the
slow removal of debris cleared the pile [11]. micro scale, smoldering takes place on the sur-
The oldest continuously burning fire on Earth is face of the pores of a solid fuel, while at the
The Burning Mountain in New South Wales, macro scale, it is a bulk phenomenon affecting
Australia, a large coal seam which has been the fuel bed at large. The fire can penetrate deep
smoldering for more than 6000 years [12]. into the bed of fuel if oxygen can be transported
Many materials can sustain a smoldering fire. from a free surface (open atmosphere, crack or
These include synthetic fuels such as charring channel). Depending on the depth of the fire, the
polymers, polyurethane foam, cellulosic insula- importance of the heat and mass transfer
tion, particleboard and sawdust, and natural fuels mechanisms vary. The rate of oxygen supply to
like wood, peat, forest litter, and coal. In chemi- the reaction zone enhances the spread. But this
cal terms, most smoldering fuels form a char on oxygen supply decreases with depth inside a fuel
heating. In physical terms, smoldering fuels con- bed. The rate of heat loss to the environment
sist of a permeable medium formed by grains, from the reaction zone hinders the spread. But
fibers or some other porous matrix. This aggre- this heat loss decreases with depth inside a
gate nature provides a large surface area per unit fuel bed. Thus, depending on the fuel bed
volume, which facilitates the heterogeneous characteristics and wind conditions, there are min-
reaction with oxygen, and permits transport of imum and maximum depths between which smol-
oxygen through the fuel bed [1]. For example, dering can spread, and an optimal depth at which
synthetic foams prone to charring, like polyure- burning is fastest. The concept of an optimal depth
thane foam (see Fig. 19.1, right), are highly is not sufficiently studied yet and could vary for
susceptible to smoldering combustion. Liquid different systems in the scale from cm to m.
fuels soaked in an inert porous matrix can also Each location of a burning fuel bed sees the
sustain smoldering fires, examples include successive arrival of four distinct thermal and
lagging fires (Drysdale [2], Chap. 18) and soil chemical subfronts that form the structure of a
remediation based on the assisted burning of smoldering fire. These are the preheating, drying,
liquid contaminants [13]. pyrolysis and oxidation subfronts [1, 14].
584 G. Rein
Forward Flow
Air
Reacted Virgin
Foam Foam
Propagation
Reaction Zone
Char & Residue
Opposed Flow
Air Flow
Air
Virgin Reacted
Foam Foam Porous Fuel
Pyrolysis
Oxidation
Reaction Zone
Fig. 19.2 (Left) Forward and opposed configurations of in forward smoldering and approximate correspondence
one-dimensional smoldering propagation inside a slab of with a burning cigarette (By G. Rein, CC BY license)
porous fuel. (Right) Structure of a one-dimensional front
The endothermic preheating, drying and approximately at 200 C for polyurethane, and
pyrolysis subfronts store or consume thermal to 250 C for cellulose [15]; subsequent heating
energy and move ahead of the oxidation subfront above this temperature increases the pyrolysis
The heat that sustains the spread is released at the rate and char production. The oxidation subfront
oxidation subfront, and from there it is trans- consumes char and oxygen, releasing heat. It
ferred via a combination of conduction, convec- involves the oxidation of the fuel and the char,
tion and radiation to the other subfronts. The but char oxidation (Equation 19.2a) is much more
preheating subfront does not involve chemical exothermic. The oxidation and pyrolysis
reactions or gas emissions in any significant subfronts may overlap in space. The extent of
quantity. The drying subfront follows with evap- this overlap depends on the propagation
oration becoming significant above 50 C, emit- conditions [16] and is discussed in the section
ting water vapour and leaving behind dry fuel. “Smoldering Kinetics”.
Drying is most significant for fuels with substan- It is convenient to characterize one-
tial moisture contents (>10 % in dry weight), so it dimensional smoldering by its direction of propa-
is negligible in most residential fire scenarios but gation relative to the direction of the oxygen
important in the natural environment (see section supply. Two one-dimensional modes exist.
“Smoldering Wildland Fires”). The pyrolysis Forward propagation occurs when the oxygen sup-
subfront (Equation 19.1) follows the preheating ply is moving in the direction of the smolder front.
and drying when the fuel temperature increases Opposed propagation (also called reverse) occurs
above a certain threshold. This threshold3 is when the oxygen supply is moving opposite to the
smolder front. These are illustrated in Fig. 19.2.
The most familiar example of forward propaga-
tion is a cigarette, as seen in Fig. 19.2 (right).
3
The onset of pyrolysis or oxidation does not occur at one Although one-dimensional spread is an idealized
fixed temperature but it is known to be a function of the situation, it can occasionally be found in fires,
heating rate and start over a range of temperatures; higher but in general, real smoldering fires are multidi-
onset temperatures are observed for higher heating rates.
See Rein et al. [15] and the section “Smoldering Kinetics” mensional and cannot be classified into a single
for evidence of this. mode.
www.ebook777.com
In forward propagation, the pyrolysis subfront and fuel kinetics, with the oxygen supply rate
is located at the leading edge of the front, and playing a secondary role. Above a critical
the oxidation subfront at the trailing edge, threshold of heat supply, the temperature
where oxygen is drawn (see Fig. 19.2, right). increase initiates endothermic pyrolysis, which
The oxygen supply flows first through the char is followed by the onset of oxidation. When the
where it is consumed. Then the hot, oxygen- heat released by oxidation is high enough to
depleted gases of combustion flow through the balance the heat required for the endothermic
virgin fuel. This convective transport results in processes (heat losses, pyrolysis, drying and
enhanced drying and preheating, but it also preheating of fuel), propagation occurs and the
results in water condensation on the virgin fuel reaction might become self-sustaining (only then
as the combustion gases cool down. oxygen supply rate will play an important role).
In opposed propagation, the oxygen supply This section discusses four types of ignition
flows first through the virgin fuel, and through the sources: radiant, conductive, ember and self-
preheating and evaporation subfronts before heating. For all four, it is proven that the critical
reaching the char where the oxidation subfront is energy condition needed for smoldering is signif-
located. Then the hot, oxygen-depleted gases of icantly lesser than that for flaming. The results
combustion flow through the char and ash residues. discussed here are on individual fuel samples.
This means that heat is transferred by convection in But note that a particularly important smoldering
the opposite direction to the virgin fuel, reducing scenario, that of upholstery and bedding fires, is a
the extend of the drying and preheating, which in composite problem with the ignition propensity
turn results in a weaker smoldering process. of both the fabric and the substrate contributing
Consequently, forward smolder is faster than to the overall behavior [18].
opposed under the same fuel and oxidizer supply,
and allows for more complete combustion of the
fuel [17]. In opposed propagation, the pyrolysis and Radiant Ignition
oxidation subfronts overlap on top of each other
from the leading edge to the trailing edge [16]. The effects of exposing polyurethane foam to an
external radiant heat flux of increasing magni-
tude is illustrated in Fig. 19.3. The heat flux
Ignition needed to initiate smoldering is significantly
lower than that for flaming (see Table 19.1).
The process of smoldering ignition requires the For instance, the critical radiation heat flux
supply of heat, and is governed by heat transfer for smolder ignition of polyurethane foam is
Fig. 19.3 Images of polyurethane foam samples exposed smoldering front did not propagate, (c) sample in which
to increasing levels of radiation (from left to right): (a) virgin smoldering propagated, and (d) sample which underwent
foam (not exposed to radiation), (b) charred foam in which a flaming ignition (By R. Hadden [19], CC BY license)
586 G. Rein
7 kW/m2, while for spontaneous flaming is smoldering ignition by radiation depends on the
30 kW/m2 [19]. For piloted flaming ignition, heat flux level and ranges from 1 to 20 min for a
the critical level is 13 kW/m2 (Chap. 36). heat flux of 45–7 kW/m2 respectively. Flaming
It has been shown that the onset of smoldering ignition was observed for heat fluxes above
involves a range of threshold temperatures. A 30kW/m2 and was fast, less than 1 min. The
single threshold temperature is not a valid crite- dependence of the time to ignition with radiant
rion for ignition but it is part of a multi-criteria heat flux shown in Fig. 19.4 (right) resembles a
signature [20]. Several experimental studies have an inverse square-root law with the incident heat
found that the minimum temperature measured flux. This suggests that smoldering ignition can
during ignition of polyurethane foam is in range be explained in terms of heat transfer, in the same
from 300 C to 450 C [19–21]. This is much way that a an inverse square-root law explains
lower than the threshold surface temperature of the flaming ignition of a thermally thick fuel
600 C required for spontaneous flaming ignition (Chap. 21). This law originates from the time it
of the same material [19]. Figure 19.4 (left) takes for heat conduction to result in a critical
shows the experimental data on the peak temper- temperature at a key location. For flaming, the
ature reached inside a slab of polyurethane foam key location is the free surface of the fuel, but for
for the cases of no ignition, smoldering or smoldering the key location is inside the fuel bed,
flaming ignition. The same experiments (see at a sufficient depth such that an insulating layer
Fig. 19.4 (right)) also show that the time to of char is formed over the oxidation front
[21]. The concept of a critical depth for ignition
is not sufficiently studied yet but could be
Table 19.1 Critical heat fluxes found experimentally for
the ignition of smoldering and spontaneous flaming com-
expected to vary for different materials and
bustion in samples of different sizes (Data from Hadden external conditions in the scale from 1 to 10 cm
et al. [19]) roughly.
Critical heat flux for Critical heat flux
Sample size smoldering ignition for flaming ignition
(mm) (kW m2) (kW m2) Conductive Ignition
50 18–19 32–45
100 8–9 32–37 The heat source that can start a smoldering fire
140 7–8 30–31 with the lowest heat flux is the conductive type.
900
No ignition 20
800 Smouldering
18
Flaming
Maximum temperature, °C
700 16
Time to ignition, min
14
600
12
500 10
400 8
6
300
4
200
2
100 0
0 10 20 30 40 50 0 10 20 30 40 50
Heat Flux, kW/m2 Heat Flux, kW⋅m−2
Fig. 19.4 Radiant ignition of smoldering and flaming in ignition. Red, green and blue represent 50, 100 and
polyurethane samples of different sizes in still air. (Left) 140 mm side square samples respectively (Data from
Maximum temperatures observed. (Right) Time to Hadden et al. [19])
www.ebook777.com
This scenario occurs when a large and hot object conduction rate decreases resulting in a longer
is in direct contact with the fuel bed. Heat is time to ignition.
transferred by conduction, but in porous fuels,
convection also plays a role. Anthenien and
Fernandez-Pello [21] studied the initiation of Ignition by Embers
smoldering using an electric heater in contact
with a sample of polyurethane foam under forced In close connection with conductive sources, a fuel
airflow. Ignition was reported at heat fluxes as bed can also be ignited by hot embers. This is
low as 3 kW/m2 and the relationship between related to ignition by hot works and also to the
time to ignition and heat flux was shown to phenomenon of spotting in wildland fires, when
follow a an inverse square-root law. Ignition lofted embers land downwind, leading to second-
was found to be weakly dependent on the airflow. ary fires in the wildland or in urban areas remote
Conductive ignition has also been studied on a from the originating flame front. Ignition by
bed of particles [22]. A series of beds of anthra- embers is a transient phenomenon involving the
citic coal particle of uniform diameter ranging loss of heat from the ember to the fuel by conduc-
from 7 to 45 mm was investigated inside a cubic tion and convection. Embers can be classified by
box (side of 100 mm) with the top side open to material (metal or biomass) and thermal state (hot,
the atmosphere and multiple perforations on the smoldering or flaming).
other sides. The heat source was an electric wire Manzello et al. [23] compared the ember igni-
that delivered 80 W. Figure 19.5 shows that the tion of three fuel types and found that a bed of
relationship of the time to ignition with particle shredded paper was much more prone to smol-
diameter has a ‘U’ shape. It was not possible to dering than pine needles or mulch. The experi-
ignite a bed which particle diameter was smaller mental study of Hadden et al. [24] found a
than 7 mm. For very small particle sizes, the bed relationship between ember size and the critical
exhibits poor internal convection which limits initial temperature required for ignition of a bed
the airflow and a long ignition time is required. of cellulose powder. They used steel spheres with
But as the particle sizes increase, the porosity and diameters in the range 0.8–19 mm at initial
the flow permeability of the fuel bed increase and temperatures between 500 C and 1300 C.
a minimum time to ignition of 130 min (average) Smaller embers require higher temperatures to
is required for a particle diameter of 25 mm. As initiate combustion. Their data shows two dis-
particles become larger, the inter-particle tinct boundaries as the ember temperature
Fig. 19.5 Experimentally 300

observed relationship
between time to ignition
and particle size in a bed of
250
anthracite coal (Data from
Time to ignition, min
Hadden and Rein [22])
200
150
100
50
10 20 30 40 50
Particle size, mm
588 G. Rein
increases, one separating no-ignition from smol- generation and heat loss from the system. The
dering ignition, and the other from smoldering to rate of heat loss scales with the surface area, and
flaming ignitions. The minimum ember tempera- the rate of heat generation scales with the vol-
ture at which smoldering could be initiated was ume. Consequently, as the size of a sample
550 C, and the minimum temperature for decreases, the surface -to-volume ratio4 of the
flaming ignition occurred at 650 C. Krause and smoldering front increases. Below a certain
Schmidt [25] also observed a decrease of the size, heat losses overwhelm heat generation and
critical temperature with ember size for a series ignition will not occur.
of organic powder samples (cork, beech and Palmer [27] found experimentally that the min-
cocoa), and reported a minimum temperatures imum thickness for smoldering of horizontal
of 400 C for smoldering ignition. layers of sawdust was around 10 mm. Ohlemiller
and Rogers [28] found the minimum thickness for
cellulosic insulation to be 35 mm. A more recent
Self-Heating Ignition experimental investigation of the effect of sample
size [19] is reported in Table 19.1 and Fig. 19.4.
Self-heating of a solid fuel can initiate smolder- Both the critical heat fluxes for smoldering and
ing fire deep within a pile of fuel without any flaming ignition increase with decreasing sample
external source, even at ambient temperatures. size, with smoldering ignition being significantly
Self-heating refers to the tendency of certain more sensitive to the sample size than flaming.
porous solid fuels to undergo spontaneous exo- Krause and Schmidt [25] studied the ignition of
thermic reactions in oxidative atmospheres at organic dust samples by embers, and found that
low temperatures ([2, 26], Chap. 20). It is a the larger the samples, the lower the critical ember
well-known problem in the store of large temperature. The fact that large samples are easier
amounts of carbon-rich materials (e.g., waste, to ignite than small samples has implications for
coal heaps, organic powders) and in the process testing standards and the translation of results
industries. The process is as follows: initially, the from small-scale testing to real scale.
pile of the material releases small amounts of The process of ignition is related to self-
heat by very slow oxidation at ambient tempera- sustained propagation (as discussed in the section
ture. Somewhere near the center of the pile, “Smoldering Spread”) which allows an approxi-
where the fuel is most insulated, the heat mate analytical treatment. The critical size Lc for
accumulates in the long term and results in a self-sustained propagation in a prismatic sample
sustained increase of the local temperature, of square cross-section can be estimated by
which in turn accelerates the oxidation rate. Equation 19.3 provided by Rein [9] based on
Large pile sizes and poor ventilation conditions the energy balance by Torero and Fernandez-
facilitate the buildup of heat. The process self- Pello [14] and Bar-Ilan et al. [29, 30].
accelerates for some time (hours, days or weeks)
4δU ðT s T 0 Þ
and above a certain threshold temperature it leads Lc ¼ 00 ð19:3Þ
to a thermal runaway. This results in a smolder- Qs m_ O2
ing fire that can spread from the inside to the
where δ is the smolder-front thickness perpendic-
outside, and may undergo transition to flaming
ular to the propagation direction, Ts is the peak
at a later stage when it reaches the free surface
temperature, and Qs is the heat of smoldering,
(discussed in section “Transition to Flaming”).
which all depend on the fuel. The overall heat
loss coefficient U, the ambient temperature T0 and
Size Effects and Ignition
4
The surface-to-volume ratio of a sample is inversely
There is a minimum size below which a fuel
proportional to its characteristic length (e.g., thickness
sample will not undergo ignition. This is deter- for a very wide layer, diameter for a cylinder, side length
mined by the balance between the rates of heat for a prism or diameter for a sphere).
www.ebook777.com
00
the supply of oxidizer m_ O2 depend on the geome- smoldering front [1, 29, 30]. Some of these
try and external conditions. For polyurethane results are presented in Fig. 19.6. Torero and
foam, Equation 19.3 says the critical size is around Fernandez-Pello [14] reported than this linear
160 mm [19]. If a sample is below the critical size increase breaks down for opposed propagation
Lc, sustained smoldering will not be achieved. at high airflows and stars decreasing (at ~3 mm/
Smoldering will only spread if the heat losses are s for polyurethane foam slabs of 150 mm square
reduced or the rate of heat generation is increased, cross section). In general, air is transported to the
or both. The former would involve insulating the reaction front by convection and diffusion. Con-
reaction front or supplying additional heat from an vection can be natural (buoyant) or forced. In the
external source, and the latter would involve absence of forced flow, buoyancy tends to domi-
increasing the supply of oxidizer. nate over diffusion in regions of lesser flow resis-
tance, for example near the free surface or in a
bed of large particles. The diffusion flux is domi-
Smoldering Spread nant when deep layers of a pile of fine particles is
ignited [27]. Forced convection in smoldering
The spread of smoldering is controlled by the fires usually takes the form of an air current
oxygen supply and heat transfer [1]. Conditions flowing through the fuel bed or over the free sur-
sufficient to yield smolder initiation, especially face of the bed. Forced convection directly
near an external heat source, might not be suffi- flowing through the bed of the fuel is often used
cient for self-sustained spread away from the igni- in laboratory studies (e.g., see Fig. 19.6). The
tion region. If the external heat supply continues, advantage is that it provides controllable
assisted propagation is possible. Otherwise, once conditions to investigate the phenomena also
the external heat supply ceases, smoldering reac- deeper into the bed. These studies have found a
tion will be self-sustained or lead to extinction. critical air supply rate required for self-sustaining
Experimental and modeling work has propagation (around 0.6 g/m2s of oxygen flow for
demonstrated that the smolder spread rate is lin- opposed mode in a 120 mm diameter slab of
early dependent on the total air supply rate to the polyurethane foam [30]). This critical supply
Fig. 19.6 Spread rate of smoldering assisted by airflow forced through the bed in opposed mode for a variety of
materials (After Ohlemiller [7])
590 G. Rein
rate is known to be a function of the heat losses. spread is dominated by forward smoldering
For example, microgravity experiments on the (Fig. 19.7, left), and creates a growing layer of
same polyurethane samples [30] have shown that ash that only decreases if wind carries the
the removal of lateral heat losses by natural con- particles away. The horizontal spread is
vection allows self-sustained propagation at sig- enhanced by a direct supply of atmospheric oxy-
nificantly lower oxygen supplies (0.3 g/m2s). gen, which is readily available, and the rate is
When a bed of fuel is ignited locally, in gen- significantly greater than downward spread
eral the spread will be multidimensional and where oxygen transfer is limited by the layer of
include both horizontal and vertical spread. ash and char. Combustion of the uppermost layer
Each front will be dominated by forward or is typically quenched due to large convective and
opposed propagation (or a combination) radiant heat losses; this leaves a very thin layer of
depending on the roles of buoyancy, wind and charred material on top while smoldering
diffusion. Two important configurations are continues below. The fact that horizontal spread
presented in Fig. 19.7 where the spread is either is faster a few cm below the surface leads to the
downwards or upwards though the fuel bed. formation of an overhang pointing inwards at
If initiation occurs on the top surface of the the rim of the void (overhang not shown in
fuel bed, the fire will spread laterally and down- figure 19.7). Horizontal spread accelerates in
ward. It leads to a void or hole in the general response to an increased wind. Palmer [27] exam-
shape of a semi-ellipsoid or pan. Downward ined this in thin horizontal layers (3–57 mm) of
Fig. 19.7 Diagrams of downward (left) and upward (right) propagation in a porous fuel bed Huand and Rein [31]
Fig. 19.8 Smoldering

spread rate through
horizontal layers of
sawdust as a function of the
horizontal airflow over the
topmost layer. Circles:
120 mm particle size;
triangles: 190 μm particle
size; squares: 480 mm
particle size (Data from
Palmer [27], after
Ohlemiller [7])
www.ebook777.com
various cellulosic particles (cork, pine, beech,

400
grass). The results in Fig. 19.8 show that the
horizontal airflow has a greater effect in forward
propagation, but that airflow in opposed propaga- 200
tion and the particle size have a weaker effect. In
the absence of any forced flow over the fuel layer,
the flow induced by the buoyant plume supplies
Time for smoldering to penetrate heap (hr)

100
oxygen for the horizontal spread at the topmost 70
fuel layer. Oxygen then penetrates into the layer
mostly by diffusion [32].
40
If initiation occurs deep within a layer of fuel
and the nearest free surface is on the top, the fire
will slowly spread upwards dominated by 20
opposed smoldering (Fig. 19.7, right). The thick
layer of virgin fuel above the reaction front
hinders the oxygen supply, but also reduces the 10
rate of heat losses. Ultimately, the spread is faster 7
towards the free surface driven by oxygen
diffusion, thus leading to fronts in the shape of 4
an elongated bell. The reaction front usually
spreads without fully consuming the char left
behind. The upward case was studied in Palmer’s 2
work [27], which consists of a collection of
observations from simple experiments involving
the initiation of a smoldering front at the base of 1
4 7 10 20 40 70 100
sawdust heaps (cork, elm and mixed wood). Depth of fuel heap (cm)
Some of the results are shown in Fig. 19.9.
Note the scales reported in this data; the time to Fig. 19.9 Upward spread of smouldering through a bed
smolder up through a layer 1 m deep is about driven by air diffusion. Ignition at base of the fuel bed and
spread upwards in heaps of wood sawdust. Squares: layer
2 weeks, and the process gave little hint of its 0.025 m deep in 0.3 m square box; diamonds: layer
presence until it was close to the surface of the 0.052 m deep in 0.3 m square box; triangles: layer
fuel heap (smoke gets trapped inside the porous 0.052 m deep in 0.6 m square box; circles: layer
bed). The slope of the curve indicates that the 0.052 m deep in 0.9 m square box (Data from Palmer
[27], after Ohlemiller [7])
time for upward smolder to penetrate a fuel layer
is proportional to the square of the layer depth.
Palmer showed that such dependence suggests governs the front structure and is ultimately
the smolder front spread is proportional to the responsible for determining the conditions
diffusion rate of oxygen from the free surface, under which a material ignites and extinguishes.
through the unburned fuel, to the reaction front. Smoldering combustion of a solid fuel involves
Data from other experiments on a variety of multiple pathways to chemical reactions, and
fuels and air supply conditions are summarized in these pathways are not yet fully understood. In
Table 19.2. spite of the complex kinetic behavior, experimen-
tal evidence suggests that mechanisms consisting
of only a few global reactions capture the most
Smoldering Kinetics important characteristics of the chemical process
and allow an approximate analysis. Smoldering
The spread rate of self-sustained smoldering is chemistry in its simplest form can be understood
typically controlled by oxygen transport and heat as a two-step process: pyrolysis of fuel (Equa-
transfer. Yet, heterogeneous chemical kinetics tion 19.1) produces the char that is then oxidized
Table 19.2 Experimental data on smoldering in various fuels and configurations (After Ohlemiller [7])
Peak
Fuel Configuration Air supply Spread rate temperature Reference Comments
Pressed fiber 13 mm thick Natural 0.8–1.3 NA Palmer [27] Smolder velocity
insulation horizontal strip, convection mm/min increased 50 %
board, width large for strips with
230–290 kg/m3 compared to width thickness
thickness
Pressed fiber 13 13 mm Natural 1.6–2.8 NA Palmer [27] Smolder velocity
insulation strip at varied convection mm/min highest for upward
board, vertical spread; lowest for
230–290 kg/m3 inclinations horizontal spread
Fiber insulation 13 50 mm Airflow 2.1 mm/min 790 C Palmer [27] Some samples
board strip forward over (20 cm/s (900 cm/s extinguished due
smolder 20–1500 airflow) airflow) to cooling at
cm/s 7.8 mm/min airflows
(1400 cm/s >1450 cm/s
airflow)
Fiber insulation 13 50 mm Airflow 1.7–2.1 NA Palmer [27] Extinguishment
board strip opposed over strip, mm/min >900 cm/s
smolder 80–900
cm/s
Fiber insulation 13 mm thick Airflow 1 mm/min NA Brenden and
board (pine or sheet, horizontal, over sheet, Schaffer [33]
aspen) forward smolder 10–18
cm/s
Cardboard Vertical rolled Natural 3–5 mm/min NA Kinbara et al Small dia.
cardboard convection [34] 2 faster than
cylinder, large dia.; ambient
downward temp. effect
propagation, measured
diameter
1.9–3.8 mm
Shredded 8 mm diameter Natural 1.8–3 mm/min 820 C Egerton
tobacco cigarette, convection et al. [35]
horizontal,
in open air
Cotton Double fabric Airflow ~6 mm/min 770 C Donaldson and Smolder behavior
upholstery layer, 2 mm over Yeadon [36] dependent on
fabric thick, horizontal, fabric, alkali metal
forward smolder 10 cm/s content
Cellulosic Various weight Natural 1.8–45 mm/min Reported Donaldson and Smolder fastest on
upholstery fabrics convection (depends on values Yeadon [36], inert fiberglass
fabric on horizontal on substrate) suspiciously Stiefel substrate
substrates fiberglass, PU low et al. [37]
foam, cotton
Cotton batting, 15 cm cube, hold Natural 2–4 mm/min 690 C Hagen Lower densities or
densities together by metal convection (decrease as et al. [38] repeated heating of
5–100 kg/m3 mesh and open to density sample result in
the air on all increases) higher ignition
sides temperature
Wood char, 22 cm tall Natural Upward Upward He and Downward peak
densities packed bed of convection ~0.4 mm/min. 530 C. Behrendt [39] temperature
290–435 kg/m3 particles, Downward Downward decreases as the
diameter ~0.05 mm/min 800 C height of ash layer
1–3 mm increases
Cork, beech Mesh wire Natural 0.1–1.5 260–375 C Krause and Ignition sources
and cocoa baskets with convection mm/min and and Schmidt [25] tested include hot
powders volumes of decreasing decreasing body, glowing nest
0.8–200 l relationship relationship and electric coil
with basket with basket
volume volume
www.ebook777.com
in situ (Equation 19.2a). In this section we explore forms a significant amount of char during ther-
more comprehensive kinetics. mal decomposition. The fuel is oxidized as well
To begin with, the simple two-step scheme (Equation 19.4) but the most exothermic reaction
proposed in Equations 19.1 and 19.2a can be is that of char, and therefore the simplest overall
extended to include the direct oxidation of the mechanisms does not include fuel oxidation.
fuel. Char oxidation (Equation 19.2a) is the prin- Fuel oxidation is also related to self-heating
cipal heat source in most self-sustained smolder when it takes place at low temperature close to
propagation processes; the potential for smolder- ambient (see Chap. 20).
ing combustion thus exists with any material that
Direct heterogeneous oxidation of the fuel:
Fuel ðsolidÞ þ O2 ! Heat þ CO2 þ H2 O þ other gases þ Char ðsolidÞ þ Ash ðsolidÞ
ð19:4Þ
The materials for which smoldering kinetics mechanism for polyurethane consisting of two
are best known are polyurethane foam and cellu- foam pyrolysis, two foam oxidations and one
lose. Kashiwagi and Nambu [40] provided a char oxidation reaction, and accounting for four
quantified three-step mechanism for cellulose, solid species (foam, β-foam, char and residue).
including cellulose pyrolysis, cellulose oxidation This mechanism was developed and the kinetic
and char oxidation and accounting for three solid constants found from thermogravimetric
species; cellulose, char and ash. In flexible poly- experiments, as shown in Fig. 19.10. This
urethane foam, the presence of oxygen during multi-step mechanism allows explaining the
degradation plays another key role, because different contributions of the pyrolysis and
without oxygen, many foams do not form char oxidation reaction to the degradation of the
[41]. Rein et al. [15] provided a five-step foam in the presence of air, as seen in Fig. 19.11.
a b 6
6
(Air Atmosphere) 20°C/min
(Nitrogen Atmosphere) 20°C/min
5 5
Mass-loss rates [1/s] x 103
Mass-loss rates [1/s] x 103
4 4
3 3
10°C/min 10°C/min
2 2
1 1 5°C/min
5°C/min
0 0
100 200 300 400 100 200 300 400
Temperature [°C] Temperature [°C]
Fig. 19.10 Thermogravimetric mass loss rate of poly- rates (Symbols are data from experiments of Chao and
urethane foam in; (a) nitrogen atmosphere, and (b) air Wang [42], and lines are data from numerical simulations
atmosphere, as a function of temperature for three heating of Rein et al. [15])
594 G. Rein
a b
12 12
10 10
foam
Oxidation rates ω [1/s] x 103

Pyrolysis rates ω [1/s] x 103
8 8
β-foam
char
6 6
4 4
foam
2 β-foam 2
0 0
100 200 300 400 100 200 300 400
Temperature [°C] Temperature [°C]
Fig. 19.11 Numerical results for the different reactions This simulates the experimental thermogravimetric data
rates (pyrolysis on the left, oxidation on the right) in of Fig. 19.10 (right) and allows explaining the origin of
polyurethane foam decomposition in air atmosphere. the three mass-loss peaks (Data from Rein et al. [15])
15
foam
20
β-foam
Heat released
rate [kJ / m3]
10 char
10
5
0
0
400 t = 900s t = 600s
450
0.2
Temperature [°C]
0.2
200
Oxygen
0.1 0.1
350
0
55 65 75 40 60
Distance from Igniter [mm] Distance from Igniter [mm]
Fig. 19.12 Numerical results for the front structure dur- released rate of each reaction (positive for oxidation,
ing self-propagation; (left) opposed smoldering; and negative for pyrolysis). Bottom figures show the tempera-
(right) forward smoldering. Top figures show the heat- ture and oxygen profiles (Data from Rein et al. [16])
The separation of pyrolysis from oxidation is right), the oxidation and the pyrolysis reactions
essential in any smouldering kinetics scheme. form two distinct propagating sub fronts. The
It was the work of Rein et al. [16] that proved pyrolysis sub front arrives first to the virgin
that the same kinetic mechanism was able to foam and then followed by the oxidation sub
explain both forward and opposed smoldering. front. In opposed smoldering (Fig. 19.12, right),
In forward smoldering propagation (Fig. 19.12, the oxidation and the pyrolysis reactions overlap
www.ebook777.com
to form a single front. Previously to the work of Hadden and Rein [22] investigated three water
Rein et al. [16], smoldering chemistry had been suppression methods (pipe, shower and spray) on
described as a function of the propagation mode: a small-scale coal bed. They showed that the most
forward smoldering with two-step chemistry, and efficient method with respect to total water
opposed smoldering with one-step chemistry. required is the shower. However, using a spray
Smoldering kinetics is an immature field of results in less water runoff and thus offers a
solid phase chemistry due to is complexity and higher efficiency. The injection pipe is signifi-
secondary role in fire spread. It has been the cantly less efficient, requiring three times more
objective of few studies to date. Despite the water than a spray, and resulting in >80 % lost as
recent advances reviewed in this section, the run-off. Tuomisaari et al. [45] tested a number of
topic remains mostly undeveloped. extinguishing agents (liquids: water, water with
additives; gases: N2, CO2, Ar and Halon) in series
of tests on a small bed of smoldering wood chips.
Suppression The result was that gaseous CO2, injected from
the bottom, was found to be the most effective.
A smoldering fire can be extraordinarily difficult
to suppress. Experiments on heaps of coal show
that smoldering requires large amounts of water.
For example, the amount of water required to Gas Emissions
suppress smoldering coal was measured to be in
the range from 1 to 2 l of water per kg of burning Gas emissions from smoldering fires differ sig-
fuel. Moreover, smoldering requires lower oxy- nificantly to those from flaming fires. First, the
gen concentration to be smothered, around 10 % emissions rate per unit area is much lower but
O2, compared to 16 % O2 for flaming [43, 44]. also the chemistry is different. Smoldering is
Oxygen removal is insufficient unless it is characteristically an incomplete combustion,
continued until the whole fuel bed is cooled to a releasing species and quantities that substantially
point where oxygen readmission will not cause depart from that in stoichiometric and complete
re-ignition. Because volumetric cooling of a fuel combustion. For example, the CO/CO2 ratio
bed is a very slow process (long thermal response which can be thought of as an index of the
time), this means that the holding time for smoth- incompleteness of combustion is ~0.4 in smol-
ering are much longer for smoldering than for dering but ~0.1 in flaming combustion [46]. The
flaming5 (months vs. hours) [22]. presence of pyrolysate in the products of smoul-
One practical problem in suppressing a large dering, significantly contributes to of a complex
fuel bed is the tendency of the extinguishing fluid gaseous mixture including volatile organic
agents to follow higher permeability channels and compounds (VOC), polyaromatic hydrocarbons
thereby miss significant in-depth burning zones. (PAH), other hydrocarbons and particulate mat-
Channeling arises when a substantial fraction of ter (PM). While the yield of toxic species is
the fluid takes the same flow path through the bed, larger in smoldering fires than in flaming fires
resulting in limited contact surface area between [47], the production rate, which is proportional to
the agent and the burning fuel. This, coupled with the spread rate, is much lower. This means that
the lower residence times in regions of high per- inside an enclosure, a smoldering fire of long
meability because of the high flow velocities, duration (in the range from 1 or 3 h for a single
requires large quantities of water for suppression. bedroom size compartment [48]) can lead to a
lethal dose of toxicity, especially CO. But there
are not as yet sufficient data on the toxicity of
smoldering materials to definitively understand
5
Avoidance of flaming re-ignition of a non-porous fuel the issue of life safety. Some more information is
requires cooling of the surface layer only. presented in Chap. 62 and in [47].
596 G. Rein
Limited information is available on the aero- responsible for up to 50 % or more of the total
sol emitted by a smolder source. The residual burned biomass during wildfires [51–53].
char left behind the smolder front and the origi- But it is the soil rich in organic matter, in
nal porous bed act as filters for aerosol. This particularly that in peatlands, which is most
finding explains the observation by Palmer [27] affected ecosystem by smoldering fires, both in
that upward smoldering in a thick layer of fuel frequency and size. Peat soils are made by the
was not detected until it neared the surface (like natural accumulation of partially decayed bio-
in Fig. 19.7 right). The mean particle size of the mass and are the largest reserves of terrestrial
aerosol in smoldering cellulose was measured to organic carbon [54]. Because of this vast accu-
be in the range from 2 to 3 μm [49]; this is about mulation of fuel, once ignited, smoldering peat
50–200 times larger than the sooty particulates fires burn for very long periods of time (e.g.,
produced by flaming combustion. months, years) despite extensive rains, weather
The low heat release rate of smoldering fires changes or fire-fighting attempts. These are the
means that the buoyant plume is weak, which has largest fires on Earth and large contributors of
implications on the location of smoke detectors in greenhouse gases [5, 54]. Peat fires occur with
rooms. The morphology of the smoke from some frequency worldwide in tropical, temper-
smouldering is different to smoke from flaming, ate and boreal regions (e.g., in Indonesia,
and this affects smoke detection. This means Australia, Alaska, Canada, Florida, British
smoke is slow to reach the ceiling, or it may Isles, Siberia). Droughts, drainage and changes
never reach it, and often the building mechanical in land use are thought to be main causes lead-
ventilation controls the smoke movement [50] ing to the high flammability conditions of dry
(see also Chap. 13). peatlands. Possible ignition events can be natu-
In the natural environment, the low buoyant ral (e.g., lightning, self-heating, volcanic erup-
strength of large smoldering fires leads to haze tion) or anthropogenic (land management,
episodes because the plume accumulates near accidental ignition, arson).
the ground and can slide into populated areas, The most studied peat megafire took place in
choking towns and cities for weeks [5]. The Indonesia in 1997 and led to an extreme haze
composition of haze measured by Bertschi event. The smoke covered large parts of South-
et al. [51] in the tropical savanna shows it East Asia, even reaching Australia and China,
contains 130 % more CO and 670 % more and induced a surge of respiratory emergencies
hydrocarbons in mass basis, but 15 % less CO2 in the population and disruption of shipping
and no NOx when compared to the flaming and aviation routes for weeks. It was estimated
wildfires. that these fires released between 0.81 and
2.57 Gton of carbon gases [55]. This is equiva-
lent to 13–40 % of global man-made emissions of
Smoldering Wildland Fires the year 1997. The 1997 megafire was not an
isolated case in the region. Analysis of
In the natural environment, smoldering fires 1960–2006 data shows that smoldering haze
burn two types of biomass: thick fuels, like tree episodes have drifted to South East Asian
branches or logs, and organic soil, like the duff countries once every 3 years on average
layer or peat [5]. These are characterized by hav- [56]. Rough figures at the global scale estimate
ing a significantly greater thermal time compared that the average greenhouse gas emissions from
to fine fuels like foliage. Thick fuels favor the peat fire is equivalent to >15 % of man-made
slow burning of smoldering combustion. The per- emissions. Moreover, because peat is ancient
sistent smoldering of thick fuels is typically carbon, and smoldering is enhanced under
observed for a few days after a flaming wildfire warmer and drier climates, it creates a positive
has passed, and it is often referred to as residual feedback mechanism in the climate system, a
smoldering combustion. Overall, smoldering is self-accelerating global process [5].
www.ebook777.com
Fig. 19.13 The data

separates the ignition
(bottom) from the no-
ignition (top) limits for a
mixture of peat, moisture
and mineral contents.
Circles and sequre symbols
are experimental data by
Frandsen [58]. Lines are
computational
simulations by [59].
Because the water content of wildland fuels because the inert content is a heat sink to the fire.
like peat can vary naturally over a wide range of The results are shown in Fig. 19.13. This rule can
values (from dry to flooded in water), and be applied to most organic soils or fuel beds to
because water represents a significant energy determine if they are susceptible to smoldering.
sink, moisture content is the single most impor- Any soil which composition is more than 80 %
tant property governing the ignition and spread mineral, cannot be ignited [58, 59]. After mois-
of smoldering wildfires. The critical moisture ture and mineral contents, other important
content for ignition (related to the moisture of properties are bulk density, porosity, flow perme-
extinction [57]) of boreal peat has been measured ability and organic composition.
in the range 110–120 % in dry basis6 [16, Because the fuel layers found in the natural
58]. Any peat drier than this is susceptible to environment (soil depths from 0.5 to 30 m) can
smoldering. The prominent role of moisture is be much thicker than those in the built environ-
such that natural or anthropogenic-induced ment (~0.1 m), smoldering wildfires can be clas-
droughts are the leading cause of smoldering sified in shallow or deep fronts. Each has
megafires. significantly different dynamics because of the
The second most important property is the different role played by the controlling
mineral content.7 As found experimentally by mechanisms of oxygen supply and heat losses.
Frandsen [58] and computationally by Huang Organic material located close to the surface
et al [59], there is a decreasing relationship of the soil burns in shallow fires (roughly <1 m
between the mineral content and the critical under the surface). These have a good supply of
moisture content: higher mineral loads mean atmospheric oxygen, but are exposed to large
soil can only ignite at lower moistures. This is convective heat losses. They propagate laterally
and downwards along the organic layers of the
ground, leave voids or holes in the soil (see
6
The water content in dry basis is the mass of water Fig. 19.7 left). This has prompted that fuel con-
divided by the mass of a dried sample expressed as a %.
7 sumption can be estimated using the depth of burn
The mineral content is the % of the fuel mass (on dry
basis) that will not burn or react at high temperatures. It to calculate the volume of the void. Depth of burn
results in ash. is the vertical distance between the original soil
598 G. Rein
location and the post-fire soil location. A typical wildfires. Due to the higher density and higher
value for the depth of burn reported in several fraction of carbon in coal, the fire spreads slower
field studies is around 0.5 m, which means that the (~0.1 mm/min) but burns hotter (~1000 C peak
average fuel consumptions per unit area is around temperature) compared to other smoldering fuels.
75 kg/m2. This value is more than two orders of Because of these characteristics, coal particles
magnitude larger than that in flaming fires [5]. For smoldering near the free surface are more fre-
this reason and in terms of fuel consumption, quently accompanied by flames than any
these are classed as megafires. other fuel.
Deep fires take place in organic the subsurface There is evidence to suggest that burning coal
layers fed by oxygen infiltrating the ground via seams are the oldest continuously burning fires
large cracks, piping systems or channels. For exam- on Earth. The best examples is The Burning
ple, it is known that peatlands have an in-built Mountain in New South Wales, Australia is a
natural piping system for water movement; when large coal seam which has been smoldering for
drained this system becomes the delivery network more than 6000 years [12]. Thousands of under-
for oxygen transport to deeper locations and smoke ground coalmine fires have been identified
exhaust to the atmosphere. Deep fires have a poorer around the world, especially in China, the United
supply of atmospheric oxygen but are better States, and India [63]. Elusive, unpredictable and
insulated from heat losses than shallow fires. costly, coal fires burn indefinitely while there is
Smoldering fires have detrimental effects on fuel, choking the life out of a community and the
the forest soil, its microflora and microfauna. environment while consuming a valuable energy
This is because it consumes the soil (>90 % resource. The associated financial costs run into
mass loss) and also because the long residence millions of dollars including the loss of coal,
time of smoldering means that heat penetrates closure of mines, damage to the environment
deep into the soil layers [5]. On the contrary, and fire-fighting efforts. There are some well-
flames produce high temperatures above the documented cases. In 1962, an abandoned mine
ground for short periods of time (in the order of pit in Centralia, Pennsylvania, USA was acciden-
15 min). This results in minimal heating of the tally lit. Several unsuccessful attempts were
soil below depths of a few cm, reaching peak made to extinguish it, letting the fire continue to
temperatures of 300 C at superficial layers burn after more than 40 years. It is currently
(<10 mm) and below 80 C at layers more than being monitored with the front advancing at
40 mm deep [60]. This superficial heating can ~20 m/year [64].
leave the soil system relatively unharmed.
However, smoldering fires lead to enhanced
heat transfer into the soil (see the discussion on Transition to Flaming
the depth of burn above) for much longer
durations (i.e. in the order of 1 h) and peak Smoldering and flaming combustion are
temperatures of 500 C [61]. As a comparison, closely related, and one can lead to the other.
these thermal conditions are more severe than Figure 19.14 shows snapshots of an experiment
medical sterilization treatments, and mean that on the transition to flaming in a foam slab. The
the soil is exposed to conditions that are lethal to transition from smoldering to flaming is a severe
biological agents. threat in residential fires, and together with the
slow buildup of CO and other incapacitating
gases, they pose the largest life hazards of bed-
Coal Seam Fires ding or upholstery fires. The transition to flaming
is also a concern in wildfires, since it offers the
Coal fires are a smoldering phenomenon that has means for flaming fronts to re-establish in unex-
attracted a larger amount of research, especially pected locations (e.g., across a fire break) and at
from geologists and petrologists [37, 62] and unexpected times (e.g., long after burn out of the
share many similarities with deep smoldering initial flaming front).
www.ebook777.com
Fig. 19.14 Series of photographs of a combustion exper- only half the sample has burned (Photo 1, far left). When
iment illustrating the transition to flaming in a smoldering the rapid transition to flaming takes place (Photos 2 and
polyurethane slab 40 cm high under external forced flow 3), the whole sample is engulfed in flames in just a few
(Photo by group of Prof. Carlos Fernandez-Pello, Univer- seconds (Photos 4 and 5, far right)
sity of California at Berkeley). After 1 h of smoldering
However, it has received little attention so the flame (Equation 19.2b) [2, 32]. The transition
current understanding of this transition is limited. occurs not at the foam/air interface, but rather
In particular, whether the transition is possible or within the depth of the char left by the smolder
not or when it will occur cannot be predicted front [66, 67]. This is the location where both
because of a lack of valid models. Babrauskas critical conditions of flammability of the pyroly-
and Krasny [65] did a literature survey of fire sate mixture and the net excess of heat released
tests on a variety of upholstered furniture items by char oxidation are met. The heat driving the
(chairs, sofas and mattresses) where a smoldering transition is released mostly by the secondary
fire was ignited and allowed to progress. Out of char oxidation [14, 66], which is more exother-
the 102 fire tests examined, 64 % did transition to mic and takes places at higher temperatures
flaming. The time to transition observed ranged (~700 C for polyurethane [68]) than char oxida-
from 22 to 306 min, with an average of 88 min. tion in Equation 19.2a. This temperature range is
Revisiting the data from previous experiments on towards the highest temperatures typical of smol-
smoldering of furniture, Quintiere et al. [48] dering combustion and hence secondary char
showed that the likelihood of having transition oxidation is only of importance during the tran-
to flaming occur peaks with a 36 % probability at sition to flaming [66].
50–100 min after ignition. They also showed that A smoldering fuel responds to an increased
within the time period from 50 to 200 min, the oxygen supply (e.g., wind) by becoming faster
likelihood of a transition event is comparable and hotter until, eventually, flames might erupt.
with the likelihood of reaching a lethal CO dose This effect was first studied quantitatively by
inside a single bedroom compartment. Palmer [27] for airflows over horizontal layers
The transition is a spontaneous gas-phase of wood sawdust. Depending on the material, the
ignition supported by the smoldering reaction, transition occurred at airflow velocities from
which acts both as the source of gaseous fuel, about 0.9 to 1.7 m/s. Ohlemiller [69] did obtain
the pyrolysate produced in Equation 19.1, and transition to flaming in layers of fibrous insula-
the source of the heat required to initiate the tion materials in forward smolder at air velocities
600 G. Rein
conductive, embers and self-heating) and is more

3 difficult to suppress than flaming combustion.
flow velocity (m/s)
Transition not This makes smoldering the most persistent com-

2 possible bustion mode. Large samples are easier to ignite
than small samples, which affects the translation
1 of laboratory testing results to the real scale.
Transition
Smoldering fuels include synthetic materials
is possible
such as charring polymers or cellulosic insula-
0.3 0.35 0.4 tion, and natural fuels like peat or coal. In chem-
oxygen fraction ical terms, most smoldering fuels form a char on
heating. In physical terms, smoldering fuels con-
Fig. 19.15 Transition map from smoldering to flaming sist of a permeable medium formed by grains,
on polyurethane samples (Data from [70]) fibers or some other porous matrix. The chemis-
try involves multiple pathways to chemical
of about 2 m/s. Bar-Ilan et al. [70] conducted reactions, but in its simplest form can be under-
experiments on small samples of polyurethane stood as a two-step process: the pyrolysis of fuel
foam and showed that the transition can be trig- and char oxidation. It is characteristically an
gered by a combination of increasing the airflow incomplete combustion reaction, releasing
velocity, the oxygen concentration or the exter- pollutants in greater yields than flaming but at a
nal radiant heat. A map of airflow vs. oxygen lower rate. The transition from smoldering to
fraction is shown in Fig. 19.15. A similar map flaming is a severe threat and can be triggered
is presented in Bar-Ilan et al. [70] for airflow by a combination of increasing airflow, oxygen
vs. external heat. They also provide an analytical concentration and external heat.
expression for the transition boundary. An emerging topic is smoldering wildland
The transition to flaming has only been fires of organic soils, like peat, which are
observed in forward propagation. Ohlemiller megafires in terms of the fuel consumption, and
[69] argued that this is because in forward together with coal seam fires are the largest and
mode, the hot gases of combustion preheat the oldest continuously burning fires on Earth.
fuel ahead and thereby enhancing the combus- Regrettably, the state of the art is incomplete,
tion; while in opposed mode, the cool air reduces and accurate calculations on smoldering phe-
preheating. However, upward propagation inside nomena are not possible. Current understanding
a pile of fuel (Fig. 19.7 right), which is mostly in is fragmented in different areas, as suggested by
opposed mode, can lead to the transition to the many different names applied to it:
flaming once the reaction breaks through to the non-flaming combustion, oxidative pyrolysis,
free surface and finds a good supply of oxygen. glowing combustion, filtering combustion, low
oxygen combustion, deep seated fire, hidden
fire, lagging fire, smoking material, in-situ com-
Concluding Remarks bustion, fireflood or underground gasification
[9]. More experimental and theoretical studies,
We have seen that smoldering fires are a source and a multidisciplinary research approach to
of safety and environmental concerns. It is a smoldering combustion are needed.
type of combustion quite distinct from that in
flaming fires, but as diverse and complex. The Acknowledgements I am most grateful to Thomas
most studied fuels to date are cellulose and J. Ohlemiller whose pioneering research and reviews on
polyurethane foam. smoldering combustion provided not only the best sources
The most important concepts to bear in mind of knowledge and inspiration for my own research and
understanding but also contributed to this chapter in the
are the following. Smoldering combustion can be form of material that I reused from his previous version
initiated with weaker ignition sources (radiant, [7]. This chapter is the culmination of 15 years of work on
www.ebook777.com
smoldering combustion that I have conducted at four 11. J. Beard, Ground Zero’s fires still burning,
universities. These are, in chronological order University 3 December 2001, New Scientist, New York. http://
of Texas at Austin, University of California at Berkeley, www.newscientist.com/article/dn1634-ground-zeros-
University of Edinburgh and Imperial College London. fires-still-burning.html
This work would not have been possible without the 12. CD Ellyett, Fleming, A.W. (1974) Thermal infrared
contributions and encouragements of my mentors, imagery of The Burning Mountain coal fire. Remote
collaborators and PhD students. Of these, I am especially Sensing of Environment 3: 79–86; doi: 10.1016/0034-
thankful to Carlos Fernandez-Pello, José Torero, Rory 4257(74)90040-6.
Hadden, Claire Belcher, Chris Lautenberger, Matt Davies 13. C. Switzer, P Pironi, G Rein, JL Torero, JI Gerhard,
and Xinyan Huang. A series of institutions have funded Self-Sustaining Smoldering Combustion: A Novel
this work along the way: Royal Academy of Engineering, Remediation Process for Non-Aqueous-Phase Liquids
Leverhulme Trust, UK Engineering and Physical Science in Porous Media, Environmental Science and Technol-
Research Council, IFIC Forensics, Met Office and NASA ogy 43, pp. 5871–5877, 2009. doi: 10.1021/es803483s.
Space flight Program. 14. J. Torero and C. Fernandez-Pello, Forward Smolder
of Polyurethane Foam in a Forced Air Flow, Combus-
tion and Flame, 106, pp. 89–109 (1996).
15. G Rein, C Lautenberger, AC Fernandez-Pello, JL
References Torero, DL. Urban, Application of Genetic
Algorithms and Thermogravimetry to Determine the
1. TJ. Ohlemiller, “Modeling of Smoldering Combus- Kinetics of Polyurethane Foam in Smoldering Com-
tion Propagation,” Progress in Energy and Combus- bustion, Combustion and Flame 146 (1-2), pp 95–108,
tion Science, 11, p. 277 (1985). 2006. doi:10.1016/j.combustflame.2006.04.013.
2. D. Drysdale, An Introduction to Fire Dynamics, 3rd 16. G Rein, AC Fernandez-Pello, DL Urban, Computa-
ed. Wiley, UK, 2011. tional Model of Forward and Opposed Smoldering
3. M. Ahrens. U.S. Home Structure Fires, National Fire Combustion in Microgravity, Proceedings of the
Protection Association Fire Analysis and Research, Combustion Institute 31 (2), pp. 2677–2684, 2007.
Quincy, MA April 2013. doi:10.1016/j.proci.2006.08.047.
4. J.R. Hall, The Smoking-Material Fire Problem, (Fire 17. TJ. Ohlemiller and D. Lucca, “An Experimental Com-
Analysis and Research Division of The National Fire parison of Forward and Reverse Smolder Propagation
Protection Association, November 2004, in Permeable Fuel Beds,” Combustion and Flame,
Massachusetts). 54, p. 131 (1983).
5. G Rein, Smouldering Fires and Natural Fuels, 18. M. Ortiz-Molina, T-Y. Toong, N. Moussa, and
Chapter 2 in. Fire Phenomena in the Earth System – G. Tesoro, 17th Symposium (International) on Com-
An Interdisciplinary Approach to Fire Science, C bustion, Combustion Institute, Pittsburgh, PA (1979).
Belcher (editor). Wiley and Sons, 2013. http://dx. 19. R Hadden, A Alkatib, G Rein, JL Torero, Radiant
doi.org/10.1002/9781118529539.ch2 Ignition of Polyurethane Foam: the Effect of Sample
6. T’ien, J.S., Shih, H., Jiang, C., Ross, H.D., Miller, F. Size, Fire Technology 50 (3), pp. 673–691 (2014)
J., Fernandez-Pello, A.C., Torero, J.L., and Walther, doi:10.1007/s10694-012-0257-x.
D.C., Mechanisms of Flame Spread and Smolder 20. M. Anderson, R. Sleight, and J. Torero, “Downward
Wave Propagation, Chapter 5 in. Microgravity Com- Smolder of Polyurethane Foam: Ignition Signatures,”
bustion: Fire in Free Fall, H.D. Ross, Editor, Aca- Fire Safety Journal, 35, pp. 131–147 (2000).
demic Press, 2001, pp. 299–417. 21. R. Anthenien and C. Fernandez-Pello, A Study of
7. T. J. Ohlemiller, Smoldering Combustion, Chapter 9; Forward Smolder Ignition of Polyurethane Foam,
Section 2; SFPE Handbook of Fire Protection Engi- Proceedings 27th Symposium (International) on Com-
neering. 3rd Edition, DiNenno, P. J.; Drysdale, D.; bustion, Vol. 2, Combustion Institute, Pittsburgh, PA,
Beyler, C. L.; Walton, W. D., Editor(s), 2/200-210 p., pp. 2683–2690 (1998).
2002. http://fire.nist.gov/bfrlpubs/fire02/art074.html 22. R Hadden, G Rein, Burning and Suppression of
8. V. Babrauskas, Ignition Handbook, 2003 Fire Smouldering Coal Fires, Chapter 18 in. Coal and
Science Publishers, Issaquah WA, USA. ISBN-10: Peat Fires: A Global Perspective, Volume 1, pp.
0-9728111-3-3. 317–326, Stracher, Prakash and Sokol (editors),
9. G Rein, Smouldering Combustion Phenomena in Sci- Elsevier Geoscience, 2011. ISBN 9780444528582.
ence and Technology, International Review of Chem- doi: 10.1016/B978-0-444-52858-2.00018-9.
ical Engineering, 1, 1, pp. 3–18, 2009. 23. Manzello, S. L.; Cleary, T. G.; Shields, J. R.;
10. TJ. Ohlemiller and W. Shaub, “Products of Wood Maranghides, A.; Mell, W. E.; Yang, J. C., Experi-
Smolder and Their Relation to Wood-Burning mental Investigation of Firebrands: Generation and
Stoves,” NBSIR 88-3767, National Bureau of Ignition of Fuel Beds, Fire Safety Journal, Vol.
Standards, Washington, DC (1988). fire.nist.gov/ 43, No. 3, 226–233, April 2008.
bfrlpubs/fire88/PDF/f88017.pdf 24. R Hadden, S Scott, C Lautenberger and AC
Fernandez-Pello, Ignition of Combustible Fuel Beds
602 G. Rein
by Hot Particles: An Experimental and Theoretical Pages 406–413, ISSN 0379-7112, http://dx.doi.org/
Study, Fire Technology 47 (2), pp. 341–355, 2011, 10.1016/j.firesaf.2011.06.007.
doi:10.1007/s10694-010-0181-x 40. T. Kashiwagi, H. Nambu, Global kinetic constants for
25. U Krause, Schmidt M (2000) Propagation of smoul- thermal oxidative degradation of a cellulosic paper,
dering in dust deposits caused by glowing nests or Combustion and Flame, Volume 88, Issues 3–4,
embedded hot bodies. J Loss Prev Process Indus 13 March 1992, Pages 345–368, http://dx.doi.org/10.
(3-5):319–326. doi:10.1016/S0950-4230(99)00031-5 1016/0010-2180(92)90039-R.
26. P. Bowes, Self-Heating: Evaluating and Controlling 41. F. Rogers and T. Ohlemiller, “Smolder Characteristics
the Hazards, Elsevier, New York, Chap. 7 (1984). of Flexible Polyurethane Foams,” Journal of Fire and
27. K. Palmer, “Smoldering Combustion in Dusts and Flammability, 11, p. 32 (1980).
Fibrous Materials,” Combustion and Flame, 42. C.Y.H. Chao, J.H. Wang, 2001a, Comparison of the
1, p. 129 (1957). Thermal Decomposition Behavior of a Non-Fire
28. TJ. Ohlemiller and F. Rogers, “Cellulosic Insulation Retarded and a Fire Retarded Flexible Polyurethane
Material II. Effect of Additives on Some Smolder Foam, Journal of Fire Science 19, pp. 137–155.
Characteristics,” Combustion Science and Technol- 43. C Belcher, J Yearsley, R Hadden, J McElwain, G Rein,
ogy, 24, p. 139 (1980). Baseline intrinsic flammability of Earth’s ecosystems
29. A. Bar Ilan, Rein G, Fernandez Pello AC, Torero JL, estimated from paleoatmospheric oxygen over the past
Urban DL (2004a) Forced forward smoldering 350 million years, Proceedings of the National Acad-
experiments in microgravity. Exper Thermal and emy of Sciences 107 (52), pp. 22448–22453, 2010.
Fluid Science 28 (7):743–751. doi:10.1016/j. doi:10.1073/pnas.1011974107.
expthermflusci.2003.12.012 44. R Hadden, G Rein, C Belcher, Study of the competing
30. A. Bar Ilan, Rein G, Walther DC, Fernandez chemical reactions in the initiation and spread of
Pello AC, Torero JL, Urban DL (2004b) The effect smouldering combustion in peat, Proceedings of the
of buoyancy on opposed smoldering. Combust Combustion Institute 34, pp. 2547–2553, 2013.
Sci Technol 176(12):2027–2055. doi:10.1080/ doi:10.1016/j.proci.2012.05.060.
00102200490514822 45. M. Tuomisaari, D. Baroudi, and R. Latva,
31. X Huang, G Rein, “Smouldering Combustion of “Extinguishing Smoldering Fires in Silos,” Publica-
Peat in Wildfires: Inverse Modelling of the tion 339, VTT Technical Research Centre of Finland,
Drying and the Thermal and Oxidative Decomposi- Espoo, Finland (1998).
tion Kinetics,” Combustion and Flame 161 (6), pp. 46. G Rein, S Cohen, A Simeoni, Carbon Emissions from
1633–1644, 2014. doi:10.1016/j.combustflame.2013. Smouldering Peat in Shallow and Strong Fronts,
12.013. Proceedings of the Combustion Institute 32, pp.
32. TJ. Ohlemiller, “Smoldering Combustion Propagation 2489–2496, 2009.
Through a Permeable Horizontal Fuel Layer,” Com- 47. A. Stec, T.R Hull, Assessment of the fire toxicity of
bustion and Flame, 81, p. 341 (1990a). building insulation materials, Energy and Buildings
33. J. Brenden and E. Schaffer, “Wavefront Velocity in 43 (2011) 498–506.
Smoldering Fiberboard,” Research Paper FPL 367, 48. J. Quintiere, M. Birky, F. McDonald, and G. Smith,
U.S. Forest Products Laboratory (1980). An Analysis of Smoldering Fires in a Closed Com-
34. T. Kinbara, H. Endo, and S. Sega, Proceedings of the partment and Their Hazard due to Carbon Monoxide,
Combustion Institute, p. 525 (1967). Fire and Materials, 6, p. 99, 1982.
35. A. Egerton, K. Gugan, and F. Weinberg, “The Mech- 49. G. Mulholland and T. Ohlemiller, “Aerosol Charac-
anism of Smoldering in Cigarettes,” Combustion and terization of a Smoldering Source,” Aerosol Science
Flame, 7, p. 63 (1963). and Technology, 1, p. 59 (1982).
36. D. Donaldson and D. Yeadon, “Smoldering Phenom- 50. H. Hotta, Y. Oka, and O. Sugawa, “Interaction
ena Associated with Cotton,” Textile Research Jour- Between Hot Layer and Updraft from a Smoldering
nal, March, p. 160, (1983). Source. Part 1. An Experimental Approach,” Fire
37. W. Stiefel, R. Bukowski, J. Hall, and F. Clarke, “Fire Science and Technology, 7, p. 17 (1987).
Risk Assessment Method: Case Study 1, Upholstered 51. I Bertschi, Yokelson, R.J., Ward, D.E., et al. (2003)
Furniture in Residences,” NISTIR 90-4243, National Trace gas and particle emissions from fires in large
Institute of Standards and Technology, Gaithersburg, diameter and belowground biomass fuels. Geophys-
MD (1990). ical Research 108 (D13): 8472; doi: 10.1029/
38. Bjarne C. Hagen, Vidar Frette, Gisle Kleppe, Bjørn 2002JD002100.
J. Arntzen, Onset of smoldering in cotton: Effects of 52. G. M. Davies, A Gray, G Rein, CJ Legg, Peat con-
density, Fire Safety Journal, Volume 46, Issue sumption and carbon loss due to smouldering wildfire
3, 2011, Pages 73–80, http://dx.doi.org/10.1016/j. in a temperate peatland, Forest Ecology and Manage-
firesaf.2010.09.001. ment 308, pp. 169–177, 2013. doi:10.1016/j.foreco.
39. H Fang, Frank Behrendt, Experimental investigation 2013.07.051
of natural smoldering of char granules in a packed 53. E.R.C. Rabelo, C.A.G. Veras, J.A. Carvalho,
bed, Fire Safety Journal, Volume 46, Issue 7, 2011, E.C. Alvarado, D.V. Sandberg, J.C. Santos, Log
www.ebook777.com
smoldering after an Amazonian deforestation fire, 64. M.A. Nolter, D.H. Vice, Looking back at the Centralia
Atmospheric Environment 38 (2004) 203–211. coal fire: a synopsis of its present status, International
54. M. Turetsky, B. Benscoter, S. Page, G. Rein, Journal of Coal Geology 59 (2004) 99–106.
G.R. van der Werf, A. Watts, Global vulnerability of 65. V. Babrauskas and J. Krasny, “Upholstered Furniture
peatlands to fire and carbon loss, (invited progress Transition from Smoldering to Flaming,” Journal of
paper), Nature Geoscience 8 (1), pp. 11–14, 2015. Forensic Sciences, Nov., pp. 1029–1031 (1997).
doi:10.1038/NGEO2325. 66. O. Putzeys, A. Bar-Ilan, G. Rein, A.C. Fernandez-
55. S.E. Page, Siegert, F., Rieley, J.O., Boehm, H.D.V., Pello, D.L. Urban, 2007, The role of the Secondary
Jaya, A. & Limin, S. (2002) The amount of carbon Char Oxidation in Smoldering and its Transition to
released from peat and forest fires in Indonesia during Flaming by Ultrasound Probing, Proceedings of the
1997. Nature 420: 61–65. Combustion Institute 31 (2007) 2669–2676
56. W Field, Shen, Nature Geoscience 2, 185–188 (2009) doi:10.1016/j.proci.2006.08.006.
Human amplification of drought-induced biomass 67. S.D. Tse, A.C. Fernandez-Pello, K. Miyasaka,
burning in Indonesia since 1960, doi:10.1038/ Controlling Mechanisms in the Transition from Smol-
ngeo443 dering to Flaming of Flexible Polyurethane foam,
57. X. Huang, G. Rein, Computational Study of Critical Proceedings of the Combustion Institute 26 (1996)
Moisture and Depth of Burn in Peat Fires, Interna- 1505–1513.
tional Journal of Wildland Fire 24 (in press), (2015). 68. TJ. Ohlemiller, 1991. Smoldering Combustion Propa-
doi:10.1071/WF14178. gation On Solid Wood. Fire Safety Science 3:
58. W Frandsen, Ignition probability of organic soils, 565–574. doi:10.3801/IAFSS.FSS.3-565
Can. J. For. Res. 27(9): 1471–1477 (1997). 69. TJ. Ohlemiller, “Forced Smolder Propagation and the
59. X. Huang, G. Rein, H. Chen, Computational Smolder- Transition to Flaming in Cellulosic Insulation,” Com-
ing Combustion: Predicting the Roles of Moisture and bustion and Flame, 81, p. 354 (1990b).
Inert Contents in Peat Wildfires, Proceedings of the 70. A. Bar-Ilan, Putzeys OM, Rein G, et al., 2005,
Combustion Institute 35, pp. 2673–2681, (2015). Transition from forward smoldering to flaming in
doi:10.1016/j.proci.2014.05.048. small polyurethane foam samples, Proceedings of
60. RA Hartford, Frandsen W.H. (1992) When it’s hot, the Combustion Institute, 30 (2) pp. 2295–2302,
it’s hot. . . or maybe it’s not! (surface flaming may not 2005. doi:10.1016/j.proci.2004.08.233.
portend extensive soil heating). International Journal
of Wildland Fire 2: 139–44. doi: 10.1071/
WF9920139. Guillermois Rein is Reader (Associate Professor) in
61. G. Rein, N. Cleaver, C. Ashton, P. Pironi, JL. Torero, Mechanical Engineering at Imperial College London,
The Severity of Smouldering Peat Fires and Damage and Editor-in-Chief of Fire Technology. His professional
to the Forest Soil, Catena 74, 304–309, 2008 activities are centered on research in fire and combustion,
62. Stracher, Prakash and Sokol, Coal and Peat Fires: A and teaching of thermofuild sciences to engineers. He has
Global Perspective, Elsevier Geoscience, 2010, ISBN studied a wide range of fire dynamics topics in the built
9780444528582 and the natural environments, including pyrolysis, fire
63. G.B. Stracher, T.P. Taylor, Coal fires burning out of modeling, wildfires, structures and fire, and forecasting
control around the world: thermodynamic recipe for techniques. Over the course of the last 15 years he has also
environmental catastrophe, International Journal of specialized in smoldering combustion, conducting both
Coal Geology 59 (2004) 7–17. computational and experimental studies on a variety of
fuels like polyurethane foam, cellulose, peat and coal.
Spontaneous Combustion
and Self-Heating 20
Brian F. Gray
Introduction boundary between the two types of behavior is

referred to as the critical condition, and it plays
The term spontaneous combustion will be used an absolutely central role in the area, both
here to refer to the general phenomenon of an conceptually and pedagogically. It can crudely
unstable (usually oxidizable) material reacting but pictorially be thought of as a watershed.
and evolving heat, which to a considerable extent The critical condition is actually a whole set
is retained inside the material itself by virtue of of combinations of parameters that affect the
poor thermal conductivity of either the material behavior. The most important of these are the
or its container. Under some circumstances this ambient (surrounding) temperature, and the size
process can lead to flaming combustion and overt and shape of the body of material involved. Thus
fire, in which case it is properly called spontane- for a given body of a particular material we
ous ignition, which here is regarded as a special would normally talk about the critical ambient
case of spontaneous combustion. This has been temperature (CAT). If we were dealing with a
responsible for significant losses of life and enor- situation where the size of the body were always
mous losses of property. Fire loss statistics from fixed by commercial practice, for instance, this
many sources show that spontaneous ignition is would be the normal statement of the critical
quoted as the cause in a much greater proportion condition. However, in the case of storage of a
of cases with multimillion-dollar losses than in variable amount of material in a constant temper-
smaller fires. Of course, one should also note that ature environment, then one would talk about
the proportion of “cause unknown” results the critical size or the critical diameter of the
follows a similar trend, probably due to the body for a given fixed temperature. The CAT is
greater degree of destruction, and hence evidence the most commonly used and stated critical
loss, in larger fires. condition.
In other circumstances, clearly delineated For both fire prevention and fire cause inves-
from the former, only relatively mild self-heating tigation, it is essential to be able to identify the
occurs. This may be referred to as self-heating, critical condition if spontaneous ignition is a
spontaneous combustion, or by research possibility either before or after the event. It is
scientists as subcritical behavior. By the same also important to be aware of other possible
token, spontaneous ignition would be referred to factors operating in particular cases, such as
as supercritical behavior. The well-defined solar irradiation in outdoor storage and
preheating if recently manufactured or processed
goods are involved. In such cases as hot laundry;
B.F. Gray (*)
School of Mathematics and Statistics, University of hot new chipboard; hot, oily, porous food
Sydney, Sydney 2006, Australia products (instant noodles, fried fish scraps);

www.ebook777.com
20 Spontaneous Combustion and Self-Heating 605
bagasse1; and the like, the temperature of the The purpose of this article is to expound the
material itself is a most important parameter detailed nature of the situations described above
affecting criticality in addition to the usual in a manner that approaches the principles
ones. In such cases we have to deal with and involved in a way that minimizes mathematical
determine a critical stacking temperature formulation as far as is reasonable. The subject
(CST), which refers to the temperature of the will be approached from the point of view of its
material itself not the ambient temperature. relevance to fire cause and fire investigation and
The CST is dependent on the CAT and the size as such will refer mainly to solid systems. Many
of the body so such cases are a degree more of the basic principles used were actually
complicated than the traditional ones involving clarified by experimental work on gaseous
usually agricultural materials stacked at ambient systems; such systems still play a central role in
temperature. In addition, in such cases with current research on this topic, particularly ones
preheated materials the time to ignition (defined where the chemical kinetics are simple and well
precisely later) is usually very much shorter than understood in their own right.
it is where the material is stacked at ambient A closely related aspect to be discussed here is
temperature. the subject of runaway reaction, or thermal
Because the basic processes competing with runaway. In the past two decades this topic
each other in spontaneous combustion are heat has developed a literature of its own [1] and
generation by chemical reaction and heat loss to threatened to lose contact with the extensive
the surroundings mainly by conduction, it is easy literature on spontaneous combustion. These
to see qualitatively why both a larger body and a two terms, which can be taken as synonymous,
higher ambient temperature will favor ignition are applied to supercritical conditions as defined
rather than subcritical behavior as they both above but only in the context of a chemical
decrease the rate of heat loss. Generally the tem- reactor. The reactor may be of batch, semibatch,
perature profile across the body itself is roughly or continuous flow type, but it will almost invari-
parabolic in shape with a peak at the center. Most ably be well stirred either mechanically or by
chemical reaction rates increase almost exponen- deliberate turbulent mixing. Therein lies the
tially with temperature, whereas heat loss pro- attraction from a pedagogical point of view of
cesses such as conduction increase only linearly. such studies because the main difficulties in
Thus the center of the body where the tempera- mathematical modeling of solid spontaneous
ture is highest is the region where ignition, or combustion arising from spatial temperature var-
thermal runaway, will commence if it is going to iation and gradually decreasing concentration of
take place at all. Many bodies that have reacting material are not present. Thus a mathe-
undergone spontaneous ignition show this tell- matical theory describing such processes exactly
tale signature of charring or complete destruction serves as a first approximation, and a tractable
to ash in the center while retaining an almost one at that, to the more complex topic of solid
pristine appearance on the outside, sometimes spontaneous combustion. In addition, the diffi-
presenting rather dangerous situations for fire cult and messy “corrections” to the simplest
fighters in large-scale examples such as bagasse, possible theories due to Semenov [2] and
woodchip, or peat piles. Similarly, the deep- Frank-Kamenetskii [3] are often impossible to
seated nature of the burning started by spontane- apply in practical situations due to the dearth of
ous ignition can be difficult to extinguish data and/or their numerical uncertainties.
completely, often reigniting days after apparent In addition, in the rare event that precise input
extinction. data are available and detailed chemical kinetics
are known, it is now entirely feasible for particu-
lar cases to invoke numerical integration of the
1
Bagasse is the residue from sugar cane after extraction, relevant equations directly without use of the
usually containing 50 % water. empirical and semiempirical curve fits involved
606 B.F. Gray
in the classical corrections to the simplest We will then move on to discuss experimental
theories. At the time of writing, average laptop testing methods, both on a laboratory and a larger
computers are quite capable of such calculations scale where possible. A large array of calorimet-
for all but the most irregularly shaped bodies ric methods can be used to obtain relevant infor-
where finite element methods need to be invoked mation, but not all of them, particularly
and custom written. differential scanning calorimetry (DSC) and dif-
Accordingly we will spend some time here ferential thermal analysis (DTA), can give other
expounding the simplest possible theory than very general information and therefore can
(Semenov), which contains all the essential often be misleading. Nevertheless, such methods
concepts for the understanding of criticality, the have their purpose when material of unknown
tangency between heat release and heat loss origin and composition is involved. Sometimes
curves, and the existence (or otherwise) of stable one needs to know whether the unknown is capa-
and unstable steady states. We then move on ble of exothermic reaction at all as postulation of
briefly to the application of such ideas to more spontaneous ignition because a fire cause looks
complex chemistry and the idea of thermal run- rather silly in its absence (this happens!). How-
away in continuous stirred tank reactors (CSTR). ever, activation energies, in particular those
We then discuss the Frank-Kamenetskii ver- obtained from DSC tests, should be treated with
sion of thermal explosion theory, which great suspicion.
considers temperature gradients within the self- A characteristic of fires where spontaneous
heating body (thereby generalizing Semenov) ignition is suspected as the cause is that they
and often gives better agreement with experi- often occur on premises that have been closed
ment for solid bodies with low thermal conduc- up or unoccupied for a significant period of time.
tivity. For this reason it is much used in fire A question of very great interest in such a context
investigations, particularly when it is necessary is, What is the time scale expected for a body of a
to predict the CAT for a large-scale industrial given size in a given ambient temperature to
body from small-scale laboratory tests. However, reach ignition, that is, the appearance of overt
this type of extrapolation requires great care in its flame? As one would expect, by application of
application to all but the simplest chemistry. Murphy’s law, this question is very difficult to
We then present some ways in which answer with confidence except in the simplest of
corrections can be made to the predictions of cases. The time to ignition is a parameter that is
the Frank-Kamenetskii theory occurring under not only extremely sensitive to many factors that
conditions where some of its assumptions are are often unknown but is also extremely sensitive
not sufficiently accurate. This occurs when the to the degree of supercriticality, that is, how far
heat of reaction is relatively small and/or when the body is from the watershed. Not only does it
the resistance to heat flow in the boundary of the depend on how far the body is from the water-
body (or container wall) is relatively large com- shed, but it depends sensitively on the direction
pared to that inside the body itself (case of small as well. In other words, the term degree of
Biot number). Corrections are also necessary supercriticality needs to be refined before any
when more than one chemical reaction generates idea of time to ignition can be properly
heat and when oxygen diffusion into the interior formulated.
of the body is rate limiting. A number of investigations of this problem
All of these factors are difficult to handle have been carried out, and it is essential to rec-
quantitatively, but fortunately none of them ognize that most of the earlier ones addressed the
really alter the qualitative conceptual nature of question of time to ignition for the initial temper-
what is going on. It is important in gaining an ature of the body equal to the ambient tempera-
understanding of spontaneous combustion not to ture—such as would be the case in the building
be confused by these corrections, although in of a haystack. Hot stacked material requires
certain cases they can be quite large. totally different considerations for the evaluation
www.ebook777.com
of times to ignition, and classical formulas can- materials generally. Many advances have been
not be used in such situations. Such bodies can made in the field of gaseous autoignition over the
ignite in times that may be an order of magnitude last decade or so, stemming from accurate and
shorter than predicted by uncritically using clas- detailed kinetic measurements and considerable
sical formulas. advances in computing power. The critical con-
In the penultimate section of this article, we dition for gaseous systems is a very complex
move on to discuss the actual fire scene where locus in the parameter space characterized by
spontaneous ignition has been the cause, or ambient temperature (as for solids), pressure,
suspected cause, of the fire. We discuss factors and composition. Many organic materials, such
that would be either positive or negative as hydrocarbons, exhibit more than one
indicators of spontaneous ignition, and also the autoignition temperature, and many also exhibit
appropriate examination of the aftermath of the the phenomenon of igniting on decreasing ambi-
fire for pointers as to whether or not spontaneous ent temperature. Many older tabulations of
ignition was the cause. We then proceed to illus- autoignition temperatures do not recognize
trate all of the above with a number of case these peculiarities and should be used with
histories, some of them common and illustrative great caution. A detailed description of the
of the basic principles expounded here, others of reasons for such complexities and their impor-
a novel nature involving quite subtle and detailed tance in a hazard context is given by Griffiths and
investigations that nevertheless can give very Gray [6] in the twenty-fourth Loss Prevention
definite results. Symposium of the American Institute of Chemi-
cal Engineers (1990). A comprehensive list of
references up to 1990 can be found in this article.
The Literature Reference to liquid reactions and related
spontaneous ignitions and thermal instabilities
There is a large and varied literature on the topic will be given later in this article in the section
of spontaneous combustion ranging from sophis- on spatially homogeneous or “well-stirred”
ticated mathematical theory to technical systems. Otherwise, references will be given at
measurements on industrial and agricultural points throughout this text resulting in a
products. It is scattered over a very wide range reasonably complete bibliography.
of journals, magazines, and disciplines. The most
comprehensive publication is probably the book
written by Bowes [4], Self-Heating: Evaluating Concept of Criticality
and Controlling the Hazards. This book was
published in 1984 and contains references to Over the last two decades the concept of critical-
work published up to 1981, so at the present ity, which has been present in the thermal context
time it is in need of updating. However, it is the for many years [7], has been recognized as a
most useful reference available for those work- branch of bifurcation theory [8], an area of non-
ing, or commencing work, in the field from either linear applied mathematics that has grown rap-
an academic or a technical viewpoint. The Igni- idly and proven to be extremely powerful in
tion Handbook by Babrauskas [5], published in solving nonlinear problems. In our case the non-
2003, contains a very useful chapter on self- linearity comes from the temperature depen-
heating and has become an indispensable refer- dence of the chemical reaction (and therefore
ence for anyone working in the area of ignition. heat production) rate. The Arrhenius form for
Although much of the understanding of spon- this for a single reaction is ZeE/RT, where E is
taneous combustion has come from the basic the activation energy and R is the universal gas
study of gas-phase reactions, where it is gener- constant. T is the absolute temperature, of course.
ally referred to as autoignition, this article will be At temperatures rather less than E/R (which can
limited to spontaneous combustion of solid typically be 10,000 K or more), the Arrhenius
608 B.F. Gray
Fig. 20.1 Typical

Arrheniusheat release and
loss curves
Heat release rate/heat loss rate

Ta,1 Ta,critical Ta,2
Temperature
function is very convex; that is, it curves upward In the region of the lower intersection at Ta,1 it
rather rapidly with temperature. In contrast, the can be seen from the diagram that the tempera-
rate of heat loss from a reacting body is generally ture of the body will increase up to the balance
only a linear function of temperature, for exam- point from below as heat release is greater than
ple, conduction. Although radiation losses are heat loss in this region. On the other hand, just
nonlinear functions of temperature, they are above this balance point the temperature of the
much more weakly nonlinear than the Arrhenius body will move down to it because the heat
function and also generally rather small at the release is lower than the heat loss in this region.
low temperatures involved in solid spontaneous Thus the lower balance point occurring at ambi-
combustion although they are important in flame ent temperature Ta,1 is recognized as a stable
extinction. Typical heat generation per unit vol- balance, or stationary, point. Small perturbations
ume and heat loss (proportional to surface vol- from it will be nullified, and the body in this
ume ratio as plotted) loci are shown in Fig. 20.1. region will tend to stay at the balance point.
The low temperature range of the Arrhenius Note that the temperature of the balance point is
curve is seen here to be rather convex and rapidly not Ta,1 but slightly above it, usually by 5–20 C.
increasing with temperature. The three straight It represents subcritical self-heating and can
lines represent the rate of heat loss from a cause loss of the material but not by overt igni-
body of fixed given size at various ambient tion or fire. It can appear as degradation or dis-
temperatures Ta,1, Ta,critical, and Ta,2. coloration of many materials, making them
At Ta,1 it can be seen that the heat production useless for their required purpose. For example,
and loss curves intersect at two points. At Ta,2 woodchips degraded in this way are not suitable
they do not intersect at all, and at Ta,critical they for paper or cardboard production, and dried milk
intersect at only one point and, in fact, touch powder when discolored is unacceptable.
tangentially. The second balance point at the ambient tem-
Because intersections represent conditions perature Ta,1 can be seen by a similar simple
where heat production and loss balance exactly, analysis to be unstable in the sense that, in the
we expect them to represent some sort of “equi- temperature region just below it, the heat produc-
librium” or stationary point where the tempera- tion is lower than the heat loss, so the tempera-
ture of the body remains constant in time. It is ture tends to drop. In the temperature region just
important to remember that they do not represent above it, the converse is true, so the body tem-
equilibrium in any thermodynamic sense. perature tends to rise and leave the balance point.
www.ebook777.com
Fig. 20.2 Disappearance

of balance points with body rsub
size increase
Heat release rate/heat loss rate

rcritical
rsuper
Ta
Temperature
The latter acts as a watershed between two totally now always above the loss line and therefore
distinct types of behavior, that is, the temperature the temperature can only increase. Subsequent
of the body dropping to the lower balance point ignition will then occur. It will occur after some
or running away to the right of the diagram and delay because the rate of temperature increase in
much higher temperatures, representing ignition. this simple model is proportional to the imbal-
Here the temperature at the higher balance ance between heat production and loss (i.e., the
point would actually be the critical stacking tem- vertical distance between the two curves). This is
perature, or CST, for this particular body when initially quite small, increasing as the tempera-
stored at ambient temperature Ta,1. We can ture rises. In this observation lie the seeds of the
immediately see that if the ambient temperature calculation of the ignition delay or time to igni-
is increased, that is, the straight line is moved to tion (TTI) to be examined later.
the right with fixed slope (which is determined by Even more insights can be obtained from this
the size and shape of the body as we shall see simple type of reasoning. As we shall see later,
later), the CST will decrease, a physically rea- the slope of the heat loss line is dependent on the
sonable and intuitive result. surface area/volume ratio of the body in ques-
Thus this oversimplified but extremely useful tion. Thus for a body of given shape the surface/
model gives a simple understanding of what volume ratio increases as the body gets smaller
Bowes refers to as thermal ignition of the second and decreases as the body gets larger. In Fig. 20.2
kind, that is, what is probably better referred to as we can see the effect of increasing the size of a
the hot stacking problem, a much more descrip- body at a fixed ambient temperature. For this
tive term. Not only that, but it also gives us a fixed ambient temperature we can speak of sub-
qualitatively correct picture of the more common critical, critical, and supercritical sizes for the
or “normal” type of thermal ignition when the body, depending on whether any balance points
body self-heats from ambient to ignition without exist.
any preheating. At Ta,1 if we very slowly increase Thus for a body with characteristic dimension
the ambient temperature after the steady state has rsub we see the existence of both a CST and a
been reached, we can see that the now “quasi- balance point. For a larger body with dimension
steady state” will also slowly increase until at Ta, rsuper we see that neither exists and we expect
critical the quasi-steady state and the CST merge at temperature to rise to ignition. The critical con-
the point of tangency. Beyond this ambient tem- dition, in this case expressed as a radius or body
perature there is no balance point, and in this dimension, is given again by the tangency condi-
temperature region the heat release curve is tion. This critical condition, of course, is
610 B.F. Gray

CST and stable subcritical
temperature with ambient CST (unstable)
temperature and fixed
body size
Body temperature
Balance point (stable)
0(K) Ta,critical
Bifurcation parameter (ambient temperature)
identical with that obtained by thinking of the link to the mathematical developments of bifur-
quasi-static variation of the ambient temperature cation theory.
as well. The critical radius for a given ambient Figure 20.3 shows what happens to the balance
temperature will be identical with the CAT for a point temperature and the CST when Ta is varied
body of that same radius. How we describe it is continuously from below its critical value to above
simply a matter of where we are coming from. it. This takes place at constant body size. In this
Of course, we do not usually continuously case the ambient temperature is known as the
vary the size of a body but we do often stack bifurcation parameter. We should note that, even
bodies together, for example, bales of cotton, at very low ambient temperatures, the CST tends to
bales of hay, and so on, and allow larger than a finite limit. In fact it becomes very insensitive
normal quantities to accumulate, for example, to the ambient temperature, and no matter how cold
coal stockpiles. Even from the point of view of the ambient temperature, there is no corresponding
this very rudimentary theory, it is obvious that rise in the CST. Storing hot products in a cold
the CAT of two bales in contact will be consid- warehouse does not help the problem much!
erably less than that for a single bale. Thus tests Conversely Fig. 20.4 shows how the CST
of the CATs of single bodies that are going to be rises indefinitely as the size of the body decreases
stacked in groups for either transport or storage at fixed ambient temperature. Regardless of
are useless unless a theory is available enabling ambient temperature it does pay to keep hot
calculation of the dependence of CAT on body stored bodies small! Figure 20.4 also shows
size. The theory allowing this is thus extremely how, for sizes above the critical radius, there is
useful in relating practical tests on small bodies no alternative but ignition. Of course the critical
to be applied to storage of large numbers of them radius depends on the ambient temperature, and
(with certain caveats to be discussed later). as the latter goes down the critical radius goes
To conclude this section it remains to show a up. It is sometimes very useful to draw a critical
convenient method of representing the behavior radius versus critical ambient temperature graph,
of the stable balance point and the unstable CST and we will see how to do this later.
as a control parameter is varied (i.e., the ambient The whole discussion above assumes that we
temperature or size of the body). This method are dealing with a given material so that the
enables a quick and convenient representation of thermal and chemical properties do not vary.
the discussion given above on a single diagram The effects of varying thermal conductivity,
(a bifurcation diagram) and also gives us a useful heat transfer coefficients, and density on the
www.ebook777.com

CST and stable subcritical
temperature with body
dimension and fixed
ambient temperature CST (unstable)
Body temperature
Stable body
temperature
Ta
Critical dimension
Body dimension
critical condition are also important but only Assumptions of the Semenov Theory
when comparing different materials. The depen-
dence of the critical condition on these properties The three assumptions of the Semenov theory are
will be enunciated in a later section. as follows:
One final point needs to be mentioned here. 1. The temperature within the reacting body
The Arrhenius function does actually level out to is spatially uniform: A spacially uniform
an asymptote at very high temperatures, which temperature implies that either the material
are off the scale in Figs. 20.1 and 20.2. Thus of the body is well stirred (i.e., it would have
theoretically there is another balance point at to be liquid or gas) or the resistance to heat
very high temperatures, but in fact this point is flow within the body is so low compared to
not physically significant as it usually occurs at that within the container or boundary that it
many thousands of degrees, well beyond the can all be assumed to be concentrated within
region where the assumptions of the model are the boundary. The latter results in a tempera-
valid. It also gives rise to a high temperature ture discontinuity at the boundary of the
branch of the curves in Figs. 20.3 and 20.4, material and is a good approximation in
which is disjoint from the curves shown. Again deliberately stirred fluids [9].
it can be ignored from the point of view of It is not a good approximation for materials
low-temperature spontaneous ignition. of vegetable origin where thermal con-
ductivities of materials such as cellulose are
low and of the order of 0.05 W/mK. Never-
The Semenov (Well-Stirred) Theory theless, even for such materials semiquantita-
of Thermal Ignition tive conclusions can be drawn from this
theory if the spatially averaged temperature
The Semenov theory represents the simplest of the body is used.
mathematical formulation of the ideas presented 2. The heat generation is assumed to be due to
above in qualitative form. As such it is a valuable a single chemical reaction: This assumption
introduction to quantitative aspects of spontane- is often a reasonably good approximation,
ous ignition without introducing the technical particularly when a “lumped” or empirically
difficulties associated with more elaborate forms determined rate law has been measured inde-
of theory where spatial variations of temperature pendently. It does not mean that the chemical
and reaction rate within the body are considered. reaction taking place is only a single-step
612 B.F. Gray
reaction. In fact this empirical approximation c decreases as the fuel is used up. The commonest
works quite well in many cases that are not and simplest form for f(c) is Zc where Z is known
single-step reactions. as the pre-exponential factor, a constant. This
3. Both the heat of reaction and activation case is known as a first-order reaction. These
energy are assumed to be sufficiently large two terms are shown graphically in Fig. 20.1 for
to support ignition behavior: The reasons any particular value of c.
for these assumptions will become clearer Despite their apparent simplicity these two
later, but it is intuitively obvious that if there equations are not soluble by classical methods,
is zero heat of reaction, ignition cannot occur. so we cannot write down their solution. Neverthe-
Likewise with zero activation energy (accel- less, we can in fact write down the critical condi-
eration of reaction rate with temperature tion exactly (and other important quantities)
increase), ignition cannot occur either. using bifurcation theory. We will illustrate this
With these assumptions we can write down for the simplest possible case only, remembering
two equations that determine the temperature and that it can also be done for more realistic and
fuel concentration as functions of time (but complicated cases as well within the confines of
uniform in space). These are simply the conser- the Semenov theory.
vation of energy and the kinetic rate law, respec- First we write Equations 20.1 and 20.2 in
tively. They are dimensionless form (see nomenclature for
details),
dT E
Cv ρV ¼ VQ f ðcÞe RT SχðT T a Þ ð20:1Þ
dt du 1
¼ νe u ‘ðu ua Þ ð20:3Þ
dc E
dτ
¼ f ðcÞe RT ð20:2Þ
dt
dν 1
¼ ενe u ð20:4Þ
where dτ
Cv ¼ Heat capacity at constant volume
ρ ¼ Density where
V ¼ Volume u ¼ Dimensionless temperature
T ¼ Temperature of the reacting material (in K) ν ¼ Fuel concentration
Ta ¼ Ambient temperature of the surroundings ε ¼ A dimensionless version of the ratio Cν/Q
(assumed constant in time) (i.e., a measure of the amount of fuel decom-
Q ¼ Heat of reaction per unit concentration of position required to produce a temperature
fuel rise of 1 C)
f(c) ¼ Kinetic rate law τ ¼ A dimensionless time
c ¼ Concentration of fuel ‘ ¼ A dimensionless heat transfer coefficient
E ¼ Activation energy of the reaction The most frequently used version of this the-
R ¼ Universal gas constant ory, without fuel consumption, corresponds to
S ¼ Surface area of the interface across which taking the limit ε ! 0, thus maintaining ν at its
heat is lost to the surroundings initial value ν0. We have only a single equation
χ ¼ Heat transfer coefficient to deal with now, that is,
The independent variable is time.
du 1
The first term on the right-hand side of ¼ ν o e u ‘ ð u ua Þ ð20:5Þ
dτ
Equation 20.1 represents the rate of heat genera-
tion by the self-heating reaction. The second term Even this much-simplified equation is not analyt-
represents the heat lost to the surroundings. The ically soluble. However, it relates exactly to
left-hand side represents the difference between Fig. 20.1 and can be used to calculate the critical
these two. Equation 20.2 simply expresses the fact condition readily. We first note that the balance
that as the reaction proceeds, the concentration points in Fig. 20.1 must satisfy the equation
www.ebook777.com
1
ν0 e us ‘ðus ua Þ ¼ 0 ð20:6Þ as the only physically variable parameter, and ν0
is proportional to the bulk density of the material,
For subcritical values of the ambient tempera- we can take this equation to give us the critical
ture, this equation will have three solutions for size body for a given ambient temperature and
a given set of parameter values, ν0, ua, and ‘. bulk density. ua,critical is the only parameter here
From Fig. 20.1 it can be seen that at the critical that cannot easily be made the argument of the
condition (Ta,critical corresponding to ua,critical) equation.
not only do the two terms of Equation 20.6 bal- Converting Equation 20.9 into dimensional
ance, but their slopes also balance at this condi- form quickly gives us the relationship:
tion. Mathematically this means that their
differential coefficients with respect to tempera- T s, critical T a, critical ΔT critical ffi RT 2a, critical =E
ture must also be equal, that is, ð20:11Þ
2 1 3
∂ ν0 e u at the critical condition. ΔTcritical would typically
4 5 1
¼ ν0 e us =u2s ¼ ‘ ð20:7Þ
∂u be 20–30 C for ambient temperatures around
u¼us 30–40 C. Not surprisingly, it is independent of
The critical value of us is then obtained by solv- the body shape, being dependent only on the total
ing Equations 20.6 and 20.7 simultaneously, surface area through which heat is lost. Never-
which interestingly can be done in closed form theless, even this oversimplified result can be
simply by eliminating the exponential, leaving a very useful in an emergency situation. If E is
quadratic equation: not known, it is a useful rule of thumb (especially
for agricultural cellulosic materials) that self-
u2s, critical us, critical þ ua , critical ¼0 ð20:8Þ heating of more than 30 above ambient, that is,
typically a body temperature of more than
From our definition of u ¼ RT/E and the gen- 60–70 C, represents imminent spontaneous igni-
eral knowledge that R/E ffi 0.0001 for most com- tion, whereas an internal body temperature of
bustion reactions, we can see that at normal 35–40 C represents subcritical heating unlikely
ambient temperatures for ignition we will have to run away but quite likely to lead to degradation
ua,critical ffi 0.02, or in any case ua,critical 1. of the material.
Using the standard formula for the solution of a Two further points need to be made before
quadratic equation and expanding the radical leaving this simplified model. First, in order for
occurring, we can derive Equation 20.8 to have real roots, it is necessary to
require that
us, critical ¼ ua, critical þ u2a, critical þ . . . ð20:9Þ
E 4RT a, critical ð20:12Þ
which is the lower of the two roots (the upper one
is unphysical).
Physically this means that the chemical heat gen-
If we substitute this back into either Equa-
eration rate is sufficiently accelerative to produce
tion 20.6 or 20.7 we get a relationship between
the phenomenon of criticality. If it is not
the parameters of the problem, that holds at
satisfied, there is only a single stable balance
criticality only. Thus if we use Equation 20.6,
point for all conditions and no abrupt change in
we obtain, after some rearrangement,
behavior can occur.
1 Second, if we examine Equation 20.10, the
v0 e a, critical ð a, critical Þ
u 1þu
‘critical ¼ ð20:10Þ critical condition, we should note that where
u2a, critical ð1 þ ua, critical Þ2 the concentration ν0 appears, in the case of
gases we would normally convert this to pres-
We can interpret this equation in a number of sure. Thus, in this case, Equation 20.10 gives a
ways. Since ‘critical involves the size of the body relationship between ambient temperature and
614 B.F. Gray
pressure at the critical condition. This is the attained and the physical condition of the mate-
familiar explosion limit curve extensively used rial itself after the test is finished. Typically at
in the study of gaseous explosions. point 2 the material is hardly different visually
from the initial condition, whereas at point
1 there is usually no more than a small amount
Inclusion of Fuel Consumption of ash remaining. The temperature attained at
point 1 is often of the order of hundreds of
If we do not make the assumption ε ! 0 in degrees above ambient compared with probably
Equation 20.4, the clear distinction between sub- 30 above ambient at point 2.
critical and supercritical behavior no longer It is impossible to get points between 1 and
exists. We can no longer define the critical con- 2 experimentally without wasting a great deal of
dition as the disappearance of two balance time due to the extreme sensitivity in this region,
points. Equations 20.3 and 20.4 possess only a so the convention is to define the CAT as the
single balance point, u ¼ ua and c ¼ 0 for all arithmetic mean of Ta,1 and Ta,2. With good
possible parameter values; and this refers to the equipment these will be only 3 or 4 apart at
equilibrium state when all fuel has been the most.
exhausted and nothing is happening—clearly a From the point of view of theoretical calcula-
condition of no interest. For the definition of tion of the CAT in this case, we note that the points
criticality in such a case it is helpful to examine in Fig. 20.5 can be joined by a smooth curve with a
the experimental or phenomenological defini- very steep region around an inflection point. It has
tion. The experimentalist determines the critical been shown (Gray [10]) that this definition of the
condition by performing various tests at differing CAT, when fuel consumption is significant, leads
ambient temperatures (we will outline the details to a relation between the usual parameters and
of test procedure in a later section) and by mea- this relation passes over smoothly to the one
suring the temperature-time history at the center derived from the tangency condition as ε ! 0.
of the sample. He or she will plot the maximum For ε ffi 0:05 or less, which is the case for most
temperature attained against ambient tempera- practically important materials, the corrections
ture and will find there is a very steep increase arising from fuel consumption are not usually
in slope over a narrow region of ambient temper- significant. This is especially the case in fire
ature. This is illustrated in Fig. 20.5. investigations where a posteriori numerical
The distinction between points 1 and 2 is very knowledge of parameter values is rather limited,
clear in terms of both the maximum temperature and this correction (and others) is not justified.
Fig. 20.5 Typical

experimental results
for criticality tests
(1)
Supercritical
T (center)
Ta,critical
(2)
Subcritical
Ta
www.ebook777.com
Extensive discussion of earlier work on the to prove (Gray [11]) that if the heat release rate is
fuel consumption correction is given in Bowes’s defined as the sum of the heat release rates of all
book [4]. Many empirical and semiempirical reactions taking place in the system, then the
corrections were devised based on approximated critical condition can be defined as the tangency
integration of Equations 20.3 and 20.4. These of this quantity with the heat loss line. Thus a
corrections will not be discussed here because diagram like Fig. 20.1 can be drawn and the same
the advent of powerful PC and laptop computa- constructions used, provided the total heat
tional capabilities has rendered them irrelevant. release curve for all the reactions is used.
Equations 20.3 and 20.4 can be integrated with The heat release curve in this case can have a
great speed and precision if accurate parameter complex shape, and thus more than one critical
values are available. Even so, it is necessary to condition can occur. This state of affairs is
have a definition of criticality when a computed extremely important in the ignition of most
or experimental version of Fig. 20.5 has been organic vapors, particularly hydrocarbons [6]
obtained. With the definition given in Gray [10] where some critical conditions occur on decreas-
allied with numerical integration, the problem ing the ambient temperature. Also in the ignition
can be regarded as solved for all practical of some commonly occurring solids, particularly
purposes. when wet, more than one heat-generating reac-
tion can be important, for example, in the spon-
taneous ignition of moist bagasse [12]. In this
Extension to Complex Chemistry case there are two critical conditions, one where
and CSTRs a jump from virtually no self-heating to self-
heating of 35 C occurs, and a second critical
Complex Chemistry condition where this intermediate state jumps to
full-fledged ignition. Modelling of such
Other than elementary gas-phase reactions, very situations is possible but beyond the scope of
few examples of chemical change occur via a this chapter; however, similar behavior is likely
single step as assumed above. As already to occur in other moist cellulosic materials,
remarked, the simple theory is more useful than including hay, chipboard, and so forth.
might be expected because many complex chem- At this stage it is worth pointing out that for
ical reactions behave as if they were a single step, bagasse at least, microbial “heat production”
over limited temperature ranges. This is usually is not a factor in these phenomena. Although
because a single step does dominate the heat natural bagasse contains large numbers of
production rate, for example, when two reactions microorganisms, sterilization by various
occur in parallel. If the activation energies are methods does not affect heat production or
rather different, they will each in turn dominate self-heating at all, as measured by Dixon [13]
the heat generation in two different temperature and predicted on the basis of bacterial microca-
ranges, and in each of these ranges the simple lorimetric data by Gray [14]. Similar work on
theory will hold. Of course, it will not hold in the hay is under way.
changeover region.
Another case where the simple theory can
hold unexpectedly is when a number of reactions CSTRs and Thermal Runaway
are in series and one is particularly slow. The
slow reaction will determine the overall Strangely, this topic has become uncoupled from
heat generation rate and its parameters will work on spontaneous ignition over recent years
dominate the critical condition. If none of the even though the basic principles and mathemati-
above conditions hold, it is still possible to cal methods used are similar. It is a huge problem
derive a generalization of the theory that is in the chemical process industry and receives
conceptually very closely related. It is possible much attention. For example, in 1998 the Joint
616 B.F. Gray
Research Centre of the European Commission,

Institute for Systems Informatics and Safety, pro- The Frank-Kamenetskii Theory
duced a book describing the proceedings of a of Criticality
European Union seminar held in Frankfurt in
1994 that managed to avoid almost completely In its original form, the Frank-Kamenetskii
any reference to the fundamentals of the problem theory included a more realistic model of heat
or related material. Risk analysis appears to have transfer within the reacting solid, that is, by
replaced fundamental scientific understanding in incorporating the heat conduction law of Fourier.
some aspects of this problem. This law allows a calculation of the variation of
We will confine ourselves here to writing temperature within the self-heating body itself
down the basic equations governing a single and allows comparison of measured and calcu-
exothermic chemical reaction taking place in a lated self-heating to take place. However, it
CSTR (continuously stirred tank reactor) to sacrifices the simple description of time-
exhibit their similarity to the equation describing dependent behavior given by the Semenov
a spontaneously ignitable material, that is, model because such considerations involve the
Equations 20.3 and 20.4. solution of partial differential equations. This is
The appropriate equations for this case are in now much faster than even a few years ago,
fact 3 and 4 with terms representing inflow and in terms of numerical computation, and improv-
outflow of reactants and products, that is, ing day by day. Nevertheless, such numerical
solutions do not lend themselves to simple
dT E interpretation even with the use of rapidly devel-
VρCv ¼ QV f ðcÞe RT SχðT T a Þ oping visualization techniques. Construction of
dt
appropriate meshes for finite element computa-
FCv ρ T T f ð20:13Þ
tion, necessary for practically occurring three-
dimensional shapes, is also far from trivial.
dc E
V ¼ V f ðcÞe RT þ F c f c ð20:14Þ As a result, the Frank-Kamenetskii theory is
dt
still mainly used for interpretation of testing
F is a volumetric flow rate and the subscript experiments on self-heating and subsequent eval-
f refers to feed values. These equations can be uation of parameters for individual systems. This
cast in dimensionless form also. Here we simply is a viable proposition for materials with suffi-
note that they possess steady-state (balance point) ciently large heats of reaction and activation
solutions without making any approximations energies. In such cases we shall see that the
at all (such as neglect of fuel consumption), and stationary (in time) conditions assumed in
Fig. 20.1 can be applied directly in slightly the Frank-Kamenetskii theory are indeed well
modified form. The critical condition referred to approximated for the duration of typical tests in
earlier occurs here also, but it can now be stated practical cases. In its original form this theory
in terms of the CAT or a critical feed temperature also neglects fuel consumption, as does the
or, indeed, a critical flow rate. Semenov theory, with similar consequences.
A critical size also occurs and this is particu- With these assumptions, the equation describing
larly prominent in CSTR considerations where the theory is
“scaleup” from prototype size to commercially E
viable size has resulted in exceeding the critical κ∇2 T þ Q f ðc0 Þe RT ¼ 0 ð20:15Þ
condition. Some references to this are given in
Safety and Runaway Reactions [1], and there are with the boundary condition T ¼ Ta on the wall(s)
many more in the chemical engineering literature of the body. Ta is the ambient temperature of the
and the study of self-heating in catalyst particles. surroundings. This boundary condition assumes
See Aris [15] for an excellent discussion of instantaneous transfer of heat from the surface of
this area. the body to the surrounding medium (usually air).
www.ebook777.com
When this is not approximately correct, very Table 20.1 Values of dcritical for various geometries
important consequences follow, as we shall see in Geometry Dimensions δcritical
a later section on the interaction of self-heating Infinite plane slab Width 2r 0.878
bodies with each other. In this formulation the Rectangular box Sides 2 l, 2r, 2 m; 0.873
shape of the body and its size both enter the math- r < l, m (1 + r 2/l2 + r2/m2)
ematical formulation through the boundary Cube Side 2r 2.52
condition only. Infinite cylinder Radius r 2.00
As usual, Equation 20.15 is not analytically Equicylinder Height 2r, radius r 2.76
soluble. However for a one-dimensional infinite Sphere Radius r 3.32
Infinite square Side 2r 1.700
slab of material by using an approximation to the
rod
Arrhenius function (Frank-Kamenetskii [3]), the
modified equation can be solved analytically.
The same approximation was later shown to be
analytically soluble for an infinite cylinder by The tabulation of figures for infinite slab or
Chambre [16]. infinite square rod is useful insofar as they are
With this approximation, Equation 20.15 often rather good approximations for real bodies,
takes the form provided one or more of their dimensions are
much larger than the others. Thus for the rectan-
∇2 θ þ δeϑ ¼ 0 ð20:16Þ gular box, if we take r ¼ l ¼ 1, m ¼ 10, we get
δcritical ¼ 1.75 compared to 1.700 for the infinite
with θ ¼ 0 on the boundary. θ is a dimensionless
square rod. If we now look at Equation 20.18
temperature defined by
for the particular case of a cube as an example,
Eð T T a Þ we get
θ¼ ð20:17Þ
RT 2a E
QE f ðc0 Þr 2 eRTa, critical
¼ 2:52 ð20:19Þ
that is, it is a measure of the temperature excess κRT 2a, critical
within the body at various points. The dimension-
at the critical condition. We have a number of
less parameter δ is defined by Equation 20.18:
choices as to interpretation of this equation
E depending on which parameter can be made the
QEr 2 f ðc0 ÞeRT a
δ¼ ð20:18Þ argument. If r is chosen as the argument, then the
κRT 2a
equation would be interpreted as giving a critical
size for the body at a fixed ambient temperature
where the symbols are already defined apart from
Ta. Because c0 depends on the density of the
r, which is usually one-half of the smallest dimen-
material, Equation 20.19 could be rearranged to
sion of the body, that is, the radius of a cylinder,
give a critical density for that particular size body
the radius of a sphere, or the half-width of a slab.
at ambient temperature Ta. What is not possible
Mathematical treatment of Equation 20.16,
is isolation of Ta as the argument of the equation,
whether it is exactly soluble or not, indicates that
and this is often the most easily varied parameter
a solution satisfying the boundary conditions
in a typical test oven.
exists only when δ is less than or equal to δcritical
This complex dependence of the critical
where δcritical is some number depending on the
condition on Ta is dealt with by rearranging
shape of the body only. For an infinite slab of
Equation 20.18 and taking natural logarithms as
material δcritical ¼ 0.878, and for an infinite
follows:
cylinder it has the value 2.000. For other shape
" #
bodies, the critical value has to be obtained δcritical T 2a, critical QE f ðc0 Þ
either numerically or by semiempirical methods ln ¼ ln E=RT a, critical
r2 Rκ
outlined in some detail by Bowes [4]. For conve-
nience, a few of the values are listed in Table 20.1. ð20:20Þ
618 B.F. Gray
from which it can be seen that a plot of inside the gauze baskets is relatively low, as it
2
ln[δcriticalTa,critical /r2] against 1/Ta,critical will be is with many agricultural materials containing
a straight line with slope -E/R and intercept cellulose (κ 0.05 W/mK). The efficacy of the
ln[QEf(c0)/κRδcritical]. The traditional and boundary condition is determined by the heat
recommended test protocol for spontaneous transfer rate from the gauze to the oven air rela-
ignitions makes explicit use of this logarithmic tive to the conduction rate within the material
form of the critical condition. Not only does it itself. This ratio (χr/κ) is known as the Biot
yield the activation energy from the slope, but the number, and the larger it gets, the more accurate
occurrence of a straight line plot assures us that the Frank-Kamenetskii boundary condition
the assumption of an Arrhenius temperature (T ¼ Ta) becomes. In practice a Biot number
dependence for the heat-generating reaction is greater than 30 is effectively infinite as the
correct over the temperature range investigated. CAT becomes extremely insensitive to it. We
Equation 20.20 can also be regarded as a scal- will return to this topic in a later section where
ing law, in principle enabling the prediction of the dependence of the critical condition on the
CATs for large-scale bodies from measured Biot number will be outlined.
CATs for much smaller laboratory-sized samples. The test procedure involves starting with the
However, as we shall see, it is necessary to ensure smallest basket and a trial oven temperature. The
that the same chemical kinetics applies over the sample is equipped with one or more fine
whole temperature range involved, that is, f(c0) thermocouples placed at the center of the sample
does not vary. Finally, if it becomes necessary to and, if desired, at various places along a radius if
estimate the CAT for a complex shape, not a spatial profile is wanted (this is generally not
included in Table 20.1, an excellent and compre- necessary). The sample is placed in the preheated
hensive discussion of approximation methods is oven and the center temperature followed as a
given by Boddington, Gray, and Harvey [17]. function of time. If the oven temperature is well
below the CAT, the sample will simply approach
the oven temperature asymptotically. If it is
Experimental Testing Methods slightly below, but getting close, it will cross
above the oven (ambient) temperature and attain
Experimental testing methods are traditionally a maximum of the order of 1–30 C above ambi-
based on the scaling relationship (Equa- ent before declining. This represents the subcriti-
tion 20.20). Appropriate containers (usually cal condition.
stainless steel gauze baskets) of various The sample is discarded and replaced with a
dimensions are used, being limited only by the fresh, similar one. If the previous run was
size and heating capability of an accurately subcritical, the oven temperature will be
thermostatted oven, which must also have a spa- increased by usually 20 C or less depending
tially homogeneous ambient-temperature distri- on the experience of the operator. The run is
bution ( 0.5 C is recommended). The gauze then repeated. If it is still subcritical, the proce-
containers may be any convenient shape, dure is again repeated until a supercritical oven
equicylindrical or cubic being preferred due to temperature is attained. The arithmetic mean of
ease of construction. The gauze does not restrict the lowest supercritical temperature and the
oxygen ingress through the boundary, nor does it highest subcritical temperature is taken as the
restrict egress of carbon dioxide and other prod- first estimate of the CAT. The uncertainty may
uct gases during combustion. If the air inside the be quite large at this stage so the process is
oven is sufficiently turbulent, usually the bound- usually continued by testing at the estimated
ary conditions of the Frank-Kamenetskii theory CAT. The process is repeated, halving the
will hold quite well. difference between highest subcritical and low-
The boundary condition is easier to satisfy est supercritical temperatures each time until
when the thermal conductivity of the material the desired errors are obtained. Typical
www.ebook777.com
Fig. 20.6 Two subcritical 200

and one supercritical center
temperature/time traces for 180
Ambient
a 0.175-m-radius
160 56.6°C
equicylinder of hydrated
calcium hypochlorite [17]
140
Temperature (°C)
120
100
Ambient
80 53.9°C
60
Ambient
40
49.8°C
20
0
0 10 20 30 40 50 60 70 80 90 100
Time (hr)
Fig. 20.7 Recalculated 20

Frank-Kamenetskii plot for
the data of Uehara et al. on
anhydrous calcium E/R = 12514 (827)
19 A=51.77 (2.23)
hypochlorite [19]
2ln(Ta/r)+lnδcritical
18
17
16
15
0.0025 0.00255 0.0026 0.00265 0.0027 0.00275 0.0028 0.00285 0.0029 0.00295
1/Ta, cr
temperature-time plots showing the critical containers, giving four or five CATs at various
separation are shown in Fig. 20.6. radii, then the next step is to construct the Frank-
This reaction is an exothermic decomposition Kamenetskii plot of the scaling Equation 20.20.
evolving oxygen [18]. From these measurements A typical plot is shown in Fig. 20.7.
one would conclude that the CAT was This plot shows a range of CATs for cylinders
55.2 1.34 C. For greater accuracy the next ranging in radius from 0.191 m down to 0.026 m,
test would be run at an ambient temperature of the larger radii corresponding to commercial
55.2 C. After at least four or five such sets of containers. From the slope of this line, E/R can
runs have been carried out in different size be read off directly, and, from the intercept, so
620 B.F. Gray
can the dimensionless group occurring in the agreement with measured results for
scaling equation. Sometimes components of commercial-size piles of this material (minimum
this group may be known from independent dimension 5–10 m).
measurements, for example, Q from calorimetry, The detailed nature of the wet reaction with a
κ from direct measurement, or f(c0) from kinetic rate maximum around the 50–60 C mark has led
measurements, in which case all the parameters to false identification with microbial activity. In
can be obtained. bagasse at least it has been shown [25] that
microbial activity does not contribute to self-
heating to any significant degree. Piles sterilized
Special Cases Requiring Correction by various methods showed self-heating rates
indistinguishable from those of nonsterile piles.
Presence of Water Microbial counts were carried out in all cases and
large decreases did not affect the self-heating
When water is present in spontaneously combus- rates. It would be rather surprising if similar
tible material, special considerations apply. First results were not obtained from tests on hay and
it is necessary to note that endothermic evapora- straw where microbiological activity (but not
tion would be expected to partly offset some of necessarily heating) are known to occur, and it
the heat generation by the exothermic reactions is surprising that such tests have not yet been
taking place. Although this is true, it is often the carried out.
case that at the high oven temperatures used in
testing small samples, the low activation energy
for evaporation (40 kJ/mol) leads to rapid Parallel Reactions
evaporation before the exothermic process has
got under way fully. Many spontaneous combus- If more than one exothermic reaction can take
tion reactions have activation energies around place in the material, and these reactions have
100 kJ/mol, particularly the group of reactions rather different activation energies, then each
of cellulosic materials. will dominate in its own temperature range.
As a result, the high-temperature CATs Thus the higher activation energy reaction will
reflect the properties of the dry material, in par- cut in at higher temperatures and be insignificant
ticular the thermal conductivity. Consequently, at lower temperatures when the low activation
extrapolations to temperatures well below energy reaction will dominate the heat genera-
100 C will be questionable for this reason tion. The wider the divergence in activation
alone. In the lower temperature range the heat energies, the sharper the discontinuity in slope,
transfer will be significantly affected by the pres- that is, the narrower the temperature range over
ence of water and its transport from the hotter to which both will contribute. Hydrated calcium
the cooler regions of the body by evaporation, hypochlorite shows a clear example of this, and
diffusion, and condensation. it is reflected in a sharp break in the slope of the
Many cellulosic materials are known to Frank-Kamenetskii plot where the changeover
exhibit a “wet reaction” [20, 21] in addition to occurs. Figure 20.8 shows this plot. The low
the dry exothermic reaction. This reaction temperature activation energy for this system is
involves liquid water as a reactant and further about 48 kJ/mol while that of the higher temper-
complicates the picture as far as high- ature reaction is around 125 kJ/mol, the transi-
temperature testing is concerned. Simultaneous tion temperature being around 120 C
evaporation, diffusion, condensation, and reac- [17]. Extrapolation of the high temperature line
tion involving water have been modeled recently in this case gives CATs for large commercial-
in connection with bagasse [22, 23], using an size containers that are seriously in error; that is,
experimentally measured rate law for the wet they are predicted to be much higher than they
reaction [24] giving results that are in good actually are. In the general case of two reactions
www.ebook777.com
Fig. 20.8 Frank- 22

Kamenetskii plot for
hydrated calcium
hypochlorite with reaction
In δcritical + 2 In (Ta,critical/r)
mechanism change 20
18
16
14
2.4 2.6 2.8 3.0 3.2
1000/Ta,critical
with different activation energies, this will some manufacturers have taken anything up to
always be the case as the high activation energy 5.5 % to be admissible resulting in large
reaction is “frozen out” at low temperatures, and variations in the thermal stability of this product.
the low activation energy reaction is “swamped” On the other hand the P and I (Protection and
at higher temperatures. Indemnity) clubs have taken a proactive stance
The spontaneous decomposition of calcium and largely refused to insure such cargoes unless
hypochlorite has caused extensive container they are shipped in refrigerated containers
ship losses, particularly in the late nineties. (reefers). The resultant increase in cost is proba-
Some due to hydrated calcium hypochlorite bly leading to failure to declare some cargoes of
(UN 2880), defined by the International Mari- this material as dangerous goods.
time Organization [26] to contain not less than Also strategies have yet to be worked out to
5.5 % moisture and not more than 16 % moisture minimize the possibility of thermal runaway in
result from faulty extrapolation as can be seen the event of a power failure on board ship. The
from Fig. 20.8. heat transfer coefficients of reefers are a fraction
The discontinuity in slope of the F-K plot is of those for normal containers, reducing the
undoubtedly due to the effect of moisture effective CATs of reefers without power to very
mediating chain reactions, the decomposition low figures indeed. An obvious emergency strat-
becoming extremely sensitive to trace metal egy would be to open the doors of reefers in such
concentrations. a situation, but this would put serious restrictions
This is confirmed by the work of Uehara on stowage possibilities.
et al. [19] on UN 1748 (“anhydrous”) calcium Other examples of mechanism change are
hypochlorite where the lower section of the line known and discussed by Bowes [4]. In such
is missing. Thus the samples with lower moisture cases accurate predictions of CATs can still be
content (<1 %) are much less prone to thermal made within each temperature range. This type
runaway. So are samples of UN 2880 tested in of example emphasizes the need for tests
gauze baskets which allow rapid evaporation of covering as wide a range of temperatures as
moisture during the tests themselves, behaving in possible. Recent methods put forward as viable
a very similar manner to Fig. 20.7. alternatives to the standard method, for example,
Unfortunately the IMO has been inactive in Jones [27] and Chen [28], are restricted to either
defining moisture content limits for UN 1748 and measurement at a single temperature or over a
622 B.F. Gray
limited temperature range and can give danger- of an infinite cylinder) have been treated and the
ously flawed results. Empirical tests such as modified critical condition obtained [33, 34].
the Mackey test [29] and the crossover test [30] Similarly modified boundary conditions must be
are not reliable and cannot be properly related to used when surface reactions occur due to cataly-
the basic principles of spontaneous ignition sis by surface material.
theory. The values of the critical parameter δ quoted
for the Frank-Kamenetskii theory are all for the
limiting case Bi ! 1, and both Thomas and
Finite Biot Number Barzykin have given semiempirical functions
exhibiting the dependence of δcritical on Bi,
The Biot number is defined as which are detailed in the book by Bowes. As
the Biot number decreases so does δcritical and
χr
Bi ¼ ð20:21Þ hence so does the CAT, all compared with the
κ standard Frank-Kamenetskii theory. For Biot
numbers greater than 30, the correction is rather
where
small but is significant for smaller values. Typical
χ ¼ Surface heat transfer coefficient
heat transfer coefficients from smooth solid
r ¼ Smallest physical dimension of the body
surfaces to rapidly stirred air (in a test oven, for
κ ¼ Thermal conductivity of the material
example) are of the order of 20 W/m2K, and ther-
It is the dimensionless measure of the ratio of
mal conductivities of typical cellulosic materials
the resistance to heat transfer within the body to
(such as sawdust) are around 0.05 W/mK, giving
that from the surface to the surroundings. Thus
a ratio of 400/m. Clearly for laboratory-size
the Semenov theory is often referred to as zero
test bodies (r 0.1 m), the Biot number is rather
Biot number and the Frank-Kamenetskii theory
large.
as infinite Biot number. They are both special
For this reason a significant amount of work
cases of a more general (and more exact) formu-
has simply assumed a sufficiently large Biot
lation, as was originally pointed out by Thomas
number without investigation of its actual numer-
[31, 32].
ical value. Sometimes the assumption is not
In general the boundary condition at the edge
justified, particularly where inorganic materials
of a self-heating body has the form of a continu-
are involved, as their thermal conductivities can
ity condition, which refers to the energy flux
be quite large. For example, typical inorganic
across the boundary. It states that the energy
salt thermal conductivities lie in the range
flux within the body (given by Fourier’s law)
0.2–3.0 W/m
K, giving for the ratio (χ/κ) a
and the energy flux from the body surface to the
value of 7–100/m. Clearly for test bodies with
surrounding air must be equal, that is,
r 0.1 m, the Biot number will be only 0.7–10.
∂T The effect of the small Biot number on δcritical is
κ ¼ χð T T a Þ ð20:22Þ
∂n to reduce it by a factor ranging from 0.21 to 0.83,
respectively. Clearly for such materials, the more
In dimensionless form this becomes general boundary condition suggested by
Thomas must be used, and it is good practice
∂T
¼ Biðu ua Þ ð20:23Þ for all but the most strongly insulating materials
∂n
to estimate the thermal conductivity (particularly
This boundary condition does not hold if there in the presence of water) independently of the
exist any heat sources on the boundary of the standard testing regime.
body itself, as can occur when there is incidence A further important feature of self-heating
of radiation or when there is heat generated by bodies with a finite Biot number is that their
friction as can occur in pulverization of materials CATs will be sensitive to the heat transfer coef-
capable of self-heating. Such cases (in the shape ficient from their surface to the surrounding air.
www.ebook777.com
Thus the value of the CAT obtained may well be 1. The theoretical treatment is much more diffi-
test-oven sensitive and be strongly influenced by cult than that of criticality itself.
air movement. For example, it has been shown 2. The actual definition has been greatly con-
for hydrated calcium hypochlorite [18] that in fused from case to case.
stirred air in a typical test oven the CAT is 3. The TTI, however defined, can be extremely
60 C for a 0.175-m-radius container, but in sensitive to quantities that have hardly any
still air the CAT is 55 C. effect on the position of the critical condition.
This observation raises serious questions
about the value of empirical testing methods
such as the SADT test for shipping of self- Theoretical Treatment
heating materials [35] that determines
criticality-related parameters under vaguely We refer the reader to Bowes [4] and Babrauskas
defined conditions of forced airflow in a test [5] for discussion of earlier treatments. For illus-
oven. The results are then used to determine trative purposes we will initially follow Bowes
“safe” conditions for shipping such materials in and define TTI from Equation 20.5 by integration
still air inside, for example, a shipping container. from ambient temperature to some value u1, say
Almost invariably many self-heating bodies are
uð1
stacked inside the same still air inside a con- h 1
i1
tainer, and they will interact with each other to τi ¼ ν0 e u ‘ðu ua Þ du ð20:24Þ
a very significant extent if the transfer of heat ua
through the container wall is not very rapid. In
practice such transfer is rather slow, involving This is, of course, in dimensionless form. Our
two successive air-metal transfers. As a result the present interest is the implicit use of ua as the
self-heating bodies collectively heat the air lower limit; that is, it is the time for the sample
inside the container and produce a “cooperative to go from ambient temperature to some
CAT,” which can be tens of degrees lower than predetermined arbitrary figure, possibly the max-
the CAT of a single body. The Semenov-type imum temperature attained (it turns out that the
theory for this collective ignition has been integral is not sensitive to this limit, provided it is
formulated by Gray [36]. A more accurate ver- sufficiently high).
sion, where the individual bodies are assumed to Although the maximum temperature attained
obey the general boundary condition put forward is a meaningful figure for laboratory tests under
by Thomas, has also been formulated. The some circumstances, it does not always corre-
predictions of this theory have been compared spond to practical large-scale circumstances.
to the experimental CAT for eighteen 14 kg For example, it requires recording the time
equicylinders packed in a rectangular steel box taken for the center of the sample to heat up in
with good agreement [37]. The CAT was reduced a test oven to ambient temperature and using this
from 62.5 C for a single keg in still air to 54 C as the reference time for TTI. Unfortunately,
for 18 kegs in still air. when the center has reached this point, other
parts of the body have often attained rather
higher temperatures [38], and the subsequent
Times to Ignition (Induction Periods) TTI will be reduced compared to a large-scale
body that may well have been built at ambient
The terms time to ignition and induction periods temperature and be quite uniform initially.
tend to be used synonymously. Here we will Extrapolations of such laboratory tests will not
abbreviate using TTI. This represents the most then be reliable since the initial condition will not
difficult area of spontaneous combustion insofar be appropriate.
as prediction is concerned. There are three prin- The TTI for the hot stacking problem is quali-
cipal reasons for this: tatively different from that in which the body is
624 B.F. Gray
formed uniformly at ambient temperature. Gen- crucial. In many cases this leads to extremely
erally this time is much shorter than the TTI for irreproducible TTIs without similar variation of
the more common case of initially ambient tem- CATs or other properties. In case this list of
perature throughout the body. The reasons have difficulties leads to an overly pessimistic view
been given, with a comparison of the two cases, of the topic of TTI, there are some things that can
by Gray and Merkin [39]. Similar considerations generally be relied on as far as the practical
apply when part of the body is at a high tempera- situation of fire investigation is concerned.
ture (hot spot), and this case has been discussed Very crudely speaking, notwithstanding the
in detail by Thomas [40]. above discussion, the larger the body, the longer
With the ready availability of powerful and the TTI will usually be. Thus a fire thought to
fast numerical techniques, it is now feasible to have been caused by spontaneous ignition of a
integrate routinely the time-dependent heat con- pile of linseed oil–contaminated rags contained
duction equation for this problem, which is prob- in a wastepaper basket will usually appear within
ably the best solution. Zinn and Mader [41] were a few hours of the rags being placed there. On the
early participants in this effort, and more recently other hand, a fire resulting from spontaneous
Gray, Little, and Wake [38] have noted that ignition of thousands of metric tons of
such numerical results can be usefully used to woodchips would occur only after some months
predict a very good lower bound to the TTI. of assembly, assuming the pile was assembled at
These results are desirable as they err on the ambient temperature. For such bodies it is gener-
side of safety. ally true that the TTI increases with size in this
Very close to criticality, perturbation manner. Accordingly haystacks tend to ignite
treatments have been formulated [42–46], but (if they are supercritical) after a few weeks and
these are mainly of theoretical interest. At the coal stockpiles after a few months. However, the
critical condition the TTI becomes infinite, and TTI can decrease dramatically if the body is very
close to this condition it is extremely sensitive to far beyond the CAT.
the degree of criticality, so unless this is known For hot stacked bodies, on the other hand,
accurately (hardly ever the case), use of such times are generally much shorter and not partic-
formulas is not advised. ularly sensitive to the ambient temperature. Thus
stacks of freshly manufactured chipboard with a
volume of a few cubic meters can ignite much
Other, Largely Chemically Kinetic, more quickly—that is, hours rather than days—
Difficulties than a similarly sized body self-heating from
ambient. Beyond these general comments one
In addition to the difficulties discussed above, has to treat each separate case on its merits with
which apply even when only a single simple a careful eye for exceptions to any general rules.
reaction is assumed, there are others that are For example, the presence of any catalytic mate-
largely chemically kinetic. It has long been rial, such as rusty metal (a common contaminant
known that chain reactions, whether branching of many materials), can dramatically decrease
or not, can exhibit very long induction periods the TTI. This indicates the presence of free-
followed by very rapid onset of (sometimes non- radical or chain reactions and is fairly common,
explosive) reaction. Many exothermic spontane- although the CATs and CSTs are only slightly
ous ignition reactions do possess some chain affected.
characteristics even though these do not manifest In summary, in fire cause investigation, where
themselves once the reaction is well underway. spontaneous ignition is suspected, it is wise to be
Thus it is feasible for complex chain mechanisms very circumspect about time factors without very
to determine the details of the TTI but not be at thorough investigation and detailed knowledge
all important in determining the critical condition of the initial conditions likely to have existed
where gross heat balance considerations are when the body was put in place. Even the
www.ebook777.com
traditional linseed-oil rag example can be thrown Fibrous Nature and Porosity of
out of the normal pattern by the presence of Material Fibrous or porous materials allow
mineral turpentine, a very common diluent for greater access of air than otherwise (solid wood
oil-based stains. The evaporation of this from the is not subject to spontaneous ignition at normal
rags can greatly prolong the TTI by virtue of the ambient temperatures, but woodchips and saw-
consequent cooling effect and also the exclusion dust certainly are!). The concept that packing
of air by the vapor. Depending on the such porous materials by compression will
circumstances, these factors could add 2 or increase the CAT by oxygen exclusion is badly
3 days to a TTI that would normally be no more flawed. This procedure increases the density
than a few hours. (thus lowering the CAT) and has virtually
no effect on the availability of oxygen. During
the preflame development, the oxygen require-
Investigation of Cause of Possible ment is very low; by the time overt flame is
Spontaneous Ignition Fires observed, there are usually broad channels of
destroyed material (chimneys) that will allow
From the investigative point of view, it is helpful ready access.
to list the practical factors that enhance the pos-
sibility of spontaneous ignition as a possible fire Pure cotton in a test oven with a nitrogen
cause. atmosphere has been shown to undergo sponta-
neous ignition but with a longer induction period
The Size of the Body of Material The larger than in the presence of air [47]. This could be due
the size of the body of material, the greater the to adsorbed oxygen on the cellulose fibers or due
likelihood of spontaneous ignition. By size of the to exothermic decomposition of the cellulose in
body we mean the parts that are in thermal con- the absence of air [48].
tact. A large pile of cotton bales with aisles Otherwise “harmless” materials (i.e., liquids
through it would not necessarily be a large body with very high flashpoints) can undergo sponta-
in the thermal sense used here. This classification neous ignition at temperatures more than a hun-
would be true even if (as often happens), once dred degrees below either their flashpoints or
ignited, fire could spread easily from one section their so-called autoignition temperatures. The
to the next. familiar drying oils (flashpoints around 230 C)
spread on cotton afford such an example, igniting
High Ambient Temperatures Because the air sometimes at room temperature under the appro-
around the body in question has to act as a heat sink, priate conditions. In bulk such oils pose little
the higher the ambient temperature, the more inef- threat of fire causation.
ficient is the air as a coolant. Also direct placement Similarly, hydraulic fluids, specifically
underneath a metal roof or adjacent to a northwest- designed for nonflammability and with
(southern hemisphere) or southeast- (northern extremely high flashpoints, can undergo sponta-
hemisphere) facing wall is a positive factor. neous ignition if allowed to leak onto thermal
lagging, such as mineral wool, fiberglass, and
Thermal Insulation Sometimes spontaneously so forth, which are characterized by having par-
ignitable materials are stored in chemical ticularly high surface areas. Practical cases of
warehouses or elsewhere packed against inert this and experimental tests have been reported
solids that prevent free airflow over the surface, by Britton [49], with particular reference to
thus reducing heat losses. This effect is evidenced ethylene oxide fires. More recently a modeling
by the appearance of maximum charring or project has been carried out [50, 51] based on
self-heating that is off center and closer to the adaptation of the Semenov theory of ignition to a
insulated side of the body. It also results in a porous solid that was wetted with combustible
reduced CAT. liquid.
626 B.F. Gray
Temperature of Stacking The factor of tem- ash is very characteristic in cellulose materials.
perature of stacking is simple—the hotter the Combustion starts in the well-insulated internal
worse! The main question is, How hot? The areas of the body, and warm or hot combustion
CST (critical stacking temperature) is only products rise by convection through the path of
weakly dependent on the ambient temperature at least resistance (which is not always vertically
low ambient temperatures, but it is sensitive to the upward), forming a “chimney” of discolored and
size of the hot body. This situation arises with partially combusted material. Because large bod-
freshly manufactured products such as foodstuffs ies of material are rarely uniform in density or
(milk powder, flour, instant noodles, fried batter, porosity, there can be more than one chimney
etc.), synthetic materials such as chipboard, cot- formed and this is the norm. The occurrence of
ton bales straight from the ginning process, multiple chimneys and consequent discovery of
bagasse straight from the sugar mill, fresh laundry more than one heavily charred or ashed area
(usually in commercial quantities), and so on. inside the body have led to erroneous charges
of arson on the basis of the myth that more than
To evaluate the CST requires full testing to one fire seat means the fire was deliberately lit.
obtain the parameters for the material (such as E, When a chimney reaches the edge of the body,
Q, κ, etc.) and then application of one of the smoke first becomes visible, then ingress of air
methods in the literature for its calculation. causes flame. The latter may engulf more flam-
Thomas [40] has given a method for hot spots mable materials in the building, and the whole
of material, and Gray and Scott [52] have given a structure can be destroyed while the spontane-
generalization of this, removing the approxima- ously combusting material may well be chugging
tion to the Arrhenius function made by Thomas. away slowly throughout most of its volume.
A simpler method of calculation of the CST has This can even be the case after the fire has been
been given by Gray and Wake [53]. It uses a extinguished. The result is then plenty of
spatially averaged temperature in the Arrhenius evidence as to the cause and origin of the fire.
function and then obtains exact results for this The author has measured temperatures as high
simplified problem. as 200 C in buried, spontaneously ignited
material more than 2 weeks after the extinction
Length of Time Undisturbed Material that has of the fire!
been in place for longer than usual is reason to The internal burning of large piles or stacks of
suspect spontaneous ignition as a fire cause. material can cause mechanical instability, and
Many industrial procedures involve the tempo- often the body collapses inward in the later
rary storage of materials that are normally above stages of ignition. This inward collapse can
their CAT but that are not left undisturbed for a cause some confusion in excavations, which
period longer than or equal to their TTI. Thus should always be carried out if spontaneous igni-
under normal circumstances fire does not occur tion is suspected along with photographic and
even though the TTI is regularly exceeded. If thermocouple temperature probe records at all
processes are slowed down for some reason, or stages.
storage is prolonged due to vacation, fire can It should be emphasized that the occurrence of
occur even though no other parameters have significant amounts of unconsumed, spontane-
been changed. ously ignitable material does not mean that spon-
taneous ignition was not the cause of the fire.
Frequently, oily rags are recovered almost intact
The Aftermath from the bottom of waste bins that have been the
seat of very large fires. The lower rags tend to be
There are very often very characteristic signs of protected from incineration by a layer of char and
spontaneous ignition even after it has been the also by lack of oxygen in the lower reaches of
cause of a very large fire. Internal charring and the bin.
www.ebook777.com
and cool the cooked noodles for packing and

Case Histories and Examples palleting. Increasing the speed of the conveyer
certainly increased the throughput in proportion,
Cottonseed Meal: Living Dangerously but the smaller length of time the noodles spent
in the hot oil resulted in incomplete cooking.
A transit warehouse temporarily storing cotton- Thus the oil-bath temperature was increased sub-
seed meal to a depth of about 3 m burned down stantially to compensate for this and again pro-
and was completely destroyed. The length and duce fully cooked noodles. However, the faster
breadth of the building were much larger than the moving belt was now conveying cooked noodles
depth of the meal, so the relevant physical to the packing area in a shorter time than before,
dimension (for substitution into the formula and they were also coming out of the fryer hotter
for δcritical) was 3 m. Spontaneous ignition was than before. The result was that they were packed
suspected because of the known presence of and palleted at a significantly higher temperature
unsaturated fatty acids prone to this. Standard than under previous management.
CAT tests for small laboratory samples were Although the scientific and technological lit-
carried out, and the extrapolation to life size erature contained no reference to spontaneous
was expected to be reasonably accurate because ignition of noodles, their porous and oily nature
only small amounts of water were present and indicated a possibility that this could occur. This
wet reaction was not suspected. was confirmed by laboratory tests obtaining the
The body of meal in the warehouse turned out CAT for a particular size noodle block. On this
to be supercritical for the average ambient tem- basis a full series of tests was carried out, and
perature in the area. The unusual factor in this the parameters for the noodles obtained from the
particular case was the fact that the meal had Frank-Kamenetskii plot in the usual way. With
been left undisturbed for much longer than these parameters available, it was possible to
usual due to a transport strike. It remained in calculate the CST for a pallet full of noodle
place for longer than the TTI, although under packages as these were shrink-wrapped onto the
normal circumstances it would have been pallets and completely encased in plastic, that is,
moved on to customers well before significant the whole pallet full of noodles was in fact the
self-heating could take place. body in question. The calculated CSTs (for a
In this case an enlightened management range of feasible ambient temperatures) turned
installed underfloor ducting to produce a high- out all to lie above the temperatures reached with
pressure air blast capable of rearranging the meal the old process parameters but well below the
substantially from time to time. A similar solu- temperatures reached with the new high-
tion has long been practiced for coal stockpiles, productivity parameters. The “bean counters”
although in that case the disturbance is usually managed to achieve a productivity of zero until
caused by a front-end loader. the factory was rebuilt.
Flaming Instant Noodles Bagasse Storage: Some Complex

Chemistry
Some years ago an instant noodle factory burned
down soon after new management had taken The sugar industry in Australia wished to use
over. New management was not satisfied with bagasse containing the usual 50 % moisture as a
the throughput of the production line and wanted biomass for cogeneration of electricity as large
higher productivity. The latter was dependent on excess tonnages are produced biannually.
the speed of a single conveyer belt that conveyed Removal of moisture increases the calorific
the raw noodles through a hot oil bath, then under (and hence monetary) value of the material as a
a number of powerful fans to remove excess oil fuel, provided it can be removed at no energy
628 B.F. Gray
cost. At the same time it has been known for through the pile, results in probably the most
some time that large piles of bagasse are prone complex modeling yet of ignition phenomena.
to spontaneous ignition and self-heating with Nevertheless, this model describes quantita-
consequent loss of value and also considerable tively the behavior of real bagasse piles and
pollution from the combustion products. An answers the questions that led to its creation,
obviously desirable aim would be to create piles that is, How does one choose a pile size in
of bagasse that are not large enough to be super- order to maximize the water removal without
critical but nevertheless large enough to self-heat losing the pile to spontaneous ignition? The
significantly and hence drive off some of the modeling is described in a number of
moisture at no cost. Thus one would turn a publications (e.g., Gray et al. [12]) and shows
dangerous energy release into a benefit. Clearly that present-day computing power coupled with
the balance would have to be just right. appropriate knowledge of physical parameters
Consequently, a major research project was enables quantitative or at worst semiquantitative
undertaken, both experimental and theoretical. modeling of spontaneous ignition situations with
Application of the standard laboratory test input of realistic chemistry and transport pro-
methods to bagasse [54] results in a prediction cesses. Such developments have also taken
of critical radius for a pile at ambient temperature place in the modeling of realistic chemistry in
30 C, which is an order of magnitude greater gas-phase ignition of hydrocarbons and related
than the observed value. This is now known to be organic materials.
due to the fact that laboratory test CATs are It seems that we are not far from a situation
above 100 C and simply drive off the moisture where the simplified theories that have been use-
before the self-heating can get under way. The ful tools for so long (with their empirical
extrapolated results are therefore only good corrections) will be superseded by more detailed
predictors for dry piles of material. In practice calculation of the required properties such as
the water content of bagasse is close to 50 % on a CATs and CSTs. Nevertheless, the simplified
dry-weight basis, and this has recently been theories will never lose their pedagogical value
shown to be instrumental in partaking in a heat- and will remain a firm conceptual foundation for
producing reaction in addition to the one more sophisticated models.
predominating in the dry material at higher
temperatures [20, 21]. This wet reaction has
been characterized in isothermal calorimetric Milk Powder: A Numerical Example
measurements over the temperature range
30–90 C, and in this range the high-activation- This example is due to Beever [55]. In a milk-
energy dry reaction is almost completely shut drying plant, air entering the spray dryer was
down by the negative exponential in the heated to 200 C, and it was thought that surfaces
Arrhenius function. in the region of the inlet would also reach this
The wet reaction does not follow an Arrhenius temperature. Any collection of powder on hot
temperature dependence at all, rather having a surfaces could cause spontaneous ignition,
maximum rate at about 55–60 C. It also has a which would not only spoil the product but also
sharp, almost discontinuous dependence on act as a source of ignition for a dust explosion.
water concentration, cutting out completely These have occurred in milk-drying plants with
below 20 % moisture. These characteristics are devastating consequences. Farther down the
probably responsible for its occurrence being dryer, where there was deemed to be a greater
mistaken for microbiological activity. Inclusion likelihood of powder accumulation, surface
of such complex chemistry in a generalization of temperatures of 80 C occurred. Three laboratory
the Frank-Kamenetskii theory for distributed basket sizes were tested with half side-lengths of
temperatures, as well as the evaporation, conden- 0.025, 0.0375, and 0.050 m. The CATs of these
sation, and diffusive movement of water vapor were 171 C, 156 C, and 141.5 C, respectively.
www.ebook777.com
For a cube we can substitute the value 2.52 for Definition 1. Test Method H1—U.S. SADT Test
δcritical in Equation 20.20: The SADT is defined as the lowest oven tem-
" # perature at which the sample temperature
δcritical T 2a, critical 9497 exceeds the oven temperature by 6 C or more.
ln ¼ 41:85
r2 T a, critical The test is run for 7 days, the time origin being
when the sample temperature is 2 C below the
We can make r the argument of this equation and oven temperature, assuming the body is at room
then substitute for Ta,critical as required. If we temperature when inserted. The occurrence of
require the critical temperature for a layer of mate- three completely arbitrary numbers (7 days,
rial, we would use the value for δcritical appropriate 2 C, and 6 C) should arouse suspicion in even
to an infinite slab, that is, 0.88. For such a flat layer the most scientifically illiterate person that this
with ambient temperature on each side of 200 C, a SADT definition cannot be a fundamental prop-
critical thickness of 0.017 m is obtained. For the erty of the body tested. Detailed considerations
cooler regions of the dryer at 80 C, a critical using fundamental thermal ignition theory show
thickness of 0.4 m is obtained. It was decided that that the SADT definition does not indicate that
these critical thicknesses were sufficiently realistic the test body will not blow up below the SADT.
to require regular cleaning inside the dryer to Some materials could do so, and similarly some
remove buildup. This problem is actually more may not do so when placed in an ambient tem-
complicated than indicated here because the criti- perature above the SADT.
cal parameters are rather sensitive to moisture con- On the other hand the critical ambient temper-
tent, the critical thickness increasing significantly ature (CAT) defined and discussed in this article is
with moisture content, which can be up to 4 % [56]. a rigorous dividing value between ignition and
subcritical self-heating. Why any version of the
SADT would have found its way into the scientific
Technical and Legal Matters fringe literature of regulation and litigation is
rather incomprehensible, but things get worse
SADT when we consider the UN tests H2 (adiabatic stor-
age test) and H3 (isothermal storage test). We will
Many definitions, particularly those published not go into the experimental details of these tests
by the United Nations (UN), the International but go straight to the definitions of SADT arising
Maritime Organisation (IMO), and the out of them. Both tests make use of a diagram
U.S. Department of Transportation (DOT), are (plotted from the test results) that is in fact identi-
incorporated in toto in various regulations and cal to Fig. 20.1. This diagram is given as Fig-
statutes, thus attaining a rigorous legal standing. ure 28.4.2.2 and again as Figure 28.4.3.2 in the
As a result, they assume a status that scientifically Manual of Tests and Criteria. Both of these
they do not deserve by virtue of being wrong, diagrams show and explicitly refer to the critical
ambiguous, confusing, or all three and more. ambient temperature but unbelievably go on to
The centerpiece of such consideration is the define the SADT as follows.
self-accelerating decomposition temperature, or
SADT. Some of the definitions of SADT (there Definition 2. Test Methods H2 and H3—The
could be more) are presented here. The UN’s Adiabatic Storage Test and Isothermal Storage
Manual of Tests and Criteria [35], in its section Test, Respectively
“Recommendations on the Transport of Danger- The SADT is the CAT rounded up to the next
ous Goods,” contains at least four definitions. higher multiple of 5 C.
Some are inconsistent with each other and some It follows from this that
are actually defined so as to be up to 5 C Every material can blow up below its SADT
above the critical ambient temperature. The as defined by UN tests H2 and H3.
materials could therefore blow up well below The U.S. Department of Transportation uses
the SADT! SADT values measured by any of these tests to
630 B.F. Gray
impose requirements for temperature-controlled Such errors in activation energies are disas-
transportation (CFR 49). Clearly it does not err trous in the prediction of critical ambient
on the side of safety. temperatures because these are extremely sensi-
tive to the activation energy. Generally speaking,
the very factors that endow a substance with the
ASTM E698-01 propensity to ignite spontaneously also tend to
increase the “correction factor” for the standard
ASTM E698-01 is entitled Standard Test Method method of data extraction from DSC results.
for Arrhenius Constants for Thermally Unstable These are well known to be low thermal conduc-
Materials, and the abstract states [57]: tivity, low activation energy, large heat of reac-
The kinetics of exothermic reactions are important tion, and high pre-exponential factor. The fact
in assessing the potential of materials and systems that necessary corrections to classical DSC
for thermal explosion. This method provides a results are directly proportional to the Frank-
means for determining Arrhenius activation Kamenetskii parameter very clearly illustrates
energies and pre-exponential factors using differ-
ential thermal methods. this problem for the use of DSC in estimating
self-ignition propensities in self-heating
A paper has recently been published [58] that materials.
throws very considerable doubt on the activation
energies obtained by this method notwithstand-
ing the caveats in the standard itself. Calcium Nomenclature
hypochlorite can be used to illustrate a difficulty Cν Heat capacity at constant volume per
arising from the lack of sensitivity of typical unit mass (J/K
mol)
differential scanning calorimetry (DSC) results c Concentration (mol/m3)
when two reactions are involved and they have cf Feed concentration in CSTR (mol/m3)
widely differing kinetic parameters. The CAT Critical ambient temperature (K)
Kissinger plot gives a good straight line but in CST Critical stacking temperature (K)
fact the presence of a low activation energy, E Activation energy (J/mol)
low-temperature reaction is not detected. The F Feed rate in CSTR (m3/s)
exotherm obtained relates to the high- f(c) Chemical reaction rate (mol/m3
s)
temperature, high activation energy reaction, Q Heat of reaction (J/mol)
albeit somewhat inaccurately. In the large-scale R Universal gas constant (J/mol
K)
practical situation the low-temperature reaction r Characteristic radius
is in fact the cause of thermal ignition, and the S Surface area (m2)
conclusions drawn from DSC results are danger- T Temperature (K)
ously in error. Ta Ambient temperature (K)
Modeling experiments in which either reac- Ta,critical Critical ambient temperature (CAT) (K)
tion is left out show that the low-temperature Tf Feed temperature in CSTR (K)
reaction does not show an exotherm until a tem- TTI Time to ignition (s)
perature about 200 C above the position of the u Dimensionless temperature (RT/E)
exotherm for the high-temperature reaction ua Dimensionless ambient temperature
under the conditions of DSC tests. V Volume of self-heating body (m3)
The practically dangerous low-temperature ν Dimensionless concentration (c/c0)
reaction does not produce heat at a sufficiently δ Frank-Kamenetskii parameter
high rate to be detected in a typical DSC until θ Frank-Kamenetskii dimensionless
very high temperatures. Before such high temperature
temperatures can be reached during a tempera- ρ Bulk density (mol/m3)
ture ramp, the material is consumed by the high- κ Thermal conductivity (W/m
K)
temperature reaction, and the low-temperature χ Heat-transfer coefficient (W/m2
K)
reaction is missed completely. ε Inverse dimensionless heat of reaction
www.ebook777.com
τ Dimensionless time 18. B.F. Gray and B. Halliburton, “The Thermal Decom-
l Dimensionless heat transfer position of Hydrated Calcium Hypochlorite UN
2880,” Fire Safety Science, 35, pp. 223–239 (2000).
coefficient 19. Y. Uehara, H. Uematsu, and Y. Saito, “Thermal Igni-
Bi Biot number (χr/κ) tion of Calcium Hypochlorite,” Combustion and
∂()/∂n Differential coefficient in a direction Flame, 32, p. 85 (1978).
normal to the boundary of the body 20. I.K. Walker, W.J. Harrison, and F.H. Jackson,
New Zealand Journal of Science, 21, p. 487 (1978).
21. R.A. Sisson, A. Swift, G.C. Wake, and B.F. Gray,
“Critical Conditions for the Exothermic Combustion
References of Damp Cellulose, Part 1,” IMA Journal of Applied
Mathematics, 50, p. 285 (1993).
22. R.A. Sisson, A. Swift, G.C. Wake, and B.F. Gray,
1. Safety and Runaway Reactions, Institute for Systems “Critical Conditions for the Exothermic Combustion
Informatics and Safety, Joint Research Centre of Damp Cellulose, Part 2,” IMA Journal of Applied
European Commission, EUR 17723 EN (1998). Mathematics, 49, p. 273 (1992).
2. N.N. Semenov, Zeitschrift für Physikalische Chemie, 23. B.F. Gray and G.C. Wake, “The Ignition of Hygro-
48, p. 571 (1928). scopic Combustible Materials by Water,” Combustion
3. D.A. Frank-Kamenetskii, “Diffusion and Heat Trans- and Flame, 79, p. 2 (1990).
fer” in Chemical Kinetics, Plenum Press, New York 24. B. Halliburton, Ph.D. Dissertation, Macquarie Univer-
(1969). sity, Sydney, Australia, p. 2109 (2002).
4. P.C. Bowes, Self-Heating: Evaluating and 25. T. Dixon and N. Ashbolt, Private communication,
Controlling the Hazards, HMSO, London (1984). Sugar Research Institute, Mackay, Queensland,
5. V. Babrauskas, Ignition Handbook, Fire Science Australia (1985).
Publishers, Issaquah, WA (2003). 26. International Maritime Dangerous Goods Code, Inter-
6. J.F. Griffiths and B.F. Gray, “Fundamentals of national Maritime Organization, London, 2010.
Autoignition of Hydrocarbon and Other Organic 27. J.C. Jones, “On the UN test for the Spontaneous
Substrates in the Gas Phase,” in 24th Loss Prevention Heating of Solids,” Loss Prevention in the Process
Symposium, American Institute of Chemical Industries, 13, pp. 177–178 (2000).
Engineers, San Diego, 92b (1990). 28. X.D. Chen and L.V. Chong, “Self-Ignition Kinetics of
7. N.N. Semenov, Some Problems in Chemical Kinetics Combustible Solids,” Transactions of Institution of
and Reactivity (English translation), Princeton Uni- Chemical Engineers, 76B, p. 90 (1998).
versity Press, Princeton, NJ (1959). 29. W. Mackey, “On a Spontaneous Ignition Test,” Jour-
8. G.C. Wake, J.B. Burnell, J.G. Graham-Eagle, and nal of the Society of Chemical Industry, 15, p. 90
B.F. Gray, Reaction Diffusion Equations (1896).
(K.J. Brown and A.A. Lacey, eds.), Oxford, UK, 30. N. Kirov, “The Crossover Test,” CSIRO Technical
pp. 25–37 (1990). Note, Chatswood, Australia (1954).
9. J.F. Griffiths and J.A. Barnard, Flame and Combus- 31. P.H. Thomas, “On the Thermal Conduction Equation
tion, Blackie, Glasgow, Scotland (1994). for Self-Heating Materials with Surface Cooling,”
10. B.F. Gray, “Critical Behaviour in Chemically Transactions of the Faraday Society, 54, p. 60 (1958).
Reacting Systems III—An Analytical Criterion for 32. P.H. Thomas, Transactions of the Faraday Society,
Insensitivity,” Combustion and Flame, 24, p. 43 56, p. 833 (1960).
(1975). 33. B.F. Gray and G.C. Wake, “Criticality in the Infinite
11. B.F. Gray, “Unified Theory of Explosions, Cool Slab and Cylinder with Surface Heat Sources,” Com-
Flames and Two Stage Ignitions,” Transactions of bustion and Flame, 55, p. 23 (1984).
the Faraday Society, 65, p. 1603 (1969). 34. B.F. Gray, A. Gomez, and G.C. Wake, “Friction and
12. B.F. Gray, M.J. Sexton, B. Halliburton, and Localised Heat Initiation of Ignition,” Combustion
C. Macaskill, “Wetting Induced Ignition in Cellulosic and Flame, 61, p. 177 (1985).
Materials,” Fire Safety Journal, 37, pp. 465–479 35. “Transport of Dangerous Goods,” Manual of Tests
(2002). and Criteria, 2nd ed., United Nations, New York
13. T. Dixon, “Spontaneous Combustion in Bagasse and Geneva (1995).
Stockpiles,” in Proceedings of the Australian Society 36. B.F. Gray, “On the Critical Conditions for an Assem-
of Sugar Cane Technologists, 53 (1988). bly of Interacting Thermons, Series B,” Journal of the
14. B.F. Gray, unpublished. Australian Mathematical Society, 43, pp. 1–12 (2001).
15. R. Aris, The Mathematical Theory of Diffusion and 37. B.F. Gray, “Interpretation of Small Scale Test Data,”
Reaction in Permeable Catalysts, Oxford, UK (1975). Interflam 2001, 9th International Fire Science and
16. P.L. Chambre, Journal of Chemical Physics, Engineering Conference, Edinburgh, Scotland,
20, p. 1795 (1952). pp. 719–729 (2001).
17. T. Boddington, P. Gray, and I. Harvey, “Thermal 38. B.F. Gray, S.G. Little, and G.C. Wake, “The Predic-
Theory of Spontaneous Ignition,” Philosophical tion of a Practical Lower Bound for Ignition Delay
Transactions of the Royal Society, A270, p. 467 Times,” 24th International Combustion Symposium,
(1971). Combustion Institute, Pittsburgh, PA, p. 1785 (1992).
632 B.F. Gray
39. B.F. Gray and J.H. Merkin, “Thermal Explosion 50. A.C. McIntosh and B.F. Gray, “Self Heating of Com-
Escape Times in the Uniform Temperature Approxi- bustible Vapor in Porous Material When Fibres Are
mation,” Mathematical Engineering in Industry, Completely Covered by Fluid,” Combustion Science
4, p. 13 (1993). and Technology, 113, p. 503 (1996).
40. P.H. Thomas, “An Approximate Theory of Hot Spot 51. A.C. McIntosh, B.F. Gray, and G.C. Wake, “Analysis
Criticality,” Combustion and Flame, 21, p. 99 (1973) of the Bifurcational Behaviour of a Simple Model of
41. J. Zinn and L. Mader, “Thermal Initiation of Vapor Ignition in Porous Material,” Proceedings of
Explosives,” Journal of Applied Physics, 31, p. 323 the Royal Society, A453, p. 281 (1997).
(1960). 52. B.F. Gray and S.K. Scott, “The Influence of Initial
42. T. Boddington, C. Feng, and P. Gray, “Thermal Temperature Excess on Critical Conditions for Ther-
Explosion, Times to Ignition and Near Critical mal Explosion,” Combustion and Flame, 61, p. 227
Behaviour in Uniform Temperature Systems, Part (1985).
2,” Journal of the Chemical Society, Faraday 53. B.F. Gray and G.C. Wake, “Critical Conditions for
Transactions, 2, 79, p. 1299 (1983). Thermal Ignition,” Mathematical and Computer
43. T. Boddington, C. Feng, and P. Gray, “Thermal Modelling, 18, pp. 65–75 (1993).
Explosion, Times to Ignition and Near Critical 54. B.F. Gray, J.F. Griffiths, and S.M. Hasko, “Spontane-
Behaviour in Distributed Temperature Systems, Part ous Ignition Hazards in Stockpiles of Cellulosic
3,” Journal of the Chemical Society, Faraday Materials,” Journal of Chemical Technology and Bio-
Transactions, 2, 80, p. 1155 (1984). technology, 34A, p. 453 (1984).
44. T. Boddington, C. Feng, and P. Gray, “Thermal 55. P. Beever, “Self Heating and Spontaneous Combus-
Explosion and Times to Ignition I, Reactant Con- tion,” in SFPE Handbook of Fire Protection Engi-
sumption Ignored,” Proceedings of the Royal Society, neering, 2nd ed. (P.J. DiNenno et al., eds.),
A385, pp. 289–311 (1982). pp. 2-180–2-189 (1988).
45. T. Boddington, C. Feng, and P. Gray, “Thermal 56. C.M. Rivers, Master’s Thesis, Massey University,
Explosion and Times to Ignition II, Reactant Con- Palmerston North, New Zealand (1994).
sumption Included,” Proceedings of the Royal Soci- 57. ASTM E698-01, Standard Test Method for Arrhenius
ety, A391, p. 269 (1984). Constants for Thermally Unstable Materials, ASTM
46. B.F. Gray and J.H. Merkin, “Thermal Explosion: International.
Escape Times in the Uniform Temperature Approxi- 58. B.F. Gray and C. Macaskill, “The Role of Self-
mation: I Effects of Parameter Perturbations,” Journal Heating in the Estimation of Kinetic Constants for
of the Chemical Society, Faraday Transactions, 2, 86, Unstable Materials Using DSC,” Interflam 2004,
p. 597 (1990). 10th International Fire Science and Engineering Con-
47. S.G. Little, Master’s Thesis, School of Chemistry, ference, Interscience Communications, London, UK
Macquarie University, Sydney, Australia, p. 2109 (2004).
(1991).
48. Y.I. Rubtsov, A.I. Kazakov, L.P. Andrienko, and
S.B. Manelis, “High Temperature Pyrolysis of Cellu- Brian Gray has published extensively on spontaneous
lose,” Combustion, Explosion, and Shock Waves, combustion and ignition of gases and solids, gas ignition,
29, p. 710 (1993). flame propagation, and related topics for many years. He
49. L.G. Britton, “Spontaneous Ignition of Liquids on now runs a combustion consulting company in Sydney,
Porous Media,” 24th Loss Prevention Symposium, Australia, and was professor of chemistry at Macquarie
American Institute of Chemical Engineers, San University, Sydney, from 1976 to 1998.
Diego (1990).
www.ebook777.com
Flaming Ignition of Solid Fuels

21
José Torero
Introduction • Simplifications where the required precision

does not warrant the inclusion of higher levels
This chapter will describe how heating of a solid of complexity
fuel leads to flaming ignition. The discussion will • etc.
be centred on flaming ignition of solid fuels but To highlight the impact of simplifications,
will not address smouldering or spontaneous whenever possible, a comparison between the
ignition since these subjects will be covered in comprehensive description and the reduced for-
Chaps. 19 and 20 respectively. Thus, the pres- mulation will be made to allow the reader to
ence of a source of heat decoupled from the solid assess potential errors. As this chapter progresses
and fuel gasification will be assumed throughout the resulting formulations become simpler and of
the chapter. greater practical use, nevertheless the impact
The main focus of this chapter is to assist the of the assumptions strengthens increasing the
reader in understanding the phenomena, potential for error or misuse of the information.
assumptions and simplifications embedded in The chapter closes with a presentation of the
different models and tests that attempt to predict simplest methodologies that correspond to clas-
ignition phenomena or to extract the parameters sic treatments and are mainly associated to
controlling it. The methodology to be followed standard tests.
goes from the general to the specific. Therefore, Most of the existing data on ignition is inti-
the problem will be initially formulated in as mately related to the methodology used to
general a manner as possible. A series of com- extract it. Therefore, it is always conditioned
mon simplifications will then be made leading to by the nature of the test procedures, the hardware
reduced formulations. These simplifications are used and by the data analysis method. Given that
introduced for many reasons that include: the objective of this chapter is to provide a
• Simplifications where the nature of the mate- phenomenological description of flaming igni-
rial studied allows the exclusion of some spe- tion of solid fuels emphasis will be given to the
cific phenomena different processes and not to reviewing avail-
• Simplifications where some processes disap- able data. Throughout this chapter the reader will
pear due to the characteristics of the test used be directed to other chapters and references
to assess the material where data will be presented in the context of
the testing protocols used to obtain it. As an
example, some of the most comprehensive
compilations of ignition data can be found in
J.Torero (*)
School of Civil Engineering, The University Chap. 36, in textbooks such as Ref. [1] or pro-
of Queensland, St Lucia, QLD, Australia fessional guidelines [2].

634 J. Torero
presented in Fig. 21.1. Figure 21.1 also shows all

The Process of Ignition the different variables that evolve through the
heating process. These variables will be described
When a solid material, initially at ambient tem- in detail later.
perature, is subject to an external source of For simplicity, all processes involved will be
energy the temperature of the exposed surface divided in two groups, those associated with the
starts to increase. This moment will be defined solid phase and those with the gas phase. The
as the onset of the ignition process, t ¼ 0. A solid phase treatment will lead to a description of
00
series of physical and chemical phenomena are the production of gas phase fuel (ṁP ) and the gas
initiated as the energy reaches the surface of the phase analysis will focus on how the ensemble of
material. This chapter will attempt to describe gaseous fuel and oxidizer lead to a flame. The
these phenomena. solid phase will be treated first (section “The
Without loss of generality and for simplicity, Solid Phase”), then the boundary conditions
the ignition process will be described in a between both phases will be established, to final-
one-dimensional frame of reference with a single ize with a description of what happens in the gas
coordinate, x. Only one surface of the material phase (section “The Gas Phase”).
will be heated and the origin of the coordinate
system, x ¼ 0, will be placed at the exposed sur-
face of the material. This frame of reference will
move with a velocity VR as the fuel consumes and The Solid Phase
the surface regresses. For some materials regres-
sion rates are very small and can be neglected, but The temperature of the solid, initially at ambient
this will not be assumed at this stage. A schematic (To), increases as the heat reaches the surface of
of a generic solid material undergoing heating is the material. Highest temperatures will be
⋅
qe″ (t)
Pilot
⋅
q″Cv (0, t)
m⋅ p″ q⋅ SR
″ (0, t)
r
YF,g(0,t) YF,s(0,t) x=0 m⋅ ″O χ (0,t) T,χ
YO,YF Yo,(0,t) To T(0,t)
VR
x=εo x=εCH
VCH χ (x,t)
x=εF
x=εP
VP
x=εT
VT
YF,s(x,t)
x=L
q⋅ ″N (L, t)
x
Fig. 21.1 Schematic of the different processes occurring as a material undergoes degradation prior to ignition induced
by an external source of heat
www.ebook777.com
21 Flaming Ignition of Solid Fuels 635
achieved close to the surface, but energy transfer generally named the pre-exponential constant. In
in-depth will result in an increase in temperature the case of a polynomial description, “A” will be
of a significant part of the solid. Therefore, the replaced by another constant that is here defined
temperature will vary in-depth and in time, thus as “C.” These constants are a characteristic of
temperature needs to be represented as a function each specific chemical reaction. “E” is the acti-
of both variables, T(x,t). Figure 21.1 shows a vation energy whose magnitude is also specific to
generic representation of the temperature distri- each reaction and “R” has the value of
bution at a particular instant in time, t. The evo- 8.314 103 kJ mol1 K1.
lution of the temperature is defined by an energy The process of pyrolysis can be extremely
balance in control volumes between both complex and depending on the fuel and heating
surfaces of the solid (x ¼ 0 and x ¼ L). The characteristics can follow distinctively different
surfaces will define the heat transfer in/out of paths. These paths can be a compendium of
the solid fuel or mathematically, the boundary numerous reactions that could be sequential or
conditions. It is important to note, that if other compete against each other. Furthermore, the
dimensions were to be considered, similar chemical pathways followed can be strongly
boundary conditions will have to be established influenced by the presence of oxygen as
at each surface of the material. indicated in Equations 21.1 and 21.2. In both
equations YO and YS are intended to be generic
representations of oxygen and solid fuel mass
Pyrolysis Process fractions participating in the solid degradation
and “m” and “n” are constants.
The process by which the solid transforms into It is important to note that while degradation
gas phase fuel is called pyrolysis and generally of some fuels will show dependency on the oxy-
implies the breakdown of the molecules into gen concentration many others will not [3, 4]. In
different, most of the time smaller, molecules. those cases “m” is assumed to be zero.
This is an important difference between solid The chemical pathways leading to the pyroly-
and liquid gasification. In the case of liquids a sis of most solid fuels of interest in fire are
change of phase is not necessarily accompanied fundamentally incomplete as much as the
by a chemical change (see Chap. 18 for details constants associated to the equations that will
on ignition of liquids). Pyrolysis tends to be an serve to quantify the rate of each reaction step.
endothermic process generally controlled by Many studies have evaluated reduced chemical
many chemical reactions (some time hundreds) mechanisms for the pyrolysis of different solids
which are a strong function of the temperature. [5–7] but there is still great uncertainty on the
Most pyrolysis reaction rates tend to be chemical pathways, the number of steps required
described by Arrhenius type functions of the and the constants associated to them [8, 9].
temperature Figure 21.2 shows an example of chemical
kinetic compiled obtained for PMMA by using
ω_ ¼ AYO m YS n eE=RT ð21:1Þ an expression similar to Equation 21.1.
But could also be described by other simple Thermo-Gravimetric Analysis (TGA) has
expressions like polynomials such as been used in the past to establish reduced
chemical reaction mechanisms as well as the
ω_ ¼ CYO m YS n ðT=To Þb ð21:2Þ associated constants. The principles behind
TGA studies and some applications to materials
The reaction rate is generally defined in units of relevant to fire are presented in Chap. 7. As an
inverse seconds ðω_ ½1=sÞ and only when example, a reduced kinetic mechanism for poly-
multiplied by the fuel density gives a gasification urethane (PU) can be found in Ref. [6]. The
000
rate per unit volume (ω_ , kg/s.m3). The constant authors propose a four step mechanism of the
“A” is also given in inverse seconds (1/s) and form:
636 J. Torero
Fig. 21.2 Kinetic parameters of the pyrolysis decomposition of PMMA as reported in the literature [8]
Step i ¼ 1 PU ! νβ, p β‐PU þ νg, p Gas

Step i ¼ 2 β‐PU ! νc, pβ Char þ νg, pβ Gas
(
PU þ νO2, o O2 ! νc, o Char þ νg, o Gas
Step i ¼ 3
β‐PU þ νO2, o O2 ! νc, o Char þ νg, o Gas
Step i ¼ 4 Char þ νO2, c O2 ! νr, c Residue þ νg, c Gas
where the reaction rate for each step ðωÞ _ is Figure 21.3 shows two curves extracted from
presented by an expression of the form of Equa- Ref. [6] where the model is compared to experi-
tion 21.1. The first two steps encompass purely mental data for inert and air atmospheres. The
thermal degradation, while the last two steps figures show the sample mass loss rate as a func-
include oxidation. Two intermediate products tion of time. The effect of oxygen and the impact
are formed from the initial degradation of the of the heating rate are evident from the data. The
polyurethane, β-PU and Char. While the terms results show very good agreement with the four
Gas and Residue represent the gaseous and solid step model for all conditions studied.
products of the degradation. It is important to Despite the generalized use of TGA data,
note that there are sequential and competing there is increasing recognition that pyrolysis
reactions; while steps 1 and 2 are sequential, reaction pathways are sensitive to the heating
step 3 competes with both previous steps. The rate. The basic nature of TGA studies requires
authors use independent TGA data [10] to obtain heating rates of the order of 1–20 C/min which
all 12 constants thus establishing a complete is generally an order of magnitude slower that
model for the degradation of polyurethane. heating rates typical of fires. Recent studies have
www.ebook777.com
Fig. 21.3 TGA data for polyurethane extracted from reference [10] (symbols) superposed to the reduced kinetic model
from reference [7] (lines) for inert atmosphere (a) and air (b)
established methodologies that use standard test material will not follow any transformation. As
methods and advanced optimization techniques the temperature increases the reaction rates
to establish reduced reaction schemes and increase and the solid fuel starts changing. Given
their associated constants [7, 11–14]. Although, the temperature distribution within the material,
these procedures allow exposing the materials to the rates of decomposition are a function of “x,”
heating rates typical of fires and obtaining with larger production of pyrolyzates close to the
comprehensive sets of constants they have only surface and lower production in-depth.
been applied to a reduced number of materials. Local production of fuel is not the only impor-
Currently, these methodologies remain funda- tant variable. The gas phase fuel produced might
mentally research tools. be the result of a combination of pyrolysis and
It is important to note that the qualitative and oxidation reactions, thus its composition might
quantitative agreement described in the above include large quantities of fully oxidized
example is not usual for materials commonly pres- compounds such as carbon dioxide (CO2), par-
ent in fire and the problem of establishing the tially oxidized gases such as carbon monoxide
chemistry of pyrolysis is far from being solved. (CO) and other molecules that can have all levels
Therefore, when studying flaming ignition of of partial oxidation. Therefore, together with the
solids it is common to make strong simplifications reaction rates, the mass fraction of inert gases
to handle chemical degradation as the solid fuel is needs to be subtracted leaving the remaining
heated. Such simplifications will be made later reactive gases. As an example, Kashiwagi and
and their impact will be assessed. Nambu [15] studied the degradation products of
cellulosic paper showing that there is a signifi-
cant presence of inert gases like water vapour,
The Production of Gaseous Fuel fully oxidized gases like CO2, partially oxidized
products like CO and fuel like CH4 and H2.
Before flaming ignition can occur, fuel in the gas There is very little data available on the deg-
phase needs to be produced. Solid materials that radation products of most materials relevant to
are not susceptible to spontaneous ignition will fire, therefore, the mass fraction of flammable
show very little evidence of chemical reactions gases present in the local products of degradation
at ambient temperatures, thus can be deemed as will be described here by means of a single
inert. The reaction rates associated to pyrolysis variable, YF,s(x,t), which represents a global con-
can be considered negligible and therefore the tribution of all compounds that can be further
638 J. Torero
oxidized. Figure 21.1 represents YF,s(x,t) as an permeability (χ(x,t)) will be introduced and
increasing function with a minimum at the sur- assumed to describe in a generic manner the frac-
face (YF,s(0,t)). This is based on the assumption tion of the fuel produced that can flow through the
that where there is a higher presence of oxygen solid material. It has to be noted that χ(x,t) is not
there is higher levels of oxidation. strictly a permeability function (as per Darcy’s
Oxygen can migrate inside a fuel resulting also law) but a combination of permeability, porosity
in an in-depth distribution (YO(x,t)) that reaches and any fractures within the material.
ambient values at the surface (YO(0,t)). In-depth Oxygen and fuel concentrations will be con-
oxygen and fuel diffusion is controlled by the trolled by the local permeability and by produc-
structure of the solid. Some materials are highly tion/consumption rates, thus indirectly by the
permeable and allow unrestricted transport of temperature distribution (T(x,t)). This makes nec-
species in and out of the solid. For other materials essary to treat them independently, therefore two
oxidation will occur only very close to the surface independent variables emerge, εF(t) and εO(t). The
and could be potentially neglected. The perme- former represent the region where fuel is being
ability of the fuel can be a function of many produced while the latter represents the region
variables including the degradation and con- where oxygen is present in relevant quantities.
sumption of the material and has deserved very If all the reactions occurring can be
little attention in the fire literature. In the absence represented in an Arrhenius form (Equation 21.1)
000
of a well defined permeability function, here, a then the local mass production (ṁP (x, t)) can be
simple variable associated to the fuel the summarized into a function of the form:
i¼N h
X i
000
Ei i =RTðx;tÞ
_ P ðx; tÞ ¼ YF, s ðx; tÞ
m O ðx; tÞYS ðx; tÞe
Ai Ymi ni
ð21:3Þ
i¼1
where the summation is not truly a sum of all the in-depth does not have to come out, and in many
different “N” reaction steps but just some global cases pressure increases within the fuel structure
combination of them that includes sequential and can be observed. The effects of permeability and
competitive reactions. pressure are combined in a complex manner to
To obtain the total fuel production at the sur- define the flow within the porous medium. This
00
face per unit area (ṁP (0, t)) it is necessary to remains an unresolved problem, thus the use of a
integrate Equation 21.3 across the entire depth simple variable such as χ(x,t) is justified.
including the permeability function described Integrating Equation 21.3 we obtain the follow-
above. It is important to note that fuel produced ing expression
ðL !
i¼N h
X i
00
Eii =RTðx;tÞ
_ P ð0; tÞ ¼
m χðx; tÞ YF, s ðx; tÞ O ðx; tÞYS ðx; tÞe
Ai Ym i ni
dx
0 i¼1
Assuming that any production of fuel is negligi-

ble for x > εF then the boundaries of integration
can be changed to
ð εF !
i¼N h
X i
00
Eii =RTðx;tÞ
_ P ð0; tÞ ¼
m χðx; tÞ YF, s ðx; tÞ O ðx; tÞYS ðx; tÞe
Ai Ym i ni
dx ð21:4Þ
0 i¼1
www.ebook777.com
Where the chemical reactions are left in a generic temperatures that can be achieved. Carbonaceous
form while recognizing that, due to the absence chars can reach much higher temperatures, lead-
of oxygen, the reactions occurring between εO < ing in many cases to vigorous oxidation (surface
x < εF might differ significantly from those glowing) that can be the catalyser for gas phase
occurring between 0 < x < εF. ignition. This will be part of the gas phase dis-
To summarize, the production of fuel is con- cussion. In what concerns the production of fuel,
trolled by the following variables: the differences appear mostly in-depth where
temperature is controlled by heat transfer through
Temperature T(x,t) the char and fuel production is affected by an
Local fuel concentration YS(x,t) overall permeability function. The effects of per-
Local oxygen concentration YO(x,t)
meability were described above and temperature
Residual fuel fraction YF,s(x,t)
effects on fuel production will be discussed in the
Permeability function χ(x,t)
context of the calculation of the temperature
Oxygen penetration depth εO(t)
distributions.
Reactive depth εF(t)
Kinetic constants Ai, mi, ni, Ei
Charring The Thermal Depth («T)
For the purpose of ignition of a solid fuel the When a heat flux is applied to one of the solid
process of charring has an impact on both heat surfaces, the heat travels across the solid fuel.
and mass transport therefore needs to be briefly Initially only a very small area is affected, but as
addressed. A general summary of the chemical the thermal wave travels through material a larger
processes leading to charring can be obtained in and larger fraction of the solid is heated. The
Chap. 7, and more details form Cullis and velocity of the thermal wave is represented in
Hirschler [16] for polymers and in the case of Fig. 21.1 by VT(t). VT(t) is a function of time
wood from Drysdale [17], thus will not be because it will decrease as the thermal wave
described here. Instead an attempt will be made moves away from the heating source and towards
to explain the role of charring in ignition. the cold back surface. The region that has been
For charring materials pyrolysis leads to the heated is quantified by the characteristic length
production of gaseous fuel (pyrolyzate) and a εT(t). It is important to note that, given that tem-
residual solid phase char. The char is mainly a perature is a continuous function, εT(t) has to be
carbonaceous solid that could be further arbitrarily defined simply as the end of the heated
decomposed. The secondary decomposition region. There is no exact mathematical definition
could be complete, leading to an inert ash or to for this length but physically it means that the
a secondary char that can be further decomposed temperature is approaching ambient temperature
in a single or multiple steps. Non-charring (T T0) or the gradient of the temperature is
materials decompose leaving no residue behind. approaching zero ( dT=dx 0 ). The proximity
From the perspective of ignition, the exposed that temperature or the gradient have to achieve
surface represents the boundary between the gas when approaching these targets is only a matter of
and the solid. This boundary moves as the mate- what precision is required by those making the
rial is completely removed. The rate at which the analysis.
surface moves is the regression rate (VR). For The length scale εT(t) is extremely important
charring and non-charring materials, this will be because it characterizes solids into different
the boundary where complete consumption of the groups. This breakdown enables the simplifica-
fuel is achieved. Although, regression rates can tion of the energy equation and the generation of
be very different between charring and simple analytical expressions for the temperature
non-charring materials, at the surface, the main distribution. For the purpose of ignition, solid
difference between the two material types is the fuels are classified in:
640 J. Torero
Semi-infinite Solid (L > εT) If the thermal boundary condition at x ¼ L, so if εT L then

wave is far from the end of the sample, the heat the boundary condition is defined as:
coming from the exposed surface has still not
x¼L ð21:6Þ
migrated to the back end. The temperature at

the back end is ambient (T0) and there are no dT
k
00
heat losses through this surface. The thickness of ¼ q_ N ðL; tÞ
dx x¼L
the sample is no longer a relevant quantity and
00
therefore the fuel can be treated as a semi-infinite where q_ N ðL; tÞ will be left as a generic heat loss
solid (L ! 1). Materials do not show semi- term at the back end of the solid fuel.
infinite solid behaviour forever, as time
progresses the thermal wave will eventually
reach the end of the sample. In many cases The Pyrolysis («P) and Charring
materials will behave as semi-infinite solids for Depths («CH)
the period of interest, in which case the assump-
tion of L ! 1 is valid. The boundary condition Within the region where the temperature has
for the energy equation becomes: increased above ambient significant chemical
x¼L!1 ð21:5Þ activity can occur. The chemical activity leads
to the production of fuel at a rate specified by
00
q_ N ð1; tÞ ¼ 0 equations of the type of (21.1) or (21.2). The
T ¼ T0 depth at which the chemistry can be assumed to
be significant is commonly defined as a pyrolysis
depth (εP) which propagates at a velocity VP.
Thermally-Thick and Thermally-Thin Solid
As with the thermal depth, there is no mathemat-
(εT L) The thermal wave has reached the end
ical function that describes the location of the
of the sample and therefore heat losses at the
pyrolysis front, x ¼ εP because the reaction
back end need to be quantified. The thickness of
equations are also continuous functions. Never-
the sample, L, becomes a relevant dimension of
theless, if the assumption is made that pyrolysis
the problem and a boundary condition for x ¼ L
reactions have high activation energy then the
needs to be defined. This group can be
transition between the zones of significant and
sub-divided into two different cases, thermally
negligible reactivity can be considered as being
thick and thermally thin. A solid can be defined
abrupt [18]. This permits the definition of critical
as thermally thick if a significant thermal gradi-
parameters that can be considered to define the
ent exists within the solid through the period of
onset of pyrolysis. The most common parameter
ignition. In contrast, in a thermally thin solid the
is a pyrolysis temperature, TP, below which the
gradient is negligible for most of the time before
solid fuel can be considered inert. It is important
ignition. A simple criterion based on the Biot
to note that the pyrolysis temperature is not a true
number (Bi) is generally used for the purpose of
physical parameter but a simple way to track the
establishing if a material is thermally thin or
onset of high activation chemical reactions.
thick. The Biot number is defined as Bi ¼ hL/k,
As described above, for x > εP the solid can
where “h” is a global heat transfer coefficient
be considered inert, thus thermal properties can
(W/m2K) and “k” is the thermal conductivity
be defined as those of the original solid fuel.
(W/mK). If Bi < <1 then temperature gradients
The thermal properties relevant to ignition are
inside the solid are negligible, while if the Biot
number is not much smaller than unity then tem- Density ρ(x,t) Kg/m3
perature gradients need to be considered. While Thermal conductivity k(x,t) W/m.K
this is an important distinction for the energy Specific heat C(x,t) J/kg.K
equation, it does not have an effect on the
www.ebook777.com
which are all functions of temperature. Since the phase changes that can have a significant effect
temperature varies in-depth they are also on the temperature distribution in the solid.
functions of “x.” The evolution of these Numerous models have been built in the past to
properties with temperature for common describe the heat sinks associated to melting and
materials can be found in most heat transfer several studies have attempted to quantify the
book [19], nevertheless, for materials typically impact of melting on practical situations such as
present in fires (wood, complex plastics, dripping.
composites, etc.) these properties are in many Phase changes are generally incorporated to
cases unknown [20, 21]. the energy equation as heat sinks where some
For x > εP the chemical reactions have rate function is created to describe the conversion
initiated the decomposition of the material. The from one phase to the other. The simplest proce-
relevant properties remain the same, nevertheless dure is to assign a critical temperature to the
pyrolysis introduces further changes to the phase change (i.e. 100 C for water) and a heat
properties. The gasification of the fuel and its of melting or evaporation (ΔHM). Once the fuel
transport towards the surface will strongly affect or water reaches this temperature it is converted
the density, while any potential voids will force to the high temperature phase. The phase change
to redefine thermal conductivity and specific heat process is assumed to be infinitely fast and there-
to account for the existence of at least two fore the rate is defined by the available energy
phases. reaching the location where the phase change is
The process of pyrolysis can lead directly to occurring. All the energy is then used for the
gasification with no residue (non-charring) or to phase change and the thermal wave can only
a carbonaceous residue (charring). Figure 21.1 proceed once the transition has been completed.
shows the case of a charring material where a This approach is inappropriate if the available
second front for charring (x ¼ εCH) is formed energy is very low, in this case thermodynamic
behind the pyrolysis front. The charring front equilibrium equations will define the rate of
will propagate at a velocity VCH and will leave vaporization or melting. Other more complex
behind a residue that will have a new set of models that include processes such as
properties that are potentially very different to re-condensation can be found in the literature
those of the fuel. The properties are still the but will not be discussed here.
permeability, the density, thermal conductivity The consequences of melting or water evapo-
and specific heat but precise values are mostly ration are various. Phase changes can affect the
unknown for most chars issued of materials rele- thermal properties of the fuel significantly and
vant to fires. can result in motion of the molten fuel or water
It is common to see in the char region large vapour. This leads to convective flow of energy
voids and cracks that compromise the or mass transfer.
one-dimensional treatment provided here. Understanding the physical processes behind
These have been considered when addressing phase change does not represent a great chal-
materials such as wood but will not be lenge. Furthermore, the potential impact of
described here. phase change on ignition is clear. Thus it is
evident that any predictive tool for ignition
should attempt to quantify the impact of phase
Melting and the Evaporation of Water change on ignition. Nevertheless, the formula-
tion of a model that can describe these processes
Melting or water evaporation have not been con- in a comprehensive manner is extremely com-
sidered in the description of the ignition until this plex and the measurements that could serve for
point. These two processes are endothermic its validation are mostly non-existent.
642 J. Torero
6.0E+05
5.0E+05
4.0E+05
kρC [W2/sm4K2]
3.0E+05
2.0E+05
1.0E+05 Glass Transition

Temperature
0.0E+00
0 50 100 150 200 250 300
o
Temperature [ C]
Fig. 21.4 Evolution of the product of the thermal conductivity, density and specific heat (kρC) for PMMA as a
function of temperature
Given that, phase change is fundamentally an The Temperature Distribution

additional heat sink that will have to be
incorporated to the energy equation in an arbi- As explained in section “The Production of
trary manner, it is justifiable to exclude the treat- Gaseous Fuel”, to determine the fuel production
ment of this subject from the present analysis. it is necessary to define the evolution of the
Nevertheless, this is done with the clear warning temperature inside the solid fuel. This can be
that its exclusion will have a significant impact achieved by defining a comprehensive energy
on any quantitative assessment of the ignition equation. Figure 21.4 represents a typical control
process. volume for x < εP where all the main heat trans-
Other processes that deserve to be addressed fer mechanisms are incorporated.
are softening or glass transition. Many materials For the purposes of this description the coor-
such as thermoplastics will undergo gradual or dinate system will be anchored to the regressing
drastic property changes with temperature. These surface, thus “x” will move with a velocity VR.
property changes are not endothermic but will A mass flow of fuel will therefore cross the
affect the progression of heat through the sample control volume presented in Fig. 21.5 carrying
00
and could lead to dripping. Softening or glass energy in and out (q_ S ). The gaseous products of
transition will be directly incorporated in the pyrolysis and oxygen diffusion will also carry
analysis through the variable properties 00 00
energy in and out of the control volume (q_ P , q_ O
described in section “The Pyrolysis (εP) and
respectively) and the generic expression for the
Charring Depths (εCH)”. An example of how 00 00
mass flow of these gases (ṁP , ṁO ) incorporates
these properties change with temperature is
the regression rate. Heat is conducted in and out
shown in Fig. 21.4. Figure 21.4 presents the 00
of the control volume ( q_ CND ) and for generality
evolution of the product of all three thermal 000
properties (kρC) for PMMA as a function of in-depth radiative absorption is allowed ( q_ RAD ).
temperature, indicating the abrupt change occur- Since for x < εP the temperature is sufficiently
ring at the glass transition temperature. high to allow for chemical reactions all heat
www.ebook777.com
⋅
q″P (x, t)
⋅ ⋅
q″O ( x, t) q″CND (x,t)
⋅
q″S ( x, t)
x=x
q⋅ ″′
g ( x, t) q⋅ ″′
RAD ( x, t)
x+=x +dx
q⋅ ″O ( x+ , t) q⋅ ″S ( x+ , t)
x q⋅ ″CND (x+ , t)
q⋅ ″P ( x+ , t)
Fig. 21.5 Typical control volume for x < εP showing the main heat transfer mechanisms
Table 21.1 Summary of all energy transport within a generic control volume for x < eP. DHP,i is the net heat resulting
from each individual chemical reaction. The net heat will be endothermic for most pyrolysis processes and exothermic
for oxidative reactions. The summation is not truly a summation, but as explained earlier, is the overall set of chemical
reactions where some could be sequential and others competing
Description In Out Formulation
00 00
Energy transported by gaseous q_ P ðxþ ; tÞ _ p ðxþ ; tÞCP, P ðxþ ; tÞTP ðxþ ; tÞ
m
fuel traversing the control volume 00 00
q_ P ðx; tÞ ṁp (x, t)CP,P(x, t)TP(x, t)
00 00
Energy transported by oxygen q_ O ðxþ ; tÞ m_ O ðxþ ; tÞCP, O ðxþ ; tÞTO ðxþ ; tÞ
traversing the control volume 00 00
ṁO (x, t)CP,O(x, t)TO(x, t)
q_ O ðx; tÞ
Energy transported by solid 00
q_ S ðxþ ; tÞ ρS ðxþ ; tÞVR ðtÞCS ðxþ ; tÞTðxþ ; tÞ
fuel traversing the control volume 00
ρS(x, t)VR(t)CS(x, t)T(x, t)
q_ S ðx; tÞ

Heat conduction 00
q_ CND ðx; tÞ -kS dT
00

dx x¼x
dT
þ
q_ CND ðx ; tÞ -kS dx x¼xþ
000 000
Radiative absorption q_ RAD ðx; tÞ:dx q_ RAD ðx; tÞ:dx
000
X
Chemical energy q_ g ðx; tÞ:dx i¼1 ΔHP, i ρS ðx; tÞ Ai YO ðx; tÞYF ðx; tÞe -
i¼N mi ni
(generation/sink) Ei =RTðx; tÞ
sources and sinks associated to all chemistry ∂ECV h 00 00 00 00

i
¼ q_ S ðxþ ; tÞ þ q_ P ðxþ ; tÞ þ q_ 0 ðx; tÞ þ q_ CND ðx; tÞ
need to be included. Table 21.1 summarizes all ∂t h 00 i
00 00 00
q_ O ðx þ , tÞ þ q_ CND ðx þ , tÞ þ q_ S ðx; tÞ þ q_ P ðx; tÞ þ
terms incorporated in Fig. 21.5. 000 000
q_ RAD ðx; tÞ dx þ q_ g ðx; tÞ dx
Estimation of the net heat transfer will lead to
a change in the energy accumulated within the where ECV ¼ ρS ðx; tÞCS ðx; tÞTðx; tÞ dx, which
control volume. The following expression after appropriate substitutions results in the gen-
summarizes the energy balance: eral energy equation for the control volume.
644 J. Torero
00 00
∂½ρS CS T ∂ ∂T ∂ m_ P CP, P TP _ O CP, O TO
∂ m ∂½ρS VR CS T 000
¼ kS þ þ þ q_ RAD
∂t ∂x ∂x ∂x ∂x ∂x
X h i
mi ni Ei =RT
þ i¼N
i¼1 ΔHP , i ρ S Ai Y Y
O S e ð21:7Þ
Given the differential nature of the equation all Density of the solid ρ(x,t)
variables are assumed to be functions of “x” and Regression rate VR(t)
00
“t” so these dependencies are no longer Mass flow ṁP 00
indicated. Many of the terms are left in a generic ṁO
form and not quantified. Their quantification is Temperature of the gas phase TP
TO
complex, thus a more detailed discussion will be
Radiative properties of the solid αS (x,t)
provided later in those cases where it is (absorptivity, αS(x,t))
necessary. Heat of reaction ΔHP,i
The solution to Equation 21.5 will provide the
evolution of the temperature distribution along
the sample and as a function of time (T(x,t)).
The Surface Boundary Conditions
This solution can then be incorporated in Equa-
(x ¼ 0 and x ¼ L)
tion 21.4 to establish the fuel production rate. It is
important to note that thermal equilibrium
Figure 21.1 shows all the different modes of heat
between phases has not been assumed, thus
transfer through the surface control volumes. In
there are three different temperatures in Equa-
theory, control volumes at x ¼ 0 and x ¼ L
tion 21.5, T, TP and TO. Expressions similar to
could be represented in a generic manner that
Equation 21.5 can be defined for each phase and
makes them identical. In practise this is generally
will have to be solved in a simultaneous manner.
not the case because materials tend to have an
The boundary condition will be the exchange of
exposed face and one that is in contact with some
heat between phases, this is generally done using
backing. The backing will define a conductive
empirical correlations for heat transfer in porous
boundary condition while the open face a con-
media [22]. The alternative approach is to dem-
vective/radiative one. For illustration purposes,
onstrate thermal equilibrium between the phases
this distinction will be made here and the
(heat transfer is much faster than mass transfer
exposed face will be defined as an open bound-
within the pores), in which case all temperatures 00
ary, thus q_ N ð0; tÞ will include convection and
will be the same and only Equation 21.5 will 00
have to be solved. radiation, while the back-face, q_ N ðL; tÞ, will be
To summarize, and in addition to the variables attached to a substrate, thus will be defined as an
established in sections “The Production of Gas- impermeable conductive boundary condition. It
eous Fuel” and “The Pyrolysis (εP) and Charring needs to be emphasized that this is an arbitrary
Depths (εCH)”, the temperature distribution is simplification that is only done to illustrate two
controlled by the following variables: different types of boundary conditions because
they are mutually exclusive. In many cases a
Thermal conductivity kS(x,t) material might be sandwiched between two
Specific heat CS(x,t) solids or exposed at both ends. The appropriate
CP,P(x,t)
choice of boundary conditions needs to be made
CP,O(x,t)
www.ebook777.com
q⋅ P″ (0, t) q⋅ Cv
″ (0, t)
q⋅ O″ (0, t)
q⋅ S″(0, t) q⋅ e″ (0, t) q⋅ SR
″ (0, t)
x=0
q⋅ g″′( x, t) q⋅ RAD
″′ (0, t)
x=ε
q⋅ O″ (ε, t) q⋅ S″(ε, t) q⋅ e″(ε, t)

x q⋅ CND
″ (ε, t)
q⋅ P″ (ε, t)
Fig. 21.6 Boundary control volume for x ¼ 0 showing the main heat transfer mechanisms
but the processes to be described will not be in Table 21.1 leading to a very similar expression
different. for the energy balance as that presented in sec-
Figure 21.6 shows the open boundary condition “The Temperature Distribution”. So at the
tion (x ¼ 0) at a specific point in time. The x ¼ 0 surface
different components are mainly those described
∂ECV ð0; tÞ h 00 00 00
i
¼ q_ S ðε; tÞ þ q_ P ðε; tÞ þ q_ 0 ð0; tÞ
∂t h 00 i
00 00 00 00 00
q_ O ðε; tÞ þ q_ CND ðε; tÞ þ q_ S ð0; tÞ þ q_ P ð0; tÞ þ q_ SR ð0; tÞ þ q_ Cv ð0; tÞ þ
000 000
q_ RAD ðx; tÞ ε þ q_ g ðx; tÞ ε
where the terms that remain undefined are For the boundary control volume the charac-
described in Table 21.2. Radiation absorption teristic thickness ε ! 0, which eliminates all
within the surface control volume is represented energy transported by mass flow, radiation
00 00 00
as q_ RAD ð0; tÞε ¼ q_ e ð0; tÞ-q_ e ðε; tÞ to remain absorption and energy generation. The final
consistent with the notation of the previous expression for the exposed boundary condition
section. is then:

∂T
0 ¼ kS εS ð0; tÞσ T4 ð0; tÞ T40 hCv ðtÞðTð0; tÞ T0 Þ ð21:8Þ
∂x x¼0þ
646 J. Torero
Table 21.2 Summary of all energy transport within the surface control volume. Only terms not presented in Table 21.1
are described here. The Stefan-Boltzmann Constant is: s ¼ 5.670 108 W/m2K4, eS(0,t) is the surface emissivity
and hCv is the convective heat transfer coefficient. Only for illustration purposes two different approaches are used to
describe radiation, absorption is allowed to happen in-depth while emission is treated as a surface process. The spectral
emissivity and absorptivity of the material will define the most appropriate treatment for each specific case
Description In Out Formulation
00
Radiation from the exposed surface to the environment q_ SR ð0; tÞ εS ð0; tÞσ T4 ð0; tÞ-T40 Þ
00
Convective losses from the surface q_ Cv ð0; tÞ hCv ðTð0; tÞ-T0 Þ
00 00
External radiative heat-flux q_ e ð0; tÞ q_ e ð0; tÞ
A similar treatment can be followed with the Global thermal conductivity of the backing kB(x,t)
back end boundary condition (x ¼ L). In this material
case the back surface is assumed to be in direct Temperature of the backing material TB(x,t)
contact with another solid. Mass transfer, con- Emissivity of the solid εS(x,t)
vection and radiative losses to the environment Convective heat transfer coefficient hCv(t)
are therefore precluded. The boundary condition Ambient temperature T0
will only include conductive terms and can be
described as:
The Gas Phase
∂T ∂TB
0 ¼ kS þ kB ð21:9Þ
∂x x¼L ∂x x¼Lþ
The sequence of events leading to the ignition of
where kB is a global thermal conductivity of the a gas phase flame will be described in this sec-
backing material that could include the thermal tion. It will be assumed that gaseous fuel emerges
resistance between the two solids. In most cases from the solid following the description provided
the contact between both solids is not perfect, in section “The Solid Phase”.
leaving air gaps or requiring adhesives, in these After the onset of pyrolysis gas begins to
cases it is important to define the thermal con- emerge from the fuel surface, initially in very
ductivity in a manner that includes the contact small quantities, but as εF and T(x,t) increase
resistance. The variable TB is the temperature of Equation 21.4 shows that the fuel mass flux will
the backing solid, these temperature will come increase. The emerging fuel will encounter the
out of a solution to an additional energy balance ambient oxidizer and eventually produce a flam-
of the form of Equation 21.5. Note that if kB is mable mixture. Given that fuel is migrating into
very small the backing can be assumed as an the oxidizer flow, the definition of a flammable
insulator and the boundary condition can be mixture is not a simple one. In standard test
summarized to no losses at the back. This methods the ambient flow is well defined,
eliminates the need to solve a second energy mixed convection generated by a horizontal
equation for TB. heated surface and the extraction system in the
To summarize, and in addition to the variables cone calorimeter [23], natural convection
established in sections “The Production of Gas- resulting from a vertical heated surface in the
eous Fuel”, “The Pyrolysis (εP) and Charring LIFT apparatus [24] and forced convection over
Depths (εCH)” and “The Temperature Distribu- the fuel surface (horizontal or vertical) in the FM
tion”, the temperature distribution is controlled Global Fire Propagation Apparatus [25]. In real
by the following variables: fires, flow fields are defined by the flames
www.ebook777.com
themselves and by the geometry of the environ- studied using a stagnation point flow over a
ment (obstacles, fuel geometry, etc.) with the solid fuel surface. In these experiments the heat
possibility of complex flow patterns. The only to initiate the combustion reaction is provided by
mechanisms to establish the fuel distribution a hot flow impinging on a fuel surface that acts as
within the gas phase are detailed measurements a heat sink. Niioka [32] identifies an induction
or modelling [26–28]. Nevertheless, from a phe- time and a pyrolysis time. The pyrolysis time
nomenological perspective, to achieve ignition, corresponds to the time required to attain a flam-
what is required is to achieve a flammable condi- mable mixture while the induction time is the
tion in at least one location in the gas phase. time for the mixture to reach a temperature at
The definition of a flammable mixture is for which ignition can occur. Given the specific con-
the fuel concentration to be found between the figuration, the pyrolysis time decreases with the
Lower or Lean Flammability Limit (LFL) and flow velocity (enhanced heat transfer to the fuel
the Upper or Rich Flammability Limit (UFL). surface) while the induction time increases
Although the LFL and UFL are apparatus depen- (reduced residence time in the gas phase).
dent measurements, it is clear that the precision Although these observations are not universally
required for flaming ignition of solids does applicable, they serve to illustrate the process of
not require a more universal description of auto-ignition. Fernandez-Pello [30, 31] describes
flammability. For a more detailed discussion on Niioka’s conclusions graphically by means of the
flammability limits and their limitations the schematic, this schematic is simplified and
reader is referred to Chap. 12. presented in Fig. 21.7. Figure 21.7 shows how
the summation of the pyrolysis and induction
times leads to an ignition time.
Auto-ignition In auto-ignition there is no hot spot that will
serve as an initiation point for the reaction, thus
Once a flammable mixture has been attained, this the mixture has to absorb enough energy to reach
mixture needs to increase in temperature until a ignition. The exact amount of energy required for
combustion reaction can occur. This process is ignition can be associated to a Damköhler num-
described in great detail by Torero [29] and by ber [18]. The Damköhler number corresponds to
Fernandez-Pello [30, 31], who cites a series of the ratio between a local residence and chemical
experiments by Niioka [32] where ignition is time. The chemical time represents the necessary

the characteristic times
Time
involved in the ignition of a
flat plate subject to a hot
stagnation point flow. This
schematic is based on the
work by Niioka [32] and
adapted from Fernandez-
Pello [30, 31]
Ignition
Time
Induction
Time
Pyrolysis
Time
Flow Velocity
648 J. Torero
time for the reaction chemistry to occur and is surface temperature will change following Equa-
expressed as the inverse of the reaction rate. tion 21.5 while the gas will absorb heat based on
Combustion reactions can be described by its absorptivity and dissipate it in a manner
expressions like that presented in Equation 21.1 governed by the flow field. The absorptivity of
thus the chemical time is directly affected by the the gas is a strong function of the fuel type and
temperature of the reactants. The higher the tem- concentration, thus also requires detailed knowl-
perature, the greater the reaction rates and the edge of the flow field. The two possible outcomes
shorter the chemical time. The residence time is a are that the gas phase heats faster than the solid
measure of the strain (or dissipation rates) or the phase or the opposite. In the former case ignition
time the reactants remain together at a specific will occur away from the fuel surface, since the
location thus it is directly related to the velocity fuel will act as a heat sink for the gas. In the latter
field. The faster the flow or the velocity case, ignition will occur closer to the fuel surface
gradients, the shorter the residence time. If the since the fuel acts as a heat source. This latter
chemical times are shorter than the residence scenario is common with charring materials
times the reaction has enough time to proceed where oxidation of the char contributes to
and a flame can exist. A critical Damköhler num- increase the surface temperature [37].
ber for ignition can then be established, above It is clear that auto-ignition is a complex pro-
which a combustion reaction can proceed [18]. cess that fully involves interactions of the solid
In the schematic presented in Fig. 21.7, critical and gas phases. Therefore, to characterize
Damköhler numbers will be attained at both sides auto-ignition of solid fuels it is necessary to
of the ignition curve preventing ignition. This is established well defined experimental conditions
probably the most precise way to describe igni- and simplifications to the analysis. Data obtained
tion but it requires the full resolution of the flow from different experimental conditions and with a
and temperature fields as well as comprehensive specific analysis will generally not be compatible
knowledge of the kinetic constants associated to with other data that was obtained from a different
the combustion reaction. While the flow field can experiment or deduced by means of an alternative
be resolved by means of Computational Fluid analysis. Thus, scatter in the reported data is
Dynamics (CFD) the chemistry of most fire common for auto ignition.
related fuels still remains uncertain. Qualitative Data on auto-ignition is generally reported as
assessment of the Damköhler number for ignition Auto-Ignition Temperatures (AIT) which
has only been achieved for a few very well corresponds to a recorded temperature at the
defined experimental conditions such as stagna- moment where ignition of a flame is first
tion flows [6, 32, 33] or boundary layers observed. A summary of much of the data avail-
[34]. Other alternative representations of the able is presented in Chap. 14 of Ref. [1] together
ignition conditions that rest on the same funda- with a series of references to relevant papers and
mental approach have been discussed by textbooks [38, 39]. Given the complexity of the
Quintiere [35] and by Gray and Lee [36]. processes leading to auto-ignition, these values
An important aspect of the ignition process can only be taken as reference values that are a
that remains to some extent unresolved is the direct function of the specific test conditions.
origin of the heat that is necessary for the gaseous Generally, significant discrepancy is found in
fuel to reach the critical Damköhler number. If the literature where reported Auto-Ignition
the air flow is hot, like in Niioka’s experiments Temperatures can vary in more than 150 C for
[32], then the energy will come from the oxidizer the same material. The greatest discrepancies
and the problem is immensely simplified. If the tend to be found when the orientation of the
oxidizer is cold and there is an external radiative solid fuel is varied and the fluid mechanics and
heat source, then solid and gas will heat at differ- heat transfer are significantly altered [1]. Auto-
ent rates. The solid will absorb heat and its Ignition Temperatures are most consistent for
www.ebook777.com
gaseous mixtures (Chap. 12) and liquid fuels To attain the LFL at the pilot location it is
(Chap. 18) where tests are conducted in enclosed necessary to resolve the momentum and mass
vessels where the fuel has been fully evaporated. transport equations simultaneously with the sur-
face boundary conditions explained above.
Figure 21.1 shows an arbitrary distribution of
Piloted Ignition the fuel concentration external to the sample,
YF,g. A similar representation could be made
As discussed in the previous section, the process for the oxygen concentration (YO,g). The charac-
of auto-ignition is extremely difficult to describe teristic equation that describes the flow field is as
in a quantitative manner, even under simple follows:
experimental configurations. Therefore, as an !
example, it is not practical to rely on auto- Du ! !
ρ0 ¼ ∇P þ ρ0 g þ μ0 ∇2 u ð21:10Þ
ignition to describe the susceptibility of solid Dt
materials to ignite. A mechanism to simplify !
the process is to include a pilot flame or a hot Where u is the velocity field, ρ0 the density of the
!
spot. This is a practical experimental simplifica- air, P the pressure field, g the gravity vector and
tion that has a basis on reality, since in most μ0 the viscosity of the air. Temperature
ignition scenarios there will be a region of high dependencies of the properties have been omitted
temperature. The presence of a pilot strongly for simplification assuming that air is the main
simplifies the gas phase processes and reduces constituent and it will remain close to ambient
the influence of environmental variables. While temperature. Conservation of fuel and oxygen
characterization of the flow field is still required concentrations can then be defined by:
to establish the presence of a flammable mixture,
DYF, g
it is no longer necessary to resolve heat transfer ρ0 ¼ ρ0 DF, O ∇2 YF, g ð21:11Þ
Dt
between phases or to define the absorption of
energy by the gas. In the presence of a pilot, DYO, g
ρ0 ¼ ρ0 DF, O ∇2 YO, g ð21:12Þ
ignition can be assumed at the moment where a Dt
flammable mixture (LFL) is attained at the loca-
where species transport is assumed to be
tion of the pilot.
non-reactive, thus the source/sink has been omit-
Currently, all standard test methods that
ted. This is an adequate assumption for pure
attempt the description of the ignitability of
mixing. To obtain the solution of Equations 21.8,
solids use some form of a pilot. In some cases,
21.9 and 21.10 it is necessary to add the follow-
the pilot is a large flame [24] while in others is
ing variables to those established in sections
either a small pilot flame [25] or a high energy
“The Production of Gaseous Fuel”, “The Pyroly-
spark [23]. Both methods have their advantages
sis (εP) and Charring Depths (εCH)”, “The Tem-
and disadvantages, sparks produce only local
perature Distribution” and “The Surface
heating thus have a weaker tendency to influence
Boundary Conditions (x¼0 and x¼L)”:
the solid phase by acting as a heat source. -
Nevertheless, given their small volume,
Density of air ρ0
ignition is strongly influenced by the spark loca- !
Velocity field u
tion. The flow field has to establish a flammable
mixture at exactly the location of the pilot. In Pressure field P
contrast, large pilot flames have a tendency to Viscosity of air μ0
Diffusivity of fuel in air DF,0
supply heat to the fuel surface, but cover a large !
Pilot location r
volume, therefore are less sensitive to the flow
field. Because of its practical relevance, all
subsequent discussion will concern piloted At this point, there is no need to specify a
ignition. critical Damköhler number for ignition because
650 J. Torero
of the presence of the pilot, although in absolute number appears, but this time is one of extinc-
rigour, this assumes that the flow conditions are tion. This concept has been described many
such that blow-off of the flame kernel does not times explicitly in the combustion literature
occur, thus the pilot will allow the establishment [18] but only implicitly in the fire literature.
of a flame across the flammable mixture. There are only few studies where a critical
extinction Damköhler number has been
presented to describe the “fire point” but in all
“Flash Point” and “Fire Point” cases they concern idealized flow fields that
allow establishing a direct correlation between
Once ignition has been achieved a flame can fuel production and flame temperature [33,
propagate through the regions where a flammable 34]. In most discussions simplifications have
mixture is present consuming the reactants. Inde- been assumed leading to simpler parameters
pendent of the flow field, it is most likely that a that can serve as surrogates for the Damköhler
flammable mixture will be established close to number. Williams [40] discusses a critical gas
the solid fuel surface. The pyrolysis rates at the phase temperature below which extinction will
moment when the flame is established will deter- occur. If the residence time remains unchanged,
mine if a flame can continue to exist or if the then extinction is only associated to the chemical
combustion reaction will cease after the gas time, thus can be directly linked to a critical gas
phase mixture is consumed. The feedback from phase temperature. It can be further argued that
the flame will enhance pyrolysis, but usually, the extinction is much more sensitive to temperature
relatively large thermal inertia of the solid will than to flow, thus only radical changes in the
result in a slow response, therefore it will be residence time need to be addressed making this
necessary for pyrolysis rates to be sufficient criterion a robust one. A more practical surrogate
even in the absence of the flame heat feedback. to the Damköhler number is a critical fuel mass
If pyrolysis rates are not sufficient, the flame will flux criterion. Under specific testing conditions
extinguish and continuous pyrolysis will lead the flow field will remain invariable. In this case
once again to the formation of a flammable mix- the attainment of a critical mass flux of fuel will
ture and subsequent ignition. This manifests be the single parameter defining the flame
itself as a sequence of flashes that precede the temperature and thus the Damköhler number
establishment of a flame over the combustible [41, 42]. Furthermore, under more restrictive
solid. This process is identical to the “flash conditions the critical mass flux can be
point” generally associated to liquid fuels associated to a critical solid phase temperature
(Chap. 18) and for solid fuels has been described [43]. Drysdale [17] and Beyler [44] provide a
in detail by Atreya [37]. detail description of the classic approaches to
The transition between the “flash point” igni- this subject while Quintiere and Rangwala
tion and the established flame, which could also address some of the more current studies [45].
be named the “fire point” in an analogy with The sequence of events relating “flash” and
liquid fuels, deserves especial attention. The “fire” points is not trivial because they represent
characteristics of the diffusion flame established distinctively different processes. For piloted igni-
on a solid fuel surface are defined by the flow tion, the “flash point” only requires a flammable
field and the supply of fuel. The rate at which mixture while for the “fire point” the rate of fuel
both reactants reach the flame zone defines the supply has to be enough to achieve a chemical
flame temperature and thus the characteristic time shorter than the residence time. Thus a
chemical time. If the amount of fuel reaching number of different scenarios can be observed
the flame is small, then the flame temperature that in many cases can affect the consistency of
will be low and the chemical time will be long. different ignition studies. A simple example will
As described above, the flow field defines the be used to illustrate this. For example; if the pilot
residence time. A second critical Damköhler is very close to the fuel surface then a flammable
www.ebook777.com
mixture will be achieved at the pilot location number of authors have attempted the solution
soon after the onset of pyrolysis. In this case to these equations for a number of materials.
fuel supply will be far from that required to Furthermore, they have in some cases added
sustain a flame. A significant delay will exist further complexity by including phenomena
between flash and fire points where several such as intumescent behaviour [49] or
flashes will be observed. If the pilot is distanced bubbling [50]. Extensive reviews of these
from the fuel surface it will take longer to attain a modelling efforts can be found in Refs. [4,
flammable mixture and therefore at the moment 51–54] and some of the more recent modelling
of the first flash the fuel supply would have exercises have achieved significant success
increased and a smaller number of flashes will [55–59]. In most cases some simplifications
be observed before the flame is fully established. have been necessary and in general the critical
Greater separation of the pilot from the fuel sur- limitation of these models is associated to the
face might result in the flammable mixture being inadequate definition of many of the relevant
attained at the pilot location at the same time as variables and parameters listed in the previous
the fuel supply is sufficient to sustain a flame. sections. As mentioned before, the current
In this case the fire point will correspond with the trend is to optimise parameters by fitting com-
first flash. A further increase in the distance plex models to specific experimental results by
between pilot and fuel will not change the physi- means of sophisticated optimization
cal manifestation but will continue to delay igni- techniques. The optimization process results
tion. In this case ignition will occur when a in ranges of possible values for all parameters
flammable mixture is attained at the pilot but stipulated. The results have then been
will not be related to the flash or fire points. extrapolated to other experimental conditions.
This example has been presented to illustrate While success has been reported [6, 7], these
the sensitivity of ignition studies to different optimization processes are only as good as the
variables and the importance of detailed models whose parameters they optimize. It is
observations to the validity of conclusions and therefore important to note that even in the
comparisons. In this case pilot location was used most complex models some simplifying
as the example, but a similar analysis could be assumptions have been made. Currently, the
made with the heat flux, the oxygen concentra- use of such models remains a research subject
tion, the flow field [31, 46] or the ambient with increasing applicability to the modelling
pressure [47]. of flaming ignition of solid fuels.
The only added variable required to model the This section will take the equations presented
“fire point” will be the critical Damköhler num- in previous sections and suggest simplifications
ber for extinction (Dae,cr) or any equivalent way that will lead to models commonly used in the
to represent the extinction condition. As men- analysis of standard test methods evaluating the
tioned above, other criteria can be used to estab- flaming ignition of solid fuels.
lish the extinction condition and that are partially
equivalent to the critical Damköhler number.
Such criteria are a critical mass transfer numbers The Inert Solid Assumption
(Bcr) [34, 48], critical mass fluxes [11, 28, 30, 42]
or critical temperatures (Tcr) [17, 35, 40, 43, 45]. The assumption that the solid remain inert until
ignition is probably the most far reaching of all
proposed simplifications. As a result of this
Simplifications and Standardization assumption the energy equation is dramatically
simplified. Despite the far reaching implications
To predict flaming ignition of a solid fuel is of assuming that the solid remains inert until
necessary to solve Equations 21.1, 21.2, 21.3, ignition there is very limited work that assesses
21.4, 21.5, 21.6, 21.7, 21.8, 21.9, and 21.10. A the validity of this assumption.
652 J. Torero
Fig. 21.8 Characteristic 800

ignition delay times (tig)
700 Pyrolysis Time
and times to the onset of
pyrolysis (tP) for PMMA Ignition Time
and a wide range of 600
external heat fluxes
500
extracted from Ref.
[60]. Onset of pyrolysis or t [s] 400
ignition did not occur
below 11 kW/m2 300
200
100
0
0 10 20 30 40 50 60
q⋅ e″ [kW/m2]
To the knowledge of the author, the only onset of pyrolysis (could be up to 100 %) there-
explicit studies that discuss the importance of fore the assumption that the fuel remains inert
assuming that the material is inert are those by until ignition might not be justified.
Fereres et al. [42], Dakka et al. [60] and Beaulieu The breakdown of the inert solid heating
and Dembsey [61]. In the first two studies trans- assumption is further discussed by Beaulieu and
parent Poly(methyl methacrylate) (PMMA) was Dembsey [61] who show that an analysis follow-
used while on the latter work the detailed analy- ing this approximation will lead to shorter igni-
sis is done with black PMMA but a number of tion delay times for realistic heat fluxes. The
other materials serve to confirm the conclusions. biggest errors were observed at the higher heat
Despite the bias towards PMMA, the discussion fluxes. Their tests were done for a comprehensive
is appropriate here to illustrate the potential array of materials and with heat fluxes up to
errors associated to this simplification. 200 kW/m2.
Figure 21.8 presents characteristic ignition Despite these experimental results, this
delay times (tig) and pyrolysis delay times (tP) assumption still remains the backbone of all stan-
for PMMA. The ignition delay time was recorded dard test method analyses for ignition [23–25].
as the first flash while the pyrolysis delay time as If this approach is followed and the regression
the moment when the fuel initiates its endother- rate is assumed to be negligible, VR 0, Equa-
mic degradation. The onset of pyrolysis was tion 21.5 is reduced to
characterized by means of mass loss
∂½ρS CS T ∂ ∂T 000
measurements, flow visualization and ¼ kS þ q_ RAD ð21:13Þ
IR-Thermography. These results show that for ∂t ∂x ∂x
these particular experiments there is a significant
And the boundary conditions to
difference between the “flash point” and the

∂T
x¼0 0 ¼ kS εS ð0; tÞσ T4 ð0; tÞ T40 hCv ðtÞðTð0; tÞ T0 Þ ð21:14Þ
∂x x¼0þ
www.ebook777.com

∂T ∂TB solid can be considered invariant (ρS ðx; tÞ ρS ,
x¼L 0 ¼ kS þ kB
∂x x¼L ∂x x¼Lþ CS ðx; tÞ CS , and kS ðx; tÞ kS ). These
ð21:15Þ assumptions further simplify Equation 21.11
because it allows neglecting in-depth radiative
absorption. The thermal properties can then be
Absorption of Radiation and Global extracted from the differential terms and external
Properties radiation now appears in the exposed boundary
condition:
The next major simplifications that are com- 2
∂½T ∂ T
monly accepted are to assume that most of the ρS CS ¼ kS ð21:16Þ
∂t ∂x2
incident heat flux is absorbed at the surface
(α(t) 1) and that the thermal properties of the

∂T 00
x¼0 0 ¼ kS þ q_ e σ T4 ð0; tÞ T40 hCv ðtÞðTð0; tÞ T0 Þ ð21:17Þ
∂x x¼0þ

∂T ∂TB
x¼L 0 ¼ kS þ kB data. An example of a comprehensive assessment
∂x x¼L ∂x x¼Lþ of the impact of variable thermal properties is
ð21:18Þ provided by Steinhaus [66].
There is little true justification in the literature to

support these assumptions, nevertheless they are The Boundary Conditions
of practical use since for many fire related
materials the absorptivity (or emissivity) will To standardize the ignition process it is important
approach unity [62], or in the case of testing the to provide a controlled environment, so that test
material surface can be treated with a coating that results can be consistent between laboratories
has these properties [25]. and different users of the standard. Therefore,
A series of recent studies have explored the standard test methods provide clear definition
absorptivity [63] of PMMA and the interaction of the environmental conditions, thermal
between the heat source and PMMA [64, 65]. characteristics of the backing material and pilot
Figure 21.9 shows that when using an electrical location [23–25]. Equations 21.8, 21.9 and 21.10
resistance (cone heater [23]) the transmissivity of do not have to be solved to obtain the fuel con-
PMMA can be neglected and the absorption can centration at the pilot location. Instead the impact
be assumed to occur at the surface. Instead, when of the gas phase on the results is ignored. This is
using tungsten lamps (from the Fire Propagation done on the basis that flow conditions are the
Apparatus [25]) in-depth absorption cannot be same between tests thus their impact on the trans-
neglected. This information allowed explaining port of fuel and oxidizer to the pilot is the same.
significant differences in the piloted ignition Standardization of the flow conditions has a
delay times obtained with both tests but mostly deep effect on the meaning of the results. The
emphasize the potential importance of assuming thermal properties associated to the analysis are
an absorptivity of unity. no longer true thermal properties of the material
Furthermore, thermal properties vary with but global properties that are a combination of
temperature, but a global set of properties can the solid and the standardized gas phase
be established to provide a good fit to ignition
654 J. Torero
Fig. 21.9 Transmitted to incident heat flux ratio for clear providing 10 and 20 kW/m2 for thicknesses ranging
PMMA samples (Lucite and Plexiglass) exposed to a between 0.375 and 51 mm [65]
radiative source (conical resistance and tungsten lamp)
conditions. This is of critical importance, Different test methods will use different flow
because, as a product of standardization, test fields therefore values for the convective heat
results can be compared among themselves transfer coefficient vary with the authors. A com-
(if the same method is used), nevertheless can monly cited value is 15 W/m2K. Furthermore, it
not be extrapolated to conditions different to is common to linearize surface radiation to define
those of the test. This applies to other standard a single total heat transfer coefficient (hT 45
tests or to real fire conditions. Cordova et al. [46] W/m2K). More precise values and models are
provides a graphical assessment of the effect of present in the literature [26–28, 31] but they
varying the flow conditions on the resulting ther- correspond to very specific scenarios and there-
mal properties showing that small variations in fore are hard to generalize.
the flow field can result in drastic variations of Most test methods define the backing material
the resultant thermal properties. as a good insulator ( kB ! 0 ) neglecting heat
It is common to apply ignition data from stan- losses through the back end of the sample.
dard tests to fire models and is only recently that Finally, characteristic ignition delay times can
CFD models such as the Fire Dynamics Simula- be considered much shorter than the time
tor (Version 5 and above) allow realistic required for the thermal wave to travel through
representations of the solid phase that include the sample therefore L > εT and the solid is
true thermal properties [67]. It is important to generally assumed as semi-infinite.
note that extrapolation is not necessarily incor- These last set of simplifications are truly not
rect. Nevertheless, it has to be done with great necessary because a simple numerical solution
care to guarantee that the effect of the environ- can be obtained without linearizing surface radi-
ment on the thermal properties can either be ation or assuming a semi-infinite solid. Many
neglected or an appropriate correction is studies have attempted to establish the impact
provided. of these simplifications by means of numerical
www.ebook777.com
solutions that relax these assumptions, the most occur at the onset of pyrolysis and that these
recent of these papers is by Mowrer [68]. If sur- process can be simply characterized by the
face radiation is described by means of constant attainment of a characteristic surface temperature
heat transfer coefficient, then a correction is nec- that is commonly labelled the ignition tempera-
essary to account for the growth of this coeffi- ture, Tig. If the sample is suddenly exposed to an
cient as the surface temperature increases. external heat flux, then the time delay between
Mowrer [68] showed that a correction to the exposure and ignition is named the ignition delay
global thermal properties could be made to time, tig. These two parameters represent then the
account for this effect. The back end boundary entire process of ignition.
condition is a more difficult problem. For low A final link can be made to establish a critical
heat fluxes the thermal wave reaches the end of ignition condition. If the ignition delay time is
the sample, L < εT, before ignition occurs and infinitely long, then there will be no gradients of
heat is exchanged between the sample and the temperature within the solid and surface heat
insulating material. Quantification of this heat losses will be equivalent to the heat input. This
exchange can be done numerically, as indicated represents the minimum heat flux required to
in section “The Surface Boundary Conditions achieve Tig, and thus flaming ignition of the
(x¼0 and x¼L)”, but this is not a simple process solid fuel. This heat flux is named the minimum
00
because it needs to properly describe the differ- heat flux for ignition, q_ 0, ig . Since surface
ent components associated to the way the sample temperatures are more difficult to measure than
is arranged during tests. The alternative solution heat fluxes, the minimum heat flux for ignition
of providing a well defined insulating boundary can be used to establish the ignition temperature.
and neglecting back end losses has been pre- Equation 21.18 can then be re-written to
ferred and detailed analyses have been conducted
00
to characterize the physical arrangements of q_ 0, ig
sample and insulating material. Among the Tig ¼ T0 þ ð21:22Þ
hT
most comprehensive of these studies is presented
in Ref. [69]. Equation 21.14 is an idealized expression that
If all these assumptions are made, Equa- assumes that no temperature gradients exist in
tions 21.12, 21.17, and 21.18 can be reduced to: the solid, this can lead to errors in the calculation
2 of Tig. To establish a relationship between exter-
∂T ∂ T nal heat fluxes and surface temperature that
ρS CS ¼ kS ð21:19Þ
∂t ∂x2 includes in-depth heat transfer a sample can be
allowed to reach thermal equilibrium and the
∂T
x ¼ 0 0 ¼ kS
00
þ q_ e hT ðTð0; tÞ T0 Þ surface temperature recorded. The obtained
∂x x¼0þ
relationship represents a more accurate represen-
ð21:20Þ tation of Equation 21.14 and can be used to
extract ignition temperatures from measured
∂T
x!1 0 ¼ kS ð21:21Þ heat fluxes. A graphic representation of this rela-
∂x x¼L tionship can be found in Ref. [34].
Again, both minimum heat flux for ignition
The Ignition Condition and ignition temperature are not material
properties but a combination of the material and
If the solid is assumed to be inert until ignition the specific environmental conditions associated
and the gas phase can be summarized into a to the test [46]. Extrapolation to realistic
single total heat transfer coefficient (hT) this scenarios and fire models has to be done with
amounts to the assumption that ignition will significant care.
656 J. Torero
The Solution the flaming ignition of a solid fuel by Quintiere

[70] and incorporated in ASTM E-1321
00
Imposing a constant external heat flux ðq_ e ¼ [24]. Alternate solutions have been postulated
constantÞ and using all the above assumptions for other test methods and will be briefly
allows for an analytical solution to Equa- discussed in Chaps. 28 and 36. More detailed
tion 21.13. This solution establishes the evolu- discussion of methodologies and nomenclature
tion of the solid temperature as a function of can be found in the description of the standard
time. This solution can be found in any heat tests [23, 25].
transfer book [19] but was first postulated for The solution for T(x,t) is given by
2 0 13
00
pðhT Þ
ffiffiffiffiffiffiffiffiffi 2
q_ e 6 x pffiffiffiffi ð hT Þ B ð hT Þ 1 x C7
Tðx; tÞ T0 ¼ 4erfc pffiffiffiffiffiffiffiffiffiffi e αD kS ρ S C S
xþ terfc@qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffit2 þ pffiffiffiffiffiffiffiffiffiffiA5
ð hT Þ 4α Dt kS ρ S C S k ρ C 4α Dt
S S S
ð21:23Þ
"
12 !#
Where αD ¼ kS =ρS CS is the global thermal diffu- tig tig
Tig ¼ T0 þ T 1 e tc erfc ð21:27Þ
sivity and “erfc” is the complement to the error tc
function. To obtain the surface temperature (Ts),
x is set equal to 0 and T ¼ T(0,t) ¼ Ts. There- To avoid the complex form of the error function
fore Equation 21.21 simplifies to: simplified solutions have been proposed in the
2 2 0 13 literature [70, 71]. In order to solve for the ignition
ðhT Þ delay time (tig) a first order Taylor series expansion
q_ 6 B ð hT Þ 1 C 7
00
t
Ts ¼ T0 þ e 41 e kS ρS CS erfc@qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffit2 A5 of Equation 21.18 is conducted. The range of
ðh T Þ
kS ρS CS validity of this expansion is limited, thus cannot
be used over a large range of incident heat fluxes.
ð21:24Þ
Thus, the domain has to be divided at least in two.
from Equation 21.15, The first domain corresponds to high incident
00
heat fluxes where the ignition temperature (Tig)
q_ e is attained very fast, thus tig < < tc. Application
T¼ ð21:25Þ
ðhT Þ of the first order Taylor Series Expansion to
Equation 21.18 around tig =tc ! 0 yields the
can be defined as a characteristic temperature
following formulation for the ignition delay
and,
time (tig):
k S ρS C S 00
tc ¼ ð21:26Þ 1 2 q_ e
ð hT Þ2 pffiffiffiffiffi ¼ pffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð21:28Þ
π kS ρS CS Tig T0
tig
is defined as a characteristic time. Equation 21.15
is the general solution to the surface temperature As can be seen from Equation 21.19, the short
at all levels of incident heat flux. To obtain the time solution for the ignition delay time (tig) is
ignition delay time (tig) the surface temperature independent of the total heat transfer coefficient
(Ts) is substituted by Tig and Equation 21.15 can term (hT). Thus the ignition delay time (tig) is
be rewritten as:
www.ebook777.com
4.5
4
Equation (29)
3.5
3
[1/s ]
1/2
2.5 Equation (28)

1/2
2
1/t
1.5
0.5
0 ⋅″
0 5 10 q0ig 15 20 25 30 35 40 45
⋅q ″ [kW/m2]
e
Fig. 21.10 Ignition delay time (1/tig0.5) for different external heat fluxes using PMMA as a solid fuel (Data extracted
from Ref. [60])
00
only a function of the external heat flux (q_ e ) and When describing ignition propensity of solid
the global properties (kS , ρS , CS) of the solid fuel fuels is customary to summarize the description
and the ignition temperature (Tig). of the materials on the basis of only two
For low incident heat fluxes tig tc , the parameters, the ignition temperature, Tig, and
Taylor series expansion is made around the thermal inertia, kS ρS CS . Several tables have
tig =tc ! 1, where the first order approximation been produced in the past with comprehensive
yields: lists of materials typical of fires. As an example,
" Table 21.3 presents the data as compiled by
pffiffiffi #
1 π hT hT T ig T 1 Quintiere [70]. A comprehensive list is not
pffiffiffiffiffi ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 ð21:29Þ
tig
kS ρS CS q_ }e presented here because a comprehensive compi-
lation of data is provided in Appendix 3 or in
Refs. [1] and [2].
Equations 21.19 and 21.20 establish the relation-
ship between ignition delay time and external
heat flux. It is convenient to express the ignition
pffiffiffiffiffi Thermally Thin Materials
delay time data presented in Fig. 21.7 as 1= tig
where Tig is obtained from the experimental min-
A very similar analysis can be conducted for
imum heat flux for ignition and Equation 21.14.
thermally thin materials where in the absence of
Such a plot is presented in Fig. 21.8. Substituting
thermal gradients and after all relevant
Tig in Equation 21.19 allows extracting the prod-
simplifications Equations 21.13, 21.20, and
uct of the three thermal properties (kS ρS CS ) as a 21.21 can be combined into a single energy
single experimental parameter representing the equation and a boundary condition
global material properties controlling flaming
ignition of solid fuels that can be considered ∂T 00
ρS C S L ¼ q_ e hT ðTðtÞ T0 Þ ð21:30Þ
semi-infinite. Quintiere terms this product the ∂t
thermal inertia [70] (Fig. 21.10).
658 J. Torero
Table 21.3 Ignition data from ASTM E-1321 per applied. Analytical solutions are obtained
Quintiere [72] showing that the description of the ignition pro-
kS ρS CS cess can be summarized to two material related
Material Tig [ C] [(kW/m2K)2.s] parameters and two specified environmental
00
Wood fiber board 355 0.46 conditions (T0, q_ e ). The material related
Wood hardboard 365 0.88
parameters are as follows:
Plywood 390 0.54
PMMA 380 1.00 Thermally thin materials Tig ρS CS
Flexible foam plastic 390 0.32 Thermally thick materials Tig kS ρ S C S
Rigid foam plastic 435 0.03
Acrylic carpet 300 0.42 It is important to insist that these parameters
Wallpaper on plasterboard 412 0.57 are a function of the material and the environ-
Asphalt shingle 378 0.70 mental conditions at which they were obtained.
Glass reinforced plastic 390 0.32 They can be directly applied for comparison
between materials (flammability assessment)
00 but extrapolation to conditions beyond the tests
x¼L q_ N ðL; tÞ ¼ 0 ð21:31Þ where they were obtained is not always possible
and if performed, has to be done with great care.
when the external heat flux is much larger than
the losses this equation can be integrated to
deliver Equation 21.22 [73]. References
1. Babrauskas, V., “Ignition Handbook,” Fire Science
ρS CS L Tig T0
tig ¼ 00 ð21:32Þ Publishers & Society of Fire Protection Engineers,
q_ e 2003.
2. Engineering Guide: Piloted Ignition of Solid
This is once again not a necessary assumption but Materials Under radiant Exposure, Society of Fire
Protection Engineers, Bethesda, Maryland, USA,
has the practical advantage of leaving the product 2002.
ρS CS as a single experimental parameter that can 3. Hirata, T., Kashiwagi, T. and Brown, J.E., “Thermal
be extracted from the slope of a simple plot and oxidative degradation of Poly (methyl methacry-
00 late): Wight loss,” Macromolecules, 18, 1410–1418,
presenting 1/tig vs q_ e . ρS CS represents then the 1985.
global material properties controlling flaming 4. Di Blasi, C., “Modeling and Simulation of Combus-
ignition of thermally thin solid fuels. A compre- tion Processes of Charring and Non-Charring Solid
Fuels,” Progress in Energy and Combustion Science,
hensive data review of this product is provided in 19, 71–104, 1993.
Refs. [1] and [2]. 5. Ohlemiller, T.J., “Modeling of Smoldering Combus-
tion Propagation,” Progress in Energy and Combus-
tion Science, 11, 277–310, 1986.
6. Rein, G., Lautenberger, C., Fernandez-Pello, A.C.,
Summary Torero, J.L. & Urban, D.L., “Application of Genetic
Algorithms and Thermogravimetry to Determine the
A review of flaming ignition of solid fuels has Kinetics of Polyurethane Foam in Smoldering Com-
been presented. Emphasis has been given to a bustion,” Combustion and Flame 146 95–108 (2006).
7. Lautenberger, C., Rein, G. & Fernandez-Pello, A.C.,
comprehensive description of all processes “The Application of a Genetic Algorithm to Estimate
involved. Some minor simplifications have been Material Properties for Fire Modeling from Bench-
made to the original formulation leading to Scale Fire Test Data,” Fire Safety Journal 41 204–214
approximately 30 variables and parameters (2006).
8. Bal, N., “Uncertainty and complexity in pyrolysis
controlling flaming ignition of a solid fuel. modelling,” PhD Dissertation, University of
A section follows where the common Edinburgh, 2012.
simplifications associated to the methodologies 9. Bal, N. and Rein, G., “Uncertainty and Calibration in
of interpretation of standard test methods are Polymer Pyrolysis Modelling,” Recent Advances in
www.ebook777.com
Flame Retardancy of Polymeric materials, vol. 25. ASTM E-2058-03, “Standard Method of Test for
23, C. Wilke (Editor), BCC, May 2012. Measurement of Synthetic Polymer Material Flamma-
10. Chao, Y.H. and Wang, J.H., “Comparison of the bility Using the Fire propagation Apparatus (FPA),”
Thermal Decomposition Behavior of a Non-Fire American Society for Testing and Materials,
Retarded and a Fire Retarded Flexible Polyurethane Philadelphia, 2003.
Foam,” Journal of Fire Science, 19, pp. 137–155, 26. Staggs, J.E.J., “Convection heat transfer in the cone
2001. calorimeter,” Fire Safety Journal, 44, 469–474, 2009.
11. Lautenberger C. and Fernandez-Pello, A.C., “Optimi- 27. Staggs, J.E.J., “A reappraisal of convection heat trans-
zation algorithms for material pyrolysis property esti- fer in the cone calorimeter,” Fire Safety Journal,
mation,” Fire Safety Science, 10, 751–764, 2011. 46, 125–131, 2011.
12. Chaos, M. Khan, M.M., Krishnamoorthy, N., De Ris, 28. Zhang, J. and Delichatsios, M.A., “Determination of
J.L. and Dorofeev, S.B. “Evaluation of optimization the convective heat transfer coefficient in three-
schemes and determination of solid fuel properties for dimensional inverse heat conduction problems,” Fire
CFD fire models using bench-scale pyrolysis tests,” Safety Journal, 44, 681–690, 2009.
Proceedings of the Combustion Institute, 29. Torero, J.L. “Scaling-Up Fire,” Proceedings of the
33, 2599–2606, 2011. Combustion Institute, 34 (1), 99–124, 2013.
13. Bruns, M.C., Koo, J.H. and Ezekoye, O.A., “Popula- 30. Fernandez-Pello, A.C., “The Solid Phase,” In Com-
tion-based models of thermoplastic degradation: bustion Fundamentals of Fire, Ed. G. Cox, Academic
Using optimization to determine model parameters,” Press, New York, pp. 31–100, 1995.
Polymer degradation and stability, 94, 1013–1022, 31. Fernandez-Pello, A.C. “On fire ignition,” Fire Safety
2009. Science, 10, 25–42, 2011.
14. Lyon, R.E., Safronava, N. and Oztekin, E., “A simple 32. Niioka, T., Takahashi, M., Izumikawa, M., 1981,
method for determining kinetic parameters for “Gas-phase ignition of a solid fuel in a hot stagnation
materials in fire models,” Fire Safety Science, point flow”, 18th Symposium on Combustion, The
10, 765–777, 2011. Combustion Institute, Pittsburgh, PA, pp. 741–747.
15. Kashiwagi, T. and Nambu, H., “Global Kinetics 33. Delichatsios M A and Delichatsios M M, “Critical
constants for thermal oxidative degradation of a cel- Mass Pyrolysis rates for Extinction of Fires over
lulosic paper,” Combustion and Flame, 88, 345–368, solid Materials” Fifth Symposium on Fire Safety Sci-
1992. ence, 153–164, 1996.
16. Cullis, C.F. and Hirschler, M.M., “The Combustion of 34. Torero, J.L., Vietoris, T., Legros, G., Joulain,
Organic Polymers,” International Series of P. “Estimation of a Total Mass Transfer Number
Monographs in Chemistry, Oxford Science from Stand-off Distance of a Spreading Flame,” Com-
Publications, Oxford, United Kingdom, 1981. bustion Science and Technology, 174 (11–12),
17. Drysdale, D., An Introduction to Fire Dynamics. Sec- pp. 187-203, 2002.
ond Edition. John Wiley and Sons, New York, 1999. 35. Quintiere, J.G., “Fundamentals of Fire Phenomena,”
18. Williams, F.A., Combustion Theory, 2nd Edition, John Wiley and Sons, 2006.
Addison-Wesley Publishing Company, Inc., 1985. 36. Gray, P. and Lee, P. R. “Thermal Explosion Theory,”
19. Incropera, F.P., Dewitt, D.P., Bergman, T.L., Lavine, Oxidation and Combustion Reviews, 2, 3–180, 1967.
A.S., Fundamentals of Heat and Mass Transfer, 6th 37. Atreya, A., “Ignition of Fires,” Philosophical
Edition, John Wiley and Sons, 2006. Transactions of the Royal Society A: Mathematical,
20. Oztekin, E.S., Crowley, S.B., Lyon, R.E., Stoliarov, S. Physical, and Engineering Sciences 356 2787–2813
I., Patel, P. and Hull, T.R., Sources of variability in (1998).
fire test data: a case study on poly(aryl ether ether 38. Horrocks, A.R., Gawande, S., Kandola, B. and Dunn,
ketone) (PEEK), Combustion and Flame, K. W., “Recent Advances in Flame Retardancy of
159, 1720–1731, 2012. Polymeric Materials,” Business Communications
21. Stoliarov, S.I., Safronava, N. and Lyon, R.E., “The Co., Norwalk, Connecticut, USA, 2000.
effect of variation in polymer properties on the rate of 39. Backer, S., Tesoro, G.C., Toong, T.Y. and Moussa, N.
burning,” Fire and Materials, 33, 257–271, 2009. A., “Textile Fabric Flammability,” The MIT Press,
22. Nield, D.A. and Bejan, A., “Convection in Porous Cambridge, Massachusetts, USA, 1976.
Media,” Springer-Verlag, 1992. 40. Williams, F.A., “A Review of Flame Extinction,” Fire
23. ASTM E-1354-03, Standard Test Method for Heat Safety Journal, 3, 163–175, 1981.
and Visible Smoke Release Rates for Materials and 41. Rasbash D J, Drysdale D D, and Deepak D, “Critical
Products Using an Oxygen Consumption Calorimeter, Heat and Mass Transfer at Pilot Ignition and Extinc-
American Society for Testing and Materials, tion of a Material,” Fire Safety Journal, 10, 1–10,
Philadelphia, 2003. 1986.
24. ASTM 1321-97a, Standard Test Method for Deter- 42. Fereres, S., Lautenberger, C., Fernandez-Pello, A.C.,
mining Material Ignition and Flame Spread Urban, D. and Ruff, G., “Mass flux at ignition in
Properties, American Society for Testing and reduced pressure environments,” Combustion and
Materials, Philadelphia, 1997. Flame, 158, 1301–1306, 2011.
660 J. Torero
43. Thomson H E, Drysdale D D, and Beyler C L, “An polymers,” Combustion and Flame, 156, 1068–1083,
Experimental Evaluation of Critical Surface Temper- 2009.
ature as a Criterion for Piloted Ignition of Solid 58. Bal, N. and Rein, G., “Numerical investigation of the
Fuels,” Fire Safety Journal, 13 185–196, 1988. ignition delay time of a translucent solid at high radi-
44. Beyler, C., “A Unified Model of Fire Suppression,” ant heat fluxes,” Combustion and Flame,
Journal of Fire Protection Engineering, 4 (1), 5–16, 158, 1109–1116, 2011.
1992. 59. Wasan, S.R., Rauwoens,P., Vierendeels, J. and Merci,
45. Quintiere, J.G. and Rangwala, A.S., “A theory for B., “An enthalpy-based pyrolysis model for charring
flame extinction based on flame temperature,” Fire and non-charring materials in case of fire,” Combus-
and Materials, Volume 28, Issue 5, September/ tion and Flame, 157, 715–734, 2010.
October, Pages: 387–402, 2004. 60. Dakka, S.M., Jackson, G. S. and Torero, J.L.,
46. Cordova, J.L., Walther, D.C., Torero, J.L. and “Mechanisms Controlling the Degradation of Poly
Fernandez-Pello, A.C. “Oxidizer Flow Effects on the (methyl methacrylate) Prior to Piloted Ignition”
Flammability of Solid Combustibles,” Combustion Proceedings of the Combustion Institute, 29,
Science and Technology, 164, No. 1–6, pp. 253–278, 281–287, 2002.
2001. 61. Beaulieu, P.A., and Dembsey, N.A., “Flammability
47. McAllister, S., Fernandez-Pello, A.C., Urban, D. and Characteristics at Applied Heat Flux Levels up to
Ruff, G., “The combined effect of pressure and 200 kW/m2”, Fire and Materials, 2007.
oxygen concentration on piloted ignition of a solid 62. Hallman. J., “Ignition Characteristics of Plastics and
combustible,” Combustion and Flame, Rubber,” Ph. D. Thesis, University of Oklahoma,
157, 1753–1759, 2010. Norman, OK, USA, 1971.
48. Roberts, A.F. and Quince, B.W., “A Limiting Condi- 63. Jiang, F., deRis J.L. and Khan, M.M. “Absorption of
tion for Burning of Flammable Liquids,” Combustion thermal energy in PMMA by in-depth radiation,” Fire
and Flame, 20, 245–251, 1973. Safety Journal, 44, 106–112, 2009.
49. Lautenberger, C. and Fernandez-Pello, A.C. “A 64. Girods, P., Bal, N., Biteau, H., Rein, G. and Torero, J.
generalized pyrolysis model for combustible solids,” L., “Comparison of pyrolysis behaviour results
5th International Seminar on Fire and Explosion between the Cone Calorimeter and the Fire Propaga-
Hazards, April, 23–27, Edinburgh, UK. tion Apparatus heat sources,” Fire Safety Science,
50. Butler, K. M. Mixed Layer Model for Pyrolysis of 10, 889–901, 2011.
Bubbling Thermoplastic Materials, National Institute 65. Bal, N., Raynard, J., Rein, G., Torero, J.L., Försth, M.,
of Standards and Technology, Gaithersburg, MD, Boulet, P., Parent, G., Acem, Z. and Linteris, G.,
NISTIR 6242; October 1998. “Experimental study of radiative heat transfer in a
51. Kashiwagi, T., “Polymer Combustion and translucent fuel sample exposed to different spectral
Flammability-Role of the Condensed Phase,” sources,” International Journal of Heat and Mass
Proceedings of the Combustion Institute, 25, Transfer, (in press), 2013.
1423–1437, 1994. 66. Steinhaus, T. 1999 “Evaluation of the
52. Di Blasi C., “The state of the art of transport models Thermophysical Properties of Poly(Methyl Methacry-
for charring solid degradation,” Polymer Interna- late): A Reference Material for the Development of a
tional 49 1133–1146, 2000. Flammability Test for Micro-Gravity Environments,”
53. Moghtaderi, B., “The State-of-the-Art in Pyrolysis Masters Thesis, University of Maryland.
Modeling of Lignocellulosic Solid Fuels,” Fire and 67. McGrattan, K., Klein, B., Hostikka, S., Floyd, J.,
Materials 30 1–34, 2006. “Fire Dynamics Simulator (Version 5), User’s
54. Lautenberger, C. & Fernandez-Pello, A.C., “Pyrolysis Guide,” NIST Special Publication 1019–5, October
Modeling, Thermal Decomposition, and Transport 1, 2007.
Processes in Combustible Solids,” to appear in Trans- 68. Mowrer, F.W., “An analysis of effective thermal
port Phenomena in Fires, Ed. M. Faghri & B. Sunden, properties of thermally thick materials,” Fire Safety
WIT Press, 2008. Journal, Volume 40, Issue 5, Pages 395–410, July
55. Lautenberger, C., Kim, E., Dembsey, N. and 2005.
Fernandez-Pello, A.C., “The role of decomposition 69. deRis, J. L. and Khan, M. M., “A Sample Holder for
kinetics in pyrolysis modelling – Application to a Determining Material Properties,” Fire and
fire retardant polyester composite,” Fire Safety Sci- Materials, 24, 219–226, 2000.
ence, 9, 1201–1212, 2009. 70. Quintiere, J.G., “A Simplified Theory for
56. Stoliarov, S.I., Crowley, S., Walters, R.N. and Lyon, R. Generalizing Results from a Radiant Panel Rate of
E., “Prediction of the burning rates of charring Flame Spread Apparatus,” Fire and Materials, Vol.
polymers,” Combustion and Flame, 157, 2024–2034, 5, No. 2, 1981.
2010. 71. Wickman, I. S., “Theory of Opposed flame Spread,”
57. Stoliarov, S.I., Crowley, S., Lyon, R.E. and Linteris, Progress in Energy and Combustion Science,
G.T., “Prediction of the burning rates of non-charring 18, 6, pp. 553–593, 1993.
www.ebook777.com
72. Quintiere, J.G., “Principles of Fire Behavior,” Delmar José Torero is the Head of the School of Civil Engineer-
Publishers, 1997. ing at The University of Queensland. He holds a BSc for
73. Lautenberger, C. Torero, J.L. and Fernandez-Pello, the Pontificia Universidad Cat olica del Perú (1989), and
A.C., “Understanding Materials Flammability,” an MSc (1991) and PhD (1992) from the University of
Chapter 1, Flammability Testing of Materials California, Berkeley. Jose is a Chartered Engineer (UK), a
in Building, Construction, Transport and fellow of the Australian Academy of Technological
Mining Sectors, V. B. Apte Editor, pp. 1-21, CRC Sciences and Engineering, the Royal Academy of Engi-
Press, 2006. neering (UK) and the Royal Society of Edinburgh (UK).
Electrical Fires
22
Vytenis Babrauskas
Introduction capacity. The traditional segregation of electrical

fires as Class C fires has been based primarily on
An electrical fire is generally understood to be a concerns of potential shock hazard to fire-
fire that is caused by the flow of an electric current fighting personnel. But research studies [1, 2],
or by a discharge of static electricity. It is not show that this would be a realistic concern only
defined as a fire involving an electrical device or for high-voltage installations. Even for these, the
appliance. For example, a fire on an electric range hazard is minimal. For example, the most recent
that occurs due to overheating and ignition of the study on this topic [3] showed that a fire fighter
oil in a deep-fry pan is not classed as an electrical would have to be holding a straight-stream noz-
fire, even though it involves an electrical appli- zle 1.0 m from a 45 kV power line for a shock
ance. Conversely, an electrical device or appli- hazard, defined as 10 mA passing through the
ance is not always needed for an electrical fire to body, to be created. For a fog nozzle, no hazard
occur. Lightning-caused fires are a form of elec- could be produced at a 2.0 m distance even with a
trical fires and these can ignite, for example, a dry 200 kV line (smaller distances were not
bush, which is not an electrical device. measured) and current with the 200 kV line was
As with other categories of fires, there are only 0.35 mA.
three main aspects of electrical fires to be con- This chapter discusses the differences
sidered: ignition, combustion behavior, and sup- between electric current and static electricity,
pression. This chapter deals mostly with the outlines the various forms of heating that can
ignition aspects. Combustion behavior of electri- take place due to electrical activity, and discusses
cally initiated fires is normally dominated by the how ignition of various substances may take
fuel characteristics of the primary combustibles place due to electrical activity. The electrical
that are involved. These will most commonly be characteristics of lightning are also discussed,
cellulosic or plastic fuels, although with some since this is a form of electrical discharge. Sev-
sustained electrical faults combustion of alumi- eral reviews of electrical fires have been
num (e.g., busbars) can play a significant role. published by Babrauskas [4–6]; these provide
Copper and steel generally do not burn even in additional details on the status of research and
worst-case electrical fires. Sustained burning of practical applications.
aluminum generally does not take place except in
installations of at least 480 V and of high current
Static Electricity and Electric Current
V. Babrauskas (*) In the simplest terms, electricity is a form of

Fire Science and Technology Inc., San Diego, CA, USA energy associated with the movement of

www.ebook777.com
22 Electrical Fires 663
electrons. The movement can be sustained (i.e., A more detailed explanation [7] of the break-
electric current) or not. Sustained movement down process is as follows. Due to cosmic radia-
requires a conductive path to be established, but tion and other factors, a small number of free
the limited movement of electrons possible in electrons are always present in air. If an electric
insulators can allow charge separation to occur. field is applied, the electrons move in the direc-
Once this happens and a certain amount of tion opposite to the electric field (i.e., to the posi-
charge is accumulated, a discharge may be pos- tive electrode). If the electric field is sufficient, an
sible and this whole process is known as static electron can travel only a short distance before it
electricity. Static electricity, however, does not collides inelastically with an atom/molecule and
mean an absence of electric current, since current ionizes it, now leaving two free electrons. Both of
(flow of electrons) occurs both in charging and in these electrons now continue to travel and each
discharging. Instead, static electricity is a some- one will again collide, and create a new pair of
what imprecise concept implying that the pri- electrons (original electron, plus electron
mary flow of electrons is either in insulators or removed from an atom) at this collision. It can
in conductors that are not connected into the be seen that this process leads to exponential
form of a closed circuit. growth and one electron, starting at the cathode,
will result in n electrons reaching the anode,
where n ¼ eαd , with d ¼ gap distance and α ¼
Townsend’s first ionization coefficient, with has
Electrical Discharges
the units of 1/distance. The value 1/α then
represents the distance between successive ioniz-
Breakdown Phenomena
ing collisions. The generation of electrons is fur-
ther augmented by the positive ions which are
Electrical breakdown means that a substance that
created in the process and which move, much
is normally an insulator suddenly (and possibly
more slowly, towards the cathode. When a posi-
just temporarily) breaks down, and becomes a
tive ion collides with the cathode, it then liberates
conductor. The process is somewhat different in
γ electrons, with γ being known as Townsend’s
gases, liquids, and solids. In a gas, the medium is
second ionization coefficient. If only electrons
self-healing—if the driving force is removed, the
that naturally get liberated from the cathode
medium restores itself essentially to its original
enter into this process, augmented by electrons
condition, although a slight chemical change
liberated due to collisions with neutral species
may occur (e.g., some ozone can be created by
along the path, then the discharge (a discharge is
an electrical discharge in air). Liquids are also
the flow of a detectable amount of current)
largely self-healing, but the chemical changes
is called a Townsend discharge, named after
entailed may have some long-term implications.
J. S. Townsend, an early researcher of gas
The best example of this is oil-filled transformers
discharges. A Townsend discharge is non-
that can withstand a certain amount of discharges
luminous and the current flow is small. If the
if these are not too energetic. But each discharge
process increases so that a sizable current starts
causes degradation of the liquid and eventually
to flow, term breakdown is applied, and the two
the transformer may suffer a catastrophic failure
main types of breakdown modes are: electric arc
due to this degradation. Discharges in solids, on
(if sustained) and electric spark (if not). An elec-
the other hand, are usually highly destructive.
tric arc requires that a sufficiently high current
With most solid materials, an electrical discharge
(more than approximately 0.1 A) be available.
creates a path that is permanently damaged or
As soon as a conducting path gets established
destroyed. The majority of insulating solids are
across the gap, the delivery of energy into the arc
organic substances and a discharge through an
channel rapidly increases, but the rate of current
organic solid has the effect of carbonizing the
growth is largely determined by the external cir-
material, but a portion of the material may also
cuit parameters. The actual arc channel starts out
be ablated.
664 V. Babrauskas
small, then grows rapidly in diameter. This radial W 13:5 1:6022 1019
growth is so fast that it is modeled as a shock E¼ ¼
eL 1:6022 1019 0:4 106
wave propagation [8]. The shock front, however,
does not correspond identically with the channel ¼ 34 106 Vm1 ¼ 34MVm1
boundary. The channel expands in two stages.
Initially it expands by means of the expanding This estimate is about a factor of 10 too high, and
shock wave until the shock wave detaches from this is because this simplest effort at estimating
the luminous core. After that point it detaches ignored the electron avalanche effect. In actual
more slowly until an ultimate value is reached. fact, breakdown in air at 1 atm requires a field of
During the initial expansion phase, the channel roughly 3 MV m1, and Paschen’s paper of 1889
conductivity [9] is 1.5 104 S m1 and the [11] is credited with defining a relation between
channel radius r (mm) increases [10] according breakdown voltage, spacing of electrodes, and
to: gas pressure, which has become known as
Paschen’s Law. According to Paschen’s Law,
I 1=3 t1=2 the pressure gap distance product is the
r ¼ 294
ρo
1=6 controlling variable and the breakdown voltage
V is given by:
where I ¼ current (A), t ¼ time (s), and ρo ¼
ambient density of air (kg m3). c1 pd
V¼
c2 þ lnð pd Þ
where p ¼ pressure, d ¼ gap distance, and c1

Paschen’s Law
and c2 are constants. Thus, when the electric
field exceeds about 3 MV m1, breakdown is
The kinetic energy W gained by an electron when
estimated to occur. Modern measurements [12]
accelerated by an electric field E over a distance
of Paschen’s Law curves for air and for nitrogen
L is:
are shown in Fig. 22.1. For gaps greater than
W ¼ EeL about 0.1 mm, it can be seen that the curve is
essentially a straight line. For smaller gaps, how-
where e is the charge of the electron ever, the breakdown voltage does not go to zero
(1.6022 1019 C). The simplest estimate of and, instead, a minimum breakdown voltage is
breakdown would be that a self-sustained ava- found. This minimum of the Paschen curve is
lanche is created when W attains a sufficient approximately 340 V, and it occurs at a Pd prod-
value so that a traveling electron can knock off uct of 0.007 atm-mm. In other words, in ambient
an outer-shell electron from molecules of the gas air, the minimum breakdown voltage occurs for a
through which it is traveling. The distance L is the gap of 0.007 mm (7 μm). This is an exceedingly
mean free path (average distance between two small distance, and two conductors this far apart
collisions) for an electron in ambient air, which would appear to be touching to the naked eye.
is experimentally found to be about 0.4 μm. This For practical applications, it can be more
value is not to be confused with the mean free convenient to present results in the form of the
path of the molecules of air, which is around dielectric strength (MV m1), which is the break-
0.068 μm, a smaller distance to the fact that air down voltage divided by the distance between
molecules are much larger than electrons. The the electrodes. This is shown in Fig. 22.2 and
energy required to ionize an oxygen molecule is indicates that for larger distances, the dielectric
13.5 eV, while for nitrogen it is somewhat higher strength of air is approximately 3.0 MV m1; the
at 15.8 eV. An electron volt (eV) is a unit of value for nitrogen is quite similar.
energy equal to 1.6022 1019 J. Thus, the sim- Paschen’s Law is not absolute but rather
ple estimate would be that breakdown would depends on experimental conditions. These
occur when
www.ebook777.com
Fig. 22.1 The breakdown 100,000

voltage (Paschen’s Law)
between spherical Nitrogen
electrodes in air and Air
nitrogen. The voltage refers
Breakdown voltage (V)

either to DC voltage or to
the peak value, for AC 10,000
voltage (From Ignition
Handbook [7], used by
permission)
1,000
100
0.001 0.010 0.100 1.000 10.000 100.000
Pressure-distance product (atm-mm)
Fig. 22.2 The dielectric 35

strength at 1 atm, as a
Nitrogen
function of distance (From
30 Air
Ignition Handbook [7],
Dielectric strength (MV m–1)
used by permission)
25
20
15
10
0
0.01 0.10 1.00 10.00 100.00
Distance (mm)
include electrode shape, the material of the (rms) values and not the peak values, if results
electrodes, electrode surface contamination, are to be applied to AC voltages, then the values
pffiffiffi
humidity, and the polarity (if electrodes are not indicated in Fig. 22.1 need to be divided by 2.
identical). The law also assumes that the Thus, on an rms basis, the minimum AC break-
impressed electric field is uniform and deviations down voltage is 340/1.414 ¼ 240 V(rms). It must
occur if the field has nonuniformities. be emphasized that Paschen’s Law is not used in
Paschen’s Law curves are identical for both the design of low-voltage equipment. There are
AC and DC voltages. But since AC voltages are many standards worldwide that govern gap sizes
normally described by their root-mean-square (clearances) required for low- or medium-
666 V. Babrauskas
Fig. 22.3 Breakdown 100,000

voltage for mica as a
function of gap distance Austen and Whitehead8
Lewis et al.9
Fitted line
Breakdown voltage (V)

10,000
1,000
100
0.0001 0.001 0.01 0.1 1
Distance (mm)
voltage equipment, but they all mandate values poly(vinyl chloride), PVC. A more recent review
much larger than the minimum that would suffice paper [19], however, indicated that no break-
to prevent breakdown. This is for serviceability down data for PVC are available below
reasons and also takes into account surges. 0.07 mm, at which thickness the breakdown
voltage is still in the kV range (approximately
7 kV).
Dielectric Strength of Solid or Liquid The available data suggest that plastics
Insulators most likely show a relationship where the
breakdown voltage approaches zero as the elec-
The dielectric strength of solids and liquids can trode spacing becomes infinitesimal. But the
also be characterized by similar graphs, but only available data are too few to establish this
limited specialized references exist [13]. Most of quantitatively.
this literature covers solely the HV regime, and
data for voltages below 1 kV are extremely
scarce. Mica is one of the rare insulators for Arcs
which low-voltage data are available, with the
results of Austen and Whitehead [14] and Lewis Definitions of Arc and Spark Both an electric
et al. [15] being shown in Fig. 22.3. For arc and an electric spark fall under the general
polymers, some data obtained by Abed [16] on definition of a continuous, luminous discharge of
polystyrene, PVC, and PTFE are shown in electric current crossing a gap or an insulating
Fig. 22.4. Also shown are data on polyethylene surface between two conductors [7]. They are
obtained by Mason [17] and a single data point distinguished in that an arc is a sustained event
given by Austen [18]. The latter indicates that while a spark is transient. Matters are clouded by
breakdown occurs at 150 V when the insulation the fact that some researchers refer to a “spark
thickness is reduced to 0.003 mm. The most phase” of an arc, but this type of definition is not
common insulator for low-voltage1 wiring is widely held and is not used in fire safety engi-
neering. Spark, however, also has another very
different definition: a small incandescent parti-
1
Low voltage is defined by various institutions as being cle. For clarity, the latter can be referred to as a
lower than 600, 660, or 1,000 V. mechanical spark.
www.ebook777.com
Fig. 22.4 Breakdown

voltage for several
polymers as a function of Polystyrene
gap distance PVC
100
PTFE
Breakdown voltage (kV)

PE (Mason)11
PE (Austen)12
10
0.1
0.001 0.01 0.1 1 10
Distance (mm)
Abnormal glow
Dark discharge
Normal glow discharge
Arc
theoretical predictions, are shown in Fig. 22.6.

Also shown is the following empirical data fit:
T ¼ 6, 500I a 4:5 A
Voltage (V)
T ¼ 4, 010 þ 1, 658 ln I a I a > 4:5 A
The theoretical predictions are only loosely

obeyed, so the empirical data fit should be suffi-
cient for calculation purposes.
Means of Creating Arcs An arc can be created

10–10 10–5 10–4 10–3 10–2 10–1 1 10 102 by a variety of means, primarily the following:
Current (A)
• Raising the voltage across a fixed pair of
Fig. 22.5 Schematic representation of steady-state volt- electrodes until breakdown occurs
age and current for several discharge types (From Ignition • Opening or closing the contacts in a current-
Handbook, used by permission) carrying circuit
• Transitioning from arcing across a carbonized
Characteristics of an Arc An arc is actually path (arc tracking)
only one of several types of steady electrical • Glow-to-arc transition
discharges that are possible, as shown in • Introducing ionized gases in between two
Fig. 22.5. But, of these, only the arc is important electrodes (e.g., from a flame)
from a safety viewpoint, and it corresponds to
discharges of the highest current and the lowest Creating an arc by raising the voltage across a
voltage. fixed pair of electrodes is very common for test-
The temperature of an arc can vary widely. Under ing purposes. It also occurs in some accidental
ambient pressure conditions, it is commonly circumstances, as discussed below.
6,500–12,000 K but can reach 50,000 K. The Contact arcs regularly occur in electric
primary factor governing arc temperature is the switches, relays, and similar devices. They also
arc current. Experimental data, along with some occur inadvertently, when, for example, two bare
668 V. Babrauskas
30,000 The voltage across the gap at the moment of

rupture is only approximately 1 V. The reason
25,000 Paschen’s Law does not apply is that it describes
the characteristics of room-temperature,
20,000 nonionized gases, and the space between the
Temperature (K)
contacts is ionized and at high temperature.

15,000 Even though 10–15 V is needed for the steady-
state operation of an arc, the arc is able to initiate
10,000
with only a 1 V drop due to inductive effects of
the wiring.
In closing switch contacts (a make-arc), the
5,000
sequence of events is very similar. Contact is
initially made at only a few high spots. These
0
1 10 100 1,000 10,000 100,000 have limited current-carrying capacity and pro-
Arc current (A) ceed to melt and rupture, at which point an arc
develops. That arc is normally extinguished by
Fig. 22.6 Temperature of arcs in ambient-pressure air,
heat losses when the contacts close together more
along with predictions from theory (light lines) and an
experimental data fit (bold line) (From Ignition Hand- tightly.
book, used by permission) Arcing across a carbonized path (arc tracking)
is arcing that is supported by a carbonized path
current-carrying conductors are accidentally on a solid, as discussed below, and this is a
shorted. When considering electric switch low-current process. If this process continues
contacts, the arc caused by closing the switch is and accelerates, one possible outcome is a nor-
called a make-arc (or closing arc), whereas the mal, high-current arc across air. The conditions
arc caused in opening the switch is a break-arc leading to this have not been explored in
(or opening arc, or parting arc). The process of detail, however. Glow-to-arc transition is a rare
creating an arc (at voltages that may be much less phenomenon not normally encountered in
than Paschen’s Law minimum of 340 V) is quite fires [20].
similar for both types of contact arcs. In the case The dielectric strength of a hot, ionized gas is
of a break-arc the steps involved are the tiny compared to that of normal ambient air.
following: Thus, when the distance between two conductors
1. The electrodes that were originally touching is such that there would be no possibility of
at numerous spots start to touch at only a few breakdown in normal air, introducing ionized
very small spots. gases in between two electrodes (e.g., from a
2. A high current density passes through the flame) can lead to arcing. This is why, in many
small metal diameter of contact area that is fires, a large number of artifacts are found
available. suggesting that an arc occurred there. It is not
3. The metal bridge joining the two contacts because several different places arced simulta-
starts to melt. neously and each erupted into fire [21]. Instead, a
4. The bridge elongates and rises in temperature. single fire introduced flames into various locales
5. The bridge reaches the metal’s boiling point, where conductor-to-conductor spacing was such
becomes unstable, and ruptures. that arcing could not be supported in ambient air
6. Voltage rises rapidly across the gap, therm- but could be supported in an ionized, hot
ionic emission from the hot cathode starts, and medium.
eventually the gap becomes ionized and an arc
forms. Arc Flash The thermal radiation from an arc is
7. The diameter of the arc expands from that of referred to as arc flash and a sustained, high-current
the bridge to its eventual free-burning arc can lead to severe injury or death of an individ-
diameter. ual so exposed. IEEE Std 1584 [22] provides a
www.ebook777.com
calculational method for estimating arc flash radi- than 150 V tend to extinguish and not reignite
ation, whereas ASTM F 1506 [23] and ASTM F when the waveform goes through the zero cross-
1959 [24] provide procedures for assessing the ing. Arcs in circuits of over 600 VAC tend
actual effectiveness of protective clothing. to draw very high currents and, consequently,
may be relatively safer since a circuit
Arc Extinguishment A DC arc will extinguish protection device is likely to open. Voltages
only if the power supply is removed or if enough between 150 and 600 V are considered the
material erodes to make the gap too large. In an most hazardous in regard to fires being
AC circuit, an arc will self-extinguish 2 60 ignited from arcing [25]. This is because
times (2 50 times with 50 Hz power) per the arcs tend to not be extinguished, yet the
second, each time the current goes to zero. But current flows are small enough that circuit pro-
it may reignite thereafter if conditions are right. tection devices operate slowly. Typical
Whether or not the arc reignites afterwards waveforms [26] for arcing in 120 VAC
depends on whether the arc channel can recover circuits are shown in Fig. 22.7. Note the inter-
so that it will not break down again with the mittent nature of the arc as it extinguishes and
new imposed voltage. Arcs in circuits of less reignites.
Voltage
Current
Fig. 22.7 Typical waveforms during a branch-circuit arcing event

670 V. Babrauskas
Table 22.1 Minimum ignition energy (MIE) of some

Ignition Modes Involving Electric common gases and vapors
Current Substance Energy (mJ)
Acetone 2.15
Sparking or Arcing Acetylene 0.03
Ammonia 680
Electric sparking or arcing can ignite materials in Benzene 0.91
all phases: gases, liquids, solids, liquid aerosols, Butane 0.26
and dust clouds. Ignition in gaseous and dust Carbon disulfide 0.039
cloud media has been studied extensively. Igni- Cyclohexane 2.65
tion in bulk liquids is rare, apart from oil-filled Ethane 0.42
transformers and other HV devices. Ignition of Ethylene 0.114
liquid sprays, fogs, or aerosols is problematic in Ethylene oxide 0.105
some industries, however. Ignition of solids from Furan 0.328
arcing or sparking is common but has not been Heptane 1.15
Hexane 0.29
researched to any satisfactory degree.
Hydrogen 0.03
Hydrogen sulfide 0.077
Gases If an atmosphere exists where a flamma-
Iso-octane 2.9
ble gas has been dispersed into an oxidant gas
Methane 0.71
(commonly air) and the mixture is within its Methanol 0.3
flammable limits, spark ignition is generally Pentane 0.82
very easy. Arc ignition has normally not been Propane 0.5
studied. Since a very low-energy transient energy Propylene 0.418
discharge ignites such mixtures, a sustained Toluene 2.5
energy discharge will be much more capable of Vinyl acetylene 0.095
ignition. For this reason, the phenomenon is p-Xylene 0.2
referred to as spark ignition rather than arc igni- From Ignition Handbook, used by permission
tion. Flammability limits for a number of gases
are given in Chap. 17. A larger collection of data errors, leading to reported values for many
is provided in the Ignition Handbook [7]. Some substances being much lower than their true
values for minimum ignition energy (MIE) are value [7]. But, as a very rough rule, dust clouds
given in Table 22.1; more extensive tables are will not reach their LFL unless visibility is down
available in the Ignition Handbook [7]. It should to zero in that location. The upper flammability
be observed that these energies are exceedingly limit (UFL) for dust clouds is rarely measured,
small and are tabulated in millijoules. To appre- simply because it is generally very difficult to
ciate the magnitude, one can consider the fact generate a dust cloud that exceeds the UFL.
that if a coffee mug is raised by 0.3 m, its poten-
tial energy is increased by roughly 1.0 J. Minimum ignition energies have been
tabulated and it is believed that those are more
Dust Clouds Dust clouds are significantly more reliable. Some typical values are shown in
difficult to ignite than gases, but explosions due Table 22.2. Unlike gases, where MIE values are
to this cause remain an important concern in typically below 1 mJ, MIE values for dust clouds
manufacturing, mining, and agricultural are typically some two orders of magnitude
industries. The lower flammability limit of dust higher. Nonetheless, these are all still low, even
clouds has generally been erroneously reported though they are higher than for gases.
in most data compilations because apparatuses
used to measure the lower flammability limit Solids Ignition of solids from electric sparks or
(LFL) of dust clouds have had gross, systematic arcs is unfortunately common. The cause can be
www.ebook777.com
Table 22.2 Minimum ignition energy for various dust igniting it. There are two primary factors
clouds operating in such cases: (1) The arc impingement
Substance MIE (mJ) may be very brief, many combustible materials
Aluminum 50 can resist enormous heat fluxes if these are
Aspirin 25–30 sustained only briefly. (2) The material may
Black powder 320 ablate too rapidly to allow ignition. These
Coal 250 mechanisms, however, are understood only qual-
Cocoa 100–180 itatively—there have not been research studies to
Coffee 160 successfully quantify them.
Cornstarch 30–60
Cotton linters 1,920
Dextrin 40
Arcing Across a Carbonized Path
Flour, cake 25–80
Grain dust 30
Many electrical fires are due to arcing across a
Magnesium 40
carbonized path. If a carbonized path is created
Manganese 305
Nitrostarch 40
where current may potentially flow, arcing may
Nylon 20–30 then occur along this path, possibly leading to
Paper dust 20–60 ignition either of the combustible insulator itself,
Phenol formaldehyde 10–6,000 or some other nearby fuel. A carbonized path can
Polyethylene 70 be created in at least three ways [27], such as the
Polyethylene terephthalate 35 following:
Polystyrene 40–120 1. Arc tracking
Rice 40–120 2. Overheating (by electrical overcurrent, exter-
Silicon 100 nal radiant heating, etc.)
Soap powder 60–120 3. Impingement of fire on solid electrical insula-
Sugar, powdered 30 tion material
Tantalum 120
Tin 80 Arc Tracking Of these three possibilities, sub-
Titanium 25 stantive research has been done only on arc
TNT 75
tracking. Arc tracking is a progressive creation
Urea formaldehyde 80–1,280
by electrical means of a carbonized path along
Wheat starch 25–60
the surface of an insulator that separates two
Wood flour 30–40
current-carrying conductors. Arc tracking is
Zinc 960
subdivided into two types: dry tracking and wet
From Ignition Handbook, used by permission
tracking. Dry tracking can be induced by causing
an electric arc to impinge onto the surface of an
either static electricity or circuits carrying an organic material. Wet tracking can occur if a film
electric current. Some of the mechanisms have of water covers the surface of the insulator and
been studied, and these are discussed later. How- spans between two conductors.
ever, the problem of understanding the response
of solid materials to a spark or arc ignition source The electric conductivity of pure water is very
has been neglected. Apart from metals and some low, but when ionic contaminants are dissolved
other rare substances, there are no combustible in water, its conductivity increases and it
solids with an ignition temperature over becomes possible to create a current flow if the
1,000 C. Figure 22.6 shows that the temperature layer of moisture has access to conductors from
of an electric arc is at least 6,500 K and may be both sides of the line. The flow of current then
much higher. Yet, an electric arc impinging onto has a drying effect on the moisture layer. The
a combustible solid is not necessarily assured of drying is nonuniform, and eventually dry patches
672 V. Babrauskas
start to be formed along the current path. With tracking will normally not be initiated unless a
buildup of carbonization along the path, small conductive moisture film exists that has electrical
electrical discharges, called scintillations, can contact to two conductors that have a voltage
then occur. Since part of the current flow is difference between them. This may happen if a
through an electrolyte of significant resistance, cable is mechanically damaged so that two
these scintillations represent a very small current current-carrying conductors are exposed. Mois-
flow and would not trip any overcurrent- ture then collects on the damaged area, and
protection devices. The ultimate event, if it pollutants are present that ionize the layer. But
occurs, is the actual flaming ignition of the mate- on some materials, arc tracking does not require a
rial over which tracking is occurring. direct contact between an electrode and the sur-
If the tracking is dry, the processes that occur face of the insulator; tracking over phenolic and
along the surface are similar. Although an melamine surfaces can be initiated even when the
overcurrent-protection device cannot be expected electrodes are separated by gaps of about
to protect against arcing across a carbonized path 0.25 mm from the insulator surface. In general,
(unless this escalates to a high-current fault), arc in low-voltage circuits, a carbonized path is
fault circuit interruption (AFCI) devices, which probably most commonly created by a poor con-
have been developed in recent years [28], are nection or other source of locally elevated
intended to respond to this condition; because temperatures, but moisture or pollutants can
they are relatively new, however, field- also be of significant importance.
performance data do not yet exist, although it is Oba [32] conducted experiments where he
known that some models of first-generation damaged the insulation on Japanese
devices have not been highly successful [29]. PVC-insulated power cable to expose the
Surprisingly, temperatures up to 1,000 C can conductors and then sprayed electrolyte onto
be generated by such surface leakage discharges. the area to initiate arc tracking. By varying the
These elevated temperatures then continue the AC voltage supplied to the cord, he obtained a
process of polymer carbonization. Thus, in the char length relation as a function of voltage
tracking process, a carbon track is laid down (Fig. 22.8). Below 50 V, progress of charring
along the surface, and that track has a low
enough resistivity that current can subsequently
start to flow along the carbonized track, which, in 8
turn, causes more carbonization and more
7
heating. A runaway situation can then develop.
Scintillations can vary widely in their discharge
Char length developed (mm)
6
energy, with the low end of the scale being mild
events that would not be expected to damage 5
metals (although they might ignite some poten-
4
tial targets). Nakamura et al. [30] measured
scintillations on PVC and reported values that 3
ranged from 100 to 3,000 J. But, they did not
endeavor to set up their experiments to elicit the 2
lowest possible discharge energy.
1
To create arc tracking, a vastly lower voltage
can suffice than for breakdown in air between 0
two electrodes. For example, with many plastics, 0 50 100 150 200 250
Yoshimura et al. [31] found that 600 VAC was Voltage (V)
sufficient to cause an arc discharge across a 4 mm
Fig. 22.8 Char length developed after 70 h in the wet
gap. By contrast, breakdown across a 4 mm gap arc-tracking experiments of Oba (From Ignition Hand-
in air requires about 10,000 VAC. On cables, wet book, used by permission)
www.ebook777.com
was very limited. He repeated the experiments by d1

depositing powdered carbon onto the damaged
area and found, in that case, for arc tracking to Metal
occur, the applied voltage had to exceed 24 V. conductor
Below 50 V, small incandescent spots could be d2 Plastic

produced but not flaming ignition. Flaming igni- insulator
tion was readily possible for voltages of
100–150 V. Above 200 V, flaming ignition read-
ily occurred, but increased char lengths were not
Fig. 22.9 Simplified view illustrating clearance and
obtained since the events were explosively force- creepage distances (From Ignition Handbook, used by
ful and blew off the carbonized material and the permission)
melted conductor portions. Under other
conditions, much less than 24 V is sufficient to effective in reducing the arc-tracking propensity
cause arc tracking. Bernstein [33] reports that arc in certain polymers [36, 37].
tracking can occur in 6 V battery circuits, Most research studies have not focused on the
provided the battery has sufficient current capac- time element and none have done so in a system-
ity to sustain the arc. Arc-tracking problems have atic way. The minimum current needed, how-
been troublesome in a variety of models and ever, has been studied. Wilkins and Billings
types of motor vehicles in circuits operating at [38] obtained the following minimum values:
around 14 VDC (nominal 12 V) [7]. PVC 0.15–0.20 mA, PVA 0.3 mA, Ebonite
For polymers, the molecular structure is the (butadiene/methylstyrene rubber) 1.1 mA, phe-
main determinant of arc-tracking propensity nolic/paper 1.15 mA, polycarbonate 1.2 mA, and
[34]. Aliphatic polymers (e.g., polyethylene, PTFE 2.3 mA.
PTFE) tend not to undergo arc tracking, whereas
aromatic ones or ones containing alternating dou- Creepage In design, the resistance to arc track-
ble bonds (e.g., phenolic, polyethylene tere- ing is controlled by two means: (1) selection of
phthalate, polystyrene) do because the latter, well-performing insulation materials, and
when pyrolyzed, leave residues that are electri- (2) observing adequate creepage distances. The
cally conductive or semiconductive. Also latter concept is illustrated in Fig. 22.9. For arcing
tending to exhibit arc tracking are polymers in air from metal to metal, the governing distance
that, although lacking aromatic rings or double is called the clearance distance, d1. But since arc
bonds originally, form rings or double bonds tracking proceeds only along solid surfaces, the
during thermal degradation; PVC and polyacry- distance across which arc tracking must travel, if
lonitrile are examples. failure is to occur, is called the creepage distance,
It was also found that oxygen is not a requisite d2. Creepage distances are set down in numerous
for the formation of arc tracks and that materials military and industrial specifications, but the
can be made to arc-track in a nitrogen atmo- rationale is usually empirical and not much sci-
sphere. Generally, arc tracking can only happen entific research is available on the topic.
if a polymer can char, since a conductive track
can only be established in char. Practical
difficulties arise in evaluating this because char- Surface Flashover
ring is not an absolute property of polymers but,
rather, depends on environmental details. It has Fire protection engineers need to be aware that
also been found [35] that arc tracking is promoted the term flashover is used in a very different way
by the presence of halogen atoms in the polymer. in electrical engineering, where it means “a dis-
Conversely, alumina trihydrate (Al2O33H2O), a charge which occurs over the surface of a solid
common filler for many polymers, is highly dielectric in a gaseous or liquid medium.” [39]
674 V. Babrauskas
The dielectric strength of air is lower than that of Ignition in the excessive-overload mode is
any commonly used electrical insulators, so if the unlikely to occur if the cable is in a circuit that
path through air and the path through a solid is protected by a circuit breaker/fuse correctly
insulator are of similar length and breakdown matched to the rating of the cable, since tripping
occurs, it will go through air, not through the would occur rapidly under 3 and greater
solid insulator. The surface of an insulator may overloads. But ignitions can readily occur if a
become polluted so that its breakdown strength much smaller gauge cable is used than
becomes low; this problem is common in locales corresponds to the rating of the circuit breaker.
exposed to salty air near the sea. The material An overload may not directly ignite an insulated
with the lowest breakdown strength may be this wire but may significantly raise the temperature
pollutant film and, if breakdown occurs along of both the wire and the insulator. Old-style
this film, it is referred to as “surface flashover.” rubber-insulated wires used to be prone to a
This reaction does not constitute arcing across a sleeving effect, whereby insulation closest to
carbonized path, since the path, although of low the wire is thermally degraded and shrinks back
breakdown strength, is not carbonized. The prob- from the conductor.
lem is relevant only to high-voltage (HV) circuits For wires insulated with thermoplastic insu-
and would be a source of ignition only in the lation (including the majority of today’s com-
vicinity of HV installations. mon cable types), a somewhat different effect is
found. Elevated temperatures cause copper to
elongate but the insulation to shrink. As a result,
Overloads and Related Phenomena copper wires readily “pop out” of the softened
insulation. A direct metallic contact can
Gross Overloads Excessive overload can lead then occur, with this short circuit being a
to fires, but this condition is much rarer than is an localized place of ignition [7]. By contrast, if a
arcing fault. It can arise if either a circuit breaker PVC-insulated cable is externally heated
is faulty or a cable is used that is of much smaller (by fire or otherwise), it usually chars
gauge than is the rating of the circuit breaker. rather than melts. But melting, rather than char-
Both of these situations are relatively uncommon. ring, may occur if the external heating is with a
Ampacity ratings of wires and cables are conser- very low heat flux [40], below about
vative enough that an overload of roughly 2 is 15 kW m2. Bubbling of thermoplastic insula-
not expected to create any significant problems, at tion has been experimentally found only to
least in the short term (long-term thermal degra- occur from overcurrent and not due to external
dation of insulation material is a separate issue). heating [40].
If a sufficiently overloaded condition persists, Excessive Thermal Insulation Ampacity

then cables may be able not just to ignite but also ratings for cables are based on a certain amount
to create a propagating, self-sustained fire. of convective cooling being available, which can
Experimental studies [7] indicate that, for this be defeated by thermal insulation. Thus, even if
event to occur, the current carried must be the current passed through a cable is within its
300–700 % of the rated current (ampacity). How- rating, embedding it in thermal insulation can
ever, all existing tests have been short-term. cause the temperature to rise to values that are
Even a current at 200 % of rated ampacity, if no longer reasonable for the particular class of
sustained for a protracted period of time, may insulation used. If, in addition, an overload con-
deteriorate the insulation enough so that carboni- dition is created, the heating can be greatly
zation can begin. Eventually, failure may not be exacerbated. Bunching of cables together can
the melting and shorting commonly involved in also lead to overheating, since ampacity ratings
ignitions from short-term overloads but, rather, envision only a limited aggregation of adjacent
some form of tracking damage. conductors. Goodson et al. [41] observed a house
www.ebook777.com
during construction where charring damage was The current can cause overheating of the metal
already found on bunched NM cables. Thus, they and lead to a rupture of the pipe [45].
ran wall-cavity tests using bunched, 90 C-rated In recent years, a related problem has been
NM cables and obtained charring when the stud fires or explosions in houses due to use of CSST
cavity was insulated with polyurethane foam (lightweight corrugated stainless steel tubing)
insulation. gas piping instead of black iron [46]. These
products proved to be particularly susceptible
Prior to World War II, knob-and-tube wiring to puncturing by lightning strikes, which do not
was common in the United States. This form of need to be direct enough to cause other damage.
wiring uses two separate conductors that are not Gas escapes from the small holes created and a
grouped into a cable but are individually strung gas explosion or a house fire ensues. In cold
on widely spaced porcelain knobs. The current- climates, it is not rare for individuals to thaw a
carrying capacity of this form of wiring is depen- frozen water pipe by attaching a welding trans-
dent on there being unobstructed air cooling of former and passing current through it. Fires
the wires. Fires have occurred when the wires have resulted due the very large currents that
were buried in thermal insulation. A similar are involved [47].
problem can be encountered when extension
cords, which are rated for exposed-air use, are Overvoltage, Floating Neutrals, and Surges
buried under thermally insulating objects or else Ignitions from an overvoltage is relatively rare,
are coiled in multiple layers on a cord reel while as concerns branch-circuit wiring. The materials
carrying a high current but one still within the used for wires and wiring devices are well able to
nominal rating. withstand the normal surges that are a regular
event in a power distribution system. To experi-
Stray Currents and Ground Faults Stray ence ignitions, one of the following events is
currents occur when circumstances cause cur- generally needed:
rent to flow through paths not intended to carry 1. Lightning strike
current. Ground faults are a well-known exam- 2. Accidental delivery of high voltage into
ple [42]. They can occur if a conductor is low-voltage wiring
abraded or damaged and contacts metal siding, 3. A floating neutral
roofing, and so on. Kinoshita et al. [43] 4. A large voltage spike (surge)
documented that a current of only 5 A was
required for ignition when a three-conductor, Lightning strikes can result in massive
PVC-insulated cable contacted a galvanized ignitions, not just of wiring but also of all sorts
iron roof. The Consumer Product Safety Com- of combustibles.
mission (CPSC) has hypothesized that the Occasional fire reports are encountered in
Beverly Hills Supper Club fire, one of the which, due to some malfunction in the power
deadliest U.S. fires of the twentieth century, distribution network, high voltage got applied to
was made possible by improper wiring of the wiring intended to carry only 120/240 V. A
neutral conductors and triggered by a ground case is documented [48] in which a utility trans-
fault that occurred [44]. former fault caused all the ground-fault circuit
interrupter devices in a house to fail, along
If a building has conductive components with igniting a fire due to an explosion of a TV
throughout it, such as metal lath, aluminum set. In another case, a faulty transformer caused
siding, an electric fault can result in an electrified the service entrance wires to ignite and burn
house, leading to multiple ignitions [7]. The fault inside a house. A systematic study of such
is commonly contact with HV wiring. fires does not exist, but ignition should be con-
An unusual mode of ignition from a ground sidered very likely any time that such a failure
fault is where current flows through a gas line. occurs.
676 V. Babrauskas
L1 one is fed to it. Conversely, most devices will not

ignite if the voltage delivered to them is too low.
120 V R1 Rx
Break in N Electric motors, however, are an exception, and
neutral flaming fires can result from certain motors run-
120 V R2 ning at a sufficiently undervoltage condition.
An ignition due to undervoltage can also
L2
occur if one hot leg of a 240 V circuit is discon-
Fig. 22.10 Floating neutral (From Ignition Handbook, nected. If the circuit has any 240 V appliances
used by permission) and these are energized, then they can transfer
power from the live leg to the disconnected leg.
A floating neutral (sometimes called open But the delivery will be through a sizeable resis-
neutral) can lead to ignition in 120/240 V wiring tance and much less than 120 V will be delivered.
circuits due to a special nature of that circuit The preceding discussion ignored the pres-
topology.2 In a single-phase service entrance, ence of the ground wire. According to The
there are three current-carrying wires: two hot National Electrical Code [49], a grounding elec-
wires and one neutral. Figure 22.10 illustrates trode must be connected from the neutral to an
the normal feed from an outdoor transformer to earth ground at the service entrance. But,
a building. Inside the building, the system provided that the neutral is functioning properly,
becomes effectively a four-wire system, since a this ground wire serves no observable function.
safety grounding wire is also run that is Consequently, there may be little to prevent its
connected to the neutral and terminates at a deterioration or abuse over the years. If a break in
ground rod.3 the neutral then occurs, a sizable current can flow
All 240 V loads are directly connected across through the ground wire. If the ground wire
L1 and L2 and do not depend on the presence of passes near or through combustibles, and an
the neutral. But 120 V loads are connected across excess current ends up flowing through it, then
N and either L1 or L2. If a neutral is in place, the an ignition might occur at that place. Fires have
loads will receive the intended 120 V voltage. also been reported [50] in installations using
However, if a break occurs in the neutral, the armored cable when a floating neutral occurred
voltage delivered to 120 V loads can swing and current that would normally flow through the
widely, in principle from barely above 0 to almost neutral instead flowed through the armor.
240 V, although in practice the range is not quite Surprisingly large voltage spikes can be found
as large. Figure 22.10 shows the circuit arrange- on 120/240 V systems. Without any overt fault
ment. The voltage present across a particular load conditions, simply the operation of a motor con-
Rx will be determined by the voltage divider troller can create a 2,000–3,000 V spike
action of other loads in the system, designated [51]. The majority of voltage surges, however,
as R1 and R2. The voltage across Rx will be are due to external—not in-premises—factors.
240 The primary causes are lightning, electrical util-
Vx ¼ ity switching transients, and failures of
R2 R2
1þ þ components in the high-voltage electrical trans-
R1 Rx
mission system [19].
Most electrical or electronic equipment can To evaluate the role of voltage surges, it is
ignite if a voltage much in excess of the intended essential to recognize that there is a fundamental
dividing-line voltage. Surges above approxi-
mately 6,000 V(peak) lead to a “sparkover of
2 clearances” widely throughout the house. In
The discussion here is based on electrical practice in
North America. other words, the householder will typically find
3
Mobile homes normally have a four-wire service from the majority of outlets and other electrical
pole to building. devices have suffered calamitous damage, and
www.ebook777.com
Fig. 22.11 Surge voltages 100

experienced in branch-
circuit wiring
10
Occurrences per year

1
0.1
0.01 Breakdown over normal

clearance distances
0.001
100 1,000 10,000
Peak voltage (V)
possibly multiple fires were ignited. Conversely, electrical damage. The fires erupted about
surges below about 6,000 V(peak) will generally 4 months after the lightning strike [19]. In the
appear benign. But appearances may be deceiv- case of breakdown of insulation due to below-
ing in the case of surges that are in the kilovolt 6,000 V surges or spikes, a delayed fire can occur
range but below 6,000 V. The 6,000 V value in the following two ways:
corresponds to the level at which wiring devices4 1. The initial fault clears itself and the circuit
that are properly designed, installed, and breaker is not tripped (or fuse not blown).
operated will typically suffer a breakdown. But 2. The circuit breaker trips, but the householder
devices that have a manufacturing or installation resets it, and operation seems “normal.”
defect may break down at less than 6,000 V. The In either case, after a certain period in which
hazard comes about if that breakdown is not arc tracking progresses unnoticed, fire
visually obvious (e.g., concealed inside a wall breaks out.
or within plastic). It should be noted that applying Paschen’s
A breakdown of plastic insulation will lead to Law to clearances typically specified by electri-
the formation of a carbonized path along which cal standards, it would be expected that wide-
the discharge occurred. As explained earlier, spread sparkover of clearances would require
once a carbonized path is formed that extends about 6,500 V(rms) or 9,200 V(peak). The
electrode to electrode, arc tracking can start. The empirical observation that 6,000 V(peak) is typi-
process is slow, although an exact time frame has cally sufficient evidently reflects the fact that
not been established. In one documented case, devices in the field do not quite behave as ideally
electrical fires due to this cause occurred after a as their laboratory testing would suggest.
modest lightning strike struck a house that did The 6,000 V(peak) value is vastly greater than
pffiffiffi
not initially lead to fire or widespread visible the operating voltage of 120 2 ¼ 170 V(peak).
But such peak voltages are not rare, as indicated
by several field studies. Figure 22.11 indicates
4
This refers to outlets, plugs, and similar devices. Electric that an individual house would expect to suffer
and electronic appliances are often designed to much less such a surge roughly once every 100 years
stringent standards and may fail or start burning at signifi-
cantly smaller surge levels. [19]. Thus, the risk for an individual house is
678 V. Babrauskas
Overheating
Cu2O breeding
Increase of
temperature
Initiation Increase of
of glowing contact resistance
Surface oxidation Creep and

Oxidation on contact spot relaxation
Vibration Breathing Current flow
Electric arc Migration of

Softening
erosion contact
Surging Thermal expansion

Work hardening
in current contraction
Current cycling
Heat cycling Stress
ON/OFF
Fig. 22.12 Mechanisms for overheating at electrical connections, as outlined by Kuroyanagi et al [52] (From Ignition
Handbook, used by permission)
low, but within a given community there can be a discussed later—there currently exist no useful
number of such events every year. It must be studies on this point.5
noted that the field studies were all completed The physics and chemistry of electrical
before the current era of widespread use of surge connections are very complex, as illustrated by
suppressors. Thus, houses where a sizable num- the phenomenological flowchart put forth by
ber of surge-suppression devices are used can be Kuroyanagi and coworkers [52]. From
expected to be at lower risk than these statistics Fig. 22.12 it can be seen that numerous phenom-
indicate. ena are involved, but not all have been studied
systematically and in detail. In the simplest
terms, failure can be understood to involve the
Overheating Connections following factors:
• Localized heating takes place, due to smaller
Failures of electrical connections are generally effective area available for current flow com-
due to manufacturing defects, installation pared to a bulk, undamaged conductor.
defects, design defects, abuse, damage, or envi-
ronmental effects. In addition, it can be expected 5
that much like any other mechanical device, an The NFPA Research Foundation is currently conducting
a study on aging electrical wiring and there may be
electrical connection will have a finite lifetime, conclusions obtained from it concerning the potential
but—apart from the aluminum-wiring problem lifetime of electrical connections.
www.ebook777.com
• Heating accelerates oxidation and promotes and the time constant τ (s) is given by
creep.
• Creep causes relaxation of the mechanical r ρC
τ¼
forces restraining the connection, leading to 2 h
fewer micro-areas through which effective where ρ ¼ density (kg m3) and C ¼ heat
current flow can take place. capacity of copper (J kg1 K1). The equilibrium
• This relaxation further raises the temperature, value of the temperature rise is
which further accelerates oxidation. rffiffiffiffiffiffiffi
• Oxidation diminishes the area through which 00 r
lim ΔT ðtÞ ¼ q
current can readily flow. t!1 2λh
• Expansion and contraction from thermal
cycling may cause further loosening. This As an example, for a copper wire of 14 AWG,
may cause certain areas of micro-contact to r ¼ 1.63/2 ¼ 0.815 mm, ρ ¼ 8,890 kg m3,
make and break, while a more severe effect C ¼ 385 J kg1 K1, and λ ¼ 400 W m1 K1.
entails irrecoverable plastic deformations. Assuming that h ¼ 50 W m2 K1, and that
Thermal cycling may be due to ambient tem- 10 W is dissipated in the connection, giving
perature fluctuations, or due to fluctuations in q00 ¼ 10/pr2 ¼ 4.8 106 W m2. Then
current, leading to changes in I2R heating. sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
• The presence of moisture or corrosive gases in 0:815 103
the environment can accelerate failure due to ΔT ð1Þ ¼ 4:8 106 ¼ 685
2 400 50
additional chemical degradation. In the case
of PVC insulation, once sufficient overheating If the ambient temperature ¼ 20 C, then the
takes place, HCl gas will get liberated from temperature at the overheating connection will
the plastic, and this is highly corrosive. be 20 + 685 ¼ 705 C, which is much higher
• The presence of vibrations also serves to make than the ignition temperature of most
and break micro-contact areas, resulting in combustibles. This theoretical treatment is
worsening of the connection. highly simplified; nonetheless, it indicates that
The simplest theoretical model of an very high temperatures can be anticipated.
overheating connection is obtained by assuming In the early stages of failure, bad connections
that heat is produced at a constant rate in an give little external evidence of their deteriorating
infinitesimally thin plane section across the condition. It is sometimes considered the infrared
wire. The wire is represented as a cylinder with- (IR) thermal imaging can be used as a preventive
out any change of geometry at the point of con- maintenance operation, since the technique can
nection [53]. The solution for the temperature of graphically show hot spots. However, research
that cross section is studies have shown that this is not possible until
very late in the process [55]. In the earlier stages,
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

00r 1 the cool metalwork surrounding the overheated
ΔT ðtÞ ¼ q γ , t=τ spot essentially preclude finding and identifying
2πλh 2
the spot.
Evidence of overheating is clear when
where mechanical connections between two current-
q00 ¼ Power density at the plane section (W m2) carrying conductors start to show glowing. Nor-
r ¼ Radius of the wire (m) mally, good electric connections should not be
λ ¼ Thermal conductivity of the wire subject to a temperature rise much in excess of
(W m1 K1) that for the conductors themselves. This depends
h ¼ Effective heat transfer coefficient from the on the connection having a very low resistance.
surface of the wire (W m2 K1) Most metals that are used for carrying electrical
γ ¼ Incomplete gamma function [54] current are subject to oxidation when exposed to
680 V. Babrauskas
atmosphere. The metal oxide film formed on the depends only on the materials involved and not
surface has a very high resistivity. Thus, a con- on the nominal size of the contacts.
nection where the mating parts are oxidized will The Cu2O breeding process at a glowing con-
be a high resistance connection and will overheat nection has been studied by several Japanese
if significant current is passed through research groups [60, 61], that provided numerous
it. Temperatures of a glowing connection vary details of this complex process [7]. One of the
widely, but peak values at the hottest point have things learned was that the process is primarily
been measured from 1,100 C [56] to 1,500 C confined to solid conductors; significant currents
[57, 58], for copper connections. Temperatures are hard to sustain at a glowing connection of a
up to approximately 300 C have been measured stranded wire—the wire tended to break at the
on metal parts some distance away from the hot point of heating.
point [54]. Even though copper melts at Overheating can occur in electrical
1,083 C, much higher temperatures can be connections of all types, but historically the
found, since the hottest portion is on the metal most notorious case was that of aluminum wiring
oxide which is being formed and not on the during the early 1970s. In the late 1960s some
metallic copper. U.S. house builders introduced a cost-cutting
One of the earliest studies on glowing measure whereby aluminum was substituted for
connections was published [59] in 1961 copper conductors in house-wiring branch
(Fig. 22.13) and it was found that the connection circuits. This substitution was done without ade-
acts as a nonlinear circuit element. For currents quate research or field-testing and the outcome
over 10 A, drops of around 2 V were found. But was a rash of house fires. The Consumer Products
for small currents, voltage drops in the tens of Safety Commission and the National Institute of
volts can be found. At a maximum current of Standards and Technology (NIST) conducted
20 A, 50 W is dissipated in a copper/brass con- extensive studies on the phenomenon and found
nection and around 35 W for copper/iron. The that the problem occurs due to a combination of
study also noted that the power dissipation metallurgical factors including creep, hardness,
Fig. 22.13 Power 60 30

dissipation and voltage
drop across glowing
connections of two types 50 Cu/Brass 25
(From Ignition Handbook,
used by permission)
Power dissipated (W)
40 20
Voltage drop (V)
Cu/Fe
30 15
20 10
10 5
Cu/Brass
Cu/Fe
0 0
0 5 10 15 20
Current (A)
www.ebook777.com
and oxide characteristics [7]. The outcome is that ignition of flammable atmospheres, but it does
small-diameter aluminum wires6 cannot be reli- not appear that any actual incidents have been
ably terminated by a screw connection and show reported [7].
rapid deterioration and failure. Details
concerning the failures of various specific types
of electrical connections can be found in the Time for Fire to Initiate from
Ignition Handbook [7]. an Electrical Cause
One area in which research is seriously lacking is

Ejection of Hot Particles for the time frame associated with electrical fires.
In some cases, such as a major lightning strike or
Electrical short circuits and arcs sometimes eject a high-current arcing fault in a 480 VAC busbar,
incandescent metal particles (i.e., mechanical the ignition may be essentially immediate. But in
sparks, sometimes called ‘ejecta’). These the case of the most common faults—a bad con-
particles can then ignite nearby combustible nection or arcing across a carbonized path—
materials, especially if the material is low density overt ignition usually takes a long time after the
or smolder-prone. The particles can be propelled initial conditions were established for the fault.
a modest distance in a residential wiring situa- These processes are qualitatively known to be of
tion; for instance, in one study [62] particles were long term, but means for quantification do not
found up to 1.5 m away. It has also been currently exist.
documented [7] that ignition-causing particles
can be ejected from openings in a receptacle
(which can superficially appear to be undam-
Static Electricity
aged) and from within a metal box with a metal
cover (since small holes are invariably
General Principles
contained).
Static electricity represents electric charges that
are notionally static; that is, they are collected on
Miscellaneous Phenomena
a surface and are not continuously flowing in an
electrical current. The steps involved in a static
Some additional phenomena have been
electricity discharge are schematically illustrated
documented but are rare. These include harmonic
in Fig. 22.14. For significant charge separation
distortion-caused overload, eddy currents, and
(sometimes loosely called charge generation) to
dendrites [7]. Slightly less rare is the formation
of adventitious batteries, which involves a poten-
tial difference created by electrochemical action Charge separation
when an electrolyte is present in conjunction
with two dissimilar metals. This process some-
times leads to a hydrogen explosion, since the Charge accumulation
electrolysis process separates water into hydro- Dissipation

gen and oxygen [7]. Numerous studies have been of charge
published examining the possibility of very Ground

Discharge
strong radio-frequency fields causing sparks and
Ignition
6
The problem is only pertinent to small-diameter, Flammable mixture
10 AWG (2.588 mm) or smaller, conductors; service-
entrance cables and other large-diameter aluminum Fig. 22.14 Static electricity fundamentals (From Igni-
conductors can generally be reliably terminated. tion Handbook, used by permission)
682 V. Babrauskas
occur, at least one material must be involved that rank-order listing of materials according to how
is an electrical insulator. An electrical insulator is much of a negative or positive charge they tend
considered to be a substance that has a resistivity to collect [63]. The triboelectric series is deter-
above about 108 Ω-m, which includes most mined by the material’s dielectric constant [64],
organic substances. however, current understanding is that electrifi-
cation is not precluded in contact between
objects made of the same material. Thus, plastic
Means by Which Charge Separation chips falling down a chute made of the same
Occurs plastic are known to be able to undergo electro-
static charging [65]. It is believed that this may
Electric charge may be separated by the follow- involve both physical factors (e.g., stresses at the
ing means [7]: surface) and chemical factors (contamination).
1. Contact and separation or friction between Electrification due to ionized gases flowing
solids past surfaces can arise, but if the gas is at normal
2. Relative motion at a phase interface temperatures (i.e., not a hot plasma) and is not
(liquid–solid, liquid–gas, or between two liq- contaminated with solid or liquid particles, then
uid phases) the charging that can be achieved is trivial,
3. Induction (whereby charge is moved due to amounting to less than a volt [66]. Gases that
the presence of an electric field), also some- contain solid–liquid aerosols or gas streams
times termed polarization that generate liquid or solid particulates (e.g.,
4. Ion collection from a discharge process (e.g., the discharge of a CO2 extinguisher, the rapid
from corona discharge) evaporation of liquid propane) can pick up a
5. Double-layer charging sizable charge, however.
6. Fragmentation of solids having nonuniform For certain materials, moisture in the air
surface charge densities promotes the dissipation of charge since it
7. Mechanical fracture (electron emission due to decreases the electrical resistivity of some
strained or ruptured bonds within solids), also materials; it never affects the separation of
termed piezoelectrification charge. For many other materials, however, the
8. Thermal cycling (e.g., charging by freezing), resistivity is not lowered due to atmospheric
also sometimes termed pyroelectrification moisture. Adding vapor-phase moisture does
Contact or friction between two dissimilar not actually change the electrical conductivity
substances can produce a charge separation if of the air; adding a mist or spray, however, raises
either of the two substances is an insulator. The the conductivity [67].
contact may be solid-solid, solid–liquid, or
liquid-liquid. The most common modes pertinent
to fire ignitions are the following: Discharge Types
1. Contact and separation between dissimilar
solids Discharges of static electricity can involve the
2. Flowing powders following geometries:
3. Flowing liquids • Discharge between two conductive electrodes
A mild amount of charging can be created • Discharge involving one conductive electrode
simply when two surfaces come into firm contact and a diffuse insulating medium
and are then separated. Friction merely enhances • Discharge from one mist or cloud to another
the charging by increasing the effective area of Apart from events taking place solely in the
contact. Traditional wisdom is that not only must atmosphere (which are considered later in the
the materials be dissimilar, but that sizable section called “Lightning”), discharges involved
charging will take place only if they are far in accidental ignitions are classified as the
apart on the triboelectric series, which is a following:
www.ebook777.com
1. Spark about 2–6 kV is necessary for a corona discharge

2. Corona discharge to occur, with smaller potentials needed for finer-
3. Brush discharge pointed needles. The maximum energy normally
4. Powder heap discharge realizable from a corona discharge—not much
5. Propagating brush discharge (Lichtenberg over 0.01 mJ—would even theoretically suffice
discharge) to ignite only the most ignitable of gases, such as
6. Lightning-like discharge CS2 or H2. However, actual ignition requires that
Additional details on discharges are given the energy be delivered into a small kernel, and
in the Ignition Handbook [7] and by Britton [68]. the diffuse nature of a corona discharge
precludes that. Corona discharges are often used
Spark in processes and machinery as a safety measure
A normal electric spark discharge occurs for lowering charge accumulation.
through the air separating two electrodes when
the electric field reaches a value of approxi- Brush Discharge
mately 3 MV m1. Thus, for a gap distance d, When a grounded conductor is brought into an
the voltage V required is 3d, where V is in electric field that is near its dielectric breakdown
megavolts and d is in meters. For a spark to be strength, a gas discharge can occur in the form of
incendive, the gap distance normally must be a brush discharge. The discharge is able to
equal to or greater than the quenching distance. occur because of electric field distortion
Considering 2 mm as a typical quenching dis- introduced by the electrode, which locally raises
tance, the voltage required is on the order of the field above its breakdown value. The name
6 kV. Up to about 1,000 mJ can be delivered comes from the brush-like shape of the dis-
in a static-discharge spark. This is a sizeable charge. It differs from a corona discharge, in
amount of energy, well beyond the minimum that the latter is visually observed to be diffuse.
ignition energy (MIE) of most substances. A brush discharge is similar to corona discharge
Spark discharges are distinguished from other in being a low-energy, one-electrode discharge,
electrostatic discharges in that breakdown but whereas a corona discharge requires a
occurs across the whole gap separating two needlelike electrode, a brush discharge occurs
electrodes. when electrodes have a radius of 5–50 mm.
The incendivity of brush discharges is propor-
Corona Discharge tional to the radius—larger-radius conductors are
A corona discharge (sometimes called point dis- more likely to lead to ignition than ones of
charge) is a slow, diffuse discharge that smaller radius. Commonly, the high electric
originates at a metallic electrode and branches field will occur due to the presence of a charged
out in a diffuse manner into space or towards insulator. It is estimated that the energy from a
poorly conducting surfaces. A corona discharge brush discharge will not exceed about 4 mJ. In
requires an electrode that has a needle-like point, addition, most of the energy released during a
typically less than 5 mm diameter. Charging such brush discharge does not contribute to
an electrode results in an electric field which is incendivity, since the energy is not just localized
distorted, being generally low, but much greater at the place where a flame kernel is formed. The
near the point. When the electric field exceeds high electric fields necessary for a brush dis-
the local dielectric strength, breakdown occurs. charge are readily found in many powder
Corona discharge has the lowest energy of the operations, in mists, and also with movement of
electrostatic discharge types. It is visible as a plastic films.
violet glow in a darkened room. A corona dis- Circumstances leading to a brush discharge
charge can also occur in the presence of second can include:
electrode, but is still considered a ‘one-electrode’ • approaching a highly-charged insulator such
discharge since the discharge does actually reach as plastic films or plastic pipes with a finger or
the second electrode. A minimum voltage of a metal tool;
684 V. Babrauskas
• discharging of solids from plastic bags in the involved polymeric resin particles in the
vicinity of metal parts; 1–10 mm range.
• filling a tank at a high velocity with an Early recommendations used to state that up to
insulating liquid, with the charged liquid sur- 10 mJ can be delivered in a single discharge step.
face approaching an internal fitting that can Some indirect evidence suggests that discharges
act as an electrode; as large as 1,000 mJ may be anticipated for large
• lowering a conductive cup, etc., into a highly particles flowing into a large silo. A minimum
charged liquid; product feed rate is needed for a powder heap
• projection of metal parts into a cloud of highly discharge to occur. This has been estimated at
charged dust or aerosol; 3,000–5,000 kg h1 for 3 mm particles, rising to
• pouring of insulating powders into silos when 25,000–30,000 kg h1 for 0.8 mm particles [81].
the fill surface approaches a conductive fitting; Powders having a resistivity of less than 1010
• projection of ships’ masts, flagpoles, or Ω-m are conservatively judged to not be suscep-
antennas into a powerful atmospheric electric tible to explosions from powder heap discharges
field—this is known as St. Elmo’s fire. [72]; powders which have caused explosions
Even though about 3.6 mJ can be delivered in have had resistivities > 1012 Ω-m.
a brush discharge [69] and there are dust clouds
that have an MIE 1 mJ, most studies have Propagating Brush Discharge
concluded that brush discharges will not ignite A very vigorous discharge can occur when
dust clouds [70], provided that the cloud is not a certain conditions are met for the charging of a
hybrid dust/gas mixture. surface. There is a limit to the amount of surface
charging that can be sustained on a surface with-
Powder Heap Discharge out discharging by ionizing the air (2.65 105
In some cases, when rapidly filling large C m-2). This limit can be increased if a double
containers such as silos or flexible intermediate layer of charges of opposite polarity is
bulk containers (FIBC) with powders, a much accumulated.
higher charge can build up in the settled powder A way for this to occur is when an insulating
than was present in the air through which the layer is directly on top of a grounded metallic
material moved and a discharge can then take layer. This allows opposite polarity charges to
place. This occurs because a growing volume of build on the second side of the insulating layer
powder is aggregated, plus when the powder is (Fig. 22.15). Under those conditions, the
compacted, its charge likewise gets compacted if maximum surface charge is governed by the
the powder is insulating and charge cannot get breakdown strength of the insulator, which
dissipated. The powder heap discharge is also may be on the order of 20–40 MV m-1, instead
called a cone discharge or a bulking discharge. of the 3 MV m-1 for air.
It occurs along the exposed surface of the pow- In addition, the dielectric constant of many
der. A minimum particle size [71] of ca. 0.1 mm common insulators is 2–4 times that of air.
is needed for powder heap discharge to occur, These two factors combine to give maximum
but the majority of the actual incidents have surface charges of around 5 104 C m-2, and
Fig. 22.15 Double-layer

charging (charge pairing)
occurring when a charged
insulator is adjacent to a
grounded conductor
www.ebook777.com
it is considered that 2.5 104 C m-2 is the accidents have been blamed on such discharges,
minimum surface charge needed for a but the details of the circumstances have never
propagating brush discharge. With very thin been clear. Lightning-like discharges would pre-
films of certain plastics, surface charge densities sumably be able to ignite almost any combustible
up to 8 103 C m-2 have been measured [73]. matter, so the conditions—if any—that might
A propagating brush discharge can ignite lead to such lightning would be important to
most flammable mixtures, including dust clouds. quantify.
A discharge occurs in one of two ways: (1) a
grounded electrode is brought near the charged
insulator surface; or n a dielectric breakdown of
the insulating layer, resulting in a local puncture. Electrostatic Charging and Discharging
A minimum voltage of ca. 4 kV is needed for a of Solids
propagating brush discharge to occur with very
thin films (10–20 μm) rising to 8 kV for 0.2 mm For simple geometries, the amount of charge that
thick ones. Up to ca. 1,000 mJ can be delivered in an isolated solid can accumulate and the voltage
a propagating brush discharge. to which it can be charged can be computed
Circumstances leading to a propagating brush [7]. The maximum charge density, σmax
discharge can include: (C m2), that can be built up on the isolated
• conveying an insulating powder at high veloc- object is given by:
ity through plastic pipes or bins that are σmax ¼ εεo Emax
grounded on their exterior;
• conveying an insulating liquid at high velocity where ε ¼ dielectric constant (–), εo ¼ permit-
through plastic pipes that are grounded on the tivity of vacuum (8.854 1012 S s m1), and
outside, or metallic pipes that have an the abbreviation C denotes coulombs, whereas S
insulating interior coating; denotes Siemens. Since the dielectric constant
• loading of insulating powders into large, for air ¼ 1.0 and Emax, the breakdown strength
non-conductive silos; of air, is approximately 3 MV m1, then σmax, the
• high velocity operation of conveyor belts that maximum charge density possible for an isolated
have metallized outer surfaces and an object in air ¼ 26.5 μC m2. If the object is
insulating core; spherical, then its area ¼ 4πr2, where r ¼ radius
• repeated collisions of dust particles on an (m), and the maximum charge that can accumu-
insulating surface atop a grounded layer. late on it is
• In some cases, a propagating brush discharge
can occur without an overt grounded layer, for Qmax ¼ 4πr 2 26:5 ¼ 333 r 2 μC
instance when rain is falling on a plastic pipe Now, since voltage V is, by definition, equal to
conveying an insulating powder [74]. the field strength E times distance,
Lightning-Like Discharge V ¼ Emax r ¼ 3:0 106 r

Lightning in the atmosphere can occur when
water droplets and ice particles are charged to which gives the maximum voltage to which the
very high potentials. Since particles in dust spherical object can be charged in air. But capac-
clouds will also pick up an electric charge, itance C is defined as ¼ Q/V,
lightning-like discharges have been observed to Q Q
occur in the dust clouds formed during volcanic V¼ ¼ ¼ Emax r ¼ 3:0 106 r
C 4πεεo r
eruptions. What is unanswered is whether such
discharges can occur on a smaller scale, to wit, in Then the capacitance with respect to ground of
connection with storage silos. A number of an isolated sphere can evaluated as
686 V. Babrauskas
C ¼ 4πεεo r Electrostatic Charging and Discharging

of Persons
where C ¼ capacitance (farads; F), and the sym-
bol C here must not be confused with the unit C A mild shock due to electrostatic discharge is
denoting coulombs. Assuming that no losses common for human beings. Such discharges can
occur, the energy W (joules) that can be delivered also be incendive. Shoes charge the wearer
from a spark discharge is because each time the foot is raised, the capaci-
1 tance to ground is decreased and charge
W ¼ CV 2 accumulates on the person. Charging readily
2
occurs when apparel is worn that is highly
Substituting the above gives insulating and the apparel contacts and separates
from external objects. The charge picked up on
W ¼ 501r 3
the apparel then induces a similar charge in the
For many hydrocarbon vapors, a value of MIE body. Although standard values are sometimes
0.25 mJ is applicable. Then, to cause an proposed for the capacitance of a human being,
incendive discharge from a charged, isolated the actual capacitance [76] varies with the thick-
body, its radius must be at least ness of the footwear, as shown in Fig. 22.16.
The resistance of the human body [77],
1=3
0:25 103 measured to a fingertip, is about 1,300–2,000 Ω,
r¼ ¼ 0:008 m but if the person undergoes a discharge via a
501
grasped metallic object, the body’s effective
Since this limit is only 8 mm, it would be very resistance may be only 360–700 Ω.
difficult to develop a safety measure on limiting
the physical size of bodies that pick up a charge.
In practice, somewhat higher minimum sizes will Table 22.3 The estimated capacitance of some objects
pertain, since ideally efficient conditions will not Object Capacitance (pF)
be present for discharge. For a propane-air mix- Buckets, small drums 5
ture (MIE ¼ 0.26 mJ), a minimum radius of 55-gal drum 100
12 mm was found necessary in order to have an Automobile 500
incendive discharge [75]. Capacitance values for Tank truck 1,000
some common objects are given in Table 22.3. Large tractor-trailer 3,000
Fig. 22.16 Effect of 10,000

footwear thickness on the
capacitance of a person
(assumed standing on a
floor of moderate
conductivity) (From
Capacitance (pF)
Ignition Handbook, used by

permission) 1,000
100
0.1 1 10 100
Distance between bare feet and floor (mm)
www.ebook777.com
In dry air, the body can charge up to 5–25 kV, Electrostatic Charging and Discharging
although voltages toward the high end of this of Granular Materials
range are uncommon and are limited by corona
discharge. Thus, under the worst circumstances, When granular materials—powders, dusts,
the energy that is stored and is available for grains, and so on—are in motion, they can pick
release in a spark is up a charge. Insulating powders—those with a
2 resistivity greater than about 1012 Ω-m—do not
1 1
W¼ CV 2 ¼ 300 1012 25 103 easily dissipate a surface charge they may
2 2
acquire and, thus, can be prone to spark
¼ 94 mJ discharges. This is especially a problem if they
are conveyed or stored in insulating pipes or
If this much energy could be effectively applied,
silos. The tendency for powders to pick up a
it would be enough to ignite all common ignit-
charge is roughly proportional to their surface/
able vapors and also many dusts. However, more
mass ratio (or inversely proportional to particle
typical values of stored energy due to friction of
diameter). The resistivity of powders changes
apparel are 5–20 mJ. A value of 25 mJ has been
drastically with moisture. At conditions of RH
adopted by a British standard [78] as the maxi-
greater than 60–65 %, any charge formed is
mum practical value needing to be considered.
rapidly leaked away and hazardous conditions
At an RH of 50 %, a person walking on a carpet
would not be expected [82].
will generate no more than about 3 kV, and for
Discharging of dry chemical (sodium bicar-
RH greater than or equal to 60 % it is impossible
bonate or ammonium phosphate/ammonium sul-
to create a significant voltage [79]. Thus, the
fate) fire extinguishers can cause static electricity
problem is limited to dry atmospheric conditions.
buildup. It was found experimentally that this
Guidelines are also available (Table 22.4)
can result in charging voltages that would corre-
which relate the energy discharge from a person
spond to discharge energies of up to 54 mJ
to the sensation [68, 80]. A perceptible sensation
[83]. Energies of this magnitude may ignite
corresponds requires that the person be charged
many dusts, not just gases.
to about 2 kV [64]. In view of the above results,
Pneumatic transport systems cause a buildup
discharges that are perceptible but not severe are
of charge largely due to bends in the pipeline, but
likely to lead to ignition if the person is in a space
within a few meters of travel distance a steady-
containing a gas in its flammable range.
state value of charge is reached [84]. For a given
However, since people do not generally walk
air velocity, increasing the product mass flow
around in flammable atmospheres, it is found
rate decreases the charging tendency. Charging
that electrostatic discharge from humans is actu-
tendency is also reduced by reducing air velocity
ally a rare cause of unwanted ignition of
and by increasing particle size of the granular
gases [81].
material.
Electrostatic discharges commonly occur
Table 22.4 Human responses corresponding to various
levels of discharge energy
whenever granular materials are pneumatically
conveyed. These are typically nonincendive
Energy Response
corona discharges. It is the possibility of spark
1 mJ Perceptible
discharges or other more energetic forms that
10 mJ Prick
forms the crux of the fire safety problem in
30 mJ Sharp prick
these applications. Silos can build up very high
100 mJ Slight jerk
potentials when granular materials are conveyed
250 mJ Severe shock
>1J Possible unconsciousness
into them; in one study [84], up to 150 kV was
> 10 J Possible cardiac arrest measured.
688 V. Babrauskas
During loading or conveying of powders, or the interface with another liquid. If a

local non-incendive discharges (corona and pos- sufficiently high charge is accumulated, an elec-
sibly brush discharges) may occur which are tric discharge may occur. This discharge may
helpful, rather, than deleterious, since they cause an ignition under appropriate fuel-air
serve to reduce the charge buildup. Based on ratio conditions. Charge relaxation readily occurs
this observation, to reduce the incendivity of if the liquid has a high electric conductivity and
bulking discharges, it is commonly for such liquids a high charge does not build
recommended that a grounded wire be strung up. Unfortunately, many organic liquids (i.e.,
through the center of a container receiving aliphatic, aromatic, and cyclic hydrocarbons;
insulating powders. This causes small corona- ethers; some silicones) are good insulators
like discharges to occur to the grounding wire, (Table 22.5). Liquids with conductivities less
instead of large sparks to the container itself. A than 5 1011 S m1 are considered to be of
ground wire is equally effective if the container low enough conductivity that electrostatic
is insulating, instead of conductive [85]. The hazards must be carefully guarded against. How-
ground wire must be thin (around 1 mm) in ever, if the conductivity is extremely low, then
order to ensure a corona-like discharge. ionized species that could cause charge buildup
are also largely absent. Liquids with a conductiv-
ity of less than 1013 S m1 are considered to be
Electrostatic Charging and Discharging in the latter category. Thus, the peak hazard
of Liquids involves liquids with conductivities from 1013
to 5 1011 S m1. Pure hydrocarbons do not
Many liquids are prone to undergo charge sepa- exhibit electrostatic charging; however, even
ration when they move past either a solid surface impurities at the 0.001 ppm level change this
Table 22.5 Electrical conductivity of common liquids [7]

Hazard Example substances Typical electrical conductivity (S m1)
Hexane 1017
Low: Carbon disulfide 8 1016
Conductivity less than 1013 Benzene 5 1015
Heptane 3 1014
Xylene 1013
Dioxane 1013
Toluene 1012
High: Cyclohexane 2 1012
Conductivity of 1013–5 1011 Styrene 1011
Kerosene 1.5 1011
Hexamethyldisilazane 2.9 1011
Jet-A fuel 2–3 1011
Gasoline 1010
Turpentine 4 1010
Crude oils 109 to 107
Halogenated hydrocarbons 108
Low: Methyl alcohol 107
Conductivity greater than 5 1011 Ethyl alcohol 1.4 107
Cetones 105
Water: deionized 105
Iso-propanol 104
Water: acid rain 102
www.ebook777.com
situation [86]. The presence of a small amount of buoyed up by turbulence or by foaming of the
water in the product can increase the electrostatic product. The combination of filtering and splash
charging effect up to 50-fold. loading was found to be highly conducive to
For insulating liquids flowing in conducting discharges; removing either of the two factors
pipes, the charge density (C m3) picked up is greatly diminished the potential for a discharge.
linearly proportional to the liquid’s flow velocity, In low-viscosity liquids, splash loading by itself
and the charge density reaches a steady-state value produces a charged foam in the tank which can
after a certain distance down the pipe, with the lead to discharges even when the inflowing liquid
charging voltage being roughly proportional to has little charge on it.
the flow velocity. However, when liquids flow in Liquid sprays can cause intense static electri-
insulating pipes, little streaming current occurs fication; this was first observed near waterfalls in
because the charge induced in the pipe walls does the nineteenth century and is called spray electri-
not get dissipated to ground. Flows that consist of fication [90]. The process occurs due to the pres-
two-phase liquids, liquids with suspended solids, ence of a double layer of charge at the liquid
or mixtures of immiscible liquids tend to build up surface. Small pieces of material in the form of
higher charges than single-phase liquids. Charge droplets removed from the surface can then pos-
buildup can especially increase if the liquid flows sess an unbalanced charge. By this process, a
through a fine-pore filter. water stream may ignite a flammable atmo-
When a low-conductivity liquid is in motion sphere, and this concern arises in operations, for
in a conducting pipe, not just separation but also example, where a water spray is used to clean
an actual flow of charge occurs. This flow is equipment that was used to store flammable
called a streaming current and it arises because liquids.
ions in the liquid tend to move with the flow,
while the opposite charge on the wall dissipates
to earth. For it to occur, the liquid must have a
conductivity in the range of about Lightning
1013–107 S m1. Experimental studies [87]
indicate that, for these liquids, the streaming Electrical Characteristics
current I (amperes) can be estimated as
Lightning becomes possible because electric
I ¼ 9:42 106 ðuDÞ2 charge can become separated and accumulated
in clouds. Clouds are highly complex entities
where u ¼ velocity (m s1) and D ¼ diameter (m).
and, even today, the physicochemical environ-
An empirical expression [88] for the charge
ment of clouds is by no means comprehensively
density is
understood. One theory by Ermakov and
000
Q ¼ 5 107 u=D Stozhkov [91] is illustrated in Fig. 22.17.
Thundercloud formation begins when a cold
Moisture and impurities can greatly increase air mass meets a warm one. Ionized, warm, moist
the charge density, but experiments have to be set air rises but is then progressively cooled at higher
up carefully to illustrate this. The effect is not elevations and condensation of water vapor on
found unless the liquid is pumped through filters nucleation centers begins. In the initial phase,
[89]; charge generation associated with tank- condensation proceeds faster on negatively
loading in the absence of filtering does not charged nuclei than on positive ones, and the
show deleterious effects of impurities. Metallic upward air flux produces large-scale separation
trash in tanks can act as ‘charge collectors’ and of charge and a resultant electric field. The latent
greatly reduce the charge density necessary to heat released in condensation assists the buoy-
cause an electrostatic discharge [89]. The effect ancy of the upward air current. Cosmic rays
takes place since small metallic objects can be produce ionized showers of particles.
690 V. Babrauskas
J–
6 5
5
1 4 1 2
2
3
3
7
8
10
Initial phase Maturity Decay
Fig. 22.17 Formation of thunderclouds as described by lightning; 7, cloud-to-ground lightning; 8, ground-to-

Ermakov and Stozhkov [91]. 1, warm air front; 2, cold air cloud lightning; 9, negative screen layer; 10, positive
front; 3, ascending flux of wet ionized air; 4 and 5, exten- charge at cloud base; J current of negative ions from
sive air showers produced by primaries with energies over the ionosphere to the top of the cloud
1014 eV or 1015 eV, respectively; 6, cloud-to-cloud
When the electric field exceeds 0.2–0.3 stepped leader induces a positive charge in the
MV m1, electron avalanches occur; ionized earth below. Protruding grounded objects start to
tracks link with each other and form a conducting conduct heavier point-discharge currents.
tree structure, which allows cloud-to-cloud A streamer then arises from one of these
discharges to occur. In the mature phase, droplets objects or from the earth itself, connects to the
grow and coagulate. There are ascending and leader, and starts a return stroke. The return
descending airflows and the cloud becomes stroke is the brightly visible lightning stroke.
asymmetric with an excess of negative charge After the first return stroke, a dart leader may
at its base. The electric field between the cloud descend directly to the ground without stair-
and earth’s surface increases, leading to cloud- stepping. This dart leader is ball shaped. It will
to-ground discharges. Thunder is an acoustic be followed by a second return stroke. There may
shock wave that originates in the gas breakdown be three or four, but occasionally many more,
region and then propagates out through the air. strokes per the total event, comprising the light-
The origins of a lightning strike are due to a ning flash. The total lightning flash may last from
separation of charges in clouds. Lightning becomes 0.01 to 2 s, with 0.2 to 0.4 s being typical, but
possible when a potential of 10–100 MV each individual stroke only lasts about 30 μs. The
with respect to the ground has been reached. A interval between strokes may be around 40 ms.
lightning flash is composed of several events. The current carried by the stepped leader is
The actual discharge begins with the formation small, only on the order of 100 A. But each return
of the first stepped leader, which is a localized stroke will typically carry 10–20 kA of current,
gas breakdown of about 50 m length. The process and peak currents in excess of 100 kA are occa-
continues in a stair-step fashion until a leader sionally recorded. A cloud-to-ground stroke may
gets to within about 50 m of the ground (or an discharge about 25 C per stroke. The average
object on the ground). The negative charge of the length of a stroke is 3 km, and the average energy
www.ebook777.com
released is 105 J m1, making an average energy Ignitions from Lightning

release of 3 108 J per stroke [92].
In electrical circuit terms, a lightning stroke The primary damages [95] to residences from
can be considered a constant-current source. lightning are the following:
Therefore, the energy dissipated in the object • Brick, concrete, and other solid surfaces
along its path is moved or cracked
ð • Plumbing pipe punctures
I 2 Rdt • Holes burned or punctures in roofs
• Arc damage to metal structures such as win-
where I ¼ current (A), R ¼ resistance (Ω), and dow frames
t ¼ time (s). This accounts for the much greater • Arcs across wiring
damage found for objects of poor conductivity The last three of these, of course, may also be
than for metals. Design values for the current of accompanied by ignition of combustible
100–200 kA are commonly used [93] since only materials. Because the temperature rise in an
about 1 % of lightning strokes give currents in object is proportional to its resistance, a metallic
excess of 200 kA [92]. object (e.g., a lightning rod) may sustain limited
The current from a second or subsequent stroke temperature rise, whereas a poor conductor such
is typically less than half that of the first one. When as wood may become ignited.
a lightning strike occurs, nearby metallic objects Multiple ignitions from a single strike are not
can have a current induced in them, including not rare. Whether combustibles will be ignited from
only electrical wiring but also other metallic a lightning flash or not depends critically on
objects such as building beams. The electric field whether there is a flow of continuing current in
induced by a lightning strike of a known current the channel after the stroke. About 25–50 % of
value can be estimated [94] according to: lightning strikes exhibit this characteristic—
these are sometimes called hot bolts. Lightning
I
E ¼ 33 strikes that are positive (i.e., the cloud being
r
positive with respect to the ground) are much
where E ¼ electric field (V m1), I ¼ current rarer than the converse, but these are precisely
(kA), and r ¼ distance from strike (km). Thus, the ones that are most likely to cause ignitions,
for example, a strike with a current of 100 kA is since their peak currents and total charge transfer
expect to induce a field of 1,100 V m1 at a are much larger. Positive flashes do not have the
distance of 3 km. stepped-leader characteristic of the common,
A direct strike to a building or structure is one negative strikes, and consist of a single stroke,
where the ground-side termination of the light- followed by a period of continuing current flow.
ning bolt attaches to any part of the building. The probability [96] of igniting a house fire
A side flash is a strike where the lightning bolt from a lightning strike is much higher if the
terminates on a nearby object, but a secondary house has plastic plumbing pipes as opposed to
flash occurs from that locale to the building. metallic ones. This is because the lightning
A side flash is considered to also be a type of current may flow to ground through a metallic
direct strike. An indirect strike is one where the pipe network, but if electric wiring is the only
main bolt terminates elsewhere, but some energy substantive metallic path, the current is likely to
from the bolt is delivered to the building by go through electric wiring, where heating will be
means of power lines, metallic underground much greater due to the smaller area of the
pipes, or other conductive paths. conductors.
692 V. Babrauskas
Safety Measures Against Lightning Franklin recommended that each air terminal
provides a downward ‘cone of protection,’ with
The lightning rod was invented by Benjamin the cone’s angle from vertical being 58 . The
Franklin in 1752. This is a metallic rod which is origin of this recommendation is unclear and it
grounded at one end and raised in the air at the was evidently not evolved from experiments.
other. When the initial streamers from the cloud Subsequent experience suggested that this angle
start to form, there is not a highly specific place is much too large, and during the nineteenth
along the ground level where the initial return century the recommendations slowly went down-
stroke is preferentially located. By providing a ward to about 30 . However, even using a 30
ground-potential conductor in the air, a prefer- angle of protection, failures were documented
ence is established, and the lightning current [100]. In more recent times, Lee [100]
flows down the rod (which must be of adequate synthesized a design method for protecting
dimension in order not to overheat). buildings, evolved from advanced calculations
Franklin recommended that the tip of the rod used by electric utilities for protection of power
(air terminal, in the jargon of the lightning pro- lines. The method is applicable only to structures
tection industry) be pointed, because this leads to of 45 m height or less and is described in the
a point discharge. In earlier times, this point following way: Imagine a rolling sphere of 45 m
discharge (corona discharge) was considered radius (Fig. 22.18). The sphere starts rolling
necessary to ‘attract’ the lightning stroke. More along the ground from a distance far away from
recently, experiments showed that a smoothly the structure in question, then roll up to and over
rounded tip is more successful in attracting light- the structure and its protective air terminal(s).
ning to itself and avoid strikes to nearby objects If the sphere only ends up touching the air termi-
[97]. The first comprehensive engineering guide nal(s) and the ground and cannot come into con-
to proper installation of lightning protection tact with the structure to be protected, then air
systems was published by Anderson [98] in terminals of sufficient height and quantity have
1879 and, perhaps surprisingly, few of his been erected, otherwise additional protection is
recommendations have been overturned by needed. This more realistic protection concept is
more modern research. Müller-Hillebrand [99] more liberal than the 30 fixed-angle scheme for
reviewed some of the early concepts of lightning low structures and more conservative for high
protection. structures. The 45 m dimension is used because
A lightning protection system basically it corresponds to the typical length of the stepped
comprises three main components: leader, which is about 50 m.
• air terminals In view of the enormous currents of around
• downconductors 20,000 A that are involved in a lightning strike, it
• ground terminals. is perhaps surprising that gigantic-size
Fig. 22.18 Sphere of protection from an air terminal: structure A is protected, structure B is not
www.ebook777.com
downconductors are not required to safely and it can be noted that the actual length of the wire
conduct the electricity. The minimum size of sensibly cancels out of the equation. This gives
conductor does not need to be huge since the A2 ¼ 1.4 1010 m4, or A ¼ 1.18 105 m2 ¼
flow of current is so brief—much less than 11.8 mm2. If the temperature criterion were the
1 s—and heating is proportional to the time of melting point of copper (1,083 C), then A ¼
current flow. The simplest model that can predict 4.9 mm2, using the same ‘action integral’ of
this is the adiabatic lumped-capacitance model: 5 106 A2 · s.
In US practice, however, a more conservative
W ¼ I 2 Rt ¼ ρ C p V T f T o approach is taken, with conductor areas being
greater than 11.8 mm2, in order to allow for
where W ¼ energy (J) flowing into a piece of
mechanical damage, some unusually potent
metal, I ¼ current (A), R ¼ resistance (Ω), t ¼
lightning strikes, etc. The most commonly
time (s), ρ ¼ density (kg m-3), Cp ¼ heat capac-
followed guidance is that published by NFPA
ity (J kg1 K1), V ¼ volume (m3), Tf ¼ final
[101]. NFPA 780 divides structures into two
temperature ( C), and To ¼ initial temperature
Classes, Class I being those up to 75 ft
( C). This assumes that the current flow is con-
(22.9 m), with Class II being those higher. For
stant over the time t; if the current flow is vary-
Class I structures, the required downconductor
ing, then the expression becomes:
area is 29 mm2 for copper and 50 mm2 for alumi-
ð
num. The cross-sectional area required for cop-
W ¼ R I 2 dt ¼ ρ C p V T f T o
per conductors in Class II service is 2 that for
Class I. Generally, stranded or braided
where R is assumed to be time-invariant and has conductors are used, to minimize loss of
been taken outside the integral. Applying the current-carrying capacity due to skin effect (this
above relation to copper wire, ρ ¼ 8,890 kg m3, electromagnetic effect pertains to transient cur-
Cp ¼ 385 J kg1 K1, V ¼ A · L, where A ¼ rent flows and leads to current flowing dispropor-
cross-sectional area (mm) and L ¼ length (m). tionately near the surface). A lead coating is
The resistance of a copper wire can be often used to minimize loss of metal due to
expressed as: corrosion from flue gases. Properly-installed
L and maintained lightning protection systems are
R ¼ ρe highly effective, with one report [102] quoting
A
old US studies giving 99.3 % and 99.9 % effec-
where ρe ¼ electrical resistivity of copper tiveness values.
¼ 1.7241 108 Ω · m. The initial temperature For configuration of the air terminal, modern
To can be taken as 20 C, while the final (allow- studies by Moore et al. [103] concluded that the
able) temperature Tf must be set to some reason- optimum tip-height to tip-radius ratio is about
able value below the melting point. Since the 680. Thus, a rod erected at 10 m height ought to
downconductor may come into contact with have a tip of 14.7 mm radius.
combustibles such as dry leaves, it seems appro-
priate to limit Tf to 200 C. Based on studies of
ð
lightning discharges, I 2 dt ¼ 5 106 A2 · s is Ignition and Values of Voltage,
Current, or Power
commonly used. The equation can then be
evaluated as:
Are there minimum values of voltage, current, or

8 L
power that must be exceeded for ignition to be
1:7241 10 5 106 possible? Sometimes the stance is taken that,
A
under certain circumstances, an energized
¼ 8890 385 ðA LÞ180 electrical device cannot be the source of a fire.
694 V. Babrauskas
This stance is usually couched in terms of power

but has also been couched in terms of current or Electrical Explosions
voltage. For example, some authors [104] claim
that ignition is not possible for devices that can- Most explosions in which electricity plays a role
not consume more than 20 W. Certain UL can be grouped into three categories:
standards [105] consider the limit to be 15 W. 1. Pure arc explosions
A more conservative view [106] is “several 2. Pure fuel explosions
watts.” Such a limit would be attractive in that 3. Mixed mode explosions.
it can simplify such standards, but there does not In a pure electrical arc explosion, the main
exist a research basis that would support these source of explosion energy is the arc itself. Com-
notions. bustion, i.e., oxidation, plays only a modest role,
Instead, research findings indicate that minus- if any. In the case of copper or steel conductors,
cule values may suffice, ones that are so low as to oxidation is generally negligible, while in the
not form a useful criterion. For example, some case of aluminum conductors some of
flammable gas mixtures can be ignited by arcing the exothermicity comes from oxidation of the
from a resistive circuit having only 4–5 V, aluminum.
whereas in inductive circuits, a 0.5 V power For such explosions to be damaging, signifi-
supply may suffice [7]. In terms of power, it has cant arc power is normally required. Thus pure
been documented that breaking an 0.95 W incan- arc explosions are mostly of interest in high
descent lamp can suffice to cause an explosion of voltage transmission and distribution equipment,
a methane/air mixture, whereas a broken 3 W in industrial facilities, or in commercial
lamp can cause an explosion of coal dust in installations where significant short-circuit
air [107]. currents are available.
Although solids are generally harder to ignite Pure fuel explosions are ones where the only
than gases or dust clouds, very limited testing role of electricity is to provide a spark or a
indicates that exceedingly low values can also low-energy arc, with the explosion energy
suffice for solids. For example, a Dacron com- corresponding to burning of fuel. The most com-
forter was ignited from a 6 W night-light lamp, mon example is spark from an electrical switch
bamboo decorations from a 25 W lamp, and causing an explosion in a house which filled with
cellulose attic insulation from a 50 W lamp natural gas. This category also includes electri-
[7]. Notebook paper was ignited [7] from a resis- cally initiated explosions of solid or liquid
tor rated at 1/8 W that was dissipating 1.25 W at explosives, for example, by use of an exploding
the time. Power values associated with bridgewire. Pure fuel explosions are treated in
arc-tracking ignitions have rarely been explored, depth in the Ignition Handbook [7] but are out-
but in one study it was found that 4.8 W sufficed side the scope of this Chapter. In addition, there
to ignite PBT plastic [108]. This should be are numerous monographs which discuss the
interpreted in the context that PBT is one of the physical, chemical, or civil engineering aspects
more arc-tracking resistant plastics and plastics of explosions of all types [109–113].
more prone to arc-tracking (such as PVC) would Mixed mode explosions are explosions which
presumably require less. are initiated by an electric spark, arc, or hot
Finally, it must be noted that the values cited surface, and where the fuel was delivered due to
above are values that sufficed to cause an ignition an electrical fault. The best known mixed mode
or explosion during limited testing. These values explosions are explosions of oil-filled
are not intended to imply lower-value limits. transformers and explosions of underground
Ignitions, in fact, may occur under conditions or electric distribution cables, including manhole
regimes not encompassed in the research that has explosions. In these explosions, an electrical
been published to date. fault gasifies the dielectric liquid or solid,
www.ebook777.com
which is then ignited by a localized source of centers. The best-known incident was in an
energy. Mixed mode explosions are also experi- Atlanta high-rise building that took place on
enced when lead-acid batteries explode and in 30 June 1989. The fumbling of an electrician
other cases of adventitious generation of hydro- replacing a fuse caused a 480 VAC bus duct
gen. In recent years, explosions of residential explosion [121] and the explosion and
dishwashers have been identified as an additional subsequent fire led to five fatalities.
example of mixed mode explosions. (Explosions Lightning strikes can lead to arc explosions in
in a purposive hydrogen-production facility any type of premises. In 1773, Lind
initiated by an electric spark are categorized as demonstrated that if a conductor from a lightning
pure fuel explosions, since an electric fault does arrester is run down through a house, but with a
not play a role in generating the fuel.) small gap in this conductor, this can form a spark
gap and a strike to the arrester can result in an arc
explosion capable of destroying the house
Basic Phenomena [122]. Individuals have been bodily knocked
over when in proximity both to electrical fault
Electric arc explosions involve complicated phe- arcs and lightning strikes, although interestingly
nomena, and none of the standard monographs often there have been negligible injuries to the
on explosions cover this specialized topic. The individual knocked over [123]. But in cases
only review of the topic has been by Babrauskas where roofs collapse, the outcome may be trau-
[114], and here some of the main findings will be matic if persons are present underneath.
summarized. Eardrum rupture can be expected at explosion
Electric arcing in circuits with sizable maxi- overpressures of 19 kPa (10 % probability) or
mum short-circuit current capacity can be a 45 kPa (50 % probability), while death due to
highly energetic effect. In fact, buildings have lung damage is 120 kPa (10 % probability) or
collapsed due to arc pressure, since in an enclosed 141 kPa (50 % probability). The above values
space some surprisingly large pressures can be come from an extensive statistical study by
built up. For instance, in one test explosion Eisenberg et al. [124]; older data are somewhat
overpressures of 83 atm were obtained. The mag- different, but not greatly. In any case, they indi-
nitude of this can best be appreciated by consid- cate that it does not take large overpressures for
ering that a fuel-air deflagration will typically injury or death to result from explosion pressures.
attain only around 7–8 atm, barring pressure- An arc explosion arises due a very rapid
piling effects or other turbulence enhancements. heating of air or other medium. In the process,
During normal operation of a circuit breaker, arc electrical energy is converted into other forms of
pressures of roughly 3 atm magnitude can be energy: dissociation, ionization, and heating of
expected [115], but these devices are designed the gas, including its compression; thermal radi-
to sustain the pressures generated by the normal ation; and conduction losses into adjacent solids
arcing associated with circuit opening. such as electrodes.
Arc explosions are not rare in industry, and in In addition, some electrode metal is vaporized
other situations where 480 V, or higher, voltages and this contributes to the total volume which is
are utilized, but published case histories are being explosively heated, yet, the role of chemi-
scarce. Neither of the two large electrical acci- cal reactions has only recently been explored.
dent compilations [116, 117] mentions the sub- When an arc breakdown is initiated, energy gets
ject. Lee [118] published four brief case deposited into the arc channel at a rate much
histories, Crawford et al. [119] documented greater than can be removed from the area by
seven case histories of arc explosions inside the shock wave that is created. This causes a
motor terminal boxes, including one fatality, rapid pressure rise and, if the arc energy is suffi-
while Heberlein et al. [120] described two ciently high, this will be perceived as an
non-fatal explosions inside motor control explosion.
696 V. Babrauskas
For a low-energy arc, the perceived sound may wave does not constitute a detonation, which
simple be a ‘snap,’ ‘crackle,’ or ‘pop.’ But within would require that the shock wave be supported
the scientific community there is not an agreed- by an exothermic reaction occurring behind the
upon, quantitative definition of the term ‘explo- shock front. For subsonic sound waves in air, the
sion,’ nor are there studies to quantify the fraction decay in pressure with distance from the source
pffiffi
of the arc energy that gets delivered as sound goes as 1= r for the infinite-cylinder geometry
energy, i.e., vibrations in the 20–20 kHz range. and 1/r for the sphere. But for shock waves, these
In an open environment, arc pressures will simple wave-equation relationships are not
rarely be highly destructive. Theoretical applicable.
modeling suggests that very high pressures may A reasonably short arc will be represented by
be created, but experimental studies do not bear a short cylinder, but this is not a geometry that
this out, which only show overpressures below lends itself to simple theoretical solutions. Baker
about 10 kPa. However, if arc explosions occur presented calculated data on a point sources
within enclosures which are sealed, or nearly so, [125], along with experimental data on bursting
then, as mentioned above, huge overpressures explosions of short cylindrical vessels and spher-
may be found. ical vessels. His results showed that, unless only
Electric arc explosions are not combustion examined over small intervals, the actual rela-
phenomena—they are predominantly physical tionship is not even of a power-law type.
explosions, due to very rapid conversion of elec- In an electric arc, when breakdown is
trical energy into heat. Chemical reactions play a initiated, a narrow conducting filament first
role, but only a supporting role, in such bridges the gap, and then grows rapidly in diam-
explosions. Recent studies suggest that chemical eter until it reaches an ultimate value and the ‘arc
reactions are mainly ones which convert air to channel’ is fully established. In the course of this,
species such as O3 and NOx. While these may there are two disturbances that a propagated: a
comprise oxidation, they are very different from sound wave propagating at the speed of sound
a fuel-air explosion of a normal sort. Some of the and a shock wave propagating at two or three
electrode metal is also vaporized in an arc explo- times that.
sion, and arc temperatures are high enough so Flowers [126] made detailed measurements
that presumably much of this metal vapor may and found velocities of 1,000–2,000 m s1 for
get oxidized. However, the electrode oxidation the radial expansion of the arc channel; these are
effect is quantitatively only a small part of the well above the 340 m s1 of sound speed,
heat balance in an arc explosion. Thus, first-order indicating that shock waves are being generated
estimates of arc explosions treat the process in all cases. The arc channel eventually reaches a
solely as converting electrical energy to heat steady-state diameter and no longer expands, and
and ignore chemical reaction contributions. Flowers found that a time of 3–35 μs was
required for the final diameter to be attained in
his experiments, but the actual value is depen-
Shock Waves from Electric Arcs dent on external circuit parameters which limit
the current growth rate.
If in a gaseous medium there is an abrupt change When the final diameter is achieved, Flowers
in pressure, temperature, or volume created at found that the arc channel cross-sectional area is
some location, a wave will be generated which linearly proportional to the current, with the
will propagate through the medium. The wave proportionality being 11 A mm2. Later
can be a sound wave, a shock wave, or both, Vanyukov et al. showed that the expanding
depending on the characteristics of the source. shock front and the channel are initially of the
In the case of an electric arc, while a shock wave same diameter, but subsequently the channel typ-
will be generated and it is audibly perceived as an ically approaches a maximum diameter, while
explosion (unless of very small scale), the shock the shock front continues expanding outwards
www.ebook777.com
Fig. 22.19 Results of Vanyukov et al. on the expansion of the arc channel and growth of the shock wave
[127] (Fig. 22.19). Intermediate between these is diatomic gas, giving γ ¼ 7/5 ¼ 1.4. This equa-
the ‘envelope,’ which is the boundary between tion predicts that the pressure rise will vary line-
compressed gas (outside the boundary) and arly with P_ and decrease with distance
rarefied gas (inside). In some cases, however, proportionally to 1/r.
arc channel diameter growth continues for a The only extensive series of experiments
protracted period, especially at higher current available to examine the relationship between P_
values [128]. and the pressure rise has been that of Drouet and
Nadeau [129]. Unfortunately Fig. 22.20 suggests
that there was a systematic error in the
Pressures from Arcs in the Open measurements. The slightly different than
expected slope for long arcs could well be due
Baker [125] treated arc explosions in the open to a limitation of theory, but the short-arc results
using results from acoustical theory. The pres- are most likely erroneous, not only because they
sure rise Δ p ¼ ð p1 po Þ is assumed to be due to have a dissimilar slope, but mainly because they
the arc effectively generating a certain volume imply that in excess of 100 % of the power of the
V of air at the ambient density ρo. Then from arc is realized as compression of air.
acoustical theory, the pressure rise at any partic-
ular distance r (m) from the arc is:
γ 1 P_ Pressures from Arcs in an Enclosure

Δp ¼
γ 4π r po
If an arc discharge occurs within an enclosure
where P_ ¼ first derivative of electrical power which is of modest size, then the whole enclosure
developed in the arc (W s1), and γ is defined as: will get measurably pressurized. In addition, of
γ ¼ c p =cv . Air can assumed to be an ideal course, there will be local shock wave propagation,
698 V. Babrauskas
Fig. 22.20 The results of

Drouet and Nadeau
compared to Baker’s theory
and also reflections of shock waves from compart- Consistent with the theoretical prediction, when
ment walls. Modeling such details in the system using very small enclosures and large arc
would require numerical calculations. But, for currents, some exceedingly large pressures can
many practical purposes, what is of most interest be obtained. Graneau [122] conducted
is the peak, quasi-steady overpressure that is experiments in a tiny cubical cavity, 12.7 mm
achieved, and this can be approximated in a simple on a side (2 106 m3), with the electrode gap
way. Neglecting all transient and hydrodynamic also being 12.7 mm. For a spark discharge of
effects, the discharge of an arc in a single, closed 40 kV and a peak current of 38 kA, he measured
compartment can be treated as an ideal gas within a an average pressure of 409 atm in the cavity, with
isolated, isochoric (constant volume) system. If an even higher pressures during the peak of the
amount of heat or energy ΔW is injected into the discharge. Baldrey and Hudson [131] conducted
volume, the change in pressure Δp is [130]: tests within a small pressure vessel and got an
overpressure of 83 atm in the worst case. Numer-
R ΔW
Δp ¼ ous other studies have also been published [114].
Mcv V The analysis is complicated since not all of the
where M ¼ molar mass, R ¼ universal gas con- electrical energy delivered actually goes into
stant ¼ 8.314 J mol1 K1, and cv constant- heating the gas. Conversely, in the case of alumi-
volume heat capacity (J kg1 K1). Using the num electrodes, an additional energy term comes
relations between cv, cp, and γ, the relation is from the combustion of the aluminum. Thus,
more usefully written as: because of differing experimental conditions,
published results tend to lack generality. An addi-
ΔW tional complication is that many studies were done
Δ p ¼ ð γ 1Þ
V not on fully sealed enclosures, but on ones with
This shows that, all else being equal, the over- certain small openings. This makes the results
pressure is inversely proportional to the volume. highly dependent on the specifics of the geometry.
www.ebook777.com
Fig. 22.21 Arc pressure

rise, as function of
electrode material,
measured by Tanaka
et al. in an 0.32 m3
enclosure for 0.1 s duration
AC arcs
In addition, many studies were at very low measuring a peak of 0.26 atm. Other researchers
overpressures. Such results are more readily [136] obtained reasonable agreement, but only
amenable to analysis, but do not necessarily by using a custom-designed CFD code.
reflect on more damaging explosion incidents.
An example of results where at least moderately
high overpressures were achieved is the study of Summary
Tanaka et al [132]. Figure 22.21 shows their
results obtained for AC arcs of fixed 0.1 s dura- Fires arising due to static electricity or electric
tion. The higher values obtained for aluminum current can be difficult to understand for the
electrodes reflect the contribution of the combus- nonspecialist because, apart from fire science,
tion of electrode metal. the separate discipline of electrical science is
Efforts have been made to use commercial invoked. Partly because of this discipline-
CFD codes for computing pressure rises in straddling nature of the phenomenon, research
enclosures, but the validity of such efforts has has not been as vigorously pursued as in some
been problematic. Capelli-Schellpfeffer et al. other aspects of fire science. But at the present
published two CFD modeling studies intended time, the fundamentals have been well enough
to simulate an experimental arc fault in established that general guidance can be given.
480 VAC switchgear in which an arc energy of The ways by which the “heat” leg of the fire
20 kJ was delivered and a peak overpressure of triangle is produced in electrical fires has been
2.6 atm was measured [133]. In the first study outlined in this chapter. The most highly
[133], a peak of 0.2 atm was computed, while in specialized of these mechanisms is the electrical
the second [134] a peak of 16,000 atm was arc, and its characteristics have been reviewed in
predicted. Interestingly, the same 20 kJ value of this chapter. Electric arc explosions are a highly
arc energy was used by Caillard et al. [135] but in specialized phenomenon and are primarily phys-
a capacitive discharge circuit and they predicted ical explosion, rather than chemical, although
a peak overpressure of 1,700 atm while some chemical reactions may occur.
700 V. Babrauskas
ρe Electrical resistivity of copper (Ω · m)

Nomenclature σ Charge density (C m2)
τ Time constant (s)
A Area (m2)
C Capacitance (F)
Cp Heat capacity, constant-pressure References
(J kg1 K1)
Cv Heat capacity, constant-volume 1. J.A. Frank, “Characteristics and Hazards of Water
(J kg1 K1) and Water Additives for Fire Suppression,” in Fire
Protection Handbook, 19th ed., National Fire Pro-
c1 Constant (-) tection Association, Quincy MA, pp. 10-12–10-15
c2 Constant (-) (2003).
D Diameter (m) 2. J.F. Casey, “Handling Utility Fires,” in The Fire
d Distance (m) Chief’s Handbook, 4th ed., Technical Publishing
Co., New York, pp. 264–270 (1978); note that this
d1 Clearance distance (m) material does not appear in more recent editions of
d2 Creepage distance (m) the handbook.
e Charge of the electron (C) 3. T. Verlo, “The Use of Water as an Extinguishing
E Electric field (V m1) Agent for Live Electrical Installations” (Report EFI
TR A3866), SINTEF EFI, Trondheim, Norway
h Heat transfer coefficient (W m2 K1) (1991).
I Current (A) 4. Babrauskas, V., How Do Electrical Wiring Faults
L Distance (m) Lead to Structure Ignitions? Fire and Arson Investi-
M Molar mass (kg/mol) gator 52:3, 39–45, 49 (Apr. 2002).
5. Babrauskas V., Research on Electrical Fires: The
p Pressure (Pa) State of the Art (The Emmons Plenary Lecture),
P Power (W) pp. 3–18 in Fire Safety Science—Proc. 9th Intl.
P_ First derivative of electrical power Symp., Intl. Assn. for Fire Safety Science, London
(W s1) (2009).
q00 Power density (W m2) 6. Babrauskas, V., Electrical Fires: Research Needed to
Improve Fire Safety, Fire Protection Engineering
Q Charge (C) No. 46, 20–22, 24–26, 28–30 (2nd Q. 2010).
Q000 Charge density (C m3) 7. V. Babrauskas, Ignition Handbook, Fire Science
r Radius (m) Publishers/Society of Fire Protection Engineers,
R Issaquah, WA (2003)
Resistance (W)
8. Somerville, J. M., The Electric Arc, Methuen,
R Universal gas constant (8.314 J mol1 London (1959).
K1) 9. Uman, M. A., Lightning, Dover Publications,
t Time (s) New York (1984).
T 10. Braginskii, S. I., Theory of the Development of a
Temperature (K)
Spark Channel, pp. 188–200 in Electrical Break-
u Velocity (m s1) down in Gases, J. A. Rees, ed., MacMillan, London
V Potential difference (“voltage”; V) (1973).
V Volume (m3) €
11. Paschen, F., Uber die zum Funkenübergang in Luft,
W Energy (J) Wasserstoff und Kohlensäure bei verschiedenen
Drucken erforderliche Potentialdifferenz [On the
α Townsend’s first ionization coefficient, Required Potential Difference for Spark Discharges
(m1) in Air, Hydrogen, and Carbon Dioxide at Various
ε Dielectric constant (–) Pressures], Annalen der Physik 37, 69–96 (1889).
εo Permittivity of vacuum (S s m1) 12. T.W. Dakin et al., “Breakdown of Gases in Uniform
Fields—Paschen Curves for Nitrogen, Air, and Sul-
γ Ratio Cp/Cv (—) fur Hexafluoride,” Electra (CIGRE, Paris), 32, pp.
γ Incomplete gamma function (–) 61–82 (Jan. 1974).
γ Townsend’s second ionization 13. T.W. Dakin, “Insulating Materials—General
coefficient (-) Properties,” in Standard Handbook for Electrical
Engineers, 13th ed. (D.G. Fink and H.W. Beaty,
λ Thermal conductivity (W m1 K1) eds.), McGraw-Hill, New York, pp. 4-117–4-160
ρ Density (kg m3) (1993).
www.ebook777.com
14. A.E.W. Austen and S. Whitehead, “The Electric Hydroxide and Aluminium Oxide,” in 4th Interna-
Strength of Some Solid Dielectrics,” in Proceedings tional Conference on Dielectric Materials,
Royal Society A, 176, pp. 33–49 (1940). Measurements and Applications, IEEE, New York,
15. A.B. Lewis, E.L. Hall, and F.R. Caldwell, “Some pp. 295–298 (1984).
Electrical Properties of Micas,” Journal of Research 31. N. Yoshimura, M. Nishida, and F. Noto, “Light
NBS, 7, pp. 403–418 (1931). Emission from Tracking Discharges on Organic
16. Y. Abed, “Experimental and Digital Determination Insulation,” IEEE Transactions on Electrical Insula-
of Breakdown Voltage of Different Synthetic tion, EI-19, pp. 149–155 (1984).
Insulating Materials,” in Conference Record of the 32. K. Oba, “Identification of Melting Marks of Electric
1982 I.E. International Symposium on Electrical Wires” (unpublished report), Yamagata Prefecture
Insulation, IEEE, New York, pp. 123–126 (1982). Police Headquarters, Criminal Scientific Laboratory,
17. J.H. Mason, “Breakdown of Solid Dielectrics in Japan (1980).
Divergent Fields,” IEE Proceedings, 102C, 33. T. Bernstein, “Electrical Fires: Causes, Prevention,
pp. 254–263 (1955). and Investigation,” in Electrical Hazards and
18. A.E.W. Austen and H. Pelzer, “The Electric Strength Accidents: Their Cause and Prevention
of Paraffins and Some High Polymers,” Journal of (E.K. Greenwald, ed.), Van Nostrand Reinhold,
IEE, 93, pp. 525–532 (1946). New York, pp. 116–134 (1991).
19. V. Babrauskas, “Mechanisms and Modes for Ignition 34. M.J. Billings, A. Smith, and R. Wilkins, “Tracking in
of Low-Voltage, PVC-Insulated Electrotechnical Polymeric Insulation,” IEEE Transactions on Elec-
Products,” Fire & Materials (2005). trical Insulation, IE-2, pp. 131–137 (Dec. 1967).
20. J.D. Cobine, Gaseous Conductors: Theory and Engi- 35. K.N. Mathes and E.J. McGowan, “Electrical Insula-
neering Applications, Dover, New York (1958). tion Tracking—A Design-Engineering Problem,”
21. L. West and D.A. Reiter, “Full-Scale Arc Mapping Electro-Technology (New York), 69, 4, pp. 146–151
Tests,” in Fire & Materials 2005, Interscience (1962).
Communications Ltd., London, pp. 325–339 (2005). 36. R.S. Norman and A.A. Kessel, “Internal Oxidation
22. Guide for Performing Arc-Flash Hazard Calculations Mechanism for Nontracking Organic Insulation,”
(IEEE Std 1584), IEEE, New York, 2002. AIEE Transactions, 77, p. 632 (1958).
23. Standard for Performance Specification for Flame 37. F. Noto and K. Kawamura, “Tracking and
Resistant Textile Materials for Wearing Apparel for Ignition Phenomena of Polyvinyl Chloride
Use by Electrical Workers Exposed to Momentary Resin under Wet Polluted Conditions,” IEEE
Electric Arc and Related Thermal Hazards (ASTM F Transactions on Electrical Insulation, EI-13,
1506), ASTM, West Conshohocken, PA, 2012. pp. 418–425 (1978).
24. Standard Test Method for Determining the Arc Ther- 38. R. Wilkins and M.J. Billings, “Effect of Discharges
mal Performance Value of Materials for Clothing between Electrodes on the Surface of Organic
(ASTM F 1959), ASTM, West Conshohocken, PA, Insulation,” Proceedings IEE, 116, pp. 1777–1784
2006. (1969).
25. T. Bernstein, “Electrical Fires: Causes, Prevention, 39. Standard Dictionary of Electrical and Electronics
and Investigation,” in Electrical Hazards and Terms, IEEE, New York (1998).
Accidents: Their Cause and Prevention 40. J.L. Ferrino, “An Investigation of Fire Phenomena in
(E.K. Greenwald, ed.), Van Nostrand Reinhold, Residential Electrical Wiring and Connections,”
New York, pp. 116–134 (1991). M.S. thesis, University of Maryland, College Park,
26. J. Engel, M. Walz, and J. McCormick, “Arc Fault MD (2002).
Circuit Interrupters,” Third Joint FAA/DoD/NASA 41. M. Goodson, T. Perryman, and K. Colwell, “Effects
Conference on Aging Aircraft, Albuquerque, NM of Polyurethane Foam Systems on Wiring
(1999). Ampacity,” Fire & Arson Investigator, 52, 4, pp.
27. T. Bernstein, “Electrocution and Fires Involving 47–50 (July 2002).
120/240-V Appliances,” IEEE Transactions on 42. B. Béland, “Fires of Electrical Origin,” Fire & Arson
Industry Applications, IA-19, pp. 155–159 (1983). Investigator, 43, pp. 35–41 (Dec. 1992).
28. G. Gregory, “AFCIs Target Residential Electrical 43. K. Kinoshita, T. Hagiwara, and J. Kinbara, “Ignit-
Fires,” NFPA Journal, 94, pp. 69–71 (Mar./Apr. ability of VVF Cable in Contact with Grounded
2000). Object,” Journal of the Japanese Association of
29. A.C. Eberhardt, “Testing and Performance of Line Fire Science and Engineering, 28, 3, pp. 30–37
Interrupters in Structure Fires,” presented at Defense (1978).
Research Institute, Product Liability Conference, 44. U.S. Consumer Product Safety Commission Staff
New Orleans (2004). Report on the Beverly Hills Supper Club Fire of
30. S. Nakamura, H. Kaneda, M. Ieda, and G. Sawa, May 28, 1977, CPSC, Washington [n.d.].
“Correlation between Resistance to Tracking and 45. M. Goodson, D. Sneed, and M. Keller, “Electrically
Energy of Scintillation Discharge for Induced Fuel Gas Fires,” Fire & Arson Investigator,
Polyvinylchloride Doped with Aluminium 49, 4, pp. 10–12 (1999).
702 V. Babrauskas
46. Rousseau, A., Validation of Installation Methods Copper Wires,” National Research Institute of
for CSST Gas Piping to Mitigate Lightning Related Police Science Reports—Research on Forensic Sci-
Damage - Phase 1, Prepared by SIFTIM, Fire ence, 41, pp. 30–37 (Aug. 1988).
Protection Research Foundation, Quincy MA 62. J. Aronstein, “Fire Due to Overheating Aluminum-
(2011). Wired Branch Circuit Connections,” Wright Malta
47. B. Béland, “Some Fires of Electrical Origin,” Fire & Corp., Ballston Spa NY (1983).
Arson Investigator, 37, 2, pp. 37–38 (Dec. 1986). 63. Guest, P. G., Static Electricity in Nature and Industry
48. K.F. Davis, “Building Fires Attributed to Be (Bulletin 368), Bureau of Mines, US Government
Caused by Electrical Wiring Faults,” in Proceedings Printing Office, Washington (1933).
of the 27th International Conference on Fire Safety, 64. Lüttgens, G., and Glor, M., Understanding and
Product Safety Corp., Sissonville, WV, pp. 101–110 Controlling Static Electricity, Expert Verlag,
(1999). Ehningen (1989).
49. NFPA 70, National Electrical Code, National Fire 65. Pratt, T. M., Electrostatic Initiation of Explosions in
Protection Association, Quincy, MA, 2014. Dusts, Cereal Foods World 23, 601–605 (1978).
50. P. Cushing, “BX Cable and Fire Cause Determina- 66. L.B. Loeb, “Static Electrification-I,” in Progress in
tion,” Fire Engineering, 153, p. 46 (July 2000). Dielectrics, vol. 4, Academic, New York,
51. S. Turkel, “In Search of Transients,” EC&M, 99, 16, pp. 249–309 (1962).
p. 18 (Oct. 2000). 67. E.M. Cohn and P.G. Guest, “Influence of Humidity
52. T. Kuroyanagi, S. Inoue, and H. Suzuki, “Glowing upon the Resistivity of Solid Dielectrics and upon
Phenomena of Copper and Copper Materials and the Dissipation of Static Electricity” (IC 7286),
Their Electrical Characteristics,” Copper Promotion Bureau of Mines, Pittsburgh (1944).
Technical Research Group Journal, 20, pp. 198–204 68. Britton, L. G., Avoiding Static Ignition Hazards in
(1981). Chemical Operations, AIChE (1999).
53. O. Keski-Rahkonen and J. Mangs, “Electrical Igni- 69. Glor, M., Ignition of Gas/Air Mixtures by
tion Sources in NPPs: Statistical, Modelling, and Discharges between Electrostatically Charged Plas-
Experimental Studies,” International Atomic Energy tic Surfaces and Metallic Electrodes,
Agency, Technical Committee Meeting on “Fire J. Electrostatics 10, 327–333 (1981).
Experience in NPPs and Lessons Learned” (J7-TC- 70. Bartknecht, W., Dust Explosions: Course, Preven-
2001.7), Vienna (July 9–13, 2001). tion, Protection, Springer-Verlag, Berlin (1989).
54. M. Abramovitz and I.A. Stegun, Handbook of Math- 71. Glor, M., Conditions for the Appearance of
ematical Functions (AMS55), NBS, Gaithersburg, Discharges during the Gravitational Compaction of
MD (1964). Powders, J. Electrostatics 15, 223–235 (1984).
55. Wilson, C., McIntosh, G., and Timsit, R. S., Contact 72. Glor, M., Overview of the Occurrence and
Spot Temperature and the Temperature of External Incendivity of Cone Discharges with Case Studies
Surfaces in an Electrical Connection, Proc. ICEC- from Industrial Practice, J. Loss Prevention in the
ICREPEC2012 – Joint Conf. of the 26th Intl. Conf. Process Industries 14, 123–128 (2001).
on Electrical Contacts and 4th Intl. Conf. on Reli- 73. Maurer, B., Glor, M., Lüttgens, G., and Post, L.,
ability of Electrical Products and Electric Contacts, Hazards Associated with Propagating Brush
Beijing, China (2012). Discharges on Flexible Intermediate Bulk
56. W.J. Meese and R.W. Beausoliel, “Exploratory Containers, Compounds and Coated Materials,
Study of Glowing Electrical Connections” (NBS pp. 217–222 in Electrostatics ‘87—7th Intl. Conf.
BSS 103), NBS, Gaithersburg, MD (1977). on Electrostatic Phenomena (Conf. Series No. 85),
57. D. Newbury and S. Greenwald, “Observations on the Institute of Physics, London (1987).
Mechanisms of High Resistance Junction Formation 74. Cross, J. A., Electrostatics: Principles, Problems
in Aluminum Wire Connections,” Journal of and Applications, Adam Hilger, Bristol, England
Research NBS, 85, pp. 429–440 (1980). (1987).
58. Shea, J. J., Glowing Contact Physics, pp. 48–57 in 75. Y. Tabata and S. Masuda, “Minimum Potential of
52nd IEEE Holm Conference on Electrical Contacts, Charged Insulator to Cause Incendiary Discharges,”
IEEE, New York (2006). IEEE Transactions on Industry Applications, IA-20,
59. “Tests of Insulating Materials for Resistance to Heat pp. 1206–1211 (1984).
and Fire,” Report of CEE Working Group “Hot 76. P.G. Guest, V.W. Sikora, and B. Lewis,
Mandrel Test,” CEE (031) D126/61, Deutsches “Static Electricity in Hospital Operating Suites:
Komitee der CEE beim Verband Deutscher Direct and Related Hazards and Pertinent
Elektrotechniker, Frankfurt am Main (1961). Remedies” (RI 4833), Bureau of Mines, Pittsburgh
60. T. Kawase, “The Breeding Process of Cu2O,” IAEI (1952).
News, 47, pp. 24–25 (July/Aug. 1975); Second 77. Fisher, R. J., A Severe Human ESD Model for Safety
Report, 49, pp. 45–46 (Nov./Dec. 1977). and High Reliability System Qualification Testing
61. Y. Hagimoto, K. Kinoshita, and T. Hagiwara, (SAND89-0194C), Sandia Natl. Labs., Albuquerque
“Phenomenon of Glow at the Electrical Contacts of NM (1989).
www.ebook777.com
78. Code of Practice for Control of Undesirable Static 95. Recommended Practice for Protecting Residential
Electricity (BS 5958 Part 1), British Standards Insti- Structures and Appliances against Surges (Docu-
tution, London (1980). ment # PEAC.0545.R), EPRI PEAC Corp.,
79. G.W. Brundrett, “A Review of Factors Influencing Knoxville, TN (1999).
Electrostatic Shocks in Offices,” Journal of Electro- 96. M.M. Frydenlund, Lightning Protection for People
statics, 2, pp. 295–315 (1976/77). and Property, Van Nostrand Reinhold, New York
80. Klinkenberg, A., and van der Minne, J. L., Electro- (1993).
statics in the Petroleum Industry, Elsevier, 97. Frydenlund, M. M., Lightning Protection for Peo-
Amsterdam (1958). ple and Property, Van Nostrand Reinhold,
81. Lüttgens, G., and Wilson, N., Electrostatic New York (1993).
Hazards, Butterworth-Heinemann, Oxford (1997). 98. Anderson, R., Lightning Conductors: Their His-
82. P. Boschung and M. Glor, “Methods for tory, Nature and Mode of Application, E&FN
Investigating the Electrostatic Behaviour of Spon, London (1879).
Powders,” Journal of Electrostatics, 8, pp. 205–219 99. Müller-Hillebrand, D., The Protection of Houses by
(1980). Lightning Conductors—An Historical Review,
83. S.J. Collocott, V.T. Morgan, and R. Morrow, “The J. Franklin Institute 274, 34–54 (1962).
Electrification of Operating Powder Chemical Fire 100. Lee, R. H., Protection Zone for Buildings against
Extinguishers,” Journal of Electrostatics, 9, pp. Lightning Strikes using Transmission Line Protec-
191–196 (1980). tion Practice, IEEE Trans. Industry Applications
84. S. Singh, P. Cartwright, and D. Thorpe, “Silo Elec- IA-14, 465–470 (1978).
trostatic Hazards” (SMS-84-052), National Grain 101. Standard for the Installation of Lightning Protection
and Feed Association, Washington, DC (1984). Systems (NFPA 780), NFPA.
85. Blythe, A. R., and Reddish, W., Charges on Powders 102. The Basis of Conventional Lightning Protection
and Bulking Effects, pp. 107-114 in Electrostatics Technology: A Review of the Scientific Develop-
1979 (Conf. Series No. 48), The Institute of Physics, ment of Conventional Lightning Protection
London (1979). Technologies and Standards, Federal Interagency
86. A. Klinkenberg and J.L. van der Minne, Electrostat- Lightning Protection User Group, [n.p] (2001).
ics in the Petroleum Industry, Elsevier, Amsterdam 103. Moore, C. B., Rison, W., Mathis, J., and Aulich, G.,
(1958). Lightning Rod Improvement Studies, J. Applied
87. L.B. Britton and J.A. Smith, “Static Hazards of Meteorology 39, 593–609 (2000).
Drum Filling,” Paper 54e, 21st Loss Prevention Sym- 104. T.E. Eaton, Notes on Electrical Fires, 3rd ed. and
posium, AIChE (1987). 1990 supplement, Eaton Engineering Co.,
88. H.L. Walmsley and J.S. Mills, “Electrostatic Ignition Nicholasville, KY (1989).
Hazards in Road Tanker Loading: Part 1. Review 105. Safety of Information Technology Equipment,
and Experimental Measurements,” Journal of Elec- Including Electrical Business Equipment
trostatics, 28, pp. 61–87 (1992). (UL 1950), Underwriters Laboratories Inc.,
89. Bachman, K. C., Variables Which Influence Spark Northbrook IL, 2006.
Production Due to Static Electricity in Tank Truck 106. J. Sletback, R. Kristensen, H. Sundklakk, G. Nåvik,
Loading, Lightning and Static Electricity Conf., and R. Munde, “Glowing Contact Areas in Loose
Royal Aeronautical Society, London (1975). Copper Wire Connections,” in Proceedings 37th
90. L.B. Loeb, “Static Electrification-I,” in Progress in IEEE Holm Conference on Electrical Contacts,
Dielectrics, vol. 4, Academic, New York, IEEE, New York, pp. 244–248 (1991).
pp. 249–309 (1962). 107. E.L. Litchfield, T.A. Kubala, T. Schellinger,
91. V.I. Ermakov and Y.I. Stozhkov, “New Mechanism F.J. Perzak, and D. Burgess, “Practical Ignition
of Thundercloud Electricity and Lightning Produc- Problems Related to Intrinsic Safety in Mine Equip-
tion,” in 11th International Conference on Atmo- ment: Four Short-Term Studies” (RI 8464), Bureau
spheric Electricity (NASA/CP-1999-209261) of Mines, Pittsburgh (1980).
(H.J. Christian, ed.), NASA, Marshall Space Flight 108. M. Saito and W. Sakurai, “An Evaluation Method of
Center, AL, pp. 242–245 (1999). Electrical Fire Hazard Caused by Tracking Break-
92. W.C. Hart and E.W. Malone, Lightning and Light- down,” in Proceedings Electrical/Electronics Insu-
ning Protection, Don White Consultants, lation Conference, IEEE, New York, pp. 137–141
Gainesville, VA (1979). (1979).
93. M.M. Frydenlund, Lightning Protection for People 109. Médard, L. A., Accidental Explosions, 2 vols., Ellis
and Property, Van Nostrand Reinhold, New York Horwood, Chichester, England (1989).
(1993). 110. Crowl, D. A., Understanding Explosions, CCPS/
94. Frey, O., The Origin, the Effects and the Simulation AIChE, New York (2003).
of Transients, as Well as Their International 111. Bodurtha, F. T., Industrial Explosion Prevention
Standardization, Electro 82, Boston (1982). and Protection, McGraw-Hill, New York (1980).
704 V. Babrauskas
112. Hattwig, M., and Steen, H., eds., Handbook of 126. Flowers, J. W., The Channel of the Spark Discharge,
Explosion Protection and Prevention, Wiley- Physical Review 64:7/8, 225–235 (1943).
VCH, Weinheim, Germany (2004). 127. Vanyukov, M. P., Isaenko, V. I., and Khazov, L. D.,
113. Baker, Wilfred E., Cox, P. A., Westine, P. S., Kulesz, Investigation of Light Phenomena Related to the
J. J., and Strehlow, R. A., Explosion Hazards and Development of the Spark-Discharge Channel,
Evaluation, Elsevier, Amsterdam (1983). Zhurnal tekhnicheskoι fiziki 25, 1248 (1955).
114. Babrauskas, V., Electric Arc Explosions, 128. Higham, J. B., and Meek, J. M., The Expansion of
pp. 1283–1296 in Interflam 2010—Proc. 12th Intl. Gaseous Spark Channels, Proc. Physical Society B
Conf., Interscience Communications Ltd, London 63, 649–661 (1950).
(2010). 129. Drouet, M. G., and Nadeau, F., Pressure Waves due
115. Lindmayer, M., and Paulke, J., Arc Motion and Pres- to Arcing Faults in a Substation, IEEE Trans. on
sure Formation in Low Voltage Switchgear, IEEE Power Apparatus and. Systems PAS-98,
Trans. on Components, Packaging, and 1632–1635 (1979).
Manufacturing Technology A21, 33–39 (1998). 130. Dasbach, A., and Pietsch, G., Investigation of the
116. Nabours, Robert E., Fish, R. M., and Hill, P. F., Power Balance of High Current Faults, pp. 15–18
Electrical Injuries: Engineering, Medical and in Proc. 9th Intl. Conf. on Gas Discharges and Their
Legal Aspects, 2nd ed., Lawyers & Judges Publish- Applications (GD88), Venice (1988).
ing Co. (2004). 131. Baldrey, H. W., and Hudson, A. A., Pressures
117. Mazer, W. M., Electrical Accident Investigation Generated by Fault Arcs in Small Enclosures (Ref.
Handbook, 3 vols., Electrodata, Inc., Glen Echo Z/T134), The British Electrical and Allied Industries
MD (var. dates). Research Assn., Leatherhead, Surrey, England (1961).
118. Lee, R. H., Pressures Developed by Arcs, IEEE Trans. 132. Tanaka, S., Miyagi, T., Ohtaka, T., Iwata, M.,
on Industry Applications IA-23, 760–764 (1987). Amakawa, T., and Goda, Y., Influence of Electrode
119. Crawford, K. S., Clark, D. G., and Doughty, R. L., Material on Pressure-rise due to Arc in a
Motor Terminal Box Explosions due to Faults, IEEE Closed Chamber, IEEJ Trans. PE 128, 1561–1568
Trans. on Industry Applications 29, 257–267 (1993). (2008).
120. Heberlein, G. E. jr., Higgins, J. A., and Epperly, R. 133. Capelli-Schellpfeffer, M.,, Miller, G. H., and
A., Report on Enclosure Internal Arcing Tests, IEEE Humilier, M., Thermoacoustic Energy Effects in
Industry Applications Magazine 2:3, 35–42 (May/- Electrical Arcs, pp. 19–32 in Occupational Electri-
June 1996). cal Injury: An Intl. Symp., / Annals of the New York
121. Jennings, C., Fire-Fatality High-Rise Office Building Academy of Sciences, Vol. 888, New York Academy
Fire, Atlanta, Georgia (June 30, 1989), US Fire of Sciences, New York (1999).
Admin., [Emmitsburg MD], (1989). 134. Bowen, J. E., Wactor, M. W., Miller, G. H., and
122. Graneau, P., The Cause of Thunder, J. Physics D: Capelli-Schellpfeffer, M., Catch the Wave:
Applied Physics 22, 1083–1094 (1989). Modeling the Pressure Wave Associated with Arc
123. Lee, R. H., The Shattering Effect of Lightning— Fault, IEEE Industry Applications Magazine 10:4,
Pressure from Heating of Air by Stroke Current, 59–67 (Aug. 2004).
IEEE Transactions on Industry Applications 22, 135. Caillard, J., de Izarra C., Brunet, L, Vallée, O., and
416–419 (1986). Gillard, P., Assessment of the Blast Wave Generated
124. Eisenberg, N. A., Lynch, C. J., and Breeding, R. J., by a Low-Energy Plasma Igniter and Spectroscopic
Vulnerability Model: Assessing Damage from Mari- Measurements, IEEE Trans. on Magnetics 39,
time Spills by Computer Simulation (CG-D-136-75), 212–217 (2003).
US Coast Guard, Washington (1975). 136. Friberg, G., and Pietsch, G. J., Calculation of Pres-
125. Baker, W. E. et al., Explosion Hazards and Evalu- sure Rise due to Arcing Faults, IEEE Trans. on
ation, Elsevier, Amsterdam (1983). Power Delivery 14, 365–370 (1999)
www.ebook777.com
Surface Flame Spread

23
Yuji Hasemi
Introduction 1. Vaporization of solid or liquid due to the

heating from flame over the fuel’s surface
Surface flame spread is a process of a moving 2. Mixing of the pyrolized gas and oxygen in the
flame in the vicinity of a pyrolyzing region on the vicinity of the fuel surface
surface of a solid or liquid that acts as a fuel 3. Combustion of the pyrolized gas and forma-
source. It is distinct from flame propagation in a tion of the diffusion flame
premixed fuel and oxygen system in that the 4. Heating of the unburnt fuel surface to ignition
surface spread of flame occurs as a result of the temperature from the diffusion flame
heating of the surface due to the direct or remote The oxygen and fuel concentrations together
heating by the flame generated from the burning with the heat transfer between the flame and the
surface. The surface flame spread is very often solid phase strongly affect the process. Taking
critical to the destiny of fires in natural and built the surface flame spread as a successive ignition
environments. This spread applies whether the front over a combustible object, the speed of
fire is an urban conflagration or is the first growth spread and its sustainability are controlled by
after ignition of a room’s draperies. This chapter the balance of the flame heating and the rise of
provides fire safety engineers with an overview surface temperature (see Chap. 21). If the flame
of surface flame spread during the growth of a ignites the virgin surface of the area larger than
fire and the modeling of different modes of flame that of the burning surface before local extinc-
spread to improve understanding of their effects tion, the spread will accelerate. From this point
on the outcomes of fires. of view, relative configuration of the flame to the
surface is critical for the dynamics and the hazard
of surface flame spread.
Flame spread in the direction of the mean flow
Surface Flame Spread Basics due to wind or buoyancy is called wind-aided or
concurrent spread, and, on the contrary, flame
Flame Spread Process spread occurring in a direction opposite to that
is called opposed-flow spread.
The surface flame spread is caused as a result of Difference of the significance of the mode of
the cycle of the following processes: flame spread for fire safety can be illustrated
through the typical process of fire growth in a
This chapter is based in part on material by Professor
J. Quintiere appearing in previous editions of this
room with highly combustible lining ignited on
handbook. the floor, as shown in Fig. 23.1. The spread of fire
Y. Hasemi (*)
over the floor is a typical opposed-flow spread
Wasada University because the flame induces unheated air along the

706 Y. Hasemi
a b c
Fig. 23.1 Orientation of combustible surface and the modes of flame spread over (a) floor, (b) wall, and (c) ceiling
carpet surface in the direction against the spread mass fires. In mass fires such as urban
of the flame (Fig. 23.1a). The surface flame conflagrations and forest fires, fire brands have
spread at the initial stage of a room fire is gener- an additional but sometimes essential role in
ally slow and can even be unstable; but once the enhancing fire spread in the wind direction. The
wall lining is ignited, the buoyancy makes the importance of opposed-flow flame spread can
flame develop along its surface and exposes the become significant when the temperature of the
lining surface above the burning region to the fire environment has been raised enough to pre-
flame (Fig. 23.1b). This condition generally heat the wide range of surfaces of combustibles.
makes the flame spread much faster than on the Although the two modes of flame spread are
floor, even if the floor and the wall are lined with apparently distinct from each other, there is nota-
identical material. ble ambiguity in fires on an inclined combustible
If not extinguished at this stage, fire may surface. On an upward inclined surface, the angle
finally reach and ignite the ceiling. The buoyancy between the flame flow and the surface is reduced
makes the flame spread laterally beneath the with an increase of the angle of inclination, and
ceiling toward the opening, exposing the overall finally the flame begins to crawl over the surface
surface of the ceiling to the flame heating [1] (Fig. 23.2). This change occurs generally at
(Fig. 23.1c). As widely recognized, flame spread 15–30 , depending on the width, side
beneath the ceiling is generally fast and could confinements, and other conditions; inclination
cause flashover by igniting remaining furniture of roofs, escalators, and slopes for wheelchairs
and wall surface far from the ignition source are generally within the range where wind-aided
within the room. The Bradford Stadium fire flame spread is anticipated. The escalator fire at
disaster in the United Kingdom (1985) is one of the Kings Cross Subway Station Fire Disaster
the most significant examples of lateral flame (1987) is a significant example of wind-aided
spread beneath a combustible roof or ceiling. flame spread along an inclined configuration
During lateral flame spread beneath the ceiling, under the enhancement of buoyancy due to the
downward flame spread over the wall lining is confinement of air supply by the side walls [2].
often observed within the smoke layer. It is an
opposed-flow spread, yet its spread velocity is
generally fast due to the additional heating from Research Background
the smoke layer.
As seen in the growth of room fire, the wind- Large numbers of experiments on and models of
aided flame spread presents a number of key surface flame spread with diverse levels of theo-
processes escalating the hazard of building and retical sophistication and practical relevance
www.ebook777.com
23 Surface Flame Spread 707
Fig. 23.2 Change of the Vertical

mode of surface flame θ = 90°
spread according to the
θ = 60°
slope angle
θ = 30°
Horizontal
θ = 0°
have been conducted, and it is impossible to around one-story-high wall specimens or with
make a precise review of all the important standard and larger-scale room tests. These stud-
research on this subject. It is recommended to ies have revealed the importance of the surface
seek comprehensive reviews such as Drysdale configurations on the general behavior of flame
[3], Quintiere [4], Fernandez-Pello and Hirano spread; examples of acceleration of flame spread
[5], Hirano [6], and Williams [7]. However, as on grooved combustible surfaces, in corner
long as surface flame spread is discussed for the walls, and in a vertical, inclined, or downward
assessment of fire safety of natural or artificial channel lined with combustible material show
composites, the phenomena are generally the significance of the configuration effect.
modeled as a thermal process causing the succes- Together with the substantial progress in the
sive piloted ignition on the surface due to the measurement technology of the combustibility
heating from the burning of the material itself of materials through the 1980s and the 1990s,
under the gravity and the atmospheric conditions these studies have made it possible to predict
of the earth. surface flame spread in a room to some extent
Active research on thermal modeling of sur- using material properties obtained from bench-
face flame spread has occurred since the late scale tests.
1960s, and significant progress was made in the
basic understanding of flame spread through the-
oretical sophistication and laboratory Wind-Aided Flame Spread
experiments throughout the 1970s. Later, during
the 1980s, the approach was extended to turbu- The wind-aided flame spread is the most impor-
lent flames for the application to the fire hazard tant mode of surface flame spread for fire hazard
assessment of lining materials. In the 1990s, fur- assessment. In this section, modeling and assess-
ther studies were conducted to analyze and assess ment are introduced regarding the two most sig-
the general behavior of room fires through input nificant examples of wind-aided flame spread in
from and validation against large-scale tests with building fires: upward wall flame spread and
708 Y. Hasemi
flame spread beneath the ceiling. Upward wall direction normal to the wall surface can be
flame spread along a wall has drawn the particu- reasonably assumed and evolution of the surface
lar interest of fire safety engineers for its primary temperature at x above the fire origin at time t,
importance in the determination of the destiny of T(x,t), is formulated as
a room fire and for the rich variety of
ðt
configurations of wall surfaces that may influ- 00
ence the general behavior of flame spread. In T ðx; tÞ T o ¼ q_ f ðx; τÞϕðt τÞdτ ð23:1Þ
this section, theories of wind-aided flame spread 0
developed essentially for wall fires are applied to

provide technical insight in the assessment of where
room fires or the combustibility of lining T0 ¼ Initial surface temperature
materials in general. q_00 ¼ Flame heat flux
f
ϕ(t τ) ¼ Impulse response of the surface tem-
perature at timet t to the surface heat flux at τ
Upward Turbulent Wall Flame Spread ϕ(t τ) ¼ [πkρc(t τ)]1/2 and ϕ(t τ) ¼
[ρcδ(t τ)]1 apply for a thermally thick wall
Figure 23.3 is a schematic of upward flame spread and for a thin wall with thickness of δ respectively
typical of wind-aided flame spread. The flame with k, ρ, and c thermal conductivity, density, and
spread is thus perceived in two different manners: specific heat of the solid. Equation 23.1 can be
first by the advancement of flame front and second solved for flame spread velocity, Vp ¼ dxp /dt, for
by the advancement of the ignition front of the simple conditions by further introducing engi-
solid surface. The rate of movement of the ignition neering relations on the flame heat flux,
front is normally defined as the flame spread 00
q_ f ¼ Fðx; τÞ.
velocity for modeling purpose. The location of
Flame heat flux is normally represented as
the ignition front of a burning surface, xp(t), is
relative location to the flame length, xf; that is,
identified as the location where surface tempera- 00
ture has reached the ignition temperature, Tig. q_ f ¼ F x=x f , Applying the Froude modeling
for the flame height (see Chap. 21) to a line fire
Flame Heat Transfer For Fig. 23.3, along a wall, xf is generally expressed as function
one-dimensional thermal conduction in the of heat release rate per unit width as
Fig. 23.3 Upward wall

flame spread [9]
Tw (x,y )
Flame
Preheat
δfc
distance
Ignition
front
xp xp x xp
Tig
qf ″(x )
•
Tw (x,0) Y
Wall temperature Heat flux Model
www.ebook777.com
2=3 2=3
x f =D ¼ kQ_ ‘
*n
ð23:2Þ x f ¼ k3=2 =c p T o ρ1 g1=2 Q_ ‘
¼ k Q_ ‘
0 2=3
ð23:3Þ
Where
D ¼ Characteristic lengthof the burning area
The k0 value is roughly 0.01 k; for Fig. 23.3, k0
Q_ ‘ ¼
*
(normally height or depth), value is found to be k0 ¼ 0.057 m1/3 kW2/3,
Q_ ‘ =cT 0 ρ1 g1=2 D2=3 : whereas k0 values in literature range from
k ¼ Empirical coefficient 0.043 m1/3 kW2/3 to 0.067 m1/3 kW2/3 [8, 9,
Q_ ‘ ¼ Heat release rate per unit width of fire 12, 13]. For a limited range of Q_ ‘ , flame length
source can be linearized against heat release rate or the
Figure 23.4 illustrates a relation between heat characteristic length of burning region for engi-
release rate per unit width and flame length for neering purposes as
line burners against a constant temperature inert 00
wall. It shows that k ¼ 6.0 for intermittent flame x f k f Q_ ‘ ¼ k f Q_ ‘ D ¼ aD ð23:4Þ

and n ~ 2/3 for Q_ ‘ 1:0, as anticipated for an 00
ideal line source by dimensional analysis where Q_ ‘ is the characteristic heat release rate
00
[8]. The value of k depends on the definition per unit width, kf, and ak f Q_ ‘ depends on heat
and measurement of flame length, and experi- release rate, but kf ~ 0.006–0.01 m2/kW is suit-
mental k values range from 4.65 to 7.0 able for lining materials in practice [14–16].
[8–11]. Newman and Wieczorek reviewed Figure 23.5 is a general relation between
reported values for k and n [12]. With n ¼ 2/3 the total incident heat flux from wall flame to
in Equation 23.2, flame length is found to be wall surface and the distance from the lower
independent of the dimension of the burning edge of the burning surface normalized by xf.
region and is represented as It is a summary of experiments on wall flames
Fig. 23.4 Relation 100

between the length of
intermittent wall flame and
dimensionless heat release 0.0375D × 0.270
rate per unit width [9] 0.082D × 0.270
0.075D × 0.270
Flame length / fuel length, xf /D (–)
Solid; laminar 3
10
1
0.1 1 10
Dimensionless heat release rate per unit width,
Q */(–)
710 Y. Hasemi
10 Large fire
Small fire
PMMA wall fire x < x0
Laminar methanol fire
x0 = 3.56 m, xf = 6.06 m
x0 = 1.1 cm, xf = 5.3 cm xf
x >x
x <x 0
0 Pyrolysis x
region x0
Turbulent
Total incident heat flux, qw ″ (W/cm2)
moderate fires
1 x > x0
2
•
5
Ahmad and Faeth19
•
qw ″ × PR
( μ∞ LB (
= k(Grx *)n Hasemi9
Grx * = Lgx 3/(4CpT∞v∞2)
0.1 k n
1
Laminar, x < x0 0.24–0.28 0.25
Laminar, x0 < x < xf 0.5–0.6 0.25 1.3
Turbulent, x0 < x < xf 0.035–0.04 0.4
x < 10 cm Laminar
x > 10 cm Turbulent
Data region
(xf : 0.3–1.4 m)
0.01
0.01 0.1 1.0 10
x /xf
Fig. 23.5 Wall flame incident heat flux for materials [17], for laminar flames [19], and for a large PMMA wall fire [18]
from steady porous burners, vertical wicks, depend on thickness and on the substrate material
and burning of specimens of finite-surface adjacent to the solid. Based on these factors, it is
area of materials in practice [9, 17–19]. For apparent that the thermally thick case is more
more detailed flame heat flux data and significant.
correlations on different configurations, see With the unique dependence of flame heat
Chap. 25. flux on dimensionless height, x/xf, Equation 23.1
As thermal conduction theories suggest, can be solved to provide a characteristic steady
response of surface temperature to heat input flame spread velocity Vp ¼ xp /t for a thermally
depends significantly on the thickness of the thick solid as
solid; modeling of surface flame spread requires ð 1
different approaches for thermally thick solid and 00 pffiffiffiffiffiffiffi 2 2
Vp ¼ q_ f ξ þ x p = ξdξ =πkρc T ig T o
thin materials. Except for items such as paper, 0
garments, or draperies in a room, in practice most ð23:5Þ
solids should behave as thermally thick under
The numerator has the dimension of
flame spread conditions. Engineering treatment
(kW/m2)2 m, and Equation 23.5 can be
of surface flame spread might appear to regard 00
solids thicker than 1 mm as thermally thick. represented more simply as q_ fc 2δfc with charac-
00
Up to a thickness of 1–2 cm, flame spread could teristic flame heat flux, q_ fc , and characteristic
www.ebook777.com
preheat distance, δfc. Equation 23.5 is thus characteristic time to ignition, it will become
expressed as difficult to sustain successive thermal ignition
. . on the surface. Taking the linearized flame length
00
Vp ¼ 4q_ fc2 δfc 2 ¼ xfc x p approximation (Equation 23.4), we introduce the
πkρc T ig T o t*ig
following dimensionless flame spread accelera-
ð23:6Þ tion factor: [14]
where
b ¼ ða 1Þ tig *=tb ð23:7Þ
xfc ¼ x p þ δfc
2 00 where (a 1) represents the significance of
t*ig πkρc T ig T o =4q_ fc2
flame heating, and if tig*/tb is large, it will
Equation 23.6 is essentially in the same form become difficult for successive ignition to sus-
with the flame spread velocity obtained for tain. Obviously, b ¼ 0 stands for Equation 23.6
flame heat flux decaying exponentially with dis- and is the condition for the achievement of
tance from the ignition front [20]. The preheat steady-state spread. The sign of b dictates the
distance, δfc, is essentially the distance between general behavior of wind-aided flame spread:
the flame front and the ignition front. b > 0 will lead to the acceleration of spread,
whereas b < 0 will result in the deceleration of
Controlling Parameters of Upward Wall spread and finally autonomous extinction. There
Flame Spread In Equation 23.6, it is important is more sophisticated discussion on the general
that the characteristic preheat distance, δfc, is behavior of wind-aided flame spread [15, 16,
essentially controlled only by the heat release 21–24], which may still await future validation.
rate in a one-dimensional flame configuration, But every kind of thermal model and analysis
and kρc(Tig T0)2 is the central part of material based on the linearized flame length approxima-
property to control ignitability. tion (Equation 23.4) finally eventuates the recog-
Another important parameter that may control nition of b as the central parameter for the
the general behavior of flame spread is the local assessment of hazard of flame spread.
burnout. Consider local burnout occurring at tb Figure 23.7 is a summary of the correlation
after the local ignition (Fig. 23.6), D ¼ Vptb. If between the time to flashover in ISO 9705 room-
the local burnout time is short compared to the corner tests with combustible linings and
b [14]. The elements of b, namely a, tig*, and tb,
are all material properties that can be quantified
with bench-scale tests such as the cone calorime-
ter. Although there is still some discussion on
what external heat flux level should be chosen
Heat release rate, Qc″
for the quantification, use of 30–50 kW/m2 exter-

nal radiant flux seems to lead to reasonable
explanation of the growth of room fire from
Equation 23.7. For a charring material, heat
release rate generally decays with time after the
sharp peak just after the ignition and can be more
suitably represented by an exponential function
00 00
of time, namely Q_ ðtÞ ¼ Q_ expðt=tc Þ. Anal-
max
ysis of the results of room-corner tests and the
tig tb
cone calorimeter on wood-based materials
suggests use of k f Q_ max and tc for a and tb
Time
Fig. 23.6 Time history of heat release rate from fixed respectively leads to a result consistent with
area burning surface Fig. 23.7 [15, 22] for such materials.
712 Y. Hasemi
Fig. 23.7 Time to

flashover in the ISO 9705 ∞
room-corner test versus the 800
Time to flashover, tig (s)

flame spread acceleration
700
factor [14]
600
500
400
300
200
100
0
–8 –6 –4 –2 0 2 4 6 8
Flame spread acceleration factor, b(–)
Equation 23.7 is simple but provides useful design, commodities or cargoes piled in
insights in the overall understanding of wind- warehouses or in mass merchandise outlets
aided flame spread. It is especially noteworthy often make “valleys with combustible cliffs,”
that material properties are not the single factor which can be considered as parallel walls from
to quantify this equation. Time to ignition and the flame spread point of view. A number of fire
time to burnout are dependent on the ambient tests of cargoes and commodities in typical ware-
temperature and external heating; and a, essen- house configuration demonstrate significant
tially the flame-length to pyrolysis-length ratio, acceleration of fire growth in such a configura-
can be augmented not only by external heating tion [24, 26]. A cavity within a wall or a roof
through enhancement of vaporization but also by sandwiched by combustible surfaces is another
increase of pyrolizing surface area due to any significant example capable of showing the
finishing treatment such as roughness and parallel-wall effect. Surface flame spread can be
grooves. Existence of the source of external further enhanced in a vertical or inclined cavity
heating has thus two implications for the acceler- by the stack effect due to general temperature
ation of flame spread: first through the increase of rise within the cavity. The fast fire spread
the temperature of the unburnt surface and second throughout whole high-rise building at the
through the promotion of vaporization of fuel due Beijing Television Cultural Center fire (2009) is
to additional heating of the burning surface partly attributed to these configuration effects.
[25]. Grooves or trench-like configurations of The vertical long cavity lined with polymer insu-
the burning surface can reduce air entrainment lation within the facades on both sides of the
to the vaporizing region and extend the flame building should have helped acceleration of fire
length. Configuration effect is as important as spread once the fire penetrated into the cavity.
material properties in the fire safety assessment Also, in a corner of walls, flame is generally
of any assembly with a combustible surface. prolonged due to the restriction of entrainment
Flame spread can be significantly enhanced in and the mutual radiation between the walls,
parallel wall configuration, where the flame on which can result in faster flame spread. There
either wall stimulates pyrolysis and preheating of are experiments and measurements of upward
the other surface and generates mutual accelera- turbulent flame spread on nearly full-scale
tion of flame spread. If the wall distance is small specimens of various materials in practice [11,
compared to the wall width, the air supply to the 18, 26–29]. These reports would provide valu-
burning surface will be restricted and further able information necessary for consideration in
prolong the flame length. Even though parallel- running large-scale flame spread tests with
wall configuration is not common in building materials in practice.
www.ebook777.com
Fig. 23.8 Area covered by 3

ceiling flame versus
effective heat release rate
[30, 31] 2.5
d = 0.09 m
Sf = 0.069 Q
Flame area, Sf (m2)

2
1.5
d = 0.16 m
Sf = 0.062 Q
1
One-dimensional ceiling fire
Sf = 0.023 Q
0.5
0
0 10 20 30 40 50
Heat release rate (oxygene consumption method, Q(kW)
Surface Flame Spread Beneath Ceiling represented as a unique function of distance

from the windward edge of the burning surface
Although fewer studies have been conducted on divided by the flame length. From these facts, the
the ceiling fires, the mechanism of the surface engineering framework for the assessment of
spread is essentially similar to the upward wall surface flame spread on a wall applies suitably
spread. Flame spreading beyond the pyrolysis for the flame spread beneath a combustible ceil-
front is the dominant force for the successive ing, whereas such physical constants as a should
ignition of unburnt ceiling surface. According be different from the wall fire configuration.
to the measurements of flame size and flame Under an unconfined ceiling, heat flux within
heat flux for one-dimensional ceiling flames in the solid ceiling flame is decreased weakly with
a corridor-like configuration and for circular ceil- distance within the range of 20–30 kW/m2, and is
ing flames from downward porous propane generally weaker than in upward wall flames of
sources [30] (Fig. 23.8), the area covered by a similar heat output, because of the buoyancy
visible ceiling flame is nearly proportional to the reducing the thickness of ceiling flame.
heat release rate, and the heat release rate per unit Ambient thermal conditions and surface
flame area is significantly larger for configurations of the ceiling could also affect
one-dimensional flames than for circular flames. the general behavior of flame spread beneath
This situation indicates uniform entrainment of the ceiling. One of the important issues to be
air beneath a ceiling flame, but the rate of considered in the understanding and fire hazard
entrainment could depend on the configuration. assessment of flame spread beneath the ceiling in
Flame length is thus proportional to heat release a room is the effect of preheating due to ceiling
rate per unit width in corridor configuration, jet and smoke layer, which normally come in
xf ¼ 0.0122 Q_ le with Q_ le as the effective heat contact with the ceiling earlier than flame in a
release rate per unit width (kW/m) for corridor likely fire growth scenario in a room. Corridor
ceiling configuration [31] and proportional to the ceiling flame spread tests lined with medium-
half power of heat release rate beneath an uncon- density fiberboard (MDF) with uniform external
fined ceiling. radiation to the ceiling surface demonstrate sig-
As shown in Chap. 25, total heat flux from the nificant sensitivity of the ultimate burn length to
ceiling flame to the ceiling surface can be pilot flame length ratio, xpoff/xpo, to the ceiling
714 Y. Hasemi
Fig. 23.9 Relation Initial surface temperature versus xpoff ≠ xpo

between the surface 14
temperature due to external
heating at piloted ignition 12
to ceiling and the ultimate 10
xpoff / xpo (–)

burnt length to pilot flame
length ratio for MDF [31] 8
6
4
2
0
0 100 200 300 400
Initial surface temperature (°C)
surface temperature just prior to the ignition by reason, in the modeling of opposed-flow flame
pilot flame, To (Fig. 23.9) [31]. To was controlled spread, spread velocity is normally assumed as
by the upward radiant panel. In the tests resulting steady state. Consider that the pyrolysis front
in xpoff/xpo > 10, almost the whole ceiling sur- traverses Δ within the time interval τ on a com-
face was finally pyrolyzed and flame spread itself bustible solid of the thickness δ small enough to
was strongly accelerated. Flame spread was less ensure practically uniform temperature across
sensitive to heat flux within the range of the thickness with no heat loss from the back
0–10 kW/m2. The remarkably fast flame spread surface. The flame spread velocity can be given
beneath a nearly unconfined roof/ceiling by Vp ¼ Δ/τ, and the energy conservation for the
observed at the event of the Bradford Stadium control volume Δ distance from the pyrolysis
fire disaster (1985) is attributed partly to the use front can be described as
of roof material of low ignition temperature.
Increase of the effective combustible surface ρcδV p T ig T o ¼ q_00 Δ ð23:8Þ
area due to beams, decorations, and so forth can
The net flame heat flux due to the gas-phase
also increase a value and accelerate flame spread.
conduction can be given by
Stenstad and Karlsson have demonstrated a sig-

nificant example of such effect by a large-scale q_00 kg T f T r =Δ ð23:9Þ
experiment [32].
where kg, Tf, and Tr are gas-phase conductivity,
flame temperature, and reference temperature for
the control volume. Tr can be correlated with
Opposed-Flow Flame Spread either Tig or To. From these equations, the flame
spread velocity can be represented as a function
Mechanism of Opposed-Flow Flame of thermal properties and the configuration con-
Spread dition as

In the opposed-flow spread, the front of the V p ¼ kg T f T r =ρcδ T ig T s ð23:10Þ
pyrolyzing region moves in the opposite direc-
tion of the flame flow. As seen in Fig. 23.10, the The flame temperature, Tf, should ideally be
unburnt surface beyond the pyrolysis front is taken as that due to adiabatic stoichiometric
heated by remote flame; the flame spread veloc- combustion but, in general, could be thought of
ity is less dependent on flame length or heat as less due to heat losses and chemical kinetic
release rate, and the distance that the flame effects. Under these ideal theoretical
heating covers should be quite limited. For this considerations, it can be shown that
www.ebook777.com
Fig. 23.10 Energy Boundary layer

conservation in opposed-
flow spread Vg Flame
Control volume
Pyrolysis region
Ts
δ V Tig
2
T 1 T ig þ Y ox, 1 =rcg ðΔH LÞ V p ¼ q_00 Δ=kρc T ig T o
2
ð23:14Þ
T f T ig ¼
1 Y ox, 1 =r
ð23:11Þ In Equations 23.13 and 23.14, the flame spread
velocity is apparently independent of the
where ΔH and L are heat of combustion and heat opposed-flow velocity, Vg. However, the inde-
of gasification of the solid fuel, respectively. T1 pendence is only the case as long as chemical
and Yox,1 are gas-phase ambient temperature and effects are unimportant. Chemical kinetic effects
oxygen concentration, respectively, and r and cg become important when the time for chemical
are stoichiometric mass ratio of oxygen to fuel reactions to be completed in the flame, tchem,
and specific heat of the gas phase, respectively. becomes long compared to the fluid flow transit
Because ΔH/L and Yox,1/r are large, and heat of time through the flame, tflow . If the flow is too
combustion per unit mass of consumed oxygen, fast, chemical reaction will be incomplete.
ΔHox, is nearly constant for most of combustible Because the flow transit time is proportional to
solids in practice (13 kJ/g), the flame temperature V g -1 and mixing should be enhanced by the
can be approximated as
decrease of Vg, the flow transit time to chemical
T f T ig Y ox, 1 ΔH ox =cg ð23:12Þ reaction time ratio, normally referred to as the
Damköhler number, Da, can be represented as
and we realize that the flame temperature is pri-
marily sensitive to only the ambient oxygen con- Da ¼ tflow =tchem / 1=V 2g ð23:15Þ
centration. This suggests that flame spread over a
ceiling would be reduced in a room as the oxygen The flame spread velocity is reduced with an
within the smoke layer near it is reduced. increase of Da. Taking the flame spread veloc-
For a thermally thin solid, δ can be taken as ity given by Equation 23.10 as Vp,ideal, the
constant. Combining Equations 23.8 and 23.9, actual flame spread velocity to Vp,ideal ratio
flame spread velocity is given as can be illustrated qualitatively as shown in
Fig. 23.11.
V p ¼ Δ=τ ¼ kg T f T o =ρcδ T ig T r However, for a thermally thick solid, the
ð23:13Þ relation between flame spread velocity and
opposed-flow velocity is more complicated.
For a thermally thick solid, δ should represent Figure 23.12 is a significant such example
the depth of thermal penetration, which depends showing either increase or decrease of V with
on time; that is, from the heat conduction equa- Vg depending on the ambient oxygen concentra-
pffiffiffiffiffiffiffiffiffiffiffiffi
tion for a semi-infinite thick slab as δ ¼ kτ=ρc. tion [33]. This illustration suggests the depen-
Substituting this into Equation 23.8, we have dence of Vp,ideal on Vg in the flame spread over a
716 Y. Hasemi
Fig. 23.11 Qualitative 10

dependence of opposed-
flow flame spread with
Damköhler number, Da
l Fast
ica High
1 h em tics
C ne
ki Y ox,∞ Low
V/ Vpideal
Vg
Low w
Slo
10–1
h
Hig
Extinction
10–2
10–2 10–1 1 10
Da
Fig. 23.12 Effect of

opposed-flow velocity and
oxygen concentration of
100
flame spread speed for
thick PMMA (Taken from
Fernandez-Pello et al.) [33]
10–1
Vp (cm/s)
Y0
1.00
0.727
10–2 0.533
0.432
0.329
0.276
0.247
0.233
0.211
10–3
101 102 103
Vg (cm/s)
www.ebook777.com
Fig. 23.13 Outline of the

heating and flame spread in
the LIFT apparatus
Radiant Sample
V 155
panel
mm
280
mm
483 800
mm mm
thermally thin solid. Attributing this effect to 2

V p Φ=kρc T ig T o ð23:18Þ
the dependence of the preheat distance, which
is the length of the control volume shown in with Φ, a parameter depending on Vg and Yox,1
Fig. 23.10, and assuming the balance of that could be quantified with a bench-scale test
gas-phase heat conduction with convection due for practical materials. Quintiere and Harkleroad
to the opposed flow has the velocity of Vg, [14] examined this approach using the lateral
ρgcgVg∂T/∂x kg∂2T/∂x2, then the following ignition and flame spread test (LIFT) apparatus
estimate can be derived: (Fig. 23.13) and have quantified the effective
opposed-flow flame spread properties as
Δ kg =ρg cg V g ð23:16Þ
summarized in Table 23.1. See Chap. 21 for
By substituting Equations 23.9 and 23.16, Equa- more detail on the ignitability parameters. Equa-
tion 23.14 yields the expression first derived by tion 23.18 also suggests the importance of the
de Ris: [34] ignitability parameter, kρc(Tig–To)2 for the fire
2 2
hazard assessment of any material. Existence of
V p V g kg ρg cg T f T r =kρc T ig T o a source of external heating, such as exposure to
smoke layer, flame sheet spreading beneath the
ð23:17Þ ceiling, and the like, may have a significant influ-
ence on the spread velocity.
Modeling of Opposed-Flow Flame

Spread Mass Fires
Significant progress has been made on the under- Urban fires, wildland fires, and forest fires are
standing and modeling of opposed-flow flame serious fire disasters in various districts. The
spread, but most of the research has concentrated conflagrations at the Great Hanshin earthquake
on poly(methyl methacrylate) (PMMA) and (1995) revealed significant risk of earthquake
other rather ideal materials from the viewpoint urban fires in Japan and other districts featuring
of combustion and pyrolysis modeling. Few urban areas densely inhabited with buildings
models have been attempted on the opposed- with relatively weak fire protection. Land devel-
flow turbulent flame spread over charring opment of wildland near urban districts in North
materials and composites in practice. Because it America and Australia has caused a risk of occur-
is the numerator of Equation 23.17 that is hard to rence of urban–wildland interface fire disasters.
quantify for materials in practice, Equation 23.17 Albini reviewed research resources for forest and
can be rewritten as wildland fires [35]. Recent investigation
718 Y. Hasemi
Table 23.1 Effective opposed-flow flame spread properties [14]

Material Tig ( C) kρc (kW2s/m4K2) Φ (kW2/m3) Ts,min ( C) Φ/kρc (mK2/s)
PMMA polycast (1.59 mm) 278 0.73 5.4 120 8
Polyurethane (S353M) 280 — — 105 82
Hardboard (6.35 mm) 298 1.87 4.5 170 2
Carpet (acrylic) 300 0.42 9.9 165 24
Fiberboard, low density (S119M) 330 — — 90 42
Fiber insulation board 355 0.46 2.2 210 5
Hardboard (3.175 mm) 365 0.88 10.9 40 12
Hardboard (S159M) 372 — — 80 18
PMMA type g (1.27 cm) 378 1.02 14.4 90 14
Asphalt shingle 378 0.70 5.3 140 8
Douglas fir particle board (1.27 cm) 382 0.94 12.7 210 14
Wood panel (S178M) 385 — — 155 43
Plywood, plain (1.27 cm) 390 0.54 12.9 120 24
Chipboard (S118M) 390 — — 180 11
Plywood, plain (0.635 cm) 390 0.46 7.4 170 16
Foam, flexible (2.54 cm) 390 0.32 11.7 120 37
GRP (2.24 mm) 390 0.32 9.9 80 31
Mineral wool, textile paper (S160M) 400 — — 105 34
Hardboard (gloss paint) (3.4 mm) 400 1.22 3.5 320 3
Hardboard (nitrocellulose paint) 400 0.79 9.8 180 12
GRP (1.14 mm) 400 0.72 4.2 365 6
Particle board (1.27 cm stock) 412 0.93 4.2 275 5
Gypsum board, wallpaper (S142M) 412 0.57 0.79 240 1
Carpet (nylon/wool blend) 412 0.68 11.1 265 16
Carpet #2 (wool, untreated) 435 0.25 7.3 335 30
Foam, rigid (2.54 cm) 435 0.03 4.0 215 141
Polyisocyanurate (5.08 cm) 445 0.02 4.9 275 201
Fiberglass shingle 445 0.50 9.0 415 18
Carpet #2 (wool, treated) 455 0.24 0.8 365 4
Carpet #1 (wool, stock) 465 0.11 1.8 450 17
Aircraft panel epoxy Fiberite 505 0.24 * 505 *
Gypsum board, FR (1.27 cm) 510 0.40 9.2 300 23
Polycarbonate (1.52 mm) 528 1.16 14.7 455 13
Gypsum board, (common) (1.27 mm) 565 0.45 14.4 425 32
Plywood, FR (1.27 cm) 620 0.76 * 620 *
Polystyrene (5.08 cm) 630 0.38 * 630 *
Note: Values are only significant to two places
*Flame spread was not measurable
witnesses the increase of the frequency and the summary of the fire spread velocity for urban and
hazard of forest fires in polar Russia, Alaska, and wildland fires superimposed on the correlation
Canada since about the 1990s [36]. between fire spread velocity and wind velocity
Flame spread in a forest or in an urban district summarized on Japanese urban fires from the
depends on radiant heat transfer, convective 1910s to the 1950s [37, 38]. The significant fire
heating, and leeward spread of fire brands due to spread velocity recorded in past urban and wild-
wind or the slope of the terrain. Figure 23.14 is a land fires, from the order of 100 m/h to 10 km/h,
www.ebook777.com
Fig. 23.14 Comparison of

rates of fire spread for
urban and wildland fires as Cribs
a function of wind speed Gorse and heather
20 Urban fires in Japan
(From Kawagoe [38] and
Thomas [37])
15
Spread velocity (m/mm)

10
0 5 10 15 20 25 30
Wind velocity (m/s)
cannot be explained by convective or radiant

heating and suggests the importance of the role Flame Spread Over Liquids
of fire brands in the wind-aided spread of mass
fires. Flame spread over a horizontal pool of liquid
In urban fires, collapse of wooden buildings fuels is essentially opposed flow, but its spreading
due to fire is the typical and most significant velocity is very often significantly larger than
source of fire brands. In wildland and forest estimated from the thermal theory [39]. The dif-
fires, the porous bush along the terrain is ference is attributed to the convective flow within
involved, and dried bush, fallen bark and pine the liquid moving concurrently with the flame.
needles, and so forth can be typical sources of fire The convection is due to the surface tension,
brands. Even though weight per area of bush, which is reduced with increase of temperature
bark, and dried leaves is small and their flaming and pulls the flame toward the unburnt surface
might not last so long, they are quick to ignite by of the liquid. This is illustrated in Fig. 23.15 for a
fire brand due to the large surface area to weight thin liquid layer, δ. Under steady conditions, the
ratio of these materials. In that sense, fire brands viscous forces on the control volume are balanced
should have primary importance in the growth of by the surface tension forces. Thus, the shear
forest fires. However, in more severe fires, the stress, τ, at the bottom surface equals the surface
crowns of the trees may also be involved. tension gradient (dσ/dx) along the free surface:
720 Y. Hasemi
Fig. 23.15 Enhanced

flame spread speed in Flame
Tig, Flashpoint Control volume
liquids due to surface-
tension induced flow V
x
Liquid fuel
δ Ts
layer y
Fig. 23.16 Relationship Liquid-phase Gas-phase

between the liquid effects effects
temperature and the rate
of plane flame spread
of methanol, based on 100
Akita [40]
Stoichiometric condition
Flashpoint
V (cm/s)
10
Pulsating spread
1
–20 –10 0 10 20 30
Ts (°C)
τ ¼ dσ=dx ¼ ðdσ=dT Þ ð∂T=∂xÞ ð23:19Þ provided σ(T ), the surface tension, is known as a
function of temperature for the liquid, and Δ can
be estimated for the conditions of speed. Also, δ,
For a thin liquid layer, the surface tension effect as in the thermally thick case for solids, is only
results in nearly a Couette flow (constant shear) the physical liquid depth for pools less than about
over the layer thickness, δ. Hence, it can be 1 mm and, therefore, must be reinterpreted for
approximated that pools of larger depth. For example, one might
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
τ ¼ ðμ∂u=∂yÞ y¼0 μV=δ ð23:20Þ estimate δ as ðμ=ρÞ ðΔ=V Þ for the deep-
pool case.
where μ is the liquid viscosity. By further Typical flame spread characteristics over a
approximating the surface tension gradient as a liquid fuel are sketched in Fig. 23.16 for liquid
difference over length Δ, the flame speed can be methanol from the data of Akita [40]. Below the
estimated as flashpoint, Ts < Ti g ~ 11 C, the spread is

governed by transport phenomena within the liq-
σðT s Þ σ T ig δ uid fuel. For initial liquid bulk temperatures
V¼ ð23:21Þ
μΔ above the flashpoint, a flammable mixture
www.ebook777.com
always exists everywhere above the surface so δ fuel thickness

that propagation is governed by gas-phase δfc characteristic preheat distance
effects. Above a liquid temperature, which D characteristic length of the burning area
corresponds to stoichiometric conditions above (height, etc.)
the surface, the flame speed remains constant and Da Damköhler number
usually above the normal premixed laminar Δ distance from the pyrolysis front
flame speed. A study by Ito and Kashiwagi [41] ΔH heat of combustion
used holographic interferometry to examine the ΔHox ΔH/r
liquid phase for subflashpoint liquid bulk g gravitational acceleration
x f =Q_ ‘
temperatures. They examined the pulsating *n
k 2=3 2=3
region depicted in Fig. 23.16, and the adjacent k0 constant k3=2 =c p T o g1=2 Q_ ‘
uniform region of spread just below the flame; k x f =Q_ ‘
both appear to contribute to flame spread rate in kg gas thermal conductivity
the uniform region. L heat of gasification
μ viscosity
q00 heat flux due to gas-phase conduction
00
Summary qf flame heat flux
00
qfc characteristic flame heat flux
This chapter has provided the practicing engineer Q_ ‘ heat release rate per unit width
Q_
*
with some insight into the nature of fire spread ‘ dimensionless heat release rate per unit
over materials. In general, surface flame spread width
Q_ ‘e
*
depends on the heat transfer processes at the effective heat release rate per unit width
flame front. These transport processes depend not r stoichiometric mass ratio oxygen/fuel
only on the fuel but also on the fuel’s configuration ρ density
and orientation and on ambient environmental ϕ impulse response function
conditions. Thus, estimates of flame spread require Φ opposed-flow preheat factor (numerator
complex analysis and specific material data. The in n)
current state of knowledge does provide limited σ surface tension
formulas and material data to make some t time
estimates. In this chapter, the full scope of flame tb time to local burnout
spread phenomena has not been addressed. For tc characteristic decay time of pyrolysis
example, flame spread in mines, ducts, and tig* characteristic time to ignition
buildings presents an entirely new and complex τ time
array of conditions. Thus, flame spread on Tf flame temperature
materials must be evaluated in the context of Tig ignition temperature
their use, and appropriate data must be made avail- To initial surface temperature
able for proper assessments of materials. Tr reference temperature
Ts surface temperature
T1 ambient temperature
Nomenclature Vg gas velocity
00
Vp flame spread velocity
a k f Q_ ‘ xf flame length
b flame spread acceleration factor xp pyrolysis front length
(¼ (a 1) tig*/tb) xpo pilot flame length
c specific heat of solid xpoff maximum pyrolysis front length
cg specific heat of gas Yox,1 mass fraction of oxygen
cp specific heat of air x, y coordinates
722 Y. Hasemi
Superscripts Standards and the Society of Fire Protection

Engineers, ASTM STP 882, ASTM International,
W. Conshohocken, PA, pp. 239–267 (1985).
· per unit time 15. B. Karlsson, “Models for Calculating Flame Spread
0
per unit length on Wall Lining Materials and the Resulting Heat
00 Release Rate in a Room,” Fire Safety Journal,
per unit area
23, pp. 365–386 (1994).
16. D. Baroudi, “A Discrete Dynamical Model for Flame
Spread over Combustible Flat Solid Surfaces Subject
References to Pyrolysis with Charring—An Application Example
to Upward Flame Spread,” Fire Safety Journal,
1. T. Hirano, S. Noreikis, and T. Waterman, “Measured 38, pp. 53–84 (2003).
Velocity and Temperature Profiles of Flames Spread- 17. J.G. Quintiere, M. Harkleroad, and Y. Hasemi, “Wall
ing over a Thin Combustible Solid,” Combustion and Flames and Implications for Upward Flame Spread,”
Flame, 23, p. 83 (1974). Combustion Science and Technology, 48, 3–4,
2. D.D. Drysdale, A.J.R. Macmillan, and D. Shilitto, pp. 191–222 (1986).
“King’s Cross Fire: Experimental Verification of the 18. L. Orloff, A.T. Modak, and R.L. Alpert, “Burning of
“Trench Effect,” Fire Safety Journal, 18 (1992). Large-Scale Vertical Surfaces,” in Proceedings of the
3. D.D. Drysdale, An Introduction to Fire Dynamics, 16th Symposium (International) on Combustion, Com-
2nd ed., John Wiley and Sons, New York (1999). bustion Institute, Pittsburgh, PA, pp. 1345–1354 (1976).
4. J.G. Quintiere, Fundamentals of Fire Phenomena, 19. T. Ahmad and G.M. Faeth, “Turbulent Wall Fires,” in
John Wiley and Sons, New York (2006). Proceedings of the 17th Symposium (International) on
5. A.C. Fernandez-Pello and T. Hirano, “Controlling Combustion, Combustion Institute, Pittsburgh, PA,
Mechanism of Flame Spread,” Combustion Science pp. 1149–1160 (1979).
and Technology, 32, pp. 1–31 (1983). 20. M. Sibulkin and J. Kim, “The Dependence of Flame
6. T. Hirano, “Physical Aspects of Combustion in Fires,” Propagation on Surface Heat Transfer ii. Upward
in Proceedings of the 3rd International Symposium on Burning,” Combustion Science and Technology,
Fire Safety Science, International Association for Fire 17, pp. 39–49 (1977).
Safety Science, Boston, MA, pp. 27–44 (1991). 21. K. Saito, J.G. Quintiere, and F.A. Williams, “Upward
7. F. Williams, “Mechanism of Fire Spread,” in Turbulent Flame Spread,” in Proceedings of the 1st
Proceedings of the 16th Symposium (International) International Symposium on Fire Safety Science,
on Combustion, Combustion Institute, Pittsburgh, International Association for Fire Safety Science,
PA, pp. 1281–1294 (1976). London, pp. 75–86 (1985).
8. M.A. Delichatsios, “Modeling of Aircraft Cabin 22. D. Baroudi and M. Kokkala, “Analysis of Upward
Fires,” Technical Report, Factory Mutual Research Flame Spread,” VTT Publications, 89 (1992).
Corp. (1984). 23. Y. Hasemi and N. Yasui, “A Strategy to Develop
9. Y. Hasemi, “Thermal Modeling of Upward Wall Engineering Upward Flame Spread Evaluation Meth-
Flame Spread,” in Proceedings of the First Interna- odology Based on the Linearized Flame Height
tional Symposium on Fire Safety Science, Interna- Approximation,” Fire Science and Technology,
tional Association for Fire Safety Science, Boston, 15, 1–2, pp. 17–28 (1995).
MA, pp. 87–96 (1985). 24. G. Grant and D.D. Drysdale, “Numerical Modelling
10. T.I. Eklund, “A Vortex Model for Wall Flame of Early Flame Spread in Warehouse Fires,” Fire
Height,” Journal of Fire Science, 4, pp. 4–14 (1986). Safety Journal, 24, pp. 247–278 (1995).
11. M. Kokkala, D. Baroudi, and W.J. Parker, “Upward 25. A.C. Fernandez-Pello, “Upward Laminar Flame
Flame Spread on Wooden Surface Products: Spread Under the Influence of Externally Applied
Experiments and Numerical Modelling,” in Thermal Radiation,” Combustion and Flame,
Proceedings of the 5th International Symposium on 17, p. 87 (1977).
Fire Safety Science, International Association for Fire 26. H. Ingason and J. de Ris, “Flame Heat Transfer in
Safety Science, Boston, MA, pp. 309–320 (1997). Storage Geometries,” Fire Safety Journal, 31, pp.
12. J.S. Newman and C.J. Wieczorek, “Chemical Flame 39–60 (1998).
Heights,” Fire Safety Journal, 39, pp. 375–382 27. Y. Hasemi, M. Yoshida, N. Yasui, and W.J. Parker,
(2004). “Upward Flame Spread along a Vertical Solid for
13. K.-M. Tu and J.G. Quintiere, “Wall Flame Heights Transient Local Heat Release Rate,” in Proceedings
with External Radiation,” Fire Technology, of the 4th International Symposium on Fire Safety
pp. 195–203 (Aug. 1991). Science, International Association for Fire Safety Sci-
14. J.G. Quintiere and M. Harkleroad, “New Concepts for ence, Boston, MA, pp. 385–396 (1994).
Measuring Flame Spread Properties,” in Proceedings 28. M.M. Delichatsios, P. Wu, M.A. Delichatsios,
of Fire Safety: Science and Engineering, a symposium G.D. Lougheed, G.P. Crampton, C. Qian, H. Ishida,
sponsored by ASTM Committee E-5 on Fire and K. Saito, “Effect of External Radiant Heat Flux on
www.ebook777.com
Upward Fire Spread: Measurements on Plywood and Combustion, Combustion Institute, Pittsburgh, PA,
Numerical Predictions,” in Proceedings of the 4th pp. 241–252 (1968).
International Symposium on Fire Safety Science, 35. F.A. Albini, “An Overview of Research on Wildland
International Association for Fire Safety Science, Fire,” in Proceedings of the 5th International Sympo-
Boston, MA, pp. 421–432 (1994). sium on Fire Safety Science, International Association
29. B.Y. Lattimer, S.P. Hunt, M. Wright, and C. Beyler, for Fire Safety Science, Boston, MA, pp. 59–74
“Corner Fire Growth in a Room with a Combustible (1997).
Lining,” in Proceedings of the 7th International Sym- 36. H. Hayasaka, “Recent Large-Scale Fires in Boreal and
posium on Fire Safety Science, International Associa- Tropical Forests,” Journal of Disaster Research, 2, 4,
tion for Fire Safety Science, Boston, MA, pp. 276–283 (2007).
pp. 419–430 (2002). 37. P.H. Thomas, “Rates of Spread of Some Wind-Driven
30. Y. Hasemi, D. Nam, and M. Yoshida, “Experimental Fires,” Forestry, 44, 2 (1971).
Flame Correlations and Dimensional Relations in 38. K. Kawagoe (ed.), Fire Safety in Buildings, Architec-
Turbulent Ceiling Fires,” in Proceedings of the 5th tural Studies and Engineering Series, 21, Shokokusha
Asia Oceania Symposium on Fire Science and Tech- (1970) (in Japanese).
nology, International Association for Fire Safety Sci- 39. I. Glassman and F.L. Dryer, “Flame Spreading Across
ence, Boston, MA, pp. 379–390 (2001). Liquid Fuels,” Fire Safety Journal, 3, pp. 123–138
31. Y. Hasemi, M. Yoshida, Y. Yokobayashi, and (1980).
T. Wakamatsu, “Flame Heat Transfer and Concurrent 40. K. Akita, “Some Problems of Flame Spread Along a
Flame Spread in a Ceiling Fire,” in Proceedings of the Liquid Surface,” in Proceedings of the 14th Sympo-
5th International Symposium on Fire Safety Science, sium (International) on Combustion, Combustion
International Association for Fire Safety Science, Institute, Pittsburgh, PA, pp. 1075–1083 (1973).
Boston, MA, pp. 379–390 (1997). 41. A. Ito and K. Kashiwagi, “Characterization of Flame
32. V. Stenstad and B. Karlsson, “Flame Spread on a Spread over PMMA Using Holographic Interferome-
Solid Wooden Ceiling,” Conference Proceedings of try Sample Orientation Effects,” Combustion and
Interflam 2007, 1, Interscience Communications, Flame, 71, pp. 189–204 (1988).
London, UK, pp. 45–57 (2007).
33. A.C. Fernandez-Pello, S.R. Ray, and I. Glassman, in
Proceedings of the 18th Symposium (International) on Yuji Hasemi is professor in the Department of Architec-
Combustion, Combustion Institute, Pittsburgh, PA ture, Faculty of Science and Engineering, Waseda
(1981). University, Tokyo, Japan.
34. J. de Ris, “Spread of a Laminar Diffusion Flame,” in
Proceedings of the 12th Symposium (International) on
Smoke Characterization and Damage

Potentials 24
Jeffrey S. Newman, Geary G. Yee, and Paul Su
Introduction water damage such as in the case of fire sprinkler

protection). In general, fire damage in the “near-
Smoke is a mixture of (1) particulates consisting field” is dominated by heat/water damage with
of soot, semi-volatile organic compounds smoke of a much lesser impact, while smoke can
(SVOC), and solid inorganic compounds; and (2) - often be the governing damage mechanism in the
non-particulates consisting of very volatile “far-field” and, due to extent of travel and area
organic compounds, volatile organic compounds, coverage, of far greater impact.
and liquid and gaseous inorganic compounds. Figure 24.1 illustrates, for example, the vari-
Soot creates bridging between electrical ous components that are necessary to evaluate
conductors and conveys corrosive products, smoke damage potentials, especially in the “far-
resulting in damage to electronics and electrical field”. As shown in the figure, the impact of the
circuits through leakage current and corrosion, smoke deposition profile resulting from smoke
while SVOC and non-particulates stain and impart release, typically from a fire event, is quantified
malodor to surfaces. Soot is also a very effective by the comparison of a defined damage function
adsorbent and transport mechanism for SVOC, with its associated damage threshold. As will be
non-particulates and inorganic compounds. subsequently discussed, the damage function can
The smoke problem (exclusive of toxicity or represent a variety of types of smoke damage
escape potential considerations) is ultimately including leakage current, corrosion and stain/
characterized by the quantification of damage odor. The deposition profile is assessed through
due to the deposition of combustion products the coupling of smoke generation, the character-
onto building surfaces and contents (e.g., equip- ization of pertinent smoke properties and the
ment, furnishings, and commodities). It is instruc- transport of smoke resulting in time and spatially
tive to categorize the assessment of smoke resolved concentration profiles. Detection and
damage potentials into two regimes: “far-field” active response both affect and are affected by
at some distance away from the fire/smoke source these concentration profiles. Knowledge of the
and “near-field” close to the source, where it is smoke deposition velocity is also an important
likely to have simultaneous heat damage (and component for the quantification of the resulting
smoke deposition profile.
For reference, the solid gray-shaded blocks
J.S. Newman (*)
shown in Fig. 24.1 will be covered in detail in
Retired from FM Global, 1151 Boston Providence
Turnpike, Norwood, MA 02062 USA this Chapter while the three unshaded blocks
labeled Concentration Profile, Transport and
G.G. Yee • P. Su
FM Global, 1151 Boston Providence Turnpike, Generation are place holders that are outside of
Norwood, MA 02062 USA this Chapter’s scope. In should be noted that

www.ebook777.com
24 Smoke Characterization and Damage Potentials 725
Fig. 24.1 Outline of components to evaluate smoke damage potentials
Generation refers to heat and gas species’ pro- thermodiffusion, and electrostatic precipitation
duction and is a function of the prescribed fire [4]. For electrically neutral aerosols, the
scenario with the content of Chap. 36 of this governing mode of smoke deposition is primarily
Handbook providing useful specific smoke and dependent on whether the transport flow is lami-
heat generation properties for a variety of nar or turbulent and the particle size.
materials. Transport and Concentration Profile The flow regime for smoke particle motion
data can be obtained from any number of physi- can be characterized by the magnitude of the
cal correlations [1] or computer models [2, 3] particle Reynolds number, i.e.,
coupled with the prescribed Generation informa-
ρg d p V
tion. Finally, Detection/Response is shown in Rep ¼ ð24:1Þ
light gray as it is covered only as an overview η
of important considerations as relating to the where ρg is the gas density, dp the particle diam-
smoke Deposition Profile with details on detec- eter, V the particle velocity and η is the gas
tion given in Chap. 40 of this Handbook. viscosity.
Laminar Flow (Rep < 1)

Deposition Profile 1. For smoke particles >1 μm in diameter, the
deposition is primarily due to gravitational
Deposition Velocity settling. Stokes’s Law applies to particle
motion when inertial forces are negligible
The process by which smoke can deposit on compared with viscous forces, and gives the
various surfaces is often complex and can result particle terminal settling velocity, VTS as
from either a single dominant physical mecha-
ρ p d2p g
nism or a combination of mechanisms including V TS ¼ ð24:2Þ
18η
particle inertia, sedimentation or gravitational
settling, Brownian diffusion, thermophoresis or where g is gravitational acceleration.
726 J.S. Newman et al.
2. For smoke particles <1 μm in diameter, both the turbulent region is given sufficient lateral
Brownian diffusion as well as gravitational velocity toward the wall to be transported
settling are important deposition mechanisms. through the laminar layer and deposit on the
An important assumption in Equation 24.2 is wall (surface). The particle settling velocity
that the relative velocity of the gas at the transitions from a dp2 dependence in the Stokes
surface of the particle is zero, which is not regime to a dp½ dependence in the Newton drag
met for particle sizes that approach the mean regime. This transition can be accounted for by
free path of the gas. The particles move faster the introduction of a drag coefficient, CD, in the
than predicted by Stokes’s law and a slip calculation of settling velocity, i.e.,
correction factor, Cc, is needed [5]: " #1=2
4ρ p d p g
2:52Λ V TS ¼ ð24:4Þ
Cc ¼ 1 þ ð24:3Þ 3CD ρg
dp
where Λ is mean free path, which is 0.066 μm Figure 24.2, for example, illustrates the
for air at 1 atm and 20 C. The slip correction functional relationship between CD and
factor is a multiplicative factor applied to the Rep for spherical particles. Numerous
right-hand side of Equation 24.2, increasing correlations have been developed to describe
the terminal settling velocity. the drag coefficient and Reynolds number
functional relationship. A particularly useful
Turbulent Flow (Rep > 1) formulation covers the transition region (1 <
1. For smoke particles >1 μm in diameter, the Rep < 1000) [4]:
deposition begins to be influenced by both 24
particle inertia as well as viscous effects, CD ¼ 1 þ 0:15Re p 0:687 ð24:5Þ
Re p
which was first shown by Friedlander and
Johnstone in 1957 [6]. Inertial deposition 2. For smoke particles <1 μm in diameter,
occurs when a particle near a wall (surface) in Brownian diffusion becomes important. The
105
104
Drag Coefficient, CD
103
102 24
CD =
Rep
101
100 CD = 0.44
Stokes's Law Transition Region Newton's Law

-1
10
10-3 10-2 10-1 100 101 102 103 104 105
Reynolds Number, Rep
Fig. 24.2 Drag coefficient versus particle Reynolds number
www.ebook777.com
diffusion coefficient of an aerosol/smoke par- Table 24.1 Smoke deposition velocities [9]
ticle can be expressed in terms of particle Deposition velocity (104 m/s)
properties by the Stokes-Einstein derivation 1.0 m3 1200 m3
[7]. In this derivation, the diffusion force on Material smoke box enclosure
the particles, which causes the particles’ net Nylon 2.5 –
motion along the concentration gradient, is Polycarbonate 1.9 –
equated to the force exerted by the gas Polyethylene – 1.2
resisting the particles’ motion, i.e., Polymethylmethacrylate 2.1 1.9
Polypropylene – 1.3
kB TCc Polyvinylchloride 1.4 –
D¼ ð24:6Þ
3πηdp Polystyrene 6.9 7.3
where D is the particles’ diffusion coefficient,

wall, the primary deposition mechanism is gen-
kB Boltzmann’s constant and T is the absolute
erally accepted as thermophoretic, with the depo-
temperature.
sition velocity dominated by the local
Thermophoresis can be important where large
temperature gradient. Thus, the thermophoretic
temperature gradients exist, e.g., from a smoke
smoke deposition velocity [10–12] is given by:
hot layer to a cold surface such as a wall or

ceiling in an enclosure. Thermophoresis is the K TH η dT
aerosol particle movement that results from the V TH ¼ ð24:7Þ
ρg T dx
force exerted by a temperature gradient in a gas,
resulting in a flow in the direction of decreasing where the gas properties are evaluated at the film
temperature. This would be the dominant mech- temperature, i.e., the average between the wall
anism for very small particles <0.1 μm, even at and gas temperature at the location of interest.
small gradients of 1 C/cm [4]. dT/dx is the gas phase temperature gradient at the
In addition, another phenomenon can occur at surface. KTH is the thermophoretic velocity coef-
relatively high smoke concentrations. Cloud ficient which approaches a constant value of 0.55
settling or mass subsidence occurs when the as the Knudsen Number, given by the ratio of the
smoke concentration is sufficient to cause the molecular mean free path length Λ to the smoke
entire cloud to move as an entity at a velocity particle diameter dp, becomes large (>10)
significantly greater than the individual particle [11]. Examples of the characterization and quan-
settling velocity [8]. (An aerosol or smoke cloud tification of fire-induced thermophoretic deposi-
is defined as a region of high smoke concentration tion are given in Refs. [13–16].
having a definite boundary in a much larger region
of smoke-free air.) The difference in density
causes the bulk motion of the smoke particles. Detection Overview
Table 24.1 gives illustrative examples of
experimentally measured smoke deposition While the detailed design of smoke detection
velocities for “far-field” conditions, i.e., systems is given elsewhere in this Handbook
locations away from the combustion/smoke (Chap. 40), it is instructive to present here a
source, where the primary observed deposition brief overview of detection criteria as well as
mechanism was characterized as gravitational the potential need for specific smoke
settling [9]. As indicated in the table, the deposi- characteristics. Fire detection is achieved by
tion velocities are on the order of 104 m/s and using various types of detectors: (1) heat
are quite similar in two different scale enclosures detectors (e.g., fixed-temperature, rate-of-rise
(1 m3 smoke box and a 1200 m3 enclosure). For sensors); (2) chemical compound – smoke
“near-field” conditions, e.g., on the ceiling above detectors (e.g., ionization, photoelectric sensors
a fire or for direct flame impact on an adjacent and gas detectors such as CO or CO2 sensors);
(3) flame detectors (e.g., ultraviolet and/or infra- Smoke Characterization

red sensors), etc. For effective detection of a fire,
the most important parameter to evaluate is the The characterization of smoke requires increas-
total time associated with: ingly more specific information concerning
1. the occurrence of a specified hazard to people, instantaneous smoke properties, particularly
buildings and contents created by the fire, tH; when assessing fire environments in enclosures.
2. the transit time of the fire product(s) to the These properties include volume fraction, mass
detector location, tt; concentration, generation efficiency, particle
3. the fire growth time to reach a detectable level size and yield. Smoke characteristics and yields
of fire product(s) at the detector location, tf; of chemical compounds can change dramati-
4. the detector response time once tf has cally with a change in ventilation conditions,
occurred, tD; and such as in an enclosure environment where the
5. the “effective” response time once the fire has supply of air may be restricted. Quantifying
been detected, tE. The relationship between changes in smoke properties have been shown
these times can be expressed as [17, 18]: to be useful in addressing the impact of ventila-
tion on fire chemistry, which in turn relates to
tr ¼ tH tt þ tf þ tD þ tE ð24:8Þ the smoke damage potential of a given fire
scenario.
where tr ¼ residual time, which must be greater Optical techniques to characterize smoke are
than or equal to zero. As an example, tH depends especially attractive since they provide continu-
on the defined hazard, which is a function of ous measurements without disturbing the envi-
material properties and configuration and, for ronment. When a monochromatic beam of light
the smoke problem, would be the time to reach passes through smoke, the fraction of light trans-
a specified damage threshold as outlined in mitted can be expressed as
Fig. 24.1. Similarly, tf is identified with the gen-
eration step in the figure and tt with the I
¼ expðODλ lÞ ð24:10Þ
transport step. Io
In general, the response of a smoke (or gas
species) detector can be treated as a first-order where I is the transmitted intensity; Io is the
system coupled with a time lag, i.e., initial intensity; ODλ is the optical density (also
referred to as the turbidity or extinction coeffi-
dCs 1 cient) of the smoke (m1); and l is the optical
¼ ½Co ðt tl Þ Cs ðtÞ ð24:9Þ
dt τ path length (m). The optical density is an intri-
cate function of the smoke particle diameter, the
where Cs is the instantaneous smoke concentra-
smoke size distribution, the incident light wave-
tion as measured by the detector at time t; Co,
length and the complex refractive index of the
identified with the concentration profile step in
smoke. It has previously been shown that
Fig. 24.1, is the “true” of reference concentration
the smoke volume fraction (i.e., the ratio of the
(outside the detector) at time t tl; τ is the
smoke particle volume to the total gas volume)
detector time constant; and tl is the lag time.
for a variety of materials can be related to the
Two apparent detector characteristics, τ and tl,
optical density through the following expression
can be noted in Equation 24.9. τ can be thought
[19]:
of as the time constant for the specific sensor
used in the detector, while tl is associated with a ODλ λ
delay in smoke (or a specific gas species) trans- fv ¼ ð24:11Þ
c
port to the sensor such as through filters, sam-
pling lines, etc. Equation 24.9 can be used to where λ is the incident wavelength; and c is
assess the detection time associated with the defined as an average coefficient of smoke
detection/response step in Fig. 24.1. extinction. For over-ventilated conditions, the
www.ebook777.com
X
smoke mass concentration, Cs (kg/m3) can be ni ln di
related to the smoke volume fraction through ln d g ¼ ð24:13Þ
N
the smoke particle density, i.e.,
where ni is the number of particles with diameter
ρ ODλ λ di and N is the total number of particles. The
Cs ¼ ρs f v ¼ s ð24:12Þ
c geometric standard deviation, σ g, is given by:
Reported values [19–23] for ρs typically "X 2 #1=2

ni ln di ln dg
range from 800 to 1800 kg/m3, with 1100 to lnσ g ¼ ð24:14Þ
1500 kg/m3 common values. This perceived N1
discrepancy in density is primarily related to
The geometric standard deviation is dimension-
the measurement technique, although combus-
less with a value equal to or greater than 1.0. One
tion conditions and the specific material smoke
geometric standard deviation represents the
source can influence the “apparent” density. For
range of particle sizes from (dg/σ g) to (dg σ g).
example, a value of 1100 kg/m3 has been shown
A variety of representative diameters can be used
to be the resulting “apparent” density for a wide
based on particle volume or mass. (For example,
range of materials during over-ventilated
Refs. [15] and [30] tabulate the geometric parti-
flaming combustion and is recommended when
cle diameters and distributions for a number of
optical methods are used to assess the relation-
materials under both flaming combustion and
ship represented by Equation 24.12 with a value
pyrolysis.) A particularly useful particle repre-
for c of 7.0 [19]. This value for the density has
sentation uses the concept of an aerodynamic
also been shown to derive consistent smoke
equivalent diameter (da), which is the diameter
yields for various materials. However, it is
of a unit-density sphere having the same
more important to maintain the ratio of ρs to
gravitational-settling velocity as the measured
c as a constant than to use a specific value for
particle. The aerodynamic equivalent diameter
the smoke particle density. Thus, if value of 8.0
takes into account the shape, roughness, and
is used for c, a value of 1300 kg/m3 should be
aerodynamic drag of the particle [4]. For exam-
used for the corresponding density. It should be
ple, Stokes’s law, given in Equation 24.2,
noted that the ratio of c/ρsλ is equivalent to the
rewritten in terms of da and a standard particle
often referenced mass specific extinction coeffi-
density, ρ0 (1000 kg/m3) gives the following for
cient, which has been extensively tabulated and
the terminal settling velocity, VTS in air:
generally shown to be relatively independent of
fuel type [24–27]. ρ0 d a 2 g
V TS ¼ ð24:15Þ
18η
Smoke Particle Size Table 24.2 presents the aerodynamic equivalent

mass mean diameter, dag for smoke generated
Smoke particle sizes are typically characterized from flaming plastic fires in various scale
by specifying some representative particle diam- enclosures [9] and ducts/low velocity wind
eter with a corresponding size distribution. Parti- tunnels [28, 29]. The width of the distribution,
cle size characteristics are often represented by a σ g is also given for reference. References [15]
log-normal particle size distribution function. offers a detailed summary review of aerody-
For this type of distribution, the log of the parti- namic diameters from a number of additional
cle diameter follows a Gaussian distribution. materials including gases and liquids under vari-
Therefore, the geometric mean diameter, dg, is ous other configuration and combustion
given by: conditions.
Table 24.2 Aerodynamic mass mean diameter of smoke from flaming plasticsa
Material dag, μm σg Environment
Nylon 0.4 2.0 1.0 m3 smoke box
Polycarbonate 3.0 3.4 1.0 m3 smoke box
Polyethylene 1.0 2.5 1200 m3 enclosure
Polymethylmethacrylate 2.3 4.4 1200 m3 enclosure
0.7–1.0 NR 0.37 m2 duct [28]
Polypropylene 1.2 2.0 1200 m3 enclosure
Polyurethane 2.0 1.8 0.18 m2 duct [29]
Polyvinylchloride 1.1 1.8 1.0 m3 smoke box
Polystyrene 2.0 2.6 1.0 m3 smoke box
2.4 2.1 1200 m3 enclosure
1.5–2.5 NR 0.37 m2 duct [28]
NR- Not Reported
a
Data from Ref. [9] except as noted
humidity (RH) is shown in Fig. 24.4 [34]. Values

Damage Functions of measured LC somewhat surprisingly remain
nearly constant at 1.8 0.3 104 A over the
Leakage Current range of relative humidity from 30 % to 90 %.
For reference, measured LC data for targets with-
Leakage current between two conducting out exposure are also plotted in the figure, and as
elements occurs as a result of circuit bridging expected are much smaller than for targets
due to the presence of a conducting medium exposed to polycarbonate smoke. For example,
between the elements, such as water, conductive at 90 % RH, the LC of an unexposed target at
ions, soot, or dust. The increase of leakage cur- 1.1 106 A is much smaller than the LC of
rent on electronic circuit boards can change over- 1.8 104 A for a target exposed to the poly-
all circuit characteristics, e.g., degrade and/or carbonate smoke. This indicates that the effect of
damage circuit properties. This principle has soot bridging is the major contributor to the
been used to design a leakage current target for increase of LC for targets exposed to polycar-
the measurement of smoke corrosivity bonate smoke and the contribution due to
[31–33]. The target, as shown in Fig. 24.3, has increase of RH is relatively minor. Reference
a comb-like pattern made of thin strips of copper [34] also gives similar data for nylon and
with 40 insulating spaces between them. The polyvinylchloride. Figure 24.5 presents
dimensions of the target are shown in the figure. normalized smoke leakage currents for nylon,
Specifically, for example, with respect to the polycarbonate and polyvinylchloride. The data
manufacture of semiconductors, leakage current in the figure were normalized by dividing the
is uncontrolled (“parasitic”) current flowing leakage current results as illustrated in Fig. 24.4
across region(s) of semiconductor structure/ by the average smoke deposition (g/m2) which
device in which no current should be flowing. occurred over the duration of the LC target
Leakage is one of the main factors limiting exposure. As indicated in Fig. 24.5, the average
increased computer processor performance. The normalized LC ranges from 7.2 0.5 105
presence of ionic compounds and soot in smoke A-m2/g for nylon to 6.1 0.9 104 A-m2/g
deposited on the surface of a semiconductor is for polycarbonate. Relative humidity appears to
expected to damage the processor’s functionality have a limited effect, with a trend to slightly
through an increase in leakage current. lower normalized LC above 60 % RH. Table 24.3
An example of leakage current (LC) data for summaries the average results for the three
polycarbonate smoke deposition versus relative materials studied.
www.ebook777.com
1.6 mm thick
fiber-glass
substrate
51±0.2 18.5±0.1
26.5±0.1
23±0.5
28±0.5
25±0.1
6±0.5
10±0.5
37±0.2
21.5±0.1
2.5 mm
diameter, 5
mounting
51±0.2 holes
Tinned circuit
elements All dimensions in mm
Fig. 24.3 Sketch of leakage current target (Taken from Ref. [32])
Fig. 24.4 Leakage current 100

versus RH for LC targets
exposed to polycarbonate Polycarbonate
smoke Background LC
10-2
Leakage Current (A)
10-4
10-6
10-8
10-10
10-12
0 20 40 60 80 100
Relative Humidity (%)
Corrosion the gradual deterioration of metal surface by

moisture and corrosive chemicals. In aqueous
Corrosion involves the reaction between a or humid environments, corrosion is an electro-
metal or alloy and its environment. It is an chemical reaction in nature; it involves elec-
irreversible interfacial reaction, which causes tron (e) transfer between anodic and cathodic
Fig. 24.5 Normalized 100

leakage current versus
Polycarbonate
Normailized Leakage Current (A-m2/g)

relative humidity
Polyvinylchloride
Nylon
10-2
6.1±0.9 x10-4
2.1±0.1 x10-4
10-4
7.2±0.5 x10-5
10-6
0 20 40 60 80 100
Relative Humidity (%)
Table 24.3 Leakage currents [34] Cathodic reactions:

Average Average Average Hydrogen evolution 2Hþ þ 2e ) H2ðgÞ [5]
leakage smoke normalized Oxygen reduction O2 þ 4Hþ þ 4e ) 2 H2 O [6]
current deposition LC (104
(acid solutions)
Material (105 A) (g/m2) A-m2/g)
Oxygen reduction O2 þ 2H2 O þ 4e ) 4OH [7]
Nylon 18 3 0.30 0.11 6.1 0.9
(neutral or basic
Polycarbonate 9.6 0.5 0.45 0.02 2.1 0.1 solutions)
Polyvinylchloride 0.94 0.07 0.13 0.02 0.72 0.05
In general, corrosion caused by smoke is due

reaction sites. For corroding metals, the anodic to the presence of inorganic anions in smoke such
reaction is the oxidation of a metal to its ionic as chloride (Cl), bromide (Br), and fluoride
state: (F) plus moisture in the environment. In fires,
corrosive combustion products such as HCl, HBr
and HF are emitted along with the other combus-
Anodic reaction: M ) Mnþ þ ne [1]
tion products, which are present as gases, liquids
Specific examples Cu ) Cu2þ þ 2e [2]
and solids. The corrosive combustion products
of anodic reactions: Al ) Al3þ þ 3e [3]
are generally emitted as gases and liquids
Sn ) Sn2þ þ 2e [4]
with inorganic anions in the structure. The
non-corrosive combustion gases and liquids are
The cathodic reaction is a reduction pro- emitted as inorganic and organic compounds and
cess. For metallic corrosion, cathodic water, whereas solids are emitted as soot and
reactions like Reactions [5–7] are frequently inorganic metals and dust. The solid combustion
encountered. In acid solutions, hydrogen evo- products are broadly defined as particulates and
lution and oxygen reduction reactions the gaseous and liquid combustion products are
(Reactions [5] and [6]) are the main cathodic broadly defined as non-particulates. The mixture
reactions. In neutral or basic solutions, oxygen of particulates and non-particulates that include
reduction reaction (Reaction [7]) is the pri- products with inorganic atoms is defined as
mary cathodic reaction. smoke. The main hazards regarding the exposure
www.ebook777.com
Table 24.4 Failure mechanisms and causes for electrical/ Table 24.5 Smoke corrosion rates of copper [34]
mechanical equipment damage
Average Average
Failure smoke normalized
mechanism Failure cause deposition corrosion
Corrosion on filters rate (105
Corrosion Metal contacts, cause open circuits
Material rate (mpy) (105 g/m2) mpy-m2/g)
Shorts Circuit bridging between contacts, cause
Nylon 0.78 0.17 0.38 0.13 2.40 1.50
leakage of current and shorts
Polycarbonate 0.03 0.02 1.24 0.43 0.03 0.03
Contact Coating of electrical contacts
resistance Polyvinylchloride 13.31 4.3 2.19 0.95 6.25 0.76
Binding Mechanical equipment (timers, hard disk
drives, etc., impeded) averaging 13.31 mpy, while solutions from the
polycarbonate tests demonstrated the lowest
corrosion rates averaging 0.03 mpy. Similarly,
of electrical/mechanical equipment to smoke is solutions of polyvinylchloride tests demonstrated
the damage due to circuit bridging, corrosion, the highest normalized corrosion rates among
and binding, defined as smoke corrosivity. Cir- these three materials tested averaging 6.25
cuit bridging occurs in reducing surface insula- 105 mpy-m2/g; solutions of polycarbonate tests
tion and increasing leakage current for digital exhibited the lowest normalized rates averaging
safety systems, multiplexers and functional cir- 0.03 105 mpy-m2/g.
cuit boards. In contrast, corrosion damage by
acids and anions from smoke can be observed
either short term or long after the fire. Smoke Smoke Stain
contamination also leads to other types of elec-
trochemical corrosion degradation of circuit In the characterization of staining of a surface
boards, such as dendrite metal migration between from smoke deposition, it is useful to understand
conduction lines, localized corrosion of uncoated the mechanism by which the surface can undergo
metal wires and contacting areas, etc. Failure a discernable change in appearance. When light
mechanisms and causes for electrical/mechanical strikes an opaque object, as shown in Fig. 24.6,
equipment as a result of exposure to smoke are the total amount of reflected light is char-
listed in Table 24.4 [35, 36]. acterized as two distinctly different light
The linear polarization resistance (LPR) and reflections from the surface: specular reflection
electrical resistance techniques based on electro- and diffuse reflection. Specular reflection is light
chemistry can be used to evaluate smoke-induced that is directed at an angle opposite to the inci-
corrosion on metals. The LPR technique enables dent light and is perceived by the observer to be
the corrosion rate of metals in solution to be glare caused by the shininess or glossiness of the
measured as in milli-inch per year (mpy). For sample. To see the apparent color of the sample,
example, the smoke and acid gases produced observers must move their eyes away from the
during material combustion can be captured and glare (specular) and concentrate on examining
passed through a water-based solution. The the diffuse (scattered) reflectance from the sam-
corrosivity (or corrosion rate) of the solution ple. Any changes in this diffuse reflectance due
can then be measured by the LPR technique. to deposition of smoke would be an indication of
Table 24.5 shows LPR data for nylon, polycar- staining, i.e., potential smoke damage. Bright-
bonate and polyvinylchloride from Ref. [34]. ness is defined as the diffuse reflectivity of an
The normalized corrosion rate is defined as opaque surface to light in the blue portion of the
corrosion rate normalized with the amount of spectrum. Brightness is typically measured at an
soot deposited on filters prior to acid gas collec- effective wavelength of 457 nm. Light of this
tion in a water-based solution. Solutions from the wavelength appears blue to the human eye. In
polyvinylchloride tests demonstrated the highest the appearance of paper, for example, brightness
corrosion rates among these three materials is an especially important property not only
Fig. 24.6 Reflectance

of light
Fig. 24.7 Measurement

of brightness
because of its aesthetic value, but also because of is mounted on the normal and receives light
its effect on legibility and contrast between print reflected along the normal-conditions sometimes
and paper. Brightness should not be confused expressed by the shorthand notation (45 illumi-
with “whiteness”. The degree of whiteness of a nation, 0 observation). An equivalent geometry
paper conveys information on its color neutrality. is shown on the right in the figure, where illumi-
It should be noted that the measurement of nation is provided on the normal and the detector
brightness is also an important tool in the char- is located on the incident at 45 and can be
acterization of the appearance of textiles, espe- described as (0 illumination, 45 observation).
cially in the effectiveness of detergent cleaning While the difference in optical geometry
agents. between the two configurations leads to equiva-
A widely accepted method of brightness mea- lent brightness readings, the 0 /45 geometry in
surement, designated “Brightness 457”, is general provides a somewhat better measurement
described by TAPPI (Technical Association of for surfaces that may be coated non-uniformly.
the Paper Industry) Standard T 452 [37]. The Figure 24.8 plots data from Ref. [38] for the
method evaluates brightness at an effective characterization of smoke deposition from the
wavelength of 457 nm and is illustrated by combustion of polystyrene, polymethyl-
Fig. 24.7. Brightness 457 measures brightness methacrylate and paper (i.e., liner board used in
with directional light incident at 45 with respect the manufacture of corrugated boxes) onto
to the normal to the sample as shown by the 47 mm dia filter targets. In the figure, the bright-
illustration on the left in the figure. The detector ness change is plotted as a percent difference
www.ebook777.com
Fig. 24.8 Change 100

in brightness versus Polystyrene
deposition
Paper
80 Polymethylmethacrylate
%Brightness Change
60
40
20
0
0.001 0.01 0.1 1 10
Deposition (g/m2)
Fig. 24.9 Classification

SVOC
of volatile organic
compounds by boiling VOC
point
VVOC
0 100 200 300 400

Boiling Point (°C)
from an uncontaminated reference of the filter compounds is responsible for stain damage,
substrate (polytetrafluoroethylene) versus the non-particulates in deposited smoke are the
smoke deposition for each of the tested materials. likely source for any potential odor damage.
The black line through the data in the figure is Smoke deposited from fires is highly porous
given by [20] and can contain substantial quantities of
adsorbed organic compounds [39, 40]. These
% Brightness Change ¼ 116 þ 26:2 non-particulates typically consist of volatile
ln Deposition (VOC) and very volatile (VVOC) organic
ð24:16Þ compounds that have boiling points between
0 C and about 260 C. Figure 24.9 illustrates
where Deposition has units of g/m2. See Ref. [38] the approximate boiling point ranges for each of
for additional experimental details including a the three classes of volatile organic compounds.
calibration of the brightness measurement tech- Potential odors derived from smoke deposited
nique with carbon black. after a fire can result from either dislodged
smoke particles containing odor causing VOCs
and VVOCs or direct desorption of organics from
Smoke Odor surfaces contaminated by smoke. The quantifica-
tion of odor is highly complex and can require
While the particulate portion of smoke deposition the quantification of five basic properties:
consisting of solid soot particles, semi-volatile (1) intensity, (2) degree of offensiveness, (3) char-
organic compounds (SVOC) and solid inorganic acter, (4) frequency and (5) duration.
Table 24.6 Total adsorbed volatiles onto deposited Langmuir [43], exists between adsorbed gaseous
smoke [38] molecules and the free gaseous molecules:
Volatile mass fraction
adsorption!
(g volatiles/g deposited Ag þ S ←desorption AS ð24:17aÞ
Material smoke)
Paper 0.31 0.08 where A is a gas molecule and S is a smoke
Polystyrene 0.35 0.06 particulate adsorption site. At equilibrium, the
Polymethylmethacrylate 0.11 0.02 overall rate constant K is given by:
kad ½AS
K¼ ¼ ð24:17bÞ
The combination of these properties characterizes kd Ag ½ S
the perception of odor and ultimately any poten-
tial damage or risk contributed by smoke where kad is the adsorption rate constant, kd is the
particles [41]. desorption rate constant and the bracketed
It can be useful to take a global approach to quantities are the molecular concentrations.
quantify odor from smoke due to the complexity Using benzene as a representative VOC present
of characterizing the perception of odor. For in typical smoke from polymeric materials, the
example, all volatile and semi-volatile organic relative adsorption and desorption rate processes
compounds (VOCs and SVOCs) can be treated can be assessed. Figure 24.10, for example, plots
as potential odorants and consolidated into a the ln of the rate constant versus the inverse of
single value associated with the deposition of the absolute temperature (1/K) [44]. As shown
smoke from the combustion of a given material. in the figure, the desorption rate of benzene gas
A useful technique [38] is to pass a known vol- molecules has a much stronger dependency on
ume of smoke through quartz filter targets. Any temperature than the adsorption rate. For exam-
deposited smoke is then analyzed by standard ple, desorption rates reduce by a factor of about
Thermal Desorption-Gas Chromatography-Mass 100 for a 100 C drop in temperature, while
Spectroscopy (TD-GC-MS) methods to assess adsorption rates reduce by less than a factor of
the quantity and character of any adsorbed 10 for the same temperature drop. The diver-
volatiles. Table 24.6 summarizes the total gence in adsorption/desorption rates is illustrated
measured volatiles using this methodology further in Fig. 24.11, where the ln of equilibrium
normalized by the total mass of deposited constant is plotted versus 1/K. Therefore,
smoke for paper, polystyrene and polymethyl- devolatilization rates of organics from deposited
methacrylate for data presented in Ref. [38]. smoke particulates are expected to be low at
While little is known about the quantitative normal ambient temperatures.
adsorption of organics onto smoke particulates, Figure 24.12 gives the adsorption capacity of
the adsorption capacity of activated carbon has benzene onto activated carbon versus the concen-
been widely investigated and has been shown to tration of benzene relative to the saturation con-
be very close to smoke particulates as well as centration. As shown in the figure, the adsorption
characterized with similar adsorption capacity reaches a plateau for a mass fraction of
mechanisms. [42] Therefore, the following analy- about 0.38 (g of benzene adsorbed per g of
sis and discussion is directed toward activated activated carbon). This is similar to the mass
carbon as a viable surrogate for the understanding fractions of VOCs found to be adsorbed on
of adsorption of organics onto smoke particulates. smoke deposited from paper and polystyrene,
The adsorption process of volatile organics 0.31 and 0.35, respectively. This would suggest
onto activated carbon can be evaluated by assum- that the deposited smoke from these two
ing that a dynamic equilibrium, as proposed by materials is nearly saturated with volatile organic
compounds. An explanation for lower mass
www.ebook777.com
Fig. 24.10 Adsorption- 0

desorption rate constants
for benzene onto activated
carbon
-2
ln(rate constant)
-4
adsorption
-6
-8
desorption
-10
0.0025 0.0030 0.0035 0.0040
1/ T (K-1)
Fig. 24.11 Equilibrium 6

constants for benzene onto
ln[Equilibrium Constant (m3/mol)]
activated carbon
0
0.0025 0.0030 0.0035 0.0040
1/ T (K-1)
fraction of 0.11 found for polymethyl- and toluene [46]. Similar capacities as previously
methacrylate may reside in the somewhat unique discussed for benzene are noted and appear to be
combustion mechanism for polymethyl- independent of the moisture condition of the
methacrylate, which initially decomposes into a activated carbon.
monomer leading to the production of small
highly combustible molecules [45]. This process
could likely result in the production of fewer Damage Thresholds
intermediate stable VOC products. Finally, for
comparison purposes, Table 24.7 gives activated The specification of damage thresholds can be
carbon adsorption capacities for two other com- rather arbitrary depending on the type of
mon VOCs found in deposited smoke – acetone anticipated damage. For example, leakage
Fig. 24.12 Adsorption 0.5

capacity of benzene onto
activated carbon
0.4 Plateau = 0.38 g/g
Adsorption Capacity (g/g)

0.3
0.2
0.1
0.0
0 5 10 15 20 25
Relative Concentration, Co /Csat (10-3)
Table 24.7 Activated carbon adsorption capacities [46] currents between 108 and 106 A are classified
as low, while currents greater than 106 A are
Activated carbon classified as high. This classification system can
Adsorption
be used as one of the criteria for evaluating
VOC State capacity (g/g)
smoke damage of electrical circuits by measur-
Acetone Dry 0.33
Wet 0.34
ing the value of LC for targets exposed to
Toluene Dry 0.41 different materials as given previously in
Wet 0.41 Table 24.3. The specific application would be
used to denote the target LC levels, with smoke
exposure to highly sensitive electronic
Table 24.8 Leakage current classifications components most likely in the very low range
(e.g., semiconductor fabrication facilities and
Leakage current (A) Class
data centers) and machine components in the
LC < 108 Very low
low range (e.g., machine shops and printing
106 > LC > 108 Low
LC > 106 High
facilities).
As noted above, stain and odor damage
thresholds are driven by human sensory percep-
current damage thresholds, while defined tion unlike damage thresholds for leakage cur-
broadly, are relatively well-specified. Con- rent. The field of psychophysics [48], for
versely, stain and odor thresholds are primarily example, attempts to quantitatively study per-
driven by perception. Suggested specific ception through the functional relationships
thresholds are covered in the following discus- between the physical properties of stimuli and
sion for leakage current and stain. A methodol- the psychological responses to them. In particu-
ogy for setting of odor thresholds is also lar, one important quantifier is the “Difference
proposed. Threshold” (or “Just Noticeable Difference”
The classification or ranking of leakage cur- often abbreviated as JND). The JND is the min-
rent (LC) has been proposed in an ASTM draft, imum amount by which stimulus intensity must
as shown in Table 24.8 [47]. Leakage currents be changed in order to produce a noticeable
less than 108 A are classified as very low, variation in sensory experience. Weber’s Law
www.ebook777.com
Table 24.9 Weber fractions for various stimuli VOC concentrations have been found for other
Stimuli Weber fraction types of papers and plastic packaging materials
Brightness 0.079 [52–54]. The food and pharmaceutical industries
Loudness 0.048 are particularly concerned with odor and/or taste
Finger span 0.022 transfer from VOCs contained in product pack-
Heaviness 0.020 aging. For example, the sources of VOCs in
Line length 0.029 paper packaging can be from the original paper
Taste 0.083 manufacturing process, including the paper itself,
Electric shock 0.013 inks, binders, adhesives and coatings. Recycled
Odor 0.25 paper, especially from newspaper, can have
VOC contents of up to 4000 ppm (or 4.0 103 g
[48] (also known as the Weber–Fechner law) VOC per g paper) [55].
states that the JND depends on a percentage of As a further illustration, the baseline odor
change in a stimulus rather than on a fixed threshold for a typical paper boxed commodity
amount of change: stored in a warehouse can be determined by a
combination of three factors: (1) the packaging
kW ¼ ΔS=S ð24:18Þ VOC content, (2) the paper density (often
referred to as basis weight or grammage in
where ΔS represents the difference threshold g/m2) and (3) the odor difference threshold or
(JND), S represents the initial stimulus intensity Weber fraction (i.e., 0.25). Typical paper
and kW signifies that the proportion on the right densities for liner board used to construct the
side of the equation remains constant despite paper box range from a low of 125 g/m2 to a
variations in the S term. kW is typically referred high of over 440 g/m2. Using 100 ppm as a
to as the Weber fraction and is given in typical average VOC content with a paper den-
Table 24.9 for sensory perception response to a sity of, for instance, 200 g/m2, the baseline vola-
number of different stimuli [49, 50]. tile organic content of the target paper surface
The Weber fractions for brightness and odor, can be estimated as:
i.e., 0.079 and 0.25, respectively, are of particu-
lar interest for assessing smoke damage 100 x 106 g VOC 200 g paper

thresholds for stain and odor. Applying a bright- g paper m2 paper
ness threshold of 0.079 [corresponding to a %
0:020 g VOC
Brightness Change of 7.9 % in Equation 24.16] ¼
m2 paper
to the smoke deposition data in Fig. 24.8, results
in a smoke damage threshold of ~0.015 g/m2 for The odor threshold corresponding to the JND
smoke stain. would be 25 % higher or 0.025 g VOC/m2.
Odor thresholds for smoke damage are some-
what more difficult to assess than those for stain.
A useful approach is first to establish a reason- Example Application
able odor baseline without smoke deposition for to Semiconductor Fabrication
the target surface. If, for example, the target Facilities
surface is a typical packaging material, then
the inherent concentration of volatile organic Semiconductor fabrication facilities are high-
compounds (VOCs) of the packaging would be value properties that contain very expensive pro-
relevant. For instance, odor in recycled packag- cess equipment and related support equipment.
ing papers has been related to several VOCs such During semiconductor fabrication, process liquid
as phenols and aldehydes [51]. The typical aver- heating and other electrical sources present
age concentration is about 100 ppm (or 1.0 potential ignition hazards. Note that semiconduc-
104 g VOC per g paper). Similar or higher tor fabrication cleanrooms are always provided
with sprinkler protection, and the likely fire sce-

nario involves combustible plastics resulting in
complete equipment loss due to heat damage in
the vicinity of the fire origin, and partial or com-
plete loss at some distance from the fire origin
due to smoke damage. The majority of smoke
damage is due to surface contamination, leading
to leakage current and long-term corrosion on
electronic products and equipment. As previ-
ously described, this type of damage can be
linked to smoke deposition by the use of damage
functions. Therefore, the estimation of smoke
damage in semiconductor cleanrooms requires a
methodology that connects the fire source and the
damage function, and allows for mapping the
damage area for a given facility size.
Damage Estimation Model
The general model for estimating smoke damage

is described in detail in Ref. [56], and includes Fig. 24.13 General methodology for modeling smoke
the definition of the fire source based on damage
experiments, simulation of smoke transport and
deposition using numerical models, and calcula-
tion of smoke damage based on measurements. middle of the left half of the cleanroom. The
When fire scenarios involve fire growth and clean bays with wet bench and etch tools (two
sprinkler suppression, current theories and right bays) were duplicated to create a four-bay
models can not predict either the burning rate or long module. The simulated fire source consists
the smoke generation rate and the fire source of a growing polycarbonate fire within a
needs to be defined based on experimental data. cleanroom stocker which peaks at a heat release
Figure 24.13 illustrates the procedures and data rate of about 300 kW. (Stockers are self-
sources for this general model. It should be noted contained units used for the storage of
that an important assumption and limitation of in-process and finished semiconductor wafers,
the transport model is that CO2 is an adequate which are commonly stored in plastic boxes in
surrogate for smoke, with smoke particles open shelves within the stocker.) Figure 24.15
transporting as the gas phase. This assumption plots the maximum expected smoke deposition
may not hold for larger smoke particles. versus radial distance from the stocker fire. The
damage functions for polycarbonate (i.e., leak-
age current from Table 24.3 and corrosion from
Fire Scenario and Results Table 24.5) can be used to convert the smoke
deposition values in the figure into the
The selected fire scenario [56] uses a numerical corresponding expected leakage currents and
simulation following the geometry of a typical corrosion rates. These results are given in
bay-and-chase cleanroom configuration. Fig. 24.16. Finally, applying the proposed leak-
Figure 24.14 shows a cleanroom module used age current damage thresholds as given in
in the numerical simulation. This module stands Table 24.8, yields expected smoke damage
for a section of four pairs of clean bays in the potentials as illustrated in Fig. 24.17.
www.ebook777.com
Fig. 24.14 Cleanroom

module used in numerical
simulation [56]
Fig. 24.15 Smoke 0.04

deposition versus radial
Smoke Deposition (g/m2)
distance [56]
0.03
0.02
0.01
0.00
0 10 20 30
Radial Distance from Fire (m)
Fig. 24.16 Leakage 2.0 10

current and corrosion rate
Corrosion Rate (10-12mpy)
Leakage Current (10-8A)
versus radial distance [56] 8

1.5
Corrosion Rate 6
1.0
4
Leakage Current
0.5
2
0.0 0
0 10 20 30
Radial Distance from Fire (m)
Fig. 24.17 Estimated

smoke damage
potentials [56]
σg Geometric standard deviation

Nomenclature Rep Particle Reynolds number
ODλ Optical density at wavelength λ
c Average coefficient of smoke extinction S Stimulus
Cc Slip correction factor ΔS Change in stimulus
Co Reference concentration t Time
Cs Smoke mass concentration τ Time constant
da Aerodynamic equivalent mass diameter T Temperature
dag Aerodynamic equivalent geometric mean V Particle velocity
mass diameter VTH Thermophoretic deposition velocity
dg Geometric mean diameter VTS Terminal settling velocity
dp Particle diameter
D Diffusion coefficient
fv Smoke volume fraction References
g Gravitational acceleration
η Gas viscosity 1. J.S. Newman and Y. Xin, “Characterization of Room
I Transmitted light intensity Environments in Growing Enclosure Fires,” Fire
Io Initial light intensity Safety Journal, 39, pp. 239–253 (2004).
2. K. McGrattan, S. Hostikka, J. Floyd, H. Baum and
kad Adsorption rate constant R. Rehm, “Fire Dynamics Simulator (Version 5)
kB Boltzmann’s constant Technical Reference Guide,” National Institute of
kd Desorption rate constant Standards and Technology, NIST Special Publication
kW Weber fraction 1018–5, October 2007.
3. Y. Wang, P. Chatterjee, and J.L. de Ris, “Large Eddy
K Overall rate constant Simulation of Fire Plumes,” Proceedings of the Com-
KTH Thermophoretic velocity coefficient bustion Institute, 33, pp. 2473–2480 (2011).
l Pathlength 4. W.C. Hinds, Aerosol Technology – Properties, Behav-
LC Leakage current ior, and Measurement of Airborne Particles, 2nd edi-
tion, John Wiley & Sons, Inc., New York, 1999.
λ Wavelength of light 5. E. Cunningham, “On the Velocity of Steady Fall
Λ Mean free path of Spherical Particles through Fluid Medium,”
ni Number of particles with diameter di Proceedings of the Royal Society A, 83, pp. 357–365
N Total number of particles (1910).
6. S.K. Friedlander and H.F. Johnstone, “Deposition of
ρg Gas density Suspended Particles from Turbulent Gas Streams,”
ρo Standard particle density Industrial and Engineering Chemistry, 49, 7, pp.
ρs Smoke particle density 1151–1156 (1957).
www.ebook777.com
7. A. Einstein, “On the Kinetic Molecular Theory of Combined Mobility and Aerodynamic Diameter
Thermal Movements of Particles Suspended in a Qui- Measurements. Part 2: Application to Combustion-
escent Fluid,” Annalen der Physik, 17, pp. 549–560 Generated Soot Aerosols as a Function of Fuel Equiv-
(1905). alence Ratio,” Aerosol Science and Technology,
8. W.C. Hinds, A. Ashley, N.J. Kennedy, N.J. and 38, pp. 1206–1222 (2004).
P. Bucknam, “Conditions for Cloud Settling and 23. F-X. Ouf, J. Vendel, A. Coppalle, M. Weill and
Rayleigh-Taylor Instability,” Aerosol Science and J. Yon, “Characterization of Soot Particles in the
Technology, 36, 12, pp. 1128–1138 (2002). Plumes of Over-Ventilated Diffusion Flames,” Com-
9. J.S. Newman, P. Su. and G.G. Yee, “Smoke Deposi- bustion Science and Technology, 180, pp. 674–698
tion Velocity in Industrial Fire Environments,” Fire (2008).
Safety Science: Proceedings of the Tenth Interna- 24. I. Colbeck, B. Atkinson and Y. Johar, “The Morphol-
tional Symposium, International Association for Fire ogy and Optical Properties of Soot Produced by Dif-
Safety Science, London, UK, pp. 655–668 (2011). ferent Fuels,” Journal of Aerosol Science, 28, 5, pp.
10. J.R. Brock, "On the Theory of Thermal Forces Acting 715–723 (1997).
on Aerosol Particles," Journal of Colloid Science, 25. G.W. Mulholland and M.Y. Choi, “Measurement of
17, pp.768-780 (1962). the Mass Specific Extinction Coefficient for Acety-
11. L. Waldmann and K.H. Schmitt, "Thermophoresis lene and Ethene Smoke Using the Large Agglomerate
and Diffusiophoresis of Aerosols," In Aerosol Science Optics Facility,” Proceedings of the 27th Interna-
(C. N. Davies, ed.), Academic Press, London, tional Symposium on Combustion, pp. 1515–1522
pp. 137–162 (1966). (1998).
12. L. Talbot, R.K. Cheng, R. Schefer, R. and D. Willis, 26. G.W. Mulholland and C. Croarkin, “Specific Extinc-
“Thermophoresis of Particles in a Heated Boundary tion Coefficient of Flame Generated Smoke,” Fire
Layer,” Journal of Fluid Mechanics, 101, 4, pp. and Materials, 24, pp. 227–230 (2000).
737–758 (1980). 27. S. Riahi, “New Tools for Smoke Residue and Deposi-
13. S. Riahi, C.L. Beyler and J. Hartman, “Wall Smoke tion Analysis,” PhD Dissertation, George Washington
Deposition from a Hot Smoke Layer,” Fire Technol- University, Washington D.C., January 2011.
ogy, (2012). 28. R.A. Martin and D.L. Fenton, “Full-Scale
14. S. Suzuki, K. Kuwana and R. Dobashi, “Effect of Measurements of Smoke Transport and Deposition
Particle Morphology on Thermophoretic Velocity of in Ventilation System Ductwork,” Los Alamos
Aggregated Soot Particles,” International Journal of National Laboratory, Los Alamos, NM, NUREG/
Heat and Mass Transfer, 52, pp. 4695–4700 (2009). CR-4321 (LA-10478-MS), 1985.
15. K.M. Butler and G.W. Mulholland, “Generation and 29. D.W. Weinert, T.G. Cleary, G.W. Mulholland and
Transport of Smoke Components,” Fire Technology, P.F. Beever, “Light Scattering Characteristics and
40, pp. 149–176 (2004). Size Distribution of Smoke and Nuisance Aerosols,”
16. H. Ono, R. Dobashi and T. Sakuraya, Fire Safety Science: Proceedings of the Seventh Inter-
“Thermophoretic Velocity Measurements of Soot national Symposium, International Association for
Particles under a Microgravity Condition,” Fire Safety Science, London, UK, pp. 209–220
Proceedings of the Combustion Institute, 29, pp. (2003).
2375–2382 (2002). 30. G.W. Mulholland, “Smoke Production and
17. J.S. Newman, “Prediction of Fire Detector Response,” Properties,” SFPE Handbook of Fire Protection Engi-
Fire Safety Journal, 12, pp. 205–211 (1987). neering, 4th edition, National Fire Protection Associ-
18. J.S. Newman, “Principles for Fire Detection,” Fire ation, Quincy, MA, pp. 2–291 to 2–302, 2008.
Technology, 24, 12, pp. 116–127 (1988). 31. T.J. Tanaka, “Measurements of the Effects of Smoke
19. J.S. Newman and J. Steciak, “Characterization of on Active Circuits,” Fire and Materials, 23, pp.
Particulates from Diffusion Flames,” Combustion 103–108 (1999).
and Flame, 67, pp. 55–64 (1987). 32. IEC/TS 60695-5-3 Ed. 1, “Fire Hazard Testing-Part
20. K.J. Rockne, G.L. Taghon and D.S. Kosson, “Pore 5.3: Corrosion Damage Effects of Fire Effluent-
Structure of Soot Deposits from Several Combustion Leakage Current and Metal Loss Test Method,”
Sources,” Chemosphere, 41, pp. 1125–1135 (2000). 89/545/DTS, 2002.
21. P.F. DeCarlo, J.G. Slowik, D.R. Worsnop, 33. R.P. Frankenthal, D.J. Siconolfi and J.D. Sinclair,
P. Davidovits and J.L. Jimenz, “Particle Morphology "Accelerated Life Testing of Electronic Devices by
and Density Characterization by Combined Mobility Atmospheric Particles: Why and How," Journal of the
and Aerodynamic Diameter Measurements. Part 1: Electrochemical Society, 140, 11, pp. 3129–3134
Theory,” Aerosol Science and Technology, 38, pp. (1993).
1185–1205 (2004). 34. J.S. Newman, P. Su, G.G. Yee and S. Chivukula,
22. J.G. Slowik, K. Stainken, P. Davidovits, “Development of Smoke Corrosion and Leakage Cur-
L.R. Williams, J.T. Jayne, C.E. Kold, D.R. Worsnop, rent Damage Functions,” Fire Safety Journal,
Y. Rudich, P.F. DeCarlo and J.L. Jimenz, “Particle 61, pp. 92–99 (2013) J.S. Newman, P. Su, G.G. Yee,
Morphology and Density Characterization by K.L.T. Jamison and S. Chivukula, “Strategic Smoke
Damage Program: Development of Smoke Damage VOC Adsorption,” Journal of Environmental Engi-
Functions for the Semiconductor Industry,” FM neering, 125, pp. 1160–1167 (1999).
Global Technical Report, Project ID 0003038685, 47. BSR/ASTM Z0334Z-200x Draft Standard, “Test
September 2010. Method for Measurement of the Leakage Currents
35. T.J. Tanaka and S.P. Nowlen, “Results and Insights on from Smoke Deposited on Electric Circuits,” ASTM
the Impact of Smoke on Digital Instrumentation and International, West Conshohocken, PA, 2003.
Control”, Sandia National Laboratories, Albuquerque, 48. G.A. Gescheider, Psychophysics: The Fundamentals,
NM, NUREG/CR-6597 (SAND99-1320), 2001. 3rd edition, Psychology Press, London, 1997.
36. R.D. Peacock, T.G. Cleary, P.A. Reneke and 49. R. Teghtsoonian, “On the Exponents in Steven’s Law
D.C. Murphy, “A Literature Review of the Effects of and the Constant in Ekman’s Law,” Psychological
Smoke from a Fire on Electrical Equipment,” Review, 78, pp. 71–80 (1971).
National Institute of Standards and Technology, 50. H. Stone, “Factors Influencing Behavioral Responses
Gaithersburg, MD, NUGEG/CR-7123, 2012. to Odor Discrimination – A Review,” Journal of Food
37. TAPPI Technical Association of the Paper Industry Science, 31, pp. 784–790 (1966).
Test Method T 452 om-08, “Brightness of Pulp, Paper 51. E. Martinez Martin and D. Ramirez Martin-
and Paperboard (Directional Reflectance at 457 nm),” Corbalan, “Chemical Substances Content in
Norcross, GA, 1998. Recycled Packaging Papers,” AIDIMA Furniture,
38. J.S. Newman, G.G. Yee and P. Su, “Development of Wood and Packaging Technology Institute, Paterna,
Smoke Damage Functions for Warehouse Spain, 2005.
Applications,” Fire and Materials, Proceedings of 52. P.A. Tice, and C.P. Offen, “Odors and Taints from
the 13th International Conference Exhibition, Paperboard Food Packaging,” Tappi Journal,
San Francisco, CA (2013); J.S. Newman, G.G. Yee 77, p. 149–154 (1994).
and P. Su, “Strategic Smoke Damage Program: Devel- 53. H. Kim-Kang, “Volatiles in Packaging Materials,”
opment of Smoke Damage Functions for Warehouse Critical Reviews in Food Science and Nutrition,
Applications,” FM Global Technical Report, Project 29, 4, pp. 255–271 (1990).
ID 0003043053, April 2012. 54. P. Landy, S. Nicklaus, E.Semon, P. Mielle and
39. C.C. Austin, D. Wang, D.J. Ecobichon and E. Guichard, “Representativeness of Extracts of Off-
G. Dussault, “Characterization of Volatile Organic set Paper Packaging and Analysis of the Main Odor-
Compounds in Smoke at Experimental Fires,” Journal Active Compounds,” Journal of Agricultural and
of Toxicology and Environmental Health, Part A,” Food Chemistry, 58, pp. 2326–2334 (2004).
63, pp. 191–206 (2001). 55. M. Biedermann, Y. Uematsu and K. Grob, “Mineral
40. B-J. De Vos, M. Froneman, and E.R. Rohwer, Oil Contents in Paper and Board Recycled to Paper-
“Organic Vapors Emitted from the Plumes of Pool board for Food Packaging,” Packaging Technology
Fires on Carpet Materials,” Journal of Fire Sciences, and Science, 24, 2, pp. 61–73 (2011).
17, pp. 383–420 (1999). 56. Y. Xin and J.S. Newman, “Numerical Simulation of
41. P. Dalton, “Odor Perception and Beliefs about Risk,” Smoke Damage in a Semiconductor Cleanroom,”
Chemical Senses, 21, pp. 447–458 (1996). Proceedings of the 9th International Conference on
42. V.I. Berezkin, I.V. Viktorovsii, L.V. Golubev, Performance-Based Codes and Fire Safety Design
V.N. Petrova and L.O. Khoroshko, “A Comparative Methods, the Excelsior, Hong Kong, June 20–22,
Study of the Sorption Capacity of Activated Charcoal, 2012.
Soot, and Fullerenes for Organochlorine
Compounds,” Technical Physics Letters, 28, pp.
885–888 (2002). [Translated from Pis’ma v Zhurnal Jeffrey S. Newman, retired, was an assistant vice presi-
Tekhnicheskoi Fiziki, 28, pp. 11–21 (2002).] dent and principal engineer for FM Global. He specialized
43. I. Langmuir, “The Constitution and Fundamental in characterization of fire environments including
Properties of Solids and Liquids. Part I. Solids,” Jour- modeling, flammability of materials, smoke damage,
nal of the American Chemical Society, 38, pp. full-scale fire testing, and smoke and fire detection.
2221–2295 (1916).
44. C.L. Chuang, P.C. Chiang, E.E. Chang and Dr. Geary G. Yee is a physical chemist and senior
C.P. Huang, “Adsorption-Desorption Rate of Nonpo- research specialist for FM Global. He has specialized in
lar Volatile Organic Compounds onto Activated Car- polymer pyrolysis, microbial corrosion, small lab-scale
bon Exemplified by C6H6 and CCl4,” Practice fire testing, smoke characterization, and material failure
Periodical of Hazardous, Toxic, and Radioactive analysis.
Waste Management, 7, pp. 148–155 (2003).
45. W.R. Zeng, S.F. Li and W.K. Chow, “Preliminary Dr. Paul Su is a senior research specialist and technical
Studies on Burning Behavior of Polymethyl- team leader for FM Global. He has specialized in materials
methacrylate (PMMA),” Journal of Fire Sciences, and corrosion research including corrosion testing and
20, pp. 297–317 (2002). control, failure analysis, smoke corrosivity, nanomaterials
46. F. Delage, P. Pre and P. Le Cloirec, “Effects of Mois- research, and chemical product development.
ture on Warming of Activated Carbon Bed during
www.ebook777.com
Heat Transfer from Fires to Surfaces

25
Brian Y. Lattimer
Introduction produced by a burning wall or ceiling. Window

flames are flames extending outside of a com-
The heat transfer from fires to adjacent surfaces is partment containing a fire.
an important consideration in many fire analyses. The heat transfer from fires has been
Some example applications that may require characterized for a range of different surface
knowledge of the heat transfer from a flame include geometries. The geometries included in this
heating and failure of structural beams, heat trans- chapter are
fer through walls and ceilings, and the ignition and • Flat vertical wall
flame spread along combustible surfaces. • Flat unconfined and confined ceilings
Flames transfer heat to adjacent surfaces • Parallel flat vertical walls
primarily through convection and radiation. • Corner walls at 90
Techniques for efficiently modeling the heat • Corner walls at 90 with a ceiling
transfer from flames are still being developed; • Horizontal I-beams beneath a ceiling
however, experimental data and empirical The majority of the data presented in this
correlations have been generated to predict flame chapter is from water-cooled heat flux gauge
heat transfer for a number of common geometries. measurements. Using these data, correlations
This chapter will focus on the data and empirical were developed from tests where important
correlations that have been developed. parameters were varied (i.e., heat release rate,
Empirical correlations for predicting heat fire base dimension, etc.). The range of the data
transfer from flames are typically simple to use; and the correlating parameters need to be taken
however, their use is usually limited to a particu- into consideration before applying the
lar type of fire or the geometry of the surface correlations. For example, the heat flux along
being heated. The types of fires considered in this the length of the flame has historically been
chapter include correlated with flame length measured in that
• Exposure area fires (burning objects) particular study. Measured flame lengths can vary
• Wall and ceiling fires depending on the measurement technique,
• Window flames definition, and surrounding geometry. For the
Exposure area fires are burning objects studies considered in this chapter, the data were
located adjacent to or near the surface being nondimensionalized with either the average (50 %
heated. Wall and ceiling fires are those fires intermittent) flame length or the flame tip length.
Therefore, heat flux correlations should be applied
using either the flame length correlation developed
B.Y. Lattimer (*)
Virginia Tech, Mechanical Engineering, 635 Prices Fork in the study or one that has been demonstrated
Road, Goodwin Hall413C, Blacksburg, VA 24060, USA to predict the flame length in that study.

746 B.Y. Lattimer
the radiation from the fire in Equation 25.2, the

Heat Transfer Boundary Condition equation for the boundary condition in Equa-
tion 25.1 becomes
The heat flux boundary condition for a material
surface exposed to a fire includes the exposure 00 εs h 00 i
qs ¼ qh f g h T f T h f g þ εh f g σT 4h f g
heat flux from the fire and the reradiation losses εh f g
from the surface. The exposure heat flux from the
þ h T f T s εs σT 4s
fire is composed of a radiative heat flux plus a
ð25:3Þ
convective heat flux. The heat flux boundary
condition for a fire heating an adjacent surface is Assuming the heat transfer coefficient at the heat
dT 00 00 flux gauge is the same as the heat transfer coeffi-
k ¼ qs ¼ εs qrad þ h T f T s εs σT 4s cient at the material surface, Equation 25.3 can
dx
be reorganized resulting in
ð25:1Þ

00 εs 00 εs
assuming negligible heating from the surroun qs ¼ q þ 1 hT f
εh f g h f g εh f g
ding environment.

εs
h Ts T h f g εs σ T 4s T 4h f g
εh f g
Heat Flux Gauges
ð25:4Þ
The radiation and convective heat flux terms in As a result, the need to determine the radiation
Equation 25.1 are difficult to accurately calculate heat flux from the fire has been removed. How-
due to the dependence of these terms on geome- ever, this expression still requires the gas temper-
try and fire properties. As a result, water-cooled ature near the surface to be measured, Tf, and the
total heat flux gauges are commonly used to emissivity of the adjacent surface, εs, to be
measure the maximum total exposure heat flux known in order to calculate the net heat flux
(or cold surface heat flux) from fires in different into the surface. Assuming the emissivity of the
configurations. The maximum total exposure adjacent surface is equal to the heat flux gauge
heat flux measured using the gauge can be used
emissivity εs ¼ εh f g , Equation 25.4 reduces to
to quantify the heat flux into the material surface. the following:
The total heat flux onto a water-cooled heat
00 00
flux gauge is described by the following qs ¼ qh f g h T s T h f g
equation:
00 00 εh f g σ T 4s T 4h f g ð25:5Þ
qh f g ¼ εh f g qrad þ h T f T h f g εh f g σT 4h f g :
ð25:2Þ Through Equation 25.5, the net heat flux
into an adjacent surface can be determined
These gauges are cooled so that their surface using the heat flux from a water-cooled gauge,
temperatures remain near ambient (20–80 C), gauge temperature, gauge emissivity, and heat
and they are coated with a high emissivity paint transfer coefficient. The gauge temperature and
ðε 0:95Þ to maximize the absorbed radiation. emissivity are typically known; therefore, the
Cooling the gauge surface maximizes the con- only unknown is the convective heat transfer
vective heat transfer and minimizes the radiative coefficient.
losses; thus, the cooled heat flux gauges measure An estimate of the local heat transfer coeffi-
the maximum total exposure heat flux. cient, h, is needed to calculate the heat flux into
The heat flux measured using the gauge can be the material. The heat transfer coefficient is
used to determine the heat flux to an adjacent dependent on the local velocity, gas temperature,
surface being heated by a fire. By solving for and geometry. For natural convection on
www.ebook777.com
25 Heat Transfer from Fires to Surfaces 747
Fig. 25.1 Magnitude Surface temperature (°F)

of the radiative and 200 400 600 800 1000 1200 1400 1600 1800
convective terms in 60
Equation 25.5: radiation
(—); convection with
50
h ¼ 0.015 kW/(m2-K)
( ); and convection
Heat flux (kW/m2)

with h ¼ 0.050 kW/(m2-K) 40
( )
30
20
10
0
0 100 200 300 400 500 600 700 800 900 1000
Surface temperature (°C)
horizontal and vertical surfaces, the heat transfer Example 1 A water-cooled heat flux-gauge is
coefficient varies from approximately 0.010 to used to measure the total incident heat flux
0.020 kW/m2-K. These coefficients apply to from a fire against a wall painted black. The
fires flowing against walls or along ceilings. measured heat flux is 30 kW/m2 and the water
Higher heat transfer coefficients are expected in cooling the gauge is measured to be 350 K. Both
areas where fires impinge on surfaces. Based on the wall emissivity and the heat flux gauge have a
data from Kokkala [1, 2] and You and Faeth surface emissivity of 0.95, and the heat transfer
[3, 4], the local convective heat transfer coeffi- coefficient is 0.01 kW/m2-K. Determine the net
cient where a diffusion flame impinges on a heat flux into the wall when the wall surface
ceiling is on the order of 0.050 kW/(m2-K). temperature is 600 K, 700 K, and 800 K.
Figure 25.1 contains a plot of the radiative and
convective heat flux terms in Equation 25.5 that Solution Equation 25.5 can be used to determine
are subtracted from the measured heat flux. the heat flux into the wall when the wall is at
The convective term is plotted using convective different temperatures.
heat transfer coefficients of 0.015 kW/m2-K 00 00
and 0.050 kW/m2-K. The radiative term is larger qs ¼ qh f g h T s T h f g εh f g σ T 4s T 4h f g
than the convective term at temperatures higher 00
qs ¼ 30 0:01ðT s 350Þ ð0:95Þ 5:67 1011
than 300 C with a heat transfer coefficient 4
T s 3504
of 0.015 kW/m2-K and at temperatures greater
than 600 C with a heat transfer coefficient • Wall surface temperature of 600 K
of 0.050 kW/m2-K. Based on results in this
00
plot, a non-conservative boundary condition qs ¼ 30 0:01ð600 350Þ ð0:95Þ 5:67 1011

will result if the heat transfer coefficient is over 6004 3504
estimated. 00
qs ¼ 21:3 kW=m2
The following examples are provided to illus-
trate how the heat flux into the material varies as • Wall surface temperature of 700 K
the material surface temperature increases and 00
qs ¼ 30 0:01ð700 350Þ ð0:95Þ 5:67 1011
how different assumptions (i.e., surface emissiv-
ity, heat transfer coefficient) affect the heat flux 6004 3504
00
into the material surface. qs ¼ 14:4 kW=m2
748 B.Y. Lattimer
• Wall surface temperature of 800 K heat transfer coefficient is 0.01 kW/m2-K. With a
00 wall surface temperature of 700 K, determine the
qs ¼ 30 0:01ð800 350Þ ð0:95Þ 5:67 1011 net heat flux into the wall if the surface emissivity

8004 3504 is 0.94, 0.90, 0.70, and 0.50. In each case, what is
00
qs ¼ 4:2 kW=m2 the percent error associated with assuming the
wall surface emissivity is equal to the heat flux
Example 2 A water-cooled heat flux gauge is gauge surface emissivity?
used to measure the total incident heat flux from
a fire against a wall. The heat flux gauge measured Solution Due to the surface emissivity of the
a heat flux of 30 kW/m2 while the gas temperature wall being different from that of the heat flux
was measured to be 1173 K. The water cooling the gauge, the heat flux into the wall is determined
gauge was measured to be 350 K. The heat flux using Equation 25.4:
gauge has a surface emissivity of 0.95, and the

00 εs
00 εs εs
qs ¼ qh f g þ 1 þ hT f h T s T h f g εs σ T 4s T 4h f g
εh f g εh f g εh f g
00 εs εs εs
qs ¼ 30 þ 1 0:01ð1173Þ 0:01 700 350
0:95 0:95 0:95

εs 567 1011 7004 3504
The heat flux into the surface where the wall and Example 3 A water-cooled heat flux gauge is
the gauge have the same emissivity is taken from used to measure the total incident heat flux from
00
Example 1b and is qs ¼ 14:4 kW=m2 . a fire against a wall painted black. The measured
• Surface emissivity of 0.94 heat flux is 30 kW/m2 and the water cooling the
gauge is measured to be 350 K. Both the wall
00
qs ¼ 14:2 kW=m2 emissivity and the heat flux gauge have a surface
emissivity of 0.95. With a wall surface tempera-
The assumption of equal surface emissivity ture of 700 K, determine the net heat flux into the
results in a heat flux 0.7 % higher. wall with heat transfer coefficients of 0.01 kW/
• Surface emissivity of 0.90 m2-K, 0.015 kW/m2-K, and 0.02 kW/m2-K.
00
qs ¼ 13:9 kW=m2 Solution Equation 25.5 can be used to determine
the heat flux into the wall when the wall is at
The assumption of equal surface emissivity
different temperatures.
results in a heat flux 3.8 % higher.
• Surface emissivity of 0.70 00 00
qs ¼ qh f g h T s T h f g εh f g σ T 4s T 4h f g
00 00
qs ¼ 11:8 kW=m2 qs ¼ 30 hð700
350Þ
ð0:95Þ 5:67 1011 7004 3504
The assumption of equal surface emissivity
results in a heat flux 17.7 % higher. • Heat transfer coefficient of 0.01 kW/m-K
• Surface emissivity of 0.50 00
00
qs ¼ 14:4 kW=m2
qs ¼ 9:8 kW=m 2
• Heat transfer coefficient of 0.015 kW/m-K

The assumption of equal surface emissivity 00
results in a heat flux 31.8 % higher. qs ¼ 12:6 kW=m2
www.ebook777.com
• Heat transfer coefficient of 0.02 kW/m-K transfer coefficient and surface emissivity. The
00
adiabatic surface temperature has been success-
qs ¼ 10:9 kW=m2 fully used as an effective gas temperature to
quantify the thermal boundary condition for
thermo-structural analysis [6, 8]. Analysis has
Adiabatic Surface Temperature not been reported on whether the boundary con-
dition using the adiabatic surface temperature
The adiabatic surface temperature has been provides the same results as using cold surface
00
proposed as a means for quantifying the thermal heat flux measurement, qhfg , and Equation 25.5.
boundary condition in fire environments [5–8]. Plate thermometers have been used to
The adiabatic surface temperature is the surface measure the adiabatic surface temperature in fur-
temperature that would exist if the surface were nace environments [5–8]. Due to the time
perfectly insulated. From Equation 25.1, the adi- constant of the devices [5–8], the adiabatic
abatic flame temperature is defined as surface temperature measurement provided by
00 plate thermometers need to be carefully consid-
0 ¼ εs qrad þ h T f T ast εs σT 4ast ð25:6Þ ered in applications where the fire environment
is transient.
Combining with Equation 25.1, a relationship
between the heat flux at the surface and the
adiabatic surface temperature is
Objects Immersed in Flames
00
qs ¼ εs σ T 4ast T 4s þ hðT ast T s Þ ð25:7Þ
Some of the highest heat fluxes measured from
The expression provides a relationship between diffusion flames have been measured in tests
the adiabatic surface temperature and the heat with objects immersed in large, open hydrocar-
flux to the surface. Equation 25.7 has a form bon pool fires. In these tests, small and large
similar to Equation 25.1 when the radiation objects (relative to the fire size) were placed
00
term, qrad , is taken as the radiation from a black- within the pool fires. These tests were performed
body source, σTrad 4
. Based on this, adiabatic sur- to evaluate the heat transfer from fires to large
face temperature can be thought of as an objects such as fuel tanks, weapons, and nuclear
effective gas temperature that embodies the containers.
radiation and convection gas temperatures. The maximum heat fluxes measured in these
The adiabatic surface temperature can then be tests are summarized in Tables 25.1 and 25.2.
used as the boundary surface temperature for From data in these tables [9–19], the size of the
calculating the thermal response of materials object relative to the pool fire has a significant
exposed to fire conditions, knowing the heat impact on the incident heat flux to the object.
Table 25.1 Heat fluxes to objects immersed in large pool fires [9–19]
Test Pool size Fuel Peak heat flux (kW/m2)
AEA Winfrith [9] 0.5 9.45 m Kerosene 150
US DOT [9] Not listed Kerosene 138
USCG [9] Not listed Kerosene 110–142
US DOT [9] Not listed Kerosene 136–159
Sandia [9] Not listed Kerosene 113–150
HSE Buxton [9] Not listed Kerosene 130
Shell Research [9] 4.0 7.0 m Kerosene 94–112
Large cylinder [10] 9.1 18.3 m JP-4 100–150
Small cylinder [10] 9.1 18.3 m JP-4 150–200
Russell and Canfield [10] 2.4 4.8 m JP-5 175
750 B.Y. Lattimer
Large calorimeters were measured to be exposed

to heat fluxes of 100–150 kW/m2. McLain [17] Exposure Fires
and Taylor et al. [18] measured slightly lower heat
fluxes (75–85 kW/m2) in their tests with large Fires Adjacent to Flat Walls
items that were similar in size to the pool fire. In
studies with small calorimeters, peak heat fluxes Heat fluxes from exposure fires adjacent to flat
were measured to range from 150 to 200 kW/m2. walls have been experimentally studied using
The difference in the heat fluxes measured for propane sand burners and characterized for vari-
small and large items immersed in pool fires has ous burning objects. The experimental study
been attributed to the difference in the convective provides a systematic approach of calculating
heat transfer coefficient, the flame thickness, and heat fluxes for this geometry.
the impact of the object on the flame temperature. An extensive experimental study was
Small-scale calorimeter data provide a bound for performed by Back et al. [20] to characterize the
heat fluxes to an item immersed in a pool fire. heat transfer from a fire to a directly adjacent wall.
Based on the available data, a bounding heat In this study, fires were generated using square
flux of 175 to 200 kW/m2 is possible. propane sand burners with edge lengths of 0.28,
0.37, 0.48, 0.57, and 0.70 m. Heat flux fields were
measured for fires ranging from 50 to 520 kW.
Table 25.2 Heat fluxes to different size objects A plot of the peak heat fluxes measured for
immersed in fires [10–19]
each type of fire evaluated is shown in Fig. 25.2.
Object Peak heat flux (kW/m2) Peak heat fluxes for the different fires evaluated
Large calorimeter [10] 100–150 were determined to be a function of fire heat
Large calorimeter [12] 85
release rate. This dependence was attributed
to the larger size fires resulting in thicker bound-
ary layers, which is related to the radiation
pathlength. Based on gray-gas radiation theory,
Large calorimeter [17] 85a
the authors found the following relation ade-
Large calorimeter [18] 75a
quately represented the data:
Large wall (3.0 0.6 m) [19] 80–120 00 00
qcl ¼ qpeak z=L f 0:4 ð25:8Þ
Small calorimeter [11] 175
Small calorimeter [10] 150–200
These peak heat fluxes were measured in the
a
Object size comparable to pool fire size lower part of the fire (z/Lf 0.4) along the
Fig. 25.2 Peak heat 120

release rates measured in
Maximum wall heat flux (kW/m2)
square propane burner fires

100
against a flat wall [20]
80
60
40
Aspect ratio ~ 3
Aspect ratio ~ 2
20
Aspect ratio ~ 1
0
0 100 200 300 400 500 600
Heat release (kW)
www.ebook777.com
Fig. 25.3 Vertical heat 1000

flux distribution along the
centerline of a square
propane burner fire
adjacent to a flat wall [20]
100
q øcl (kW/m2)
Q ≈ 59 kW
10 Q ≈ 121 kW
Q ≈ 212 kW
Q ≈ 313 kW
Q ≈ 523 kW
Correlation for Q = 59 kW
Correlation for Q = 523 kW
1
0.01 0.1 1 10
z /Lf

centerline, with the flame length taken from 00 00 x 2 x
q ¼ qcl exp 1:0 ð25:11aÞ
Heskestad [21]: 0:5D 0:5D
x 1=7 x
> 1:0 ð25:11bÞ
00 00
L f ¼ 0:23Q2=5 1:02D ð25:9Þ q ¼ 0:38qcl
0:5D 0:5D
Above this region, the heat fluxes were measured Heat fluxes from burning objects to an adja-
to decrease with distance above the fire. The heat cent wall have been measured for a variety of
flux data measured along the centerline is shown items; however, limited data have been published
in Fig. 25.3. Lines in this plot are a general on this work [22, 23]. Heat fluxes at the rim of
correlation of the centerline data: wastebasket fires were reported by Gross and
00 00 Fang [22]. At the rim, heat fluxes as high as
qcl ¼ qpeak z=L f 0:4 ð25:10aÞ
50 kW/m2 were measured; however, the authors
noted that peak heat fluxes for these fires
00 5
00 00
qcl ¼ qpeak z=L f 2=5 qpeak 20 occurred approximately 0.22 m above the rim.
3 Mizuno and Kawagoe [23] performed
0:4 < z=L f 1:0 experiments with upholstered chair fires against
ð25:10bÞ a flat wall. In these tests, Mizuno and Kawagoe
measured heat fluxes to the wall of 40–100 kW/m2
00 5=3
qcl ¼ 20 z=L f z=L f > 1:0 ð25:10cÞ over the continuous flaming region (~z/Lf < 0.4).
All of these tests were performed using foam-
Heat fluxes were measured to decrease with padded chairs.
horizontal distance from the centerline, as shown
in Fig. 25.4. The normalized lateral heat flux
distribution data shown in Fig. 25.4 was found Fires in a Corner
to be half-Gaussian in shape over the half width
of the burner. The line in the plots is a fit to the Fires in a corner of a room lined with a combus-
data in Fig. 25.4a: tible material have been shown to cause more
752 B.Y. Lattimer
a b
1.4 1.4
6
1.2 1.2 6
6
5
5
3
3
1.0 1.0 3
3
5
5
3
6 7
q / qcl
6
0.8 0.8
q / qcl
2 8 5
2 6
6
2
7
2
9 7 8 6
9
9
9 5
6 2 8 5
0.6 9 83
9 5 0.6 7 7 5
5
83 2 5 2
3
2 5
7 85 2 8 3
9 5 8
1 5
7
5
0.4 7 55
9 0.4 2 7
2
8
1
1 3
9
9 8 37 5
9 8 9
9 6 7 3
8
8 5 2 8
4
7 8 6
6
5 4
7 7 8 65
5
4
7 7
2 7 6
0.2 7 5
5
7 8
8
8
5
6
0.2 2 7
2
6
6
6
7
6 89 6
6 2
2
9 8
5
7 9 98 2 9
9 6
2
7 9
9 5
6
6 98
8 6
64 8 2 3
7
2 21
8 6 8
6 8
5 9
91 64 8 4 2 4
2 2 64
64 1
1 4
4 2 1 4
0.0 7 7 8 8 0.0 7 7 527 3 8
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
x / 0.5D x / 0.5D
Fig. 25.4 Lateral heat flux distribution with distance from the centerline of square propane burner fires against a flat
wall [20] (a) in the flaming region and (b) in the plume
rapid flame spread and growth to flashover com- heating the top part of the wall. This resulted in
pared to cases with fires in other locations within higher heat fluxes farther out from the corner
the room. For these reasons, a significant amount along the top part of the wall.
of work has been performed to characterize the A series of fire tests were performed by
heat fluxes produced by corner fires. Heat flux Lattimer and Sorathia [29] to develop empirical
measurements have been performed both in an correlations to estimate heat fluxes from an expo-
open environment to quantify the heat flux due to sure fire to the walls and ceiling of a corner. Tests
the exposure fire alone and within rooms to mea- were performed using 0.17-, 0.30-, and 0.50-m
sure the heat flux due to the exposure fire and the square propane burners placed directly against
room environment. the corner. Heat flux fields were measured for
The heat flux from the exposure fire has been fires ranging from 25 to 300 kW.
quantified in several studies performed in an Correlations were developed for three regions
open laboratory environment [24–29]. All the in the corner: along the height of the walls in the
studies were performed in a noncombustible cor- corner, along the top of the walls near the ceiling,
ner with a ceiling except the study of Kokkala and along the ceiling. The region containing the
[26], which was performed in a noncombustible walls in the corner extended from the top of the
corner without a ceiling. A comparison of the fire to approximately 1.8 m above the floor, which
heat flux fields measured in the study with a is approximately the ceiling height minus twice
ceiling [29] and the study without a ceiling [26] the ceiling jet thickness (δ ¼ 0.1H ). Correlations
is shown in Fig. 25.5. Note that the contour plot for the top part of the walls, which are heated by
of Lattimer and Sorathia is relative to the floor, the ceiling jet, were developed using data at
while the plot of Kokkala is relative to the top of locations greater than 1.8 m above the floor.
the burner. Lattimer et al. used a burner 0.15 m Along the height of the walls in the corner, the
high. peak heat fluxes were typically measured near
Up to approximately 1.8 m above the floor, the base of the fire. The peak heat fluxes along
the heat flux distributions are similar. In the case the height of the walls in the corner were
with the ceiling, the ceiling jet and the radiation measured to be a function of the fire diameter,
from the fire flowing along the ceiling were as shown in Fig. 25.6. The curve in Fig. 25.6 is a
www.ebook777.com
a b
1.8 4
1.6 10
•
Q = 300 kW
1.4 10 D = 0.17 m
Distance from corner, y (m)
10
1.2
20 10
1.0
30
40
20
0.8
50 20
60 30
3
0.6
40
70
0.4 50 30 20
80 70 60
40
0.2
Burner
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
Distance above burner, z (m)

Distance from corner, x (m) 50 40 30 20 10 kW/m2
Ceiling
60
2
2.2
80 70
70 50
40 30 20
60 10
2.0
1.8 80
20
70 10
50 40 30
Distance above floor, z (m)
1.6
60
5
1.4 60
50
10
20
1.2
50
40 30 5
60 1
1.0
60
0.8 10
20
5
50 30
40
0.6 60
50
0.4
40 10
20
5
30
0.2
Burner
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 0
Distance from corner, x (m) 0 0.2 0.4 0.6
Wall Distance from corner, x (m)
Fig. 25.5 A comparison of the heat flux fields produced Lattimer and Sorathia [29] are plotted relative to the floor,
in a corner (a) with a ceiling [29] and (b) without a ceiling and the data of Kokkala [26] are plotted relative to the top
[26]. The fire was produced by a 0.17-m-square propane of the burner
burner with a heat release rate of 300 kW. Note data of
correlation to the data and is expressed using the L f , tip =D ¼ 5:9QD

*1=2
ð25:13Þ
following relation:
00
where
qpeak ¼ 120½1 expð4:0 DÞ ð25:12Þ
Q
Q*D ¼ pffiffiffi ð25:14Þ
The vertical distribution in the maximum heat ρ1 C p T 1 gD5=2
flux along the walls near the corner is shown in
Fig. 25.7 plotted with the vertical distance Peak heat flux levels were measured in the lower
normalized with respect to the flame tip, part of the flame (z/Lf,tip 0.4) and decreased
754 B.Y. Lattimer
Fig. 25.6 Peak heat flux 120

along the height of the
walls in the corner [29] 110
100
90
Peak heat flux, q ″peak (kW/m2)

80
70
60
50
40
30
20
10
0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Length of area burner side, D (m)
Fig. 25.7 Maximum heat 1000

fluxes to the walls near the
corner with square burner
Maximum heat flux in corner, q ″max (kW/m2)
sides of 0.17 m (○), 0.30 m

(Δ), 0.30 m (elevated) (∇),
and 0.50 m (□), and fire
sizes ranging from 50 to 100
300 kW [29]
10
1
0.01 0.1 1 10
z / Lf,tip
with distance above z/Lf,tip ¼ 0.4. A general cor- 10=3

00 z
relation to represent this behavior is qmax ¼ 7:2 z=L f , tip 0:65
L f , tip
00 00 ð25:15cÞ
qmax ¼ qpeak z=L f , tip 0:4 ð25:15aÞ
This is similar to the form used by Back

00 00 z 2 00 et al. [20] to correlate heat fluxes from an expo-
qmax ¼ qpeak 4 qpeak 30 sure fire to a wall (see Equations 25.10a, 25.10b,
L f , tip 5
and 25.10c), except the constants are different.
0:4 < z=L f , tip 0:65 The horizontal distribution in the heat flux
ð25:15bÞ along the wall is shown in Fig. 25.8 to best
www.ebook777.com
Fig. 25.8 Lateral 1.1

distribution in the heat flux
along the walls with 1.0
distance from the corner
0.9
with square burner sides of
0.17 m (○), 0.30 m (Δ), 0.8
0.30 m (elevated) (∇), and
0.50 m (□), and fire sizes 0.7
ranging from 50 to 300 kW
q ″/q ″max
[29] 0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Distance from corner, x (m)
Fig. 25.9 Maximum heat 1000

flux along the top of the
walls during a corner fire
test with square burner
sides of 0.17 m (○), 0.30 m
Maximum heat flux, q ″max (kW/m2)
(Δ), 0.30 m (elevated) (∇),

and 0.50 m (□), and fire 100
sizes ranging from 50 to
300 kW [29]
10
1
0.1 1 10
(x + H ) /Lf,tip
00 00
correlate with actual distance from the corner [29]. q ¼ 0:058qmax x1:8 x > 0:4 m ð25:16bÞ
This was attributed air being entrained in the cor-
ner, pushing the fire into the corner. Near the It has not been established whether this correla-
corner the shape is half-Gaussian; however, heat tion holds for fire sources larger than 0.50 m in
fluxes outside of this decrease slower. The trend in length on a single side.
the data, which is shown as the line in Fig. 25.8, Along the top part of the wall the maximum
can be represented using the following relations: heat fluxes were measured at locations less than
00 00
0.15 m below the ceiling. The maximum heat
q ¼ qmax exp½7:5x2 x 0:4 m ð25:16aÞ fluxes are shown in Fig. 25.9 plotted against the
756 B.Y. Lattimer

normalized distance along the flame (x + H)/Lf, 00 rþH
q ¼ 120 0:52 ð25:18aÞ
tip, where x is the distance from the corner. These L f , tip
heat fluxes can be estimated using the following

relations: 00 x þ H 3:5 r þ H
q ¼ 13:0 > 0:52
L f , tip L f , tip
00 xþH
qmax ¼ 120 0:52 ð25:17aÞ ð25:18bÞ
L f , tip
Again, the assumed plateau in the correlation

00 x þ H 3:5 x þ H was based on the maximum heat flux expected
qmax ¼ 13:0 > 0:52
L f , tip L f , tip from a flame, according to Equation 25.12.
ð25:17bÞ Similar levels were measured by Hasemi
et al. [25] with an exposure fire in the corner,
The assumed plateau in the correlation was based simulated burning corner walls, and an exposure
on the maximum heat flux expected from a flame, fire and simulated burning corner walls in the
according to Equation 25.12. corner.
The heat fluxes to the ceiling were determined
to be a function of normalized distance along the Room Environment Effects Corner fires are
flame length, (r + H)/Lf,tip. All of the ceiling heat currently used to evaluate fire growth potential
flux data taken in the study with a square burner of a combustible lining material. As such, several
in the corner are shown in Fig. 25.10. Heat fluxes studies have been conducted to characterize the
along the ceiling due to the exposure fire were heat flux from an exposure fire inside a room
similar to those measured along the top of the [30–33]. In these cases, the heat flux to the sur-
wall. This resulted in similar correlations to esti- face will be due to both the exposure fire and the
mate the heat flux to the ceiling: room environment.
Fig. 25.10 The heat flux 1000

along the ceiling above a
fire located in a corner for
tests with square burner
sides of 0.17 m (○), 0.30 m
(Δ), 0.30 m (elevated) (∇),
100
Heat flux to ceiling, q ″ (kW/m2 )
and 0.50 m (□), and fire

300 kW [29]
10
0.1
0.1 1 10
(r + H ) /Lf,tip
www.ebook777.com
The effect of the room environment on the boundaries. A room or fire different from that
heat fluxes was clearly demonstrated through used to produce the data in Fig. 25.11b may
the work performed by Dillon [33] in an ISO result in a different gas layer temperature,
9705 room [34]. The incident heat fluxes from which will result in a different heat flux contri-
the fire were determined by measuring the tem- bution due to the room environment.
perature rise at several locations on an insulated Heat fluxes due to the hot layer environment
steel plate. Heat fluxes were calculated using a inside a room were measured by Tanaka
two-dimensional heat balance on the plate. Heat et al. [35]. In tests conducted in a 3.3-m-wide,
fluxes included contributions from both the expo- 3.3-m-deep, 2.35-m-high room with the propane
sure fire and the room environment. Using the fire in the center of the room, heat fluxes were
surface temperature measurements and initial measured at different locations on one of the side
heat flux measurements after the burner was walls. The average heat flux measured in the
ignited, the heat fluxes to the hot steel plate upper layer formed inside of the room is shown
were corrected for both reradiation from surfaces in Fig. 25.12 versus the layer temperature for
in the room and heating by the hot gas layer. different compartment door widths. The line
The effects of the room environment on the in the plot represents the blackbody heat flux
00
heat fluxes to the corner boundaries is discussed using the layer gas temperature, q ¼ σT 4g . As
here for the case with a 300 kW fire in the corner, seen in the Fig. 25.12, the blackbody heat flux
produced using a 0.17-m square burner. The heat using the layer gas temperature provides a rea-
fluxes shown in Fig. 25.11a represent the heat sonable estimate of the incident heat flux to the
flux from the fire only, as measured using a heat walls inside a compartment; however, the
flux gauge (i.e., cold surface). Note that the top measured heat fluxes are generally higher than
of the burner is 30 cm above the floor. In general, the blackbody heat flux. A more detailed investi-
the heat fluxes in Fig. 25.11a compare well gation of the heat flux to compartment fire
with the total heat flux data shown in Fig. 25.5a. surfaces was performed by Toflio et al. [36]
Heat fluxes shown in Fig. 25.11b correspond Through this study, it was determined that the
to heat fluxes due to the fire and the room envi- higher heat flux was attributed to convection
ronment (i.e., hot gas layer and reradiation from between the hot gas layer and the wall. In addi-
walls), as measured using a heat flux gauge. For tion, as fires became large in size, radiation
this room environment, the heat fluxes including exchange between the fire and the walls could
the room environment were higher than the heat also increase the heat flux to wall surfaces.
fluxes from the exposure fire to a cold wall. The
magnitude of the increase depends on the eleva- Effects of Fire Standoff Distance Several
tion inside the room. Measurements in the lower researchers have investigated the effects of
part of the room showed less of an increase moving the exposure fire away from the corner
compared with those near the ceiling. Heat fluxes (i.e., standoff distance) [24, 30, 37]. As one
in the upper part of the room increased by as might expect, moving the fire away from the
much as 20 kW/m2, an increase largely attributed corner decreases the heat fluxes to the room
to the hot gas layer that forms inside the boundaries. Tests were performed by Williamson
room during the fire. For the 300-kW fire inside et al. [30] in a full-scale ISO 9705 room using a
the ISO 9705 compartment, average gas 0.30-m-diameter burner. Heat fluxes to the wall
temperatures in the upper part of the room were were strongly dependent on whether the flame
measured to be approximately 680 K. Note that was attached to the corner walls or burned freely
the heat flux due to the room environment is near the wall. At a heat release rate of 40 kW,
dependent on the gas layer temperature, which with the burner against the corner walls, the
is dependent on the fire size, room geometry, flame was attached to the walls and heat fluxes
ventilation, and thermal properties of the were measured to be as high as 50 kW/m2. When
758 B.Y. Lattimer
a b
232.5 232.5
217.5 217.5
80 100
70
202.5 202.5
60
50 90
187.5 187.5
40
30 80
172.5 172.5
20
70
157.5 157.5
60
50
Distance above floor, z (cm)

Distance above floor, z (cm)
40
142.5 142.5
30
10 kW/m2 20
127.5 127.5
10 kW/m2
112.5 112.5
97.5 97.5
82.5 82.5
67.5 67.5
52.5 52.5
37.5 37.5
22.5 22.5
Burner Burner
7.5 7.5
112.5 97.5 82.5 67.5 52.5 37.5 22.5 7.5 112.5 97.5 82.5 67.5 52.5 37.5 22.5 7.5
Distance from corner, x (cm) Distance from corner, x (cm)
112.5 112.5
97.5 97.5
Distance from corner, y (cm)

Distance from corner, y (cm)
30 kW/m2
82.5 82.5
40 40 kW/m2
67.5 67.5
50
50 60 70
52.5 80 52.5
60 90
70 110
100
37.5 37.5
80
90
22.5 22.5
Burner
Burner 90 100
110 90
7.5 7.5
112.5 97.5 82.5 67.5 52.5 37.5 22.5 7.5 112.5 97.5 82.5 67.5 52.5 37.5 22.5 7.5
Distance from corner, x (cm) Distance from corner, x (cm)
Fig. 25.11 Heat fluxes to corner boundaries from (a) 300-kW, 0.17-m square propane sand burner exposure fire alone
to a “cold” surface and from (b) the 300-kW exposure fire and the room environment [33]
the fire was moved 50 mm from the walls, the be attached to the walls and heat fluxes of
flames were observed to be detached from the 40–60 kW/m2 were measured at the walls. Addi-
walls with the highest heat fluxes measured to be tional work needs to be performed to investigate
approximately 25 kW/m2. In tests with a heat distances at which fires attach to nearby surfaces,
release rate of 150 kW, the fire was observed to such as a flat wall or walls in a corner.
www.ebook777.com
Fig. 25.12 Heat flux 200

to the walls inside a
compartment containing
a hot gas layer [35]
100
80
50
Heat flux to wall (kW/m2)

30
20
10
8
5 Width (cm)
29
45
3
59
89
2
1
400 500 600 800 1000 1400
Hot layer average temperature (K)
between the ceiling and fire, H, plus the virtual

Fires Beneath Unconfined Ceilings source location, z0 . The virtual source location
for this geometry was determined using the fol-
There have been several experimental and theo- lowing relations:
retical studies performed on fires impinging on
0

*2=5 *2=3
an unbounded ceiling [1–4, 38–41]. Total heat z ¼ 2:4D QD QD Q* < 1:0
fluxes from fires and fire plumes impinging on
ð25:19aÞ
the ceiling were measured by Hasemi et al. [38],
You and Faeth [3, 4], and Kokkala [1, 2].
0

*2=5
Hasemi et al. [38] conducted a series of fire z ¼ 2:4D 1 QD Q* 1:0 ð25:19bÞ
tests using propane gas burners located at differ-
ent distances beneath a noncombustible ceiling. where QD* is defined as in Equation 25.14 with
Fires as large as 400 kW (approximated) were D being the diameter of the exposure fire. The
considered in the study. Heat flux gauges were length of the flame, Lf,tip , in this geometry is
used to measure the incident heat flux along the defined as the distance between the fire and the
ceiling at different distances away from the fire ceiling, H, plus the radial extension of the flame
centerline, or stagnation point. The measured out from the center of the fire, LH. The location of
heat flux at the stagnation point is shown in the flame tip in this geometry was found to cor-
Fig. 25.13 to plateau at approximately 90 kW/m2. relate with QH*, which is defined the same as in
In order to collapse the data, the flame tip length Equation 25.14 except D is replaced by H. The
was normalized with respect to the distance flame tip correlation was determined to be
760 B.Y. Lattimer
Fig. 25.13 Stagnation 100

point heat fluxes on an
unbounded ceiling with a
fire impinging on it [38]
80
Heat flux, q s (kW/m2)

60
D = 1.0 m H = 1.0 m
H = 1.2 m
H = 0.8 m
40
H = 0.6 m
H = 0.4 m
D = 1.0 m H = 0.64 m
H = 0.8 m
20
H = 1.0 m
D = 0.3 m H = 1.0 m
H = 0.8 m
0
0 1 2 3 4 5 6 7 8 9 10 11
Lf /(H + z )
Fig. 25.14 Heat fluxes to 100

a ceiling due to a propane D = 0.5 m H = 1.0 m
fire impinging on the H = 1.2 m
surface [38] H = 0.8 m
H = 0.6 m
H = 0.4 m
D = 1.0 m H = 0.64 m
H = 0.8 m
Heat flux, q (kW/m2)
H = 1.0 m
D = 0.3 m H = 1.0 m
H = 0.8 m
10
1
0.1 1 10
(r + H + z) / (LH + H + z)
ðLH þ H Þ *1=3 Figure 25.14 contains a plot of the heat flux

¼ L f , tip =H ¼ 2:89QH ð25:20Þ to the ceiling as a function of location within
H
the flame. The correlation recommended by
The heat flux was measured to decrease with Wakamatsu [42] can be used to predict the heat
distance from the fire stagnation point. fluxes:
www.ebook777.com
Fig. 25.15 A comparison 1000

of the best fit curve
proposed by Wakamatsu
[42] ( ) and a bounding
fit to the data (—). The

unbounded ceiling data of 100
Hasemi et al. [38] is
represented as the
outlined area
10
1
0.1 1 10
w = (r + H + z )/(LH + H + z )
00
q ¼ 518:8e3:7w ð25:21aÞ Kokkala [1, 2] and You and Faeth [4]. Kokkala
used natural gas as a fuel and measured heat
where fluxes to plateau at 60 kW/m2 at the stagnation
0
0
point (Fig. 25.16). Due to the small burners used
w ¼ r þ H þ z = LH þ H þ z ð25:21bÞ in this study (D ¼ 0.064 m), Kokkala’s data
collapsed without applying a virtual source ori-
Figure 25.15 contains a plot of Equa- gin correction. Heat fluxes measured in the natu-
tion 25.21a (dashed line) along with a represen- ral gas fire tests were lower than those measured
tation of the data of Hasemi et al. [38] for a flat by Hasemi et al. [38], an effect that can partly be
unbounded ceiling. Equation 25.21a adequately attributed to the higher radiation levels from the
estimates the data when w is greater than 0.45 propane flames. The heat fluxes measured by
but significantly overestimates heat flux levels Kokkala [1] at different locations from the stag-
for smaller values of w. Based on the data from nation point are plotted in Fig. 25.17. The distri-
Hasemi et al. [38] and other data from fires bution in the heat flux along the ceiling radially
impinging on I-beams mounted to a ceiling out from the stagnation point was measured to be
[43], a correlation was developed to predict the similar to that measured by Hasemi et al. [38]
bounding heat flux levels to an unconfined with larger propane gas fires.
ceiling:
00
q ¼ 120 w 0:5 ð25:22aÞ
Fires in Corridors
00
q ¼ 682expð3:4wÞ w > 0:5 ð25:22bÞ
There are few reported studies that have
where w is defined in Equation 25.21b. This measured heat fluxes to the ceiling or walls of a
correlation is shown in Fig. 25.15 as the solid corridor. Hinkley et al. [44–46] performed
line. The peak heat flux of 120 kW/m2 at w less experiments with a town gas burner located at
than or equal to 0.5 bounds nearly all of the heat the end of a hallway. Tests with a noncombusti-
flux measurements made in this range for the ble ceiling were performed in a 1.2-m-wide hall-
studies of Hasemi et al. [38] and Myllymaki way using a range of heat release rates
and Kokkala [43]. (170–600 kW) with the burner at various
Heat flux measurements with smaller fires distances (0.37–1.20 m) below the ceiling. Heat
(<11 kW) beneath a ceiling were made by fluxes were determined through temperature
762 B.Y. Lattimer
Fig. 25.16 Heat fluxes

at the stagnation point on
a ceiling for tests with
natural gas fires impinging 60
on the ceiling [1]
Heat flux, qs (kW/m2)

40
20
2.9 kW
4.0 kW
6.3 kW
8.6 kW
10.5 kW
0
0.5 1 1.5 2 2.5
Lf /H
Fig. 25.17 Heat fluxes 100

to a ceiling due to a natural
gas fire impinging on
the surface [1]
Heat flux (kW/m2)
10
1
0.1 1 10
(r + H )/Lf
measurements on exposed and unexposed sides scenarios. The increase in the heat flux over the
of the ceiling material. unconfined ceiling data of Hasemi et al. [38],
The heat fluxes determined in these with peak heat fluxes of approximately 90 kW/
experiments are provided in Fig. 25.18, plotted m2, may be due in part to the thicker layer of
as a function of (x + x0 )/L, where x0 is the virtual flames formed in the corridor. However, due to
source and L is the flame length from the virtual the method used to develop the heat fluxes, the
source. This plot indicates that heat fluxes as high accuracy of the heat flux data reported by
as 160 kW/m2 can exist in these types of Hinkley et al. is uncertain [44, 45].
www.ebook777.com
Fig. 25.18 Heat fluxes 1.0

to the ceiling in a corridor 0.8
with an exposure fire at 0.6 Distance of burner
the closed end [44] beneath ceiling (cm)
0.4 37
66
Normalized heat transfer, q /q 0

90
0.2
120
0.1
0.08
0.06
0.04
0.02
0.01
0.008
0.006
0.004
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.5
Normalized distance, x /
Lattimer et al. [47] conducted a separate study the tests and were reported as total flame lengths,
to measure and correlate heat fluxes from a fire at which include the part of the height of the flame
the closed end of a corridor. The apparatus and below the ceiling and the flame extension along
burner were similar to that used by Hinkley the ceiling.
et al. [45] The corridor was 2.44 m (8 ft) long, Figure 25.19 contains a plot of the flame
1.22 m (4 ft) wide, and 2.1 m (6.8 ft) high, with length data from this study along with data
one end of the corridor blocked. The bottom from Hinkley et al. [45] in a similar corridor
portion of the corridor was open to allow air to apparatus. The flame length data from these
flow freely into the corridor along its length. tests agree well with flame lengths reported by
Tests were conducted with the interior surface Hinkley et al. [45] The correlation in the plot is
of the corridor apparatus lined with a 25-mm described using the following relation:
(1.0 in.) thick, 96 kg/m3 (6 lb/ft3) Unifrax
0 2=3
Durablanket noncombustible ceramic blanket. L f , tip ¼ 0:075Q ð25:23Þ
A rectangular propane gas sand burner was
located at the blocked end of the corridor. The where Lf,tip is the total flame length determined
burner had dimensions of 1.15 m (3.8 ft) wide by adding flame height below a ceiling with
and 0.46 m (1.5 ft) deep. Tests were conducted flame extension along a ceiling (m) and Q0 is
with the burner located 0.60 m (2.0 ft) and 1.1 m the heat release rate of the fire per unit width of
(3.6 ft) below the ceiling and using heat release a hallway (kW/m). The correlation represents a
rates ranging from 100 to 400 kW. best fit to the corridor data as well as larger-scale
Thermal characterization included the mea- test data from fires in tunnels [48], which is a
surement of flame lengths, total heat flux to the similar geometry. The correlation with the corri-
ceiling, and gas temperatures just below the ceil- dor and tunnel flame length data is shown in
ing. Total heat flux was measured using water- Fig. 25.20. The data are a reasonable fit for all
cooled Schmidt-Boelter-type heat flux gauges. the data but do underpredict the higher heat
Flame lengths were visually determined during release rate data by Hinkley et al. [45]
764 B.Y. Lattimer
Fig. 25.19 Flame length 10

data from tests in corridor
tests by Lattimer et al. [47] Hinkley et al.40
and Hinkley et al. [45] 8 H = 1.1m
Flame length, Lf,tip (m)

H = 0.6m
Lf = 0.075Q ⴕ 2/3
6
0
0 100 200 300 400 500 600
Heat release rate per unit width, Qⴕ (kW/m)

correlation compared
with corridor and tunnel Tunnel data
100 Corridor data
data [48]
Lf = 0.075Q 2/3
Flame length, Lf,tip (m)
80
60
40
20
0
0 10,000 20,000 30,000 40,000 50,000
Heat release rate per unit width, Q (kW/m)
A plot of the average heat fluxes to the corri- where H is the distance between the fire and the
dor ceiling is provided in Fig. 25.21 as a function corridor ceiling (m), x is the distance along the
of dimensionless distance along the flame length. corridor (m), and Lf,tip is the flame tip length (m).
Also shown in the plot are some of the data from Close to the impingement point of the fire, the
Hinkley et al. [45] The data from these tests heat fluxes were measured to reach 160 kW/m2.
match well with data from the tests by Hinkley This is 70 kW/m2 higher than similar fires in an
et al. [45] A correlation for the heat flux along the unconfined ceiling geometry.
ceiling is also shown in the figure and can be All of these tests are with the gas burner width
predicted through these equations: being the same as the corridor width. As the
width of the fire is decreased relative to the
00
q ¼ 160 ðx þ H Þ=L f , tip < 0:25 ð25:24Þ corridor width, heat fluxes are expected to pro-
vide results closer to the unconfined ceiling data.
00 1=3 Heat release rate and fire location below the
q ¼ 26:39½ðx þ hÞ=L f , tip
ð25:25Þ ceiling will also affect heat fluxes to the ceiling
ðx þ HÞ=L f , tip 0:25 in these situations.
www.ebook777.com
Fig. 25.21 Heat fluxes Hinkley et al.40

to a corridor ceiling with
q = 160 (x + H )/Lf,tip < 0.25
dimensionless distance
q = 26.39*[(x + H )/Lf,tip–1.3 (x + H )/Lf,tip > = 0.25
along the flame length
1000 H = 1.1 m
H = 0.6 m

100
10
1
0.01 0.1 1 10
(x + H )/Lf, tip
Fires Beneath I-Beams

Ceiling
Three studies have evaluated the heat flux inci- Downward face
dent onto an I-beam mounted to a ceiling with an of upper flange
exposure fire impinging on the beam [38, 43, 49]. Web Upward face
These studies all measured the heat flux to the of lower flange
four surfaces shown in Fig. 25.22 on the I-beam:
downward face of the lower flange, upward face
of the lower flange, the web, and downward face Downward face
of lower flange
of the upper flange. For each of these surfaces,
Fire
heat fluxes were measured from the stagnation
point of the fire (centerline of the fire) past the Fig. 25.22 Location of heat flux measurements on
location of the flame tip. I-beams
The study by Wakamatsu et al. [49] provides a
framework for determining heat fluxes to differ-
near the upper flange, LC. Heat fluxes along the
ent parts of the I-beam. The I-beam evaluated in
lower flange were taken to be a function LB while
the study was 3.6 m long, a web 150 mm high
heat fluxes to other surfaces were related to LC.
and 5 mm thick, and flanges 75 mm wide and
Flame lengths were related to the dimensionless
6 mm thick. Tests were performed using fires
Q*, as defined in Equation 25.14, with D being
from 0.5- or 1.0-m propane burners with heat
replaced by the appropriate distance between the
release rates ranging from 100 to 900 kW. The
fire and the flange,
distance between the fire source and I-beam was
also varied. Q
When the fire impinges on the I-beam, the Q*HB ¼ pffiffiffi 5=2 ð25:26Þ
ρ1 C p T 1 g H B
flame length is different on the lower flange
compared to the flame length on the upper flange Q
(Fig. 25.23). Flame lengths along the lower Q*HC ¼ pffiffiffi 5=2 ð25:27Þ
flange, LB, were shorter than those observed ρ1 C p T 1 g H C
766 B.Y. Lattimer
Fig. 25.23 Flame lengths a

(a) along the lower flange 10.0
(Equation 25.28a) and (b)
along the upper flange in
I-beam tests performed by y = 2.3x 0.3
Wakamatsu et al. [49]
(LB + HB) /HB

1.0
0.1
0.0 0.1 1.0 10.0
Q*HB
b
10.0
(LC + HC) /HC
1.0
0.1
0.0 0.1 1.0 10.0
Q*HC
Correlations were developed to predict the flame virtual origin, z0 , was determined using Equation
tip length along the lower and upper flanges: 25.19. The variation in the heat flux along
the downward face of the lower flange with hori-
ðLB þ H B Þ=H B ¼ 2:3Q*0:3
HB ð25:28aÞ zontal distance, r, from the stagnation point is
shown in Fig. 25.25. The data appear to fall
ðLC þ H C Þ=H C ¼ 2:9Q*0:4 ð25:28bÞ between the range of the data measured in the
HC
unconfined ceiling tests, which are represented
The heat flux measured at the stagnation point by the dashed and solid lines. These heat fluxes
on the downward face of the lower flange was were the highest measured on the I-beam assem-
found to be the same as that measured for a fire bly and can be estimated using the following
beneath a ceiling (Fig. 25.24). The location of the correlation:
www.ebook777.com
Fig. 25.24 Heat flux at the 100

stagnation point on the
Lower flange downward
downward face of the lower
flange [49] Free flame H = 1.0 (m)
H = 0.6 (m)
80
Ceiling test H = 1.0 (m)
H = 0.6 (m)
Beam test HB = 1.0 (m)
Heat flux, qs (kW/m2)

HB = 0.6 (m)
60 HB = 1.2 (m)
40
20
0
0 2 4 6 8 10
Lf /(H + z ¢ )

along the downward face
of the lower flange [49]
10
Flame tips
1
0.1 1.0 10.0
(r + HB + z¢ )/(LB + HB + z¢ )
H = 1.0 m Q = 100 kW H = 1.2 m Q = 540 kW

Q = 150 kW Q = 750 kW
Q = 200 kW Q = 900 kW
H = 0.6 m Q = 95 kW
Q = 130 kW Flat ceiling maximum
Q = 160 kW Flat ceiling minimum
768 B.Y. Lattimer

along the upward face
of the lower flange (key
same as in Fig. 25.25) [49]

10
1
0.1 1.0 10.0
(r + HC + z ¢ )/(LC + HC + z¢ )

along web (key same
as in Fig. 25.25) [49]
10
1
0.1 1.0 10.0
(r + HC + z¢ )/(LC + HC + z¢ )
00
q ¼ 518:8e3:7w ð25:29aÞ where
0
0

where w ¼ r þ HC þ z = LC þ HC þ z ð25:30bÞ
0
0

w ¼ r þ H B þ z = LB þ H B þ z ð25:29bÞ The lowest heat fluxes on the I-beam were
measured on the downward facing part of the
The heat fluxes to the upward face of the upper flange. As seen in Fig. 25.28, heat fluxes
lower flange and the web are shown in to this part of the I-beam are slightly less than
Figs. 25.26 and 25.27 to be lower than those on those measured on an unconfined ceiling. Heat
the downward face of the lower flange. This was fluxes to the downward face of the upper flange
attributed to the lower flange shielding these can be estimated using the following fit to the
parts of the I-beam from radiative and convective data:
heat transfer. These data can be represented by
00
the following expression: q ¼ 100:5e2:85w ð25:31aÞ
00
q ¼ 148:1e2:75w ð25:30aÞ where
www.ebook777.com

along the downward face
of the upper flange (key
same as in Fig. 25.25) [49]

10
1
0.1 1.0 10.0
(r + HC + z ¢ )/(LC + HC + z ¢ )

0 0
w ¼ r þ HC þ z = LC þ HC þ z ð25:31bÞ using the appropriate expression for w provided
in Equations 25.29b, 25.30b, and 25.31b.
Figure 25.30 provides a plot of this correlation
Myllymaki and Kokkala [43] evaluated the along with the I-beam data [43].
use of the approach and data of Wakamatsu
et al. [49] to estimate heat fluxes onto I-beams
exposed to fires as large as 3.9 MW. They found Burning Walls and Ceilings
that for fires over 2.0 MW, the correlations
suggested for the upward face of the lower Fires from burning boundaries typically produce
flange, web, and downward face of the upper thinner flames than those generated by exposure
flange underestimate the heat flux to these areas fires. As a result, heat fluxes from burning
on the I-beam. For these large fires, the I-beam boundary flames are typically lower than those
becomes completely engulfed in fire. As a result, measured for exposure fires in a similar geome-
heat fluxes on all parts of the I-beam follow the try. As was the case with heat fluxes from expo-
correlation suggested for the downward face of sure fires, heat fluxes from burning boundaries
the lower flange provided in Equation 25.29. are dependent on the geometry of the burning
Heat fluxes to the downward face of the lower surfaces.
flange, the upper flange, and the web are shown
in Fig. 25.29, along with the correlations
recommended by Wakamatsu [42]. The highest Wall Fires
heat fluxes measured in the tests performed by
Myllymaki and Kokkala [43] were approxi- Heat fluxes from a burning wall flame back to the
mately 130 kW/m2 and were along the down- surface have been studied fairly extensively.
ward face of the upper flange. Most of the work in this area has been performed
Data from these studies demonstrate that the with smaller fires. Though the data indicate that
heat flux to the I-beam can be conservatively these heat fluxes are dependent on both fire size
estimated using the bounding heat flux correla- and smoke production, no reported study has
tion in Equation 25.32: fully characterized this behavior.
00
ð25:32aÞ Much of the detailed heat flux measurements
q ¼ 120 w 0:5
for fires produced by burning flat surfaces have
00 been done with smaller-scale fires (<100 kW).
q ¼ 682expð3:4wÞ w > 0:5 ð25:32bÞ
Through this work, the heat fluxes from flames
770 B.Y. Lattimer
a Lower flange downward

b Ceiling/Upper flange
c Web
160 160 160
140 140 140
120 120 120
100 100 100

Heat flux (kW/m2)
Heat flux (kW/m2)
Heat flux (kW/m2)

80 80 80
60 60 60
40 40 40
20 20 20
0 0 0
0.00 0.50 1.00 1.50 2.00 0.00 0.50 1.00 1.50 2.00 0.00 0.50 1.00 1.50 2.00
w = (r + HB + z ¢ )/(LB + HB +z ¢ ) w = (r + HC + z ¢ )/(LH + HC + z ¢ ) w = (r + H web + z ¢ )/(L web + H web + z ¢ )
C
Test 1 Test 2 Test 1 Test 2 Test 1 Test 5

518.8 exp (–3.7w) 518.8 exp (–3.7w) 518.8 exp (–3.7w)
100.48 exp (–2.8549w) 148.13 exp (–2.7w)
Fig. 25.29 Heat fluxes measured at the (a) downward face of the lower flange, (b) upper flange, and (c) the web on an
I-beam in the study by Myllymaki and Kokkala [43], with fires as large as 3.9 MW

measured on the bottom
flange (○), web (□), upper
flange (Δ) of an I-beam
100
mounted below an
unbounded ceiling for fires
565–3870 kW [43]. The
line in the plot is the curve
10
given in Equation 25.32
0.1
0.1 1 10
w (–)
produced by a variety of different burning wicks soaked in different alcohols. Ahmad and
materials have been characterized [50–53]. All Faeth performed flat wall fire experiments using
of these studies were conducted with fires over a 0.66-m-wide, 0.81-m-high flat wall test appara-
flat, solid surfaces except the study of Ahmad tus, with the lower part of the wall being an
and Faeth [50, 51], which was performed using alcohol-soaked wick. Different wick heights
www.ebook777.com
Fig. 25.31 The heat 100

fluxes from burning flat
surfaces for fires less than Particle board
100 kW [50–52] Flexible foam
E EEMM M
M Carpet (nylon/wool)
P P
PP EP Rigid foam
M EE PMMA
10 Aircraft panel
Heat flux (kW/m2)

M
M M Methanol
M E Ethanol
P 1-Propanol
P
Equation 32
Equation 33
0.1
0.1 1 10 100
z /Lf
and different types of alcohol were included in bottom part of the fire and the empirical
the study. Data from Quintiere et al. [52] were constants that govern the decay. Similar to that
attained using samples 0.28 m by 0.28 m exposed proposed by Hasemi [55], the line in the plot is an
to different external heat fluxes to generate dif- average fit to the data:
ferent heat release rate fires from the same sam- 00
ple. Experiments performed by Orloff et al. [53] q ¼ 25 z=L f 0:5 ð25:34aÞ
were conducted using a 0.41-m-wide, 1.57-m- 2:5
00
high sample of PMMA. q ¼ 4:4 z=L f z=L f > 0:5 ð25:34bÞ
Data from the studies of Ahmad and Faeth
[50, 51] and Quintiere et al. [52] are shown in A more conservative fit that bounds this data set
Fig. 25.31. Heat fluxes are approximately was developed:
20–30 kW/m2 in the lower part of the flame 00
q ¼ 30 z=L f 0:7 ð25:35aÞ
(z ¼ 0.5Lf) for a wide range of fuels. Peak heat
fluxes measured by Orloff et al. [53] (22 kW/m2) 2:5
00
were also in this 20–30 kW/m2 range. The value q ¼ 12:3 z=L f z=L f > 0:7 ð25:35bÞ
of Lf can be determined by using a flame height
correlation for a line fire, such as that proposed Line burners have been used by some
by Delichatsios: [54] researchers to simulate a fire produced by a burn-
ing surface such as a wall. Hasemi [55, 59, 60]
0 2=3
L f ¼ 0:052Q ð25:33Þ measured the heat flux from a methane line
burner fire to an incombustible wall. In this
where Q0 is the heat release rate per unit length of study, the fire heat release rate per unit length
burning wall. of burner (0.30 m) was varied from 16.7 to
Several empirical correlations have been pro- 218.2 kW/m and two different liner burner
posed in the literature [55–58] to predict heat widths (0.037 m and 0.082 m). For the test
fluxes to walls. All correlations assume a con- conditions considered, the heat fluxes along the
stant heat flux in the lower part of the fire and a flame are seen in Fig. 25.31 to be similar for each
power law decay above this. The difference in test condition. In addition, heat fluxes measured
these correlations is the peak heat flux over the in this study are shown in Fig. 25.32 to be similar
772 B.Y. Lattimer
Fig. 25.32 The heat 100

fluxes from methane line
burners against a flat wall Methane57
Propane63
[29, 55, 61]
Propane31
Equation 34
Equation 35
10
Heat flux (kW/m2)

1
0.1
0.01 0.1 1 10 100
z /Lf
to those shown in Fig. 25.31. The correlations Figure 25.33 provides the heat flux data for
presented in Equations 25.34 and 25.35 ade- the different materials included in the study.
quately bound the data. Line burner experiments Peak heat fluxes measured for the different
using propane as fuel have resulted in higher heat materials were measured to range from 25 to
fluxes than those measured with methane as 60 kW/m2. Heat fluxes from burning masonite
the fuel. In tests using propane with Q0 ¼ 83 board, cardboard, and white pine board were in
167 kW/m, Kokkala et al. [61] and Lattimer [29] the 20 to 30-kW/m2 range, similar to that
both measured heat fluxes of approximately measured in experiments by Ahmad and Faeth
45 kW/m2 in the lower half of the flame (z ¼ 0.5 [50, 51] and Quintiere et al. [52]. However, fires
Lf). Though not shown on the plot, Foley and involving PMMA, polyurethane foam, and
Drysdale [62] measured 40–50 kW/m2 from velour fabric were all measured to produce heat
propane line burners with Q0 ¼ 11.6 and fluxes greater than 30 kW/m2. The PMMA and
20.9 kW/m. These data indicate that the radiation polyurethane foam had the highest flame lengths
from the fire to the surface is dependent on fuel of all the materials (~1.75 m), which is compara-
smoke production. ble to the flame lengths reported by Quintiere
Slightly larger-scale fire tests were performed et al. [52] for similar materials (PMMA and
by Kulkarni et al. [63, 64] In this study, heat flux flexible foam). This indicates that the heat
measurements were made along the length of release rates for the PMMA and polyurethane
different 0.3-m-wide, 1.2-m-high samples of foam are comparable in the two studies. The
solid combustibles. Fires were initiated using a reason for the differences in the peak heat fluxes
line burner at the bottom of the sample, and heat (e.g., 30–60 kW/m2 in tests by Kulkarni
fluxes were continuously measured during the et al. [63, 64] with PMMA, while 20–26 kW/m2
test. Heat fluxes and flame lengths were continu- in tests by Quintiere et al. [52] ) is not known.
ously monitored as the fire spread along the Less detailed heat flux measurements have
combustible material. These transient heat flux been reported in the literature for larger fires.
and flame length measurements were averaged Orloff et al. [65] and Delichatsios [54] reported
over particular time periods and plotted to deter- data on heat fluxes from flames produced by a
mine the heat flux at different locations along the 3.6-m-high burning PMMA wall. Total heat
flame length. fluxes incident on the PMMA were calculated
www.ebook777.com

for different materials
[64, 64] Cardboard
PMMA (black)
Masonite board
Polyurethane foam
White pine board
10 Velour fabric
Heat flux (kW/m2)

Equation 34
Equation 35
0.1
0.1 1 10 100
z /Lf
Fig. 25.34 Heat flux from 60

a PMMA wall flame
back to the fuel surface
50
[53, 54, 65]
Heat flux (kW/m2)
40
30
20
H = 1.2 m55
10 H = 3.6 m67
H = 3.6 m56
0
0 1 2 3 4 5 6
Distance along burning wall, z (m)
using theory and mass loss rate data. Heat fluxes for methane, ethane, ethylene, and propylene
are shown in Fig. 25.34 to increase with height. fires were reported. The impact of fire size on
All the data in this plot were at positions where the heat flux distribution along the height of the
z/Lf is less than 0.5. This behavior is different panel is shown in Fig. 25.35. Similar to the
than that observed with smaller fires, where heat PMMA results, the heat fluxes were measured
flux is relatively constant over this region. to increase with height in the test with the higher
Markstein and de Ris [66] also explored the heat release rate (816 kW/m).
effects of larger fire size and soot production on The heat flux from the flame is shown in
the heat flux incident on the burning surface. The Fig. 25.36 to also be a function of fuel smoke
apparatus used in the study was 0.38 m wide and production rate. Methane and ethane have low
1.98 m high, with the bottom 0.79 m of the wall smoke yields (less than 0.013 g/g) [67] and are
being a sintered metal gas burner. Heat flux data measured to produce heat fluxes as high as
774 B.Y. Lattimer

along a propylene gas wall
fire at different heat release 60 Q = 88 kW/m
rates per unit width Q = 816 kW/m
[66]. Burning wall height 50
Heat flux (kW/m2)

was 0.79 m
40
30
20
10
0
0.0 0.5 1.0 1.5 2.0 2.5
Height, z (m)
Fig. 25.36 Heat flux at 70

a height of z ¼ 1.25 m
for different size fires
Heat flux at z = 1.25 m (kW/m2)
60
and different fuels
[66]. Burning wall height 50
was 0.79 m
40
30
Methane
20 Ethane
Ethylene
10 Propylene
0
0 200 400 600 800 1000
Heat release rate per unit width, Q (kW/m)
35–38 kW/m2. The smoke yield of ethylene Table 25.3 Peak heat flux from flames measured in
(0.043) [67] is less than that of propylene 2.4-m-high, 0.60-m-wide plywood wall experiments [68]
(measurements up to 1.8 m above floor)
(0.095), but similar heat fluxes were measured
with height along the apparatus. Peak heat fluxes Radiant Heat release rate Peak
of 59 kW/m2 were measured for the largest pro- exposure per unit width Q0 heat flux
Fuel (kW/m2) (kW/m) (kW/m2)
pylene fire considered in the study.
Plywood 4.8 175 38
Heat fluxes were measured in tests on large
(Finished side 5.2 197 40
(2.4-m-high, 0.60-m-wide) plywood walls [68]. exposed)
The peak heat fluxes measured in these tests are 7 292 45
provided in Table 25.3 for various preheat Plywood 7.5 217 45
levels. As the heat release rate per unit width (Unfinished 11 417 50
increases, the heat flux from the fire to the wall side exposed)
increases. Though heat fluxes are not as high
as those measured for a burning PMMA wall, Similar experiments were performed by
the heat flux is 8–20 kW/m2 higher than the Ohlemiller and Cleary [69] on composite
30-kW/m2 peak level measured in the smaller- panels. The peak heat fluxes measured in this
scale tests. study are provided in Table 25.4. Similar to
www.ebook777.com
the results of Delichatsios et al. [68], heat fluxes Corner Wall Fires
were measured to increase with an increase in
heat release rate (i.e., increase in external heat Limited work has been performed to quantify the
flux). heat fluxes from burning boundaries in a corner.
Data presented in this section demonstrate In general, the heat fluxes produced by burning
that both heat release rate and smoke production corner walls are higher than those produced by a
rate of the fuel can influence the heat flux levels wall flame.
produced by wall flames back onto the burning Qian et al. [70, 71] measured heat fluxes pro-
surface. Larger fires with high smoke production duced in a corner of burning PMMA walls
rates can result in heat fluxes to the walls of beneath an incombustible ceiling. In these
approximately 60 kW/m2. Additional research experiments, a 1.6-m-high corner was lined
needs to be performed to better quantify the with 12.7-mm-thick PMMA 0.20 m in width.
transition between the smaller-fire experiments During the tests, the walls were ignited using a
and the large-fire results. torch at the bottom of the corner and were
allowed to burn until flames had spread to the
top of the walls. The peak heat release rate of the
Table 25.4 Heat fluxes from 1.2-m-High, 0.3-m-wide fire was estimated to be 80 kW. The heat fluxes
Composite panel fires [69] measured during the growing fire are shown in
Radiant Heat release rate Peak heat Fig. 25.37. In the lower half of the flame,
exposure per unit width Q0 flux heat fluxes were measured to be, on average,
Fuel (kW/m2) (kW/m) (kW/m2) 33 kW/m2. Above this, the heat fluxes were
Fire-retarded 2.5 — 35 measured to decay similarly to heat fluxes
vinyl ester 7.5 — 48 measured for wall fires (see Equations 25.35
11 — 52 and 25.36).
Polyester 0 — 35
A series of experiments were conducted by
Hasemi et al. [25] using L-shaped sintered

burning PMMA corner
•
walls (1.6 m high and q = 0.822 (x /xf )–2.294
0.20 m wide) [71]
R ^2 = 0.926
1
q ~ (x /xf)–1.8
0.1
0.01
0.1 1 10
Normalized height, z /Lf
776 B.Y. Lattimer
a b
100 100
Q = 15 kW Q = 40 kW
Q = 20 kW Q = 50 kW
Q = 25 kW Q = 60 kW
Q = 30 kW
Q = 40 kW
10 Q = 50 kW 10

Q = 60 kW
1 1
Z/d = 2 Z/d = 4
0.1 0.1
0.1 1 10 100 0.1 1 10 100
z/Lf z/Lf
d = 0.225 m, Z = 0.45 m d = 0.225 m, Z = 0.90 m
Fig. 25.38 Heat fluxes to a corner from a simulated burning corner fire using propane as fuel [25]
metal burners mounted to the walls of the corner corner. Burners were placed in a 2.4-m-high cor-
to simulate burning corner walls. Using propane ner with a ceiling, with all surfaces constructed
gas as fuel, experiments were conducted using of noncombustible materials. In this study, heat
two different burner sizes (0.23 m wide and fluxes were measured for different size burners
0.45 m high, 0.23 m wide and 0.90 m high) (single side length of 0.17 m, 0.30 m, and 0.50 m)
mounted to an open corner of walls with no and various heat release rates (50–300 kW).
ceiling. The heat fluxes above the burners in Similar to the approach used to develop the
these fires are provided in Fig. 25.38 for fire heat flux correlations for area burners, burning
heat release rates of 15–60 kW. The line on the boundary correlations were developed for three
plots represents the decay in the heat flux of a regions in the corner: along the height of the
wall fire. Peak heat fluxes in the lower part of walls in the corner, along the top of the walls
the flame were measured to range from 28 to near the ceiling, and along the ceiling. The region
38 kW/m2 and were constant up to approxi- containing the walls in the corner extended from
mately half the flame length. Above this, heat the top of the fire to approximately 1.8 m above
fluxes were measured to decay in a manner simi- the fire, which is approximately the ceiling
lar to that determined for burning walls. height minus twice the ceiling jet thickness
Hasemi et al. [72] also performed tests in a (δ ¼ 0.1H ). Above 1.8 m was considered to be
1.8-m-high corner with a ceiling. Tests were the region along the top of the wall, or the wall-
performed with the top 1.35 m of the corner ceiling interface region.
lined with 0.23-m-wide sintered metal burners Heat flux data for these fires were normalized
and with the top 0.45 m of the corner lined with with respect to the flame tip location. The flame
sintered metal burners. Heat fluxes to the ceiling tip was the farthest distance at which flaming was
were measured to be as high as 40 kW/m2, while visually observed. In cases where the fire
heat fluxes as high as 60 kW/m2 were measured impinged and flowed along the ceiling, the
along the top of the walls near the ceiling. flame tip length was taken to be the corner height
Lattimer et al. [29] performed a detailed study plus the flame extension along the ceiling.
using L-shaped propane line burners in the Lattimer et al. [29] developed the following
www.ebook777.com
correlation to predict the flame tip of a burning normalized with respect to the flame tip. Peak
boundary fire: heat fluxes were measured over the initial half of
the flame length. Above this, heat fluxes decayed
*1=2
L f , tip =d ¼ 5:9Qd ð25:36Þ in a fashion similar to that observed for wall fires.
The line in the plot represents a fit to the data,
where dimensionless Qd* is
which can be described by the following
Q expressions:
Q*d ¼ pffiffiffi ð25:37Þ
ρ1 C p T 1 gd 5=2 00
qmax ¼ 70 z=L f < 0:5 ð25:38aÞ
Equations 25.35b and 25.36 are similar to those 2:8
00
used in predicting flame heights from area qmax ¼ 10 z=L f z=L f > 0:5
burners in a corner except the length scale is d, ð25:38bÞ
which is the width of the burning area on the wall
or the side of a single L-shaped burner. In the Heat fluxes in the decay region (z/Lf ) > 0.5
L-shaped line burner tests, d is the length of a decrease with dimensionless height raised to the
single side; however, in a burning corner d was 2.8 power, which is a slightly lower power than
found to be the average width of the burning on the decay for wall fires (2.5).
the walls. For fires in a 2.4-m-high corner, the Peak heat fluxes in the corner are shown in
width of the burning 0.90 m above the floor was Fig. 25.40 to have some dependence on the heat
found to adequately represent the average burn- release rate of the fire. The increase in the peak
ing width [29]. heat flux with increase in fire size was attributed
The vertical distribution in the maximum heat to an increase in radiative pathlength. Assuming
flux along the walls near the corner is shown in the gases to be gray, the following curve fit was
Fig. 25.39 plotted with the vertical distance developed:

simulated corner wall fires
back to the corner walls at a
height less than 1.8 m
above the floor. L-shaped
Maximum heat flux in corner (kW/m2)
line burner with single side

lengths of 0.17 m (○), 100
0.30 m (Δ), 0.30 m
(elevated) (∇), and 0.50 m
(□) and fire sizes ranging
from 50 to 300 kW [29]
10
0.1
0.01 0.1 1 10
z/Lf,tip
778 B.Y. Lattimer
Fig. 25.40 Peak heat flux 140

to the walls in the corner as
a function of wall heat 130 q peak = 70
release rate less than 1.8 m 120 q peak = 120 [1 – exp(–0.1Q1/2)]
above the floor. L-shaped
line burner with single side 110
Peak heat flux in corner (kW/m2)

lengths of 0.17 m (○), 100
0.30 m (Δ), 0.30 m
(elevated) (∇), and 0.50 m 90
80
from 50 to 300 kW [29]
70
60
50
40
30
20
10
0
0 50 100 150 200 250 300 350 400
Heat release rate, Q (kW)
00
h i 00 h i x
qpeak ¼ 120 1 exp 0:1Q1=2 ð25:39Þ q
¼ exp 1:0ðx=dÞ2 < 1:3 ð25:41aÞ
00
qmax d
Based on Equation 25.38b, a more conserva- 00 x 1:8
tive fit to the data in Fig. 25.39 was developed: q x
00
¼ 0:30 1:3 ð25:41bÞ
qmax d d
00 00
qmax ¼ qpeak z=L f 0:5 ð25:40aÞ
Burning boundary beneath a ceiling will form a
00 00 00
ceiling jet that will heat the top part of the walls
qmax ¼ qpeak 5 z=L f 0:5 qpeak 27
and the ceiling. The maximum heat flux along the
0:5 < z=L f 0:7 top part of the walls is shown in Fig. 25.42. The
ð25:40bÞ line in the plot represents a fit to the data, which
can be represented by the following expressions:
2=8
00
qmax ¼ 10:0 z=L f z=L f > 0:7 ð25:40cÞ
00 xþH
qmax ¼ 120 0:52 ð25:42aÞ
The maximum heat fluxes along the height of L f , tip
the corner shown in Figs. 25.39 and 25.40 were
measured approximately 0.05–0.10 m outside of 00 x þ H 3:5 x þ H
qmax ¼ 13:0 > 0:52
the corner. Heat fluxes decrease with horizontal L f , tip L f , tip
distance from the corner. The horizontal heat flux ð25:42bÞ
distributions at heights less than 1.8 m above the
floor are shown in Fig. 25.41 to be half-Gaussian The assumed plateau in the correlation is based
in shape over the flame, but decays slower than on the maximum heat flux expected from a flame
predicted by a half-Gaussian curve outside of the in this configuration. This correlation is the same
flaming region. The line in this plot is a fit to the as that determined for area fires in a corner.
data, which can be represented by the following The heat flux to the ceiling was correlated to
expressions: the dimensionless distance away from the burner,
www.ebook777.com
Fig. 25.41 Horizontal 1.2

heat flux distribution on the
walls out from the corner at
a height less than 1.8 m
above the floor [29] 1.0
0.8
max
Heat flux, q /q
0.6
0.4
0.2
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
x /d
Fig. 25.42 Maximum 1000

heat flux along the top of
the walls with a simulated
burning boundary fire in the
corner. L-shaped line
burner with single side
Maximum heat flux, q max (kW/m2)
lengths of 0.17 m (○),

0.30 m (Δ), 0.30 m 100
(elevated) (∇), and 0.50 m
from 50 to 300 kW [29]
10
1
0.1 1 10
(x + H ) /Lf,tip
780 B.Y. Lattimer

along the ceiling with a
simulated burning
boundary fire in the corner.
L-shaped line burner with
single side lengths of
Heat flux to overhead, q (kW/m2)

0.17 m (○), 0.30 m (Δ), 100
0.30 m (elevated) (∇),
and 0.50 m (□) and fire
300 kW [29]
10
0.1
0.1 1 10
(r + H ) /Lf,tip
(r + H )/Lf,tip. A plot of the heat flux versus this ceilings range from 20 to 30 kW/m2, which is
dimensionless parameter is shown in Fig. 25.43. similar to those measured for small wall fires.
The line in this plot is a fit to the data, which are
represented through the following relations: Unconfined Ceiling Fires Heat fluxes from
unconfined ceiling fires were measured by Hasemi
00 rþH
q ¼ 120 0:52 ð25:43aÞ et al. [73] using different sizes of sintered metal
L f , tip propane gas burners mounted into a 1.8-m-square
incombustible ceiling. Using two different circu-
00 r þ H 3:5 r þ H lar burner sizes (D ¼ 0.09 and 0.16 m), heat flux
q ¼ 13:0 > 0:52
L f , tip L f , tip to the ceiling was measured for fire heat release
ð25:43bÞ rates of 2.5–38 kW.
This is the same relation used for the top of the The radius of the flame (intermittent)
corner walls, except the length scale in the over- measured using the two burners is shown in
head data is r. In addition, this is the same rela- Fig. 25.44 to be slightly dependent on burner
tion determined using the ceiling heat flux data size, with the larger burner having a lower radius.
from tests with an area burner. However, as the fires become larger, the depen-
dence on burner diameter becomes small. Flame
lengths are proportional to the heat release rate
Ceiling Fires raised to the one-half power.
Hasemi et al. [73] also measured the heat
Heat fluxes from burning ceilings have been fluxes as a function of distance from the center
evaluated for both unconfined ceilings and of the burner. The measured heat fluxes are shown
ceilings in a corridor. Due to buoyancy effects, in Fig. 25.45 to be at peak levels in the first 0.4Lf
flames from burning ceilings tend to be relatively and then decay with distance from the burner.
thin. As a result, peak heat fluxes from burning Peak heat fluxes were measured to range from
www.ebook777.com
Fig. 25.44 Flame radius 0.8

produced by a simulated
burning ceiling in an 0.7
unconfined area [73] φ = 0.09 m
0.6 y = 0.1324x 0.4465
Flame radius (m)

0.5
0.4 φ = 0.16 m
y = 0.0848x 0.5644
0.3
0.2
0.1
0
0 5 10 15 20 25 30 35 40 45
Heat release rate (kW)

from fires beneath an
unconfined ceiling [73] d = 0.09 m
d = 0.16 m
One dimensional
ceiling flames
10
Total heat flux (kW/m2)
Wall fires
1
0.1
0.1 1 10
r /Lf (calculation)
16 to 27 kW/m2, with the smaller burner produc- Ceiling Fires in a Corridor Heat fluxes from
ing higher heat fluxes. These peak heat fluxes flames produced by burning ceilings in a corridor
were similar to those measured for burning were investigated by Hasemi et al. [72]. Tests were
ceilings in a corridor (i.e., one-dimensional ceil- performed beneath a 2.73-m-long ceiling with two
ing flames) and for small wall fires. Heat fluxes 0.10-m-high soffits mounted along the length
from the unconfined ceiling fires were measured of the ceiling to form a 0.30-m-wide channel.
to decay at a rate between that measured for wall At the closed end of the channel, a 0.30-m-wide,
fires and that observed for a burning ceiling in a 0.04-m-long porous propane burner was mounted
corridor. in the ceiling. Heat flux distributions along the
782 B.Y. Lattimer

produced by a burning
ceiling in a corridor [72]
LH = 0.0122QL
1
LH (m)
1
0
0 50 100 150 200 250
Qᐉ (kW/m)
corridor were measured for fire heat release rates

ranging from 10 to 50 kW (33–166 kW per meter Burning Parallel Vertical Surfaces
of corridor width).
A common configuration encountered when
The intermittent flame lengths from these fires commodities are being stored in rooms or
are seen in Fig. 25.46 to increase linearly with warehouses is parallel vertical surfaces. As a
heat release rate per unit hallway width. A fit to result, several studies have focused on both
these data produced the following relation to experimentally and analytically characterizing
predict flame length due to a burning ceiling in this configuration [62, 74, 75]. Ingason and de
a corridor: Ris [76] also performed experiments in a rack
0
storage configuration with a fire between four
L f ¼ 0:0122Q ð25:44Þ equally spaced storage towers.
Part of the work by Tamanini [74, 75]
The heat flux distributions along the center of investigated the effects of wall spacing on the
the corridor are shown in Fig. 25.47 for the fuel mass loss rate of combustible parallel verti-
different fires considered in the study. The line cal walls. Walls were 0.94 m high and 0.460 m
in the plot represents a best fit to the methane line wide with the spacing varied from 0.470 to
burner data of Hasemi [55]. Heat fluxes were 0.025 m and no floor at the base of the walls.
measured to be constant at approximately The average fuel mass loss rate was measured to
20 kW/m2 up to 0.4 Lf. Above this, heat fluxes increase (i.e., the average heat flux to the wall
were measured to decay at a slower rate than that increased) with an increase in spacing until the
previously measured for wall fires. Heat fluxes spacing was less than 0.076 m. At a spacing of
along a flame from a burning ceiling in a corridor 0.038 m or less, the average mass loss rate was
(not shown in the figure) can be determined using less than that measured with no parallel wall. At a
the following expressions: spacing of 0.038 m (or wall height divided by
00 spacing of 25 with a fire size of approximately
q ¼ 20 x=L f 0:4 ð25:45aÞ
180 kW), the flames from the two burning
00 5=4 surfaces were observed to merge together approx-
q ¼ 6:36 x=L f x=L f > 0:4 imately two-thirds the distance up the walls.
ð25:45bÞ
www.ebook777.com

the ceiling of a corridor
[72] Solid line: wall fire correlation Q = 10 kW
Q = 15 kW
Q = 20 kW
Q = 25 kW
Q = 30 kW
Q = 35 kW

Q = 40 kW
Q = 45 kW
Q = 50 kW
10
1
0.01 0.1 1 10
x /Lf
Though not evaluated in this study, the presence The heat flux distributions measured with the
of a floor may cause the flames to merge together fire against the instrumented wall are shown in
at larger spacings. Fig. 25.48. As seen in Fig. 25.48a, heat fluxes
Heat fluxes due to a fire between two parallel reached as high as 80 kW/m2 with an open base
vertical surfaces were measured by Foley and (no floor between the walls). Heat fluxes on the
Drysdale [62]. The study was performed using panel can be estimated using the following
two 0.61-m-wide, 0.81-m-high walls separated expression:
by a gap of 0.06, 0.10, or 0.14 m. The fire was a 0 0:38 1:47
00 *2=3
0.60-m-long propane line burner that had either a q ¼ 67:38 zða=d Þ0:36 = Qd d y =d
11.6-kW/m or a 20.9-kW/m heat release rate per
unit length. One of the walls was instrumented ð25:46Þ
with four heat flux gauges that could be moved to where y0 ¼ 0.5d y with y being the horizontal
measure the heat flux distribution on the walls. distance from the burner centerline.
Heat fluxes were measured as far as 0.150 m With the base of the walls closed (a floor
from the centerline of the wall. For the different between the walls) and the fire against the
gap and heat release rate fires, heat fluxes were instrumented wall, the heat flux data in
measured with the burner against the Fig. 25.48b were seen to be as high as 70 kW/m2.
instrumented wall and with the fire in the center Heat fluxes for this case are slightly lower than the
of the gap between the two walls. The effect of open-base case. A similar expression to that in
air entrainment flow path was also evaluated by Equation 25.45b was developed by Foley and
performing tests with and without a floor Drysdale [62] to predict heat fluxes with the base
between the panels. Results were correlated of the walls closed:
using a/d where the spacing between the walls, 0 2=3 1:2
00 *2=3
a, is divided by the burner length, d, and the q ¼ 23:31 zða=dÞ0:905 = Qd d y =d
dimensionless quantity Qd*, as defined in Equa-
tion 25.14, with d being the burner length. ð25:47Þ
784 B.Y. Lattimer
Heat transfer from flames
a b
100 100
8 8
6 6
4 4
2 2

10 10
8 8
6 6
4 4
2 2
1 1
8 8
6 6
4 4
8 2 4 6 8 2 2 4 6 8 2 4 6 8
1 10 0.1 1 10
0.36 * 2/3 0.38 0.905 2/3 0.67
[z(a/d ) /Q d d(y¢/d ) ] [z(a/d ) /Q d * d(y¢/d ) ]
Fig. 25.48 Heat fluxes measured with the fire against the spacing (✯), 0.10-m spacing (○), 0.060-m spacing (Δ);
instrumented wall with (a) an open base (no floor in the open symbols, Q0 ¼ 11.6 kW/m; closed symbols,
gap) and (b) a closed base (a floor in the gap): 0.140-m Q0 ¼ 20.9 kW/m [62]
Heat fluxes were also measured with the fire in into the fire through the sides of the gap. The
the center of the gap between the two walls. In the following expression can be used to estimate the
case with an open base (no floor), the heat fluxes heat flux to the walls for this case:
were measured to be 50 % lower than those 0 1:34 1:04
00
measured with the fire against the instrumented q ¼ 23:94 zða=d Þ1:7 = Q*D d y =d
wall. As seen in Fig. 25.49a, the peak heat flux
was measured to be approximately 30 kW/m2. ð25:49Þ
This decrease was attributed to the air being
drawn up at the base of the walls, preventing the Additional research needs to be performed
fire from attaching to the instrumented wall. The with this configuration to further validate the
line in the figure is the best fit to the data, which results. Larger-scale tests need to be conducted
are given by the following expression: to verify the results of Foley and Drysdale
[62]. In addition, the transition from wall fire
0 0:806 0:797
00 heat fluxes to gap fire heat fluxes needs to be
q ¼ 22:71 zða=d Þ1:04 = Q*d d y =d
identified. Heat fluxes produced by area
ð25:48Þ fires between parallel walls also need to be
quantified.
The case with the base closed and the fire in the
center of the gap resulted in the highest heat
fluxes measured in the study. As seen in Exposure Fires and Burning Walls
Fig. 25.49b, heat fluxes greater than 100 kW/m2 and Ceilings
were measured in this case. In the tests with the
high heat fluxes, the flames were observed to A series of tests were performed by Lattimer
occupy the width of the gap. This behavior was et al. [77] to investigate the use of steady-state
attributed to only allowing air to be entrained heat flux correlations, developed using burners
www.ebook777.com
a b
2 2
100 100
8 8
6 6
4 4

2
10 10
8 8
6 6
4 4
2 2
1 1
8 8
6 6
4 4
6 8 2 4 6 8 2 4 68 2 4 68 2 4 68 2
1 10 0.1 1 10
[z(a/d ) 1.04 /Q*d d(y/d ) 0.86 ] [z(a/d ) 1.7 /Q*d d(y/d) 1.34 ]
Fig. 25.49 Heat fluxes measured with the fire in the spacing (✯), 0.10-m spacing (○), 0.060-m spacing (Δ);
center of the gap with (a) an open base (no floor in the open symbols, Q0 ¼ 11.6 kW/m; closed symbols,
gap) and (b) a closed base (a floor in the gap): 0.140-m Q0 ¼ 20.9 kW/m [62]
and noncombustible boundaries, for estimating comparison between predicted and measured
the heat fluxes in growing fires. Three tests heat fluxes was done for the region along the
were performed in an 2.4-m-high, 2.0-m-wide top of the wall.
open corner lined with a combustible material. Transient data were averaged every 30 s to
A single test was performed on three different create a reasonable amount of data to compare
lining materials: 12-mm-thick Douglas fir ply- to the developed correlations. A comparison of
wood, 12-mm-thick E-glass fire-retarded vinyl the flame length predicted using Equations 25.13
ester, and 88-mm-thick sandwich composite and 25.36 and the measured flame length is
(76-mm-thick balsa wood with 6-mm-thick shown in Fig. 25.50. The dimensionless length
E-glass fire-retarded vinyl ester facings). The used in this calculation was the width of the
initiating fire in the test was a square propane burner, D, while the burning had spread
sand burner with a 0.17 m side length and a heat laterally less than the width of the burner.
release rate of 100 kW for 10 min followed by When the average lateral flame spread 0.90 m
300 kW for 10 min, total test time of 20 min. above the floor exceeded the burner width, the
Total heat release during the test was measured dimensionless length was taken to be the hori-
by performing oxygen calorimetry on the gases zontal flame front location 0.9 m above the
collected in an exhaust hood, and flame lengths floor. The flame front at 0.9 m above the floor
were measured through visual observation. Heat was approximately the average flame front on
fluxes were measured 0.075 m from the corner the wall.
along at eight different elevations, 0.15 m below Heat fluxes to the walls near the corner are
the ceiling along the top of the wall, and along provided in Fig. 25.51. Measured heat fluxes
the ceiling on a 45 diagonal out from the corner. were slightly higher than values predicted by
Due to mounting the heat flux gauges along the both the initiating fire correlation and the burning
top of the wall too far below the ceiling, no boundary correlation (assuming the heat flux is
786 B.Y. Lattimer
Fig. 25.50 Flame lengths 28

measured in combustible
corner fire tests compared
with the flame length 24
correlations developed for
initiating and burning
boundary fires in Equations
20
25.13 and 25.36: plywood
(□), E-glass FR vinyl ester
(○), sandwich composite
(Δ) [77] 16
Lf,tip /d
12
0
0 2 4 6 8 10 12 14 16 18 20 22
1/2 5/2
Q *d = Q /(ρ∞CρT∞g d )

along the height of the
corner in tests with different
combustible boundaries
compared with the heat flux
predicted using Equation
100
Heat flux in corner, q (kW/m2)
25.38 (—) and Equation

25.15 ( ): plywood
(Δ) [77]
10
0.1
0.01 0.1 1 10
z /Lf,tip
www.ebook777.com

the ceiling during open
corner tests with the corner
lined with a combustible
material compared with the
heat flux predicted using
Equation 25.43: plywood 100
Heat flux to ceiling, q (kW/m2)

(Δ) [77]
10
0.1
0.1 1 10
(r + H ) /Lf,tip
independent of the wall heat release rate). window of a postflashover compartment fire have
Inspection of the data indicates better agreement been performed by Oleszkiewicz [78, 79], Thomas
between the data and the correlations can be and Bullen [80], and Beitel and Evans [81]. In
achieved using the initiating fire correlation up these studies, heat fluxes as high as 200 kW/m2
to when ignition occurs in the corner. After this, have been measured.
the corner wall heat flux correlations in Experiments performed by Oleszkiewicz
Equations 25.38b and 25.40 can be used to esti- [78, 79] were conducted using two differently
mate heat fluxes in the corner. sized full-scale rooms with a wall above the
A comparison of the heat fluxes along the window that extended as much as two stories
ceiling and the heat fluxes predicted using Equa- above the burn room (Fig. 25.53). The effects
tion 25.43 is shown in Fig. 25.52. In general, heat of window size, window aspect ratio, and fire
fluxes are adequately predicted by the correla- size inside the compartment were evaluated in
tion, with heat fluxes as high as 130 kW/m2 the study. Heat fluxes from the flames extending
measured during a test. This indicates that Equa- outside the burn room for different door sizes and
tion 25.43 can be used to estimate heat fluxes to different fires sizes are shown in Figs. 25.54 and
the ceiling near the corner containing the fire. 25.55 for propane gas fires. Note that the heat
release rate of the fires stated in Figs. 25.54 and
25.55 is the ideal heat release rate of the com-
Fires from Windows partment fire, which was determined from the gas
flow rate and the heat of combustion for propane.
Fires that have reached flashover conditions typi- Data in Fig. 25.54 show the effect of fire heat
cally result in burning outside of the actual burn release rate and window size on the heat flux
room. Flames from postflashover fires extending 0.5 m above the window. The distribution in the
out of a building through a window will buoyantly heat flux along the height exterior wall is shown
rise along the exterior of the building. Experiments in Fig. 25.55 for the case with a window 2.6 m
characterizing the heat fluxes to the wall above the wide and 1.37 m high.
788 B.Y. Lattimer
Vertical channel Quintiere and Cleary [32] found that flame

test apparatus lengths for this situation can be estimated using
the relation developed by Yokoi [82]. With Lf
Target being the distance from the bottom of the open-
wall ing to the average flame height, the heat release
rate outside of the compartment, Q, and the
effective diameter of the window, D, can be
used to predict the flame length above the win-
dow with the following expression:
2=3
Q
L f ¼ 0:0321 ð25:50aÞ
D
where
rffiffiffiffiffiffiffiffiffiffiffiffi
Ho W o
D¼2 ð25:50bÞ
2π
Window
Heat Fluxes in Standard Tests
Heat fluxes in some standard tests are provided in

this section to compare with heat fluxes
measured in realistic geometries presented in
Burn room
this chapter. Heat fluxes in room-corner tests
such as ISO 9705 and NFPA 286, Standard
Fig. 25.53 Exterior wall fire test facility used by Methods of Fire Tests for Evaluating Contribu-
Oleszkiewicz [79] tion of Wall and Ceiling Interior Finish to Room

from a window flame 0.5 m 200 0.94 × 2.00
above the top of the 0.94 × 2.70
window for different size 2.60 × 1.37
Heat flux (kW/m2)
propane fires inside the 150 2.60 × 2.00

compartment [79] 2.60 × 2.70
(width × height)
100
50
0
5 6 7 8 9 10 11
Heat release rate (MW)
www.ebook777.com

from window flames along 200 10.3 MW
the exterior wall above a 8.6 MW
2.6-m-wide, 1.37-m-high 6.9 MW
window [79]
Heat flux (kW/m2)

150 5.5 MW
100
50
0
0 1 2 3 4
Height above window (m)
Fire Growth, can be determined using heat flux These furnace exposures are utilized to evaluate
data previously presented in the section on heat the fire resistance of structural elements on
fluxes from exposure fires in a corner. This sec- buildings, on ships, and in some transportation
tion will focus on heat fluxes produced in other applications (e.g., railcars). ASTM E119 is pri-
tests including fire resistance test furnaces and marily used in North America whereas ISO 834 is
the ASTM E84 flame spread test. Note that these used more internationally (e.g., Europe and
heat fluxes, along with most data previously Australia). As seen in Fig. 25.56, the two time-
presented for room-corner tests, were typically temperature curves are similar, with the ISO
measured with a noncombustible, insulating sur- 834 temperatures being slightly higher at times
face mounted to the test apparatus. The heat flux greater than 1 h. The ASTM E119 furnace expo-
to actual test specimens could be different sure is measured using shielded thermocouples,
depending on specimen thermal properties, the whereas the ISO 834 furnace exposure is
occurrence of sample ignition and burning, as measured using sheathed thermocouples.
well as other factors. Though the time-temperature curves in these
tests are similar, the actual heat flux exposure
early in the ASTM E119 fire exposure is more
Fire Resistance Tests severe due to the type of thermocouples used to
control the furnace [85, 86]. The European stan-
Several furnace fire exposures are used through- dard EN1363-1 [87] uses the ISO 834 time-
out the world to evaluate the fire resistance of temperature curve, but the furnace is controlled
products. These fire exposures have peak using plate thermometers, which provide a more
temperatures ranging from 1050 C to 1350 C severe exposure compared with ISO 834
after a 3-h exposure (Fig. 25.56). The type of thermocouples for the test duration [88, 89]. Com-
exposure used depends on the end-use applica- pared with ASTM E119 shielded thermocouples,
tion of the product. Tunnel and offshore oil rig Sultan [90] measured that plate thermometers
applications have the highest temperature, most resulted in a slightly less severe exposure during
severe fire exposures, whereas less severe the first 10 min of the test, but thereafter the thermal
exposures are used for different building exposures were the same.
applications. The total heat flux measured in an ASTM
The ASTM E119 [83] and ISO 834 [84] time- E119 furnace test is provided in Fig. 25.57 for a
temperature curves are perhaps the most common wall and floor furnace. Total heat fluxes were
furnace exposures used in fire resistance testing. measured using a water-cooled Gardon gauge.
790 B.Y. Lattimer
Fig. 25.56 Furnace time- 1400

temperature exposure
curves 1200
Furnace temperature (°C)

1000
UL 1709
EN1363-2 HC
800
Modified HC
RABT-ZTV(train)
600 RABT-ZTV(car)
RWS
ASTM E119
400 ISO 834
200
0
0 1 2 3 4
Time (hr)

measured during ASTM
E119 furnace exposure in 160
floor and wall furnaces. 140
Blackbody heat flux was
Heat flux (kW/m )
calculated from the ASTM 120

2
E119 furnace temperature

curve 100
80
60
Measured-floor
40 Measured-wall
20 Blackbody
0
0 20 40 60 80 100 120 140
Time (min)
In this test, gaseous fuel was used and the tem- hydrocarbon pool fire exposure and the
perature was controlled with ASTM E119 EN1363-2 [93] hydrocarbon curve (HC) are typ-
shielded thermocouples [91]. The wall furnace ically used for offshore oil platform applications,
was lined with ceramic fiber while the floor fur- whereas the other higher-temperature curves are
nace was lined with brick. The same furnace used to represent a large hydrocarbon fire inside a
controlled with a plate thermometer provided tunnel.
similar heat flux levels at times after 10 min. The UL 1709 and EN1363-2 [93] both have a
Also provided in the plot is the blackbody heat maximum gas temperature of 1100 C; however,
flux based on the furnace temperatures specified the UL 1709 exposure reaches 1100 C faster
in ASTM E119. As seen in the figure, the black- than does the EN1363-2 exposure (in 5 min ver-
body heat flux is similar to heat fluxes measured sus after 25 min, respectively). Unique to this fire
in the furnace except during the initial 10 min. exposure curve, the UL 1709 fire exposure also
The higher temperature fire exposure curves has a heat flux requirement. During a calibration
in Fig. 25.56 are used to evaluate products used test with a UL 1709 exposure, the heat flux as
in petrochemical, offshore oil platform, and measured from a water-cooled heat flux gauge
some tunnel applications. The UL 1709 [92] mounted to a calibration specimen must be
www.ebook777.com
204
16 kW/m2 while the furnace temperature Runehamar test series, where temperatures rang-
is 1093
111 C. This heat flux is approxi- ing from 1280 C to 1365 C were measured [48].
mately equal to the blackbody heat flux at the
furnace temperature (i.e., 1093 C results in a
blackbody flux of 197 kW/m2). ASTM E84 Tunnel Test
The curves for tunnel applications have peak
temperatures that range from 1200 to 1350 C. The ASTM E84 test is a “tunnel” test that
The RABT-ZTV curves were developed in provides flame spread and smoke production
Germany to represent different vehicle fires in data from wall and ceiling lining materials. The
tunnels. These curves reach a peak temperature test chamber is approximately 18 in. (0.46 m)
of 1200 C in 5 min and remain at that tempera- wide, 12 in. (0.30 m) high, and 25 ft (7.63 m)
ture for 30–60 min. Thereafter, the temperatures long, with a gas burner located at one end and
decrease linearly with time to ambient conditions exhaust ducting located at the other. The test
after 2.5–3.0 h. Estimated peak heat fluxes, as the material is oriented on the “ceiling” of the tunnel
blackbody flux using the peak furnace tempera- by attaching a 24-ft (7.32-m) long sample of
ture, in these tests are 267 kW/m2. A modified the test material to the underside of the remov-
version of the EN1363-2 HC curve has been used able lid of the test chamber. A flow of 240 ft/min
in France to represent fires in tunnels. The (1.22 m/s) is established through the test cham-
Modified HC curve peaks at 1300 C instead of ber. The initiating fire is an 88 kW gas burner
1100 C. Estimated peak heat flux in this test, located at one end of the sample. The flames
based on the blackbody flux using the peak fur- from the two burner pipes impinge on the sample
nace temperature, is 347 kW/m2. The RWS fire at two off-center locations, producing a flame
curve was developed by the Rijkswaterstaat, Min- that flows 1.2 m down the sample.
istry of Transport, in the Netherlands based on Parker [94] measured heat fluxes from the
results from testing conducted by TNO in the initiating fire with a noncombustible ceiling in
Netherlands. The RWS curve peaks at a tempera- the tunnel. The highest heat fluxes were measured
ture of 1350 C, which is the highest of all time- where the burner flames impinge on the ceiling.
temperature curves. Estimated peak heat flux in A plot of heat flux, measured using a water-
this test, based on the blackbody flux using the cooled heat flux gauge, at this location is shown
peak furnace temperature, is 393 kW/m2. The in Fig. 25.58. The heat flux during the initial
potential for these temperatures in tunnel fires 2 min of the test was 20–30 kW/m2. By 4 min,
was validated through vehicle testing in the the heat flux increased to 50–60 kW/m2 where it
Fig. 25.58 Heat flux at the 70

ASTM E84 burner
impingement point
60
Heat flux (kW/m2)
50
40
30
20
10
0 2 4 6 8 10 12
Time (min)
792 B.Y. Lattimer
Fig. 25.59 Calculated 50

incident heat flux along the
length of the ASTM E84
tunnel 40
Heat flux (kW/m2)

30
20
10
0
0 2 4 6 8
Distance along tunnel (m)
remained for the duration of the test (10 min). The period are shown in Fig. 25.60. The peak heat
increase in heat flux with time was attributed to flux was measured to be 49 kW/m2 approxi-
reradation from the tunnel surfaces. mately 1.0 m down the tunnel. The heat fluxes
Incident heat fluxes along the center of the are similar to those determined by Parker [94],
tunnel length were calculated using surface tem- except the Gandhi et al. [95] measured the peak
perature measurements. Figure 25.59 provides the heat flux location 0.5 m further down the tunnel.
heat fluxes after a 20-min exposure, which is Gandhi et al. [95] stated that this difference may
10 min longer than the actual test. Heat fluxes be due to using actual heat flux gauges instead of
near the burner are approximately 40 kW/m2 and using an inverse method as well as the sample
then decrease rapidly with distance along the tun- location differences. Transient heat fluxes
nel. In a test with flames along the entire length of measured by Gandhi et al. [95] determined that
the tunnel, heat fluxes were calculated to be the heat fluxes increase during the test by approx-
70 kW/m2 at 2.0 m from the burner and 30 kW/ imately 10 kW/m2 at locations 0.13–1.65 m
m2 at the end of the tunnel 7.3 m from the burner. along the tunnel.
A UL 1666 test is used to evaluate cables used
in high rise buildings installed in riser shafts or
Cable Tests floor-to-floor installations. The sample is in a
vertical orientation in this test adjacent to a dif-
Gandhi et al. [95] measured heat fluxes due to the fusion burner. The exposure fire in this test
exposure fire in three different standard cable produces a peak heat flux of 43 kW/m2 0.30
tests : UL 910, UL 1666, and UL 1685. Heat above the burner and decays to 6 kW/m2 by
fluxes were measured using water-cooled Gardon 1.5 m above the burner.
gauges. A UL 1685 test is performed to evaluate
The UL 910 test is conducted in the ASTM cables used in applications other than air
E84 tunnel apparatus to evaluate low power handling or floor-to-floor. The sample is located
cables without conduit in air handling spaces. in a vertical orientation during the test with a
The sample is in a horizontal orientation for this propane gas-air premix burner impinging on the
test with the flame impinging on the underside of bottom of the cables. The exposure fire produced
the cables. Average heat flux measurements peak heat fluxes of 46 kW/m2 at the burner ele-
along the length of the ASTM E84 tunnel vation and decayed to 2 kW/m2 by 1.5 m above
where the cable would be located over the test the burner.
www.ebook777.com
Fig. 25.60 Test average 60

heat fluxes measured in the
UL 910 experiment by ASTM E84
Gandhi et al. [95] 50 UL 910
compared with the ASTM
E84 measurements made
Heat Flux (kW/m2)

by Parker [94] 40
30
20
10
0
0 2 4 6 8
Distance Along Tunnel (m)
Effects of Other Variables Nomenclature
The environment in which a fire is burning can a spacing between parallel walls (m)
affect the heat flux levels incident on the surface. Cp specific heat capacity of air at 300 K
Studies have been conducted by Atreya and (0.998 kJ/[kg-K])
Mekki [96], Santo and Tamanini [97], Mekki d length of single side on L-shape burner,
et al. [98], and Chao and Fernandez-Pello to length of line burner, width of burning
evaluate the impact of oxygen concentration on area on corner wall (m)
the heat fluxes transferred by flames to surfaces. D length of single side of square burner,
In tests with methane fires, Atreya and Mekki diameter (m)
[96] found that flame radiation (and the total g acceleration of gravity (9.81 m/s2)
heat flux to the surface) was increased by increas- H distance between fire and ceiling (m)
ing the oxygen concentration. HB distance between fire and lower flange of
More important for most problems in fire is I-beam (m)
the effect of decreasing the oxygen concentration HC distance between fire and upper flange of
on heat fluxes from the flame. Santo and I-beam (m)
Tamanini [97] found that decreasing the Ho height of room window (m)
surrounding oxygen concentration from 20.9 % h convective heat transfer coefficient
to 18.0 % the radiative flux to an external target (kW/[m2 -K])
was decreased to an external target by 40 %. This k thermal conductivity (kW/m-K)
decrease was attributed to a decrease in lower LB flame tip length along lower flange of
soot concentrations in flames in lower oxygen I-beam (m)
environments. Chao and Fernandez-Pello [99] LC flame tip length along upper flange of
found that this reduction in heat transfer to the I-beam (m)
surface reduces the flame spread rate along com- Lweb flame tip length along center of web on
bustible panels. I-beam (m)
794 B.Y. Lattimer
Lf average flame length (m) web defined using Hweb as length scale
Lf,tip flame tip length (m) inc incident
LH flame extension along ceiling away from m measured
stagnation point (m) max max level
Q fire heat release rate (kW) net net
Q0 fire heat release rate per unit width peak peak
(kW/m) rad radiative
Q* dimensionless parameter, Q*D ¼ rr reradiated
Q
pffiffi 5=2 , with D being length scale
s material surface
ρ C T
1 p 1 gD
r distance from corner or stagnation point
to measurement location (m)
References
00
q heat flux (kW/m2) 1. M. Kokkala, “Heat Transfer to and Ignition of Ceiling
Tf local gas temperature (K) by an Impinging Diffusion Flame,” VTT Research
Tg room gas temperature (K) Report 586, Technical Research Centre of Finland,
Ts material surface temperature (K) Escopo, Finland (1989).
2. M. Kokkala, “Experimental Study of Heat Transfer to
T1 ambient temperature (300 K) Ceiling from an Impinging Diffusion Flame,” Fire
Wo width of room window (m) Safety Science—Proceedings of the 3rd International
w dimensionless distance along ceiling or Symposium, Elsevier Applied Science, New York,
I-beam, pp. 261–270 (1991).
0 0 3. H.Z. You and G.M. Faeth, “Ceiling Heat Transfer
w ¼ r þ HB þ z = LHB þ HB þ z During Fire Plume and Fire Impingement,” Fire and
x horizontal coordinate (m) Materials, 3, 3, pp. 140–147 (1979a).
y horizontal coordinate (m) 4. H.Z. You and G.M. Faeth, “An Investigation of Fire
Impingement on a Horizontal Ceiling,” NBS-GCR-
y0 distance from center of line burner, 79-188, U.S. Department of Commerce,
0
y ¼ 0:5d y ðmÞ Washington, DC (1979b).
Z burner height (m) 5. Wickstrom, U., “Adiabatic Surface Temperature and
the Plate Thermometer for Calculating Heat Transfer
z vertical coordinate (m)
and Controlling Fire Resistance Furnaces,” Fire
z0 virtual source location (m) Safety Science -Proceedings of the Ninth Fire Safety
Science, 2008, pp.1227–1238
Greek Letters 6. Wickstrom, U. Dathinh, D., and McGrattan, K., “Adi-
abatic Surface Temperature for Calculating Heat
Transfer to Fire Exposed Structures,” Proceedings of
ε material surface emissivity () the 11th International Conference on Fire Science
ρ1 ambient density of air (1.2 kg/m3) and Engineering Interflam, 2007
π constant (3.14159) 7. Wickstrom, U., “The Plate Thermometer-A Simple
Instrument for Reaching Harmonized Fire Resistance
σ Stefan-Boltzman constant 5:67 Tests,” Fire Technology 2:195–208
11
2
10 kW= m K 4 8. Duthinh, D., McGrattan, K., and Khaskia, A., (2008)
“Recent Advances in Fire-Structural Analysis,” Fire
Safety Journal 43:161–167
Subscripts 9. L.T. Cowley, “Behaviour of Oil and Gas Fires in the
Presence of Confinement and Obstacles,” Miscella-
cl centerline neous Report TNMR.91.006, Shell Research Limited,
Thornton Research Center, Combustion and Fuels
conv convective Department, Chester, UK (Feb. 1991).
d defined using d as length scale 10. J.J. Gregory, R. Mata, and N.R. Keltner, “Thermal
D defined using D as length scale Measurements in a Series of Large Pool Fires,”
H defined using H as length scale Sandia Report Number SAND85-0196, Sandia
National Laboratories, Albuquerque, NM (1987).
hfg heat flux gauge 11. L.H. Russell and J.A. Canfield, “Experimental
B defined using HB as length scale Measurements of Heat Transfer to a Cylinder
C defined using HC as length scale
www.ebook777.com
Immersed in a Large Aviation Fuel Fire,” Journal of Wall-Ceiling Configurations,” in Proceedings of

Heat Transfer, pp. 397–404 (Aug. 1973). Interflam ‘96 (Franks and Grayson, eds.), Interscience
12. G. Wachtell and J. Langhaar, “Fire Test and Thermal Communications Ltd., London, pp. 179–188 (1996).
Behavior of 150-Ton Lead-Shielded Casks,” DP 1070, 26. M. Kokkala, “Characteristics of a Flame in an Open
Engineering and Equipment, TID-4500, E.I. DuPont Corner of Walls,” in Proceedings from Interflam ‘93,
De Nemours and Co., Wilmington, DE (1966). Interscience Communications, Ltd., London,
13. C. Anderson et al., “Effects of a Fire Environment on pp. 13–24 (1993).
a Rail Tank Car Filled with LPG,” Report 27. T. Ohlemiller, T. Cleary, and J. Shields, “Effect of
No. FRA-OR&D 75–31, U.S. Department of Trans- Ignition Conditions on Upward Flame Spread on a
portation, Federal Railroad Administration, Composite Material in a Corner Configuration,” Fire
Washington, DC (1974). Safety Journal, 31, pp. 331–344 (1998).
14. National Academy of Science, Committee on Hazard- 28. T.J. Ohlemiller and J.R. Shields, “The Effect of Sur-
ous Materials, Division of Chemistry, and Chemical face Coatings on Fire Growth Over Composite
Technology (National Research Council), Pressure- Materials in a Corner Configuration,” Fire Safety
Relieving Systems for Marine Cargo Bulk Liquid Journal, 32, 2, pp. 173–193 (1999b).
Containers, National Academy of Sciences, 29. B.Y. Lattimer and U. Sorathia, “Thermal
Washington, DC (1973). Characteristics of Fires in a Noncombustible Corner,”
15. K. Moodie et al., “Total Pool Fire Engulfment Trials Fire Safety Journal, 38, pp. 709–745 (2003).
on a 5-Tonne LPG Tank,” HSE Internal Report 30. R.B. Williamson, A. Revenaugh, and F.W. Mowrer,
No. IR/L/FR/87/27, Health and Safety Executive, “Ignition Sources in Room Fire Tests and Some
London, UK (1987). Implications for Flame Spread Evaluation,” Fire
16. M. Tunc and J. Venart, “Incident Radiation from an Safety Science—Proceedings of the 3rd International
Engulfing Pool Fire to a Horizontal Cylinder, Part I Symposium, Elsevier Applied Science, New York,
and II,” Fire Safety Journal, 8, pp. 81–95 (1985). pp. 657–666 (1991).
17. W. McLain, “Investigation of the Fire Safety 31. H. Tran and M. Janssens, “Modeling the Burner
Characteristics of Portable Polyethylene Tanks Source Used in the ASTM Room Fire Test,” Journal
Containing Flammable Liquids,” Report No. CG-M- of Fire Protection Engineering, 5, 2, pp. 53–66
1-88, U.S. Coast Guard, Washington, DC (1988). (1993).
18. A. Taylor et al., “Engulfment Fire Tests on Road 32. J.G. Quintiere and T.G. Cleary, “Heat Flux from
Tanker Sections,” Rarde Technical Report 7/75, Con- Flames to Vertical Surfaces,” Fire Technology,
troller HMSO, London (1975). 30, 2, pp. 209–231 (1994).
19. M. Schneider and L. Kent, “Measurement of Gas 33. S.E. Dillon, “Analysis of the ISO 9705 Room/Corner
Velocities and Temperatures in a Large Open Pool Test: Simulations, Correlations and Heat Flux
Fire,” Fire Technology, pp. 51–81 (Feb. 1989). Measurements,” NIST-GCR-98-756, U.S. Department
20. G. Back, C.L. Beyler, P. DiNenno, and P. Tatem, of Commerce, National Institute of Standards and
“Wall Incident Heat Flux Distributions Resulting Technology, Washington, DC (1998).
from an Adjacent Fire,” Fire Safety Science— 34. International Standards Organization, ISO 9705:1993
Proceedings of the 4th International Symposium, (E), International Standard for Fire Tests—Full-Scale
International Association of Fire Safety Science, Room Test for Surface Products, International Orga-
Ottawa, Canada, pp. 241–252 (1994). nization for Standardization (ISO), Geneva,
21. G. Heskestad, “Luminous Heights of Turbulent Dif- Switzerland (1993).
fusion Flames,” Fire Safety Journal, 5, pp. 103–108 35. T. Tanaka, I. Nakaya, and M. Yoshida, “Full Scale
(1983). Experiments for Determining the Burning Conditions
22. D. Gross and J.B. Fang, “The Definition of a Low to Be Applied to Toxicity Tests,” Fire Safety Sci-
Intensity Fire,” in NBS Special Publication 361, Vol- ence—Proceedings of the 1st International Sympo-
ume 1: Performance Concept in Buildings, Proceed- sium, Hemisphere Publishing, Gaithersburg, MD,
ing of the Joint RILEM-ASTM-CIB Symposium, pp. 129–138 (1985).
National Bureau of Standards, Washington, DC, 36. Tofilo, P., Delicatsios, M.A., and Silcock, G.W.H.,
pp. 677–686 (1972). (2005), “Effect of Fuel Sootiness on the Heat
23. T. Mizuno and K. Kawagoe, “Burning Behaviour of Fluxes to the Walls in Enclosure Fires,” Fire Safety
Upholstered Chairs, Part 2: Burning Rate of Chairs in Science-Proceedings of the Eighth International Sym-
Fire Tests,” Fire Science and Technology, 5, 1, posium, Beijing, China, pp. 987–998.
pp. 69–78 (1985). 37. W. Takashi et al., “Flame and Plume Behavior in and
24. M. Daikoku and K. Saito, “A Study of Thermal Near a Corner of Walls,” Fire Safety Science—
Characteristics of Vertical Corner Wall in Room Proceedings of the 5th International Symposium
Fire,” Proceedings of the ASME/JSME, Thermal (Y. Hasemi, ed.), International Association for Fire
Engineering, Book No. H0933C-1995 (L.S. Fletcher Safety Science, Melbourne, Australia, pp. 261–271
and T. Aihara, eds.), pp. 83–90 (1995). (1997).
25. Y. Hasemi, M. Yoshida, S. Takashima, R. Kikuchi, 38. Y. Hasemi, S. Yokobayashi, T. Wakamatsu, and
and Y. Yokobayashi, “Flame Length and Flame Heat A. Ptchelintsev, “Fire Safety of Building Components
Transfer Correlations in Corner-Wall and Corner- Exposed to a Localized Fire—Scope and Experiments
796 B.Y. Lattimer
on Ceiling/Beam System Exposed to a Localized Combustion Insititute, Pittsburgh, PA, pp. 183–192
Fire,” Proceedings from ASIAFLAM, Kowloon, (1975).
Hong Kong, pp. 51–361 (1995). 54. M.A. Delicatsios, “Flame Heights in Turbulent Wall
39. R.L. Alpert, “Convective Heat Transfer in the Fires with Significant Flame Radiation,” Combustion
Impingement Region of a Buoyant Plume,” Science and Technology, 39, pp. 195–214 (1984).
Transactions of ASME, 109, pp. 120–124 (1987). 55. Y. Hasemi, “Experimental Wall Flame Heat Transfer
40. H.Z. You, “An Investigation of Fire-Plume Correlations for the Analysis of Upward Wall Flame
Impingement on a Horizontal Ceiling 2— Spread,” Fire Science and Technology, 4, 2, pp. 75–90
Impingement and Ceiling-Jet Regions,” Fire and (1984).
Materials, 9, 1, pp. 46–56 (1985). 56. H. Mitler, “Predicting the Spread Rates on Vertical
41. L.Y. Cooper, “Heat Transfer from a Buoyant Plume to Surfaces,” in 23rd Symposium (International) on
an Unconfined Ceiling,” ASME Journal of Heat Combustion, Combustion Institute, Pittsburgh, PA,
Transfer, 104, pp. 446–452 (1982). pp. 1715–1721 (1990).
42. T. Wakamatsu, personal communication (Sept. 1999). 57. C.L. Beyler, S.P. Hunt, N. Iqbal, and F.W. Williams,
43. J. Myllymaki and M. Kokkala, “Thermal Exposure to “A Computer Model of Upward Flame Spread on
a High Welded I-Beam Above a Pool Fire,” First Vertical Surfaces,” in Fire Safety Science—
International Workshop on Structures in Fires, Proceedings of the 5th International Symposium
Copenhagen, pp. 211–226 (2000). (Y. Hasemi, ed.), International Association for Fire
44. P.L. Hinkley, H.G.H. Wraight, and C.R. Theobald, Safety Science, Melbourne, Australia, pp. 297–308
“The Contribution of Flames under Ceilings to Fire (1997).
Spread in Compartments,” Fire Safety Journal, 7, pp. 58. F.W. Williams, C.L. Beyler, S.P. Hunt, and N. Iqbal,
227–242 (1984). “Upward Flame Spread on Vertical Surfaces,” NRL/
45. P.L. Hinkley, H.G.H. Wraight, and C.R. Theobald, MR/6180—97-7908, Navy Technology for Safety and
“The Contribution of Flames under Ceilings to Fire Survivability, Chemistry Division (1997).
Spread in Compartments, Part I: Incombustible 59. Y. Hasemi, “Thermal Modeling of Upward Wall
Ceilings,” Fire Research Note No. 712, Fire Research Flame Spread,” Fire Safety Science—Proceedings of
Stations, Borehamwood, Herts, UK (1968). the 1st International Symposium, Hemisphere Pub-
46. P.L. Hinkley, H.G.H. Wraight, and C.R. Theobald, lishing, Gaithersburg, MD, pp. 87–96 (1986).
“The Contribution of Flames under Ceilings to Fire 60. Y. Hasemi, “Deterministic Properties of Turbulent
Spread in Compartments, Part II: Combustible Ceiling Flames and Implications on Fire Growth,” Interflam
Linings,” Fire Research Note No. 743, Fire Research ’88, John Wiley and Sons, Cambridge, UK, pp. 45–52
Stations, Borehamwood, Herts, UK (1969). (1988).
47. B. Lattimer, J. Beitel, and C. Mealy, “Heat Fluxes to a 61. M. Kokkala, D. Baroudi, and W.J. Parker, “Upward
Corridor Ceiling,” unpublished data (2006). Flame Spread on Wooden Surface Products:
48. A. Lonnermark and H. Ingason, “Fire Spread and Experiments and Numerical Modelling,” Fire Safety
Flame Length in Large-Scale Tunnel Fires,” Fire Science—Proceedings of the Fifth International Sym-
Technology, 42, pp. 283–302 (2006). posium, International Association for Fire Safety Sci-
49. T. Wakamatsu, Y. Hasemi, Y. Yokobayashi, and ence, Melbourne, Australia, pp. 300–320 (1997).
A.V. Ptchelintsev, “Experimental Study on the 62. M. Foley and D.D. Drysdale, “Heat Transfer from
Heating Mechanism of a Steel Beam Under Ceiling Flames Between Vertical Parallel Walls,” Fire Safety
Exposed to a Localized Fire,” in Proceedings from Journal, 24, pp. 53–73 (1995).
Interflam ’96 (Franks and Grayson, eds.), Interscience 63. A.K. Kulkarni, C.I. Kim, and C.H. Kuo, “Heat Flux,
Communications, Ltd., London, pp. 509–518 (1996). Mass Loss Rate and Upward Flame Spread for Burn-
50. T. Ahmad and G.M. Faeth, “Fire Induced Plumes ing Vertical Walls,” NIST-GCR-90-584, U.S. Depart-
Along a Vertical Wall, Part III: The Turbulent ment of Commerce, Washington, DC (1990).
Combusting Plume,” NBS Report for Grant 64. A.K. Kulkarni, C.I. Kim, and C.H. Kuo, “Turbulent
No. 5–9020, U.S. Department of Commerce, Upward Flame Spread for Burning Vertical Walls
Washington, DC (1978). Made of Finite Thickness,” NIST-GCR-91-597, U.S.
51. T. Ahmad and G.M. Faeth, “Turbulent Wall Fires,” in Department of Commerce, Washington, DC (1991).
17th Symposium (International) on Combustion, Com- 65. L. Orloff, A.T. Modak, and R.L. Alpert, “Burning of
bustion Institute, Pittsburgh, PA, pp. 1149–1160 (1979). Large-Scale Vertical Surfaces,” in 16th Symposium
52. J.G. Quintiere, M. Harkelroad, and Y. Hasemi, “Wall (International) on Combustion, Combustion Institute,
Flames and Implications for Upward Flame Spread,” Pittsburgh, PA, pp. 1345–1354 (1977).
AIAA-85-0456, American Institute of Aeronautics and 66. G.H. Markstein and J. de Ris, “Wall-Fire Radiant
Astronautics, Reno, NV (1985). Emission, Part 2: Radiation and Heat Transfer from
53. L. Orloff, J. de Ris, and G.H. Markstein, “Upward Porous-Metal Wall Burner Flames,” in 24th Sympo-
Turbulent Fire Spread and Burning of Fuel Surface,” sium (International) on Combustion, Combustion
in 15th Symposium (International) on Combustion, Institute, Pittsburgh, PA, pp. 1747–1752 (1992).
www.ebook777.com
67. Kahn, M., et al. “Combustion Characteristics of 81. J.J. Beitel and W.R. Evans, “Multi-Story Fire Evalua-
Materials and Generation of Fire Products” SFPE tion Program,” SwRI Project 01–6112, Final Report,
Handbook of Fire Protection Engineering, 5th Volume 1, Southwest Research Institute, San
ed. (M. J. Hurley, ed.), Springer (2015). Antonio, TX, and Society of the Plastics Industry,
68. M.M. Delichatsios, P. Wu, M.A. Delichatsios, Inc., New York (1980).
G.D. Lougheed, G.P. Crampton, C. Qian, H. Ishida, 82. S. Yokoi, “Study on the Prevention of Fire Spread
and K. Saito, “Effect of External Radiant Heat Flux on Caused by Hot Upward Current,” Report No. 34,
Upward Flame Spread: Measurements on Plywood Building Research Institute, Tokyo, Japan (1960).
and Numerical Predictions,” Fire Safety Science— 83. ASTM E119, Standard Test Method for Fire Tests of
Proceedings of the 4th International Symposium, Building Construction and Materials, American Soci-
International Association of Fire Safety Science, ety for Testing and Materials, West Conshohocken, PA.
pp. 421–432 (1994). 84. ISO, ISO 834, Fire-Resistance Tests—Elements of
69. T.J. Ohlemiller and T.G. Cleary, “Upward Flame Building Construction, International Organization
Spread on Composite Materials,” Fire Safety Journal, for Standardization, Geneva, Switzerland (1999).
32, pp. 159–172 (1999a). 85. T. Harmathy, M. Sultan, and J. MacLaurin, “Compar-
70. C. Qian, H. Ishida, and K. Saito, “Upward Flame ison of Severity of Exposure in ASTM E119 and
Spread Along PMMA Vertical Corner Walls, Part II: ISO 834 Fire Resistance Tests,” Journal of Testing
Mechanism of M Shape Pyrolysis Front Formation,” and Evaluation, pp. 371–375 (Nov. 1987). In Hand-
Combustion and Flame, 99, pp. 331–338 (1994a). book of Experimental Mechanics (A.S. Kobayashi,
71. C. Qian and K. Saito, “An Empirical Model for ed.), Society for Experimental Mechanics, Prentice-
Upward Flame Spread over Vertical Flat and Corner Hall, Inc., Englewood Cliffs, NJ (1987).
Walls,” in Fire Safety Sceince—Proceedings from the 86. V. Babrauskas and B. Williamson, “Temperature
5th International Symposium (Y. Hasemi, ed.), Measurement in Fire Test Furnaces,” Fire Technol-
Melbourne, Australia, pp. 285–296 (1994b). ogy, 13, 3, pp. 226–238 (1978).
72. Y. Hasemi, M. Yoshida, Y. Yokobayashi, and 87. EN1363-1, Fire Resistance Tests, Part 1: General
T. Wakamatsu, “Flame Heat Transfer and Concurrent Requirements, European Committee for
Flame Spread in a Ceiling Fire,” in Fire Safety Sci- Standardization (CEN), Brussels, Belgium (1999).
ence—Proceedings from the 5th International Sympo- 88. P. Fromy and M. Curtat, “Application of a Zone
sium (Y. Hasemi, ed.), International Association for Model to the Simulation of Heat Transfer in Fire
Fire Safety Science, Melbourne, Australia, Resistance Furnaces Piloted with Thermocouples or
pp. 379–390 (1997). Plate Thermometers,” in Fire Safety Science—
73. Y. Hasemi, M. Yoshida, and R. Takaike, “Flame Proceedings of the 6th International Symposium,
Length and Flame Heat Transfer Correlations in Ceil- International Association for Fire Safety Science,
ing Fires,” poster at Fire Safety Science—6th Inter- pp. 531–542 (1999).
national Symposium, International Association for 89. P.H. van de Leur and L. Twilt, “Thermal Exposure in
Fire Safety Science, Poitiers, France (1999). Fire Resistance Furnaces,” Fire Safety Science—
74. F. Tamanini, “Calculations and Experiments on the Proceedings of the 6th International Symposium,
Turbulent Burning of Vertical Walls in Single and International Association for Fire Safety Science,
Parallel Configurations,” FMRC J.I.OAOE7.BU-2, pp. 1087–1098 (1999).
FMRC Technical Report, Factory Mutual Research 90. M. Sultan, “Fire Resistance Furnace Temperature
Corporation, Norwood, MA (1979). Measurements: Plate Thermometers vs. Shielded
75. F. Tamanini and A.N. Moussa, “Experiments on the Thermocouples,” Fire Technology, 42, pp. 253–267
Turbulent Burning of Vertical Parallel Walls,” Com- (2006).
bustion Science and Technology, 23, pp. 143–151 91. M. Sultan, N. Benichou, and Y. Byung, “Heat Expo-
(1980). sure in Fire Resistance Furnaces: Full-Scale
76. H. Ingason and J. de Ris, “Flame Heat Transfer in vs. Intermediate-Scale,” Fire and Materials, 27, pp.
Storage Geometries,” Fire Safety Journal, 31, pp. 43–54 (2003).
39–60 (1998). 92. UL 1709, “Rapid Rise Fire Tests of Protection
77. B.Y. Lattimer and H. Sorathia, “Thermal Materials for Structural Steel,” Underwriters
Characteristics of Fires in a Combustible Corner,” Laboratories, Northbrook, IL (1991).
Fire Safety Journal, 38, pp. 747–770 (2003). 93. EN1363-2, Fire Resistance Tests, Part 2: Alternative
78. I. Oleszkiewicz, “Heat Transfer from a Window Fire and Additional Procedures, European Committee
Plume to a Building Façade,” ASME HTD, for Standardization (CEN), Brussels, Belgium
23, pp. 163–170 (1989). (1999).
79. I. Oleszkiewicz, “Fire Exposure to Exterior Walls and 94. W. Parker, “An Investigation of the Fire
Flame Spread on Combustible Cladding,” Fire Environment in the ASTM E84 Tunnel Test,” NBS
Technology, 26, 4, pp. 357–375 (1990). Technical Note 945, U.S. Department of Commerce,
80. P.H. Thomas and M.L. Bullen, “Compartment National Bureau of Standards, Washington, DC
Fires with Non-Cellulosic Fuels,” in 17th Symposium (1977).
(International) on Combustion, Combustion Institute, 95. Gandhi, P., Caudill, L., Hoover, J., and Chapin, T.,
Pittsburgh, PA, pp. 1139–1148 (1979). (1996), “Determination of Fire Exposure Heat Flux in
798 B.Y. Lattimer
Cable Fire Tests,” Fire Safety Science-Proceedings of Non-Charring Solids: An Experimental Investiga-
the Fifth International Symposium, Portier, France, tion,” in 23rd Symposium (International) on Combus-
pp. 141–152. tion, Combustion Institute, Pittsburgh, PA,
96. A. Atreya and K. Mekki, “Heat Transfer During pp. 1701–1707 (1990).
Wind-Aided Flame Spread on a Ceiling Mounted 99. Y.H. Chao and A.C. Fernandez-Pello, “Flame Spread
Sample,” in 24th Symposium (International) on Com- in a Vitiated Concurrent Flow,” Heat Transfer in Fire
bustion, Combustion Institute, Pittsburgh, PA, and Combustion Systems, ASME HTD, 199, pp.
pp. 1677–1684 (1992). 135–142 (1992).
97. G. Santo and F. Tamanini, “Influence of Oxygen
Depletion on the Radiative Properties of PMMA
Flames,” in 18th Symposium (International) on Com- Brian Y. Lattimer is a Professor in Mechanical Engi-
bustion, Combustion Institute, Pittsburgh, PA, neering at Virginia Tech. His research areas include fire
pp. 619–631 (1981). dynamics, heat transfer from fires, and material response
98. K. Mekki, A. Atreya, S. Agrawal, and I. Wichman, to fires.
“Wind-Aided Flame Spread over Charring and
www.ebook777.com
Heat Release Rates

26
Vytenis Babrauskas
HRR ¼ Δhc MLR ð26:1Þ

Introduction
where hc is the effective heat of combustion
Calculations of fire behavior in buildings are not (kJ kg1) and MLR is the mass loss rate
possible unless the heat release rate of the fire is (kg s1). Such an equation implies that HRR
known. This chapter on heat release rates and MLR are simply related by a constant. This
provides both theoretical and empirical informa- is not in general true. Figure 26.1 shows the
tion. The chapter is organized so that theory and results obtained from a test on a 17 mm sample
basic effects are considered first, then a compen- of Western red cedar. It is clear that the effective
dium of product data is provided, which is heat of combustion is not a constant; it is roughly
arranged in alphabetic order. 12 MJ kg1 for the first part of the test, but
increases to around 30 MJ kg1 during the char-
ring period at the end of the test.
Definitions In principle, the effective heat of combustion
can be determined by theory or by testing. In
The essential characteristic that describes quanti- practice, if the effective heat of combustion is
tatively How big is the fire? is the heat release not a constant, then experimental techniques nor-
rate. This is so important that it has been mally involve directly measuring the HRR,
described as the single most important variable rather than using Equation 26.1.
in fire hazard [1]. The heat release rate (HRR) of
a burning item is measured in kilowatts (kW). It
is the rate at which the combustion reactions Measuring the HRR, Full-Scale
produce heat. The term “burning rate” is also
often found. This is a less specific term, and it The simplest case is when full-scale HRR can be
may either denote the HRR or the mass loss rate. directly measured. This can be grouped into two
The latter is measured in units of kg s1. It is best types of techniques:
to reserve ‘burning rate’ for non-quantitative fire • Open-burning HRR calorimeters
descriptions and to use either HRR or mass loss • Room fire tests.
rate, as appropriate. The relationship of these two Open-burning HRR calorimeters were devel-
quantities can be expressed as: oped in the early 1980s at NIST by Babrauskas
and colleagues [2] and at FMRC by Heskestad
[3]. The operating principles of these
calorimeters are described in Chap. 27. Based
V. Babrauskas (*) on this work, a large number of different test
Fire Science and Technology Inc.

800 V. Babrauskas
Fig. 26.1 Effective heat 40

of combustion for 17 mm
thick Western red cedar,
35
tested at an irradiance of
Effective heat of combustion (MJ kg-1)

65 kW m2
30
25
20
15
10
0
0 120 240 360 480 600 720
Time (s)
Baffles
To exhaust
gas cleaning
Gas measuring
instrumentation
in hood Test specimen
Variable
on load cell
Front view Side view
Fig. 26.2 NORDTEST NT FIRE 032 calorimeter
standards have been issued, for example [4–8]. A normal-sized door opening) then actual room
discussion of a number of other standards can be HRR values post-flashover can be drastically
found [9]. different from their open-burning rates. This is
The NORDTEST furniture calorimeter [7] is due primarily to additional radiant heat flux con-
shown in Fig. 26.2. Open-burning HRR tribution from the hot gas layer and the hot room
measurements are simpler to make since a test surfaces, although ventilation effects can also
room does not need to be constructed. The HRR play a role.
within a room and under open conditions are, For upholstered chairs, extensive studies have
clearly, identical at very low HRR. What shown that room effects are only at the 20 %
happens at higher values of HRR depends on level up to a 1 MW fire [10]. The same study,
the situation at hand. If the fire is so large that however, showed that for mattresses, a room
room flashover can be reached (about presence effect shows up at much lower HRR
1.5–1.75 MW if ventilation is through a single values. For liquid pools, the HRR is strongly
www.ebook777.com
26 Heat Release Rates 801
Gas and smoke instrumentation

in duct
Exhaust duct
400 dia
To exhaust
gas cleaning
Hood 3000 x 3000

Doorway
2400 800 x 2000
3600 Front view

Side view
Fig. 26.3 ISO 9705 room HRR test
affected by the surrounding room [11]. For most Institute [16]. Room test standards include [17,
other commodities, this issue has not been 18] and also [4, 5]. A typical standard room fire
studied. test, ISO 9705 is shown in Fig. 26.3; a similar
The degree by which the room affects the room fire test is ASTM E 2257 [19]. This test
HRR is largely determined by how ‘open’ the equipment is available for commercial testing in
fuel package itself is. A liquid pool on the floor North America, Europe, Asia, and other places.
has a view factor of 0 to itself and 1.0 to the
room. By contrast, the reason that chairs tend to
be little-affected by the room is that the chair Measuring the HRR, Bench-Scale
‘sees’ its own surfaces to a significant extent,
rather than being fully-exposed to the room. To measure the HRR in a bench-scale test is
Some useful error analyses of large open calo- nowadays an easy task. Most commonly, the
rimeter measurements have been reported [12]; a Cone Calorimeter [20] developed at NIST by
theoretical discussion of the ‘ideal’ large scale Babrauskas will be used (Fig. 26.4). These
calorimeter has also been presented [13]. instruments are available at commercial and
Room fire tests should be commissioned when research laboratories worldwide. The
room effects are anticipated to be strong, or when procedures for conducting Cone Calorimeter
a more precise estimate is needed. Apart from tests are described in ASTM E 1354 [21] and
cost, there is a drawback to room fire testing. ISO 5660 [22]. Other HRR calorimeters, such
This is because the as the Ohio State University apparatus or the
HRR measured in a room fire cannot be Factory Mutual Research Corp. Flammability
extrapolated to any rooms with larger Apparatus are also in use at some laboratories.
ventilations. Open-burning HRR data could, by A textbook is available which discusses many
contrast, be applicable to such well-ventilated of the details of HRR measuring technology
rooms. [23]. Thus, the modeler can assume that if at
The development of the modern room HRR least enough material is available to run several
test took place at several institutions, including small samples (100 mm 100 mm, in the case
Fisher and Williamson at the University of of the Cone calorimeter), an empirical HRR
California [14], Lee at NIST [15], and Sundström curve can be obtained by running bench-scale
at the Swedish National Testing and Research tests.
802 V. Babrauskas
Fig. 26.4 The cone Laser photometer beam

calorimeter including temperature measurement
Temperature and differential
pressure measurements taken here
Soot sample tube
Exhaust
Exhaust hood
blower
Soot collection filter Gas samples Cone heater

taken here
Controlled Spark igniter
flow rate
Specimen
Load cell
Vertical orientation
question (and not some possibly very differently

Measuring the HRR, Intermediate- performing ‘similar’ item), there are some addi-
Scale tional concerns. Supposing one finds full-scale
test results on one’s exact commodity, can the
The newest experimental technology for deter- data simply be used unquestioningly? The
mining the HRR is intermediate-scale calorime- answer, of course, is not. There are two main
try. Various earlier efforts have been made, but issues:
the first instrument to receive standards support • The available data may be open-burning calo-
is the ICAL, developed at Weyerhaeuser [24] rimetry data. One must then determine if there
(Fig. 26.5). It has been standardized as ASTM is an enclosure effect to be accounted.
E 1623 [25]. This test method accommodates • The available data may be room fire data, but
1.0 m by 1.0 m specimens, which allows for the test enclosure may not correspond to the
complex or highly non-homogenous room for which modeling is to be done.
constructions to be tested. However, since the The first of these issues was briefly touched on
data are still not of full scale, some additional above already. The availability of quantitative
analysis is needed to be able to utilize the test guidance is lacking. For upholstered chair fires
data in fire modeling. in a room of about the size of the ISO 9705 room,
one can estimate a 20 % augmentation over the
open-burn rates when considering fires in the
Modeling Implications for Using Full- 100–1000 kW range. For mattresses, the effect
Scale HRR Data is large and without adequate guidance. For liq-
uid pools, a pool sub-model must be specifically
If access is available to full-scale HRR data, then present in the fire model used, since no simple
the task of defining the fire is on a solid basis. approximation is adequate. For wood cribs, there
Even here, however, there are a number of are formulas for guidance [26], although of
problems and caveats. Apart from the obvious course wood cribs are hardly a feature of most
issue that the available full-scale data must be real fires. For other combustibles, neither data
known to describe the specific fuel source in nor guidance is available.
www.ebook777.com
Fig. 26.5 Intermediate

scale (ICAL) calorimeter
A very similar problem is faced when the nonetheless substances of similar heat of com-
modeler has available full-scale HRR data, but bustion and state of aggregation can readily be
the test was run in a room of rather different size found in shops, storage rooms, and various other
or ventilation conditions than is the intended places in diverse building types. The test data
application. Only two studies on this topic have showed that the whole-pallet test had to be
been published in the literature. Kokkala and extinguished at about 4500 kW m2; the fire
colleagues compared [27] some room wall/ceil- was still growing, and its ultimate HRR would
ing linings in a large room to the values obtained have been higher.
in the ISO 9705 room. Also, during the CBUF Growth curves for the FM data listed in
project some furniture fires were done in rooms Table 26.8 are not available; nonetheless the
of two scales [10]. Neither of these studies peak values of roughly 2,000–20,000 kW m2
looked at this issue comprehensively enough to are sobering. Goods of this kind cannot occupy
yield numerical guidance. anywhere close to 100 % of the floor area, of
Some European designers have proposed that course, but even assuming coverage at ¼ to ½,
250 or 500 kW m2 of floor area is an appropri- the actual HRR values are enormous. Now, there
ate peak value of HRR according to which to are clearly occupancies where it is impossible to
design buildings of almost any kind [28]. It is not introduce high fuel loads—swimming facilities
clear how these values were obtained, but one may be an example. But other facilities, even if
must consider whether they are conservative. designed to be spartan in actual use (e.g., cere-
Figure 26.48 gives HRR data for one pallet and monial lobbies) may sustain large fuel loads dur-
half a pallet loads of some elastomer pellets. ing construction, remodeling, expansion, and
While these are ‘industrial’ materials, similar activities.
804 V. Babrauskas

ignition source on the HRR
of PVC foam wall
900
coverings 30
800
Product heat release rate (kW)

25
700
Total heat release (MJ)

600
20
500
15
400
300 10
200
5
100 Peak HRR (kW)
Total heat released (MJ)
0 0
0 20 40 60 80 100 120 140 160
Ignition source HRR (kW)
Effect of Ignition Source the extent of obtaining a V-0 classification in the

on Full-Scale HRR UL 94 [30] test gave no heat release when using a
10 kW burner, but burned well when exposed to a
Full-scale tests for HRR usually do not impose an 30 kW burner. Dembsey [31] conducted room
overall radiant heat flux and are ignited with tests on rooms partially lined with a PVC-foam
localized flame sources. But locally, the heat wall covering. His results are shown in Fig. 26.6.
fluxes from various ignition sources will differ Note that the curve is very steep and could be
both in their magnitude and in the size of the area represented reasonably by a step-function. Apart
subjected to the heat flux. Most plastic from a few examples, this type of data, unfortu-
commodities that do not contain fire retardants nately, is very rarely available for practical
(and are not made from an intrinsically-FR plas- commodities of engineering interest.
tic) can be ignited with very small flame sources,
often no bigger than a paper match. FR
commodities, however, will resist ignition from Effects of Other Variables
small flames, but may be ignited from a large-
flame ignition source. Commonly, such products Some thermoplastic materials have a highly pro-
show an all-or-nothing behavior. That is, ignition nounced tendency to melt and flow. Conse-
sources below a certain size will cause essen- quently, commodities made from these
tially no heat release from the test article, while materials, when burning, will often exhibit object
a larger ignition source may cause a large frac- burning above the floor and an accompanying
tion of, if not the total, combustible mass of the pool fire at the floor, formed by the melt material.
article. Sherratt and Drysdale [32] studied the problem in
For example, it was shown [29] that a televi- intermediate scale, by burning vertical polypro-
sion cabinet made from a plastic fire retarded to pylene sheets above various floor materials.
www.ebook777.com
Major differences were found both in the peak ejection at the surface. Oxygen may or may not
HRR and in the time-resolved HRR curve, directly interact with the surface to create a
depending on the floor type. The differences glowing combustion.
were largely attributable to thermal The chemical reactions being undergone are
characteristics (thickness, density, thermal con- commonly several in number and occurring at
ductivity, etc.) of the floor materials. Upholstered different temperature regimes. Finally, the mate-
furniture using plastic foam padding often burns rial may undergo large-scale cracking, buckling,
with a secondary pool fire underneath, however, or sloughing. Each of these physical phenomena
this behavior occurs only in some cases. may significantly affect the rate of specimen
decomposition. From even this very brief
description, it is clear that computing the pyroly-
Modeling with Bench-Scale HRR Data sis of a material may be a difficult task. Thus,
today for any fire hazard analysis purposes, HRR
If full-scale data on HRR are available, then is invariably measured, rather than being
these are simply used in the fire model. In many computed from more fundamental theory.
cases, however, such data are not available, often Readers wishing to look more closely at the
due to cost of testing or unavailability of large type of modeling needed to represent the pyroly-
size specimens. In such cases, it is desirable to be sis process can refer to the dissertation of Parker
able to use bench-scale data, denoted as and [33] as a good example of how charring materials
measured in units of kW m2. With the bench- need to be treated. Some half-dozen other
scale HRR, there are two main questions: (1) can dissertations have been written on the same
it be predicted from some more fundamental topic. Melting type materials have proven to be
measurements? and (2) how can the full-scale even more interesting as a subject of advanced
HRR be predicted from the bench-scale HRR? research. Several hundred of papers have been
published on various aspects of modeling the
pyrolysis behavior of just one common material,
Predicting Bench-Scale HRR from poly(methylmethacrylate). References [34–40]
Fundamental Considerations can provide an introduction to this research.
The former question has been of considerable

interest to fire researchers, but practical engineer-
Predicting Full-Scale HRR from
ing methods are not yet at hand. This task is often
Bench-Scale Data: Overview
described as creating a ‘pyrolysis model,’ since
the degradation of a material when it is exposed
Prediction of full-scale HRR is probably the
to heat is known as pyrolysis. When a material
single most important engineering issue in
heats up, degrades, ignites, and burns, some very
successful modeling of fires. Schematically, we
complicated physical and chemical phenomena
may write that:
are taking place. In addition to a change of phase,
there is often flow of moisture simultaneously ð
with heat flow. q_ ¼ q_ 00 dA ð26:2Þ
The material may undergo several different
types of phase changes during the decomposition This representation does not fully reveal the
process, each accompanied by changes in density difficulties involved. More explicitly,
and porosity. Bubbles may be created within the ð
bulk of the material and migrate to the surface. q_ ðtÞ ¼ q_ 00 ðt; x; y; zÞdAðtÞ ð26:3Þ
These may be accompanied by molten flow
806 V. Babrauskas
This makes more clear that the instantaneous Predicting Full-Scale HRR from Bench-
per-unit-area HRR is a function of time and Scale Data: The Role of Irradiance
also of the location of the burning element. The
instantaneous burning area, A(t), is also a func- Engineering variables such as HRR, ignitability,
tion of time. In addition, while we have not flame spread, etc. are sometimes viewed as mate-
00
written this explicitly, q_ ðtÞ depends on the rial fire properties. This is a useful view, but it
heating boundary conditions to the element. must be kept in mind that such ‘properties’ are
This quantity usually identified as the heat flux not solely properties defined by the physical/
or irradiance incident upon the element. The chemical nature of the substance. Instead, they
latter term is commonly used since in full-scale are also determined by the boundary conditions
fires the heating is dominated by the radiant of exposure. The boundary conditions can be
component. divided into two types: (1) intended, and (2) unin-
By examining the nature of dA(t), we can tended. The intended boundary conditions
also identify the role of flame spread in include irradiance (since the heat fluxes in room
characterizing the HRR of full-scale fires. A fires are dominated by the radiant component, the
bench-scale HRR test specimen is usually terms irradiance and imposed heat flux are used
ignited nearly-instantaneously over its entire interchangeably) and thickness. Unintended
surface. Full-scale fire, by contrast, nearly boundary conditions, sometimes known as appa-
always exhibit finite spread rates. The flame ratus-dependencies, include such factors as edge
spread velocity in a full-scale fire can be effects, perturbations due to non-uniform
identified with the movement of the boundaries heating, drafts and uncontrolled air velocities,
of the flame-covered area dA(t). Flame spread etc. The latter are usually small if a well-
may occur in several directions over walls, designed test apparatus was used for measuring
ceilings, floors, and over individual surfaces of the response of the specimen.
discrete commodities burning in a space. Con- The most significant intended boundary con-
sequently, it can be seen that tracking flame dition is the heat flux imposed on the specimen.
spread and dA(t) is a major undertaking. This This variable is crucial and no reduced-scale
task, by its nature, is incompatible with zone- HRR results have meaning without knowing the
type of fire models, since it presumes that a irradiance. A test apparatus can impose a very
mechanism is in place to track very small sur- wide range of specimen irradiances. For exam-
face elements. Such mode line is variable with ple, the Cone Calorimeter is capable of
CFO models [41] the quality of production is irradiances from zero to 100 kW m2. For the
dependent on fuel type and the user needs to user of the data, the crucial question becomes
verify the permanent details. what irradiance to select when requesting a test.
Our approach will have to be restricted to There are no simple answers to this, but we
identifying some of the attempts which have summarize here the main conclusions of an
been made to simplify the problem in order to extensive study [42].
make it tractable for zone modeling. The major consideration in the selection of the
Simplifications are not yet possible for the ‘gen- test irradiance must come from a knowledge of
eral’ case. Instead, we must examine specific heat fluxes associated with real fires. In theory,
combustibles, for which appropriate flame spread this could range from zero to an upper value

representations have been established. This is which would be εσ T 4f T 4o , where ε ¼ emis-
illustrated in a number of the sections below.
sivity (–), σ ¼ Stefan-Boltzmann constant (5.67
Before we do this, however, it is important to
1011 kW m2 K4), Tf ¼ flame temperature
examine in more detail some of the variables
(K), and To ¼ ambient temperature (K). But the
which influence the HRR.
www.ebook777.com
ε 1 for larger flames, and the ambient temper- For porous square-faced gas burners, the wall
ature contribution is insignificant, since heat flux was found to depend on the burner
To << Tf. face size [46]. In some cases, fluxes up to
The adiabatic flame temperature for most 65–80 kW m2 were noted, although for most
organic fuels [43] is approximately 2300 K. cases fluxes of 30–50 kW m2 are considered
This would give a maximum irradiance limit of appropriate [47].
some 1500 kW m2. This limiting value is, of At the extreme, ignition can occur due to a
course, nearly 10 times the actual maximum that high velocity jet, such as from a failure on an
is found in building fires of normal types. Thus, it oil-drilling rig. There, heat fluxes in the vicinity
is evident that the theoretical bounds to possible of 150–300 kW m2 have been observed [48].
heat fluxes do not offer any guidance for testing. Such situations, however, are very specialized.
Instead, it is necessary to look at experimental For ignitions from small wood cribs or other
data on heat fluxes found in actual building fires. solid-fuel ignition sources, it can be estimated
We divide this into several types of building fires that the heat flux to adjacent objects in the same
to be examined. More detailed data on heat fluxes 35 kW m2 range as for small flaming sources.
from various objects and ignition sources is The picture is more complicated, however, for
contained in the Ignition Handbook [44]. the heat flux from these sources to the object
underneath. These heat fluxes may be much
(a) Heat fluxes in the vicinity of ignition higher [42], but they are highly non-uniform
sources and difficult to model.
First, we must be clear by what is meant by ‘igni-
tion source.’ The innate definition of the term does (b) Heat fluxes in preflashover room fires
not have limits—a burning building can be the After ignition, the combustibles in a room can be
ignition source to its neighboring building, as can considered to be exposed to preflashover
a fire bomb. For discussion here, however, ignition conditions. Heat fluxes occurring in preflashover
sources can be limited to those that are small with room fires will vary widely. Away from the
respect to a fully-developed room fire. Since the initial source of fire there will be essentially no
latter will be in the range of over 1 MW, the range heating at all. Near a small initial fire source, heat
of fires considered to be ignition sources might be fluxes of the sort described in the preceding
taken as < ca. 300 kW. section will be seen. With increasing fire spread
A NIST study examined various ignition and involvement, a hot gas layer will build
sources, ranging from 5 W to over 300 kW up below the ceiling. The heat fluxes will be
[45]. The sources included both realistic igniting significantly hotter within this layer than in
objects (cigarettes, matches, burning paper lunch lower spaces. Söderbom [49] found values
bags, etc.) and schematic ones (small gas burners typically < 45 kW m2 at the center of the ceil-
and wood cribs). It was found that, as the power ing during preflashover fires. The value at
output of the ignition source increased, the peak the floor level is, of course, always < 20 kW m2
heat flux generally did not increase. Instead, only prior to flashover, since attaining 20 kW m2 at
the area covered by the peak heat flux progres- floor level is one definition of flashover
sively increased. For flames ranging from a [50]. Since there is surprisingly little general
0.3 kW Bunsen-type burner to a 50 kW waste- guidance on this point, the user will have
basket, the peak fluxes were remarkably constant to make some assumptions or ad hoc
at 30–40 kW m2. Thus, for HRR from objects calculations.
being ignited with a small ignition source, a test
irradiance of 35 kW m2 can be selected. There (c) Heat fluxes on burning walls
are some unusual sources having a much higher Heat fluxes from burning items of larger types
flux, and these are discussed elsewhere [42]. have, in general, not been studied in enough
For larger burners, such as used in room fire detail to be systematically known. The notable
tests, higher heat fluxes may have to be assumed. exception is for upward flame spread on vertical
808 V. Babrauskas
surfaces. For this configuration, a number of One might reasonably conclude that a heat
studies have explored the heat fluxes from the flux of ca. 150 kW m2 would be needed to
flame to the yet-unignited portion of the surface. properly represent the environment of the post-
Hasemi studied this problem in detail [51] and flashover room fire. Today’s bench-scale HRR
provided correlations. For his experiments, peak apparatuses, however, can only go to about
values of ca. 25 kW m2 were seen for the region 100 kW m2 or less. Interestingly, the inability
downstream of the ignited area, but before the tip to realistically create the heat fluxes of the post-
of the flames; beyond the flame tip, fluxes were flashover fire has not been seen to be a problem in
no longer constant, but dropped off further down- fire testing. Often, the situation is avoided in its
stream. Additional similar data have also been entirety by assuming that the maximum burning
presented in a summary form [52]. Work by rate that will occur within the room is consistent
Kulkarni and co-workers has enlarged the diver- with the available oxygen supply [55]. Nonethe-
sity of material types to have been studied less, if for more detailed fire modeling the HRR
[53]. The value of 25 kW m2 is seen from of individual items in the post-flashover fire
these more extensive studies to be the lower would be required, such high heat flux values
bound of where data are clustered—most of the would be required.
data are in the interval from 25 to 45 kW m2.
Thus, a value of 35 kW m2 might better capture
the mean behavior. The Dependence of the HRR
A 35 kW m2 heat flux, then, can be used to on the Heat Flux
characterize the peak level of heating to a vertical
surface element from its own upstream flame, In the simplest case, the relationship of the HRR
just prior to its ignition. This value will need to to the irradiance is very simple, as shown in
be increased if the material is so situated as to be Fig. 26.7. Here, we see that the HRR depends
in a hot gas layer that is accumulating in the in a linear manner on the irradiance. The curve
upper reaches of the room. Apart from the data does not pass through the origin due to the exis-
of Söderbom, discussed above, this additional tence of flame flux. The total heat flux seen by the
heating has not been studied in detail. specimen can be viewed as comprised of two
components: the external irradiance, and the
(d) Heat fluxes in post-flashover room fires flux from its own flame. Only if the flame flux
The maximum temperatures actually seen in were zero would the curve pass through the ori-
post-flashover room fires are ca. 1100 C. A gin. Otherwise, the x-axis intercept is equal to
perfect black-body radiator at that temperature (minus) the flame flux.
would produce heat fluxes of approximately Flame flux is very difficult to measure experi-
200 kW m2. Actual heat fluxes measured in mentally, as decomposing materials tend to foul
post-flashover room fires can come close to this the instrumentation and invalidate the readings.
value, but are usually somewhat lower. For A value of ca. 35 kW m2 has been reported for
instance, examining the extensive room burn the flame flux of PMMA burned in the horizontal
data of Fang [54], one finds the following ranges orientation in the Cone Calorimeter [56]. In
of experimental results shown in Table 26.1. another study, estimates of flame flux were
made for several plastics burned in a similar
manner [57]. These showed 30, 25, and
Table 26.1 Heat fluxes measured in postflashover room
fires
14 kW m2, respectively, for nylon, polyethyl-
ene, and polypropylene. The furniture research
Heat flux (kW m2) program CBUF [10] determined that the flame
Ceiling Walls Floor fluxes in the Cone Calorimeter associated
Maximum 106–176 116–229 119–143 with fabric/foam composites are in the range
Average 68–147 91–194 –
20–25 kW m2. Finally, some data are
www.ebook777.com
Fig. 26.7 The simplest

form of HRR dependence
on irradiance
available [58] for liquids in containers of similar of fire-retardant grades of polyurethane foam are
size as a Cone Calorimeter specimen holder. shown in Fig. 26.9. Of the five formulations
Flame fluxes of about 10–15 kW m2 are seen shown, three show somewhat linear behavior,
for alcohols and about 15–20 kW m2 for some whereas two clearly do not. For most categories
hydrocarbons (heptane, methylmethacrylate, tol- of specimens, however, substantially linear
uene, styrene). behavior can be seen.
The value appears to depend only slightly on
the chemical nature of the fuel. Gore et al. [58]
specifically determined that this value does Predicting Full-Scale HRR from
not increase with increasing fuel sooting Bench-Scale Data: The Effect
tendencies. All of the above data refer specifi- of Thickness
cally to the horizontal specimen orientation.
There is very little data for the vertical orienta- The same material may be used in different
tion, although Janssens deduces that for wood applications in varying thicknesses. Thickness
products the vertical-orientation flame flux is does affect the HRR response. In general, a thin
ca. 10–15 kW m2, of which only about material will show a spike of HRR, whereas a
1 kW m2 is due to radiation [59]. thick product will commonly (but not always)
With regards to linearity, the following very show some quasi-steady period of burning.
broad generalization can be made: for many This variable has not been extensively explored,
products, over a substantial heat flux range, the and there is not much guidance available.
HRR is linearly proportional to the heat flux. Figure 26.10 shows results from Paul [62] on a
This generalization, however, will be seen to thermoplastic, PMMA. This illustrates that near-
have only limited utility, since it is rarely steady burning behavior can be seen when the
known a priori whether or not it will be obeyed. thickness approaches ca. 20 mm. Some similar
Furthermore, there is a distinct tendency for data on polyethylene [63] have been published,
most materials and products to deviate from but the maximum thickness specimens examined
linearity at very high and at very low heat in that study, 10 mm thick, did not reach steady
fluxes. burning.
This behavior is best illustrated by an exam- Data for medium-density wood fiberboard
ple. Some data obtained by Sorathia and obtained by Tsantaridis [64] are shown in
co-workers [60] on advanced composites are Fig. 26.11. If tested over the standard substrate
shown in Fig. 26.8. It is clear that the results are (lightweight mineral fiber blanket), wood-family
somewhat linear, but not precisely so. Some old, materials show a second HRR peak which
but still suggestive data were obtained in the corresponds to the accelerated burning when the
1970s by Parker [61]. His results for a number specimen becomes nearly burned through.
810 V. Babrauskas
350
GL/BMI
GL/EP
300
GL/PH
GL/PMR
250
GL/PPS
Peak HRR (kW m-2)
GL/VE
200 GR/BMI
GR/EP
150 GR/PAS
GR/PEEK
100 GR/PH
50
0
0 25 50 75 100
Irradiance (kW m-2)
Fig. 26.8 Response to irradiance of some advanced polyaryl sulfone, PEEK poly(ether ether ketone), PH
composite materials. Reinforcements: GL glass, GR phenolic, PMR monomer-reactant polyimide, PPS
graphite. Resins: BMI bismaleimide, EP epoxy, PAS polyphenylene sulfide, VE vinyl ester
Fig. 26.9 HRR of FR 250

polyurethane foams, as
measured in the NBS I
calorimeter by Parker
200
Q
HRR (kW m–2)
150
AC
Z
100
Q+Z
50 C
0
0 10 20 30 40 50 60
Irradiance (kW m–2)
For foams, by contrast, no reasonable amount structure and become a thin liquid film. This
of thickness will normally show steady-state occurs before ignition takes place. Thus, after
burning. Of special interest are polystyrene ignition what is burning is a thin coating on
foams. These are normally very low density whatever was the substrate. This is the reason
foams of around 16 kg m3. When exposed to why the HRR of PS foams tends to be so
heat, PS foams tend to collapse their cell apparatus-dependent that it is hard to discern
www.ebook777.com
1200
Heat release rate (kW.m−2)
900
600
300
1.5 3 6 10 12.5 20 25 mm
0
0 100 200 300 400
Time (s)
Fig. 26.10 Effect of thickness on the HRR for PMMA (heat flux ¼ 35 kW m2)

thickness on the HRR for
medium-density fiberboard
(heat flux ¼ 50 kW m2)
10 mm
300
Heat release rate (kW m-2)
19 mm
200
28 mm
50 mm
100
0
0 1000 2000 3000 4000 5000
Time (s)
any ‘intrinsic’ response of the material at all: its only in the horizontal orientation. This is because
performance is totally dominated by the (1) many products show serious testing
specimen-holder and edge conditions [65]. difficulties (e.g., melting) when tested in the ver-
tical orientation. (2) Conversely, the vertical ori-
Predicting Full-Scale HRR from entation does not provide ‘a better simulation’ of
Bench-Scale Data: The Effect the burning of vertical objects. This is because
of Orientation there is no direct connection between flame
fluxes in a bench-scale test and in a real-scale
Routine testing in the Cone Calorimeter is fire. The actual fluxes occurring in a real-scale
specified by the ASTM standard to be done fire are determined by many factors, including
812 V. Babrauskas
Fig. 26.12 Orientation 1200

effect on the peak HRR, as
determined from two Cone
Calorimeter round robins
1000
Peak HRR (kW m-2), vertical orientation

800
600
400
200 45 degree line

Data correlation
0
0 200 400 600 800 1000 1200
-2
Peak HRR (kW m ), horizontal orientation
size of room, thickness of hot gas layer, flame For the peak HRR, a least-square regression
spread occurring over other surfaces, etc. None gives that:
of these are subject to the control of the bench-
scale apparatus but, rather, must be specifically q_ pk00 ðV Þ ¼ 0:71 q_ pk00 ðH Þ ð26:4Þ
modeled.
While for the 180 s average HRR, the
Orientation effects will also make a difference
corresponding relation is:
during the bench-scale testing of specimens.
Even though routine testing is done only in the q_ 18000 ðV Þ ¼ 0:72 q_ 18000 ðH Þ ð26:5Þ
horizontal orientation, a small body of work
exists where both orientations were explored. Both can be adequately approximated by the
This is best illustrated by the results of two general relation that:
round robins which were conducted on the q_ 00 ðV Þ ¼ 0:7 q_ 00 ðH Þ ð26:6Þ
Cone Calorimeter, one under the auspices of
ASTM and one under ISO. The data were taken This clearly verifies that the thin, boundary-layer
at two irradiances, 25 and 50 kW m2, and the type flames occurring in the vertical orientation
results are briefly summarized in the Appendix to provide a lower heat flux than the pool-like
ASTM E 1354 [20]. Such results are especially flames in the vertical orientation.
valuable since the values tabulated are the ‘best
estimate’ values and are not subject to the spe-
cific errors of any one particular laboratory. A Predicting Full-Scale HRR from
comparison for the peak HRR is shown in Bench-Scale Data: Other Controlling
Fig. 26.12, while the comparison for the 180 s Variables
average value of HRR is given in Fig. 26.13. In
both cases, the data points plotted represent all of Numerous other variables can, in principle, affect
the data analyzed within the two round robins for the HRR of specimens. This can include local
which horizontal and vertical orientation results velocities, scale and intensity of turbulence, etc.
were obtained on a product. For room fire modeling purposes, such effects
www.ebook777.com
Fig. 26.13 Orientation 700

effect on the 180-s-average
HRR, as determined from
two Cone Calorimeter 600
round robins
HRR (180 s avg.), vertical orientation

500
400
300
200
100
45 degree line
Data correlation
0
0 100 200 300 400 500 600 700
HRR (180 s avg.), horizontal orientation
can be assumed to be small. Two effects which _ ¼ 5:32 þ 3:97x

m} ð26:7Þ
are often of specific interest, however, are scale
and vitiation effects. Scale effects are, in princi- where x is the vertical distance (m). Note that this
ple, normalized out when the per-unit area vari- result implies that there is but little variation for
able is computed. These effects will not be zero, specimens with height < 0.5 m, but significant
however. increases for very large specimens.
One factor affecting them is the flame flux In the case of objects burning in the horizontal
found in the bench-scale test apparatus. This orientation, large ‘pool’ flames surmount the
will have some scale effect. The studies in this specimen. The flux from such flames will vary
area are not extensive. A study using a custom greatly with scale. Guidance provided for
Cone Calorimeter with 200 mm 200 mm spec- estimating burning rates of pools (addressed
imen size tested horizontally found only a very later in this chapter) can be directly applied to
small scale effect, when compared to standard this case.
Cone data [66]. A comparison between the ICAL
and the Cone Calorimeter for a series of wood
products showed that systematic differences HRR for Real Products
were surprisingly small, despite the 10 differ-
ence in linear dimension of the specimens For many objects and commodities, published
[67]. Note, however, that in this case the HRR are not available, thus, laboratory tests
specimens were tested in the vertical orientation. will have to be run if an answer is needed. For
In such orientation, the specimen flames are thin some commodities, however, exemplar data have
and there is little variations with scale. Of addi- been published and are available to the public.
tional guidance is a study by Orloff [68] where a The tabulated test data can be very useful as
vertical 3.56 m high PMMA slab was burned. generic representatives of items constructed of
The mass loss rate, per unit area was found to be: these materials, and with this general geometry.
814 V. Babrauskas
Fig. 26.14 HRR of a 350

small air conditioner with a
plastic housing
300
250
HRR (kW)
200
150
100
50
0
0 300 600 900 1200 1500 1800 2100
Time (s)
Where the analysis is intended to evaluate a Air Conditioners

specific product, that product should be tested
in a suitable calorimeter and the data then used Beard and Goebeldecker [70] tested a small
in the analysis. It must be strongly emphasized European in-room air conditioner 466 406
that in no case should generic database informa- 855 mm high. The unit had an ABS plastic
tion be used when the purpose of the analysis is housing, polystyrene foam inside, and a mass of
to seek regulatory approval for a product or to 35 kg, of which 26 kg remained post-test; the
demonstrate the performance of a specific prod- total HR was 212 MJ (Fig. 26.14).
uct in a court of law. In all such cases, actual
laboratory testing on the item in question must
be done. Audio Equipment
In the case of a few product categories,
methods are available for estimating large-scale EFRA [70] tested two bookshelf-size micro-ste-
HRR on the basis of bench-scale HRR data. The reo systems, each comprising a receiver and a
question then becomes: where can bench-scale pair of stereo speakers. The receiver enclosures
HRR data be found? For a few product were made HIPS plastic, but one system had
categories, some data are provided in the sections fiberboard speaker cabinets (P), while the other
below. For the user interested in a more compre- had HIPS cabinets (G). The systems were both
hensive look at bench-scale HRR data, the text- very small, with the mass before test being only
book Heat Release in Fires [23] and the Cone 4.1 kg (specimen G) and 4.9 kg (specimen P).
Calorimeter Bibliography [69] are good sources. Figure 26.15 shows the HRR results for the two
Also, Chap. 36, “Combustion Characteristics of tests.
Materials and Generation of Fire Products,”
provides some data on pure chemicals.
For convenience, the sections below are Bedding
arranged alphabetically by type of product. How-
ever, many of the ideas are an offshoot of Ohlemiller et al. [71] tested inert beds (twin-size)
pioneering studies on pool fires. Thus, it is with 12 different bedding combinations, with the
recommended that the user first read through peak HRR values found ranging from 38 to
the section on “pools” before progressing to 200 kW. Detailed HRR curves are shown in
other product categories. Fig. 26.16 for one bedding combination. This
www.ebook777.com
Fig. 26.15 HRR values 80

for two small stereo
G
systems
70 P
60
50
HRR (kW)
40
30
20
10
0
0 500 1000 1500 2000
Time (s)
Fig. 26.16 HRR results 250

for two replicate tests for a
Test 14
bedding set on an inert,
twin-size bed Test 16
200
150
HRR (kW)
100
50
0
0 200 400 600 800 1000
Time (s)
combination involved 2 polyester/cotton sheets, fiber filling in for both combinations and were
a mattress pad, a pillow, an acrylic blanket, and a covered with a polyester/cotton pillow case.
medium-weight comforter. The HRR (average Detailed HRR curves were not published for
for the replicates) was exceeded by only one other combinations. The lowest peak HRR
other combination, which gave values about values were for a combination with two sheets
5–10 kW higher. The latter comprised two poly- and a pillow only, which showed 38 and 73 kW
ester/cotton sheets, a mattress pad, a pillow, a for the two replicates.
polyester blanket and a medium-weight com- NRCC [72] conducted four tests on bedding
forter. The pillows were filled with polyester and got peak HRR results up to 388 kW.
816 V. Babrauskas
Fig. 26.17 HRR of 2000

storage units
1800
Metal office storage units,
clear aisle
1600
Metal office storage units,
aisle with fuel

1400
3-shelf metal bookcase with
X-ray films
1200
3-shelf wood bookcase with
1000 files
800
600
400
200
0
0 200 400 600 800 1000 1200
Time (s)
Bookcases, Casegoods and Storage shelf. When fire attacks the front, folders pro-
Units gressively fall out and burn in the aisle. While
well-known, this effect has not been documented
In most cases, for storage furniture the fire hazard with HRR testing.
is created by the contents, not by the furniture
item itself. An exception is modular storage units
made of thermoplastic materials, which tend to Boxes and Packaging
burn very vigorously [73], but quantitative HRR
data have not been published. Storage furniture Full-scale tests were run at Western Fire Center
made of wood or wood covered with thin layers [74] to measure the HRR of fruit/berry baskets
of thermosetting plastic tend to resist ignition (i.e., small plastic containers), packaged in card-
unless filled with combustible contents. Some board shipping cartons, and assembled into pallet
data are illustrated in Fig. 26.17. The metal office loads. In each case, no fruit goods were actually
storage units tests [147] arrangement involved included, the boxing material being packaged as
two tiers of shelving with an 0.76 m aisle in would be delivered from the manufacturer. For
between. Each test contained 480 kg of paper all tests, only a single pallet was used. Identifica-
fuel load in shelving units totaling 1.67 m2 of tion of materials is given in Table 26.2, while
floor area. For the configuration with fuel in the HRR results are given in Fig. 26.18. Southwest
aisle, only 3 kg was placed in the aisle, but this Research Institute [75] tested pallets similar to
extra fuel provided a major difference in fire Sample A, but assembled as a 2 2 2 array of
severity. The data on X-ray film shelves and pallets. This test gave a peak HRR of 8695 kW
wooden bookcase are from Ref. 170. For storage and the results are shown in Fig. 26.19.
of paper files, it is known that the arrangement is
more important than the quantity of fuel. Espe-
cially, storing files in cardboard boxes so that Carpets and Other Floor Coverings
they can exfoliate exacerbates burning. Exfolia-
tion occurs when paper folders are placed paral- Carpets which are in the room of fire origin are
lel, rather than perpendicular to the front of the not likely to contribute significantly to fire
www.ebook777.com
Table 26.2 Packaged fruit cartons tested in shipping pallets
Overall dimensions Mass before Mass after Peak HRR Eff. heat of
Sample of pallet load (m) test (kg) test (kg) (kW) comb. (MJ/kg)
A 0.75 1.14 1.83 393 307 4923 17.3
B 1.02 1.26 1.83 308 222 3553 14.0
C 0.99 1.19 1.87 421 393 3044 12.1
D 1.33 0.80 1.17 430 344 896 11.9
E 1.18 1.07 2.29 461 319 3894 11.0
F 1.00 1.22 2.00 254 192 4280 13.9
Fig. 26.18 HRR of single 5000

pallet-loads of packaged
fruit/berry baskets 4500
A
B
4000 C
D
3500 E
F
3000
HRR (kW)
2500
2000
1500
1000
500
0
0 300 600 900 1200 1500 1800
Time (s)
Fig. 26.19 HRR of a 9000

2 2 2 pallet array
of packaged fruit/berry 8000 A
baskets
7000
6000
HRR (kW)
5000
4000
3000
2000
1000
0
0 300 600 900 1200
Time (s)
818 V. Babrauskas
Fig. 26.20 HRR of 2500

stackable chairs,
Single chair
polypropylene with steel
5 chairs in 1 row
frame, no padding
2000 8 chairs in 4 rows
6 chairs in 1 stack

12 chairs in 2 stacks
1500
1000
500
0
0 500 1000 1500 2000 2500
Time (s)
growth. This has been demonstrated experimen- been measured to produce a peak HRR of
tally [76]. It is also consistent with modeling 3 MW [78]. The test carpet was 80 % acrylic/
considerations: the floor area is convectively 20 % nylon; no other types of carpeting were
cooled and has normally the smallest view factor explored.
to the hot regions, which tend to be in the upper
regions. The same material may be much more
hazardous if installed on wall surfaces, although Chairs, Stackable
it must be pointed out that commercial textile
wall coverings are normally similar, but not iden- Stackable chairs are most commonly used in
tical to carpeting. hotels and banqueting facilities. These chairs
The hazard from floor coverings arises when typically have metal legs and frame and only a
an unsuitable product is used in a corridor, espe- small amount of combustible padding or struc-
cially if this is an escape path. In such situation, tural material. Thus, a single chair can be
very rapid flame spread and high HRR can result expected to represent negligible hazard. How-
due to the fact that the corridor floor covering ever, when not in active use, they are stored in
becomes involved due to a room fire feeding tall piles and many of these piles may be
it. Not only carpeting, but solid materials such aggregated together. The hazard of even a single
as linoleum and wood parquet flooring are also pile of modest height can be notable. Figure 26.20
subject to becoming fully involved down the illustrates some typical data on non-upholstered,
length of a corridor. A recent study has quantified molded chairs [169]. Figure 26.21 illustrates
this behavior and has also provided a predictive some data on lightly-upholstered chairs [79].
method [77]. It is shown that floor coverings with For the latter, the effect of radiant augmentation
a peak HRR of less than 200 kW m2, measured from burning in a corner is also illustrated.
in the Cone Calorimeter under an irradiance of
25 kW m2 tend not to show accelerating flame
spread down a corridor. Clothing Items
Some carpeting materials can present a rapid
fire spread hazard when installed on stairs. A Two men’s jackets (anoraks) were tested by SP
residential carpet installed over a stairway has [10] as potential ignition sources. One was a
www.ebook777.com
Fig. 26.21 HRR of metal- 1000

frame, upholstered stacking
chairs Single chair
4 chairs in 1 stack
800
8 chairs in 1 stack
8 chairs in 1 stack in corner

600
400
200
0
0 500 1000 1500
Time (s)
Fig. 26.22 HRR of men’s 100

jackets (laid loose) Polyester jacket
90 Acrylic jacket
80
70
60
HRR (kW)
50
40
30
20
10
0
0 100 200 300 400 500 600
Time (s)
‘polyester’ jacket with an outer fabric comprising wadding. The total weight was 618 g. The HRR
65/35 cotton/polyester, an inner fabric of 100 % of these jackets are shown in Fig. 26.22. Stroup
polyamide, and a filling of 100 % polyester et al. [80] measured the HRR of racks of men’s
wadding. The total weight was 739 g. The second suits, such as might be found in a retail shop.
jacket tested was an ‘acrylic’ jacket with a fabric Each rack held 48 suits, made of polyester and
of nylon/Taslan and a filling of 100 % acrylic wool and arranged in two rows vertically. The
820 V. Babrauskas
Fig. 26.23 HRR of 500

hanging, cotton shirts
10 shirts
5 shirts
400
300
HRR (kW)
200
100
0
0 50 100 150 200 250 300
Time (s)
Fig. 26.24 HRR of racks 2000

of men’s suits; the rack was Test 1
1.8 m long (arranged as two 1800 Test 2
rows side-by-side) Test 3
1600
1400
1200
HRR (kW)
1000
800
600
400
200
0
0 100 200 300 400 500 600 700 800
Time (s)
results of three replicate tests are shown in authors also tested a single hanging cotton shirt,
Fig. 26.24. There is a lot of scatter, since the which gave a peak HRR of 70 kW, and a pile of
suits fall as they are burning. Japanese results 10 folded shirts, which only showed a peak of
[81] were reported for cotton shirts, hanging on 35 kW. Additional data are given under “Shop
a rack. These results are given in Fig. 26.23. The displays” later in this chapter.
www.ebook777.com
Fig. 26.25 HRR of coffee 45

makers
40
35

30
25
20
15
10
0
0 300 600 900 1200 1500 1800
Time (s)
Coffee Makers Computer mice. Edenburn [84] tested two

brands of computer mice. Both were ignitable
The HRR of several coffee makers weighing by a needle flame and one brand showed a
0.8–1.4 kg is shown in Fig. 26.25 [70, 169]. peak HRR of 3.6 kW and a total heat release
The material for the unit showing the highest of 1.20 MJ. The second brand was tested in
HRR was identified as polypropylene [70], the the Cone Calorimeter with an applied external
others were not specified. heat flux. Using a 25 kW m2 flux, a peak
HRR of 5.0 kW was found and a heat release
of 1.35 MJ; at a 50 kW m2 heat flux, a peak
Computers and Electronic Equipment HRR of 6.1 kW and a heat release of 1.45 MJ
were found.
Here are given results on HRR testing of Computer monitors. Bundy and Ohlemiller
computers and electronic equipment, set up as [83] tested at NIST a series of 480 mm
intended to be used. Additional results are given (19 in.) computer monitors of the CRT type.
under Industrial commodities for packaged Three ignition sources of progressively
goods. greater intensity were used: a needle flame, a
Computer CPUs. Two computer CPUs were burning polystyrene keyboard, and a radiant
tested by SP [82]. One was made by IBM, panel providing a heat flux of 21 kW m2 onto
using a plastic facing rated V0 according to the specimen (Table 26.3). Selected results
UL 94 [30]. This could not be ignited from a are shown in Figs. 26.28, 26.29, 26.30,
small ignition source. The second unit, of HP 26.31, and 26.32.
manufacture, could be ignited by a small igni- Computer printers. Three computer printers
tion source and its HRR is shown in were tested by SP [85]. All were of the per-
Fig. 26.26. sonal type, manufactured by Epson, HP, and
Computer keyboards. Bundy and Ohlemiller Lexmark; the results are given in Fig. 26.33.
[83] tested at NIST three polystyrene com- The printers were tested without paper or
puter keyboards weighing 580 g. These were toner.
ignited with a needle flame and the results are Computer tapes. A test was conducted on a set
shown in Fig. 26.27. of open steel shelves holding 90 computer
822 V. Babrauskas
Fig. 26.26 HRR of 12

computer CPU tested by SP
HP CPU
10
HRR (kW) 6
0
0 300 600 900 1200 1500 1800
Fig. 26.27 HRR of 25

polystyrene computer
keyboards tested at NIST Keyboard 9
Keyboard 24
20 Keyboard 29
15
HRR (kW)
10
0
0 500 1000 1500 2000
Time (s)
tapes [86]. The tapes were 300 mm diameter Racks with computer equipment. Zicherman
and the total mass of 99 kg was distributed on and Stevanovic [87] tested stainless steel
four shelves, two tiers deep. The results are mesh-type racks containing computer and
given in Fig. 26.34. electronic equipment. The rack size was
www.ebook777.com
Table 26.3 Computer monitors tested at NIST

Specimen ID Material UL 94 rating Test # Ignition source CRT Peak HRR (kW) Total mass loss (g)
7 ABS V0 6 Needle flame No DNI
7 ABS V0 11 Needle flame No DNI
7 ABS V0 6a Keyboard No 43.6 797
7 ABS V0 11a Keyboard No 31.3 831
7 ABS V0 27 Keyboard Yes 34.5 830
7 ABS V0 13 Radiant panel No 0.0 24
1 PC V0 7 Needle flame No DNI
1 PC V0 8 Needle flame No DNI
1 PC V0 7a Keyboard No 45.8 768
1 PC V0 8a Keyboard No 120.2 2048
1 PC V0 28 Keyboard Yes 54.7
1 PC V0 31 Keyboard No 54.4 1626
1 PC V0 18 Radiant panel No 124.0 1504
1 PC V0 20 Radiant panel No 117.2 1441
18 HIPS V1 2 Needle flame No DNI
18 HIPS V1 10 Needle flame No DNI
18 HIPS V1 2a Keyboard No 114.5 1483
18 HIPS V1 10a Keyboard No 88.8 1607
18 HIPS V1 25 Keyboard Yes 72.4
18 HIPS V1 16 Radiant panel No 87.7 1267
18 HIPS V1 21 Radiant panel No 94.2 1329
13 PP V2 5 Needle flame No DNI
13 PP V2 12 Needle flame No DNI
13 PP V2 5a Keyboard No 205.1 2469
13 PP V2 12a Keyboard No 198.5 2545
13 PP V2 30 Keyboard Yes 180.0 3303
13 PP V2 17 Radiant panel No 192.6 1776
13 PP V2 22 Radiant panel No 166.2 1849
3 HIPS HB 1 Needle flame No 207.2 2401.0
3 HIPS HB 4 Needle flame No 199.8 2478.0
3 HIPS HB 26 Needle flame Yes 143.8 3309.0
3 HIPS HB 14 Radiant panel No 239.2 2475
3 HIPS HB 19 Radiant panel No 189.8 2413
1.73 m high, 0.92 m wide, and 0.61 m deep, barbeque lighter was used to ignite the com-
with each rack having six shelves. The top puter monitor in Test 1, the stack of paper in
shelf contained a CRT monitor and a personal Test 2, and likewise the paper in Test 3. Test
computer, the next four shelves each 3 differed from the others in that each of the
contained two small data acquisition units top five shelves also contained a 0.60 0.90
(each fully metal-cased), while the bottom cardboard sheet, which was treated with an
shelf held a dot-matrix printer and a 75 mm antistatic treatment. The sheets were located
high stack of computer paper. A keyboard and directly on top of each shelf and underneath
a power strip were hung from the top shelf. the electronic equipment. The HRR results are
Three tests were run on replicate units. A shown in Fig. 26.35. They have an important
824 V. Babrauskas
Fig. 26.28 HRR results of 40

ABS monitors, rated UL
94 V0 Test 11a
35 Test 27
30
25
HRR (kW)
20
15
10
0
0 500 1000 1500 2000
Time (s)

polycarbonate monitors,
rated UL 94 V0 Test 8a
120 Test 28
100
HRR (kW)
80
60
40
20
0
0 500 1000 1500 2000 2500 3000
Time (s)
www.ebook777.com
Fig. 26.30 HRR of high- 140

impact polystyrene (HIPS)
monitors, rated UL 94 V1 Test 2a
120 Test 25
100
HRR (kW)
80
60
40
20
0
0 500 1000 1500 2000 2500 3000
Time (s)
Fig. 26.31 HRR of 250

polypropylene computer
monitors, rated UL 94 V2 Test 5a
Test 30
200
150
HRR (kW)
100
50
0
0 500 1000 1500 2000 2500 3000
Time (s)
826 V. Babrauskas
Fig. 26.32 HRR of high- 250

impact polystyrene (HIPS)
monitors, rated UL 94 HB Test 1
Test 26
200
150
HRR (kW)
100
50
0
0 500 1000 1500 2000 2500 3000
Time (s)
Fig. 26.33 HRR of two,

personal-type printers
100 Epson ink jet
tested at SP
HP 690C
Lexmark Z11
80
HRR (kW)
60
40
20
0
0 300 600 900 1200 1500 1800
Time (s)
instructive value in demonstrating that minor 528 kW (Test 3). Conversely, changing the
changes in the fuel loading or fuel arrange- ignition location had a major effect on the
ment can have drastic influences on the HRR. time of the peak, but essentially no effect on
In this case, introducing the cardboard sheets the HRR peak value (146 kW in Test
raised the peak HRR from 155 kW (Test 2) to 1, 155 kW in Test 2).
www.ebook777.com
Fig. 26.34 HRR of a rack 2000

of computer tapes
1800
1600

1400
1200
1000
800
600
400
200
0
0 300 600 900
Time (s)
Fig. 26.35 HRR of 600

computer equipment racks
Test 1
Test 2
500 Test 3
400
HRR (kW)
300
200
100
0
0 300 600 900 1200 1500 1800
Time (s)
Miscellaneous electronic equipment cabinets. Two series of tests on steel cabinets used for
Babrauskas et al. [88] tested two types of housing nuclear power-plant control electronics
plastic business machine cabinets. The were conducted by VTT [89, 90]. These showed
cabinets were tested as pairs (two identical HRR peaks of 100–200 kW. The authors also
units) and ignited with a 50 kW burner. The proposed computation formulas for predicting
cabinets had 3 mm wall thickness and each the HRR level to cause internal cabinet ‘flash-
pair of cabinets weighed 3.5 kg. The HRR over’ and for burning to reach a ventilation limit
results are shown in Fig. 26.36. [91]. Such computations are based on the
828 V. Babrauskas
Fig. 26.36 HRR of 700

business-machine cabinets
made from polyphenylene
oxide Non-fire-retarded
600
Fire retarded
500

400
300
200
100
0
0 100 200 300 400 500 600
Time (s)
assumption that air flow occurs only through flammability behavior, in their tests with actual
fixed cabinet openings. In such cases, the peak electronic equipment—as opposed to PMMA—
HRR (or, a quasi-steady-state plateau) can be they did not find any instances of such external
computed and actual testing would not be burning.
needed. However, some cabinets may react to
fire by effectively increasing their air inflow
area, e.g., if doors warp open or fall from the Cribs (Regular Arrays of Sticks)
cabinet. Researchers at Institut de Radioprotec-
tion et de Sûreté Nucléaire (IRSN) extended the Cribs here are taken to mean regular, three-
VTT theory and conducted numerous validation dimensional arrays of sticks. Each stick is of a
experiments [92]. They found that, in most cases, square cross-section and of a length much greater
predictions based on ventilation-controlled burn- than its thickness. The sticks are placed in
ing were quite closely borne out by experiments. alternatingly oriented rows, with an air space
In a few cases, fires did not develop sufficiently separating horizontally adjacent sticks. (See
to cause internal flashover, and the theory con- Fig. 26.37). Wood crib burning rates have been
servatively over-predicts the HRR for such studied longer than any other product, with early
instances. With an extremely flammable fuel data available from the 1930s [93]. Different
(PMMA), but one which is probably unrealistic analysis formulas have been presented over the
for actual industrial electronics cabinets, they did years by numerous authors. Here we present a
note that actual HRR can exceed the prediction, method of analysis [26] based largely on the
since some of the pyrolysis gases which lack voluminous experimental data of Nilsson [94]
sufficient oxygen to burn inside the cabinet can on wood cribs and the functional form
leave the cabinet and burn as a fire plume out- suggestions of Yamashika and Kurimoto
side. However, since industrial electronics equip- [95]. The scant available data on plastic cribs
ment is usually selected with at least some are from Harmathy [96] and Quintiere and
attention being paid to avoidance of excessive McCaffrey [97]. The conditions of most interest
www.ebook777.com
Table 26.4 Fuel type versus regression velocity vp for

cribs
Material vp
Wood 2.2 106D0.6
Polymethylmethacrylate 1.4 106D0.6
Thermosetting polyester 3.1 106D0.6
hc (Crib height)
Rigid polyurethane foam 3.8 106D0.6
X
t
m ¼ mo m_ i ðti ÞΔt ð26:9Þ
D (Stick thickness) i
S (Clear spacing) Crib porosity control:

S mo
Fig. 26.37 General arrangement of a wood crib m_ ¼ 4:4 104 ð26:10Þ
hc D
Room ventilation control:
are when cribs are ignited instantaneously, as pffiffiffiffiffi
with the use of a small amount of combustible m_ ¼ 0:12Av hv ð26:11Þ
liquid underneath. The first group of equations The least of Equations 26.8, 26.10, or 26.11 is
below represents this case. There is occasionally to be taken as the governing rate (Equation 26.11
an interest in a crib fire where only one end of a is discussed later in this chapter). Equation 26.8a
crib is ignited, and a slow fire propagation is is necessary instead of the simpler Equation 26.8
seen. An analysis for this situation has also when a switch of burning regime occurs during
been made [98]. A similar analysis is also avail- the course of the fire, e.g., the burning changes
able for the center-ignited, fire-spreading crib from porosity control to fuel surface control at
scenario [99]. some point. This can happen since Equation 26.8
For cribs ignited uniformly overall, the burn- or (26.8a) is a time-dependent expression. Thus,
ing rate can be governed by one of three a crib may start burning under porosity or room
conditions: (1) the natural limit of stick surfaces ventilation-controlled conditions, then later
burning freely; this limit applies to cribs with switch to fuel surface control.
wide inter-stick spacings; (2) the maximum In the above equations, D is the stick thick-
flow rate of air and combustion products through ness, m0 is the crib initial mass, t is the time since
the air holes in the crib; this governs for tightly ignition, hc is crib height, S is the clear spacing
packed cribs; and (3) the maximum oxygen that between sticks, and room ventilation variables
can be supplied to the room; this effect is are Av, the ventilation opening area, and hv, the
discussed separately. The numerical expressions ventilation opening height. The fuel surface
are as follows: regression velocity, vp, depends on the stick
Fuel surface control: thickness and on the fuel type, as shown in
Table 26.4. The experimental data for the plastic
4 2v p t materials are extremely scant, however, so the
m_ ¼ mo v p 1 ð26:8Þ
D D values should be viewed as indicative rather than
quantitative.
or
For the case of the center-ignited crib, the
1=2 burning regimes are divided according to
4 m
m_ ¼ mo v p ð26:8aÞ whether at a particular time the flame spread
D mo
has reached the edge of the crib. This time is
with defined as t0.
830 V. Babrauskas
to ¼ 15:7n ð26:12Þ Curtains

where n ¼ the number of sticks per row. For Thermoplastic curtains often do not sustain any
time t < to, the following relation holds [99]: appreciable burning when ignited by a flame.
v p t2 Instead, a small piece ignites, but it falls off and
m_ ¼ 0:0254 mo ð26:13Þ the rest of the material still in place does not
n2 D
burn. The dropped-down material will usually
For t > to, Equations 26.8 through 26.11 are continue burning, but its HRR will be trivial.
used. The heat release rate is determined from There is no systematic study available that
Equation 26.1. For plastics, the heat of combus- would elucidate under what conditions curtains
tion is commonly fairly constant and can be will burn in place (and release a significant
taken from tabulations or from Cone Calorimeter amount of heat), versus burning only to a trivial
testing. For wood cribs, commonly the heat of extent.
combustion is taken to be 12 103 kJ kg1. Even if curtains ignite and burn in place, the
However, as illustrated in Fig. 26.1, the heat of heat content and HRR are generally moderate,
combustion of wood is a varying function of but curtains can contribute to the severity of a fire
time. A better procedure would be to either pre- by quickly propagating fire over large surfaces.
dict the HRR of wood cribs directly, without Moore has done the most extensive study of
going through Equation 26.1, or else to be able curtains and draperies [100]. His test specimens
to have recourse to a realistic value of Δhc (t). were ignited with a match along the bottom. The
Neither of these possibilities have currently been results are summarized in Table 26.5 and
developed. Fig. 26.38. His results show primarily the effect
of fabric weight. Lightweight fabrics, of weight
Room Fire Effects Experimentally, it has long around 125 kg m2, can show heat release rate
been observed [94] that, unlike a pool fire, which peaks almost as high as heavy ones (around
can burn in a room in a highly fuel-rich manner, a 300 kg m2); however, their potential to ignite
wood crib does not burn more than approxi- surrounding objects is much smaller, as
mately 30–40 % fuel rich. Conditions more fuel demonstrated in Fig. 26.38. These conclusions
rich than that are not sustained, presumably, hold for both thermoplastic and cellulosic
because of the highly vitiated air being supplied materials, but not for constructions using foam
to the crib under those conditions. The stoichio- backings, for which insufficient data were avail-
metric fuel pyrolysis rate can be estimated able. Whether the curtain was in the closed or in
as [11] the open position seemed to make little differ-
pffiffiffiffiffi ence. The reason for the more severe fire perfor-
1
m_ p ðstÞ ¼ 0:5Av hv ð26:14Þ mance of the heavyweight curtains was largely
r
due to their increased burning time, which was
where the stoichiometric air/fuel mass ratio, r, typically about twice that for the lightweight
for wood can be taken as r ¼ 5.7. Comparing, curtains. Additional data on the HRR of curtains
then, the maximum pyrolysis rate given by have been published by VTT [156] and by
Equation 26.11 with the stoichiometric rate SP [101].
given by Equation 26.14, it can be seen that a Yamada et al. [102] conducted full-scale tests
limit of approximately 37 % fuel rich is reached on curtains of 0.9–1.2 m width and 2.0 m length.
when Equation 26.11 becomes the governing They tried 10, 30 and 50 kW square burners and
limit to the burning rate. Similar limits may found that generally at least the 30 kW burner
possibly exist for other classes of combustibles, needed to be used if full flame development was
but experimental data are only available for to be reached. Polyester curtains, both FR and
wood cribs. non-FR, melted and failed to show a sustained
www.ebook777.com
Table 26.5 HRR data for curtains. Nominal curtain size: two curtains each, 2.13 m high by 1.25 m wide. Wall area
covered: 2.13 m high by 1.0 m wide (in closed position)
Number of
Weight Peak wall and ceiling
Type of fiber (g/m2) Configuration HRR (kW) panels igniteda
Cotton 124 Closed 188 1
Cotton 124 Open 157 0
Cotton 260 Open 152 7
Rayon/cotton 126 Closed 214 0
Rayon/cotton 126 Open 176 0
Rayon/cotton 288 Open 191 2
Rayon/acetate 296 Closed 105 4
Acetate 116 Closed 155 0
Cotton/polyester 117 Closed 267 1
Cotton/polyester 328 Closed 338 5
Cotton/polyester 117 Open 303 0
Rayon/polyester 367 Closed 658 2
Cotton/polyester 328 Open 236 7
Polyester 108 Closed 202 0
Acrylic 99 Closed 231 0
Acrylic 354 Closed 1177 8
Acrylic 99 Open 360 0
Acrylic 354 Open NA 7
Cotton/polyester/foam 305 Closed 385 1
Rayon/polyester/foam 284 Closed 326 0
Rayon/fiberglass 371 Closed 129 5
Rayon/fiberglass 371 Closed 106 5
Maximum possible number of panels to ignite ¼ 10
a
fire, as did FR cotton and FR rayon. Acrylic, and various all-plastic constructions. For samples
modacrylic, non-FR rayon and non-FR cotton sized 0.61 0.61 m, a redwood deck gave a
showed sustained burning, attaining peak HRR of 12 kW. Wood/plastic composites
100–250 kW peak HRR values when subjected ranged between 10 and 394 kW, while all-plastic
to the 50 kW ignition source. products ranged from 10 to 1055 kW.
Decks Desks
The California Office of State Fire Marshal Chow et al. [104] measured the HRR of a small
reported some HRR tests [103] done on outdoor wooden office desk. The desk was
decks, comparing wood, wood/plastic composite, 0.6 1.2 0.8 m high The ignition source
832 V. Babrauskas

fabric weight on number of
curtain panels ignited
300
Weight (g/m2)
200
Construction type
100 Fabric only
Fabric and foam
0
0 1 2 3 4 5 6 7 8
Number of panels ignited
Fig. 26.39 HRR of a 700

wooden desk tested by
Chow et al.
600
500
HRR (kW)
400
300
200
100
0
0 500 1000 1500 2000
Time (s)
was a pool of 0.5 L gasoline which, by itself, described in Table 26.6, while test results are
produced a peak HRR of 40 kW. These results shown in Fig. 26.40. These results must not
are shown in Fig. 26.39. be applied to appliances used in North America,
since European appliance styles are different
from North American ones and also because
Dishwashers local standards are such as to permit appliances
of greater flammability in Europe. HRR data
VTT tested [105] European dishwashers using a on North American dishwashers are not
propane burner of 1 kW. The specimens are available.
www.ebook777.com
Dressers Electric Cable Trays
A test of a wooden dresser has been conducted by Cable tray fires present almost an endless pleth-
NIST [106], see Fig. 26.41. ora of combinations of cable materials, tray con-
struction, stacking, ignition sources, etc. Only a
very few of these have been explored. The most
Dryers systematic studies available are those from
Tewarson et al. [107] and Sumitra [108]. A use-
Results for a small European clothes dryer ful engineering analysis of their data has been
(40 kg) have been published [70]. Even though prepared by Lee [109]. Lee provided a basic
use of plastics in North American clothes dryers correlation of Tewarson’s and Sumitra’s data
has been increasing, nonetheless it would appear (see Fig. 26.43), which shows that the peak full-
that the unit was more typical of the European scale heat release rate q_ fs (kW m2) can be
market than the American one. In the test predicted according to bench-scale heat release
(Fig. 26.42), 11 kg of mass was lost and rate measurements:
253 MJ of heat was released.
q_ fs ¼ 0:45q_00bs A
where q_00bs is the peak bench-scale HRR

Table 26.6 European dishwashers tested by VTT
(kW m2), measured under 60 kW m2 irradi-
Specimen D1 D2 ance, and A is the exposed tray area actively
Initial mass (kg) 35.6 47.5 pyrolyzing (m2). The active pyrolysis area, in
Mass loss (kg) 6.1 8.4 turn, is estimated from Fig. 26.44, which gives
dA/dt as a function of q_00 . Thus, at any given
Peak HRR (kW) 476 347
bs
Total heat (MJ) 165 206
time, t,
Fig. 26.40 HRR of 600

European dishwashers
tested by VTT D1
D2
400
HRR (kW)
200
0
0 600 1200 1800 2400 3000
Time (s)
834 V. Babrauskas
Fig. 26.41 HRR of 2000

wooden dresser
1800
1600

1400
1200
1000
800
600
400
200
0
0 300 600 900
Time (s)
Fig. 26.42 HRR for a 600

small European clothes
dryer
500
400
HRR (kW)
300
200
100
0
0 300 600 900 1200 1500 1800
Time (s)
dA types of snacks were tested in a single test—

Að t Þ ¼ Ao þ t
dt potato chips and cheese nibbles. A total fuel
load of 275 kg was set up in a tightly-packed,
Finally, Table 26.7 gives a selection of measured
three-shelf high shelving unit, 5.4 m long. The
values of q_00 for various cable types.
bs HRR results are shown in Fig. 26.45. Visual
observations indicated that potato chips burned
more vigorously than cheese nibbles.
Foodstuffs NIST [111] ran two full-scale tests on bags of
potato chips on a rack with open-wire-mesh
SP reported on a test [110] to simulate the burn- shelves. Each shelf had 20 bags of potato chips.
ing of snack foods in a shop. Retail bags of two The bags were arranged five across and four
www.ebook777.com
Fig. 26.43 HRR

prediction for cable trays 15
(numbers at data points 10 Cable tray correlation
n 17
identify full-scale tests) qfs = 0.45 qbs • A
9 16
Measured qfs (MW)

7
7 8
6
4 14
13
3 5 6
2 PE/PVC
PE, PP/CI • S • PE (Hypalon)
1 Silicone/asbestos
0
0 1 2 3 4 5 6 7 8 9 10
Predicted qfs (MW)

bench-scale cable heat
release rate on full-scale 2.2
rate of flame coverage
2.0
Rate of flame coverage (m2/min)
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0
0 100 200 300 400 500 600
Bench-scale rate of heat release (kW m2)
836 V. Babrauskas
Table 26.7 Heat release rates of typical cables in bench- deep, with a total fuel load of 27.1 kg. Each bag
scale tests of chips was approximately 200 mm wide by
Specimen IEEE q_00bs 100 mm (thick) by 360 mm high. Each bag
number Cable sample 383 test (kW m2) weighed 33.8 g, of which chips accounted for
20 Teflon Pass 98 32.5 g and the plastic bag for the rest. The potato
21 Silicone, glass braid Pass 128 chip ingredients were listed by the manufacturer
10 PE, PP/Cl · S · PE Pass 177 as: potatoes, sunflower oil and salt. Two replicate
14 XPE/XPE Pass 178 heat release rate experiments were conducted
22 Silicone, glass braid Pass 182 (Fig. 26.46). It is interesting to note that the
asbestos
NIST tests showed the same peak HRR (6 MW)
16 XPE/Cl · S · PE Pass 204
a as the SP test, albeit with a much shorter duration
18 PE, nylon/PVC, nylon 218
19 PE, nylon/PVC, nylon a
231
time due to the fact that the fuel load was 1/10 of
15 FRXPE/Cl · S · PE Pass 258 SP amount.
11 PE, PP/Cl · S · PE Pass 271
8 PE, PP/Cl · S · PE Pass 299
17 XPE/Neoprene Pass 302 Industrial Stored Commodities
a
3 PE/PVC 312
12 PE, PP/Cl · S · PE Pass 345 Pallet loads of plastic-based commodities are
a
2 XPE/Neoprene 354 commonly stored in factories, warehouses, and
a
6 PE/PVC 359 wholesale establishments. Most tests have
4 PE/PVC Fail 395 involved multiple pallets being tested, and most
13 XPE/FRXPE Pass 475 of these have also involved some manner of
5 PE/PVC Fail 589 water application being done during the test.
a
1 LDPE 1071 But there have been a few tests reported where
20 Teflon Pass 98 single pallet-loads were tested, without water. SP
a
Test not conducted [112] tested single pallet-loads of four kinds:
Fig. 26.45 HRR of potato 7000

chips and cheese nibbles
set up in a shop display unit
6000
5000
HRR (kW)
4000
3000
2000
1000
0
0 600 1200 1800 2400
Time (s)
www.ebook777.com
Fig. 26.46 Potato chip

bags tested at NIST
• FM Group A plastic standard commodity (see total weight of 680 kg of pellets. The pallet was
Table 26.10) over-wrapped with clear plastic film and spillage
• CEA standard commodity. Each corrugated- did not occur during the test. The full-pallet test
cardboard box is 450 550 370 mm and was ignited with a propane torch at the bottom.
each (wood) pallet holds 12 boxes in a The half-pallet test was ignited with a propane
2 2 3 array. Each box weighs 805 g torch at the top. The full-pallet test (Fig. 26.48)
and is filled with 340 g of polystyrene chips. showed a HRR of close to 7 MW when
The pallet-load is 800 1200 mm with a conditions required that the commodity be
height of 1110 mm, excluding the pallet itself. extinguished; peak HRR conditions had not
• SCEA standard commodity. This is a Swedish been reached.
version of the CEA, with each box being Heskestad [114, 115] analyzed a large series
380 570 380 mm. Each box weighs of palletized1 storage tests conducted at FM in
700 g and holds 420 g of chips. The pallet- 1975 by Dean [116]. These experiments
load is 800 1200 mm with a height of pre-dated the availability of HRR calorimeters,
1140 mm, excluding the pallet itself. so Heskestad obtained peak HRR values by using
• Large SCEA standard commodity. This is a mass loss rate data and values of effective heat of
variant where the box is 800 600 500 combustion. The test arrangement was
mm. Each box weighs 1470 g and contains 2 2 3 pallets high, with a flue space running
1220 g of chips. Each pallet holds a in only one direction. Heskestad also analyzed a
1 2 2 array of boxes. later series of rack-storage tests by Yu and Kung
The HRR results for these tests are shown in [117, 118]. The test arrangement was 2 2, with
Fig. 26.47. heights being two, three, or four pallets, and with
Despite the intention being that Group A flue spaces running in both directions.
plastics represent a severe fire hazard, some plas-
tic commodities produce significantly more
HRR. In tests by Babrauskas [113], pellets of 1
‘Palletized’ denotes a storage configuration where
SBR (styrene-butadiene rubber) were packed in pallets are stored directly on top of each other, without
paper bags and loaded on a wooden pallet, with a use of shelving.
838 V. Babrauskas
Fig. 26.47 HRR of single 4500

pallet-loads of various Group A std.
commodities tested at SP 4000 CEA std.
SCEA std.
Large SCEA std.
3500
3000
HRR (kW)
2500
2000
1500
1000
500
0
0 300 600 900 1200 1500 1800
Time (s)
Fig. 26.48 HRR of pallets 7000

holding bags of SBR pellets Extinguished
Full pallet
6000 Half pallet
5000
HRR (kW)
4000
3000
Extinguished
2000
1000
0
0 300 600 900 1200
Time (s)
Heskestad’s tabulated peak HRR values are area of 2.44 2.59 m, while the rack storage
given in Table 26.8. The peak HRR values were tests were 2.29 2.29 m. The cardboard cartons
obtained by dividing the value in kilowatts by the with metal liner are ‘FM Standard Class II Com-
floor area occupied by the commodity. The modity’ (Table 26.10 [119, 122]) while the PS
palletized test commodities occupied a floor cups are ‘FM Standard Plastic Commodity’
www.ebook777.com
Table 26.8 HRR values of palletized and rack-storage commodities tested at FM

Storage Peak HRR Time of
Test Commodity ht. (m) (kW m2) peak (s)
SP-4 PS jars in compartmented CB cartons 4.11 16,600 439
SP-13 PS foam meat trays, wrapped in PVC film, in CB 4.88 10,900 103
cartons
SP-23 PS foam meat trays, wrapped in paper, in CB cartons 4.90 11,700 113
SP- PS toy parts in CB cartons 4.48 5,210 120
30A
SP-35 PS foam insulation 4.21 26,000 373
SP-44 PS tubs in CB cartons 4.17 6,440 447
SP-15 PE bottles in compartmented CB cartons 4.20 5,330 434
SP-22 PE trash barrels in CB cartons 4.51 28,900 578
SP-43 PE bottles in CB cartons 4.41 4,810 190
SP-6 PVC bottles in compartmented CB cartons 4.63 8,510 488
SP-19 PP tubs in compartmented CB cartons 4.26 5,870 314
SP-34 PU rigid foam insulation 4.57 1,320 26
SP-41 Compartmented CB cartons, empty 4.51 2,470 144
RS-1 CB cartons, double tri-wall, metal liner 2.95 1,680 260
00 00
RS-2 2.95 1,490 89
00 00
RS-3 2.95 1,680 180
00 00
RS-4 4.47 2,520 120
00 00
RS-5 4.47 2,250 240
00 00
RS-6 5.99 3,260 210
RS-7 PS cups in compartmented CB cartons 2.90 4,420 95
00 00
RS-8 2.90 4,420 100
00 00
RS-9 2.90 4,420 120
00 00
RS-10 4.42 6,580 100
00 00
RS-11 5.94 8,030 148
CB cardboard, PE polyethylene, PP polypropylene, PS polystyrene, PU polyurethane
Table 26.9 Miscellaneous stored commodities tested [120] are listed in Table 26.9. These have not
by FM been re-analyzed by Heskestad.
Storage Peak HRR The effect of storage height [121] on the HRR
Commodity ht. (m) (kW m2) growth curve for Class II commodities is shown
Fiberglass (polyester) 4.6 1,400 in Fig. 26.49. An initial period of limited fire
shower stalls, in cartons
growth has been removed from these curves.
Mail bags, filled 1.52 400
These results are from FM testing in the 1980s.
PE letter trays, filled, 1.5 8,500
stacked on cart Also shown is the HRR curve for a 2 2 2
PE and PP film in rolls 4.1 6,200 array tested in 2005. For much of the time, the
HRR exceeded the earlier results. This is because
FM identified that the standard Class II commod-
ity supplied in 2005 is somewhat different than
(Group A Plastic). Note that there does not exist a that supplied earlier [125]. The early fire growth
scaling rule that would enable HRR values to be period [122] for Class I, III, and IV commodities
computed for stack/rack heights other than those is shown in Fig. 26.50. The early fire growth
tested. Thus, the reported values could conserva- period for the FM Standard Plastic Commodity
tively be applied to shorter heights, but cannot be is shown in Fig. 26.51. These results are based on
extrapolated to greater heights. Some older data early FM studies [123, 124] which were
Table 26.10 FM Commodities and standard test commodities
840
Class Products Examples Test commodity

Class I Essentially noncombustible; may be in Glass, minerals, Single-wall corrugated cardboard carton measuring 2100 (0.533 m) on side,
light cardboard cartons and may be on metals, ceramics divided into five horizontal layers by corrugated cardboard sheets. Each
wood pallets layer was divided by interlocking cardboard partitions forming a total of
125 compartments. Each compartment occupied by one 16-oz (0.47 l)
glass jar, without lid, open side facing down to prevent collection of water.
A pallet load consists of one wood pallet and eight of the above-described
cartons.
Class II Class I products with more or heavier Class I products in multiwall Double triwall (approx. 25 mm thick total) corrugated cardboard carton
packaging and containers cartons, boxes, or barrels. measuring 42" (1.07 m) on a side containing a 24 ga. (0.56 mm) sheet
metal liner box measuring 3800 3800 3600 (1.07 1.07 1.02 m)
high. A pallet load consisted of one wood pallet and one the above
described cartons.
Class III Combustible products in combustible Products of wood, paper, Single-wall corrugated cardboard carton measuring 2100 (0.533 m) on a
wrapping or containers on wood leather, and some foods side, divided into five horizontal layers by corrugated cardboard sheets.
pallets. May contain a limited amount Each layer divided by interlocking corrugated cardboard partitions
of plastic. forming a total of 125 compartments. Each compartment occupied by one
16-oz (0.95 l) paper jar (wide mouth container/ cup), without lid, open side
facing down to prevent the collection of water. A pallet load consists of
one wood pallet and eight of the above described cartons.
Class IV Class I, II, or III with considerable Typewriters and cameras Single-wall corrugated cardboard carton measuring 2100 (0.533 m) on a
plastic content in product, packaging of metal and plastic parts side, divided into five horizontal layers by corrugated cardboard sheets.
or pallets Each layer divided by interlocking corrugated cardboard partitions
forming a total of 125 compartments. Each compartment occupied by forty
www.ebook777.com
16-oz (0.95 l) polystyrene and eighty-five 16-oz (0.95 l) paper jars (wide
mouth container/ cup), without lids, open side facing down to prevent the
collection of water. A pallet load consists of one wood pallet and eight of
the above described cartons.
Standard plastic Commodities containing a greater Single-wall corrugated cardboard carton measuring 2100 (0.533 m) on a
(Group A Plastic) amount of plastic than would be side, divided into five horizontal layers by corrugated cardboard sheets.
permitted in Class IV commodities Each layer divided by interlocking corrugated cardboard partitions
forming a total of 125 compartments. Each compartment occupied by one
16-oz (0.95 l) polystyrene jar, without lids, open side facing down to
prevent the collection of water. A pallet load consists of one wood pallet
and eight of the above described cartons.
V. Babrauskas
Fig. 26.49 Effect of FM Class II Commodity on 4 Tier Rack Storage (5.99 m.high)
storage height for Class II
FM Class II Commodity on 3 Tier Rack Storage (4.47 m. high)
commodities
FM Class II Commodity on 2 Tier Rack Storage (2.95 m. high)
FM Class II commodity on 2 Tier (2005 data)
10,000
9,000
8,000
7,000
6,000
HRR (kW)
5,000
4,000
3,000
2,000
1,000
0
0 50 100 150 200 250
Time (s)
Fig. 26.50 The early fire-

FM Class I Commodity on 2 Tier Rack Storage (2.95 m. high)
growth period for Class I,
III, and IV commodities FM Class III Commodity on 2 Tier Rack Storage (2.95 m. high)
FM Class IV Commodity on 2 Tier Rack Storage (2.95 m. high)
250
200
150
HRR (kW)
100
50
0
0 50 100 150 200
Time (s)
842 V. Babrauskas
FM Plastic Commodity on 5 Tier (2x2x5) Rack Storage (7.50 m. high)

FM Plastic Commodity on 2 Tier (2x2x2) 2005 data
20000
18000
16000
14000
12000
HRR (kW)
10000
8000
6000
4000
2000
0
0 50 100 150 200 250
Time (s)
Fig. 26.51 The early fire-growth period for FM Standard Plastic commodity, as a function of storage height
conducted in their Norwood MA facility. Also high radiant heat release fraction, thus, to
shown are the results obtained in 2005 at their account for the radiant fraction and for the dimi-
West Gloucester RI facility for the 2 2 2 nution due to water spraying, the total HRR
configuration [125]. curve shown in Fig. 26.52 was estimated by
Additional FMRC data for different multiplying the measured convective portion by
commodities loaded onto wooden pallets are a factor of 2. The polyurethane foam results
shown in Fig. 26.52. The egg carton test [126] [127] are for a three-tier (4.27 m high) stack of
used foam-polystyrene egg cartons of 12-egg foam in cardboard boxes and used a PUR foam of
capacity. Polyethylene bags were used to hold high HRR; other results (not shown) were also
200–216 of these egg cartons, open and nested obtained by FM for fire-retardant grades. The
into each other. Each pallet held about 20.4 kg of PET (polyethylene terephthalate) bottles test
egg cartons. Each pallet contained about 22.7 [128] used 46 bottles of a 2 L size packed into
wood, and the load also contained about 0.4 kg single-wall corrugated cardboard boxes. Each
polyethylene. In this test, a low density of water box contained 2.55 kg of plastic and 1.29 kg of
extinguishment was applied, but this did not cardboard. Total test arrangement comprised
appear to significantly reduce the HRR of the eight pallet loads arranged in a 2 2 2
commodity. Only the convective portion of the arrangement. Each pallet contained eight cartons
HRR was measured. Polystyrene shows a very of the size 0.53 0.53 0.53 m. The
www.ebook777.com
Fig. 26.52 FMRC HRR PS egg cartons

results for several PUR Foam in Cardboard Boxes 3 Tier (4.27m. high)
additional commodities
PET bottles
Newsprint
20,000 FM Class II commodity (1.07 x 1.07 x 1.02 m high)
18,000
16,000
14,000
12,000
HRR (kW)
10,000
8,000
6,000
4,000
2,000
0
0 300 600 900 1200
Time (s)
Table 26.11 Boxed computer items tested by Hasegawa et al.

Code Items Peak HRR (kW)
P1 Boxed monitors, one pallet of 12 4700
P8 Boxed monitors, one pallet of 12, point-source ignition 5030
P5 Boxed monitors, one pallet of 12 (stabilized from collapse) 6400
P6 Boxed monitors, two pallets (side-by-side) of 12 each 17,300
P10 Boxed monitors, stack of two pallets high, 10 per pallet 14,100
P3 Boxed desktop computers, one pallet of 16 1400
P7 Boxed desktop computers, pallet of 16 + boxed accessory boxes on top 8190
P9 polystyrene foam in boxes 6730
P11 Monitor boxes, one pallet of 12 4600
newspaper test [129] comprised 8.2 kg of Packaged computers and computer

shredded newsprint placed in a accessories were tested by Hasegawa
0.53 0.53 0.51 m single-wall corrugated et al. [131, 132]. They tested pallet-loads of
cardboard box of 2.73 kg. Eight cartons packaged goods and also individual items, as
comprised one pallet load. The pallets were packaged and boxed in individual cardboard
arranged in a 2 2 2 arrangement. The news- boxes. The items were ignited using a line burner
print test [130] used a 2 2 2 arrangement of placed near the bottom edge of the package or
pallets, each load being 1.07 1.07 1.02 m stack. Ignition sources in the range of
high. The Class II commodity results are from 50–200 kW were used. Table 26.11 identifies
Khan [130]. the specimens tested, while Figs. 26.53 through
844 V. Babrauskas
Fig. 26.53 HRR of single, 400

packaged and boxed
Desktop (boxed)
computers and monitors
350 Laptop (boxed)
Monitor (boxed)
300
250
HRR (kW) 200
150
100
50
0
0 120 240 360 480 600
Time (s)
Fig. 26.54 HRR of a stack 2000

of polystyrene foam boards
1800
1600
1400
1200
HRR (kW)
1000
800
600
400
200
0
0 300 600 900
Time (s)
26.57 show the results. The monitors were by Särdqvist [97]. The total stack size was
16.8 kg each, while the desktop computers were 1.2 1.2 1.2 m, with a mass of 1.4 kg. Igni-
4.9 kg ea. The pallet load in test P1 collapsed tion was with a 1 MW burner at the side of the
during test and the full HRR was not registered, stack. The HRR curve is shown in Fig. 26.54.
consequently, it was re-tested with supported Numerous other example data are tabulated by
sides. Särdqvist [97].
A stack of expanded polystyrene boards was Dillon et al. [133] tested several commodities
burned by Dahlberg at SP and results are reported in a furniture calorimeter: acrylic yarns in boxes,
www.ebook777.com

of packaged, boxed
P1
computer monitors
P5
6000
P8
P11
5000
HRR (kW)
4000
3000
2000
1000
0
0 60 120 180 240 300 360 420
Time (s)

of packaged, larger arrays
of computer monitors 18000 P6
P10
16000
14000
12000
HRR (kW)
10000
8000
6000
4000
2000
0
0 10 20 30 40 50 60
Time (s)
computer monitors (US models, 430 mm [1700 ] polyurethane foam and had polyethylene outer
screen) packed in shipping boxes, plastic coolers, shells; the #1 sample had a polystyrene liner
and potato chip bags packed in cardboard boxes. while the #2 sample had a polypropylene liner.
The coolers with both insulated with The computer monitors were padded with
846 V. Babrauskas

of miscellaneous computer
items P3
8000
P7
P9
7000 P11
6000
HRR (kW)
5000
4000
3000
2000
1000
0
0 20 40 60 80 100 120
Time (s)
Table 26.12 HRR of packaged household commodities effect. The storeroom test [135] comprised a
tested by Dillon et al. mocked-up small storeroom in a retail shop,
Mass Peak HRR Time to Total HR with miscellaneous goods boxed in cardboard
Commodity (kg) (kW) peak (s) (MJ) boxes, placed on shelving 2.4 m high. A small
Acrylic yarn 8.7 263 210 127 amount of additional shelving was provided
skeins across an aisle 1.4 m wide. The FMRC test
Computer 24.6 140 398 70 involved pallets in a 2 2 2 arrangement. In
monitor
the storeroom test, ignition was at the base of the
Cooler #1 6.4 400 648 147
face of the ‘main’ storage rack. The FMRC test
Cooler #2 5.2 276 702 128
[136] used the standard FMRC procedure
Potato chips 8.3 322 230 139
whereby an igniter is also placed at the base,
but is located internally, at the two-way intersec-
expanded polystyrene foam, as is customary for tion of flue spaces between piles. The data for the
shipping. Their results are summarized in storeroom test are plotted as real time, while the
Table 26.12. FMRC test data were shifted 470 s to make the
A study has been reported on burning pallet steep HRR rise portions coincide. From a com-
loads of organic peroxides [134]. Liquids were parison of this kind, one can roughly estimate
packaged in plastic containers within cardboard that igniting a rack at the front face causes events
boxes, while solids were packaged in cardboard to occur 470 s later than would happen if ignition
drums. The data are given only for a few packag- were at the center of the flue spaces.
ing configurations with sufficient data not being
available to generalize HRR predictions to other
configurations. Kiosks
For all rack storage tests, the times are very
strongly affected by the ignition source location. NIST have reported [137] some HRR results on
Not enough data exist to make general full-scale tests of kiosks. These are the booths
correlations, but Fig. 26.58 illustrates the basic used in shopping malls, exhibitions, and other
www.ebook777.com
Fig. 26.58 Effect of 18,000

ignition source location on
fire development 16,000 Storeroom test
FMRC plastic commodity,

14,000 shifted by 470 s
12,000
HRR (kW)
10,000
8,000
6,000
4,000
2,000
0
300 350 400 450 500 550 600
Time (s)
Fig. 26.59 HRR of 2000

display kiosks
Test 1
1800
Test 2
Test 3
1600
Test 4
Test 5
1400
1200
HRR (kW)
1000
800
600
400
200
0
0 600 1200 1800 2400
Time (s)
places wherein a small amount of merchandise but refer to various configurations of the open-
display or sales occur. Some HRR curves are able panels. Test 5 appears to have been more
illustrated in Fig. 26.59 for a kiosk, built largely severe since all the panels were closed. Test
of wood, which measured 1.2 m 1.2 m 2.1 m 1 involved the same kiosk placed in a room,
high. Tests 2–5 are all of the same sized kiosk, rather than in the furniture calorimeter.
848 V. Babrauskas
Luggage Fig. 26.61. The larger ignition source used Test

2 led to much greater HRR, despite the fact that
At the LSF Laboratories, Messa [138] tested the the mass of paper goods was smaller than in Test
HRR of two suitcases filled with clothes. Ignition 3. While all tests were conducted in an ISO 9705
was with a square-ring burner applying approxi- room, the large HRR in Test 2 was evidently
mately 5.5 kW. The test articles are described in attributable to room-effect radiant heat flux rein-
Table 26.13 and the results are given in forcement, which was of less significance for the
Fig. 26.60. other tests. Thus, for design purposes, only Tests
1 and 3 should be considered, unless the end-use
environment is a relatively small room.
Magazine Racks
Chow et al. [139] conducted full-scale tests on Mattresses

several steel magazine racks, holding magazines,
newspapers, and books. Ignition was with a small Despite the relatively simple shape of mattresses,
pool of gasoline. Test details are given in the prediction of mattress HRR from bench-scale
Table 26.14, while HRR results are shown in data is difficult. Even the use of full-scale HRR
data is problematic, due to a peculiarity of mat-
tress fires. Most other combustibles interact only
Table 26.13 Test description for suitcases tested at LSF modestly with their environment, until large
Condition Soft suitcase Hard suitcase HRR values are reached or until room flashover
Mass empty (kg) 0.98 5.20 is being approached. Liquid pools on the other
Mass filled (kg) 3.06 10.34 hand, as discussed below, interact very strongly
Burner HRR (kW) 5.5 5.5 with a room, if either the room size or the avail-
Burner application time (s) 180 240 able ventilation are not very large in comparison
Total heat released (MJ) 33.4 139.0 to the pool’s HRR. The identical phenomenon is
Fig. 26.60 HRR of 140

suitcases Hard suitcase
Soft suitcase
120
100
80
HRR (kW)
60
40
20
0
0 600 1200 1800 2400 3000 3600
Time (s)
www.ebook777.com
Table 26.14 Details of magazine rack tests

Size of each rack Location of racks Mass of paper Ignition source, quantity
Test no. (WxH), m in room goods (kg) of gasoline (L)
1 1 2.2 Left, back 15 2
2 2 2.2 Back, right 60 15
3 2 2.2 Left, back, right 90a 3
a
Of which 15 kg was placed on floor, in front of racks
Fig. 26.61 HRR of 9000

magazine racks loaded with Test 1
magazines, newspapers, Test 2
8000
and books Test 3
7000
6000
HRR (kW)
5000
4000
3000
2000
1000
0
0 300 600 900 1200 1500 1800
Time (s)
Table 26.15 Some mattress HRR data; full-scale data are for small or no room effect, bench-scale data are peak
values, taken at 25 kW m2 irradiance
Ticking Combustible Peak HRR, Bench-scale HRR
Padding material material mass (kg) full-scale (kW) (kW m2)
Latex foam PVC 19 2720 479
Polyurethane foam PVC 14 2630 399
Polyurethane foam PVC 6 1620 138
Polyurethane foam Rayon 6 1580 179
Polyurethane foam Rayon 4 760 NA
Neoprene FR cotton 18 70 89
Cotton/jute FR cotton 13 40 43
observed with mattresses. Thus, there may not be HRR values that are found for common material
a single value of the HRR of a mattress, the HRR combinations [45]. The full-scale test protocol
having to be considered related to the room itself. used a complete set of bedding; ignition was
Some example data are compiled in achieved with a wastebasket. Figure 26.62
Table 26.15 to illustrate the peak full-scale illustrates the relation of bench-scale to full-scale
850 V. Babrauskas
Fig. 26.62 HRR of 3000

mattresses predicted from
bench-scale results. Full-
scale tests under conditions 2500
of negligible room effect;
Peak full-scale HRR (kW)

bench-scale HRR measured
at 25 kW m2 irradiance 2000
1500
1000
500
0
0 100 200 300 400 500
Peak bench-scale HRR (kW m−2)
Table 26.16 Some mattress HRR data; full-scale data include room effect of small bedroom
Combustible Peak HRR, 180 s avg HRR,
Padding material Ticking material mass (kg) full-scale (kW) bench-scale (kW m2)
Polyurethane foam Unidentified fabric 8.9 1716 220
Melamine-type PUR/cotton Polyester/ NA 547 169
batting/polyester fiber pad polypropylene
Polyurethane foam/cotton Unidentified fabric NA 380 172
batting/ polyester fiber pad
Polyurethane foam/polyester PVC NA 335 195
fiber pad
Melamine-type PUR FR fabric 15.1 39 228
FR cotton batting PVC NA 17 36
FR cotton batting Polyester 15.7 22 45
Neoprene PVC 14.9 19 31
data from the same data set, where full-scale cotton padding. Additional bedding included two
testing was done under conditions not leading to pillows, pillowcases, two sheets, and a comforter.
significant room fire effect. Not enough specimens Two tests were run in an open calorimeter—in
were tested to develop a usable correlation, so one test, an electric match was used to ignite the
the results should be taken only as indicative. bed, while in the other test a newspaper-filled
King-size mattresses dating from before the wastebasket was the ignition source. Unlike the
Federal HRR regulations can produce very high typical findings in the case of upholstered furni-
HRR values, even absent a room effect. NIST ture, here the ignition source type had a major
[140] tested a king-size bed assembly which effect, with the larger ignition source resulting in
contained box springs and an innerspring mat- a peak HRR over 5000 kW, while the smaller
tress consisting of polyurethane foam and felted only showed about 3500 kW and burner a longer
www.ebook777.com
Fig. 26.63 HRR of 1750

mattresses predicted from
bench-scale results. Full-
scale tests under conditions 1500
of significant room effect;
Peak full-scale HRR (kW)

bench-scale HRR measured 1250
at 35 kW m2 irradiance
1000
750
500
250
0
0 50 100 150 200 250
180 s avg. bench-scale HRR (kW m−2)

ignition source on king-size Wastebasket ignition
bed assemblies
Match ignition
5000
4000
HRR (kW)
3000
2000
1000
0
0 200 400 600 800 1000 1200
Time (s)
time at a slower rate (Fig. 26.64). In either case, The full-scale test setup was different for this
however, the HRR values would suffice to cause data set, in that no bedding was used and ignition
flashover in a bedroom environment, especially was with a burner flame at the edge of the mat-
in view of the fact that the HRR would be much tress. Thus, some mattresses were able to show
higher due to room effect augmentation. essentially zero HRR since bedding was not
Some full-scale data obtained under available to sustain burning, and the ignition
conditions where a strong room interaction effect source could be ‘evaded’ by receding specimens.
was seen are shown in Table 26.16 [141, 142]. A relation between full-scale and bench-scale
852 V. Babrauskas
Table 26.17 Results on mattress from the CBUF study

Pk. HRR furn. Pk. HRR Thick. Thick. Prop.
00 00 00 00
calor. (kW) room (kW) Springs (mm) factor q_ 60 q_ 180 q_ 180 th. fac. qtot fire
26 42 Sofabed 22 0.44 162 135 59 50 N
31 45 N 50 1.00 136 82 82 21 N
47 61 Y 10 0.20 225 227 45 43 N
47 NA Y 20 0.40 111 118 47 45 N
275 NA N 90 1.00 111 118 118 45 Y
348 471 Y 20 0.40 327 159 64 30 Y
313 1700 N 100 1.00 256 191 191 62 Y
917 2550 N 140 1.00 232 198 198 37 Y
results from this study is shown in Fig. 26.63. would be desired. As a first cut, it was concluded
The behavior in that study was found to be: that mattresses can be grouped into two groups—
• Mattresses with a bench-scale HRR (180 s those leading to propagating fires (the mattress
average value) of < 165 kW m2 led to being consumed in flaming combustion during a
room fires of less than 100 kW. relatively short time), and those that do not. The
• Mattresses with a bench-scale HRR (180 s former can be considered to be of the highest
average value) of > 165 kW m2 generally hazard, while the latter present only trivial haz-
led to room fires on the order of 1–2 MW. ard. Since, for practical reasons, all mattress
• The transition between those extremes was composites must be tested in the Cone Calorime-
very abrupt. ter using a 50 mm thickness, to take into account
The sharp transition between trivial fires and effects due to thin mattresses, a thickness factor
room flashover conditions can be attributed to the is defined:
details of the test room, but also to the use of an

ignition source which specimens of intermediate thickness, mm
Th: f ac: ¼ min , 1:0
characteristics could ‘evade.’ 50
Additional data on mattress HRR have been
published by SP [143], Lund University [144], For mattresses where the innersprings are used,
and in the CBUF project [10]. The CBUF study the thickness is measured from the top of the
included full-scale room fire tests, open-burning mattress down to top of the metal springs; it is
furniture calorimeter tests, and Cone Calorimeter not the total thickness. To determine whether the
tests. The mattress results are given in mattress fire will propagate or not, the following
Table 26.17. In both of the full-scale test rules were developed:
00
environments, no bedding was used, but a If q_ 180 ðTh: f ac:Þ < 100 kW m2
square-head burner was applied to the top surface and
of the specimen, precluding complications from
00
any receding-surface behavior. The bench-scale q_ 60 < 250 kW m2
test data presented were obtained at a 35 kW m2
irradiance. The results indicate that, when tested then,
in the standard ISO 9705 room, a very drastic Q_ < 80 kW (non-propagating fire)
room effect occurs for open-air HRR values over else,
about 300 kW. Q_ > 80 kW (propagating fire)
The bench-scale data indicated that when The HRR values over 80 kW in fact are flash-
widely varying mattress thicknesses exist, a sim- over values of up to 2.5 MW, but the scheme
ple relation of bench-scale to full-scale HRR does not assign a specific HRR number. Qualita-
cannot be sought, even if only predictions of tively, this scheme reflects the type of abrupt
open-burning (furniture calorimeter) results behavior change found in earlier studies
www.ebook777.com
(Fig. 26.63), but here some more refined rules showed an extreme radiant feedback effect, since
were developed that avoid non-predictions which mattresses not made to the Federal standard typi-
would occur from simple correlation. During the cally showed HRR values in excess of 3000 kW
same CBUF project, a more sophisticated mat- even for small mattress sizes, while a bunk
tress fire model has been developed by Baroudi bed attained > 6000 kW in the room test.
et al.; this model is not easy to use, but details are
available [10, 145].
In the US, mattresses made after July 1, 2007 Mining Equipment
have been required by law to conform to the
16 CFR 1633 standard of the Consumer Product Hansen and Ingason [146] tested two pieces of
Safety Commission. The latter augments the pre- mining equipment, burning them in an under-
vious standard (16 CFR 1632) for smoldering by a ground mine facility. The first item was a Toro
flaming test procedure. The primary requirement 501 DL diesel-powered wheel loader. The
for the new standard is that the peak HRR not machine weighs 36,000 kg and stands 2.85 m
exceed 200 kW; in addition the total heat release tall. The structure is steel, but it also contains
during the first 10 min of test must not exceed rubber tires, hydraulic oil, diesel fuel, and
15 MJ. NRCC [72] tested an example of one such smaller components, including driver’s seat,
mattress and did confirm a peak HRR < 200 kW. cables, etc., for an estimated fuel content of
However, a room fire test run with this same 76 GJ, the majority of this being the giant tires.
model of mattress, an equally-conforming mat- The second item was a Rocket Boomer 322 dril-
tress foundation, and a set of bedding produced a ling rig. This item weighs 18,400 kg and stands
peak HRR of 1812 kW. This would likely lead to 2.95 m tall. Its fuel content was estimated at
flashover in a room of the ISO 9705 room size and 46 GJ, with the fuel comprising hydraulic oil,
doorway dimensions. Another test run by NRCC hoses, tires, diesel fuel, cables, and
Fig. 26.65 HRR for two 30,000

types of mining equipment
Drilling rig
Loader
25,000
20,000
15,000
10,000
5,000
0
0 2000 4000 6000 8000 10000
Time (s)
854 V. Babrauskas
miscellaneous smaller items. With both items, person’s workstation. The identification of the
the fuel tank was partly emptied and poured out main conditions in these tests is given in
to create a pool fire under the specimen, and this Table 26.18. In one test series [147] replicates
pool was ignited to start the fire. Figure 26.65 were tested in an open furniture calorimeter, then
shows the HRR results, with loader achieving a the configuration was tested again in a room test;
peak value of 15.9 MW, while the drilling rig this is illustrated in Fig. 26.67.
showing 29.4 MW. In 2004, NIST [150] reported results of some
tests of modern office furniture, i.e., primarily
plastics-based. Two full-scale tests were
Office Furniture conducted, a single person cubicle, and a four-
person cluster of cubicles. The one-person cubi-
Office worker cubicles (‘workstations’) have cle was tested in an open environment, while the
been tested in several projects at NIST four-person cluster was in a semi-open arrange-
[147–149]. Figure 26.66 show that severe fire ment: three walls and a ceiling were present, but
conditions can be generated by these not the fourth wall. The results (Fig. 26.68) indi-
arrangements. In some cases, fires of nearly cate both a radiant augmentation due to the ceil-
7 MW were recorded from the burning of a single ing and an augmentation due to multiple fuel
Table 26.18 Workstations tested by NIST in 1988 and 1992

Combustible Number of sides
Code mass (kg) Description w. acoustic panels Ref.
A 291 Mostly old-style wood furniture 0 146
B 291 Semi-modern furniture 1 146
C 335 Modern furniture 2 148
D NA Modern furniture 3 148
E 291 Modern furniture 4 146
F NA Modern furniture 4 148
Fig. 26.66 HRR of office 7000

workstations tested by
NIST in 1988 and 1992 A
6000 B
C
5000 D
E
F
4000
3000
2000
1000
0
0 500 1000 1500 2000
Time (s)
www.ebook777.com
Fig. 26.67 NIST results 3000

for workstation tests of
1988 and 1992
Open test
2500 Open test, repeat
Room test

2000
1500
1000
500
0
0 300 600 900 1200 1500 1800 2100
Time (s)
Fig. 26.68 NIST results 20,000

for workstation tests
1 workstation
of 2004 18,000
4 workstations
16,000
14,000
12,000
HRR (kW)
10,000
8,000
6,000
4,000
2,000
0
0 200 400 600 800 1000 1200 1400
Time (s)
loads being present in direct proximity. In that steel components. The swivel chair had major
same study, NIST also ran open calorimeter tests components comprising hard-plastic shell
on two office chairs, a swivel chair and a chair material, and the fire involvement of these
with a fixed metal frame (Fig. 26.69). The gross components was the cause of the second HRR
mass for the chairs were 20.5 kg, and 11.8 kg, peak.
respectively, but the mass of the combustible Additional tests were conducted by Kakegawa
portions was not evaluated, although the major et al. [151] at Japan’s National Research Institute
fraction of the total mass was the mass of the of Fire and Disaster. Each test was started by a
856 V. Babrauskas

for the two office chairs Swivel chair
tested 450 Metal frame chair
400
350
300
HRR (kW) 250
200
150
100
50
0
0 200 400 600 800 1000 1200 1400
Time (s)
Fig. 26.70 HRR of four-

unit workstations tested at 3000 Test 1
NRIFD Test 2
Test 3
2500 Test 4
2000
HRR (kW)
1500
1000
500
0
0 300 600 900 1200 1500
Time (s)
polypropylene wastebasket filled with 0.2 kg of plastic trim parts. In addition, the workstations
paper. The wastebasket, by itself, was found to contained small filing cabinets, telephones,
show a peak HRR of 50–60 kW. The desks chairs, computers, and a modest amount of
were of modern metal-frame construction, with office paper. The HRR results for the four-unit
www.ebook777.com
Table 26.19 Workstations tested by NRIFD

Combustible No. of Partition Peak HRR Time to
Test mass (kg) Type of workstation desk units panels (kW) peak (s)
1 570 Clerical 4 N 3035 508
2 597 Clerical 4 Y 2476 616
3 1054 Engineering 4 N 2957 793
4 1086 Engineering 4 Y 2271 732
11 272 Engineering 1 Y 1602 441
12 264 Engineering 1 N 1870 412
14 263 Engineering 1 N 1219 601
Fig. 26.71 HRR of 2000

one-unit workstations Test 11
tested at NRIFD 1800 Test 12
Test 14
1600
1400
1200
HRR (kW)
1000
800
600
400
200
0
0 300 600 900 1200 1500
Time (s)
workstations are shown in Fig. 26.70, while those identical rows of square-section sticks, pallets
for the one-workstation units are shown in are made up of rectangular elements in a tradi-
Fig. 26.71. Even though the four-unit tionally dimensioned configuration as shown in
workstations had a very high fuel load, the Fig. 26.72. The fire safety concern with pallets
HRR values were lower than the American arises when they are idle and stacked many units
units studied at NIST. This is presumably due high. Krasner [152] has reported on a number of
to a more protected arrangement of the fuel, plus tests where the burning rate of pallets was
the fact that only short (0.45 m high) partition measured. A typical experimental heat
panels were used (Table 26.19). release rate curve is shown in Fig. 26.73. This
curve shows that, much like for a wood crib, a
substantially constant plateau burning can be
Pallets seen if the stack is reasonably high. The results
for a standard pallet size of 1.22 1.22 m can
Conceptually, a wood pallet is a similar arrange- be given as a general heat release rate
ment to a wood crib. The geometry, however, is expression
different. Instead of being composed of
858 V. Babrauskas
Fig. 26.72 The geometric

arrangement of a stack of
wood pallets
hc
1.2 P.
2m TY
TY 2m
P. 1.2
Fig. 26.73 HRR of a 4000

typical wood pallet stack
(1.22 1.22 1.22 m 3500
high)
3000
2500
2000
1500
1000
500
0
0 200 400 600 800
Time (s)
00
q_ ¼ 1368 1 þ 2:14h p ð1 0:03MÞ q_ ¼ 919 1 þ 2:14h p ð1 0:03MÞ
where hp is stack height (m), M is moisture The agreement between the above equations and
(%), and a net heat of combustion of experimental data is seen to be good over a wide
12 103 kJ kg1 has been assumed. For conve- range of pallet heights (Fig. 26.74), but the
nience in applying to nonstandard pallet sizes, expressions do somewhat overpredict the burn-
this can be expressed on a per-unit-pallet-flooring rates if applied to short stacks, with stack
area basis as: height hp < 0.5 m.
www.ebook777.com
Fig. 26.74 Dependence of

pallet HRR on stack height 14
Wood pallet stacks
1.22 m × 1.22 m × 0.14 m each pallet
12 Assumed: Δhc (net) = 12 MJ/kg
10
Peak HRR (MW)

4 points
8
4 Mean of 8 points
0
0 1 2 3 4 5 6 7
Pile height (m)
Fig. 26.75 HRR of 120

pillows PUR, 527 g
Polyester, 430 g
100 Latex, 1003 g
PUR, 650 g
PUR, 630 g
Polyester, 602 g
80
Feathers, 966 g
60
40
20
0
0 300 600 900
Time (s)
Pillows a variant of ISO 9705 especially configured for

pipe insulation testing [154]. Data on this config-
Pillow tests have been reported by NIST [153] uration have been published by Wetterlund and
and SP10. The results are given in Fig. 26.75. Göransson [155] and by Babrauskas [156].
Pipe Insulation Plants and Vegetation
The available data are from the configuration Trees, Natural

where pipe insulation is used to entirely cover Some tests on Christmas trees were reported by
the ceiling of a test room. The test method used is VTT [157] and by Damant and Nurbakhsh [158].
860 V. Babrauskas
Fig. 26.76 The peak HRR 400

for Douglas-fir Christmas
trees, as a function of
moisture and mass 350
300
Peak HRR/mass (kW kg−1)

250
200
150
100
50
0
0 20 40 60 80 100 120 140 160
Foliar moisture content (%)
Newer studies, however, indicated that these dry basis, so values can readily exceed 100 %;
tests, which examined only a few trees, did not also note that it is the needle (foliar) moisture
capture the full range of HRR values associated that governs the burning behavior—trunk mois-
with Christmas trees. The main variables that ture is not a relevant variable. The mass of the
govern the HRR of Christmas trees are the tree used here is the entire mass; Evans
following: et al. [160] suggested that if data are available
• Moisture content of the needles only for the foliar mass, but not the mass of the
• Mass of the tree entire tree, the approximation be used:
• Species
• Ignition source used mass ¼ 2 massfoliar
Moisture is the dominant variable and this had
not been studied previously. The results of an To ignite trees with a small flame requires that
extensive series of fire tests [159] on Douglas-fir the moisture content be below 50–60 %. Other-
(Pseudotsuga menziesii) trees are shown in wise, ignition is still possible if using larger
Fig. 26.76, while the HRR of a typical test is combustible objects. In the work reported, the
illustrated in Fig. 26.77. The trees were about trees which could not be ignited by a small flame
2.1 m tall, had an average mass of 11 kg. The were all ignited by first igniting wrapped gift
trees were cut, placed in a watering stand, and packages placed under the tree. For design
watered according to various watering programs. purposes, it should be adequate to assume that
The average tree was kept for 10 days prior to the heat release curve is a triangle. This requires
testing. The relation of the curve fit in Fig. 26.76 is: knowing only the peak HRR and the total heat
released. To estimate the latter, it was found in
400 the tests that the Christmas trees showed an
_
q=mass ¼
1 þ 0:0538MC effective heat of combustion of 13.1 MJ kg1.
Thus, from knowing the mass of the tree and the
where MC ¼ foliar moisture (%) and the units of effective heat of combustion, the total heat
_
q=mass are kW kg1. Moisture is measured on a
www.ebook777.com
release may be estimated. The needle moisture The author also ran one interesting test where
may not be known for design purposes. It is a tree was burned lying on its side, instead of
governed both by the watering program and by vertical. This gave a much lower peak HRR
the innate biology, e.g., the species, of the tree. value of 1603 kW, indicating a major role of
No model is available at the present time that can geometry.
predict the moisture. However, the research Jackman et al. [163] tested Noble fir and
indicated that Douglas-firs are a notably short- Norway spruce specimens. For a relatively dry
lasting species. The data points shown in (26 % MC) Noble fir (16.3 kg, 3.07 m tall) they
Fig. 26.76, with one exception, represent trees got a peak HRR of 2880 kW. A similar Norway
that had been on display for less than 16 days; spruce (20 % MC, 14.9 kg, 3.02 m high) showed
some were watered carefully and regularly, only 1590 kW. Because very few tests were run,
while others were not. Other species of Christ- this should not be taken to indicate an intrinsic
mas trees, such as Noble fir or Fraser fir are species effect.
considered to be longer-lasting, but are less com-
monly bought. Bushes, Natural
A smaller test series on Scotch pine trees was Stephens et al. [164] tested Tam juniper
tested at NIST by Stroup et al. [161]. They exam- (Juniperus savina tamariscifolia) shrubs of vari-
ined trees of 2.3–3.1 m height and mass between ous moisture contents, ignited by a medium gas
9.5 and 20.0 kg; with one exception, the trees were flame. For MC < 50 %, rapid combustion
of mass 12.7 kg or greater. Apart from one tree, resulted and samples showed 1800–2100 kW
which is not considered here since it was not peak HRR. Specimens of 50–80 % MC typically
successfully ignited, the trees were left without showed 600–800 kW, while specimens of higher
water for 3 weeks in a room at 50 % RH and 23 C. MC did not burn significantly. Unfortunately,
Ignition was with an electric match to a lower neither the mass nor the size of the specimens
branch of the tree. The Scotch pines were substan- were specified, except that they were denoted as
tially taller and heavier than the Douglas-firs, so it “mature.” Shrubs of this species in general reach
is not surprising that higher peak HRR values about 0.45 m height and cover about 1 m of
were attained. The peak HRR values ranged width.
from 1620 to 5170 kW. Normalized per mass, Etlinger [165] conducted more extensive tests
the average was 183 kW kg1, with the range on a series of decorative-shrub species:
being 103–259 kW kg1. The moisture of the Armstrong juniper (Juniperus chinensis
branches was not recorded, but presumably ‘Armstrongii’), hedge saltbush (Rhagodia
was <20 % in all cases. Comparing to the spinescens), milkflower cotoneaster (Cotoneas-
above results, Douglas-firs showed about ter lacteus), mountain lilac (Ceanothus ray
160–330 kW kg1. This would suggest that there ‘Hartman’), oleander (Nerium oleander), purple
is a species effect and that Scotch pines show a rockrose (Cistus purpureus), quail bush (Atriplex
HRR/mass ratio approximately 0.75 of that found lentiformis), sageleaf rockrose (Cistus
for Douglas-firs. This conclusion is very tentative, salvifolius), Santa Barbaras ceanothus (Ceano-
however, since the test programs did not use the thus impressus ‘Eleanor Taylor’), trailing rose-
same test protocol. Part of the difference might mary (Rosmarinus officinalis ‘Prostrata’), and
also be attributed to a height effect, since this vine hill manzanita (Arctostaphylos densiflora
cannot separately be taken into account. ‘Howard McMinn’). The bushes were typically
Madrzykowski [162] ran more recent tests at in the range of 0.5–1.0 m tall and weighed
NIST on 2.05–2.54 m tall Fraser fir trees with a 1–3 kg. He ignited the bushes first with a
moisture content of 6–9 %. The peak HRR values 40 kW burner, which did not cause the specimens
ranged from 3231 to 4344 kW, while the mass of to show a significant HRR output, followed by an
the specimens ranged from 10.97 to 13.81 kg, exposure to a 150 kW burner. Some typical
giving an average peak HRR/mass value of results upon exposure to a 150 kW burner are
286 kW kg1, with a range of 218–348 kW kg1. shown in Fig. 26.78.
862 V. Babrauskas
Fig. 26.77 Typical HRR 3500

curves of Douglas-fir
Christmas trees
3000
MC=20%
2500
2000
HRR (kW)
MC=38%
1500
1000 MC=70%
500
0
0 100 200 300 400
Time (s)
Fig. 26.78 Typical HRR 900

for bushes obtained by
Etlinger 800
Juniper 2.07 kg,
31% MC
700 Rockrose 2.03 kg,
14% MC
Manzanita 1.11 kg,
600 87% MC
Ceanothus 1.68 kg,
HRR (kW)
500 17% MC
400
300
200
100
0
0 100 200 300
Time (s)
A summary of Etlinger’s peak HRR values is be derived which is suitable for both trees and
shown in Fig. 26.79. There is a significant bushes:
amount of scatter, but the results for the bushes
are not systematically different from those for 700
_
q=mass ¼
trees, which are also shown on the same plot. 1 þ 0:1295MC
Thus, a single expression for the peak HRR can
www.ebook777.com
Fig. 26.79 HRR of bushes 400

and trees, plotted on a
Bushes
unified basis
350 Trees
300
Peak HRR/mass (kW kg−1)

250
200
150
100
50
0
0 50 100 150 200 250 300 350
Foliar moisture content (%)
Trees, Plastic heights imply peak HRR values of

UL had published a procedure for testing artifi- 2800–3100 kW. The values actually measured are
cial Christmas trees. This was identified as Sub- systematically low due to two reasons:
ject 411 [166]. It was eventually withdrawn due (1) mixing dilution effects due to use a large,
to lack of activity in this area, but the basic test room-sized calorimeter hood; and (2) inability
procedure is sound. In the UL procedure, the of instrumentation to respond to a fast-growing
ignition source comprises 454 g of shredded fire.
newspaper, conditioned at 35–40 % RH, and Jackman et al. [162] tested three artificial trees
dispersed in a 610 mm diameter circle around (2.0–2.5 m height) and obtained peak HRR
the base of the tree. The newspaper is ignited at values of 100–400 kW. It is not clear whether
four points around the perimeter of the circle. these low values represent an intrinsically low
This ignition source is realistic, since it HRR of these trees (of unspecified plastic) or
represents the effect of burning gift packages or whether it simply reflects the fact that an ignition
decorations at the base of the tree. Babrauskas source was used which is much less serious than
[167] tested PVC trees of 4.2–4.7 kg mass and the one in the UL procedure.
1.96–2.01 m height.
Figure 26.80 shows the results for two Bushes, Plastic
replicates using the exact using the UL Some HRR data on plastic house plants are
procedure and a third test where the newspaper shown in Fig. 26.81 [170].
was ignited at one point only. The specimens
proved impossible to ignite with a small flame,
but using the UL procedure they produced rapidly Pools, Liquid or Plastic
developing fires, with 500 kW being attained
11–20 s after ignition. The peak flame heights of Possibly the simplest geometric arrangement of
the two specimens ignited in four places was fuel is a liquid (or thermoplastic) pool. Over the
4.8–5.1 m. Using the Zukoski [168] or McCaffrey last four decades, an enormous number of studies
[169] flame height/HRR correlations, such flame have been conducted where pool burning was
864 V. Babrauskas
Fig. 26.80 Recorded 1000

HRR for PVC Christmas Ign. at 4 points
trees (actual HRR was 900 Ign. at 4 points
greater, due to very fast rise Ign. at 1 point
time) 800
700

600
500
400
300
200
100
0
0 60 120 180
Time (s)
considered theoretically or measured empiri- Table 26.20 The burning regimes for liquid pools
cally. The most systematic early study was by Diameter (m) Burning mode
two Russian researchers, Blinov and Khudiakov < 0.05 Convective, laminar
[171]. Their results were analyzed by Hottel 0.05–0.2 Convective, turbulent
[172], who pointed out that conservation of 0.2–1.0 Radiative, optically thin
energy can be applied to the pool: > 1.0 Radiative, optically thick
00
00 00 00 00

q_ ¼ q_ A ¼ q_ r þ q_ c q_ rr q_ loss
heat of evaporation Δhv, which is the enthalpy
Δhc
A required to change a unit mass of liquid to a gas
Δhg at 25 C. The relation between these two
where q_ is the heat release rate of the pool; quantities is:
double-prime denotes per unit area; A is the
00
area of the pool (m2); q_ r is the radiant heat flux Δhg ¼ Δhv þ ðT b 25Þ Cpv
00
absorbed by the pool; q_ c is the convective heat where we have taken the simplification that Cpv,
00
flux to the pool; q_ rr is the heat flux re-radiated the heat capacity of the vapor (kJ kg1 K1) is
00
from the surface of the pool; and into q_ loss are a constant. An extensive tabulation of these
lumped wall conduction losses and non-steady constants is provided by Babrauskas [173].
terms. The heat of gasification is Δhg (kJ kg1), Hottel’s analysis of Blinov and Khudiakov’s
while the (lower, or net) heat of combustion is data showed two basic regimes are possible:
Δhc. Note that some authors use the symbol L for radiatively dominated burning for large pool
the heat of gasification. The heat of gasification is diameters, D, and convectively dominated burn-
defined as the enthalpy required to bring a unit of ing for small D. Furthermore, in the convective
mass of liquid-phase substance at 25 C to the regime the flow can be either laminar or turbulent
gaseous state at the temperature Tb, its boiling (being always turbulent for radiatively driven
point. It should not be confused with the latent pools), while in the radiative regime the flames
www.ebook777.com

can be optically thin or thick. These distinctions 00 Δhc
q_ ¼ q_ c A
can, in the simplest analysis, be made solely on Δhg
the basis of pool diameter. Such a simple classi-
fication is possible if the pool is strictly circular, however, the values of q_00c to be taken are not
radiant heating is only from the pool’s flames and easily determined. Some additional details are
not augmented by external sources, and there are given in [174]. For fire hazard analysis purposes,
no interferences to the flow streamlines which liquid pool fires will rarely be significantly dan-
could trip the onset of turbulence. In such a gerous if they are smaller than about 0.2 m in
simplified case, the regimes can be identified as diameter. Thus, it will often only be necessary to
in Table 26.20. treat pools burning in the radiative regime. In the
In the convective limit (small pools), one may radiative regime, it is found that data for most
make the following approximation: organic liquids can be well correlated by:
Table 26.21 Pool burning: thermochemical and empirical constants for a number of common organic fuels
00
Material Density (kg m3) Δhg (kJ kg1) Δhc (MJ kg1) m_ 1 (kg m2 s1) kβ (m1)
Cryogenics
Liquid H2 70 442 120.0 0.017 (0.001) 6.1 (0.4)
LNG (most CH4) 415 619 50.0 0.078 (0.018) 1.1 (0.8)
LPG (mostly C3H8) 585 426 46.0 0.099 (0.009) 1.4 (0.5)
Alcohols
a a
Methanol (CH3OH) 796 1195 20.0
a a
Ethanol (C2H5OH) 794 891 26.8
Simple organic fuels
Butane (C4H10) 573 362 45.7 0.078 (0.003) 2.7 (0.3)
Benzene (C6H6) 874 484 40.1 0.085 (0.002) 2.7 (0.3)
Hexane (C6H14) 650 433 44.7 0.074 (0.005) 1.9 (0.4)
Heptane (C7H16) 675 448 44.6 0.101 (0.009) 1.1 (0.3)
Xylenes (C8H10) 870 543 40.8 0.090 (0.007) 1.4 (0.3)
Acetone (C3H6O) 791 668 25.8 0.041 (0.003) 1.9 (0.3)
Dioxane (C4H8O2) 1035 552 26.2 0.018 5.4
Diethyl ether (C4H10O) 714 382 34.2 0.085 (0.018) 0.7 (0.3)
Petroleum products
Benzine 740 – 44.7 0.048 (0.002) 3.6 (0.4)
Gasoline 740 330 43.7 0.055 (0.002) 2.1 (0.3)
Kerosene 820 670 43.2 0.039 (0.003) 3.5 (0.8)
JP-4 760 – 43.5 0.051 (0.002) 3.6 (0.1)
JP-5 810 700 43.0 0.054 (0.002) 1.6 (0.3)
Transformer oil, hydrocarbon 760 – 46.4 0.039 0.7
Fuel oil, heavy 940–1000 – 39.7 0.035 (0.003) 1.7 (0.6)
Crude oil 830–880 – 42.5–42.7 0.060 0.62
Solids
Polymethylmethacrylate 1184 1611 24.9 0.020 (0.002) 3.3 (0.8)
Polyoxymethylene (CH2O)n 1425 2430 15.7
Polypropylene (C3H6)n 905 2030 43.2
polystyrene (C8H8)n 1050 1720 39.7
a
See text
866 V. Babrauskas
00 00
q_ ¼ Δhc m_ 1 1 ekβ D A constant values of m_ , independent of diameter.
Based on some newer test results [175], it is clear
This requires determining two empirical that a diameter effect does exist, although it
00
constants: m_ 1 and the term (kβ); the first of cannot be expressed in standard form. Thus, it
these is the asymptotic mass loss rate per unit is recommended that for methanol or ethanol the
00 00
area as the pool diameter increases towards infin- values be used: m_ ¼ 0.015 (D < 0.6 m); m_ ¼
00
ity; the second is the product of the extinction- 0.022 (0.6 < D < 3.0 m); and m_ ¼ 0.029
absorption coefficient k and the beam-length (D > 3.0 m).
corrector β. These constants are given in The above discussion implicitly assumed that
Table 26.21 for a number of common fuels. the pool depth is at least several millimeters. If
The net heat of combustion, Δhc, is also listed liquids are spilled on a horizontal surface that has
in the table. In principle, a slightly lower value, no low spots and no diking, then a liquid layer
the effective heat of combustion, should be used will form that is less than 1 mm thick. Thin-layer
instead of the net heat of combustion that is pools of this nature (which can occur in arson
determined with oxygen bomb calorimetry. cases) show a lower HRR than do pools of
Some bench-scale values of a combustion greater depths. Putorti et al. [176] studied
efficiency factor to convert oxygen bomb values gasoline spills on wood parquet, vinyl floor tiles
into experimentally-measured values are given in and carpeting. When a specified volume of liquid
Chap. 36, “Combustion Characteristics of is spilled, the problem to be solved can be
Materials and Generation of Fire Products.” For separated into two components: (1) determining
most liquids, however, the bench-scale values are the area of the spill, or, equivalently, the spill
not greatly below unity and realistic large-scale thickness; and (2) determining the HRR per
measurements are not available, thus the unit area.
improvement in accuracy by extrapolating from For wood floors, Putorti found the A ¼ 1.5 V,
bench-scale results may be nil. where A ¼ area (m2) and V ¼ volume (L).
Alcohol fuels show minimal radiative flux, in For vinyl tile, a similar relation was also found,
comparison to other fuel types. Thus, the best but the constant being 1.8. Converted into layer
recommendation previously had been to use thicknesses, the thickness for wood was 0.67 mm
Fig. 26.81 HRR of plastic 200

house plants
Palm, bushy
Palm, slim
Ficus
150
100
50
0
0 60 120 180 240
Time (s)
www.ebook777.com
and for vinyl tile it was 0.56 mm. Earlier work Table 26.22 HRR of European refrigerators
has indicated that a relation of this kind should Specimen R1 R2 R3
only be applied to smooth floor surfaces. For Initial mass (kg) 70.0 67.2 43.7
rough, absorptive surfaces a constant thickness Mass loss (kg) 18.0 14.3 18
is not obtained, and larger spill volumes produce, Peak HRR (kW) 2125 1816 852
effectively, greater layer thicknesses [177]. Extinguishment time (s) 925 722 –
Putorti’s study with carpets both indicated Total heat (MJ) 537 404 432
large differences between carpet types and also
showed that the data could not be represented as
a constant layer thickness. The HRR per unit area simple formulas exist for computing the
values are shown in Fig. 26.82. For the solid enhanced burning rates when a pool receives
surface pours, the spill areas were in the range significant room radiation. If computations
0.4–1.8 m2. As presented above, pools of large under these conditions are necessary, the theoret-
depths in this size range would show HRR values ical study of Babrauskas and Wickström [11]
of 1900–2400 kW m2. Thus, the carpet-surface should be consulted. The computer program
values are about 70–80 % of values that would COMPF2 [180] can also be used to treat
have been computed using the normal pool fire this case.
formulas. The smooth-surface values, however, The problem of pool burning is interesting
are only about 1/5 of the values that would be from a combustion science point of view, and
found for pools of sizable depths. over the years there has been a very large number
A similar study by Gottuk et al. [178] also of studies which attempted to go beyond empiri-
describes HRR values for spills on hard surfaces cal predictions [181–184]. In addition, work is
that are, very roughly, about 1/5 of those for occurring to provide more detailed experimental
‘normal’ pools. The relationships found by measurements for specific fuels [185, 186].
Putorti can only be expected to hold on dead-
flat surfaces. If surfaces are crooked, then
ponding at low spots will occur and uniform Refrigerators
spill depths should never be anticipated.
DeHaan [179] conducted two tests using 1.9 L VTT tested [105] two European refrigerators
of Coleman camping fuel. This is a straight-run using a propane burner of 1 kW (designated R1,
petroleum distillate containing normal and R2), while EFRA tested a single refrigerator
iso-alkanes ranging from hexane to undecane. (R3), ignited with a needle-flame burner. The
When poured on an unpadded carpet, a HRR specimens are described in Table 26.22, while
peak of 1150 was found, with a burning time of test results are shown in Fig. 26.83. The VTT
roughly 3 min. When poured upon a carpet that specimens were extinguished before the ultimate
had an pad underneath it, a lower HRR peak peak burning would have occurred, while the
(890 kW) was found, the peak was slightly EFRA specimen was not. These results must
delayed (85 s, versus 65 s) and there was a long not be applied to appliances used in North Amer-
tail to the HRR curve. ica, since European appliance styles are different
The discussion above pertains only to open- from North American ones and also because
burning fires. Thus, the literature-derived burn- local standards are such as to permit appliances
ing rates can be used only in the case of a very of greater flammability in Europe.
large, well-ventilated room (compared to the size
of the fire). If calculations show that the ‘free-
burning’ pool would cause a temperature rise of Shop Displays
more than, say, 100 C, then it is clear that
radiative feedback will start being important Chow [187] tested shop displays of three types:
and such an approximation cannot be made. No clothing display, compact disc (CD) display, and
868 V. Babrauskas
2000
1800
1600
HRR per unit area (kW m−2)
1400
1200
Wood
Vinyl
1000
Carpet 1
800 Carpet 2
600
400
200
0
0 0.25 0.5 0.75 1 1.25
Spill volume (L)
Fig. 26.82 HRR of thin pools of gasoline over various surfaces
Table 26.23 Shop-display commodities tested by Chow

Display type Combustible mass (kg) Size (m) Ignition source (kW) Peak HRR (kW)
Clothing 470 2400
Compact discs 2 ea, 1.5 m wide 1.6 m high 1100
Newsstand 15 2 ea, 1 m wide 2.2 m high 400 3600
Fig. 26.83 HRR of 2500

European refrigerators
tested by VTT and EFRA R1
R2
2000 R3
1500
HRR (kW)
1000
500
0
0 500 1000 1500 2000 2500
Time (s)
www.ebook777.com
Fig. 26.84 HRR of 4000

various shop-display
Newsstand
commodities tested CD rack
by Chow 3500
Clothes rack
3000
2500
HRR (kW)
2000
1500
1000
500
0
0 200 400 600 800 1000
Time (s)
newsstand. The clothing display comprised Babrauskas et al. [88] tested at NIST small
all-cotton T-shirts arranged on four small display polystyrene television cabinets of two types, fire-
racks. The CD display contained a total of retarded and not. Since the circuit components
240 discs. The ignition source in each case was contribute negligible HRR in comparison to the
a small pool of gasoline, to represent an arson outer shell, only the cabinets were tested. Two
fire. The results are shown in Table 26.23 and very small (“personal size”) units were tested
Fig. 26.84. side-by-side in each test. This can represent either
two appliances or simply the mass of one larger set.
SP tested two television sets [29], a
Television Sets US-market set with housing having a V-0 rating,
and a Swedish set with a housing having an HB
The burning characteristics of TV sets depend rating. The US set was a 690 mm (27 in.) model,
greatly on whether they have been made for while the Swedish one was 710 mm (28 in.). The
the North American market, following the US set had a total combustible mass of 6.5 kg,
requirements of UL, or not. In countries where with 2.9 kg comprising the enclosure, while the
UL standards do not apply, plastic TV cabinets Swedish set had 6.0 and 2.7 kg, respectively. The
are generally highly flammable, commonly being Swedish set was successfully ignited and burned
made of plastics that only have an HB rating with a small flame the size of a match flame. The
according to the UL 94 [30] procedures. These US set resisted ignition from this source and was
are readily ignitable from small-flame ignition then subjected to a 10 kW burner. With this
sources and burn vigorously when ignited challenge, the set burned, but showed little
[44, 188]. By contrast, sets made for the North HRR beyond the 10 kW of the source. Finally,
American market have to obtain a V-0 classifica- the test protocol chosen was a 30 kW burner. The
tion under UL 94 and will resist ignition from burner HRR was subtracted out from the data
small flame sources. shown in Fig. 26.85.
870 V. Babrauskas
Fig. 26.85 HRR of 600

various television sets NIST, non-FR
tested by NIST and SP NIST, FR
SP, Swedish TV
500
SP, US TV
400

300
200
100
0
0 300 600 900 1200 1500 1800
Time (s)
Table 26.24 European televisions tested by VTT entertainment center. The ignition source was a
Specimen TW1 TW3 TP1 TP2 TP3 small amount of alcohol for HB-rated cabinets.
Type Wood Wood Plastic Plastic Plastic For the V-0 rated cabinets, some small consumer
Size (inches) 24 26 28 25 28 goods, HB rated, were first ignited and these were
Initial mass (kg) 32.7 39.8 31.8 24.4 30.5 then used to ignite the test TV sets (Tables 26.24
Mass loss (kg) 10.2 10.2 5.2 4.6 5.3 and 26.25). After the initial peak (Fig. 26.87),
Peak HRR (kW) 230 290 274 239 211 the burning involved the wood entertainment cen-
Total heat (MJ) 146 150 140 116 137 ter, thus the latter portion of these HRR curves is
not germane to TV sets per se.
VTT conducted two projects where TV sets
were tested. In the first study [156], they tested
two old, 1960s vintage (black-and-white) Transport Vehicles and Components
televisions with large wood cabinets; these
were ignited with a small cup of alcohol. In a Passenger car HRR was measured at the Fire
newer study [105], they tested modern plastic- Research Station [191] and VTT [192]. The
cabinet televisions using a propane burner of FRS laboratory examined a 1982 Austin Maestro
1 kW. The specimens are described in and a 1986 Citroën BX, while VTT examined a
Table 26.22, while test results are shown in Ford Taunus, a Datsun 160, and a Datsun 180.
Fig. 26.86. Nam et al. [189] tested a modern The dates of manufacture were only stated as late
TV set (plastic cabinet) together with a wood 1970s. These results are shown in Fig. 26.88.
stand for it. They obtained peak HRR values of Additional tests were reported by MFPA [193]
200–300 kW, although the peak took 20–40 min and SP [194]. MFPA tested a Citroën, a Trabant,
to reach. and a Renault Espace, while SP tested a Fiat
The most recent results come from Hoffmann 127 of unspecified vintage. These results are
et al. [190] who tested TV sets in a wooden shown in Fig. 26.89. The peak values range
www.ebook777.com
Table 26.25 Characteristics of TV sets tested by Hoffmann et al.
TV
Test No. Screen Size Rating Other HB Devices Ignition Source Peak HRR (kW) Time to Peak (s)
1A 510 mm (20 in.) V0 1 cordless Phone, 5 mL IPA 363 273
1 small radio
2A 510 mm (20 in.) V0 1 telephone 5 mL IPA adjacent 199 594
to phone
3A 480 mm (19 in.) HB None 5 mL IPA >1450 615
1B 510 mm (20 in.) V0 1 cordless phone, 5 mL IPA >1000 216
1 small radio
2B 510 mm (20 in.) V0 1 telephone 5 mL IPA adjacent 299 975
to phone
Fig. 26.86 HRR of 300

European television sets TP1
tested by VTT TP2
TP3
250
TW1
TW3
200
HRR (kW)
150
100
50
0
0 300 600 900 1200 1500 1800
Time (s)
from 1.5 to 8.5 MW. These numbers are rather tests were run in a two-car configuration, involv-
widely disparate and it is not fully clear why, ing one small car (790 kg) side-by-side to a large
except that this is not due to the fraction of car (1306 kg). These results are shown in
polymer content onboard. Fig. 26.91, but test details were not published.
Some very extensive testing was conducted at The mass loss values are shown in Table 26.26.
CTICM, as shown in Fig. 26.90. Test 2 was a Okamoto et al. [195] ran a series of
Renault 18 (951 kg), Test 3 a Renault 5 (757 kg), experiments where they tested replicates of the
Test 4 another Renault 18 (955 kg), while the same vehicle (Toyota Cressida, also known as
specimens for the remaining tests were only Mark2 GX81) but varied the test conditions
identified as a “Large car, 1303 kg” (Test 7), (Table 26.27). Figure 26.92 shows the HRR
and “Small car, 830 kg” (Test 8). Additional results; spikes judged to be spurious were
872 V. Babrauskas
Fig. 26.87 HRR of TV 800

sets tested by Hoffmann
Test 1A
et al.
Test 2A
700
Test 3A
Test 1B
600 Test 2B
500
HRR (kW)
400
300
200
100
0
0 200 400 600 800 1000 1200
Time (s)
Fig. 26.88 HRR of cars 9000

tested at FRS and VTT
8000 Citroen BX
Austin Maestro
7000 Ford Taunus
Datsun 160
6000 Datsun 180
HRR (kW)
5000
4000
3000
2000
1000
0
0 500 1000 1500 2000 2500 3000
Time (s)
removed from these data. In Test B, an explosion many details of fire development in these
occurred at 1517 s, when pyrolysates experiments. The results suggest that small
accumulated in the passenger compartment sud- differences in test conditions can affect the time
denly ignited. Explosions did not occur with the scale of fire development in an automobile quite
other tests. The tests are especially valuable notably, also that windows should be open if
since, in their paper, the authors documented maximum HRR conditions are to be elicited. It
www.ebook777.com
Fig. 26.89 HRR of cars 7000
tested at MFPA and SP
Citroen Steinert
6000 Trabant Steinert
Renault Espace
Fiat 127
5000
HRR (kW)
4000
3000
2000
1000
0
0 1000 2000 3000 4000 5000
Time (s)

CTICM one-car tests
Test 2
8000
Test 3
Test 4
7000 Test 7
Test 8
6000
HRR (kW)
5000
4000
3000
2000
1000
0
0 1200 2400 3600
Time (s)
Table 26.26 Results of CTICM car tests

Test Peak HRR (kW) Total heat released (MJ) Mass loss of car #1 (kg) Mass loss of car #2 (kg)
2 1208 1758 185 –
3 3476 2100 138 –
4 2159 3080 145 –
7 8310 6670 278 –
8 4073 4090 184 –
9 7500 8890 124 172
10 8230 8380 175 166
874 V. Babrauskas
Table 26.27 Test conditions for sedan vehicles tested by Okamoto et al.
Amount of fuel
Test Windows in tank (L) Ignition point Peak HRR (kW) Total HR (MJ)
A Open 10 Rear wheel splashguard 3512 4950
00
B Closed 10 3034 4860
00
C Closed 20 1856 4930
D Closed, exc. part 10 Left front seat 2395 5040
of left-front window

CTICM two-car tests
8000 Test 9
Test 10
7000
6000
HRR (kW)
5000
4000
3000
2000
1000
0
0 600 1200
Time (s)
Table 26.28 Test conditions for minivan vehicles tested by Okamoto et al.
Amount of fuel
Test Windows in tank (L) Ignition point Peak HRR (kW) Total HR (MJ)
A Closed 10 Rear wheel splashguard 3603 5367
B Closed 10 Right front bumper 3144 5006
C Closed 10 Center of the second row seat – –
D Closed, exc. part 10 Center of the third row seat 4094 5153
of left-front window
is also noteworthy that the total HR values were was ragged and not approximately triangular or
nearly identical for all tests. constant (Fig. 26.93). In Test C, the fire self-
Okamoto et al. [196] later ran tests on minivan extinguished due to dropping oxygen levels
type vehicles, using only one model of vehicle since no windows broke.
(Nissan Serena), but four different test conditions Ohlemiller and Shields [197] tested a number
(Table 26.28). The vehicle weighed 1440 kg and of individual components from a passenger vehi-
had a 2.0 L gasoline-powered engine. Same as cle (a minivan). The components that has a mass
for the sedan vehicles, the HRR development of around 2 kg or less all showed small HRR
www.ebook777.com
Fig. 26.92 HRR results

for automobiles tested by
Okamoto et al.

for minivans tested by Test A
Okamoto et al. 4000 Test B
Test D
3500
3000
2500
2000
1500
1000
500
0
0 1000 2000 3000 4000 5000 6000 7000
Time (s)
values, typically less than 80 kW. Three agents. The non-FR version showed HRR in
components, however, showed substantial HRR excess of 200 kW, while the FR versions devel-
values—an empty plastic fuel tank (8.5 kg), a oped only about 5 kW.
passenger seat (8 kg), and an instrument panel Railway car results were reported by SP [197]
(10.6 kg). The HRR curves for these items are and by Steinert [198]. Figure 26.95 shows a passen-
shown in Fig. 26.94. In a separate study, ger railway car (European type IC train) reported by
Ohlemiller [198] tested one production version SP and an ICE train car by Steinert, who also
of an automotive HVAC unit, along with two published the data labeled as “two halves.” The
experimental versions containing fire-retardant latter comprised two half cars, one being aluminum
876 V. Babrauskas
Fig. 26.94 HRR of some 800

larger components from a Instrument
passenger vehicle panel
700
Passenger
600 seat
500
HRR (kW) 400 Plastic fuel

tank (empty)
300
200
100
0
0 200 400 600 800 1000 1200 1400
Time (s)
Fig. 26.95 HRR of 50,000

railway cars Railway car (ICE train)
45,000 Railway car (IC train)
Railway car (two halves)
40,000
35,000
30,000
HRR (kW)
25,000
20,000
15,000
10,000
5,000
0
0 1000 2000 3000 4000 5000 6000 7000
Time (s)
and the other steel. These were abutted to form one subway cars [205] and half a tram car [169]; these
test specimen. A fire was ignited in the aluminum results are shown in Fig. 26.96. Data on school
car, but did not become rapid until windows failed buses from SP [199] and Steinert [200] are shown
at around 40 min. SP also reported results on two in Fig. 26.97.
www.ebook777.com
Fig. 26.96 HRR of 40,000

subway cars and half a Subway car, F4
tram car
35,000 Subway car, F42
Half of a tram car
30,000
25,000
HRR (kW) 20,000
15,000
10,000
5,000
0
0 500 1000 1500 2000 2500 3000 3500
Time (s)
Fig. 26.97 HRR of school 35,000

buses School bus, SP
School bus, Steinert
30,000
25,000
HRR (kW)
20,000
15,000
10,000
5,000
0
0 1000 2000 3000 4000 5000 6000 7000
Time (s)
A number of researchers have tested portions polyurethane, while the cover was a
of various heavy vehicles. Tests on transport viscose/wool/polyester/polyamide blend for the
seating were done at SP [201]. They measured bus seats and 100 % wool fabric for the
an array of four double bus seats and a similar train seats. These HRR results are shown in
arrangement of train seats. The foam was HR Fig. 26.98.
878 V. Babrauskas
Fig. 26.98 HRR of 2500

seating components of Bus seating
heavy vehicles Train seating
2000
1500
HRR (kW)
1000
500
0
0 200 400 600 800 1000 1200 1400 1600
Time (s)
600
Test 11, TB 133 burner ignition
Test 12, TB 133 burner ignition
500 Test 13, Trash bag ignition
Test 14, Round gas burner ignition
400
HRR (kW)
300
200
100
0
0 200 400 600 800 1000 1200 1400
Time (s)
Fig. 26.99 HRR of Amtrak seats (pair), as tested by NIST, exposed to various ignition sources
NIST conducted tests [202] on a pair of from Amtrak wall/soffit carpeting tested in the
Amtrak seats, presented with various ignition same study (Fig. 26.101). These test specimens
sources; these results are shown in Fig. 26.99. were only 1.0 m wide by 1.5 m high for wall
In the same research study, NIST also tested carpeting, while the test that also added soffit
sleeping Amtrak berths; these results are shown carpeting had an 0.5 m deep carpeted soffit.
in Fig. 26.100. Quite high HRR values were seen Additional test results were obtained for Amtrak
www.ebook777.com
Fig. 26.100 HRR of 1000

Amtrak sleeping berths Test 15, Lower berth
Test 16, Lower berth
800 Test 17, Lower and upper berths
600
HRR (kW) 400
200
0
0 200 400 600 800 1000
Time (s)
Fig. 26.101 HRR of 800

Amtrak wall carpet and Test 18, Wall carpet
wall/soffit carpet Test 19, Wall carpet
specimens Test 20, Wall and soffit carpet
Test 21, Wall and soffit carpet
600
HRR (kW)
400
200
0
0 100 200 300 400 500
Time (s)
window drapes (Fig. 26.102) and compartment mounted on a tandem wheel arrangement. The
door privacy curtains (Fig. 26.103). Amtrak win- HRR curve is given in Fig. 26.105.
dow assemblies are made from polycarbonate Vehicle tires are also prone to be ignited and
glazing material and also have polymeric to burn in tire dumps. The HRR will depend
gasketing and trim; these show substantial HRR directly on the geometry and on the amount of
(Fig. 26.104). tires involved. Some quantitative HRR
Vehicle tires can ignite from an overheated experiments have been reported [204] on
axle and can release a substantial amount of heat experiments done at the Fire Research Station.
if they burn. There is one study in the literature These experiments were for flaming tires, but
which documents such a fire. Hansen [203] most recent tire dump problems have been
burned a pair of 285/80 R22.5 truck tires associated with a smoldering condition and no
880 V. Babrauskas
180
Test 22, Extended
160 Test 23, Contracted
Test 24, Contracted
140
120
HRR (kW)
100
80
60
40
20
0
0 100 200 300 400 500
Time (s)
Fig. 26.102 HRR of Amtrak window drapes
180
Test 25, Extended
160 Test 26, Contracted
Test 27, Contracted
140
120
HRR (kW)
100
80
60
40
20
0
0 100 200 300 400 500
Time (s)
Fig. 26.103 HRR of Amtrak privacy curtains
HRR quantification under these conditions has 100 kW burner, while the HRR for the
been reported. second specimen is shown in Fig. 26.105. The
Tests were also reported on two plastic mud ignition source was a 100 kW burner, and its
guards [205], as used on large tanker trucks. HRR has not been subtracted from the results
One specimen failed to get ignited from a shown.
www.ebook777.com
Fig. 26.104 HRR of 500

Amtrak coach window
Test 28, 50 kW gas burner
assembly
Test 29, Trash bag
400
300
HRR (kW) 200
100
0
0 200 400 600 800
Time (s)
Table 26.29 Characteristics of the SP Runehamer Tunnel tests

Fire growth rate during
Total Peak HRR Time to Total heat linear-growth period
Test Load mass (kg) (MW) peak (s) release (MJ) (MW s1)
T1 380 wood pallets, 74 polyethylene 11,010 201.9 1110 242,000 0.335
pallets
T2 216 wood pallets, 240 PUR foam 6,930 156.6 846 141,000 0.438
mattresses
T3 Mixed goods, comprising plastic and 8,550 118.6 600 131,000 0.273
wood furniture, fixtures, and toys; also
10 large tires
T4 600 cardboard cartons with 18,000 2,850 66.4 444 57,000 0.282
polystyrene cups, 40 wood pallets
For heavy-goods vehicles, the heat content of For all except T4, the goods themselves
the combustibles being hauled is likely to greatly were wrapped with polyethylene film. The
exceed the heat content of the vehicle itself. authors especially noted that the primary period
Thus, a recent research program at SP conducted of fire growth in each case, up to ca. 100 MW
by Ingason and Lönnermark [206] (“Runehamer (66 MW in the case of test T4), was linear and
Tunnel tests”) characterized the HRR of not of a t2 type. These linear-growth rates are
some typical commodities of this type. Four given in Table 26.29. These results are espe-
large-scale tests were conducted (Table 26.29), cially noteworthy since they represent the
with the results shown in Fig. 26.106. The highest HRR fires, of realistic products thus
commodities were arranged as volume 10.45 m far studied. An earlier European research pro-
long, 2.9 m wide, and 4.5 m high, but were not gram [207–209] estimated the HRR of a truck
enclosed by a trailer body. In many cases, the loaded with 2,000 kg of modern upholstered
trailer body is aluminum or tarpaulin, thus furniture; however, these estimated HRR
nearly-free burning may be expected in such values, as derived by several investigators,
worst-case situations. varied widely.
882 V. Babrauskas
Fig. 26.105 HRR for 1000

truck tires and mud guard
900
800
700
600
HRR (kW)
500
400
300
Truck tires
200 Mud guard
100
0
0 500 1000 1500 2000 2500
Time (s)
Table 26.30 Some data obtained at VTT on 14 L polyethylene wastebaskets showing effect of packing density and
basket construction
Basket Filling Filling Peak Total heat
Basket sides mass (kg) Filling type mass (kg) density (kg m3) HRR (kW) released (MJ)
Solid 0.63 Shredded paper 0.20 14 4 0.7
Netted 0.63 Milk cartons 0.41 29 13 3.0
Solid 0.53 Shredded paper 0.20 14 18 7.3
Netted 0.53 Milk cartons 0.41 29 15 5.8
Trash Bags and Containers significantly lower HRR rates. For example,
Mehaffey et al. [210] tested a similar wastebasket
Bench-scale measurements of trash are not read- filled with mixed paper/plastic fuel load and obtain
ily feasible, due to the naturally irregular arrange- a HRR curve which can be approximated as being
ment of these combustibles. There are full-scale 30 kW for 60 s. NIST [140] tested slightly larger,
test results available, however, that can suggest 8.5 L “office style” round polypropylene
appropriate values to be used in different wastebaskets, filled with sheets of newspaper,
circumstances. A small “bathroom size” (6.6 L) totaling about 300 g of newspaper in a 315 g con-
plastic wastebasket stuffed with 12 milk cartons tainer. These gave peak HRR values of 28–35 kW
used at NIST as an ignition source in early HRR and an active burning time of ca. 800 s. Table 26.30
testing [45] was found to show a HRR of about shows some additional data [156], where, over a
50 kW, sustained for about 200 s. certain range, increasing packing density is seen
This value evidently represents a worst-case to increase the heat release rate. Some typical
condition, since most researchers have measured trash-bag fires are shown in Fig. 26.107 [109].
www.ebook777.com
Fig. 26.106 HRR of

representative heavy-goods
vehicle cargo, as 200 T1
determined in SP’s T2
T3
Runehamer Tunnel tests T4
150
HRR (MW) T3
T2 T1
100
T4
50
0
0 500 1000 1500 2000 2500 3000
Time (s)
Fig. 26.107 HRR of 400

trash bags
350 Three sacks, 3.51 kg
Two sacks, 2.34 kg
One sack, 4.1 kg
300
One sack, 1.17 kg
250
200
150
100
50
0
0 100 200 300 400 500 600
Time (s)
Lee has correlated the peak heat release values that when the results are normalized per unit base
according to the effective base diameter and pack- area, a downward trend is seen. The correlations
ing density [109]. Figure 26.108 shows that the total according to packing density should only be con-
burning rate (kW) increases with effective base sidered rough observations, and not firm guidelines.
diameter, but decreases with the tighter packing For design purposes, the range of 50–300 kW
densities. Figure 26.109, conversely, illustrates appears to cover the bulk of the expected fires
884 V. Babrauskas
Fig. 26.108 Peak HRR of 600

trash fires
Packing density values (kg/m3)
500 about 30 kg/m3
about 100 kg/m3
30 kg/m3
400 UC R
Peak qfs (kW)

35
UC A3
300 UC A2 29
29
200
100 kg/m3
UC A1
29
100 UC6.6 SNL 4, 11
102 109 SNL 5,
NBS–C 51
10 84
NBS–F 30
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Effective diameter (m)
Table 26.31 Peak HRR of small wastebaskets 110 sheets (2.7 kg) of crumpled newspaper;
Fuel load these results are shown in Fig. 26.111.
Wastebasket material PS paper NIST also tested [211] 30-gal size (136 L)
Steel 12 8 plastic trash containers made from high-density
Polyethylene 50 30 polyethylene (HDPE) and filled with
Polypropylene 50 40 construction-site debris. The debris included cut
Polystyrene 37 22 pieces of lumber, sawdust, cardboard, paper,
cups, food wrappers and pager bags. The
from normal residential, office, airplane, or similar containers were 515 mm diameter, 700 mm tall
occupancy trash bags and trash baskets. and had a mass of 3.6 kg. The debris totaled
Yamada et al. [102] measured the HRR of 10 kg for each test. Figure 26.112 shows the
6.5–11.8 L wastebaskets made of steel and plas- results for two test replicates.
tic and filled with paper and polystyrene foam Tests have been reported on some very large
trash. The peak HRR values found are shown in (364 L, 96 gal) polyolefin garbage cans
Table 26.31. The authors concluded that the (wheeled, household type) [212]. These were
HRR characteristics could be reasonably well tested empty, and they were ignited with the
represented by one of two paradigms: wood crib specified in UL 1975 [213]. That par-
(1) 30 kW for 600 s; or (2) 50 kW for 300 s. ticular crib weighs 340 g and is ignited with 20 g
NIST conducted tests [200] on trash bags col- of excelsior. Three tests were conducted; two
lected from Amtrak overnight trains. The bags gave fairly similar peak HRR values (2383 and
were about 450 mm diameter and 800 mm high 1942 kW), while the third one was much lower at
and were ignited with a 25 kW burner. Test 977 kW (Fig. 26.113). Such variability is typical
results are shown in Fig. 26.110. Based on these of polyolefin products, when they are tested in an
results, NIST researchers endeavored to create a arrangement where the product can melt and
‘standard’ trash bag by filling the bag with recede from the ignition source.
www.ebook777.com
Fig. 26.109 Trash heat

release rates, normalized 500
Trash bag fires
per unit base area
400
UC 6.6
Peak qfs″ (kW/m2)

300 UC R
UC A2
·
UC A1
200 UC A3
100 SNL 5,
NBS–C
10
NBS–F
SNL 4, 11
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Effective diameter (m)
300
Test 1, 1.8 kg
250 Test 2, 9.5 kg
Test 3, 5.4 kg
Test 4, 6.8 kg
200 Test 5, 7.3 kg
Test 6, 5.0 kg
HRR (kW)
150
100
50
0
0 200 400 600 800 1000 1200 1400
Time (s)
Fig. 26.110 HRR of Amtrak trash bags in NIST tests
Upholstered Furniture method to predict the full-scale HRR. Of all the

occupant goods that can be found in a normal
The HRR of upholstered furniture can be deter- residence, upholstered furniture normally has the
mined in three different ways: (1) by room fire highest HRR, thus knowledge of its performance
testing; (2) by testing in the furniture calorimeter; is essential for many applications.
(3) by conducting bench-scale tests in the Cone Until the 1970s, upholstered furniture used to
Calorimeter and then using a mathematical be made from ‘traditional’ materials. Thus,
886 V. Babrauskas
300
Test Data
250 Average of 3 replicates
200
HRR (kW)
150
100
50
0
0 200 400 600 800 1000 1200 1400
Time (s)
Fig. 26.111 HRR of ‘standard’ Amtrak trash bag, based on crumpled newspaper
500
450 Trash Container 1

Trash Container 2
400
350
300
HRR (kW)
250
200
150
100
50
0
0 200 400 600 800 1000
Time (s)
Fig. 26.112 HRR of 136-L HDPE trash containers filled with construction-site debris
www.ebook777.com
Fig. 26.113 HRR of 2500

364 L (96 gal) PE
garbage cans Test 1
Test 2
2000 Test 3
1500
HRR (kW)
1000
500
0
0 200 400 600 800 1000 1200 1400 1600
Time (s)
during the 1950s and 60s, in the US widely bought by consumers. Specimen F32 is a
furniture commonly had a wood frame, steel sofa made from the same materials. Chair F24
springs, cotton batting padding, and an uphol- illustrates the large improvement in HRR when
stery fabric which was commonly a natural cotton fabric is substituted for polyolefin fabric.
fiber such as wool, silk, or cotton. A fraction of The peak HRR decreases by about 2/3, from
the furniture used latex foam padding instead of 2 MW to 700 kW. Further improvements, at
cotton batting. In earlier-yet times, furniture was present, are not readily available on the retail
commonly stuffed with rubberized horse hair. By market. Contract furniture can be procured to
the 1970s, however, the predominant advanced specifications, however, notably
padding material became polyurethane foam, California TB 133 [213]. The latter limits the
and fabric selection became very wide, including peak HRR to values less than 80 kW, which
both thermoplastic synthetics and natural fibers. will present negligible fire hazard in almost any
The HRR of the modern furniture were found to circumstance.
be many times that of traditional types [214], In the case of the tests discussed above, igni-
apart from the special case of latex foam. The tion was from the flame of a 50 kW burner
latter shows HRR values distinctly higher than placed at the side of the specimen, representing
for polyurethane foam, but the material has a the burning of a small trash can. Such an ignition
finite life and few specimens would survive to source provides the minimum time between igni-
this day. tion and peak HRR. The effect of ignition source
Figure 26.114 illustrates several furniture on the HRR curve has been found to be almost
items tested at NIST [2]. Chair F21 used poly- exclusively that of time shifting—use of smaller
urethane foam complying with the 1975 flames, non-flaming sources, or placing of
California TB 117 standard [215] and polyolefin ignition sources in less vulnerable locations
fabric. A specimen using ordinary polyurethane results in an increase of time to peak HRR
foam gave essentially identical results. This level (Fig. 26.115), but otherwise does not have a
of performance represents a very common, but statistically significant effect on the HRR curve
unfortunately worst-performance furniture item [216–218].
888 V. Babrauskas
Fig. 26.114 HRR of 3500

several upholstered
furniture items tested F32
at NIST 3000
F21
F24
2500

2000
1500
1000
500
0
0 200 400 600 800 1000 1200
Time (s)
Back (C2) Seat (C1) Side (C3) Front (C4)

820 s 880 s 1280 s 2520 s
1237 kw 1349 kw 1346 kw 1271 kw
Heat Release Rote (Kw)
1000
500 (C2)
(C1)
0
0 1000 2000 3000 4000
Time (s)
Fig. 26.115 Effect of ignition source location on the HRR curve of upholstered chairs [214]
Foams with fire retardant chemical additives [219]. A recent study with a very small ignition
(FR) improve the fire performance only if large source compared the performance of furniture
loadings are used. Furniture made for the State of with non-FR foams and with TB117 foams
California had been required to use FR foams using cotton upholstery [220]. Using specially
since 1975, but the loading of FR chemicals constructed, non-commercial furniture for test-
used was very small (3–5 %). For furniture with ing, no effect was found for three-seater sofas,
a HRR high enough to be a room fire hazard, and an effect was only seen for single-seat chairs.
such minimal FR levels have no effect on HRR But the latter were of a design where even the
www.ebook777.com
non-FR version showed HRR values so low Since that time, the materials in use by the
(approx. 400 kW) as to not comprise a room furniture makers changed substantially and,
fire hazard. especially, some highly improved materials
Interestingly, the same study reported test became available to the contract furniture mar-
results for a large number of commercial chairs ket. In addition, predictive techniques readily
and sofas burned for comparison. With few available in the early 1980s were less sophisti-
exceptions, the latter showed peak HRR values cated than those developed more recently. Thus,
in the range of 900–2500 kW (Fig. 26.115), during the course of the European fire research
indicating that the custom-made furniture was program CBUF, two new predictive models
not representative of the retail residential furni- were developed [10, 145]. ‘Model I’ is a rela-
ture market. Furniture made to the 1975 TB117 tively simple model and is described below
standard was actually not intended to have lower briefly. A more advanced model was also devel-
HRR values but, rather, to resist small-flame oped and its details are provided in the above
ignitions. However, studies also showed that it references.
was ineffective for that purpose [221]. During To use the CBUF Model I, Cone Calorimeter
recent years, concerns have emerged that the data must first be obtained at an irradiance of
1975 TB117 standard, while ineffective from a 35 kW m2. A well-controlled specimen prepa-
fire safety point of view, resulted in use of nox- ration method is needed, and this is provided in
ious chemicals which have been found to have ASTM E 1474 [224]. Then, one determines if the
environmental toxicology problems [222]. Con- furniture item is likely to sustain a propagating
sequently, in 2012 the State of California fire, or whether a moderate external flame
replaced the 1975 TB117 regulation with source will simply result in limited burning and
TB117-2012. The latter is a cigarette-ignition no propagation. This is determined from the
(smolder resistance) test and will not require 180 s average of Cone Calorimeter HRR results.
00
use of toxic FR chemicals to meet test If q_ 180 < 65 kW m2, then no propagation is
requirements. assumed to occur; otherwise further calculations
A Cone Calorimeter-based prediction method are made to estimate the peak HRR. The scheme
was proposed by Babrauskas and Walton, based is as follows:
on data obtained in 1982 [223]. This was the If
earliest effort, and was based on a data set com-
prising materials primarily from the 1970s.
00

ðx1 > 115Þ or q35tot > 70 and x1 > 40 or ðstyle ¼ f3; 4g and x1 > 70Þ
then q_ fs ¼ x2 combustible parts of the item (kg); it includes

Else, fabric, foam, interliner, dust cover, etc., but
If x1 < 56 does not include the frame nor any rigid support
then q_ fs ¼ 14:4 x1 pieces.
0:7
Else, q_ fs ¼ 600 þ 3:77x1 And, x2 ¼ 880 þ 500 msoft ðstyle factor AÞ

Δhc, eff 1:4
1:25 00
q35tot
where x1 ¼ msoft ðstyle factor AÞ

q_0035 pk þ q_0035300 0:7 15 þ tig35 0:7 Here, Δhc,eff is the test-average effective heat
of combustion in the Cone Calorimeter
and the subscript 35 denotes that the Cone 00
Calorimeter HRR tests run at a 35 kW m2 (MJ kg1), and q35tot is the total heat released
irradiance. The msoft is the mass of the ‘soft’ ¼
890 V. Babrauskas
Table 26.32 Style factors used in the CBUF model for predicting upholstered furniture heat release rates
Type of furniture Style factor A Style factor B
Armchair, fully upholstered, average amount of padding 1.0 1.0
Sofa, 2-seat 1.0 0.8
Sofa, 3-seat 0.8 0.8
Armchair, fully upholstered, highly padded 0.9 0.9
Armchair, small amount of padding 1.2 0.8
Wingback chair 1.0 2.5
Sofa-bed (convertible) 0.6 0.75
Armchair, fully upholstered, metal frame 1.0 0.8
Armless chair, seat and back cushions only 1.0 0.75
Two-seater, armless, seat and back cushions only 1.0 1.0
at a flux of 35 kW m2. Another correlation mass of the item (kg), that is, everything except
predicts the total heat release: metal parts.
1:5 Finally, the time to peak, tpk (s) for the full-
qtot ¼ 0:9 msoft Δhc, eff þ 2:1 mcomb, tot msoft scale item is estimated as:
where mcomb,tot denotes the total combustible
0:3 00 0:5 00 0:5 0:2

t pk ¼ 30 þ 4900ðstyle factor BÞ msoft q_ pk#2 q_ trough t pk# 1 þ 200
where the ‘peak’ and ‘trough’ notations refer to Wall/Ceiling Lining Materials
the fact that, in the general case, the Cone Calo-
rimeter HRR of furniture composites shows two Combustible interior finish materials are substan-
main peaks and one trough in between them. The tially more difficult to treat than free-standing
style factors are obtained from Table 26.32. combustibles. They cannot be measured in a
With these values computed, a triangular device such as the furniture calorimeter, and
HRR curve can then be constructed. The peak require any full-scale study to be a room fire.
HRR and the time to peak are given directly, The materials cover a large area, but the area of
while the base width of the triangle is determined active flame involvement is generally not pre-
from the calculated total heat release of the dictable, except after flashover, when in many
furniture item. cases it can be assumed that all surfaces are
involved. In the early 1980s, a series of wall
materials was studied by Lee at NIST [15] in
Video Games full-scale room fires, and also in bench-scale,
with the Cone Calorimeter. This work comprised
Edenburn [225] tested the joystick controller
the first attempted correlation between bench
from video game console having a plastic enclo- scale and full scale for wall lining materials.
sure made from ABS (UL 94 V-2 rated). When
For several materials in the test series, which
ignited with a needle flame, the unit showed a
included both cellulosics and plastics, it was
peak HRR of 6.7 kW and a total heat release of
found that, after flashover, the per-unit-area
2.52 MJ. HRR results for the main portion (con-
full-scale heat release rates, were approximately
sole) were not provided.
www.ebook777.com
the same as the values obtained from the Cone Classification [228]. The primary assumptions
Calorimeter. Lee’s work did not yet lead to a in the model are:
predictive method, since no technique for 1. The burning area growth rate and the HRR are
estimating the flame-covered area, A(t) was decoupled.
found. 2. The burning area growth rate is proportional
At about the same time, Babrauskas found to the ease of ignition, i.e. the inverse of the
that full-scale fire development on wall/ceiling time to ignition in small scale.
00
linings could be approximated [226] by the 3. The history of q_ at each location in the full
00
expression q_ bspk =tig , where the HRR value and scale is to be the same as in the Cone
the ignition time were obtained from the Cone Calorimeter test.
Calorimeter. The 1/tig factor effectively The model pays mind to the observation that
represented the growth of A(t), but such a scheme burning patterns on wall/ceilings can be very
was only semi-quantitative. different and, especially, that some products
The first successful quantitative method came stop spreading fire under certain conditions,
with the work of Wickström and Göransson in while others continue. The basic area growth
1987 [227]. The model was based on the premise regimes are illustrated in Fig. 26.117, where the
that the full-scale scenario involves the combus- regimes are marked in Roman numerals. The fire
tible materials located on the walls and ceiling of spread may follow three different routes. At
the ISO 9705 room. Note that the same material points ‘A’ and ‘B’ fire spread may or may not
is expected to be placed on both walls and ceil- continue, based on whether a calculated fictitious
ing. The model uses the principle of area convo- surface temperature is higher than a critical
lution and elaborates on Babrauskas’ assumption value. The calculation is based on data from the
that 1/tig controls the growth of the burning area. Cone Calorimeter. Within the different flame
The model was later extended and extensively spread regimes, the burning area growth rate
validated in the European research program depends on ignitability, i.e. time to ignition in
EUREFIC, EUropean REaction to FIre the Cone Calorimeter. Once the flame spread rate
Fig. 26.116 SwRI test results on commercial residential furniture showing that peak HRR values are primarily in the
range of 900–2500 kW
892 V. Babrauskas
is determined, the HRR is calculated assuming correspond to the 900 kW burner level. In addi-
00
that q_ is the same in small and large scale. This is tion, it was found that the constant had to be
understood to be a simplification. The HRR modified for the 100 and 300 kW time periods.
depends on the actual heat flux level received The agreement between model and prediction
by the product as a function of time. Experience was very good, but only five tests were available
showed, however, that the errors average out and for validation at the huge scale.
can be included in empirical constants. The A second extension was developed by
method is only of moderate difficulty to apply, Sumathipala and coworkers [230, 231]. This
but the description is somewhat lengthy. Details model extends the applicability to the case of
are available [23]. This reference also contains the room fire test studied by Lee [15]. The
graphs illustrating the kind of agreement that is dimensions of that room are almost identical to
obtained between predictions and experiments. the ISO room. The differences arise because
While highly successful for its intended pur- (a) the two burner regimes are 40 and 160 kW,
pose, the EUREFIC model does have notable (b) the burner face size is different, and (c) the
limitations. It: product is normally mounted on walls only,
• Can only treat the standard ISO 9705 room, rather than walls and ceiling. The authors, how-
with the standard doorway for ventilation ever, in their development work, included tests
• Only predicts the ISO 9705 100/300 kW of both rooms in both mounting configurations.
burner The success of these extension confirms that the
• Requires that the material be on both walls basic ideas behind the EUREFIC model
and ceiling are sound and can potentially have flexibility.
• Cannot deal with products that do not ignite in On the other hand, it must be borne in mind
the Cone Calorimeter at a 25 kW m2 that even the extensions are ‘hard-wired’
irradiance. configurations and do not yet approach a tech-
It must be remembered that the primary pur- nique which could be applicable towards user-
pose for developing this model was to predict selected room sizes, burner levels, and product
product performance categories to be obtained configurations.
in the ISO 9705 test, while only using bench- Perhaps the most ambitious model so far for
scale Cone Calorimeter data. For its intended wall/ceiling products has been one developed by
purpose, it has been an unquestionable success. Karlsson and coworkers [232–234]. Karlsson’s
The above limitations indicate that the model incorporates much more of current
EUREFIC model, while a major breakthrough, concepts of plumes, flame length calculations,
was certainly not the final answer to modeling ceiling jets, and similar constructs than does the
needs for wall/ceiling products. Two extensions EUREFIC model. The model has the same ‘hard-
have been proposed to generalize the applicabil- wired’ limitations that the EUREFIC model has
ity of this model. Göransson, one of the in terms of ignition sources, product configura-
developers of the EUREFIC model, proposed tion, and room size being fixed. Another wall/
an extension [229] to encompass a ‘huge-scale’ ceiling model was developed by Quintiere and
room. Such a test room was constructed at VTT. Cleary [235–237] and extended by Janssens and
Its dimensions were 6.75 m by 9.0 m, with a coworkers [238].
ceiling height of 4.9 m. The door opening, 0.8
by 2.0 m high, however, was the same as for the
ISO 9705 room. The burner operation was at the Wardrobes
100 kW level for 10 min, then at 300 kW for
another 10 min, finally at 900 kW for 10 more Information on the HRR of wardrobes is avail-
minutes. An extended model was created for this able from a NIST study [239]. The test
situation by introducing a new set of regimes to wardrobes are illustrated in Fig. 26.118; data
www.ebook777.com
Fig. 26.117 EUREFIC 15
fire spread regimes
10
Burning area (m2)

II V
B VI
5
IV
A
III
2
I
0
0 5 10 15 20
Time (min)
Table 26.33 The HRR properties of wardrobes

Wardrobe Avg. heat of
combustible Clothing and Peak Total heat combustion
Test No. Construction mass (kg) paper (kg) HRR (kW) released (MJ) (MJ kg1)
21 Steel 0 1.93 270 52 18.8
43 Plywood, 12.7 mm thick 68.3 1.93 3100 1068 14.9
41 Plywood, 3.2 mm thick, unpainted 36.0 1.93 6400 590 16.9
42 Plywood, 3.2 mm thick, 1 coat FR 37.3 1.93 5300 486 15.9
paint
44 Plywood, 3.2 mm thick, 2 coats FR 37.3 1.93 2900 408 14.2
paint
61 Particleboard, 19 mm thick 120.3 0.81 1900 1349 17.5
1.22 m 0.61 m
Door
Hanger
rod
1.78 m
Hinged Hinged
door door
Front Side
Fig. 26.118 Configuration of the tested wardrobes

894 V. Babrauskas
7000
Test 21
6000
Test 41
Test 42
5000 Test 43
Test 44
Test 61
4000
3000
2000
1000
0
0 300 600 900 1200
Time (s)
Fig. 26.119 HRR of various wardrobes
Table 26.34 European washing machines tested specimen mass is seen). Thus, while the total
by VTT
heat content of the 19 mm particleboard speci-
Specimen W1 W2 W3 men is high (see Table 26.33), its peak HRR is
Ignition source (kW) 1 1 300–550 quite low, since flame spread and fire involve-
Initial mass (kg) 69.3 69.9 63.3 ment proceed more slowly over a thick material
Mass loss (kg) 10.1 10.4 12.3 (Fig. 26.120).
Peak HRR (kW) 345 431 221
Total heat (MJ) 259 245 383
Washing Machines
are given in Table 26.33 and Fig. 26.119. The
wardrobes were outfitted with a small amount of VTT tested [105] European washing machines.
clothing, or simulated clothing, and some paper. The specimens are described in Table 26.34,
Tests were not run on the clothes items by them- while test results are shown in Fig. 26.120. The
selves. However, since in the case of the steel specimens were extinguished before the ultimate
wardrobe, the only other combustible present peak burning would have occurred. These results
was the paint on the metal, it is reasonable to must not be applied to appliances used in North
assign a value of about 270 kW peak for the America, since European appliance styles are
1.93 kg clothes load. The most important conclu- different from North American ones and also
sion, however, was that, for combustible because local standards are such as to permit
constructions, the peak HRR is inversely depen- appliances of greater flammability in Europe
dent on wardrobe panel thickness (and, by con- (Fig. 26.120). HRR data on North American
trast, no simple connection to combustible washing machines are not available.
www.ebook777.com
Fig. 26.120 HRR of 500

European washing
machines tested by VTT W1
W2
400 W3
300
HRR (kW)
200
100
0
0 600 1200 1800 2400 3000 3600
Time (s)
Windows, Plastic some very simplified assumptions, especially

that flame spread could, in the first approxima-
In applications where vandal resistance is tion, be ignored. Further experience gained with
needed, polycarbonate windows are sometimes additional classes of combustibles, as discussed
used. This material is combustible, and limited above, suggests that such a condition will only
testing was reported by Peacock very rarely hold. Furthermore, the user has no
et al. [240]. The tests indicated that it is hard way of knowing when it might hold. Thus, pru-
to derive an ‘innate’ HRR value. The windows dent design practice should now demand that first
do not burn unless a sustained flame or heat recourse be made to the specific sections above
source is applied. In that case, the HRR of the which may address the modeler’s needs. If they
product increases with increasing severity of the do not, then testing is indicated. For the modeler
ignition source. For a 50 kW exposure source, a wishing to start up a major research activity, the
test window showed an additional 50 kW HRR, schemata outlined for upholstered furniture,
with a burning time of ca. 80 s. For a 200 kW mattresses, and wall/ceiling lining should serve
exposure source, the window peak HRR as illustrations of appropriate starting points in
was about an additional 250 kW, but with a theory and practice. It must be pointed out, how-
longer duration of about 200 s, at progressively ever, that such research programs have proven to
diminishing HRR values. be complex and that quick or inexpensive results
cannot be expected.
Estimating the HRR for General

Combustibles Uncertainty of HRR Measurements
The previous edition of the Handbook suggested As in any engineering measurement, uncertainty
a hypothetical method for estimating the HRR in HRR measurements can be subdivided into:
for general combustibles. This was based on 1. Bias,
896 V. Babrauskas
Table 26.35 The 95 % confidence limits for HRR test apparatuses as determined from recent round robins
Peak HRR Total HRR

Apparatus Year Labs Levels r (%) R (%) r (%) R (%)
Cone calorimeter 2000 4 16 17 23 8 15
ICAL 1999 3 8 56 67 72 118
SBI 1997 16 30 38 54 47 71
Room calorimeter 1994 12 5 65 79 25 41
2. Random error, sometimes termed ‘precision 2. Babrauskas, V., Lawson, J. R., Walton, W. D., and
uncertainty.’ Twilley, W. H., Upholstered Furniture Heat Release
Rates Measured with a Furniture Calorimeter
Bias is properly minimized by use of calibra- (NBSIR 82-2604), U. S. Natl. Bur. Stand. (1982).
tion standards; for HRR testing this often 3. Heskestad, G., A Fire Products Collector for
comprises a metered flow of a calibration gas of Calorimetry into the MW Range (FMRC J. I.
high purity. Another source of bias that can be OC2El.RA), Factory Mutual Research Corp.,
Norwood (1981).
minimized, when appropriate, is specific to 4. Standard Test Method for Fire Testing of Real Scale
oxygen-consumption calorimetry bases Upholstered Furniture Items (ASTM E 1537),
measurements. For most testing, a standard oxy- ASTM, West Conshohocken PA.
gen consumption constant value of 13.1 MJ per 5. Standard Test Method for Fire Testing of Real Scale
Mattresses (ASTM E 1590), ASTM, West
kg of oxygen consumed is used. A small number Conshohocken PA.
of substances of fire-safety interest show oxygen 6. Pipe Insulation: Fire Spread and Smoke Production--
consumption constants substantially different Full-scale Test (NT FIRE 036), NORDTEST, Espoo
from this standard value. If the molecular com- (1988).
7. Upholstered Furniture: Burning Behaviour--Full
position of the substance is known, a correction Scale Test (NT FIRE 032), 2nd ed., NORDTEST,
can always be made to eliminate this source Espoo, Finland (1991).
of bias. 8. Standard Fire Test of Limited-Smoke Cables
Most of the instruments in which the HRR (UL 1685), Underwriters Laboratories, Northbrook,
IL (1991).
measurements are made have been subjected to 9. Hirschler, M. M., Use of Heat Release Calorimetry
round robins (“inter-laboratory trials”) to quan- in Standards, pp. 69-80 in Fire Calorimetry
tify the magnitude of random error that can be (DOT/FAA/CT-95/46), Federal Aviation Adminis-
expected. Comparative values have been com- tration, Atlantic City Intl. Airport, NJ (1995).
10. Sundström, B., ed., Fire Safety of Upholstered Fur-
piled by Janssens [241], as shown in Table 26.35. niture--The Final Report on the CBUF Research
For a number of them, several round robins have Programme (Report EUR 16477 EN). Directorate-
been conducted, thus the data shown are General Science, Research and Development
identified by year. SBI denotes the European Sin- (Measurements and Testing). European Commis-
sion. Distributed by Interscience Communications
gle Burning Item test [242], which is a regulatory Ltd, London (1995).
HRR test for building products that uses two wall 11. Babrauskas, V., and Wickström, U. G., Thermoplas-
panels in a corner configuration, without ceiling. tic Pool Compartment Fires, Combustion and Flame
The values tabulated refer to the 95 % confidence 34, 195-201 (1979).
12. Dahlberg, M., Error Analysis for Heat Release Rate
intervals; standard deviations can be obtained by Measurement With the SP Industry Calorimeter
dividing the figures shown by 2.8. (SP Report1994:29), Swedish National Testing and
Research Institute, Borås (1994).
13. Cooper, L. Y., Some Factors Affecting the Design of
a Calorimeter Hood and Exhaust, J. Fire Prot. Engi-
References neering 6, 99-112 (1994).
14. Fisher, F. L., and Williamson, R. B., Intralaboratory
1. Babrauskas, V., and Peacock, R. D., Heat Release Evaluation of a Room Fire Test Method (NBS-GCR-
Rate: The Single Most Important Variable in Fire 83-421), U.S. Natl. Bur. Stand. (1983).
Hazard, Fire Safety J. 18, 255-272 (1992).
www.ebook777.com
15. Lee, B.T., Standard Room Fire Test Development at 30. Tests for Flammability of Plastic Materials for Parts
the National Bureau of Standards, pp. 29-44 in Fire in Devices and Appliances (UL 94), Underwriters
Safety: Science and Engineering (ASTM STP 882), Laboratories, Northbrook IL.
T. Z. Harmathy, ed., American Society for Testing 31. Dembsey, N. A., Compartment Fire Measurements
and Materials, Philadelphia (1985). and Analysis for Near Field Entrainment, Model
16. Sundström, B., Room Fire Test in Full Scale Validation and Wall Lining Fire Growth (Ph.D. dis-
for Surface Products (Rapport SP-RAPP 1984:16). sertation), Univ. California, Berkeley (1995).
Statens Provningsanstalt, Borås, Sweden (1984). 32. Sherratt, J., and Drysdale, D. D., The Effect of the
17. Surface Products: Room Fire Tests in Full Scale Melt-Flow Process on the Fire Behaviour of
(NORDTEST Method NT FIRE 025). NORDTEST, Thermoplastics, pp. 149-159 in Interflam 2001—
Espoo, Finland (1986). Proc. 9th Intl. Conf., Interscience Communications
18. International Standard--Fire Tests--Full scale room Ltd., London (2001).
test for surface products. ISO 9705:1993(E). Interna- 33. Parker, W. J., Prediction of the Heat Release Rate of
tional Organization for Standardization, Geneva Wood (Ph.D. dissertation). George Washington Uni-
(1993). versity, Washington, DC (1988).
19. Standard Test Method for Room Test of Wall and 34. Hirata, T., Kashiwagi, T., and Brown, J. E., Thermal
Ceiling Materials Assemblies (ASTM E 2257), and Oxidative Degradation of Poly
ASTM Intl., West Conshohocken PA. (methylmethacrylate): Weight Loss,
20. Babrauskas, V., Development of the Cone Calorim- Macromolecules 18, 1410-1418 (1984).
eter--A Bench Scale Heat Release Rate Apparatus 35. Kashiwagi, T., Hirata, T., and Brown, J. E., Thermal
Based on Oxygen Consumption, Fire and Materials and Oxidative Degradation of Poly
8, 81-95 (1984). (methylmethacrylate): Molecular Weight,
21. Standard Test Method for Heat and Visible Smoke Macromolecules 18, 131-138 (1985).
Release Rates for Materials and Products using an 36. Vovelle, C., Delfau, J. L., Reuillon, M., Bransier, J.,
Oxygen Consumption Calorimeter (ASTM E 1354), and Laraqui, N., Experimental and Numerical Study
ASTM, West Conshohocken PA. of the Thermal Degradation of PMMA, pp. 43-66 in
22. International Standard -- Fire Tests -- Reaction to Papers of ITSEMAP International Meeting of Fire
Fire -- Part 1: Rate of Heat Release from Building Research and Test Centers, Avila, Spain (October
Products (Cone Calorimeter method). ISO 5660- 7-9, 1986).
1:1993(E). International Organization for 37. Holland, K. A., and Rae, I. D., Thermal Degradation
Standardization, Geneva (1993). of Polymers. Part 3. Thermal Degradation of a Com-
23. Babrauskas, V., and Grayson, S. J., eds., Heat pound Which Models the Head-to-Head Linkage in
Release in Fires, Elsevier Applied Science Poly(Methyl Methacrylate), Australian J. Chemistry
Publishers, London (1992). 40, 687-692 (1987).
24. Urbas, J., and Luebbers, G. E., The Intermediate 38. Manring, L. E., Thermal Degradation of Saturated
Scale Calorimeter Development, Fire and Materials Poly(methylmethacrylate), Macromolecules
19, 65-70 (1995). 21, 528-530 (1988).
25. Standard Test Method for Determining of Fire and 39. Inaba, A., Kashiwagi, T., and Brown, J. E., Effects of
Thermal Parameters of Materials, Products, and Initial Molecular Weight on Thermal Degradation of
Systems using an Intermediate Scale Calorimeter Poly(methyl methacrylate). Part 1, Polymer Degra-
(ICAL), (ASTM E 1623), ASTM, West dation and Stability 21, 1-20 (1988).
Conshohocken PA. 40. Steckler, K. D., Kashiwagi, T., Baum, H. R., and
26. Babrauskas, V., A Closed-Form Approximation for Kanemaru, K., Analytical Model for Transient
Post-Flashover Compartment Fire Temperatures, Gasification of Noncharring Thermoplastic
Fire Safety J. 4, 63-73 (1981). Materials, pp 895-904 in Fire Safety Science—
27. Kokkala, M., Göransson, U., and Söderbom, J., Five Proc. 3rd Intl. Symp., International Association for
Large-Scale Room Fire Experiments. Project Fire Safety Science. Elsevier Applied Science,
3. EUREFIC Fire Research Program (VTT New York, (1991).
Publications 104), VTT-Technical Research Center 41. McGrattan, K., Hostikka, S., McDermott, R., Floyd,
of Finland, Espoo (1992). R., Weinschenk, C., and overholt, K., Fire Dynamics
28. Schleich, J.-B., and Cajot, L.-G., Natural Fire Safety Simulator Technical Reference G uide. Vol. 1:
for Buildings, pp. 359-367 in Interflam 2001—Proc. Mathematical Model (NISTSP 1018) NIST,
9th Intl. Conf., Interscience Communications Ltd., Gaithersburg MD (2013).
London (2001). 42. Babrauskas, V., Specimen Heat Fluxes for Bench-
29. Simonson, M., Blomqvist, P., Boldizar, A., Möller, scale Heat Release Rate Testing, Fire and Materials
K., Rosell, L., Tullin, C., Stripple, H., and Sundqvist, 19, 243-252 (1995).
J. O., Fire-LCA Model: TV Case Study, Swedish 43. Basic Considerations in the Combustion of Hydro-
National Testing and Research Institute, Borås carbon Fuels in Air (NACA Report 1300), National
(2000).
898 V. Babrauskas
Advisory Committee for Aeronautics, Washington 58. Gore, J., Klassen, M., Hamins, A., and Kashiwagi,
(1957). T., Fuel Property Effects on Burning Rate and Radi-
44. Babrauskas, V., Ignition Handbook, Fire Science ative Transfer from Liquid Pool Flames, pp. 395-404
Publishers/Society of Fire Protection Engineers, in Fire Safety Science—Proc. 3rd Intl. Symp.,
Issaquah WA (2003). Elsevier Applied Science, London (1991).
45. Babrauskas, V., and Krasny, J. F., Fire Behavior of 59. Janssens, M., Cone Calorimeter Measurements of
Upholstered Furniture (NBS Monograph 173), the Heat of Gasification of Wood, pp. 549-558 in
U.S. Natl. Bur. Stand. (1985). Interflam ‘93: Sixth Intl. Fire Conf. Proc.,
46. Kokkala, M., and Heinilä, M., Flame Height, Tem- Interscience Communications Ltd., London (1993).
perature, and Heat Flux Measurements on a Flame in 60. Sorathia, U., Dapp, T., Kerr, J., and Wehrle, J.,
an Open Corner of Walls, Project 5 of the EUREFIC Flammability Characteristics of Composites
fire research programme, Valtion Teknillinen (DTRC SME 89/90), US Navy, David Taylor
Tutkimuskeskus, Espoo, Finland (1991). Research Center, Bethesda MD (1989).
47. Quintiere, J. G., A Simulation Model for Fire 61. Rowen, J. W., and Lyons, J. W., The Importance of
Growth on Materials Subject to a Room-Corner Externally Imposed Heat Flux on the Burning
Test, Fire Safety J. 20, 313-339 (1993). Behavior of Materials, J. Cellular Plastics
48. Parker, A. J., Wenzel, A. B., and Al-Hassan, T., 14, 25-32 (1978).
Evaluation of Passive Fire Protection by Jet Fire 62. Paul, K., unpublished data, RAPRA Technology,
Test Procedure, paper 4-d in 29th Loss Prevention Shawbury, England.
Symp., American Institute of Chemical Engineers, 63. Elliot, P., Whiteley, R. H., and Staggs, J. E., Steady
New York (1995). State Analysis of Cone Calorimeter Data, pp. 35-42
49. Söderbom, J., EUREFIC--Large Scale Tests in Proc. 4th IIntl. Fire and Materials Conf.,
according to ISO DIS 9705. Project 4 of the Interscience Communications Ltd., London (1995).
EUREFIC fire research programme (SP Report 64. Tsantaridis, L., Reaction to Fire Performance of
1991:27). Statens Provningsanstalt, Borås, Sweden Wood and Other Building Products (Ph.D. disserta-
(1991). tion), Kungliga Tekniska Högskolan, Stockholm
50. Lee, B.T., Standard Room Fire Test Development at (2003).
the National Bureau of Standards, pp. 29-44 in Fire 65. Cleary, T. G., and Quintiere, J. G., Flammability
Safety: Science and Engineering (ASTM STP 882), Characterization of Foam Plastics (NISTIR
T. Z. Harmathy, ed., American Society for Testing 4664), Natl. Inst. Stand. Technol., Gaithersburg,
and Materials, Philadelphia (1985). MD (1991).
51. Hasemi, Y., Experimental Wall Flame Heat Transfer 66. Nussbaum, R. M., and Östman, B. A.-L., Larger
Correlations for the Analysis of Upward Wall Flame Specimens for Determining Rate of Heat Release in
Spread, Fire Science and Technology 4, 75-90 the Cone Calorimeter, Fire and Materials
(1984). 10, 151-160 (1986); and 11, 205 (1987).
52. Quintiere, J. G., The Application of Flame Spread 67. Janssens, M., and Urbas, J., Comparison of Small
Theory to Predict Material Performance, J. of and Intermediate Scale Heat Release Data,
Research of the National Bureau of Standards pp. 285-294 in Interflam ‘96, Interscience
93, 61-70 (1988). Communications Ltd, London (1996).
53. 53., Kulkarni, A. K., Kim, C. I., and Kuo, C.H., 68. Orloff, L., Modak, A. T., and Alpert, R. L., Burning
Turbulent Upward Flame Spread for Burning Verti- of Large-Scale Vertical Wall Surfaces, pp. 1345-54
cal Walls Made of Finite Thickness (NIST-GCR-91- in 16th Symp. (Intl. ) on Combustion, The Combus-
597), Natl. Inst. Stand. and Technol., Gaithersburg, tion Institute, Pittsburgh (1976).
MD (1991). 69. Babrauskas, V., Cone Calorimeter Annotated
54. Fang, J. B., and Breese, J. N., Fire Development in Bibliography, 2003 edition, Fire Science Publishers,
Residential Basement Room (NBSIR 80-2120), Issaquah WA (2004).
U.S. Natl. Bur. Stand., Gaithersburg, MD (1980). 70. Beard, A., and Goebeldecker, S., Fire Behaviour of
55. Babrauskas, V., and Williamson, R. B., The Histori- Household Appliances towards External Ignition,
cal Basis of Fire Resistance Testing, Part I, Fire European Fire Retardants Assn., Brussels (2007).
Technology 14, 184-194, 205 (1978). Part II, Fire 71. Ohlemiller, T. J., Shields, J. R., McLane, R. A., and
Technology 14,304-316 (1978). Gann, R. G., Flammability Assessment Methodol-
56. Rhodes, B. T., Burning Rate and Flame Heat Flux for ogy for Mattresses (NISTIR 6497), Nat. Inst. Stand.
PMMA in the Cone Calorimeter (M.S. thesis, Uni- and Technol., Gaithersburg MD (2000).
versity of Maryland). NIST-GCR-95-664. Natl. Inst. 72. Bwalya, A. C., Characterization of Fires in Multi-
Stand. and Technol., Gaithersburg (1994). Suite Residential Dwellings: Phase 1 – Room Fire
57. Hopkins, D. jr., and Quintiere, J. G., Material Fire Experiments with Individual Furnishings (IRC-RR-
Properties and Predictions for Thermoplastics, Fire 302), National Research Council Canada, Ottawa
Safety J. 26, 241-268 (1996). (2010).
www.ebook777.com
73. Klitgaard, P. S., and Williamson, R. B., The Impact 89. Mangs, J., and Keski-Rahkonen, O., Full Scale
of Contents on Building Fires, J. Fire and Flamma- Experiments on Electronic Cabinets (VTT
bility/Consumer Product Flammability Supplement Publications 186), Valtion Teknillinen
2, 84-113 (1975). Tutkimuskeskus, Espoo, Finland (1994).
74. White, J. A. jr, Western Fire Center, Inc., Kelso WA, 90. Mangs, J., and Keski-Rahkonen, O., Full Scale
unpublished test results (2003). Experiments on Electronic Cabinets II (VTT
75. Huczek, J. P., Southwest Research Institute, San Publications 269), Valtion Teknillinen
Antonio TX, unpublished test results (2003). Tutkimuskeskus, Espoo, Finland (1996).
76. Tu, K. M., and Davis, S., Flame Spread of Carpet 91. Keski-Rahkonen, O., and Mangs, J., Maximum and
Systems Involved in Room Fires (NBSIR 76-1013), Minimum Rate of Heat Release during Flashover in
U. S. Natl. Bur. Stand., Washington (1976). Electronic Cabinets of NPPs. Paper presented at Fire
77. Vandevelde, P., and Van Hees, P., Wind Aided Safety in Power Plants and Industrial Installations,
Flame Spread of Floor Coverings, Development SMiRT 13 Post Conference Seminar No. 6,
and Evaluation of Small and Large Scale Tests, Gramado, Brazil. Valtion Teknillinen
pp. 57-67 in Interflam ‘96, Interscience Tutkimuskeskus, Espoo, Finland (1995).
Communications Ltd., London (1996). 92. Rigollet, L., and Mélis, S., Fires of Electrical
78. Ames, S., Colwell, R., and Shaw, K., The Fire Cabinets, Paper no. 023 in 11th Intl. Topical Meeting
Behaviour and Fire Testing of Carpet Used as a on Nuclear Reactor Thermal-Hydraulics (NURETH-
Stair Covering, pp. 69-77 in Interflam ‘96, 11), Avignon, France; publ. by American Nuclear
Interscience Communications Ltd., London (1996). Society, LaGrange Park, IL (2005).
79. Williamson, R. B., and Dembsey, N. A., Advances in 93. Folke, F., Experiments in Fire Extinguishment,
Assessment Methods for Fire Safety, Fire Safety J. NFPA Quarterly 31, 115 (1937).
20, 15-38 (1993). 94. Nilsson, L., The Effect of Porosity and Air Flow on
80. Stroup, D. W., DeLauter, L., Lee, J., and Roadarmel, the Rate of Combustion of Fire in an Enclosed Space
G., Fire Tests of Men’s Suits on Racks (FR 4013), (Bulletin 18), Lund Institute of Technology, Lund,
Nat. Inst. Stand. and Technol., Gaithersburg MD Sweden (1971).
(2001). 95. Yamashika, S., and Kurimoto, H, Burning Rate of
81. Satoh, H., and Mizuno, T., Fire Source Model Based Wood Crib, Rept. of Fire Res. Inst. Japan,
on the Ignited Material and Its Burning Property in No. 41, 8 (1976).
the Early Stages of Fire in Residential Accommoda- 96. Harmathy, T.Z., Experimental Study on the Effect of
tion, Fire Science & Technology 25, 163-188 (2006). Ventilation on the Burning of Piles of Solid Fuels,
82. Simonson, M., Report for the Fire Testing of One Combustion and Flame 31, 259 (1978).
Printer and Two CPUs, (P008664), Swedish 97. Quintiere, J.G., and McCaffrey, B.J., The Burning of
National Testing and Research Institute, Borås Wood and Plastic Cribs in an Enclosure, Vol.
(2000). 1 (NBSIR 80-2054), [U.S.] Natl. Bur. Stand.,
83. Bundy, M., and Ohlemiller, T., Full-Scale Flamma- Washington (1980).
bility Measures for Electronic Equipment (Tech. 98. Fons, W.L., Clements, H.B., and George, P.M.,
Note 1461), Nat. Inst. Stand. and Technol., Scale Effects on Propagation Rate of Laboratory
Gaithersburg MD (2004). Crib Fires, in 9th Symp. (Intl.) on Combustion, The
84. Edenburn, D., Burning Mouse, Albemarle Corp. [n. Combustion Institute, Pittsburgh (1962).
p.] (2003). 99. Delichatsios, M.A., Fire Growth Rates in Wood
85. Bliss, D., and Simonson, M., Fire Performance of IT Cribs, Combustion and Flame 27, 267 (1976).
Equipment Studied in the Furniture Calorimeter, 100. Moore, L. D., Full-scale Burning Behavior of
pp. 171-179 in Interflam 2001—Proc. 9th Intl. Curtains and Drapes (NBSIR 78-1448), [U.S.] Nat.
Conf., Interscience Communications Ltd., London Bur. Stand., Washington (1978).
(2001). 101. Wetterlund, I., and Göransson, U., A Full Scale Fire
86. Steel, J. S., unpublished data, Natl. Inst. Stand. and Test Method for Free-Hanging Curtain and Drapery
Technol., Gaithersburg (1985). Textiles (SP Report 1988:45),Swedish National
87. Zicherman, J. and Stevanovic, A., unpublished test Testing Institute, Borås (1988).
results, Fire Cause Analysis, Inc., Richmond CA, 102. Yamada, T., Yanai, E., and Naba, H., A Study of
(20035). Full-Scale Flammability of Flame Retardant and
88. Babrauskas, V., Harris, R. H., Jr., Gann, R. G., Non-Flame Retardant Curtains, pp. 463-473 in
Levin, B. C., Lee, B. T., Peacock, R. D., Paabo, Proc. 4th Asia-Oceania Symp. on Fire Science &
M., Twilley, W., Yoklavich, M. F., and Technology, Asia-Oceania Assn. for Fire Science &
Clark, H. M., Fire Hazard Comparison of Fire- Technology/Japan Assn. for Fire Science & Engi-
Retarded and Non-Fire-Retarded Products (NBS neering, Tokyo (2000).
Special Publication SP 749), U. S. Natl. Bur. Stand. 103. Urban Wildland Interface Building Test Standards
(1988). (12-7A-5), Fire Resistive Standards for Decks and
900 V. Babrauskas
Other Horizontal Ancillary Structures, California 120. Guide for Smoke and Heat Venting (NFPA 204),
Office of State Fire Marshal, Sacramento (2004). National Fire Protection Assn., Quincy MA (1998).
104. Chow, W. K., Han, S. S., Dong, H., Gao, Y., and 121. Kung, H.-C., Spaulding, R. D., and You, H.-Z.,
Zou, G. W., Full-Scale Burning Tests on Response of Sprinkler Links to Rack Storage Fires.
Heat Release Rates of Furniture, Intl. J. of Engineer- FMS J.I.0G2E7.RA(2). FMRC (1984).
ing Performance-Based Fire Codes 6, 168-180 122. Chicarello, P. J., and Troup, J. M. A., Fire Collector
(2004). Test Procedure for Determining the Commodity
105. Hietaniemi, J., Mangs, J., and Hakkarainen, T., Classification of Ordinary Combustible Products.
Burning of Electrical Household Appliances—An FMRC J.I. 0R0E5.RR. FMRC (1990).
Experimental Study (VTT Research Notes 2084), 123. Yu, H.-Z., and Stavrianidis, P., The Transient Ceil-
Valtion Teknillinen Tutkimuskeskus, Espoo, ing Flows of Growing Rack Storage Fires. FMRC
Finland (2001). J.I. 0N1J0.RA(3). FMRC (1989).
106. NIST, unpublished data. 124. Yu, H.-Z., A Sprinkler-Response-Prediction Com-
107. Tewarson, A., Lee, J.L., and Pion, R.F., Categoriza- puter Program for Warehouse Applications. FMRC
tion of Cable Flammability. Part I. Experimental J.I. 0R2E1.RA. FMRC (1992).
Evaluation of Flammability Parameters of Cables 125. Newman, J., and Troup, J. M. A., The Building
Using Laboratory-scale Apparatus, EPRI Project Calorimeter: FM Global’s Novel Approach to
RP 1165-1, Factory Mutual Research Corp., Large-Scale Fire Testing, NFPA World Safety
Norwood (1979). Conf. and Expo., Las Vegas (2005).
108. Sumitra, P.S., Categorization of Cable Flammability. 126. Yu, H.-Z., RDD and Sprinklered Fire Tests for
Intermediate-Scale Fire Tests of Cable Tray Expanded Polystyrene Egg Crates, FMRC
Installations, Interim Report NP-1881, Research J.I. 0R2E3.RA(1), Factory Mutual Research Corp.,
Project 1165-1, Factory Mutual Research Corp., Norwood MA (1990).
Norwood (1982). 127. Sleights, J. E., SPRINK 1.0—A Sprinkler Response
109. Lee, B.T., Heat Release Rate Characteristics of Computer Program for Warehouse Storage Fires
Some Combustible Fuel Sources in Nuclear Power (M.S. thesis), Worcester Polytechnic Institute,
Plants, NBSIR 85-3195, [U.S.] Nat. Bur. Stand., Worcester MA (1993).
Washington (1985). 128. Lee, J. L., and Dean, R. K., Fire Products Collector
110. Arvidson, M., Potato Crisps and Cheese Nibbles Tests of Polyethylene Terephthalate (PET) Plastic
Burn Fiercely, Brandposten [SP] No. 32, 10-11 Bottles in Corrugated Carton, FMRC J.I. 0N0J6.
(2005). RA070(A), Factory Mutual Research Corp.,
111. Madrzykowski, D., unpublished test results (2012). Norwood MA (1986).
112. Persson, H., Evaluation of the RDD-measuring 129. Lee, J. L., Combustibility Evaluation of Shredded
Technique. RDD-Tests of the CEA and FMRC Stan- Newsprint Commodity and an Improved Polyure-
dard Plastic Commodities (SP Report 1991:04), SP, thane Foam Packaging Product Using the Fire
Borås, Sweden (1991). Products Collector, FMRC J.I. 0K0E6.RANS, Fac-
113. Babrauskas, V., unpublished test results (1997). tory Mutual Research Corp., Norwood MA (1984).
114. Heskestad, G., Flame Heights of Fuel Arrays with 130. Khan, M. M., Evaluation of the Fire Behavior of
Combustion in Depth, pp. 427-438 in Fire Safety Packaging Materials, presented at Defense Fire Pro-
Science—Proc. 5th Intl. Symp., Intl. Assn. for Fire tection Symp., Annapolis (1987).
Safety Science (1997). 131. Hasegawa, H., Alvares, N. J., and White, J. A., Fire
115. Heskestad, G., Flame Heights of Fuel Arrays Tests of Packaged and Palletized Computer
with Combustion in Depth, FMRC J.I. 0Y0J3.RU Products, Fire Technology 35, 294-307 (1999).
(2), Factory Mutual Research Corp., Norwood MA 132. Hasegawa, H., private communication (2000).
(1995). 133. Dillon, S. E., Janssens, M. L., and Garabedian, A. S.,
116. Dean, R. K., Stored Plastics Test Program (FMRC A Comparison of Building Code Classifications and
Serial No. 20269), Factory Mutual Research Corp., Results of Intermediate-Scale Fire Testing of Stored
Norwood MA (1975). Plastic Commodities, pp. 593-604 in Interflam
117. Yu, H.-Z., and Kung, H.-C., Strong Buoyant Plumes 2001—Proc. 9th Intl. Conf., Interscience
of Growing Rack Storage Fires, pp. 1547-1554 in Communications Ltd., London (2001).
20th Symp. (Intl.) on Combustion, Combustion Insti- 134. Roberts, T. A., Merrifield, R., and Tharmalingam, S.,
tute, Pittsburgh PA (1984). Thermal Radiation Hazards from Organic Peroxides,
118. Yu, H.-Z., and Kung, H.-C., Strong Buoyant Plumes J. Loss Prevention in the Process Industries
of Growing Rack Storage Fires, FMRC J.I. 0G2E7. 3, 244-252 (1990).
RA(1), Factory Mutual Research Corp., Norwood 135. Babrauskas, V., unpublished test data (1997).
MA (1984). 136. Yu, H.-Z., and Stavrianidis, P., The Transient Ceil-
119. Commodities and Storage Arrangements, Record ing Flows of Growing Rack Storage Fires, FMRC
66:3, 13-18 (May/June 1989). J.I. 0N1J0.RA(3), Factory Mutual Research Corp.,
Norwood MA (1989).
www.ebook777.com
137. Mitler, H. E., Input Data for Fire Modeling, 151. Kakegawa, S., et al., Design Fires for Means of
pp. 187-199 in Thirteenth Meeting of the UJNR Egress in Office Buildings Based on Full-scale Fire
Panel on Fire Research and Safety, March 13-20, Experiments, pp. 975-986 in Fire Safety Science—
1996 (NISTIR 6030, vol. 1), Nat. Inst. Stand. and Proc. 7th Intl. Symp., International Association for
Technol., Gaithersburg MD (1997). Fire Safety Science (2003).
138. Messa, S., Designing Fires for FIRESTARR, LSF 152. Krasner, L. M., Burning Characteristics of Wooden
Fire Laboratories, Montano Lucino, Italy (2000). Pallets as a Test Fuel (Serial 16437), Factory Mutual
139. Chow, W. K., Zou, G., Dong, H., and Gao, Y., Research Corp., Norwood (1968).
Necessity of Carrying out Full-Scale Burning Tests 153. Babrauskas, V., Pillow Burning Rates, Fire Safety J.
for Post-Flashover Retail Shop Fires, Intl. J. on Engi- 8, 199-200 (1984/85).
neering Performance-Based Fire Codes 5, 20-27 154. Pipe Insulation: Fire Spread and Smoke Production--
(2003). Full-scale Test (NT FIRE 036), NORDTEST, Espoo,
140. Madrzykowski, D., and Kerber, S, Fire Fighting Finland (1988).
Tactics under Wind Driven Conditions: Laboratory 155. Wetterlund, I., and Göransson, U., A New Test
Experiments (TN 1618), Nat. Inst. Stand & Technol., Method for Fire Testing of Pipe Insulation in Full
Gaithersburg MD (2009). Scale (SP Report 1986:33), Swedish National Test-
141. Babrauskas, V., Bench-Scale Predictions of Mattress ing Institute, Borås (1986).
and Upholstered Chair Fires, pp. 50-62 in Fire and 156. Babrauskas, V., Toxic Fire Hazard Comparison of
Flammability of Furnishings (ASTM STP 1233). Pipe Insulations: The Realism of Full-scale Testing
American Society for Testing and Materials, Contrasted with Assessments from Bench-scale
Philadelphia (1994). Toxic Potency Data Alone, pp. 439-452 in Asiaflam
142. Damant, G. H., and Nurbakhsh, S., Heat Release ‘95, Interscience Communications Ltd, London
Tests of Mattresses and Bedding Systems, State of (1995).
California, Bureau of Home Furnishings and Ther- 157. Ahonen, A., Kokkala, M. and Weckman, H., Burn-
mal Insulation, North Highlands, CA (1991). ing Characteristics of Potential Ignition Sources of
143. Holmstedt, G., and Kaiser, I., Brand I vårdbäddar Room Fires (Research Report 285), Valtion
(SP-RAPP 1983:04), Swedish National Testing and Teknillinen Tutkimuskeskus, Espoo, Finland (June
Research Institute, Borås, Sweden (1983). 1984).
144. Andersson, B., Fire Behaviour of Beds and 158. Damant, G., and Nurbakhsh, S., Christmas Trees--
Upholstered Furniture--An Experimental Study What Happens When They Ignite? Fire and
(LUTVDG/ISSN 0282-3756),Lund University, Materials 18, 9-16 (1994).
Dept. of Fire Safety Engineering, Lund, Sweden 159. Babrauskas, V., Chastagner, G., and Stauss, E.,
(1985). Flammability of Cut Christmas Trees, IAAI Annual
145. Babrauskas, V., Baroudi, D., Myllymäki, J., and General Meeting and Conference, Atlantic City NJ
Kokkala, M., The Cone Calorimeter Used for (2001).
Predictions of the Full-scale Burning Behaviour of 160. Evans, D. D., Rehm, R. G., Baker, E. S., McPherson,
Upholstered Furniture, Fire and Materials E. G., and Wallace, J. B., Physics-Based Modeling of
21, 95-105 (1997). Community Fires, pp. 1065-1076 in Interflam 2004,
146. Hansen, R., and Ingason, H., Heat Release Rate Interscience Communications Ltd., London (2004).
Measurements of Burning Mining Vehicles in an 161. Stroup, D. W., DeLauter, L., Lee, J., and Roadarmel,
Underground Mine, Fire Safety J. 61, 12-25 (2013). G., Scotch Pine Christmas Tree Fire Tests
147. Walton, W. D., and Budnick, E. K., Quick Response (FR 4010), Nat. Inst. Stand. and Technol.,
Sprinklers in Office Configurations: Fire Test Gaithersburg MD (1999).
Results (NBSIR 88-3695), [U. S.] Natl. Bur. Stand., 162. Madrzykowski, D., Impact of a Residential Sprinkler
Gaithersburg, MD (1988). on the Heat Release Rate of a Christmas Tree Fire
148. Madrzykowski, D., and Vettori, R. L., Sprinkler Fire (NISTIR 7506), Nat. Inst. Stand & Technol.,
Suppression Algorithm for the GSA Engineering Gaithersburg MD (2008).
Fire Assessment System (NISTIR 4833), Natl. Inst. 163. Jackman, L., Finegan, M., and Campbell, S., Christ-
Stand. Technol., Gaithersburg, MD (1992). mas Trees: Fire Research and Recommendations
149. Madrzykowski, D., Office Work Station Heat (LPR 17:2000), Loss Prevention Council, London
Release Rate Study: Full Scale vs. Bench Scale, (2000).
pp. 47-55 in Interflam ‘96, Interscience 164. Stephens, S. L., Gordon, D. A., and Martin, R. E.,
Communications Ltd., London (1996). Combustibility of Selected Domestic
150. Madrzykowski, D., and Walton, W. D. Cook County Vegetation Subjected to Desiccation, pp. 565-571
Administration Building Fire, 69 West Washington, in Proc. 12th Intl. Conf. on Fire and Forest
Chicago, Illinois, October 17, 2003: Heat Release Meteorology, Society of American Foresters,
Rate Experiments and FDS Simulations (NIST SP Bethesda MD (1994).
1021), Nat. Inst. Stand. & Technol., Gaithersburg
MD (2004).
902 V. Babrauskas
165. Etlinger, M. G., Fire Performance of Landscape 182. Hamins, A., Fischer, S. J., Kashiwagi, T., Klassen,
Plants (M.S. thesis), Univ. California, Berkeley M. E., and Gore, J. P., Heat Feedback to the Fuel
(2000). Surface in Pool Fires, Combustion Science and Tech-
166. Outline of Investigation for Artificial Christmas nology 97, 37-62 (1994).
Trees (Subject 411), 2nd ed., Underwriters 183. Adiga, K. C., Ramaker, D. E., Tatem, P. A., and
Laboratories Inc., Northbrook IL (1991). Williams, F. W., Modeling Pool-Like Gas Flames
167. Babrauskas, V., to be published. of Propane, Fire Safety J. 14, 241-250 (1989).
168. McCaffrey, B., Flame Height, pp. 2-1 to 2-8 in SFPE 184. Adiga, K. C., Ramaker, D. E., Tatem, P. A., and
Handbook of Fire Protection Engineering, 2nd ed., Williams, F., Modeling Thermal Radiation in Open
National Fire Protection Assn., Quincy MA (1995). Liquid Pool Fires, pp. 241-250 in Fire Safety Sci-
169. McCaffrey, B. J. Momentum Implications for Buoy- ence—Proc. 2nd Intl. Symp., International Associa-
ant Diffusion Flames, Combustion and Flame 52, tion for Fire Safety Science, Hemisphere Publishing
149-167 (1983). Corp., New York (1989).
170. Särdqvist, S., Initial Fires: RHR, Smoke Production 185. Koseki, H., and Mulholland, G. W., Effect of Diam-
and CO Generation from Single Items and Room eter on the Burning of Crude Oil Pool Fires, Fire
Fire Tests (LUTVDG/TVBB-3070-SE), Lund Uni- Technology 27, 54-65 (1991).
versity, Dept. of Fire Safety Engineering, Lund, 186. Koseki, H., Boilover and Crude Oil Fire, J. Applied
Sweden (1993). Fire Science 3, 243-272 (1993/1994).
171. Blinov, V. I., and Khudiakov, G. N., Diffusion Burn- 187. Chow, W. K., Necessity of Testing Combustibles
ing of Liquids. U.S. Army Translation. NTIS under Well-developed Fires, J. Fire Sciences (2005).
No. AD296762 (1961). 188. Troitzsch, J. H., Flammability and Fire Behaviour of
172. Hottel, H.C., Review Certain Laws Governing Dif- TV Sets, pp. 979-990 in Fire Safety Science—Proc.
fusive Burning of Liquids, by V. I. Blinov and G. N. 6th Intl. Symp., Intl. Assn. of Fire Safety Science
Khudiakov, Fire Research Abstracts and Reviews (2000).
1, 41-44 (1958). 189. Nam, D.-G., Hasemi, Y., and Kamikawa, D., Inves-
173. Babrauskas, V., Tables and Charts, pp. A-1 to A-17 tigation of an Apartment Fire—Experiments for
in Fire Protection Handbook, 18th ed., National Fire Estimating the Cause and Mechanism of the Fire,
Protection Assn., Quincy, MA (1997). pp. 389-400 in Fire & Materials 2005, Interscience
174. Babrauskas, V., Estimating Large Pool Fire Burning Communications Ltd., London (2005).
Rates, Fire Technology 19, 251-261 (1983). 190. Hoffmann, J. M., Hoffmann, D. J., Kroll, E. C., and
175. Gosse, A., BG Technologies Ltd., private communi- Kroll, M. J., Full Scale Burn Tests of Television Sets
cation (2000). and Electronic Appliances, Fire Technology
176. Putorti, A. D. jr., Flammable and Combustible Liq- 39, 207-224 (2003).
uid Spill/Burn Patterns (NIJ 604-00), National Insti- 191. Shipp, M., and Spearpoint, M., Measurements of the
tute of Justice, U.S. Department of Justice, Severity of Fires Involving Private Motor Vehicles,
Washington (2001). Fire and Materials 19, 143-151 (1995).
177. Modak, A. T., Ignitability of High-Fire-Point Liquid 192. Mangs, J., and Keski-Rahkonen, O., Characteriza-
Spills (EPRI NP-1731), Electric Power Research tion of the Fire Behaviour of a Burning Passenger
Inst., Palo Alto, CA (1981). Car. Part I: Car Fire Experiments, Fire Safety J.
178. Gottuk, D. T., Scheffey, J. L., Williams, F. W., Gott, 23, 17-35 (1994).
J. E., and Tabet, R. J., Optical Fire Detection (OFD) 193. Steinert, C., Experimentelle Untersuchhungen zum
for Military Aircraft Hangars: Final Report on OFD Abbrand-und Feuerubersprungsverhalten von
Performance to Fuel Spill Fires and Optical Stresses Personenkraftwagen, VFDB-Zeitschrift,
(NRL/MR/6180--00-8457), Naval Research Lab., No. 4, 63-172 (2000).
Washington (2000). 194. Ingason, H., Gustavsson, S., and Werling, P.,
179. DeHaan, J. D., The Dynamics of Flash Fires Involv- Brandförsök i en bergtunnel—Naturlig ventilation.
ing Flammable Hydrocarbon Liquids, Amer. Delrapport II (SP AR 1995:45), Swedish National
J. Forensic Medicine and Pathology 17, 24-31 Testing and Research Institute, Borås (1995).
(1996). 195. Okamoto, K., Watanabe, N., Hagimoto, Y., Chigira,
180. Babrauskas, V., COMPF2—A Program for Calculat- T., Masano, R., Miura, H., Ochiai, H., Tamura, Y.,
ing Post-Flashover Fire Temperatures (Tech Note Hayano, K., Maeda, Y., and Suzuki, J., Burning
991), [U. S.] Natl. Bur. Stand., Gaithersburg MD Behavior of Sedan Passenger Cars, Fire Safety J.
(1979). 44, 301-310 (2009).
181. Gore, J. P., Klassen, M., Hamins, A., and Kashiwagi, 196. Okamoto, K., Otake, T., Miyamoto, H., Honma, M.,
T., Fuel Property Effects on Burning Rate and Radi- and Watanabe, N., Burning Behavior of Minivan
ative Transfer From Liquid Pool Flames, Passenger Cars, Fire Safety J. 62, 272-280 (2013).
pp. 395-404 in Fire Safety Science—Proc. 3rd Intl. 197. Ohlemiller, T. J., and Shields, J. R., Burning Behav-
Symp., International Association for Fire Safety Sci- ior of Selected Automotive Parts from a Minivan
ence, Elsevier Applied Science, New York (1991).
www.ebook777.com
(NISTIR 6143), Nat. Inst. Stand. & Technol., Nat. Inst. Stand. & Technol., Gaithersburg MD
Gaithersburg MD (1998). (2003).
198. Ohlemiller, T. J., Influence of Flame-Retarded 212. Zicherman, J. B., Fire Cause Analysis, Berkeley CA;
Resins on the Burning Behavior of a Heating, unpublished tests conducted at the Western Fire
Ventilating and Air Conditioning Unit from a Sports Center, Inc. (2008).
Coupe (NISTIR 6748), Nat. Inst. Stand. & Technol., 213. Fire Tests for Foamed Plastics Used for Decorative
Gaithersburg MD (2003). Purposes (UL 1975), Underwriters Laboratories Inc.,
199. Ingason, H., Gustavsson, S., and Dalhberg, M., Heat Northbrook IL.
Release Rate Measurements in Tunnel Fires 214. Babrauskas, V., Upholstered Furniture Heat Release
(SP Report 1994:08), Swedish National Testing & Rates: Measurements and Estimation, J. Fire
Research Institute, Borås (1994). Sciences 1, 9-32 (1983).
200. Steinert, C., Smoke and Heat Production in Tunnel 215. Flammability Information Package (Contains Tech-
Fires, pp. 123-137 in Proc. Intl. Conf. on Fires in nical Bulletins 116, 117, 121, 133, 106 and 26).
Tunnels (SP Report 1994:54), Swedish National Bureau of Home Furnishings, Dept. of Consumer
Testing & Research Institute, Borås (1994 Affairs, State of California, North Highlands (1987).
201. Göransson, U., and Lundqvist, A., Fires in Buses and 216. Babrauskas, V., Full-Scale Burning Behavior of
Trains: Fire Test Methods (SP Report 1990:45). Upholstered Chairs (Tech Note 1103), [U. S.] Natl.
Swedish National Testing and Research Institute, Bur. Stand., Gaithersburg MD (1979).
Borås (1990). 217. Mitler, H. E., and Tu, K.-M., Effect of Ignition
202. Peacock, R. D., Reneke, P. A., Averill, J. D., Location on Heat Release Rate of Burning
Bukowski, R. W., and Klote, J. H., Fire Safety of Upholstered Furniture, pp. 121-122 in Annual
Passenger Trains. Phase II: Application of Fire Haz- Conf. on Fire Research. Book of Abstracts. October
ard Analysis Techniques (NISTIR 6525), Nat. Inst. 17-20, 1994 (NISTIR 5499), Nat. Inst. Stand. &
Stand. & Technol., Gaithersburg MD (2002). Technol., Gaithersburg MD (1994).
203. Hansen, P. A., Fire in Tyres: Heat Release Rate and 218. Collier, P. C. R., and Whiting, P. N., Timeline for
Response of Vehicles (STF25 A95039). SINTEF Incipient Fire Development (Study Report 194),
NBL, Norwegian Fire Research Laboratory, BRANZ, Judgeford, New Zealand (2008).
Trondheim (1995). 219. Babrauskas, V., Lawson, J. R., Walton, W. D., and
204. Shipp, M. P., Fire Spread in Tyre Dumps, pp. 79-88 Twilley, W. H., Upholstered Furniture Heat Release
in Interflam ‘96. Interscience Communications Ltd., Rates Measured with a Furniture Calorimeter
London (1996). (NBSIR 82-2604), [U. S.] Natl. Bur. Stand.,
205. Murrell, J., and Briggs, P., Developments in Gaithersburg MD (1982).
European and International Fire Test Methods for 220. Janssens, M. L., Gomez, C., Huczek, J. P., Overholt,
Composites Used in Building and Transport K. J., Ewan, D. M., Hirschler, M. M., Mason, R. L.,
Applications, pp. 21-32 in Proc. 2nd Intl. Conf. on and Sharp, J. M., Reducing Uncertainty of
Composites in Fire, Conference Design Consultants, Quantifying the Burning Rate of Upholstered Furni-
Newcastle upon Tyne, England (2001). ture (SwRI Project No. 01.15998), Prepared for
206. Ingason, H., and Lönnermark, A., Heat Release National Institute of Justice, Southwest Research
Rates from Heavy Goods Vehicle Trailer Fires in Institute, San Antonio TX (2012).
Tunnels, Fire Safety J. 40, 646-668 (2005). 221. Medford, R. L., and Ray, D. R., Upholstered Furni-
207. Fires in Transport Tunnels. Report on Full-Scale ture Flammability: Fires Ignited by Small Open
Tests (EUREKA Project EU 499: FIRETUN), Flames and Cigarettes, CPSC, Washington (Oct.
Studiengesellschaft Stahlanwendung e.V., 24, 1997).
Düsseldorf, Germany (1995). 222. Babrauskas, V., Blum, A., Daley, R., and Birnbaum,
208. Proceedings of the International Conference on Fires L., Flame Retardants in Furniture Foam: Benefits
in Tunnels, SP - Swedish National Testing and and Risks, pp. 265-278 in Fire Safety Science—
Research Institute, Borås (1994). Distributed by Proc. 10th Intl. Symp., Intl. Assn. for Fire Safety
Interscience Communications Ltd, London. Science, London (2011).
209. Ingason, H., An Overview of Vehicle Fires in 223. Babrauskas, V., and Walton, W. D., A Simplified
Tunnels, pp. 425-434 in Intl. Conf. on Tunnel Fires Characterization for Upholstered Furniture Heat
and Escape from Tunnels, Madrid (2001). Release Rates, Fire Safety J. 11, 181-192 (1986).
210. Mehaffey, J. R., Craft, S. T., Richardson, L. R., and 224. Standard Test Method for Determining the Heat
Batista, M., Fire Experiments in Furnished Houses, Release Rate of Upholstered Furniture and Mattress
pp. 163-174 in Proc. 4th Intl. Symp. on Fire and Components or Composites Using a Bench-Scale
Explosion Hazards, FireSERT, Univ. Ulster, North- Oxygen Consumption Calorimeter (E 1474-96a).
ern Ireland (2004). American Society for Testing and Materials,
211. Stroup, D. W., and Madrzykowski, D., Heat Release Philadelphia (1996).
Rate Tests of Plastic Trash Containers (FR 4018), 225. Edenburn, D., Burning Video Game System (Tech-
nical Report), Albemarle Corp., [n.p.] (2003).
904 V. Babrauskas
226. Babrauskas, V., Bench-Scale Methods for Prediction 235. Cleary, T. G., and Quintiere, J. G., A Framework for
of Full-Scale Fire Behavior of Furnishings and Wall Utilizing Fire Property Tests, pp. 647-656 in Fire
Linings, SFPE Technical Report 84-10, Society of Safety Science--Proc. of the 3rd Intl. Symp., Elsevier
Fire Protection Engineers, Boston (1984). Applied Science, London (1991).
227. Wickström, U., and Göransson, U., Prediction of 236. Quintiere, J. G., A Simulation Model for Fire
Heat Release Rates of Surface Materials in Large- Growth on Materials Subject to a Room-Corner
Scale Fire Tests Based on Cone Calorimeter Results, Test, Fire Safety J. 20, 313-339 (1993).
J. Testing and Evaluation 15, 364-370 (1987). 237. Quintiere, J. G., Haynes, G., and Rhodes, B. T.,
228. Proceedings of the International EUREFIC Seminar Applications of a Model to Predict Flame Spread
1991, Interscience Communications Ltd, London over Interior Finish Materials in a Compartment,
(1991). J. Fire Prot. Engineering 7, 1013 (1995).
229. Göransson, U., Model, Based on Cone Calorimeter 238. Janssens, M., Grexa, O., Dietenberger, M., and
Results, for Explaining the Heat Release Rate White, R., Predictions of ISO 9705 Room/corner
Growth of Tests in a Very Large Room, pp. 39-47 Test Using a Simple Model, pp. 73-83 in Proc. 4th
in Interflam ‘93: Sixth Intl. Fire Conf. Proc., Intl. Fire and Materials Conf., Interscience
Interscience Communications Ltd., London (1993). Communications Ltd., London (1995).
230. Sumathipala, K., Kim, A. K., and Lougheed, G. D., 239. Lawson, J. R., Walton, W. D., and Twilley, W. H.,
A Comparison of ASTM and ISO Full-scale Room Fire Performance of Furnishings as Measured in the
Fire Test Methods, pp. 101-110 in Proc. Fire and NBS Furniture Calorimeter. Part 1 (NBSIR 83-2787),
Materials, 2nd Intl. Conf., Interscience U.S. Natl. Bur. Stand., Gaithersburg MD (1983).
Communications Ltd, London (1993). 240. Peacock, R. D., Reneke, P. A., Averill, J. D.,
231. Sumathipala, K., Kim, A. K., and Lougheed, G. D., Bukowski, R. W., and Klote, J. H., Fire Safety of
Configuration Sensitivity of Full-scale Room Fire Passenger Trains, Phase II: Application of Fire Haz-
Tests, pp. 237-246 in Proc. Fire and Materials, 3rd ard Analysis Techniques (NISTIR 6525), Nat. Inst.
Intl. Conf., Interscience Communications Ltd, Stand. and Technol., Gaithersburg MD (2002).
London (1994). 241. Janssens, M. L., Heat Release Rate, FORUM Work-
232. Karlsson, B., and Magnusson, S.-E., An Example shop on Measurement Needs for Fire Safety, Nat.
Room Fire Model, pp. 159-171 in Heat Release in Inst. Stand. and Technol., Gaithersburg MD (2000).
Fires, op cit. 242. Smith, D. A., and Shaw, K., Single Burning Item
233. Karlsson, B., Models for Calculating Flame Spread (SBI) Test: The Euroclasses and Transitional
on Wall Lining Materials and the Resulting Heat Arrangements, pp. 1-9 in Interflam ’99, Interscience
Release Rate in a Room, Fire Safety J. 23, 365-386 Communications Ltd., London (1999).
(1994).
234. Magnusson, S. E., and Sundström, B., Combustible
linings and room fire growth – A first analysis, Dr. Vytenis Babrauskas is the President of Fire Science
pp. 45-69 in Fire Safety Science and Engineering and Technology Inc., Issaquah, WA, a company
(ASTM STP 882), American Society for Testing specializing in fire safety research, fire testing issues,
and Materials, Philadelphia (1985). and fire science applications to fire investigations and
litigations.
www.ebook777.com
Calorimetry
27
Marc Janssens
Introduction hazard assessment methods were far from perfect

and received major criticism [6, 7]. Nevertheless,
Heat release rate is the single most important Smith deserves recognition as one of the pioneers
variable in fire hazard assessment [1]. Various of heat release rate calorimetry.
test methods for measuring the heat release With compartment fire hazard assessment
rate of materials and products under different as the primary application, there is a need for
conditions have therefore been developed. This high-quality heat release rate data and, conse-
chapter is dedicated to these test methods. An quently, for devices and methods to measure it
apparatus used for measuring heat release rate accurately. The first of two basic approaches to
is referred to as a calorimeter and the measure- assess the fire hazard of a material consists of an
ment of heat release rate is called calorimetry. experimental evaluation in full scale. Typically,
The importance of heat release rate in fire this approach requires multiple large-scale fire
hazard assessment was first recognized in the tests covering all relevant fire scenarios and
early 1970s by Smith at Ohio State University end-use conditions. The second option is the
[2]. Smith and coworkers developed one of the use of small-scale data, primarily heat release
first small-scale test methods for measuring the rate, in conjunction with a calculation procedure
heat release rate of planar products exposed to to estimate full-scale fire performance. The sec-
radiant heat [3]. They also proposed various ond approach is significantly more versatile, and
procedures to assess compartment fire hazard time- and cost-effective. With the continuous
on the basis of the small-scale data. These improvement of the predictive capability and
procedures ranged from simple calculation accuracy of fire models and calculation
methods [4] to a relatively complex computer methods, the latter has become the preferred
model [5]. This work was initiated at a time approach.
when the most accurate measuring techniques This chapter begins with a brief discussion of
for heat release rate were not available and the oxygen bomb calorimeter, which is used to
when computer fire modeling was still in its measure the maximum amount of heat that can
infancy. Moreover, Smith advocated a practical be released from combustion of a material.
approach based on engineering judgment rather The oxygen bomb calorimeter has significant
than detailed science. Hence, his test and fire limitations. For example, materials and products
are not evaluated under realistic fire conditions.
Also, the total heat released is measured
as opposed to the heat release rate as a function
M. Janssens (*)
Southwest Research Institute, San Antonio, of time, that is, no information is obtained
TX 78238, USA concerning the dynamic behavior of the material.

906 M. Janssens
Various heat release tests have been devel-

oped to address these limitations. These test
methods all rely on one of four measuring
techniques, which are described in detail in the
next section. This is followed by a discussion of 19
the effect on the measurements of various small-

scale calorimeter features and construction
details. In the next section a brief description is 18 14
provided of commonly used calorimeters ranging
in size from small to industrial scale. The chapter 17
concludes with a discussion of the use and appli-
cation of heat release rate data and a section on
uncertainty of heat release rate measurements.
2
3
Oxygen Bomb Calorimetry 1
Oxygen Bomb Calorimeter Test
The maximum amount of heat that can be 9 15

released during combustion of a material can be 11 13
determined in an oxygen bomb calorimeter. An 8 10
oxygen bomb calorimeter consists of a sealed 6 7
4
stainless steel container (the “bomb”) in which 5
a small quantity of material (approximately 1 g) 12

16
is combusted at high pressure (30 bar) in
pure oxygen (Fig. 27.1). The heat released is
1 High pressure cylinder 11 Water container
measured on the basis of the temperature rise of 2 Oxygen admission valve 12 Calorimeter mount
the surrounding water vessel (adiabatic method) 3 Pressure release valve 13 Wire
4 Base cap 14 Plug and socket
or the heat loss needed to keep the water temper-
5 Rubber washer 15 Outer can
ature constant (isoperibol method). Standard 6 Pillar 16 Wooden platform
procedures for measuring the gross heat of com- 7 Pillar 17 Cover plate
8 Support ring 18 Stirrer
bustion of solid materials are described in ASTM 9 Fuse wire 19 Thermometer
D5865 and ISO 1716. 10 Quartz crucible
The building and life safety codes
promulgated by the National Fire Protection Fig. 27.1 Oxygen bomb calorimeter
Association (NFPA) make a distinction between
noncombustible and limited combustible
materials. Limited combustible materials must Gross Versus Net Heat of Combustion
have a potential heat of 8.2 MJ/kg or less as
determined by NFPA 259, Standard Test Method The gross heat of combustion of a solid or liquid
for Potential Heat of Building Materials. fuel is measured in an oxygen bomb calorimeter
According to this method, the potential heat of a as described in the previous section. Because the
material is determined as the difference between cooling water temperature remains close to
the gross heat of combustion of the material ambient during a test, all water vapor generated
measured with an oxygen bomb calorimeter and in the combustion process fully condenses. The
the gross heat of combustion of its residue after measured gross heat of combustion therefore
heating in a muffle furnace at 750 C for 2 h. includes the heat released due to condensation
www.ebook777.com
27 Calorimetry 907
of the water vapor. In practice, combustion

products are usually removed from the system me = ma + mv, Te
at a temperature above the dew point. It is there-
fore more realistic to quantify the potential heat
released in a fire assuming that all water vapor
remains in the gaseous state. The corresponding
heat released per mass unit of fuel burnt is
ma, Ta Qf,l
referred to as the net heat of combustion. It is
equal to the gross heat of combustion measured mv Tv
in an oxygen bomb calorimeter minus the heat of
vaporization of the water in the products of com-
bustion, which is a function of the moisture and
hydrogen content of the fuel: Fig. 27.2 Gas-phase energy balance
Δhc, net ¼ Δhc, gross ð8:936Y H þ Y W ÞΔhv

ð27:1Þ
Sensible Enthalpy Rise Method
where
Δhc,net ¼ Net heat of combustion (kJ/g) Consider the energy balance of a gas-phase con-
Δhc,gross ¼ Gross heat of combustion (kJ/g) trol volume enclosing the flame of a burning
YH ¼ Mass fraction of hydrogen in the fuel (g/g) specimen (Fig. 27.2). Air enters the control vol-
YW ¼ Moisture content of the fuel (g/g) ume at a flow rate ṁa and temperature Ta. The
Δhv ¼ Latent heat of vaporization of water enthalpy of this air can be written as
(2.442 kJ/g at 25 C)
ASTM has developed test standards to deter- ha ¼ h0a þ cp ðT a T 0 Þ ð27:2Þ
mine the moisture content and hydrogen content
where
in a variety of solid fuels, e.g., ASTM D3173 and
ha ¼ Enthalpy of air at temperature T a ðkJ=gÞ
ASTM D5373, respectively.
h0a ¼ Enthalpy of air at reference temperature
The gross heat and net heat of combustion are
T 0 ðkJ=gÞ
usually reported at a standard temperature of 25 C.
cp ¼ Average specific heat of air between T 0
Gross heat and net heat of combustion values for a
and T a ðkJ=g KÞ
wide range of materials can be found in Appendix
T a ¼ Temperature of the air entering the
C, “Fuel Properties and Combustion Data.” combustion zoneðKÞ
T 0 ¼ Reference temperature ðKÞ
Part of the heat flux that strikes the exposed
Techniques for Measuring Heat surface is conducted into the specimen. The
Release Rate resulting heat flow raises the temperature of
the solid and decomposes some fraction into
The development of the oxygen consumption
combustible fuel vapors. The vapors are
technique in the late 1970s was a major break-
generated at a rate ṁv and enter the control
through in the accurate measurement of heat
volume at temperature Tv. Assuming that the
release rate in fire tests. Inferior methods had
specific heat of all gases is approximately con-
been used prior to that. The most practical of the
stant and temperature independent (a reasonable
older methods is still used today for applications
approximation), the enthalpy of the fuel vapors
that do not require the highest accuracy. The older
can be written as
methods and the oxygen consumption technique
(and the related carbon oxide generation tech- hv ¼ h0v þ cp ðT v T 0 Þ ð27:3Þ
nique) are described in this section.
908 M. Janssens
where Q
hv ¼ Enthalpy of volatiles at temperature Tv (kJ/g)
hv0 ¼ Enthalpy of volatiles at reference tempera-
ture T0 (kJ/g) ma ,T0
Tv ¼ Temperature of volatiles entering the com- Combustion chamber me ,T0
bustion zone (K) Gas phase
mv ,T0
The fuel vapors mix with air and are
converted in the flame to products of combustion.
The total flow rate, ṁe, of combustion products,
which includes some excess air, has a tempera- Fig. 27.3 Hypothetical combustion chamber
ture Te and enthalpy given by
Suppose now that the same flow rates of air and
he ¼ h0e þ cp ðT e T 0 Þ ð27:4Þ
volatiles, both at temperature T0, are mixed in a
where hypothetical combustion chamber. Furthermore,
he ¼ Enthalpy of combustion products at tem- assume the combustion reactions in the chamber
perature Te (kJ/g) are identical to those in the flame in Fig. 27.2,
he ¼ Enthalpy of combustion products at refer-
0 and the products of combustion are cooled down
ence temperature T0 (kJ/g) to the reference temperature T0 without condens-
Te ¼ Temperature of combustion products leaving water. This hypothetical situation is shown
ing the control volume (K) in Fig. 27.3.
Te is higher than the mass-weighted average of Application of the first law of thermodynam-
Ta and Tv because of the heat released by combus- ics for the combustion chamber control volume
_ However, only a fraction of in Fig. 27.3 leads to
tion in the flame, Q.
this heat contributes to the temperature rise of the Q_ ¼ m_ a h0a þ m_ v h0v m_ e h0e ð27:6Þ
gases. This fraction is referred to as the convec-
tive fraction of the heat release rate. The where Q_ is the total rate of heat released by
remaining fraction of Q_ is lost and is denoted as combustion in the flame (kW).
Q_ f , 1 . For the most part Q_ f , 1 is lost in the form of Q_ is identical in Figs. 27.2 and 27.3 but is
thermal radiation to the walls of the calorimeter distributed in different ways. By expressing the
(closed configuration) or to the environment heat released per unit mass of volatiles, an effec-
(open configuration). A small part of Q_ f , 1 consists tive heat of combustion can be defined as
of convective and radiative feedback to the fuel
surface. Assuming that gas-phase transients and m_ v Δhc, eff Q_ ð27:7Þ
pressure gradients can be neglected, application or per unit exposed area
of the first law of thermodynamics for the control
00
volume in Fig. 27.2 results in
m_ v Δhc, eff Q_
00
ð27:8Þ
Q_ f , 1 ¼ m_ a ha þ m_ v hv m_ e he ð27:5Þ
where
00
where m_ v ¼ Generation rate of fuel volatiles per unit
Q_ f , 1 ¼ Convection and radiation heat loss rate area of fuel surfaceðg=m2 sÞ
from the flameðkWÞ Δhc, eff ¼ Effective heat of combustionðkJ=gÞ
m_ a ¼ Mass flow rate of air entering the Q_
00
¼ Total rate of heat released per unit area
combustion zoneðg=sÞ
of fuel surfaceðkW=m2 Þ
m_ v ¼ Generation rate of fuel volatilesðg=sÞ
m_ e ¼ Mass flow rate of combustion products Δhc,eff is for the combustion reactions as they take
leaving the control volumeðg=sÞ place in the calorimeter. Δhc,eff must be
www.ebook777.com
27 Calorimetry 909
distinguished from the net heat of combustion, Stack with

Δhc,net. The difference between Δhc,eff and Δhc,net TC hot junction
is very significant for charring materials such as
wood. In an oxygen bomb calorimeter; nearly all
the mass of wood is consumed, leaving a small
fraction of noncombustible ash (usually less than
1 % by mass). The net heat of combustion, Δhc,net,
of dry wood is in the range of 16–18 MJ/kg. When
ΔT
exposed under real fire conditions, only 70–80 %
of the mass is converted to volatiles that burn Specimen
almost completely. The heat of combustion of Heater
the volatiles, Δhc,eff, measured in a small-scale
calorimeter is only 12–13 MJ/kg. A solid char
residue remains, primarily consisting of carbon,
with a net heat of combustion of approximately
30 MJ/kg. In an oxygen bomb calorimeter, most TC cold junction
of this char is also burnt, explaining why Δhc,net
exceeds Δhc,eff by 25–50 %. Even for materials
that do not form a char, Δhc,eff can be significantly Fan and flow control
lower than Δhc,net if combustion of the volatiles in
Fig. 27.4 Sensible enthalpy rise calorimeter
the small-scale calorimeter is incomplete. In this
case, the products of combustion contain measur-
able amounts of combustible components such as calorimeter) can be calculated as 10 kW/3 kJ
CO, soot, unburnt hydrocarbons, and so forth. per g of air ¼ 3.3 g/s. Thus, the air supply in
The ratio of Δhc,eff to Δhc,net is defined as com- the cone calorimeter is at least nine times stoi-
bustion efficiency, χ. For clean-burning gaseous chiometric, or at least 9 3 ¼ 27 times the gen-
fuels, such as methane, χ is close to unity. For eration rate of volatiles. Usually, the ratio is
fuels that produce sooty flames, including gases, much greater. Hence, ṁv is negligible compared
χ can be significantly lower. For example, χ for to ṁa and Equation 27.9 can be approximated as
acetylene is approximately 0.75. χ values for a
number of gases, liquids, and solids are listed in Q_ Q_ f , 1 m_ a cp ðT e T a Þ ð27:10Þ
Chap. 36, “Combustion Characteristics of
Materials and Generation of Fire Products.” This equation is the basis for the sensible
Substitution of Equations 27.2, 27.3, 27.4, and enthalpy method. Heat release rate is calculated
27.6 into Equation 27.5 leads to from the temperature rise Te Ta of the gases
flowing through a calorimeter. A schematic of a
Q_ Q_ f , 1 ¼ c p m_ e ðT e T 0 Þ cp m_ a ðT a T 0 Þ calorimeter based on this principle is shown in
c p m_ v ðT v T 0 Þ Fig. 27.4.
There are a few problems with the practical
ð27:9Þ
implementation of this technique. The main con-
The stoichiometric air-to-fuel ratio ranges cern is that only a fraction of the heat released in
between 3 and 16 for most fuels. Moreover, the flame is used to raise the sensible enthalpy or
small-scale calorimeters are usually operated temperature of the gases. Therefore, it is neces-
with excess air. For example, the standard initial sary to recover or measure the loss term Q_ f , 1 .
flow rate in the cone calorimeter is 30 g/s. Based Some calorimeters have water-cooled walls that
on the oxygen consumption principle (see trap most of the losses. These losses can be
below), the stoichiometric flow rate of air for a estimated by measuring the enthalpy rise of the
10 kW fire (practical upper limit in the cone cooling water. However, due to the additional
910 M. Janssens
hardware and instrumentation, such calorimeters sensible enthalpy rise method. After ignition,
are rather complex and difficult to operate. part of the heat released by a burning specimen
A more popular method relies on a gas burner is transferred by radiation to the enclosure walls.
calibration to determine Q_ f , 1 in the assumption A fraction of this heat is stored in the walls,
that the losses are fuel independent. Defining the causing an increase of their temperature, in turn
loss fraction, χr, by resulting in an enhanced heat transfer with the
air flowing through the box. The result is that, for
Q_ Q_ f , 1 ð1 χr ÞQ_ ð27:11Þ a material that quickly reaches steady burning
conditions, there is a delay for Te to reach the
where χr is the fraction of total heat release rate corresponding steady temperature. A similar
lost by radiation. phenomenon occurs when the heat release rate
The symbol χr is chosen for this fraction from the specimen decreases or after the speci-
because Q_ f , 1 consists primarily of radiation. men burns out and the heat release rate goes back
to zero. Under unsteady burning conditions, Te
If the calorimeter is operated with a constant
constantly lags behind the temperature
airflow rate ṁa, Equation 27.11 can be written as
corresponding to the instantaneous heat release
m_ e c p rate. Several methods have been suggested to
Q_ ðT e T a Þ kðT e T a Þ ð27:12Þ
1 χr mathematically address this problem, but none
are completely satisfactory [8–12].
where k is the calibration constant (kW/K). The substitution method was developed to
The calibration factor, k, is determined from a eliminate problems associated with thermal lag.
gas burner calibration with known Q. _ By repeat- The method requires two runs to determine the
ing the calibration over a range of heat release heat release rate of a material under a given set of
rate levels, k can be determined as a function of Q_ conditions. The first run uses a similar arrange-
or Te. If the specimen is enclosed with the heater, ment as shown in Fig. 27.4. The temperature
Equation 27.12 is still valid, provided a reference difference Te Ta is measured as a function of
temperature Tr is used instead of Ta. The temper- time. The second run uses the same apparatus,
ature difference Tr Ta results from the heat airflow rate, and radiant heat flux. However,
transfer between the heater and the airflow the specimen is replaced by a noncombustible
through the enclosure. Tr is therefore a function dummy specimen and a substitution gas burner.
of heater setting, to be determined via calibration. The flow of gas to the burner is controlled in such
Ed Smith’s rate of heat release test developed a way that the temperature difference Te Ta
at Ohio State University is the most well-known closely follows the curve measured during the
and most widely used calorimeter based on first run. Figure 27.5 shows a schematic of the
the sensible enthalpy rise method [3]. The test substitution run.
method is described in detail in a separate Presumably, the dynamics are identical in both
section. runs. Hence, problems with thermal lag have been
eliminated, and the heat release rate of the speci-
men can be determined from the fuel flow rate to
Substitution Method the burner in the second run. Unfortunately,
implementation of this method is not trivial,
For practical reasons, calorimeters based on the because a sophisticated control system is needed
sensible enthalpy rise method use a closed con- for the second run. Moreover, due to the addition
figuration. The specimen and heater(s) are of substitution runs, the number of tests required
located inside a metal box, which may be (partly) to evaluate a material is doubled.
insulated. The dynamic response of the enclosure The substitution method was first
to changes in the thermal environment creates implemented at Factory Mutual [13]. The appa-
problems in the practical implementation of the ratus was designed to measure the heat release
www.ebook777.com
27 Calorimetry 911
Stack with Stack with

TC hot junction TC hot junction
Compensation
Constant gas burner
ΔT
Noncombustible Heater
ΔT dummy specimen
Substitution
Heater gas burner
Specimen
TC cold junction
TC cold junction
Fan and flow control

Fan and flow control
Fig. 27.6 Compensation calorimeter

Fig. 27.5 Second run with substitution burner
In practice, however, the specimen and burner

rate from roof assemblies. A small-scale substi-
have to be separated to avoid that radiation from
tution calorimeter was developed at the Forest
the burner flame enhances radiant heat flux to the
Products Laboratory [14].
specimen. Hence, the calorimeter enclosure is
not truly isothermal, and the problem remains
unresolved. As with substitution calorimeters,
Compensation Method
the burner flow control system makes compensa-
tion calorimeters rather complex and difficult to
A compensation calorimeter is similar to a sub-
operate. As a result, they are suitable only for
stitution calorimeter, except that the burner is
research and not for routine testing.
operated while a specimen is exposed. A sche-
Compensation calorimeters were developed at
matic is shown in Fig. 27.6. Initially, the burner
the National Bureau of Standards [15, 16] and
flow rate is chosen so that the corresponding heat
Stanford Research Institute [17].
release rate exceeds that of any material to be
tested. During a test, the gas flow rate to the
burner is controlled so that Te Ta remains con-
stant. The heat release rate corresponding to the
Oxygen Consumption Method
reduction in flow rate to the burner is equal to the
In 1917, Thornton showed that for a large num-
heat release rate from the specimen.
ber of organic liquids and gases, a nearly con-
The compensation method also eliminates
stant net amount of heat is released per unit mass
problems with the dynamic response of the calo-
of oxygen consumed for complete combustion
rimeter enclosure. In theory, a compensation
[18]. Huggett found this also to be true for
calorimeter is operated at a constant temperature.
organic solids and obtained an average value
This would resolve another problem associated
for this constant E of 13.1 MJ/kg of oxygen
with the assumption that Q_ f , 1 is fuel indepen-
[19]. This value may be used for practical
dent, whereas in reality it is not (Q_ f , 1 is a strong applications and is accurate with very few
function of the sootiness of the flame). exceptions to within 5 %. Thornton’s rule
912 M. Janssens
implies that it is sufficient to measure the oxygen (E0 and E00 , respectively). Although there is more
consumed in a combustion system in order to variation between different categories, these
determine the net heat released. This is the values are also reasonably constant within a
basis for the oxygen consumption method for given category of fuels or polymers.
measuring heat release rate in fire tests.
The first application of the oxygen consump- Volatiles or Condensed Phase?
tion technique in fire research was by Parker on An interesting question is whether the oxygen
the ASTM E84 tunnel test. [20] During the late consumption technique measures heat release
1970s and early 1980s, the oxygen consumption rate for the volatiles or the solid fuel. Thermal
technique was refined at the National Bureau of methods approximately measure heat release rate
Standards (NBS, currently the National Institute from the volatiles. However, Huggett’s constant
of Standards and Technology, or NIST). The of 13.1 kJ/g is based on the average net heat of
oxygen consumption method is now recognized combustion for a large set of materials. Hence,
as the most accurate and practical technique for one would expect that oxygen consumption cal-
measuring heat release rates from experimental orimetry gives the heat released by the fuel in its
fires. It is widely used throughout the world, for natural state at ambient temperature, because that
both small-scale and large-scale applications. is how the fuel is supplied in an oxygen bomb
calorimeter.
Thornton’s Rule The question can be examined in more detail
The exact value of E for a specific fuel is equal to for some synthetic polymers by comparing the
the net heat of combustion of the fuel divided by net heat of combustion of the polymer to that of
the mass of oxygen needed for complete com- the corresponding monomer. If one were to burn
bustion of a mass unit of fuel. The mass of a monomer in an oxygen consumption calorime-
oxygen required for complete combustion of a ter, the products of complete combustion would
mass unit of fuel can be determined from the be the same as for the corresponding polymer,
stoichiometry of the combustion reactions. provided test conditions are identical. Therefore,
Consider, for example, the following equation measured heat release rate would be the same in
to describe complete combustion of methane: the two cases. However, the net heat of combus-
tion is higher for the monomer. The difference
CH4 þ 2O2 ! CO2 þ 2H2 O ð27:13Þ with the net heat of combustion of the polymer
is the net heat released in the polymerization
This equation indicates that 64 g of oxygen process.
are required for complete combustion of 16.04 g Table 27.2 gives values for the net heat of
of methane. Hence, the mass of oxygen needed to combustion of nine polymers and their
burn 1 g of methane is r0 ¼ 64/16.04 ¼ 3.99 g monomers. The former are taken from Huggett;
O2/g CH4. Because the net heat of combustion of [19] the latter are obtained by adding the heat of
methane is 50.04 kJ/g, the net heat released polymerization as reported in the literature [23].
per mass unit of oxygen consumed is equal to Table 27.2 confirms that the oxygen consump-
E ¼ Δhc,net/r0 ¼ 50.04/3.99 ¼ 12.54 kJ/g O2. tion technique measures net heat release rate of a
An extensive list of E values can be found in solid fuel. The heat release rate from the
Chap. 36; in Tables C2–C4 and in the literature volatiles is always higher, but not by as much
[21, 22]. A summary of average values for dif- as indicated in the last column of the table,
ferent categories of fuels and polymers based on because only a fraction of polymeric fuels
the data in Chap. 36 is given in Table 27.1. This decomposes back into the monomer (see
table also lists values for the amount of heat Chap. 7, “Thermal Decomposition of Polymeric
released per mass unit of CO2 and CO generated Materials”).
www.ebook777.com
27 Calorimetry 913
Table 27.1 E, a, E0 and E00 values for different fuel and polymer categories
Category E (kJ/g O2) α E0 (kJ/g CO2) E00 (kJ/g CO)
Fuels containing C and H
Normal alkanes 12.7 1.079 14.6 12.9
Substituted alkanes 12.6 1.076 14.6 12.8
Cyclic alkanes 12.7 1.069 13.8 11.6
Normal alkenes 13.2 1.070 14.2 12.4
Cyclic alkenes 13.0 1.062 13.4 11.1
Dienes 13.5 1.057 13.5 11.2
Normal alkynes 13.3 1.060 14.0 12.0
Arenes 13.0 1.049 12.4 9.4
Fuels containing C, H and O
Alcohols 13.3 1.104 14.5 12.8
Aldehydes 14.2 1.108 13.3 10.6
Ketones 13.2 1.088 13.2 11.1
Acids 14.2 1.245 9.7 5.4
Esters 13.0 1.118 12.5 9.7
Others 13.9 1.076 12.2 8.9
Fuels containing C, H, N and S
C-H-N fuels 11.5 1.063 15.4 14.1
C-H-S fuels 11.3 1.055 13.1 11.5
Polymeric materials
C and H in the structure 12.5 1.051 12.4 9.5
C, H, O and N in the structure 12.5 1.085 10.9 7.2
C, H and Cl in the structure 12.8 1.124 12.1 9.6
C, H and F in the structure 11.3 1.293 9.2 –
C, H and Si in the structure 13.7 1.083 14.8 13.3
Table 27.2 Net heat of combustion of some polymers and their monomers
Polymer Δhc,net (kJ/g fuel) E (kJ/g O2) Monomer (state) Δhc,net (kJ/g fuel) E (kJ/g O2) ΔE (%)
Polyethylene 43.3 12.65 C2H4 (g) 47.2 13.78 8.9
Polypropylene 43.3 12.66 C3H6 (g) 45.8 13.39 5.8
Polybutadiene 42.8 13.14 C4H6 (L) 44.1 13.56 3.2
Polystyrene 39.9 12.97 C8H8 (L) 40.5 13.19 1.7
Polyvinylchloride 16.4 12.84 C2H3Cl (g) 18.0 14.10 9.8
Polyvinylidene chloride 8.99 13.61 C2H2Cl2 (L) 9.77 14.79 8.7
Polyvinylidene fluoride 13.3 13.32 C2H2F2 (g) 15.6 15.61 17.2
Polymethylmethacrylate 24.9 12.98 C5H8O2 (L) 25.4 13.26 2.2
Polyacrylonitrile 30.8 13.61 C3H3N (L) 32.2 14.25 4.7
Average 13.09 13.99 6.9
Implementation of the Oxygen exhaust duct. At a distance downstream suffi-

Consumption Method cient for adequate mixing, both flow rate and
composition of the gases are measured. A sche-
The basic requirement to use the oxygen con- matic of an oxygen consumption calorimeter is
sumption technique is that all combustion shown in Fig. 27.7. It is not necessary to mea-
products are collected and removed through an sure the inflow of air, provided the flow rate is
914 M. Janssens

measured in the exhaust duct. Therefore, Q_ ¼ E m_ a Y Oa 2 m_ e Y Oe 2 ð27:14Þ
oxygen consumption calorimeters are typically
open, to avoid that part of Q_ f , 1 that is reflected where
by the calorimeter walls and reaches the speci- E ¼ Heat release per mass unit of oxygen
men surface. This would result in an uncon- consumed ð 13:1kJ=gÞ
trolled radiant heat flux, in addition to that Y Oa 2 ¼ Mass fraction of oxygen in the combustion
from the heater. airð0:232 g=g in dry airÞ
The practical implementation of the oxygen Y Oe 2 ¼ Mass fraction of oxygen in the combustion
consumption technique is not straightforward. productsðg=gÞ
Application of Thornton’s rule to the combustion There are three problems with the practical
system shown in Fig. 27.8 leads to the following implementation of Equation 27.14. First, oxy-
equation for the heat release rate: gen analyzers measure the mole (volume) frac-
tion and not the mass fraction of oxygen in a
Flow measurement and fan gas sample. Mole fractions can be converted to
mass fractions by multiplying the mole fraction
with the ratio between the molar mass of oxy-
gen and the molar mass of the gas sample. The
latter is usually close to the molar mass of air
O2 analyzer (29 g/mol). Second, water vapor is removed
from the sample before it passes through a
paramagnetic analyzer, so that the resulting
mole fraction is on a dry basis. Third, flow
meters measure volumetric rather than mass
Specimen flow rates. The volumetric flow rate in the
Heater
exhaust duct, normalized to the same pressure
Fig. 27.7 Oxygen consumption calorimeter and temperature, is usually slightly different
Fig. 27.8 Mass flow rates in oxygen consumption calorimeter
www.ebook777.com
27 Calorimetry 915
from the inflow rate of air because of expan- where

sion due to the combustion reactions. ϕ ¼ Oxygen depletion factor
Parker [24] and Janssens [25] solved these α ¼ Volumetric expansion factor
problems and developed equations for calculat- m_ e ¼ Mass flow rate in the exhaust duct of
ing rate of heat release by oxygen consumption the calorimeterðg=sÞ
for various applications. The equations are a MO2 ¼ Molecular mass of oxygenð28 g=molÞ
Ma ¼ Molecular mass of the combustion
function of the extent of the gas analysis.
airð29 g=mol for dry airÞ
The oxygen concentration must be measured as XHa 2 O ¼ Actual mole fraction of water vapor in
a minimum. However, the accuracy can be the combustion air
improved by adding instrumentation for measur- XCOa
¼ Actual mole fraction of carbon dioxide in
2
ing the concentration of other gases. the combustion air
a
Equations for the two most common XAO2 ¼ Measured mole fraction of oxygen in the
configurations of gas analyzers are discussed combustion air
c
below. Detailed derivations are not repeated XAO2 ¼ Measured mole fraction of oxygen in the
here and can be found in the aforementioned exhaust flow
references. Modified equations to address spe- As the composition of the fuel is usually not
cific circumstances or problems, such as heat known, some average value has to be used for α.
release rate measurements during suppression Complete combustion of carbon in dry air results
experiments, from fires with significant soot in α ¼ 1. If the fuel is pure hydrogen, α is equal
yields, or during experiments conducted in a to 1.21. A commonly-used value for α is
vitiated (reduced O2) environment, can also therefore 1.105. The average value and standard
be found in the literature [26–28]. Derivation deviation for the fuels in Table 27.1 is
of detailed equations for carbon oxide calo- 1.093 0.066. XHa 2 O can be calculated from the
rimetry, a technique that is used extensively relative humidity and temperature in the labora-
by FM Global, can also be found in the tory. Typically it is of the order of 1–2 % in a
literature [29, 30]. Carbon oxide calorimetry temperature-controlled laboratory.1 XCOa
in dry
2
is discussed in section “Carbon Oxide 2
air is approximately 390 ppm. Note that the
Calorimetry”. symbols for oxygen mole fraction measured in
the combustion air (prior to a test) and the
Only O2 Measured exhaust flow include a superscripted A. This is
In this case all water vapor (by a chiller and to make a distinction between the actual and
moisture sorbent) and CO2 (by a chemical sor- measured mole fractions of oxygen, because the
bent) must be removed from the exhaust gas latter are for a dry gas sample.
sample stream before O2 is measured. This Equation 27.15 is expected to be accurate to
leads to the assumption that the sample gas within 10 %, provided combustion is complete;
consists of only O2 and N2. The resulting equa- for example, all carbon is converted to CO2. The
tion for calculating heat release rate is error may be larger if CO or soot production is
ϕ MO a
Q_ ¼ E m_ e 2 1XHa 2 O XCO
a
XAO2
1 þ ϕ ðα 1 Þ Ma 2
ð27:15Þ
1
For example, air at 20 C, 1013 mbar and a relative
humidity of 50 % contains 1.2 % of water vapor by
volume.
with 2
The concentration of carbon dioxide in the atmosphere is
a e measured at the Mauna Loa Observatory in Hawaii. The
XA þ XA average concentration measured in 2010 was 390 ppm.
ϕ ¼ O2 e O2 a ð27:16Þ
The concentration varies annually by about 3–9 ppm, but
1 XAO2 XAO2 the annual average has steadily increased by about 74 ppm
since 1958, when the measurements were first recorded.
916 M. Janssens
considerable, or if a significant amount of com- the need for an expensive sorbent to scrub CO2
bustion products consists of species other than from the gas sample.
CO2 or H2O (e.g., HCl). The error is partly due to If a significant fraction of carbon in the fuel is
the uncertainty of E and α. If more exact values converted to CO instead of CO2, the equations can
are available, accuracy can be improved by using be corrected to take incomplete combustion into
those instead of the generic values of 13.1 kJ/g account. Heat release rate is then calculated from
and 1.105. " e
#
_ 1 ϕ XACO
Q ¼ Eϕ ðECO EÞ e
O2, CO2, and CO Measured 2 XAO2
ð27:18Þ
In this case, only water vapor is trapped before m_ e MO2 e
1 XH2 O XO2
a A
the exhaust gas sample reaches the analyzers. CO 1 þ ϕ ðα 1Þ M a
in many cases is negligible. The rate of heat
release in those cases can be calculated from with
Equation 27.15 with the minor modification that a

Ae Ae Ae Aa
a
XCO is not included in the expression inside O2 1 XCO2 XCO XO2 1 XCO2
XA
2
ϕ¼ a
parentheses. In addition, ϕ is slightly different 1 XA
e
Ae Ae
O2 XCO2 XCO XO2
A
and follows from
ð27:19Þ
a e e a
XAO2 1 XACO2 XAO2 1 XACO2 where
ϕ¼ e
e e ECO ¼ Heat release per mass unit of oxygen
1 XAO2 XACO2 XAO2
consumed for COð17:6 MJ=gÞ
c
ð27:17Þ XACO ¼ Measured mole fraction of carbon monoxide
in the exhaust flow
where One might wonder under what conditions the
a
XACO2 ¼ Measured mole fraction of carbon dioxide CO correction becomes significant. Figure 27.9
in the combustion air shows the ratio of heat release rate obtained by
c
XACO2 ¼ Measured mole fraction of carbon dioxide ignoring CO to the actual heat release rate, as a
in the exhaust flow function of the ratio of measured CO to CO2
Generally, adding CO2 does not greatly mole fractions in the exhaust flow for methane
improve the accuracy of measuring heat release and for a gaseous fuel of composition (CH2O)n.
rate. However, adding a CO2 analyzer eliminates According to Roberts, the molecular formula of

ignoring CO on Q_ error
CH2O
1.04
Ratio of measured to actual Q
CH4
1.03
1.02
1.01
1.00
0.99
0.00 0.05 0.10 0.15 0.20 0.25 0.30
Ae Ae
Ratio of X CO to X C O2
www.ebook777.com
27 Calorimetry 917

the latter represents the thermal degradation Q_ ¼ E m
0 00
_ CO
e
2
_
m CO2 þ E m
a
_ CO
e
products of beech wood [31]. For the CO effect
0 00
examined here, this fuel represents a “worst ¼E m _ e YCOe
2
m _ a YCO
a
2
þE m _ e YCO
e
case” because it contains enough oxygen for

combustion of all hydrogen. Methane gives a ð27:20Þ
practical lower limit for the error, because it is where
the hydrocarbon with the highest hydrogen-to- 0
E ¼ Heat released per mass unit of carbon
carbon ratio. dioxide generatedðkJ=gÞ;
0
There is some experimental evidence that the E ¼ Heat released per mass unit of carbon
yield of CO in underventilated fires reaches an monoxide generatedðkJ=gÞ;
upper limit approximately equal to 0.2 kg of CO _ CO
m e
2
¼ Exhaust flow rate of carbon dioxide ðg=sÞ;
per kg of fuel, when the equivalence ratio _ CO
m a
2
¼ Inflow rate of carbon dioxide from ambient
exceeds unity [32]. For the fuels considered environment ðg=sÞ;
_ CO
m e
¼ Exhaust flow rate of carbon monoxide ðg=sÞ;
here, the limit corresponds to a CO/CO2 mole
fraction ratio of 0.27. Figure 27.9 indicates that,
e
YCO 2
¼ Carbon monoxide mass fraction in the
exhaust flowðÞ;
even under the worst conditions, the error by a
YCO ¼ Ambient carbon monoxide mass
ignoring CO generation is less than 5 %. 2
fraction ðÞ; and
e
YCO ¼ Carbon monoxide mass fraction in the
Carbon Oxide Calorimetry exhaust flowðÞ:
An alternative method for measuring heat release Khan et al. give values for E0 and E00 of
rate is based on the fact that amount of heat 13.3 kJ/g 11 % and 11.1 kJ/g 18 %, respec-
released per mass unit of carbon dioxide and tively (Chap. 36). Practical implementation of
carbon monoxide generated is also relatively Equation 27.20 faces the same challenges as
constant within a category of fuels or polymers that of Equation 27.14. The following equations
(see Table 27.1). This method is particularly can be used to calculate the generation rates of
useful for oxidizers [33]. Application to the com- carbon dioxide and carbon monoxide [34]:
bustion system shown in Fig. 27.8 leads to the
following equation for the heat release rate:

MCO2
e
Aa Aa Ae Ae
XA
CO2 1 X O2 X CO2 1 XO2 X CO _e
m
_ CO
m e
_ CO
m a
¼ 1 XHa 2 O
1 þ ϕð α 1 Þ M a
e e e
2 2
1 XA
O2 XCO2 XCO
A A
ð27:21Þ
with

MCO
e
Aa Aa
XA
CO 1 X O2 XCO2 _e
m
_ CO
m e
¼ 1 XHa 2 O ð27:22Þ
1 þ ϕðα 1Þ Ma
e e e
1 XA
O2 XCO2 XCO
A A
Practical Considerations methods for measuring heat release rate based

The equations presented in the previous sections on oxygen consumption calorimetry [35]. They
are rather complex and it is very easy to make found errors in the equations for 12 of the
mistakes. To illustrate the point, Lattimer and 17 standards. In total, 22 equations were found
Beitel reviewed 17 different standard test to be in error. Because it is difficult to avoid
918 M. Janssens
errors in transcribing to a spreadsheet program supplied to one side of the box, and combustion
even with the correct equations, it is essential to products are removed from the opposite side.
do so very carefully and to double-check the Specimen, heater, and ignition device typically
results. are located inside the box. Advantages of a
Improper setup and maintenance of the gas closed configuration are that airflow rate can be
sampling and analysis system is another common measured at the inlet under clean and soot-free
cause for errors. The proper sorbent columns conditions, the combustion air can be heated, and
have to be installed depending on the configura- the oxygen concentration in the air can be
tion of gas analyzers used. It is necessary to use increased (by adding O2) or decreased
three columns in series when only oxygen is (by adding N2) from ambient. Disadvantages
measured. The first column contains a drying are thermal lag due to heating or cooling of the
agent. The second column contains a sorbent to enclosure walls and uncontrolled radiation feed-
scrub the CO2 from the gas sample. The third back from the enclosure walls to the specimen.
column also contains a drying agent and is nec- To address the first disadvantage, various
essary to remove water vapor that is generated by numerical procedures have been proposed for
the CO2 scrubber. If carbon oxide analyzers are correcting the temperature signal measured with
used, a drying column is all that is needed. calorimeters based on the sensible enthalpy rise
Silica gel is a commonly used drying agent in method [8–12]. These procedures are based on a
chemistry labs. However, it is not suitable for mathematical model of the calorimeter
most oxygen consumption calorimetry consisting of two first-order systems in series.
applications because of the generation of CO2 The first system has a rather small time constant
[36]. Drierite® is commonly used instead of sil- (between 8 and 30 s for various calorimeters) and
ica gel, but some batches seem to have the same is related to the heat capacity of the gases flowing
problem. Ascarite® is the most commonly used through the calorimeter. The second system has a
CO2 scrubbing agent. large time constant (200–930 s for various
Finally, the analyzers are calibrated with calorimeters), which is associated with the heat
certified zero and span gases at the start of each capacity of the calorimeter walls. The correction
testing day and sometimes more than once a day. procedures adjust the output signal for thermal
It is absolutely essential that flows and pressures lag, using discrete forward and inverse Laplace
during these calibrations are the same as during transform techniques.
testing. In spite of the complex calculations, the
resulting correction may not always be accurate
due to the crude mathematical model for the
Factors Affecting Small-Scale Heat calorimeter. A more convenient, and perhaps
Release Measurements equally accurate, correction method relies on an
electronic compensator as described in ASTM
This section examines the effects of some calo- E906 and ASTM E1317. The compensator elec-
rimeter construction details on quality and accu- tronically corrects the output signal of the
racy of the measurements. The discussion results exhaust thermocouples, based on the negative
in some guidelines for building the “ideal” small- feedback of a wall temperature signal.
scale calorimeter for a specific application. The oxygen consumption method also has a
time delay, but with properly adjusted sampling
flows and oxygen analyzer, this delay consists
Open or Closed Configuration almost entirely of the transport time for a gas
sample from the combustion zone to the analyzer
Calorimeters that utilize a measuring technique [37]. Because flow rates in the exhaust duct and
other than oxygen consumption consist of a sampling lines do not change significantly during
closed “box” configuration. Combustion air is a test, this delay time is approximately constant.
www.ebook777.com
27 Calorimetry 919
It can be determined with gas burner calibrations specific heat flux level because a major fraction
and can be easily addressed by shifting the gas of the heat transfer is convective. Moreover, the
analysis data over the appropriate time interval. burner gas and combustion products mix with
To obtain accurate measurements for thin fuel volatiles, which affects burning behavior.
materials that produce a heat release vs. time In short, impinging flames are not desirable as
curve that has the form of a single sharp peak, the external heat source in heat release rate
it is necessary to make corrections for the calorimeters.
response time of the oxygen analyzer [12]. An It is much easier to create constant and
oxygen analyzer is modeled as a first order sys- uniform exposure conditions if the incident heat
tem and its time constant is obtained from gas flux is primarily radiative. Porous gas panels as
burner step response measurements. well as electrical heating elements are used for
The second disadvantage can be eliminated this purpose. The radiant heat flux can be
only by using blackened water-cooled calorime- adjusted by changing the power of the heater or
ter walls. If the walls are allowed to heat or cool by changing the distance between heater and
freely, they emit radiation, which varies with specimen. If the second method is used, practical
time. Part of this radiation reaches the specimen upper and lower limits to the range of radiant
surface and enhances the radiant heat flux from heat flux levels can be created. If the heater is too
the heater in an uncontrolled fashion. Obviously, close to the specimen, convective heat transfer
the need for water-cooled walls makes the appa- becomes significant. Therefore, the upper limit
ratus much more complex and costly. corresponds to the minimum distance that has to
Problems with thermal lag and radiation feed- be maintained in order to ensure predominantly
back to the specimen can be eliminated by using radiative heat transfer. The lower limit is deter-
an open configuration. Solid objects must be mined by the uniformity of the incident radiant
water cooled or sufficiently remote from heater heat flux, which drops with increasing distance
and specimen so that they do not interfere with between heater and specimen. The exact limits
the controlled radiant heat flux to the specimen. depend on the geometrical configuration, the
A closed configuration can be recommended power of the heater, and the degree of nonunifor-
only for specialized applications, for example, mity of the incident heat flux profile that is
to study the effect of oxygen concentration or deemed acceptable.
temperature of the combustion air on heat release Another important aspect is the ability of the
rate and burning behavior. heater to maintain the radiant heat flux at a con-
stant level during a test. If the heater is operated
at a constant power level, incident radiant heat
Type of Heater flux changes during testing. At the start of a test,
a cold specimen is inserted. The specimen acts as
Heat release rates must be measured at constant a heat sink, resulting in a decrease of the heater
heat flux levels over a range that is relevant for temperature and consequently a decrease of the
the fire scenario of interest. The heat flux can be incident radiant heat flux. After ignition, the heat
provided with a gas burner flame in contact with released by the specimen results in an increase of
the specimen or with a radiant panel remote from the heater temperature and incident radiant
the specimen. heat flux.
Incident heat flux from impinging gas burner To maintain incident radiant heat flux during a
flames can be adjusted only over a narrow range. test, it is therefore necessary to keep the temper-
To increase the heat flux from a gas burner, either ature of the heater constant. This is very difficult
the flame size has to be increased, or a fuel with with a gas panel, but relatively straightforward
higher soot yield has to be used. Usually, these for electrical heating elements. With the oxygen
parameters can be adjusted only slightly or not at consumption method, another drawback of using
all. It is very difficult to set and maintain a a gas panel is that its products of combustion
920 M. Janssens
result in an oxygen depletion that is usually much sometimes extinguished by fire retardants or
larger than the oxygen consumed for combustion halogens in the fuel volatiles. A glowing wire is
of the specimen. Thus, small fluctuations in panel not an efficient method for igniting fuel volatiles,
flow can result in significant error of the sometimes leading to poor repeatability. An elec-
measured heat release rate. This “baseline” prob- tric spark remains stable when fire retardants or
lem can be avoided by using a separate exhaust halogens are present. However, it occupies a
system for the heater. small volume, so that the positioning of the
It is clear from the preceding discussion that spark plug is more critical than with other types
an electrical heater is preferable over a gas panel. of ignition pilots.
Two types of electrical heaters are used: high and
low temperature. The former are commonly
tungsten filament lamps that operate at Specimen Size
temperatures close to 2600 K. According to
Wien’s displacement law, peak radiant heat flux The ideal situation would be if small-scale heat
from such lamps is at a much shorter wavelength release rate data could be used directly to predict
than for real fires, with temperatures in the range burning rate in real-scale fires. Unfortunately, the
of 600–1400 K. Piloted ignition studies on minimum specimen size that is required to allow
plastics and wood have shown that these for such a straightforward prediction is not prac-
materials absorb much less radiation in the visi- tical. As described earlier, the burning rate of a
ble and near-infrared range than at higher specimen is a direct function of the net heat flux
wavelengths [38, 39]. On the basis of these transferred to the fuel. The net flux is equal to the
findings, it can be concluded that commercially total of external heat flux, flame convection, and
available low-temperature elements are prefera- flame radiation, minus radiative heat losses from
ble over high-temperature lamps. Such elements the fuel surface and heat losses (or gains) at the
typically operate between 800 and 1200 K, a specimen edges.
range that is representative of real fire exposure The Russian work on the effect of diameter on
conditions. pool fire burning rate by Blinov and Khudiakov
gives some insight into this problem. A detailed
discussion of this work and its implications is
Type of Ignition Pilot given by Drysdale [40]. If the pool diameter is
less than 0.03 m, flame convection is laminar and
Heat release rate tests are usually conducted with burning rate increases with decreasing diameter.
an ignition pilot. The use of a pilot reduces the If the pool diameter exceeds 1 m, flame convec-
variation in time to sustained flaming between tion is turbulent and burning rate is independent
multiple tests conducted under identical test of diameter. There is a transition region between
conditions. Because the duration of the preheat these two limits, with a minimum burning rate
period prior to ignition affects burning rate after for a pool diameter of approximately 0.1 m. This
ignition, use of a pilot also improves repeatabil- work indicates that specimen size in a heat
ity of heat release rate measurements. Further- release rate calorimeter must be at least 1 m for
more, piloted ignition is used because it is the results to be independent of scale. This is
representative of most real fires and conservative indeed not feasible in practice. The Russian
in other cases. pool fire data also indicate that heat transfer at
The ignition pilot in small-scale fire tests the edges becomes excessive at diameters below
consists of a small gas burner flame, a glowing 0.1 m. Therefore, specimen size in small-scale
wire, or an electric spark. An impinging flame calorimeters should be at least 0.1 m.
should not be used because it locally enhances To predict real-scale burning rates,
the incident heat flux to the specimen. Another differences in flame heat transfer, and up to a
problem with pilot flames is that they are lesser extent heat transfer at the edges, have to
www.ebook777.com
27 Calorimetry 921
be considered. Östman and Nussbaum reported polymethylmethacrylate (PMMA), the specimen

ignition and heat release data for 13 materials is to be wrapped with aluminum foil on the sides
and two specimen sizes [41]. Increase in speci- and bottom. The main purpose of the foil is to
men size from 0.1 m 0.1 m to 0.2 m 0.2 m eliminate mass transfer along all boundaries
resulted in a slight reduction of piloted ignition except the exposed face of the specimen. Fur-
time. Average heat release rate over the first thermore, the ISO standard requires all tests be
minute after ignition on a per-unit-area basis conducted in the horizontal orientation with the
increased by approximately 12 % at exposure stainless steel retainer frame.
levels exceeding 25 kW/m2. Larger increases Toal et al. tested several materials with and
were observed at the 25 kW/m2 exposure level without foil wrapping, and with and without the
and for peak heat release rate. retainer frame [43, 44]. They found that the
Janssens and Urbas presented a comparison of retainer frame reduces peak heat release rate,
heat release rate data for nine wood products and lengthens the burning time. This is to be
obtained in the cone and an intermediate-scale expected because the retainer frame is a rela-
calorimeter [42]. A 100-fold increase in speci- tively large mass of steel that acts as a heat
men size resulted in only a 10 % increase of the sink, reducing the energy transferred to the
heat release rate. This modest effect can be specimen.
explained by the fact that the heat feedback Urbas and Sand were also concerned with the
from the flame is relatively insensitive to speci- heat sink effect of the retainer frame [45]. They
men area for testing in the vertical orientation, in designed an alternative retainer frame, composed
particular for materials, such as wood, that do not of an insulating collar made of medium-density
produce very luminous flames. or high-density refractory material. Their conclu-
Depending on the specimen size in a small- sion was that the best edge conditions were
scale test, there is a limit on the degree of non- obtained using the insulating frame with insula-
uniformity and irregularity of the product being tion material that most closely resembles the
tested, if the test conditions are to be representa- specimen in thermal properties. Researchers at
tive of end-use conditions. Therefore, there FM Global proposed a similar approach to
might be some merit in choosing a specimen address edge effects in a small-scale
size that exceeds the minimum of 0.1 m. How- calorimeter [46].
ever, the main trade-off is that a larger specimen Babrauskas et al. conducted a very extensive
requires a larger and more powerful heater to study of the effects of specimen edge conditions
achieve uniform incident radiant heat flux to the on heat release rate [47]. The objective of this
specimen. It should be recognized that, no matter study was to further examine the issues raised
what the specimen size is, there are assemblies by Toal et al. and by Urbas and Sand, and to
and composites for which it is not possible to develop definitive recommendations. Specimens
prepare representative small-scale specimens. of 10 materials were tested in the horizontal
Intermediate-scale or full-scale tests are needed orientation at 50 kW/m2 using three
to evaluate the fire performance of such configurations: without retainer frame, with
assemblies and composites. retainer frame, and with an insulated retainer
frame akin to that developed by Urbas and
Sand [45]. All specimens were wrapped in alu-
Edge Effects minum foil. The study concluded that the use of
an insulated frame gives heat release rate values
An issue that is closely related to specimen size is that are slightly closer to the expected true
that of edge effects. These effects have been values. However, the insulated frame makes
studied extensively in the cone calorimeter. the test procedure significantly more compli-
ASTM and ISO standards of the cone calorimeter cated, so that it is not recommended for routine
prescribe that, except for calibrations with testing.
922 M. Janssens
If the standard retainer frame is used, oxygen. However, specialized studies have been
Babrauskas et al. recommended that heat release conducted to evaluate the effect of ventilation
rate data be expressed on the basis of an effective and vitiation and to determine the ‘Limiting Oxy-
exposure area of 0.0081 m2. The standard gen Concentration’, which is an important
retainer frame reduces the actual exposed area parameter in the design of fire protection systems
from 0.1 m 0.1 m to 0.094 m 0.094 m, or that rely on a reduction of oxygen concentration
from 0.01 to 0.0088 m2. The recommendation by in the room [49, 50]. Such studies require a
Babrauskas et al. [47] to further reduce the closed configuration.
exposed area to an effective value of 0.0081 m2
indicates that the heat sink effect of the retainer
frame reduces heat release rate values by approx- Other Measurements
imately 8 %.
Östman and Tsantaridis tested 11 products in Heat release rate calorimeters often include addi-
the cone calorimeter in the horizontal orientation tional instrumentation to measure parameters that
at 50 kW/m2, with and without the retainer frame are important in characterizing the fire perfor-
[48]. They also found that the use of the retainer mance of materials. Perhaps the most important
frame results in a reduction of heat release rate additional measurement is that of mass loss rate.
greater than what can be explained by the reduc- Most calorimeters can be provided with a load
tion of the exposed area. For the average heat cell to measure specimen mass loss, but this can
release over the first 3 min following ignition, be very difficult in a closed configuration. Mass
they found an average reduction of 8 %, identical loss rate is obtained from numerical differentia-
to Babrauskas et al. [47] However, for maximum tion of the mass loss measurements. Smoke
heat release rate, they found reductions as high as meters are added to measure smoke obscuration
25 %. in the exhaust duct. Both white light and laser
It can be concluded from these studies that the light systems are being used. Toxic gas species
specimen holder configuration in a small-scale can be measured in the exhaust duct with addi-
heat release rate test may have a significant effect tional gas analysis equipment. Such equipment
on the measurements. This effect should be ranges from standard infrared CO and CO2
addressed if the test data are used to predict analyzers to complex online Fourier transform
performance in real fires. infrared (FTIR) instrumentation. Whether instru-
mentation can be added depends mainly on the
design and construction details of the calorimeter.
Specimen Orientation
Products do not necessarily have to be tested in Commonly Used Small-Scale

the same orientation as they are used. For practi- Calorimeters
cal reasons, the preferred orientation for small-
scale testing is horizontal facing upward. The Ohio State University (OSU) Calorimeter
vertical orientation might be preferable for
collecting specialized data for research purposes. This apparatus, originally designed by Ed Smith
at Ohio State University, is one of the most
widely used and best-known small-scale
Airflow calorimeters [3]. The test method was first
published as a proposed ASTM standard in
Standard rate of heat release test methods are 1980. In 1983, it was adopted as ASTM E906.
operated under overventilated conditions. Plenty The standard was recently amended to include
of excess air is supplied, so that the two configurations of the test apparatus. Config-
measurements are not affected by lack of uration A is that which the Federal Aviation
www.ebook777.com
27 Calorimetry 923
Fig. 27.10 Ohio State

10
University (OSU)
calorimeter
8
ΔT
7 3
6
5
4
2
1
1 Air supply fan 6 Heating elements

2 Main flow control 7 Gas pilot
3 By-pass flow control 8 Specimen and holder
4 TC cold junctions 9 Baffle plate
5 Air distributor plate 10 TC hot junctions
Administration (FAA) uses for assessing aircraft located symmetrically along a diagonal of the
cabin materials at a radiant heat flux of 35 kW/m2. stack cross section, above the baffle plate. The
The test procedure in this configuration is also cold junctions are located below the air distribu-
described in the FAA Aircraft Material Fire Test tor plate. An electrical compensator is used to
Handbook [51]. Configuration B is the original correct the temperature signals for thermal lag.
configuration. Both configurations rely on ther- The factor k in Equation 27.12 is obtained from
mopile measurements. Several laboratories have line burner calibration runs.
modified the OSU calorimeter to measure heat
release rate based on oxygen consumption. Configuration. Heater and specimen are located
inside a box with approximate dimensions of
Thermopile Versions 0.2 m 0.41 m 0.64 m. The side walls of
A schematic view of the apparatus is shown in the box are insulated, and the hollow top wall
Fig. 27.10. The apparatus consists of an insulated section is cooled with air.
metal box. The conical wall section between the
combustion chamber and the stack is hollow. Air Heater. The vertical radiant heat source
flows through this cavity and mixes with the com- measures approximately 0.3 m 0.3 m and
bustion products downstream of the thermocouple consists of four silicon carbide heating elements.
hot junctions. However, recovery of the wall heat A steel masking plate is located in front of the
losses is not fully accomplished. The main features elements to improve uniformity of the incident
of the OSU apparatus are described below. heat flux distribution over the specimen. The
maximum incident heat flux to a vertical speci-
Measuring technique. Heat release rate is deter- men is approximately 65 kW/m2.
mined by the sensible enthalpy rise method. The
temperatures of inflowing air and outflowing Ignition pilot. The optional ignition source is a
gases are measured with a thermopile of three pilot flame of 2 mL/s methane, premixed with
type K thermocouples. The hot junctions are 14 mL/s air. The pilot flame either impinges on
924 M. Janssens
the specimen at the bottom (point ignition), is material qualification. Improved flammability
located in the gas phase at the top of the speci- standards and requirements for airplane cabin
men (pilot ignition), or is not used. interior materials based on ASTM E906 first
went into effect in 1986 [52]. The limits for
Specimen size and orientation. For testing in the acceptance were based on heat release rate
vertical orientation, specimens with an exposed measured at a radiant heat flux level of
area of 0.15 m 0.15 m are positioned parallel 35 kW/m2. Peak heat release rate could not
to the heating elements. Specimens can be tested exceed 100 kW/m2, and average heat release
in the horizontal orientation with the aid of an rate over the first 2 min following ignition had
aluminum reflector foil, which reflects the radia- to be 50 kW/m2 or less.
tion from the heating elements to the specimen. Originally, the test method used by the FAA
In this case, the maximum radiant heat flux is was identical to ASTM E906. More recently,
reduced to 50 kW/m2 and the specimen size is some significant modifications have been made
0.11 m 0.15 m. The use of the reflector plate is [11]. The FAA method now uses a thermopile of
awkward and cumbersome, so that testing in the five thermocouples, a lighter specimen holder,
horizontal orientation with the OSU apparatus is and a modified test procedure to minimize
not recommended. problems associated with thermal lag [52]. The
FAA criteria for acceptance were revised in 1990
Airflow. Total airflow rate is 40 L/s, of which to 65 kW/m2 for peak heat release rate during the
only 10 L/s passes through the combustion cham- 5 min test and to 32.5 kW/m2 for average heat
ber and the remaining 30 L/s flows through the release rate over the first 2 min following
upper hollow wall section. Nevertheless, the air- ignition [52].
flow rate through the combustion chamber
contains enough oxygen to feed a 36 kW fire. Oxygen Consumption Versions
Because the heat release rate from test specimens When oxygen consumption calorimetry became
rarely exceeds 20 kW, burning conditions in the the preferred method for measuring heat release
OSU apparatus are always overventilated. The rate, fire research laboratories in the United
airflow rates are measured accurately with stan- States, Canada, and Sweden modified their OSU
dard orifices. apparatus. These modifications typically
consisted of the elimination of the original ther-
Additional measurements. The ASTM E906 mopile, the addition of a gas sampling probe and
standard does not include a mass loss measure- gas analysis equipment, and some adjustments to
ment but has a smoke measuring system with a the airflow rates [6, 53–55].
white light source in the stack. The Forest Products Laboratory (FPL) made
The FAA established a committee in 1978 to two additional significant modifications [54]. An
examine the factors affecting the ability of aircraft auxiliary heat flux meter was added beneath the
cabin occupants to survive in a postcrash environ- specimen to monitor incident radiant heat flux
ment. The committee recommended research to during a test. Measurements obtained with this
evaluate the fire performance of cabin materials auxiliary meter indicated that the incident radiant
and development of a method using radiant heat heat flux to a burning wood specimen increases
for testing cabin materials. As a result, the FAA significantly during a test. For example, the inci-
conducted an extensive series of full-scale fire tests dent radiant heat flux to a Douglas fir plywood
and evaluated numerous small-scale tests for their specimen at the end of a 10-min burning period
capability to provide results that correlate well increased by 20 % over the 35 kW/m2 baseline.
with full-scale performance. This is due to the fact that the heater elements in
The OSU apparatus, standardized as ASTM the OSU calorimeter are supplied with constant
E906, was found to be the most suitable for power and are not temperature controlled, and
www.ebook777.com
27 Calorimetry 925
that the calorimeter walls are allowed to heat up Configuration. Cone heater, spark igniter,
(or cool down) during testing. specimen holder, and load cell are located
The fact that exposure conditions in the OSU beneath the hood. The standard configuration is
calorimeter are not constant is a major weakness of open, with free access of air to the
the apparatus. It is nearly impossible to remedy combustion zone.
this problem. The addition of an auxiliary heat flux
meter is highly recommended to record and Heater. The heater consists of a 5 kW electri-
account for the time-varying exposure conditions. cal heating element wound inside an insulated
The second modification at FPL was the addi- stainless steel conical shell. The heater can be
tion of a load cell to measure specimen mass loss oriented horizontally or vertically to perform
during a test. This was a rather difficult task due tests in either orientation. When tests are
to the geometry of the apparatus and the mecha- performed in the horizontal orientation, the
nism for inserting specimens. The FPL load cell specimen is positioned approximately 25 mm
design seemed to be satisfactory, demonstrating below the bottom plate of the cone heater.
the feasibility of measuring mass loss in the OSU Flames and products of combustion rise and
apparatus. emerge through a circular opening at the top
of the heater. Maximum radiant heat flux to the
specimen exceeds 100 kW/m2.
Cone Calorimeter
Ignition pilot. An electric spark is used as the
The cone calorimeter was developed at the ignition pilot at the top of vertical specimens and
National Bureau of Standards (NBS) by over the center of horizontal specimens.
Dr. Vytenis Babrauskas in the early 1980s
[37]. It is presently the most commonly used Specimen size and orientation. Specimen size in
small-scale calorimeter. The apparatus and test both orientations is 0.1 m 0.1 m. The optional
procedure are standardized in the United States as retainer frame in the horizontal orientation and
ASTM E1354 and NFPA 271, Standard Method the standard specimen holder in the vertical ori-
of Test for Heat and Visible Smoke Release Rates entation reduce the exposed area to
for Materials and Products Using an Oxygen 0.094 m 0.094 m.
Consumption Calorimeter, and internationally
as ISO 5660. Standard cone calorimeter Airflow. Combustion products and dilution air
specimens are exposed in an open environment are extracted through the hood and exhaust duct
with abundant supply of ambient air for combus- by a high-temperature fan. The initial flow rate
tion. Some laboratories have used a modified can be adjusted between 10 and 32 L/s. Volumet-
version of the standard apparatus to conduct stud- ric flow rate remains relatively constant during
ies in vitiated or oxygen-enriched atmospheres. testing. Some cone calorimeters include addi-
tional instrumentation to optionally control and
Standard Version maintain the mass flow rate through the
A schematic view of the apparatus is shown in exhaust duct.
Fig. 27.11. The main features are summarized
below. Additional measurements. The specimen is
Measuring technique. Heat release rate is deter- mounted on a load cell. Most cone calorimeters
mined by the oxygen consumption method. The include instrumentation for measuring light
gas flow rate in the exhaust duct is calculated extinction in the exhaust duct (using a laser
from the pressure drop across and temperature at light source, described in ASTM E1354 and
an orifice plate in the duct. A methane burner ISO 5660-2). Instrumentation to measure
calibration is performed to determine the orifice concentrations of soot, carbon dioxide, carbon
constant. monoxide, and other gases is commonly added.
926 M. Janssens
Orifice plate flowmeter
Exhaust duct
Blower
Hood
Laser smoke meter

Gas sampling port Cone heater Spark plug
Specimen
Specimen holder
Load cell
Fig. 27.11 Cone calorimeter
Modified Versions Originally, only convective heat release rate

A number of laboratories have used the cone was measured on the basis of enthalpy rise of the
calorimeter to study the effect of reduced or exhaust gases. Test results reported since the late
increased oxygen on the burning behavior of 1970s also include total heat release rates calcu-
materials [28, 49, 56–59]. An enclosure was lated from oxygen consumption or carbon diox-
built around the heater and load cell and a mix- ide generation. Several industrial laboratories in
ture of nitrogen and oxygen, or air, was supplied France, Germany, and the United States
to create the desired environment. constructed the apparatus in the 1980s.
Tewarson used the apparatus to determine fire
Fire Propagation Apparatus hazard indices [61] and material properties for
fire modeling [62]. He also investigated the
Factory Mutual Research Corporation (FMRC, effect of environmental conditions (such as oxy-
currently FM Global Research) developed the gen concentration in the combustion air) on heat
fire propagation apparatus (originally referred to release rate and burning behavior. The results of
as the combustibility apparatus) to measure heat his extensive research are summarized Chap. 36.
release rate and generation rates of smoke and The fire propagation apparatus is standardized as
combustion products [60]. A schematic of the ASTM E2058 and has the following main
apparatus is shown in Fig. 27.12. features:
www.ebook777.com
27 Calorimetry 927
Combustion products gases through the exhaust duct is determined by

measuring pressure drop across a precalibrated
Product orifice. The original apparatus had a vertical
sample
analysis
exhaust duct. A commercial version with a hori-
zontal duct was recently developed.
Heater: Four heaters, which are located coaxially

outside the quartz tube, are used to generate an
Collection
hood incident heat flux to the specimen with a maxi-
mum of 65 kW/m2. The electrical heaters operate
at high temperatures so that the spectral distribu-
tion of the emitted radiation is not representative
Quartz
of that present in most fires. This problem is
Infrared
tube heaters (4) discussed elsewhere in this chapter.
Ignition pilot: The ignition pilot burner consists

of a 6.35-mm stainless steel tube. The burner is
Test supplied with a mixture of 60 % ethylene and
Aluminum
sample 40 % air by volume, at a flow rate to create a stable
support flame length of 10 mm. The flame is located
cylinder
Sample 10 mm above the horizontal specimen surface or
support
(on load
10 mm from the vertical specimen surface.
cell)
Specimen size and orientation: Horizontal
Air + oxygen Air distribution box
specimens measure 0.1 m 0.1 m or 0.1 m in
Fig. 27.12 Fire propagation apparatus diameter. ASTM E2058 also describes a proce-
dure for upward flame spread experiments on
Measuring technique: Total heat release rate is 0.1 m 0.3 m vertical planar specimens or 0.8-
determined by the oxygen consumption method. m-long cable specimens in an environment of
Tewarson also used carbon dioxide generation to 40 % oxygen in nitrogen.
calculate heat release rate. However, the amount
of energy generated per mass unit of carbon Airflow: Total gas flow rate supplied to the bot-
monoxide generated is much more fuel depen- tom of the quartz tube is 3.3 L/s. Oxygen content
dent than the amount of energy produced per of the combustion air can varied between 0 %
mass unit of oxygen consumed. Therefore, this and 60 %. Oxygen concentrations below ambient
technique is not as universally accepted as the are used for simulating ventilation-controlled
oxygen consumption method. fires. Oxygen concentrations above ambient are
used to increase flame radiation simulating larger
Configuration: Tests are conducted in a fires [63]. Pure nitrogen is used to determine the
semiopen environment. The specimen is located heat of gasification.
inside a segmented quartz tube, 0.66 m in length
and 0.17 m in diameter. A mixture of oxygen and Additional measurements: The apparatus
nitrogen is supplied at the bottom of the tube. A includes instrumentation to measure specimen
stainless steel funnel and vertical exhaust duct mass loss, smoke obscuration, and
are located at some distance above the tube. concentrations of O2, CO2, and CO in the exhaust
Dilution air is entrained in the area between the flow. An optional hydrocarbon analyzer can be
tube and the exhaust system. The total flow of attached as well.
928 M. Janssens
FAA Microscale Combustion gases. Oxygen consumption calorimetry with

Calorimeter E ¼ 13.1 kJ/g is used to measure heat
release rate.
Recently, the U.S. Federal Aviation Administra- The MCC is different from the calorimeters
tion (FAA) developed the Microscale Combus- described in the previous three subsections
tion Calorimeter (MCC) to assist with the because the primary result is the heat release
development of fire-resistant polymers for use rate per mass unit of fuel volatiles as a function
in commercial passenger aircraft. A schematic of time or pyrolysis chamber temperature
of this micro-scale calorimeter is shown in (as opposed to the heat release rate per unit
Fig. 27.13. exposed specimen area as a function of time).
The apparatus is described in ASTM D7309. The heat release rate per mass unit of fuel
A milligram-size specimen is heated at a con- volatiles is referred to as the specific heat release
stant rate between 0.2 and 2 K/s. Decomposition rate, Q(t), and is expressed in W/g. A typical
can take place in nitrogen (method A) or in a result of an MCC test is shown in Fig. 27.14.
mixture of nitrogen and oxygen (method B). The following five parameters are calculated
When method A is used, char-forming when method A is used:
specimens do not decompose completely and 1. The heat release capacity ηc Qmax/β in
leave a solid residue. In this case the volatiles J/g∙K, where Qmax is the maximum value of
are mixed with a metered supply of oxygen in Q(t) and β is the heating rate in K/s.
the combustor to obtain the heat release rate 2. The heat release temperature Tmax in K as
of the volatiles. When method B is used, the the pyrolysis chamber temperature at which
specimen is completely consumed. The temper- Q(t) ¼ Qmax.
ature of the combustor is set at approximately 3. The specific heat release hc in J/g as the area
900 C to completely oxidize all specimen under the Q(t) curve.
Fig. 27.13 Fire

propagation apparatus
www.ebook777.com
27 Calorimetry 929
Fig. 27.14 Specific heat release rate versus MCC pyrolysis chamber temperature for PMMA
4. The pyrolysis residue Yp mp/mo in g/g, average heat release rate over the first minute
where mp is the residual mass of the specimen following ignition.
at the end of the test. Babrauskas compared peak heat release rate
5. The specific heat of combustion of the speci- from various calorimeters for five aircraft wall
men gases hc,gas hc/(1Yp) in J/g. paneling materials [65]. He found good agree-
For method B only three parameters are ment between the fire propagation apparatus and
calculated: the cone calorimeter. However, he also found that
1. The combustion temperature Tmax in K as the the peak heat release rate from the OSU apparatus
pyrolysis chamber temperature at which the was approximately 50 % of the peak from the
specific heat release rate is a maximum, i.e., cone calorimeter. Whether thermopile or oxygen
QðtÞ ¼ Q0max . consumption were employed seemed to have only
2. The combustion residue Yc mc/mo in g/g, a minor effect on the results from the OSU appa-
where mc is the residual mass of the specimen ratus. Unfortunately, correlation of average heat
at the end of the test. release rate was not reported, so that a compari-
3. The net calorific value h0c in J/g as the area son with the work of Östman et al. is not possible.
under the Q(t) curve. Tran compared heat release rate curves for
Douglas fir plywood from the cone calorimeter,
and the OSU apparatus modified for oxygen con-
Comparative Studies Between Different sumption [54]. First and second peaks agreed
Small-Scale Tests well, but the OSU data exceeded the cone calo-
rimeter data by up to 20 % between the peaks.
A number of comparisons are reported in the The increased burning rate can be explained by
literature on how results obtained with different the enhanced radiant heat flux to the specimen
calorimeters for the same material compare. due to temperature rise of the calorimeter walls
Östman et al. reported on a comparison of heat and heater during a test. Tran tested the same
release data for 13 building materials obtained material in the OSU apparatus with the vertical
with the modified OSU, the cone calorimeter, specimen holder from the cone calorimeter and
and an open calorimeter developed by Sensenig found no effect.
at NBS [64]. Agreement was remarkably good Kandola et al. tested several aircraft interior
with a correlation coefficient exceeding 90 % for fabrics in the OSU calorimeter according to the
930 M. Janssens
FAA specification and in the cone calorimeter

[66]. They found that the specimens ignited Intermediate- and Large-Scale
much earlier in the OSU apparatus and that the Calorimeters
heat release rates were significantly lower in the
cone calorimeter. The heat flux in the cone calo- This section covers commonly used intermedi-
rimeter had to be increased to 50 kW/m2 to ate- and large-scale calorimeters. Collection and
obtain comparable heat release rates as measured extraction of the products of combustion
in the OSU calorimeter at 35 kW/m2. generated in these calorimeters are usually
Two comparative studies were conducted accomplished in a specific manner. The standard
involving electrical cables. Gandhi et al. hood and exhaust duct are described before the
measured shorter ignition times and lower heat main features of different intermediate- and
release rates in the fire propagation apparatus large-scale calorimeters are summarized.
than in the cone calorimeter for communication
cables [67]. Carman et al. compared oxygen con-
sumption and thermopile measurements for six Standard Hood and Exhaust Duct
different types of cables [68]. Good agreement
was obtained between the two measurement To measure heat release rate in a fire test
techniques under flaming conditions. based on the oxygen consumption technique, it
The thermal combustion properties measured is necessary to collect all combustion products
in the MCC are related to flammability and to measure the oxygen concentration and
characteristics of the material [69–73]. For flow rate of the effluents. A properly designed
example, the heat release temperature from hood and exhaust duct with the necessary instru-
method A approximates the surface temperature mentation are used for this purpose. Various
at ignition. The net calorific value from method B intermediate- and large-scale calorimeters
approximates the net heat of combustion described in subsequent sections use the same
measured in an oxygen bomb calorimeter. standard hood and exhaust duct setup shown in
Heat release data from small-scale Fig. 27.15.
calorimeters are always apparatus dependent. The square opening of the hood is approxi-
Differences in geometry, test conditions, and mately 2.4 m 2.4 m and the bottom of the
mounting methods explain discrepancies hood is 2.4–3.0 m above the floor of the labora-
between the results from different calorimeters. tory. Skirts can be hung down from the hood to
Apparatus-specific factors must be considered minimize spilling. Baffle plates in the plenum or
and addressed in a comparison between different an orifice plate at the entrance of the exhaust duct
calorimeters or when the data are used to predict are used to provide proper mixing of the exhaust
performance in real fire conditions. gases.
Fig. 27.15 Standard Gas sampling probe

calorimeter hood and Mixing orifice
exhaust duct Plenum
Exhaust duct
Hood
Bidirectional probe
and thermocouple Smoke photometer
www.ebook777.com
27 Calorimetry 931
The instrumentation section is located at a highest and most accurate measurement of oxy-
distance from the entrance to the exhaust duct gen depletion because no additional air above
of at least 10 times the diameter, that is, 4.0 m for what is entrained in the plume is drawn into the
the standard 0.4-m-diameter exhaust duct. The exhaust duct. The plume flow rate at a height of
measuring section consists of a combination of 2.2 m above a fire with an effective diameter of
a bidirectional probe and thermocouple, a gas 1.5 m can be estimated from the simple equation
sampling probe, and a light extinction measure- developed by Thomas et al. [75] (see Chap. 13):
ment system. A straight section of pipe of a π 2 3=2
length at least six times the diameter is located m_ p ¼ 0:188 D z
4
downstream of the measuring section. 3:14159
The distance between the base of the fire and ¼ 0:188 1:52 2:23=2 ¼ 1:08 kg=s
4
the bottom of the hood determines the peak heat ð27:24Þ
release rate that can be measured for a burning
object located beneath the hood [74]. On the one where
hand, the hood has to be sufficiently elevated m_ p ¼ Plume flowðkg=sÞ
above the fire to avoid flame impingement. z ¼ Height above the fireðmÞ
Flame impingement has two undesirable effects. Assuming a radiative loss fraction of 30 %,
First, flames impinging on a relatively cold sur- which is a typical value for many fuels (see
face are quenched, which adversely affects com- Chap. 36), this corresponds to a volumetric flow
pleteness of combustion. Second, radiation from rate of 2.8 m3/s at a temperature of 650 C.
hot metal surface enhances the burning rate of an Intermediate- and large-scale calorimeter test
object located beneath the hood above that of a standards that specify the hood and exhaust sys-
free burn. tem described in this section typically require an
The maximum heat release rate without flame exhaust fan with a capacity of 3.5 m3/s.
impingement on the standard hood is approxi-
mately 1 MW. This is consistent with the fact
that the flame height of a 1 MW fire with an Intermediate-Scale Calorimeter (ICAL)
effective diameter of 1.5 m is approximately
2.2 m based on Heskestad’s flame height correla- One of the limitations of the cone calorimeter is
tion (see Chap. 13, “Fire Plumes, Flame Height, that only relatively small samples can be
and Air Entrainment”): evaluated. As a result, products that have joints
or layered materials with a thickness exceeding
L f ¼ 0:235Q_ 1:02D
2=5
50 mm can generally not be tested in the cone
¼ 0:235ð1000Þ2=5 1:02ð1:5Þ 2:2m calorimeter in a representative manner. For those
ð27:23Þ types of products or assemblies, a larger calorim-
eter, such as the intermediate-scale calorimeter
where (ICAL) described in ASTM E1623, is required.
L f ¼ Flame lengthðmÞ The ICAL apparatus consists of an array of
Q_ ¼ Heat release rateðkWÞ gas heaters, forming a vertical radiant panel with
D ¼ Effective diameter of the fireðmÞ an approximate height and width of 1.33 m and
On the other hand, the hood cannot be located 1.54 m, respectively (Fig. 27.16). The standard
too high above the burning object to avoid spill- test specimen measures 1 m 1 m and is posi-
ing. The skirts are helpful in this respect. tioned parallel to the radiant panel. The heat flux
An ideal exhaust system extracts combustion to the specimen is preset in the range of
products at the same rate as the flow rate in the 10–60 kW/m2 by adjusting the distance to the
plume when it enters the hood. This results in the panel. Gas flow to the panel is controlled to
932 M. Janssens
Fig. 27.16 ICAL

apparatus
Collection
Gas sampling
hood
port
Wire igniter
Water-cooled
supporting frame Top cap of the
specimen holder
Radiant heat Specimen

units
Wire Weighing
igniter platform
Trolley
Radiant panel Specimen holder
maintain the temperature to the panel and conse- For example, it is very hard to determine the
quently the heat flux to the specimen during a burning behavior of upholstered furniture on the
test. The products of pyrolysis from the specimen basis of the fire characteristics of the foam, fab-
are ignited with hot wires located close to, but ric, and framing materials and to account for the
not in contact with, the specimen at its top and geometry and configuration of the furniture and
bottom. The specimen is placed in a holder that is how it is ignited. It is much more practical to
put on a load cell to measure mass loss during measure the heat release rate and related
testing. properties for the complete object.
Panel and specimen are positioned beneath Furniture calorimeters were developed in the
the standard hood described in the previous sub- 1980s in several laboratories to obtain this kind
section. Measurements of oxygen concentration, of data [76, 77]. The first furniture calorimeter
flow rate, and light transmission in the exhaust test standard was published in 1987 in the Nordic
duct are used to determine the heat release rate countries as NT Fire 032.
and smoke production rate from the specimen as A furniture calorimeter consists of a weighing
a function of time. Because the combustion platform that is located on the floor of the labo-
products from the radiant panel are also captured ratory beneath the standard hood (Fig. 27.17).
in the hood, it is necessary to subtract the The object is placed on the platform and ignited
corresponding heat release rate or smoke produc- with the specified ignition source. The products
tion rate to determine the contribution from the of combustion are collected in the hood and
specimen. extracted through the exhaust duct.
Measurements of oxygen concentration, flow
rate, and light transmission in the exhaust duct
Furniture Calorimeter are used to determine the heat release rate and
smoke production rate from the object as a func-
Often it is very difficult to determine the burning tion of time.
behavior of complex objects on the basis of the Furniture calorimeter test standards have been
fire performance of their individual components. developed in ASTM for chairs (ASTM E1537),
www.ebook777.com
27 Calorimetry 933
Fig. 27.17 Furniture Gas sampling probe

calorimeter Mixing orifice
Plenum
Exhaust duct
Hood
Bidirectional probe
and thermocouple Smoke photometer
Weighing platform
Table 27.3 Ignition sources specified in fire tests for chairs and mattresses
Gas burner ignition source

Test method Specimen No. Type Heat output Location of application
ASTM E1537 CAL TB 133 Single chair 1 Square 19 kW for 80 s Horizontal seating surface
ASTM E1822 Stacked chairs 1 Line 18 kW for 80 s Bottom chair front edge
ASTM E1590 Mattress (set) 1 Line 18 kW for 180 s Front bottom edge
CAL TB 603 Mattress (set) 2 Line 19 kW for 70 s Top surface
16 CFR 1633 Line 10 kW for 50 s Vertical along side
mattresses (ASTM E1590), and stacked chairs specified in 16 CFR 1633. The method and
(ASTM E1822). The California Bureau of requirements are identical to CAL TB 603, except
Home Furnishings and Thermal Insulation that the 10-min total heat release limit is reduced
(CBHFTI) published California Technical to 15 MJ. The fire test method for mattresses and
Bulletins (CAL TB) 133 and 603. These mattress/box spring sets described in this docu-
documents describe fire test procedures to qual- ment was developed at the National Institute of
ify seating furniture and mattresses, respectively, Standards and Technology (NIST) in a research
for use in public occupancies in the state of program sponsored by the Sleep Products Safety
California. Acceptance is primarily based on a Council (SPCS), an affiliate of the International
peak heat release rate and 10-min total heat Sleep Product Association (ISPA) [79–81].
release limits of 80 kW and 25 MJ for chairs The original version of CAL TB 133 specified
(CAL TB 133) and 200 kW and 25 MJ for that the test specimen be located in a corner
mattresses (CAL TB 603). The primary differ- against the back wall of a 3.7 m 3.0 m 2.4 m
ence between the different fire test methods for room with a 1.0 m 2.1 m open doorway in the
chairs and mattresses is the ignition source, front wall. The ignition source consisted of five
which has been demonstrated to affect heat double sheets of loosely-wadded newspaper
release rate [78]. The main features of the gas inside a chicken wire cage placed on the back
burner ignition source specified in different of the seat. Acceptance was based primarily on a
methods are given in Table 27.3. maximum temperature rise of 111 C just below
All mattresses sold in the United States must the ceiling above the chair. Subsequent NIST
comply with the heat release requirements research resulted in the development of the
934 M. Janssens
presently used gas burner as an alternative igni- consumption measurements were made at that
tion source [82]. The gas burner flame generates time. A heat balance was obtained by comparing
an equivalent thermal insult but is much more the theoretical heat release from combustion of
repeatable and reproducible than the original gaseous fuel to the sum of the heat losses. The
ignition source. heat losses consisted of convection through the
The NIST research project also resulted in an doorway, conduction through the walls and ceil-
equivalent heat release rate criterion and ing, and radiation through the doorway.
demonstrated that the room effects are negligible Heat convection through the doorway was
for heat release rates below 600 kW [83]. Later estimated by measuring gas velocity and temper-
studies found significant effects at lower heat ature at many points in the doorway. Heat con-
release rates and proposed a threshold of 460 kW duction through the surrounding surfaces was
[84]. Because this is still much higher than peak calculated based on total heat flux, radiation,
heat release rate limits specified in regulations, all and surface temperature data. Heat loss by radia-
furniture and mattress test standards discussed tion through the door was calculated from radi-
here permit the use of an open furniture calorime- ometer measurements.
ter configuration as an alternative to the room Initially, a series of quasi-steady calibration
configuration. In fact, all test standards except tests were conducted in an inert room. Three
16 CFR 1633 allow two room configurations: the different circular propane gas burners were used
original CAL TB 133 room and a smaller 3.7 m with diameters of 0.2, 0.3, and 0.4 m, respec-
2.4 m 2.4 m room commonly used for room/ tively. Heat balance calculations showed reason-
corner testing (see below). able agreement, with convection losses being
dominant. In subsequent tests with surface
finishes, a heptane pool fire with a heat release
Room/Corner Test rate of approximately 50 kW was used as the
ignition source.
Room/corner tests are by far the most frequently Ahonen et al. at the Technical Research Cen-
conducted large-scale fire experiments through- ter of Finland (VTT) studied the effects of differ-
out the world. This section provides a historical ent gas burner ignition sources on room/corner
overview of the development of room/corner test fire growth [90]. Tests were conducted for each
protocols and summarizes the resulting test combination of three burner sizes (0.17 m 0.17
standards in use today. m, 0.305 m 0.305 m, and 0.5 m 0.5 m) and
three square wave heat release rates (40, 160, and
Historical Overview 300 kW). Oxygen consumption calorimetry was
Much of the work toward the development of a used for measuring heat release rate. The burner
standard room/corner test was performed in the was placed in a corner in the back of the room.
United States in the late 1970s and early 1980s. Ceiling and all walls except the front wall were
The need for a standard room fire test and some lined with 10-mm-thick particleboard with a den-
aspects of its design were discussed by Benjamin sity of 720 kg/m3.
in 1977 [85]. Subsequent research in North The following six criteria were used to deter-
America to arrive at a standard full-scale test mine the time to flashover:
was conducted primarily by Fisher and • Flames emerging through the door
coworkers at the University of California (flameover)
(UCB) [86] and by Lee at the National Bureau • Total heat release rate of 1 MW
of Standards (NBS) [87]. • Total heat flux to the floor of 20 kW/m2
Considerable seminal research was also • Specified rate of smoke production
performed in the Nordic countries. An extensive • Temperature of 600 C at the geometric center
project to construct a full-scale room calorimeter of the room
was conducted in Sweden [88, 89]. No oxygen • Total heat flux to the floor of 50 kW/m2
www.ebook777.com
27 Calorimetry 935
With the time to flashover defined as the compartments meet specific performance
average of the six criteria, the following remark- requirements when tested according to NFPA
able results were obtained: 265, Standard Methods of Fire Tests for
• At the 40 kW level, the medium-sized burner Evaluating Room Fire Growth Contribution of
resulted in flashover first, followed by the Textile Coverings on Full Height Panels and
smaller burner, and then the larger burner. Walls. The principal requirement of these tests
• At 160 kW, the largest burner resulted in is that flashover does not occur. The same codes
flashover first, quickly followed by the other also require that all other interior wall and ceiling
two configurations. finish materials comply with requirements based
• At 300 kW, the trend was the same as at on NFPA 286, including a limit on the total
160 kW with an even smaller spread between smoke released.
the three results. The Safety Of Life At Sea (SOLAS) conven-
The effect of burner size was most significant tion, promulgated by the International Maritime
at the lowest heat release rate, with the medium- Organization (IMO), permits the use of combus-
sized burner being the most severe. At higher tible bulkhead and ceiling linings on high-speed
exposure levels, the size of the burner had no craft, provided they meet stringent fire perfor-
significant effect. Radiative and convective heat mance requirements based on assessment
transfer from the burner flame were shown to according to ISO 9705. ASTM E2257 is the
depend on burner size and heat release rate and American version of ISO 9705.
had a significant effect on the performance of the The test apparatus and instrumentation
material tested. On the basis of the results, the described in the NFPA and ISO room/corner
medium burner size and heat release rate were test standards are very similar (Fig. 27.18). How-
recommended. ever, some significant differences exist in terms
of specimen configuration and ignition source.
Room/Corner Test Standards The apparatus consists of a room measuring
Several standard room/corner test protocols are 3.6 m deep by 2.4 m wide by 2.4 m high, with a
now available and are specified in codes and single ventilation opening (open doorway) mea-
regulations for qualifying interior finishes. For suring approximately 0.8 m wide by 2.0 m high
example, U.S. model building codes require that in the front wall. Walls and ceiling are lined for
textile wall coverings for use in unsprinklered tests according to ISO 9705. For tests according
Fig. 27.18 Room/ Optical density

corner test (lamp/photocell)
Gas analysis (O2, CO, CO2)
Exhaust gases
Volume flow temperature
and differential pressure
Exhaust hood
2.4 m
Gas
burner
3.6 m
Doorway
0.8 m × 2.0 m
936 M. Janssens
to the NFPA standards, the interior surfaces of all thermocouples be installed in the upper part of
walls (except the front wall) are covered with the the room and doorway to measure the tempera-
test material. NFPA 286 is also suitable for ture of hot gases that accumulate beneath the
evaluating ceiling finishes (see below). ceiling and exit through the doorway. In addition
The test material is exposed to a propane to quantitative heat release and smoke produc-
burner ignition source, located on the floor in tion rate measurements, time to flashover (if it
one of the rear corners of the room opposite the occurs) is one of the main results of a room/
doorway. The burner is placed directly against corner test. Different criteria are commonly
(ISO 9705 and NFPA 286) or at a distance of used to define flashover; for example,
50 mm from (NFPA 265) the walls. The ISO upper layer temperature of 600 C, flames
burner consists of a steel sandbox measuring emerging through the doorway, heat flux to the
0.17 m 0.17 m 0.145 m, with the top sur- floor of 20 kW/m2, heat release rate of 1 MW,
face 0.145 m above the floor of the room. Pro- and so forth.
pane is supplied to the burner at a specified rate
such that a net heat release rate of 100 kW is
achieved for the first 10 min of the test, followed Single Burning Item Test
by 300 kW for the remaining 10 min (20-min test
duration unless terminated when flashover The European reaction-to-fire classification sys-
occurs). The NFPA burner consists of a steel tem for construction products except floor
sandbox measuring 0.305 m 0.305 m 0.152 coverings (EN 13501) is based primarily on per-
m, raised so that the top surface is 0.305 m above formance in this test. An SBI test in progress is
the floor of the room. Propane is supplied at a shown in Fig. 27.19. Two specimens of the mate-
specified rate so that a net heat release rate of rial to be tested are positioned in a specimen
40 kW is achieved for the first 5 min of the test, holder frame at a 90 angle to form an open
followed by 150 kW (NFPA 265) or 160 kW corner section. Both specimens are 1.5 m high.
(NFPA 286) for the remaining 10 min (15-min One specimen is 1 m wide and is referred to as
test duration unless terminated when flashover the long wing. The other specimen is 0.5 m wide
occurs). A fundamental difference between and is referred to as the short wing. During a test,
NFPA 265 and NFPA 286 is the fact that the
flame from the burner alone just touches the
ceiling in NFPA 286. This makes it suitable for
assessing the fire performance of interior ceiling
finish, an application for which NFPA 265 is
unsuitable. This effect is partly due to the higher
energy release rate of the NFPA 286 burner, but
primarily because of the burner being in direct
contact with the walls, thereby reducing the area
over which the flames can entrain air and increas-
ing the overall flame height.
All combustion products emerging from the
room through the open doorway are collected in
the standard hood. Instrumentation is provided in
the exhaust duct for measuring heat release rate
based on the oxygen consumed (ISO and NFPA
standards) and smoke production rate (ISO 9705
and NFPA 286 only). The room contains a single
heat flux meter located in the center of the floor.
The NFPA standards also specify that seven Fig. 27.19 SBI test in progress (Photo courtesy SwRI.)
www.ebook777.com
27 Calorimetry 937
the specimens are exposed for 20 min to the of 15 m3/s. A larger calorimeter and higher fan
flame of a triangular-shaped diffusion propane capacity are needed to handle more severe exper-
gas burner operating at 30 kW. The specimen imental fires.
holder and primary gas burner are mounted on a Cooper presented useful guidelines to address
trolley that can be moved in and out of an enclo- the special challenges associated with the design
sure of 3 m 0.6 m wide, 3 m 0.6 m deep, of an industrial-scale calorimeter [74]. ASTM
and 2.4 m 0.1 m high. The enclosure walls E2067 is a standard practice for conducting accu-
consist of noncombustible materials (concrete rate heat release rate measurements at the multi-
block, calcium silicate board, etc.) and/or gyp- megawatt level.
sum board, and have windows to allow the oper- The first industrial-scale calorimeter for fires
ator to observe the test. The mean height and into the multi-megawatt range was built at Fac-
maximum heat flux from the 30 kW burner tory Mutual around 1980 [93]. This calorimeter,
flame are approximately 0.8 m and 35 kW/m2, also referred to as the FM fire products collector,
respectively [91]. was designed to measure heat and other fire
Prior to a test, the specimens are placed in the products from test fires up to a size associated
holder, and the trolley is rolled into the enclosure with sprinkler activation in commodity ware-
and positioned under an insulated hood. During a house storage and other representative
test, the products of combustion are collected in occupancies.
the hood and are extracted through an exhaust Approximately 10 years later, a similar
duct. Instrumentation is provided in the duct to industrial-size calorimeter for heat release rate
measure temperature, velocity, gas composition measurements up to 10 MW was constructed at
(O2, CO2, and CO), and smoke obscuration. the National Testing Laboratory (SP) in Sweden
The velocity and gas composition data are [94]. Since then several other laboratories—such
used to determine heat release rate on the basis as the National Research Council of Canada, the
of the oxygen consumption technique. Materials Fire Research Station in the United Kingdom, the
are tested in triplicate. Classification is based Bureau of Alcohol, Tobacco, Firearms and
primarily on fire growth (FIGRA) and smoke Explosives (ATF) Fire Research Laboratory, the
development (SMOGRA) indices that are equal National Institute of Standards and Technology
to the peak heat release and smoke production (NIST), Southwest Research Institute and
rate, respectively, divided by the time to reach Underwriters Laboratories in the United
the peak. FIGRA and SMOGRA limits were States—developed the capability of measuring
established based on performance in the ISO heat release rate from large fires into the mega-
9705 room/corner test as the reference watt range.
scenario [92].
Use and Application of Heat Release

Industrial-Scale Calorimeters Rate Data
Industrial-scale calorimeters consist of a larger The primary use of heat release rate data is in
version of the hood and exhaust duct shown in support of fire hazard assessment of materials
Fig. 27.15. To handle fires up to 10 MW in size and products. Small- and intermediate-scale
for a short duration, the hood must be at least data can be used in conjunction with mathemati-
6 m 6 m in size or 6 m in diameter and is cal models to predict the performance of
typically located 6.5 m or higher above the floor materials and products in real fire scenarios.
of the laboratory. The fan must be capable of Heat release rates measured in large- and
extracting combustion products through a industrial-scale calorimeters can be used directly
0.9-m-diameter exhaust duct at a minimum rate in support of a fire hazard assessment.
938 M. Janssens
ig ρ
Small-Scale Calorimeter Data t0:25 1:72
tfo ¼ 0:0716 0 0 1:30
þ 57:4 ð27:25Þ
Q300
Predictive models that are available for fire haz-
ard assessment on the basis of small- or where
intermediate-scale heat release rate data vary tf0 ¼ Time to flashover (s)
widely in complexity. The extent of the heat tig ¼ Time to ignition in the cone calorimeter at
release rate data that are needed varies according 50 kW/m2 (s)
to the complexity of the model. Room/corner ρ ¼ Density of the material (kg/m3)
testing is a widely accepted approach to assess Q30000 ¼ Total heat released at 50 kW/m2 during
the fire hazard of interior finish materials. The 300 s following ignition (MJ/m2)
room/corner scenario is used here to illustrate the This equation can be used to predict room/
different types of predictive models and corner test performance on the basis of small-
corresponding requirements for input data from scale data for materials that have not been tested
small-scale calorimeters. in the room. With this approach the dynamics of
There are three distinct types of room/corner the heat release curve are lost entirely. There is
test models: regression models, analytical no difference in the prediction of full-scale per-
models, and physics-based models. Regression formance for two materials with heat release
models express a relationship between a particu- curves of very different shapes, provided the
lar room/corner test performance characteristic, heat release rate parameter and ignition time
for example, the time to flashover, and small- used in the correlation are identical. The
scale fire test data for the same product. Regres- predictions are valid for one scenario and geom-
sion models are based on a statistical analysis of etry only. For example, Equation 27.25 cannot be
room/corner and small-scale test data for a set of used to predict the time to flashover for the
products and can be used as a screening tool. NFPA 265 and NFPA 286 room/corner tests.
Analytical models predict fire growth but do not The main advantage of this approach is that a
simulate the room environment. Physics-based minimal amount of small-scale testing is needed
models predict how the room environment varies to obtain the necessary input data.
as a function of time and how flames spread over
the walls and ceiling of the compartment. There
is a strong interaction between regression and Analytical Models
physics-based models because the conditions in
the room determine the heat that is transferred The approach outlined in the preceding subsec-
back to the wall and ceiling surfaces, which tion can be improved by using a heat release
affects the flame spread and the heat release and curve at a single radiant heat flux level. Thus,
smoke production rate of burning wall and ceil- the dynamic effects of the room fire on the expo-
ing sections. sure level are ignored while the dynamics of the
heat release curve are largely maintained. The
radiant heat flux level is chosen so that it is a
Regression Models representative average (over space and time) for
the fire scenario that is being modeled [96]. Rep-
The least sophisticated models are based on resentative heat flux levels for the room/corner
regression analyses. The following equation, test scenario range from 25 to 50 kW/m2.
obtained by Östman et al. based on an analysis The single heat release curve is used in com-
of test data for 28 materials in the cone calorim- bination with a flame spread algorithm to predict
eter and according to ISO 9705, serves as a good heat release rate as a function of time in the
example of this type of room/corner test room/corner test. The flame spread algorithm
modeling [95]. can be very simple but needs at least some
www.ebook777.com
27 Calorimetry 939
ignition data for the material. The burning area completely consume the material than at higher
expands as time proceeds and new sections of the heat fluxes. To address this problem, heat release
material are ignited. The heat release rate is cal- rate has to be expressed as a function of a prog-
culated at discrete time increments, Δt. The ress variable that is consistent at different radiant
flame spread algorithm calculates the expansion heat flux levels.
of the burning area at every time step. The heat Smith and Green conducted experiments in
release rate in the room/corner test at a particular the OSU calorimeter at different heat flux levels
time is then obtained by adding the contributions and tested the same material at time-varying
from all incremental areas to the heat release rate heat fluxes [98]. They were able to reconstruct
from the ignition burner at that time: the heat release rate curve measured under
dynamic exposure conditions from interpolation
N X 00
Q_ t ðtN Þ ¼ Q_ b ðtN Þ þ Ai Q_ ðtN ti Þ ð27:26Þ between the curves obtained at fixed radiant
i¼0 heat fluxes using total heat release as the prog-
ress variable. Mitler used total mass loss as the
where
progress variable [99]. Janssens suggested using
Q_ t ¼ Total heat release rate in the room=corner char depth as a suitable progress variable for
testðkWÞ wood [100].
tN ¼ Time after N time increments Δt from the The following modification to Equation 27.26
start of the test, equal toNΔtðsÞ represents an improved room/corner test model
Q_ b ¼ Heat release rate from the ignition that accounts for the effect of incident heat flux
burnerðkWÞ on heat release rate. Total heat release rate is
Ai ¼ Incremental area ignited at time ti ðm2 Þ used as the progress variable.
00
Q_ ¼ Heat release rate measured in the cone
calorimeterðkW=m2 Þ Q_ t ðtN Þ ¼ Q_ b ðtN Þ
ti ¼ Time afteritime increments Δt from the start XN 00 h i
þ Ai Q_ q_ e, i ðtN Þ, Qi ðtN Þ
00 00
of the test, equal to i ΔtðsÞ
i¼0
This method automatically accounts for burn-
out. The most widely known room/corner test ð27:27Þ
model of this type was developed by Wickström where
and Göransson [97].
00
q_ e, i ¼ Incident heat flux to Ai ðkW=m2 Þ
00
Q_ i ¼ Total heat release from Ai ðMJ=m2 Þ
Physics-Based Models
The model calculates the incident heat flux
Direct Use of Heat Release Rate and keeps track of the total heat release rate for
Measurements at Multiple Heat Fluxes each incremental area. A room/corner test model
The modeling approach described in the preced- of this type was developed by Smith et al. at Ohio
ing subsection can be refined by using heat State University [5].
release rate curves obtained at multiple heat The primary limitations of using this type of
fluxes. This makes it possible to account for the model are that (1) it is based on the assumption
fact that the incident heat flux to each incremen- that the heat flux from the material’s own flame
tal area varies with time. The problem, however, in the small-scale calorimeter is comparable to
is that the heat release rate of an incremental area that in the room/corner test, and (2) Equa-
at a particular time cannot be determined from tion 27.27 is assumed to be valid regardless
direct interpolation of the heat release rate curves whether the incident heat flux is purely radiative
measured in the calorimeter. This is because it (as in the small-scale calorimeter) or partly con-
takes more time at lower heat fluxes to vective (as, for example, in areas of the room
940 M. Janssens
where the material is exposed to the flame of the during which the heat release rate is at least
ignition source). These limitations can be 80 % of the first peak heat release rate. The
addressed, at least in an approximate manner, third value is equal to the ratio of total heat
but not without making the approach much released and total mass loss over the entire
more complex. flaming period.
The heat of gasification is defined as the net
Use of Heat Release Properties heat flow into the material required to convert
A more fundamental approach to account for the one mass unit of solid material to volatiles. The
effects of time-varying heat fluxes on heat net heat flux into the material can be obtained
release rate is based on two material properties from an energy balance at the surface of the
that can be measured in a small-scale calorime- specimen. Typically, a sample exposed in a
ter. These properties are the effective heat of small-scale calorimeter is heated by external
combustion, Δhc,eff, and the heat of gasification, heaters and by its own flame. Heat is lost from
Δhg. Both properties have the units of kJ/g or the surface in the form of radiation. A schematic
MJ/kg. of the heat balance at the surface of a burning
The effective heat of combustion is the ratio specimen in the cone calorimeter is shown in
of heat release rate to mass loss rate measured in Fig. 27.20. Hence, Δhg is defined as
a small-scale calorimeter. 00 00 00 00
q_ q_ e þ q_ f q_ l
00 Δhg net00 ¼ ð27:29Þ
Q_
00
m_ m_
Δhc, eff 00 ð27:28Þ
m_
where
00
where q_ e ¼ Heat flux to the specimen surface from
00
Q_ ¼ Heat release rate per unit exposed external sourcesðkW=m2 Þ
00
areaðkW=m2 Þ q_ f ¼ Heat flux to the specimen surface from the
00
m_ ¼ Mass loss rate per unit exposed flameðkW=m2 Þ
00
areaðg=m2 sÞ q_ l ¼ Heat losses from the exposed surface
The effective heat of combustion at a particu- ðkW=m2 Þ
lar time t can be calculated by substituting the
00
values for Q_ and ṁ00 at that time in Equa-
tion 27.28. A curve of Δhc,eff as a function of
time can be determined in this manner. Unfortu-
Cone heater surface
nately, mass loss rate data are often very noisy
and the calculated time-varying heat of combus-
tion values may not have any physical meaning.
More meaningful values are obtained by calcu-
lating an average Δhc,eff, over a specified time
00
period by substituting average values of Q_ and
ṁ00 over that time period in Equation 27.28. Flame
Dillon et al. proposed three methods to calcu-
late the effective heat of combustion [101]. The qe⬙ qt⬙ qf⬙
first value is equal to the ratio of the first peak
heat release rate and mass loss rate at the same Specimen
time. The second value is obtained as the ratio of

the average heat release rate over the peak burn- Ceramic fiber blanket
ing period and the mass loss rate over the same
period. The peak burning period is defined as the Fig. 27.20 Heat balance at the surface of a burning cone
period around the first peak heat release rate calorimeter specimen
www.ebook777.com
27 Calorimetry 941
00 00
If the flame is approximated as a homoge- q_ l q_ f . Tewarson et al. [62] and Petrella [102]
neous gray gas volume, the heat flux from the have used this technique to obtain average Δhg
flame can be expressed as follows: values for a large number of materials. Tewarson
00 00 00 et al. also conducted tests in vitiated O2-N2
q_ f ¼ q_ f , c þ q_ f , r ¼ h* Tf T s þ σε f T 4f 00
mixtures and found q_ f to decrease linearly with
ð27:30Þ decreasing oxygen concentration. Analysis of
these additional experiments made it possible to
where 00 00
00
separate q_ f and q_ l .
q_ f , c ¼ Convective fraction of the flame flux Many materials, in particular those that form
ðkW=m2 Þ an insulating char layer as they burn, take a long
00
q_ f , r ¼ Radiative fraction of the flame flux time to reach steady burning conditions or may
ðkW=m2 Þ never reach steady conditions. Equation 27.29 is
h* ¼ Convection coefficient corrected for still valid for such materials, but the heat and
blowingðkW=m2 KÞ mass fluxes and resulting Δhg values vary with
Tf ¼ Flame temperatureðKÞ time. Tewarson and Petrella have used the
Ts ¼ Surface temperatureðKÞ method described in the preceding paragraph to
σ ¼ Boltzmann constant 5:67 1011 kW=m2 K4 determine average Δh values for nonsteady
g
εf ¼ Emissivity of the flame burning materials using average mass loss rates.
The flow of combustible volatiles emerging They found that average ṁ00 is still an approxi-
through the exposed surface of the specimen mately linear function of q_ 00 . However, the aver-
e
adversely affects the convective heat transfer
age heat of gasification values obtained in this
between the flame and the surface. This effect
manner may not have any physical meaning. For
is referred to as “blowing.” The flame flux in a
example, Janssens demonstrated that the values
small-scale calorimeter is primarily convective,
based on average mass loss rates are too high for
in particular in the vertical orientation, and flame
wood and suggested a method to determine Δhg
absorption of external heater and specimen sur-
as a function of char depth [103].
face radiation can be neglected.
Dillon et al. proposed six methods to calculate
The heat losses from the surface can be
the heat of gasification [101]. The first three
expressed as
values are derived from Equation 27.29 and are
00 4 equal to the reciprocal of the slope of a linear fit
q_ l ¼ σεs T s T 14
ð27:31Þ
through data points of peak mass loss rate, aver-
where age mass loss rate over the peak burning period,
εs ¼ Surface emissivity of the specimen and average mass loss rate over the entire
T1 ¼ Ambient temperature (K) flaming period respectively plotted as a function
Some materials exhibit nearly steady mass of heat flux. The other three values are equal to
loss rates when exposed to a fixed radiant heat the appropriate heat of combustion times the
flux. Ts for these materials reaches a steady value reciprocal of the slope of a linear fit through
after a short initial transient period and all terms data points of first peak heat release rate, average
in Equation 27.29 are approximately constant. heat release rate over the peak burning period,
Δhg can then be obtained by measuring steady and average heat release rate over the entire
mass loss rates at different radiant heat flux flaming period respectively plotted as a function
00
levels and by plotting ṁ00 as a function of q_ e . of heat flux.
The reciprocal of the slope of a straight line fitted Physically meaningful nonsteady values for
through the data points is equal to Δhg. The Δh g can be obtained from Equation 27.29, with
00 00
intercept of the line with the abscissa is equal to q_ f and q_ l calculated from Equations 27.30 and
942 M. Janssens
27.31, respectively. These calculations require flux estimates are relatively less important. How-
values for h*, εf, Tf, εs, and Ts. All parameters, ever, for practical reasons, it is often preferable
except Ts, do not vary greatly during a test and to run small-scale tests in the horizontal orienta-
can be estimated relatively easily [104]. Ts, how- tion to avoid problems with, for example, melt-
ever, may change significantly as a function of ing and dripping of the specimen.
time. The surface temperature can be measured, The equation to calculate the total heat release
but that presents some major challenges. rate in a room/corner test based on a model that
Janssens obtained Ts as a function of time for relies on heat release rate properties has the fol-
wood specimens exposed in the cone calorimeter lowing form:
by solving the equation for heat conduction 00
N X q_ net, i ðtN Þ
through the char layer using an integral tech- Q_ t ðtN Þ ¼ Q_ b ðtN Þ þ Ai Δhc, eff
nique. [100] The resulting values for Δhg are i¼0 Δhg
consistent with a theoretical analysis [105] and ð27:32Þ
calculated Δhg values based on measured surface
00
temperatures [106]. A drawback of this approach where q_ net, i is the net heat flux to Ai (kW/m2). A
is that thermal properties of the material’s char room/corner test model of this type was devel-
are needed. oped by Quintiere [107].
The discussion in this subsection is useful in
clarifying a common misconception. Often it is
believed that materials used in a particular Intermediate-Scale Calorimeter Data
end-use orientation should be tested in that ori-
entation. This is not necessarily correct. Heat Products that have joints or layered materials
release rate is independent of specimen orienta- with a thickness exceeding 50 mm can generally
tion. However, the heat release rate in a calo- not be tested in a small-scale calorimeter in a
rimeter under otherwise identical conditions is representative manner. The ICAL is suitable to
higher in the horizontal than in the vertical obtain heat release rate data for these products.
orientation. This is because the heat feedback The ICAL also has some practical advantages
from the flame is much greater in the horizontal over small-scale calorimeters for measuring
orientation. In that orientation, the flame is a Δhc,eff and Δhg [106]. The four approaches
relatively large volume of hot gas located above discussed in the previous section can be used in
the specimen. The flame is a thin sheet in front support of a hazard assessment of such products
of a vertical specimen, leading to a much lower based on ICAL data.
heat feedback. However, neither of these
situations is comparable to that in a real fire,
where burning areas and flame volumes are Furniture Calorimeter Data
much larger and heat flux from the flame is
much higher regardless of orientation of the The primary application of furniture calorimetry
fuel surface. is to obtain heat release rate data for input into
Hence, the best approach is to interpret small- zone models such as CFAST. [108] Furniture
scale measurements in terms of material calorimeters are designed to obtain data under
properties that are independent of the test appa- free burning conditions. However, the heat
ratus. These material properties can then be used release rate of a burning object inside a room
to predict full-scale performance using a method might be higher due to heat feedback from the
that accounts for the effect of the enhanced heat hot upper smoke layer and heated walls and
flux from large flames. The preceding reasoning ceiling. Although this effect is not very signifi-
indicates that small-scale testing in the vertical cant during the early stages of a compartment
orientation is preferable, because the heat feed- fire, it can become significant as the fire
back to the flame is smaller and errors of flame approaches flashover. Zone models are typically
www.ebook777.com
27 Calorimetry 943
not capable of accounting for this effect, except • Class I: noncombustible products on wood
for some simple geometries for which the effect pallets
of external heat flux on burning rate can be • Class II: Class I products in slatted wooden
calculated. crates, solid wooden boxes, multiple thickness
corrugated cartons, or equivalent combustible
packaging material on wood pallets
Industrial-Scale Calorimeter Data • Class III: packaged or unpackaged wood,
paper, natural fiber cloth or products there-
Industrial-scale calorimeters can be used to from on wood pallets
obtain heat release rate data from large objects • Class IV: Class I, II, or III products containing
in support of the design of passive fire protection no more than 25 % by volume of expanded
of structures. For example, heat release rates plastic or polyurethane or 15 % by weight of
from motor vehicles have been measured in sev- unexpanded plastic or polyurethane in ordi-
eral laboratories throughout the world nary corrugated cartons on wood plastic
[109–115]. The data obtained in these tests pallets
were used to develop guidelines for passive fire • Cartoned Group B unexpanded plastics
protection of structural steel in parking garages. • Cartoned Group A expanded or unexpanded
Industrial-scale calorimeters are also used to plastics
determine the hazard classification of • Idle wood pallets
commodities. The level of active fire protection
required to protect a warehouse is based on the
hazard classification of the commodity stored in
Uncertainty of Heat Release Rate
the warehouse. Standard methods for commodity
Measurements
classification have been developed in the United
States and Sweden [116]. For example, FM 3995
The objective of a measurement is to determine
describes a protocol to determine the hazard class
the value of the measurand, that is, the physical
of plastic pallets and other products. The com-
quantity that needs to be measured. Every mea-
modity configuration consists of eight pallet
surement is subject to error, no matter how
loads of products, each measuring 1.1 m 1.1 m
carefully it is conducted. The true value of a
1.1 m. The commodities are placed on a
measurand is therefore unknowable because it
double-row rack segment in a standard
cannot be measured without error. However, it
2 2 2 array with each pallet separated by
is possible to estimate, with some confidence,
150 mm. The commodities are ignited at the
the expected limits of error. This estimate,
bottom in the center of the array. Water is applied
referred to as the uncertainty of the measure-
to the fire by a special applicator located 200 mm
ment, provides a quantitative indication of its
above the top surface of the commodity. The
quality.
water is applied at the time that a sprinkler sys-
The value of the measurand is generally not
tem located in a warehouse at 3 m above the
obtained from a direct measurement but is deter-
commodity would activate. The activation time
mined as a function ( f ) from N input quantities
is calculated based on the convective heat release
X1, X2, . . ., XN:
rate measured during the test and a sprinkler
activation program akin to DETACT-QS Y ¼ f ðX 1 ; X 2 ; . . . ; X N Þ ð27:33Þ
[117]. Tests are performed at three different
water application rates in the range of where
0.11–0.39 gpm/ft2 (4.5–15.9 mm/min) and the Y ¼ True value of the measurand
commodity is classified based on the heat release f ¼ Functional relationship between measurand
rates measured in the tests. The FM Global com- and input quantities
modity classification system in ascending order Xi ¼ True values of the input quantities (i ¼ 1
of hazard is as follows: . . . N)
944 M. Janssens
The input quantities may be categorized as estimates x1, x2, . . ., xN for the values of the
• Quantities whose values and uncertainties are N input quantities:
directly determined from single or repeated
observation y ¼ f ðx 1 ; x 2 ; . . . ; x N Þ ð27:34Þ
• Quantities whose values and uncertainties are The standard uncertainty of y is obtained by
brought into the measurement from external appropriately combining the standard uncertainties
sources such as reference data obtained from of the input estimates x1, x2. . ., xN. If all input
handbooks quantities are independent, the combined standard
An estimate of the value of the measurand, y, uncertainty of y is given by
is obtained from Equation 27.34 using input
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u " #2
uX ∂f X
uc ð y Þ ¼ t
N N
u2 ðxi Þ ½ c i uð x i Þ 2 ð27:35Þ
∂X
i¼1 i xi i¼1
where • A type B evaluation of standard uncertainty of

u ¼ Standard uncertainty xi is not based on repeated measurements but
uc ¼ Combined standard uncertainty on an a priori frequency distribution. In this
ci ¼ Sensitivity coefficients case, the uncertainty is determined from pre-
The standard uncertainty of an input estimate vious measurements, experience or general
xi is obtained from the distribution of possible knowledge, manufacturer specifications, data
values of the input quantity Xi. There are two provided in calibration certificates,
types of evaluations depending on how the dis- uncertainties assigned to reference data taken
tribution of possible values is obtained: from handbooks, and so on.
• A type A evaluation of standard uncertainty of Equation 27.35 is referred to as the law of
xi is based on the frequency distribution, propagation of uncertainty and is based on a
which is estimated from a series of first-order Taylor series approximation of Y ¼ f
n repeated observations xi,k (k ¼ 1 . . . n): (X1, X2, . . ., XN). When the nonlinearity of f is
rffiffiffiffiffiffiffiffiffiffiffi significant, higher-order terms must be included.
pffiffiffiffiffiffiffiffiffiffiffi s2 ðx i Þ When the input quantities are correlated, Equa-
uðxi Þ s ðxi Þ ¼
2
n tion 27.35 must be revised to include the covari-
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uX ance terms. The combined standard uncertainty
u ðx x Þ2
n
u i, k i of y is then calculated from
t
¼ k¼1 ð27:36Þ
nð n 1Þ
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
XN X X
N1 N
uc ð y Þ ¼ ½ci uðxi Þ 2 þ 2 ci cj uðxi Þu xj r xi ; xj ð27:37Þ
i¼1 i¼1 j¼iþ1
where r(xi, xj) is the estimated correlation coeffi- estimated on the basis of the measured values
cient between Xi and Xj. of the input quantities. The combined standard
Because the values of the input quantities are uncertainty in Equation 27.37 is usually
not known, the correlation coefficient is multiplied by a coverage factor to raise the
www.ebook777.com
27 Calorimetry 945
confidence level. A multiplier of 2 is often used, measurement errors and obtained an uncertainty
which corresponds to a confidence level of of 11 % for the peak heat release rate of a glass
approximately 95 %. fiber–reinforced plastic measured in the cone
Equation 27.37 can be used to calculate the calorimeter at a heat flux of 50 kW/m2 [121].
uncertainty of heat release rate measurement
based on oxygen consumption calorimetry. For
example, Equation 27.15 provides the functional Summary
relationship between the measurand (heat release
rate) and the input quantities. Assuming the mass Heat release rate is the single most important
flow rate in the exhaust duct is calculated from variable in fire hazard assessment. Various test
the differential pressure of and temperature at an methods have therefore been developed for mea-
orifice plate or bidirectional probe, the output suring heat release rate of materials and products
and input quantities are defined as follows: under different conditions. This chapter dealt
with calorimeters of various sizes and the use
_
Y Q,
e
X1 E, X2 XAO2 , X3 α, and application of heat release measurements.
X4 C, X5 ΔP, X6 T e The discussion started with a description of
ð27:38Þ the oxygen bomb calorimeter. The most signifi-
cant limitation of this test is that it does not
C is the calibration coefficient, which relates provide a quantitative measure of heat generation
the mass flow rate in the exhaust duct to the under realistic fire conditions.
differential pressure and gas temperature The next section described four techniques
measurements. In a test Q_ is calculated as a that have been used to measure heat release rate
function of time based on the input quantities in fire tests. The sensible enthalpy rise method is
measured at discrete time intervals Δt. The the least complicated. The substitution and com-
uncertainty of the heat release rate measured at pensation methods partly address the problem of
each time interval is estimated from thermal lag associated with sensible enthalpy rise
Equation 27.37. measurements but require sophisticated control
Dahlberg used this approach to determine the instrumentation. The oxygen consumption
uncertainty of heat release rate measured in the method, based on Thornton’s rule, was devel-
industrial-size calorimeter at SP and reported oped in the late 1970s. It is currently the most
values of 7 % and 12 % depending on the popular method for measuring heat release rate in
use of the CO correction, that is, for fire tests.
Equations 27.18 and 27.15, respectively The effects of some calorimeter construction
[118]. Enright and Fleischmann reported an details on quality and accuracy of small-scale
uncertainty of 5 % for the cone calorimeter heat release rate measurements were discussed.
[119]. These uncertainties are significantly Factors examined include configuration (open
below the precision obtained from vs. closed), type of heater, type of ignition pilot,
interlaboratory trials involving oxygen consump- specimen size and orientation, edge effects, and
tion calorimeters. airflow. Four commonly used small-scale
For example, a cone calorimeter round robin calorimeters were briefly described: the Ohio
resulted in estimates for the peak heat release rate State University calorimeter, the cone calorime-
repeatability and reproducibility of 17 % and ter, the fire propagation apparatus, and the micro-
23 %, respectively [120]. The discrepancies can scale combustion calorimeter. This discussion
be explained by the fact that the uncertainty was followed by a review of studies comparing
analyses did not account for dynamic errors and different small-scale calorimeters.
specimen, operator, and heat flux variations. This The chapter continued with a description of
is consistent with the calculations performed by the hood and exhaust duct that is specified in
Janssens, who accounted for the contribution many intermediate- and large-scale calorimeter
from specimen variations and heat flux standards. This was followed by a description of
946 M. Janssens
different calorimeters for measuring the heat 4. E. Smith, “Application of Release Rate Data to Haz-
release rate from chairs, mattresses, and other ard Load Calculations,” Fire Technology, 10, pp.
181–186 (1974).
objects and from wall and ceiling finishes in a 5. E. Smith and S. Satija, “Release Rate Model for
corner configuration. Some historical back- Developing Fires,” ASME Journal of Heat Transfer,
ground was provided for the room/corner test. 105, pp. 282–287 (1981).
Industrial-scale calorimeters that can be used to 6. V. Babrauskas, “Performance of the OSU Rate of
Heat Release Apparatus Using PMMA and Gaseous
measure multimegawatt heat release rates from Fuels,” Fire Safety Journal, 5, pp. 9–20 (1982).
large objects and commodities were also briefly 7. M. Janssens, “Critical Analysis of the OSU Room
discussed. Fire Model for Simulating Corner Fires,” in Fire and
The next section discussed common Flammability of Furnishings and Contents, ASTM
STP 1233, American Society of Testing and
applications of heat release rate data. Heat Materials, Philadelphia, PA, pp. 169–185 (1994).
release rate data are used primarily in support 8. D. Evans and J. Breden, “Time Delay Correction for
of fire hazard assessment of materials and Heat Release Rate Data,” Fire Technology, 14, pp.
products. Small- and intermediate-scale data 85–96 (1978).
9. D. Bluhme and R. Getka, “Rate of Heat Release
must be used in conjunction with a mathematical Test—Calibration, Sensitivity and Time Constants
model to predict performance of materials and of ISO RHR Apparatus,” NORDTEST Project
products in real fire scenarios. General concepts 115-77, National Institute for Testing of Materials,
of four types of models were discussed using the Copenhagen, Denmark (1979).
10. P. Vandevelde, “An Evaluation of Heat Release
room/corner test as an example of a real fire Criteria in Reaction-to-Fire Tests,” Fire and
scenario: Materials, 4, pp. 157–162 (1980).
• Correlations 11. Abramowitz and R. Lyon, “Effect of Instrument
• Models based on heat release rate data Response Time on Heat Release Rate
Measurements,” Fire and Materials, 19, pp. 11–17
obtained at a single heat flux (1995).
• Models based on heat release rate data 12. R. Filipczak, and R. Lyon, “The Correlation of Heat
obtained at multiple heat fluxes Release Calorimetry Measurements,” DOT/FAA/
• Models based on heat release rate properties AR-TN02/104, FAA William J. Hughes Technical
Center, Atlantic City International Airport, NJ
Heat release rates measured in large- and (2002).
industrial-scale calorimeters can be used directly 13. N. Thompson and E. Cousins, “The FM Construction
in support of fire hazard assessment. For exam- Materials Calorimeter,” NFPA Quarterly, 52, pp.
ple, furniture calorimeter measurements can be 186–192 (1959).
14. J. Brenden, “Apparatus for Measuring Rate of Heat
used to generate heat release rate curves for input Release from Building Products,” Journal of Fire
in zone models. Industrial-scale calorimeter data and Flammability, 6, pp. 50–64 (1975).
can be used to support the design of passive fire 15. W. Parker and M. Long, “Development of a Heat
protection for structures or to obtain a hazard Release Rate Calorimeter at NBS,” in Ignition, Heat
Release and Noncombustibility of Materials, ASTM
classification of a commodity. STP 502, American Society of Testing and
The chapter concluded with a brief discussion Materials, Philadelphia (1972).
of uncertainty of heat release rate measurements. 16. J. Tordella and W. Twilley, “Development of a Cal-
orimeter for Simultaneously Measuring Heat
Release and Mass Loss Rate,” NBSIR 83-2708,
National Bureau of Standards, Gaithersburg, MD
References (1983).
17. S. Martin, “Characterization of the Stanford
1. V. Babrauskas and R. Peacock, “Heat Release Rate: Research Institute Large-Scale Heat Release Rate
The Single Most Important Variable in Fire Hazard,” Calorimeter,” NBS-GCR 76-54, National Bureau of
Fire Safety Journal, 18, pp. 255–272 (1992). Standards, Gaithersburg, MD (1975).
2. E. Smith, “An Experimental Determination of Com- 18. W. Thornton, “The Relation of Oxygen to the Heat
bustibility,” Fire Technology, 7, pp. 109–119 (1971). of Combustion of Organic Compounds,” Philosoph-
3. E. Smith, “Heat Release Rate of Building Materials,” ical Magazine and Journal of Science, 33 (1917).
in Ignition, Heat Release and Noncombustibility of 19. C. Huggett, “Estimation of the Rate of Heat Release
Materials, ASTM STP 502, American Society of by Means of Oxygen Consumption,” Fire and
Testing and Materials, Philadelphia, PA, Materials, 12, pp. 61–65 (1980).
pp. 119–134 (1972).
www.ebook777.com
27 Calorimetry 947
20. W. Parker, “An Investigation of the Fire Environ- Award No. 2010-DN-BX-K221, National Institute
ment in the ASTM E84 Tunnel Test,” NBS Technical of Justice, Washington, DC (2012).
Note 945, National Bureau of Standards, 35. B. Lattimer and J. Beitel, “Evaluation of Heat
Gaithersburg, MD (1977). Release Rate Equations Used in Standard Test
21. R. Walters, S. Hackett and R. Lyon, “Heats of Com- Methods,” Fire and Materials, 22, pp. 167–173
bustion of High Temperature Polymers,” Fire and (1998).
Materials, 24, pp. 245-252 (2000). 36. V. Babrauskas and P. Thureson, “Short Communica-
22. C. Gomez and M. Janssens, “Thornton’s Constant tion: Drying Agents’ Effects on CO2 Readings,” Fire
Revisited,” in Proceedings of the 11th International and Materials, 18, pp. 201–268 (1994).
Fire and Materials Conference, Interscience 37. V. Babrauskas, “Development of the Cone Calorim-
Communications Limited, London, England, (2009). eter—A Bench-Scale Heat Release Rate Apparatus
23. H. Sawada, Thermodynamics of Polymerization, Based on O2 Consumption,” Fire and Materials,
Marcel Dekker, New York (1976). 8, pp. 81–95 (1984).
24. W. Parker, “Calculations of the Heat Release Rate by 38. J. Hallman, Ignition of Polymers by Radiant Energy,
Oxygen Consumption for Various Applications,” University of Oklahoma, Norman, OK (1971).
NBSIR 81-2427, National Bureau of Standards, 39. A. Koohyar, Ignition of Wood by Flame Radiation,
Gaithersburg, MD (1982). University of Oklahoma, Norman, OK (1967).
25. M. Janssens, “Measuring Rate of Heat Release by 40. D. Drysdale, An Introduction to Fire Dynamics, 2nd
Oxygen Consumption,” Fire Technology, 27, pp. ed., John Wiley and Sons, Chichester, UK (1998).
234–249 (1991). 41. B. Östman and R. Nussbaum, “Larger Specimens for
26. S. Brohez, C. Delvosalle, G. Marlair, and Determining Rate of Heat Release in the Cone Calo-
A. Tewarson, “Soot Generation in Fires: An Impor- rimeter,” Fire and Materials, 10, pp. 151–160
tant Parameter for Accurate Calculation of Heat (1986).
Release,” in Proceedings of the 6th International 42. M. Janssens and J. Urbas, “Comparison of Small and
Symposium, International Association of Fire Safety Intermediate Scale Heat Release Rate Data,” in
Science, London, UK (2000). Proceedings of Interflam ‘96, Interscience
27. B. Dlugogorski, J. Mawhinney, and V. Duc, “Mea- Communications, London, UK (1996).
surement of Heat Release Rates by Oxygen Con- 43. B. Toal, T. Shields, and G. Silcock, “Observations
sumption Calorimetry in Fires under Suppression,” on the Cone Calorimeter,” Fire and Materials,
in Proceedings of the 4th International Symposium, 14, pp. 73–76 (1989).
International Association for Fire Safety Science, 44. B. Toal, T. Shields, and G. Silcock, “Suitability and
London, UK (1994). Preparation of Samples for the Cone Calorimeter,”
28. M. Werrel1, J. Deubel, S. Krüger, A. Hofmann and Fire Safety Journal, 16, pp. 85–88 (1990).
U. Krause, “The Calculation of the Heat Release 45. J. Urbas and H. Sand, “Some Investigations on Igni-
Rate by Oxygen Consumption in a Controlled- tion and Heat Release of Building Materials Using
Atmosphere Cone Calorimeter,” Fire and Materials, the Cone Calorimeter,” in Proceedings of Interflam
38, pp. 204-226, (2014). ‘90, Interscience Communications, London, UK
29. P. Beaulieu and N. Dembsey, “Enhanced Equations (1990).
for Carbon Dioxide and Oxygen Calorimetry,” in 46. J. deRis (2000). “Sample Holder for Determining
Proceedings of the 9th Fire and Materials Confer- Material Properties.,” Fire and Materials, 24, pp.
ence, Interscience Communications, London, 219-226 (2000).
England., (2005). 47. V. Babrauskas, W. Twilley, and W. Parker, “The
30. S. Brohez and C. Delvosalle, “Carbon Dioxide Gen- Effect of Specimen Edge Conditions on Heat
eration Calorimetry-Errors Induced by the Release Rate,” Fire and Materials, 17, pp. 51–63
Simplifying Assumptions in the Standard Test (1993).
Methods,” Fire and Materials, 33, pp. 89-97 (2009). 48. B. Östman and L. Tsantaridis, “Communication:
31. A. Roberts, “Ultimate Analysis of Partially Retainer Frame Effects on Cone Calorimeter Results
Decomposed Wood Samples,” Combustion and for Building Products,” Fire and Materials, 17, pp.
Flame, 8, pp. 345–346 (1964). 43–46 (1993).
32. V. Babrauskas, “The Generation of CO in Bench- 49. V. Babrauskas, W. Twilley, M. Janssens, and
Scale Fire Tests and the Prediction for Real-Scale S. Yusa, “A Cone Calorimeter for Controlled Atmo-
Fires,” in Proceedings of the 1st International Fire sphere Studies,” Fire and Materials, 16, pp. 37–43
and Materials Conference, Interscience (1992).
Communications, London, UK (1992). 50. Y. Xin and M. Khan, “Flammability of Combustible
33. E. Buc, “Oxidizer Classification Research Project: Materials in Reduced Oxygen Environment,” in
Tests and Criteria,” Fire Protection Research Foun- Proceedings of the 10th Fire and Materials Confer-
dation, Quincy, MA (2009). ence, Interscience Communications, London,
34. M. Janssens, D. Ewan, C. Gomez, M. Hirschler, England (2007).
J. Huczek, R. Mason, K. Overholt and J. M. Sharp, 51. Aircraft Material Fire Test Handbook, DOT/FAA/
“Reducing Uncertainty of Quantifying the Burning CT-89/15, U.S. Department of Transportation,
Rate of Upholstered Furniture”, Final Report for
948 M. Janssens
Federal Aviation Administration, Atlantic City, NJ Communication Cables,” in Proceedings of the 5th
(1990). International Fire and Materials Conference and
52. In Federal Register, 51, Federal Aviation Adminis- Exhibition, Interscience Communications, London,
tration, Washington, DC, pp. 26206–26221 (1986). UK (1998).
53. J. Blomqvist, “Rate of Heat Release of Building 68. J. Carman, D. Price, G. Milnes, and D. Purser,
Materials, Experiments with an OSU Apparatus “Comparison of Heat Release Rates as Measured
Using Oxygen Consumption,” LUTVDG/(TVBB- by Oxygen Depletion and Thermopile Techniques,”
3017), Lund University, Lund, Sweden (1983). in Proceedings of Interflam’99, Interscience
54. H. Tran, “Modifications to an Ohio State University Communications, London, UK (1999).
Apparatus and Comparison with Cone Calorimeter 69. R. Lyon, “Heat Release Kinetics,” Fire and
Results,” in HTD, Vol. 141, Proceedings of the Materials, 24, pp. 179-186 (2000).
AIAA/ASME Thermophysics and Heat Transfer Con- 70. R. Lyon, “Plastics and Rubber,” in Handbook of
ference, American Society of Mechanical Engineers, Building Materials for Fire Protection, (C. Harper,
New York (1990). ed.), McGraw-Hill.: New York, NY, pp. 3.1-3.51
55. Y. Tsuchiya, “Methods of Determining Heat Release (2004).
Rate,” Fire Safety Journal, 5, pp. 49–57 (1982). 71. R. Lyon and M. Janssens, “Polymer Flammability”
56. F. Hshieh and R. Buch, “Controlled-Atmosphere in Encyclopedia of Polymer Science & Engineering
Cone Calorimeter Studies of Silicones,” Fire and (On-line Edition), John Wiley & Sons: New York,
Materials, 21, pp. 265-272 (1997). NY (2005).
57. J. Leonard, P. Bowditch and V. Dowling, “Develop- 72. R. Lyon, R. Walters and S. Stoliarov, “A Thermal
ment of a Controlled-Atmosphere Cone Calorime- Analysis Method for Measuring Polymer Flamma-
ter,” Fire and Materials, 24, pp. 143-150 (2000). bility,” Journal of ASTM International, 3, pp. 1-18
58. C. Gomez, A. Zalkin and M. Janssens, “Using the (2006).
Cone Calorimeter for Quantifying Toxic Potency,” 73. R. Lyon, R. Walters, N. Safronava and S. Stoliarov,
in Proceedings of Interflam 2010, Interscience “In A Statistical Model for the Results of Flamma-
Communications, London, England (2010). bility Tests,” in Proceedings of the 11th Interna-
59. C. Gomez, M. Janssens and A. Zalkin, “Measuring tional Fire and Materials Conference, Interscience
Yields of Toxic Gases from Materials during Differ- Communications Limited, London, England,
ent Stages of Fire Development,” in Proceedings of pp. 141-159 (2009).
the 12th Fire and Materials Conference, Interscience 74. L. Cooper, “Some Factors in the Design of a Calo-
Communications, London, England (2011). rimeter Hood and Exhaust,” Journal of Fire Protec-
60. A. Tewarson, “Heat Release Rates from Burning tion Engineering, 6, pp. 99–112 (1994).
Plastics,” Journal of Fire and Flammability, 8, pp. 75. P.H. Thomas, P.L. Hinkley, C.R. Theobald, and
115–130 (1977). D.L. Simms, Fire Technical Paper No. 7,
61. A. Tewarson, “Reliable Small-Scale Fire Testing H.M. Stationary Office, Joint Fire Research Organi-
Apparatus,” Modern Plastics, 57, 11, pp. 58–62 zation, London, UK (1963).
(1980). 76. V. Babrauskas, J. Lawson, W. Walton, and
62. A. Tewarson and R. Pion, “Flammability of Plastics. W. Twilley, “Upholstered Furniture Heat Release
I. Burning Intensity,” Combustion and Flame, Rates Measured with a Furniture Calorimeter,”
26, pp. 85–103 (1976). NBSIR 82-2604, National Bureau of Standards,
63. A. Tewarson, J. Lee, and R. Pion, “The Influence of Gaithersburg, MD (1982).
Oxygen Concentration on Fuel Parameters for Fire 77. S. Ames and S. Rogers, “Large and Small Scale Fire
Modeling,” in Proceedings of the 18th Symposium Calorimetry Assessment of Upholstered Furniture,”
(International) on Combustion, Combustion Insti- in Proceedings of Interflam ‘90, Interscience
tute, Pittsburgh, PA (1981). Communications, London, UK (1990).
64. B. Östman, G. Svensson, and J. Blomqvist, “Com- 78. J, Ezinwa, J. Rigg, D. Torvi and E. Weckman
parison of Three Test Methods for Measuring Rate “Effects of Ignition Location on Flame Spread and
of Heat Release,” Fire and Materials, 9, pp. 176–184 Heat Release Rates in Furniture Calorimeter Tests of
(1985). Polyurethane Foams,” in Proceedings of the 11th
65. V. Babrauskas, “Comparative Rates of Heat Release International Fire and Materials Conference,
from Five Different Types of Test Apparatuses,” pp. 645-656 (2009).
Journal of Fire Sciences, 4, pp. 148–159 (1986). 79. T. Ohlemiller, “Flammability Tests of Full-Scale
66. B. Kandola,, A. Horrocks, K. Padmore, J. Dalton and Mattresses: Gas Burners versus Burning Bed-
T. Owen, “Comparison of Cone and OSU Calorimet- clothes,” NISTIR 7006, National Institute of
ric Techniques to Assess the Flammability Standards and Technology, Gaithersburg, MD
Behaviour of Fabrics Used for Aircraft Interiors,” (2003).
Fire and Materials, 30, pp. 241-255 (2006). 80. T. Ohlemiller and R. Gann, “Effect of Bed Clothes
67. P. Gandhi, L. Caudill, and T. Chapin, “Comparison Modifications on Fire Performance of Bed
of Cone Calorimeter Data with FM 3972 for Assemblies,” NIST Technical Note 1449, National
www.ebook777.com
27 Calorimetry 949
Institute of Standards and Technology, Gaithersburg, 94. M. Dahlberg, “The SP Industry Calorimeter—For
MD (2003). Rate of Heat Release Rate Measurements up to
81. T. Ohlemiller, T. Shields, A. McLane, and 10 MW,” SP Report, 43, National Testing Institute
R. Gann, “Flammability Assessment Methodology (SP), Borås, Sweden (1992).
for Mattresses,” NISTIR 6497, National Institute of 95. B. Östman and L. Tsantaridis, “Correlation Between
Standards and Technology, Gaithersburg, MD Cone Calorimeter Data and Time to Flashover in the
(2000). Room Fire Test,” Fire and Materials, 18, pp.
82. T. Ohlemiller and K. Villa, “An Investigation of the 205–209 (1994).
California Technical Bulletin 133 Test, Part II: 96. V. Babrauskas, “Specimen Heat Fluxes for Bench-
Characteristics of the Ignition Source and a Compa- Scale Heat Release Rate Testing,” Fire and
rable Gas Burner,” NBSIR 90-4348, National Bureau Materials, 19, pp. 243–252 (1995).
of Standards, Gaithersburg, MD (1990). 97. U. Wickström and U. Göransson, “Full-Scale/
83. W. Parker, K. Tu, S. Nurbakhsh, and G. Damant, Bench-Scale Correlations of Wall and Ceiling
“An Investigation of the California Technical Bulle- Linings,” Fire and Materials, 16, pp. 15–22 (1992).
tin 133 Test, Part III: Full Scale Chair Burns,” 98. E. Smith and T. Green, “Release Rate Tests for a
NBSIR 90-4375, National Bureau of Standards, Mathematical Model,” in Mathematical Modeling of
Gaithersburg, MD (1990). Fires, ASTM STP 983, American Society of Testing
84. J. Krasny and W. Parker, “Impact of the Room and Materials, Philadelphia (1987).
Enclosure on the Peak Heat Release Rates of 99. H. Mitler, “Predicting the Spread Rates of Fires on
Upholstered Furniture,” in Proceedings of the Fire Vertical Surfaces,” in Proceedings of the 23rd Sym-
and Materials 4th International Conference and posium (International) on Combustion, Combustion
Exhibition, Interscience Communications, London, Institute, Pittsburgh, PA (1991).
UK (1995). 100. M. Janssens, “Cone Calorimeter Measurements of
85. I. Benjamin, “Development of a Room Fire Test,” in the Heat of Gasification of Wood,” in Proceedings
Fire Standards and Safety, ASTM STP 614, Ameri- of Interflam ‘93, Interscience Communications,
can Society of Testing and Materials, Philadelphia London, UK (1993).
(1977). 101. S. Dillon, W. Kim, and J. Quintiere, “Determination
86. F. Fisher and R. Williamson, “Intralaboratory Eval- of Properties and the Prediction of the Energy
uation of a Room Fire Test Method,” NBS GCR Release Rate of Materials in the ISO 9705 Room-
83-421, National Bureau of Standards, Gaithersburg, Corner Test,” NIST-GCR-98-753, National Institute
MD (1983). of Standards and Technology, Gaithersburg, MD
87. C. Lee, “Standard Room Fire Test Development at (1998).
the National Bureau of Standards,” in Fire Safety: 102. V. Petrella, “The Mass Burning Rate of Polymers,
Science and Engineering, ASTM STP 882, American Wood and Liquids,” Journal of Fire and Flammabil-
Society of Testing and Materials, Philadelphia, ity, 11, pp. 3–21 (1980).
pp. 29–42 (1985). 103. M. Janssens, Thermophysical Properties of Wood
88. B. Sundström and U. Wickström, “Fire: Full Scale and Their Role in Enclosure Fire Growth, University
Tests,” SP-RAPP, 14, National Testing Institute of Ghent, Ghent, Belgium (1991).
(SP), Borås, Sweden (1980). 104. J. Urbas and W. Parker, “Surface Temperature
89. B. Sundström and U. Wickström, “Fire: Full Scale Measurements on Burning Wood Specimens in the
Tests, Calibration of Test Room,” SP-RAPP, Cone Calorimeter and Effect of Grain Orientation,”
48, National Testing Institute (SP), Borås, Sweden Fire and Materials, 17, pp. 205–208 (1993).
(1981). 105. M. Sibulkin, “Heat of Gasification for Pyrolysis of
90. A. Ahonen, C. Holmlund, and M. Kokkala, “Effects Charring Materials,” in Proceedings of the 1st Inter-
of Ignition Source in Room Fire Tests,” Fire Science national Symposium on Fire Safety Science, Interna-
and Technology, pp. 1–13 (1987). tional Association for Fire Safety Science, London,
91. J. Zhang, M. Delichatsios and, M. Colobert, “Assess- UK (1985).
ment of Fire Dynamics Simulator for Heat Flux and 106. J. Urbas, “Non-Dimensional Heat of Gasification
Flame Heights Predictions from Fires in SBI Tests,” Measurements in the Intermediate Scale Rate of
Fire Technology, 46, pp. 291-306 (2010). Heat Release Apparatus,” Fire and Materials,
92. E. Smith, N. Marshall, K. Shaw, and S. Colwell, 17, pp. 119–123 (1993).
“Correlating Large-Scale Fire Performance with the 107. J. Quintiere, “A Simulation Model for Fire Growth
Single Burning Item Test,” in Proceedings of on Materials Subject to a Room-Corner Test,” Fire
Interflam’01, 9th International Fire Conference, Safety Journal, 20, pp. 313–339 (1992).
Interscience Communications, London, UK (2001). 108. R. Peacock, P. Reneke, W. Jones, R. Bukowski, and
93. G. Heskestad, “A Fire Products Collector for Calo- G. Forney, “A User’s Guide for CFAST: Engineering
rimetry into the MW Range,” Technical Report Tools for Fire Growth and Smoke Transport,” Spe-
FMRC J.I0C2E1.RA, Factory Mutual Research Cor- cial Publication 921, National Institute of Standards
poration, Norwood, MA (1981). and Technology, Gaithersburg, MD (2000).
950 M. Janssens
109. J. Mangs and O. Keski-Rahkonen, “Characterization Consumer Products Safety Commission, Washington,
of the Fire Behavior of a Burning Passenger Car. Part DC (2006).
I: Car Fire Experiments,” Fire Safety Journal, ASTM D3173, Standard Test Method for Moisture in the
23, pp. 17–35 (1994). Analysis Sample of Coal and Coke, ASTM Interna-
110. M. Shipp and M. Spearpoint, “Measurements of the tional, West Conshohocken, PA.
Severity of Fires Involving Private Motor Vehicles,” ASTM D5373, Standard Test Methods for Instrumental
Fire and Materials, 19, pp. 143–151 (1995). Determination of Carbon, Hydrogen, and Nitrogen in
111. C. Joyeux, “Natural Fires in Closed Car Parks,” Laboratory Samples of Coal, ASTM International,
INC-96/294d-DJ/NB, Centre Technique Industriel West Conshohocken, PA.
de la Construction Métallique (CTICM), Saint- ASTM D5865, Standard Test Method for Gross Calorific
Aubin, France (1997). Value of Coal and Coke, ASTM International, West
112. C. Steinert, “Experimental Investigation of Burning Conshohocken, PA (2007).
and Fire Jumping Behavior of Automobiles ASTM D7309, Standard Test Method for Determining
(in German),” VFDB Journal, 49, pp. 163–172 Flammability Characteristics of Plastics and Other
(2000). Solid Materials Using Microscale Combustion Calo-
113. C. Joyeux, J. Kruppa, L. Cajot, J. Schleich, P. van de rimetry, ASTM International, West Conshohocken,
Leur, and L. Twilt, “Demonstration of Real Fire PA (2007).
Tests in Car Parks and High Buildings,” Centre ASTM E906, Standard Test Method for Heat and Visible
Technique Industriel de la Construction Métallique Smoke Release Rates for Materials and Products,
(CTICM), Saint-Aubin, France (2002). ASTM International, West Conshohocken, PA (2007).
114. Y. Shintani, N. Kakae, K. Harada, H. Masuda, and ASTM E1317, Standard Test Method for Flammability of
W. Takahashi, “Experimental Investigation of Burn- Marine Surface Finishes, ASTM International, West
ing Behavior of Automobiles,” in Proceedings of the Conshohocken, PA (2008).
6th Asia-Oceania Symposium on Fire Science and ASTM E1354, Standard Test Method for Heat and Visible
Technology, International Association for Fire Smoke Release Rates for Materials and Products
Safety Science, London, UK (2004). Using an Oxygen Consumption Calorimeter, ASTM
115. B. Zhao and J. Kruppa, “Structural Behavior of an International, West Conshohocken, PA (2008).
Open Car Park Under Real Fire Scenarios,” Fire and ASTM E1537, Standard Test Method for Fire Testing of
Materials, 28 (2004). Upholstered Furniture, ASTM International, West
116. H. Persson, “Commodity Classification—A More Conshohocken, PA (2007).
Objective and Applicable Methodology,” SP Report, ASTM E1590, Standard Test Method for Fire Testing of
70, National Testing Institute (SP), Borås, Sweden Mattresses, ASTM International, West
(1993). Conshohocken, PA (2007).
117. D. Evans and D. Stroup, “Methods to Calculate the ASTM E1623, Test Method for Determination of Fire and
Response Time of Heat and Smoke Detectors Thermal Parameters of Materials, Products, and
Installed Below Large Unobstructed Ceilings,” Systems Using an Intermediate Scale Calorimeter
NBSIR 85-3167, National Bureau of Standards, (ICAL), ASTM International, West Conshohocken,
Gaithersburg, MD (1985). PA (2004).
118. M. Dahlberg, “Error Analysis for Heat Release Rate ASTM E1822, Standard Test Method for Fire Testing of
Measurements with the SP Industry Calorimeter,” Stacked Chairs, ASTM International, West
SP Report, 29, National Testing Institute (SP), Conshohocken, PA (2007).
Borås, Sweden (1994). ASTM E2058, Standard Test Methods for Measurement
119. P. Enright and C. Fleischmann, “Uncertainty of Heat of Synthetic Polymer Material Flammability Using a
Release Rate Calculation of the ISO 56601—Cone Fire Propagation Apparatus, ASTM International,
Calorimeter Standard Test Method,” Fire Technol- West Conshohocken, PA (2006).
ogy, 35, pp. 153–169 (1999). ASTM E2067, Standard Practice for Full-Scale Oxygen
120. J. Urbas, “BDMC Interlaboratory Cone Calorimeter Consumption Calorimetry Fire Tests, ASTM Interna-
Test Program,” Fire and Materials, 26, pp. 29–35 tional, West Conshohocken, PA (2008).
(2002). ASTM E2257, Standard Test Method for Room Fire Test
121. M. Janssens, “Uncertainty of Fire Test Results,” in of Wall and Ceiling Materials and Assemblies, ASTM
Proceedings of Interflam ‘07, Interscience International, West Conshohocken, PA (2008).
Communications, London, UK (2007). CAL TB 133, Flammability Test Procedure for Seating
Furniture for Use in Public Occupancies, California
Bureau of Home Furnishings and Thermal Insulation,
North Highlands, CA (1991).
Codes and Standards CAL TB 603, Requirements and Test Procedure for
Resistance of a Mattress/Box Spring Set to a Large
16 CFR 1633, Standard for the Flammability (Open Open-Flame, California Bureau of Home Furnishings
Flame) of Mattresses and Mattress/Foundation Sets, and Thermal Insulation, North Highlands, CA (2004).
www.ebook777.com
27 Calorimetry 951
EN 13501, Fire Classification of Construction Products NFPA 220, Standard on Types of Building Construction,
and Building Elements—Part 1: Classification Using National Fire Protection Association, Quincy, MA
Test Data from Reaction-to-Fire Tests, European (2006).
Committee for Standardization (CEN), Brussels, NFPA 259, Standard Test Method for Potential Heat of
Belgium (2002). Building Materials, National Fire Protection Associa-
EN 13823, Reaction to Fire Tests for Building Products— tion, Quincy, MA (2003).
Building Products Excluding Flooring Exposed to the NFPA 265, Standard Methods of Fire Tests for
Thermal Attack of a Single Burning Item, European Evaluating Room Fire Growth Contribution of Textile
Committee for Standardization (CEN), Brussels, Coverings on Full Height Panels and Walls, National
Belgium (2002). Fire Protection Association, Quincy, MA (2007).
FM 4995, Approval Standard for Commodity Classifica- NFPA 271, Standard Method of Test for Heat and Visible
tion of Idle Plastic Pallets, FM Global Research, Smoke Release Rates for Materials and Products
Norwood, MA (1992). Using an Oxygen Consumption Calorimeter, National
ISO 1716, Reaction to Fire Tests for Building Products— Fire Protection Association, Quincy, MA (2004).
Determination of the Calorific Value, International NFPA 286, Standard Methods of Fire Tests for
Organization for Standardization, Geneva, Evaluating Contribution of Wall and Ceiling Interior
Switzerland (2002). Finish to Room Fire Growth, National Fire Protection
ISO 5660-1, Reaction-to-Fire Tests—Heat Release, Association, Quincy, MA (2006).
Smoke Production and Mass Loss Rate—Part 1: NFPA 5000®, Building Construction and Safety Code®,
Heat Release Rate (Cone Calorimeter Method), Inter- National Fire Protection Association, Quincy, MA
national Organization for Standardization, Geneva, (2006).
Switzerland (2002). NT Fire 032, Upholstered Furniture, Burning Behavior—
ISO 5660-2, Reaction-to-Fire Tests—Heat Release, Full-Scale Test, NORDTEST, Helsinki, Finland
Smoke Production and Mass Loss Rate—Part 2: (1991).
Smoke Production Rate (Dynamic Measurement),
International Organization for Standardization,
Geneva, Switzerland (2002). Dr. Marc Janssens is a Senior Engineer at Southwest
ISO 9705, Fire Tests—Reaction-to-Fire—Room Fire Research Institute in San Antonio, Texas. His research has
Test, International Organization for Standardization, focused on computer fire modeling, fire hazard and risk
Geneva, Switzerland (1993). assessment, fire test standards development, and the
NFPA 101®, Life Safety Code®, National Fire Protection experimental and theoretical evaluation of material flam-
Association, Quincy, MA (2006). mability with emphasis on heat release calorimetry.
The Cone Calorimeter

28
Vytenis Babrauskas
Introduction instead provides the reader with an overall feel

for the equipment. Space is not available in this
Chapter 27 describes the history and develop- handbook to fully discuss the applications of
ment of techniques for measuring heat release cone calorimeter data, apart from the review of
rate (HRR). This chapter outlines features and data given in Chap. 26. Extensive guidance on
details of today’s preferred instrument for mea- using cone calorimeter data is given in a textbook
suring bench-scale HRR—the cone calorimeter. on this subject [4]. It also provides example data
Other cone calorimeter measuring functions are compilations and information on using cone cal-
1. Effective heat of combustion orimeter data for predictions of fires.
2. Mass loss rate
3. Ignitability
4. Smoke and soot Summary of Features
5. Toxic gases
The cone calorimeter is based on the concept A schematic view of the cone calorimeter is
of oxygen consumption calorimetry, which is shown in Fig. 28.1. Figure 28.2 shows a commer-
also presented in Chap. 27. cial instrument, and Fig. 28.3 identifies some of
This chapter provides both an introduction to the major components. The more salient opera-
and description of cone calorimeter measurement tional features and limits of the apparatus are
technology. The cone calorimeter has recently
Specimen size 100 100 mm, thickness
assumed a dominant role in bench-scale fire test- of 6–50 mm
ing of various products; therefore, an emphasis Specimen orientation Horizontal, face up (standard
will be placed on the why of various design testing) or vertical (reserved
features. When conducting tests, the cone calo- for exploratory studies)
rimeter operator needs to consult several other Specimen back-face Very low loss insulating
conditions ceramic fibrous material
documents. Testing will presumably be in con-
Load cell live load 500 g
formance with either ISO 5660 [1] or ASTM capacity
E1354 [2]. In addition, the “User’s Guide for Load cell tare capacity 3.5 kg
the Cone Calorimeter” [3] should be consulted. Load cell resolution 0.005 g
This chapter does not emphasize the operational Ignition Electric spark
aspects documented in these references but Heating flux range 0–110 kW · m2
Flux uniformity, Typically 2 %
horizontal
V. Babrauskas (*) Flux uniformity, vertical Typically 7 %
Fire Science and Technology Inc. (continued)

www.ebook777.com
28 The Cone Calorimeter 953
Sensing principle Oxygen consumption, only 1. comparative evaluation of materials;

Maximum instantaneous In excess of 20 kW 2. obtaining of thermophysical constants (fire
output properties) of materials;
Normally calibrated range 0–12 kW 3. as input data to fire models or engineering
Linearity over 0–12 kW 5% calculation;
range 4. for regulatory compliance.
Noise intrinsic to oxygen 20 ppm O2
meter
Noise in HRR 2.5 %
measurement, over Comparative Evaluation of Materials
0–12 kW range
Smoke meter operating 0–20 m1 (linear) Comparative evaluation of materials is the
range easiest and simplest use of cone calorimeter
Smoke meter resolution 0.01 m1
data. This, in fact, is also where the largest
Soot sampler mass 0–1 part in 200 (of exhaust
amount of published literature involving cone
fraction range gas flow)
calorimeter data is found, of which the fire
retardants field is a prominent example. There
Uses of Cone Calorimeter Data have been hundreds of papers published examin-
ing fire retardant formulations with the use of the
Cone calorimeter data are primarily used for four cone calorimeter. For such studies, modeling or
purposes: large-scale testing is inappropriate, since the
Laser extinction beam including

temperature measurement
Temperature and differential pressure
measurements taken here
Soot sample tube location
Exhaust
blower
Exhaust
hood
Gas samples Cone heater

taken here
Soot collection filter Spark
Controlled igniter
flow rate
Sample
Load cell
Fig. 28.1 Schematic view of the cone calorimeter

954 V. Babrauskas
same polymer formulation can be used for a wide

array of products. Thus, cone calorimeter data
are normally used and a comparative evaluation
is made. Most commonly, candidate materials
are evaluated simply by comparing their peak
HRR values. This approach is not adequate if
flame spread in the real-life environment is sig-
nificant, i.e., if the material is not quickly ignited
over its entire face. For taking flame spread into
account, albeit in a simplified way, Babrauskas
00
[5] proposed in 1984 that the variable q_ =tig be
used, which is the ratio of the HRR value to the
ignition time. The ignition time was shown to be
correlated to flame spread rate, thus, this hazard
parameter increases with both increasing HRR
and increasing propensity for rapid flame spread.
A reasonable semi-quantitative prediction of the
time to flashover was possible using this ratio for
various wall lining materials. Petrella [6] later
00
proposed a modified rating system where q_ =tig
is plotted on one axis, while total heat released
is plotted on the other. Materials of better perfor-
00
Fig. 28.2 A commercial cone calorimeter (Photo cour- mance have both a low q_ =tig value and a low
tesy Fire Testing Technology, Ltd.) total heat release. The most refined scheme
Pressure ports
Orifice plate,
orifice size is Thermocouple (located
1/2 I.D. of stack on stack center line)
685* mm 57 mm* dia. orifice

Gas sample
114 mm dia. duct 127 mm
Blower 140 mm
Hood
Blower Rubber
motor vibration
530*
mm mounts
Sample
1625 mm
1680 mm 686 mm
*Indicates a critical dimension
Fig. 28.3 View of major components of the cone calorimeter
www.ebook777.com
which is still simple is the one put forth in 1991 is possible to obtain fire properties from cone
by Cleary and Quintiere [7]. They introduced a calorimeter data. This topic is treated at length
parameter b: in the Ignition Handbook [8], but here the most
useful computation will be identified. For ther-
00 tig
b ¼ 0:01q_ avg 1 ð28:1Þ mally thick materials, Janssens derived the
tb
relationship:
00
where q_ avg ¼ average HRR (kW m2) at a 2 !0:55 3
50 kW m2 irradiance, tig ¼ ignition time (s), λρC
q_ e ¼ q_ cr 41 þ 0:73 2 5
00 00
ð28:2Þ
and tb ¼ duration of flaming (s). They showed heff tig
that materials which show b < 0.4 have negli-
gible propensity to spread fire, while those with According to this, if experimental data are plot-
00
progressively higher values show increased ted (Fig. 28.4) such that q_ e is put on the x-axis
hazard in full-scale applications. The Cleary/
and t0:55
ig on the y-axis, then the data will fall in a
Quintiere b is not to be confused with Spalding’s 00
B number, sometimes use to characterize hazards straight line, with the x-axis intercept being q_ cr .
00 00
of burning liquids. Here q_ e ¼ irradiance ðkWm2 Þ, q_ cr ¼ x-axis
intercept, tig ¼ ignition time (s), and λρC is the
thermal inertia (kJ2 m4 s1 K2) of the speci-
Obtaining Thermophysical Constants men. From such a plot, the value of thermal
of Materials inertia can be computed, which is an effective
fire property of importance in both ignition and
The HRR of materials cannot be computed from flame spread problems.
some ostensibly simple material fire properties,
but is rather a complex relationship governed by
chemical (reaction kinetics), thermal (heat trans- Input Data for Fire Models or
fer properties), and mechanical (cracking, delam- Calculations
ination, etc.) properties. Thus, in general, it is not
possible to deduce some underlying material fire A number of correlational schemes for making
properties from HRR data. However, the situa- engineering calculations on various types of
tion is more amenable for ignition data, where it commodities have been developed which are
Fig. 28.4 The main

variables of the
ignitability plot
Transformed ignition time (t-0.55)
Minimum flux
0
0 Irradiance (kW m-2)
Critical flux
956 V. Babrauskas
based on Cone Calorimeter data. These are flame flux. Hostikka and Axelsson [24] showed
reviewed in Chap. 26. For more refined models, an interesting example of CFD modeling by
i.e., zone or CFD models for room fires, the predicting the flame flux in the cone calorimeter.
application is more difficult. This is because the
HRR is strongly a function of the irradiance. But
in most real fires, the irradiance received by any Regulatory Compliance
particular locale is a dynamic function of time
and is not a constant. Because of this difficulty, it The New Zealand building code specifies use of
has become more common for modern-day com- the ISO version of the cone calorimeter standard,
puter codes, e.g., FDS, to adopt a pyrolysis ISO 5660 [1], for external wall cladding
model, rather than using small-scale experimen- products. The Building Code of Australia uses
tal HRR data as an input. A pyrolysis model cone calorimeter testing to assess fire retardant
effectively is a scheme where the HRR of a treated wood for use in bushfire-prone areas. The
small area of material is computed from some building code of Japan uses cone calorimeter
sort of input data. But, as discussed above, for testing extensively, as a primary measure of the
realistic materials there generally is no simple fire performance of surface lining materials [25].
series of expression that would be able to predict In their application, tests are run at 50 kW m2
the HRR, based on the input of a modest collec- irradiance for a duration of 5, 10, or 20 min,
tion of constants. Even if the constants can be depending on the classification sought. In each
defined, they must in turn be obtained from case, a peak HRR value below 200 kW m2 must
experiments, and this is already known to be be found, with the total heat release being less
difficult in the first place. CFD models however than 8 MJ m2, with the latter being a particu-
may have an option to input small-scale HRR larly onerous requirement. Taiwan has also
data; typically in that case the HRR at a fixed adopted similar provisions. IMO, the Interna-
irradiance is used. Capote et al. [9] illustrated tional Maritime Organization, which provides
such an approach in modeling train fires with the regulations for constructing of sea-going
FDS. Aksit et al. [10] described use of cone vessels, uses cone calorimeter testing for
calorimeter input data for modeling cable acceptance of “fire restricting material” for
tray fires with SOFIE, while Andersson [11] high-speed craft in the case of furniture and
described a more general effort with SOFIE. related materials.
Tsai et al. [12] described a proprietary CFD
model using cone calorimeter data; the model
was used solely for calculating ignition behavior Operating Principle
of materials.
For zone fire models, the most successful It is emphasized at this point that the cone calo-
example has been the BRANZfire model of rimeter has been designed to use only oxygen
Wade [13–15]. Lattimer et al. [16] described a consumption calorimetry as its measurement
module for CFAST based on cone calorimeter principle [26]. Other calorimeters that on occa-
input data. Janssens and Dillon [17] described a sion use oxygen consumption principles, for
simple room fire model based on cone calorime- example, the Factory Mutual Research Corpora-
ter data. Cone calorimeter data have also been tion (FMRC) flammability apparatus (Chap. 27),
used in an application simpler than room fires, sometimes incorporate a sensible enthalpy flow
the prediction of upward flame spread on vertical measurement technique to arrive at the convec-
panels [18–23]. tive component of the heat release rate. In the
For some models, it is necessary to know the design of the cone calorimeter, such an approach
total heat flux incident on the specimen, not just was deemed to be misleading. The implicit
the external heat flux; the total heat flux is assumption behind this type of measurement is
comprised of the external heat flux, plus the that the fraction of the total heat release being
www.ebook777.com
manifest as the sensible flow enthalpy is a prop- up, and calibrating oxygen analysis systems is
erty of the material being tested. Such is not, in given in Twilley and Babrauskas [3] and in
fact, the case. The convective fraction is depen- Babrauskas and Grayson [4].
dent on details of the apparatus design and also
on the scale of the specimen [27].
Where high-quality results are required, such The Radiant Heater
as in the cone calorimeter, current-day practice
demands that a paramagnetic oxygen analyzer be After establishing the operating principle, the
used. The various manufacturers use measuring next most important feature is the type of heater.
schemes that differ in detail, but all rely on the In general, such a heater should be able to
same paramagnetic principle whereby the sens- achieve adequately high irradiances, have a rela-
ing element is sensitive to the partial pressure of tively small convective heating component, pres-
oxygen in the cell. The most significant ent a highly uniform irradiance over the entire
interferents to this detection principle are NO exposed face of the specimen, and be designed so
and NO2, both of which show a strong paramag- as not to change its irradiance when the main
netic response, but not as strong as that of oxy- voltage varies, when heater element aging
gen. Interferents are never a problem in fire occurs, or when the apparatus retains some resid-
testing, however, since O2 levels measured are ual heat from the exposure given to a prior
10–21%, whereas concentrations of NOx are specimen.
rarely above 100 ppm.
Unlike in applications where oxygen levels Range of Heat Fluxes Needed for Testing A
are monitored as simply one of many indications room fire burning near its maximum rate can
of fire hazard, in HRR work it is essential that the show gas temperatures over 1000 C, producing
instrumentation be designed for the highest pos- corresponding irradiances to walls and contents
sible resolution. Thus, both the ASTM and ISO of 150 kW · m2. Testing under such extreme
standards specify that the short-term noise + conditions may not be required; nonetheless, if
drift of the oxygen analyzer must be less than postflashover fires are to be simulated,
or equal to 50 ppm O2. The best-grade commer- irradiances of over 75 kW · m2 should be avail-
cial instruments are able to meet a 20 ppm O2 able, and preferably closer to 100 kW · m2. A
limit. In addition, the standards provide a signifi- significant convective component would negate
cant amount of detail on the layout of the gas the purpose of having a radiant ignition test.
sampling system, including desiccation, mass Rather low convective fluxes can be achieved
flow control, and bypass flows. All of these for specimens oriented horizontally, face up,
aspects have to be in conformance with the and with the prevailing airflow being upwards.
specifications for good repeatability and repro- For vertical specimens, orientation is considered,
ducibility performance (see Fig. 28.3) to be and it becomes evident that a boundary layer will
achieved. normally be expected to develop that will add
Because the detection principle responds to some convective component. The convective
oxygen partial pressure, there needs to be a com- boundary layer component is not uniform over
pensation for changes in atmospheric pressure, the height of a specimen; thus it is seen that better
either with a mechanical back-pressure regulator uniformity can also be expected under conditions
or by measuring the pressure and correcting elec- where the convective component is minimized.
trically. Without compensation, there can be sig-
nificant error in the calculated heat release rate. Choice of Heater Type In a real fire, the igni-
Carbon dioxide, the other major component tion source is, in most cases, in the vicinity of a
expected to be in the oxygen analyzer, causes combustible. The radiation spectrum depends on
less than 0.3 % error in the oxy-gen reading. the size of the fire. A very small fire can show a
Extensive practice advice on selecting, setting substantial fraction of its radiation at
958 V. Babrauskas
wavelengths characteristic of H2O, CO2, and influenced by work at NIST with earlier types of
other combustion products [28]. For larger calorimeters. One of the primary requirements of
fires—certainly for room fires reaching a hazard- the heater is that it not change the irradiance
ous condition—the radiation from the soot tends impressed on the specimen when the specimen
to dominate. The result is an approximation to a ignites. This undesired event is, of course,
graybody radiation [29]. For such a graybody exactly what happens with several of the older
radiation the temperature is typically in the types of calorimeters. The specimen’s flames
vicinity of 1000 C [30]. Experimentally, heater directly heat nearby ironwork, which, in turn,
choices for test apparatuses have included radiates to the specimen. The heater, which had
gas-fired panels, electric resistance heaters, been viewing a cold specimen prior to ignition,
flames, and high-temperature lamps. Electrical also starts to view a hot flame afterwards. The
heaters tend to have a near-graybody characteris- result is that its efficiency increases drastically,
tic and, assuming a dull or oxidized surface con- giving a rise to its radiating temperature. Based
dition, a high emissivity. Gas-fired panels derive on these observations, guidelines were
a substantial portion of their radiation from the formulated so that the specimen must, as much
ceramic face; thus, while there are discrete as possible, view only
molecular wavelength peaks, overall the radia- 1. A temperature-controlled heater
tion shows a graybody continuum, typically in 2. A water-cooled plate
the range of 700–1000 C [31]. High-temperature 3. The open-air, ambient-temperature environment
lamps, which have been used by several Reliance on item 2 increased costs signifi-
investigators [28, 32], typically have radiating cantly; thus, it was more desirable to use only
temperatures of 2200–3000 C. The spectral dis- items 1 and 3. Prior to the development of
tribution of such a source—further limited by a the cone calorimeter, fire test apparatuses
translucent enclosure—is much different from typically controlled the power (or fuel rate) into
one operating at 1000 C. Whether this change the heater, but did not maintain it at a fixed
in spectral characteristics is important depends on temperature.
the surface of the material to be ignited. For a
material with a radiant absorbance independent of The Conical Shape The cone calorimeter
wavelength, this source variation would not mat- derives its name from the conical shape of the
ter. Hallman, however, has reported data for a heater (Fig. 28.5). The decision had been made to
large number of plastics and shows that although use an electric resistance heater, running at a
there are some specimens with negligible wave- realistic maximum temperature of about
length dependence to their absorbance, the major- 950 C, but its material and shape still had to be
ity shows strong variations [28]. Hallman also determined. Based on poor experiences with
measured ignition times of plastics with both a exposed-wire resistance heaters and with silicon
flame source and high-temperature lamps. The carbide rod–type heaters, the tube heater was
effect on ignition times ranges from negligible chosen. The tube heater consists of a resistive
to more than an order of magnitude, depending wire element inside a protective tube, swaged
on the specimen. For a general-purpose test, over a packing of inorganic insulation. The tube
flames would probably be the least desirable is made of Incoloy™ and can be bent to a desired
source of heating. For a bench-scale test, flame shape.
size has to be kept small. This means that such To determine the best shape, the conical
flames are optically thin, their emissivity is low, heater used in the ISO 5657 ignitability apparatus
and higher heat fluxes cannot be achieved unless a [33] was examined. This seemed to be a
strong convective component is added. promising shape. The proper shape had to have
a hole in the middle, since otherwise a hot spot
Design Details Once an electrical radiant heater would occur at the sample center, where the
had been decided upon, design details were also radiation view factor is the highest. The same
www.ebook777.com
Fig. 28.5 Cross-sectional Outer shell

view through the cone
heater Thermocouple Inner shell
80* mm
65*
mm 46*
mm
160* mm
Spacer block Ceramic fiber packing
Heating element
Cone hinge and mount bracket
heater had to serve in both horizontal and vertical where a specimen is situated. Over the range of
orientations. In the horizontal orientation, it was fluxes of 10–90 kW · m2, Janssens determined
essential that all the products of combustion flow the ε F product to be 0.73, with F being
out the hole in the middle, and not “splash” on computed as 0.78. Then, solving for ε gives
the heater coil itself, nor escape from the under- ε ¼ 0.91. Since the temperatures of the heater
side. The original ISO 5657 design proved to be closely resemble those in room fires, and the
unsuitable in the former respect. It also had emissivity approaches 1.0, this means that the
problems with durability and assembly. Thus a spectral distribution is likely to be very close to
totally new design was created, which, however, that expected from room fires (neglecting the
looked superficially similar to the ISO 5657 molecular radiation contribution from CO2 and
cone. With the actual cone calorimeter design, H2O). It is important that the heater element be
the flames from the specimen do not splash on kept in good repair, in order that expected uni-
the heater coil. Instead, a sheath of cold air is formity be achieved. Aging may cause the coil
pulled up, surrounding the flame plume. Thus, windings to separate and sag. If this occurs,
there is not a concern that any surface reactions poorer uniformity has been shown to occur [35].
occur on the heater coil.
The space between the inner and outer cones Convective Fraction of the Heating
is packed with refractory fiber. This arrangement Flux During the development of the cone calo-
helps keep the outside of the unit cool and also rimeter at NIST, a study was conducted to deter-
helps bring the heater up to operating tempera- mine the fraction of the heating flux accounted
ture rapidly. for by the convective contribution [36]. When
measured with respect to a water-cooled heat
Emissivity of the Heater The emissivity was flux meter, the results showed that, in the hori-
characterized by Janssens [34]. The heater coil, zontal specimen orientation, the convective con-
once installed and operated a few times, becomes tribution was immeasurably small. In the vertical
essentially radiatively black. The emissivity orientation, the fraction was typically 8–12 %.
itself cannot be directly measured; however, it Janssens later remeasured the vertical configura-
is possible to compute an approximate view fac- tion [34] using a more accurately calibrated heat
tor, F, for the cone heater. The possibility of flux meter and found that, even for the vertical
measurements is based on a simultaneous deter- orientation, the convective transfer is immeasur-
mination of the heater surface temperature and ably small. Thus, it can be stated that the
the heat flux falling on the heat flux meter, with objective of having a test method where the
the meter held in place at the same location heating is primarily radiant was successfully
960 V. Babrauskas
met. For modeling of test results, however, one described [36]. Over the range of irradiances
may be more interested in the possibility of con- from 25 to 100 kW · m2, the ratio of the flux
vective heat transfer to a specimen that is heated, at the specimen center to average flux varied only
or even burning, not to a calibration meter from 1.00 to 1.06. The peak deviations from
constrained by its water-cooling jacket at near- average were typically 2 % in the horizontal
room temperature. Janssens also made some orientation and 7 % in the vertical. Deviations
determinations of such actual specimen heating. are higher in the vertical orientation, since the
The direction of the heat transfer was such as to effect of convective fluxes, due to the boundary
represent a heat loss from the specimen in all layer flow, is more pronounced there. Additional
cases. A single convective heat transfer coefficient measurements have been made in the specimen-
could not be derived, however, since the value was depth plane. Control of the surface of the speci-
dependent on the irradiance level from the heater. men was a special concern to the designers of the
Janssens’s results could be represented by: ISO apparatus, where a special compressive
loading mechanism is provided that attempts to
Irradiance from Convective heat transfer relevel the exposed surface, in case the specimen
heater (kWm2) coefficient hc (Wm2K1) recedes due to melting. In the cone calorimeter,
20 9.0 measurements have been made in the horizontal
40 18.0 orientation using a small, 6-mm-diameter
60 27.0 Gardon-type heat flux gauge. A flux mapping
was obtained starting at the initial surface, and
For practical work, Janssens recommended progressing down to the maximum depth of a
that an average value of hc ¼ 13.5 W · m2K1 specimen, which is 50 mm. A normal aluminum
should be appropriate for work over the common foil rectangular specimen wrap was used for
irradiance range of 20–40 kW · m2. The actual these tests, but without any specimen. The results
details of this small amount of convective heat show that, at heating fluxes of both 25 and
transfer are pertinent only to certain specialized 50 kW · m2, the deviations over the entire spec-
studies. For most work, it is entirely adequate to imen depth are less than 10 %, and can, therefore,
assume that the specimen heating is entirely be neglected (Fig. 28.6). At the lower depths,
radiative. reflection from the aluminum foil probably
assists in maintaining this uniformity.
Uniformity of the Heating Flux The unifor-
mity of the heating flux over the face of the Orientation of the Heater and Specimen The
specimen in the cone calorimeter has been normal orientation of the specimen should be
Fig. 28.6 Measured flux Depth of aluminum-foil boat = 50 mm

at various positions below 50
the top surface of a 49
Measured flux (kW/m2)
specimen 48
47
46
45
25
24
23
22
0 10 20 30 40 50
Depth below top (mm)
www.ebook777.com
Sample
(100 mm × 100 mm size) Chain
Cone hinge and

Aluminum foil mount bracket
Low-density
ceramic wool
Calibration
burner
Sample
pan
Flux
13 mm calcium
meter
silicate board
Sample mount assembly
13 mm calcium Flux meter

Load cell silicate heat shield mount
Fig. 28.7 Heater in the horizontal (standard) orientation
horizontal, face up, with the heater being par- The Shutter
allel, face down. This allows thermoplastics,
liquids, and other melting or dripping samples The original NIST design for the Cone Calorim-
to be successfully tested. Because it was con- eter did not include a shutter. The operator would
sidered desirable to allow testing in a vertical just quickly drop the specimen holder on top of
orientation for certain application exploratory the mount plate at the top of the load cell. This
studies, provision was made to swing the was satisfactory for most building products and
heater 90 into a vertical orientation. Vertical plastics. However, in 1993 researchers at SP
orientation testing may be preferable when (Technical Research Institute of Sweden) found
probing the flame regions or measuring that there were some reproducibility issues when
specimen surface temperatures is desired. testing upholstered furniture specimens that
Figures 28.7 and 28.8 show the comparative ignited very quickly. Thus, they designed a shut-
horizontal and vertical heater orientations, ter (originally described as “heat shield” and later
respectively. It is especially emphasized that as “radiation shield”) to be interposed between
no standard testing should be specified for the heater and the specimen surface; this was
the vertical orientation, even for products that originally described in a 1996 SP report
are normally used in a vertical orientation. The [37]. The use of a shutter makes it possible to
ASTM standard [2] was amended in 1992 to (a) get the load cell to equilibrate before com-
clarify that the vertical orientation is only for mencing exposure, and (b) provide an nearly
special research studies and not for product step-function initiation of radiant heat flux to
testing. the specimen.
962 V. Babrauskas
Spark plug
Vertical sample holder

Sample (100 mm × 100 mm)
Aluminum foil
Low-density Latching
ceramic wool mechanism
Calibration
burner
Retaining Flux meter

clip
Calcium silicate Loadcell
back-up board
Fig. 28.8 Heater in the vertical orientation
However, with the use of a shutter there is ignition results are improperly affected. It must
potentially a different type of error that is also not be so slow that products of combustion
introduced. A shutter will reflect some heat back spill out of the hood. If this were a closed system,
to the heater, and will also rise in temperature and one would also be concerned about airflow being
reradiate heat flux to the heater. Both of these so slow that the air/fuel ratio drops into the fuel-
would cause the heater’s temperature to rise. The rich regime. The standard cone calorimeter, how-
solution adopted by ASTM [2] and ISO [1] ever, has been designed for ambient air testing,
standards was that the shutter should be in place and this consideration does not apply.
for no longer than 10 s prior to start of test, and that Systematic guidance in this area was not
it be either water-cooled with a black coating, or available. However, as an example of the effect
else not water-cooled, but with made of ceramic of airflow, measurements were made at NIST
material or made of reflective metal. The reflec- using the OSU apparatus. Specimens of black
tive metal option is the least satisfactory, because, polymethyl methacrylate (PMMA) were exposed
while radiation towards the specimen gets in the horizontal orientation to a heating flux of
eliminated by reflection, the reflection towards 35 kW · m2. With an airflow rate of 12 Ls1
the heater does cause its temperature to rise. through the combustion chamber, the ignition
Thus, the best accuracy is attained with a minimal time was 209 s. When the airflow rate was
duration of the shutter’s closure. This change was doubled to 24 Ls1, the specimen ignition time
made in the 1997 edition of ASTM E 1354 and in increased to 403 s. By contrast, Table 28.1 shows
the second edition (2002) of ISO 5660-1. comparative results with the cone calorimeter;
it can be seen a flow rate of 24 Ls1 was found
to be satisfactory. That flow rate was also about a
Airflow factor of 2 greater than the minimum at which no
spill out of the hood occurs.
The feasible airflow rate through the system is The exhaust system uses a high-temperature
bound by certain limits. It must not be so fast that cast-iron blower to exhaust the gases and an
www.ebook777.com
Table 28.1 Effect of exhaust hood airflow on ignition times in the cone calorimetera
Material Thickness (mm) Orientation Fan setting Ignition timeb (s)
PMMA 13 Horizontal No fan 71
PMMA 13 Horizontal 24 Ls 1 76
PMMA 13 Horizontal 41 Ls 1 67
PMMA 13 Vertical No fan 86
PMMA 13 Vertical 24 Ls 1 84
PMMA 13 Vertical 41 Ls 1 77
Redwood 13 Horizontal No fan 23
Redwood 13 Horizontal 24 Ls 1 24
Redwood 13 Horizontal 41 Ls 1 31
Redwood 13 Vertical No fan 22
Redwood 13 Vertical 24 Ls 1 27
Redwood 13 Vertical 41 Ls 1 29
a
At an irradiance of 35 kW · m2
b
Typical ignition time scatter was on the order of 10 % (1σ, N ¼ 3)
(Not to scale) Section A-A Section B-B Section C-C

250
mm Thermocouple Gas sampling
*25 mm ring probe
A A 57 mm Hood
Orifice *25 mm I.D. hole 57 mm I.D. hole
plate
1.59 mm 75*
thick mm Soot sampler
76
350 mm probe
Smoke meter
mm Thermocouple purge tubes 6.35 mm O.D.
114 mm 7.9 mm O.D.
Use an alignment
rod when welding
tubes to duct to
ensure perfect
Thermocouple location (rear) alignment
225 mm 685* mm
111* Orifice plate
mm B C
114 mm
B C
Smoke meter
Blower Gas sampling Tube is 0.6 mm Hood
location
ring probe thick stainless steel.
(sample holes 114 mm I.D.
face blower)
Fig. 28.9 Exhaust duct
orifice plate flowmeter (Fig. 28.9). The orifice Means of Ignition

plate flowmeter is instrumented with a differen-
tial pressure transducer and a thermocouple. For In some cases no external ignition source is
specialized studies, where the entire combustion desired, and specimen testing is to be done
system is glass enclosed [38], it is possible to go solely on the basis of autoignition. In most
to flow rates below 12 Ls1. With such enclosed cases, however, an external ignition source is
systems, accurate measurements can be made desirable. This ignition source should, in gen-
down to about 9 Ls1 using the standard orifice eral, not impose any additional localized
plate. For lower flow rates, down to about heating flux on the specimen. Apparatus designs
5 Ls1, the standard orifice plate is replaced by have been developed, with impinging pilots that
one with a smaller opening. can, in some cases, produce such high localized
964 V. Babrauskas
heat fluxes as to burn a hole through the speci- tractable alternative (i.e., electric spark igni-
men at the point of impingement, yet not ignite tion). This spark plug arrangement for ignition
it outside of that region [39]. Applications for was successful, and so a similar electric pilot
such devices tend to be specialized, since the was designed for the cone calorimeter. The loca-
general objective of radiant ignition testing is to tion of the ignitor should be at the place where
produce data that can be analyzed in the context the lower flammable limit is expected to first be
of an assumed one-dimensional heat flow. A reached when the specimen begins its pyrolysis.
design using an impinging pilot has an addi- It should, however, not be so close to the speci-
tional difficulty. Since most of the specimen men surface that minor swelling of the specimen
face is not yet heated to the ignition temperature would interfere with the ignition function. In the
when ignition first begins in the vicinity of the cone calorimeter, the ignitor locations were cho-
pilot, no unique ignition time can be deter- sen so that, when testing in the horizontal orien-
mined. Instead, there is a significant time spread tation, the spark plug gap is located 13 mm
between when ignition first occurs at the initial above the center of the specimen; in the vertical
location, to when the final portions of the face orientation, the spark plug gap is located at the
have been ignited. specimen plane and 5 mm above the top of the
The ignitor should reliably ignite a combusti- specimen holder.
ble gas mixture in its vicinity. Thus, the location The actual spark plug arrangement is shown in
of the ignitor must be chosen so that it is near the Fig. 28.10. The spark plug is provided by a
place where maximum evolution of pyrolysate special-purpose 10-kV ignition transformer. The
gases is expected. Some materials are highly spark plug is moved in and out by remote control,
fire-retardant treated, and, when heated, emit operated by an air motor that rotates the shaft on
vapors that tend to extinguish a pilot flame. The which the spark plug rests. A reversible lock bar
ignitor has to be designed so as not to be is used to adjust the spark-plug-to-heater distance
extinguished by fire-retardant compounds com- when changing from the horizontal to the vertical
ing from the specimen, nor by airflows within the orientation (the spark gap is 13 mm away from
test apparatus. the heater baseplate in the horizontal orientation,
The ISO 5657 apparatus was designed with a but 25 mm away in the vertical).
“dipping” gas pilot, which is periodically thrust
for a short while down close to the specimen
face, then retracted. This solution, however, Specimen Area and Thickness
introduces an uncertainty into ignition times
and provides further complexity. A gas pilot, Both specimen area and thickness may be
based on experience, also requires oxygen expected to have some effect on the ignitability
premix to achieve a flame that is both small and the heat release rate. The main practical size
and resistant to blowout [40]. With products and thickness limitations come from the fact that
high in fire retardant, even such precautions are the specimens to be tested should exhibit primar-
not likely to lead to a reliable pilot; thus, for ily one-dimensional heat transfer. Thus, the con-
instance, the ISO 5657 apparatus uses a second figuration should be such that excessive edge
pilot to reignite the main pilot. Pilot stability effects are not seen. If the specimen thickness is
also tends to be crucially dependent on the phys- such that it is thermally thick (the heat wave
ical condition of the pilot tube tip, and signifi- penetration depth being less than the physical
cant maintenance can be necessary. Finally, depth), then further increases in thickness are
if used in a heat release apparatus, a gas pilot not expected to change ignitability results. For
can add noise to the baseline of the heat thinner specimens, however, there can be
release measurement. Experimental efforts at expected to be a thickness effect, and the backing
the National Bureau of Standards (NBS) had or substrate material’s thermophysical properties
success using the NBS-II calorimeter, a more can be of importance.
www.ebook777.com
Spark plug position lock bar Spark plug arm

Spark plug carrier (shown in position
for horizontal testing, slide to other stop
for vertical testing)
Position of arm when

spark plug not in use
Air motor
Fig. 28.10 Spark plug, carrier, and air motor
Specimen Area Janssens [34] studied in some of magnitude as the repeatability of the results.
detail the general problem of area effect on igni- Babrauskas, in commenting on these data
tion. The effect is seen to be smaller when [42], discussed tests on larger size, horizontal
irradiances are high rather than low. The exact PMMA samples, where each doubling of the
magnitude of the effect is also dependent on specimen’s area increased the heat release rate,
the specimen’s thermophysical properties. For per unit area, by about 10 %. The more general
specimens of area 0.01 m2 or larger, however, his treatment of the horizontal specimen, of course,
results show an increase in ignition time of only is as a liquid pool. Chapter 26 gives details on the
about 10 % over what would be seen with a speci- size effect for burning pools. It can be seen
men of infinite area. Later, Nussbaum and Östman that the diameter has to be greater than about
[41] studied specimens in an experimental appara- 1 m before the specimen area effect becomes
tus somewhat similar to the cone calorimeter, but negligible.
accommodating 200 200 mm specimens. Their The effect of specimen size for vertical
comparison of the ignition times of these larger samples was examined at Factory Mutual
specimens against the standard 100 100 mm Research Corporation (FMRC) in a series of
ones shows that quadrupling the specimen area experiments on PMMA walls [43, 44]. The
decreases the ignition time by about 20 %. FMRC studies showed little size effect for speci-
For heat release rate, the specimen size affects men heights up to 200 mm; beyond 200 mm there
00
the measurement, since flame volume is larger was approximately a linear dependence of q_ on
over larger specimens; consequently the flame the height. This was true up to the maximum
radiation tends to approach a value of higher height tested, that is, 3.56 m. Unlike horizontal
emissivity. Nussbaum and Östman also exam- pools, the rate of heat release was not leveling off
ined heat release rates from larger specimens; at even these sizes, and estimates suggested that
the differences were generally of the same order the specimen size would have to be increased by
966 V. Babrauskas
another order of magnitude before a leveling off material or substrate can influence the measured
would be seen. value of the ignition time. In the cone
The conclusion from the above studies was that calorimeter, the substrate is a blanket of refrac-
100 100 mm was a suitable size for bench-scale tory ceramic fiber material, having a nominal
00
testing, but that the bench-scale q_ rates will density of 65 k · gm3. In use, the material
always be somewhat lower than for full-scale fires. assumes a more compacted density of roughly
100 kg · m3. Whenever possible, materials
Specimen Thickness The cone calorimeter is whose thicknesses are less than the minimum
intended for testing actual commercial products. suggested in the above formula should be
Thus the specimen thickness should be, as much mounted on that substrate material over which
as possible, the thickness of the finished product. they will actually be used. As a practical guide
There are limitations at both ends of the scale, for testing unknown commercial samples, it is
however. The instrument is restricted to testing desirable to specify that any specimens less than
specimens not thicker than 50 mm. For products 6 mm thick should always be considered as need-
that in their finished state are greater than 50 mm ing to be tested over their in-use substrate.
thick, it can readily be seen that, for almost any Fabrics are a special case. Thin fabrics are
realizable combination of thermophysical sometimes used for constructing air-supported
properties and incident radiant fluxes, a 50 mm structures; these should be tested with an air
specimen is thermally thick, and increasing space in back, simulating the usage conditions.
thickness would not change the ignition times A special holder has been constructed that allows
[45, 46]. By making calculations for various the fabrics to be pulled taut and held above a
densities and heat fluxes, it was found that for dead-air space (Fig. 28.11).
particleboard the minimum thickness required to
ensure that the specimen is thermally thick can
be represented by Sample Testing Specifications
ρ
‘ ¼ 0:6 00 ð28:3Þ Specimen Orientation and Specimen
q_ Holders
where
The specimen holders in Figs. 28.12 and 28.13
‘ ¼ Thickness ðmmÞ show the two specimen holders, respectively.
With proper precautions, the horizontal orienta-
ρ ¼ Density ðkg m3 Þ tion can be used for testing liquids and melting
00
q_ ¼ Heat flux ðkW m2 Þ materials, whereas the vertical orientation’s small
melt trough can only catch a very small amount of
This is probably a reasonable rule of thumb molten material. Also, some specimens, when
for other materials as well. The proportionality of tested in the vertical orientation, show a tendency
00
the required thickness to ρ=q_ is derived from to lose physical strength and fall out of the
classical heat conduction theory by equating the holder, which does not happen in the horizontal
time for the front surface to reach ignition tem- orientation.
perature to the time the rear surface’s tempera- In the vertical orientation, there are several
ture begins to rise, assuming that the thermal layers of rigid millboard behind the blanket, suf-
conductivity is proportional to the density. ficient in thickness to fill out the depth of the
Numerical calculations were necessary to deter- specimen holder. The specimen is wrapped in a
mine a suitable constant because of the impact of single sheet of aluminum foil, covering the sides
front surface heat losses. and bottom. The aluminum foil serves to limit the
For materials that are not thermally thick at flow of molten material and prevent it from
the time of ignition, the nature of the backing seeping into the refractory blanket.
www.ebook777.com
All dimensions in millimeters 2.5
18 evenly spaced 2.5

90 degree cut teeth 33
20
Side view
Edge frame
111
Sample fabric
Tensioning insert
Refractory fiber
blanket
111 Horizontal
Top view Cross section specimen holder
Fig. 28.11 Special holder for testing fabrics and similar thin materials
Fig. 28.12 Horizontal

orientation specimen 106* mm
holder Spot weld, 4 corners
106* 59* 25
A mm mm mm A
2.4 mm thick
25* 30°
mm
8 mm 40
59* 40 mm
4 mm mm mm
Stainless
(mill smooth) Section A-A
Load Cell had been tried in various earlier apparatuses, but

most suffered because they were not designed for
Many ancillary measurements made in the cone purely single-axis linear motion. That is, if the
calorimeter (such as yields of various gas spe- weight of the specimen was not well balanced, or
cies) require the use of a load cell. Transducers differential heating stresses occurred, it was
968 V. Babrauskas
Section B-B
25
104* 2.4 Slot
25
73*
116
135°
94
58
A 15
2.4
25 Material: 1.59 mm stainless steel
(except base plate)
All dimensions in mm
(except where noted)
*Indicates a
critical dimension
A 4.8
94* 104*
4.8 mm S.S. base plate
B B
10 5
6
4.8 dia. pins, 24 13 4.8
round off ends, 25.4 25.4
4 places, press fit Section A-A
Fig. 28.13 Vertical orientation specimen holder
likely that a mechanical moment (or torque) mechanical tare adjustment range, along with a
would be applied to the device, with the trans- sensitive weighing range (500 g). A resolution of
ducer then being prone to jamming. For the cone 0.005 g is readily achievable.
calorimeter, a commercial-design load cell was Figures 28.7 and 28.8 show, respectively, how
found that permits only up-and-down axial the horizontal and vertical orientation specimen
motion while being insensitive to torques or holders are accommodated on the load cell. The
forces from other directions. horizontal holder has a square recess on the bot-
The load cell has to accommodate two tom and simply is placed straight down. The
differently oriented specimen holders and may vertical holder is more conveniently inserted
need to hold additional fixtures. All of these directly toward the heater, correctly locating the
can have substantial—and different—weights, specimen by four mounting pins on the bottom.
yet must allow accurate mass determination for In both cases there is a positive specimen loca-
low-density specimens. The solution adopted tion, and the operator does not have to be
was a weighing system that has a large (3.5 kg) concerned with how far to insert the holder.
www.ebook777.com
Edge Conditions
Edge Effects In an apparatus such as the cone

calorimeter, it is desired that the small-scale test
specimen would behave, as much as is possible, 94*
mm
like a correspondingly sized element of the full-
111*
scale object. If one is dealing with relatively mm
94*
large, flat, full-scale objects, then heat and mass mm
transfer will occur only in the direction perpen-
dicular to the exposed face. There will be no heat
or mass flow along either of the face directions.
The guidance to be derived from this conceptual
model in designing the bench-scale test environ-
ment is clear: there should be a minimum of heat
or mass transfer at the specimen edges. The alu-
minum foil used to wrap the specimen usually
serves to minimize any mass transfer that may
occur. The heat transfer situation, however, is 54*
mm 10-32 tapped hole,
more complicated. 4 places
In the vertical specimen orientation, the spec- 4 mm
imen has to be restrained against falling out;
therefore, the vertical specimen holder 55.5
incorporates a small lip extending 3 mm along mm
the edges. In the horizontal orientation, no spe- 111* mm
Inside dimension
cial measures need to be taken against falling out. (stainless steel, 1.9 mm thick)
Thus, for many specimens it is satisfactory to
simply cover the edges and bottom with alumi-
num foil, leaving the top exposed in its entirety. Fig. 28.14 Edge frame for the horizontal specimen
Some categories, however, present special holder
problems—specimens that either have a propen-
sity to ignite first along the outside edge or that, any losses to an edge frame may be difficult to
when ignited, burn disproportionately vigorously approach in practice. This is still a topic of active
near the edges. Such behavior is often found with study at several institutions.
wood specimens and with certain composites. This In some cases, an edge frame is needed for
problem is alleviated by using a stainless steel thermostructural reasons. Some specimens, espe-
edge frame for the horizontal orientation, which cially certain composites, can show pronounced
like the vertical holder provides a 3 mm lip around edge warping and curling when subjected to heat.
the edge of the specimen face (Fig. 28.14). The burning of such a specimen would be highly
Specimens showing unrepresentative edge nonuniform if its edges were not held down with
burning can be viewed as having a spurious an edge frame. In many cases, an edge frame is
heat gain along the edges when compared against all that is required. In some cases, however,
a hypothetical ideal situation of exactly zero heat additional measures such as a wire grid (see
loss or gain at the edges. When an edge frame is below) are required.
applied, the opposite situation can tend to result,
that is, an observed net heat loss from the speci- Intumescing Samples Intumescence is a com-
men [47]. The ideal situation of a specimen mon difficulty with fire test specimens, either
prevented from showing unrepresentative before ignition or during the burning. The sim-
increased edge burning but equally not sustaining plest solution used in the cone calorimeter,
970 V. Babrauskas
Fig. 28.15 Wire grid 10 100 mm

mm 20 mm
10 mm
20
mm
100 mm
Sample retaining
grid (optional)
for use with
samples that are
expected to
intumesce.
Material: 2 mm
stainless steel rod
Material: Stainless steel, weld all intersections
1.9 mm thick
sufficient in many, but not all, cases, is a wire caused by the large number of both practical and
grid placed on top of the specimen. Figure 28.15 theoretical difficulties were successfully resolved
shows a medium-weight grid. To minimize by developing a flow-through smoke measuring
effects on measurements, the grid weight should system, using a helium-neon laser as the light
be the smallest possible consistent with source and a sophisticated quasi-dual-beam mea-
providing adequate mechanical restraint to the suring arrangement. Figure 28.16 shows the
tested specimen. Effects on measurements will overall arrangement of the laser photometer. It
be negligible if the average grid mass is less is mounted on the exhaust duct at the location
than 0.6 kg · m2 of specimen face area. This shown in Fig. 28.9. A thermocouple is also
mass corresponds to quite a thin, small grid mounted nearby, since the calculations require a
and will practically be usable only in occasional determination of the actual volume flow rate in
cases. Additional guidance is given in the NBS the duct at the photometer location. The user
“User’s Guide for the Cone Calorimeter” [3], should consult Geake [49] for details explaining
but testing laboratories will, on occasion, be the operation of the laser photometer. Briefly,
required to devise their own special schemes for the light from the laser goes, via two beam
mounting and restraint. splitters, into two detectors. The light reaching
the compensation detector is not attenuated by
smoke; its signal serves as the reference to
Smoke Measurement cancel out fluctuations in laser output power.
The main beam detector measures a signal that
One of the most essential ancillary measurements is attenuated by the smoke. The optical path is
performed with the cone calorimeter is smoke purged by a minute flow of room air through a
obscuration. Widespread dissatisfaction with purge system. The flow is maintained by the
older, closed-box types of smoke tests [48, 49] pressure differential in the exhaust duct.
www.ebook777.com
Beam Purge air orifices Beam splitter Filterslot

splitter
Optical path 0.11 m

Cap
Opal glass
0.5 mW
Helium-neon
laser
Opal glass
Filter
Ceramic Compensation detector
slot
fiber packing
Main
detector
Fig. 28.16 Laser photometer
For certain research purposes, it is advanta- accomplished by controlling the flow of high-
geous, in addition to obtaining optical smoke purity methane going to the burner and compar-
obscuration measurement, also to record the ing it to a known value and using the net heat of
gravimetric soot yield by measuring grams of combustion for pure CH4 as 50 MJ · kg1.
soot evolved per gram of specimen burned. A The laser photometer is calibrated by neutral-
soot mass sampler is connected to the port density glass filters. These are inserted into a
indicated in section C-C of Fig. 28.9, and a filter slot in front of the main beam detector. An
known mass fraction of the exhaust duct flow is auxiliary filter slot is provided in front of the
passed through a measuring filter and is weighed laser. This serves to check the correct balancing
before and after the test. of the dual-beam system’s common mode
rejection ratio.
The NBS “User’s Guide to the Cone
Calibration Equipment Calorimeter” [3] details how calibrations are
performed.
Two basic calibrations are needed: (1) the cali-
bration of the temperature controller for the
conical heater and (2) the actual heat release Miscellaneous Details
rate calibration. The temperature controller is
calibrated using a Schmidt-Boelter-type heat Ring Sampler
flux meter equipped with a locating collar and
inserted in place of the specimen, with its face The combustion products flowing through the
where the specimen face would be located. No exhaust system can be heavily laden in soot,
specimen holder is used for this operation. which would cause rapid clogging of the oxygen
Figures 28.7 and 28.8 show the insertion of the measurement system if precautions were not
heat flux meter. taken. The most important precaution is the spe-
The heat release rate is calibrated with a cali- cially designed ring sampler (Fig. 28.18), which
bration burner inserted into the same bracket is installed in the exhaust duct with the intake
used for the heat flux meter (Fig. 28.17). The holes facing away from the direction of airflow.
calibration burner, however, instead of being A number of small holes are used so as to provide
inserted facing the heater, is inserted so that the a certain degree of smoothing with respect to
discharge opening faces upward. Calibration is duct flow turbulence.
972 V. Babrauskas
16 2.1 hole, countersunk for

#0 flathead screw, 2 places 22.5
1/4" tapered
4.65 23.88 16
pipe tap
28.83 4.65
23.88
31.19 16 22.5
4.88
2.79
5.84 All dimensions in mm (except where noted) 3.3 hole, 8 places
1.65 1.65 9.5 6.5
3.05
31.19
Screen retaining ring
(brass) End cap
3.2 square stock, silver (brass)
soldered around 3 sides
9.5
A B
A
4.7 31.19
22
4.7
1.91 28.96 3.2 walls 6.5 16

5
23.88
0-80 tapped hole, 2 places

2.79 31.19 16
150 54
220 4-40 tapped hole,
Pack burner with 8 places
Burner main body ceramic fiber batting Section B-B
(brass)
Section A-A
Fig. 28.17 Calibration burner
Additional Gas Analyzers be exerted in configuring the samples or in

testing.
Many users of cone calorimeters provide not
just an oxygen analyzer but also additional gas
analyzers to help determine combustion Liquids
chemistry and toxicity. CO and CO2 analyzers
are simply fitted into the same sampling line The HRR of liquids is generally not the quantity of
serving the oxygen analyzer. Other analyzers, interest to regulators and other individuals charged
for example, H2O, HCI, and total unburned with enforcing fire safety provisions for liquids.
hydrocarbons, require a completely separate, In addition, there is no easy way to scale from
heated sampling line system. Such a system bench-scale results to large-scale applications.
also needs to have a heated soot filter at the However, some research studies on liquids using
front. the cone calorimeter have been reported. In such
studies, use of a circular dish is generally more
convenient than using a square specimen. For
Special Issues with Product Testing example, Hayakawa et al. [50] used a 113 mm
diameter dish, while Iwata et al. [51] used a
The cone calorimeter has been used for studying 90 mm dish. Liu et al. [52] conducted a study of
a very wide range of products and materials. liquids in the cone calorimeter, accompanied by
In this section, some items of interest are water mist extinguishment. A number of other
considered where special care needs to studies [53–57] have been reported.
www.ebook777.com
All material is stainless steel

All dimensions in millimeters
(except where noted)
A
6.35
5.0 mm hole, 4 places,

evenly spaced
2.2 mm hole, 12 places,

evenly spaced
(12 spaces at 30° each)
114.3
76 30° 66.7 R
152.4
6.35 O.D. stainless

steeltube
Weld in place, face must be

38 flush and smooth
Section A-A
Fig. 28.18 Ring sampler
Electrical Cables should not be excessive, say less than 15 mm

or so.
In testing electric cables, pyrolysis gases have a ASTM has issued a standard [58] on the test-
pronounced tendency to flow along the length of ing of electrical cables with the cone calorimeter.
the cable interior and burn only at the edges In the ASTM standard, the cables may either be
rather than uniformly over the surface. For cut into sections, or else the insulation material
such specimens, it has often been found useful alone is to be tested as a flat plaque. The latter
to coat the cable ends with a sodium silicate will generally not be practical, since cable
cement, such as Insa-Lute Adhesive Cement manufacturers do not produce the plastic in this
Paste No. 1, produced by the Sauereisen Cements form. The ASTM standard also permits the ends
Co. When the ends are sealed in such a manner, a to be sealed, or unsealed, when actual cable
knife puncture must be made in the face of each sections are tested.
piece of cable to avoid pressure buildup and Grayson et al. [59] documented the results of
rupture. the FIPEC research program, where electrical
Even though electrical cables are circular cable testing and modeling was done using a
rather than flat, it has been found that they can wide array of techniques. A very extensive cone
be successfully tested in the cone calorimeter. calorimeter testing effort is described in this
Normally, 100 mm long cable sections are cut connection. The FIPEC researchers concluded
and placed side by side, filling up the specimen that the best results are obtained when the
holder. For this to be practical, the diameter ends are sealed, except for very small cables
974 V. Babrauskas
(<4 mm dia.). The sealing did not affect the Low HRR Materials
magnitude of the results, but did improve the and Noncombustibility
repeatability. They also conducted plaque tests
on some individual materials. Many building codes, including all in North
America, have long taken the stance that
materials are to be divided into two kinds: com-
Intumescing Materials bustible and noncombustible. This kind of cate-
gorical yes/no distinction basically reflects the
Intumescing materials can present special fire knowledge of the nineteenth century, and
problems for testing. Most substances do not not the twenty-first, where fire safety engineering
intumesce significantly during testing. The is acknowledged to be a profession capable of
exceptions are, of course, intumescent coatings, quantitative assessment of hazards. While codes
but also some grades of PVC show a strong have numerous and complex provisions on this
tendency to intumesce, as do some a few other subject, the most important application, by far, is
materials. The first problem encountered is if in regards to materials that are used as either
the specimen rises sufficiently to contact the structural members or as lining materials,
spark plug and cause a short. More extreme e.g., wall or ceiling linings. Babrauskas and
intumescing can actually result in contacting the Janssens recently examined this question in
heater coil, although this has rarely been seen. detail [61]. It was concluded that the concept of
Apart from problems with the spark plug noncombustibility has no reasonable relationship
(which can be avoided by improvising another to life safety, as pertains to sub-surface construc-
ignition arrangement), there is the issue of flux tion materials, that is, everything except the
non-uniformity—as the specimen surface rises, it surface materials. The HRR of surface materials,
encounters a locale with higher radiant flux. on the other, is crucial to life safety. But
The original solution for intumescing noncombustibility is not useful metric for hazard.
materials, incorporated into the ASTM [2] and It was recommended that a solution which is
ISO [1] standards, has been to provide a wire consistent with fire safety engineering concepts
retaining grid, which is placed on top of the and is presented in a practical way is the Cleary/
specimen and held down the edge frame. This Quintiere calculation of the ‘b’ parameter, in
generally sufficiently reduces the intumescence Equation 28.1. A comparison to large-scale
to eliminate technical problems. However, this room fire test results showed a very good ability
has caused concern for some manufacturers, of this parameter to distinguish materials which
who considered that their product may not led to hazardous fire conditions, versus ones
able to achieve the optimal performance that it which did not.
might in its end-use application, where a thicker In addition, low HRR materials may show up
layer of intumesced material would be in the form of composites, where the top layer is
developed. highly resistant to burning, while the layer
A procedure was suggested [60] whereby the (s) underneath are less so. Such constructions
specimen surface is placed at 60 mm below the present issues which are discussed in the next
base of the heater, instead of 25 mm, in order to section.
allow ample height for expansion. This however
also has drawbacks. The heat flux uniformity at
such a spacing becomes poor across the face of Composites
the specimen. In addition, if a tall hemispherical
or elongated configuration is allowed to arise, The cone calorimeter is inherently designed to be
there is very limited heat flux incident upon the able to test composite products, provided the
sloped sides of such a specimen. layers are can adequately be represented by an
www.ebook777.com
assembly no more than 50 mm deep. In practice, Identification Various data items must be included
this allows for a reasonable representation of here
products which are even much thicker, provided Preparation Any nonstandard specimen
that there are not additional layers which are both preparation details must be reported
deep and producing a high HRR. In some cases, Test number Serial number of test; also
information on testing laboratory,
however, special precautions need to be taken. operator, and so forth
For example, upholstered furniture Irradiance The heating flux set for the test
composites cannot be tested in a representative (kW · m2)
manner unless the outer layers are constrained to Exhaust flow rate Recorded for completeness, usually
stay in place, instead of crumpling due to expo- the standard value of 24 Ls1
sure to heat. In view of this need, a special test Orientation Horizontal or vertical
method, ASTM E 1474 [62], was developed for Spark ignition Yes or no
preparing and testing such composites. Edge frame Yes or no
Wire grid Yes or no
Some high-pressure laminates explosively
Area of specimen (m2), since may be nonstandard in
delaminate when subjected to heating. These special cases
can only be tested if the use of the wire grid Specimen initial (g)
and edge frame eliminates this problem, which mass
it normally does. Another special category of Specimen final (g)
products are ones where the surface layer has a mass
low HRR, while layers underneath show a much Time to ignition According to the ISO and ASTM
standards, “sustained flaming” (s)
higher HRR. If tested simply as a cut specimen,
Time to flameout (s)
the edges are likely to get involved early, and this
Peak q_
00
(kW · m2)
may produce unrepresentative burning. The
Peak m_ 00 (g · s1 · m2)
problem was studied by Canadian researchers
[63, 64] who developed a special holder which Total q00 (MJ · m2)
O2 consumption (kJ · kg1); this is set to a specific
protects the edges of such specimens. The latter
constant value if known, otherwise to
was incorporated into the Canadian CAN/ULC 13,100
S135 standard [65]. Effective heat of (MJ · kg1), reported for period of
combustion entire test run
Specific extinction (m2 · kg1), reported for period of
Measurements Taken with the Cone rate entire test run
Calorimeter Average mass loss Computed over period starting when
rate 10 % of the ultimate specimen mass
loss rate has occurred and ending at
The relevant ISO [1] or ASTM [2] standards the time when 90 % of the ultimate
mandate certain minimum variables to be specimen mass loss has occurred
recorded. In practice, it is normally desired to (g · s1 · m2)
00
make the data from the test be as complete as Average q_ (60 s) Computed for the first 60 s after
ignition (kW · m2)
possible. Cone calorimeter data are normally 00
Average q_ (180 s) Computed for the first 180 s after
handled as data tables and files standardized ignition (kW · m2)
according to the Fire Data Management System 00
Average q_ (300 s) Computed for the first 300 s after
(FDMS) prescription [66]. A complete set of data ignition (kW · m2)
from the cone calorimeter are illustrated there.
Here the more important of these are given, Note in the above 60, 180, and 300 s averages
somewhat augmenting the ISO and ASTM set. that, if the test is ended before having burned,
Note that most items must be reported for each say, 300 s, a proper average can still be correctly
test run, and a complete test consists of three computed (i.e., at the end of the averaging period
runs. a number of zeroes are used for data points past
976 V. Babrauskas
the end of the test). Since users are often con- representative for the test results, were tign,
00 00 00
fused by this point, it must be emphasized: it is q_ max , q_ 180 , qtot , Δhc,eff, and σf . A linear regres-
not sensible to report an “average heat-release sion model was used to describe r and R as
rate” without specifying the time interval. The functions of the mean overall replicates and
reason has to do with the question of determining overall laboratories for each of the six variables.
the end of the test. The ISO and ASTM standards The regression equations given below also indi-
specify that the end of the test is considered to be cate the range of mean values over which the fit
1. After all flaming and other signs of combus- was obtained.
tion cease The results for time to sustained flaming, tign,
2. While there may still be vestigial combustion in the range of 5–150 s were
evidence, but the mass loss rate has become
very small (less than 150 gm2 being lost r ¼ 4:1 þ 0:125tign
during any 1 min) R ¼ 7:4 þ 0:220tign
3. 60 min have elapsed 00
These rules are needed for establishing some The results for peak heat release rate, q_ max , in
uniformity among testing laboratories. They do the range of 70–1120 kW · m2 were
not, however, mean that it is technically sound to 00
compare the average of one material that may r ¼ 13:3 þ 0:131q_ max
00
have burned for 10 min with another that may R ¼ 60:4 þ 0:141q_ max
have burned for 5 min. It is technically sound,
however, to compare their burning over the first The results for 180 s average heat release rate,
00
one, three, and so forth, minutes of test. q_ 180 , in the range of 70–870 kW · m2 were
Further information on the form, units, and 00
usage of fire properties measured in the cone r ¼ 23:3 þ 0:037q_ 180

00
calorimeter can be found in Babrauskas [67]; R ¼ 25:5 þ 0:151q_ 180
specific information on the smoke and soot 00
properties measured in the cone calorimeter is The results for total heat released, q_ tot , in the
given in Babrauskas and Mulholland [48]. range of 5–720 MJ · m2 were
00
r ¼ 7:4 þ 0:068qtot
00
Repeatability and Reproducibility R ¼ 11:8 þ 0:088qtot
The repeatability, r, and reproducibility, R, of the The results for effective heat of combustion,
cone calorimeter were studied in two sets of Δhc,eff, in the range of 7–40 kJ · g1 were
interlaboratory trials, one sponsored by ISO and r ¼ 1:23 þ 0:050Δhc, eff
one by ASTM. According to the ISO instructions
[68], the definitions of repeatability and repro- R ¼ 2:42 þ 0:055Δhc, eff
ducibility were taken as The results for average specific extinction
r ¼ 2:8σr area, σf, in the range of 30–2200 m2 · kg1 were
R ¼ 2:8σR r ¼ 59 þ 0:076σ f
where σr is the repeatability standard deviation, R ¼ 63 þ 0:215σ f
σR is the reproducibility standard deviation, and
the 2.8 factor comes from specifying the proba- A comparison of the cone calorimeter repeat-
bility level of 95 %. ability and reproducibility to the values obtained
From the results of the interlaboratory trials, for the ISO 5657 radiant ignition test showed
values for r and R were calculated for six the cone calorimeter results to be about a factor
variables. These variables, chosen as being of 2 better.
www.ebook777.com
Fig. 28.19 The NIST

controlled-atmosphere
cone calorimeter
for use only under ambient pressures. There is

Special Cone Calorimeters interest at this time from at least one research
group to design and construct a unit for aerospace
The standard cone calorimeter has been designed studies that would function under nonambient
to use room air for combustion. All standard pressures.
testing is done under such conditions. For special
combustion studies, however, it can be of interest
to explore the burning of materials at oxygen Nomenclature
levels other than 21 %. Such a unit, constructed
at NIST, is described in Babrauskas et al. [38]. b Parameter
This controlled-atmosphere unit has already been F View factor (-)
used for studies of the burning of materials in air hc Convective heat transfer coefficient
supplies with less than 21 % oxygen, with N2 or (W · m–2K1)
CO2 being mixed into the air stream (Fig. 28.19). Thickness (mm)
‘
It has also been used for pyrolysis studies under 00
m_ Mass loss rate (g · s1 · m2)
pure nitrogen flow conditions. In principle, it
q_ Total energy released per unit area
could also be used for studies of enriched-oxygen
(MJ · m2)
atmospheres; however, the necessary safety
q" Total energy released per unit area
procedures for handling high-concentration oxy-
(MJ m2)
gen streams are required. A unit for handling O2 00
q_ Heat flux (kWm2)
mixtures greater than 21 % has been constructed 00
q_ 180 180 s average heat release rate
for NASA, but data are not yet available from
(kW · m2)
it. A controlled-atmosphere unit is also appropri- 00
q_ avg Average heat release rate (kW m2)
ate for use when airflow rates of less than 00
12 Ls1 are required. q_ Irradiance (kW m2)
00
All of the present cone calorimeter designs, q_ max Maximum heat release rate (kW m2)
00
both standard and otherwise, have been designed q_ tot Total heat released (MJ · m2)
978 V. Babrauskas
00
q_ cr Critical heat flux (kW m2) 9. Capote, J. A., Alvear, D., Lazano, M., and Espina, P.,
r Repeatability (units dependent on Heat Release Rate and Computer Fire Modelling vs.
Real-Scale Fire Tests in Passenger Trains, Fire &
quantity investigated) Materials 32, 213–229 (2008).
R Reproducibility (units dependent on 10. Aksit, I. M., Moss, J. B., and Rubini, P. A., CFD
quantity investigated) Simulation of Cable Tray Fires, pp. 1129–1140 in
tb Duration of flaming (s) Interflam 2001—Proc. 9th Intl. Conf., Interscience
Communications Ltd., London (2001).
tig Ignition time (s) 11. Andersson, J., and Persson, F., Datostöd simulering av
Δhc,eff Effective heat of combustion pyrolysförlopp [Computer supported simulation of
(MJ · kg1) pyrolysis], Examensarbete, Dept. of Chemical Engi-
ε Emissivity (-) neering, Chalmers Lindholmen University College,
Göteborg, Sweden (2001).
λρC Thermal inertia (kJ2 m4 s1 K2) 12. Tsai, T.-H., Li, M.-J., Shih, I-Y., Jih, R., and Wong,
ρ Density (kgm3) S.-C., Experimental and Numerical Study of
σr Repeatability standard deviation (units Autoignition and Pilot Ignition of PMMA Plates in a
dependent on quantity investigated) Cone Calorimeter, Combustion and Flame 124,
466–480 (2001).
σR Reproducibility standard deviation 13. Wade, C., A Room Fire Model Incorporating Fire
(units dependent on quantity Growth on Combustible Linings (M.S. thesis), Dept.
investigated) of Fire Protection Engineering, Worcester Polytech-
σf Average specific extinction area nic Institute, Worcester MA (1996).
14. Wade, C. A., LeBlanc, D., Ierardi, J., and Barnett,
(m2 · kg1) J. R., A Room-Corner Fire Growth and Zone
Model for Lining Materials, pp. 106–117 in Proc.
2nd Intl. Conf. on Fire Research and Engineering,
References Society of Fire Protection Engineers, Bethesda MD
(1998).
1. ISO 5660, International Standard, “Reaction-To-Fire 15. Wade, C. A., A Theoretical Model for Fire Spread in a
Tests—Heat Release, Smoke Production and Mass Room Corridor Configuration, pp. 295–306 in Proc.
Loss Rate,” International Organization for 3rd Intl. Conf. on Performance-Based Codes and Fire
Standardization, Geneva (2002). Safety Design Methods, Society of Fire Protection
2. ASTM E1354, Standard Test Method for Heat and Engineers, Bethesda MD (2000).
Visible Smoke Release Rates for Materials and 16. Lattimer, B. Y., Hunt, S. P., Wright, M., and Sorathia,
Products Using an Oxygen Consumption Calorimeter, U., Modeling Fire Growth in a Combustible Corner,
American Society for Testing and Materials, West Fire Safety J. 38, 771–796 (2003).
Conshohocken, Pennsylvania (2011). 17. Janssens, M. L., and Dillon, S. E., Balanced Approach
3. W.H. Twilley and V. Babrauskas, “User’s Guide for to the Fire Performance Evaluation of Interior
the Cone Calorimeter,” NBS Special Publication SP Finish Materials, pp. 43–50 in Fifteenth Meeting of
745, U.S. National Bureau of Standards, the UJNR Panel on Fire Research and Safety, March
Gaithersburg, MD (1988). 1–7, 2000 (NISTIR 6588), S. L. Bryner, ed., Nat.
4. V. Babrauskas and S.J. Grayson, eds., Heat Release in Inst. Stand. & Technol., Gaithersburg MD (2000).
Fires, Elsevier, London (1992). Distributed in the 18. Karlsson, B., Models for Calculating Flame Spread on
U.S. by NFPA. Wall Lining Materials and the Resulting Heat Release
5. Babrauskas, V., Bench-Scale Methods for Prediction Rate in a Room, Fire Safety J. 23, 365–386 (1994).
of Full-Scale Fire Behavior of Furnishings and Wall Published in 1995.
Linings (SFPE Technical Report 84-10), Society of 19. Grant, G., and Drysdale, D., Numerical Modelling of
Fire Protection Engineers, Boston (1984). Early Flame Spread in Warehouse Fires, Fire Safety J.
6. Petrella, R. V., Assessment of Full-Scale Fire Hazards 24, 247–278 (1995).
and Cone Calorimeter Data, J. Fire Sciences 12, 20. Kokkala, M., Baroudi, D., and Parker, W. J., Upward
14–43 (1994). Flame Spread on Wooden Surface Products:
7. Cleary, T. G., and Quintiere, J. G., Framework for Experiments and Numerical Modelling, pp. 309–320
Utilizing Fire Property Tests, pp. 647–656 in Proc. 3rd in Fire Safety Science—Proc. 5th Intl. Symp.¸ Intl.
Intl. Symp. on Fire Safety Science, Intl. Assn. for Fire Assn. for Fire Safety Science (1997).
Safety Science. Elsevier Applied Science, New York 21. Beyler, C. L., Hunt, S. P., Iqbal, N., and Williams,
(1991). F. W., A Computer Model of Upward Flame Spread
8. Babrauskas, V., Ignition Handbook, Fire Science on Vertical Surfaces, pp. 297–308 in Fire Safety Sci-
Publishers/Society of Fire Protection Engineers, ence—Proc. 5th Intl. Symp.¸ Intl. Assn. for Fire Safety
Issaquah WA (2003). Science (1997).
www.ebook777.com
22. North, G. A., An Analytical Model for Vertical Flame 38. V. Babrauskas, W.H. Twilley, M. Janssens, and
Spread on Solids: An Initial Investigation (Fire Engi- S. Yusa, “A Cone Calorimeter for Controlled-
neering Research Report 99/12), School of Engineer- Atmospheres Studies,” Fire and Materials, 16,
ing, University of Canterbury, New Zealand (1999). pp. 37–43 (1992).
23. Wright, M. T., Flame Spread on Composite Materials 39. ASTM E906, Standard Test Method for Heat and
for Use in High Speed Craft (M.S. thesis), Worcester Visible Smoke Release Rates for Materials and
Polytechnic Institute, Worcester MA (1999). Products Using a Thermopile Method, American
24. Hostikka, S., and Axelsson, J., Modeling of the Radi- Society for Testing and Materials, West
ative Feedback from the Flames in the Cone Calorim- Conshohocken, Pennsylvania (2010).
eter (TR 540), Nordtest, Espoo, Finland (2003). 40. V. Babrauskas, “Combustion of Mattresses Exposed
25. Hakkarainen, T., and Hayashi, Y., Comparison of to Flaming Ignition Sources, Part II. Bench-Scale
Japanese and European Fire Classification Systems Tests and Recommended Standard Test,” NBSIR
for Surface Linings, Fire Science & Technology 80-2186, U.S. National Bureau of Standards (1981).
(Japan) 21:1, 19–42 (2001). 41. R.M. Nussbaum and B.A.-L. Östman, “Larger
26. International Code for Application of Fire Test Specimens for Determining Rate of Heat Release in
Procedures, 2010 (2010 FTP Code), International the Cone Calorimeter,” Fire and Materials,
Maritime Organization, London (2011). 10, pp. 151–160 (1986).
27. B.J. McCaffrey and G. Cox, “Entrainment and Heat 42. V. Babrauskas, Letter to the editor, Fire and
Flux of Buoyant Diffusion Flames,” NBSIR 82-2473, Materials, 11, p. 205 (1987).
U.S. National Bureau of Standards (1982). 43. L. Orloff, J. deRis, and G.H. Markstein, “Upward
28. J.R. Hallman, “Ignition Characteristics of Plastics and Turbulent Fire Spread and Burning of Fuel Surfaces,”
Rubber,” Ph.D. Dissertation, University of Oklahoma, in Fifteenth Symposium (International) on Combus-
Norman (1971). tion, The Combustion Institute, Pittsburgh, PA,
29. B. Hägglund and L.-E. Persson, “The Heat Radiation pp. 183–192 (1974).
from Petroleum Fires,” FOA Rapport C 20126-D6 44. L. Orloff, A.T. Modak, and R.L. Alpert, “Burning of
(A3), Försvarets Forskningsanstalt, Stockholm, Large-Scale Vertical Wall Surfaces,” in Sixteenth
Sweden (1976). Symposium (International) on Combustion, The Com-
30. V. Babrauskas, “Estimating Large Pool Fire Burning bustion Institute, Pittsburgh, PA, pp. 1345–1354
Rates,” Fire Technology, 19, pp. 251–261 (Nov. (1976).
1983). 45. W.D. Weatherford, Jr. and D.M. Sheppard, “Basic
31. J.J. Comeford, “The Spectral Distribution of Radiant Studies of the Mechanism of Ignition of Cellulosic
Energy of a Gas-Fired Radiant Panel and Some Diffu- Materials,” Tenth Symposium (International) on Com-
sion Flames,” Combustion and Flame, 18, pp. bustion, The Combustion Institute, Pittsburgh, PA,
125–132 (1972). pp. 897–910 (1965).
32. A. Tewarson, “Physico-Chemical and Combustion/ 46. H.R. Wesson, J.R. Welker, and C.M. Sliepcevich,
Pyrolysis Properties of Polymeric Materials,” NBS- “The Piloted Ignition of Wood by Thermal Radia-
GCR-80-295, U.S. National Bureau of Standards, tion,” Combustion and Flame, 16, pp. 303–310
Gaithersburg, MD (1980). (1971).
33. ISO 5657, “Reaction to Fire Tests—Ignitability of 47. J. Urbas and H. Sand, “Some Investigations on
Building Products Using a Radiant Heat Source,” Ignition and Heat Release of Building Materials
International Organization for Standardization, Using the Cone Calorimeter,” in Interflam’90, Fifth
Geneva (1997). International Fire Conference Proceedings,
34. M.L. Janssens, “Fundamental Thermophysical Interscience Communications, Ltd., London,
Characteristics of Wood and Their Role in Enclosure pp. 183–192 (1990).
Fire Growth,” Ph.D. Dissertation, University of Gent, 48. V. Babrauskas and G. Mulholland, “Smoke and Soot
Belgium (1991). Data Determinations in the Cone Calorimeter,” in
35. Boulet, P., et al., Characterization of the Radiative ASTM STP 983, Mathematical Modeling of Fires,
Exchanges When Using a Cone Calorimeter for the American Society for Testing and Materials,
Study of Plywood Pyrolysis, Fire Safety J. 51 53–60 Philadelphia, pp. 83–104 (1987).
(2012). 49. P.J. Geake, “Smoke Characterisation by Laser Dif-
36. V. Babrauskas, “Development of the Cone Calorime- fraction,” Ph.D. Dissertation, Polytechnic of the
ter—A Bench-Scale Heat Release Rate Apparatus South Bank, London (1988).
Based on Oxygen Consumption,” Fire and Materials, 50. Hayakawa, T., Sakurai, Y., and Yoshida, K., Devel-
8, pp. 81–95 (1984). opment of New Test Method of Cone Calorimeter for
37. Babrauskas, V., and Wetterlund, I., The CBUF Cone Liquid Substances, National Maritime Research Insti-
Calorimeter Test Protocol: Results from International tute, Tokyo (2004).
Round Robin Testing (SP Report 1996:12), Swedish 51. Iwata, Y., Koseki, H., Janssens, M. L., and Takahashi,
National Testing and Research Institute, Borås T., Comparison of Combustion Characteristics of
(1996). Crude Oils, Fire and Materials 25, 1–7 (2001).
980 V. Babrauskas
52. Liu, J., Liao, G., Fan, W., and Yao, B., Study of 61. Babrauskas, V., and Janssens, M., Quantitative
Liquid Pool Fire Suppression with Water Mists by Variables to Replace the Concept of
Cone Calorimeter, J. Fire Sciences 20, 465–477 ‘Noncombustibility,’ pp. 77–90 in Proc. Fire &
(2002). Materials 2009, Interscience Communications Ltd,
53. Hshieh, F.-Y., and Julien, C. J., Experimental Study London (2009).
on the Radiative Ignition of Silicones, Fire and 62. ASTM E 1474, Standard Test Method for Determin-
Materials 22, 179–185 (1998). ing the Heat Release Rate of Upholstered Furniture
54. Armand, Y., Delfau, J. L., and Vovelle, C., Kinetics of and Mattress Components or Composites Using a
Thermal Degradation of Solid or Liquid Components Bench-Scale Oxygen Consumption Calorimeter,
under a Radiative Flux, pp. 205–221 in Industrial ASTM Intl., West Conshohocken PA (2010).
Fires II—Workshop Proceedings (Report EUR 63. Richardson, L. R., Determining Degrees of Combus-
15967 EN), European Commission, Luxembourg tibility of Building Materials—National Building
(1995). Code of Canada, Fire and Materials 18, 99–106
55. Breulet, H., and Desmet, S., Characterization of (1994).
Industrial Liquids by Means of the Cone Calorimeter, 64. Richardson, L. R., and Brooks, M. E., Combustibility
pp. 223–235 in Industrial Fires II—Workshop of Building Materials, Fire and Materials 15,
Proceedings (Report EUR 15967 EN), European 131–136 (1991).
Commission, Luxembourg (1995). 65. Standard Method of Test for Determination of
56. Grand, A. F., and Trevino, J. O., Flammability Degrees of Combustibility of Building Materials
Screening and Fire Hazard of Industrial Fluids Using using an Oxygen Consumption Calorimeter (Cone
the Cone Calorimeter, pp. 157–173 in Fire Resistance Calorimeter), National Standard of Canada, CAN/
of Industrial Fluids (ASTM STP 1284), ASTM, ULC-S135, Underwriters’ Laboratories of Canada,
Philadelphia (1995). Scarborough, Ont., Canada.
57. Elam, S. K., Altenkirch, R. A., Saito, K., and Arai, M., 66. V. Babrauskas, R.D. Peacock, M. Janssens, and
Cone Heater Ignition Tests of Liquid Fuels, Fire N.E. Batho, “Standardizing the Exchange of Fire
Safety J. 16, 65–84 (1990). Data—The FDMS,” Fire and Materials,
58. Standard Test Method for Using a Cone Calorimeter 15, pp. 85–92 (1991).
to Determine Fire-Test-Response Characteristics of 67. V. Babrauskas, “Effective Measurement Techniques
Insulating Materials Contained in Electrical or Optical for Heat, Smoke, and Toxic Fire Gases,” in Fire:
Fiber Cables (ASTM D 6113), ASTM Intl., West Control the Heat . . . Reduce the Hazard, QMC Fire
Conshohocken PA. & Materials Centre, London, pp. 4.1–4.10 (1988).
59. Grayson, S. J., et al., Fire Performance of Electric 68. ISO 5725, “Accuracy (Trueness and Precision) of
Cables—New Test Methods and Measurement Measurement Methods and Results,” International
Techniques (Contract no. SMT4-CT96-2059), final Organization for Standardization, Geneva (1986).
report to the European Commission. Interscience
Communications Ltd., London (2000).
60. Gensous, F., and Grayson, S., Improved Procedure for Vytenis Babrauskas is the president of Fire Science and
Testing Intumescent Materials Using the Cone Calo- Technology Inc., a company specializing in fire safety
rimeter, pp. 977–981 in Interflam’96, Interscience research, fire testing issues, and fire science applications
Communications Ltd, London (1996). to fire investigations and litigations.
www.ebook777.com
Compartment Fire Modeling

29
James G. Quintiere and Colleen A. Wade
Introduction uniform in temperature and composition. Distinct

phenomena were discerned that could be studied in
An approach for predicting various aspects of fire isolation, enabling better predictions of their roles
phenomena in compartments has been called in the compartment fire system.
zone modeling. Based on a conceptual represen- Fowkes [1], in his work with Emmons on the
tation for the compartment fire process, it is an Home Fire Project, was the first to publish a basis
approximation to reality. Any radical departure for the zone model approach in his description of
by the fire system from the basic concept of the the “Bedroom Fire” series conducted at Factory
zone model can seriously affect the accuracy and Mutual Research Corporation (FMRC). Almost
validity of the approach. The zone model simultaneously, computer models based on the
represents the system simply as two distinct com- zone model approach were produced by
partment gas zones: an upper volume and a lower Quintiere [2], Pape and Waterman [3], and Mitler
volume resulting from thermal stratification due [4] working with Emmons. Since then the devel-
to buoyancy. Conservation equations are applied opment of such computer models has been pro-
to each zone and serve to embrace the various lific, extending early efforts on single rooms to
transport and combustion processes that apply. computer codes that can address a number of
The fire is represented as a source of energy interconnected rooms, using a number of new
and mass manifested as a plume, which acts fire phenomena and computer features. These
as a pump for the mass from the lower zone to advances in fire science, together with computer
the upper zone through a process called development, have given the engineer conve-
entrainment. nient tools for investigating the fire hazard in
The zone modeling approach emerged in the buildings. A notable illustration of such a tool is
mid-1970s when the effort to study the developing the Hazard I software developed by the National
fire in a compartment intensified. Careful Institute of Standards and Technology (NIST)
measurements and observations revealed charac- [5], only one of numerous current computer
teristics of the compartment fire system. The upper codes and software packages based on the zone
and lower layers (zones) were deemed relatively model approach. In 1992, Friedman [6] cited
21 zone models in use around the world.
In 2003, Olenik and Carpenter [7] revisited the
survey and summarised the available computer
models and their general capabilities. For a
discussion of specific zone computer fire models,
J.G. Quintiere (*) • C.A. Wade see Chap. 31.

982 J.G. Quintiere and C.A. Wade
This chapter outlines the basic conservation CV1

equations for the gas zones and describes the
various transport and combustion processes that
make up the system, referred to as the submodels T,p
of the system. As such, they can contribute
subroutines to computer codes, which implement m
•
the mathematical solution. Discussion of

submodels will be limited, but the reader will
w =V
be referred to appropriate references. In most •
cases, other chapters of The SFPE Handbook of me
Fire Protection Engineering will be cited. No

discussion of a computer code or its numerical
w =0
solution algorithm will be addressed, since these
are issues more of style and mathematics. The
presentation will elucidate the mathematical
basis of the zone model, its assumptions, its • CV2
ms (Fuel)
features, and its scope of application. Each user
of this approach must sufficiently understand its Fig. 29.1 Control volumes selected in zone modeling
basis to assess its accuracy and validity. When
used correctly, zone models predict the average
macroscopic features of compartment fires. equations for mass, species, and energy conserva-
There are many examples of comparisons to tion in a combustion system has been presented
data that illustrate their level of accuracy, and by Quintiere [8] and serves as a reference for the
these will not be repeated here. The user must be equations that follow.
skilled in assessing the quality of the data and Figure 29.1 illustrates a typical zone model
submodels that directly influence the variables of for a compartment fire process. It shows a fire
the problem of interest. It is hoped that the dis- plume and a door vent. The hot combustion gases
cussion that follows will make the user more that collect in the upper space of the room and
knowledgeable or sensitive in making these qual- spill out of the vent constitute the upper-layer
ity assessments. zone. A control volume, CV1, is defined to
enclose the gas in this upper layer and the fire
plume. The lower interface of the upper layer
Conservation Equations moves with the control volume such that no
mass is transferred across this thermally stratified
The building block of the zone model is the con- region. The velocity of the control volume along
servation equations for the upper and lower gas this interface, w, is equal to the fluid velocity, v.
zones. These equations are developed either The temperature of the upper layer is greater than
(1) by using fundamental equations of energy, that in the lower layer (zone) which includes all
mass, and momentum transport in control volume the remaining gas in the room, and is delineated
form as applied to the zones, or (2) by using by a second control volume, CV2. It has been
differential equations that represent the conserva- assumed in zone modeling that the volume of
tion laws and integrating them over the zones. the fire plume is small relative to the gas layer
However, the momentum equation will not be or zone volumes, and therefore its effect has been
explicitly applied, since information needed to ignored. In general, multiple fire plumes can
compute velocities and pressures is based on occur at any height in the room, and multiple
assumptions and specific applications of momen- vents or mass transport can take place between
tum principles at vent boundaries of the compart- the zones (CV1 and CV2) and the surroundings. In
ment. An extensive review of control volume each case mass transport must be appropriately
www.ebook777.com
29 Compartment Fire Modeling 983
described in terms of the system variables; variation for a height, H ¼ 1 m, gives a

however, this may not always be easy or pressure difference of ρgH ¼ 1.2 kg/m3
known. The properties of the upper and lower 9.8 m/s2 1 m ¼ 10 kg/ms2 ¼ 10 Pa (N/m2).
zones are assumed to be spatially uniform, but 8. Mass flow into the fire plume is due to turbu-
can vary with time. Thus, temperature, T, and lent entrainment. Entrainment is the process
species mass concentration, Yi, are properties by which the surrounding gas flows into the
associated with ideal upper and lower homoge- fire plume as a result of buoyancy. Empiri-
neous gas layers. Other assumptions in the appli- cally, the inflow velocity linearly depends on
cation of the conservation laws to the zones are the vertical velocity in the plume.
listed below: 9. Fluid frictional effects at solid boundaries are
1. The gas is treated as an ideal gas with a ignored in the current models.
constant molecular weight and constant spe-
cific heats: cp and cv.
2. Exchange of mass at free boundaries is due to Conservation of Mass
pressure differences or shear mixing effects.
Generally these are caused by natural or forced The conservation of mass for a control volume
convection, or by entrainment processes. states that the rate of change of mass in the
volume plus the sum of the net mass flow rates
3. Combustion is treated as a source of mass and
energy. No mechanism from first principles is out for J flow streams is zero
included to resolve the extent of the d X
J
combustion zone. A ðρzl Þ þ m_ j ¼ 0 ð29:1Þ
dt
4. The plume instantly arrives at the ceiling. No j¼1
ðnetoutÞ
attempt is made to account for the time required
to transport mass vertically or horizontally in where
the compartment. Hence, transport times are ρ ¼ Density of the gas in the control volume
not explicitly accounted for in zone modeling. (or zone)
5. The mass or heat capacity of room contents is zl ¼ Height of the zone
ignored compared to the enclosure wall, ceil- For the illustration in Fig. 29.1, applying
ing, and floor elements; that is, heat is consid- Equation 29.1 to the upper layer (CV1) would give
ered lost to the structure, but not to the
contents. Where room contents shield bound- X
3
m_ j ¼ m_ m_ e m_ s ð29:2Þ
ary structural surfaces, some compensations j¼1
can occur in the analysis, but for cluttered
rooms this assumption may be poor. where
6. The horizontal cross section of the enclosure ṁ ¼ Mass flow rate out of the door
is a constant area, A. In most cases of zone ṁe ¼ Mass rate of entrainment into the fire
modeling, rectilinear compartments have plume
been considered. However, this is not a neces- ṁs ¼ Mass rate of gaseous fuel supplied
sary assumption, and enclosures in which Mass flows at the boundaries can occur due to
A varies with height can easily be handled. many phenomena. Therefore, the user or
7. The pressure in the enclosure is considered designer of a zone model must include the
uniform in the energy equation, but hydrostatic appropriate mass flow phenomena. For example,
variations account for pressure differences at in addition to the mass rates in Equation 29.2,
free boundaries of the enclosure; that is, p mass flows can occur due to forced convection
ρgH. In general, the enclosure pressure, p, is from wind or ventilation effects, from shear
much greater than the variations due to hydro- entrainment as flows affect layer interfaces, or
statics. For example, for p ¼ 1 atm ¼ 14.7 psi from cold plumes that could plunge through hot
¼ 102 kPa (kN/m2) ¼ 105 Pa, the hydrostatic layers.
Conservation of Species dT dp XJ
ρc p zl A zl A þ cp m_ j T j T
dt dt j¼1
The mass concentration of species i is given by
Yi. By using Equation 29.1 and applying the ¼ ω_ F ΔH Q_ net loss
conservation of mass for species i to a control ð29:4Þ
volume, it follows that
where
dY i X J T ¼ Temperature of the gases within the control
ρzl A þ m_ j Y i j Y i ¼ ω_ i ð29:3Þ
dt j¼1
volume
Tj ¼ Temperature of the gases in the j flow
Where stream crossing the control volume
Yij ¼ Mass concentration of species i leaving the boundary
control volume through the j flow stream Q_ net loss ¼ Net rate of heat transfer lost at the
ω_ i ¼ Mass concentration rate of species due to boundary
combustion ΔH ¼ Heat of combustion (taken as a positive
The production term, ω_ i , in principle, can be quantity)
described through a knowledge of the chemical ω_ F ¼ Rate at which the fuel supplied is reacted
equation of the reaction or its particular stoichi- Usually in zone models it is assumed that all
ometry. Thus, stoichiometric coefficients can be of the fuel supplied can react, provided there is
used to represent the production of species and sufficient oxygen available. One assumption on
the consumption of oxygen in terms of the mass the sufficiency of oxygen is to consider that all
rate of fuel reacted. Stoichiometry is not easily the fuel supplied is reacted as long as the oxygen
determined, and the fuel gases as they emerge concentration in that control volume is greater or
from the pyrolysis of solids can take many equal to zero, that is,
chemical forms that differ from the solid fuel’s
original molecular composition. A partial way ω_ F ¼ m_ s if Yo 0 ð29:5Þ
to overcome these complications has been to Thereafter, an excess rate of fuel can exist that
represent the mass production of species for can be transported into adjoining zones or control
fire in terms of the rate of mass loss for the volumes where a decision must be made about
pyrolyzing fuel. Hence, one must be careful to whether it can continue to react. At this condi-
distinguish between the mass of fuel lost and tion, all of the net oxygen supplied to the control
that reacted, and to relate available species yield volume is reacted, so that, as long as Yo ¼ 0
data to the particular fire conditions of the appli-
cation. Yield is defined as the mass ratio of ω_ F ¼ r ðnet mass rate of oxygen suppliedÞ
species to fuel lost. The yields or production ð29:6Þ
rates may change with fire conditions, and
therefore, in general, will not be consistent where r is the stoichiometric fuel-to-oxygen
with data from small-scale tests. For example, mass ratio. This condition when Yo ¼ 0 in com-
the production rate of CO changes markedly partment fires is termed the ventilation-limited
with air-to-fuel ratio. condition. At this moment, significant changes
take place in the nature of the chemical reaction.
Notably, incomplete combustion is more likely,
Conservation of Energy and for hydrocarbon fuels this leads to a signifi-
cant increase in the yield of carbon monoxide
The conservation of energy for the control vol- and soot. Thus, care must be used in interpreting
ume is applied along with Equation 29.1 and the the results of zone models once ventilation-
equation of state, p ¼ ρRT, to give limited conditions arise, particularly with respect
www.ebook777.com
to the prediction of species concentrations and phenomenon, when cycling between heating
the extent of burning. Material data used for and cooling, explains the breathing effect for
well-ventilated conditions will no longer apply. fires in closed buildings.
A flame extinction condition can be defined The third term of Equation 29.4 accounts for
by a flammability line that is based on a critical the enthalpy flow rates and only applies to j flow
flame temperature below which extinction occurs streams that enter the control volume, since Tj ¼
and no energy is contributed to the system T for all flow streams leaving, as long as the
[9]. Based upon a extinction flame temperature uniform temperature assumption still applies.
of 1300 C, and incorporating Equations 29.5 and
29.6, the criteria for energy release (or burning
rate) is expressed as: Summary
8
< m_ s ; Y o > 0 and T f > 1300 C The zone model for the compartment fire system
ω_ F ¼ r m_ o Y o ; Y o ¼ 0 and T f > 1300 C consists of two zones: the upper and lower gas
:
0 ; T f 1300 C layers. The solution process for the layer
ð29:7Þ properties can be visualized by considering the
conservation Equations 29.1, 29.3, and 29.4
The first term on the left side of Equation 29.4 applied to each zone. The species equation can
arises due to the change of internal energy within yield the Yi for each layer. The mass and energy
the control volume. If the temperature is not equations comprise four equations (two for each
changing rapidly with time, this term can be layer) that permit the determination of the two
small and its elimination gives rise to a quasi- layer temperatures, one layer height (since the
steady approximation for growing fires that height of the other layer is directly found by
allows a more simple analysis. The second term difference from the total height of the compart-
arises from the rate of work done by pressure as ment), and the compartment pressure (which is
the gas layer expands or contracts due to the assumed uniform by Equation 29.4). The
motion of the thermal stratification interface. densities are found from the ideal gas equation
Having been rearranged, this term now is of state in which ρT is approximately a constant.
expressed as rate of pressure, p, increase for the To complete this solution process, each source or
compartment and is essentially caused by net transport term in the equations must be given in
heat or mass additions to the compartment terms of the above layer properties, or auxiliary
gases. Except for the rapid accumulation of relationships must be included for each new var-
mass or energy, for compartments with small iable introduced. The source terms are associated
openings to the surroundings this pressure rise with the ω_ i terms, and the transport terms include
is small, and the pressure nominally remains at the j mass flow rates and the boundary heat
nearly the ambient pressure. For example, an transfer rates. The extent to which source and
addition of 100 kW to a 40 m3 gas volume in a transport relationships are included reflects the
room with a 0.1 m2 vent area gives rise to sophistication and scope of the zone model.
roughly an increase of 10 Pa in less than 102 s Some source and transport terms are essential to
over normal ambient pressure of 105 Pa [8]. Any a basic zone model, others can be specified as
increase in pressure within the compartment approximations to reality, and others can be
could give rise to a flow of mass through a vent, ignored when physically irrelevant. These source
and this term in Equation 29.4 may be associated and transport relationships can comprise
with a volumetric expansion effect. Conversely, a subroutines of a zone model computer code.
reduction in energy release rate will cause the The nature of these submodels is discussed
pressure to drop relative to the ambient. This below.
oxygen effect on the flame heat flux while the

Source Term Submodels second term is responsible for the thermal feed-
back from the smoke and compartment surfaces.
The principal source term for the zone model is The rate of energy release, ω_ F ΔH, required by
the rate of fuel supplied. In an experimental fire Equation 29.4 has already been discussed
this can be known if the fire source is simulated through Equations 29.5, 29.6, and 29.7. The
by a gas burner. At the other extreme, the mass of point should be made that the heat of combus-
fuel supply can result from a fire spreading over tion, ΔH, employed must be with respect to the
an array of different solid fuels. In general, mass of fuel gases pyrolyzed, given by such data
m_ s ¼ f ðfuel properties, heat transferÞ ð29:8Þ as Tewarson’s, and is not the theoretical oxygen
bomb value for the solid fuel (Chap. 36). Due to
in which the heat transfer to the fuel results from incomplete combustion, ΔH will be less than the
the flame configuration and the heated compart- theoretical value, in general.
ment. The fuel properties are still not completely The production of species can be described in
defined or conventionally accepted for fire terms of species yield, γi, such that
applications, since no general theory exists for
pyrolysis, and theories of flame spread and igni- ω_ i ¼ γi m_ s ð29:10Þ
tion are couched in terms of effective fire
properties, which are modeling parameters. Nev- For well-ventilated fires, γi may be reasonably
ertheless, data exist for fuel fire properties and constant for a given fuel, as tabulated by Khan
can enable approximate models for ṁs of reason- et al. (see Chap. 36). In general, it can vary with
able accuracy. For example, Tewarson describes time and can significantly vary as ventilation-
how the mass supply and energy release can be limited conditions are approached and achieved.
determined from fuel properties and tabulates For example, Tewarson shows that γi for CO can
properties for a number of solid fuels vary with equivalence ratio, Φ, where
(Chap. 36). For realistic items under well-
Mass of fuel available=Mass of oxygen available
ventilated conditions, Babrauskas has compiled Φ¼
r
results that could serve as initial estimates for ṁs
ð29:11Þ
in compartment fires (Chap. 26). In under-
ventilated compartment fires particularly during where r is the stoichiometric value for complete
the fully developed stage, a more complete combustion. Zukoski et al. [11] have shown how
submodel should address the fuel response to this relationship may be applied to compartment
the thermal feedback from the smoke and com- fires. The equivalence ratio, Φ, may be computed
partment walls, and the vitiated oxygen effect on in a zone (or upper layer) where combustion has
the flame heat flux. The mass of fuel supply (the occurred by computing the mass concentrations
fuel mass loss rate) can be given as [10], of the “available” fuel and oxygen in the zone.
00 Y o, l q_ external This is done by Equation 29.3 in which ω_ i is set
m_ s ¼ m_ s AF, b þ ð29:9Þ equal to zero for both the fuel and oxygen, since
Yo L
this yields the available YF and Yo values, not
where m_ 00 s is the free burning rate per unit area, their actual concentrations in the layer following
AF,b is the fuel burning area, Yo,l is the local mass combustion. The generality of considering
fraction of oxygen feeding the flame, Yo is the γi ¼ γi(Φ) for zone models is still under study,
mass fraction of oxygen (ambient), L is the fuel and its use must be considered as exploratory.
heat of gasification and q_ external is the total exter- Nevertheless, it currently offers the only practi-
nal heat flux feedback from the smoke and the cal approach for estimating species, such as CO,
compartment surfaces. The first term on the right under ventilation-limited conditions in compart-
hand side of Equation 29.9 represents the vitiated ment fires.
www.ebook777.com
has a fatal flaw; it simply means more systematic

Mass and Heat Transport Submodels data are necessary to expand the versatility of the
zone model and its accuracy. Moreover, if a zone
Entrainment model with its selected entrainment relationship
tracks well with data from an experimental fire
An essential feature of a zone model is the mass scenario, it can be assumed accurate for
rate of entrainment, ṁe, relationship for the fire simulating the process and can be used with
plume. This relationship allows the principal some assurance for that scenario. A catalogue
mechanism for flow between the lower and of empirical entrainment relationships for vari-
upper stratified gas layers. Considerable work ous object fires developed from specialized
has been performed to develop entrainment entrainment apparatus would help resolve the
relationships for pool fires or axisymmetric gas entrainment issue. This apparatus could be devel-
burner fires. Unfortunately both the ideal theoret- oped from the large calorimeter intended to mea-
ical plume models and correlations based on data sure energy release rate in which the fire plume is
vary widely, and no consensus exists among zone collected in a hood-duct system and the total flow
models in practice for the optimum pool fire rate is recorded. The reader is referred to
entrainment model. Rockett [12] illustrates the Chap. 13 for a more detailed discussion of
variations in results he found using different fire plume entrainment.
entrainment models in the Harvard/NIST Mark
VI compartment fire zone model. He found that
the layer height, entrainment rate, and layer gas Vent Flows Through Openings
temperature varied by roughly a factor of two in Vertical Partitions
among the various models. More useful data
rather than ideal mathematical models are clearly Classic models of fire in a room or building
needed to resolve this issue of accuracy for a represent the structure with an opening, such as
simple pool fire. Yet even a perfect entrainment a door or window, to the ambient surroundings.
relationship for an axisymmetric pool fire would Fire-induced flows through such openings have
not necessarily be perfect in a zone model, been well studied, and a widely accepted model
because a plume in a compartment can be subject exists to compute these flows based on the tem-
to nonsymmetric airflows that can bend the plume perature distribution of the gases on either side of
and thus affect its entrainment rate. These wind the opening. The theoretical basis of the compu-
effects can increase the entrainment rate by as tation is orifice flow utilizing Bernoulli’s equa-
much as two or three times [13], however it is tion along a streamline, as illustrated in Fig. 29.2.
also reported that even small ambient dis- The velocity at station two is given by
turbances could provide 20–50 % increases in
the measured plume mass flows compared to the
‘still’ conditions in a laboratory [14].
Rockett [12] has shown that the effect of the
entrainment model is crucial to predictions for
the developing fire. This research suggests
that the entrainment model must be representa-
tive of the actual object burning and its location
1 2
within the enclosure. However while correlations
have been developed for axisymmetric plumes
• • Streamline
and for plumes contacting a wall or corner, no

entrainment models exist for an item of furniture;
this illustrates a need for further research in this
area. Yet this does not mean that the zone model Fig. 29.2 Orifice flow
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Emmons presents the general equations that
2ð p1 p2 Þ
v2 ¼ : ð29:12Þ enable this computation to be included in a
ρ1
zone model (see Chap. 15). It is by far the most
where v1 is assumed to be zero. The mass flow accurate of the submodels, providing the basis
rate is computed by integration over the flow for linking rooms together in a zone model,
area, A, adjusted by a flow coefficient, C: which allows smoke and fire growth
ð computations for a large building.
The flow through an opening in a horizontal
m_ ¼ C ρ1 υ2 dA ð29:13Þ
partition can be compared to that for the vertical
partition, provided the pressure difference is
Emmons suggests that a value of 0.68 for C has large enough. If there is only a single vent from
an accuracy of 10 %, except at very low flow the fire compartment through a horizontal parti-
rates at the beginning of a fire (see Chap. 15). In tion, such as a ceiling, the flow must be
general, C will depend on the Reynolds number. oscillatory or bidirectional. The latter case
Figure 29.3 depicts examples of typical vent implies a zero pressure difference, with gravity
flows through an opening in a vertical partition. solely determining the flow. A theory for this
In both cases Equations 29.12 and 29.13 apply, case has been developed by Epstein [16] and
but the pressure distribution must be described has been implemented by Cooper [17]. For
appropriately. For example, in the pure natural orifice-like vents with zero pressure difference,
convection case shown in Fig. 29.3a, the pressure the volumetric exchange flow rate, V, given by
is determined by the static pressure with respect Epstein [16], is approximately
to the floor pressure, p(0). Actually it is the floor
pressure that applies in Equation 29.4 and in the h ρ þ ρ i1=2
V_ ¼ 0:055 D5 gðρ1 ρ2 Þ 1 2
perfect gas equation of state. 2
The assumption is that the flow velocities are ð29:15Þ
small compared to the vent flow velocities,
justifying the static pressure computation. Thus, where D is the diameter of the vent and ρ1 and
the vertical pressure distribution on either side of ρ2 are the corresponding fluid densities on either
the opening is computed as side of the vent. For vents of significant depth, L,
ðz the coefficient in Equation 29.15 depends on L/D.
pðzÞ ¼ pð0Þ ρg dz ð29:14Þ
0
McCaffrey and Rockett [15] illustrate the Convective Heat Transfer to Surfaces
accuracy of the hydrostatic assumption in
Fig. 29.4. The sign of the pressure difference The Q_ net loss term in Equation 29.4 is composed of
across the opening determines the flow direction. the convective and radiative heat loss to the
Fig. 29.3 Typical vent a b

flows
Neutral plane
Hot
Cold N
p
z
Natural convection Forced convection
www.ebook777.com
ΔP (N/m2)
−2 −1 0 1 2 3 4 5 6 7
2.0 Two temperature model
Numerical integration
H (m)
1.0
0
−.01 0 .01 .02 .03 0 50 100 150 200
ΔP (TORR) T (°C)
Fig. 29.4 Vertical pressure difference across a room vertical partition compared to a computation based on room fire
temperature distribution and a two-temperature zone model approximation using the hydrostatic pressure assumption [15]
boundary surfaces of the layer control volumes. computed. Convective heat transfer data for the
This involves both heat transfer from the gas walls and floor of a fire compartment or for
layers at their bulk temperatures and the heat rooms beyond the fire compartment have not
transfer from the flame. Consistent treatment of been developed. Hence, most zone models use
the flame and layer gas heat transfer must be estimates from natural convection correlations.
carried out for the zone model. If the flame
becomes large and fills the upper layer, one can-
not count the flame and gas heat transfer without Radiative Heat Transfer
being redundant.
Convective heat transfer to a ceiling by a fire The theory of radiative heat transfer is adequate
plume has been widely studied at modest scales, to develop the needed components for the zone
such that flame radiation may have been insignif- model. However, the theory is not sufficiently
icant. Alpert [18] specifically examined only developed to predict flame radiation from first
convective heating in contrast to studies by You principles without very sophisticated modeling
and Faeth [19] and Kokkala [20] who included of the soot and temperature distributions.
flame effects. Hence, flame radiation is relegated to empirical
In general, convective effects will vary along practices. Radiation from a smoke layer is easier
the ceiling, walls, and floor, and depend on the to deal with within the context of a uniform
nature and position of the fire. In some cases an property gas layer for the zone model. One diffi-
adiabatic wall temperature has been appropri- culty still is the availability of property data to
ately introduced since the driving force for con- determine the contribution of smoke particulates
vective heat transfer locally is not the bulk gas to the layer radiation properties. The discussion
layer temperature but rather the local boundary presented by Tien et al. can be used to begin a
layer temperature, which is not explicitly development of the radiative equations needed
by the zone model (see Chap. 4). Also, the Mixing Between the Layers
presentation by Beyler gives empirical
approaches to dealing with flame radiation (see The primary exchange of fluid between the lower
Chap. 66). The paper by Forney [21] lays out the and upper gas layers is due to the buoyant effect
theory and equations describing radiation of the fire plume. Secondary but significant
exchange between the gas layers and boundary mixing processes can occur due to the other
surfaces. effects. These are shown in Fig. 29.5 and include
three phenomena:
1. Exchange due to a cold flow injected into the
Conduction Heat Transfer hot layer
2. Exchange due to shear mixing associated with
The radiative and convective heat transfer from vent flows
the gas must be balanced by conduction heat 3. Exchange due to wall flows
transfer through the boundary surfaces. This bal- Phenomenon 1 is the inverse of the hot fire plume
ancing requires a numerical solution to a partial penetrating the upper layer. In both cases the
differential equation in conjunction with the fluid at the edge of the plume may not be
ordinary differential equations in time describing buoyant enough to penetrate the respective
the conservation of energy and mass for the gas layer. A comparable situation is a cold forced
layers. Usually zone models have considered jet introduced vertically into the lower layer.
only one-dimensional conduction, which should Depending on the relative temperatures, it
be adequate for most applications. Most may not escape the lower layer and, therefore,
multiple-compartment models do not consider may not penetrate into the upper layer. These
communication by conduction into the next com- are issues that can be resolved to some extent
partment, treating the structural elements as ther- by research available in the literature on
mally thick instead. In principle, there is no buoyant plumes and jets.
difficulty with developing an accurate algorithm Phenomenon 2 requires the near vent mixing
for conduction through the boundary elements behavior to be characterized. Data suggest that
for any conditions. For more information, the the flow rate of the mixed stream can be signifi-
reader is referred to the discussion by Rockett cant relative to the vent flow rate, especially for
and Milke (see Chap. 2). small vents [22]. A correlation for the mixing
Fig. 29.5 Secondary Cold flow

flows—mixing
phenomena. 1. A cold
plume descending from the
upper layer into the lower
layer. 2. Shear mixing of an
entering vent flow stream.
3. Wall flows due to local
buoyancy effects 3
2
1 Vent
flow
Fuel
www.ebook777.com
rate has been developed from saltwater building fires. Wind effects and the resultant
simulation experiments [23] based on the less pressure distribution around a tall building have
than ideal assumption that the cold incoming become standard data elements for structural
flow through the opening would behave like a design but have not been utilized for fire safety
point source buoyant plume. More recently a design. The movement of smoke through a build-
mixing model has been proposed by Utiskul ing due to the mechanical ventilation system has
[24] where the incoming cold air behaves like been simulated by network models that treat the
a jet entering the doorway with a characteristic compartment volume as uniform in properties,
velocity and diffusing downward because of including the pressure losses due to vents and
buoyancy. While the cold air descends, the duct friction. A two layer zone model has not
surrounding hot gas is entrained with a velocity been linked to the mechanical ventilation system
that is proportional to the incoming flow char- in a building. To create a link, one must include
acteristic velocity. An equation for the ratio of the full-pressure-flow characteristics of fans in
mass entrained to the total incoming mass flow both directions to allow for the possibility of
was given and single-vent compartment fire the backflow of smoke against the direction of
experiments were conducted to establish the airflow in the ducts. An attempt at this linkage
correlation for the mixing at the quasi-steady has been presented by Klote and Cooper [26],
state. The correlation exhibited a linear rela- who hypothesize a fan characteristic relationship.
tionship up to an apparent asymptote for the Ultimately an experimental study will be needed
mixing ratio of about 1.3. to lay a foundation for this analysis.
Phenomenon 3 has been discussed by Jaluria
[25]. He presents relationships that allow the
estimation of the rate of transfer of cold Fire Growth Rate
fluid adjacent to the wall in the hot upper gas
layer into the cold lower gas layer or vice In most all zone models, the fire source is con-
versa. sidered an input quantity, based on some experi-
All of these flows tend to blur the sharp dis- mental or empirical data. This limits the
tinction between the upper and lower gas layers, simulation capability of a zone model, since fire
reducing their degree of stratification. Obviously, growth and spread are not modeled. Also the
if sufficient mixing occurs, the layer may appear effects of compartment feedback due to thermal
to become well-mixed or destratified. Destrati- and vitiation (oxygen depletion) effects are not
fication should occur naturally in the context of taken into account. The versatility and utility of a
the zone model, and one should not have to zone model can only be improved by developing
switch to a well-mixed compartment model techniques for accommodating realistic fire
under these conditions. growth characteristics for building contents and
Relationships for all of these secondary flows architectural elements. This process will also
have not been developed with confidence nor have an impact on the use and development of
with full acceptance. However they are important flammability tests for hazard analysis and prod-
for improving the accuracy of a zone fire model, uct acceptability.
and additional research is still needed to further
establish their validity.
Embedded Submodels
Forced Flow Effects The detailed physics that one can include in a
zone model are only limited by current research
The effect of forced airflow on fire conditions and imagination. The zone model can be versatile
and smoke spread due to mechanical or natural in accommodating new phenomena, even if they
wind forces has always been an issue in large appear inconsistent with the uniform property
Fig. 29.6 Examples of Ceiling jet velocity

embedded phenomena
Plume temperature
Flame height, shape
Fuel
layer assumption. By analogy to the relationship vent flames, transient flow in corridors, shaft
between inviscid flow and boundary layer flow in flows (see Fig. 29.7). These phenomena require
the analysis of aerodynamic bodies, the layer more research and new strategies to enable them
properties can be regarded as a first-order approx- to be included into a zone model. Vent flames are
imation for higher order analysis. Flame and significant for fire growth into the next compart-
boundary layer phenomena within the compartment and usually follow flashover. Information
ment can be computed by regarding the layer about their rate of heat transfer and extent needs
properties as infinite reservoirs. These phenom- to be computed. Transient corridor flows are
ena can be computed after the primary layer important in the analysis of smoke transport
properties are computed. Examples of embedded through long corridors. The current zone model
phenomena are shown in Fig. 29.6. Although the methodology yields an instantaneous layer that
combustion region is assumed to be of negligible would descend, but the actual process produces a
volume in the zone model formulation, the flame transient ceiling jet. Flows in vertical shafts
height can be computed along with the velocity involve the interaction of plumes with walls,
and temperature distributions in an axisymmetric pressure-driven effects, and turbulent mixing.
fire plume [27]. Other potential embedded phe-
nomena are (1) the ceiling jet, (2) the computation
of temperature distributions over the ceiling,
(3) the deposition of soot and other products of Selected Reading and Comments
combustion on surfaces, (4) the heating and deg-
radation of structural elements, (5) balcony spill Zone models provide the integrating framework
plumes, (6) flame spread and fire growth for sur- for the phenomena of fire and its fire protection
face linings, and (7) the response of detectors engineering components. Many zone models
and/or sprinklers. have been constructed for fire predictions in
compartments. They involve the basic conserva-
tion equations, submodels describing the particu-
Unresolved Phenomena lar phenomena included, and the mathematical
algorithms for solutions. Some have developed
Some significant phenomena are not addressed user-friendly interfaces. Most provide documen-
by the zone modeling approach for fire, such as tation on the model and its use. The interested
www.ebook777.com
Fig. 29.7 Examples of

significant phenomena
absent from zone models
Shaft
flows
Vent flames
Corridor flows
Fuel
reader is referred to some published models for v fluid velocity

more detailed information [28–31]. Many zone- V volume
based computer models exist, similar both in w control volume velocity
substance and the ability to analyze fire effects Y mass fraction
in buildings. They can stimulate needed research. zl height of control volume or zone
However, more effort appears to have gone into ΔH heat of combustion
the computer code developments rather than the γi yield of species i
experimental research needed for improvement ρ density
in the model. Φ equivalence ratio
ω_F consumption rate of fuel
ω_i production rate of species
Nomenclature
Subscripts
A area compartment floor
C flow coefficient e entrained
cp specific heat at constant pressure f flame
cv specific heat at constant volume F fuel
g acceleration due to gravity i species
H compartment height j flow stream
J number of flow streams in control volume o oxygen
m mass s supplied
p pressure
Q heat transfer
r stoichiometric fuel-to-oxygen mass ratio
Superscripts
R ideal gas constant
( .) per unit time
t time
T temperature
14. B.M. Cetegen, E.E. Zukoski, and T. Kubota, “Entrain-

References ment in the Near and Far Field of Fire Plumes,”
Combustion Science and Technology, 39, pp.
1. N.D. Fowkes, “A Mechanistic Model of the 1973 and 305–331 (1984).
1974 Bedroom Test Fires,” in A Study of Room Fire 15. B.J. McCaffrey and J.A. Rockett, “Static Pressure
Development: The Second Full-Scale Bedroom Fire Measurements of Enclosure Fires,” Journal of
Test of the Home Fire Project (P.A. Croce, ed.), (July Research of the National Bureau of Standards,
24, 1974), Vol. II, FMRC Technical Report 82, 2, pp. 107–117 (1977).
No. 21011.4, pp. 8–50 (1975). 16. M. Epstein, “Buoyant-Driven Exchange Flow
2. J. Quintiere, “Growth of Fire in Building Through Small Openings in Horizontal Partitions,”
Compartments,” in Fire Standards and Safety (A.F Journal of Heat Transfer, 110 (1988).
Robertson, ed.), ASTM STP614, American Society 17. L.Y. Cooper, “An Algorithm and Associated Com-
for Testing and Materials, pp. 131–167 (1977). puter Subordinate for Calculating Flow Through a
3. R. Pape and T. Waterman, “Modification to the Horizontal Ceiling Flow Vent in a Zone-Type Com-
FIRES Pre-Flashover Room Fire Computer Model,” partment Fire Model,” NISTIR 4402, National Insti-
IITRI Project J6400, IIT Research Institute, Chicago tute of Standards and Technology, Gaithersburg, MD
(1977). (1990).
4. H.E. Mitler, “The Physical Basis for the Harvard 18. R.L. Alpert, “Convective Heat Transfer in the
Computer Fire Code,” Home Fire Project Technical Impingement Region of a Buoyant Plume,” Journal
Report No. 34, Harvard University, Cambridge, MA of Heat Transfer, 109 (1987).
(1978). 19. H.Z. You and G.M. Faeth, “Ceiling Heat Transfer
5. R.W. Bukowski, R.D. Peacock, W.W. Jones, and During Fire Plume and Fire Impingement,” Fire and
G.P. Forney, “HAZARD I: Technical Reference Materials, 3, 3 (1979).
Guide,” NIST Handbook 146, Vol. II, National Insti- 20. M.A. Kokkala, “Experimental Study of Heat Transfer
tute of Standards and Technology, Gaithersburg, MD to Ceiling from an Impinging Diffusion Flame,”
(1989). Proceedings of the 2nd International Symposium on
6. R. Friedman, “An International Survey of Computer Fire Safety Science, G. Cox and B. Langford, eds.,
Models for Fire and Smoke,: Journal of Fire Protec- Elsevier Applied Science, London (1991).
tion Engineering, Vol. 4., No. 3., 1992, pp. 81–92. 21. G.P. Forney, “Computing Radiative Heat Transfer
7. S.M. Olenik and D.J. Carpenter, “Updated Interna- Occurring in a Zone Model,” Fire Science & Technol-
tional Survey of Computer Models for Fire and ogy, 14, pp. 31–47 (1994).
Smoke,” Journal of Fire Protection Engineering, 22. B.J. McCaffrey and J.G. Quintiere, “Buoyancy-
13, 2, pp. 87–110 (2003). Driven Countercurrent Flows Generated by a Fire
8. J.G. Quintiere, “Fundamentals of Enclosure Fire Zone Source,” Heat Transfer and Turbulent Buoyant Con-
Models,” Journal of Fire Protection Engineering, 1, 3 vection, Vol. II (D.B. Spalding and N. Afgan, eds.),
(1989). Hemisphere Publishing, pp. 457–472 (1977).
9. J.G. Quintiere and A.S. Rangwala, “A Theory of 23. C.S. Lim, E.E. Zukoski, and T. Kubota, “Mixing in
Flame Extinction Based on Flame Temperature.” Doorway Flows and Entrainment in Fire Flames,”
Fire and Materials, 28, pp. 387–402 (2004). California Institute of Technology, Pasadena, CA
10. Y.P. Utiskul, and J.G. Quintiere, “An Application of (1984).
Mass Loss Rate Model with Fuel Response Effects in 24. Y. Utiskul, “Theoretical and Experimental Study on
Fully-Developed Compartment Fires” Proceedings of Fully-Developed Compartment Fires” NIST GCR
the 9th International Symposium on Fire Safety Sci- 07-907, National Institute of Standards and Technol-
ence, pp. 827–838 (2008). ogy, Gaithersburg, MD (2007).
11. E.E. Zukoski, S.J. Toner, J.H. Morehant, and 25. Y. Jaluria, “National Convection Wall Flows,” in
T. Kubota, “Combustion Processes in Two-Layered SFPE Handbook of Fire Protection Engineering
Configurations,” Proceedings of the 2nd International (P.J. DiNenno et al., eds.), National Fire Protection
Symposium on Fire Safety Science, Hemisphere Pub- Association, Quincy, MA, pp. 2-50–2-63 (1995).
lishing Corporation, New York, pp. 255–304 (1989). 26. J.H. Klote and L.Y. Cooper, “Model of a Simple
12. J.A. Rockett, “Using the Harvard/NIST Mark VI Fire Fan-Resistance Ventilation System and Its Applica-
Simulation,” NISTIR 4464, National Institute of tion to Fire Modeling,” NISTIR 89-4141, National
Standards and Technology, Gaithersburg, MD (1990). Institute of Standards and Technology, Gaithersburg,
13. J. Quintiere, W.J. Rinkinen and W.W. Jones, “The MD (1989).
Effect of Room Openings on Fire Plume Entrain- 27. B. McCaffrey, “Flame Height,” in SFPE Handbook of
ment.” Combustion Science and Technology, 26, pp. Fire Protection Engineering (P.J. DiNenno et al.,
193–201 (1981). eds.), National Fire Protection Association, Quincy,
MA, pp. 2-1–2-8 (1995).
www.ebook777.com
28. H.W. Emmons, H.E. Mitler, and L.N. Trefethen, “B-RISK User Guide and Technical Manual”.
“Computer Fire Code III,” Home Fire Project Tech- BRANZ Study Report No 282. BRANZ Ltd,
nical Report No. 25, Harvard University, Cambridge, New Zealand (2013).
MA (1978).
29. T. Tanaka, “A Model of Multicompartment Fire
Spread,” NBSIR 83-2718, National Bureau of James G. Quintiere is Professor Emeritus, Ph.D. of Fire
Standards, Washington, DC (1983). Protection Engineering at the University of Maryland. His
30. R.D. Peacock, W.W. Jones, R.W. Bukowski, and research has focused on fire dynamics, fire growth and
G.P. Forney, “CFAST- Consolidated Model of Fire flame spread.
Growth and Smoke Transport. (Version 6): Technical
Reference Guide,” SP-1026, National Institute of Colleen A. Wade is a senior scientist with BRANZ Ltd
Standards and Technology, Gaithersburg, MD (2005). in New Zealand. Her research and expertise includes
31. C. Wade, G. Baker, K. Frank, A. Robbins, R fire modeling, flame spread and performance-based
Harrison, M. Spearpoint, and C. Fleischmann, design.]
Estimating Temperatures
in Compartment Fires 30
William D. Walton, Philip H. Thomas,
and Yoshifumi Ohmiya
Introduction along with the growth stages. Fire safety design

in terms of evacuation and fire resistance was
The ability to predict temperatures developed examined by taking into account fire stage. The
in compartment fires is of great significance to stage of ignition and growth are very significant
the fire protection professional for protection of to estimate the time for fire detectors and sup-
human life and property. There are many uses for pression systems to activate. The stage of fully-
a knowledge of compartment fire temperatures, developed fire is important for the fire resistance
including the prediction of (1) the onset of haz- of building loadbearing elements and separating
ardous conditions, (2) property and structural elements.
damage, (3) changes in burning rate, pyrolysis 1. Ignition
rate and heat (energy) release rate, (4) ignition of 2. Growth
objects,(5) the onset of flashover and so on. 3. Flashover
The fundamental principles underlying com- 4. Fully-developed fire
partment fires are presented in Chap. 29. This 5. Decay
chapter gives a number of simplified solution Although many fires will not follow this ide-
techniques. alization, it provides a useful framework for the
discussion of compartment fires. All fires include
an ignition stage but, beyond that, may fail to
Fire Stages grow, or they may be affected by manual or
automatic suppression activities before going
In this chapter, compartment fires are defined as through all of the stages listed above.
fires in enclosed spaces, which are commonly
thought of as rooms in buildings, but may include
other spaces such as those found in transportation Growth Stage Definitions
vehicles such as ships, planes, trains, and the like.
Compartment fires are often discussed in Ignition Stage The period during which the fire
terms of growth stages [1]. Figure 30.1 shows begins in a compartment.
an idealized variation of temperature with time
Growth Stage Following ignition, the fire
initially grows primarily as a function of the
fuel itself, with little or no influence from
the compartment. The fire can be described in
W.D. Walton (*) • P.H. Thomas • Y. Ohmiya terms of its rate of energy and combustion
product generation. A discussion of energy

www.ebook777.com
30 Estimating Temperatures in Compartment Fires 997
Fig. 30.1 General

description of room fire
in absence of fire control Postflashover
Flashover
Temperature rise
Fully developed fire
Ignition Decay
Growth
Time
generation or burning rate can be found in pyrolized than can be burned with the oxygen
Chap. 26. If sufficient fuel and oxygen are avail- available in the compartment. In this case, the
able, the fire will continue to grow, causing the fire is said to be ventilation controlled. If there
temperature in the compartment to rise. Fires are openings in the compartment, the unburned
with sufficient oxygen for combustion are said fuel will leave the compartment in the gas flow
to be fuel controlled or well-ventilated. and may burn outside of the compartment. It will
cause fire spread to upper floors and neighboring
Flashover Flashover is generally defined as the buildings. During the fully developed stage, the
transition from a growing fire to a fully devel- environment within the compartment has a sig-
oped fire in which all combustible items in the nificant effect on the pyrolysis rate of the burning
compartment are involved in fire. During this objects.
transition there are rapid changes in the compart-
ment environment. Flashover is not a precise Decay Stage Decay occurs as the fuel becomes
term, and several variations in definition can be consumed, and the heat release rate and temperature
found in the literature. However the onset of within a compartment decline. The fire may change
flashover should be estimated for considering from ventilation to fuel controlled during this
fire safety. Most have criteria based on the period.
temperature at which the radiation from the
hot gases in the compartment will ignite all of
the combustible contents. Gas temperatures of Compartment Fire Phenomena
300–650 C have been associated with the
onset of flashover, although temperatures of Compartment Fire Model
500–600 C are more widely used [2]. The igni-
tion of unburnt fuel in the hot fire gases, the In order to calculate or predict the temperatures
appearance of flames from openings in a and other properties generated in a compartment
compartment, or the ignition of all of the fire, a description or model of the fire phenomena
combustible contents may actually be different must be created. This model will be described in
phenomena all related to flashover. terms of physical equations that can be solved to
predict the temperature in the compartment. Such
Fully Developed Fire During this stage, the a model is, therefore, an idealization of the com-
heat release rate of the fire is the greatest. The partment fire phenomena. Consider a fire that
fire behavior in this stage is influenced by starts at some point below the ceiling. It releases
conditions of enclosure such as the size and con- energy and products of combustion at a rate that
struction materials, size and form of openings, may change with time. The hot products of
type amount and distribution of fuel in the enclo- combustion form a plume that, due to buoyancy,
sure. Frequently during this stage more fuel is rises toward the ceiling above heat source. As the
998 W.D. Walton et al.
Fig. 30.2 Two-layer

model with no exchange
between layers except the
plume
Hot upper layer •
mg
T∞ Tg
Ho
•
ma Xd XN
plume rises, it draws in cool air from within

the compartment, decreasing the plume’s tem-
perature and increasing its volume flow rate. Convention
Radiation from
When the plume reaches the ceiling, it spreads upper layer Enthalpy
out under ceiling. When the extension of hot flow
current under ceiling reaches the walls, the flow
turns downward and forms a hot gas layer that Radiation Convention
descends with time as the plume’s gases continue Energy from flame
release
to flow into it. There is a relatively sharp inter-
face between the hot upper layer and the air in the
lower part of the compartment. The only Fig. 30.3 Energy balance of compartment fire during
assumed interchange between the air in the Two-layer condition
lower part of the room and the hot upper layer
comprised of products of combustion is through Calculation of Compartment Fire
the plume. As the hot layer descends and reaches Temperatures
openings in the compartment walls (e.g., doors
and windows), hot gas will flow out the openings The basic principle used to calculate the temper-
and outside air will flow into the openings. This ature in a compartment fire is the conservation of
description of compartment fire phenomena is energy. In order to estimate the temperature in
referred to as a two-layer or zone model. The enclosure, energy balance can be assumed as
basic compartment fire phenomena are shown shown in Fig. 30.3.
schematically in Fig. 30.2. As applied to the hot upper layer, the conser-
The two-layer concept assumes that the vation of energy can be simply stated as follows:
compositions of the layers are uniform, that is, the energy added to the hot upper layer by the fire
that the temperature and other properties are the equals the energy lost from the hot layer plus the
same throughout each layer. Although the tem- time rate of change of energy within the
perature of the lower layer will rise during the hot upper layer. From the time rate of change of
course of the fire, the temperature of the upper energy within the hot layer, the temperature
layer will remain greater and is the most impor- of the layer can be computed. Conservation of
tant factor in compartment fires. The assumptions energy can also be applied to the lower layer.
may be less valid for very large spaces or for long, Since the volume of the upper layer changes with
narrow spaces such as corridors and shafts. To time, and mass flows in and out of the upper
describe the vertical distribution, multi-layer layer, conservation of mass must be used along
models were being developed [3]. with the conservation of energy. Because the
www.ebook777.com
energy generated by the fire and the temperatures where

in the compartment vary as a function of time, ṁf ¼ Mass burning rate of the fuel (kg/s)
the application of conservation of energy will q ¼ Heat flux to the fuel surface(kW/m2)
result in a series of differential equations. For A ¼ Suraface area of the fuel(m2)
the purposes of examining the components of L ¼ Heat of gasification (kJ/kg)
the conservation of energy, the steady-state At initial stage in compartment fire, the
expressions for the conservation of energy in heat flux to surface of the fuel is due to the
the hot upper layer will be used. flame generated above the fuel. In case that com-
The transport of energy in a compartment fire partment fire reached flashover, it is due to the
is a very complex process. In order to formulate gas around fuel. If there are oxygen in the enclo-
expressions for the conservation of energy in a sure, pyrolysis products react with oxygen,
practical way, a number of assumptions must be generating heat and producing flames. The rate
made. It is possible to formulate the equations in of energy release is equal to the mass loss rate
a number of ways, based on the level of detail of the fuel times the heat of combustion of
desired. The expressions and assumptions used in the fuel:
this chapter are based on those commonly found
in the fire research literature and represent a Q_ ¼ m_ f Δhc ð30:2Þ
somewhat simplified description of the phenom-
ena. Additional details may be found in the where
references cited. Q_ ¼ Energy release rate of the fire (kW)
The steady-state conservation of energy for ṁf ¼ Mass burning rate of the fuel (kg/s)
the hot upper gas layer in a compartment can be Δhc ¼ Effective heat of combustion of the fuel
simply stated as follows: the energy generated by (kJ/kg)
the fire and added to the hot layer equals the The effective heat of combustion is the heat of
energy lost from the hot layer through radiation combustion expected in a fire where incomplete
and convection plus the energy convected out of combustion takes place. This amount is less than
the compartment openings. the theoretical heat of combustion as measured in
the oxygen bomb calorimeter [4].
The effective heat of combustion is often
Energy Generated by the Fire described as a fraction of the theoretical heat of
combustion. The effect of fluctuations is largely
The energy generated by the fire is the primary neglected.
influence on the temperature in a compartment In fuel-controlled fires, there is sufficient air
fire, and much research has been conducted in to react with all the fuel within the compartment.
predicting the energy release rate of many fuels In ventilation-controlled fires, there is insuffi-
under a variety of conditions. This discussion cient air within the compartment, and some of
will focus on flaming combustion, as it is the pyrolysis products will leave the compart-
most important in generating a significant tem- ment, possibly to react outside the compartment.
perature rise in a compartment. A discussion of For calculating the temperatures produced in
non-flaming combustion is found in Chap. 19. As compartment fires, the primary interest is the
a fuel is heated it releases pyrolysis products. The energy released within the compartment.
amount of pyrolysis products depends on fuel The pyrolysis rate of the fuel depends on the
properties, incident heat flux to surface of fuel, fuel type, its geometry, and the fire-induced envi-
oxygen mass fraction in the enclosure and so ronment. The energy generated in the compart-
on. The fortmula to estimate the mass burning ment by the burning pyrolysis products then
rate of the fuel can be given as follows, depends on the conditions (temperature, oxygen
concentration, etc.) within the compartment. The
m_ f ¼ qA=L ð30:1Þ processes involved are complex, and some are
not well understood, but for two fire types some Conservation of Mass
simplifying assumptions can lead to useful
methods for approximating the energy released The mass flow into the compartment and the flow
by the fire. out are related by
Fuel-controlled fires are defined as those in
which the pyrolysis rate and the energy release m_ g ¼ m_ a þ m_ f ð30:3Þ
rate are affected only by the burning of the fuel where ṁf is the mass burning rate of the fuel
itself and not by the room environment, analo- (kg/s).
gous to a fire burning outdoors on a calm day. The mass flow rate of hot gas out of a window
Babrauskas has provided data on free-burning or door is given by Rockett as [7]:
fires in Chap. 26. This data is most useful for

estimating burning rates of primarily horizontal 2 T1 T 1 1=2
m_ g ¼ Cd W o ρ1 2g 1 ðH o XN Þ3=2
fuels in preflashover fires, where the primary 3 Tg Tg
heating of the fuel is from the flames of the ð30:4Þ
burning item itself. Vertical fuels, such as wall
linings and fuels located in the upper hot gas where
layer, will likely be influenced by the ṁg ¼ Mass flow rate of hot gas out an opening
preflashover room environment. (kg/s)
Ventilation-controlled fires are defined as Cd ¼ Orifice constriction coefficient
those in which the energy release rate in the (typically 0.7)
room is limited by the amount of available Wo ¼ Width of opening (m)
oxygen. The mass flow rate of air or oxygen Ho ¼ Height of opening (m)
into the room through a door or window can be ρ1 ¼ Ambient air density (kg/m3)
calculated from the expressions described g ¼ Acceleration due to gravity, 9.8 m/s2
below and in Chap. 23. For most fuels [5], the XN ¼ Height of neutral plane (m)
heat released per mass of air consumed is a Tg ¼ Temperature of the hot upper gas layer (K)
constant approximately equal to 3000 KJ/kg. T1 ¼ Ambient temperature (K)
Therefore, the rate of energy release of the The mass flow rate of air into a door or
fire can be approximated from the air window is given by
inflow rate.
The amount of energy released by the fire 2 T 1 1=2
m_ g ¼ Cd W o ρ1 2g 1 ðXN Xd Þ1=2
that enters the hot upper layer is a function of 3 Tg
the fire, layer conditions, and geometry. For ðXN þ Xd =2Þ
most fires, approximately 35 % of the energy ð30:5Þ
released by the fire leaves the fire plume as
radiation [6]. (A discussion of flame radiation where
can be found in Chap. 23.) In a compartment Xd ¼ Height of the interface (m).
fire, a fraction of the radiated energy reaches the The expressions for mass flow in and mass
upper layer. The majority of the remaining flow out cannot be solved directly for Tg since
energy released by the fire is convected into the height to the neutral plane and interface are
the upper layer by the plume. As the plume unknown. The complete solution of these
rises, it entrains air from the lower layer, thus equations requires expressions for plume entrain-
reducing its temperature and increasing the ment and additional energy equations and is nor-
mass flow rate. For a first approximation, it can mally carried out only in computer fire models. If
be assumed that all of the energy generated the mass burning rate of the fuel is small com-
by the fire is transported to the upper layer. For pared with the mass flow rate of air into the
a complete discussion of fire plumes see compartment, the mass flow out of the opening
Chap. 13. may be approximated as equal to the mass inflow
www.ebook777.com
rate. Flows out of vents in the ceiling are The Conseil International du Bâtiment (CIB)
discussed in Chap. 66. experiments upon which Law [11] has based her
For preflashover fires in compartments with method shows a dependence on AT. It seems
typical doors or windows, the neutral plane and possible that the wide use of Equation 30.7 is a
interface can be approximated at the midlevel of result of a concentration of experimental fires in
the opening. This approximation can only be rooms of a limited range of
made after the initial smoke filling of the com-
partment is complete, and flow in and out of the AT
pffiffiffiffiffiffi
opening is established. Ao H o
For fires nearing flashover and postflashover
where
fires, the interface between the upper and lower
AT ¼ Total area of the compartment enclosing
layers is located near the floor, and the flow
surfaces (m2)
reaches a maximum for a given upper gas temper-
Traditionally, energy balances were often
ature. Rockett [7] has shown the temperature
stated in terms of the energy produced by the
dependence on the flow becomes small above
burning fuel and, thereby, led to an effective
150 C and the flow into the compartment can be
pffiffiffiffiffiffi heat of combustion of the fuel. However, this
approximated as a constant multiplied by Ao H o . practice in principle leads to the same result—
Rockett calculated values for this constant of the energy produced is related to the air flow
0.40–0.61 kg/s · m5/2, depending on the dis- for ventilation-controlled fires. Kawagoe [9]
charge coefficient of the opening. Thomas and and Magnusson and Thelandersson [12] used
Heselden estimate the value of this constant at 10.75 MJ/kg for the effective heat of combustion
0.5 kg/s · m5/2, which is the value most com- of wood in the flaming phase for fully developed
monly found in the literature [8]. The resulting compartment fires. With 16.4 MJ/kg for the heat
approximation is then of combustion of wood volatiles, this setup
pffiffiffiffiffiffi corresponds to a combustion efficiency of
m_ a ¼ 0:5Ao Ho ð30:6Þ 10.75/16.4, which is virtually identical to the
0.65 used in several computer models.
where
By far most data are based on experiments in
Ao ¼ Area of opening (m2)
which the fuel was cellulosic, and much of the
Ho ¼ Height of opening (m)
pffiffiffiffiffiffi experimental data are based on wood in the form
The term Ao H o is commonly known as the of cribs. For the post-flashover burning of a dif-
ventilation factor. The first use of this type of ferent fuel with a different chemistry, the burning
opening flow analysis for evaluating rate expressions may still be used, as long as the
postflashover fire test data is attributed to fuel is a hydrocarbon producing approximately
Kawagoe [9]. From early work analyzing such 3000 kJ for each kg of air consumed in the
data, the empirical observation was made that combustion process. Because different fuels
wood fires in rooms with small windows appeared react differently to the thermal environment and
to burn at an approximate stoichiometric rate. will pyrolyze at different rates according to the
Although flames emerging from the windows energy requirements to produce volatiles, one
implied that some fuel was burning outside, can only estimate temperatures by evaluating
calculations often suggested that enough air was the differences or obtain maximum temperatures
entering the fire for stoichiometric burning. by using stoichiometry. Fuels more volatile than
Empirical observations on wood fires [9] led to wood will probably produce lower temperatures
pffiffiffiffiffiffi inside a compartment, even if the excess fuel
m_ f ¼ 0:09Ao H o ð30:7Þ
produces a greater hazard outside the compart-
There is now a body of data [10] that modifies ment. The assumptions that the energy is related
this simple proportionality between ṁf and to the air flow and that the fuel is in stoichiomet-
pffiffiffiffiffiffi ric proportion will give an upper estimate of
Ao H o .
temperatures for ventilation-controlled fires.

Since Equation 30.7 is close to stoichiometric, Methods for Predicting Preflashover
it could, coupled with the effective heat of com- Compartment Fire Temperatures
bustion of wood, give results close to an upper
temperature limit for other fuels. The solution of a relatively complete set of
equations for the conservation of energy requires
the solution of a large number of equations that
Conservation of Energy vary with time. Although individual energy
transport equations may be solved, in general
The heat generated by burning materials within there is not an explicit solution for a set of these
a compartment is absorbed by the enclosing equations. As a result, one of two approaches can
surfaces of the compartment and any other be taken. The first is an approximate solution
structural surfaces, by the surfaces of the fuel, accomplished by hand using a limiting set of
and by the incoming air and any excess fuel. assumptions. The second is a more complete
Heat is lost to the exterior in the flames and hot solution utilizing a computer program. In either
gases that exit from the openings in the com- case, a number of methods have been developed.
partment enclosing surfaces and by radiation The methods presented are those with the widest
through the openings. Table 30.1 gives an apparent acceptance in the fire protection com-
example of an experimental heat balance munity, each with different assumptions and
measured in a small compartment, for which limitations that should be understood before
unglazed windows provided ventilation from employing the method. The methods presented
the start of the fire. in this chapter predict average temperatures
Table 30.1 illustrates the significant amount of and are not applicable to cases where prediction
heat loss in the effluent gases and shows that, of local temperatures are desired. For example,
with decreasing window area, a larger proportion these methods should not be used to predict
of the heat released will be absorbed by the detector or sprinkler actuation or the
enclosing surfaces. The total heat released, temperatures of materials as a result of direct
assuming a complete burnout, is directly propor- flame impingement.
tional to the amount of the fire load, but the rate
of heat release may also be controlled by the
ventilation. In this example, with the lower fire Method of McCaffrey, Quintiere,
load, both window areas give sufficient ventila- and Harkleroad
tion for the fuel to burn at its maximum (free-
burning) rate but, with the doubled fire load, the McCaffrey, Quintiere, and Harkleroad have
burning rate is not doubled, because the window used a simple conservation of energy expression
area restricts the ventilation rate. and a correlation with data to develop an
Table 30.1 Heat balance measured in experimental fires in a compartment of 29 m2 floor area with a fire load
of wood cribs
Heat loss from hot gases (%)

Fire load Window Heat release Structural Feedback Window
(kg) area (m2) (kcal/s) Effluent gas surfaces to fuel radiation
877 11.2 1900 65 15 11 9
5.6 1900 52 26 11 11
1744 11.2 3200 61 15 11 13
5.6 2300 53 26 12 9
2.6 1600 47 30 16 7
www.ebook777.com
approximation of the upper layer temperature in where

a compartment [13]. Applying the conservation Cd ¼ Orifice constriction coefficient
of energy to the upper layer yields Wo ¼ Width of opening (m)
Ho ¼ Height of opening (m)
Q_ ¼ m_ g c p T g T 1 þ qloss ð30:8Þ ρ1 ¼ Ambient air density (kg/m3)
g ¼ Acceleration due to gravity, 9.8 m/s2
where XN ¼ Height of neutral plane (m)
Q_ ¼ Energy (heat) release rate of the fire (kW) Since XN primarily depends on Tg, Q, _ and
ṁg ¼ Gas flow rate out the opening (kg/s) geometric factors (Ho and Wo), ṁg may be
cp ¼ Specific heat of gas (kJ/kg · K) pffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffi
replaced by gρ1 Ao H o in the two dimension-
Tg ¼ Temperature of the upper gas layer (K) less variables in Equation 30.11, without any loss
T1 ¼ Ambient temperature (K)
in generality. The effects of Tg and Q_ are
qloss ¼ Net radiative and convective heat trans-
incorporated into the correlation via other terms.
fer from the upper gas layer (kW)
Based on an analysis of test data, Equation 30.10
The left-hand side of Equation 30.8 is the
was written as a power-law relationship:
energy generated by the fire. On the right-hand
side, the first term is the heat transported from the !2=3
Q_
upper layer in the gas flow out an opening. The ΔT g ¼ 480 pffiffiffi pffiffiffiffiffiffi
second term is the net rate of radiative and con- g c p ρ1 T 1 A o H o
ð30:12Þ
vective heat transfer from the upper layer, which !1=3
is approximately equal to rate of heat conduction h k AT
pffiffiffi pffiffiffiffiffiffi
into the compartment surfaces. The rate of heat g c p ρ 1 Ao H o
transfer to the surfaces is approximated by
where
qloss ¼ hk AT T g T 1 ð30:9Þ Ao ¼ Area of opening (m2)
where Ho ¼ Height of opening (m)
hk ¼ Effective heat transfer coefficient The numbers 480, 2/3, and 1/3 were deter-
(kW/m2K) mined by correlating the expression with the data
AT ¼ Total area of the compartment enclosing from over 100 experimental fires. These data
surfaces (m2) included both steady-state and transient fires in
Substituting Equation 30.9 into Equation 30.8 cellulosic and synthetic polymeric materials and
yields the non-dimensional temperature rise in gaseous hydrocarbon fuels. Compartment height
terms of two dimensionless groups: ranged from 0.3 to 2.7 m and floor areas from
0.14 to 12.0 m2. The compartments contained a
ΔT g _ c p T 1 m_ g
Q=
¼ ð30:10Þ variety of window and door sizes. The term
T1 1 þ hk AT = c p m_ g raised to the 2/3 power in Equation 30.12
represents the ratio of the energy released to the
where ΔTg is the upper gas temperature rise
energy convected, and the term raised to the
above ambient (Tg– T1)(K).
1/3 power represents the energy lost divided
The mass flow rate of hot gas out of a window
by the energy convected.
or door can be rewritten from Equation 30.4:
Substituting the values for ambient conditions
of
2 T1 T 1 1=2
m_ g ¼ Cd W o H o ρ1 2g
3=2
1 g ¼ 9.8 m/s2
3 Tg Tg
3=2 cp ¼ 1.05 kJ/kg · K
1
XN ρ1 ¼ 1.2 kg/m3
Ho T1 ¼ 295 K
ð30:11Þ into Equation 30.12 yields [14, 15]
!1=3 4. The energy release rate of the fire must be

Q_
2
ΔT g ¼ 6:85 pffiffiffiffiffiffi ð30:13Þ determined from data or other correlations.
Ao H o h k AT 5. The characteristic fire growth time and ther-
mal penetration time of the room-lining
The heat transfer coefficient can be determined materials must be determined in order to eval-
using a steady-state approximation when the time uate the effective heat transfer coefficient.
of exposure, t, is greater than the thermal pene- 6. The correlation is based on data from a limited
tration time, tp, by number of experiments and does not contain
extensive data on ventilation-controlled fires
hk ¼ k=δ for t > tp ð30:14Þ
nor data on combustible walls or ceilings.
The thermal penetration time is defined as Most of the fuel in the test fires was near the
center of the room.
ρc
δ2
tp ¼ ð30:15Þ
k 2 Example of McCaffrey et al. Method Calculate
the upper-layer temperature of a room 3 3 m
where in floor area and 2.4 m high with a door opening
ρ ¼ Density of the compartment surface (kg/m3) 1.8 m high and 0.6 m wide. The fire source is a
c ¼ Specific heat of the compartment surface steady 750 kW fire. The wall-lining material is
material (kJ/kgK) 0.016 m (5/8 in.) gypsum plaster on metal lath.
k ¼ Thermal conductivity of compartment sur- Perform the calculation at times of 10, 60, and
face (kW/m · K) 600 s after ignition. Using Equation 30.12,
δ ¼ Thickness of compartment surface (m)
!2=3
t ¼ Exposure time (s) Q_
tp ¼ Thermal penetration time (s) ΔT g ¼ 480 pffiffiffi pffiffiffiffiffiffi
g c p ρ1 T 1 A o H o
When the time of exposure is less than the
penetration time, an approximation based on !1=3
hk A T
conduction in a semi-infinite solid is pffiffiffi pffiffiffiffiffiffi
g c p ρ1 A o H o
1=2
kρc
hk ¼ for t tp ð30:16Þ where
t
cp ¼ 1 kJ/kgK
If there are several wall and/or ceiling materials T1 ¼ 27 C (300 K)
in the compartment, an area-weighted average ρ1 ¼ 1.18 kg/m3
for hk should be used. Ao ¼ 1.8 m 0.6 m ¼ 1.08 m2
The limitations as stated by McCaffrey g ¼ 9.8 m/s2
et al. on the use of this method for estimating Ho ¼ 1.8 m
temperatures are as follows: Q_ ¼ 750 kW
1. The correlation holds for compartment upper- AT ¼ Awalls þ Afloor þ Aceiling Aopenings
layer gas temperatures up to approximately ¼ 4 ð3 2:4Þ þ ð3 3Þ þ ð3 3Þ 1:08
600 C. ¼ 28:8 m2 þ 9m2 þ 9m2 1:08m2
2. It applies to steady-state as well as time- ¼ 45:72m2
dependent fires, provided the primary tran-
sient response is the wall conduction The wall heat loss coefficient, hk, is a func-
phenomenon. tion of time.
3. It is not applicable to rapidly developing fires (a) Calculate the thermal penetration time, tp.
in large enclosures in which significant fire ρc
δ2
growth has occurred before the combustion tp ¼
products have exited the compartment. k 2
www.ebook777.com
where Method of Foote, Pagni, and Alvares

ρ ¼ Wall material density (1440 kg/m3)
k ¼ 0.48 103 kW/mK The Foote, Pagni, and Alvares method follows
c ¼ 0.84 kJ/kgK the basic correlations of McCaffrey, Quintiere,
δ ¼ 0.016 m and Harkleroad and adds data for forced-
tp ¼ 161.3 s ventilation fires. Using Equation 30.10 and not
(b) Calculate hk at 10, 60, and 600 s. introducing an expression for doorway flow
For t<tp (10, 60 s), results in the expression [16]
1=2 !0:72
hk ¼
kρc
kρc ¼ 0:581 ΔT g Q_ hk AT 0:36
t ¼ 0:63
T1 m_ g c p T 1 m_ g c p
1. At t ¼ 10 s, ð30:17Þ

0:581 1=2 where
hk ¼ ¼ 0:24 kW=m2K ΔTg ¼ Upper gas temperature rise above
60
ambient (K)
2. At t ¼ 60 s, T1 ¼ Ambient air temperature (K)
Q_ ¼ Energy (heat) release rate of the fire (kW)
0:581 1=2 ṁg ¼ Compartment mass ventilation rate (kg/s)
hk ¼ ¼ 0:098 kW=m2K
60 cp ¼ Specific heat of gas (kJ/kgK)
3. For t > tp (600 s) at t ¼ 600 s, hk ¼ Effective heat transfer coefficient
(kW/m2K)
k 0:48 103 AT ¼ Total area of the compartment-enclosing
hk ¼ ¼ ¼ 0:03 kW=m2 K surfaces (m2)
δ 0:016
The coefficient and exponents are based on
data from well-ventilated tests in a compartment
(c) Calculate the compartment temperature at with a 6 4 m floor area and a height of 4.5 m
the three times using Equation 30.12. with ventilation rates of 110–325 g/s. The com-
1. At t ¼ 10 s, partment exhaust was through a 0.65 0.65 m
" #2=3 duct located 3.6 m above the floor. Four air
750
ΔT g ¼ 480 pffiffiffiffiffiffiffi pffiffiffiffiffiffiffi inlet openings were 0.5 0.12 m high, with
9:8 ð1Þ ð1:18Þ ð300Þ ð1:08Þ 1:8 centerlines 0.1 m above the floor. A methane gas
" #1=3 burner fire in the center of the floor with heat
ð0:24Þ ð45:72Þ
pffiffiffiffiffiffiffi pffiffiffiffiffiffiffi release rates of 150–490 kW resulted in upper
9:8 ð1Þ ð1:18Þ ð1:08Þ 1:8
gas temperatures of approximately 100–300 C.
¼ 480ð0:47Þ2=3 ð2:05Þ1=3 Foote et al. have shown that the correlation for
¼ 227 K forced-ventilation fires agrees well with the data
presented by McCaffrey et al. for free ventilation
2. At t ¼ 60 s, fires with
pffiffiffi pffiffiffiffiffiffi
ΔT g ¼ 480ð0:47Þ2=3 ð0:837Þ1=3 m_ 0:1 ρ1 gAo H o

¼ 307 K
3. At t ¼ 600 s, Example of Foote et al. Method Estimate the

temperature in a 5 5 m floor area and 4-m-high
ΔT g ¼ 480ð0:47Þ2=3 ð0:26Þ1=3 compartment having 0.025-m (1-in.) thick
¼ 453 K concrete walls. The forced-ventilation rate is
2.4 m3/s of air (5000 cfm). Perform the calculation Quintiere, and Harkleroad for naturally
for t>tp. The fire size is given as 1000 kW; ambi- ventilated compartments. Beyler offers an
ent air conditions at 300 K. Using Equation 30.17, improved correlation for compartments where
!0:72
the forced-ventilation flow rate is known
ΔT g Q_ hk AT 0:36 [17, 18]. This method begins by applying the
¼ 0:63
T1 m_ g c p T 1 m_ g c p conservation of energy in the upper layer of a
compartment. Combining Equations 30.8 and
where 30.9 yields
Q_ ¼ 1000 kW
T1 ¼ 300 K Q_ ¼ m_ g c p T g T 1 þ hk AT T g T 1
cp ¼ 1.0 kJ/kgK ð30:18Þ
AT ¼ 4 (5 4) + 2 (5 5) ¼ 105 m2
ṁg ¼ (2.4 m3/s) (1.18 kg/m3) ¼ 2.8 kg/s where:
Calculate hk for t > tp. For 0.025-m-thick Q_ ¼ Energy (heat) release rate of the fire (kW)
concrete, ṁg ¼ Gas flow rate out the opening (kg/s)
δ ¼ 0.025 m cp ¼ Specific heat of gas (kJ/kgK)
ρ ¼ 2000 kg/m3 Tg ¼ Temperature of the upper gas layer (K)
k ¼ 1.4 103 kW/mK T1 ¼ Ambient temperature (K)
cp ¼ 0.88 kJ/kgK hk ¼ Effective heat transfer coefficient
(kW/m2K)
ρc
δ2
AT ¼ Total area of the compartment enclosing
tp ¼
k 2 surfaces (m2)
Rearranging Equation 30.18 yields
ð2; 000Þ ð0:88Þ 0:025 2
¼
1:4 103 2
Q_
¼ 196s for t > t p ΔT g ¼ ð30:19aÞ
m_ g c p þ hk AT
k
hk ¼ or
δ
1:4 103 ΔT g m_ g c p 1
¼ ¼ ð30:19bÞ
0:025 Q_ 1 þ ðhk AT Þ=m_ g c p
¼ 0:056 kW=m2 K
where ΔT g ¼ T g T 1 .
0:72
ΔT g 1, 000 A nondimensional temperature rise is
¼ ð0:63Þ
T1 ð2:8Þ ð1Þ ð300Þ defined as

ð0:056Þ ð105Þ 0:36 ΔT g m_ g c p
ΔT* ð30:20Þ
ð2:8Þ ð1Þ
Q_
ΔT g ¼ ð0:14Þ ðT 1 Þ and the ratio of the bounding surface loss to the

¼ 164 K ventilation losses is defined as
T g ¼ 164 þ 300 K ¼ 464 K
hk A T
Y* 1 þ ð30:21Þ
m_ g c p
Method of Beyler and Deal
By plotting ΔT* as a function of ΔY* for data
Beyler and Deal compared a number of methods with experiments with known ventilation rates
for naturally ventilated compartments to test data Beyler and Deal developed a correlation for the
and recommend the method of McCaffrey, effective heat transfer coefficient of
www.ebook777.com
rffiffiffiffiffiffiffi !
kρc k δ ¼ 0.25 m
hk ¼ 0:4max ; ð30:22Þ ρ ¼ 2000 kg/m3
t δ
k ¼ 1.4 103 kW/mK
where c ¼ 0.88 kJ/kgK
k ¼ Thermal conductivity of the compartment
k 1:4 103
surface (kW/mK) hk ¼ 0:4 ¼ 0:4
ρ ¼ Density of the compartment surface (kg/m3) δ 0:25
c ¼ Specific heat of the compartment surface ¼ 0:0224 kW=m2 K
material (kJ/kgK)
δ ¼ Thickness of the compartment surface (m) (b) Calculate the compartment temperature
t ¼ Exposure time (s) using Equation 30.19a.
The expression switches from transient to 1000
steady state at a thermal penetration time of T g 300 ¼
ð2:8Þ ð1:0Þ þ ð0:224Þ ð105Þ
tp ¼ (ρc/k)δ2 rather than tp ¼ (ρc/k)(δ/2)2 used
by McCaffrey et al. and Foote et al. For the data T g ¼ 494 K
set Beyler and Deal evaluated, the standard error
for their method was 29 K as compared to 51 K Method of Peatross and Beyler
for the method of Foote et al., even though the
equation uses only one fitting constant. The correlations used in the McCaffrey,
Beyler and Deal demonstrated that this Quintiere, and Harkleroad method and the Beyler
method works for ventilation to the lower part and Deal method are based on the assumption of
of the compartment (with or without a plenum) normal insulating wall materials. For highly con-
as well as for ventilation to the upper part of the ductive walls such as steel, Peatross and Beyler
compartment. The Beyler and Deal method was suggest the use of an alternative heat transfer
based on data up to 2000 s into fire tests. At coefficient [19]. Using a lumped mass analysis
longer times, the heat loss model breaks down. for heat transfer through the wall that is appro-
priate for a highly conductive wall yields
Example of Beyler and Deal Method Estimate
the temperature in a 5 5 m floor area and 00 dT w
4-m-high compartment with 0.025-m (1-in.) mw c ¼ hg T g T w h1 T w ð30:23Þ
dt
thick concrete walls. The forced-ventilation rate
is 2.4 m3/s of air (5000 cfm). Perform the calcu- where
00
lation for t > tp. The fire size is given as ṁw ¼ Mass per unit area of the wall (kg/m2)
1000 kW; ambient air conditions at 300 K. c ¼ Specific heat of the wall (kJ/kgK)
Using Equation 30.19a, Tw ¼ Wall temperature (K)
t ¼ Time (s)
Q_ hg ¼ Heat transfer coefficient on the hot side of
Tg T1 ¼
m_ g c p þ hk AT the wall (kW/m2K)
Tg ¼ Upper layer temperature (K)
where
H1 ¼ Heat transfer coefficient on the ambient
Q_ ¼ 1000 kW side of the wall (kW/m2K)
ṁg ¼ (2.4 m3/s) (1.18 kg/m3) ¼ 2.8 kg/s Solving for the wall temperature with the ini-
cp ¼ 1.0 kJ/kgK tial condition of the wall at ambient temperature
T1 ¼ 300 K yields
AT ¼ 4(5 4) + 2(5 5) ¼ 105 m2
hg T g hg þ h1
Tw ¼ 1 exp 00 t ð30:24Þ
(a) Calculate hk for t > tp. For 0.25-m-thick hg þ h1 mw c
concrete,
00
The heat transfer through the wall, q_ , may be The hk calculated with this method can be used
expressed in terms of the heat transfer to the hot directly in the Beyler and Deal method. It must
side of the wall or in terms of an overall effective be multiplied by 2.5 for use in the McCaffrey,
heat transfer coefficient, hk. Quintiere, and Harkleroad method to account for
the 0.4 fitting constant in the hk in the Beyler and
00 Deal method.
q_ ¼ hg T g T w ¼ hk T g T 1 ð30:25Þ
Solving for hk yields Example of Peatross and Beyler Method

for Forced Ventilation Estimate the tempera-
ture in a 5 5 m floor area and 4-m-high com-
h2g hg þ h1
hk ¼ hg 1 exp t partment having 0.00635-m (0.25-in.) thick,
hg þ h1 ρδc
0.5 % carbon steel walls. The forced-ventilation
ð30:26Þ rate is 2.4 m3/s of air (5000 cfm). Perform the
calculation for t ¼ 200 s. The fire size is given as
where 1000 kW; ambient air conditions at 300 K. Using
ρ ¼ Density of the wall material (kg/m3) Equation 30.19a,
δ ¼ Thickness of the wall (m)
hk ¼ Overall effective heat transfer coefficient Q_
Tg T1 ¼
W/m2K m_ g c p þ hk AT
From the above equations it can be seen that
where
hg h1 Q_ ¼ 1000 kW
hk ¼ at t¼1
hg þ h1 ṁg ¼ (2.4 m3/s) (1.18 kg/m3) ¼ 2.8 kg/s
cp ¼ 1.0 kJ/kgK
hk ¼ hg at t¼0 T1 ¼ 300 K
AT ¼ 4(5 4) + 2(5 5) ¼ 130 m2
From a number of experiments, Peatross and
Beyler found the heat transfer coefficients of (a) Using Equation 30.27, calculate hk for t
30 W/m2K for hg and 20 W/m2K for h1. ¼ 200 s. For 0.25-m-thick, 0.5 % carbon
Substituting these values yields steel,
δ ¼ 0.00635 m
50 ρ ¼ 7833 kg/m3
hk ¼ 30 18 1 exp t ð30:27Þ
ρδc c ¼ 0.465 kJ/kgK

50
hk ¼ 30 18 1 exp t
ρδc

50
¼ 30 18 1 exp 200
ð7833Þ ð0:00635Þ ð0:465Þ
¼ 23:7 W=m2 K
(b) Calculate the compartment temperature Method of Beyler

using Equation 30.19a.
For compartments with no ventilation the
1, 000
T g 300 ¼ quasi-steady approximation used in many of the
ð2:8Þ ð1:0Þ þ ð0:0237Þ ð130Þ methods is not appropriate since the conditions
T g ¼ 470 K in the compartment will not reach steady state.
www.ebook777.com
Beyler applied a nonsteady energy balance to the which include the fitting coefficient. Beyler
closed compartment expressed by the differential used data with a maximum temperature rise of
equation [17] 150 C to develop this correlation.
dT
mc p ¼ Q_ hk AT ΔT g ð30:28Þ Example of Beyler Method Estimate the
dt temperature in a 5 5 m floor area and 4-m-high
where “closed” compartment having 0.025-m (1-in.)
Q_ ¼ Energy (heat) release rate of the fire (kW) thick concrete walls. Perform the calculation for t
m ¼ Mass of the gas in the compartment (kg) ¼ 120 s. The fire size is given as 100 kW; ambient
cp ¼ Specific heat of gas (kJ/kgK) air conditions at 300 K. Using Equation 30.29,
ΔTg ¼ Tg – T1 2K 2 pffi pffi
Tg ¼ Temperature of the upper gas layer (K) ΔT g ¼ 2 K 1 t 1 þ eK1 t

K1
T1 ¼ Ambient temperature (K)
hk ¼ Effective heat transfer coefficient where
(kW/m2K) T1 ¼ 300 K
AT ¼ Total area of the compartment enclosing t ¼ 120 s
surfaces (m2)
ρ ¼ Density of the compartment surface (kg/m3) (a) Calculate K1 using Equation 30.30.
δ ¼ Thickness of the compartment surface (m) pffiffiffiffiffiffiffi
2ð0:4 kρcÞAT
t ¼ Exposure time (s) K1 ¼
In this case a “closed” compartment has suffi- mc p
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
cient leaks to prevent pressure buildup, but the 2 0:4 1:4 103 ð2; 000Þ ð0:88Þ ð130Þ
leakage is ignored. The mass of the fuel is ¼
ð118Þ ð1:0Þ
ignored, and the initial temperature is assumed
¼ 1:3834
to be ambient temperature. For constant heat
release rate, the solution to Equation 30.28 is where
2K 2 pffi pffi
m ¼ (100 m3) (1.18 kg/m3) ¼ 118 kg
ΔT g ¼ 2 K 1 t 1 þ eK1 t ð30:29Þ cp ¼ 1.0 kJ/kgK
K1
ρ ¼ 2000 kg/m3
where k ¼ 1.4 103 kW/mK
pffiffiffiffiffiffiffi c ¼ 0.88 kJ/kgK
2ð0:4 kρcÞAT AT ¼ 130 m2
K1 ¼ ð30:30Þ
mc p (b) Calculate K2 using Equation 30.31.
Q_ Q_ 100
K2 ¼ ð30:31Þ K2 ¼ ¼ ¼ 0:84746
mc p mc p ð118Þ ð1:0Þ
where where
k ¼ Thermal conductivity of the compartment m ¼ (100 m3) (1.18 kg/m3) ¼ 118 kg
surface (kW/mK) cp ¼ 1.0 kJ/kgK
c ¼ Specific heat of the compartment surface (c) Calculate the compartment temperature
material (kJ/kgK) using Equation 30.29.
ð2Þ ð0:84746Þ pffiffiffiffiffiffiffiffi pffiffiffiffiffiffi
T g 300 ¼ ð1:3834Þ 120 1 þ eð1:3834Þ 120

ð1:3834Þ2
T g ¼ 312:5 K
formula of the fuel, taking the products to be

Methods for Predicting Postflashover CO2, H2O, and N2.
Compartment Fire Temperatures
79
Cx H y Oz þ wO2 þ w N2
Method of Babrauskas 21

y 79
The following method is based on the work of ! xCO2 þ H2 O þ w N2 ð30:35Þ
2 21
Babrauskas [20, 21]. The upper gas temperature,
Tg, is expressed according to a series of factors, where
each one accounting for a different physical
phenomenon: 2x þ 2y=2 z
w¼ ð30:36Þ
2
T g ¼ T 1 þ ðT* T 1 Þ θ1 θ2 θ3 θ4 θ5
and
ð30:32Þ
½w þ wð3:76Þ 28:97
where T* is an empirical constant ¼ 1725 K, and r¼ ð30:37Þ
the factors θ are in Equations 30.38, 30.43, 30.45, 12:01x þ 1:00y þ 16:00z
30.46, 30.48 and 30.49.
At stoichiometry ϕ ¼ 1, and it is greater than
1 for fuel-rich burning and less than 1 for fuel-
Burning Rate Stoichiometry, θ1 The dimen-
lean conditions.
sionless stoichiometric coefficient ϕ is defined as
The effect of ϕ on gas temperatures was
m_ f evaluated by numerical computations using the
ϕ¼ ð30:33Þ
m_ f , st COMPF2 computer program [22]. The efficiency
factor, θ1, accounts for deviation from stoichiom-
where ṁ is the fuel mass pyrolysis rate (kg/s), etry and is shown in Fig. 30.4. It is seen that the
and ṁf,st is the stoichiometric mass burning rate fuel-lean and the fuel-rich regimes exhibit a very
(i.e., no excess fuel and no excess oxygen). different dependence. For the fuel-lean regime,
pffiffiffiffiffiffi the results can be approximated by
0:5Ao H o
m_ f , st ¼ ð30:34Þ
r θ1 ¼ 1:0 þ 0:51 ln ϕ for ϕ < 1 ð30:38Þ
where the ratio r is such that 1 kg fuel + r kg Similarly, in the fuel-rich regime a suitable
air ! (1 + r) kg products. The value of r is read- approximation is
ily computable for fuels containing carbon,
hydrogen, and/or oxygen from the chemical θ1 ¼ 1:0 0:05ðln ϕÞ5=3 for ϕ > 1 ð30:39Þ
Fig. 30.4 Effect 0.8

of equivalence ratio
Fuel lean
0.6 (1 – θ1) = 0.51 (lnφ)
(1 – θ1)
0.4
Fuel rich
(1 – θ1) = 0.05 (lnφ) 1.67
0.2
0
–1.6 –1.2 –0.8 –0.4 0 0.4 0.8 1.2 1.6
lnφ
www.ebook777.com
Fig. 30.5 Effect of pool 1.0

diameter
(1 – θ1) = 0.092 (–lnη)1.25
(1 – θ1)
0.1
0.01
0.1 1.0 10
(–lnη)
_ rather than mass loss rate,

If heat release rate, Q, σ ¼ Stefan-Boltzmann constant (5.67 1011
ṁ, is used, then kW/m2K4)
Tb ¼ Liquid boiling point temperature (K)
Q_ This expression unfortunately requires an esti-
ϕ¼ ð30:40Þ
Q_ stoich mate for Tg to be made, so for the pool fire case, a
certain amount of iteration is necessary. The
And since the stoichiometric heat release rate is
relationship above is plotted in Fig. 30.5.
pffiffiffiffiffiffi
Q_ ¼ 1, 500Ao Ho ð30:41Þ
Wall Steady-State Losses, θ2 The next effi-
then ciency factor, θ2, accounts for variable groups
of importance involving the wall surface (which
Q_
ϕ¼ pffiffiffiffiffiffi ð30:42Þ is defined to include the ceiling) properties: area
1, 500Ao H o
AT(m2), thickness L (m), density ρ (kg/m3), ther-
The value of Q_ can be determined from Chap. 26. mal conductivity k (kW/mK), and heat capacity
A separate procedure is necessary for pool cp(kJ/kgK). This factor is given as
fires, due to the strong radiative coupling. Here "
pffiffiffiffiffiffi2=3 1=3 #

Ao H o L
θ2 ¼ 1:0 0:94 exp 54
θ1 ¼ 1:0 0:092ðln ηÞ1:25 ð30:43Þ AT k
where ð30:45Þ
pffiffiffiffiffiffi
Ao H o 0:5Δh p and is shown in Fig. 30.6.
η¼
ð30:44Þ
Af rσ T 4 T 4
g b
Wall Transient Losses, θ3 For the transient
where case, Equation 30.45 predicts the asymptotic
Δhp ¼ Heat of vaporization of liquid (kJ/kg) temperature value. An additional time-dependent
Af ¼ Pool area (m2) factor, however, is needed (Fig. 30.7).
Fig. 30.6 Effect of wall 0.7

steady-state losses
0.6
Values for
0.
0.
00
0.5 01 L
(m2·0 C/W)
5
0.
0 K
02
0. .05
0.4 1
(1 – θ2)
0.
0.
2
1. 5
0.3 2. 0
0
0.2
0.1
0
0.001 0.01 0.1
A h (m1/2)
Aw
Fig. 30.7 Effect of wall 0.7

0.0 Values for
transient losses 00
0.0 2 t
0.6 00 (m4.0 C 2/W 2)
0. 5 ρC
00
0. 1
00
0.5 0. 2
00
0. 5
0. 01
0.4
(1 – θ3)
02
0.
05
0.3
0.2
0.1
0
0.001 0.01 0.1
A h (m1/2)
Aw
" pffiffiffiffiffiffi0:6 0:4 #

Ao H o t is not a serious limitation, since the method is
θ3 ¼ 1:0 0:92 exp 150 only designed for postflashover fires.
AT κρc
For transient fires, the possibility of two sepa-
ð30:46Þ rate effects must be considered. First, the wall
where loss effect, represented by Equation 30.46, in all
κ ¼ Thermal conductivity of wall (W/mK) fires exhibits a nonsteady character. Second, the
c ¼ Specific heat of wall (J/kgK) fuel release rate may not be constant. Since in the
If only steady-state temperatures need to be calculational procedure the previous results are
evaluated, then θ3 ¼ 1:0. not stored, it is appropriate to restrict consider-
Wall effects for t just slightly greater than ation to fires where ṁf does not change drasti-
zero are not well modeled with the above cally over the time scale established by θ3. This
relationships for θ2 θ3; however, this condition “natural” time scale can be determined as the
www.ebook777.com
time when the response has risen to 63 % of its Method of Law

ultimate value, that is, at θ3 ¼ 0.63, and is
1:5 The area of structural surface to which heat is lost
AT
t ¼ 2:92 106 ðκρcÞ pffiffiffiffiffiffi ð30:47Þ is expressed by (AT – Ao). For a given fire load,
Ao H o compartments with different values of AT, Ao,
and height Ho will have a different heat balance,
Opening Height Effect, θ4 The normalization and thus the temperatures in the compartments
of burning rate and wall loss quantities with the will differ. This is illustrated in Fig. 30.10, which
pffiffiffiffiffiffi
ventilation factor Ao H o does not completely shows how temperature varies with
determine the total heat balance. An opening of a ð AT Ao Þ
pffiffiffiffiffiffi
given Ao Ho can be tall and narrow or short and Ω¼ pffiffiffiffiffiffi
Ao H o
squat. For the shorter opening, the area will have
to be larger. Radiation losses are proportional For low values of Ω (i.e., high ventilation), the
to the opening area and will, therefore, be higher rate of heat release is at a maximum, but the heat
for the shorter opening. By slight simplification, loss from the window is also large and the resul-
a representation for θ4 can be made as tant temperature is low. For high values of Ω (i.e.,
low-ventilation areas), there is little heat loss to
θ4 ¼ 1:0 0:205H 0:3
o ð30:48Þ the outside, but the rate of heat release is also
small and the resultant temperature is, again, low.
as shown in Fig. 30.8. The curve in Fig. 30.11 has been derived from
many experimental fires conducted internation-
Combustion Efficiency, θ5 The fire compart- ally by CIB [10]. For design purposes, Law has
ment is viewed as a well, but not perfectly, stirred defined it as follows:
reactor. Thus a certain “unmixedness” is present.
A maximum combustion efficiency, bp, can be ð1 e0:1Ω Þ
T gðmaxÞ ¼ 6, 000 pffiffiffiffi ð CÞ ð30:50Þ
used to characterize this state. Since the model Ω
assumes infinitely fast kinetics, any limitations where
can also be included here. Data have not been
available to characterize bp in real fires, but ð AT Ao Þ
Ω¼ pffiffiffiffiffiffi
agreement with measured fires can generally be Ao H o
obtained with bp values in the range 0.5–0.9. The and
effect of bp variation can be described by AT ¼ Total area of the compartment enclosing
θ5 ¼ 1:0 þ 0:5 ln b p ð30:49Þ surfaces (m2)
as shown in Fig. 30.9. Ho ¼ Height of opening (m)

window height
(1 – θ4)
(1 – θ4) = 0.205Ho−0.3
0.1
0.1 1.0 10
Window height, Ho (m)
Fig. 30.9 Effect of bp, the 0.5
maximum combustion
efficiency
(1 – θ5) = 0.5 lnbp
(1 – θ5)
0.10
0.05
0.1 1.0
(1 – lnbp )
Fig. 30.10 Average

temperature during fully
1200
developed period measured
in experimental fires in
Temperature (°C)
compartments
600
0
30 60
At – Ao
(m–1/2)
A o Ho
Fig. 30.11 Variation of a

0.15
rate of burning during fully
(D/W ) (kg·s–1·m–5/2)
developed period measured

in experimental fires in
compartments
0.10
0.05
Ao Ho
mf
0.00
0 30 60
A t – Ao
(m–1/2)
Ao H o
www.ebook777.com
This equation represents an upper limit of fire value of ṁf depends on L and the type of fuel.
temperature rise for a given Ω. However, if the For example, domestic furniture has a free-
fire load is low, this value may not be obtained. burning fire duration of about 20 min, giving
The importance of the effect of fire load also τ ¼ 1200 s and m_ f ¼ L=1200.
depends on Ao and AT, and can be expressed as The temperatures discussed above are
averages measured during the fully developed
T g ¼ T gðmaxÞ 1 e0:05Ψ ð CÞ ð30:51Þ period of the fire. It is assumed that all fires are
ventilation controlled, with the simple relation-
where
ship for rate of burning given by Equation 30.53,
Tg ¼ Average temperature in the compartment
which is near stoichiometric burning, and it is
( C)
assumed that combustion of 1 kg of wood
where
releases 18.8 MJ in total.
L
Ψ¼
½Ao ðAT Ao Þ 0:5 Method of Delichatsios et al.
Delichatsios’ proposed a method assuming
where L is the fire load (wood) in kg. that [23];
The effect of the fire on the structure depends 1. determination of uniform maximum gas tem-
not only on the value of Tg but also on the perature in the compartment for adiabatic
duration of heating. The effective fire duration, conditions
τ, in seconds, is given by 2. determination of an average heat flux to the
compartment boundary during the develop-
L
τ¼ ð30:52Þ ment of the fire
m_ f 3. a transient thermal model for the response
where ṁf is the rate of burning measured in kg/s. of the compartment boundary to account
Equation 30.7 implies that the smaller the for heat losses to the boundary of the
pffiffiffiffiffiffi compartment.
value of Ao H o the lower the rate of burning
First, gas temperature in the compartment is
and the longer the duration. Assuming a com-
defined from the following equation at quasi-
plete burnout, therefore, the effect on the struc-
steady conditions:
ture tends to be more severe for large values of Ω
pffiffiffiffiffiffi

given for small Ao H o .

For design purposes the following equation Q_ c ¼ m_ g c p T *g T 1 þ σAo T *4
g T1
4
has been developed to express the correlation of ð30:54Þ

experimental results [11]:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi where
m_ f ¼ 0:18Ao Ho ðW=DÞ 1 e0:036Ω ξ < 60
Q_ c ¼ Heat release rate inside the compartment
ð30:53Þ (kW)
ṁg ¼ Flow rate of gases out of the compartment
where
(kg/s)
W ¼ Compartment width (m)
Tg* ¼ Adiabatic temperature leaving through the
D ¼ Compartment depth (m)
opening (K)
1=2 T1 ¼ Ambient temperature (K)
m_ f D
ξ¼ pffiffiffiffiffiffi Ao ¼ Area of the opening (m2)
Ao H o W
Assuming ventilation controlled fire, the heat
Equation 30.53 is shown in Fig. 30.11 over the release rate inside the enclosure can be calcu-
range where the data lie. Both equations are for lated using following equation.
ventilation-controlled fires. When there is ample pffiffiffiffiffiffi
ventilation, so that the fuel is free burning, the Q_ c ¼ 1, 500Ao Ho ð30:55Þ
The mass flow rate of hot gas out of the opening instances, the energy release must be less than
can be given by the equation below. stoichiometric. The method does not take into
pffiffiffiffiffiffi account that the actual mass loss rate may be
m_ g ¼ 0:5Ao Ho ð30:56Þ greater than stoichiometric, with the excess fuel
where burning outside the compartment. A computer
Ho ¼ Height of opening (m) program, SFIRE (versions 1 through 3), is avail-
Substituting Equations 30.55 and 30.56 into able to perform this method. The results from
Equation 30.54, the following energy balance the computer program have been compared with
equation is obtained for the case of ventilation a large number of full-scale fire experiments,
controlled fires. both in the fuel- and ventilation-controlled
pffiffiffiffiffiffi pffiffiffiffiffiffi

regimes, with good agreement between theory
1, 500Ao Ho ¼ 0:5Ao H o C p T *g T 1 and experiment. It should be added, however,

that most of the experiments involved wood crib
þ σAo T *4
g T 4
1 fires, which inherently burn slower and
produce less excess fuel load than furnishings
ð30:57Þ
and other combustibles found in practical
Equation 30.57 can be solved by using simple fire loads. In the Swedish method, the fire
numerical inversion to obtain the adiabatic tem- load is expressed in relation to AT as Q_ ¼
perature. As for Equation 30.57, the adiabatic 18:8 L=AT MJ=m2 .
pffiffiffiffiffiffi
temperature depends only on Ao H o . The design curves approved by the Swedish
From the dimensional analysis, authorities were computed on the basis of
0 1 systemized ventilation-controlled heat-release
pffi
Tg T1 t Q _c
¼ function@pffiffiffiffiffiffiffiffiffiffiffiffiffi
A: curves taken from Magnusson and Thelandersson
T *g T 1 ðkρcÞw AT T * T 1 [12]. Figure 30.12 shows some typical curves.
g
The curves are calculated for wall, floor, and
ð30:58Þ ceiling materials with “normal” thermal
where properties from an energy balance which assumes
AT ¼ Total area of the compartment enclosing a uniform temperature in the compartment.
surfaces (m2)
Compared to the experimental data, the model Japanese Method
for gas temperature in the enclosure is proposed
below for the growing period of fire before burn- The McCaffery’s method (Equation 30.12)
out occurs, was originally derived for fuel-controlled fires
0 11=2 [24]. However, the formula was extended to
pffi
Tg T1 t Q_c calculate the temperature of ventilation controlled
¼ 0:5@pffiffiffiffiffiffiffiffiffiffiffiffiffi
A fires as well [25]. The fire temperature in the com-
T *g T 1 ðkρcÞw AT T * T 1
g partment and fire duration are calculated by
ð30:59Þ !2=3
Q
pffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffiffi
T f ¼ 1, 280 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffi t1=6 þ T 1
AT kρc Ao H o
Swedish Method ð30:60Þ
The Swedish method, developed by Magnusson

and Thelandersson [12], is based on the conven-
tional mass and energy balance equations.
1 FL Ar
The fire itself is not modeled; heat release tD ¼ ð30:61Þ
where 60 Q
rate curves are provided as input and, in all
www.ebook777.com
1200
Ao Ho /At = 0.02 m–1/2 Ao Ho /At = 0.04 m–1/2
1000
800 Qt = 500 MJ/m2

Qt = 250 MJ/m2
600
50 100 150 200 100 200 300 400
400 37.5 75
25 50
200 12.5 25
Tg (°C)
1 2 3 4 5 6 1 2 3 4 5 6
Time (hr)
1200
Ao Ho /At = 0.08 m–1/2 Ao Ho /At = 0.12 m–1/2
1000
Qt = 1000 MJ/m2 Qt = 1500 MJ/m2
800
600
200 400 600 800 300 600 900 1200
400
225
150 150
200 100
50 75
1 2 3 4 5 6 1 2 3 4 5 6
Time (hr)
pffiffiffiffiffiffiffiffiffiffiffiffiffi
Fig. 30.12 Examples of gas temperature-time curves of area, At, and the opening factor Ao H o =At . Fire compart-
postflashover compartment fires for different values of the ment, type A—from authorized Swedish standard
fire load density Q_ t MJ per unit of total internal surface specifications [12]
Q ¼ Heat release rate by combustion (MW) The heat release rate is calculated by the
AT ¼ Internal surface area of compartment burning type index (fraction of ventilation factor
enclosure (m2) to surface area of fuel),
pffiffiffiffiffiffiffi
kρc ¼ Thermal inertia of compartment enclo- pffiffiffiffiffiffi
Ao H o
sure (kW.s1/2/m2.K) χ¼ : ð30:62Þ
Afuel
Ao ¼ Area of window opening (m2)
Ho ¼ Height of window opening (m) The fuel surface area was assumed to follow the
T1 ¼ Initial and ambient temperatures ( C) following formula [26].
FL ¼ Fire load density (MJ/m2) 1=3
Ar ¼ Floor area of the room (m2) Afuel ¼ 0:26FL Aroom ð30:63Þ
tD ¼ Fire duration (min.) Using burning type index, the heat release rate is
calculated by
8
>
> 1:6χ ðχ 0:081Þ
<
Q ¼ Afuel 0:13 ð0:081 < χ 0:1Þ ð30:64Þ
>
>
:
2:5χexpð11χ Þ þ 0:048 ð0:1 < χ Þ
Example of Japanese Method Calculate the The compartment enclosure is made of

pffiffiffiffiffiffiffi
maximum temperature of a room; 3 m width concrete with kρc ¼ 1:75 kW s1=2 =m2 K. Per-
3 m in floor depth 2.4 m high with a door form calculation until burnout.
opening installed at 1.8 m high 0.6 m wide. The fuel surface area and ventilation
The internal surface area is 31.32 m2. The fire factors are
load density is 720 MJ/m2.
1=3
Afuel ¼ 0:26 FL Aroom ¼ 0:26 7201=3 ð3 3Þ ¼ 21:0 m2
pffiffiffiffiffiffi pffiffiffiffiffiffiffi
Ao Ho ¼ 0:6 1:8 1:8 ¼ 1:45 m5=2
The burning type index is For example, the fire temperature at fire
pffiffiffiffiffiffi duration is
Ao H o 1:45
χ¼ ¼ ¼ 0:069
Afuel 21:0 T f ¼ 522 46:61=6 þ 20 ¼ 1, 009 C
As χ < 0.081, fire is ventilation-controlled. The
heat release rate and fire duration are
Predicting Flashover
Q ¼ Afuel 1:6χ ¼ 21:0 1:6 0:069 ¼ 2:32MW
1 FL Ar 1 720 3 3 One use of predicted compartment fire
tD ¼ ¼ ¼ 46:6 min:
60 Q 60 2:32 temperatures is estimating the likelihood of
flashover. The methods used are similar to those
Combining the results, the fire temperature in the
used in the prediction of temperature. In one case,
compartment can be calculated by the following
that of McCaffrey et al., the method is simply an
equation as shown in Fig. 30.13.
extension of the temperature calculation.
!2=3
Q
pffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffiffi
T f ¼ 1, 280 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffi t1=6 þ T 0
AT kρc Ao H o Method of Babrauskas
2=3
2:32
¼ 1, 280 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffi t1=6 þ 20 Babrauskas uses the energy balance for the upper
31:32 1:75 1:45
layer given in Equation 30.8, where the gas flow
¼ 522t1=6 þ 20 rate out of the opening is approximated by [27]
Fig. 30.13 Example of 1200

temperature-time curve
calculated by Japanese 1000
method
Temperature (°C)
800
600
400
200
0
0 10 20 30 40 50
Time(min)
www.ebook777.com
pffiffiffiffiffiffi
m_ g 0:5Ao Ho ð30:65Þ Q_ ¼ 0:5 Q_ stoich
The primary energy loss is assumed to be which, substituting into Equation 30.68 yields
radiation to 40 % of the wall area, which is at pffiffiffiffiffiffi
Q_ ¼ 750Ao Ho ð30:69Þ
approximately ambient temperature:

The 33 test fires used had energy release rates
qloss ¼ εσ T 4g T 41 ð0:40AT Þ ð30:66Þ from 11 to 3840 kW, with fuels primarily of wood
pffiffiffiffiffiffi
and polyurethane. Ventilation factors Ao H o
where ranged from 0.03 to 7.51 m5/2, and surface area
ε ¼ Emissivity of the hot gas to ventilation factor ratios
σ ¼ Stefan-Boltzmann constant 5.67 1011
kW/m2K4 AT
pffiffiffiffiffiffi
Combining Equations 30.8, 30.65 and 30.66, Ao H o
using a gas temperature for flashover of 873 K, a ranged from 9 to 65 m1/2.
specific heat of air of 1.0 kJ/kgK, an emissivity
of 0.5, and assuming the correlation between Example of Babrauskas’s Method Calculate
compartment wall and opening area of the heat release rate necessary to cause flashover,
AT using the method of Babrauskas. Assume the
pffiffiffiffiffiffi 50 same room as in the McCaffrey et al. method
Ao H o
example for predicting compartment fire
yields a minimum Q_ required for flashover, temperatures. From Equation 30.69
pffiffiffiffiffiffi
pffiffiffiffiffiffi Q_ ¼ 750Ao Ho
Q_ ¼ 600Ao Ho ð30:67Þ
where
The airflow into the compartment has been Ao ¼ 1.08 m2
approximated as Ho ¼ 1.8 m
pffiffiffiffiffiffi Q_ ¼ (750)(1.08)(1.8)1/2 ¼ 1087 kW
0:5Ao Ho
The maximum amount of fuel that can be burned

Method of McCaffrey, Quintiere,
completely with this air is known as the stoichio-
and Harkleroad
metric amount. For most fuels, the heat released
per mass of air consumed is a constant approxi-
The method of McCaffrey, Quintiere, and
mately equal to 3000 kJ/kg. Therefore, the
Harkleroad for predicting compartment fire
stoichiometric heat release rate can be Q_ stoich temperatures may be extended to predict the
calculated: energy release rate of the fire required to result
pffiffiffiffiffiffi in flashover in the compartment.
Q_ stoich ¼ 3, 000 m_ g ¼ 3, 000 0:5Ao H o
Equation 30.12 can be rewritten as
pffiffiffiffiffiffi
¼ 1, 500Ao H o " #1=2
ð30:68Þ pffiffiffi ΔT g 3 pffiffiffiffiffiffi
_
Q¼ g c p ρ1 T 1
2
h k AT Ao H o
480
From this derivation, it is shown that the min-
ð30:70Þ
imum Q_ required for flashover equals 0:4 Q_ stoich .
Comparing these results with fire tests, Selecting an upper gas temperature of 522 C
Babrauskas found that the data fall within a and ambient temperature of 295 K or ΔTg ¼
range of Q_ ¼ 0:3 Q_ stoich to Q_ ¼ 0:7 Q_ stoich . 500 C for flashover, and substituting values for
A best fit of the data suggests the gravitational constant (g ¼ 9.8 m/s2), the
specific heat of air (cp ¼ 1.0 kJ/kgK), and the Method of Thomas
density of air (ρ1 ¼ 1.18 kg/m3), and rounding
607.8–610 yields Thomas uses the energy balance for the upper
pffiffiffiffiffiffi
1=2 layer shown in Equation 30.8, where the gas flow
Q_ ¼ 610 hk AT Ao H o ð30:71Þ rate out of the opening is approximated by [2]
pffiffiffiffiffiffi
where m_ g 0:5Ao Ho ð30:73Þ
hk ¼ Effective heat transfer coefficient
Thomas develops an expression for q_ loss which
(kW/m2K)
assumes the area for the source of radiation for
AT ¼ Total area of the compartment surfaces (m2)
roughly cubical compartments is AT/6:
Ho ¼ Height of opening (m) AT
A
T
Using Equation 30.13 yields a slightly differ- q_ loss hc T g T w þ εσ 2T 4g T 4floor
2 6
ent value, 623.6 rounded to 620, of the leading ð30:74Þ
coefficient because of the difference in the value
used for the specific heat of air: where
pffiffiffiffiffiffi
1=2 AT ¼ Total area of the compartment-enclosing
Q_ ¼ 620 hk AT Ao H o ð30:72Þ surfaces (m2)
hc ¼ Convective heat transfer coefficient
The use of either 610 or 620 is acceptable (kW/m2K)
within the accuracy of the expression. Tw ¼ Temperature of the upper walls (K)
Tfloor ¼ Temperature of the floor (K)
Example of McCaffrey et al.’s Method Estimate From experimental data, Thomas developed
the energy release rate required for flashover of a an average for q_ loss of 7.8 AT. Using an upper
compartment. Assume the same room as in the layer temperature of 577 C or a ΔTg of 600 C
McCaffrey et al. method example for predicting for flashover criterion and cp ¼ 1.26 kJ/kgK
compartment fire temperatures. Assuming yields an expression for the minimum rate of
ΔTg ¼ 500 C as a condition for flashover, and energy release for flashover:
air properties at 295 K, use Equation 30.71 and pffiffiffiffiffiffi
assume the compartment has heated for a period Q_ ¼ 7:8AT þ 378Ao Ho ð30:75Þ
of time that exceeds the thermal penetration time.
pffiffiffiffiffiffi
1=2
Q_ ¼ 610 hk AT Ao Ho Comparison of Methods for Predicting
Flashover
where
Babrauskas has compared the effect of room wall
k 0:48 103 area on the energy release required for flashover,
hk ¼ ¼ ¼ 0:03 kW=m2 K
δ 0:016 using the above methods [28]. The results of his
comparisons, along with some experimental data
for rooms with gypsum board walls, are shown in
AT ¼ 45.72 m2
Fig. 30.14. The graph shows the energy required
Ao ¼ 1.08 m2
for flashover as a function of compartment wall
Ho ¼ 1.8 m
area, both normalized by the ventilation factor
Therefore, pffiffiffiffiffiffi
Ao Ho . The method of Babrauskas used in this
pffiffiffiffiffiffiffi1=2 graph is based on Equation 30.32 with T1 ¼ 25
Q_ ¼ 610 ð0:03Þ ð45:72Þ ð1:08Þ 1:8
C and Tg ¼ 600 C. Babrauskas observes that
¼ 860 kW over the range of compartment sizes of most
www.ebook777.com
Fig. 30.14 The effect of Data McCaffrey et al.

room wall area (gypsum 50% of stoichiometric line Thomas
walls) on the heat required Babrauskas
for flashover
1.0
Fraction of stoichiometric fuel

1400
Q /Ao Ho (kW·m–5/2)
1200 0.8
1000
0.6
800
600 0.4
400
0.2
200
0
0 10 20 30 40 50 60 70
Aw
(m–1/2)
Ao H o
interest, all of the methods produce similar k thermal conductivity of the wall
results. The method of McCaffrey et al. diverts (kW/mK)
from the others for small room sizes. Babrauskas L fire load, wood (kg)
notes that all of the methods are a conservative m mass of the gas in the compartment
representation of the data. (kg/s)
ṁa mass flow rate of air into an opening
(kg/s)
Nomenclature ṁg gas flow rate out the opening (kg/s)
ṁf mass burning rate of fuel (kg/s)
Aceiling area of compartment ceiling (m2) ṁf,st stoichiometric mass burning rate of
Af pool fire area (m2) fuel (kg/s)
Afloor area of compartment floor (m2) m_ 00w mass per unit area of the wall (kg/m2)
Ao area of openings (m2) q_ loss net radiative and convective
AT total area of the compartment heat transfer from the upper gas layer
enclosing surfaces (m2) (kW)
Awalls area of compartment walls (m2) _
Q energy (heat) release rate of the fire
bp maximum combustion efficiency (kW)
c specific heat of the wall (kJ/kgK) _
Q stoich
stoichiometric heat release rate (kW)
Cd orifice constriction coefficient
t time (s)
cp specific heat of gas (kJ/kgK)
tp thermal penetration time (s)
D compartment depth (m)
Tb liquid boiling point (K)
g acceleration due to gravity, 9.8 m/s2
Tfloor temperature of the floor (K)
hc convective heat transfer coefficient
Tg temperature of the upper gas layer
Δhc effective heat of combustion of the
(K)
fuel (kJ/kg)
Tp thermal penetration time (s)
hg heat transfer coefficient on the hot side
Tw wall temperature (K)
of the wall (kW/m2K)
T1 ambient temperature (K)
hk effective heat transfer coefficient
W compartment width (m)
(kW/m2K)
Wo width of opening (m)
h1 heat transfer coefficient on the ambient
Xd height of the interface (m)
side of the wall (kW/m2K)
XN height of neutral plane (m)
Ho height of opening (m)
7. J.A. Rockett, “Fire Induced Gas Flow in an Enclo-

Greek Letters sure,” Combustion Science and Technology, 12, pp.
165–175 (1976).
8. P.H. Thomas and A.J.M. Heselden, “Fully Developed
δ thickness of the wall (m) Fires in Single Compartments,” Fire Research Note
ε emissivity of the hot gas No. 923, Fire Research Station, Borehamwood, UK
ρ density of the wall (kg/m3) (1972).
9. K. Kawagoe, “Fire Behaviour in Rooms,” Report of the
ρ1 ambient air density (kg/m3) Building Research Institute, No.27, Building Research
σ Stefan-Boltzmann constant, 5.67 1011 Institute, Ministry of Contruction, Japan (1958).
kW/m2K4 10. P.H. Thomas and A.J.M. Heselden, “Fully Developed
Fires in Single Compartments,” CIB Report No. 20, A
Co-operating Research Programme of the Conseil
Subscripts International du Batiment, Joint Fire Research Orga-
nization Fire Research Note 923/197.
a air 11. M. Law, Structural Engineering, 61A, 1, p. 25 (1983).
b boiling 12. S.E. Magnusson and S. Thelandersson, “Temperature-
Time Curves of Complete Process of Fire Develop-
d thermal discontinuity ment. Theoretical Study of Wood Fuel Fires in
f fuel Enclosed Spaces,” Civil Engineering and Building
g gas Construction Series No. 65, Acta Polytechnica
N neutral plane Scandinavia, Stockholm, Sweden (1970).
13. B.J. McCaffrey, J.G. Quintiere, and M.F. Harkleroad,
o opening “Estimating Room Fire Temperatures and the Likeli-
stoich stoichiometric hood of Flashover Using Fire Test Data Correlations,”
T total Fire Technology, 17, 2, pp. 98–119 (1981).
w wall 14. J.G. Quintiere, “A Simple Correlation for Predicting
Temperature in a Room Fire,” NBSIR 83-2712,
1 ambient National Bureau of Standards, Washington, DC
(June 1983).
Superscripts 15. J.R. Lawson and J.G. Quintiere, “Slide-Rule
Estimates of Fire Growth,” NBSIR 85-3196,
National Bureau of Standards, Washington, DC
. per unit time (s1) (June 1985).
00
per unit area (m1) 16. K.L. Foote, P.J. Pagni, and N.J. Alvares, “Tempera-
ture Correlations for Forced-Ventilated Compartment
Fires,” in Proceedings of the First International Sym-
posium, International Association for Fire Safety Sci-
References ence, Hemisphere Publishing, Newport, Australia,
pp. 139–148 (1986).
1. D. Drysdale, “The Pre-Flashover Compartment Fire,” 17. C. Beyler, “Analysis of Compartment Fires with
An Introduction to Fire Dynamics, John Wiley and Forced Ventilation,” Fire Safety Science—
Sons, Chichester, UK, pp. 278–303 (1985). Proceedings of the Third International Symposium,
2. P.H. Thomas, “Testing Products and Materials for Elsevier Science, New York, pp. 291–300 (1991).
Their Contribution to Flashover in Rooms,” Fire and 18. S. Deal and C. Beyler, “Correlating Preflashover
Materials, 5, 3, pp. 103–111 (1981). Room Fire Temperatures,” Journal of Fire Protection
3. K. Suzuki, K. Harada, T. Tanaka, “A Multi-Layer Engineering, 2, 2, pp. 33–88 (1990).
Zone Model for Predicting Fire Behavior in a Single 19. M.J. Peatross and C.L. Beyler, “Thermal Environ-
Room”, Fire Safety Science - Proceedings of the 7th mental Prediction in Steel-bounded Preflashover
International Symposium on Fire Safety Science, Compartment Fires,” in Fire Safety Science—
Worcester, June 2002, pp. 851–862, 2003 Proceedings of the Fourth International Symposium,
4. NFPA 259, Standard Test Method for Potential Heat International Association for Fire Safety Science,
of Building Materials Boston, pp. 205–216 (1994).
5. C. Huggett, “Estimation of Rate of Heat Release by 20. V. Babrauskas, “A Closed-Form Approximation for
Means of Oxygen Consumption Measurements,” Fire Post-Flashover Compartment Fire Temperatures,”
and Materials, 4, 2, pp. 61–65 (1980). Fire Safety Journal, 4, pp. 63–73 (1981).
6. J. de Ris, Fire Radiation—A Review, Tech. Report 21. V. Babrauskas and R. B. Williamson, “Post-Flashover
FMRC, RC78-BT-27, Factory Mutual Research Cor- Compartment Fires: Basis of a Theoretical Model,”
poration, Norwood, MA, pp. 1–41 (1978). Fire and Materials, 2, 2, pp. 39–53 (1978).
www.ebook777.com
22. V. Babrauskas, “COMPF2—A Program for Calculat- 28. V. Babrauskas, “Upholstered Furniture Room
ing Post-Flashover Fire Temperatures,” NBS TN 991, Fires—Measurements, Comparison with Furniture
National Bureau of Standards, Washington, DC (1979). Calorimeter Data, and Flashover Predictions,” Jour-
23. M. Delichatsios, Y. P. Lee, P. Tofilo (2009) A new nal of Fire Science, 2, pp. 5–19 (1984).
correlation for gas temperature inside a burning enclo-
sure. Fire Safety Journal 44(8):1003–1009
24. Ministry of construction of Japan, Notification 1430, William D. Walton is retired from the Building and Fire
2000 Research Laboratory, National Institute of Standards and
25. K. Matsuyama, T. Fujita, H. Kaneko, Y. Ohmiya, Technology.
T. Tanaka, T. Wakamatsu, “A Simple Predictive
Method for Room Fire Behavior”, Fire Science and Philip H. Thomas was with the Fire Research Station,
Technology, Vol. 18 (1998) No. 1 Borehamwood, England.
26. K. Aburano, H. Yamanaka, Y. Ohmiya, K. Suzuki,
T. Tanaka, T. Wakamatsu, “Survey and Analysis on Yoshifumi Ohmiya is a professor in the department of
Surface Area of Fire Load”, Fire Science and Tech- architecture, graduate school of science and technology,
nology, Vol. 19 (1999) No. 1 Tokyo University of Science.
27. V. Babrauskas, “Estimating Room Flashover
Potential,” Fire Technology, 16, 2, pp. 94–104 (1980).
Zone Computer Fire Models

for Enclosures 31
William D. Walton, Douglas J. Carpenter,
and Christopher B. Wood
Introduction deterministic models represent the ultimate

capability: the evaluation of discrete changes
Understanding the behavior of fire in compartments in any physical parameter in terms of the effect
is of interest to the fire protection engineer for both on fire hazard. The emphasis in this chapter is on
fire safety design and postfire reconstruction. Such deterministic fire models and, more specifically,
understanding may be obtained by examining zone fire models for enclosures.
experimental fires (full or reduced scale) or by fire Although manual or “hand calculation”
models using mathematical techniques to represent methods provide reasonable estimates of specific
the processes encountered in compartment fires by fire effects (e.g., prediction of time to flashover),
interrelated expressions based on physics and they are not well suited for comprehensive
chemistry. The two major classes of fire models analyses involving the time-dependent and
for analyzing enclosure fire development are sto- highly-coupled interactions of multiple physical
chastic and deterministic. and chemical processes present in growing com-
Stochastic or probabilistic fire models gener- partment fires. With the advent of the personal
ally treat fire growth as a series of sequential computer in the 1980s came the development of
events or states. Mathematical rules are computer fire models with increasing complexity.
established to govern the transition from one Computer fire models perform large numbers of
event to another (e.g., from ignition to tedious and lengthy calculations (in a fraction of
established burning). Probabilities are assigned the time required by “hand calculations”) and can
to each transfer point based on analysis of rele- provide analytical solutions to problems that are
vant experimental data, historical fire incident impractical to solve manually or for which no
data, and computer model results. Deterministic simple closed form analytical solution exists. As
fire models represent the processes encountered early as 1985, it had become widely accepted that
in fire by interrelated mathematical expressions the future of fire engineering would include various
based on physics and chemistry. These models levels of modeling aided by the modern computer
may also be referred to as room or compartment [1]. Today, computer models are used in many
fire models or mathematical fire models. Ideally, areas of fire protection engineering, including sup-
pression system design, smoke control system
W.D. Walton (retired) design, and egress analysis that all provide an engi-
National Institute of Standards and Technology neering approximation of the time-varying
D.J. Carpenter (*) conditions associated with fires.
Combustion Science and Engineering The state of the art in computer fire modeling is
C.B. Wood evolving for two reasons: increases in fundamental
FireLink LLC knowledge or understanding of the processes and

www.ebook777.com
31 Zone Computer Fire Models for Enclosures 1025
increases in computer processing speeds. fundamental principles behind the zone fire
Increased understanding of the processes involved model formulation can be found in Chap. 29.
in fire growth improves the technical basis for the The beginnings of preflashover zone fire
models. The capabilities, documentation, and sup- modeling can be traced to the mid-1970s with
port for a given model can change dramatically the publication of a description of the fundamen-
over a short period of time. In addition, computer tal equations by Quintiere [2]. Based on these
technology itself (both hardware and software) is equations, the first zone fire model published
advancing rapidly. In the past, a large mainframe was RFIRES by Pape, Waterman, and Eichler
computer was required to use most available com- [3], followed shortly by the Harvard model by
puter fire models. Today, most computer fire Emmons and Mitler [4, 5]. Following the
models can be run on personal computers. Despite publication of these two models, a number of
the increased awareness and practical application zone fire models for mainframe computers were
of fire modeling, current computer-based models introduced. In 1985 ASET-B, the first zone
are not substantially different from their model written specifically for the newly avail-
predecessors; they are just more complex and pos- able IBM-compatible personal computer, was
sess greater capabilities. This chapter will provide introduced by Walton [6]. Since that time addi-
a description of the characteristics of zone models, tional models have been introduced, and most of
an overview of selected models, and a discussion the models written for mainframe computers
of the selection, validation, and application of such have been converted for use on personal
models. computers.
Zone models allow for relatively inexpensive
parametric studies by providing results very
Zone Models quickly on modern computers. This capability
allows for the combination of engineering judg-
The most common type of physically based fire ment and deterministic modeling to isolate par-
model is the zone or control volume model, ticular scenarios of interest when more in-depth
which solves the conservation equations (i.e., field modeling is required to answer the ultimate
conservation of mass and energy) for discrete questions posed for the modeling activity. This
and relatively large control volumes. Although section discusses more specific characteristics
many zone models use two control volumes of zone models that may assist the modeler in
corresponding to an upper (hot) layer and lower making an appropriate model selection.
(cool) layer, other zone models may have differ- The dominant characteristic of a zone fire
ent approaches for specific problems such as a model is the division of the compartment(s) or
single control volume for postflashover room(s) into a hot, upper layer and a cooler,
modeling. A complete discussion of the lower layer (Fig. 1). The model calculations
Fig. 1 Two-layer model

with no exchange between
layers except the plume
Hot upper layer
Tg
provide estimates of key conditions for each of some models support simulated heating, venting,
the layers as a function of time. Zone modeling and air conditioning (HVAC) duct work, in some
has proved to be a practical method for providing cases supporting actual interconnections between
estimates of fire effects in enclosures. rooms while in others only approximating overall
The fundamentals of zone modeling require supply and extraction without specifying inter-
that a space, room, or compartment be broken up change between rooms.
into various zones or control volumes to account Some models attempt to overcome zonal
for mass and energy driven by the fire. In addi- limitations with specialized submodels in
tion to the room or compartment of origin, some addition to the zonal approach. One example is
models provide multiroom capability. In these the implementation of detection activation
cases, in addition to accounting for control algorithms and the tracking of detection devices,
volumes in the room of origin for heat and such as fusible links, to predict activation. Other
mass transfer as affected by the fire plume, a submodels include the switching of the employed
multiroom model also accounts for flows out of governing equations based on the burning
vents from one room into another or to the regime, such as fuel versus ventilation limited.
“atmosphere.” In the room being exposed to This is an example of a conversion from a
fire conditions, where the mass and energy now two-zone to a single-zone approach.
flow from the compartment of fire origin, the Some models add a variety of combustion and
vent plume adds mass and energy to a new fire-related submodels. Pyrolysis and combus-
control volume in this compartment being tion products may be tracked in addition to oxy-
exposed. gen to determine whether combustion can occur
Heat transfer within zone models focuses on given the compartment conditions. If combus-
buoyancy-driven flows. In most cases, this flow tion is not possible due to lack of oxygen, for
calculation is implicit as the hot gases produced example, then these combustion products may
by the fire are assumed to be instantaneously ignite if they retain sufficient heat and progress
transported from the fire to the upper layer. Intrin- to another compartment where the requisite oxy-
sically, this implies that transport time details are gen is available. Similarly, models sometimes
lost in zone models. Some models also calculate calculate radiative and convective heat transfer
the convective heat transfer from the hot layer to along with ignition temperatures of specific
the surrounding compartment boundaries whereas objects to predict involvement of additional
others treat this as a lumped-mass loss coefficient, fuel packages.
demonstrating how treatment of heat loss from the
hot gases may be handled through multiple
approaches. Overview of Selected Models
Vent flow handling in zone models often only
supports vents that carry horizontal flow and thus Although the state of the art in understanding fire
do not model holes in floors and ceilings. This processes has not yet developed the ultimate
restriction does not usually cause difficulty for model, a number of available computer models
many commonly occurring fire scenarios. Very provide reasonable estimates of selected fire
early models often did not support vent flows at effects. The reader is referred to Friedman [7]
all. Vents were implicitly handled in the lower and Olenick and Carpenter [8] for a summary of
layer mathematically so that no pressure buildup available computer fire models and their general
occurred within the modeled space, which would capabilities. The following brief summaries will
have violated a variety of the basic assumptions. focus on a representative selection of zone
These early models were generally only applica- models rather than providing an exhaustive
ble during the early phases of a fire or other review of the changing state of the art in avail-
limited situations. In addition to natural vents, able computer models.
www.ebook777.com
ASET COMPF2
ASET (Available Safe Egress Time) is a program COMPF2 is a computer program for calculating
for calculating the temperature and position of the the characteristics of a postflashover fire in a
hot smoke layer in a single room with closed single building compartment, based on fire-
doors and windows [6]. ASET can be used to induced ventilation through a single door or win-
determine the time to the onset of hazardous dow. It is intended both for performing design
conditions for both people and property. The calculations and for the analysis of experimental
required program inputs are the heat loss burn data. Wood, thermoplastics, and liquid fuels
fractions, the height of the fuel above the floor, can be evaluated. A comprehensive output
criteria for hazard and detection, the room ceiling format is provided that gives gas temperatures,
height, the room floor area, a heat release rate, and heat-flow terms, and flow variables. The docu-
a species generation rate of the fire (optional). mentation includes input instructions, sample
The program outputs are the temperature, thick- problems, and a listing of the program. The
ness, and (optional) species concentration of the program was written in FORTRAN by
hot smoke layer as a function of time and the time Babrauskas [11].
to hazard and detection. ASET, written in FOR-
TRAN by Cooper and Stroup [9], can examine
multiple cases in a single run. ASET-B is a com- CONTAM
pact version of ASET written by Walton [6] that
was designed to run on personal computers. CONTAM is a multizone (nodal) indoor air qual-
Species concentrations and time to hazard and ity and ventilation analysis computer program
detection calculated by ASET are not calculated designed to predict: (1) airflows in building
in the compact ASET-B version. systems driven by mechanical means, wind
pressures acting on the exterior of the building,
and buoyancy effects induced by the indoor and
COMPBRN III outdoor air temperature difference; (2) the dis-
persal of airborne contaminants transported by
COMPBRN III is primarily used in conjunction airflows; and (3) the exposure of occupants to
with probabilistic analysis for the assessment of airborne contaminants. Unlike other fire models,
risk in the nuclear power industry. The model is zones in CONTAM are typically building areas
based on the assumption of a relatively small fire connected by airflow paths. CONTAM was
in a large space, or a fire involving large fuel developed principally to predict nonfire related
loads during the early preflashover fire growth conditions within a building and does not have a
period. The model’s strengths are (1) emphasis fire routine; however, the user may specify the
on the thermal response of elements within the temperature in a zone. CONTAM can be useful
enclosure to a fire within the enclosure, and for evaluating smoke management methods
(2) model simplicity. The temperature profile including compartmentation, dilution, pressuri-
within each element is computed, and an element zation, airflow, and buoyancy. CONTAM
is considered ignited or damaged when its sur- consists of two components, a graphical user
face temperature exceeds the user-specified igni- interface (GUI) and a simulation engine. Once a
tion or damage temperature. The model outputs building representation is developed with the
include the total heat release rate of the fire, the GUI, it is passed to the simulation engine to
temperature and depth of the hot gas layer, the calculate zone pressures, airflow rates, and con-
mass burning rate for individual fuel elements, taminant concentrations. The interface also
the surface temperatures, and the heat flux at provides a means to graphically review results.
user-specified locations. COMPBRN III was CONTAM was written in C by Walton and
written by Ho et al. [10] Dols [12].
CSTBZ1 BRANZFIRE
CSTBZ1 is a computer program for BRANZFIRE is a two-zone fire model for

postflashover fires, based on similar basic predicting the fire environment in an enclosure
assumptions as those of COMPF2. The equations resulting from a room-corner fire involving walls
of mass and energy conservation are written and ceilings. The zone fire model uses conserva-
without neglecting the fuel source term, and sev- tion equations based on those found in CFAST.
eral vertical openings in a room can be consid- BRANZFIRE predicts ignition, flame spread,
ered. The equation of heat diffusion into the and the resultant heat released by the wall and
walls is solved either by a classical explicit finite ceiling lining material subjected to a burner fire.
difference method, or by a new modal approach The model considers upward flame spread on the
that offers the capability of storing in a file a few walls and beneath the ceiling, lateral flame
numbers characterizing a given wall, leading to a spread on the walls, and downward flame spread
rapid calculation of the superficial wall from the ceiling jet. Wall and ceiling flame
temperatures. A sophisticated numerical algo- spread properties are computed from cone calo-
rithm was used to solve the equations through rimeter data. Program outputs include layer
uncoupling. The program was written by Curtat height, species concentrations, gas temperatures,
and Bodart [13]. visibility, wall temperatures, and heat release
rate. BRANZFIRE was written in Visual Basic
by Wade [18–20].
CFAST
CFAST (Consolidated model of Fire growth And JET

Smoke Transport) is an upgrade of the FAST
program [14] and incorporates numerical solution JET is a two-zone single compartment model
techniques originally implemented in the CCFM where the compartment is enclosed by a combi-
program [15]. CFAST is a multiroom model that nation of draft curtains and walls. The ceiling
predicts conditions within a structure resulting may contain fusible links and vents where the
from a user-specified fire. CFAST Version 6 can vents operate in response to the fusing of the
accommodate up to 30 compartments with multi- links. The ceiling vents remove hot, upper-layer
ple openings between compartments and to the gas from the compartment. The fire is
outside. The required program inputs are the characterized by a time-dependent heat release
geometry data describing the compartments and rate, a time-dependent radiative fraction, and
connections; the thermophysical properties of the either a constant or variable fire diameter,
ceiling, walls, and floors; the fire as a rate of mass which is determined using a heat release rate
loss; and the generation rates of the products of per unit area for the burning material. Inputs
combustion. The program’s outputs are tempera- also include the thermal properties of the ceiling.
ture, thickness, and species concentrations in the Program outputs include the ceiling jet tempera-
hot, upper layer and the cooler, lower layer in ture and velocity, link temperature, and activa-
each compartment. Also given are the surface tion times. JET was written in FORTRAN by
temperatures and heat transfer and mass flow Davis [21].
rates. CFAST also includes very limited
mechanical ventilation capabilities, a ceiling
jet algorithm, capability of multiple fires, heat FIRST
transfer to targets, detection and suppression
systems, and a flame spread model. The CFAST FIRST is the direct descendant of the Harvard V5
model was developed and is maintained by NIST program developed by Emmons and Mitler [4].
[16, 17]. The program predicts the development of a fire
www.ebook777.com
and the resulting conditions within a room response of heat-actuated detection devices,
given a user-specified fire or user-specified sprinklers, and smoke detectors; oxygen, carbon
ignition. It predicts the heating and possible monoxide, and carbon dioxide concentrations in
ignition of up to three targets. The required the smoke; and the effects of available oxygen on
program inputs are the geometrical data combustion. FPETOOL was written in BASIC
describing the rooms and openings and the by Nelson [24].
thermophysical properties of the ceiling,
walls, burning fuel, and targets. The generation
rate of soot must be specified, and the genera- FSSIM
tion rates of other species may be specified.
The fire may be entered either as a mass loss FSSIM [25] is a single-zone model originally
rate or in terms of fundamental properties of designed for fire hazard analysis of ships. It
the fuel. Among the program outputs are tem- includes most of the features of the popular
perature, thickness, and species concentrations two-zone models but within a single zone fire
in the hot, upper layer and the cooler, lower environment description. It includes HVAC sys-
layer in each compartment. Also given are tem model components, stack effect, prediction
surface temperatures and heat transfer and of compartment temperatures, and smoke and
mass flow rates. The FIRST program was writ- gas concentrations, as well as compartment to
ten in FORTRAN by Mitler and Rockett [22]. compartment fire spread, detection, and suppres-
sion. It has been applied to ships and buildings
with several thousand compartments.
FIRE SIMULATOR Calculations are slower than CONTAM but still
much faster than two-zone models and CFD
FPETOOL is the descendant of the FIREFORM models.
program [23]. It contains a computerized selec-
tion of relatively simple engineering equations
and models useful in estimating the potential fire LAVENT
hazard in buildings. The calculations in
FPETOOL are based on established engineering LAVENT (Link-Actuated VENT) is a two-zone
relationships. The FPETOOL package addresses model developed to simulate the environment
problems related to fire development in buildings and the response of sprinkler elements in com-
and the resulting conditions and response of fire partment fires with draft curtains and fusible-link
protection systems. The subjects covered include actuated ceiling vents. The zone model used to
smoke filling in a room, sprinkler/detector acti- calculate the heating of the fusible links includes
vation, smoke flow through (small) openings, the effects of the ceiling jet and the upper layer of
temperatures and pressures developed by fires, hot gases beneath the ceiling. The required pro-
flashover and fire severity predictions, fire prop- gram inputs are the geometrical data describing
agation (in special cases), and simple egress esti- the compartment, the thermophysical properties
mation. The largest element in FPETOOL is a of the ceiling, fire elevation, the time-dependent
zone fire model called FIRE SIMULATOR. heat release rate of the fire, the fire diameter or
FIRE SIMULATOR is designed to estimate the heat release rate per unit area of the fire, the
conditions in both pre- and postflashover enclo- ceiling vent area, the fusible-link response time
sure fires. The inputs include the geometry and index (RTI) and activation temperature, the
material of the enclosure, a description of the fusible-link positions along the ceiling, the link
initiating fire, and the parameters for sprinklers assignment to each vent, and the ambient tem-
and detectors being tracked. The outputs include perature. A maximum of five ceiling vents and
the temperature and volume of the hot smoke 10 fusible links are permitted in the compart-
layer; the flow of smoke from openings; the ment. The program outputs are the temperature
and height of the hot layer, the temperature of forced ventilation for ceiling and floor vents,
each link, the ceiling jet temperature and velocity and an interface to a finite difference computer
at each link, the radial temperature distribution model for the calculation of boundary surface
along the interior surface of the ceiling, the acti- isotherms and hot spots. The WPI/FIRE pro-
vation tie of each link, and the area opened. gram was written in FORTRAN by Satterfield
LAVENT was written in FORTRAN by and Barnett [29], and additions to the program
Cooper [26]. continue to be developed by graduate students at
the Center for Fire Safety Studies, Worcester
Polytechnic Institute.
MAGIC
MAGIC is a two-zone fire model for simulations BRI2

that predict the fire environment as a function of
time resulting from a fire in single- or multicom- BRI2 is a two-layer, multi-compartment model to
partment geometries, based on a combination of predict smoke spread through large buildings. In
macroscopic conservation equation and empiri- addition to conventional zone equations, it
cal correlations for specific phenomena [27]. includes radiative and convective heat transfer
The required program inputs include the between layers and wall enclosures and species
compartment geometry; natural and mechanical yield prediction using equivalence ratio. For
ventilation; parameters describing the fire practical purposes, users can input the time-
characteristics such as location, fuel type, foot- based events such as smoke vent activation,
print area, and stochiometric fuel-oxygen ratio; door opening by evacuees etc. The BRI2 program
thermophysical properties and location; and any was written in FORTRAN by Tanaka et al. [30].
inputs for sprinklers or detectors. The extensive
program outputs are customized by the user, but
typically include environmental conditions, heat Selection, Validation, and Application
transfer–related outputs, oxygen effects, and flow of Zone Models
velocities [28]. MAGIC is being extended to
include nonrectangular rooms, convex and slop- Selection
ing ceilings, rooms cluttered with objects, spread
of fire through ventilation ducts, and extinction. The appropriate selection of a zone fire model for
Although the program can be used for general a particular application requires significant atten-
fire modeling, it was originally intended for tion since no existing zone fire model is best for
nuclear power plant applications. The program all applications. Although most zone fire models
is developed and maintained by Electricité de are based on the same fundamental principles,
France (EdF). available models can vary significantly in
features and capabilities. Appropriate selection
of a zone fire model requires sufficient under-
WPI/FIRE standing of the assumptions and limitations for
any particular model. Frequently, experienced
WPI/FIRE is a direct descendant of the Harvard V users will employ more than one model to evalu-
[4] and FIRST [22] programs. It includes all of the ate a particular situation. If several models pro-
features of the Harvard program version 5.3 and vide similar results this can increase (although
many of the features of the FIRST program. not generally guaranteed) confidence in the
WPI/FIRE also includes the following additional results. A significant issue in selecting a zone
features: improved input routine, momentum- fire model is model validation, a process by
driven flows through ceiling vents, two different which to assess the predictive capabilities and
ceiling jet models for use in detector activation, accuracy of a fire model.
www.ebook777.com
The predictive capability of any model depends predictions, and actual calculations performed
on the underlying scientific knowledge and under- by users. This analysis is ideally performed by
standing of the phenomenon being modeled and the model developer but is sometimes undertaken
the translation of that knowledge and under- by other individuals or organizations, such as the
standing into a mathematical model. This knowl- Society of Fire Protection Engineers analysis of the
edge may be based upon well-established fire thermal detector actuation model DETACT [32].
science, secure empirical correlations, or, if A critical use of the methodology in ASTM 1355
these are not available, then the best engineering is to identify sensitive variables associated with
judgment that can be made [1]. Due to its com- the model and to consider those sensitive
plexity, there remain substantial voids and variables that may have a significant effect in a
shortcomings in the scientific understanding of fire safety design application. They therefore
combustion and fire and associated processes. allow an analyst to consider the commensurate
Obviously, the current level of knowledge uncertainty in the predicted values relating
supports the development of mathematical to the input uncertainty and further allow
models with sufficient predictive capabilities. It designers to determine safety factors on a more
is not necessary to understand a phenomenon in considered basis.
every respect in a pure scientific context to
exploit the current level of understanding for
design and practice purposes. In fact, zone fire Application
models take advantage of “imperfect” knowl-
edge and simplifying assumptions to yield prac- Zone models have enjoyed wide application and
tical and useful results. general acceptance due to their relatively
Despite imperfect knowledge and simplifying simplified approach to the modeling problem.
assumptions, properly applied zone models have Zone fire models have been used or
been shown to be a source of good engineering commissioned by engineers and architects, build-
approximations of fire development within ing officials, the fire service, fire investigators,
enclosures. Thus, comparison of zone fire building and fire code developers, materials and
modeling results with experimental data is valu- systems manufacturers, fire researchers, and
able for determining the applicability of a model educators. Application and use of zone fire
to a particular situation in the model selection modeling include performance-based fire safety
process. As with any fire model, published design, postfire reconstruction, and fire risk
comparisons of zone fire model results with assessment.
experimental data are somewhat limited when With respect to performance-based fire safety
compared to the range of possible applications. design, some fire protection engineers have been
able to develop alternative equivalent approaches
to meeting the prescriptive requirements of codes
Validation through the use of zone fire models. Other
applications have utilized zone fire models in
With respect to the validation of fire models, fire hazard analyses to develop engineered fire
ASTM 1355, “Standard Guide for Evaluating mitigation strategies for unique situations in
the Predictive Capability of Deterministic Fire which a prescriptive approach would not have
Models,” [31] sets forth a methodology for met specific fire loss criteria. The problem of
analyzing a computer fire model, examining postfire reconstruction lends itself readily to the
two broad conceptual categories. The first cate- use of fire models. Several fire reconstruction
gory relates to the foundation of the model—its case studies of fire incidents using computer
theoretical basis, mathematical equations, and models have been published [33–36].
numerical methods. The second category relates Although zone models have experienced the
to sensitivity, comparison between data and lion’s share of practical applications, the use of
more complex computer fire models has and will and validation of the key physics, and evaluating
continue to increase. This evolution will not ren- user effects and uncertainties. Good documenta-
der zone models obsolete in everyday practice tion of the modeler’s activities in this regard and
due to their significant advantages in solving a using the Guide’s framework, while helping
subset of engineering problems. future reviewers for users of the work, will also
In 2011, the Society of Fire Protection benefit the modeler in providing completeness
Engineers published its engineering guide titled and thoroughness of the model activities.
“Guidelines for Substantiating a Fire Model for a
Given Application” [37]. Similar to the SFPE’s
guide on performance-based design, the docu- References
ment recommends a workflow for the modeler.
This workflow is captured in a flowchart shown 1. H.W. Emmons, “The Needed Fire Science,”
in the document’s introduction. As is true with Proceedings of the 1st International Symposium on
Fire Safety Science (C.E. Grant and P.J. Pagni, eds.),
the scientific method, generally, the first step is to International Association for Fire Safety Science,
define the problem. This is, in fact, a critical step October 7–11, 1985, Gaithersburg, MD, pp. 33–54
to successful computer fire modeling. (1986).
In the absence of a well-defined problem and 2. J. Quintiere, “Growth of Fires in Building
Compartments,” ASTM STP 614, American Society
the applicable parameters that apply to that prob- for Testing and Materials, Philadelphia (1977).
lem, the next step of selecting a candidate model 3. R. Pape, T.E. Waterman, and T.V. Eichler, “Develop-
is unlikely to be properly performed. Running the ment of a Fire in a Room from Ignition to Full Room
wrong model may leave the modeler to inaccu- Involvement—RFIRES,” NBS-GCR-81-301, National
Bureau of Standards, Washington, DC (1981).
rate, unreliable, or inapplicable results. However, 4. H.E. Mitler, and H.W. Emmons, “Documentation for
one of the benefits of using zone models on CFC V: The Fifth Harvard Computer Fire Code,” NBS
modern computers is to generally be able to run GCR 81-344 (Home Fire Project Technical Report
multiple models in a very short period of time. 45), National Bureau of Standards, Gaithersburg,
MD (1981).
This can also be done to get preliminary results to 5. H.E. Mitler, “The Harvard Fire Model,” Fire Safety
better develop the parameters and boundary Journal, 9, pp. 7–16 (1985).
conditions for use in a fire model requiring 6. W.D. Walton, “ASET-B: A Room Fire Program for
much greater time to run, such as a field or Personal Computers,” NBSIR 85-3144, National
Bureau of Standards, Washington, DC (1985).
CFD model. 7. R. Friedman, “An International Survey of Computer
The guide goes further to discuss uncertainty Models for Fire and Smoke,” Journal of Fire Protec-
and user effects related to model results. Some tion Engineering, 4, 3, pp. 83–92 (1992).
key studies are discussed as representative of the 8. S.M. Olenick and D.J. Carpenter, “Updated Interna-
tional Survey of Computer Models for Fire and
type of analysis to be considered by the modeler Smoke,” Journal of Fire Protection Engineering,
during the work. Some of this work comes out of 13, 2, pp. 87–110 (2003).
verification and validation performed by 9. L.Y. Cooper and D.W. Stroup, “ASET: A Computer
modelers as well as independent studies consid- Program for Calculating Available Safe Egress
Time,” Fire Safety Journal, 9, pp. 29–45 (1985).
ering model accuracy. Beyond the specific stud- 10. V. Ho, N. Siu, and G. Apostolakis, “COMPBRN III—
ies cited, the guide provides recommendations A Fire Hazard Model for Risk Analysis,” Fire Safety
for addressing the uncertainty and user effects. Journal, 13, 2–3, pp. 137–154 (1988).
For example, sensitivity and bounding analyses, 11. V. Babrauskas, “COMPF2—A Program for Calculating
Post-Flashover Fire Temperatures,” NBS TN 991,
among others, are methods for examining user National Bureau of Standards, Washington, DC (1979).
effects and their impact on a particular analysis. 12. G.N. Walton and W.S. Dols, “CONTAM 2.4 User
Finally, the Guide suggests documenting the Guide and Program Documentation,” NISTIR 7251,
findings from this work for use by reviewers, National Institute of Standards and Technology,
Gaithersburg, MD (2005).
AHJs, and future users of the model results. It 13. M.R. Curtat and X.E. Bodart, “Simple and Not So
addresses the specific concepts of problem defi- Simple Models for Compartment Fires,” Proceedings
nition, candidate model evaluation, verification of the 1st International Symposium on Fire Safety
www.ebook777.com
Science (C.E. Grant and P.J. Pagni, eds.), Interna- 26. Y.L. Cooper, “Estimating the Environment and the
tional Association for Fire Safety Science, October Response of Sprinkler Links in Compartment Fires
7–11, 1985, Gaithersburg, MD, pp. 637–646 (1986). with Draft Curtains and Fusible-Link-Actuated Ceil-
14. W.W. Jones, “A Multicompartment Model for the ing Vents Theory,” Fire Safety Journal, 16, pp.
Spread of Fire, Smoke and Toxic Gases,” Fire Safety 137–163 (1990).
Journal, 9, pp. 55–79 (1985). 27. L. Gay, C. Epiard, and B. Gautier, “MAGIC Software
15. G.P. Forney and L.Y. Cooper, “The Consolidated Version 4.1.1: Mathematical Model,” EdF HI82/04/
Compartment Fire Model (CCFM) Computer 024/B, Electricité de France, Paris, France (2005).
Application. VENTS, Part I, Physical Basis,” NISTIR 28. L. Gay, “User Guide of the MAGIC Software V4.1.1”
4342, National Institute of Standards and Technology, EdF HI82/04/23/A, Electricité de France, Paris,
Gaithersburg, MD (1990). France (2005).
16. R.D. Peacock, W.W. Jones, P.A. Reneke, and 29. D.B. Satterfield and J.R. Barnett, “User’s Guide to
G.P. Forney, “CFAST—Consolidated Model of Fire WPI-FIRE Version 2 (WPI-2)—A Compartment
Growth and Smoke Transport (Version 6): Technical Fire Model,” Center for Fire Safety Studies, Worces-
Reference Guide,” SP-1026, National Institute ter Polytechnic Institute, Worcester, MA (1990).
of Standards and Technology, Gaithersburg, MD 30. T. Tanaka, S. Yamada, “BRI2002: Two layer Zone
(2005). Smoke Transport Model”, Fire Science and Technol-
17. R.D. Peacock, W.W. Jones, P.A. Reneke, and ogy, 23(1), special issue, Tokyo University of Sci-
G.P. Forney, “CFAST—Consolidated Model of Fire ence, 2004.
Growth and Smoke Transport (Version 6): User’s 31. “Standard Guide for Evaluating the Predictive Capa-
Guide,” SP-1041, National Institute of Standards and bility of Deterministic Fire Models,” ASTM E1355-
Technology, Gaithersburg, MD (2005). 97, American Society for Testing and Materials, West
18. C.A. Wade, D. LeBlanc, J. Ierardi, and J.R. Barnett, Conshohocken, PA (1997).
“A Room-Corner Fire Growth and Zone Model for 32. “SFPE Engineering Guide to the Evaluation of the
Lining Materials,” Second International Conference Computer Model DETACT-QS,” Society of Fire Pro-
on Fire Research and Engineering (ICFRE2), National tection Engineers, Bethesda, MD (2002).
Institute of Standards and Technology and Society of 33. H.E. Nelson, “An Engineering Analysis of Fire
Fire Protection Engineers, Gaithersburg, MD, Development in the Hospice of Southern Michigan,
pp. 106–117 (1998). December 15, 1985,” Proceedings of the 2nd Interna-
19. C. Wade and J. Barnett, “A Room-Corner Fire Model tional Symposium on Fire Safety Science, Hemisphere
Including Fire Growth on Linings and Enclosure Publishing, New York (1989).
Smoke-Filling,” Journal of Fire Protection Engineer- 34. R.S. Levine and H.E. Nelson, “Full-Scale Simulation
ing, 8, 4, pp. 183–193 (1997). of a Fatal Fire and Comparison of Results with Two
20. C. Wade, “Branzfire-Engineering Software for Multiroom Models,” NISTIR 90-4268, National Insti-
Evaluating Hazard of Room Lining Materials,” in tute of Standards and Technology, Gaithersburg, MD
Conference Proceedings on the Eighth International (1990).
Interflam Conference, Interscience Communications 35. N. Alvares, “Defining Fire and Smoke Spread
Ltd., London, pp. 1147–1152 (1999). Dynamics in the DuPont Plaza Fire of 31 December
21. W.D. Davis, “The Zone Fire Model Jet: A Model for 1986,” Proceedings of the International Conference
the Prediction of Detector Activation and Gas Tem- on Fire Research and Engineering, Society of Fire
perature in the Presence of a Smoke Layer,” NISTIR Protection Engineers, September 10–15, 1995, Bos-
6324, National Institute of Standards and Technology, ton, MA (1995).
Gaithersburg, MD (1999). 36. R.W. Bukowski, and R.C. Spetzler, “Analysis of the
22. H.E. Mitler and J.A. Rockett, “User’s Guide to Happy Land Social Club Fire with HAZARD I,”
FIRST, A Comprehensive Single-Room Fire Journal of Fire Protection Engineering, 4, 4, (1992).
Model,” CIB W14/88/22, National Bureau of 37. Guidelines for Substantiating a Fire Model for a
Standards, Gaithersburg, MD (1987). Given Application, Society of Fire Protection
23. H.E. Nelson, “FIREFORM—A Computerized Collec- Engineers, Bethesda, MD, 2011
tion of Convenient Fire Safety Computations,” NBSIR
86-3308, National Bureau of Standards, Gaithersburg,
MD (1986). William D. Walton retired from the Building and Fire
24. H.E. Nelson, “FPETOOL: Fire Protection Engineer- Research Laboratory, National Institute of Standards and
ing Tools for Hazard Estimation,” NISTIR 4380, Technology.
National Institute of Standards and Technology,
Gaithersburg, MD (1990). Douglas J. Carpenter is vice president and principal
25. J.E. Floyd, S.P. Hunt, F.W. Williams, and P.A. Tatem, engineer with Combustion Science & Engineering, Inc.,
“A Network Fire Model for the Simulation of Fire in Columbia, Maryland.
Growth and Smoke Spread in Multiple Compartments
with Complex Ventilation,” Journal of Fire Protec- Christopher B. Wood is a member of Fire Link, LLC, in
tion Engineering, 15, 3, pp. 199–229 (2005). Tewksbury, Massachusetts.
Modeling Fires Using Computational

Fluid Dynamics (CFD) 32
Kevin McGrattan and Stewart Miles
Introduction CFD provides the potential to study these

complicated problems that are only partially
It was in the early 1920s that Lewis Richardson amenable to reduced-scale physical modeling
first demonstrated the feasibility of solving, because of the large number of non-dimensional
using numerical methods, the governing groups that need to be preserved to simulate
equations of fluid flow [1] for the purpose of full-scale behavior. Furthermore, CFD has
weather prediction. It was not for another emerged as an important tool because the
50 years that what is now known as computa- assumptions employed in zone models limit
tional fluid dynamics (CFD) emerged as a gen- their range of application to relatively simple
eral analysis tool for the full breadth of fluid flow fire scenarios that can be described in terms of a
problems including those associated with fire. small number of idealized components (e.g.,
In contrast to zone models, the techniques unperturbed fire plume, unconfined ceiling jet,
of CFD evolved outside the fire community uniform hot gas layer).
and were subsequently imported into it. The The starting point for CFD models is the set of
same basic CFD technology is being developed, partial differential equations that assert conser-
applied, verified, and validated for a wide vation of mass, momentum, and energy within
range of applications. Problems and successes the fire and throughout the space surrounding
demonstrated elsewhere can often be exploited it. These equations are solved numerically to
in the fire context, although there are many issues yield time-varying predictions of temperature,
that are unique to fire that can only be the respon- gas velocity, gas species concentrations, and so
sibility of the fire community. The same tools forth, on a three dimensional mesh of control
that are used to study, for example, heat transfer volumes that spans the geometry being modeled.
in an internal combustion engine can also be Unlike two zone models, CFD models enforce
used to evaluate indoor air movement, early the conservation laws in each of the thousands or
detection of smoke, and the dispersion of com- millions of relatively small control volumes.
bustion products within the atmospheric bound- However, the exact solution of the governing
ary layer. equations, resolving fully the length and time
scales that occur in the turbulent flows associated
with fire, is still beyond the capabilities of even
K. McGrattan (*)
the largest computers currently available. To
Fire Research Division, National Institute of Standards
and Technology Gaithersburg, Gaithersburg, MD, USA capture the details of the chemical processes of
a fire would require spatial resolution of less than
S. Miles
International Fire Consultants Ltd., 1 millimeter. As a consequence, it is necessary to
Buckinghamshire, UK modify the governing equations to model the

www.ebook777.com
32 Modeling Fires Using Computational Fluid Dynamics (CFD) 1035
unresolvable turbulence. Two main approaches A recent review by Olenick and Carpenter [10]
are currently employed in CFD simulations of lists about a dozen CFD models developed spe-
fire: large eddy simulation (LES) and Reynolds- cifically for fire. Some of these models were
averaged Navier-Stokes (RANS) equations, developed for specific fire scenarios or phenom-
which are described later in this chapter. ena. Others were developed to handle a variety of
In addition to the uncertainties associated with fire scenarios, including JASMINE from the Fire
the modeling of turbulent flow, others are Research Station (UK), the Fire Dynamics Sim-
introduced by the description of combustion ulator (FDS) from NIST (US) and VTT
chemistry; radiation; and mass, momentum, and (Finland), SMARTFIRE from the University of
heat transfer at solid boundaries. Further com- Greenwich (UK), and SOFIE, the product of a
plexity is introduced in the numerical solution of European consortium. A more recent develop-
the equation set in which the choice of numerical ment is the FireFOAM model from FM Global,
schemes and the resolution of the numerical which is based on the open source CFD code
mesh strongly influences the quality of the CFD OpenFOAM. There are also general-purpose
solution. An appreciation of all these issues is CFD models that have been used for fire
important in the successful exploitation of CFD simulations. These computer programs contain
to solve fire problems. It should be recognized tens to hundreds of thousands of lines of
that this topic is rapidly evolving and that this instructions, along with manuals that contain
chapter can only represent a “snapshot” in time. hundreds of pages of documentation of the devel-
Whereas a thousand mesh points constituted a opment, algorithms, and validation of the
detailed CFD solution in the early 1980s, models.
simulations using millions of mesh points are This chapter provides an introduction to the
now routine. This number can be expected to theory and practice of computational fluid
increase further, especially as parallel processing dynamics as applied to the study of fire.
becomes more widespread. However, although Although the mathematical framework for the
the modeling of smoke transport may be consid- subject is more than 100 years old, it is only in
ered reasonably mature, there remains consider- the past 20 years that computers have become
able research and development still to be done fast enough to make the models practical.
with some of the more complex issues related to
the underlying combustion and fire science (e.g.,
flame spread, oxygen vitiation, soot formation, Governing Equations
and water suppression). Here the challenges
remain considerable and will not be satisfactorily This section presents the conservation equations
solved for some time yet. of mass, momentum, and energy that constitute
This chapter does not provide a comprehen- the core of any CFD model. The derivation of
sive description of CFD. There are already these equations can be found in any textbook on
numerous introductory textbooks on the sub- the subject and will not be included here. Instead,
ject [2, 3, 4, 5, 6]. Depending on the background a discussion of the various techniques used to
of the author(s), these books tend to emphasize make these equations tractable for numerical
techniques developed for particular fields of solution will follow. The most important of
science and engineering, such as aerospace, these is the turbulence model, a subject of con-
meteorology, or combustion. Discussion of siderable study and debate.
issues related to the modeling of fire can be
found in review articles by Cox [7, 8] and
Novozhilov [9]. Finally, a comprehensive Conservation Equations
description of the specific algorithms used by a
particular CFD model can only be found within The conservation equations for mass, momen-
the manuals that accompany the software. tum, and energy for a mixture of gases are briefly
1036 K. McGrattan and S. Miles
described here. A particularly useful reference Conservation of Energy

for a more detailed description of the equations, The conservation equation for energy is
the notation used, and the various approxi-
@ðρhÞ Dp
mations employed is Anderson et al. [2]. ¼ r ðρhuÞ ¼ þ q_ 000 r q þ ε
@t Dt
Conservation of Mass (32.4)
The conservation equation for mass is
As in the mass conservation equation, the sensi-
@ρ ble enthalpy, h, at a given point changes
þ r ρu ¼ 0 (32.1) according to the net energy flux across the
@t
boundary of a small control volume surrounding
This equation is nothing more than a statement the point. Now, however, there are additional
that mass is neither created nor destroyed. In source terms on the right-hand side of the equa-
words, the change in the density, ρ, at a given tion related to the pressure, combustion heat
point in the flow field is equal to the net mass release rate, radiation and conduction, and
flux, ρu, across the boundary of a small control kinetic energy dissipation, respectively. For fire
volume surrounding the point. For fire applications, the contributions from the pressure
simulations, it is usually necessary to account term and the dissipation term are negligible,
explicitly for various individual gaseous species, except in situations where the compartment of
for example, fuel and oxygen. Thus, the mass interest is sealed and the pressure rises
conservation equation is often written as a set substantially.
of transport equations for the mass fractions of Equations 32.1 through 32.4 constitute a set of
the individual gaseous species, Yα: partial differential equations for the density
(or component mass fractions), velocity, pres-
@ðρYα Þ sure, and sensible enthalpy of the fluid. The
þ r ðρYα uÞ ¼ r ðρDα rYα Þ þ m_ 000
α
@t momentum equation is actually three equations
(32.2) for the three components of velocity: u, v, and
w making five equations for six unknowns. To
When all of the species equations are added close the system, an equation of state is needed to
together, the mass diffusion and production relate the pressure, p, and sensible enthalpy, h.
terms on the right-hand side sum to zero, leaving The latter is a function of the specific heat and
the original mass conservation equation. temperature of the fluid:
Z T
Conservation of Momentum h¼ cp dT (32.5)
The conservation equation for momentum is T0
@ðρuÞ For most fire applications, it is sufficient to

þ r ðρuuÞ ¼ rp þ f þ r τ (32.3)
@t assume a perfect gas:
The momentum equation is written in compact ρ<T 1

p¼ ; W¼P (32.6)
notation here to emphasize that it is essentially W ðYα =Wα Þ
Newton’s Second Law of Motion, or simply
Force ¼ Mass Acceleration. The forces that where < is the universal gas constant and W is the
drive the fluid consist of the pressure gradient, average molecular weight of the gas mixture.
∇p, friction (in the form of the viscous stress
tensor, τ), and external force terms, f, such as Additional Assumptions
buoyancy. A complete expansion of the various The foregoing equations form the basis for a
terms can be found in the appendix. wide variety of engineering applications but not
www.ebook777.com
without the further application of simplifying simulation (DNS), Reynolds-averaged Navier-

assumptions unique to each field. The only Stokes (RANS), and large eddy simulation
assumptions made thus far are that the fluid is a (LES). DNS is really nothing more than a direct
perfect gas and that the viscous stress is linearly numerical solution of the governing equations.
dependent on the strain. A further approximation RANS and LES employ models of the unre-
can be made here, exploiting the fact that fire- solved subgrid dissipative processes. RANS
driven flows are substantially slower than the averages over relatively large spatial and tempo-
speed of sound. By assuming that the pressure ral scales than those that are characteristic of the
is constant (or at most a time-varying average) in given numerical grid or the fundamental fre-
the equation of state and the energy equation, the quency of the fire, whereas LES attempts to
number of unknowns is reduced from six to five, compute as much of the “resolvable” length and
as temperature can now be found from the den- time scales (i.e., the “large eddies”) as possible.
sity. More importantly, there is no longer a need
to account for pressure fluctuations that propa-
gate at the speed of sound, a fact that makes the Direct Numerical Simulation (DNS)
numerical solution of the equations more tracta-
ble because the time step size is limited only by Direct numerical simulation (DNS) means that
the flow speed, not the sound speed [11]. the governing equations are solved numerically
In the sections to follow, additional with no modifications to the form presented
assumptions will be imposed on the governing above, which implies that all of the relevant
equations to apply them to fire and other temporal and spatial scales are resolved directly
low-speed thermal processes. Most important is without invoking any models to represent the
the treatment of the diffusive and source terms diffusive terms, like the viscosity, thermal con-
that distinguishes one type of CFD model from ductivity, and material diffusivity. Because this
another. technique demands very fine spatial and temporal
resolution (less than 1 mm and 1 ms, respec-
tively), it is limited to small laminar flames and
Turbulence Modeling sometimes small turbulent jets. DNS is still not
practical for large-scale fire simulations, and
The governing conservation equations describe nothing more is included in this chapter on the
the transport of mass, momentum, and energy subject.
via convection and diffusion (material diffusivity,
viscosity, thermal conductivity). In large-scale fire
applications, convection is the primary mode of Reynolds-Averaged Navier-Stokes
transport of heat and combustion products, but (RANS)
the diffusive processes play a significant role in
the flames and in boundary layers near solid The approach to modeling turbulence in many
surfaces. Capturing the large-scale convective commercial and fire-specific CFD models is to
transport and the small-scale diffusive processes solve a statistically time-averaged form of the
in the same simulation requires too much compu- conservation equations, often referred to as the
tation, even for current-generation computers. Reynolds-averaged Navier-Stokes (RANS)1
Consequently, most practical fire simulations
employ models to describe subgrid-scale, or
“unresolvable,” phenomena. The most important 1
A density-weighted variant of Reynolds averaging
of these are the turbulence models. (Favre averaging) is often necessary for the treatment of
The most popular numerical techniques for regions where density fluctuations have a significant
effect. This variant is not usually necessary for smoke
approximating the governing equations are
movement studies but can be important close to the fire
roughly categorized according to their spatial source. This treatment is not presented here but can be
and temporal fidelity: direct numerical found in, for example, Cox [7].
equations after Osborne Reynolds who use an eddy viscosity turbulence model to
introduced the idea over a century ago. The close the set of equations. The basic idea is that
starting point in Reynolds averaging is to the unresolved, turbulent fluctuations can be
decompose the velocity components, enthalpy, modeled with diffusive terms that effectively
and species mass fractions, into a time-averaged represent the dissipation of turbulent energy:
component (denoted by an overbar) and a
fluctuating component (denoted by a prime): @ ui @ uj 2
ρu0i u0j ¼ μt þ δij ρk
@xj @xi 3
tÞ þ ϕ0 ðx; tÞ
ϕðx; tÞ ¼ ϕðx; (32.7) ( (32.10)
1 if i ¼ j
δij ¼
Notice that for transient flows the time-averaged 0 if i 6¼ j
component is still a function of time, representing
the evolution of the mean flow field. In typical @ h
ρu0j h0 ¼ λt (32.11)
fire simulations, the time scale associated with @xj
the time-averaged component is on the order of
several seconds, whereas that of the fluctuating Here μt is the turbulent (eddy) viscosity, k the
component is on the order of milliseconds. turbulent kinetic energy, and λt the turbulent
Substituting the decomposed primitive variables diffusivity, which is related to the eddy viscosity
into the conservation equations (Eqs. 32.1, 32.3, by the expression
and 32.4) and then applying the same time-
μt
averaging process to the entire system of λt ¼ (32.12)
equations yields a set of equations that is similar σt
in form to the original equations, with the mass
where σ t is known as the turbulent Prandtl or
conservation equation remaining unchanged:
Schmidt number, depending on whether the sca-
@ðρ
uÞ lar quantity is an energy or species variable. It is
þ r ðρ
uuÞ ¼ rp þ f þ r τ r ρu0 u0
@t usually assigned a constant value depending on
(32.8) the scalar variable in question.
For fire simulations, a two-equation eddy vis-

@ðρhÞ cosity model is commonly used. This model, in
uÞ ¼ Dp þ q_ 000 r q
þ r ðρh þ ε
@t Dt effect, allows the turbulence to be characterized by
r ρu0 h0 ð32:9Þ a velocity and a length scale that varies at each grid
cell in the computational domain. The two equa-
Notice that the Reynolds-averaging process has tion model that is employed in the majority of
introduced additional terms on the right-hand RANS applications in fire engineering is the k-ε
sides of the equations and increased the number model. Here two additional transport equations are
of unknowns. Thus, the system of equations is solved; one for the turbulent kinetic energy, k, and
no longer closed (i.e., there are more unknowns one for its rate of dissipation, ε. These quantities
than equations). The additional terms are referred form the turbulent viscosity in Eq. 32.10:
to as the Reynolds stresses and the turbulent
k2
scalar flux, respectively. The scalar, h, is the μt ¼ ρCμ (32.13)
ε
sensible enthalpy, and similar equations can be
derived for the other scalars (e.g., gas species The factor Cμ is an empirical constant. Details of
mass fractions). Note that the pressure and den- these equations are provided in CFD textbooks.
sity in the preceding equations refer to their time- Note that a modification of the basic k-ε model is
averaged components. often made in fire applications, in which buoy-
The majority of commercial and fire-specific ancy source terms for the k and ε equations are
CFD models that employ the RANS approach added. These terms ensure that the effects of
www.ebook777.com
Fig. 32.1 What is meant

by “resolvable.” At right is
an instantaneous map of the
flow vectors for an LES
simulation of a helium
plume. the inset shows the
smallest “eddy” that can be
supported by the numerical
grid, the spacing of which
is indicated by the distance
between the arrows
stable stratification in hot gas layers and the Eqs. 32.8 and 32.9. However, the standard k-ε
destabilizing thermal gradients in rising plumes model remains the most commonly used RANS
are adequately described. Details are discussed technique in fire applications.
by Cox [8].
The k-ε model has been used successfully for
many engineering applications. Its limitation in Large Eddy Simulation (LES)
treating, for example, swirling flows, flow sepa-
ration, and so on is due partly to the inherent The derivation of LES models is very similar to
assumption that the modeled turbulence is isotro- that of the RANS models, albeit with subtle
pic, that is, that the unresolved flow field has no differences in the interpretation of the decompo-
cohesive structure. There are a number of sition of the primitive variables. RANS
variants of the k-ε model, available in commer- emphasizes temporal averaging, whereas LES
cial CFD models, that claim to provide improved emphasizes spatial averaging, or filtering.
results for some flows (e.g., the RNG k-ε Regardless, the mechanics of the derivation are
model [12] and the k-ω model [13]). Research essentially the same, and the simplest of the LES
into nonlinear eddy viscosity turbulence models models makes the same “eddy viscosity” argu-
that attempt to incorporate directional effects is ment in replacing the unresolved convection
being undertaken within the CFD community terms with diffusion terms. The key difference
but, as far as the authors are aware, have not between the techniques lies in the magnitude of
been applied to fire problems. Another approach, the diffusive coefficient, the “eddy” viscosity.
applied mainly outside of fire, is to solve individ- With LES, one attempts to resolve the flow field
ual differential equations for Reynolds stresses as faithfully as possible on a given numerical
and turbulent scalar fluxes that appear in grid. In other words, flow structures, like swirling
eddies, can span just a few grid cells (Fig. 32.1). turbulent viscosity is calculated from finite dif-
To achieve this result, the eddy viscosity must be ference approximations of the gradients of the
small enough to avoid smoothing out these small, computed velocity components. The other diffu-
but resolvable, eddies, but large enough to ensure sive parameters, the thermal conductivity and
numerical stability and account for the dissipa- material diffusivity, are related to the eddy
tion of energy at sub-grid scales. viscosity through the expressions
To better understand the eddy viscosity
μt cp μt
model, the original Navier-Stokes equations are kt ¼ ; Dt ¼ (32.18)
examined. An evolution equation for the kinetic Prt ρ Sct
energy of the gas is formed by taking the dot
The “turbulent” Prandtl and Schmidt numbers
product of the momentum from Eq. 32.3 and
are assumed to be constant and of order unity.
the velocity vector u:
The appeal of LES is that no additional equations
Du Dðjuj2 =2Þ need be solved, as in the RANS k-ε model, and
ρ uρ ¼ E (32.14) that the dynamics of the fire are captured in the
Dt Dt
simulation, rendering a more realistic-looking
A term emerges on the right-hand side known as flow field than the time-averaged RANS model.
the dissipation rate (see appendix for details): The drawback of LES is that it demands that the
" # model user pay close attention to the choice of
@u 2 numerical grid because the accuracy of the tech-
E τ ru ¼ μ 2 þ (32.15)
@x nique can be degraded significantly if there are
not enough grid cells to describe the flow field
which appears as a source term in the energy adequately. Various perspectives on LES in gen-
conservation equation 32.4: eral can be found in Pope [15].
D
ðρhÞ ¼ þ E (32.16)
Dt Other Approaches
Extra terms have been hidden to make the point Detached eddy simulation (DES), formulated
more clearly, and no approximations have been originally by Spalart et al. [16], combines the
introduced thus far to the Navier-Stokes RANS and LES approaches discussed above. At
equations. This exercise merely explains in solid boundaries and locations where the turbu-
mathematical terms how the kinetic energy of lent length scale is less than the mesh dimension
the flow is converted into thermal energy by the the RANS treatment for turbulence is employed.
action of the viscosity, μ. In the early 1960s, At other locations, where the mesh is suitably
a meteorologist named Smagorinsky [14] resolved, LES modeling is employed. This
suggested, based mainly on dimensional approach has the potential benefit of reducing
arguments, that the effect of the viscosity could the overall mesh requirement to achieve a given
modeled by a “turbulent” viscosity coefficient: level of turbulent flow resolution. The mesh gen-
" #12 eration process, however, tends to be more
2 u 2
δ challenging.
μt ¼ ρðCs ΔÞ 2 þ (32.17)
δx A fundamentally different approach to CFD is
provided by the so-called Lattice-Boltzmann
where Cs is an empirical constant, Δ is a filter method (LBM). Instead of the Navier-Stokes
length comparable to the size of a grid cell, and equations, the discrete Boltzmann equation is
the expression in brackets has the same func- solved in which the fluid is treated as a set of
tional form as the kinetic energy dissipation particles rather than as a continuum. These
rate. The notation δ
u=δx emphasizes that the particles are tracked numerically through a
www.ebook777.com
discrete lattice mesh that encompasses the oxygen availability is an important factor or
domain of interest, and are allowed to collide where the composition of the gas mixture is
with each other and the bounding surfaces. required for the radiation model.
While LBM models offer some advantages over Most engineering models assume that the
traditional CFD, in particular in respect to com- combustion process can be represented as a sin-
plex geometries and parallel computing, the gle, one-step reaction between fuel and oxygen
authors are not aware of the application of the forming a mixture of products including the
Lattice-Boltzmann method to fire problems, and major species CO2 and H2O and minor species
it is not discussed any further. like soot and CO:
F þ s O2 ! P (32.19)
Source Terms and Boundary
Conditions This model is appropriate provided that the
detailed kinetics are not important and that the
The governing conservation equations discussed product yields are known from experiment. If
previously do not pertain only to fire scenarios. the prediction of minor product species such as
What constitutes a “fire model” are the conserva- CO is required, then the assumption of fast,
tion equations plus a set of boundary conditions single-step chemistry is no longer valid and a
and source terms describing mass, momentum, more sophisticated approach is required. An
and energy exchange between hot gases and example of such an approach is given at the end
compartment walls, the reaction of fuel and oxy- of this section.
gen, the redistribution of energy by thermal radi- The most widely used combustion models for
ation, the spray of water from a sprinkler, the fire applications are the Eddy Break-Up (EBU)
activation of a smoke detector, and dozens of model of Spalding [17], and the closely related
other phenomena that occur in a burning build- Eddy Dissipation Concept (EDC) devised by
ing. Describing these phenomena mathemati- Magnussen and Hjertager [18]. The models
cally is what modeling is all about. In the assume that the consumption of fuel, m_ 000
f , is
sections to follow are brief discussions of the controlled by the rate of molecular mixing of
boundary conditions and source terms found in reactants, which in turn is proportional to the
the governing equations. rate of dissipation of turbulent eddies:

Cρ YO
m_ 000
f ¼ min Yf ; 2 (32.20)
Combustion τmix s
The reaction of fuel and oxygen and the C is a dimensionless empirical constant and τmix
associated entrainment of air into the fire plume k=E is the mixing time. RANS and LES models
is the driving source term in the model. The treat the turbulent kinetic energy, k, and the dis-
simplest approach to modeling combustion is to sipation rate, ε, differently, but the basic idea is
ignore the chemistry and assume that the heat is the same in both types of models. The heat
released within a prescribed volume. This may release rate is obtained by multiplying the fuel
suffice for some applications (e.g., smoke move- mass consumption rate by an effective heat of
ment associated with a well-ventilated fire), combustion. Equation 32.20 may be augmented
provided a reasonable volume is selected for the by an additional term involving the products of
release of heat. However, for fire scenarios where combustion and also by an Arrhenius expression
the combustion region cannot be predefined or to limit the rate of reaction in cold mixtures. The
the chemistry becomes important, a combustion EBU and EDC models have the merit of simplic-
model should be employed. The calculation of ity while permitting heat to be released over a
chemical species is required, for example, where distributed volume determined by the enclosure
geometry and availability of air. Furthermore, numerical point of view is that radiant energy is
the phenomenon of flame lengthening as a con- propagated at the speed of light, as opposed to
sequence of ventilation control is incorporated by the speed of the gas flow or the speed of sound.
this modeling approach. It has been applied with Numerical flow solvers typically advance the
reasonable success by Miles et al. [19] and Holen solution of the governing equations using time
et al. [20] as it has been in other areas of com- steps that are constrained either by the flow speed
bustion engineering. (low Mach number LES) or time steps that are
The hazard posed by high temperatures consistent with large-scale changes in the envi-
and the loss of visibility due to fire may be ronment (RANS). In either case, the speed of
compounded by exposure to toxic gases such as light is essentially infinite; thus, radiant energy
carbon monoxide or hydrogen cyanide. Although is assumed to redistribute itself instantaneously.
advanced combustion models may include the The simplest model of radiation transport
capability to predict some of these toxic gas assumes that the gases are non-scattering (the
species, it is generally necessary in fire engineer- gases only absorb or emit thermal radiation)
ing to prescribe the species yields. In other and gray (the radiation has no spectral depen-
words, the production rate of smoke and other dence). Under these assumptions, the governing
combustion products is usually specified by the equation can be written as
user based on experimental measurements.
Predicting, rather than specifying, the generation s rIðx; sÞ ¼ κðxÞ ½Ib ðxÞ Iðx; sÞ (32.21)
of species such as CO, HCN, and soot requires
finite rate chemistry to be included in the com- Here, I is the radiant intensity, a function of both
bustion model. An approach that has been position, x, and direction, s. The gray-gas
exploited in a number of fire studies is to assume assumption neglects the fact that I is also a func-
that the flame is a statistical ensemble of thin tion of the wavelength, as are the absorption
laminar flamelets that incorporates detailed coefficient, κ, and the source term, Ib. To account
chemistry from either experimental measure- for wavelength dependence, Eq. 32.21 must be
ments or detailed kinetic calculations. Details solved over discrete “bands” of the electromag-
of this approach may be found in Peters [21] netic spectrum [23]. However, in fires, soot is the
and in the references of Cox [8] and principal emitter and absorber of thermal radia-
Novozhilov [9]. Magnussen and Hjertager [18] tion. Because its radiation spectrum is continu-
extended the eddy breakup model to include soot ous, it is often assumed that there is no spectral
formation and oxidation, using the mechanisms dependence; that is, the participating medium
suggested by Tesner et al. [22]. This involved is gray.
the solution of two further transport equations. Even in its simplest form, the radiation trans-
Soot modeling is, however, difficult and not port equation poses two challenges to the
generally included within current CFD fire modeler: (1) the prescription of the spatially
computations and remains a topic for research. dependent absorption coefficient, κ(x); and
(2) the numerical solution. As for the latter, a
number of methods have emerged that involve
Radiation Heat Transfer discretizing the equation into a finite number of
solid angles and sweeping over the numerical
The governing conservation equations of mass, grid until the radiant energy is redistributed.
momentum, and energy describe the convection The process is often done gradually over several
and diffusion of hot gases from a fire. However, time steps of the flow solver, depending on the
the redistribution of energy via thermal radiation level of temporal fidelity desired. Usually, equiv-
is very important and needs to be included in the alent or greater temporal resolution is demanded
energy transport equation 32.4. What makes radi- by the flow solver, easing the computational bur-
ation heat transfer particularly difficult from a den of the radiation solver [24].
www.ebook777.com
The prescription of the absorption coefficient, from the outside. In most cases, air and smoke
κ(x), is challenging because it combines the pass through open doors or windows, fans blow
contributions of soot and gaseous exhaust and extract gases, and items burn and introduce
products. Soot is the main emitter and absorber fuel gases into the space. For the modeler,
in practical fire scenarios; thus, any model for all these phenomena are known as boundary
must account for it. This requirement has caused conditions that supplement the governing
difficulties over the years because contributions equations.
from the combustion community often empha- Free (or open) boundaries are located where
size the spectrally dependent properties of CO2 the modeled domain interfaces with the external
and H2O rather than the spectrally independent world beyond. Generally the pressure is
properties of soot. The dominance of soot in the specified, typically to a reference or ambient
radiation calculation simplifies the prescription value, and the velocity derivatives normal to the
of κ, but at the same time it demands a better boundary are set to zero. Furthermore, where
description of soot growth and oxidation than the outflow occurs the normal derivatives of the sca-
current practical fire models are able to provide. lar fields will be set to zero, and where inflow
Regardless of the choice of absorption coeffi- occurs their ambient (atmospheric) values will be
cient and numerical method, the solution of the assumed. To achieve realistic predictions of air
radiation transport equation is added to the over- entrainment and smoke exhaust, it may be neces-
all solution via a source term in the energy sary to locate these boundaries well away from
equation 32.4: the compartment of fire origin.
Mass inlet boundaries are located where a
r qr ðxÞ ¼ κðxÞ ½UðxÞ 4πIb ðxÞ; prescribed source of air enters the computational
Z
UðxÞ ¼ Iðx; sÞ dΩ domain. One example is at mechanical ventila-
4π tion supplies. Other examples include external
(32.22) wind boundaries and the fuel bed itself if pyroly-
sis is included in the model in DiBlasi [25].
The integrated intensity, U, multiplied by κ As for free boundaries where inflow occurs, it is
represents the rate of energy absorbed per unit critical that appropriate values are assigned to the
volume, whereas 4π κ Ib is the rate of energy sensible enthalpy (temperature) and other scalar
emitted per unit volume. The source term is fields.
usually assumed to be a blackbody radiator: Ib ¼
σ T4/π. Within the fire itself, the emission term
dominates, and there is a net loss of energy from Momentum Exchange at Boundaries
the region, whereas within the relatively cool,
smoke-laden upper layer, there is a net gain of Where the fluid comes into contact with solid
energy due to absorption. Because of the uncer- objects (compartment walls, for example) bound-
tainty in the near flame temperature and soot ary conditions are required for the momentum
concentration, it is still common practice to pre- equations and, where appropriate, for the solved
scribe, rather than predict directly via the source turbulence variables. For most applications the
term, the fraction of energy lost from the fire via no-slip velocity condition is assumed at solid
thermal radiation. Typically, 30–40 % is chosen, surfaces (i.e., zero flow directly adjacent to the
depending on the fuel type and fire size. surface). However, the proper specification of
boundary conditions is complicated by the fact
that the inner region of the turbulent boundary
Mass Exchange at Boundaries layer adjacent to a solid surface has very sharp
velocity gradients. A large number of grid points
Rarely are fire simulations performed for are required to resolve this region, making the
compartments that are completely sealed off simulation computationally expensive and
impractical for realistic compartments. An alter- is likely to be an important mechanism. For

native, widely used approach to resolving the example, in large compartments or tunnels this
boundary layer is to use a so-called wall func- heat loss may account for most of the heat losses
tion [26], which assumes that the tangential from the fire, and ignoring it will generate erro-
velocity is a logarithmic function of the normal neous predictions.
distance from the surface. An empirical relation- Where there is heat transfer, it is necessary to
ship relates the wall shear-stress to the resolved provide boundary conditions for the enthalpy and
variables at the first grid point, and appropriate radiation equations, which in turn requires the
source terms can be derived for each solved temperature of the solid surface (Tw) be defined
equation. Details are provided in standard at each location adjoining the CFD grid. The
textbooks [13]. surface temperature, together with its emissivity,
Care is required in using the wall-function establishes the radiative heat transfer rate for the
approach to ensure that the first grid point is radiation transfer equation. For the enthalpy
positioned such that the empirical relationships equation the source term is the rate of convective
are valid. Guidance should be taken from the heat transfer (q_ 00c ) to the surface from the adjoin-
model developer or the model documentation ing grid point
on the appropriate numerical grid to employ at
surfaces. Commercial CFD models may offer q_ 00c ¼ hc ðTw Tg Þ (32.23)
advanced wall treatments that adapt automati-
cally depending on the grid point and the flow where Tg is the gas temperature at the nearest
regime, and these should generally be employed grid point and hc is the convective heat transfer
if available. coefficient. This coefficient can be defined as a
constant as in a zone model or, alternatively, a
more realistic approach is to compute the coeffi-
Energy Exchange at Boundaries cient as a function of the local flow regime
making use of an empirical wall function akin
Conventional CFD models devote considerable to that done for the momentum equations and
attention to the velocity boundary conditions, but turbulence model.
for fire, the energy exchange at solid surfaces is The definition of the surface temperature is
paramount, especially when the surfaces are trivial in the case of isothermal boundaries where
burning. In fact, pyrolysis modeling is a subject the surface is maintained at a fixed temperature.
unto itself and is touched on only briefly here. In general, however, the conduction of heat into
Suffice it to say that almost any treatment of a the walls needs to be modeled. If it is assumed
solid object can be incorporated in a CFD fire that conduction occurs only in the normal direc-
model because of the fairly loose coupling that tion, then it may be treated by solving a
exists between the gas and solid phases. one-dimensional heat conduction equation at
Hot gases lose heat to the structure at a rate each location on the surface adjoining the gas
determined by both the thermal properties of the phase grid. The net conduction heat flux into
bounding solids and the evolution of conditions the solid is given by the sum of the net radiation
within the gas phase. Early in the fire, the walls and convection fluxes at the surface.
are nearly at ambient temperature and the rate of A more sophisticated solution may be
heat transfer will, all other things being equal, be provided by coupling the CFD model with a
at its greatest. Later, as the walls and other separate solid phase calculation of the thermal
surfaces heat up, the rate of heat transfer conduction within the solid structure(s). This
decreases. In some situations it may be appropri- calculation could take the form of a fully coupled
ate to treat the solid surfaces as adiabatic (i.e., no gas and solid-phase model, or a coupling of
heat transfer). However, care must be taken when boundary conditions between a CFD model and
adopting this approach as heat loss to the surfaces a separate solid-phase model. In general, the
www.ebook777.com
required information to be passed from the solid- 7.6 m2/g for flaming combustion), and ms is the
phase model to the CFD model is the surface mass concentration of soot particles in units of
temperature and pyrolysis rate and, in the other g/m3. The fire model predicts this latter quantity.
direction, the convective and radiative heat Visibility distance calculated in this way
fluxes. at a CFD grid cell represents the visibility
corresponding to a homogenous smoke layer
and assumes the smoke density at the cell applies
Open Boundary Conditions everywhere in the domain. Using the CFD solu-
tion data, it is possible to perform a more realistic
The results of a CFD fire analysis can be sensi- “line of sight” calculation using the computed
tive to the location of the “free” (ambient or open data at each grid cell along this line [29].
air) boundaries at the edge of the computational
domain. These should be located at a sufficient
distance from the fire source and regions of inter- Sprinklers and Fire Suppression
est so that they do not incorrectly influence the
solution. The setting of ambient values in rela- It is fairly straight forward to include a sprinkler
tion to other boundary conditions may require activation algorithm within a CFD fire model,
special attention (e.g., it can dictate whether based on the RTI (Response Time Index) concept
smoke can be expected to vent naturally in an of Heskestad and Bill [30]. This algorithm
air-conditioned atrium). Heat losses to solid requires the gas temperature and velocity in the
boundaries can also have a significant influence vicinity of the device. Following sprinkler acti-
on smoke movement, particularly where the vation, the water spray interaction with the hot
smoke has cooled down and the amount of heat gases can be modeled in a variety of ways. It is
transferred to walls and ceilings can determine most often done using a Lagrangian particle
the degree of buoyancy and stratification of the tracking sub-model, in which statistically repre-
smoke gases. sentative water particles are tracked numerically
through the gas phase. The impact on the fire
gases from the transfer of mass, momentum and
Visibility heat associated with the trajectories and evapora-
tion of the water particles is included as source
Fire protection applications often require an esti- terms in the underlying fluid dynamics equations.
mate of visibility, the distance through smoke A good summary of this approach is given by
that a person is expected to be able to see, say, Makhviladze et al. [31]. Water suppression and
an exit sign. Visibility distance may be derived extinguishment, however, are far more difficult
from a CFD solution using the predicted con- to model, and these phenomena are not generally
centration of soot particulates. The corres- a practical option for most fire engineering
pondence between smoke concentration and applications.
visibility is based on the work of Jin [27] and
Mullholland [28] and is commonly implemented
into CFD models using the following (or similar) Fire-Structure Interface
expression for visibility distance:
While the modeling of the transport of smoke
K1 and heat within the gas phase remains the pri-
S¼ (32.24)
K2 ms mary role of a CFD fire model in the majority
of fire engineering applications, there are others
Here K1 is a constant usually set to 3 for where the coupling with the compartment
reflecting signs and 8 for illuminated ones, K2 is boundaries and supporting structures is the
the specific extinction coefficient (often taken as main focus. For example, the absorption and
transmission of heat through elements of glazing, would be impossible if not for graphical tools
possibly to predict the likelihood of failure or that have been developed specifically to visualize
to calculate the irradiance where people may be the computed results in various ways.
escaping on the unexposed side, might be
required. Another example would be to investi-
gate the response of concrete linings to fully Finite Difference Approximation
developed fire conditions, for example in a road
tunnel. A commonly used technique for solving the
The area where the coupling of CFD and governing conservation equations is known as
structural response has been considered in most finite-differencing. The three-dimensional vol-
detail is arguably in relation to the fire protection ume of interest, say a room in a building, is
of supporting steel structures. By conducting subdivided with a mesh made up of many small
CFD calculations of reasonably worst case fire grid cells, with each cell containing an average
scenarios, the thermal conditions to which the value of each flow variable. In general, more grid
steel elements would be exposed can be applied cells produce a more accurate approximation of
as a boundary condition to the solid phase model the true solution but at increasing computational
describing the transfer of heat and mechanical cost. The simplest meshes consist of cells that are
forces within the structure. In the simplest box-shaped with the dimensions of the boxes
approach, the analysis would ascertain whether either fixed or varying. Such a mesh is described
the temperature of the steel elements reaches a as rectilinear. Meshes that conform to the
critical level beyond which the structure might boundaries of complicated objects are known as
become unstable or even collapse. The method- body-fitted. Fire models that are intended for
ology adopted to transfer the information, building applications most often use rectilinear
i.e. surface temperature and convective and radi- meshes, whereas special applications, like
ative thermal fluxes, between the CFD and struc- furnaces and combustors, often employ body-
tural model is non-trivial, and is an area of fitted coordinate systems.
on-going research, see for example [32]. As an example of how finite-differencing
works, consider the discretization of the mass
conservation Eq. 32.1 on a three-dimensional
Numerical Solution rectilinear mesh. The cells are typically indexed
by the integers i, j, and k. Each flow variable is
The equations described in the previous sections represented as an average over the cell volume, δ
are all written in continuous form; that is, they x δ y δ z, and over the time step, δ t. At the start
are exact representations of the governing con- of the simulation, all the flow variables in each
servation laws. Unfortunately, they have no grid cell are assigned an initial value, and as the
closed form solution except in very limited simulation progresses in time, these values are
cases. Therefore, numerical techniques are updated at each discrete time step. A common
required to obtain approximate solutions. Even discretization technique is to define scalar
on the fastest computers available, these quantities at the center of each cell and vector
techniques can be costly in terms of the time components at their respective cell faces. For
consumed by the computer’s central processing example, ρnijk represents the average density in
unit (CPU) and the size of the computer’s the cell with indices i j k at the nth time step,
random-access memory (RAM). Calculation whereas unijk and uni1;jk represent the averages of
times can range from several hours to several the velocity component u over the faces of the
weeks. Memory requirements can range from cell that point in the positive and negative
hundreds of megabytes to tens of gigabytes. In directions of x in the standard Cartesian coordi-
the end, the calculations will produce an enor- nate system. A simple way to advance the density
mous amount of data, the processing of which of each cell in time is to write the mass
www.ebook777.com
conservation equation in an approximate form as advancing the solution in time is merely a matter
follows: of computing these terms. An implicit scheme
uses values of the flow variables at both the
ρniþ1;jk unijk ρni1;jk uni1;jk
ρnþ1 current and next time steps, meaning that there
ijk ¼ ρijk δt
n 2 2
δx is no direct way of updating the values. Rather, a

ρni;jþ1;k vnijk ρni;j1;k vni;j1;k
þ 2 2 linear system of equations must be solved.
δy Although they require more computational effort
ρnij;kþ1 wnijk ρnij;k1 wnij;k1 per time step, implicit schemes typically are
þ 2 2
δz more numerically stable and permit larger time

(32.25) steps than explicit schemes. In fact, the most
commonly used numerical scheme in commer-
Notice that even the simplest of the conservation cial CFD packages is known as SIMPLE (Semi-
equations approximated with the simplest of Implicit Method for Pressure-Linked Equations).
differencing schemes on the simplest of meshes Details may be found in Patankar [4].
is still fairly difficult to express in a succinct way.
In fact, to save space, the density at the respective Conservative Versus Nonconservative
cell faces, computed as an average of its values Schemes
in the adjoining cells, is denoted by the 12 in A numerical scheme that is described as conser-
the appropriate subscript. The momentum and vative has the following property: If the right-
energy conservation equations are far more com- hand side of Eq. 32.25 were summed over all
plicated than this one, even though the basic ideas grid cells, the intermediate mass fluxes would
are the same. It is easy to understand why CFD cancel exactly, leaving only the mass fluxes at
software consists of tens of thousands of lines of the boundary of the computational domain. This
computer instructions and that calculations can is the discrete analogue of the Divergence
take hours to weeks to complete, depending on Theorem:
the number of grid cells in the mesh and the Z I
complexity of the differencing scheme. r ρu dV ¼ ρu dS (32.26)
CFD models employ various types of V @V
differencing schemes, meshes, physical and
numerical assumptions, and so on. No two This very desirable property in a numerical
models are exactly alike, but many have common scheme ensures that mass is globally conserved,
traits. An informal classification system has regardless of inaccuracies related to the grid cell
evolved within the community of developers to size, differencing scheme, or other local effects.
distinguish one model from another. For exam-
ple, the simple scheme for updating the pre- Staggered Versus Co-located Variables
vious density would be described as a three- Assigning various flow variables to different
dimensional, explicit, conservative, staggered locations within the grid cell is a strategy adopted
finite-difference scheme that is first order accu- by the developer based on the particular applica-
rate in time and second order accurate in space. tion of the model and the convenience of pre-
Consider each descriptor in turn: scribing boundary conditions, obstructions, and
other special features. Suffice it to say that the
Explicit Versus Implicit Schemes strategy adopted in Eq. 32.25 is not unique.
To say that a numerical scheme is explicit means Other techniques are discussed in Anderson
that the flow variables may be advanced in time et al. [2] and Ferziger and Peric [3].
based solely on their values at the current time
step. In Eq. 32.25, notice that all the values of the Order of Accuracy of the Scheme
density and the velocity components on the right- A common misconception about numerical
hand side are taken at the nth time step; thus, schemes is that those with a high order of
accuracy are better than those of lower order. A running along the length of the tunnel. Care is
finite difference is an approximation of a partial required in using symmetry boundaries to ensure
derivative, and the order of accuracy relates the that the imposed condition of symmetry is physi-
degree of approximation. For example, the cally realistic. For example, in an LES calcula-
scheme described in Eq. 32.25 is first order accu- tion, it is generally undesirable to use a symmetry
rate in time because the time derivative of the boundary because the methodology does not
density at the nth time step is approximated as assume the fire to be symmetric (in a time-
averaged sense).
@ρ ρnþ1 ρn
¼ þ OðδtÞ (32.27)
@t δt
Finite Volume Method
The symbol OðδtÞ represents the discretization
error—neglected terms from the Taylor series
A popular alternative to finite-differencing
expansion of the density that are proportional to
employed in both commercial and fire-specific
the first and higher powers of the time step size.
CFD models is the finite volume method. This
Another way of characterizing this approxima-
method has some of the characteristics of the
tion is to say that it has a forward bias in time.
finite element method employed in structural
The spatial derivatives in Eq. 32.25 are known as
and solid-phase thermal analysis programs and
central differences, having neither a forward nor
can be employed with either structured or
backward bias in the respective coordinate
unstructured meshes. Rather than working with
directions. Such differences that incorporate
the conservation equations in their differential
values from the nearest adjacent grid cells are
form, as in Eqs. 32.1–32.4, the finite volume
typically second order accurate. Higher-order
method takes as its starting point the equations
accuracy requires information beyond nearest
in their integral form. For example, integrating
neighbors, adding to the cost of computing the
equation 32.2 over a control volume V and apply-
finite differences. The potential increases in
ing the Divergence Theorem (Eq. 32.26) yields
accuracy are often offset by the decrease in
an integral form of the equation:
computational efficiency and ease of implemen-
Z I
tation. A popular trade-off is to use more grid @
ρYα dV þ n ðρYα uÞ dS
cells with a second-order accurate scheme rather @t V @V
I Z (32.28)
than less grid cells with a higher-order scheme.
¼ n ðDα rYα Þ dS þ m_ 000α dV
@V V
Spatial Dimension of the Scheme
There are some fire scenarios where it may not be The solution domain is subdivided into a finite
necessary to solve the fully three-dimensional number of control volumes (cells) in much the
form of the governing equations. Sometimes, same way as in the finite difference method
flows through ducts or tunnels can be described previously. The crux of the finite vol-
approximated in two spatial dimensions. Also, ume method is in determining the surface and
where there is a symmetry in the fire scenario volume integrals in Eq. 32.28, which in turn
(e.g., a simple compartment with a fire in the requires that the surface values of the solved
middle of the room), the computational effort variables be expressed in terms of the neighbor-
can be reduced by a factor of two by modeling ing control volume (node) values, using an
one-half of the geometry. The plane of symmetry appropriate interpolation scheme. A system of
is then defined as a symmetry boundary condi- algebraic equations is developed as in the finite
tion at the corresponding edge of the modeled difference method, the solution of which
domain. Tunnels are another application in provides the mean value of the solved variable
which a vertical symmetry plane is often used, at each control volume (generally assigned to the
www.ebook777.com
node points). Although the finite volume method Computer Hardware and Software
was originally employed with structured grids,
the trend with modern commercial models has CFD solvers are written in a variety of computer
been toward unstructured grids consisting of languages, most notably Fortran or C. They will
hexahedral or tetrahedral elements. run, or execute, on any computer for which a
The finite volume method is intrinsically con- compiler is available to translate the source
servative (provided the surface fluxes are defined code into the native machine language. Usually,
appropriately). It may be explicit or implicit and CFD models consist of three basic parts: an
the mesh may be staggered or co-located. interface that allows the user to input parameters,
Depending on the numerical approximation of the flow solver, and a graphical program to dis-
the integrals, the solution may be formally any play results. It is not unusual for the user inter-
order of accuracy. For the diffusion terms, a face, flow solver, and graphics utility to be three
second order (central difference) approach is separate computer programs, sometimes written
generally adopted. For the convection terms, in different languages, and sometimes run on
matters are more complicated, and the use of different computers. For example, the flow solver
central differencing can produce erroneous itself can be run on a remote machine because it
results in convection-dominated flows when a does not require any interaction with the user.
RANS turbulence formulation is used. First While a calculation runs somewhere else, the
order, or “up-wind,” schemes remove the user is free to set up new calculations or process
instabilities caused by central differencing, but results from a previous one. What were once
they can be inaccurate if the grid is not suffi- referred to as “supercomputers” are now simply
ciently fine. A common approach has been to clusters of conventional computers that are dedi-
use the so-called “hybrid” scheme [33] in calcu- cated to running CFD calculations, either serially
lating the surface fluxes, which produces stable or in parallel.
solutions that are generally more accurate than The input interface can be a graphical user
those provided by a pure first order, upwind interface (GUI), a simple text editor, or some
approach but not formally as accurate as a second combination of the two. Some of the more com-
order approach. plex interfaces allow for input data to come from
other sources, like a computer-aided design
(CAD) program. This is an increasingly impor-
Alternative Solution Methods tant consideration in CFD, as the most time-
consuming part of a simulation (for the user,
The finite element method (FEM) is well not the computer) is the description of the geom-
established for solving the partial differential etry. It is now possible to simulate the flow of
equations of structural mechanics and solid fire-driven gases through entire buildings, but
phase heat transfer, and shares some of the without some automated method of input, the
underlying mathematics associated with finite setup process can be tedious.
difference and finite volume methods. An CFD is becoming more and more routine as
unstructured numerical mesh, often composed computers get cheaper and faster. In addition to
of tetrahedra, is employed, which makes the the availability of doubly fast CPUs every
method particularly suited to complex 18 months, an emerging trend in the industry is
geometries. While the finite element method to use more than one CPU (and the associated
has been applied to fluid flows, its application memory) for a single simulation. The technique,
in the fire field has to date been limited. Other known as parallel processing, has been made
numerical methods that have been employed by possible by the development of software that
the CFD community include spectral methods allows data to be transferred rapidly between
and boundary elements. two or more computers that are working on the
same calculation. The most popular data-passing species concentration, visibility is not defined at
protocol is message passing interface (MPI). To a point but rather integrated over a path
the CFD developer, MPI is nothing more than a connecting the virtual occupant and the virtual
set of “call” statements that can be written into exit sign. The CFD model can produce a spatially
the source code, instructing computer A to send a and temporally varying smoke density field, but
packet of numbers to computer B and vice versa. it is difficult or impossible to integrate a priori the
The formats of the call statements have been density over every possible path between the sign
standardized by a committee of interested indus- and the moving occupant. Fortunately, graphical
try experts, and several very good, and free, techniques are now available that can process the
implementations of MPI have been released smoke density data and produce at any given
into the public domain. Although writing and time and any given place in any given direction
running a good, efficient parallel CFD model is a realistic view through the smoke, allowing the
still a daunting task, parallel processing does not viewer to assess whether or not the sign
significantly increase the cost of hardware or (or whatever is of interest) would be visible in
software. It is possible to exploit the computers the event of a real fire.
that a company already owns to run CFD models
in parallel.
Verification and Validation
Visualization The use of fire models currently extends beyond

the fire research laboratories and into the engi-
As computers get faster and calculations get big- neering, fire service, and legal communities. Suf-
ger, there is a tremendous need to render the ficient evaluation of the models is necessary to
output of the simulations in some useful form. ensure that users can judge the adequacy of its
Certainly, time histories of individual quantities technical basis, appropriateness of its use, and
at discrete locations can be saved and readily confidence level of its predictions. The model
plotted, much like thermocouple data from a evaluation process consists of two main
real experiment. But this only tells part of the components: verification and validation [34]. Ver-
story. To better understand the flow dynamics, it ification is a process to check the correctness of
is necessary to render pictures or animations of the solution of the governing equations. Verifica-
the entire computational domain, either in the tion does not imply that the governing equations
form of contour plots, flow vectors, streamlines, are appropriate, only that the equations are being
or tracer particles. This type of output is invalu- solved correctly. Validation is a process to deter-
able to the numerical analyst but still only par- mine the appropriateness of the governing
tially satisfying to the lay person. Perhaps equations as a mathematical model of the physi-
spurred on by the current generation of cal phenomena of interest. Typically, validation
computer-animated films, some want to see the involves comparing model results with experi-
results of the simulations rendered in a lifelike mental measurement. Differences that cannot be
way, considerably different than conventional attributed to uncertainty in the experimental
CFD output. measurements are attributed to the assumptions
This trend to realism is more than superficial. and simplifications of the physical model.
In both design and forensic applications, the real- It is commonly assumed by model users, par-
istic rendering of smoke is very valuable, but it ticularly those who have purchased expensive
requires graphical techniques that stretch the CFD software, that verification and validation is
limits of even the best video cards available. the responsibility of the model developers. Cer-
Visibility is a key consideration in the design of tainly, CFD developers do a considerable amount
any large structure, especially given the current of this type of work, in particular verification, but
trend of open architecture. Unlike temperature or it is impossible to ensure that the model is
www.ebook777.com
“validated” for every conceivable application. tests should be performed to ensure that mass
Indeed, the very point of numerical modeling is and energy conservation laws are properly
to predict the outcome of fire scenarios that have applied in the model.
not, or cannot be, replicated in a controlled labo-
ratory environment. Thus, the burden of verifica-
tion and validation must be shared by the model Validation
developer and user. The benefit to the user is
twofold: first, it confirms that the user can use The typical way to validate a CFD fire model is
the software properly, at least for the given appli- to compare its results to experimental
cation; and second, it assures the user that the measurements. This involves two important
model can address the given fire scenario, even tasks: (1) selecting appropriate experiments and
providing the user with some estimate of its (2) quantifying the accuracy of the model’s pre-
accuracy. diction of the outcome of these experiments. The
following two sections address these two tasks.
Verification Applicability of Validation Experiments

How can one determine if a particular application
The most important task for a CFD modeler is to of the model has been validated? For example,
perform a grid resolution study for the particular suppose the problem at hand is a fire in a ware-
application at hand. The accuracy of a CFD cal- house with a 10 m ceiling, sprinklers, roof vents,
culation is dependent on the resolution of the and HVAC system. There is probably no experi-
underlying numerical grid. As discussed in mental data set that is exactly like it, and it would
Sect. 32, the order of accuracy of a numerical be too expensive to conduct new experiments.
algorithm indicates the rate at which the solution How does one determine if any validation work
of the discretized governing equations converges is appropriate for this scenario? The approach
towards the solution of the exact equations. For taken in a fire model validation study conducted
example, if the scheme is second-order accurate by the US Nuclear Regulatory Commission
in space and time, the error due to the finite (NRC) and the Electrical Power Research Insti-
difference/volume approximation will decrease tute (EPRI) was to characterize the experiments
by a factor of 4 if the grid cell and time step used in the study in terms of a handful of com-
decrease by a factor of 2. Demonstrating this for monly used non-dimensional quantities from the
a particular application of the model is known as fire literature [35]. This essentially defines
a grid resolution study. Selection of an appropri- the “parameter space” for which the model was
ate grid size should be based on whether or not a validated. The model users are warned that the
particular level of accuracy is achieved for a models should only be applied within this param-
particular grid cell size. eter space. This prevents the tendency by users
In addition to a grid resolution study, there are to simply declare that the model has been
a variety of other tests that a model user should validated and can be used for any application.
perform to ensure that the model is solving the The non-dimensionalized parameters used in the
governing equations correctly, and that the NRC/EPRI study are:
model user is running the model properly. What 1. Fire Froude Number, Q_ : A convenient way
tests to perform depends on the application. For to express the heat release rate of the fire
example, if the application involves a pressure relative to its base area is by way of the
rise in a relatively tight compartment, simple non-dimensionalized quantity:
tests should be performed to ensure that the pres-
sure increase obeys the basic laws of thermody- Q_
Q_ ¼ pffiffiffiffiffiffi (32.29)
namics. If water droplets are involved, simple ρ1 cp T1 gD D2
where Q_ is the peak heat release rate of the fire the compartment is considered “well-
and D is the equivalent diameter of the base of ventilated” and if ϕ > 1, the compartment is
the fire. The Fire Froude Number is a useful considered “under-ventilated.” In general,
quantity for plume correlations and flame under-ventilated fire scenarios are more chal-
height estimates. A large value ( Q_
1 ) lenging for the models because the combus-
describes a fire for which the energy output tion physics are more complicated.
is relatively large compared to its physical 4. Relative Distance Along the Ceiling, rcj =H :
diameter, like an oil well blowout fire. A low This ratio indicates the distance from the fire
value (Q_ 1) describes a fire for which the plume of a sprinkler, smoke detector, etc.,
energy output is relatively small compared to relative to the compartment height, H. The
its diameter, like a brush fire. maximum ceiling jet temperature, important
2. Flame Height Relative to Ceiling Height, in determining device activation, has been
Lf/H: This ratio is a convenient way to express shown to be a function of this ratio.
the physical size of the fire relative to the size 5. Relative Distance from the Fire, rrad =D: This
of the room. A value greater than one ratio indicates whether a “target” is near or far
indicates that there is flame impingement on from the fire. In general, it is more challenging
the ceiling, an important consideration when to predict the radiative heat flux to objects
evaluating devices such as sprinklers and near the fire.
smoke detectors. The Flame Height, Lf, is 6. Room Length and Width Relative to the Ceil-
the height of the visible flame, based on ing Height, L/H and W/H: These ratios are
Heskestad’s correlation [36]: useful mainly when assessing an empirical or
zone model because most of the correlations

Lf ¼ D 3:7 ðQ_ Þ2=5 1:02 (32.30) used by these models are limited in terms of
compartment aspect ratio. For CFD, extreme
values of these ratios might indicate unusual
3. Global Equivalence Ratio, ϕ: A convenient fire behavior.
way to determine whether a compartment
7. Ceiling Height Relative to the Fire Diameter,
fire is limited by fuel or oxygen is to estimate
H/D∗: This ratio is a non-dimensional mea-
the ratio of the fuel supply rate, m_ f , to the sure of the height of the fire plume. D∗ is a
oxygen supply rate, m_ O2 , divided by the stoi- length scale that incorporates the heat release
chiometric ratio, r: rate of the fire:
m_ f =m_ O2 Q_ !2=5
ϕ¼ ¼ ; Q_
r ΔhO2 m_ O2 D ¼ pffiffiffi (32.32)
8 ρ1 cp T1 g
> pffiffiffiffiffiffi
< 1 0:23 A0 H0 Natural (32.31)
_
mO2 ¼ 2
>
: 0:23 ρ V_ The larger the ratio H/D∗, the more important
Mechanical the plume becomes in the overall scenario.
For empirical and zone models, it indicates
Here, Δh is the heat of combustion, Q_ ¼ m_ f Δh whether or not the plume entrainment correla-
is the peak heat release rate of the fire, ΔhO2 tion is appropriate. For CFD, it indicates how
¼ r Δh is the heat of combustion per unit mass “high” the plume actually is, in
oxygen consumed, A0 is the area of the com- non-dimensional terms.
partment opening, H0 is the height of the
opening, ρ is the density of air, and V_ is the Quantifying Model Uncertainty
volume flow of air into the compartment Having determined the appropriateness of the
due to a ventilation system. The factor 0.23 chosen experiments, it is now necessary to quan-
is the mass fraction of oxygen in air. If ϕ < 1, tify the accuracy of the model in predicting the
www.ebook777.com
Fig. 32.2 A comparison

of measured versus
predicted wall temperature.
The off-diagonal lines
σ bounds
indicate the 2e
for the experiments (long
dash) and the model
(short dash)
outcome of these experiments. For each quantity would merge and the dash-dot bias line would
of interest, a summary plot of the results should overlap the solid diagonal line. The extent to
be constructed like the one shown in Fig. 32.2. which the data scatters outside of the experimen-
The accuracy of the model can be expressed in tal bounds is an indication of the degree of model
terms of two statistical parameters. The first, δ, is uncertainty.
the bias factor. It indicates the extent to which the The derivation of these uncertainty statistics is
model, on average, under or over-predicts the described in [37], and it is summarized here.
measurements. For example, the bias factor for First, a few assumptions are made:
the data shown in Fig. 32.2 is 1.13. This means 1. The experimental measurements are assumed
that the model over-predicts wall temperatures to be unbiased, and their uncertainty is
by 13 %, on average, and this is shown assumed to be normally distributed with a
graphically by the dash-dot line. The second constant relative standard deviation, e
σE .
statistic is the relative standard deviation of the 2. The model uncertainty is assumed to be nor-
model, e σ M. This indicates the degree of scatter of mally distributed about the predicted value
the points. Referring again to Fig. 32.2, there are divided by a bias factor, δ. The relative
two sets of off-diagonal lines. The first set, standard deviation of the distribution is
shown as long-dashed black lines, indicate the denoted as eσM .
estimated experimental uncertainty. The slopes Now, given a set of experimental
of these lines are 1 2e σ E , i.e. the 95 % confi- measurements, Ei, and a corresponding set of
dence interval for the measurements. In this case, model predictions, Mi, compute the following:
e
σ E ¼ 0:07. The second set of off-diagonal lines,
shown as short-dashed lines, indicates the model 1X n
lnðM=EÞ ¼ lnðMi =Ei Þ (32.33)
uncertainty. The slopes of these lines are δð1 2 n i¼1
e
σ M Þ. In this case, e
σ M ¼ 0:2. If the model were as
accurate as the measurements against which it is The relative standard deviation of the model, e
σ M,
compared, the two sets of off-diagonal lines can be computed from the following equation:
n h i2
1 X It must be emphasized that this estimated
e
σ 2M þ e
σ 2E ¼ lnðMi =Ei Þ lnðM=EÞ probability of failure is based only on the model
n 1 i¼1
uncertainty. It does not account for parameter
(32.34)
uncertainty; that is, the uncertainty in the input
parameters.
The bias factor is:

e
σ 2M eσ 2E
δ ¼ exp lnðM=EÞ þ (32.35)
2 2 Applications
For a given model prediction, M, the “true” value This section presents examples of how CFD fire
of the quantity of interest is assumed to be a models are used in practice. These applications
normally distributed random variable with a can be divided into three general categories—
mean μ ¼ M/δ and a standard deviation of σ ¼ research, design, and forensic. For research, the
e
σ M ðM=δÞ. Using these values, the probability of models can help explain basic fire phenomena.
exceeding a critical value, xc, is: For design, the models are used to predict the
spread of smoke and heat from a hypothetical fire
1 xc μ
Pðx > xc Þ ¼ erfc pffiffiffi (32.36) in a real or planned building. For forensics, the
2 σ 2 models aid in the reconstruction of an actual fire.
For a design application, the fire is usually
Note that the complimentary error function is
specified; that is, the ignition, growth, and even-
defined as follows:
tual decay of the fire are not predicted by the
Z 1
2 model but rather specified by the design engineer
et dt
2
erfcðxÞ ¼ pffiffiffi (32.37)
π t and reviewed by the code enforcing authority.
For a reconstruction, the model is usually used
It is a standard function in most mathematical or to explain how a small fire grew and spread to
spread sheet programs. cause serious damage or injury. Rarely are fire
As an example of the procedure, suppose that models of the type described in this chapter used
the model whose results are plotted in Fig. 32.2 to show how a fire was actually ignited, as the
has predicted that the wall temperature within a physical mechanism of this event (electrical
compartment would peak at a value of 300 ∘C short, arcing fault, arson, etc.) is usually not
due to a given design fire. Suppose also that the included in the model.
failure criterion for the wall lining material is
325 ∘C. What is the probability that the wall
temperature could reach 325 ∘C? First, it is best Fundamental Fire Dynamics
to work in terms of temperature rise. The ambient
temperature is 20 ∘C; thus, the predicted temper- CFD, in particular large eddy simulation,
ature rise, Δ Tp, is 280 ∘C and the critical tem- provides a convenient means to study basic fire
perature rise, Δ Tc, is 305 ∘C. From Eq. 32.36, behavior. The most obvious application is the
the probability that the temperature could exceed study of fire plumes; for example, predicting
the critical value is: the height of the visible flame, the centerline
! velocity and temperature, and the pulsation fre-
1 ΔTc ðΔTp =δÞ quency. Heskestad’s empirical flame height
PðΔT > ΔTc Þ ¼ erfc pffiffiffi
2 e
σ M ðΔTp =δÞ 2 correlation (Eq. 32.30) is valid for values of Q_
! between 0.1 and 10,000, characterizing intermit-
1 305 ð280=1:13Þ
¼ erfc pffiffiffi ffi 0:12 tent grass fires all the way to oil well blowout
2 0:2 ð280=1:13Þ 2
fires. Figure 32.3 compares FDS predictions
(32.38) with Heskestad’s correlation. Note that the
www.ebook777.com
Fig. 32.3 Comparison of SVN 8718

FDS predictions of flame 103
height from a 1 m square Flame Height
pan fire for Q* values
ranging from 0.1 to 10,000 102
Lf/D
101
Heskestad Correlation
100 FDS (D∗/δx = 5)
FDS (D∗/δx = 10)
FDS (D∗/δx = 20)
10−1
10−1 100 101 102 103 104
Q*
simulations were run at three different grid mechanical smoke exhaust, and pressurization
resolutions. A useful way to characterize the of protected spaces. In routine applications, the
grid resolution of a fire simulation is via the design of the smoke control measures may often
ratio D∗/δx, where D∗ (Eq. 32.32) is a measure be achieved by recourse to various guidance
of the effective fire diameter, based on heat publications and empirical correlations [40,
release rate, and δx is the size of a grid cell. In 41]. Network airflow and zone fire models are
effect, D∗/δx is the number of grid cells spanning also available to assist in the design process.
the effective fire diameter. However, where the building space is large or
The fundamental, or “puffing,” frequency is a complex in shape, or where a novel ventilation
quantity that the fire model also ought to predict system is proposed, CFD can be a useful tool.
accurately. Figure 32.4 displays sequential flame Covered shopping malls, atria in hotels and office
images for a single puff from a simulation of a buildings, leisure complexes, airport terminals,
1 m methane fire experiment conducted at and large warehouses are just some examples of
Sandia National Laboratories [38]. The dominant where CFD is being increasingly employed.
puffing mode shows good agreement with the Many of the earliest examples of the applica-
measured puffing frequency of 1.65 Hz. Higher tion of CFD to fire engineering were in smoke
frequency fluctuations from the simulation movement applications [42, 43]. Simulations
exhibit the classic 5/3 scaling of Kolmogorov were at that time restricted to a few tens of
turbulence [39]. thousands of grid cells. Although this number
has increased to hundreds of thousands or even
millions of cells, many of the modeling issues
Smoke Movement remain the same and are discussed later [44].
A significant proportion of fire fatalities can Smoke Transport in a Mechanically

be attributed to the inhalation of smoke parti- Ventilated Library
culates and toxic gases. Furthermore, the effects In collaboration with Olof Granlund Oy, ANSYS
of reduced visibility, high temperature, radiative Europe Ltd. conducted a CFD analysis of the
flux, and oxygen depletion may add to the hazard movement of smoke generated by a fast-growing
associated with the smoke generated by an fire, with a peak heat release rate of 550 kW,
enclosure fire. Means to control the movement inside a library for which the ventilation system
of smoke include physical barriers, natural or was left running in its normal mode of operation.
Fig. 32.4 Snapshots of the flame envelope from a simulation of the Sandia 1 m diameter methane pool fire using
1.5 cm grid resolution. The images span a single “puff”
The building has a pitched, vented glass roof and scenarios within the structure. The building
diffusers at various elevations. ANSYS CFX is a supervisors wished to avoid having to disrupt
general purpose, commercial CFD model that the historical fabric of the rotunda while updating
has a history in fire modeling dating back to the the building’s fire protection systems. The model
investigation of the 1987 Kings Cross fire in was used to examine a number of fire scenarios
London [45]. Because it is designed to handle and how they might impact the ability of
virtually any type of geometry, a model like this occupants to evacuate the building. Note in
one can employ a numerical mesh that conforms Fig. 32.6 the use of rectangular obstructions to
to the unusual shape of the building. In this case, approximate the very complicated geometry of
an unstructured mesh with 3.2 million mesh the building—a simple alternative to the more
elements was used. Figure 32.5 shows a contour CPU intensive body-fitted coordinate system.
map of visibility distance after 3 min in a vertical
slice through the building. Smoke Transport in a Multistory
Residential Building
Smoke Transport in a Historic Landmark To assist means of escape and fire fighter access
CFD fire modeling is commonly used during the in high rise residential buildings it is common
renovation of historic buildings. Often at issue is practice to provide some form of smoke control
the inclusion or exclusion of a fire protection to the stairwells, which could take the form of a
system (sprinklers, exhaust fans, etc.) that might sophisticated pressurization scheme or a rela-
require a variance from the local building code tively simple natural ventilation provision.
requirements. For example, as part of an overall There may be cases where smoke protection is
effort in modernizing the Rhode Island State- required also in the corridors and lobbies at each
house (the rotunda is the fourth largest self- story, possibly as a compensatory measure for
supporting dome in the world), the LES model extended travel distances. Another approach,
FDS was used to model a number of fire adopted in some parts of the world, is to provide
www.ebook777.com
Fig. 32.5 Map of visibility in a vertical plane at time ¼ 3 min for a fire in a library building (Figure courtesy: ANSYS
Europe Ltd)
Fig. 32.6 Smoke filling

analysis of the Rhode
island State capitol
(Figure courtesy: Hughes
Associates)
mechanical ventilation in the corridor. In the and lobbies have a more complicated layout or
event of a fire in an adjoining apartment smoke the ventilation inlets and outlets cannot be
is purged from the corridor, while at the same installed in ideal locations.
time the ventilation system provides smoke pro-
tection to the stairwell by the combined action of Tunnels
pressure differentials and open-door airflows Historically, smoke control inside tunnels was
akin to a stair pressurization system. CFD often considered as an afterthought to design of
modeling of the air and smoke transport in the ventilation for the provision of fresh air. Where
corridor and stairwell is often required in support natural ventilation was insufficient, mechanical
of the design, in particular where the corridors ventilation was designed using network flow
theory aided in recent years by network models. Parking Garages

As a two-layer description is generally not valid The application of CFD to the analysis and
within a tunnel environment, fire zone models design of smoke control systems in basement
have been applied to tunnels in only a limited car parks may share elements that apply also to
number of cases. CFD, however, is now finding road tunnels. Not only is the potential fire source
increasing application to fire hazard analysis for similar, the ventilation strategies and perfor-
road tunnels in particular. In recent years a num- mance criteria may well overlap. Two ventilation
ber of major tunnel fire incidents such as the strategies that might be considered in a basement
tragedy in the Mont Blanc tunnel in 1999, car park include dilution (purging by fresh air)
resulting in 39 deaths, have highlighted the and the directional control of smoke by the appli-
need to understand better the mechanism of fire cation of applied air flows.
development and spread inside tunnels. In partic- While diluting smoke and vehicle emissions
ular, heat transfer to the tunnel walls is important can in principle be achieved by ventilating at a
to account for correctly as this has a strong influ- specified air change rate, in many cases addi-
ence on the distribution of heat along the tunnel, tional ventilation provisions will be required in
the degree of stratification that can be expected, order to ensure an even mixing of fresh air and
and the threat to the integrity of the structure. the elimination of stagnant regions. This is com-
One area where CFD has been applied suc- monly achieved by the strategic location of
cessfully is in the analysis of the critical velocity impulse (jet) fans on the underside of the ceiling,
required in a longitudinally ventilated tunnel to which assist the movement of air from the inlet
control the spread of heat and smoke so that it is points to the exhaust locations. CFD may be
forced in the downstream direction, providing usefully employed to determine the number and
safe conditions upstream [46]. Another area cur- location of these fans.
rently receiving much attention is the choice of The directional control of smoke in the event
design fire for tunnel fire safety design. In the of a fire, with the objective of providing a rela-
light of the recent tunnel fire incidents and full- tively smoke free access to the location of the fire
scale fire tests [47], the size of the fire load that for fire fighting personnel, presents a greater
can be generated from what were previously challenge than simply purging smoke from the
considered as nonhazardous cargoes has been car park. Here careful design of the ventilation
revised. Heat release rates well in excess of system is required, with the impulse fans
100 MW have been measured for heavy-goods operating akin to the case of a longitudinally
vehicles carrying commercial merchandize. ventilated tunnel, directing the smoke and heat
CFD was used in the investigation into the in a direction away from the approaching
Mont Blanc tunnel fire incident in 1999. One of personnel.
the modeling studies involved the use of the
JASMINE fire model to re-construct conditions
inside the tunnel during the first 30 min, during Fire Investigation
which time most of the fatalities would have
occurred. Using information available on the CFD is increasingly used to reconstruct actual
ventilation settings and the location of vehicles, fires, providing fire service personnel and fire
the model predicted the transport of smoke and investigators with a better understanding of the
heat along the length of the tunnel. The data were events that led to injury, loss of life, or loss of the
then fed into a model for fractional effective structure. In any reconstruction, the time line of
dose to enable an assessment of when and how events provided by the first responders and other
the fatalities occurred. Subsequent parametric eyewitnesses is as crucial as the model input, but
simulations were performed to investigate it is also invaluable in assessing the results. Ren-
whether alternative tunnel ventilation measures dering the results of the simulation as realisti-
would have helped on the day of the incident [48]. cally as possible facilitates the synthesis of
www.ebook777.com
Fig. 32.7 Comparison of photographs of the Charleston furniture store fire with a numerical simulation.
Figure courtesy of NIST
model simulation with photographic and visual correlations. If the plume rise is not calculated
evidence. Most people at the fire scene are cer- correctly, substantial errors in downwind con-
tainly not experts in CFD, but they are very centration can result. In the case of smoke-stack
experienced with fire. Examples of how CFD emissions, the plume does not rise appreciably
has been used in actual fire reconstructions are high, reducing the uncertainty of the results.
available [49, 50, 51, 52, 53]. Most of the assumptions required by integral
As an example, NIST researchers used the models can be removed by taking advantage of
Fire Dynamics Simulator to analyze a fire that the advances in CFD over the past few decades.
occurred on the evening of June 18, 2007, in a For example, as part of the process of evaluating
furniture store in Charleston, South Carolina [54]. the feasibility of using in situ burning as a reme-
Using evidence collected at the scene and eye- diation tool for large oil spills, NIST developed a
witness accounts, the investigators put together numerical model, ALOFT (A Large Out-door
a plausible sequence of events that led to the Fire plume Trajectory), to predict the concentra-
deaths of nine fire fighters. Figure 32.7 presents tion of smoke and other combustion products
a snapshot of the numerical simulation compared downwind of a large fire [56]. The model is
to a photograph of the actual fire. simply a variant of the large eddy simulation
model FDS, with a simplified plume rise model
coupled with a coarsely gridded wind calculation
Outdoor Applications and Wind spanning tens of kilometers (Fig. 32.8). This
combination of models is not unusual for outdoor
Buoyant windblown plumes have been studied application, as the range of length scales spans at
since the early 1960s. A summary of the early least three orders of magnitude.
work together with a useful bibliography is given
by Turner [55]. Most of the models described in
these works are integral models, where the Virtual Experiments
profiles of physical quantities in cross-sectional
planes perpendicular to the wind direction are Many codes and standards for fire protection are
assumed, together with simple laws relating based upon simple room geometries. For exam-
entrainment into the plume to macroscopic ple, the spacing for smoke detectors has histori-
features used to describe its evolution. cally been based upon smooth, level ceilings
The potential shortcomings of these types of with some additional rules for beams, slope,
models are that they were designed for typical and height. Under those rules a single story
industrial sources, like smokestacks, that are room with a 30 m by 30 m smooth ceiling
much smaller in terms of energy output than a could be protected by a grid of nine smoke
large fire. The plume from an oil or forest fire detectors, but a ceiling of waffle concrete con-
will rise higher into the atmosphere, and it is struction (Fig. 32.9) with structural deep beams
difficult to predict the rise based on empirical 1 m on center, could, under a strict interpretation
Fig. 32.8 Simulation of smoke from a large oil fire in the Valdez narrows, Alaska (Figure courtesy of NIST)
Fig. 32.9 Simulation of smoke filling under a coffered ceiling (Figure courtesy: Aon Fire Protection Engineering
Corp.)
of the guidelines, require 900 detectors, one in costly full-scale experiments, a small set of
each beam pocket. While this is obviously unrea- full-scale experiments was combined with a
sonable, making a change to the building code large set of “virtual” experiments done with
requires evidence. In lieu of a large number of CFD [57, 58]. The researchers evaluated the
www.ebook777.com
appropriateness of the prescriptive provisions that have traditionally been applied in fire pro-
and identified ceiling structure parameters, tection engineering.
which if altered, would cause significant It is as a parametric design tool that CFD is
differences in smoke detection performance as often most gainfully employed, allowing the
compared to a smooth level ceiling. It had been impact of varying the input and boundary
believed that such beam projections would sig- conditions to be examined. For example, how
nificantly delay the activation time. The use of sensitive is the smoke control solution to the
CFD modeling showed this expectation to be design fire size, or how much influence does a
incorrect and a subsequent full-scale experimen- change in wind direction have on a natural smoke
tal study proved the general findings of the CFD ventilation strategy? The reader is encouraged to
analysis [59]. The final result of the study led to consult the guidance documents available on the
an exception, under some circumstances, to the best practice use of CFD in the various applica-
code requirement of a smoke detector in every tion areas and to consult the guidance documen-
beam pocket. tation provided with the CFD model being
employed. For example, the guidance document
prepared by the US Nuclear Regulatory Commis-
The Role of CFD in the Design Process sion and the Electrical Power Research Institute
(EPRI) on fire modeling for nuclear power
As discussed elsewhere in this chapter, CFD has plants [60] includes useful information on the
an ever increasing role to play in the develop- appropriate role and application of CFD in the
ment of fire safety science, and has an important fire safety design process, and is relevant also to
contribution to make in better understanding the fire scenarios outside the nuclear field.
fundamentals such as flame spread and chemical
species production where it is being used in par-
allel with physical experiments. However, it is as Summary
a fire engineering design tool that CFD is proba-
bly of most relevance to the majority of readers. Computational fluid dynamics modeling of fire
Here a few words of caution are worth noting. has made tremendous progress over the past few
CFD modeling has a useful, and sometimes decades as our understanding of fire improves
critical, role in developing safe and robust fire and as computers get ever faster. However,
engineering solutions where the control of smoke although it appears to many that CFD is the
and heat generated by fire forms part of the fire cutting edge of fire protection engineering,
safety strategy. It allows architectural designs to many non-modelers are surprised to learn that
be adopted that in previous eras would have been our ability to reproduce fire phenomena via com-
difficult to justify, for example in respect to puter simulation lags our empirical understand-
smoke control in large and complicated building ing by about 10 years. Indeed, current-generation
atria or where a reduced level of structural fire models address transport phenomena reasonably
protection is desired. It should, however, be seen well, making them useful for many engineering
as a contributing component to the overall design applications. However, they have not yet reached
process, and not as a “black box” that faithfully the point of reliably predicting, for large-scale
provides the correct answers regardless of the applications, such important phenomena as flame
inputs and assumptions made. A great deal of spread, extinction, suppression, and CO and
care and experience is required in order to sensi- smoke production, all of which demand more
bly use CFD in support of fire engineering detailed chemistry and physics than are currently
designs, and it should be employed alongside incorporated in the models.
simpler calculation and design methods wher- Moving forward will require a new generation
ever possible to confirm that the CFD results of engineers who have expertise in fire physics,
are comparable to the empirical correlations mathematics, and computer science to build on
the knowledge possessed by the current genera- κ Radiation absorption coefficient

tion of modelers. This chapter sets forth the basic λ Generalized diffusion coefficient
elements to lay the foundation of further study μ Dynamic viscosity
for future modelers, and it also provides the μt Turbulent viscosity (in eddy viscosity turbu-
current practitioner with a better understanding lence model)
of the models being used. ρ Density
σ Stefan-Boltzmann constant
τ Stress tensor
Nomenclature
cp Specific heat
Dα Material diffusivity of species α Appendix
f External body force
h Sensible enthalpy Much of the difficulty in learning and applying
hc Convective heat transfer coef- computational fluid dynamics is the complexity
ficient of the governing equations. In this appendix,
I Radiant intensity some of the common terms found in the mass,
k Thermal conductivity; turbulent momentum, and energy equations are expanded.
kinetic energy Many of the variables and operators can be
m_ 000
α Mass production (destruction) represented as 3 3, 1 3, or 3 1 matrices,
rate of species α per unit volume and the expansions can be carried out following
p Pressure the rules of linear algebra. For example, the
Pr Prandtl number divergence of the flow vector, ∇ u, is a scalar
q Heat flux vector formed by multiplying the 1 3 gradient opera-
q_ 00 Heat release rate per unit area tor ∇ and the 3 1 vector u. On the other hand,
q_ 000 Heat release rate per unit the product of the velocity vectors, u u, is found
volume by multiplying a 3 1 vector by a 1 3 vector:
R Universal gas constant 0 1 0 2 1
s Direction vector u u uv uw
B C B C
s Stoichiometric air requirement uu ¼ @ v Aðu; v; wÞ ¼ @ vu v2 vw A
of the fuel w wu wv w2
Si j Strain tensor
(32.39)
Sc Schmidt number
t Time
Thus, the convection term in the momentum
T Temperature
conservation equation can be expanded as
u ¼ (u, v, w) Velocity vector
follows:
W Average molecular weight
0 1
x ¼ (x, y, z) Position vector ρu2 ρuv ρuw

Yα Mass fraction of species α @ @ @ B B ρvu ρv2 ρvw C
C
r ðρuuÞ ¼
@x @y @z @ A
ρwu ρwv ρw 2
0 1T
Greek Letters ðρu Þx þ ðρvuÞy þ ðρwuÞz
2
B C
B C
¼ B ðρuvÞx þ ðρv2 Þy þ ðρwvÞz C
ε Rate of dissipation of turbulent kinetic @ A
energy ðρuwÞx þ ðρvwÞy þ ðρw2 Þz
(32.40)
www.ebook777.com
The result is a vector whose components form the 2. D.A. Anderson, J.C. Tannehill, and R.H. Pletcher.
convective terms of the three component Computational Fluid Mechanics and Heat Transfer.
Hemisphere Publishing Corporation, Philadelphia,
momentum equation. Note that here the PA, 1984.
subscripts x, y, and z denote partial derivatives 3. J.H. Ferziger and M. Peric. Computational Methods
with respect to that particular coordinate for Fluid Dynamics. Springer-Verlag, Berlin, second
direction. edition, 1999.
4. S.V. Patankar. Numerical Heat Transfer and Fluid
The term for the viscosity in the momentum Flow. Hemisphere Publishing, New York, 1980.
equation, ∇ τ, is deceptively simple. In reality, 5. R. Peyret and T.D. Taylor. Computational Methods
it is not, and because it constitutes the heart of the for Fluid Flow. Springer-Verlag, New York, 1983.
debate over turbulence models, some attention 6. K. Versteeg and W. Malalasekera. An Introduction to
Computational Fluid Dynamics – The Finite Volume
must be paid to it. Using customary tensor nota- Method. Longmann, Essex, UK, 1995.
tion, the viscous stress tensor is defined as 7. G. Cox. Combustion Fundamentals of Fire, chapter
Compartment Fire Modelling. Academic Press,
@ui @uj 2 London, 1995.
τij ¼ μ þ δij r u ;
@xj @xi 3 8. G. Cox. Turbulent closure and the modelling of
( (32.41) fire using computational fluid dynamics.
1 if i ¼ j Phil. Trans. R. Soc. Lond. A, 356:2835–2854, 1998.
δij ¼ 9. V. Novozhilov. Computational Fluid Dynamics
0 if i 6¼ j Modeling of Compartment Fires. Progress in Energy
and Combustion Science, 27:611–666, 2001.
These expressions assert that the viscous stresses 10. S. Olenick and D. Carpenter. An Updated Interna-
are linearly related to the strains, the very defini- tional Survey of Computer Models for Fire and
Smoke. Journal of Fire Protection Engineering,
tion of a Newtonian fluid. The proportionality 13:87–110, May 2003.
constant, μ, is called the dynamic viscosity of 11. R.G. Rehm and H.R. Baum. The Equations of Motion
the fluid. The viscous stress tensor can also be for Thermally-Driven, Buoyant Flows. Journal of
represented as a 3 3 matrix: Research of the National Bureau of Standards,
83:297–308, 1978.
0 1 0 1 12. V. Yakhot, S.A. Orszag, S. Thangam, T.B. Gatski, and
2ux uy þ vx uz þ wx ru 0 0
@ A 2@
τ ¼ μ vx þ uy 2 vy vz þ wy 0 ru 0 A C.G. Speziale. Development of Turbulence Models
wx þ uz wy þ vz 3 for Shear Flows by a Double Expansion Technique.
2wz 0 0 ru
Physics of Fluids A, 4:1510–1520, 1992.
(32.42) 13. D.C. Wilcox. Turbulence Modeling for CFD. DCW
Industries, La Cañada, CA, third edition, 2006.
The dissipation function, ε, is a scalar formed by 14. J. Smagorinsky. General Circulation Experiments
the dot product of two 3 3 matrices: with the Primitive Equations. I. The Basic Experi-
ment. Monthly Weather Review, 91(3):99–164, 1963.

15. S.B. Pope. Ten Questions Concerning the Large-Eddy
ε τ ru ¼ μ 2u2x þ 2v2y þ 2w2z Simulation of Turbulent Flows. New Journal of Phys-
ics, 6:1–24, 2004.
2 16. P.R. Spalart, W.H Jou, M. Stretlets, and
þ ðvx þ uy Þ2 þ ðwy þ vz Þ2 þ ðuz þ wx Þ2 ðr uÞ2
3 S.R. Allmaras. Comments on the Feasibility of LES
(32.43) for Wings and on the Hybrid RANS/LES Approach.
In Proceedings of the First AFOSR International Con-
ference on DNS/LES, Louisiana Tech University,
1997. Air Force Office of Aerospace Research.
17. D.B. Spalding. Mixing and Chemical Reaction in
Steady State Confined Turbulent Flames. In 13th Sym-
References posium (International) on Combusion, pages
649–657, Pittsburgh, PA, 1971. The Combustion
Institute.
1. P.G. Drazin, editor. Collected Papers of LF
18. B.F. Magnussen and B.H. Hjertager. On Mathemati-
Richardson, Volume 1: Meteorology and Numerical
cal Modelling of Turbulent Combustion with Special
Analysis. Cambridge University Press, Cambridge,
Emphasis on Soot Formation and Combustion. In 16th
UK, 1993.
Symposium (International) on Combustion, pages 32. S. Welch, S. Miles, S. Kumar, T. Lemaire, and
719–729, Pittsburgh, PA, 1976. The Combustion A. Chan. FIRESTRUC – Integrating Advanced
Institute. Three-dimensional Modelling Methodologies for
19. S.D. Miles, S. Kumar, and G. Cox. Comparison of Predicting Thermo-mechanical Behaviour of Steel
‘Blind Predictions’ of a CFD Model with Experimen- and Composite Structures Subjected to Natural Fires.
tal Data. In Proceedings of the 6th International Sym- In Proceedings of the 9th International Symposium on
posium on Fire Safety Science, pages 543–554. Fire Safety Science, pages 1315–1326, Karlsruhe,
International Association of Fire Safety Science, Germany, September 2008. International Association
2000. of Fire Safety Science.
20. J. Holen, M. Brostrom, and B.F. Magnussen. Finite 33. D.B. Spalding. A Novel Finite Difference Formula-
Difference Calculation of Pool Fires. In 23rd Sympo- tion for Differential Expressions Involving Both First
sium (International) on Combustion, pages and Second Derivatives. International Journal for
1677–1683, Pittsburgh, PA, 1990. The Combustion Numerical Methods in Engineering, 4:551–559, 1972.
Institute. 34. Society of Fire Protection Engineers, Bethesda,
21. N. Peters. Laminar Flamelet Concepts in Turbulent Maryland. Guidelines for Substantiating a Fire
Combustion. In 21st Symposium (International) on Model for a Given Application, 2011.
Combustion, pages 1231–1250, Pittsburgh, PA, 35. K. Hill, J. Dreisbach, F. Joglar, B. Najafi,
1986. The Combustion Institute. K. McGrattan, R. Peacock, and A. Hamins. Verifica-
22. P.A. Tesner, T.D. Snegirova, and V.G. Knorre. Kinet- tion and Validation of Selected Fire Models for
ics of Dispersed Carbon Formation. Combustion and Nuclear Power Plant Applications. NUREG 1824,
Flame, 17:253–260, 1971. United States Nuclear Regulatory Commission,
23. R. Siegel and J.R. Howell. Thermal Radiation and Washington, D.C., 2007.
Heat Transfer. Taylor and Francis, New York, fourth 36. G. Heskestad. Luminous Heights of Turbulent Diffu-
edition, 2002. sion Flames. Fire Safety Journal, 5:103–108, 1983.
24. S. Hostikka and K.B. McGrattan. Numerical 37. K. McGrattan and B. Toman. Quantifying the predic-
Modeling of Radiative Heat Transfer in Water Sprays. tive uncertainty of complex numerical models.
Fire Safety Journal, 41:76–86, 2006. Metrologia, 48:173–180, 2011.
25. C. DiBlasi. Modeling and Simulation of Combustion 38. S. R. Tieszen, T. J. O’Hern, R. W. Schefer, E. J.
Processes of Charring and Non-Charring Solid Fuels. Weckman, and T. K. Blanchat. Experimental study
Progress in Energy and Combustion Science, of the flow field in and around a one meter diameter
19:71–104, 1993. methane fire. Combustion and Flame, 129:378–391,
26. B.E. Launder and D.B. Spalding. The Numerical 2002.
Computation of Turbulent Flows. Computer Methods 39. Stephen B. Pope. Turbulent Flows. Cambridge Uni-
in Applied Mechanics and Engineering, 3:269–289, versity Press, 2000.
1974. 40. J.H. Klote and J.A. Milke. Principles of Smoke Man-
27. T. Jin. SFPE Handbook of Fire Protection Engineer- agement. American Society of Heating, Refrigerating
ing, chapter Visibility and Human Behaviour in Fire and Air Conditioning Engineers (ASHRAE), Atlanta,
Smoke. National Fire Protection Association, Quincy, GA, 2002.
MA, fourth edition, 2008. 41. National Fire Protection Association, Quincy,
28. G.W. Mulholland. SFPE Handbook of Fire Protection MA. NFPA 92B, Standard for Smoke Management
Engineering, chapter Smoke Production and Systems in Malls, Atria and Large Spaces, 2005.
Properties. National Fire Protection Association, 42. K.A. Pericleous, D.R.E. Worthington, and G. Cox.
Quincy, MA, fourth edition, 2008. The Field Modelling of Fire in an Air-Supported
29. B.P. Husted, J. Carlsson, and U. Goransonn. Visibility Structure. In Proceedings of the 2nd International
Through Inhomogeneous Smoke Using CFD. In Symposium on Fire Safety Science, pages 871–880,
Proceedings of Interflam 2004, pages 697–702, Tokyo, 1988. Hemisphere Publishing Corporation.
Edinburgh, 2004. 43. G. Cox, S. Kumar, P. Cumber, V. Thomson, and
30. G. Heskestad and R.G. Bill. Quantification of Thermal A. Porter. Fire Simulation in the Design Evaluation
Responsiveness of Automatic Sprinklers Including Process: An Exemplification of the Use of a Computer
Conduction Effects. Fire Safety Journal, Field Model. In Proceedings of the 5th Interflam Con-
14:113–125, 1988. ference, pages 55–66, Canterbury, UK, 1990.
31. G.M. Makhviladze, J.P. Roberts, O.I. Melikhov, and 44. S. Kumar and G. Cox. Some Guidance on Correct Use
V.I. Melikhov. Numerical Simulation of Sprinkler of CFD Models for Fire Applications with Examples.
Jet-Fire Interaction for Compartment Fires. In In Proceedings of Interflam 2001, pages 823–834,
Proceedings of the 2nd International Seminar on Edinburgh, 2001.
Fire and Explosion Hazard of Substances and Venting 45. S. Simcox and N.S. Wilkes. Computer Simulation of
of Deflagrations, pages 485–496, Moscow, August the Flows of Hot Gases from the Fire at King’s Cross
1997. All-Russian Institute for Fire Proection. Underground Station. Fire Safety Journal, 18:49–73,
1992.
www.ebook777.com
46. C.C. Hwang and J.C. Edwards. The Critical Ventila- 55. J.S. Turner. Buoyancy Effects in Fluids. Cambridge
tion Velocity in Tunnel Fires – A Computer Simula- University Press, Cambridge, UK, 1973.
tion. Fire Safety Journal, 40:213–244, 2005. 56. H.R. Baum, K.B. McGrattan, and R.G. Rehm. Simu-
47. H. Ingason and A. Lönnermark. Heat Release Rates lation of Smoke Plumes from Large Pool Fires. In
from Heavy Goods Vehicle Trailer Fires in Tunnels. Proceedings of the 25th (International) Symposium
Fire Safety Journal, 40:646–668, 2005. on Combustion, pages 1463–1469, Pittsburgh, PA,
48. S. Miles and S. Kumar. Computer Modelling to 1994. The Combustion Institute.
Assess the Benefits of Tunnel Sprinkler and Ventila- 57. D.J. O’Connor, E. Cui, M.J. Klaus, C.H. Lee, C. Su,
tion Fire Safety Measures. In Proceedings of 5th Z. Sun, M. He, Y. Jiang, J. Vythoulkas, and T. Al-Farra.
International Conference on Tunnel Fires, pages Smoke Detector Performance for Level Ceilings with
23–32, London, 2004. Tunnel Management Deep Beams and Deep Pocket Configurations. Fire
International. Protection Research Foundation report, National Fire
49. W.L. Grosshandler, N. Bryner, D. Madrzykowski, and Protection Association, Quincy, MA, 2006.
K. Kuntz. Report of the Technical Investigation of 58. C. Mealy, J. Floyd, D. Gottuk, and S. Riahi. Smoke
The Station Nightclub Fire. NIST NCSTAR 2, Detector Spacing Requirements for Complex Beamed
National Institute of Standards and Technology, and Sloped Ceilings. Fire Protection Research Foun-
Gaithersburg, MD, 2005. dation report, National Fire Protection Association,
50. K.B. McGrattan, C. Bouldin, and G.P. Forney. Fed- Quincy, MA, 2008.
eral Building and Fire Safety Investigation of the 59. D. Gottuk, C. Mealy, and J. Floyd. Smoke Transport
World Trade Center Disaster: Computer Simulation and FDS Validation. In Fire Safety Science –
of the Fires in the WTC Towers. NIST NCSTAR Proceedings of the 9th International Symposium,
1-5F, National Institute of Standards and Technology, pages 129–140, Karlsruhe, Germany, September
Gaithersburg, MD, 2005. 2008. International Association of Fire Safety Science.
51. D. Madrzykowski and R.L. Vettori. Simulation of the 60. D. Stroup and R. Wachowiak. Nuclear Power Plant
Dynamics of the Fire at 3146 Cherry Road NE, Fire Modeling Analysis Guidelines. NUREG 1934,
Washington, D.C., May 30, 1999. NISTIR 6510, United States Nuclear Regulatory Commission,
National Institute of Standards and Technology, Washington, D.C., 2012.
Gaithersburg, MD, 2000.
52. D. Madrzykowski and W.D. Walton. Cook County
Administration Building Fire: Heat Release Rate Kevin McGrattan is a mathematician in the Fire
Experiments and FDS Simulations. NIST Special Research Division at the National Institute of Standards
Publication 1021, National Institute of Standards and and Technology in Gaithersburg, Maryland. He is the
Technology, Gaithersburg, MD, 2004. principal developer of the Fire Dynamics Simulator
53. A.M. Christensen and D.J. Icove. The Application of (FDS)
NIST’s Fire Dynamics Simulator to the Investigation
of Carbon Monoxide Exposure in the Deaths of Three Stewart Miles, qualified originally as a physicist, is
Pittsburgh Fire Fighters. Journal of Forensic currently a practicing fire engineer at International Fire
Sciences, 49(1):1–4, 2004. Consultants Ltd. in the UK. Previously he worked in fire
54. N.P. Bryner, S.P. Fuss, B.W. Klein, and A.D. Putorti. research and engineering at the UK Building Research
Technical Study of the Sofa Super Store Fire - South Establishment, where he contributed to the development
Carolina, June 18, 2007. NIST Special Publication of the CFD fire model JASMINE
1118, National Institute of Standards and Technology,
Gaithersburg, MD, 2011.
Enclosure Smoke Filling

and Fire-Generated Environmental 33
Conditions
Frederick W. Mowrer
Introduction is addressed by Boyce and Gwynne (see

Chap. 64) and Gwynne and Rosenbaum
Fires in buildings and other structures are distin- (Chap. 59). The available safe egress time is
guished from outdoor fires by the confinement addressed in this chapter in terms of the time it
effects associated with enclosure boundaries takes for the smoke layer to descend and
and by the ventilation effects associated with immerse people located within the fire enclosure
openings in these boundaries. The confinement and in terms of the hazards associated with fire-
of heat and smoke released by a fire in an enclo- generated conditions within the smoke layer.
sure gives rise to the evolution of fire-generated The control volume, or zone modeling,
environmental conditions that can threaten approach presented by Wade (Chap. 29) is used
life safety and cause thermal and nonthermal as the basis for the analyses presented here. A
damage to the structure and its contents. For number of explicit equations for evaluating the
performance-based building fire safety analysis smoke layer interface position and the average
and design, it is important to be able to calculate conditions within a smoke layer are presented in
the environmental conditions generated by fires this chapter for certain idealized fire scenarios.
in enclosures in order to evaluate the threat levels These closed-form equations, sometimes called
posed by anticipated fire scenarios. This chapter “hand calculations” because they can be solved
addresses the enclosure smoke-filling process without the aid of a computer, are useful for
and the fire-generated environmental conditions estimating smoke layer interface position and
that develop within an enclosure during this average smoke layer conditions for the range of
process. applications for which these equations are valid.
The concept of available safe egress time For more detailed analyses and for scenarios
(ASET) has become a fundamental aspect of where hand calculations are impractical or not
performance-based analysis of life safety from valid, use of either a computer-based zone model
fire. In general, life safety from fire is achieved or a computational fluid dynamics (CFD) model
if the required safe egress time (RSET) is shorter may be warranted to evaluate fire-generated
than the available safe egress time (i.e., RSET < conditions in an enclosure. Such computer-
ASET) for the range of expected fire scenarios. based models will generally be needed to
The time it takes to evacuate a space, the RSET, evaluate multiroom fire scenarios and may be
preferred to evaluate single room scenarios.
The concepts presented here are relevant to the
F.W. Mowrer (*)
computer-based zone fire models, but the
Director of Fire Protection Engineering Programs,
California Polytechnic State University, San Luis Obispo, complexities associated with keeping track of
CA, USA multiroom fire scenarios are not addressed.

www.ebook777.com
33 Enclosure Smoke Filling and Fire-Generated Environmental Conditions 1067
Computer-based zone models are addressed by for Smoke Management Systems in Malls, Atria
Walton, Carpenter, and Wood (see Chap. 31), and Large Spaces [10] and is discussed further
whereas computer-based CFD models are by Milke (see Chap. 51). More recently, Mowrer
addressed by McGrattan and Miles (see [11] has revisited the enclosure smoke-filling
Chap. 32). process, recasting its formulation in terms of the
volumetric flow rates generally used for ventila-
tion system design. Mowrer [12] has also
Background addressed the role of mechanical ventilation on
smoke filling and management in terms of these
The first efforts to characterize the enclosure volumetric flow rates. Matsuyama et al. [13] and
smoke-filling process and the environmental Delichatsios [14] have developed closed-form
conditions generated by a fire in a closed room solutions for enclosure smoke filling, whereas
can be traced to the late 1970s and early 1980s. Delichatsios [15] has also addressed tenability
The seminal paper on this topic was published in conditions and filling times for fires in large
1978 by Zukoski [1], who applied thermody- spaces.
namic control volume concepts to evaluate
mass and energy balances within a closed room
subjected to a fire. Shortly thereafter, Cooper Stages of Enclosure Fires
applied Zukoski’s concepts to develop the avail-
able safe egress time (ASET) model [2, 3], a Enclosure fires go through a series of stages that
computer-based fire model designed to calculate depend on the size and shape of the enclosure,
the evolution of the descending smoke layer the thermal properties of the boundary materials,
interface position and the average temperature the sizes and locations of ventilation pathways
and smoke concentration conditions within the through the enclosure boundaries, and the devel-
smoke layer in response to specified fires. During opment of the fire. Mowrer [11] has identified the
the early 1980s, Walton [4] converted the origi- four stages of enclosure fires as
nal ASET model from FORTRAN to BASIC • Fire plume/ceiling jet stage
and simplified the numerical methods used in • Enclosure smoke filling stage
the model to allow its convenient application • Preflashover vented stage
on the desktop personal computers that were • Postflashover vented stage
just then starting to be used in engineering prac-
tice; this version of the model was known as
ASET-B. Since its original development, various Fire Plume/Ceiling Jet Stage
versions of the ASET model have been
incorporated into different fire modeling suites, During the first stage of an enclosure fire, the fire
such as FPETOOL [5]. Hurley [6] has compared plume/ceiling jet stage, air is entrained into the
ASET-B model predictions with large-scale flame region, where it mixes with fuel being
experimental test data. released from the fuel surface and burns, typi-
During the early 1990s, Mowrer [7] addressed cally in a nonpremixed (diffusion) flame. The
the evolution of fire and smoke conditions in a energy released by the combustion reaction
closed room as part of the development of the causes the temperature of the combustion
FIVE Methodology [8] (see Chap. 89) for products to increase and their density to
evaluating fire-induced vulnerabilities in com- decrease. Because these combustion products
mercial nuclear power plants. Subsequently, are less dense than the surrounding air, they rise
Milke and Mowrer [9] expanded this analysis through the surrounding air in a buoyant coherent
for application to smoke management systems stream known as the fire plume, as shown in
in atria and covered malls. This application has Fig. 33.1. As the buoyant gases rise in the fire
been incorporated into the NFPA 92B, Standard plume, additional air is entrained into the fire
1068 F.W. Mowrer
plume, causing the temperature and smoke con- relatively thin layer known as the ceiling jet, as
centration within the plume to decrease while shown in Fig. 33.1. These gases continue to
causing the volume of smoke, which is defined spread radially beneath the ceiling until they are
to include the actual combustion products as well confined by the enclosing walls of the fire room.
as the entrained air, to increase with increasing The impingement of the plume at the ceiling and
height above the fire source. Phenomena the confinement of flow beneath the ceiling
associated with fire plumes are addressed in constitute the first significant distinction between
more detail by Heskestad (see Chap. 13). an enclosure fire and an outdoor fire. Ceiling
When the fire plume impinges on a smooth jets are addressed in more detail by Alpert
horizontal ceiling, the buoyant gases turn and (see Chap. 14).
spread out radially beneath the ceiling in a
Enclosure Smoke-Filling Stage
Once the ceiling jet reaches the wall boundaries,

the smoky gases in the ceiling jet turn downward
and begin to accumulate beneath the ceiling, as
shown in Fig. 33.2. This begins the second stage
of enclosure fires, the smoke-filling stage. During
the smoke-filling stage, smoke is injected via the
fire plume into the developing smoke layer,
where the buoyant smoke accumulates beneath
the ceiling. The smoke layer interface descends
within the enclosure as a result of continued
smoke injection via the plume.
If no sizable vents are available in the enclo-
Fig. 33.1 The fire plume/ceiling jet stage of an sure boundaries, the smoke layer will continue
enclosure fire
Fig. 33.2 The smoke-

filling stage of an
enclosure fire
www.ebook777.com
to descend until it reaches either the elevation enclosure fires, the preflashover vented stage.
of the fire source or the floor. Enclosure During this stage, the smoke layer will descend
boundaries in buildings are generally leaky to the elevation where the flow of air into the fire
enough to prevent significant pressure increases room is balanced by the flow of smoke out of the
as a result of the gas expansion associated with fire room. In many cases, both the inflow of air
enclosure fires. In a closed room, the smoke and the outflow of smoke will be through the
layer may descend to the level of the fire and same wall vent, as illustrated in Fig. 33.3. Vent
act to suppress the fire due to oxygen depletion flows are addressed in more detail by Tanaka (see
within the smoke layer, much as a candle flame Chap. 15).
will extinguish when placed beneath an inverted
jar. On the other hand, if a vent is opened, such
as a window that breaks from heat-induced Postflashover Vented Stage
stresses or a door that is opened by fire fighters,
such a fire may rapidly redevelop due to the The fourth stage of enclosure fires, the
influx of fresh air. With an influx of fresh air, postflashover vented stage, occurs if the fire
a backdraft explosion [16] may occur if suffi- intensifies to the point where the smoke layer
cient unburned fuel vapors have accumulated reaches a temperature sufficient to cause the
within the room. radiant ignition of exposed combustible surfaces
within the fire enclosure, as illustrated in
Fig. 33.4. Either an average smoke layer temper-
Preflashover Vented Stage ature of approximately 600 C or an incident heat
flux of approximately 20 kW/m [2] at floor level
If one or more open wall vents are provided from is often used as an indication of the onset of
the fire space, such as a window to the outside or flashover. Methods for estimating the heat
a doorway to an adjacent space, then the smoke release rates necessary to cause flashover and
will flow from the enclosure into the adjacent for estimating smoke layer temperatures
space once the smoke layer descends to the resulting from pre- and postflashover vented
level where a wall vent is available, as illustrated fires are addressed by Walton, Thomas and
in Fig. 33.3. This begins the third stage of Ohmiya (see Chap. 30).
Fig. 33.3 The preflashover vented stage of an enclosure fire

1070 F.W. Mowrer
T ~ 600°C
Fig. 33.4 The postflashover vented stage of an enclosure fire
Fig. 33.5 Control

me , Vexp
volumes and phenomena
ρu, Tu, Vu (Case 2)
associated with enclosure zu Smoke
Vpl + Vexp layer
smoke filling
mpl
H Plume
mpl , Vpl
zl
Qf
Lower
zf ρl, Tl, Vl layer
me , Vexp
(Case 1)
expansion of gases within the enclosure. Of

Phenomena Associated with Modeling the heat released by the fire, a fraction, χr, is
of Enclosure Smoke Filling radiated away from the combustion zone, while
the remaining fraction, χc ¼ 1 – χr, is convected
Phenomena associated with the modeling of in the buoyant plume that rises from the fire
enclosure smoke filling are described in this source to the ceiling.
section. The basic phenomena associated with The plume entrains surrounding air as it rises
this stage of enclosure fires are illustrated in through the atmosphere. Combustion products
idealized form in Fig. 33.5. A fire located at and entrained air are transported along with
some arbitrary elevation, zf, above the floor of a convected heat to the ceiling, where the plume
room is represented as a point source of heat turns to form a ceiling jet that spreads radially
addition, Q_ f , to the space. A fraction, χl, of the beneath the ceiling. When the ceiling jet reaches
heat released by the fire is lost by heat transfer to the wall boundaries of the enclosure, it is
the boundaries of the enclosure or to other deflected downward. For purposes of modeling
surfaces within the enclosure, while the enclosure smoke filling, it is common to neglect
remaining fraction, (1 – χl), causes heating and the ceiling jet altogether and to assume that the
www.ebook777.com
enclosure begins to fill uniformly with smoke to have uniform conditions throughout. The
from the ceiling down due to the injection of primary purpose of this global analysis is to
smoke into the smoke layer via the fire plume. address pressure effects, but this global analysis
This is the approach taken here. is also instructive with respect to temperature and
The developing smoke layer is normally smoke concentration effects, providing a basis
treated as a distinct control volume with uniform for comparison of the more detailed smoke
properties for zone modeling purposes. As a layer descent analyses. This global analysis is
modeling idealization, the upper and lower followed by two smoke layer descent analyses,
layer control volumes are assumed to be designated as Cases 1 and 2, that address the
separated by a distinct thermal discontinuity at descending smoke layer explicitly in terms of
the interface between the two layers. This inter- upper and lower layer control volumes. Leakage
face is known as the smoke layer interface. The paths are assumed to be at floor level only in
smoke layer descends within the enclosure due to Case 1 and at ceiling level only in Case 2, as
the entrainment of fresh air from the lower layer illustrated in Fig. 33.5. Case 1 is the scenario
into the fire plume and, depending on the location addressed by the ASET model [2]. As shown in
of leakage paths from the enclosure to the subsequent analysis of these two cases,
surrounding spaces, the expansion of heated the location of leakage paths does not have a
gases in the upper layer. Mechanical ventilation large influence on smoke layer development or
will also influence the development of the smoke conditions.
layer and the conditions within the smoke layer; Once the smoke layer descends to the eleva-
it is not included in this discussion of general tion of the fire source, the fire source becomes
phenomena, but the influence of mechanical ven- immersed in the smoke layer, and further entrain-
tilation on the development of the smoke layer ment of fresh air from the lower layer is assumed
and the conditions within the smoke layer is to cease. After this time, the fire will entrain and
addressed in a subsequent section of this chapter. recirculate smoke from within the smoke layer,
The expansion of gases within the enclosure as illustrated in Fig. 33.6, and the smoke layer
due to heat addition pressurizes the enclosure will continue to descend due only to gas expan-
relative to adjacent spaces and forces the sion. Because fresh air is no longer being
flow of gases from the enclosure through avail- entrained into the fire, the intensity of the fire
able leakage paths. Three different cases are will eventually diminish due to oxygen depletion
addressed to consider these pressure effects for within the smoke layer.
different leakage flow cases. First, a global anal- As a fire in a closed compartment diminishes
ysis is presented where the entire enclosure is due to oxygen depletion, the rate of heat losses to
treated as a single, fixed control volume assumed enclosure boundaries will become greater than
Fig. 33.6 Smoke

me , Vexp
recirculation associated
(Case 2)
with smoke layer descent zu Vexp
to fuel surface
H Smoke
ρu, Tu, Vu layer
zl
Qf
ρl, Tl, Vl Lower

zf layer me , Vexp
(Case 1)
1072 F.W. Mowrer
the rate of heat addition due to the fire. The Q_ f , and the rate of heat loss, Q_ l , to boundaries
smoke will cool and contract as a consequence, and other solid surfaces, such as equipment
causing the fire enclosure to depressurize relative located within the space.
to adjacent spaces. This depressurization will Many fire models calculate boundary
draw air into the enclosure from surrounding heat losses explicitly, usually in terms of
spaces, which in turn may allow the fire to one-dimensional heat transfer through a slab.
reintensify and repressurize the enclosure. This For the present discussion, a constant heat loss
cycle of depressurization and repressurization, fraction, χl, is used to represent boundary heat
sometimes called puffing behavior, can repeat losses, such that
indefinitely and is one of the warning signs of
an underventilated fire, which may result in a Q_ net ¼ Q_ f Q_ l ¼ Q_ f ð1 χ l Þ ð33:2Þ
backdraft if a large ventilation opening is sud-
denly provided in an enclosure boundary. This is the approach taken by Cooper [2, 3],
who suggests values for χl in the range of 0.6–0.9
for most situations. Cooper suggests that values
Global (One-Zone) Analysis near the low end of this range are appropriate for
spaces with smooth ceilings and large ceiling
In this section, the entire enclosure gas volume is area to height (A/H2) ratios. Values near the
treated as a fixed control volume, as shown in high end of the range would be appropriate for
Fig. 33.7. Zukoski [1] addressed the pressure rise spaces with irregular ceiling shapes, with small
that would occur in both sealed and leaky ceiling area to height ratios, or where fires are
enclosures by considering this global control vol- located near walls. Mowrer [7] found that a value
ume. In both the sealed and leaky cases, a general of 0.7 for the heat loss fraction provided good
energy balance for the enclosure control volume agreement with experimental temperature data
can be written as for a series of fire tests [17] conducted in a
room with a floor area of 223 m2, a smooth
dU dV
¼ Q_ net þ m_ i hi m_ o ho P ð33:1Þ ceiling, an aspect ratio (A/H2) of 6, and a dimen-
dt dt
sionless heat release rate, Q*, defined as
where U is the total internal energy in the control pffiffiffiffiffiffi
volume and Q_ net is the net rate of heat addition Q* Q=_ ρ c p T a gH H 2
a
into the space; it is equal to the difference
between the actual heat release rate of the fire, of approximately 1.7 104.
Fig. 33.7 Single control

(Sealed
volume and leakage flows
case)
used for global analysis
H – zf
ρg, Tg, V
H
Qf
zf me , Vexp
(Leaky
case)
www.ebook777.com
It should be recognized that temperature and constant specific heat with properties of air,
predictions are sensitive to the selection of the for the sealed compartment Equation 33.1
heat loss fraction when using this approach. reduces to
Because the heat retained in the gas volume is
proportional to (1 – χl), a seemingly small change du dT cv V dP
ρV ¼ ρcv V ¼ ¼ Q_ net ð33:3Þ
in the heat loss fraction from 0.9 to 0.8 represents dt dt R dt
a twofold difference in the net heat addition term. Through manipulation of the ideal gas law,
the normalized pressure and temperature rise in
a sealed compartment subject to a net change of
Sealed Compartment energy, but without a change in mass or molar
quantity, can be expressed as
Neglecting fuel vapor addition to the enclosure
control volume associated with the fire, the mass ðt
flow rates into and out of the enclosure are Q_ net dt
assumed to be nil for the sealed compartment. ΔP ΔT Q
¼ ¼ o ¼ net ð33:4Þ
The volume of the compartment does not change Po T o ρo cv T o V Qo, v
and the total mass within the compartment
remains constant. Assuming ideal gas behavior where
Qo, v ð ρo cν T o V Þ ¼ Total internal energy of the control volume

Po ¼ Absolute ambient pressure
T o ¼ Absolute temperature of the control volume
The product of ρocνTo can be treated as a constant pressure differential of approximately 1013 Pa
with a value of approximately 252 kJ/m3, assum- (0.01 atm) before failing [18], would this pres-
ing air properties at standard temperature and sure rise be likely to cause window failure?
pressure.
Application of Equation 33.4 to representative Solution First, estimate the enthalpy of reaction
building fires demonstrates how quickly typical associated with the gasoline:
building boundaries would fail due to overpres-
surization if the boundaries were in fact hermeti- Q f ¼ m f ΔH c ¼ ρ f V f ΔHc ¼ ð760 kg=m3 Þ

cally sealed to prevent mass flow through 103 m3 ð44, 000 kJ=kgÞ ¼ 33, 440 kJ
enclosure boundaries. The following example
illustrates this point. Pressure changes may be a Then, calculate the net heat release associated
significant issue for fires in airtight vessels, with the burning of the gasoline:
including submarines and space vehicles, but
Qnet ¼ Q f ð1 χ l Þ ¼ 33, 440 kJ ð1 0:9Þ
are not usually significant for typical building
spaces, which are leaky by nature. ¼ 3344 kJ
Example 1 Determine the pressure rise and aver- Next, calculate the ambient internal energy
age temperature increase associated with com- level associated with the enclosure gases:
bustion of 1 L of gasoline within a sealed
Qo, v ¼ ρo cv T o V ¼ 252 kJ=m3 300 m3 ¼ 75, 600 kJ
enclosure with dimensions of 10 m by 10 m by
3 m. Assume a heat loss fraction of 0.90. Assum- Next, calculate the dimensionless pressure
ing ordinary window glass can withstand a and temperature changes using Equation 33.4:
1074 F.W. Mowrer
ΔP ΔT Qnet 3344 kJ dP R _ ðk 1Þ _
¼ ¼ ¼ ¼ 0:044 ¼ Q net m_ o ho ¼ Q net m_ o ho
Po T o Qo, v 75, 600 kJ dt cv V V
ð33:7Þ
Finally, calculate the dimensional pressure
and temperature changes: where k cp/cv ( 1.4 for air).
Equation 33.7 generally requires numerical
ΔP ¼ 0:044Po ¼ 0:044ð101, 325 PaÞ ¼ 4482Pa
solution because the mass outflow term on the
ΔT ¼ 0:044T o ¼ 0:044ð293 KÞ ¼ 13 K right-hand side is a function of the pressure differ-
ential between the fire enclosure and surrounding
Thus, the calculated pressure increase is more
spaces, while the net heat release rate term can
than four times the specified pressure differential
vary with time. Zukoski [1] examined the assump-
associated with window breakage, despite the
tion that the rate of pressure change is negligible
following: the fuel source is only 1 L of gasoline,
by comparing the time for the pressure to rise to
90% of the heat release is assumed to be lost to
86% of its equilibrium value with the time for the
the boundaries, the room volume is relatively
smoke layer to descend to the floor. For most
large at 300 m3, and the average temperature
scenarios, he found the ratio of these times to be
change is only 13 K. The fact that such small
on the order of 102; for relatively large fires or
fires do not routinely cause overpressure failures
relatively small leakage areas, this ratio was on the
of enclosure boundaries can be viewed as de
order of 101. Based on this analysis, Zukoski
facto evidence that real building enclosures are
concluded that an assumption of quasi-steady
leaky by nature.
pressure would be satisfactory for most fire
scenarios. In all the cases considered by Zukoski,
the pressure rise was so small that gas density and
Leaky Compartment
pressure were virtually unaffected. This quasi-
steady pressure assumption is employed here.
For the global analysis of the leaky compartment,
From a practical standpoint, the overall pres-
the entire enclosure volume is again considered
sure rise relative to atmospheric pressure, ΔP/Po,
as a fixed control volume, just as it was for the
is generally very small for fires in leaky
sealed compartment. In this case, the pressure
enclosures, typically on the order of 103 to
rise in the compartment caused by the release of
105, depending on the heat addition rate and
energy is assumed to force flow out of the enclo-
the area of leakage paths. Pressure differences
sure through available leakage paths while at the
of this magnitude are significant with respect to
same time preventing mass flow into the com-
the flows they cause through leakage paths in the
partment through these same leakage paths. Con-
enclosure boundaries, but can be considered as
sequently, for the leaky case, the energy balance
negligible with respect to the energy equation.
expressed by Equation 33.1 reduces to
Consequently, for most enclosure fire scenarios,
d ðρuV Þ the pressure can be treated as quasi-steady
¼ Q_ net m_ o ho ð33:5Þ (i.e., dP/dt ! 0 in Equation 33.7) and the
dt
quasi-steady global energy balance for the leaky
As for the sealed compartment analysis, the left- compartment can then be expressed as
hand side of Equation 33.5 can be expressed, for
an ideal gas, as Q_ net ¼ m_ o ho ¼ ρe c p T e V_ exp ð33:8Þ
dðρuV Þ cv V dP
¼ ð33:6Þ Equation 33.8 can be rearranged to solve for
dt R dt
the volumetric flow rate of gases from a compart-
Substituting Equation 33.6 into Equation 33.5 ment due to expansion. Using air properties with
permits the rate of pressure change to be calcu- the customary assumptions of constant specific
lated as heat, standard atmospheric pressure, and ideal
www.ebook777.com
2
gas behavior, this volumetric flow rate caused by 1 V_ exp
expansion can be expressed as ΔP ¼ ρe
2 Cd Aleak
2
Q_ net Q_ net ðkWÞ ð1:2 kg=m3 Þ 0:42 m3 =s
V_ exp m3 =s ¼ ffi ð33:9Þ ¼
ρe c p T e 353ðkJ=m3 Þ 2 ð0:65Þ ð0:04 m2 Þ
¼ 157Pa
As illustrated in Figs. 33.5 and 33.6, Equa-
tion 33.9 also represents the volumetric expan- Thus,
sion rate of the smoke layer.
ΔP 157 Pa
The quasi-steady pressure rise associated with ¼ O 103
this volumetric expansion rate can be calculated Po 101, 325 Pa
using classical orifice flow theory as discussed by which supports the assumption of negligible effect
Tanaka (see Chap. 15): on the energy equation. Based on measurements of
sffiffiffiffiffiffiffiffiffi pressure differentials during enclosure fire tests,
2ΔP
V_ exp ¼ Cd Aleak ve ¼ Cd Aleak ð33:10Þ this is a relatively high pressure differential,
ρe suggesting that real enclosures tend to be even
more leaky than was assumed for this example.
Equations 33.9 and 33.10 can be combined to
solve for the quasi-steady pressure rise, ΔP,
within the fire enclosure: Temperature Rise
!2
1 Q_ net The average temperature rise of the fixed control
ΔP ¼ ρe ð33:11Þ volume associated with a leaky compartment
2 ρe c p T e Cd Aleak
is considered in this subsection. First, the
Equation 33.11 can be used to check the assump- mass balance for the fixed control volume is
tion that the pressure rise in a compartment fire is introduced, recalling that the pressurization
negligible relative to the ambient pressure level. of the control volume caused by heat release
from the fire is assumed to prevent mass inflow:
Example 2 Calculate the volumetric expansion dðρV Þ dρ
rate and quasi-steady pressure rise that would be m_ o ¼ ¼ V ð33:12Þ
dt dt
associated with a fire with a heat release rate of
500 kW, a heat loss fraction of 0.7, a discharge Substituting Equation 33.12 into Equa-
coefficient of 0.65, and a leakage area of 0.04 m2. tion 33.8 yields
Assume air with a density of 1.20 kg/m3 is being dρ
expelled from the fire room. Q_ net ¼ m_ o ho ¼ c p TV ð33:13Þ
dt
Solution First, calculate the net heat release rate: For an ideal gas at constant pressure, the density
is related to the temperature as ρ ¼ ρoTo/T.
Q_ net ¼ Q_ f ð1 χ l Þ ¼ ð500 kWÞ ð1 0:7Þ Consequently, the rate of change of density can
¼ 150 kW be related to the rate of temperature change as
dρ ρ T o dT
Then, calculate the volumetric expansion rate ¼ o2 ð33:14Þ
dt T dt
associated with this net heat release rate:
Substituting Equation 33.14 into Equa-
Q_ net ð150 kWÞ tion 33.13 yields
V_ exp ¼ ¼ ¼ 0:42 m3 =s
ρe c p T e ð353 kJ=m3 Þ
dT Qo, p dT
Q_ net ¼ ρo c p T o V ¼ ð33:15Þ
Finally, calculate the quasi-steady pressure Tdt T dt
rise associated with this flow rate:
1076 F.W. Mowrer
The term Qo,p ( ρocpToV) is analogous to the Equation 33.17 permits calculation of an aver-
Qo,v term for the sealed compartment case; this age temperature rise caused by a fire within an
term represents the ambient enthalpy level for a enclosure. Equation 33.17 is particularly useful
fixed control volume at constant pressure. Assum- for fire hazard screening purposes, because it
ing that gases within the control volume have the allows thermal hazards to be estimated without
properties of air, the product of ρocpTo can be the need to track conditions in the descending
treated as a constant with a value of approximately smoke layer by numerical integration. Global
353 kJ/m3 over the temperature range of interest temperatures calculated with Equation 33.17 are
for preflashover enclosure fires. Equation 33.15 compared with smoke layer temperatures for
can be rearranged and integrated as Case 1 and Case 2 descending layer scenarios
in a subsequent section of this chapter.
ðt Tðg
dT
Q_ net dt ¼ Qo, p ð33:16Þ Example 3 Determine the average global temper-
0 T
To ature rise in an enclosure with dimensions of
18.3 m by 12.2 m by 6.1 m in response to a fire
The solution to Equation 33.16 can be expressed that grows as a t-squared fire to a heat release rate
in terms of the dimensionless average tempera- of 500 kW in 240 s, then burns at a constant heat
ture rise in the fixed control volume: release rate of 500 kW for another 360 s. Estimate
! the average temperature rise within the enclosure
ΔT Qnet at 240 s and at 600 s based on this heat release rate
¼ exp 1 ð33:17Þ
To Qo, p history, assuming a constant heat loss fraction of
0.70 and an ambient temperature of 20 C (293 K).
where
ðt Solution First, calculate the net heat release for
Qnet ¼ Q_ net dt ¼ Net energy addition to the the two selected times:
0
ðt ðt
control volume as before
Qnet ¼ Q_ net dt ¼ Q_ f ð1 χ l Þdt
T o ¼ Absolute ambient temperature o o
The relationship expressed by Equation 33.17 is During the period up to 240 s, the fire heat
illustrated in Fig. 33.8. release rate follows the specified t-squared
Fig. 33.8 Average 4.0

dimensionless temperature
rise in a closed room 3.5
as a function of the
dimensionless net energy 3.0
addition
2.5
ΔT/To
2.0
1.5
1.0
0.5
0.0
0.00 0.25 0.50 0.75 1.00 1.25 1.50
Qnet /Qo,p
www.ebook777.com
ð 600
growth, and the net heat release during this
Qnet ð@600 sÞ ¼ Qnet ð@240 sÞ þ 500 ð1 0:7Þdt
period is calculated as 240
ð 240 ! ¼ 12, 000 kJ þ 54, 000 kJ

500
Qnet ð@240 sÞ ¼ 2
ð1 0:7Þt2 dt ¼ 66, 000 kJ
o ð 240 Þ
! !
500 ð240Þ3 Next, the ambient enthalpy level within the
¼ ð0:3Þ enclosure is calculated:
ð240Þ2 3

¼ 12, 000 kJ Qo, p ¼ ρo c p T o V ¼ 353 kJ=m3 1362 m3
The net heat release at 600 s is equal to this ¼ 480, 746 kJ
value plus the net heat release during the period
from 240 s to 600 s, when the fire heat release Then, the average dimensionless temperature
rate is constant: rise at each time is calculated as
!
ΔT Qnet 12, 000
¼ exp 1 ¼ exp 1 ¼ 0:025 at 240 s
To Qo, p 480, 746
!
ΔT Qnet 66, 000
¼ exp 1 ¼ exp 1 ¼ 0:147 at 600 s
To Qo, p 480, 746
Finally, the dimensional temperature changes increase more rapidly for elevated fires than if it
are calculated at each time as is based on the entire enclosure volume, assuming
the heat loss fraction does not change significantly
ΔT ¼ 0:025 T o ¼ 0:025 293 K ¼ 7:3 K at 240 s
for elevated fires. This will produce a more con-
ΔT ¼ 0:147 T o ¼ 0:147 293 K ¼ 43:1 K at 600 s servative estimate of temperature hazards based
on application of Equation 33.17.
Note that the ambient enthalpy level has been
based on the entire volume of the enclosure, as Example 4 For the previous example, estimate
illustrated in Fig. 33.7, not just on the volume the average temperature rise within the upper
above the fire source, as illustrated in Fig. 33.6. layer at the same times (240 s and 600 s) assum-
Rather than apply Equation 33.17 based on the ing the fire source is located at an elevation of
entire enclosure volume, it makes sense to con- 3.0 m and assuming the same constant heat loss
sider the fixed control volume defined as the vol- fraction of 0.70. How do these values compare
ume between the base of the fire source and the with the results in the previous example?
ceiling, as illustrated in Fig. 33.6. Assuming that
air entrainment occurs only laterally, this control Solution The only difference in this case com-
volume will not have mass inflow across the lower pared with the previous example is that the size
control volume boundary. For this scenario, air of the control volume has decreased by a factor
entrained into the fire plume is simply recirculated of two. Therefore, the ambient enthalpy level
from within the fixed control volume, with some within the control volume is calculated as
gases forced out of this control volume through its
lower face due to expansion. With the smaller
Qo, p ¼ ρo c p T o V ¼ 353 kJ=m3 681 m3
control volume defined in this way, the tempera-
ture rise expressed by Equation 33.17 will ¼ 240, 373 kJ
1078 F.W. Mowrer
For this case, the average dimensionless

temperature rise at each time is calculated as
!
ΔT Qnet 12, 000
¼ exp 1 ¼ exp 1 ¼ 0:051 at 240 s
To Qo, p 240, 373
!
ΔT Qnet 66, 000
¼ exp 1 ¼ exp 1 ¼ 0:316 at 600 s
To Qo, p 240, 373
Finally, the dimensional temperature changes negligible. Furthermore, the left-hand side of
are calculated at each time as Equation 33.18 can be expanded to
ΔT ¼ 0:051 T o ¼ 0:51 293 K ¼ 14:9 K at 240 s dðmY i Þ dY i dm

¼m þ Yi
ΔT ¼ 0:316 T o ¼ 0:316 293 K ¼ 92:6 K at 600 s dt dt dt
dY i
¼m m_ o Y i ð33:19Þ
Note that these values for the temperature dt
rise are slightly more than twice the respective
The production of a particular species can
values in the previous example. The reason for
generally be described in terms of the product
this is that as the smoke layer heats up, more
of a species yield factor, fi, by the fuel mass loss
mass is expelled from the smoke layer, leaving
rate, ṁf, such that
less mass within the smoke layer to absorb addi-
tional heat input. That is why the temperature Q_ f
relationship expressed by Equation 33.17 m_ i, gen ¼ f i m_ f ¼ ð33:20Þ
ðΔH c = f i Þ
increases exponentially with heat input rather
than linearly as in the sealed enclosure case, Equations 33.18, 33.19, and 33.20 can then be
where the mass within the control volume combined and simplified to
remains constant.
dY i m_ i, gen m_ i, gen
¼ ¼
dt mcv ðρV Þcv
Concentrations of Smoke and Other Q_ f
Species ¼ ð33:21Þ
ðρV Þcv ðΔH c = f i Þ
The conservation of different chemical species Equation 33.21 applies under scenarios where
(e.g., O2, N2, CO2, CO, H2O, soot) within a the fire is fuel limited, such that the heat release
control volume can be expressed generally as
rate of the fire can be expressed as Q_ f ¼i m_ f ΔH c .
d ðmY i Þ The term (ΔHc/fi) in Equation 33.21 represents a
¼ ðm_ i Y i, in Þ ðm_ o Y i Þ þ m_ i, gen ð33:18Þ “species heat of combustion;” it is the quantity of
dt
heat released per unit mass of species i produced
Yi is the mass fraction of species i within the (or consumed in the case of oxygen). The species
control volume. For the fixed control volume heat of combustion can be estimated based on
shown in Fig. 33.7, it is assumed that mass inflow stoichiometry for products of complete combus-
is precluded by pressurization of the enclosure tion (i.e., CO2 and H2O) or based on experimen-
(i.e., this analysis does not address the “puffing” tal yield data for products of incomplete
behavior noted previously), so the first term combustion (i.e., CO and soot). Extensive yield
on the right-hand side of Equation 33.18 is data for a range of fuels is provided by Tewarson
www.ebook777.com
(see Chap. 36). Equation 33.21 can be integrated The volume of the space is calculated as
with appropriate limits to yield 18.3 m 12.2 m 6.1 m ¼ 1362 m3. There-
ðt fore, the heat release per unit volume is calcu-
Q_ f dt lated at each time as
Qf
ðY i Y i, o Þ ¼ 0
¼
ðρV Þcv ðΔHc = f i Þ ðρV Þcv ðΔH c = f i Þ Q f =V ð@240 sÞ ¼ 40, 000 kJ=1382 m2 ¼ 28:9kJ=m3
ð33:22Þ Q f =V ð@600 sÞ ¼ 220, 000 kJ=1382 m2 ¼ 159:2 kJ=m3
Yi,o is the initial mass fraction of species i in the The species heat of combustion is calculated
control volume. For most products of combus- from the given data as
tion, the initial species mass fraction, Yi,o, is nil.
For this case, the species mass concentration can 46:4 MJ=kg f
ΔHc = f i ¼ ¼ 488:42 MJ=kgsoot
be expressed as 0:095 kgsoot =kg f

Q f =V Finally, the soot mass concentration at each
ρY i ¼ ð33:23Þ
ðΔH c = f i Þ time is calculated as
where 28:9 kJ=m3

ρY soot ð@240 sÞ ¼
488:42 103 kJ=kgsoot
ρY i ¼ Mass concentration of species iðkgi =m Þ
3
Q f =V ¼ Fire heat release per unit volume of the ¼ 5:92 105 kgsoot =m3
control volume ðkJ=m3 Þ
ðΔH c = f i Þ ¼ Species heat of combustion ðkJ=kgi Þ 159:2 kJ=m3
488:42 103 kJ=kgsoot
Example 5 Assume propylene (C3H6) is the fuel ¼ 3:26 104 kgsoot =m3
burned in the previous examples. Assume pro-
pylene has a heat of combustion of 46.4 MJ/kg of As discussed in a subsequent subsection, the
fuel and a soot yield of 0.095 g of soot per g of visibility through smoke can be related directly
fuel. Estimate the average mass concentration of to the soot mass concentration.
soot within the 18.3 m by 12.2 m by 6.1 m For oxygen in air under standard conditions,
enclosure at 240 s and 600 s after ignition of a the initial species mass fraction is Yi,o ¼ 0.233
fire that grows as a t-squared fire to a heat release and the species heat of combustion can be
rate of 500 kW in 240 s, then burns at a constant taken as the well-known “oxygen heat of
heat release rate of 500 kW for another 360 s. combustion,” for a wide range of representative
fuels, with a value of approximately ΔH c = f O2 ¼
Solution First, calculate the fire heat released up 13, 100 kJ=kgO2 , where the negative sign
to the 240 s and 600 s times, respectively: indicates that oxygen is consumed rather than
ð 240 ! produced in the combustion reaction. For these
500
Q f ð@240 sÞ ¼ t2 dt values, the oxygen mass fraction within the
o ð240Þ2 fixed control volume can be estimated as
! !
500 ð240Þ3 Qf
¼ Y O2 ¼ Y O2 , o þ
ð240Þ2 3 ρV ΔH c = f O2

¼ 40, 000 kJ Q f =V
¼ 0:233 ð33:24Þ
ð 600 ð13; 100Þρ
Q f ð@600 sÞ ¼ Q f ð@240 sÞ þ 500 dt
240 Example 6 Assume propylene (C3H6) is the fuel
¼ 40, 000 kJ þ 180, 000 kJ burned in the previous examples. Assume pro-
pylene has a heat of combustion of 46.4 MJ/kg of
¼ 220, 000 kJ
1080 F.W. Mowrer
fuel and a soot yield of 0.095 g of soot per g of ð28:9Þ

Y O2 ð@240 sÞ ¼ 0:233 ¼ 0:231
fuel. Estimate the average oxygen mass fraction ð13; 100Þ ð1:15Þ
within the 18.3 m by 12.2 m by 6.1 m enclosure
ð159:9Þ
at 240 s and 600 s after ignition of a fire that Y O2 ð@240 sÞ ¼ 0:233 ¼ 0:220
ð13; 100Þ ð0:92Þ
grows as a t-squared fire to a heat release rate of
500 kW in 240 s, then burns at a constant heat
Thus, for these examples, the oxygen concen-
release rate of 500 kW for another 360 s.
tration is relatively close to the ambient concen-
tration and consequently would not be expected
Solution From the previous example, the fire
to have a significant effect on the fire heat release
heat released up to the 240 s and 600 s times is,
rate. As discussed in the following subsection,
respectively:
however, this is not always the case.
ð 240 !
500
Q f ð@240 sÞ ¼ t2 dt
o ð240Þ2 Oxygen Limitations on Heat Release
! ! in a Closed Room Fire
500 ð240Þ3
¼
ð240Þ2 3 There is a limit to how much heat can be released
¼ 40, 000 kJ by combustion within a closed room because the
ð 600 release of heat is coupled with consumption of a
Q f ð@600 sÞ ¼ Q f ð@240 sÞ þ 500 dt finite amount of oxygen from the air in the enclo-
240 sure. It is assumed that oxygen does not enter
¼ 40, 000 kJ þ 180, 000 kJ from outside due to pressurization of the com-
¼ 220, 000 kJ partment, so the fire must eventually die down
due to oxygen depletion, much like the familiar
The volume of the space was calculated as candle flame trapped inside an inverted jar.
18.3 m 12.2 m 6.1 m ¼ 1362 m3 and the Equation 33.17 will result in nonphysical and
heat release per unit volume was calculated at incredible temperatures if applied indefinitely
each time as because it does not account for the effect of
oxygen depletion on limiting heat release within
Q f =V ð@240 sÞ ¼ 40, 000 kJ=1382m3 ¼ 28:9 kJ=m3 an enclosed space.
Mowrer [11] has addressed the issue of oxy-
Q f =V ð@600 sÞ ¼ 220, 000 kJ=1382m3 ¼ 159:2 kJ=m3
gen limitations on heat release in unventilated
enclosure fires. The heat released by combustion
The oxygen heat of combustion is assumed
in a room fire is related directly to the oxygen
to be
consumed. This relationship can be expressed as
ΔH c = f i ¼ 13, 100 kJ=kgO2 " #
Qo , p ρo V ΔH c
The average temperature in the enclosure was Q f , lim ¼ ln 1 þ χ ð1 χ l Þ
ð 1 xl Þ Qo, p r air O2 , lim
previously calculated at each time to be
ð33:25Þ
T ¼ T o þ ΔT ¼ 293 K þ 14:9 K ¼ 307:9 K at 240 s
Equation 33.25 can be inserted into Equa-
ðso ρ ¼ ρo T o =T ¼ 353=307:9 ¼ 1:15kg=m3 Þ
tion 33.17 to yield the limiting temperature rise
T ¼ T o þ ΔT ¼ 293 K þ 92:6 K ¼ 385:6 K at 600 s associated with the oxygen-limited heat release
ðso ρ ¼ ρo T o =T ¼ 353=385:6 ¼ 0:92kg=m3 Þ in an enclosure:
Finally, the oxygen mass fraction at each time ΔH c χ O2 , lim ð1 χ l Þ

ΔT g, lim ¼ ð33:26Þ
is calculated as r air cp
www.ebook777.com
where χ O2 , lim represents the fraction of oxygen [19–21] if present, providing new pathways for
that can be consumed before extinction; it is the introduction of oxygen to the enclosure and
given in general as consequent escalation of the fire intensity.

Y O2o Y O2 , lim
χ O2 , lim ¼
Y O2o Light Attenuation and Visibility

XO2 o XO2 , lim Through Smoke
¼ ð33:27Þ
X O2 o
Light attenuation and visibility through smoke
can be estimated based on the soot mass concen-
χ O2 , lim is normally evaluated at a limiting extinc-
tration within the smoke layer. The light extinc-
tion value of XO2 or Y O2 . A representative value tion coefficient, K, is directly proportional to the
for XO2 at extinction under normal ambient soot mass concentration as
conditions is approximately 13% for a range of
K ¼ K m ρY soot ð33:28Þ
hydrocarbon fuels when the oxygen is diluted
with nitrogen (see Chap. 17). For an ambient Km is the specific light extinction coefficient.
oxygen mole fraction XO2, 0 of 21%, a representa- Seader and Einhorn [22] suggested a value of
tive value for χ O2 , lim would therefore be about 0.4. Km ¼ 7600 m2/kg for flaming combustion and
Km ¼ 4400 m2/kg for smoke produced by pyrol-
Example 7 Determine the oxygen-limited aver- ysis. These values have been widely used for
age temperature rise in an enclosure fire for heat light attenuation and visibility calculations, but
loss fractions of 0.6 and 0.9, respectively. more recently Mulholland and Croarkin [23]
Assume a value of χ O2 , lim ¼ 0:4. have suggested a value of Km ¼ 8700 m2/kg for
flaming combustion of wood and plastic fuels.
Solution Light attenuation within the smoke layer is
calculated in accordance with Bougher’s law
3000ðkJ=kgÞ for monochromatic light:
ΔT g, lim ¼ 0:4 ð1 0:6Þ
1:0ðkJ=kg KÞ
I=I o ¼ eKL ð33:29Þ
¼ 480 K for χ l ¼ 0:6
3000ðkJ=kgÞ Visibility through smoke is expected to
ΔT g, lim ¼ 0:4 ð1 0:9Þ
1:0ðkJ=kg KÞ vary inversely with the light extinction coeffi-
¼ 120 K for χ l ¼ 0:9 cient, with this inverse relationship generally
expressed as
Although these temperatures are potentially sig-
S ¼ C=K ð33:30Þ
nificant from a thermal injury or damage stand-
point, they are below the temperature rise of where
approximately 580 K commonly associated S ¼ Visibility distance ðmÞ
with flashover conditions. This simple analysis C ¼ Nondimensional constant associated with the
suggests the difficulty of attaining flashover object being viewed through the smoke
conditions in an unventilated, fully enclosed Mulholland [24] suggests a value of C ¼ 8 for
compartment fire. Lower heat loss fractions and light-emitting signs and a value of C ¼ 3 for
higher oxygen consumption fractions would be light-reflecting signs based on the work of Jin
needed to achieve temperature increases (see Chap. 61). These values suggest that light-
associated with flashover. On the other hand, emitting signs can be observed when the light
these calculated global temperature increases attenuation is I/Io ¼ e8 ¼ 3.35 104, that is,
might be sufficient to cause the fracture and the transmitted light is much less than 1% of
collapse of ordinary plate glass windows the unattenuated light intensity; whereas
1082 F.W. Mowrer
light-reflecting signs can be observed when the conditions within the descending smoke layer
light attenuation is I/Io ¼ e8 ¼ 0.05, that is, the are addressed. These can then be compared with
transmitted light is reduced to 5% of the unatten- the global analysis results presented in this
uated light level. section.
Example 8 Estimate the visibility of a light-

reflecting sign through the smoke layer based Descending Smoke Layer Analysis
on the soot mass concentrations determined in a
previous example. In this section, the descending smoke layer is
treated as a distinct control volume, as illustrated
Solution For this example, the soot mass con- in Fig. 33.5. Two limit cases are addressed based
centration at each time was calculated to be on the location of leakage paths in the enclosure
boundaries, following the approach originally
28:9 kJ=m3
ρY soot ð@240 sÞ ¼ taken by Zukoski [1]:
488:42 103 kJ=kgsoot • Case 1—Leakage paths near the floor (from
¼ 5:92 105 kgsoot =m3 the lower layer)
• Case 2—Leakage paths near the ceiling (from
159:2 kJ=m3 the upper layer)
488:42 103 kJ=kgsoot In Case 1, the expansion of gases in the upper
¼ 3:26 104 kgsoot =m3 layer pushes fresh air at ambient temperature
from the lower layer until the smoke layer
Using these values of soot mass concentration descends to the floor. At that point, smoke at
along with the specific light extinction coeffi- the upper layer temperature and composition
cient of 8700 m2/kg suggested by Mulholland would be expelled and the analysis would be
and Croarkin, the extinction coefficient is deter- the same as for Case 0, the global analysis
mined to be presented in the previous section. The Case 1 sce-
nario is the case addressed by Cooper [2, 3] in the
K ð@240 sÞ ¼ K m ρY soot
development of the ASET model. In Case 2, the
¼ ð8700 m2 =kgsoot Þ expansion of gases from the compartment is
assumed to occur directly from the upper layer.
5:92 105 kgsoot =m3
Cooper does not address this scenario, but, as
¼ 0:52 m1 demonstrated by the following analysis, the
K ð@600 sÞ ¼ K m ρY soot differences between the two scenarios are minor.
Mass balances on the lower layer for the
¼ ð8700 m2 =kgsoot Þ respective cases can be written as

3:26 104 kgsoot =m3
d ðρV Þl dV l
1 Case 1 : ¼ ρl ¼ m_ pl þ m_ e
¼ 2:83 m dt dt
ð33:31Þ
The visibility distance for a light-reflecting
sign is then estimated at each time as d ðρV Þl dV l
Case 2 : ¼ ρl ¼ m_ pl ð33:32Þ
dt dt
S ð@240 sÞ ¼ 3=0:52 m1 ¼ 5:8 m ð19:0 ftÞ
S ð@600 sÞ ¼ 3=2:83 m1 ¼ 1:1 m ð3:6 ftÞ These mass balances for the lower layer can be
converted to volumetric filling rates for the upper
This concludes the global analysis of fire- layer by noting that dVu ¼ dVl and by dividing
induced conditions in a closed room. In the next through by the lower layer density, ρl , assumed
section, methods to analyze fire-induced to remain constant at the ambient air density.
www.ebook777.com

dV u m_ pl þ m_ e where
Case 1 : ¼ ¼ V_ pl þ V_ exp
dt ρl V As H
τV ¼ ¼
ð33:33Þ V_ pl, H kv Q_ 1=3 H 5=3
c

dV u m_ pl As =H 2
Case 2 : ¼ ¼ V_ pl ð33:34Þ ¼ ð33:37Þ
dt ρl kv ðαn tn Þ1=3 =H4=3
Equations 33.33 and 33.34 show that the volumet-
ric growth rate of the smoke layer is due to both τV is a characteristic smoke filling time constant,
plume entrainment and gas expansion in Case represented as the volume above the fire source
1 and due to entrainment only in Case 2. This divided by the volumetric entrainment rate
difference occurs because the expanding gases evaluated at the ceiling height. Note that this
are being expelled directly from the smoke layer time constant is actually only constant in the
rather than the lower layer in Case 2, and conse- case of a steady fire (n ¼ 0). The elevation of
quently gas expansion does not contribute to the smoke layer interface above the fire source
smoke layer development in Case 2. derives directly from Equation 33.36:
The rate of smoke layer descent for the two
3=2
zL zu 2t
cases derives directly from Equations 33.33 and ¼1 ¼ 1þ ð33:38Þ
H H ðn þ 3ÞτV
33.34 by noting that dVu ¼ Asdzu. For rooms
with vertical walls and horizontal ceilings, the
The solution represented by Equation 33.38 is
horizontal area of the space, As, remains constant
shown in Fig. 33.9 for the cases of a steady fire
with height, assuming no physical obstructions
(n ¼ 0) and a t-squared fire (n ¼ 2). This
are located within the space. In general, Equa-
solution is also approximately accurate for
tion 33.33 must be integrated numerically to
Case 1 scenarios where V_ pl
V_ exp , such that
determine the smoke layer interface position as
a function of time because analytical solutions do V_ pl þ V_ exp V_ pl .
not exist for most realistic scenarios. Equa- Qualitatively, Fig. 33.9 illustrates the fact that
tion 33.34 does have an analytical solution for smoke layer descent is initially very rapid, then
the case of a power law fire, where the fire heat slows down and asymptotically approaches the
release rate is assumed to vary with time as fuel surface. This is because the air entrainment
rate for axisymmetric plumes varies with the 5/3
Q_ f ¼ αn tn power of the elevation between the fuel surface
and the smoke layer interface. As the smoke
and axisymmetric plume entrainment [25], where
layer descends, the plume entrains less and less
the plume volumetric flow rate varies as
air, causing the descent rate to slow down as
shown in Fig. 33.9.
V_ pl ¼ kv Q_ c z5=3
1=3
The nondimensional representation of smoke
For these conditions, Equation 33.34 can be layer descent given in Equation 33.38 and shown
rearranged and expressed as in Fig. 33.9 is useful to generalize the smoke
ð zu ð layer descent analysis, but is not as useful for
dzu kv t the computation of specific fire scenarios. This is
5=3
¼ ðαn tn Þ1=3 dt ð33:35Þ
0 ðH zu Þ A s 0 particularly so for the case of growing t-squared
fires because the characteristic time constant
The solution to Equation 33.35 can be given in Equation 33.37 is a function of time
expressed nondimensionally as for cases other than steady fires (i.e., other than

3=2 when n ¼ 0). For such computations it is more
zu 2t
¼1 1þ ð33:36Þ useful to represent the smoke layer descent in
H ðn þ 3ÞτV dimensional terms. To calculate the time for the
1084 F.W. Mowrer
Fig. 33.9 Relative smoke 1.0

layer interface position as a n=0
function of normalized 0.9 n=2
time due to entrainment
only in an axisymmetric 0.8
fire plume
0.7
0.6
ZL/H
0.5
0.4
0.3
0.2
0.1
0
0 2 4 6 8 10
t /τv
smoke layer to reach a particular elevation rela- Solution For this example, the enclosure area is
tive to the fuel to ceiling distance (zL/H ) for a As ¼ 18.3 m 12.2 m ¼ 223.3 m2 and the
t-squared fire, Equation 33.37 is substituted into enclosure height above the fire source is H ¼ 6.0
Equation 33.38 and the resulting equation is m. The fire grows as a t-squared fire to reach a
solved for t. The result is heat release rate of 500 kW in 240 s, then remains
constant at 500 kW for an additional 360 s. Thus,
(
)1=ð1þn=3Þ
ðn þ 3Þ As =H2 H 4=3 zL 2=3 during the growth stage,
t¼ 1=3
1
2 k v αn H
500 kW
αn ¼ ¼ 8:68 103 kW=s2
ð33:39Þ ð240 sÞ2
Similarly, for a t-squared fire, the relative smoke and n ¼ 2. Assuming an axisymmetric plume,
layer elevation can be expressed explicitly as a the entrainment coefficient is taken to be
function of time as
kv ¼ 0:064 m4=3 =kW1=3 s
" #3=2
zL 2kv α1=3 tð5=3Þ
¼ 1 þ n 2 4=3 ð33:40Þ For the smoke layer interface elevation of 3.0 m,
H 5 As =H H the relative smoke layer interface elevation is
zL 3:0 m
Example 9 For the previous enclosure fire ¼ ¼ 0:5
H 6:0 m
example, determine how long it would take
for the smoke layer interface to descend to Substituting these values into Equation 33.39
elevations of 3.0 m and 1.5 m above the floor yields the time for the smoke layer to reach the
for the 6.0-m-high enclosure. 3.0 m elevation in the enclosure:
( " 2=3 #)1=ð1þ2=3Þ

2þ3 223:3=6:02 6:04=3 3:0
tð3:0 mÞ ¼ 1
2 0:064 8:68 103 1=3 6:0
¼ 212 s
www.ebook777.com
Because this time is less than the growth time stop growing before the smoke layer descends to
of 240 s, the smoke layer will descend to the the 1.5 m elevation. This can be confirmed by
3.0 m elevation during the growth period. From substituting this elevation of 1.5 m into
this analysis, it appears likely that the fire will Equation 33.39:
( " #)1=ð1þ2=3Þ
ð2 þ 3Þ 223:3=6:02 6:04=3 1:5 2=3
tð1:5 mÞ ¼ 1=3 1
2 0:064 8:38 103 6:0
¼ 375 s
Because the fire stops growing at 240 s, it is determine the additional time needed for the
necessary to apply Equation 33.40 to determine smoke layer interface to reach the 1.5 m
the smoke layer interface position at the end of elevation as a result of the steady 500 kW fire.
the growth period, then use this elevation as Applying Equation 33.40 with a time of 240 s
the enclosure height H in Equation 33.39 to yields
" 1=3 #3=2

zL 2ð0:064Þ 8:68 103 240ð1þ2=3Þ
ð@240 sÞ ¼ 1 þ
H ð2 þ 3Þ 223:3=6:02 6:04=3
¼ 0:44
zL ð@240 sÞ ¼ 0:44 H ¼ 0:44 6:0 ¼ 2:64 m
This value then becomes the starting height

(i.e., H ) for the steady fire following 240 s:
( " #)1=ð1þ0=3Þ
0 þ 3 223:3=2:642 2:644=3 1:5 2=3
tð1:5 mÞ ¼ 240 þ 1
2 0:064 ð500Þ1=3 2:64
¼ 240 þ 159 ¼ 399 s
In other words, the smoke layer interface Conditions in the Descending

reaches the 2.64 m elevation at the end of the Smoke Layer
240-s fire growth period, then takes another 159 s
to reach the 1.5 m elevation during the ensuing The average temperature in the smoke layer is
steady fire period. calculated by invoking the ideal gas law relation-
ship for a constant pressure process, ρoTo ¼ ρuTu,
1086 F.W. Mowrer
and by noting that the average smoke layer Case 2:

density, ρu, is simply the mass of the upper 1 0ð t
layer divided by the volume of the upper layer. dV u
ρ Tl ρ TlVu B o dt dtC
For the two cases, the average temperature of the T u ðtÞ ¼ l ¼ l ¼ ρl T l B
@ð t dmu A
C
upper layer is then evaluated as ρu mu
dt
Case 1: o dt
0ð t 1 ðt

dV u
dt ρl T l V_ pl dt
ρ Tl ρ TlVu B o dt C
T u ðtÞ ¼ l ¼ l ¼ ρl T l B C ¼ ðt o
ρu @ð t dmu A
mu
dt ρl V_ pl ρu V_ exp dt
o dt o
ðt ð33:42Þ

V_ pl þ V_ exp dt
¼ Tl o ð t
The mass fraction of oxygen in the smoke
V_ pl dt
o layer is calculated for the two cases as
ð33:41Þ Case 1:
ðt " !#
Q_ f
ρl V_ pl Y O2 , o dt
mO2 , u o ΔH c =r O2
Y O2 , u ðtÞ ¼ ¼ ð33:43Þ
mu mu
Case 2:
ðt " ! #
Q_ f
ρl V_ pl Y O2 , o ρu V_ exp Y O2 , u dt
mO2 , u 0 ΔH c =r O2 ð33:44Þ
Y O 2 , u ðt Þ ¼ ¼
mu mu
The mass fractions of different products of

combustion in the smoke layer are calculated
for the two cases as
Case 1:
ðt " !#
Q_ f
ρl V_ pl Y i, o þ dt
mi, u 0 ΔH c = f i
Y i, u ðtÞ ¼ ¼ ð33:45Þ
mu mu
Case 2:
ðt " ! #
Q_ f
ρl V_ pl Y i, 0 ρu V_ exp Y i, u dt
mi, u 0 ΔHc = f i
Y i, u ðtÞ ¼ ¼ ð33:46Þ
mu mu
www.ebook777.com
The denominators in Equations 33.43 through within a space, for example, to prevent smoke
33.46 are evaluated in the same way as in from reaching the highest elevation of human
Equations 33.41 and 33.42, respectively. In gen- occupancy.
eral, Equations 33.41 through 33.46 do not have
analytical solutions and therefore require numer-
ical integration. Numerical methods are Global Effects of Mechanical Ventilation
discussed in a subsequent section.
Global conditions within a mechanically
ventilated enclosure are first considered in
Influence of Mechanical Ventilation terms of a one-zone analysis, as illustrated in
on Smoke Layer Conditions Fig. 33.11. This global analysis assumes that
conditions throughout the enclosure can be
The introduction of mechanical ventilation treated as uniform outside the fire-plume/ceil-
changes the analysis of enclosure smoke filling ing-jet sublayer as a result of mixing caused by
in a number of ways. As illustrated in Fig. 33.10, the mechanical ventilation and plume entrain-
with mechanical ventilation, flow may be ment. One-zone approaches have been used pre-
injected into or extracted from either the upper viously to evaluate average fire conditions in
layer or the lower layer. This will depend on the mechanically ventilated spaces [7, 26–28].
type of mechanical ventilation system employed, Volumetric flow rates associated with a fire in
the elevations of injection and extraction vents, a mechanically ventilated enclosure are
and the elevation of the smoke layer interface at a illustrated in Figs. 33.10 and 33.11. Because the
particular time. Smoke layer descent may be total volume of the enclosure remains essentially
either accelerated or retarded relative to the constant, the rate at which flow is forced through
unventilated scenario as a result of mechanical leakage paths in the boundaries of the enclosure
ventilation. Conditions within the smoke layer can be expressed as
will also be affected. A quasi-steady smoke
layer interface position will develop if the flow V_ net ¼ V_ in j þ V_ exp V_ ext ð33:47Þ
rates balance properly. Indeed, the purpose of a
dedicated mechanical smoke extraction system is V_ net can be either positive or negative, depending
normally to prevent the smoke layer interface on the values for the terms on the right-hand side
position from descending past a certain elevation (RHS) of Equation 33.47. As defined here, V_ net
Fig. 33.10 Mechanical

ventilation in a two-layer Vinj Vext
environment
Vnet
Vpl + Vext
(Case 2)
zext
Vpl
zinj
Qf, Vpl, Vexp
Vnet
(Case 1)
1088 F.W. Mowrer
Fig. 33.11 Mechanical

ventilation in a one-layer
Vinj Vext
environment
Vnet
pg , Tg , V Qf , Vext
will be positive if flow is forced from the enclo- dmgl

¼ ρo V_ inj V_ net ρgl V_ ext V_ net < 0
sure to surrounding spaces and negative if flow is dt
drawn from surrounding spaces into the ¼ ρo V_ ext þ V_ exp ρgl V_ ext
enclosure. ð33:48bÞ
The injection and extraction rates, V_ in j and
_
V ext , will depend on the design of the ventilation For any individual species to be tracked, such
system. For present purposes, it is assumed that as O2, CO2, CO, or soot, a global species balance
the user specifies the injection and extraction also depends on the sign of V_ net . This global
volumetric flow rates, although it should be species balance can be expressed as
recognized that these flow rates may be influenced dmi
by fan characteristics and space pressures. Under ¼ ρo Y i, o V_ inj ρgl Y i V_ inj þ V_ exp
dt
the quasi-steady pressure conditions assumed
here, the volumetric expansion rate of gases can þ m_ i, gen V_ net > 0
be related directly to the net rate of heat addition ð33:49aÞ
resulting from the fire, as shown in Equation 33.9.
dmi
A global mass balance for the enclosure can ¼ ρo Y i, o V_ ext V_ exp ρgl Y i V_ ext
dt
be expressed in terms of the volumetric flows þ m_ i, gen V_ net < 0
across the enclosure boundaries. For situations
ð33:49bÞ
where V_ net is positive, flow is forced from the
enclosure to adjacent spaces through available This global species balance can also be
leakage paths. For this situation, the global expressed in terms of the mass fraction for each
mass balance can be expressed as chemical species of interest:
dmgl dY i 1 _
¼ ρo V_ inj ρgl V_ net þ V_ ext V_ net > 0 ¼ ρo V inj ðY io Y i Þ þ m_ i, gen V_ net > 0
dt dt mgl

¼ ρo V_ inj ρgl V_ inj þ V_ exp ð33:50aÞ
ð33:48aÞ dY i 1 _
¼ ρo V ext V_ ext ðY io Y i Þ þ m_ i, gen
dt mgl
For situations where V_ net is negative, air is
drawn into the enclosure through leakage paths V_ net < 0
and the mass balance becomes ð33:50bÞ
www.ebook777.com
The generation (or consumption) rate of a Q_ net ð500 kWÞ ð1 0:7Þ

species, ṁi, gen, can be expressed in terms of the V_ exp ¼ ¼
ρe c p T e ð353 kJ=m3 Þ
fuel mass loss rate or the fire heat release rate, as
shown in Equation 33.20. ¼ 0:42 m3 =s
The global temperature in the enclosure is
calculated, assuming ideal gas behavior, constant The volumetric injection rate, V_ inj , is calcu-
pressure, and properties of air, as lated based on the compartment volume and the
specified air exchange rate as
ρo T o ρo T o V
T gl ¼ ¼ ð33:51Þ
ρgl mgl 10ð18:3 12:2 6:1Þm 3
V_ inj ¼ ¼ 3:78 m3 =s
3600 s
Under quasi-steady conditions, the rate of
change of mass in the enclosure goes to zero The quasi-steady global temperature rise is
as the inflow and outflow equilibrate. For calculated by applying Equation 33.54:
an injection-only mechanical system, V_ net will !
always be positive and the quasi-steady mass V_ exp 0:42
ΔT gl ¼ To ¼ 293 K
balance expressed by Equation 33.48a becomes V_ inj 3:78

ρgl V_ inj þ V_ exp ¼ ρo V_ inj ð33:52Þ ¼ 32:5 K
For this situation, the quasi-steady global For an extraction only system, the analysis
temperature can be expressed in dimensionless depends on the rate of extraction relative to the
form as expansion rate. If the extraction rate is less than
the expansion rate, then the extraction will serve
T gl ρo V inj þ V_ exp V_ exp simply to relieve a fraction of the expansion flow,
¼ ¼ ¼1þ ð33:53Þ
T o ρgl V_ inj V_ inj with the remaining fraction forced through
available leakage paths. This situation can be
Alternatively, the quasi-steady global dimen- treated using the same global analysis as for an
sionless temperature rise above ambient can be unventilated enclosure. There is not a quasi-
expressed simply as the ratio of the expansion steady solution for this situation.
flow rate to the injection flow rate: For the situation where the extraction rate is
ΔT gl V_ exp greater than the expansion rate, V_ net will be
¼ ð33:54Þ negative, so air will be drawn into the enclosure
To V_ inj
through available leakage paths. From Equa-
tion 33.48b, the quasi-steady mass balance for
Example 10 Determine the quasi-steady average this situation is
global temperature rise in a mechanically

ventilated enclosure with dimensions of 18.3 m ρgl V_ ext ¼ ρo V_ ext V_ exp ð33:55Þ
by 12.2 m by 6.1 m with an air injection rate of
ten air changes per hour in response to a fire with For this situation the temperature ratio is cal-
a constant heat release rate of 500 kW. Assume a culated to be
constant heat loss fraction of 0.70 and an ambient
temperature of 20 C (293 K). T gl ρo V_ ext
¼ ¼ ð33:56Þ
T o ρgl V_ ext V_ exp
Solution For this example, the volumetric
expansion rate, V_ exp , is calculated as in previous The dimensionless temperature rise above
examples as ambient for this extraction scenario is
1090 F.W. Mowrer
ΔT gl V_ exp only and extraction only analyses. If V_ net is

¼ ð33:57Þ
To _
V ext V_ exp positive, which will occur if V_ inj þ V_ exp is greater
than V_ ext , then the global temperature rise can be
Note the similarity between Equations 33.54 calculated with Equation 33.52. If V_ net is nega-
and 33.57. The extraction case is analogous
tive, which will occur if V_ ext is greater than
to the injection scenario with an effective injec-
tion rate equal to the difference between the V_ inj þ V_ exp , then the global temperature rise can
extraction and expansion rates. This difference be calculated with Equation 33.57.
is the rate at which air is being drawn into the Neglecting the injection of fuel into the enclo-
enclosure when the extraction rate exceeds the sure, the quasi-steady limits for different species
expansion rate. can be evaluated by setting the left-hand side of
Equation 33.50a to zero and solving for the mass
Example 11 Determine the quasi-steady average fraction, Yi:
global temperature rise in a mechanically ρo Y i, o V_ inj þ m_ i, gen
ventilated enclosure with dimensions of 18.3 m Yi ¼
ρgl V_ inj þ V_ exp
by 12.2 m by 6.1 m with an air extraction rate of
ten air changes per hour in response to a fire with m_ i, gen
¼ Y i, o þ ð33:58Þ
a constant heat release rate of 500 kW. Assume a ρo V_ inj
constant heat loss fraction of 0.70 and an ambient
temperature of 20 C (293 K).
Example 12 Assume propylene (C3H6) is the
fuel burned in the previous mechanically
Solution This example is the same as the previ-
ventilated examples. Assume propylene has a
ous example, except that the mechanical ventila-
heat of combustion of 46.4 MJ/kg of fuel and
tion system is extracting smoke at ten changes
a soot yield of 0.095 g of soot per g of fuel.
per hour instead of injecting air at this rate. For
Estimate the quasi-steady average mass concen-
this case, the volumetric extraction rate, V_ ext , is
tration of soot within the 18.3 m by 12.2 m by
the same as the volumetric injection rate from the
6.1 m enclosure for a fire that burns at a constant
previous example:
heat release rate of 500 kW, assuming the enclo-
10ð18:3 12:2 6:1Þ m 3 sure is mechanically ventilated with an injection
V_ ext ¼ ¼ 3:78 m3 =s system at ten air changes per hour. Based on this
3600 s
soot mass concentration, estimate the visibility
The quasi-steady global temperature rise is distance for a light-reflecting sign.
calculated by applying Equation 33.57:
" # Solution For this example, the ambient mass
V_ exp
ΔT gl ¼ To fraction of soot is assumed to be zero and the
V_ ext V_ exp soot mass generation rate is calculated with

Equation 33.20 as
0:42
¼ 293 K ¼ 36:6 K
ð3:78 0:42Þ
Q_ f
m_ i, gen ¼ f i m_ f ¼
The extraction scenario results in a slightly ðΔH c = f i Þ
higher average temperature rise than the injec- 500 kW
tion scenario because the airflow rate is lower for ¼
46, 400 kJ=kg f = 0:095 kgsoot =kg f
the extraction scenario. This assumes that the
ventilation rates remain constant and the heat ¼ 1:02 103 kgsoot =s
loss fractions remain the same for both scenarios.
Scenarios with both injection and extraction Then the soot mass fraction is calculated with
can be considered as variations on the injection Equation 33.58 as
www.ebook777.com
1:02 103 kgsoot =s Case 1—Leakage paths near the floor (to and
Y soot ¼ 0 þ from the lower layer)
ð1:20 kg=m3 Þ ð3:78 m3 =sÞ
Case 2—Leakage paths near the ceiling (to and
¼ 2:26 104 kgsoot =kgmix
from the upper layer)
The average soot mass concentration is In Case 1, ambient air is expelled from, or
calculated based on the average temperature drawn into, the enclosure through leakage paths
and density within the enclosure as near the floor until the smoke layer descends to
the floor. At that point, smoke at the upper layer
353 temperature and composition would be expelled
ρY soot ¼ kg=m3
ð293 þ 32:5Þ if V_ net were positive. In Case 2, smoke at the upper

2:26 104 kgsoot =kg layer temperature and composition would be
expelled from the upper layer through the leakage
¼ 2:45 104 kgsoot =m3
paths if V_ net were positive, whereas ambient air
Using this value for the soot mass concentra- would be drawn into the smoke layer through
tion, along with the specific light extinction coef- these leakage paths if V_ net were negative.
ficient of 8700 m2/kg suggested by Mulholland The rate of smoke layer descent and conditions
and Croarkin, the extinction coefficient is deter- within the smoke layer will depend on the
mined to be elevations of vents and the injection and extrac-
tion rates of the ventilation systems. For the pres-
K ¼ K m ρY soot
ent discussion, it is assumed that all injection
¼ ð8700 m2 =kgsoot Þ 2:45 104 kgsoot =m3 vents are located at one elevation, zinj, whereas
¼ 2:13 m1 all extraction vents are located at another unique
elevation, zext. Multiple elevations for either
The visibility distance for a light-reflecting injection or extraction vents are not addressed.
sign is then estimated as The elevation of the injection vents determines
whether air is being injected into the upper layer
S ¼ 3=2:13 m1 ¼ 1:4 m ð4:6 ftÞ
V_ inj, u or the lower layer V_ inj, l , depending on
As a final comment on the global analysis of the current elevation of the smoke layer interface.
mechanically ventilated enclosure fires, it is If the elevation of the injection vents is below the
worth noting that V_ net can switch during the smoke layer interface position, then air is injected
course of a fire scenario from negative to posi- into the lower layer; otherwise, it is injected into
tive, for example, for the case of a growing fire the upper layer. As the smoke layer interface posi-
tion moves during a fire scenario, injection can
where V_ exp increases with time; or from positive
shift between the upper and lower layers. Simi-
to negative, for example, for the case where an
larly, the elevation of the extraction vents
extraction system is started at some time after the
determines whether smoke is being extracted
fire starts. As a consequence, the appropriate
from the upper layer V_ ext, u or air is being
equations used to calculate transient and quasi-
steady conditions may change over the course of extracted from the lower layer V_ ext, l . For the
a fire scenario. analysis presented here, it is assumed that no
mixing occurs between the upper and lower layers.
Smoke Layer Analysis with Mechanical

Ventilation Floor Leak (Case 1) Analysis
In this section, conditions within the descending The rate of change of the upper layer volume can
smoke layer are addressed in terms of the two be expressed in terms of the volumetric flow rates
limit cases illustrated in Fig. 33.10. These are into and out of the smoke layer. For Case 1, it is
1092 F.W. Mowrer
assumed that there is no leakage from the upper Based on the Zukoski [25] entrainment
layer until the smoke layer descends to floor correlation, the value for kv will be approxi-
level. The rate of change of the smoke layer mately 0.064 m4/3 s1 kW–1/3. For other plume
volume or depth can be expressed as geometries, such as window plumes, balcony
spill plumes, or line plumes, other entrainment
dV u dzu
¼ As ¼ V_ pl þ V_ exp þ V_ inj, u V_ ext, u rate relationships exist [10]. Further information
dt dt on plume entrainment is provided by Beyler [29],
ð33:59Þ who prepared a comprehensive review of fire
The last two terms on the right-hand side of plume and ceiling jet correlations, and by
Equation 33.59 distinguish the mechanically Quintiere and Grove [30], who more recently
ventilated Case 1 from the unventilated Case reviewed the literature on different types of
1. If air is injected into the upper layer, then the plumes and developed correlations for fire
rate of smoke layer descent will be faster than for plumes of different geometries.
the unventilated case. If smoke is extracted from Once the extraction rate needed to maintain
the upper layer, then the rate of smoke layer the smoke layer interface at an elevation zi above
descent will be slower than for the unventilated the floor is determined, Equations 33.52 through
case. The injection or extraction of air from the 33.58 can be used to evaluate the global effects
of this extraction rate. These effects will included
lower layer does not influence the smoke layer
descent rate for the Case 1 scenario. These flows the makeup air requirements V_ net to balance
simply affect the net flow rate through the leak- the design exhaust rate as well as the leakage
age path in the lower layer boundary for this opening area requirements to prevent excessive
scenario. pressure drop and flow velocities across the
Provided the upper layer extraction rate is at enclosure boundaries. As indicated by Equa-
least large enough to offset the expansion and tion 33.47, mechanical injection of air into the
injection rates, the smoke layer interface will lower layer can be used to reduce the makeup air
eventually equilibrate at the elevation where a requirements and consequently the pressure drop
balance is struck between the rates of flow into and flow velocities across openings in the enclo-
and out of the smoke layer: sure boundaries.
At this point, conditions within the smoke
V_ ext, u ¼ V_ pl þ V_ exp þ V_ inj, u ð33:60Þ layer are considered. For Case 1, mass conserva-
tion for the upper layer can be expressed in terms
Equation 33.60 represents the typical situation
of the various volumetric flow rates into and out
for a smoke management system designed
of the smoke layer:
according to the exhaust method. For this situa-
tion, air is not typically injected into the upper

SFP Han of Fir Pro Eng 5 TH Edi

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

SFP Han of Fir Pro Eng 5 TH Edi

Enviado por

Direitos autorais:

Formatos disponíveis

free ebooks ==> www.ebook777.

SFPE Handbook of Fire Protection

ISBN 978-1-4939-2564-3 ISBN 978-1-4939-2565-0 (eBook)

Library of Congress Control Number: 2015953225

Springer New York Heidelberg Dordrecht London

Printed on acid-free paper

This edition of the SFPE Handbook of Fire Protection

No other single event had as significant an impact on

Acknowledgment of Past Authors

Name 1st edition 2nd edition 3rd edition 4th edition

The editors of the SFPE Handbook of Fire Protection Engineering have

1 Introduction to Fluid Mechanics . . . . . . . . . . . . . . . . . . . . 1

14 Ceiling Jet Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429

32 Modeling Fires Using Computational Fluid

47 Foam Agents and AFFF System Design

62 Combustion Toxicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2207

78 Data for Engineering Analysis . . . . . . . . . . . . . . . . . . . . . . 3073

Appendix 1 Conversion Factors . . . . . . . . . . . . . . . . . . . . . . . 3397

Introduction to Fluid Mechanics

Fluid Properties certain time is defined as in Equation 1.1, taking

M.J. Hurley (ed.), SFPE Handbook of Fire Protection Engineering, 1

In a mixture of fluids (see below), one species

Conduction Coefficient Dimensionless Groups of Fluid

(through exchange of momentum), α tries to Internal Energy

In liquids, the density is essentially constant,

The pressure (p) can be defined as the normal

high pressure or temperature, but this is not rele-

A very simple illustration of Equation 1.15 is !

τxy (x) σxx (x+dx)

‘solenoidal’. τyx (y)

The above results in the Navier–Stokes

In differential formulation, this reads:

hi ðT Þ ¼ hre f , i þ c p, i ðT ÞdT; ð1:33Þ ∇: q ¼ ∇:ðλ∇T Þ ∇: ρ hi Di ∇Y i þ D:E:

For unity Lewis number (Lei ¼ 1 for all i, Buoyancy

temperature correction. Using the thermal volu-

ðρamb ρÞg ¼ ρamb βðT T amb Þg , if βðT T 1 Þ  1: ð1:42Þ

The viscous forces tend to damp the inherent

Examination of the terms in the Navier–Stokes

how it is possible that randomness occurs when ð1:52Þ

Reynolds Averaging v,x ðtÞ ¼ 0; T , ðtÞ ¼ 0; vx ðtÞ

As mentioned in the previous section, the ¼ vx ðtÞ; T ðtÞ ¼ T ðtÞ: ð1:54Þ

Fig. 1.4 Sketch of U

the leading edge of the flat plate, and in the ð1:63Þ

Fig. 1.7 Moody diagram

(i.e. what extraction fans are required to over-

special case concerns the flow into a large

This can be interpreted as a ‘correction’ to the

In Equation 1.84, the final term reflects the pres-

z2 2 Wind is an important factor in fire protection

Fig. 1.11 Schematic

Plan Rear Eddy Wind

Conduction of Heat in Solids

Introduction scenarios. The treatise of Carslaw and Jaeger [1]

M.J. Hurley (ed.), SFPE Handbook of Fire Protection Engineering, 25

Thermodynamic Properties necessary to define the heat transfer rate in terms

heat transfer coefficient is proportional to the

tACT u1=2 ρLc

This combination of parameters is the well

Table 2.1 Characteristic quantities used in microscale carrier model of conduction

Fig. 2.2 Range of thermal

Under conditions in which there is a primary heat

Fig. 2.5 Development of pin fin equation for cylindrical rod

Simple 1D Composite Systems Example 2 A composite wall for a furnace is

ðρamb ρÞg ¼ ρamb βðT T amb Þg , if βðT T 1 Þ 1: ð1:42Þ