Experimental and Applied Mechanics Volume 4 PDF

Conference Proceedings of the Society for Experimental Mechanics Series
Series Editor
Tom Proulx
Society for Experimental Mechanics, Inc.,
Bethel, CT, USA
For further volumes:

http://www.springer.com/series/8922
Carlos E. Ventura • Wendy C. Crone • Cosme Furlong
Editors
Experimental and Applied Mechanics,

Volume 4
Proceedings of the 2012 Annual Conference on Experimental
and Applied Mechanics
Editors
Carlos E. Ventura Wendy C. Crone
University of British Columbia University of Wisconsin
Vancouver, BC, Canada Madison, WI, USA
Cosme Furlong
Worcester Polytechnic Institute
Worcester, MA, USA
ISSN 2191-5644 ISSN 2191-5652 (electronic)

ISBN 978-1-4614-4225-7 ISBN 978-1-4614-4226-4 (eBook)
DOI 10.1007/978-1-4614-4226-4
Springer New York Heidelberg Dordrecht London
Library of Congress Control Number: 2012948609
# The Society for Experimental Mechanics, Inc. 2013

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.
Exempted from this legal reservation are brief excerpts in connection with reviews or scholarly analysis or material supplied specifically for the purpose of
being entered and executed on a computer system, for exclusive use by the purchaser of the work. Duplication of this publication or parts thereof is permitted
only under the provisions of the Copyright Law of the Publisher’s location, in its current version, and permission for use must always be obtained from
Springer. Permissions for use may be obtained through RightsLink at the Copyright Clearance Center. Violations are liable to prosecution under the
respective Copyright Law.
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.
While the advice and information in this book are believed to be true and accurate at the date of publication, neither the authors nor the editors nor the
publisher can accept any legal responsibility for any errors or omissions that may be made. The publisher makes no warranty, express or implied, with
respect to the material contained herein.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)

Preface
Experimental and Applied Mechanics, Volume 4: Proceedings of the 2012 Annual Conference on Experimental and Applied
Mechanics represents one of seven volumes of technical papers presented at the Society for Experimental Mechanics SEM
12th International Congress & Exposition on Experimental and Applied Mechanics, held at Costa Mesa, California, June
11–14, 2012. The full set of proceedings also includes volumes on Dynamic Behavior of Materials, Challenges in Mechanics
of Time-Dependent Materials and Processes in Conventional and Multifunctional Materials, Imaging Methods for Novel
Materials and Challenging Applications, Mechanics of Biological Systems and Materials, MEMS and Nanotechnology, and
Composite Materials and Joining Technologies for Composites.
Each collection presents early findings from experimental and computational investigations on an important area within
Experimental Mechanics. The conference track Experimental and Applied Mechanics was organized by: Carlos E. Ventura,
University of British Columbia; Wendy C. Crone, University of Wisconsin–Madison; and Cosme Furlong, Worcester
Polytechnic Institute.
This book covers a wide variety of subjects that are related to the broad field of experimental or applied mechanics. It is
SEM’s mission to disseminate information on a good selection of subjects. To this end, research and application papers
relate to the broad field of experimental mechanics. The current volume on experimental and applied mechanics includes
studies on:
• Fracture and Fatigue
• Microscale and Microstructural Effects on Fatigue and Fracture
• Material Applications
• Composite Characterization Using Digital Image Correlation Techniques
• Multi-scale Simulation and Testing of Composites
• Residual Stress
• Inverse Problems/Hybrid Methods
• Nano-composites
• Microstructure Material Characterization
• Modeling and Uncertainty Quantification
• Impact Behavior of Composites
The organizers would like to thank the session organizers and chairs, presenters, and authors for their participation.
The opinions expressed herein are those of the individual authors and not necessarily those of the Society for
Experimental Mechanics, Inc.
Vancouver, BC, Canada Carlos E. Ventura

Madison, WI, USA Wendy C. Crone
Worcester, MA, USA Cosme Fulong
v
Contents
1 Correlation of Structural Strain to Tip Vorticity and Lift for a MAV Pliant Membrane Wing . . . . . . . . . . . . 1
Trenton Carpenter, Cody Ray, and Roberto Albertani
2 Deformation Mapping at the Microstructural Length Scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Adam Kammers and Samantha Daly
3 Molecular Tailoring of Interfacial Adhesion Using Self-Assembled Monolayers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Martha E. Grady, Mark D. Losego, Paul V. Braun, and Nancy R. Sottos
4 Micromechanical Characterization of Ductile Damage in DP Steel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
J.P.M. Hoefnagels, C.C. Tasan, F.J. Peters, and M.G.D. Geers
5 MEMS Device for Fatigue Testing of 25 mm Thick Aluminum Specimens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
L.A. Liew, D.T. Read, and N. Barbosa III
6 High Confidence Level Calibration for AFM Based Fracture Testing of Nanobeams . . . . . . . . . . . . . . . . . . . . . . . 43
Scott Grutzik, Richard Gates, Yvonne Gerbig, Robert Cook, Melissa A. Hines, and Alan Zehnder
7 High Strain Rate Characteristics of Fiber Bragg Grating Strain Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
Wun-Jheng Lin, Liren Tsai, Cia-Chin Chiang, and Shih-Han Wang
8 Dynamic Shape and Strain Measurements of Rotating Tire in Time-Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Motoharu Fujigaki and Kousuke Shimo
9 Measurement of Metallic Adhesion Force-to-Elongation Profile Under High
Separation-Rate Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
D. Bortoluzzi, M. Benedetti, C. Zanoni, and J.W. Conklin
10 Prediction of Constraint Parameters Along the 3D Crack Front Under Negative Biaxial Loadings . . . . . . . 75
Ragupathy Kannusamy and K. Ramesh
11 Stress Intensity Factors for Viscoelastic Axisymmetric Problems Applied to Wood . . . . . . . . . . . . . . . . . . . . . . . . . 89
Rostand Moutou Pitti, Claude Chazal, Florence Labesse-Jied, and Yuri Lapusta
12 Role of Viscoelasticity in Predicting the Shape Memory Effect of Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Diani Julie, Gilormini Pierre, Frédy Carole, and A.R. Ingrid
13 Damage Assessment in Syracuse Limestone Specimens by Frequency Analysis of Elastic Emissions . . . . . . 105
A. Schiavi, G. Niccolini, A. Carpinteri, and G. Lacidogna
14 Photogrammetry Measurements During a Tanking Test on the Space Shuttle
External Tank, ET-137 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Justin Littell, Tim Schmidt, John Tyson, Stan Oliver, Matt Melis, Charles Ruggeri,
and Duane Revilock
15 Experimental, Numerical and Analytical Evaluation of Structural Integrity
of Experiment Container Subjected to Reduced Gravity Flight Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Mujahid Umar, Michael G. Jenkins, and Thomas L. Nguyen
vii
viii Contents
16 Use of Fiber Bragg Grating Strain Gages on a Pipeline Specimen Repaired

with a CFRE Composite System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
J.L.F. Freire, V.A. Perrut, A.M.B. Braga, R.D. Vieira, A.S.A. Ribeiro, and M.A.P. Rosas
17 Microstructure in the FSW Butt Joint of Aluminum Alloy AA5083 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
Thaiping Chen, Wei-Bang Lin, and Chang-Mao Hung
18 Load History Dependency of Plastic Collapse Behavior of Cracked Piping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
R. Suzuki, M. Matsubara, M. Suzuki, T. Shiraishi, S. Yanagihara, K. Sakamoto, and S. Izawa
19 Production of Metal Matrix Composites Using Thixoinfiltration Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
M.H. Robert, L.M.P. Ferreira, A.F. Jorge, and E. Bayraktar
20 Use of Vision-Based Methods for Measurement of Mechanical Properties
of Exhaust Catalysts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
Andrew P. Dickinson, Paul J. Gloeckner, Randy J. Stafford, and Kevin B. Golovin
21 Review of Methods for Determining Residual Stresses in Biological Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
Drew Nelson
22 A Comparison of Residual Stress Measurements on a Linear Friction Weld
Using the Contour Method and Neutron Diffraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
Adrian T. DeWald, Daira Legzdina, Bjørn Clausen, Donald W. Brown,
Thomas A. Sisneros, and Michael R. Hill
23 Experimental Study of High Speed Ball End Milling of Titanium Alloy (Ti-6Al-4V) . . . . . . . . . . . . . . . . . . . . . . . . 191
M.-H. Baccar, E. Bayraktar, T. Rickert, M. Boujelbene, and D. Katundi
24 Very High-Cycle Fatigue Resistance of Shot Peened High-Strength Aluminium Alloys . . . . . . . . . . . . . . . . . . . . . 203
M. Benedetti, V. Fontanari, and M. Bandini
25 The Application of the Combined Creep and Rate-Independent Plasticity Model
to Turbine Components Life Prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
Alexander Staroselsky and Brice N. Cassenti
26 Modified Constitutive Relation Error Strategy for Elastic Properties Identification . . . . . . . . . . . . . . . . . . . . . . . . . 221
M. Ben Azzouna, P. Feissel, and P. Villon
27 Identification of the Anisotropic Plastic Behaviour of Sheet Metals at Large Strains . . . . . . . . . . . . . . . . . . . . . . . . 229
Marco Rossi, Fabrice Pierron, Michaela Štamborská, and František Šimčák
28 Cohesive Zone Law Extraction from an Experimental Peel Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
Christopher Kovalchick, Shuman Xia, and Guruswami Ravichandran
29 Application of a New Experimental Method to Determine Bi-Material
Interface Bonding Strength . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
M.A.K. Chowdhuri and Z. Xia
30 Calibration of Barkhausen Noise for Residual Stress Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
C. Casavola, C. Pappalettere, and F. Tursi
31 Film Residual Stress Assessment Method via Temporarily Thermal Relaxation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
Xin-Xin Cheng and Chen-Wu Wu
32 Determining Residual Stresses with the Aid of Optical Interference Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
D. Findeis and J. Gryzagoridis
33 Interfacial Fracture Toughness and Residual Stress of Thermally Sprayed Coatings . . . . . . . . . . . . . . . . . . . . . . . 285
Yoshifumi Okajima, Toshio Nakamura, and Sanjay Sampath
34 Design of Ultrasonic Probe Configuration Using Finite-Difference
Time Domain Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
Rahul Sharma, Shyamsunder Baskaran, and H. Murthy
Contents ix
35 Absorbed Versus Released Energy in the Cracking Process

of Heterogeneous Materials Under Compression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
G. Lacidogna, M. Corrado, and A. Carpinteri
36 Detailed Measurements of Thread Deformation and Failure in Thin Walled
Aluminum Alloy Joints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
Bonnie R. Antoun, Spencer Grange, Gerald W. Wellman, and J. Franklin Dempsey
37 Simulation of Interference Fitted Joint Strength as Used in Hot Rolling Work Rolls . . . . . . . . . . . . . . . . . . . . . . . 319
Martin McMillan, Julian Booker, and David Smith
38 Modelling of Nosing for the Assembly of Aerospace Bearings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
Johnpaul Woodhead and Julian Booker
39 Multi-axial Quasi-static Strength of a Clinched Sheet Metal Assembly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
S. Coppieters, P. Lava, H. Sol, P. Van Houtte, and D. Debruyne
40 Development of a Test Simulator to Perform Optimized Experiment Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
Marco Rossi, Fabrice Pierron, Lava Pascal, and Debruyne Dimitri
41 Simplified Calculation Method of the Torsion Effect on the Seismic Behaviour
of Timber Building . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
Thanh Kien Vu, Eric Fournely, Rostand Moutou Pitti, and Abdelhamid Bouchair
42 Experimental Analysis of Semi-rigidity Joint in the Standardized Timber Beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357
Eric Fournely, Rostand Moutou Pitti, Abdelhamid Bouchair, and Frédéric Dubois
43 Nano-Macro Correlation of Nano-Silica Concrete . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367
Joan Schoepfer and Arup Maji
44 Experimental Results of Green Wood Slice Under Natural Drying . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
Rostand Moutou Pitti, Frederic Dubois, and Eric Fournely
45 Thickness Effect on Cross-Ply GLARE 5 FML Beams Subjected to Ballistic Impact . . . . . . . . . . . . . . . . . . . . . . . 379
Ali Seyed Yaghoubi and B. Liaw
46 Hole-Drilling Residual Stress Measurement in an Intermediate Thickness Specimen . . . . . . . . . . . . . . . . . . . . . . . 389
Colin Abraham and Gary S. Schajer
47 Effects of Strain’s Error on Residual Stresses Calculated by HDM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395
C. Casavola, G. Pappalettera, C. Pappalettere, and F. Tursi
48 Hole-Drilling Residual Stress Measurement with Artifact Correction Using Full-Field DIC . . . . . . . . . . . . . . . 403
G.S. Schajer, B. Winiarski, and P.J. Withers
49 Anisotropic and Pressure-Dependent Plasticity Modeling for Residual Stress Prediction . . . . . . . . . . . . . . . . . . . 415
Michael B. Prime
50 Simulation of Triaxial Residual Stress Mapping for a Hollow Cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429
Mitchell D. Olson, Wilson Wong, and Michael R. Hill
Chapter 1
Correlation of Structural Strain to Tip Vorticity and Lift
for a MAV Pliant Membrane Wing
Trenton Carpenter, Cody Ray, and Roberto Albertani
Abstract The last decade has witnessed a surge of scientific interest in flight at the micro air vehicle (MAV) scale. To date,
a MAV utilizing an adaptable, flexible smart wing design has yet to come to fruition. While highly flexible aerodynamic
surfaces have repeatedly demonstrated improved performance through passive adaptation, limited structural and flow state
knowledge prevents the inclusion of active control strategies which could improve performance of such designs.
In this work, a flexible membrane wing constructed of latex was considered. The goal of estimating lift was approached
through experimental and theoretical correlation of wing strain state due to flow-induced pressure. Using visual image
correlation, elastic deformations, strains and membrane pretensions of the wing were measured in the Oregon State
University wind tunnel. In addition, a six-degree-of-freedom sting balance was used to determine aerodynamic loads.
A linear membrane wing formulation was utilized to correlate the wing’s structural strain to lift and wing-tip vorticity.
Results of the forces measured by a sting balance were then compared to those predicted by flow simulation. This work
describes experimental results that illustrate the effectiveness of low fidelity models in predicting and estimating useful
information for flexible wing designs.
Keywords MAV • Membrane wing • Tip vorticity • Load identification • Lift estimation • Visual image correlation • Digital
image correlation • Wind tunnel testing
1.1 Introduction
Scientific and engineering interest in small unmanned aerial vehicles (UAVs) and micro aerial vehicles (MAVs) has
increased substantially in the last decade. Such craft are expected to benefit military, law, and search and rescue operations
in a variety of ways. However, to date, a MAV utilizing an adaptable, robust, flexible smart wing design has yet to come to
fruition. While structural knowledge of such designs is well known and described in literature, the problems of control and
estimation of relevant states on board such small craft (with limited computational power) is, as of yet, restricting the actual
implementation of such craft in the field.
Flexible aerodynamic surfaces have repeatedly demonstrated improved performance through passive adaptation to
incident flow and disturbances. This passive behavior and proprioceptive sensing is utilized throughout the natural world
by biological fliers such as birds, bats and insects. However, from a control design perspective, such wings present a
challenge in traditional actuation approaches. Furthermore, limited structural and flow state knowledge prevent the inclusion
of active control strategies which could improve performance of such designs. In response to these difficulties, a study of a
flexible membrane wing constructed of latex was considered. Through experimental and theoretical correlation of wing strain
state due to flow-induced pressure resultants, the goal of estimating lift was approached. Using visual image correlation
(VIC), elastic deformations, strains and membrane pretensions of the wing were measured in the Oregon State University
(OSU) wind tunnel along with true values for lift and drag obtained via a six degree-of-freedom sting balance. A linear partial
differential equation relating pressure distribution to membrane deflection was used to relate and correlate wing structural
strain to lift and wing-tip vorticity via approximating lift resultant and applying basic linear aerodynamic principles.
T. Carpenter (*) • C. Ray • R. Albertani

School of Mechanical, Industrial and Manufacturing Engineering (MIME), Oregon State University, Corvallis, OR, USA
e-mail: carpentr@engr.orst.edu
C.E. Ventura et al. (eds.), Experimental and Applied Mechanics, Volume 4: Proceedings of the 2012 Annual Conference 1
on Experimental and Applied Mechanics, Conference Proceedings of the Society for Experimental Mechanics Series 34,
DOI 10.1007/978-1-4614-4226-4_1, # The Society for Experimental Mechanics, Inc. 2013
2 T. Carpenter et al.
Results from the aerodynamic forces measured by the sting balance can be compared to those predicted by flow
simulation. Preliminary results indicate that such approaches to lift/circulation estimation do indeed correlate well to the
true values. These approaches and results are discussed along with possible future research directions.
1.2 Experimental Setup
During testing our objective was to determine the out-of-plane deformation of the membrane wing under steady aerody-
namic loads, in different steady conditions of angle of attack (AOA) and free stream velocity. The low-speed wind tunnel at
OSU was used to conduct this phase of testing which can be seen in Fig. 1.1. The close loop, close test section wind tunnel is
capable of speeds from 1 to 18 meter per second (m/s) and has a 1.3 1.5 m test section. The installation of the test wing can
also be seen mounted to a one degree of freedom motion rig inside the test section in Fig. 1.1. The apparatus was used for
AOA sweeps in steady state flow conditions. The angle of the apparatus was measured by an inclinometer sensor with an
accuracy of 0.1 . Loads were measured using a six degree-of-freedom strain gauge load cell. The load cell was capable of
measuring loads up to 100 N normal force and 200 N axial force with a resolution of 0.05 N and 0.10 N respectively.
The wind tunnel flow velocity was monitored by a pitot tube probe installed in the test section ahead of the model connected
to a pressure transducer with a resolution of 0.05 mmH20. Air temperature was monitored by a J-type thermocouple sensor
mounted inside the test section. All channels were monitored simultaneously and recorded during testing.
For this test, the three main testing variables considered were membrane pre-strain (PS), wing AOA and flow velocity
(V). To design the experiment to yield results which would be readily relatable to one another, with regard to pre-strain and
out-of-plane deformation, ∏2 values as defined by [1], were considered. The values for ∏2 are defined as follows,
eps t
P2 ¼ ; (1.1)
qc
where εps is the membrane pre-strain, t is the membrane thickness, q the dynamic pressure applied to the membrane, and c
the wing chord length. Using a velocity range known to be relevant to MAV’s with a Reynolds number (Re) between 60,000
and 90,000, values of pre-strain were chosen such that the maximum value of ∏2 for a given membrane would span the
minimum range of the next membrane PS. The values of ∏2 chosen for the test can be seen in Fig. 1.1 for three different pre-
strains of 5, 3.5 and 2% strain.
Values for AOA were chosen such that the wing would not experience flow separation, i.e. remain in the linear region of
lift, thus resulting in predicable behavior easily characterized by linear aerodynamic theory. From preliminary testing
results, a linear lift behavior was observed between AOA’s of 0–10 for membranes between 2% and 5% pre-strain.
Fig. 1.1 Exterior view of the test section of the low speed closed loop, closed test section wind tunnel at OSU (left); Values of ∏2 from Eq. 1.1,
velocity and PS chosen for wind tunnel testing (right)
1 Correlation of Structural Strain to Tip Vorticity and Lift for a MAV… 3
Table 1.1 Wind tunnel test matrix

Test 1 Test 2 Test 3
PS [%] AoA V [m/s] PS [%] AoA V [m/s] PS [%] AoA V [m/s]
2 3 12 3.5 3 12 5 3 12
15 15 15
18 18 18
6 12 6 12 6 12
15 15 15
18 18 18
9 12 9 12 9 12
15 15 15
18 18 18
Fig. 1.2 Schematic showing the wing frame (dark gray), the frame’s dimensions and domain of the individual membrane areas (light gray); all
values in millimeters
Due to interest in tip vorticity, it was also desired to have sufficiently high AOA such that strong predictable vortices would
be generated. From these requirements an AOA range of 3–9 was chosen. Three values for each of the three variables were
chosen. This yielded a factorial design space of 33 or 27 tests. The resulting test matrix can be seen in Table 1.1.
For the final phase of testing a rectangular planform wing with an aspect ratio of ~4 was chosen. A rectangular domain
lends itself to linear theory structurally and is well-characterized aerodynamically. Aerodynamically, it is well-known that
rectangular wings generate strong wing-tip vortices, relative to other planforms. The tip vortices contain a significant amount
of the wake flow energy and are highly indicative of the induced drag for a rectangular shape compared to other planforms.
This allows closer comparison and correlation of strain to such aerodynamic effects, although any planform would exhibit
the same correlation behavior due to the relationship of aerodynamic effects and structural deformation in flexible wings.
An aspect ratio between 4 and 6 was chosen to be large enough to apply traditional aerodynamic analysis, such as lifting line
theory, to the model in order to immediately compare measured and estimated quantities including lift and circulation.
Structurally, a rectangular membrane under sufficient pretension exhibits behavior captured by the Poisson equation on a
rectangular domain. This simplifies computation greatly and allows for a convenient rectangular domain to be defined and
used directly with VIC data. The rectangular steel frame geometry used for this study can be seen in Fig. 1.2.
Full-field measurements of strain and deflection of the membrane wing were performed using a visual image correlation
(VIC) technique originally developed by researchers at the University of South Carolina [2, 3]. Images are captured with two
high-speed AOS S-PRI cameras, capable of 1,000 frames per second and capable saving 1,000 frames at a resolution of
1,280 1,024 in an in-camera flash-memory buffer. The cameras were positioned outside and above the wind tunnel and
viewed through a clear window. The VIC reference image to measure displacement was of the wing at the specified AOA,
with zero wind velocity. Because this condition was used to take the reference images, the pretension would not appear
directly in the VIC results. The sequence of events for a given test was first to take a picture of the model at the set angle, with
wind off. Next was to set the airspeed to the desired dynamic pressure, and when stable conditions were reached, take the
pictures of the deformed wing as well as dynamic loads. Then the wind was stopped, and the model was moved to the next
AOA. At the same time, the system was recording the aerodynamic loads, wind velocity and model AOA. This sequence was
repeated for three wings with different membrane pretensions for the selected nominal AOA of 3, 6 and 9 at the selected
nominal wind velocities of 12, 15, and 18 m/s. The aerodynamic loads of a nominally rigid wing, built from a plate of thin
sheet aluminum and attached to the same frame were also measured.
1.3 Force Estimation from Vorticity
In this approach, wind tunnel test data, numerical fluid models, and analytical structural models were used in order to
estimate lift. Using VIC strain data from testing, the out-of-plane deformation of the membrane was reconstructed to fit a
quadratic basis function using pseudo strain gauge sensors. Using the surface reconstruction information, a 3D solid model
of the deformed membrane wing was generated and a computational fluid dynamics (CFD) analysis was performed to
calculate wake behavior. Using the wake information generated from this analysis, specifically wing tip vorticity, aerody-
namic lift was estimated.
1.3.1 Surface Reconstruction
In this work, it is assumed that the Poisson equation could be used to describe the membrane wing used in experiment, and
proceed to use a least squares approach to estimate membrane deformation from pseudo strain sensors strategically placed
throughout the membrane domain. To create the strain sensor, VIC stain data was collected over the full planform of the
wing during wind tunnel testing. The full field strain domain was then partitioned into smaller individual areas, or “patches,”
representing the strains present on a sensor in that given location and over that given area; an example can be seen later in the
text in Fig. 1.9. Using this technique, any combination of sensor quantity, size and location could be used.
Assuming the wake flow from the wing can be captured by a low order dimensional model, which has shown to be
a reasonable assumption [4–6]; the pressure distribution applied to the membrane is assumed to be constant, resulting in a
deformation which can be represented by a quadratic function. Using this assumption, the out-of-plan deformation of the
membrane over the domain, 0 ≤ x ≤ a and 0 ≤ y ≤ b, where a and b represent the dimension of the membrane boundary in
the x and y direction respectively, as illustrated in Fig. 1.2. Thus, the out-of plane deformation for the membrane, w, was
modeled as,
axðx aÞyðy bÞ
wðx; yÞ ¼ ; (1.2)
wða=2; b=2Þ
where a is a constant which represents the max amplitude of the membrane out-of-plane deformation. Independently
differentiating equation (1.2) with respect to x and y, yields,
@w a
¼ ð2x aÞyðy bÞ
@x wða=2; b=2Þ
@w a
¼ xðx aÞð2y bÞ: (1.3)
@y wða=2; b=2Þ
For a membrane subject to an evenly distributed load with small deformations, it is assumed that in-plane deformations
are negligible and thus, strains are purely a result of out-of-plane deformation. Using this assumption, directional strains can
be approximated as,

1 @w 2 1 @w 2
ex ¼ ; ey ¼ : (1.4)
2 @x 2 @y
Using Eqs. 1.2 through 1.4, an analytic formulation of strain over the wing domain 0 ≤ x ≤ a and 0 ≤ y ≤ b, can be
defined as,
ex þ ey
Cðx; yÞ ¼ ;
2
a2 h i
Cðx; yÞ ¼ ðð2x aÞyðy bÞÞ2 þ ðxðx aÞð2y bÞÞ2 : (1.5)
wða=2; b=2Þ2
Assuming a strain sensor is used on the membrane surface which provides a non-directional strain output and has an area
of si, VIC strain data over that area can be used to represent the sensor output as,
e xi þ e yi
ei ¼ ; where
Z2
1
e xi ¼ ex ðx; yÞds;
si si
Z
1
e yi ¼ ey ðx; yÞds; (1.6)
si si
where εx(x,y) and εy(x,y) represent values of VIC strain measurements. Given n number of strain sensors, the measured strain
for each sensor on the wing can be defined as,
2 3
e1
6 e2 7
6 7
S ¼ 6 .. 7 (1.7)
4 . 5
en
Using Eq. 1.5, corresponding analytical values to those in Eq. 1.7 can be defined. The analytic strain at each sensor
position can be defined as,
C ¼ a2 ½ C 0 1 C0 2 C0 n (1.8)
where C0 i ¼ Cii =a2 . By observing from Eq. 1.2 that a is a constant throughout the domain, a can be factored out of each term
of Eq. 1.8. Finally, by equating Eqs. 1.7 and 1.8, a formulation is derived which can be used to solve for a,
S ¼ a2 C (1.9)
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 ffi
a¼ CT C CT S (1.10)
Solving for a in this manner yields a least squared approximation of the constant a. Given the inherent variability in the
output of the strain sensors, the variability in membrane structural properties, and irregularities in the fluid flow, this
formulation generates a quadratic surface which best describes the actual out of plane deformation of the membrane. Using
this surface reconstruction, a numeric fluid simulation can be performed to determine fluid behavior in the wake of the wing.
1.3.2 Numerical Fluid Model
A three dimensional, steady state, computational model was used to compute the fluid behavior induced by the deformed
membrane wing. For typical MAV’s, Re based on wing chord length are commonly in the range of 10e4 to 10e5. With the
given test wing described above, placed in a flow with a Re of 10e5 in air at 25 C and atmospheric pressure, yields a Mach
number of 0.052, allowing an incompressible fluid assumption to be reasonably applied to the simulation. A symmetrical
boundary condition was applied at the centerline of the wing model in order to reduce computation time. The half span
simulation volume was sized to be five times larger than the chord in the flow direction, two chord lengths above and below
the wing and one to the side of the wing. The volume was discretized into approximately 1.2 million hexahedral volume cells
with a cell width of ~0.1 mm in the near wake region, with cells growing to 10 mm near the simulation boundaries.
The simulation volume is shown in Fig. 1.3. The Reynolds Averaged Navier Stokes equations were solved for an
incompressible flow in conjunction with the Spalart-Allmaras turbulence model. As shown by [7, 8] the Spalart-Allmaras
turbulence model proves to perform well in near wake regions such as wing tip vortex flows while remaining relatively
computationally inexpensive.
The simulation used a computer-aided-design (CAD) model of the deformed membrane and the frame to which the
membrane was attached. The surface of the membrane was modeled to match the quadratic form from Eq. 1.2. For a particular
Fig. 1.3 Half span CFD simulation domain (left) and meshed membrane wing (right) shown in full span
Fig. 1.4 Speckled latex membrane wing during wind tunnel testing (left) and CAD model generated from the strain sensor surface reconstruction
dynamic pressure and AOA, the surface deformation constant from Eq. 1.10 was calculated and a model was generated
accordingly. A side-by-side comparison of the actual deformed membrane during wind tunnel testing and the corresponding
CAD model derived from the sensor surface reconstruction method discussed above, can be seen in Fig. 1.4.
1.3.3 Lift from Circulation
For the purpose of this analysis, a simplified version of lifting line theory was utilized. The horseshoe vortex is a
simplification of Prandtl’s continuous lifting line theory. In this model, all vorticity is assumed to be present in a vortex
of constant circulation bound to the wing. The bound vortex is connected at the wing-tips to two trailing vortices which
extend, in theory, back to the starting vortex itself, thus forming a loop. This results in a shape reminiscent of a horseshoe in
the local region of the wing. An illustration of this behavior can be seen in Fig. 1.5.
The Kutta-Joukowski theorem for steady flight states that lift is proportional to circulation as follows:
@L
¼ rUG (1.11)
@x
Fig. 1.5 Vorticity contours of

a 2% pre-strained membrane
wing at AOA of 6 and 18 m/s
Fig. 1.6 Contour map

showing vorticity 15mm
behind wing; velocity vector
field normal to free stream;
and circulation integration
domain of wing tip vortex
where @L=@x is the lift per unit span (if x is in the span-wise direction and y the chord-wise), r is the fluid density, U is the
free stream velocity, and Γ is circulation per unit span. Circulation can be related to vorticity by Stokes’ theorem:
I Z Z
G¼ V ds ¼ ðr V Þ dS (1.12)
C S
where V is the flow velocity vector field, S is an area containing the vortex and c is the boundary (and a closed curve) of S.
Using flow information from CFD simulations taken downstream of a wing, a velocity flow field can be extracted and
Eqs. 1.11 and 1.12 can be applied in order to derive an estimation of lift generated by the wing. Figure 1.6 shows a vorticity
contour and velocity vector field of the wake flow 15 mm behind a rectangular membrane wing at 2% pre-strain, 6 AOA and
a flow regime corresponding to a Re of 90000. Also in this figure, the area of integration over the tip vortex can be seen,
which was used to calculate the value of circulation per unit span from Eq. 1.12.
1.4 Force Estimation from Poisson Model
The linear Poisson equation, a partial differential equation, can be used to model a membrane under sufficient pretension
undergoing small deformations. This section will briefly outline how the estimation of membrane deformation from strain
outlined above can be used to obtain a rough estimate of the pressure distribution present on the membrane-wing; a quantity
indicative of lift and strength of wing-tip vortex circulation.
The two dimensional Poisson equation is obtained through linearization of nonlinear membrane dynamical equations
and/or minimization of energy of an axially deforming planar structure undergoing small transverse displacements. Such
assumptions allow rotation angles of the structure and internal tension resultants to be represented as spatial derivatives and
scalar constants, respectively. Thus, such a linear model is actually a model capturing the membrane’s resistance to changes
in its internal state of stress. Such approximations have been used to model membranes with success and become
increasingly accurate as pretension increases and maximum deformation decreases. The Poisson equation can be written
for a rectangular domain as,
@2w @2w
Tx þ T y ¼ Fðx; yÞ; (1.13)
@x2 @y2
where Tx and Ty are the pretension resultants in the x and y directions, respectively, and F(x, y) is the resultant pressure
distribution on the membrane. Associated with this equation are boundary conditions for all edges of the domain, describing
fixed edges, i.e. zero displacement.
If one makes an assumption regarding the functional form of the deformation of the membrane, i.e. quadratic in nature,
the approach simplifies greatly. Supposing the membrane deformation can be represented by a single quadratic surface
interpolating the boundaries of the membrane (thus satisfying the boundary conditions of Poisson’s equation explicitly).
Such a surface can be written as that used in Eq. 1.2. Once a is determined from Eq. 1.10, an estimate of the pressure
distribution can be immediately found by substituting the equation for w(x, y) back into Poisson’s equation using the
estimated value of a. Taking second derivatives of w(x, y) and substituting into Eq. 1.13 yields an estimate of F from
the strain values observed on the membrane.
Classical aerodynamic theory states that the derivative of lift with respect to span is proportional to circulation, as
described in Eq. 1.11. Thus, one approach to estimate circulation is to utilize the estimated force from Poisson’s equation and
integrate this result analytically with respect to chord. This would yield an expression that describes the lift per unit span.
If one were to integrate again with respect to span, the total lift resultant would be obtained. Therefore, by the fundamental
theorem of calculus, one can integrate the estimated pressure distribution with respect to x, in order to yield an expression
also describing the lift per unit span. Equating this to circulation using the classical expression for circulation as given in
Eq. 1.11, one may divide by r*U to obtain an estimate for circulation. This provides a first-order model-based approach to
estimating circulation directly from strain.
1.5 Results and Discussion
1.5.1 Aerodynamic Loads
In order to verify lift predictions from sensed strain and CFD analysis, aerodynamic loads were measured for each wing
configuration. The model was swept through AOA from 0 to 20 at an air speed of 18 m/s corresponding to a chord line Re of
90000. Four wing configurations were tested; one with a 0.75 mm thick aluminum plate attached to the wing frame, and three
with 0.13 mm thick latex membrane stretched over the frame, each with a different level of pre-strain. Membranes with an
average percent pre-strain over the domain of the membrane of 2.13 0.24, 3.36 0.25 and 5.07 0.41 were tested. Values
of coefficient of lift versus AOA can be seen in Fig. 1.7 for each lifting surface. Classical aerodynamic behavior is seen in that,
the membrane wings continue to increase in lift at higher angles of attack (i.e. above 8–10 ) where the rigid plate experiences
flow separation on the suction surface and fails to produce greater amounts of lift. It can also be seen that with increased
membrane pretension, the lift values begin to approach that of the rigid wing.
Fig. 1.7 Comparison of coefficients of lift versus AOA; wind tunnel testing for wings with various membrane pre-tensions at 18 m/s or Re
9 104, compared to rigid wing as a reference
Fig. 1.8 Comparison of membrane out-of-plane deformation for one half of the wing; VIC data (left) and quadratic surface reconstruction from
pseudo strain sensors (right); Residual error from out-of-plane deformation of VIC data and reconstructed surface (center)
1.5.2 Surface Reconstruction
For the three different membranes at 2, 3.5 and 5 nominal percent pre-strain, VIC images were taken at 3, 6 and 9 AOA and
each at a flow velocity of 12, 15 and 18 m/s. A sample of one data set for a 2% pre-strain wing at 6 AOA and 18 m/s can be
seen on the left side of Fig. 1.8. Using this data, strain sensors were created with quantity, size and positions as defined in
Table 1.2 Strain sensor configuration

Sensor w [mm] h [mm] 2x/b y/c
1 5 5 0.5 0.5
2 0.5 0.2
3 0.5 0.8
4 0.25 0.5
5 0.25 0.2
6 0.25 0.8
7 0.75 0.5
8 0.75 0.2
9 0.75 0.8

Fig. 1.9 Representation of strain sensors from Table 1.2, derived from full field VIC total strain ex þ ey =2
Table 1.2 and are visualized by the “patches” in Figs. 1.8 (right) and 1.9. Full field error of the reconstructed surface
compared to the actual membrane surface as measured using VIC, can be seen in the center of Fig. 1.8. These results are
favorable and show less than 10% error for max residual with respect to maximum deflection.
Comparing normalized camber, z/c, (where z = a*1 mm, and c being the wing chord length in millimeters) of the
reconstructed surface to that of the actual surface, as measured by VIC, shows favorable results. Values of camber for a
single membrane at different AOA and wind velocities can be seen in Fig. 1.10, which shows relatively accurate values of
reconstruction compared to that of the actual membrane. It can be seen that, while the estimate of the max deflection is quite
accurate, it begins to deviate at higher AOA due to the fact that the surface departs from a quadratic shape. It can also be seen
that an expected increase of camber with respect to velocity and with respect to AOA occurs, while an expected reduction of
camber occurs with respect to increased pre-strain.
1.5.3 Material Characterization
To derive the tension fields created from pre-strain in the membrane, the elastic modulus of the latex material was
determined empirically. A 25 mm by 100 mm speckled sample of latex was vertically suspended in front of the same
aforementioned VIC cameras system. In this test, weights were applied to one end of the latex and images were taken for
each weight. Full field VIC strains were determined for each load. Due to the hyper-elastic behavior of latex, the modulus
was not constant throughout deformation. For this reason, a quasi-linear region is considered between the values of lowest
pre-strain and max expected induced strain from loading. For this test, strains between 2% and 10% were observed, thus the
modulus of elasticity was assumed to be linear in this region and yielded a magnitude of 1.14 MPa. Since VIC data also
provides lateral strain as well as longitudinal strains, the Poisson ratio could be readily determined and was found to be 0.39.
Fig. 1.10 Comparison of normalized camber [z/c] versus AOA, between actual deflection (VIC) and strain sensor reconstruction (Est) for a 2%
pre-strain membrane at different Re (left); a comparison of normalized camber versus amount of pre-strain, between actual deflection (VIC) and
strain sensor reconstruction (Est) for a wing at 6 AOA
Fig. 1.11 Comparison of aerodynamic load generated at different AOA for a 2% pre-strained wing using various force measuring/estimating
techniques; sting balance measurement from wind tunnel tests, force calculation from CFD, force estimation from tip vorticity, and force
estimation from Poisson model
1.5.4 Lift Calculation
Using the data collected and methods described above, lift estimations were calculated for the membrane wings. Load
estimates as compared to actual loads measured during wind tunnel testing can be seen in Fig. 1.11. As can be easily
observed, all load estimates/calculations are less in magnitude than the measured values. This is to be expected due to the
low order nature of the methods employed. Comparing the force resultant from the CFD model to measured data, where
the CFD value was derived from the integration of the numerical pressure gradient over the entire meshed wing surface,
yields a result with a constant error offset of ~15% at 6 AOA. This result provides confidence that the majority of the wake
flow energy is present in the simulation and was captured by the reconstructed membrane surface.
Next, the lift estimation from vorticity is observed. It can be seen that the force estimation increases with the same slope
as that of the actual lift but is again off by a constant offset of ~50% at 6 AOA. This can be attributed to many factors, i.e.
theoretical model fidelity and numerical model fidelity, but is likely most significantly affected by the fact that the tip vortex
does not contain the complete lift circulation energy and will therefore not be completely capable of providing the full force
estimation. Nevertheless, this behavior can certainly be correlated to tangible lift values, derived from strain and vorticity,
through the inclusion of a constant offset applied to the result. This offset would likely be a constant value which would vary
for different wing geometries.
Finally, the behavior of the force estimation from the Poisson model is observed. Here it can be seen that, the estimation
behaves in a significantly different way from the previous estimation. In this case the model has a proportional error of
approximately 66%. This allows us to conclude that the force estimation is off by a constant multiple as opposed to the
constant offset seen by the vorticity estimation. This again gives rise to a correlatable lift estimation through the inclusion of
a constant multiplied to Eq. 1.13.
1.6 Conclusion
In this study, lift was estimated for a pliant membrane wing using structural strain information, analogous to proprioceptive
sensing found in biological fliers. Wind tunnel tests of three membrane wings, each at different membrane pre-strain values,
were conducted; aerodynamic loads, membrane deformations and membrane strains were measured. Using a pseudo strain
sensor configuration from VIC data and a basic analytic model, the membrane surface was reconstructed to a reasonable
order of accuracy. A numeric fluid model was applied to this reconstructed surface to solve for wing tip vorticity behavior in
the near wake field of the wing. Calculating the circulation from the tip vortex, lift could be directly derived using
assumptions from linear aerodynamic theory.
Results from this approach gave favorable lift predictions. For a given membrane wing, it was observed that changes in
the wing’s structural state, due to changes in AOA and flow velocity, generated correlatable behavior in estimated lift as
compared to actual measured lift. Absolute values of estimated lift were offset from actual values due to the assumptions
applied and due to wing tip vortices not capturing the entirety of the wing circulation energy. Nevertheless, using a low order
approximation of structural state and flow state, a distinct correlation between strain, vorticity and lift was made within the
bounds of this study.
1.7 Future Work
This work has strongly encouraged future research in a variety of ways. First, the assumptions that were used in formulating
the overall estimation problem were fairly extensive. This was the goal of the present work; however it encourages future
research which refines the assumptions made. Most significantly for this study was the use of a quadratic basis function for
the deformed membrane surface. It is known that an infinite Fourier series solution can describe such a surface, thus to
increase accuracy of the estimation, higher order terms could be incorporated into the basis function.
An investigation of the trade-off of more complex flow and pressure state solver options versus simple aerodynamic
principles should also be investigated in the future. Perhaps coupling of a panel solver to a finite element membrane model
with strain estimation could yield an algorithm which uses both aerodynamic and structural theory to minimize the estimate
error of both load and structural deformation in an iterative fashion.
Acknowledgments The authors would like to thank the support of the Air Force Office of Scientific Research, Flow Control & Aeroelasticity
under Contract FA9550-10-1-0325, with Douglas R. Smith as project monitor and the Air Force Office of Scientific Research under the
Multidisciplinary University Research Initiative grant FA9550-07-1-0540. The authors would also like to acknowledge Dr. Belinda Batten’s
continuous and critical support to the project and the University of Florida for assistance with contractual aspects.
References
1. Shyy W (2008) Aerodynamics of low Reynolds number flyers. Cambridge University Press, New York
2. Sutton MA, Cheng M, Peters WH, Chao YJ, McNeill SR (1986) Application of an optimized digital correlation method to planar deformation
analysis. Image Vis Comput 4(3):143–151
3. Sutton MA, Turner JL, Bruck HA, Chae TA (1991) Full field representation of the discretely sampled surface deformations for displacement and
strain analysis. Exp Mech 31(2):168–177
4. Schmit R, Glauser M (2005) Use of low-dimensional methods for wake flowfield estimation from dynamic strain. Am Inst Aeronaut Astronaut
J 43(5):1133–1136
5. Schmit R (2002) Low dimensional tools for flow-structure interaction problems: application to micro air vehicles. Doctoral dissertation,
Clarkson University
6. Song A, Tian X, Israeli E, Galvao R, Bishop K, Swartz S, Breuer K (2008) Aeromechanics of membrane wings with implications for animal
flight. Am Inst Aeronaut Astronaut J 46(8):2096–2106
7. Posada JA (2007) Numerical study of wingtip shed vorticity reduction by wing boundary layer control. Doctoral dissertation. Available from
ProQuest Disserations & Teses database (Publication no. 3298564), West Virginia University Libraries
8. Imamura T, Enomoto S, Yamamoto K (2006) Noise generation around NACA0012 wingtip using large-eddy simulation. Int Congr Aeronaut
Sci 25:1–10
Chapter 2
Deformation Mapping at the Microstructural Length Scale
Adam Kammers and Samantha Daly
Abstract Digital Image Correlation (DIC) experiments performed in a scanning electron microscope (SEM) allow for
deformation mapping at length scales not possible with optical cameras. This paper presents the development of a new
experimental methodology to map full-field surface deformations at the microstructural level. The results obtained through
this approach when combined with electron backscatter diffraction (EBSD) microstructure maps enable linking of the
deformation behavior to the underlying microstructure as the material is thermo-mechanically loaded. This paper will
address the challenges associated with this technique including surface patterning at the nano- and micro-scale, image
distortion correction, and dealing with stress relaxation that occurs over long SEM frame scans.
Keywords Deformation mapping • Digital image correlation • Scanning electron microscopy
2.1 Introduction
To further advance the mechanical properties of engineering materials, it is necessary to link macroscopic material behavior
to the microstructure. To fill this need, the Digital Image Correlation (DIC) technique has recently been combined with
Scanning Electron Microscopy (SEM) [1–3] to track full-field deformation at the nano-scale. This methodology (here termed
SEM-DIC) allows investigators to link the microstructural features of a material to its macro-scale mechanical behavior.
DIC is a non-contact optical technique used to measure displacements on the surface of an object, from which Lagrangian
strains can be calculated. The DIC technique was developed at the University of South Carolina in the 1980s [4, 5].
To calculate full-field displacements, random markers are first applied to the surface of the test sample as shown in Fig. 2.1.
These markers are tracked in subsets as the sample is deformed. Surface deformation is calculated by comparing the
deformed marker positions to their initial positions in the reference image. Combining the SEM-DIC strain and displacement
fields with Electron Backscatter Diffraction (EBSD) microstructure maps enables the relationship between localized strain
and microstructure to be established.
The tracking markers, or “speckles”, required for SEM-DIC are well below the capabilities of common patterning
methods such as airbrushing and sprinkling toner powder used for larger scale DIC [7]. This paper will present on the
substantial work that has been performed to create new SEM-DIC patterning techniques. It will also cover the distortion
correction routine applied to the SEM micrographs. A modification to the technique to remove displacements caused by
stress relaxation will also be presented.
2.2 Small Scale Patterning
DIC requires the test sample surface to be covered in a high contrast, random surface pattern. Very few small-scale, SEM-
DIC suitable patterning techniques exist [7–12]. This is due to the recent development of SEM-DIC and the inability of
common macro-scale patterning techniques to create a small enough pattern for SEM-DIC. The authors have recently
A. Kammers (*) • S. Daly

Department of Mechanical Engineering, The University of Michigan, 2350 Hayward, Ann Arbor, MI 48109, USA
e-mail: akammers@umich.edu
16 A. Kammers and S. Daly
Fig. 2.1 Illustration of a DIC surface pattern in the reference (left) and deformed (right) state. The box in the center of each image represents a
subset used to track the pattern. (Image courtesy of [6])
Fig. 2.2 Gold nanoparticle pattern on an aluminum substrate. The pattern was created by soaking the aluminum substrate in dilute silane for 24 h
followed by soaking in gold nanoparticles for 24 h
published a paper which presents multiple new SEM-DIC patterning techniques [6]. The two techniques from that paper that
are most capable of producing usable patterns are nanoparticle (NP) patterning and electron-beam lithography. The
following section will provide details on an improvement that has been made to the NP patterning technique to increase
the control over the pattern location and enlarge the size of the patterned area.
The new NP surface patterning technique allows for the surface of the test sample to be evenly covered with 13–30 nm
diameter gold NPs. The technique was originally developed by others [13–16] to generate a NP monolayer on a variety of
substrates. The first step to generate the pattern is to soak the test sample in silane for 24 h. The silane bonds to the surface
of the test sample with a pendant functional group extending out from the surface. This functional group is strongly attracted
to noble metals, so when the test sample is dipped into the NP solution, the NPs attach to the surface in an even distribution
as shown in Fig. 2.2. This new NP patterning technique allows for a DIC pattern to be created evenly over the entire surface
of the test sample. This fine pattern, when combined with a higher pixel resolution image setting (2048 1768 or
3584 3094) and a larger FOV, still follows the 3 3 pixels/speckle rule of thumb [17]. This allows for future
experiments to investigate strain localization within individual ultrafine-grains as well as over a representative area of
the microstructure. This patterning technique could also be combined with a coarser pattern so that an even wider FOV could
be investigated at a lower spatial resolution with specific areas being zoomed into and investigated with this fine pattern.
2 Deformation Mapping at the Microstructural Length Scale 17
2.3 Image Distortion Correction
SEM images suffer from complex distortions, which if uncorrected, will lead to inaccurate DIC data. Other researchers [1–3]
have characterized the distortion into drift distortion and spatial distortion. Drift distortion varies over time and in this work
is corrected using the technique developed in reference 1. Spatial distortion is similar to the distortion that is present in
optical lenses. However it cannot be corrected with classical parametric distortion models due to the complex electromag-
netic focusing and scanning processes utilized in a SEM. The spatial distortion correction utilized in this work differs
slightly from that presented in references 1–3.
To correct for distortion, the procedure as outlined in reference 1 was followed as a baseline and subsequently modified.
In summary, throughout a pre-test calibration phase, stationary calibration image pairs of a speckle-patterned surface are
captured using a FEI Quanta 3D 200 SEM. Between image pairs, the sample is translated by a known amount in either x or y.
Following the test, the stationary calibration image pairs are analyzed with DIC. DIC is also performed on the translated
images resulting in pure u (displacement in the x-direction) and v (displacement in the y-direction) DIC displacement data
for translations ranging from a few pixels to approximately 1/4 of the FOV. Recording the time at which each image is
captured allows for the drift-velocity at every data point to be determined from the stationary image DIC displacement data.
The drift velocity is plotted versus time, and fit with a unique quadratic function at every data point. These quadratic drift
functions are then integrated over time so that the drift distortion can be determined over any time frame. Using these
drift functions, the calibration and test image displacements are corrected for drift distortion.
The drift distortion corrected u and v DIC displacement fields from the translated images are subsequently used for spatial
distortion correction. To correct for spatial distortion, first the input displacement is subtracted from each displaced data set.
The resulting u and v displacement fields, if undistorted, would be flat and centered at zero, so any variation from zero is a
result of spatial distortion. The spatially distorted u and v displacement surfaces are plotted and fit with biquintic surfaces.
Figure 2.3 shows the u and v spatial distortion surfaces for a translation of 10 pixels in the x-direction. Equivalent surfaces
exist for translations in y. These spatial distortion surface fits are evaluated at every data point to determine the spatial
distortion at each point for the 10 pixel x translation. By translating the sample by different amounts in x and y, the
relationship between spatial distortion and input translation can be quantified and fit with a function at every data point.
A total of four functions will exist at each data point because input displacement in either x or y can cause distortion in both
the u and v displacement fields. To correct for spatial distortion in test images, the drift corrected u and v displacement values
are input into the appropriate spatial distortion functions at the specified data point. This will yield the amount of spatial
distortion that exists at each data point for the given displacement field. The u and v spatial distortions are summed and
subtracted from the drift distortion corrected displacement fields to remove spatial distortion.
2.4 Stress Relaxation
Unlike digital cameras, which capture each pixel in an image at once, SEM image capture is a scanning process, recording
each pixel sequentially. Long image scans are required in SEM-DIC to increase the signal to noise ratio. These long scans
(in this work 90 s) capture stress relaxation that occurs while the load is being held for image capture. This stress relaxation is
Fig. 2.3 Typical (a) horizontal and (b) vertical spatial distortion surfaces for a 10 pixel horizontal displacement. To correct for spatial distortion
due to this 10 pixel horizontal displacement, these surfaces would be subtracted from their respective displacement fields
Fig. 2.4 Stress relaxation in stationary image pairs shows up as a vertical gradient in the horizontal (u) displacement field. (a) shows the u
displacement field for a stationary image pair in which the sample is initially relaxing to the left. (b) shows another stationary image pair in which
the relaxation is initially to the right
Fig. 2.5 Image of the straining stage grips used for SEM-DIC experiments and the patterned Inconel element and positioning device (boxed) used
for calibration images. The height and position of the Inconel pattern can be adjusted to match that of the test sample
only seen in images captured after the sample has plastically deformed. It shows up as a vertical gradient in the horizontal
displacement fields of stationary calibration image pairs as shown in Fig. 2.4.
Since calibration images for drift distortion correction are captured throughout a test, stress relaxation will be observed in
later calibration image pairs. As shown in Fig. 2.4, the relaxation can occur in different directions and as a result interferes
with drift distortion correction, even causing the drift distortion correction routine to increase distortion in some cases. Thus
displacements resulting from stress relaxation must be removed from calibration pair DIC data before it is used for distortion
correction. Two techniques have been developed for this purpose and will be discussed in the following text.
The first technique to remove stress relaxation from images used for distortion correction utilizes an unloaded pattern for
all calibration images. This unloaded calibration pattern has been applied to an Inconel element as shown in Fig. 2.5 and is
placed next to the tensile sample through the use of an auxiliary positioning device. Inconel was chosen due to its high
strength and resistance to corrosion and creep. These properties are important for this application since the calibration
2 Deformation Mapping at the Microstructural Length Scale 19
Fig. 2.6 Surface showing

the drop in load that occurred
between when each pixel in
the first image was scanned
and when it was scanned again
in the second image for the
DIC displacement field shown
in Fig. 2.4b
Fig. 2.7 The u displacement

field from Fig. 2.4b after
removal of stress relaxation.
Relaxation is removed by
applying a scaling/conversion
factor to the change in load
surface and then subtracting
this surface from the
displacement field. The
displacement field is now
centered at 0.1 mm, but will
be shifted to 0 by subsequent
distortion correction
pattern will be required for high temperature tests. The positioning device is controlled by the external stage control knobs
and allows for rotation and vertical translation of the Inconel pattern. Vertical translation is necessary to position the
calibration pattern at the exact height of the tensile sample so that the focus does not need to be adjusted when moving
between the test and calibration patterns. Rotation will be used in high temperature tests to move the calibration pattern away
from the heater when not in use as further protection against creep.
The second technique to remove stress relaxation involves linking the displacement field showing relaxation to the load
drop that occurs during the image scans. A benefit of this technique over the previous technique is that it allows for
quantification of the relaxation so that it can be removed from test images if desired. Figure 2.6 shows a surface
representation of the load drop that occurred between the first and second image scans for the DIC displacement field
shown in Fig. 2.4b. A conversion factor is applied to the change in load surface to convert it to a displacement surface and
scale it to the u displacement field. This scaled surface is then subtracted from the u displacement field to remove the stress
relaxation prior to using the displacement data for distortion correction. For 6061-T6 Al, this conversion factor has been
found to be approximately 0.4 pixels/N. Figure 2.7 shows the displacement fields from Fig. 2.4b after the effect of stress
relaxation has been removed. This technique was developed for single image scans but has also been applied to integrated
image scans as well.
2.5 Summary
SEM-DIC is a valuable technique for deformation tracking at the scale of the microstructure but care must be taken to
address the challenges of sample patterning, distortion correction, and stress relaxation. The new nanoparticle patterning
technique introduced in this paper allows for greater surface coverage than previous techniques. Image distortion can affect
the DIC results and can be corrected by performing DIC on stationary and translated image pairs. Stress relaxation, if
neglected in the test stationary image pairs can lead to increased image distortion. It can be corrected either by using an
unloaded sample for calibration images, or by relating the relaxation to the load drop that occurs during image capture.
References
1. Sutton MA, Li N, Garcia D, Cornille N, Orteu JJ, McNeill SR, Schreier HW, Li X (2006) Metrology in a scanning electron microscope:
theoretical developments and experimental validation. Meas Sci Technol 17:2613–2622
2. Sutton MA, Li N, Joy DC, Reynolds AP, Li X (2007) Scanning electron microscopy for quantitative small and large deformation
measurements part I: SEM imaging at magnifications from 200 to 10,000. Exp Mech 47:775–787
measurements part II: experimental validation for magnifications from 200 to 10,000. Exp Mech 47:789–804
4. Peters WH, Ranson WF (1982) Digital imaging techniques in experimental stress analysis. Opt Eng 21:427–431
5. Sutton MA, Wolters WJ, Peters WH, Ranson WF, McNeill SR (1983) Determination of displacements using an improved digital correlation
method. Image Vis Comput 1:133–139
6. Kammers AD, Daly S (2011) Small-scale patterning methods for digital image correlation under scanning electron microscopy. Meas Sci
Technol 22:125501
7. Sutton MA, McNeill SR, Helm JD, Chao YJ (2000) Advances in two-dimensional and three-dimensional computer vision photomechanics.
Top Appl Phys 77:323–372
8. Tschopp MA, Bartha BB, Porter WJ, Murray PT, Fairchild SB (2009) Microstructure-dependent local strain behavior in polycrystals through
in-situ scanning electron microscope tensile experiments. Metall Mater Trans A 40A:2363–2368
9. Berfield TA, Patel JK, Shimmin RG, Braun PV, Lambros J, Sottos NR (2007) Micro- and nanoscale deformation measurement of surface and
internal planes via digital image correlation. Exp Mech 47:51–62
10. Scrivens WA, Luo Y, Sutton MA, Collete SA, Myrick ML, Miney P, Colavira PE, Reynolds AP, Li X (2007) Development of patterns for
digital image correlation measurements at reduced length scales. Exp Mech 47:63–77
11. Collette SA, Sutton MA, Miney P, Reynolds AP, Li X, Colavita PE, Scrivens WA, Luo Y, Sudarshan T, Muzykov P, Myrick ML (2004)
Development of patterns for nanoscale strain measurements: I. fabrication of imprinted Au webs for polymeric materials. Nanotechnology
15:1812–1817
12. Li N, Guo S, Sutton MA (2011) Recent progress in e-beam lithography for SEM patterning. In: Proulx T (ed) MEMS and Nano, Vol 2.
Conference Proceedings of the Society for Experimental Mechanics Series. New York, pp 163–166
13. Grabar K, Freeman RG, Hommer MB, Natan MJ (1995) Preparation and characterization of Au colloid monolayers. Anal Chem 67:735–743
14. Freeman RG, Grabar KC, Allison KJ, Bright RM, Davis JA, Guthrie AP, Hommer MB, Jackson MA, Smith PC, Walter DG, Natan MJ (1995)
Self-assembled metal colloid monolayers: an approach to SERS substrates. Science 267:1629–1632
15. Brown KR, Walter DG, Natan MJ (2000) Seeding of colloidal Au nanoparticle solutions. 2. Improved control of particle size and shape. Chem
Mater 12:306–313
16. Brown KR, Lyon LA, Fox AP, Reiss BD, Natan MJ (2000) Hydroxylamine seeding of colloidal Au nanoparticles 3. Controlled formation of
conductive Au films. Chem Mater 12:314–323
17. Sutton MA, Orteu JJ, Schreier HW (2009) Image correlation for shape, motion and deformation measurements, basic concepts, theory and
applications. Springer, New York
Chapter 3
Molecular Tailoring of Interfacial Adhesion
Using Self-Assembled Monolayers
Martha E. Grady, Mark D. Losego, Paul V. Braun, and Nancy R. Sottos
Abstract Self-assembled monolayers (SAMs) provide an enabling platform for molecular tailoring of the chemical and
physical properties of an interface in an on-demand fashion. In this work, we systematically vary SAM end-group functionality
and quantify the corresponding effect on interfacial adhesion between a transfer printed gold (Au) film and a fused silica
substrate. SAMs with two different end groups are investigated: Dodecyltriethoxysilane and 11-mercapto-undecyltri-
methoxysilane. The adhesive strength of the SAM-mediated interfaces is measured by a non-contact laser-induced spallation
method at strain rates in excess of 106 s1. A high strain rate test method is selected to facilitate comparison with forthcoming
molecular dynamics simulations of the molecular failure process. Interfacial stresses are inferred from interferometric displace-
ment measurements and finite element analysis. By making multiple measurements at increasing stress amplitudes (controlled
by the laser fluence), the adhesion strengths of Au films transfer-printed on different SAM modified substrates are compared.
Varying the end-group functionality drastically alters the adhesion strength of Au films, leading to improved adhesion over
transfer printed films on unmodified quartz. We demonstrate a spallation strength of 24.2 0.4 MPa for interfaces prepared
with dodecyltriethoxysilane and 60 11 MPa for interfaces prepared with 11-mercapto-undecyltrimethoxysilane confirming
that interfacial bonding at the Au-thiol interface is significantly stronger than at the Au-methyl interface.
Keywords Self-assembled monolayers • Laser spallation • Interfacial adhesion
3.1 Introduction
Self-assembled monolayers have wide application, but are specifically of interest in the role they play in micro and
nano-electronics [1–3]. In such device systems, SAMs are used to bridge from substrate to thin film and the ability to
control the molecular bonding of the SAMs to thin film layer has an impact on the overall adhesion. The selectivity of the
self-assembled monolayer chemistries can be useful in controlling interfacial thermal conductance as well [4–6]. Attempts
have been made to show the influence of several self-assembled monolayer chemistries on crack propagation of
M.E. Grady (*)

Department of Mechanical Science and Engineering, University of Illinois at Urbana-Champaign, 1304 W. Green St, Urbana, IL 61801, USA
e-mail: grady3@illinois.edu
M.D. Losego
Department of Chemical and Biomolecular Engineering, North Caroline State University, Raleigh, NC, USA
P.V. Braun
Department of Materials Science and Engineering, Frederick Seitz Materials Research Laboratory, University of Illinois
at Urbana-Champaign, Urbana, USA
Beckman Institute for Advanced Science and Technology, University of Illinois at Urbana-Champaign, Urbana, USA
N.R. Sottos
Department of Mechanical Science and Engineering, University of Illinois at Urbana-Champaign, 1304 W. Green St, Urbana, IL 61801, USA
Department of Materials Science and Engineering, Frederick Seitz Materials Research Laboratory, University of Illinois
at Urbana-Champaign, Urbana, USA
Beckman Institute for Advanced Science and Technology, University of Illinois at Urbana-Champaign, Urbana, USA
22 M.E. Grady et al.
Table 3.1 List of chemicals used to prepare self-assembled monolayers

Chemical Formula Notation Purchased from
Dodecyltriethoxysilane CH3(CH2)11Si(OC2H5)3 CH3-C11-Si Sigma Aldrich
St. Louis, MO
11-mercapto-undecyltrimethoxysilane SH(CH2)11Si(OCH3)3 SH-C11-Si Gelest, Inc
Morrisville, PA
substrate-epoxy interfaces [7–9]. These crack propagation studies can be strain rate dependent and are complicated by
plasticity. In this work, the adhesive strength of the SAM-mediated interface is measured by a non-contact laser-induced
spallation method at strain rates in excess of 106 s1. The benefits of laser-induced spallation methods come from the non-
contact aspect, high strain rates, and rapid evolution of the experiment. A non-contact test method allows minimal disruption
of the interface prior to testing, unlike many experiments that use a “sandwich” structure. High strain rates limit the
mechanisms for plasticity and other strain-rate dependencies. Lastly, the experiment evolves over a period of 15 ns on
average, which is on a time scale appropriate for comparison with molecular dynamics simulations.
Our experimental system consists of a Au film transfer-printed to a self-assembled monolayer (SAM) on a fused silica
substrate with two different termination chemistries. We chose the Au/SiO2 interface because bifunctional molecules with
orthogonal attachment chemistries having specificity towards either SiO2 or Au can be used. By varying the SAM’s end-
group functionalities, we control the bond strength across the interface. Silane chemistries on the quartz surface permit dense
packing and strong bonding. Diverse chemistries (e.g. –CH3, –SH) at the Au surface permit large variation in bond strength
at the Au-SAM interface. When properly formed, these chemistries do not result in multi-layer formation nor loop
attachment to the fused silica substrate.
Previous work in molecular electronics [10] has demonstrated that the “soft-deposition” approach of transfer-printing, as
compared to more energetic physical vapor deposition techniques, causes minimal damage to the SAM layer. Thus, we do
not expect the process of transfer-printing the gold layer to disrupt the uniformity and dense packing of the monolayer.
Because adhesion between the elastomeric stamp and Au film is a function only of peel rate, Au layers can be transfer-
printed to any surface chemistry regardless of adhesion strength [11]. This deposition technique also allows transfer printing
of Au films to SAMs with both strong covalent bonds (e.g. –SH) as well as weak van der Waals attractions (e.g. –CH3).
Following previous laser-induced spallation protocols [12, 13], this technique is used to characterize the macroscopic
adhesion strength of the transfer-printed Au layers. In these experiments a pulsed Nd:YAG laser generates a high amplitude
longitudinal stress wave on the backside of the substrate. Upon reflection from the Au surface, this stress wave loads the Au
film in tension, effectively testing its adhesion strength by determining the stress wave amplitude that induces film failure.
Multiple measurements at increased stress amplitudes (controlled by the laser fluence) are made to compare the adhesion
strengths of Au layers transfer-printed to different SAM chemistries. This work quantitatively shows the macroscopic
adhesion of a SAM functionalized interface on time scales applicable for comparison with molecular dynamic simulations.
Ultimately these results along with molecular dynamic simulations should give more insight into the effect of molecular
bonding on interfacial adhesion.
3.2 Specimen Preparation
3.2.1 Self-Assembled Monolayers
The fused silica substrates (SiO2, 500 mm thick, Quartz Scientific) were cleaned in piranha solution (3 H2SO4 : 1 H2O2 by
volume) at 65 C for 60 min (WARNING: piranha is a strong oxidant and must be handled according to documented
safety procedures). The substrates were then rinsed with copious amounts of water, dried under a stream of nitrogen, and
further dried in an oven at 120 C for 30 min in air. This is a standard procedure done to leave only a monolayer of water on
the surface as the activator for silane attachment. Silicon pieces with a native oxide layer were prepared in the same manner,
and SAMs were prepared on both fused silica and silicon surfaces in the same reaction vessel. The silicon substrates were
then used for ellipsometry measurements. Substrates were immersed in a toluene solution (ACS certified, Fisher Scientific)
of 10 mM silane plus 15 mM triethylamine for 24 h in a sealed container on the benchtop. Silanes used in this experiment are
listed in Table 3.1. Experimentation with the SAM deposition procedures revealed that anhydrous solvents and preparation
in an inert atmosphere did not improve monolayer formation for alkoxysilanes. SAM thickness was characterized using a
single-wavelength (633 nm) ellipsometer (Gaertner L116C). Because of the similarity in the refractive index of SiO2 and
alkyl SAMs, ellipsometry could not be conducted directly on fused silica substrates. Thus, for ellipsometric measurements,
3 Molecular Tailoring of Interfacial Adhesion Using Self-Assembled Monolayers 23
Fig. 3.1 Side profile of experimental system: Au film transfer printed on SAM-functionalized fused silica substrate. An Al absorbing layer is
e-beam deposited on the backside followed by spin casting a layer of sodium silicate (waterglass)
SAMs were concurrently deposited on silicon substrates in the same reaction vessel. A refractive index of 1.5 was used for
the SAM and the native oxide was assumed to be 22 Å thick. Other techniques in addition to ellipsometry were used to
ensure a uniform monolayer including water contact angle and atomic force microscopy. Atomic force microscopy (AFM)
was conducted in tapping mode using an Asylum Research MFP-3D system.
3.2.2 Transfer Printing
Transfer-printing of gold films to the SAM modified substrates followed the procedures of Meitl et al. [11]. Silicon substrates
with thermally grown oxide layers (~75 nm) were cleaned in a piranha solution, rinsed with copious amounts of water, and
dried under a stream of nitrogen. Gold films of nominally 185 nm thickness were deposited by e-beam evaporation
(Temescal) onto the SiO2/Si substrates. Poor adhesion at the Au/SiO2 interface makes these oxide surfaces ideal donor
substrates for the transfer-printing process. A thin layer of PVA (87–89% hydrolyzed, MW ~ 20,000 g/mol, Sigma-Aldrich,
10 wt.% dissolved in H2O) was cast onto the Au surface to impart mechanical stability during the transfer process. The PVA
layer is dried at 85 C for 5 min. A block (approx. 2 cm 2 cm 0.7 cm thick) of polydimethylsiloxane (PDMS, Sylgard 184
Silicon Elastomer, Dow-Corning) is used to transfer the film. The PDMS block is manually pressed onto the donor substrate
such that it completely wets the PVA/Au surface. It is then rapidly peeled, causing delamination and resulting in the PDMS
stamp being “inked” with the gold film. The film is laminated to the receiving, pre-functionalized fused silica substrate and
heated on a hotplate at 115 C. After 90 s of manual pressure, the PDMS stamp is slowly peeled from the surface leaving the
PVA/Au film bonded to the substrate. The film is kept on the hotplate for another 90 s to finish bonding. The PVA layer is
then rinsed away under flowing water. After transfer printing the Au film onto the functionalized substrates, a 400 nm thick
Al absorbing layer is e-beam evaporated on to the backside followed by spin casting a 1 mm sodium silicate (waterglass)
confining layer. A depiction of this test structure is shown in Fig. 3.1.
3.3 Experimental Protocol
Adhesion measurements were made with a laser-induced spallation setup shown schematically in Fig. 3.2. A rapid, high-
amplitude acoustic wave was initiated by the impingement of an Nd:YAG pulsed laser (New Wave Tempest) on the Al energy
absorbing layer on the back surface of the specimen. Because of the confinement of a sodium silicate layer, rapid expansion of
Al generated an acoustic wave that propagated through the substrate in compression. After reflection at the thin film free
surface, the wave loaded the thin film-substrate interface in tension. The magnitude of the acoustic wave is a function of the
laser fluence (mJ/mm2) controlled by the energy and spot size of the Nd:YAG pulsed laser, the substrate material and
thickness, as well as the material and thickness of the absorbing and confining layers. For all tests, the substrate, absorbing
layer and confining layer remained constant, only the fluence was incremented. An attenuator provided by the manufacturer
controls the laser energy and the distance of the focusing lens controls the spot size. A Michelson interferometer was used to
measure the displacement of the free surface. A biased silicon photodetector (Electro-Optics Technology ET-2030) connected
to a high-rate oscilloscope (LeCroy LC584 A) recorded the temporal interference pattern as a voltage trace, V(t), given by,
V max þ V min V max V min

VðtÞ ¼ þ sinð2pnðtÞÞ; (3.1)
2 2
Fig. 3.2 Schematic of laser spallation test set up. A single pulse from an Nd:YAG laser impinges on the backside of the sample generating a
compressive acoustic wave. The wave reflects off the free surface and loads the interface in tension. A Michelson interferometer is used to measure
the displacement of the free surface
Table 3.2 Material properties used to calculate substrate stress

Material Young’s modulus (GPa) Density (kg/m3) Poisson’s ratio
Fused Silica 78 2,200 0.28
Au 77 19,300 0.42
where Vmax and Vmin are the voltage maximum and minimum respectively of each interference fringe and t is time. The
interference fringe number, n(t), is unwrapped and then converted to displacement, u(t), using,
l0 nðtÞ
uðtÞ ¼ ; (3.2)
2
where the wavelength of the interferometric laser, l0, is 532 nm. The substrate stress ssub was then calculated from the
displacement using Eq. 3.3 and the substrate material properties listed in Table 3.2.
1 du
ssub ðtÞ ¼ ðrCd Þsub (3.3)
2 dt
In order to determine the substrate and interface stresses for a given laser fluence (energy per area), we carried out a set of
calibration experiments with highly reflective thin Al films (200 nm) that were electron-beam deposited on identical
substrates. A calibration set was necessary because the Au/SAM/SiO2 films partially failed at laser fluences of interest,
invalidating the in situ interferometry measurements. Because the acoustic wave is dependent only on substrate parameters,
we were able to generate the same acoustic wave in any top surface with the same laser fluence. Thus, an Al film, which has a
higher interface strength and larger reflectivity, was used to calibrate the stress wave magnitude as a function of laser
fluence. A laser spot size of diameter 2 mm was chosen and kept constant for these experiments. The rise time of the Nd:
YAG laser is typically 3–5 ns and one spallation experiment takes approximately 15 ns to transpire. The laser energy was
incremented, and the interference pattern captured at a sampling rate of 8 GS/s. The displacement of the free surface and
substrate stress are calculated for each increment in laser energy. The substrate stress pulse measured during calibration
provided input for a 1D finite element analysis that calculated the interface stress for a 185 nm Au film. The peak interface
stress for each substrate stress pulse was averaged for the corresponding laser fluence. From this data, a relationship between
the laser fluence and the interface stress was determined, as shown in Fig. 3.3. The standard error corresponding to the set of
calibration experiments at each laser fluence was calculated and is represented by error bars in the figure.
After testing over a range of fluence values, optical microscopy was used to investigate the loaded regions of the Au/SAM/
SiO2 structures. Figure 3.4a–h show representative images of Au films after laser-induced adhesion tests with corresponding
interfacial stress loading. Black areas in the optical images designate delamination where the Au film has separated from the
Fig. 3.3 Experimental calibration of interface stress for increments in laser fluence
Fig. 3.4 Optical images of delamination damage (black areas) caused by laser spallation to (a–d) Au/CH3-C11-Si interfaces and (e–h) Au/SH-
C11-Si interfaces at increasing stress amplitudes: (a, e) 24 MPa,(b, f) 30 MPa (c, g) 41 MPa, (d, h) 86 MPa. Scale bars are 500 mm
substrate. All samples that were tested at higher laser fluences and corresponding interface stresses beyond that shown in
Fig. 3.4a–h, demonstrated further delamination of the Au film. The shape of the delamination at higher laser fluences
matches the 2 mm spot size chosen for these experiments and is the maximum delamination exhibited. From the optical
images, we observe that delamination failure of the Au/CH3-C11-SiQz structures consistently occurs at significantly lower
interface stress amplitudes than the Au/SH-C11-SiQz structures. Thus, the SH terminated SAM structures exhibited higher
interfacial strength than CH3 terminated SAM structures. Unlike most previous studies using physical vapor deposited
(PVD) films, these transfer-printed Au layers show non-abrupt failure. For typical PVD deposited films, a critical laser
fluence value exists at which all films will show some delamination. For these transfer printed films, a range of fluences
exists where only a fraction of the films will show delamination. Figure 3.5 plots this behavior showing the fraction of films
Fig. 3.5 Plot of the fraction of spots tested that showed delamination for both Au/SH-C11-Si and Au/CH3-C11-Si interfaces as a function of the
laser fluence. The calibrated interface stresses are also shown. Numbers above each data point represent the number of spots tested
at each laser fluence that exhibit delamination. An example of the non-abrupt behavior is the Au/SH-C11-Si sample at
10.4 mJ/mm2 where only 40% of the 10 spots tested showed delamination. We hypothesize that this non-abrupt behavior is
due to the inherent macroscopic defects in these transfer-printed layers. However, as Fig. 3.5 demonstrates, both the onset of
delamination and full delamination occur at higher laser fluences for the Au/SH-C11-Si as compared to the Au/CH3-C11-
Si structures, confirming the stronger bonding at the thiol-Au interface.
For this work, we define the critical laser fluence as the lowest fluence value that initiates delamination 100% of the time.
The interface strength is then determined from the corresponding average interface stress value in the calibrations conducted
at this critical laser fluence. For structures with CH3-C11-Si molecules at the interface, this critical laser fluence is only
12.6 mJ/mm2, which corresponds to an interface stress of 24.2 MPa. For structures with SH-C11-Si at the interface, this
critical laser fluence is 19.0 mJ/mm2, which corresponds to an interface stress of 60 MPa.
3.5 Conclusions
Laser-induced spallation experiments are used to characterize the macroscopic adhesion strength of the transfer-printed Au
layers. In these experiments a pulsed Nd:YAG laser generates a high amplitude longitudinal stress wave on the backside of
the substrate. Upon reflection from the Au surface, this stress wave loads the Au film in tension, effectively testing its
adhesion strength. By making multiple measurements at increased stress amplitudes (controlled by the laser fluence), the
adhesion strengths of Au layers transfer-printed to different SAM chemistries can be compared. Figure 3.4a–h qualitatively
show that delamination failure of the Au/CH3-C11-SiQz structures occurs at significantly lower stress amplitudes than the
Au/SH-C11-SiQz structures. The size of the damaged area is also much larger for the Au/CH3-C11-SiQz structures when
compared at the same stress amplitude. We calculate a spallation strength of 24.2 0.4 MPa for Au/CH3-C11-SiQz and
60 11 MPa for Au/SH-C11-SiQz confirming that interfacial bonding at the Au-thiol interface is significantly stronger
than at the Au-methyl interface.
In summary, we have experimentally and quantitatively shown that the strength of a single bonding layer directly affects
the adhesion of a transfer printed Au film. We expect future experiments in parallel with molecular dynamics to expose new
understanding of the relationship of molecular bonding and strength of interfaces.
Acknowledgements We acknowledge support from NSF CMMI 07–26742. Sample fabrication and characterization was carried out in part in the
Frederick Seitz Materials Research Laboratory at UIUC, which is partially supported by the U.S. Dept. of Energy under grants DE-FG02-
07ER46453 and DE-FG02-07ER46471.
References
1. Cahill DG et al (2003) Nanoscale thermal transport. J Appl Phys 93:793–818

2. Garimella SV (2006) Advances in mesoscale thermal management technologies for microelectronics. Microelectron J 37:1165–1185
3. Bae MH, Ong ZY, Estrada D, Pop E (2010) Imaging, simulation, and electrostatic control of power dissipation in graphene devices. Nano Lett
10:4787–4793
4. Losego MD, Moh L, Arpin KA, Cahill DG, Braun PV (2010) Interfacial thermal conductance in spun-cast polymer films and polymer brushes.
Appl Phys Lett 97:011908
5. Hung MT, Choi O, Ju YS, Hahn HT (2006) Heat conduction in graphite-nanoplatelet-reinforced polymer nanocomposites. Appl Phys Lett
89:023117
6. Chiritescu C et al (2007) Ultralow thermal conductivity in disordered, layered WSe2 crystals. Science 315:351–353
7. Mello AW, Liechti KM (2006) The effect of self-assembled monolayers on interfacial fracture. J Appl Mech 73:860–870
8. Zhuk AV, Evans AG, Hutchinson JW, Whitesides GM (1998) The adhesion energy between polymer thin films and self-assembled
monolayers. J Mater Res 13:3555–3564
9. Kent MS, Reedy ED, Yim H, Matheson A, Sorenson J, Hall J, Schubert K, Tallant D, Garcia M, Ohlhausen T, Assink R (2004) Using
self-assembling monolayers to study crack initiation in epoxy/silicon joints. J Mater Res 19:1682–1695
10. Loo YL, Lang DV, Rogers JA, Hsu JWP (2003) Electrical contacts to molecular layers by nanotransfer printing. Nano Lett 3:913–917
11. Meitl MA et al (2006) Transfer printing by kinetic control of adhesion to an elastomeric stamp. Nat Mater 5:33–38
12. Wang J, Sottos NR, Weaver RL (2003) Mixed-mode failure of thin films using laser-generated shear waves. Exp Mech 43:323–330
13. Wang JL, Weaver RL, Sottos NR (2003) Laser-induced decompression shock development in fused silica. J Appl Phys 93:9529–9536
Chapter 4
Micromechanical Characterization of Ductile Damage in DP Steel
J.P.M. Hoefnagels, C.C. Tasan, F.J. Peters, and M.G.D. Geers
Abstract Weight minimization triggered the automotive industry to introduce new advanced high strength steels that show
ductile fracture by microvoid evolution under deformation, resulting in unexpected failure without significant necking.
Therefore, an extensive study was initiated to gain insight on the formation and development of microstructural damage in
dual phase (DP) steel from a mechanical point of view.
Different DP (model) microstructures were created that are very similar except for a single microstructural parameter
(ferrite grain size or martensite volume percentage). Using our miniaturized Marciniak setup, these DP microstructures were
deformed to fracture for three strain paths (uniaxial tension, plane strain tension, and biaxial tension). Exploiting the in-situ
SEM visualization capabilities of the miniaturized Marciniak setup, the relevant damage mechanisms were identified and
their evolution studied. Finally, a procedure was developed to quantitatively analyze the evolution of these damage
mechanisms (post-mortem) by semi-automatically categorizing each damage site over large SEM viewing areas to obtain
statistically relevant trends of each damage mechanism with increasing strain, for the strain paths and microstructures. This
procedure yielded rich data on the evolutions of the different damage mechanisms as a function of the DP microstructure and
deformation mode. Some preliminary observations and hypotheses were formulated.
Keywords DP steel • Ductile damage • Damage characterization • Strain path dependence
4.1 Introduction
Weight minimization triggered the automotive industry to introduce new advanced high strength steels that show ductile
fracture by microvoid evolution under deformation, resulting in unexpected failure without significant necking. Therefore,
an extensive study was initiated to gain insight on the formation and development of microstructural damage in dual phase
(DP) steel from a mechanical point of view.
4.2 Experimental Procedure, Results, and Discussion
4.2.1 Model Microstructures
First, much effort was put to create different DP model microstructures were created that are very similar except for a single
microstructural parameter (ferrite grain size or martensite volume percentage). The final microstructure should resemble that
of the commercial steel, but production artifacts from the hot rolling process should be removed. To this end, different heat
treatments were designed, based on the following general heat treatment strategy starting with commercial DP600 steel
J.P.M. Hoefnagels (*) • F.J. Peters • M.G.D. Geers

Department of Mechanical Engineering, Eindhoven University of Technology, Den Dolech 2, Eindhoven 5612AZ, The Netherlands
e-mail: j.p.m.hoefnagels@tue.nl
C.C. Tasan
Max-Planck-Institut f€
ur Eisenforschung, D€
usseldorf, Germany
30 J.P.M. Hoefnagels et al.
Fig. 4.1 The three (model) microstructures investigated in this study: (a) the ‘control’ microstructure, (b) the ‘down-scaled’ microstructure, and
(c) the ‘more martensite’ microstructure. All images are taken at the same magnification using bright field optical microscopy, where ferrite is
resembled by the lighter and martensite by the darker colored areas
(from Tata steel). First, the sample is heated in a furnace to a temperature of 960 C, at which temperature the microstructure
is fully austenized. The sample remains at his temperature for 10 min to allow the austenite grains to grow. After 10 min, the
sample is cooled slowly in a second furnace to room temperature. Thereafter, the sample is put back in the first furnace and
heated to a temperature of 775 C, to undergo an intercritical annealing step in which the material separates into a ferrite and
an austenite phase, which will be turned into martensite after quenching to room temperature. The temperature of the
intercritical annealing step determines the weight percentage of both phases as well as the carbon content of both phases and
thus the final martensite phase. The martensite volume percentage for this microstructure is found to be 32.5 0.2% as
compared to 25% for the commercial DP600 steel. The above described heat treatment yields the so-called ‘control’
microstructure, as shown in Fig. 4.1a.
By decreasing the austenization time from 10 min to only 1 min, the growth of austenite grains is mostly prohibited while
the martensite volume fraction and martensite morphology are also largely unaffected, as shown in Fig. 4.1b. Therefore, this
microstructure resembles a scaled down microstructure of the control microstructure and is hence named the ‘down-scaled’
microstructure.
Finally, a heat treatment was designed to create a microstructure that is the same as the control microstructure, except for
the fact that the martensite volume percentage is higher. This treatment is exactly the same as the control microstructure
treatment, but now the DP800 grade steel is used instead of DP600. For these two commercially available steels, only the
carbon content in DP800 is higher than in DP600, whereas all the other alloying contents are the same. Therefore, when
applying the same intercritical annealing temperature, the DP800 steel with higher carbon content will end up with a higher
martensite volume percentage compared to the DP600 steel with lower carbon content, while the carbon content of the
martensite phase and the ferrite phase will be the same (this can be seen by applying the well-known lever rule to the steel
phase diagram). Figure 4.1c shows indeed that a difference in martensite volume percentage is established, whereas ferrite
grain size remains approximately constant. The martensite volume percentage for this microstructure is found to be
41.3 0.2%, significantly higher that of the control and down-scaled microstructures, therefore, it is named the ‘more
martensite’ microstructure.
4.2.2 In-situ Marciniak Experiments
The above-mentioned three DP microstructures were each deformed up to fracture for three different strain paths: uniaxial
tension (UAT), plane strain tension (PST), and biaxial tension (BAT). This was done using our miniaturized Marciniak
setup, shown in Fig. 4.2, which was developed previously in house [1].
Figure 4.3a–c shows digital image correlation overlays of the von Mises strain fields measured in-situ under optical
microscopy, for the three deformation modes. By tracking the strain state over time up to fracture, the full deformation path
can be followed. It is shown in Fig. 4.3d–f that by correctly choosing the geometry of the steel specimen and that of the
underlying steel ‘washers’ (these washers are used in the Marciniak concept to control that the fracture occurs at the center of
the specimen), a deformation paths very close to uniaxial tension (Fig. 4.3d), plane strain tension (Fig. 4.3e), and biaxial
tension (Fig. 4.3f) can indeed by enforced.
4 Micromechanical Characterization of Ductile Damage in DP Steel 31
Fig. 4.2 The miniaturized Marciniak setup (a), designed and built in house [1], which allows for deformations of steels from uni-axial tension to
biaxial tension. The setup can be place inside the sample chamber of a scanning electron microscope (b) or under an optical microscope to allow
for in-situ visualization of the deformation mechanisms
Fig. 4.3 (a–c) Real-time measured optical images for UAT, PST, and BAT tested with the miniaturized Marciniak setup, overlaid with the von
Mises strain field obtained through digital image correlation (DIC). From the strain fields evolutions have been used to calculate the actual major-
minor strain curves of the three strain paths (d–f)
Fig. 4.4 The most relevant damage mechanisms encountered in the DP microstructures, as observed with in-situ scanning electron microscopy:
(a) severe plastic straining of ferrite near martensite (PSM), (b) severe plastic straining of ferrite near ferrite grain boundaries (PSBG),
(c) martensite cracking (MC), (d) damage inside a ferrite grain (FED) and (e) damage around inclusions (INC)
4.2.3 Relevant Damage Mechanisms
Exploiting the in-situ SEM capabilities of the miniaturized Marciniak setup, the damage mechanisms in our DP
microstructures were studied meticulously. This study will be reported elsewhere. Here, only the five most relevant damage
mechanisms are presented in Fig. 4.4: (a) severe plastic straining of ferrite near martensite (PSM), (b) severe plastic straining
of ferrite near ferrite grain boundaries (PSBG), (c) martensite cracking (MC), (d) damage inside a ferrite grain (FED) and (e)
damage around inclusions (INC). These five mechanisms will serve as categories for the quantitative damage analysis,
discussed next.
4.2.4 Quantitative Analysis of Damage Mechanisms
A procedure was developed to quantitatively analyze the evolution of the damage mechanisms (post-mortem) by semi-
automatically categorizing each damage site over large SEM viewing areas to obtain statistically relevant trends of each
damage mechanism with increasing strain and for the different strain paths and microstructures. This program, to be called
the Matlab Statistical Damage Identification algorithm (MSDI) from here on, identifies each individual damage incident and
prompts the incident to the user for classification, as shown in Fig. 4.5. Furthermore, for each incident, the size in pixels is
logged.
The analysis process in which the program is used consists of the following steps:
1. Per sample that is deformed up to fracture, five cross sections representing five different strain levels are prepared.
2. Per strain level, five images are taken by SEM in order to provide a statistically relevant analysis. Images are taken with a
constant magnification of 456 times. This is an ideal magnification in order to image a representative (i.e. large) region of
the microstructure with sufficiently high resolution.
3. Contrast/brightness settings are equalized for all images in order to provide reproducible results.
4. The image is loaded into the program.
5. A loop is run over the total amount of pixels in the image. For each pixel the gray value is compared to a certain threshold
identifying the level of ‘blackness’ of a pixel. If the threshold is satisfied, the pixel is stored as being part of a possible
damage incident.
6. To check if the stored pixels are indeed valid damage incidents, the surrounding pixels are subjected to a second
threshold. If satisfied, the group of pixels is stored as a singular damage incident.
7. Each of the stored damage incidents is prompted to the user for classification. The user can then select the appropriate
damage mechanism that best describes the particular incident. This process is shown in Fig. 4.5.
8. For each damage incident the classification and other important properties such as size and location are logged.
Since steps 4 through 8 have been completed 25 times per sample, this analysis process yields an enormous amount of
valuable data.
During deformation, the area of the cross section of the sample changes due to thickness strain. As an effect, the square
area observed by SEM needs to be scaled in order to accommodate for this effect. Therefore, the total amount of incidents
registered after analysis is scaled by a factor describing this cross sectional area difference.
Fig. 4.5 Screen capture of the Matlab Statistical Damage Identification algorithm. The user interface automatically prompts a damage incident for
classification by the user, for which a selection box is shown in the middle of the screen. To assist the user in his/her assessment of the damage
incident, the damage site is simultaneously shown at low magnification (top right image), at higher resolution (top left image), and at high
magnification with original contrast/brightness (bottom left image), enhanced contrast/brightness (bottom middle image), and different color map
4.2.5 Quantitative Evolutions of the Damage Mechanisms; Preliminary Results
The resulting quantitative evolutions of the damage mechanisms are presented in Fig. 4.6 for the ‘control’ microstructure
and in Fig. 4.7 for the ‘down-scaled’ microstructure. It is noted that these are preliminary results and most details of this rich
data have yet to be well understood. Nevertheless, some preliminary observations and hypotheses can be formulated:
• For both microstructures and all three deformation mode, severe plastic straining of ferrite near martensite (PSM) is the
most frequently occurring damage mechanisms. As martensite is stronger than ferrite, it is no surprise that most
deformation occurs in the ferrite. Around martensite islands, the ferrite needs to overcompensate the limited martensite
deformation, which may explain that damage development due to severe plastic straining in ferrite occurs primarily near
martensite islands.
• Plastic straining of ferrite near ferrite grain boundaries (PSBG), damage inside a ferrite grain (FED), and damage around
inclusions (INC) probably do not occur frequently enough to have a significant impact on the (limit of the) final failure
mechanism of the steels.
• Martensite cracking (MC) is important, but only for biaxial loading. This may be due to the higher stress triaxiality under
biaxial loading, which may ‘lock’ the deformation mechanisms in the ferrite grains (e.g. dislocation glide) surrounding
martensite grains, and as a consequence may put higher stresses on the martensite islands, triggering martensite cracking.
– This mechanism may also explain the increase in PSM going from UAT to PST to BAT for the down-scaled
microstructure. It is, however, surprising that PSM is constant for all three deformation mode for the control
microstructure.
• It is surprising the many of the damage evolutions show an initial increase with increasing strain for low strains, followed
by a saturation in the number of damage sited for higher strains. This may be indication that damage sites start to interact
which may be the onset of final failure. In regions outside these damage concentration regions, the stress may be
somewhat relaxed, preventing new damage sites to be formed there.
Fig. 4.6 Quantitative trends in damage mechanisms for the ‘control’ microstructure, for the three deformation modes (uniaxial tension (UAT),
plane strain tension (PST), and biaxial tension(BAT)) and for the five damage mechanisms (severe plastic straining of ferrite near martensite
(PSM), severe plastic straining of ferrite near ferrite grain boundaries (PSBG), martensite cracking (MC), damage inside a ferrite grain (FED), and
damage around inclusions (INC)). All curves and each data point within the curve are shown with their 68% confidence interval as defined by the
standard deviation
Fig. 4.7 Quantitative trends in damage mechanisms for the ‘down-scaled’ microstructure, for the three deformation modes (uniaxial tension
(UAT), plane strain tension (PST), and biaxial tension(BAT)) and for the five damage mechanisms (plastic straining of ferrite near martensite
(PSM), plastic straining of ferrite near ferrite grain boundaries (PSBG), martensite cracking (MC), damage inside a ferrite grain (FED), and
damage around inclusions (INC)). All curves and each data point within the curve are shown with their 68% confidence interval as defined by the
standard deviation
– Interaction of PSM damage sites for higher strains was a few times directly observed, as for those occasions the PSM
damage sites stretched over a large area.
4.3 Conclusions
Different DP (model) microstructures were created that are very similar except for a single microstructural parameter (ferrite
grain size or martensite volume percentage). Using our miniaturized Marciniak setup, these DP microstructures were
deformed up to fracture for three different strain paths (uniaxial tension, plane strain tension, and biaxial tension). Exploiting
the in-situ SEM visualization capabilities of the miniaturized Marciniak setup, the relevant damage mechanisms were
identified and their evolution studied. Finally, a procedure was developed to quantitatively analyze the evolution of these
damage mechanisms (post-mortem) by semi-automatically categorizing each damage site over large SEM viewing areas to
obtain statistically relevant trends of each damage mechanism with increasing strain and for the different strain paths and
microstructures. This procedure yielded rich data on the evolutions of the different damage mechanisms as a function of the
DP microstructure and deformation mode. Some preliminary observations and hypotheses were formulated.
Acknowledgements This research was carried out under project number MC2.05205 in the framework of the Research Program of the Materials
innovation institute M2i (www.m2i.nl), which is the former Netherlands Institute for Metals Research. The contribution of Marc van Maris is also
gratefully acknowledged.
Reference
1. Tasan CC, Hoefnagels JPM, Dekkers ECA (2011) Multi-axial deformation setup for microscopic testing of sheet metal to fracture. Experimen-
tal Mechanics. doi:10.1007/s11340-011-9532-x
Chapter 5
MEMS Device for Fatigue Testing of 25 mm Thick
Aluminum Specimens
L.A. Liew, D.T. Read, and N. Barbosa III
Abstract We report the development and demonstration of a MEMS (micro-electro-mechanical system) test instrument to
perform fatigue testing on 25 mm thick aluminum 1145 H19 foil. Both the foil material and the specimen fabrication were
obtained from commercial sources, while the MEMS device was fabricated in-house. The loading was in bending with a
ratio R of minimum to maximum force equal to zero. The gauge section of the specimens had a minimum width of about 120
mm. The bending force was applied at a distance of about 3 mm from the gage section, through a lever arm fabricated as
part of the specimen. The forces used were on the order of 5 mN, while the displacements at the load point were up to 50 mm.
The loading waveform was a 70 Hz square wave. The S-N (stress versus number of cycles to failure) curve exhibits the
expected behavior, with the shortest lives occurring at stresses near the tensile strength and the longest at stresses near half
that value. The cracks exhibited a very abrupt transition from the incubation stage, marked by increasing surface
deformations and other evidence of local microstructural damage, to the rapid growth stage. The failure criterion used for
the S-N curve was the beginning of the rapid-growth stage.
Keywords Aluminum • Bending • Comb drive • Electrostatic • Fatigue • Foil • MEMS • S-N
5.1 Introduction
We seek to leverage the advantages of MEMS (micro electro mechanical systems) technology to provide a new MEMS-
based platform for measuring the fatigue properties of any material from any source, in particular, ‘bulk’ materials, meaning
materials that are not vapor-deposited and are not co-fabricated with the MEMS device. We have developed chip-sized
fatigue test instruments for bending fatigue of separately fabricated specimens that are tens to hundreds of micrometers in
size (smaller than current state of the art in conventional-scale fatigue testing). This approach offers the key advantages of
MEMS technology: adaptability to in-situ environmental testing, high throughput enabled by economical replication of the
test device, and low power consumption. But unlike the MEMS fatigue testing approaches reported to date (e.g., [1]), which
use part of the MEMS device as the fatigue specimen, the present device is a standalone ‘loadframe on a chip’ into which
specimens made of any material may be inserted and tested. Therefore it is suited to testing specimens cut from bulk
materials, rather than being limited to thin films. In this paper, we briefly describe the design and construction of the fatigue
test instrument, the design of the specimen, and the test protocol developed, and we demonstrate proof of concept by
producing an S-N curve for 25 mm thick Al 1145 H19 foil.
Contribution of the U.S. National Institute of Standards and Technology. Not subject to copyright in the U.S.A.
L.A. Liew
Materials Reliability Division, National Institute of Standards and Technology, Boulder, CO 80305, USA
Department of Mechanical Engineering, University of Colorado, Boulder, CO 80309, USA
D.T. Read (*) • N. Barbosa III
Materials Reliability Division, National Institute of Standards and Technology, Boulder, CO 80305, USA
e-mail: read@boulder.nist.gov
38 L.A. Liew et al.
5.2 Experiment
MEMS device The MEMS device (Fig. 5.1) includes a comb drive actuator [2] for applying the forces and another set of
combs for sensing actuator displacement. Our comb drive actuator differs from those in the literature primarily in its
significantly larger force capacity (millinewtons as opposed to micronewtons) and larger displacement capacity (up to 100
mm with potential for longer range, as opposed to a few to tens of mm). These are accomplished by its larger overall size; the
entire silicon wafer thickness is used as the structural material, rather than silicon-on-insulator (SOI) [3] layers, surface
micromachined films, or electrodeposited films. The overall chip size is 23 by 13 mm. The MEMS device is driven to
oscillate following a commonly used actuation scheme [2]. An AC (alternating current) drive voltage referenced to the
instrument ground is applied to the fixed drive comb, and a DC (direct current) bias voltage, also referenced to the instrument
ground and equal to the amplitude of the AC drive voltage, is applied to the movable combs (both drive and sense). The fixed
side of the sense comb is connected to an off-chip preamplifier input. The periodic motion of the DC-biased movable sense
comb induces an alternating current across the sense combs due to the time varying capacitance between them. The sensing
circuit converts this current into a DC voltage proportional (at fixed frequency) to the amplitude of the push rod’s oscillatory
displacement.
5.2.1 Specimen Design and Gripping
We chose a generic material–aluminum (Al 1145 H19)–obtained commercially as nominally 25 mm thick (0.001 in) foil.
Our material had a thickness of 23.7 mm. The alloy designation indicates chemical composition and temper [4]. The
specimens were fabricated by use of lithographic wet etching by a commercial vendor. As a result of the vendor’s wet
etching limitations, however, the resultant specimen geometry deviated substantially from the design and the etched surfaces
were very rough, which could contribute to experimental uncertainty of our S-N curve. The design of the specimens, shown
in Fig. 5.2, is based on in-plane, fixed-free beam bending. The specimen includes a fixed grip section, a gage section, and a
longer lever arm. The specimen geometry was designed to maximize the stress in the gage section within the constraints of
the actuator’s range of force and displacement, to produce a near pure-bending stress field, and to avoid stress concentrations
in the specimen. Specimen holders, or grips, were fabricated from a separate silicon wafer 250 mm thick.
The grips are large enough to be easily handled with tweezers. Each grip has a T-shaped ‘tab’ that fits into a
complementary-shaped numbered slot in the MEMS device’s outer frame. These indexed slots facilitate easy, yet repeatable,
positioning and alignment of the grip/specimen assembly with respect to the push rod. Like the MEMS devices, each grip
may be re-used indefinitely.
Fig. 5.1 Essential design

features of the MEMS fatigue
testing machine and scanning
electron microscope images
of key parts of a device: 250
mm-thick bulk-silicon comb
drive actuator, removable
specimen holder (grip), and
25 mm-thick aluminum foil
specimen. The drive combs
have a total of 840 individual
electrodes (‘comb teeth’), and
the sense combs have 125
electrodes
5 MEMS Device for Fatigue Testing of 25 mm Thick Aluminum Specimens 39
Fig. 5.2 (a) and (b): Details

of specimen geometry. (c) and
(d) Optical micrographs of a
specimen mounted in a silicon
grip. In (d), the grip is
engaged to a test frame
Fig. 5.3 Schematic of the

drive and data acquisition
circuitry of the fatigue testing
system showing the MEMS
chip, the off-chip drive
electronics, sensing circuit,
and optical microscope and
data acquisition systems
5.2.2 Fatigue Test Protocol
Figure 5.3 shows the drive and sensing circuitry. A typical fatigue test is carried out under an optical microscope equipped
with a digital camera, as follows. A specimen is mounted to the device (Fig. 5.2). The push rod, at its rest position, is in slight
but firm contact with the specimen’s lever arm. The DC drive voltage is then ramped up (ranging from 80 to 220 V) to cause
the push rod to apply increasing static loads to the specimen’s lever arm. The specimen’s displacements at both the loadpoint
and the free end are measured with the optical microscope and accompanying imaging software. The objective here is to
experimentally determine the displacement at which a particular specimen starts to deform irreversibly (to account for
variability across specimens), as a guide for selecting the fatigue test parameters. A target loadpoint displacement is then
selected, and the actuator’s AC drive signal (set to 1 Hz) and the DC bias are adjusted to produce the desired loadpoint
displacement. We used a square wave as the AC drive signal, because it improved the optical displacement-amplitude
measurements and also gave a better signal from the displacement amplitude measurement circuit.
40 L.A. Liew et al.
The drive frequency is then increased to 70 Hz to begin the fatigue test. The sense circuit’s output signal is recorded at 1-s
intervals, and the optical microscope, which is focused on the gage section through a 20X long-working-distance objective
lens, acquires images at time intervals ranging from 10 s for low-cycle fatigue tests to 30 min for high-cycle tests. Because
the drive voltage remains unchanged throughout a fatigue test, and because the push rod pushes in only one direction on the
specimen rather than applying fully reversed loading, this is a quasi-stress controlled fatigue test with R ratio ¼ 0. We use
the onset of fast crack growth as the failure criterion. We end the fatigue test when we observe a crack in the specimen gage
section that exceeds roughly half the gage width, and the sense signal is significantly larger than it was at the start of the test.
Data from a typical fatigue test are shown in Fig. 5.4. The recorded sense circuit signal is smoothed and plotted against time.
As the specimen accumulates fatigue damage its stiffness gradually decreases, resulting in larger displacement amplitude of
both the specimen’s lever arm and the actuator’s push rod, and thus a gradual increase in the sense signal. During this period,
a fatigue crack will initiate and grow slowly. The time at which the sense signal suddenly and dramatically increases
corresponds to the fast crack growth stage of fatigue, which is accompanied by a significant increase in the specimen’s
compliance. This is our failure criterion. The number of cycles to failure N is obtained from the time to failure and the drive
frequency of 70 Hz.
The time lapse micrographs of the gage section provide a second, independent measure of the average lever arm
displacement, as well as of crack initiation and growth. We analyze these images to obtain the specimen bending and
crack length over time, as plotted in Fig. 5.4. From the time-lapse images, we also obtain the time of crack initiation
(to within the temporal resolution of the time-lapse interval); however, this requires manual interpretation of the images.
Finite element analysis, with the known individual specimen geometry and imposed displacements, provided a value for the
maximum tensile stress in the gage section.
Figure 5.5 shows the S-N curve for the Al 1145 H19 specimens. The solid symbols on the plot indicate fatigue failures.
The open symbols represent tests that were terminated before fatigue failure (runouts). No literature data were found on the
fatigue performance of Al 1145 H19, probably because this alloy is used predominantly in foil [5] and not as a structural
material. Furthermore this material has a highly directional rolled microstructure, as verified by optical metallography. Here
crack propagation is in the rolling direction, and the loading is transverse to the rolling direction. Kaufman [6] presents
tabular data on the fatigue performance of several commercially pure aluminum alloys including 1100 H18 and 1350 H19;
the H codes indicate that the microstructure is heavily rolled, similar to the present specimen material. The chemistries, as
indicated by the alloy designations, are similar to the present material. Kaufman’s specimens were, of course, much larger
Fig. 5.4 Data from a typical

fatigue test, consisting of the
sensing circuit signal, plotted
on the left ordinate, and the
specimen’s in-plane
displacement as obtained via
DIC analysis of the time-lapse
optical micrographs, and the
crack length, also from the
micrographs, both plotted on
the right ordinate
5 MEMS Device for Fatigue Testing of 25 mm Thick Aluminum Specimens 41
Fig. 5.5 Present S-N data

compared with data tabulated
by Kaufman [6] for
conventional fatigue tests
of similar alloys
than those used in the present study. These data are compared with the present data in Fig. 5.5. The figure shows that the
Kaufman data form a close lower bound to the present data. The higher scatter in the present data, compared to that of the
literature data, is attributed to the irregularities in the specimen geometry produced by the specimen fabrication process. It is
interesting to note that the present specimens have a very high initial (i.e. before crack initiation) stress gradient compared to
conventional-sized smooth (i.e. uncracked and unnotched) fatigue specimens. The initial stress gradient in the present study
is over 2 MPa/mm in some specimens. The present technique appears to be well-suited for use in studies of stress gradient
effects.
In conclusion, we have developed a MEMS based fatigue test method for micrometer-sized specimens of any material,
fabricated by bulk micromachining of single-crystal silicon wafers. We demonstrated the device by generating an S-N curve
for Al 1145 H19. The failure criterion was the onset of fast crack growth, which was detected by use of capacitive sensing of
imposed displacement as well as analysis of time-lapse optical images. The new capability of this technique is fatigue testing
of small specimens that may be extracted from critical regions of bulk materials. This test method may enable fundamental
study of fatigue at the microscale in a variety of bulk materials. It may also be used as a diagnostic tool where there are
limited volumes of material available for analysis, such as heat-affected zones with a small form factor, and will facilitate
testing in hazardous environments.
References
1. Muhlstein CL, Howe RT, Ritchie RO (2004) Fatigue of polycrystalline silicon for microelectromechanical system applications: crack growth
and stability under resonant loading conditions. Mech Mater 36(1–2):13–33
2. Tang WC, Nguyen TCH, Judy MW, Howe RT (1990) Electrostatic-comb drive of lateral polysilicon resonators. Sens Actuators A: Phys 21
(1–3):328–331
3. Noworolski JM, Klaassen E, Logan J, Petersen K, Maluf NI (1996) Fabrication of SOI wafers with buried cavities using silicon fusion bonding
and electrochemical etchback. Sens Actuators A Phys 54(1–3):709–713
4. Boyer HE, Gall TL (1984) Metals handbook, desk edition. American Society for Metals, Metals Park
5. Hatch JE (1984) Aluminum: properties and physical metallurgy. American Society for Metals, Metals Park
6. Kaufman JG (2008) Properties of aluminum alloys: fatigue data and the effects of temperature, product form, and processing. ASM
International, Materials Park
Chapter 6
High Confidence Level Calibration for AFM Based
Fracture Testing of Nanobeams
Scott Grutzik, Richard Gates, Yvonne Gerbig, Robert Cook, Melissa A. Hines, and Alan Zehnder
Abstract When designing micro- or nanoelectromechanical systems, (MEMS and NEMS), it is important to consider
whether structural elements will withstand loads experienced during operation. Fracture behavior at length scales present in
MEMS and NEMS is much different than at macro- and mesoscopic scales. Due to a smaller probability of crystal defects
and a high surface to volume ratio, fracture is controlled by surface characteristics rather than volumetric ones. Prior
measurements using doubly clamped Si beams loaded with an atomic force microscope (AFM) showed that fracture of Si
nanobeams is highly affected by surface roughness (Alan T et al., Appl Phys Lett 89:091901, 2006) and oxidation (Alan T
et al., Appl Phys Lett 89:231905, 2006). In experiments of this type, calibration of the system, particularly the AFM
cantilever stiffness, is critical to the accuracy of both the force and displacement results. A new set of experiments are
underway in which the tests are performed by adapting a direct, traceable method for calibrating the AFM cantilever
stiffness (Ying ZC et al., Rev Sci Instrum 78:063708, 2007). The improved calibration should not only improve the accuracy
of the strength results but will allow linear stiffness measurements of the sample to be used to back out sample thickness, a
key parameter in interpretation of the data.
6.1 Introduction
The rapid development of micro- and nanoelectromechanical systems (MEMS and NEMS) necessitates the ability to predict
and control strength of micro- and nanoscale structures. Previous results have shown that these types of structures can have
strengths approaching the ideal strength of the material and that strengths are strongly dependent on both the surface
roughness [1] and surface oxidation [2]. These tests were performed using doubly clamped silicon beam structures which
were under-etched so that they were freely suspended across a trench. An atomic force microscope (AFM) cantilever was
then used to deform the beam structures until failure. Because the AFM cantilever deforms as well as the beam, calculating
the stress at failure relies explicitly on knowing the stiffness of the AFM cantilever.
It is often necessary to know the stiffness of an AFM cantilever when one is interested in quantifying the forces between
the tip and the sample. Several methods have been suggested for measuring stiffness. If the length L, width w, elastic
modulus E, density r, and resonant frequency f0 are known to high enough precision, the stiffness may be calculated using
standard beam theory by Cleveland et al. [4]
pffiffiffi3 pffiffiffi
k ¼ 2p3 w f 0 L r = E: (6.1)
S. Grutzik • A. Zehnder (*)

Field of Theoretical and Applied Mechanics, Cornell University, Ithaca, NY 14853, USA
e-mail: atz2@cornell.edu
R. Gates • Y. Gerbig • R. Cook
Nanomechanical Properties Group, Material Measurement Laboratory, National Institute of Standards and Technology,
Gaithersburg, MD 20899, USA
M.A. Hines
Department of Chemistry and Chemical Biology, Cornell University, Ithaca, NY 14853, USA
44 S. Grutzik et al.
The Sader method involves calculating stiffness by observing the effects of dynamic loading and damping by the viscous
fluid surrounding the cantilever (typically air) on the frequency and quality factor of resonance [8]. Cleveland et al.
developed a method in which a small mass is added to the end of the cantilever and the stiffness is calculated from the
change in resonant frequency [4]. Hutter and Bechhoefer suggested relating the stiffness to magnitude of thermal vibrations
using the equipartition theorem by Hutter and Bechhoefer [6]

1=2k z2t ¼ 1=2kB T; (6.2)
where hz2t i is the mean square displacement of thermal motion, kB is the Boltzmann constant, and T is the absolute
temperature. Butt and Jaschke added a correction factor to Eq. (6.2) by considering the shape of each mode of vibration [3]

k z2t ¼ 0:817kB T : (6.3)
This method is used for automated calibration of cantilever stiffness in many commercial AFM systems [9]. Ohler has
provided a concise review of these methods and a direct comparison of the results of each for a range of cantilever
stiffnesses [7].
Although these techniques are sufficient for most AFM applications, they are poorly suited to the strength measurements
developed by Alan et al. [1, 2]. For typical AFM applications, a fairly compliant cantilever is used so that weak sample-tip
interactions can be observed. In the case of strength measurements, the cantilever must be much stiffer. The cantilevers used
in the beam fracture experiments have stiffnesses of around 200–250 N/m. In contrast, the Sader, added mass, or thermal
calibration techniques are typically used to calibrate cantilevers with stiffness of 0.01–35 N/m. To have confidence in the
resulting fracture strength data, the stiffness of the cantilever should be measured with a method well suited to very stiff
cantilevers.
In this paper, we will compare two methods, that used by Alan et al. [1, 2] and a new one based on work by Gates et al.
[5, 10]. In the method of Alan et al., the length, width, and tip position of the cantilever was first quantified using a scanning
electron microscope (SEM). The cantilever thickness was then estimated by comparing the measured resonant frequency of
the cantilever to that found using a variable thickness thickness finite element model using density and elastic modulus
values from the literature. The method based on the work by Gates et al. involves measuring the stiffness of a reference
cantilever using an instrumented nanoindentor. That reference cantilever can then be used to measure the stiffness of the
cantilever for use in beam fracture experiments. Following a more thorough description of each of these methods, we will
compare stiffness values obtained with each method and the accuracy of each.
6.2 Methods
The resonant frequency method used by Alan et al. starts with taking an SEM image of the cantilever to be calibrated with
the cantilever oriented so that thickness direction is normal to the plane of the image as shown in Fig. 6.1a. Using this image,
the length and width of the cantilever can be obtained as well as the position of the tip. These dimensions are then used to
create a finite element model such as that shown in Fig. 6.2. Because the thickness of the cantilever could not be obtained
directly from the SEM image, the thickness of the finite element model is varied until the resonant frequency of the model
matches the experimental resonant frequency. At this point a static load can be applied to the tip of the cantilever in the finite
element model. By dividing that load by the resulting deflection one obtains the cantilever stiffness.
The second method is drawn from the methods of Gates et al. who have developed a method in which the cantilever to be
calibrated displaces a reference cantilever of known stiffness [5]. In order to accommodate a range of cantilever stiffnesses,
Gates et al. suggests using an array of reference cantilevers which span a range of stiffnesses. Unfortunately we again run
into a problem that the cantilever to be used for beam fracture experiments is far stiffer than any of those used by Gates et al.
Ying et al. have also developed a second method in which an instrumented nanoindentor is used to directly measure
cantilever stiffness [10]. This has the disadvantage that each cantilever that would be used for nanobeam fracture tests must
be calibrated with the nanoindentor and the nanoindentor is not at the same facility where the rest of the tests are to be done.
The method we have developed combines the two methods proposed by Gates et al.
The new stiffness calibration method uses a batch of reference cantilevers which are only slightly more compliant
than the cantilevers to be used for fracture testing. An SEM image of one of these reference cantilevers is shown in Fig. 6.3.
The stiffness of these reference cantilevers is measured directly with an instrumented nanoindentor. Each of these reference
6 High Confidence Level Calibration for AFM Based Fracture Testing of Nanobeams 45
Fig. 6.1 (a) SEM image of AFM cantilever from which cantilever dimensions would be taken to construct a finite element model from which one
can obtain the stiffness. This is referred to as the test cantilever in the text. (b) SEM image of the same AFM cantilever from the side. The tip height
and cantilever may be obtained from this image
Fig. 6.2 A finite element model of an AFM cantilever with the dimensions obtained from an SEM image such as that shown in Fig. 6.1a
Fig. 6.3 SEM image of a reference cantilever used to calibrate the stiffness of cantilevers used for nanobeam fracture tests. The white cross
represents the location at which the cantilever is calibrated
cantilevers may then be used to calibrate the stiffness of a large number of other cantilevers. Notice that the reference
cantilever shown in Fig. 6.3 does not have a tip. Since the reference cantilever is used only to calibrate the stiffness of other
cantilevers, it will never interact with a sample surface and so a tip is not needed. A tip would only get in the way when used
with the instrumented nanoindentor and with other other cantilevers.
Fig. 6.4 Plot showing the measurement of the AFM cantilever sensitivity S1 as the cantilever is pushed against a hard surface
Instrumented Cantilever to be
Reference
nanoindentor with NIST Calibrated calibrated
cantilever to
traceable calibration reference
be calibrated
cantilever
Cantilever to be calibrated
is attached to AFM
Reference cantilever is
attached to rigid substrate
Fig. 6.5 The reference cantilever is first calibrated using a NIST traceable instrumented nanoindentor. During calibration of the test cantilever, it
is pushed against the end of the now calibrated reference cantilever. Because the stiffness of the reference cantilever is known, the stiffness of the
test cantilever can be calculated
To use the reference cantilever to calibrate the stiffness of a second cantilever, the reference cantilever is placed on the
stage of an AFM and the cantilever to be calibrated is mounted normally. The test cantilever is first pushed against the bulk
silicon at the base of the reference cantilever. This bulk silicon is considered to be approximately rigid and so the test
cantilever deflects an amount equal to the displacement of the AFM head. The output of this is data in volts and nanometers
as shown in Fig. 6.4. The horizontal axis, in nanometers, is the displacement of the base of the cantilever as applied by
the AFM head. The vertical axis, in volts, is proportional to the deflection of the tip of the test cantilever relative to its base.
The trend starts out flat as the cantilever approaches the surface. Once the tip contacts the surface, the deflection signal
begins to increase. From this we extract a value called the sensitivity S1 which is the inverse of the slope of this plot after the
tip has contacted the surface. It has units of nm/volt and relates the signal in volts to the deflection of the cantilever tip
relative to the cantilever base.
The test cantilever is next aligned over the location at which the reference cantilever was calibrated and a second force
curve is performed. The test cantilever pushes on the reference cantilever and both deflect, as shown in Fig. 6.5. Because
both cantilevers are deflecting, they act as springs in series and so the AFM head must travel through a greater distance to
produce the same tip deflection relative to the cantilever base. This means that the slope of the resulting plot is less. Although
it is not a true sensitivity, we will call the inverse of this slope S2.
In order to find the cantilever stiffness k in terms of S1, S2, and the reference cantilever stiffness kref, note that
d ¼ S1 =V; (6.4)
where d is the deflection of the test cantilever and V is the deflection signal in volts. Similarly,
S2 ¼ D=V ¼ ðdref þ dÞ=V; (6.5)
where dref is the deflection of the reference cantilever and D is the head displacement when the test cantilever is in contact
with the reference cantilever and is the sum of the displacements of both cantilevers. Eliminating the deflection signal V from
Eqs. (6.4) and (6.5),

dref þ d dref S2 S1
S2 ¼ S1 ! ¼ ; (6.6)
d d S1
Since the two cantilevers are in static equilibrium, we also know that
dref
kref dref ¼ kd ! k ¼ kref : (6.7)
d
Finally, in order to get the cantilever stiffness k in terms of S1, S2, and kref, we substitute Eq. (6.7) into Eq. (6.6) to get

S2 S1
k¼ kref : (6.8)
S1
The same cantilever was calibrated with both the reference cantilever method and the resonant frequency/finite element
model method. The calibrated cantilever was an uncoated TAP525 from Bruker AFM Probes, model number MPP-13100-10
with a nominal stiffness of 200 N/m. The reference cantilever was an uncoated AppNano ACL-TL cantilever with a nominal
stiffness of 45 N/m, although the group of reference cantilevers was handpicked to have a higher than average stiffness of at
least 60 N/m. The nanoindentation was done with a Hysitron Triboindentor at the National Institute for Standards and
Technology1 and the AFM work was done with a Dimension Icon 3100. All finite element analysis was done using
ABAQUS. For the finite element simulations, the density of silicon was set to 2.329 g/cm3 and the C11, C12, and C44 elastic
constants were set to 166, 64, and 79.6 GPa respectively. The crystal orientation was set so that the top and bottom surfaces
of the cantilever were (100) planes and the long axis of the beam was in the [110] direction. This orientation was confirmed
in the physical cantilevers using Laue back-reflection of X-rays.
6.3 Results
The stiffness of the reference cantilever was measured with the instrumented nanoindentor ten times, producing an average
value of 79.3 N/m with a standard deviation of 0.467 N/m. Using these values in the reference cantilever method, the
stiffness was found to be 233 N/m. Using the finite element model with the model thickness adjusted so that the resonant
frequency of the model matched the measured resonant frequency, the stiffness was calculated to be 183 N/m. A second
stiffness calculation using the finite element model was also performed except the thickness was measured directly from a
side view SEM image, Fig. 6.1b. In this case, the resulting stiffness was 231 N/m.
The Hysitron nanonindentor calibration has an uncertainty of about 3% and repeated measurements with the
nanoindentor probe repositioned each time showed an uncertainty in repeatability of about 2%. These combine to give an
uncertainty in reference cantilever stiffness of 3.6%. The uncertainty in placement of the test cantilever is about 2%.
Combining this with the uncertainty in reference cantilever stiffness gives a potential error for the reference cantilever
method of 4.1%.
In the resonant frequency method, errors in measurements of cantilever dimensions lead to errors in resonant frequency
which, which produces an error in thickness, which ultimately results in an error in stiffness. Notice from Fig. 6.1 that the
cantilever may be partitioned into three parts. The first and largest part, which we will call the arm of the cantilever, is from
1
Certain instruments and materials are identified to adequately specify the experimental procedure. Such identification does not imply recom-
mendation or endorsement by the National Institute of Standards and Technology, nor does it imply that the materials or instruments identified are
necessarily the best available for the purpose.
the base to the tip and has a constant trapezoidal cross section. The pyramidal tip which interacts with the sample is a second
part and the third is the somewhat irregularly shaped triangular point at the end of the cantilever. Because the mass of the
second and third parts are difficult to determine from Fig. 6.1a and they do not contribute much to the cantilever stiffness, we
will lump them together into a single mass m In a similar fashion as the added mass method [4], the resonant frequency may
be expressed as
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
k
o¼ ; (6.9)
m þ M
where k is the stiffness of the finite element model before the thickness is adjusted, m is the combined mass of the cantilever tip
and the triangular point region, and M ∗ ¼ 0. 24M where M is the mass of the cantilever arm. M and k can be expressed as

aþb Et2 ð3abÞ
M ¼ rLt ; k¼ ; (6.10)
2 12l3
where r is the density of silicon, t is the thickness, L is the cantilever length, and a and b are the short and long parallel sides
respectively of the cantilever’s trapezoidal cross section. These dimensions are illustrated in Fig. 6.2. From Eq. (6.9),
uncertainty in the resonant frequency o as calculated using the finite element model relates to uncertainties in m, L, a, and b by
2 2 2 " 2 # 2
do 1 m=M dm 9 1 M =m dL
¼ þ þ þ
o 4 1 þ m=M
m 4 4 1 þ M =m
L
" 2 2 2 # 2
1 1 1 1 M =m da
þ þ
4 1 þ b=3a 4 1 þ b=a 1 þ M =m a
" 2 2 2 # 2
1 1 1 1 M =m db
þ : (6.11)
4 1 þ 3a=b 4 1 þ a=b 1 þ M =m

b
If we make the approximation that m is small compared to M*, then thickness is linearly proportional to resonant
frequency and their uncertainties are equal. From Eq. (6.10), we see that
dk dt do
¼3 ¼3 ; (6.12)
k t o
where do/o can be found using Eq. (6.11). Errors in length measurements using SEM are estimated to be at most 5%. Using
measurements taken from SEM images such as those in Fig. 6.1, dm/m is estimated to be about 10%. Using these values in
Eqs. (6.11) and (6.12), the maximum error in cantilever stiffness using the resonant frequency method is estimated to be
dk=k ¼ 25 %:
6.4 Discussion
It is immediately obvious that the stiffness calculated using the finite element with the model thickness adjusted by resonant
frequency does not match very closely with that found using the reference cantilever. The fact that the finite element
generated stiffness agrees with the reference cantilever method when the model thickness is measured directly by a side view
SEM image suggests that the reference cantilever method is accurate. It is also apparent that if the thickness can be
accurately determined the finite element model can produce good results as well. The arm region of the cantilever, extending
from the base to the tip, has a very simple geometry which is easy to accurately represent in the finite element model. The
pyramidal tip has a simple geometry but extends in a direction normal to the plane of the image and it is impossible to
determine the height of this tip from Fig. 6.1a. The thickness of various portions of the triangular point region at the end of
the cantilever also cannot be determined directly from Fig. 6.1a. The data indicates that the potentially large error in the
mass of the triangular point and tip regions of the cantilever cause an error in resonant frequency. This leads to the model
thickness being adjusted to the wrong value, which results in an incorrect cantilever stiffness. Additional analysis few the
relative uncertainties and ac curacies of the two methods is in progress.
6.5 Summary
We have compared two methods for calibrating the stiffness of very stiff AFM cantilevers. The first method involved
measuring cantilever dimensions in a SEM, finding the cantilever thickness by matching model resonant frequency to actual
resonant frequency, and using the model to calculate stiffness. The second method involved measuring the stiffness of a
reference cantilever using an instrumented nanoindentor and then using that as a transfer artifact to calibrate the stiffness of
the test cantilever via the reference cantilever method. We have found that due to difficulties in accurately finding the
volume of the cantilever tip, the method of finding thickness with resonant frequency does not result in a stiffness that
matches with that found using the reference cantilever. By using a finite element model with a thickness obtained using a
side view SEM image the model produces a stiffness very close to that of the reference cantilever method. Uncertainty for
the reference cantilever method has been shown to be considerably smaller than that of the resonant frequency method.
Acknowledgements This work was funded by the National Science Foundation (NSF) grant CMMI-0856488 and the NSF IGERT graduate
traineeship in Nanoscale Control of Surfaces and Interfaces, (NSF Grant DGE-0654193). It was performed in part at the Cornell NanoScale
Facility, a member of the National Nanotechnology Infrastructure Network, which is supported by NSF Grant ECS-0335765, and made use of
the X-ray diffraction facility of the Cornell Center for Materials Research (CCMR) with support from the National Science Foundation Materials
Research Science and Engineering Centers (MRSEC) program (DMR-1120296). The usage of the Hysitron Tribondentor was completed at the
National Institute of Standards and Technology in Gaithersburg, MD.
References
1. Alan T, Hines MA, Zehnder AT (2006) Effect of surface morphology on the fracture strength of silicon nanobeams. Appl Phys Lett 89:091901
2. Alan T, Zehnder AT, Sengupta D, Hines MA (2006) Methyl monolayers improve the fracture strength and durability of silicon nanobeams.
Appl Phys Lett 89:231905
3. Butt H, Jaschke M (1995) Calculation of thermal noise in atomic force microscopy. Nanotechnology 6:1–7
4. Cleveland JP, Manne S, Bocek K, Hansma PK (1993) A nondestructive method for determining the spring constant of cantilevers for scanning
force microscopy. Rev Sci Instrum 64(2):403–405
5. Gates RS, Reitsma MG (2007) Precise atomic force microscope cantilever spring constant calibration using a reference cantilever array. Rev
Sci Instrum 78:086101
6. Hutter JL, Bechhoefer J (1993) Calibration of atomic-force microscope tips. Rev Sci Instrum 64(7):1868–1873
7. Ohler B (2007) Cantilever spring constant calibration using laser Doppler vibrometry. Rev Sci Instrum 78:063701
8. Sader JE, Chon JWM, Mulvaney P (1999) Calibration of rectangular atomic force microscope cantilevers. Rev Sci Instrum 70(10):3967–3969
9. Veeco Instruments (2010) Inc. NanoScope V Controller Manual
10. Ying ZC, Reitsma MG, Gates RS (2007) Direct measurement of cantilever spring constants and correction for cantilever irregularities using an
instrumented indentor. Rev Sci Instrum 78:063708
Chapter 7
High Strain Rate Characteristics of Fiber Bragg
Grating Strain Sensors
Wun-Jheng Lin, Liren Tsai, Cia-Chin Chiang, and Shih-Han Wang
Abstract Fiber Bragg Grating sensors (FBGs) have been utilized in various engineering fields because of their lightweight
and good environment tolerance. In this research, FBG strain sensors were embedded inside carbon fiber reinforced polymer
composites (CFRP) to study the FBG wave spectrums at high strain rate. The FBG embedded CFRP specimens were
machined to dog-bone shape and a foil strain gauge was attached at its gauge section. The dynamic response of FBG sensors
were then examined using split Hopkinson tension bar (SHTB). By comparing the strain measurements from FBGs, foil
gauges, and SHTB measurements, the high strain rate behavior of FBG strain sensors was able to be explored at strain rates
between 130 and 2,100 s1.
Keywords SHTB • Dynamic responses • Fiber Bragg Grating sensor
7.1 Introduction
Today, various researchers forced on composites material in the past decades, because of its excellent characteristics in
engineering applications such as low density, high thermal conductivity rate, and high elastic strength [1]. CFRP was widely
utilized in many engineering fields, it is composed of polymer matrix with carbon fibers, and CFRPs were usually composed
of 10–70% of carbon fibers [2, 3]. Polymer matrix provides support to cohere with each carbon fiber and increased the
toughness and strength of the composite. The matrix could also prevent the environmental erosion and oxidation [3, 4].
Meanwhile, optical fiber has evolved in many technology fields in modern telecommunication system and photonics [5].
Optical fiber grating was found in 1978s, and flow production in 1993s [6]. When periodic gratings were applied on the optical
fiber, the fiber possesses fiber Bragg grating characteristics. When FBG was subjected to a perturb, the refractive index would
change hence resulted in a coupling phenomenon, i.e. when external force acts on the optical fiber, the grating structure of
FBG would affect the light signal transmit, and the applied force or strain of the optical fiber could be determined [6].
FBG sensors were very suitable for measuring in situ strain behavior of composite materials. K.S.C. Kuang and R. Kenny
et al., studied various composite materials using embedded FBG strain sensors, and the results showed that FBGs exhibit
linearly strain measurement response under quasi-static tension tests [7]. Y. Okabe and S. Yashiro et al. found that when
embedding FBG sensors in polymer matrix, the residual stress would occur [8]. In this research, we embedded the long
period fiber Bragg grating into the carbon fiber composite material, and the dynamic response of this embedded FBG sensor
under shock loading is examined. Moreover, the specimen was attached foil strain gauge to measure the surface strain of
each specimen. As the result, the strain history and sensitive of FBG were discussed.
W.-J. Lin (*) • L. Tsai • C.-C. Chiang

Graduate Institute of Mechanical and Precision Engineering, National Kaohsiung University of Applied Sciences,
Chien Kung Campus 415 Chien Kung Road, Kaohsiung 807, Taiwan, R.O.C
e-mail: wen9255@gmail.com; liren@cc.kuas.edu.tw
S.-H. Wang
Department of Chemical Engineering & Institute of Biotechnology and Chemical Engineering, I-Shou University,
No.1, Sec. 1, Xuecheng Rd,Dashu Dist, Kaohsiung City 840, Taiwan, R.O.C
52 W.-J. Lin et al.
7.2 Method and Material
7.2.1 SHTB
A split Hopkinson tensile bar facility at Kaohsiung University of Applied Sciences was utilized in this research. The SHTB setup
included striker, incident bar, transmission bar and momentum trap bar, and the bar lengths were 400, 2,000, 1,000 and 400 mm,
respectively. The semi-conductor strain gauges were utilized to measure the wave signal. While elastic wave propagated through
incident bar, the elastic wave had two types. One was passed through to specimen, and the other was reflected back to
incident bar, and by using one dimensional wave theory, the resultant stress vs. strain curves of CFRPs were determined [9–12].
7.2.2 Fiber Bragg Grating Sensor
The FBG sensing system utilized in this research consists of photodiode (PD), light source, optical tunable filter and couple,
as show in Fig. 7.1a. The optical fiber were expanded under tensile forces, then its light passes through core grating (likes
Bessel function) section and its wavelength were displaced. After that, the signal were recorded and analyzed, its signal were
compared with strain signal of strain gauge to calculate the sensitivity coefficients. The L/W ratios were controlled between
two and four rates [13] and FBG sensors of data and foil strain gauges were acquired through static testing, and each FBG
was calculated to ensure its high accuracy.
7.2.2.1 Principal of Optical Fiber Gratings
Theory of grating was defined in geometrical optics of diffraction: from periodic slit structures. While light sources emitted
in a transverse slit plane, it resulted in a shade of diffractive stripe. And this periodic slit structure was called grating structure
[6, 14].
l
n sin y2 ¼ n sin y1 þ m (7.1)
L
Fig. 7.1 Schematic diagram

of (a) specimen [0/45/0/-45/
0]s, and (b) Fiber Bragg
Grating sensor of principle
7 High Strain Rate Characteristics of Fiber Bragg Grating Strain Sensors 53
Fig. 7.2 The specimen of

(a) and (b) CFRP were before
and after dynamic loading
In Eq. 7.1, n is the refractive index of medium, l is the wavelength, L is the grating of period, m is the diffraction of order,
y1 is the light of incident angle and y2 is the diffractive angle for 1 order.
In Fig. 7.1b, optical fiber grating structure was engraved onto optical fiber. The light emitted in optical fiber and resulted
in diffractive phenomenon. The propagation constant was defined given by [6, 14]:

2p
b¼ neff (7.2)
l
7.2.2.2 Theory of Fiber Bragg Grating
Fiber Bragg Grating were called reflective fiber grating, furthermore, it utilized reflection to monitor light reflective
spectrums. Forward incident light were coupled with backward reflective light, and it was reflective via grating section.
The Fiber Bragg Grating equation was given by [6, 14]:
l ¼ 2neff L (7.3)
The l is the Bragg wavelength, neff is the effective reflective index and L is the periodic grating [6, 14].
7.2.3 Specimen Preparation
In this study, the material of pre-preg was combined with 63% fibrous and 37% matrix. The width and thickness of
specimens were designed as 4 and 2 mm, and its layers were folded up with [0/45/0/-45/0]s. In the specimen curing process,
it was cured under a vacuum situation, and was heated to 240 C. The pre-pregs were air-cooled until its temperature
decreased to room temperature. Then each specimen was machined to dog-bone shape by water-jet cutting as shown in
Fig. 7.2.
7.3 Result and Discussion
In this research, Split Hopkinson Tensile Bar (SHTB) was utilized to study the dynamic behavior of FBG and materials.
The specimens were fixed in incident bar and transmission bar, and then the air pressure projected the impact bar to strike the
incident bar. The strain waves propagated through the specimen, between incident bar and transmitted bar. The part of stress
54 W.-J. Lin et al.
Table 7.1 Mechanical property Strain rate (s1) Young’s modulus (GPa) Yield stress (MPa)
of CFRP under high strain rate
135 158.7 455.3
323 275.6 953.5
492 223.6 1,333.8
721 147.3 1,445.9
983.5 165.5 1,833
Fig. 7.3 (a) Normal FBG sensor surfaced, (b) FBG sensor, and (c) CFRP was surfaced feature under high-speed loading
a b
0.0016
Y = 0.004577828496 * X - 8.16E-005
R-squared = 0.953789
0.0012
Strain
0.0008
0.0004
0
0 0.1 0.2 0.3 0.4
Voltage
Fig. 7.4 (a) The FBGs of date and strain of value were calculated for linear equations, (b) Signal of FBG sensor and Strain gauge under high strain
rate 492 s1
wave was reflected to incident bar because there were impedance mismatch. In order to understand the FBG sensor wave
spectrum under dynamic situation, it FBGs were embedded inside CFRP materials, and foil strain gauges were attached on
each specimen. In the dynamic responses of CFRP, stress–strain curve, FBG sensors and strain gauges measurements were
presented in Fig. 7.5a–c, respectively. Young’s modulus and yield stress were shown in Table 7.1.
Strain rates between 135 s1 ~ 983.5 s1 were applied to examine the dynamic response of FBGs and CFRPs. Young’s
modulus was direct proportional to strain rates. The specimens exhibit tensile fracture at the neck of specimen under high
stain rate loadings. According to scanning electron microscopy, as shown in Fig. 7.3, after dynamic loading, fibers inside
specimen were cracked at certain degrees. After shock loading the FBG sensors were cracked, it’s conceivable to understand
the brittle material in a destroyed situation.
Young’s modulus of CFRP in this research have shown the results of 158.7 GPa at strain rate 135 s1 and 165.5 GPa at
strain rate 983.5 s1, yield stress of CFRP was 455.3 MPa at strain rate 1351 and 1,833 MPa at strain rate 983.5 s1,
respectively. When comparing the recorded strain signal from FBG and surface strain gauge, as shown in Fig. 7.4b, FBG
sensor exhibit higher precision.
In Fig. 7.4a, the gauge factors of FBG were calibrated using a MTS tensile test before each experiment. Figure 7.4b shows
the raw data of both FBG and surface strain gauge. From this voltage vs. time profile, it is quite clear that the FBG strain
sensor has higher sensitivity and quicker response time. CFRP specimen was strain measured by FBG sensor and foil strain
gauge increased sharply under shock loadings. The strain vs. time as well as stress vs. strain curves were presented in
7 High Strain Rate Characteristics of Fiber Bragg Grating Strain Sensors 55
Fig. 7.5 (a) FBGs and (b) strain gauge of strain-time curve under strain rate 135 to 982.5 s1, (c) the stress–strain curve of CFRP under strain rate
135 to 982.5 s1
Fig. 7.5. In Fig. 7.5a, b, strain vs. time profile of FBGs and surface strain gauges were presented, respectively. At lower strain
rate conditions, both FBG and strain gauge curves exhibit J-shape curve feature, however, when strain rates increased to
more than 700 s1, strain gauge signals started to vibrate severely at higher strain range. As shown in Fig. 7.5c, the SHTB
results showed clear yield strength increments with increasing strain rates between 135 and 982.5 s1.
7.4 Conclusion
The stresses of specimen were direct proportional to strain rate. The results of SHTB measures and FBG sensors and foil
strain gauges were investigated under dynamic loading from 135 to 983.5 s1. FBG sensors were more environment
tolerance than foil strain gauges, and in addition, FBG sensors were embedded inside materials and monitored the material
strain conditions inside composite materials.
Acknowledgments This study was supported from National sciences council, and plan of no. NSC-99-2221-E-151-014 National Kaohsiung
University of applied sciences in Taiwan. It’s deeply appreciated Professor L.R. Tsai and Professor C.C. Chiang to help in this research.
References
1. Suk SK, Bhowmik SR, Windhorst T, Blount G (1997) Carbon-carbon composites: a summary of recent developments and applications. Mater
Des 18(1):11–15
2. Lee HC (2010) The influences of residual stress in epoxy carbon-fiber composites under high strain-rate. Master thesis, National Kaohsiung
University of Applied Sciences, Kaohsiung Taiwan
3. Devi GR, Rao KR (1993) Carbon-carbon composites – an overview. Def Sci J 43(4):369–383
4. Lifshitz JM, Leber H (1998) Respond of fiber-reinforced polymers to high strain-rate loading interlaminar tension and combined tension/shear.
Compos Sci Technol 58(6):987–996
5. Kashyap R (2010) Fiber Bragg grating. Photon design. Elsevier Inc, pp 596, ISBN 978-0-12-372579-0
6. Liou J (2010) Curing monitor of the composite material by optical fiber grating sensors. Master thesis, National Kaohsiung University of
Applied Sciences, Kaohsiung Taiwan
7. Kuang KSC, Kennyb R, Whelanb MP, Cantwell WJ, Chalker PR (2001) Embedded fibre Bragg grating sensors in advanced composite
materials. Compos Sci Technol 61(10):1379–1387
8. Okabe Y, Yashiro S, Kosaka T, Takeda N (2000) Detection of transverse cracks in CFRP composites using embedded fiber Bragg grating
sensors. Smart Mater Struct 9(6):832–838
9. Griffiths LJ, Martin DJ (1974) A study of the dynamic behaviour of a carbon-fibre composite using the split Hopkinson pressure bar. J Phys D:
Appl Phys 7(17):2329–2341
10. Harding J, Welsh LM (1983) A tensile testing technique for fibre-reinforced composites at impact rates of strain. J Mater Sci 18(6):1810–1826
11. ASM International (2000) Mechanical testing and evaluation, ASM handbook, vol 8. ASM International, Materials Park
12. Kolsky H (1949) An investigation of the mechanical properties of materials at very high rates of loading. Proc Phy Soc B 62(11):676–700
13. Yang X, Nanni A, Dharani L (2001) Effect of fiber misalignment on FRP laminates and strengthened concrete beams. Struct Faults Repair :4–6
14. Jhong LW (1999) The research of Fiber Gratings theory computation and actual formation. Master thesis, Chung Cheng Institute of
Technology National Defense University, Taiwan
Chapter 8
Dynamic Shape and Strain Measurements of Rotating
Tire in Time-Series
Motoharu Fujigaki and Kousuke Shimo
Abstract Shape measurement and strain distribution measurement are important to analyze the behavior of a rotating tire.
A dynamic shape and strain measurement system by the sampling moire method was developed. Recently, phase analysis
method using phase-shifting moire patterns generated in a computer from a grating image is proposed. This method is very
suitable for dynamic measurement because the phase can be obtained from a one-dimensional or two-dimensional grating
image with high accuracy. It is possible to analyze strain distribution and shape measurement of a moving object using a
grating pattern attached on the specimen surface. A dynamic shape and strain measurement system by the sampling moire
method is developed. In this paper, we applied high-speed cameras to this method. The behavior when a tire got over a
projection with high-speed rotating was observed as time-series data. The principle and experimental results are shown.
Keywords Dynamic shape and strain measurements • Rotating tire • Time-series • Sampling moire method • High-speed
camera
8.1 Introduction
Shape measurements and strain distribution measurements are important in analyzing the behavior of a rotating object. The
results can be applied to computational analysis. The stereoscopic method is often used for measuring three-dimensional (3-
D) coordinates on the surface of an object. The 3-D coordinates are obtained from two images of an object recorded from
different directions by two cameras. It is important to find corresponding points accurately between the two images, and in
particular the accuracy of the 3-D coordinates is important when using these coordinates to analyze strain distribution before
and after deformation of the object.
As a conventional method, digital image correlation (DIC) is a well-known method used to find corresponding points
between two images using random patterns [1, 2]. This is a very useful method to use when measuring displacement and
strain distributions. In the case of a rotating tire, measurement methods using Fourier transform have been proposed [3–5].
These measurement methods have some problems, including the time required for analysis, and measurement accuracy.
A spatial fringe analysis method using a sampling technique was proposed by Arai et al. [6, 7]. The method is used to
analyze the phase distribution of phase-shifted moire images produced computationally from a grating image. The method is
very useful and accurate. It is often used for electronic speckle pattern interferometry [8–10]. We applied this method to
analyze a two-dimensional (2-D) grating image and the method is called a sampling moire method [11]. The sampling moire
method applied for a 2-D grating image can be used to find corresponding points between two images highly accurately
using regular 2-D grating patterns.
We applied this method to the dynamic shape and strain measurement for a rotating tire [12]. In this paper, we applied
high-speed cameras to this method. The principle and experimental results are shown.
M. Fujigaki (*)
Department of Opto-Mechatoronics, Faculty of Systems Engineering, Wakayama University, 930, Sakaedani, Wakayama 640-8510, Japan
e-mail: fujigaki@sys.wakayama-u.ac.jp
K. Shimo
Graduate School of Systems Engineering, Wakayama University, 930, Sakaedani, Wakayama 640-8510, Japan
58 M. Fujigaki and K. Shimo
8.2 Principle
8.2.1 Phase Analysis of 2-D Grating Patterns with Sampling Moire Method
The principle of the phase analysis method for a one-dimensional grating in a sampling moire method is essentially the same
as the spatial fringe analysis method using a sampling technique [6, 7].
Figure 8.1 shows an x-dimensional phase analysis using the sampling moire method. Figure 8.1a shows the 2-D grating
pattern image captured by a CCD camera. Figure 8.1b shows a grating image obtained after the y-dimensional smoothing
process with two times the sampling pitch. The x-dimensional phase distribution can be analyzed because the 2-D grating is
transformed into a 1-D grating pattern image. Figure 8.1c shows the phase-shifted moire patterns produced from Fig. 8.1b.
Figure 8.1d shows the phase distribution produced from Fig. 8.1c by the phase-shifting method. Figure 8.1e shows the
phase distribution of the x-directional reference grating. The period of the reference phase distribution is the same as
the pitch of the sampling pixels in the sampling moire method. Figure 8.1e shows the phase distribution obtained for the
x-directional grating component in the 2-D grating image shown in Fig. 8.1a. Figure 8.1g shows the x-directional unwrapped
phase distribution.
In the same process, the y-dimensional unwrapped phase distribution can be obtained.
8.2.2 Calibration with Two Reference Planes
Figure 8.2 shows a schema of the arrangement of cameras, reference planes and an object. Both the left and right CCD
cameras capture the reference planes and the object. In this method, 3-D coordinates corresponding to each pixel of the
camera are obtained with two reference planes for calibration.
A liquid crystal display (LCD) is used as the reference plane. The reference plane is placed at two positions, R0 and R1.
The reference plane can display grating patterns in the x- and y- directions with shifting phases. An initial phase for the
grating patterns corresponding to each pixel of both cameras can be obtained accurately with a phase-shifting method.
The unwrapped phases for the x and y directions (fnx, fny) (n ¼ 0, 1) are also obtained, with an unwrapping process, at each
corresponding pixel of both cameras.
Fig. 8.1 Phase analysis of 2-D grating patterns with sampling moiré method for x-direction. (a) 2-D grating pattern image captured by a CCD
camera, (b) Grating image obtained after the y-dimensional smoothing process, (c) Phase-shifted moire patterns produced from (b), (d) Phase
distribution produced from (c) by the phase-shifting method, (e) Phase distribution of the x-directional reference grating, (f) Phase distribution for
the x-directional grating component in the 2-D grating image shown in (a), (g) x-directional unwrapped phase distribution
8 Dynamic Shape and Strain Measurements of Rotating Tire in Time-Series 59
Fig. 8.2 Arrangement of

cameras, reference plane and
object
The 3-D coordinates (x, y, z) of a point on the reference plane are calculated from the x directional grating pitch px, the
y-directional grating pitch py and the z position of the reference plane zn (n ¼ 0, 1), as shown in Eq. 8.1.
px fnx py fny
x¼ þ x0 ; y¼ þ y0 ; z ¼ zn (8.1)
2p 2p
Here py and py are grating pitches for the x- and y- directions, respectively, and (x0, y0) are the coordinates at the point
where (fnx, fny) ¼ (0, 0). 3-D coordinates at a point on the reference plane where the pixel records an image are known for
every pixel from each camera. A pixel from each camera has two corresponding points on the two reference planes. A line
can be obtained by connecting the two points; that is, a line corresponding to each pixel of each camera can be determined
from the results of the calibration method.
8.2.3 Shape and Strain Distribution Measurements
An object is placed between two reference planes, and the 2-D grating attached to the object is captured by the same set of
CCD cameras.
Line l1 passes through point S on the object as recorded by CCD camera 1 and line l2 also passes through point S on the
object as recorded by CCD camera 2 as shown in Fig. 8.2. Line l1 is the line of sight of a pixel from CCD camera 1 and line l1 is
the line of sight of a pixel from CCD camera 2. Points Clb and Clf are the intersections of line l1 with reference planes R0 and
R1, respectively. Points Crb and Crf are the intersections of line l2 with reference planes R0 and R1, respectively. The positions
of reference planes R0 and R1 are z ¼ 0 and z ¼ Dz, respectively. An expression for the straight line l1 and the straight line l2
can be obtained from each point Clb, Clf, Crb and Crf. The 3-D coordinates of point S are obtained as the intersection of the two
lines l1 and l2.
Strain can be calculated from the 3-D coordinates of two points, both before and after deformation.
Fig. 8.3 Relationship between the 2-D grating on the object and the images recorded by the left and right cameras
Fig. 8.4 Relationship between the 2-D grating on the object and the phase analyzed image of the x- and y-directions. (a) 2-D grating on the object,
(b) x-directional unwrapped phase distribution, (c) y-directional unwrapped phase distribution
8.2.4 Finding Corresponding Points
Figure 8.3 shows the relationship between the 2-D grating on an object and the recorded images taken by the left and right
cameras. Let us explain how to find corresponding points on the two cameras. Every grating line in the x- and y-directions has
a line number in each direction, and every intersection has its own address, such as (i, j), composed of the x- and y-directional
line numbers i and j, respectively. These line numbers increase or decrease sequentially on the surface of the object.
Therefore, we are able to number every grating line on the images in order. The point on the image whose address is (i, j)
is the corresponding point on the surface of the object whose address is (i, j). Accordingly, correspondence between the image
and the object is obtained by matching the x- and y- directional line numbers on the images.
Figure 8.4 shows the relationship between the 2-D grating on the object and the phase analyzed image of the x- and
y- directions. The corresponding points are found from the unwrapped phases (fx, fy).
8.3 Experiment
8.3.1 Experimental Setup
Figure 8.5 shows the experimental setup. A tire is set on a tire testing machine with a rotating drum as shown in Fig. 8.6. This
test machine can control the rotational speed and the load on the axis. The movement of the drum produces the rotation of the
tire. The rotating drum has a projection whose cross-sectional shape is a semicircle with radius 10 mm is fixed on the drum.
The tire is rotated at 80 km per hour (circumferential velocity). The load on the axis is 4 kN.
The rotating tire has a 2-D grating with 2.0 mm pitch on its surface as shown in Fig. 8.7. The 2-D grating pattern is printed
on a thin rubber sheet. The images of the tire are captured by two high-speed cameras (FASTCAM-1024PCI, PHOTRON
LIMITED). The frame rate is 2,000 fps with exposure time 0.02 ms. Image size is 1,024 512 pixels. The distance between
the two cameras is 560 mm and the distance between the tire and the center of the two cameras is 1,350 mm.
Fig. 8.5 Experimental setup
Fig. 8.6 Rotating drum and a projection

Fig. 8.7 2-D Grating pattern

attached on a tire surface
Two markers for producing a camera trigger are put on both the rib of the wheel and the drum as shown in Fig. 8.6. Each
marker can be caught with a corresponding optical trigger device. A camera trigger pulse is generated when both markers are
caught concurrently. The high-speed cameras take images of 30 frames before the trigger and 30 frames after the trigger.
8.3.2 Experimental Results
Figure 8.8a, b shows time-series images captured by camera 1 without a projection and with a projection, respectively.
Figure 8.9a, b shows the results of height distributions (z-coordinate distributions) in time-series without a projection and
with a projection, respectively. The results show that a part near contact area is deformed by the projection.
Figure 8.10a, b shows the results of strain distributions for the x-direction and for the y-direction in time-series,
respectively. In this case, the initial 3-D coordinates for the strain calculation are the measured 3-D coordinates in the
corresponding position of the tire without a projection in the corresponding time. Figure 8.10a shows that the high strain area
for the x-direction appears ahead of the contact area. Figure 8.10a shows that high strain area for the y-direction appears on
the contact area.
8.4 Conclusions
The sampling moire method applied for a 2-D grating image can be used to find corresponding points between two images
highly accurately using regular 2-D grating patterns. We applied this method to the dynamic shape and strain measurement
with high-speed cameras for a rotating tire. The 3-D shape and strain distribution of a tire rotating at 80 km/h can be
measured with this method. The behavior when a tire got over a projection with high-speed rotating was observed as time-
series data.
Fig. 8.8 Time-series images

captured by camera 1,
(a) Without a projection,
(b) With a projection
Fig. 8.9 Height distributions

in time-series. (a) Without
a projection, (b) With a
projection
Fig. 8.10 Strain distributions

in time-series. (a) x-direction,
(b) y-direction
References
1. Bruck HA, McNeill SR, Sutton MA, Peters WH (1989) Digital image correlation using Newton–Raphson method of partial differential
correction. Exp Mech 29–3:261–267
2. Yoneyama S, Kitagawa A, Kitamura K, Kikuta H (2006) In-plane displacement measurement using digital image correlation with lens
distortion correction. JSME Int J Ser A 49–3:458–467
3. Motimoto Y, Fujigaki M (1993) Automated analysis of 3-D shape and surface strain distribution of a moving object using Stereo vision. Opt
Lasers Eng 18–3:195–212
4. Fujigaki M, Yang IH, Morimoto Y (1996) Strain analysis of moving objects using a Fourier Transform grid method. NDT&E Int
29–4:197–203
5. Iwase M (1997) Measurement of tire shape using image processing. Tire Technol Int 1997:40–43
6. Arai Y, Kurata T (1995) Improvement of spatial fringe analysis method using sampling technique on pixel of CCD, Jpn J Optics (in Japanese)
24–9:581–586
7. Arai Y, Yokozeki S, Shiraki K, Yamada T (1997) High precision two-dimensional spatial Fringe analysis method. J Mod Opt 44–4:739–751
8. Arai Y, Yokozeki S (1999) Spatial Fringe analysis method in ESPI. Proc SPIE 3744:222–230
9. Arai Y, Yokozeki S (2004) In-plane displacement measurement using electronic-speckle-pattern-interferometry-based on spatial Fringe
analysis method. Opt Eng 43–9:2168–2174
10. Arai Y, Hirai H, Yokozeki S (2008) Electronic speckle pattern interferometry based on spatial Fringe analysis method using two cameras.
J Mod Opt 55–2:281–296
11. Ri S, Fujigaki M, Morimoto Y (2010) Sampling Moiré method for accurate small deformation distribution measurement. Exp Mech
50–4:501–508
12. Fujigaki M, Shimo K, Masaya A, Morimoto Y (2011) Dynamic shape and strain measurements of rotating tire using a sampling Moire method.
Opt Eng 50(10):101506
Chapter 9
Measurement of Metallic Adhesion Force-to-Elongation Profile
Under High Separation-Rate Conditions
D. Bortoluzzi, M. Benedetti, C. Zanoni, and J.W. Conklin
Abstract The Laser Interferometer Space Antenna (LISA) is a space mission, which will reveal gravitational waves. Some
of its key technologies will be tested by means of the LISA Pathfinder mission, in order to reduce the risk correlated with
LISA challenging requirements. Among its more critical subsystems, the Gravitational Reference Sensor (GRS) has been
designed to host a 2 kg Au-Pt cubic Test Mass (TM) in free-fall, with the tightest requirement ever set on the purity of a
geodesic trajectory. The GRS shows a specific criticality related to the need to constrain the TM during the spacecraft launch
and subsequently inject it in the geodesic, breaking the adhesion arisen at the contact surfaces with its mechanical
constraints. The injection strategy is dynamic: by means of a quick retraction of the release-dedicated holding tips the
TM is detached from the constraint, relying on its inertia. The experiment realized through the Transferred Momentum
Measurement Facility reproduces the injection of the LISA TM into geodesic, aiming at quantifying the residual velocity of
the mass after separation from the release tip. In the field of dynamic behaviour of materials, the TM injection constitutes a
novel area of research, where the strength of metallic adhesion is explored under high separation-rate conditions.
Keywords Dynamic rupture of adhesion • Momentum transfer • Injection into geodesic motion
9.1 Introduction
The Laser Interferometer Space Antenna (LISA [1]) is a joint ESA-NASA mission, which will reveal gravitational waves in
space. Some of its key technologies will be tested by means of the ESA LISA Pathfinder mission, in order to reduce the risk
correlated with LISA challenging requirements. Among its more critical subsystems, the Gravitational Reference Sensor
(GRS) has been designed to host a 2 kg Au-Pt cubic Test Mass (TM) in free-fall, with the tightest requirement ever set on the
purity of a geodesic trajectory. LISA shows a specific criticality related to the need to design a mechanism to constrain
the TM during the launch phase and subsequently inject it in the geodesic. On one side, high constraining forces (on the order
of 103 N) need to be applied to lock the TM against the launch vibration, while on the other side a very weak capacitive force
authority (on the order of 10 6 N) is available to detach it from its constraint against the unavoidable adhesive bonds.
Due to the conflicting requirements between high holding/adhesion force and low control authority, the design approach
followed for the LISA GRS is to split the caging and releasing functions into two different mechanisms, the Caging
Mechanism Sub-System (CMSS) and the Grabbing Positioning and Release Mechanism (GPRM). While the CMSS needs to
secure the TM during the launch, the GPRM is designed to grab it from any position and attitude inside its housing, hold and
D. Bortoluzzi (*) • C. Zanoni

Department of Mechanical and Structural Engineering, University of Trento, via Mesiano 77, 38123 Trento, Italy
e-mail: daniele.bortoluzzi@ing.unitn.it; carlo.zanoni@ing.unitn.it
M. Benedetti
Department of Materials Engineering and Industrial Technologies, University of Trento, via Mesiano 77, 38123 Trento, Italy
e-mail: matteo.benedetti@ing.unitn.it
J.W. Conklin
Department of Aeronautics and Astronautics, Stanford University, 496 Lomita Mall, Stanford, CA, USA
e-mail: johnwc@stanford.edu
68 D. Bortoluzzi et al.
release it in absence of inertial loads. This makes it possible to design the GPRM with limited contact surfaces, reducing
adhesion with the TM. Still, there is no confidence that the residual adhesion is lower than the mN-order force authority.
The injection strategy is then dynamic: by means of a quick retraction of the release-dedicated holding tips the TM is
detached from the contact, relying on its inertia. The released TM may be controlled by the capacitive actuation to its
geodesic if its residual velocity is lower than 5 mm/s.
Due to the criticality of the injection into geodesic phase, a specific testing activity is put in place at the University of
Trento. The experiment realized through the Transferred Momentum Measurement Facility (TMMF) reproduces the
injection of the LISA TM into geodesic motion, aiming at quantifying the residual velocity of the mass after dynamic
separation from the release tip [2, 3]. In the TMMF, the injection of the LISA test mass into a geodesic trajectory is
reproduced in a representative environment of the in-orbit one. The key element in the transfer of momentum to the released
TM is adhesion arising at the contact between the TM and the release-dedicated tip. As a consequence, the TM injection
experiment is focused on a deeper understanding of the behaviour of adhesive bonds under the conditions of dynamic failure
upon quick separation. The retraction velocity of the tip is intentionally high, in order to minimize the impulse developed by
adhesion. The resulting separation rate of the mating surfaces is on the order of 10 2 m/s. This paper presents the main
challenges in the preparation of the test set-up, the test results achieved, and discusses the relevance of earth-bound tests with
respect to what will happen in space.
9.2 Experimental Apparatus
The release experiments are performed using the Transferred Momentum Measurement Facility (TMMF), an apparatus
designed and realized at the University of Trento (Italy) to provide the characterization of the momentum transfer to a
released test mass by the dynamic failure of the adhesive bonds [2–4]. A schematic illustration thereof is given in Fig. 9.1.
The TM mock-up is a prismatic (36 36 4 mm3) tungsten-frame hosting at the centre an Au-Pt gold coated cylindrical
insert. The rear surface is mirror finished for reflecting the laser beams used to measure the TM position and attitude. The
total weight of the TM is about 90 g. The release tip dummy is an Au/Pt lenticular disk with 2 mm diameter, 1 mm thickness,
10 mm radius of curvature. The TM mock-up is suspended by a wire, forming a simple pendulum. The adhesive patch
formed between the two bodies is pulled to failure producing a force impulse on the TM. From the following swing
oscillations of the TM it is possible to estimate the pendulum initial velocity, i.e. the net momentum transferred to the TM by
the adhesive pull.
Fig. 9.1 Schematic overview

of the last version of the
Transferred Momentum
Measurement Facility
(TMMF) operated at the
University of Trento (Italy)
9 Measurement of Metallic Adhesion Force-to-Elongation Profile Under. . . 69
Fig. 9.2 Three-dimensional

solid model of the release
tip positioning and retraction
subsystem
The facility can be schematically subdivided into six subsystems, whose detailed description is given in [3]:
• Vacuum system. It is devoted to guaranteeing the high vacuum level (10 7 mbar) needed for the experiment. In order to
minimize the detrimental effect exerted by the micro seismic activity on the adhesion strength, the vacuum chamber is
installed on the top of an active 6-dof anti-vibration table.
• Cleaning system. It consists of a plasma source able to perform ion etching (typically with Ar+ or O ions) of the mock-up
surfaces for UHV cleaning.
• Suspension and positioning stage. It is aimed at suspending and positioning the TM mock-up with respect to the release
tip (along x, y, z and j-axes, where j-rotation is about z-axis).
• Measurement system. This subsystem is devoted to measuring the position and attitude of the TM mock-up.
• Release tip subsystem. It is devoted to (1) the actuation of the release tip and (2) the application of the preload between the
TM mock-up and the release tip. This subsystem will be described in detail below.
In the more recent versions of the TMMF, the release tip mock-up is actuated through a rigid connection to the linear
stage (Fig. 9.2). The actuation direction can be adjusted by varying both yaw and pitch angular positions of the actuation
systems through a rotational stage and a tilting platform, respectively. The optimal tip actuation direction, maximizing the
transferred momentum to the TM, is tentatively searched for by exploring several pitch and yaw angular positions. The
purpose is twofold: (1) to minimize the spurious actions that limit the conversion of adhesion rupture into linear momentum;
(2) to identify the angle between the pull direction of the release tip and the TM contacting patch that maximizes the strength
of the adhesive junction. In fact, this angle may differ from that of ideal alignment due to deviation from the ideal spherical
tip shape, non-uniform surface roughness, and small ground tilting.
The preload system is schematically illustrated in Fig. 9.3. The TM mock-up is approached and kept still by a blocking
system, composed of three screws with a needle edge, carrying an anti-adhesive coating. The x-position as well as the
attitude in terms of pitch and yaw of the blocking system is adjusted through a linear position and a double-tilting platform.
To apply the preload, the release tip is actuated against the TM. The measurement of the preload force is performed
indirectly from the commanded forward stroke of the retraction system. The calibration is performed through the measure-
ment of the applied force vs. commanded stroke relationship. In order to enhance the forward stroke sensitivity to the applied
force, the release tip is connected to the linear stage through very compliant harmonic steel foils. The attitude of the blocking
system is adjusted so as to keep the TM aligned with respect to both gravity and torsional equilibrium position of the
suspending wire during the loading and unloading phase. This is performed to reduce as much as possible spurious forces
and torques, acting on the contact patch, which can weaken the strength of the adhesive bond.
The main goal of the experiment is the prediction of the residual velocity of the LISA Pathfinder TM after the in-flight
injection into geodesic. However, the results of the TMMF cannot be directly transposed to the in-flight conditions, since the
Fig. 9.3 Test mass blocking

system with pitch and yaw
adjustments
Fig. 9.4 The differential

optical shadow sensor used
to measure the position of
the release tip
actuator used in the present experiment is far slower (three orders of magnitude in terms of initial jerk) than that adopted in the
LISA Pathfinder release mechanism. A possible solution to overcome this obstacle is to assume that the strength of
the adhesive junction between TM and release finger is strain little rate-sensitive, and hence to derive, from the TM and
release finger equations of motion, the adhesive force-to-elongation curve. This can be used to simulate the injection of the
TM into geodesic by means of the mathematical model of the release mechanism, which has been characterized separately
[5, 6] For this purpose, in a recent paper [7], we assumed that adhesion produces a force as a function of elongation of the
bonds, i.e. separation between the contacting surfaces. Adhesion force as a function of time was measured through
the estimation of the second derivative of the TM displacement profile during the incipient motion, when it is dragged by
the actuator through the bonds arisen at the contact. However, since no measurement of the tip motion profile was available,
no direct measurement of the elongation could be performed. This limitation was overcome by separately characterizing the
motion of the tip positioner and assuming it as a known input to the system composed of the blades, release tip and adhered
TM. The adhesion force-to-elongation law was measured indirectly by fitting the dynamic model of the system to the
measured TM motion profile. Following this approach, the force profiles were estimated with fair accuracy (around 10%), but
the fitting procedure required a careful and time-consuming data analysis, also due to the lack of synchronization between the
positioner and the TM displacement signals.
Therefore, the latest version of the release tip subsystem has been equipped with a differential optical shadow sensor
(DOSS), illustrated in Fig. 9.4, which measures the position of the release tip. Two light beams are emitted by two pairs of
Fig. 9.5 Residuals of the

calibration procedure of the
differential optical shadow
sensor. The linear
interpolation performed on a
limited retraction is sufficient
to minimize the residuals to
the noise of the sensor. The
plot on the right shows the
histograms of the residuals
LEDs tangent to a slotted target which is rigidly connected to the release tip such that the target partially blocks the beams
[8]. Any displacement of the target produces a motion of its shadow on the photodiodes and therefore a difference of their
outputs. The differential voltage signal is amplified by a low-noise amplifier and acquired through a low-noise high-
frequency acquisition board, which simultaneously acquires the signal of the laser interferometer pointing the TM. In this
way, the measurements of the TM and Tip displacements are synchronized and from their difference the adhesive bonds
elongation is directly estimated.
The DOSS is calibrated by measuring the displacement of the release tip with the SIOS interferometer and with the
DOSS, at the same time. The ratio between the output voltage and the measured position of the interferometer is known,
thus, it’s possible to correlate the output voltage of the shadow sensor with the position of the tip. If the region of calibration
is small enough (i.e 300 mV), the relation is linear (see Fig. 9.5). This procedure is possible if the calibration is performed
after the adhesion measurement. Once the most interesting or worst-case release tests have been identified, the DOSS is
calibrated around the voltage representing the nominal 0 of elongation for that retraction (i.e. when the system is idle).
9.3 Experimental Procedures and Results
The experimental procedure may be summarized as follows. After the stabilization of the pendulum, the TM is approached
on opposite sides by the blocking needles and the tip. Then, the preload is gradually increased up to the maximum value (400
mN), kept constant for a dwell time of about 10 s and then gradually reduced to a minimum value equal to about 3–4 mN.
The lowest residual preload is left which guarantees a stable contact between the surfaces, against the detrimental effect
exerted by the micro-seismic vibration on the adhesion strength. Finally, the release tip is retracted 2,000 mm away from the
TM mock-up by commanding the linear stage at the highest acceleration (100 m/s2). During the first 0.5 s preceding and
following the release, the TM position and attitude are acquired. In order to evaluate the momentum transfer to the test mass,
the pendulum has been designed to provide large and easily measurable swing oscillation with a resonant frequency of about
0.47 Hz. More details on the measurement performance of the TMMF may be found in [9–11].
The direction of retraction, which maximizes the imparted impulse, is accurately searched. For this purpose, a two-
parameter test matrix has been explored by varying the tilt and yaw of the tip with respect to the nominal alignment between
the pulling direction and the normal to the TM surface. The matrix has 25 elements spaced by 0.3 and contains pairs of pitch
and yaw angles ranging from 0.6 to 0.3 and from 0.6 to 0.6 with respect to the orthogonal, respectively. The release tests
have been performed in the same conditions in terms of maximum preload before release and maximum commanded
acceleration of the tip. Five of these tests have been carried out per each pair of pitch and yaw angles. The results of the test
matrix in terms of mean impulse are illustrated in Fig. 9.6. It can be noted that there is a clear correlation between the pull
Fig. 9.6 Maximum impulse in [Ns] over five tests per each of the couples of tilt and yaw angle explored. The mesh is colour interpolated over
steps smaller than the real one (0.3 )
Fig. 9.7 Displacement (left) velocity (center) and acceleration (right) time profiles of the experiments leading to the highest imparted impulse
direction of the tip and the mean impulse. It is worth underlying that the peak of impulse reaches the requirement of
maximum release velocity, which is 10 5 Ns.
The release experiment, which led to the highest imparted impulse, was selected to derive the adhesive force-elongation
curve. Figure 9.7 illustrates the displacement signals of the TM (left), the corresponding velocity (centre) and acceleration
(right) signals obtained by displacement differentiation. For this purpose, the derivative is estimated using a Savitzky-Golay
filtering technique based on a polynomial fit applied to the TM displacement signal [4, 5]. A second-order fit is sufficient to
estimate up to the second derivative of the displacement signal, avoiding ill-conditioned fitting problems related to higher
order fitting. The bandwidth of the filter is set so as to attenuate the high-frequency noise superimposed to the adhesion pull,
in order to avoid over – estimation of adhesion strength. The minimum bandwidth (3 dB attenuation at 2,575 Hz) is chosen
which leaves the velocity rise time (i.e. the adhesion pull time, about 1 ms) unaffected.
Fig. 9.8 Estimate of the

adhesion force versus
elongation
Figure 9.8 shows the force-to-elongation function for the direction of maximum impulse (yaw 0.3 , pitch 0.3 ).
The elongation reaches a very high value, which may be a critical factor for the GPRM. As there is no clear sign of the
time instant (and similarly, of the elongation) when the adhesion is broken, the disturbances have to be further minimized in
order to allow a deeper investigation of this issue.
Possible sources of error are:
1. The preload is nominally equal to 3–4 mN, however, the difficulty in identifying the 0 elongation position while
performing the tests may lead to higher values. Moreover, the preload requires longer times than the adhesion in order
to become 0. All this may give a long time to the elongation.
2. The transferred vibrations from the blocking needles to the test mass.
In the scope of the LISA-Pathfinder ground testing activities, these results have to be specified to the in-flight conditions.
This allows the prediction of the real release velocity, the estimation of the failure probability and the identification of
possible solutions.
9.4 Conclusions
In this work the on-ground testing activity for the qualification of the LISA-Pathfinder Grabbing Positioning and Release
Mechanism concerning the test mass release phase is described. This activity aims at characterizing the residual velocity of
the released test mass, in order to assess its compliance with the requirement. The criticality of the release phase lies on the
presence of adhesion between the metallic surfaces of the TM and the release-dedicated tip.
By means of the Transferred Momentum Measurement Facility the TM release experiment is performed on-ground and,
through analysis of the measured data, the metallic adhesion force-to-elongation profile under high separation-rate condition
is obtained. Including this force-to-elongation profile in the dynamic model of the GPRM, the in-flight TM release may be
simulated.
References
1. Bell TE (2008) Hearing the heavens. Nature 452:18–21

2. Bortoluzzi D, De Cecco M, Vitale S, Benedetti M (2008) Dynamic measurements of impulses generated by the separation of adhered bodies
under near-zero gravity conditions. Exp Mech 48:777–787
3. Bortoluzzi D, Benedetti M, Baglivo L, De Cecco D (2011) Measurement of momentum transfer due to adhesive forces: on-ground testing of in-
space body injection in geodesic motion. Rev Sci Instrum 82:125107
4. Bortoluzzi D, Benedetti M, Baglivo L, Vitale S (2010) A new perspective in adhesion science and technology: testing dynamic failure of
adhesive junctions for space applications. Exp Mech 50:1213–1223
5. Bortoluzzi D, Foulon B, Garcı́a Marirrodriga C, Lamarre D (2010) Object injection in geodesic conditions: in-flight and on-ground testing
issues. Adv Space Res 45:1358
6. Bortoluzzi D, Mausli PA, Antonello R, Nellen PM (2011) Modeling and identification of an electro-mechanical system: the LISA grabbing
positioning and release mechanism case. Adv Space Res 47:453–465
7. Bortoluzzi D, Benedetti M, Conklin J, Baglivo LL, Da Lio M (2011) Testing the LISA pathfinder test mass injection into a geodesic. In:
Ubertini F, Viola E, de Miranda S, Castellazzi G (eds) Proceedings of the 20th AIMETA conference, Bologna, 12–15 Sept 2011.
Publi&Stampa, Conselice Italy
8. Remo JL (1997) High-resolution optic displacement measurement using a dual – photodiode sensor. Opt Eng 36:2279–2286
9. Benedetti M, Bortoluzzi D, Vitale S (2008) A momentum transfer measurement technique between contacting free-falling bodies in the
presence of adhesion. J Appl Mech 75:011016
10. Benedetti M, Bortoluzzi D, Baglivo L, Vitale S (2011) An optimal two-input approach for impulse measurements in the nanoNs range
produced by contact forces. Mech Syst Sign Proc 25:1646–1660
11. De Cecco M, Bortoluzzi D, Baglivo L, Benedetti M, Da Lio M (2009) Measurement of the momentum transferred between contacting bodies
during the LISA Test Mass release phase – uncertainty estimation. Meas Sci Technol 20(5):1–15
Chapter 10
Prediction of Constraint Parameters Along the 3D Crack
Front Under Negative Biaxial Loadings
Ragupathy Kannusamy and K. Ramesh
Abstract In the recent past, the effect of constraint parameter along the 3D crack front under different biaxial ratios has
received huge attention due to its practical importance for complex structures. A large amount of information is available in
the literature for predicting in-plane and out-of-plane constraint effects along the 3D crack front subjected to different biaxial
loadings but majority of them are based on positive biaxial or uniaxial loading conditions which has considerable amount of
plane strain effect. Linear elastic fracture mechanics (LEFM) from finite element(FE) analysis using J-integral method
confirms that far-field stress applied parallel to the crack front does not contribute anything to the stress intensity factor (SIF)
as biaxiality has no effect on singular term. However the biaxiality affects the non-singular terms and it can be predicted by
using elastic–plastic FE analysis. Some aircraft components operate under negative biaxial loadings up to 0.5. In this work,
attempts have been made to quantify the effect of in-plane and out-of-plane constraints along the 3D crack front under
different biaxial loadings using elastic–plastic FE analysis approach for AA2014-T6 alloy. Aluminum alloy is used for
current study considering its wide range of applications in the aerospace industry. In the aircraft industry, more than 40% of
the components are manufactured with various aluminum alloys. J-Q two-parameter fracture mechanics approach has been
considered to quantify the constraint effect along the 3D crack front. A comprehensive study was carried out on constraint
parameters with 3D Flat plat models with different a/c and a/t ratios under three different biaxial loadings from 0.5 to +1.
Also the effect of negative biaxiality on constraint parameter was investigated using three different far-field loadings to
address from small scale yielding to large scale yielding (20%, 50% and 70% of yield strength). In order to determine Q
parameter, the reference stress has to be calculated under plane strain condition. Therefore the modified boundary layer
(MBL) approach is modeled with T ¼ 0 condition to determine the reference stress. It was observed from the study that
negative biaxiality has significant influence on constraint parameters along the 3D crack front even at lower loading cases
due to the nature of excess yielding. According to the yielding criteria, tension-compression is the worst load combination
and it reduces the hydrostatic stress to the lowest level as compared to the positive and uniaxial loadings. Plotting of
photoelastic fringe patterns using commercial FE package like ANSYS is demonstrated and numerical validation has been
done with the cruciform specimen. A new methodology is proposed to quantify the impact of constraint parameters under
negative biaxial loadings along the 3D crack front.
Keywords Constraint parameter • MBL • J-Q • Q • H • Biaxial loading • Plane strain • J-integral
10.1 Introduction
Surface cracks in pressure vessel components in power and aerospace industries are often encountered as a major origin of
potential catastrophic failure. The application of conventional fracture mechanics is mainly based on a single parameter
approach like Stress intensity factor (SIF) and the J-integral which quantifies the resistance of a material to fracture [1–3] and
this is applicable if geometry is limited to a high constraint crack geometry resulted from deep surface crack with high
thickness where plane strain condition is significantly high. If the structure contains shallow surface crack, the plastic
deformation along the crack front is very high due to the nature of plane stress condition, particularly if geometry is made of
R. Kannusamy (*) • K. Ramesh

Applied Mechanics Department, Indian Institute of Technology, Madras 600036, India
e-mail: Ragupathy.Kannusamy@honeywell.com
76 R. Kannusamy and K. Ramesh
high ductile materials like aluminum alloys. In such a situation, the single parameter approach like J-integral alone may not
be sufficient to address the stress field around the crack tip. Therefore, it is necessary to consider a two-parameter fracture
mechanics approach like J-Q theory [4]. Two-parameter approach is widely used for prediction of fracture toughness where
geometry is very sensitive with fracture toughness value. The two-parameter approach was originally developed by O’Dowd
[5] based on initial work done by Li and Wang [2–4]. Q parameter measures the constraint level at the crack front and it
indicates the degree of triaxiality field around the crack front. The Q parameter is difference between the stress field from the
geometry of the interest with crack and plane strain reference stress solution [1]. Some of the structural integrity assessment
procedures like Failure Assessment Diagram (FAD) included the methodologies in their recent versions to account the effect
of crack front constraints while assessing the remaining life of the component [5]. A defect like shallow surface cracks
induces a complex stress field on the crack front due to the nature of three-dimensional effects and plastic deformations from
negative biaxial loadings. Hence, it is necessary to carry out a detailed elastic-plastic fracture mechanics (EPFM) analysis
using FE approach under different biaxial loadings.
Constraint parameter in fracture mechanics is an indication of resistance to the plastic deformation of the geometry.
The literature [4–6] indicates that crack front constraints are mainly influenced by variables like,
1. Biaxial/Multiaxial loadings
2. Crack shape and size (a/c, a/t ratios)
3. Magnitude of loadings
4. Plastic zone size and shape
5. Geometries
6. Thermal loadings
A large amount of information is available in the literature [7–11] for predicting the effect of in-plane (normal to the crack
front) and out-of-plane constraints (parallel to the crack front) on 2D and 3D cracks but majority of them are based on plane
strain conditions. Also recently, there is considerable amount of publications [4, 6] available about the effect of crack front
constraints on 3D cracks under different biaxial loadings but most of them are based on either uniaxial or positive biaxial
loadings. Some of the aerospace components operate under negative biaxial loadings up to 0.5. Lacking with limited
amount of information in the literature, understanding the application of two-parameter fracture mechanics approach for
structures operating under negative biaxial loading remains questionable. Therefore, it is necessary to investigate the effect
of negative biaxial loadings on crack front constraint parameters with different 3D surface cracks configurations.
10.2 Material Properties
AA2014-T6 material is considered for current analysis [12, 13]. The true stress–strain curve of the material, shown in
Fig. 10.1, is characterized by a Ramberg-Osgood power law hardening model as shown in Eq. 10.1.
n
e s s
¼ þa (10.1)
eO s O sO
where s0 is taken as reference stress or the 0.2% yield strength. 20 is s0/E, E is the Young’s modulus of the material, a is the
yield offset and n is the material hardening exponent which represents a moderate hardening material used in aerospace
industry. The values of materials properties and constants are as follows.
pffiffiffiffi
sO ¼ 390 MPa; E ¼ 72; 398 MPa; a ¼ 0:372; n ¼ 18: KIC ¼ 32 MPa m
In this study, the isotropic material hardening model is considered.
10.3 Geometries and Finite Element Modeling
Volume integral technique available in ANSYS V13.0 [14] is used to calculate J-integral values. The 3D models with surface
cracks are meshed with 20 noded hexahedral elements at the crack front and remaining regions are meshed with 10 noded
tetrahedral elements as shown in Figs. 10.2 and 10.3. This kind of element combinations tends to reduce the computational
10 Prediction of Constraint Parameters Along the 3D Crack Front Under Negative Biaxial Loadings 77
Fig. 10.1 True stress strain curve for AA2014-T6 material
Fig. 10.2 3D Geometry and FE model (Half symmetry FE model)
Fig. 10.3 FE model with elliptical surface crack (a/c ¼ 0.6, a/t ¼ 0.3) with a refined mesh at the crack front and coarse mesh away from the
crack front
Fig. 10.4 Spider web mesh patterns for 2D crack tip and 3D crack front region
Table 10.1 3D crack geometries and biaxial loading cases

Biaxial loadings and geometries
3D Geometries Biaxial loadings Crack angle
a/c ¼ 1 B ¼ +1 a¼0
a/t ¼ 0.15 B¼0
B ¼ 0.5
a/c ¼ 1 B ¼ +1
a/t ¼ 0.55 B¼0
B ¼ 0.5
a/c ¼ 0.6 B ¼ +1
a/t ¼ 0.3 B¼0
B ¼ 0.5
a/c ¼ 0.5 B ¼ +1
a/t ¼ 0.55 B¼0
B ¼ 0.5
time and also easy to model complex geometries with 3D crack. The first layer of 3D elements along the crack front contain
degenerated elements with one face collapsed but the mid-side nodes are retained at the mid-position to simulate
elastic–plastic FE analysis with 1/r crack tip behavior. Here r represents the distance from the crack tip. In order to capture
the steep strain gradient at the crack front, a very fine spider web mesh pattern is introduced at the crack front as shown in
Fig. 10.4. To ensure the path independent behavior of J-integral approach, the element should be small enough at the crack and
also more number of contours should be extracted [14, 15] to ensure identical J-integral values from each contour. For the
current study, the half symmetry model is considered for FE analysis.
In order to capture the variation of constraint level along the 3D crack front, four different surface crack geometries are
considered as shown in Table 10.1. It is to be noted that Mode I crack geometries were modeled for the current study.
10.4 Background of J-Q Methodology
Crack front constraint is an indication of structural resistance to plastic deformation which arises mainly due to the nature of
geometry and loading conditions [2]. A single parameter fracture mechanics like J-integral, K and CTOD is applicable to
characterize the crack tip field for highly constrained geometries. However, the single parameter fracture mechanics loses its
applicability when there is excess plastic deformation at the crack front. In the case of negative biaxial loading, the geometry
undergoes large scale yielding which leads to significant reduction in triaxiality at the crack front. For negative biaxial
loading, the loss of J-integral dominance is controlled by far-field compressive stress applied parallel to the crack front
which introduces a second order term into the nonlinear asymptotic expansion. The second order term affects the stress
pattern ahead of the crack front and reduces the constraint level. Therefore it is imperative to consider two-parameter
fracture mechanics methodologies like J-T, J-Q and J-A2 [7] to quantify the stress field near to the crack front. In the current
study, J-Q methodology is considered to quantify the effect of constraint along the 3D crack front considering its practical
application for large scale yielding conditions. In the case of non-linear deformation, the crack stress field can be evaluated
by asymptotic series in a similar way as the Williams expansion for linear elasticity [6]. The first term in the series represents
the HRR field as shown below Eq. 10.2
1þn
1 t
J r
sij ¼ s0 sðy; nÞ þ Q sðy; nÞ þ :::::: (10.2)
aso eo I n r J=so
where r and y represent the polar coordinates at the crack tip. a is a material constant.
O’ Dowd et al. [3, 4] studied the variation of crack tip constraint and investigated the effect of second order and higher
order in the series and they proposed a single term called Q-Parameter which collectively accounts the effect of higher order
terms [8]. Q-parameter measures the degree of crack tip constraint due to higher order terms in the asymptotic series. They
also simplified Eq. (10.3) and rewritten as:

sij ¼ sij HRR þ Qso dij (10.3)
where dij is the kronecker delta. Since this equation was defined based on small deformation condition, O’Dowd and Shih [3]
carried out detailed FE analysis using 2D geometries to validate the above equation near to the crack front. They defined the
second term as:
syy ðsyy ÞHRR 2J

Q¼ for y ¼ 0; r ¼ (10.4)
so so
where syy refers to the hoop stress at the distance of r ¼ 2 J/s0 from the crack tip. Later Wang [8] used the hydrostatic
(mean) stress instead of hoop stress proposed by Dodds
sm ðsm ÞT¼0 2J
Q¼ for y ¼ 0; r ¼ (10.5)
so so
s1 þ s2 þ s3
where sm ¼
3
and sm is the hydrostatic stress from the geometry of the interest with crack. (sm)T¼0 is also the hydrostatic stress
calculated using MBL approach under plane strain condition. In order to get (sm)T¼0 term, a detailed elastic–plastic FE
analysis should be carried out to calculate principal stress components under T ¼ 0 condition using MBL approach. In this
study, Q parameter was determined along the crack front nodes from 0 to 90 for different surface cracks under three
different biaxial loadings.
10.5 Variation of J Values Along 3D Crack Front
The J-integral based on energy approach, is a measure of the intensity of the stress and deformation of the crack front for a
nonlinear material obtained by Hutchinson, Rice and Rosengren [4, 15] and is popularly known as the HRR singularity field.
In order to predict the variation of J-integral values along the 3D crack front, two different geometries (a/c ¼ 1, a/t ¼ 0.15
and a/t ¼ 0.55) were considered. The J-integral values are calculated using FE analysis along the crack from 0 to 90 and
results are plotted for three different biaxial loadings subjected to three different far-field stresses (20%, 50% and 70% of yield
stress) as shown in Figs. 10.5, 10.6, and 10.7. From the comparison study, it is confirmed that shallow surface crack has
less variation of J along the crack front for all three different biaxial loadings as compared to a deep surface crack as shown in
Figs. 10.5 - 10.7. In the case of shallow surface crack (a/t ¼ 0.15), the highest J-integral value occurs around 15–20 as shown
in Figs. 10.5a, 10.6a and 10.7a. The results also confirms with earlier investigations [8] on J-integral values under biaxial
loadings. However, In the case of deep crack (a/t ¼ 0.55), the maximum J-integral value occurs around 40–50 . The main
reason could be that a major portion of crack experiences plane strain condition which has resulted in higher J-integral values
a 1.6 b 1.6
a/c=1 B=1
1.4 a/t=0.15 B=0 1.4
a=3mm B=-0.5
1.2 1.2
J-Values, N/mm
J-Values, N/mm
1 1
0.8 0.8
0.6 0.6
a/c=1 B=1
0.4 0.4 a/t=0.55 B=0
a=11mm B=-0.5
0.2 0.2
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90
Crack front, degrees Crack front, degrees
Fig. 10.5 J values along 3D crack front for two different geometries (a/c ¼ 1, a/t ¼ 0.15, a/t ¼ 0.55). (a) J values for a shallow surface crack for
externally applied load equal to 20% of syield, (b) J values for a deep surface crack for externally applied load equal to 20% of syield
a b
20 20
18 a/c=1 B=1 18
16 a/t=0.15 B=0 16
a=3mm B=-0.5
J-Values, N/mm
14
J-Values, N/mm
14
12 12
10 10
8 8
a/c=1 B=1
6 6
a/t=0.55 B=0
4 4 a=11mm B=-0.5
2 2
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90
Fig. 10.6 J values along the3D crack front for two different geometries (a/c ¼ 1, a/t ¼ 0.15, a/t ¼ 0.55). (a) J values for a shallow surface crack
for externally applied load equal to 50% of syield, (b) J values for a deep surface crack for externally applied load equal to 50% of syield
a 50 b 50
45 a/c=1 B=1 45
a/t=0.15 B=0
40 40
a=3mm B=-0.5
35
J values, N/mm
J-Values, N/mm
35
30 30
25 25
20 20
a/c=1 B=1
15 15
a/t=0.55 B=0
10 10 a=11mm B=-0.5
5 5
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90
Fig. 10.7 J values along the 3D crack front for two different geometries (a/c ¼ 1, a/t ¼ 0.15, a/t ¼ 0.55). (a) J values for a shallow surface crack
for externally applied load equal to 70% of syield, (b) J values for a deep surface crack for externally applied load equal to 70% of syield
The results also clearly reveal that biaxial loadings have pronounced effect on J-integral values for the geometry where plane
strain condition is very siginifcant.
10.6 Variation of Plastic Zone Size and Shape Along 3D Crack Front
In order to evaluate the effect of biaxiality on plastic zone size and shape along the 3D crack front, a 3D model with a surface
crack (a/c ¼ 1, a/t ¼ 0.55) is simulated for externally applied equal to 50% of yield strength for three different biaxial
loadings as shown in Fig. 10.8a. To capture the variation of plastic zone size and shape along the 3D crack front, plastic zone
deformation contours are captured at every 15 from the free surface (0 ) to deepest point (90 ) as shown in Figs. 10.8 and
10.9. From the study, it is confirmed that negative loading has huge variation along the crack front for all three biaxial ratios.
It is also noticed that the size of the plastic zone along 3D crack front is substantially increased under negative biaxial
Fig. 10.8 Variation of plastic zone size and shape along the 3D crack front for flat plate model with a/c ¼ 1 and a/t ¼ 0.55 for externally applied
load equal to 50% of syield. (a) Focused mesh along 3D crack front (b) Variation of plastic zone size and shape for B ¼ 0
Fig. 10.9 Variation of plastic zone size and shape along the 3D crack front for flat plate model with a/c ¼ 0.5 and a/t ¼ 0.55 for externally
applied load equal to 50% of syield. (a)Variation of plastic zone size and shape for B ¼ 0.5, (b) Variation of plastic zone size and shape
for B ¼ 1
loading as compared to uniaxial and positive conditions [11]. Therefore the magnitude of constraint under negative biaxial
loading is significantly reduced nearest to the free surface where the crack front has significant amount of plane stress effect.
Positive biaxiality adds more resistance to the plastic deformation which contributes to the magnitude of constraints along
the crack front and also the orientation of the plastic zone shape is turned into the backward direction near to the free surface
(0 ) as shown in Fig. 10.9b.
10.7 Modeling of Modified Boundary Layer (MBL) Method
In order to simulate the plane strain condition, a highly refined spider web mesh using plane strain elements are generated as
shown in Fig. 10.10. To avoid the boundary effect, the maximum radius (rmax) of the disc is taken equivalent to the thickness
of plate or greater than ten times of the crack size [7–9].The FE model has a sharp crack at the crack tip with symmetric
boundary conditions along the crack plane.
The displacement based external loads (Ux and Uy) are calculated using Eqs. 10.6 and 10.7 and the same is applied on the
outer edge of the model. Since MBL method is used for calculating reference stress under plane strain condition, the T-stress
is taken as zero. A small script is written in ANSYS [14] to calculate Ux and Uy at different angular positions from 0 to 180 .
rffiffiffiffiffiffiffi
KI r y 2 y 1n
Ux ¼ cos k 1 þ 2sin þ Tr cos y (10.6)
2m 2P 2 2 2m
rffiffiffiffiffiffiffi
KI r y y ðnÞ
Uy ¼ sin k þ 1 2cos2 þ Tr sin y (10.7)
2m 2P 2 2 2m
where KI is a mode I stress intensity factor (SIF), T is the far-field T-stress, r is a distance from the crack tip, m is shear
modulus and n is the poisson’s ratio. J-integral value is calculated for actual geometry using FE approach and converted into
KI under plane strain condition.
1 n2 2
J¼ K (10.8)
E
where E is young’s modulus.
10.8 Quantification of Q Parameter Using FE Approach
Earlier investigations confirm [8, 9] that J-integral alone can be used to characterize the crack field if the stress normal to the
crack front is within 90% of the HRR field at the distance of 2 J/so from the crack front. J. Hancock and Betegon [1] proved
that single term like J-integral will be dominant if Q parameter value is greater than 0.1 [8]. In the case of negative biaxial
Fig. 10.10 FE modeling of

Modified Boundary Layer
with symmetric boundary
conditions
a 0.2 b 0.2
0.1 0.1
0 0
-0.1 -0.1
-0.2 -0.2
-0.3 -0.3
Q
Q
-0.4 -0.4
-0.5 -0.5
-0.6 a/c=1 B=1 -0.6 a/c=1 B=1
-0.7 a/t=0.15 B=0 -0.7 a/t=0.55 B=0
-0.8 a=3mm B=-0.5 -0.8 a=11mm B=-0.5
-0.9 -0.9
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90
Fig. 10.11 Q parameter along the 3D crack front for two different geometries (a/c ¼ 1, a/t ¼ 0.15, a/t ¼ 0.55. (a) Q for a shallow surface crack
for externally applied load equal to 20% of syield, (b) Q for a deep surface crack for externally applied load equal to 20% of syield
a 0.2 b 0.2
0.1 0.1
0 0
-0.1 -0.1
-0.2 -0.2
-0.3 -0.3
Q
-0.4 -0.4
-0.5 -0.5
a/c=1 B=1
-0.6 -0.6 a/c=1 B=1
a/t=0.15 B=0
-0.7 -0.7 a/t=0.55 B=0
a=3mm B=-0.5
-0.8 -0.8 a=11mm B=-0.5
-0.9 -0.9
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90
Fig. 10.12 Q parameter along the 3D crack front for two different geometries (a/c ¼ 1, a/t ¼ 0.15, a/t ¼ 0.55). (a) Q for a shallow surface crack
loadings, the Q-parameter reaches to this condition even under very less far-field load conditions due to the nature of excess
yielding comes from tension-compression loading combinations [16]. For instance, AA2014-T6 alloy reaches to this
condition if externally applied load is equal to 20% of yield strength as shown in Figs. 10.11–10.13. However, the earlier
investigations [9] confirm that this condition can be reached only if externally applied load is equal to 50% of yield strength
but this study was done with steel alloy under uniaxial and positive biaxial loading conditions.
10.9 Quantification of H Parameter Using FE Approach
Kim et al. [6, 10] used the local stress triaxiality parameter [H] to predict in-plane and out-of-plane constraint on 3D crack
front instead of Q parameter as shown below in Eq. 10.9.
sh s1 þ s2 þ s3 2J
H¼ ¼ for y ¼ 0; r ¼ (10.9)
syield 3syield so
a 0.2 b 0.2
a/c=1 B=1
0.1 0.1
a/t=0.55 B=0
0 0
a=11mm B=-0.5
-0.1 -0.1
-0.2 -0.2
-0.3 -0.3
Q
Q
-0.4 -0.4
-0.5 -0.5
-0.6 a/c=1 B=1 -0.6
-0.7 a/t=0.15 B=0 -0.7
-0.8 a=3mm B=-0.5 -0.8
-0.9 -0.9
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90
Fig. 10.13 Q parameter along the 3D crack front for two different geometries (a/c ¼ 1, a/t ¼ 0.15, a/t ¼ 0.55). (a) Q for a shallow surface crack
a 1.4 b 1.4
1.2 1.2
1 1
σH/σyield
σH/σyield
0.8 0.8
0.6 0.6
a/c=1 B=1
0.4 0.4
a/c=1 B=1 a/t=0.55 B=0
0.2 a/t=0.15 B=0 0.2
a=11mm B=-0.5
a=3mm B=-0.5
0 0
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90
Fig. 10.14 H parameter along the 3D crack front for two different geometries (a/c ¼ 1, a/t ¼ 0.15, a/t ¼ 0.55). (a) H for a shallow surface crack
for externally applied load equal to 50% of syield, (b) H for a deep surface crack for externally applied load equal to 50% of syield
where si (i ¼ 1,2,3) are the principal stresses and sh is the hydrostatic stress. The H parameter is easier to calculate and it
indicates the magnitude of hydrostatic stress along the front. It is observed from earlier publications [5–7] that in some
cases, the von-Mises stress is used in Eq. 10.9 instead of yield strength. They claimed that the constraint parameter (H) is
fundamentally equivalent to Q parameter. However it is to be noted that Q parameter was derived based on plane strain
condition which is assumed to be more conservative for 3D crack problems. In reality, the constraint level along the 3D
crack front is a combination of plane stress and plane strain conditions. In order to account the constraint level, H
parameter is calculated for three different surface crack configurations (a/c ¼ 0.3, a/c ¼ 1, a/t ¼ 0.15 and a/t ¼ 0.55,)
and the results are plotted along the crack front from 0 to 90 as shown in Figs. 10.14 and 10.15. The results indicate that
the hydrostatic stress level for shallow surface is greatly reduced as compared to deep surface crack as shown in
Figs. 10.14b and 10.15b. The study reveals that the crack aspect ratio (a/c) and crack to thickness ratio (a/t) have
significant impact on constraint level in addition to the nature of biaxial loadings. The results are confirmed with the earlier
investigations [6].
a 2
b 2
1.8 a/c=1 B=1 a/c=1 B=1
1.8
a/t=0.15 B=0 a/t=0.55 B=0
1.6
a=3mm B=-0.5 1.6 a=11mm B=-0.5
1.4
1.2 1.4
σH/σyield
σH/σyield
1 1.2
0.8 1
0.6
0.8
0.4
0.2 0.6
0 0.4
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90
Crack front , degrees Crack front, degrees
Fig. 10.15 H parameter along the 3D crack front for two different geometries (a/c ¼ 1, a/t ¼ 0.15, a/t ¼ 0.55). (a) H for a shallow surface crack
for externally applied load equal to 70% of syield, (b) H for a deep surface crack for externally applied load equal to 70% of syield
10.10 Photoelasticitic Fringe Patterns Using FE Technique
Photoelasticty is an optical experimental technique which provides the information of principal stress difference (N) in the
form of fringe contour called isochrornatics [17]. Ramesh and his-coworkers [17] demonstrated a simplified technique to
plot photoelasticty fringes using commercial FE packages like ABAQUS which enables visual comparison with experimen-
tal results. Using the maximum and minimum in-plane principal stresses from the FE analysis for the 2D cruciform
specimen, the total fringe order (N) is computed using Eq. 10.10.
ðs1 s2 Þ
N¼ h (10.10)
Fs
where, s1 is the maximum in-plane principal stress, s2 is the minimum in-plane principal stress, h is the specimen thickness
and Fs is the material stress fringe value. In this work, photoelasticty fringes are plotted for 2D half symmetry cruciform
specimen under plane stress condition for three different biaxial loadings (B ¼ 0.5, 0, 1) using ANSYS V13 as shown in
Fig. 10.16. The results indicate that the variation of fringe field also an indication of plastic zone at the crack tip is very
significant for three different biaxial loadings and plotting methodology was validated with earlier investigations [17].
10.11 Conclusions and Recommendations
In order to quantify the magnitude of constraint parameter along the 3D crack front, a series of elastic–plastic FE analyses
are carried out for four surface crack configurations under three different biaxial loadings (B ¼ 0.5, 0 and 1) using
AA2014-T6 material.
Earlier publications [16] have claimed that the stress field along the crack front is significantly influenced when the
externally applied load applied parallel to the crack front exceeds 40% of yield strength for any material [18]. However in
the current study, it is confirmed that negative biaxiality affects the stress field when externally applied load exceeds 20% of
yield strength for AA2014-T6. Therefore, the effect of biaxiality cannot be quantified in a generalized manner as it varies
with respect to material properties.
1. Under LEFM case, negative biaxiality has little or no influence on stress fields along the crack front. J values are the same
for both uniaxial and positive biaxial load cases. Elastic–plastic FE analysis approach has to be used to capture the effect
of biaxiality on constraint parameters.
2. Two-parameter fracture mechanics theory like J-Q approach can be used for structures operating under negative biaxial
loadings to calculate constraint level along the 3D surface crack front and also this approach can be used to predict when
J-integral (single-parameter approach) deviates from the reference stress solutions.
Fig. 10.16 Photoelasticitic fringe patterns (N) for half symmetry cruciform specimen for three different biaxial loadings (B ¼ 0.5, 0, 1) for
externally applied equal to 30% of yield strength. (a) B ¼ 1, (b) B ¼ 0, (c) B ¼ 0.5
3. The constraint level for shallow surface crack (a/t ¼ 0.15) is nearly the same from all three different biaxial loadings as
compared to deep surface cracks (a/t ¼ 0.55). The study reveals that variation of constraint level is along the crack front
is resultant of biaxial loadings and crack configurations like crack aspect ratio (a/c) and crack to thickness ratio (a/t).
4. Crack front constraint for deep 3D surface cracks, as measured by Q and H parameters, appears to be very sensitive under
negative biaxial loadings as compared to uniaxial and positive biaxial loadings. The main reason is that a large portion of
the crack front experiences the plane strain condition for a deep surface crack (a/t ¼ 0.55) as compared to shallow crack
(a/t ¼ 0.15).
5. The variation of Q parameter along 3D shallow crack from three different biaxial loadings are not significant for the load
case with 20% of yield strength as compared to 50% and 70% of yield strength.
6. J-integral values along the 3D surface crack front for a deep surface crack under negative biaxial loading are significantly
increased as compared to positive biaxial and uniaxial loadings. It is also observed that the magnitude of J-integral value
is highest for a shallow surface cracks when Y reaches around 15–25 under biaxial loadings. The results from this study
confirmed with the earlier investigations [9].
References
1. Betegon C, Hancock JW (1991) Two-parameter characterization of elastic–plastic crack-tip fields, Transaction of ASME. J Appl Mech
58:104–110
2. Larsson C (1973) Influence of non-singular stress terms and specimen geometry on small scale yielding at crack tips in elastic–plastic
materials. J Mech Phy Solid 21:17–26
3. O’Dowd NP, Shih CF (1991) Family of crack tip fields characterized by a triaxiality parameter – I. Structure of fields. J Mech Phy Solid
39:989–1015
4. O’Dowd NP, Kolednik O, Naumenko VP (1999) Elastic–plastic analysis of biaxially loaded center-cracked plates. Int J of Solid Struct
36:5639–5661
5. Ainsworth RA, Sattari-Far I, Sherry AH, Hooton DG, Hadley I (2000) Methods for including constraint effects within SINTAP procedures.
Eng Fract Mech 67:563–571
6. Kim YJ, Chung KH, Kim JS, Young JK (2004) Effect of biaxial loads on elastic–plastic J and crack-tip constraint for crack plate: finite element
study. Int J Fract 130:803–825
7. Yuan H, Brocks W (1998) Quantification of constraint effects in elastic–plastic crack front fields. J Mech Phys Solid 46(2):219–241
8. Wang X (2009) Two-parameter characterization of elastic–plastic crack front fields: surface cracked plates under tensile loading. Eng Fract
Mech 76(7):958–982
9. Wang X (2008) On the quantification of the constraint effect along a three-dimensional crack front. J ASTM Int 5(6):1–12
10. Kim Y-J, Kim J-S, Cho S-M, Kim Y-J (2004) 3D-Constraint effects on J testing and crack tip constraint in M(T), SE(B), SE(T) and C(T)
specimens: numerical study. Eng Fract Mech 71:1203–1218
11. Cravero S, Ruggieri C (2003) A two-parameter framework to describe effects of constraint loss on cleavage fracture and implications for
failure assessments of crack components. J Braz Soc of Mech Sci Eng XXV(4):403–412
12. NASMAT (2005) Material database. NASGRO 5.1, SWRI
13. MIL-HDBK-5H (1998) Material behavior for aerospace structures, USA
14. ANSYS V12 (2009) Ansys Manual, USA
15. Brocks W, Scheider I (2001) Numerical aspects of the path-dependence of the J-integral in incremental plasticity. Technical report GKSS/
WMS/01/08, GKSS Germany
16. Rhodes D, Radon JC (1982) Effect of local stress biaxiality on the behavior of fatigue crack growth test specimens, Multiaxial Fatigue, ASTM
STP 853
17. Simon N, Ramesh K (2010) A simple method to plot photelastic fringes and phasemaps from finite element results. J Aero Sci Tech
62:174–181
18. Lam YC (1993) Fatigue crack growth under biaxial loading. Fatigue Fract Eng Mater 16(4):429–440
Chapter 11
Stress Intensity Factors for Viscoelastic Axisymmetric
Problems Applied to Wood
Rostand Moutou Pitti, Claude Chazal, Florence Labesse-Jied, and Yuri Lapusta
Abstract Many materials used in engineering applications obey to time-dependent behaviours and the mechanical fields are
affected by the time effects. As a result, the evolution of the stresses and strains in these materials appear still very complex
and difficult to study. Among such cases is the situation when the material has an axisymmetric shape and when it is
submitted to a complex fracture loading. In this paper, the creep loading is applied on an axisymmetric viscoelastic
orthotropic material and the stress intensity factors are computed in the opening mode, in the shear mode and in the
mixed mode using to a finite element approach. The uncoupling method is based on M integral, combining the virtual and
real mechanical fields. In the same time, the viscoelastic effects are introduced according to the generalized Kelvin-Voigt
model composed by four branches. The numerical solution is obtained with an incremental viscoelastic formulation in the
time domain. Using a Compact Tension Shear (CTS) specimen, the evolutions of stress intensity factor versus time are
posted in each fracture mode configuration. The obtained results demonstrate the efficiency of the proposed model.
Keywords Stress intensity factors • Axisymmetric problems • Viscoelasticity • Wood materials
11.1 Introduction
Modern advanced materials like composites, bi-materials, polymers and also soft materials are strongly present in the
mechanical and civil engineering industries and play an important role in the integrity and resistance of structures.
An important factor in their serviceability is the crack initiation and growth (see e.g. [1–3]). In industrial applications, one
of the important cases is the symmetry of revolution. It is often present in nuclear power plants [4], engines or flying
equipments [5, 6]. Also, the symmetry of revolution is typical in the wood used in individual or industrial houses [7]. In this
case, the material is submitted to fracture under mixed-mode loading combined with viscoelastic behaviour [8, 9], which
remains a complex problem.
R. Moutou Pitti (*)

Clermont Université, Université Blaise Pascal, Institut Pascal, BP 10448, F-63000 Clermont-Ferrand, France
CNRS, UMR 6602, Institut Pascal, F-63171 Aubière, France
CENAREST, IRT, BP 14070, Libreville, Gabon
e-mail: rostand.moutou.pitti@polytech.univ-bpclermont.fr
C. Chazal
GEMH Laboratory, Université de Limoges, Centre Universitaire Génie Civil, 19300 Egletons, France
F. Labesse-Jied
Y. Lapusta
French Institute of Advanced Mechanics, Institut Pascal / IFMA / Clermont Université, F-63175 Aubière, France
90 R. Moutou Pitti et al.
Fig. 11.1 (a) Crack in cylindrical peace of wood. (b) Cross section of wood
In the literature, some developments have been suggested to approach such problems. Among them, numerical solutions
have been proposed to resolve the time-dependent material behaviour with Maxwell equations [10]. Also, according to non-
dependent integrals, a generalization of the M integral [11] and T and A integrals to viscoelastic materials have been
proposed in order to resolve the axisymmetric problem submitted to environmental loading [12]. The main advantage of this
last approach is to compute the real mechanical fields in a virtual configuration and uncoupling fracture and viscoelastic
parameters. In this paper an adaptation of these integrals to the case of viscoelastic problems in axisymmetric configurations
like wood materials is developed.
According to Fig. 11.1a, the fracture process is an important fact of the collapse of peace of wood using in the timber
building. It observed that after the cut, the wood is often exposed in open air and submitted to environmental and climate
loading during a long time [13]. With the drying cycle actions, the micro cracks are initiated and can propagate under the
creep loading when the wood is used as timber structures. Also, the cracks are often oriented from the center to the boundary
of the sample, see Fig. 11.1a. The observed cracks can be explained by the orthotropic character, the cellular composition of
the specie, but also by its viscoelastic behaviour combined with the axisymmetric property. This paper is strongly motivated
by the necessity to study such an interesting behaviour.
Firstly, the conservative laws and the Langrangian variation is recalled in order to introduce the bilinear form of free
energy density [14, 15] and the virtual extension crack in cylindrical coordinates [16]. Also, the independent path integrals M
[11] is defined and generalized to axisymmetric configuration. Secondly, the integral is defined to viscoelastic behaviour
according to the generalized Kelvin Voigt chain. In order to introduce the numerical fracture process in mixed-mode
configuration, the Compact Tension Shear specimen is applied. The last part proposes the viscoelastic incremental law
routine and the evolutions of stress intensity factors and viscoelastic energy release rate versus the each mixed mode ratio.
11.2 Background of Axisymmetric Integral Parameters
11.2.1 Lagrangian Conservation
According to Moutou Pitti et al. [16], for cracked domain V, the Noether’s theorem [15] translates that the Langrangian
variation is equal to zero for all time t chosen arbitrary and for all real du and virtual dv displacement fields:
ðð
dL ¼ dF dt dV ¼ 0 with Fðu; vÞ ¼ F (11.1)
t V
In the case of Arbitrary Euleurian Langrangian configuration, we have the following notations [11]
d~
vi ¼ dvi ; d~
vi ¼ 0; vi ¼ dvi þ dvi ;
d~ d~
ui ¼ dui ; d~
ui ¼ 0; ui ¼ dui þ dui
d~ (11.2)
where v~ and v are virtual Euleurian and Lagrangian (ALE) displacement fields, respectively. Using relation (11.2), and
considering the virtual extension vector ~
kda, where a is a crack length, the Langrangian (11.1) becomes:
11 Stress Intensity Factors for Viscoelastic Axisymmetric Problems Applied to Wood 91
ðð
@F @F @F @F
dL ¼ dui;j þ du þ dvi;t þ dv kk da dt dV (11.3)
@ui;j @ui;j i;j @vi;t @vi;t i;t
t V
According to Eq. 11.2, the virtual displacement gradient can be written as:
@ui;j @vi;j @ui;t @vi;t

dui;j ¼ kk da ; dvi;j ¼ kk da ; dui;t ¼ kk da ; dvi;t ¼ kk da (11.4)
@xk @xk @xk @xk
Using Eq. 11.4, the bilinear free energy density (11.1) becomes

@F @F @F @F @F @F
dui; j ¼ dui dui ; dui;t ¼ dui dui
@ui;j @ui; j ;j @u i; j ;j @u i;t @u i;t ;t @u i;t ;t

@F @F @F @F @F @F
dvi; j ¼ dvi dvi ; dvi;t ¼ dvi dvi (11.5)
@vi; j @vi; j ;j @v i; j ; j @v i;t @v i;t ;t @v i;t ;t
By introducing expressions 11.5 in the Lagrangian (11.3), and applying the Gauss – Ostrogradski theorem leads to
Z Z Z Z
@F @F @F
dL ¼ kk da dt dV þ ui;k þ vi;k kk nj da dt dS
@xk @ui; j @vi; j
t V t @V
Z Z Z !
@F @F @F @F
ui;k þ vi;k kk da dV þ ui;k þ vi;k kk da dt dV
@ui;t @vi;t @ui; j ; j @vi; j ; j
V t V
Z Z ! Z Z !
@F @F @F @F
þ ui;k þ vi;k kk;j da dt dV þ ui;k þ vi;k kk da dt dV (11.6)
@ui; j ; j @vi; j ; j @ui;t ;t @vi;t ;t
t V t V
@V designates the boundary curve of V or a closed contour around the crack tip, composed by the specific contours [11, 16].
11.2.2 Axisymmetric Integral
According to the Eq. 11.6 and the non-dependent domain and without pressure on the crack lips as proposed by [16], we can
write after mathematical transformation:
Z
1
Maxi ¼ svij;k ui suij vi;k kk nj dG1
2
G1
Z
1
þ svij ui; j ;k þ suij vi; j ;k svij ui; j þ suij vi; j kk;j dV (11.7)
2 ;k ;k
V
The integral (11.7) is defined with a curvilinear integration domain. In order to implement this integral in a finite element
software, it is easier to take into account a surface domain integral. In this context, the curvilinear domain must be
transformed by introducing a vector field ~ y [11, 12]. After simplification, we obtain the modelling form of Maxi integral
called Myaxi adapted to axisymmetric problem.
Z
1
Myaxi ¼ suij vi;k svij;k ui yk;j kk dV
2
O
Z
1
þ sij ui; j ;k þ sij vi; j ;k
v u
sij ui; j þ sij vi; j
v u
yk kk;j dV (11.8)
2 ;k ;k
O
The first term of Eq. 11.8 provides the mixed mode separation for a stationary crack and integrates the axisymmetric
aspect. The second term traduces the dissipated energy induced by the crack-growth process.
Fig. 11.2 Generalized Kelvin Voigt model
11.3 Generalization to Viscoelastic Behaviour
In the case of creep loading, the linear viscoelastic behaviour is based on a generalized Kelvin Voigt model composed by N
cells of Kelvin Voigt associated with a spring in series, see Fig. 11.2.
In this case, the Eq. 11.9 can be generalized for each elastic property as follows:
Z
1 ðpÞ u ðpÞ ðpÞ
Myvaxi ¼ sij vij ðpÞ svij;k ui yk;j kk dV
2
O
Z ðpÞ ðpÞ
1 ðpÞ v ðpÞ ðpÞ
þ sij ui; j þðpÞ suij vi; j ðpÞ v
sij ui; j þ ðpÞ suij vi; j yk kk;j dV (11.9)
2 ;k ;k ;k ;k
O
ðpÞ u
sij and ðpÞ svij indicate the real and virtual stresses of the pth spring respectively, induced by mechanical and thermal fields.
ðpÞ ðpÞ
ui and vi are real and virtual displacement fields in the pth spring respectively. For an orthotropic media, virtual field vðpÞ is
given by the Sih’s singular form. According to Eq. 11.9, the real stress intensity factors are given by:

ðpÞ v ðpÞ ðpÞ ðpÞ v ðpÞ ðpÞ
Myvaxi KI ¼ 1; v KII ¼ 0 Myvaxi KI ¼ 1; v KII ¼ 1
ðpÞ ðpÞ
u
KI ¼8 ðpÞ
and u
KII ¼ 8 ðpÞ
(11.10)
C1 C2
Finally, the viscoelastic energy release rate in each fracture mode are given by

u ðpÞ u ðpÞ
KI KII
ðpÞ ðpÞ
GðpÞ 1 ðpÞ 2 ðpÞ
v ¼ Gv þ Gv ¼ C1 þ C2 (11.11)
8 8
ðpÞ ðpÞ ðpÞ ðpÞ
1
Gv and 2 Gv are the energy release rate of the pth spring in mode I and II respectively. C1 and C2 designate the associate
viscoelastic compliance.
11.4 Numerical Results
11.4.1 Axisymmetric Plan and Compact Tension Shear Specimen
In the axisymmetric configuration, the circular crack evolution and the integration domain are posted in Fig. 11.3 [17]. The
crack tip is oriented in the radial direction and the continuum map y surrounds the cracked surface.
In order to compute the stress intensity factors in the axisymmetric material, a modified CTS specimen, Fig. 11.4,
proposed by Moutou Pitti et al. [16] for orthotropic material has been considered. This specimen is used in computational
finite elements to generate the different mixed-mode ratios. The initial square form specimen is replaced by the cylindrical
Fig. 11.3 Crack growth

position in axysimmetric
plan [17]
Fig. 11.4 Compact Tension

Shear specimen [16]
form with the diameter d ¼ 100 mm and the initial crack length a ¼ 50 mm. The loading points Aa and Ba with a ∈ (1 . . . 7)
are oriented according to the angle solicitation b. The pure mode I (b ¼ 0 ) is obtained by using opposite forces in A1 and B1,
the pure mode II by the loading points A7 and B7 (b ¼ 90 ).
11.4.2 Numerical Routine
Figure 11.5 presents the viscoelastic incremental and the fracture routine implemented in finite element software Castem.
• In the beginning, the elastic mechanical fields are computed at time tn1 with the initial circular crack length a ¼ 50 mm,
posted in Fig. 11.4. Also, the representative circular mesh in the axisymmetric configuration is shown in Fig. 11.3.
IN
Mechanical fields
at time tn-1
Viscoelastic incremental routine

mechanical fields at time tn
Virtual routine Real stress intensity factors Virtual routine

in mode I, vk(p)
I
uk (p)
I and uk (p)
II
(p)
in mode II, vkII
Axisymmetric Mqvaxi
routine
if a ≤ d - 30 if a ≤ d - 30
Kv1 and Kv2
if a = d - 30
STOP
Fig. 11.5 Viscoelastic incremental and fracture routine
• At the following crack length a ¼ 51 mm, the viscoelastic incremental formulation proposed by [11, 16, 18] is applied in
order to compute the viscoelastic mechanical fields at time tn . In the fact, the Boltzmann equation introduced with the
generalized Kelvin Voigt model (Fig. 11.2) is resolved step by step in finite element process.
ðpÞ
• Simultaneously, the compliance factors, the virtual stress intensity factors in opening v KI and in shear mode [15] have
ðpÞ ðpÞ
calculated according to the crack virtual procedure. Then, the real stress intensity factor u KI and u KI obtained by
considering the Eq. 11.10.
• The real stress intensity factors are introduced in the axisymmetric Myvaxi routine according to the relation 11.9. Then,
using the Eq. 11.11, the viscoelastic stress intensity factors in opening mode kv1 and in shear mode kv2 for the
axisymmetric materials are evaluated. The precedent steps are repeated until the final collapse of the wood material
with the crack length a ¼ 70.
11.4.3 Stress Intensity Factor and Energy Release Rate
In the numerical finite element calculation, the stress plan configuration is considered. The creep function, the elastic
orthotropic moduli of the pine spruce are considered with a constant Poisson coefficient v ¼ 0:3, see Moutou Pitti et al. [16]
and three Kelvin Voigt cells have been used according to the Fig. 11.2.
Figure 11.6 presents the evolutions of the viscoelastic stress intensity factors versus time in opening mode kv1 and in
shear mode kv2 for the mixed-mode loading b ¼45. It observed that the kv1 is around 0.50 N.mm3/2 and kv2 is around
2.20 N.mm3/2; this values increase slowly after 6 days. The low differences of values prove that the time calculation must
be increase in order to obtain the important stress intensity factors. Also, the value of kv2 is six time more important that kv1;
Fig. 11.6 Stress intensity

factor versus time for
mixed-mode b ¼45
Fig. 11.7 Stress intensity

factor versus crack growth
position in mixed-mode b ¼45
this fact is justified by the important energy required by the material to propagate in shear mode comparatively to the
opening mode.
Figure 11.7 presents the evolutions of the stress intensity factors versus crack length a in opening mode kv1 and in shear
mode kv2 for the mixed-mode loading b ¼45. The final crack length is a ¼ 69 mm and the considered time increment is
Dt ¼ 1h. First, the increasing of stress intensity factors versus crack length is observed. The same evolutions as proposed by
Moutou Pitti et al. [16] about viscoelastic energy release rate are obtained. The maximum values of kv2 vary between
1.77–5.27 N.mm3/2, and 0.36–1.19 N.mm3/2 in opening mode, respectively. Note that the initial values at the crack length
a ¼ 50 mm define the critical energy realise rate on the considered material.
11.5 Conclusion
The evolution of stress intensity factors in viscoelastic orthotropic materials like wood have been investigated in this work.
The analytical form of the axisymmetric integral has been developed according to the conservative law and the bilinear free
energy density. In order to resolve this integral with finite element software, a modelling form of Maxi integral, called Myaxi
has also been proposed and generalized to time dependent materials by Myvaxi . The numerical mixed-mode crack growth has
been introduced according the Compact Tension Shear specimen and a specific axisymmetric crack growth plan. The stress
intensity factors in mixed-mode b ¼ 45 have posted versus time and crack growth for the opening and the shear mode.
It observed that the model is efficiency to compute the stress intensity factors in these types of material, but the
computational time must be increase. In the coming works, the all mixed-mode ratios must be calculation by taking into
account the pressure on the crack lips.
Acknowledgements The work at the Institut Pascal and IFMA is supported by Laboratoire d’excellence IMobS 3.
References
1. Moutou Pitti R, Alaa C, Chazal C (2011) Reliability analysis of mixed mode cracking with viscoelastic orthotropic behaviour. Mechanics of
time-dependent materials and processes in conventional and multifunctional materials, vol 3. In: Conference proceedings of the society for
experimental mechanics series, 2011, vol 99999, pp 249–256, doi: 10.1007/978-1-4614-0213-8_36
2. Yasniy P, Maruschak P, Lapusta Y (2006) Experimental study of crack growth in a bimetal under fatigue and fatigue-creep conditions. Int J
Fract 139(3–4):545–552. doi:10.1007/s10704-006-0102-7
3. Lapusta YN, Henaff-Gardin C (2000) An analytical model for periodic a -layer cracking in composite laminates. Int J Fract 102(3):73–76.
doi:10.1023/A:1007696725543
4. Sarler B (1998) Axisymmetric augmented thin plate splines. Eng Anal Bound Elem 21:81–85. doi:0955-7997/981519.00
5. Jiang Q, Gao CF (2010) Axisymmetric stress in an electrostrictive hollow cylinder under electric loading. Acta Mech 211:309–321.
doi:10.1007/s00707-009-0228-6
6. Yosibash Z, Hartmann S, Heisserer U, D€ uster A, Rank E, Szanto M (2007) Axisymmetric pressure boundary loading for finite deformation
analysis using p-FEM. Comput Methods Appl Mech Eng 196:1261–1277. doi:10.1016/j.cma.2006.09.006
7. Thibauta B, Gril J, Fournier M (2001) Mechanics of wood and trees: some new highlights for an old story, Mécanique du bois et biomécanique
des arbres: nouveaux regards sur une vieille question. CR Mecanique 329:701–716
8. Atkinson C, Eftaxiopoulos DA (1992) Crack tip stress intensities in viscoelastic anisotropic bimaterials and the use of the M-integral. Int J
Fract 57:61–83
9. Chalivendra VB (2009) Mixed-mode crack-tip stress fields for orthotropic functionally graded materials. Acta Mech 204:51–60. doi:10.1007/
s00707-008-0047-1
10. Assous F, Ciarlet P Jr, Labrunie S, Segré J (2003) Numerical solution to the time-dependent Maxwell equations in axisymmetric singular
domains: the singular complement method. J Comput Phys 191:147–176. doi:10.1016/S0021-9991(03)00309-7
11. Moutou Pitti R, Dubois F, Petit C, Sauvat N, Pop O (2008) A new M-integral parameter for mixed-mode crack growth in orthotropic
viscoelastic material. Eng Fract Mech 75:4450–4465. doi:10.1016/j.engfracmech.2008.04.021
12. Moutou Pitti R, Dubois F, Petit C (2010) Generalisation of T and A integrals to time dependent materials: analytical formulations. Int J Fract
161:187–198. doi:10.1007/s10704-010-9453-1
13. Ormarsson S, Dahlblom O, Petersson H (1998) Numerical study of the shape of Sawn timber subjected to moisture variation, Part 1: theory.
Wood Sci Technol 32:325–334
14. Chen FMK, Shield RT (1977) Conservation laws in elasticity of J-integral type. J Appl Mech Phys 28:1–22
15. Noether E (1983) Invariant variations problems. Trans Theory Stat Phys 1:183–207
16. Moutou Pitti R, Chazal C, Labesse-Jied F, Lapusta Y (2011) A generalization of Mv integral to axisymmetric problems for viscoelastic
materials. Acta Mech 220:365–373. doi:10.1007/s00707-011-0460-8
17. Dubois F, Moutou Pitti R, Picoux B, Petit C (2012) Finite element model for crack growth process in concrete bituminous. Adv Eng Softw
44(1):35–43. doi:10.1016/j.advengsoft.2011.05.039
18. Chazal C, Moutou Pitti R (2011) Incremental constitutive formulation for time dependent materials: creep integral approach. Mech Time
Depend Mater 15(3):239–253. doi:10.1007/s11043-011-9135-z
Chapter 12
Role of Viscoelasticity in Predicting the Shape Memory
Effect of Polymers
Diani Julie, Gilormini Pierre, Frédy Carole, and A.R. Ingrid
Abstract Torsional tests are used to study the shape memory property of an amorphous thermoset epoxy. The effect of the
thermal history on the shape recovery is emphasized since most existing models are unable to reproduce the effect of
the heating rate without using adjusted parameters. Here, the shape memory effect is modeled based only on the polymer
viscoelasticity and its time–temperature superposition. The model parameters are solely determined from testing the
material in dynamic mechanical analysis (DMA) and uniaxial tension at high temperatures. The parameters are easily
introduced in a standard finite element code (ABAQUS) to accurately and precisely simulate and predict the shape memory
behavior of the material submitted to various heating history. The simulation work highlights the crucial need for accurate
temperature measurements in order to precisely predict the shape recovery of such a material.
Keywords Shape memory polymer • SMP • Modeling • Thermomechanical behavior
12.1 Introduction
The multifunctionality that shape memory polymers (SMPs) can offer through the combination of their sensing and
actuating capabilities to industries spanning from the (bio)medical to aerospace areas is the main driver for the research
activity witnessed in the field. Yet, to date, SMP research remains primarily focused on material/chemical design for which
performances are compared via experimental monitoring of the materials’ physical properties and shape memory behavior.
Unfortunately, efficient predicting tools of such behavior that would also enable targeted product engineering and design are
still lacking. As a result, SMPs have yet to be implemented in applications with strict requirements and/or long term
applications. Moreover, real product design remains a challenge. Therefore, modeling the thermomechanical behavior of
SMPs becomes essential.
12.2 Experiments
12.2.1 Material
The epoxy network was synthesized by reacting a bisphenol-A-based epoxy monomer (Dow D.E.R. 383) purchased from the
Dow chemical company with a diamine crosslinker (Jeffamine D-230) from Huntsman company and a monoamine
(N-decylamine) from Aldrich. All chemicals were used as received. The monomer mix was injected in a glass mold sealed
D. Julie (*) • G. Pierre • F. Carole

Laboratoire PIMM, CNRS, Arts et Métiers ParisTech, Paris, France
e-mail: julie.diani@ensam.eu
A.R. Ingrid
Research and Development Center, General Motors Company, Warren, MI, USA
98 D. Julie et al.
Fig. 12.1 Torsion angle 360

recovery versus temperature 1.1 °C/min
for various heating ramps 3.2 °C/min
300
Torsion angle recovery (degree)

5.6 °C/min
240
180
120
60
0
35 40 45 50 55 60 65 70
Temperature (°C)
for curing. A two-step curing process was performed in an oven at 80 C for 2 h and 120 C for 2 h. Epoxy plates of 1.6 mm
thick were obtained from which the experimental samples were punched.
12.2.2 Tests
Material viscoelasticity and time–temperature superposition were obtained by DMA analysis performed on a RDA III
Rheometric Scientific® torsion rheometer at a constant strain of 0.2%. First, the material was submitted to a temperature
sweep at 1 Hz evidencing a tan d peak at 50 C characteristic of its glass transition temperature, Tg. Then, frequency sweeps
from 0.01 to 63 Hz were run at temperatures ranging from 40 C to 60 C with temperature increments of 5 C. The material
time–temperature superposition was shown to satisfy well to the WLF equation [1] and therefore was characterized by the
following parameters C1 ¼ 10.17 and C2 ¼ 47.35 C with Tref ¼ 50 C. For additional details on this characterization see [2].
Dumbbell samples were submitted to uniaxial tension using an Instron machine at a temperature above Tg to determine
the polymer’s elastic behavior. A Young modulus of approximately 7.2 MPa was measured.
Shape storage and recovery tests were performed in torsion using a house-built device [3]. The device allows performing
changes of shapes that involves small strain (6%) and large deformation which combination is more relevant to potential
industrial applications. The shape storage-shape recovery thermocycles were applied on a slender rectangular sample of
dimension 100 10 1.6 mm3. The material was heated above its glass transition and submitted to 360 rotation at high
temperature above Tg. Then, the material was cooled down while the torsion angle was maintained. Once at a temperature
below the glass transition, stresses were released while the torsion was naturally maintained by the material glassy state.
Finally, the torsion angle recovery was measured upon heating the sample. In order to bring to light the heating rate effect on
the material shape recovery, the deformed samples were submitted to three different heating ramps, 1.1, 3.2 and 5.6 C/min.
Figure 12.1 shows the torsion angle recovery with respect to temperature for the three heating ramps.
Moreover in order to highlight the effect of a temperature gradient along the sample during the shape recovery, one
sample was placed in a metal body made of thick walls opened at the bottom-end and on its two opposite sides. The air
convection within the body was observed to be poor creating a significant gradient of temperatures from the clamped top-end
of the sample to its free bottom-end. Figure 12.2 shows the samples placed inside the thermal chamber when submitted to
thermal shape recovery. On the left hand side, a dummy sample is used to record the temperature. In the center, a free-
hanging, twisted (deformed) sample is submitted to an homogeneous temperature upon heating. On the right hand side, a
sample within the metal body is submitted to a strong temperature gradient from top to bottom. The temperature gradient
encountered by a sample when incased in the metal body during heating is shown in Fig. 12.3 by the temperature
measurements recorded at the top, middle and bottom of the sample during a 5 C/min heating ramp. Let us note that the
temperature recorded at the bottom free-end of the sample is lower but close to that recorded on the dummy sample
(Fig. 12.2, left). The torsion angle recoveries during heating of both deformed samples were measured simultaneously to
highlight the effect of the temperature gradient on shape recovery. Results are plotted in Fig. 12.4.
In the next section, the intrinsic viscoelastic and time–temperature superposition properties of the material are used to run
original finite element simulations aimed at reproducing the above experimental results.
Fig. 12.2 Experimental set
up for measure of torsion
recovery upon sample heating
Fig. 12.3 Measurements 100

of temperatures in the thermal Thermal chamber
chamber (dummy sample) and 90 Bottom−end
within the metal body at the Center
bottom-end (open on the 80 Top−end
thermal chamber), at the
Temperature (°C)
center and at the top-end,

70
for an applied heating ramp
of 5 C/min
60
50
40
30
20
0 200 400 600 800 1000 1200
Time (s)
Fig. 12.4 Torsion recovery of 350

samples submitted to a heating Outside body
ramp of 5 C/min for a sample Inside body
300
placed in the metal body

(temperature gradient) and
a free-hanging sample (no 250
thermal gradient) (see samples
Fig. 12.2) 200
150
100
50
0
0 200 400 600 800 1000 1200
Time (s)
100 D. Julie et al.
12.3 Modeling
12.3.1 Constitutive Equations
The material linear viscoelasticity is modeled by a generalized Maxwell model,
X
n
ð2pf ti Þ2
G0 ¼ Ge þ Gi (12.1)
i¼1 1 þ ð2pf ti Þ2
The elastic modulus Ge is extracted from the values of G0 recorded at high temperatures. Calculating the pairs of
parameters (ti,Gi) is a well-known problem in rheology that may be solved using a Tikhonov regularization method. For this
purpose, we used the NLREG program disclosed by J. Weese [4]. The representation of the experimental data by the
generalized Maxwell model is shown in Fig. 12.5.
The bulk modulus (B) of the material is difficult to reach experimentally but it is required during 3D simulations,
therefore, the value of B ¼ 3.1 GPa was selected as it provided a Poisson ratio of 0.38 in the glassy state and of
approximately 0.5 in the rubbery state, both values being satisfactory for epoxy networks [5].
12.3.2 Simulations
Finite element simulations were run on Abaqus standard (2010) finite element code [6] using its existing material libraries.
First, the parameter of viscoelasticity characterized by the relaxation times and shear moduli pairs that define the master
curve (Fig. 12.5) are introduced using the Simo [7] finite strain extension of the generalized Maxwell model. This allows the
use of finite strain and large deformation. Second, the reference elastic behavior was represented by a neo-Hookean model
which parameter was chosen to fit our experimental data obtained in uniaxial tension up to 50% at high temperatures
(E ¼ 7.2 MPa).
Our parallelepiped sample was meshed by 2000 8-node linear brick hybrid elements. The upper end of the mesh was fixed
while the lower end was allowed to rotate. The entire thermomechanical cycle was simulated by applying the recorded
temperature history. For samples submitted to the same temperature along the sample (samples outside the metal body), the
same temperature was applied on all nodes of the mesh at every time. For samples placed inside the metal body (Fig. 12.2),
the sample was divided in three parts, bottom, middle and top, each with a different temperature history which was applied
on all nodes of each part. Figure 12.6 illustrates how the temperature is applied along the sample according to the sample
considered.
In the next section some selected results obtained from the above simulations are shown.
109
Experimental data
Generalized Maxwell approx.
Storage modulus (Pa)
108
107
Fig. 12.5 Approximation

of the viscoelastic
behavior of the epoxy 106
by a 12-relaxation time −6 −4 −2 0 2 4 6
generalized Maxwell model log10(frequency)
12 Role of Viscoelasticity in Predicting the Shape Memory Effect of Polymers 101
Fig. 12.6 Torsion sample

meshes with homogeneous
temperature applied on the
whole sample (top) and
gradient temperature along the
sample (bottom)
12.4 Results
12.4.1 Effect of the Heating Rate
The effect of the heating ramp on the sample recovery was shown in Fig. 12.1. The temperature histories were applied in the
finite element simulations as recorded on the dummy sample. In Fig. 12.7, the theoretical recoveries are plotted against
the experimental data. The simulations reproduce the experimental data well without any adjusting parameters.
As mentioned earlier, the model parameters entered in Abaqus were obtained from mechanical behavior characterization
tests. This result does not only support the largely spread idea that viscoelasticity coupled with time–temperature depen-
dence is the source of the shape storage/recovery property of crosslinked networks, it also provides the necessary modeling
tool for accurately predicting shape storage and recovery of these materials.
12.4.2 Effect of the Temperature Gradient
When the temperature is not homogeneous along or within the samples, the recovery cannot be plotted in terms of
temperature and has to be plotted in terms of time. In such a case the accuracy of the temperature measurements is crucial
to precisely reproduce the torsion recovery. To illustrate the latter assessment, simulations were run for the samples
submitted to homogeneous temperature fields (i.e., thermal chamber temperature) and to the temperature measured at the
bottom-end of the body (Fig. 12.3). The temperature fields applied and the simulation results are shown in Fig. 12.8.
Figure 12.9 (left) shows the torsion angle recovery with respect to time for both temperature histories. One notices a
relatively significant shift in the torsion angle recovery with respect to time. In fact, the duration to reach the temperature that
triggers the beginning of the sample shape recovery is shorter for temperature ramp T1. The torsion angle recoveries are also
plotted with respect to temperature (Fig. 12.9, right). One notices that the shape recoveries with respect to temperature are
similar for both temperature histories. According to the results shown in Fig. 12.8 and acknowledging the fact that both
heating histories have similar heating ramps, the latter result was expected.
The accuracy of the temperatures measured by the thermocouples is of 1 C. This accuracy was shown to be insufficient
to simulate the angle recovery versus time without observing a time shift between the simulation result and the experimental
data. Therefore, to illustrate the effect of a temperature gradient on the torsion angle recovery with respect to time, we have
compared theoretical results obtained by the finite simulations only. We have considered three samples submitted to
different temperature histories. One sample was submitted to the temperature gradient defined by the bottom-end, the center
and top-end temperature histories (Fig. 12.3), and two other samples were submitted to homogeneous temperature fields, one
102 D. Julie et al.
360
Data − 1.1 °C/min
Simulation − 1.1 °C/min
300 Data − 3.2 °C/min

Simulation − 3.2 °C/min
Data − 5.6 °C/min
240 Simulation − 5.6 °C/min
180
120
60
0
35 40 45 50 55 60 65 70
Temperature (°C)
Fig. 12.7 Theoretical estimates of torsion recovery versus temperature according to the sample heating rates
80
Abaqus simul. T1
Abaqus simul. T2
70 Measure themal chamber
Measure cage bottom end
Temperature (°C)
60
50
40
30
20
0 200 400 600 800 1000
Time (s)
Fig. 12.8 Temperature histories applied for recovery simulations
350 350
Simulation T1 Simulation T1
Simulation T2 Simulation T2
300 300
Recovery angle (degree)
Recovery angle (degree)
250 250
200 200
150 150
100 100
50 50
0 0
200 300 400 500 600 700 800 900 1000 20 30 40 50 60 70 80
Time (s) Temperature (°C)
Fig. 12.9 Torsion angle recovery versus time and versus temperature for the temperature histories shown in Fig. 12.8
12 Role of Viscoelasticity in Predicting the Shape Memory Effect of Polymers 103
Fig. 12.10 Applied 90

temperature fields along Bottom−end
the sample submitted to Mid−end
80
a temperature gradient Top−end
70
Temperature (°C)
60
50
40
30
20
0 200 400 600 800 1000 1200 1400
Time (s)
Fig. 12.11 Torsion angle 350 Homogen. bottom−end temp.

recovery versus time for Gradient temp.
the temperature histories Torsion angle recovery (degree)
300 Homogen. top−end temp.
displayed in Fig. 12.9
250
200
150
100
50
0
200 400 600 800 1000 1200 1400
Time (s)
corresponding to the sample’s bottom-end temperature history and the other to the sample’s top-end temperature history.
The applied temperature ramps are shown in Fig. 12.10. Figure 12.11 compares the torsion angle recovery. One notices two
main effects of the existence of a temperature gradient along the sample: First, the onset of the recovery is legitimately
affected by the temperature gradient, second, the torsion recovery rate depends on the three temperature ramps applied
simultaneously.
12.5 Conclusion
In this study, it was shown that the linear viscoelasticity and the time–temperature superposition rule the shape storage/
recovery of an amorphous polymer network under small strain and large deformation. The intrinsic material linear
viscoelasticity and time–temperature superposition were shown to be the only properties needed to quantitatively and
accurately predict the heating rate (i.e., temperature and time) dependent shape recovery.
The importance of recording the temperature precisely during the heat activated recovery step was demonstrated.
Both temperature shifts and temperature gradients were shown to significantly modify the material shape recovery with
respect to time.
104 D. Julie et al.
References
1. Williams ML, Landel RF, Ferry JD (1955) The temperature dependence of relaxation mechanisms in amorphous polymers and other glass-
forming liquids. J Am Chem Soc 77:3701–3707
2. Diani J, Gilormini P, Fredy C, Rousseau I (2012) Predicting thermal shape memory of crosslinked polymer networks from linear viscoelasticity.
Int J Solid Struct 49:793–799
3. Diani J, Fredy C, Gilormini P, Merckel Y, Régnier G, Rousseau I (2011) A torsion test for the study of the large deformation recovery of shape
memory polymers. Polym Test 30:335–341
4. Weese J (1993) A regularization method for nonlinear ill-posed problems. Comput Phys Commun 77:429–440
5. Smith A, Wilkinson SJ, Reynolds WN (1974) The elastic constant of some epoxy resin. J Mater Sci 9:547–550
6. Abaqus/Standard (2010) Dassault Systèmes Simulia Corp., Providence, RI, USA
7. Simo JC (1987) On a fully three-dimensional finite strain viscoelastic damage model: formulation and computational aspects. Comput Methods
Appl Mech Eng 60:153–173
Chapter 13
Damage Assessment in Syracuse Limestone Specimens
by Frequency Analysis of Elastic Emissions
A. Schiavi, G. Niccolini, A. Carpinteri, and G. Lacidogna
Abstract This paper investigates the mechanical behaviour and the damage evolution on four different size scale Syracuse
limestone specimens, under compressive loading through the frequency spectra analysis of elastic emissions (ELE).
ELE are detected in a very low frequency range (from 20 kHz down to few hertz) and are characterized by high levels of
released energy. In the damage process, approaching to the large fractures and the final collapse of the material, bursts
of ELE are observed. Amplitude, cumulate number, rate in time and frequency peak distribution of ELE indicate the
specimen degradation, from the integrity state to its irreversible plastic deformations.
This study focuses on the estimation of the ELE amplitude and frequency distribution in order to achieve informations
and analytical previsions about damage evolution both on the laboratory scale and in situ, such as Syracuse Athena Temple.
Keywords Acoustic emission • Low frequency • Fractal analysis
13.1 Introduction
The study of fracture mechanics and damage evolution by using the acoustic emission analysis in solids in compression is a
well established field of research [1–8]. In particular, several statistical analyses based, as an example, on rate, amplitude and
temporal distribution of acoustic emission events can be considered as effective tools to evaluate the mechanical properties
degradation [9–13].
Recently it has been observed that a selective analysis in frequency of the acoustic emissions allows to distinguish
micro-damage from macro-damage phenomena. In particular, the detection and analysis of Elastic Emissions (ELE), or
acoustic emissions at very low frequencies (between 20 kHz down to a few Hz, in which half-wave length of elastic waves is
greater than the maximum size of the sample), allow the behavior of materials in compression during the formation and
distribution of macro-cracks to be highlighted. The energy released from meso- and macro-fractures is able to stimulate both
the vibration natural modes of the specimen and the so called “quasi-rigid-body vibration”, due to dislocation of material
[14]. These vibrations and dislocations are detected on the specimen surface as ELE and can be considered as indicators of
incoming irreversible damage, namely the stage in which the diffuse microcracks in the bulk of the material coalesce to form
fracture surfaces, often anticipating the classic changes of slope (or jumps) on the stress–strain curve. Several aspects of
damage such as the typology of fracture (ductile, quasi-fragile and fragile) and damage evolution, such as the incipit of
irreversible plastic deformation and the energy released quantification, are also quantifiable through the quantitative and
qualitative analysis of ELE [15–18]. All these informations, both quantitative (based on data measured through accele-
rometers calibrated), and statistical, provide a very accurate evaluation of the mechanical properties and the compressive
behavior of solid materials.
In this work, lying in the framework of an advanced research field [19–23], both the time series of frequency peaks and a
fractal analysis of time/frequency distribution of ELE events, in order to investigate the temporal evolution of macro-cracks
A. Schiavi (*) • G. Niccolini

INRiM – National Institute of Metrological Research, Strada delle Cacce 91, Torino 10135, Italy
e-mail: a.schiavi@inrim.it
A. Carpinteri • G. Lacidogna
Department of Structural Engineering and Geotechnics, Politecnico di Torino, C.so Duca degli Abruzzi 24, Torino 10129, Italy
106 A. Schiavi et al.
on four specimens in Syracuse limestone are presented. The Syracuse limestone is of particular interest in the conservation
of national cultural heritage since it is the building material of the temple of Athena in the City of Siracusa (V Century b.C.).
Since the geographical area of Siracusa is particularly seismic is of interest determining the mechanical properties of this
building materials in order to provide useful and accurate information for the preservation of this structure through
appropriate monitoring.
13.2 Experimental Set Up and Materials
Four samples of Syracuse limestone of three different size have been subjected to compression tests. The test is performed in
displacement control at constant piston velocity of 0.5 mm/s, using a servo-hydraulic press with a maximum capacity of
500 kN, equipped with control electronics. The applied force is determined by measuring the pressure in the loading cylinder
by means of a transducer. A “delta shear” accelerometer, with the upper frequency limit of 20 kHz, for detection of ELE
events is used. The events are characterized by the output response of the calibrated transducers (charge sensitivity
0.33 pC/m·s2), expressed in mm/s2. The accelerometer transducer is rigidly fixed to the specimens in order to detect the
surface acceleration of the vibrations due to the fracture. The environmental background noise has been cut off using a
detection thresholds of 60 dB (referred to 1 mm/s2) and a pass-high filter in order to filter out any residual background
vibration under 1 kHz.
In order to define the field of investigation of the ELE is necessary to exclude all the elastic waves (longitudinal,
transverse and surface) whose half-wavelength is less than maximum size of the specimen dimensions. In this way only
vibration modes of the body and quasi-rigid-body vibration pulses are detected. The upper limit for the ELE frequencies is
defined as fELE 0.265 cL/dMAX, where cL is the longitudinal velocity wave in the material and dMAX is the maximum size of
the specimen (Table 13.1).
13.3 Data Analysis
Elastic Emission signals are detected during the compression tests. In Figs. 13.1, 13.2, 13.3, and 13.4 experimental data and
the time series of ELE frequency peaks united with the fractal analysis are reported. Blue points represent the ELE events as
a function of frequency, i.e. the corresponding maximum values of amplitude in the frequency spectrum of each single event;
purple points are the ELE event recorded after the main failure. The black line is the load versus time. The red line represent
the damage parameter , from which bt has been calculated united with the cumulative number of ELE until failure. Graphs
with red points represent the fractal analysis (in the time/frequency domain).
We analyze the time series of frequency peaks (in the frequency spectra between 1 kHz and fELE) of each single event and
the cumulative number N of ELE events as a damage parameter ¼ N/Ncrit, where Ncrit is the total number of ELE events up
to failure time tf. Both time series are correlated with the load history up to the specimen failure, which is revealed by the
peak load. Subsequent ELE events, corresponding to frictional modes, are disregarded in the present analysis.
The damage parameter can be also expressed as a function of time t:
bt
t
¼
tf
where the exponent bt can be obtained from the ELE data of the specimen.
Keeping constant the displacement rate during compression tests, it can be noted that specimens with increasing size
respond with accelerated damage accumulation before reaching the peak load. This effect is described by bt, ranging from
Table 13.1 Tested materials and features

Longitudinal
Specimen Material Density (kg/m3) Max size (m) wave speed (m/s) fELE (kHz)
1 Syracuse limestone 2,550 2.5 102 ~4,500 47.7
2 Syracuse limestone 2,550 5 102 ~4,500 23.9
3 Syracuse limestone 2,550 5 102 ~4,500 23.9
4 Syracuse limestone 2,550 101 ~4,500 11.9
Fig. 13.1 Syracuse limestone 1 (dMAX ¼ 2.5 cm)
Fig. 13.2 Syracuse limestone B (dMAX ¼ 5 cm)
Fig. 13.3 Syracuse limestone C (dMAX ¼ 5 cm)

108 A. Schiavi et al.
Fig. 13.4 Syracuse limestone D (dMAX ¼ 10 cm)
low values (bt ¼ 0.47) for decreasing damage accumulation rate in the smallest size (specimen 1), up to high values
(bt ¼ 1.50) for the largest size (specimen 4), describing sudden damage accumulation just before the specimen failure.
We wish to describe the organization of the ELE events, by estimating the fractal dimension of the set formed by their
representative points in the time-frequency domain. The generic ith ELE event will be described by its occurrence time ti and
the dominant (peak) frequency fi of its spectrum, and thus it can be represented by a point xi ¼ (ti, fi) in the time-frequency
domain. As suggested by Grassberger and Procaccia, we measure the distance between every pair of points and then
compute the correlation integral:
2 X
N1 X N
CðdÞ ¼ H d xk xj
N ðN 1Þ k¼1 j¼kþ1
Where H(x) is the Heaviside

step function, and N is the number points. The summation counts the number of pairs (xk, xj) for
which the distance xk xj is less than d.
For a fractal set with dimension D, the correlation integral scales like dD for small d. To find this D, Grassberger and
Procaccia prescribe that a log-log plot of C(d) versus d be constructed and that the dimension be read off as the slope of the
rectilinear part of the curve.
At the beginning of the tests on specimens 1, 2 and 3, the patterns of frequency peaks form straight lines in the time-
frequency domain. This evidence suggests that well-defined vibrational modes hold in the specimen experiencing damage,
as revealed by the ELE phenomenon, but still structurally intact. Furthermore, the frequency increase in time of each mode is
correlated with the increasing stress acting on the specimen.
We can observe also the role of failure precursor played by the ELE, whose onset precedes any observable load drop (see
Figs. 13.2 and 13.3).
As shown Fig. 13.1, in the specimen 1 three vibrational modes (~1.5, ~4 and ~7.5 kHz), increasing in time until to about
500 s and followed by a chaotic distribution of peak frequencies, can be clearly observed. The transition from ordered to
more chaotic behaviour can be captured calculating the fractal dimension of the two sets. We obtain D ¼ 0.84 for
t < t1 ¼ 500 s, describing a set of points which tend to organize along lines, and D ¼ 1.18 for t > t1, describing a set of
points trying to fill the time-frequency plane.
A similar behaviour can be observed for the specimen 2, where at t1 ¼ 60 s a ~5 kHz- single mode (D ¼ 0.78) split in
two, illustrating a more chaotic behaviour (D ¼ 1.10) in the time-frequency plane. For the specimen 3 (two vibrational
modes at ~5.5 kHz and ~11 kHz for t < t1 ¼ 85 s) we observe the transition from an organized distribution (D ¼ 1) to an
extremely chaotic distribution without a unique fractal dimension (see Fig. 13.3). Therefore, in all cases the transition from
organized to chaotic distribution of vibrational modes in the time-frequency domain is a precursor of the specimen failure.
13 Damage Assessment in Syracuse Limestone Specimens by Frequency Analysis. . . 109
In the sample 4, the points are randomly distributed over the time-frequency domain through all the test duration.
The complete lack of organization in vibrational modes for ELE events may be signature of pre-existing damage for the
specimen.
13.4 Conclusion
In this work both the time series of frequency peaks and a fractal analysis of time/frequency distribution of ELE events have
been proposed. The time series of frequency peaks and the cumulative number N of ELE events allow to define the damage
parameter ¼ N/Ncrit. The fractal analysis (in the time/frequency domain) allow to focus the transition from order to
disorder of the vibrational mode of the specimen during the formation of macro-crack and fracture surface. The transition
from organized to chaotic distribution of vibrational modes in the time-frequency domain is a precursor of the specimen
failure.
The principal aim of this work is focused on finding a proper tool for damage monitoring in real time in particular for
preserving national cultural heritage.
References
1. Anzani A, Binda L, Carpinteri A, Lacidogna G, Manuello A (2008) Evaluation of the repair on multiple leaf stone masonry by acoustic
emission. Mater Struct (RILEM) 41:1169–1189
2. Scholz CH (1968) Microfracturing and the inelastic deformation of rock in compression. J Geophys Res 73(4):1417–1432
3. Ohnaka M, Mogi K (1982) Frequency characteristics of acoustic emissions in rocks under uniaxial compression and its relation to the
fracturing process to failure. J Geophys Res 87(B5):3873–3884
4. Khair AW (1984) Acoustic emission pattern: an indicator of mode of failure in geologic materials as affected by their natural imperfections. In:
Hardy HR, Jr., Leighton FW (eds) Proceedings of the 3rd conference on acoustic emission/microseismic activity in geologic structures and
materials, 1981. University Park, PA. Trans Tech Publications, Clausthal, pp 45–66
5. Lockner DA, Byerlee JD, Kuksenko V, Ponomarev A, Sidorin A (1991) Quasi-static fault growth and shear fracture energy in granite. Nature
350:39–42
6. Lockner D (1993) The role of acoustic emissions in the study of rock fracture. Int J Rock Mech Min Sci Geomech Abstr 7:883–889
7. Ohtsu M (1996) The history and development of acoustic emission in concrete engineering. Mag Concr Res 48:321–330
8. Eberhardt E, Stead D, Stimpson B, Read RS (1998) Identifying crack initiation and propagation thresholds in brittle rock. J Can Geotech
35:222–233
9. Colombo S, Main IG, Forde MC (2003) Assessing damage of Reinforced Concrete Beam using “b-value” Analysis of Acoustic Emission
signals. J Mater Civil Eng (ASCE) 15:280–286
10. Rao MVMS, Lakschmi PKJ (2005) Analysis of b-value and improved b-value of acoustic emissions accompanying rock fracture. Curr Sci
(Bangalore) 89:1577–1582
11. Rundle JB, Turcotte DL, Shcherbakov R, Klein W, Sammis C (2003) Statistical physics approach to understanding the multiscale dynamics of
earthquake fault systems. Rev Geophys 41:1–30
12. Carpinteri A (1994) Scaling Laws and renormalization groups for strength and toughness of disordered materials. Int J Solids Struct
31:291–302
13. Shiotani T, Yuyama S, Li ZW, Othsu M (2000) Quantitative evaluation of fracture process in concrete by the use of improved b-value. 5th
international symposium non-destructive testing in civil engineering, Eds. Elsevier Science, Amsterdam, pp 293–302
14. Niccolini G, Schiavi A, Tarizzo P, Carpinteri A, Lacidogna G, Manuello A (2010) Scaling in temporal occurrence of quasi-rigid body vibration
pulses due to macrofractures. Phys Rev E 82:046115/1–046115/5
15. Schiavi A, Niccolini G, Tarizzo P, Lacidogna G, Manuello A, Carpinteri A (2009) Analysis of acoustic emissions at low frequency in brittle
material under compression. In: Proceedings of the SEM annual conference 2009, Albuquerque
16. Schiavi A, Niccolini G, Tarizzo P, Carpinteri A, Lacidogna G, Manuello A (2011) Acoustic emissions at high and low frequencies during
compression tests in brittle materials. Strain 47(s2):105–110
17. Schiavi A, Niccolini G, Tarizzo P, Lacidogna G, Manuello A, Carpinteri A (2011) Analysis of energy released by elastic emission in brittle
material under compression. In: Experimental Mechanics on Emerging Energy Systems and Materials 5:103–108
18. Schiavi A, Niccolini G, Tarizzo P, Carpinteri A, Lacidogna G, Manuello A (2011) Waveforms and frequency spectra of elastic emissions due
to macrofracture in solids. In: Experimental and Applied Mechanics 6:609–617
19. Carpinteri A, Lacidogna G (2006) Damage monitoring of an historical masonry building by the acoustic emission technique. Mater Struct
39:161–167
20. Carpinteri A, Lacidogna G (2006) Structural monitoring and integrity assessment of medieval towers. J Struct Eng (ASCE) 132:1681–1690
21. Carpinteri A, Lacidogna G (2007) Damage evaluation of three masonry towers by acoustic emission. Eng Struct 29:1569–1579
22. Carpinteri A, Invernizzi S, Lacidogna G, Manuello A, Binda L (2008) Numerical simulation and monitoring of the Cathedral of Siracusa in
Sicily. In: Proceedings of the 6th international conference on structural analysis of historical construction, Bath, 2–4 July 2008
23. Niccolini G, Carpinteri A, Lacidogna G, Manuello A (2011) Acoustic emission monitoring of the Syracuse Athena temple: scale invariance in
the timing of ruptures. Phys Rev Lett 106:108503/1–108503/4
Chapter 14
Photogrammetry Measurements During a Tanking Test
on the Space Shuttle External Tank, ET-137
Justin Littell, Tim Schmidt, John Tyson, Stan Oliver, Matt Melis, Charles Ruggeri, and Duane Revilock
Abstract On November 5, 2010, a significant foam liberation threat was observed as the Space Shuttle STS-133 launch
effort was scrubbed because of a hydrogen leak at the ground umbilical carrier plate. Further investigation revealed the
presence of multiple cracks at the tops of stringers in the intertank region of the Space Shuttle External Tank. As part of an
instrumented tanking test conducted on December 17, 2010, a three dimensional digital image correlation photogrammetry
system was used to measure radial deflections and overall deformations of a section of the intertank region.
This paper will describe the experimental challenges that were overcome in order to implement the photogrammetry
measurements for the tanking test in support of STS-133. The technique consisted of configuring and installing two pairs
of custom stereo camera bars containing calibrated cameras on the 215-ft level of the fixed service structure of Launch
Pad 39-A. The cameras were remotely operated from the Launch Control Center 3.5 miles away during the 8 h duration test,
which began before sunrise and lasted through sunset.
The complete deformation time history was successfully computed from the acquired images and would prove to play a crucial
role in the computer modeling validation efforts supporting the successful completion of the root cause analysis of the cracked
stringer problem by the Space Shuttle Program. The resulting data generated included full field fringe plots, data extraction time
history analysis, section line spatial analyses and differential stringer peak-valley motion. Some of the sample results are included
with discussion. The resulting data showed that new stringer crack formation did not occur for the panel examined, and that large
amounts of displacement in the external tank occurred because of the loads derived from its filling. The measurements acquired
were also used to validate computer modeling efforts completed by NASA Marshall Space Flight Center (MSFC).
Keywords Photogrammetry • Space shuttle • Digital image correlation • Stringer crack • STS-133
J. Littell (*)
ATK Space Systems, NASA Langley Research Center, 12 W. Bush Rd MS 495, Hampton, VA 23681, USA
e-mail: Justin.D.Littell@nasa.gov
T. Schmidt • J. Tyson
Trilion Quality Systems, 500 Davis Drive St 200, Plymouth Meeting, PA 19462, USA
e-mail: Schmidt@trilion.com; Tyson@trilion.com
S. Oliver
EV31, NASA Marshall Space Flight Center, Huntsville, AL 35812, USA
e-mail: Stanley.T.Oliver@nasa.gov
M. Melis • C. Ruggeri • D. Revilock
NASA Glenn Research Center, 21000 Brookpark Road MS 49-8, Cleveland, OH 44135, USA
e-mail: Matthew.E.Melis@nasa.gov; Charles.R.Ruggeri@nasa.gov; Revilock@nasa.gov
112 J. Littell et al.
14.1 Background
Space Transportation System 133 (STS-133) was the 133rd mission in the Space Shuttle Program and the final mission for
the Discovery orbiter. Discovery carried a Multipurpose Logistics Module (MPLM) “Leonardo”, and an ExPRESS1 logistics
carrier (ELC) to the International Space Station [1]. STS-133 was scheduled to lift off from NASA Kennedy Space
Center (KSC), Launch Pad 39-A, on September 16, 2010. A series of scrubs pushed the launch back to November 5,
2010. On November 5, 2010, the launch was scrubbed because of leaks in the ground umbilical carrier plate (GUCP), which
occurred during the filling process of the external tank (ET), ET-137. During the post-scrub walk down inspection, cracks
were discovered in the thermal protection system (TPS) foam insulating the flange connecting the liquid oxygen (LO2) tank
and the intertank near the top of the ET on Panel 2. These cracks were unrelated to the GUCP leak issue.
The intertank is an unpressurized cylinder which connects the liquid hydrogen tank (LH2) to the LO2 tank. It connects the
two tanks via bolting flange rings in both its forward and aft locations. It is comprised of eight panels, six of which are thin
skins with bolted on stiffeners. In some locations, additional panels, called doublers, are added to the thin skin to provide
additional support. The setup of the ET can be seen in Fig. 14.1, along with the location where the TPS cracks developed.
Closer inspections indicated that the TPS cracks initiated from an area where the riveted hat stiffeners (stringers) mounted
to the ET Intertank, providing additional structural support to the Intertank near the LO2/intertank interface. Inspections of
the stringers necessitated cutting the TPS foam away and revealed that cracks were present in more of the aluminum
stringers in regions outside the one highlighted in Fig. 14.1. The issue was mitigated by removing the cracked portions of the
original stringers and replacing with new stringers and z-doublers. The remaining un-cracked stringers received a small
reinforcement on their feet [2]. New TPS foam was reapplied to the repaired region to provide insulation for the ET.
Figure 14.2 shows a final repaired Panel 2 region.
The Space Shuttle Program concluded that a comprehensive root cause analysis for the stringer cracks would be
necessary, prior to flying the STS-133 mission, to insure flight safety for the vehicle. This investigation would require
months of testing and analysis, and would include a tanking test of the external tank, ET-137, with an extensive array of
strain gages and thermocouples added in an attempt to characterize the thermal and deformation state of the tank under pre-
launch loads. The tanking test, lasting nearly 8 h, effectively simulated all of the procedural steps in the launch countdown
process including a complete fill, pressurization and drain with over 1.5 million pounds of the liquid oxygen and liquid
hydrogen propellants.
The External Tank Project Office inquired into the feasibility of using stereo digital image correlation techniques to
capture full-field deformations of the tank external surface at two separate locations during the tanking test, a measurement
Fig. 14.1 Cracks in the ET

foam at the LO2 flange
on Panel 2 due to stringer
cracking
1
Expedite the processing of experiments to the Space Station.
The use of trademarks or names of manufacturers in this report is for accurate reporting and does not constitute an official endorsement, either
expressed or implied, of such products or manufacturers by the National Aeronautics and Space Administration.
14 Photogrammetry Measurements During a Tanking Test on the Space. . . 113
Fig. 14.2 A repaired region

of the LO2 Flange on ET
Panel 2
which has never been attempted previously. As a background, stereo digital image correlation is the acquisition of
engineering data from digital stereo image pairs for the determination of object shape and deformation. The typical
procedure involves the application of a dot or speckle pattern onto an object and tracking the motion of the applied pattern
as loading occurs. For three dimensional digital image correlation, at least two cameras are needed such that the third, out-of-
plane direction can be triangulated between camera locations. The digital image correlation that was used for the tanking test
was a commercial system which had previously been used to validate LS-DYNA models of foam impact on the Orbiter wing
reinforced carbon-carbon panels during the Columbia Return to Flight program [3–5]. Three dimensional digital image
correlation has been successfully used to capture motion in a wide range of research areas from material characterization
tests [6–9] to structural tests [10–12], to biomechanics [14–16].
The area of interest at both of these locations would include stringers and the flange structure joining the intertank to the
LO2 tank. A rapid assessment from a NASA photogrammetry team and industry consultants established that it was
reasonable to use digital image correlation techniques for the test, given enough time to assemble and set up the equipment.
As a consequence, the photogrammetry team was assembled and tasked to move forward supporting the tanking test.
14.2 Test Methodology
After a walk down of Launch Pad 39-A conducted on December 6, 2010, photogrammetry engineers decided on a plan of
action for two independent views requiring two camera systems. The first would focus on Panel 2 on the repaired region,
while the second would focus on Panel 6, a similar, non-repaired panel that would serve as a baseline.
14.2.1 Final Camera Positions
The Panel 6 camera pair was mounted on the railing of the GUCP walkway located on the 215-ft level of the fixed service
structure (FSS) and aimed at ET Panel 6, focusing on the region near the LO2 flange. This particular region was chosen
because it allowed technician and engineer access to both the cameras and the ET via the GUCP walkway, and also because
Panel 6 was structurally similar to Panel 2 in terms of geometry, including similar skin thickness and doubler regions. A pair of
Vision Research Phantom v10.1 4 megapixel (MP) cameras was attached onto a 13 ft long, vertically mounted 6-in. 6-in.
extruded aluminum beam at a base separation of approximately 7 ft. They were enclosed in nitrogen gas purged bags to meet
the safety requirements and mitigate the risk for a spark, fire or explosion during the test. The aluminum beam was covered in
a thermal protection blanket to minimize thermal expansion due to temperature change throughout the day. The thermal
expansion of the beam was a major concern for the integrity of the calibration of the cameras. This issue will be discussed
further in the Calibration Procedures section. Figure 14.3 shows the Panel 6 test setup.
Fig. 14.3 Details and camera setup for ET Panel 6
Fig. 14.4 Details and camera setup for ET Panel 2
A second camera pair was constrained to be used on the repaired region on ET Panel 2. The first option was to use the
removable service structure (RSS) for the mounting of the camera pair because the RSS was very close to the Panel 2 region
while also having ample room to setup and operate the camera equipment. However, the tanking test required that the RSS
be in the open position in order to accurately simulate the launch countdown, which made it impractical to use. With the
removable service structure rotated into the open position and therefore unusable, the closest available mounting location
was the walkway which connected the FSS to the RSS. This walkway was located approximately 60 ft away from the ET.
The 60-ft distance constrained the camera base separation to be between 20 and 30 ft. As such, Baumer 5MP cameras were
mounted on a custom 20-ft long, 6-in diameter woven composite hollow beam. A composite beam was chosen for the beam
as alternative aluminum or steel beams were rejected because of their large weight and possible large expansion properties
under changing weather conditions. As with the Panel 6 cameras, the Panel 2 cameras were enclosed in nitrogen purged gas
bags. Figure 14.4 shows the setup for ET Panel 2.
Fig. 14.5 Calibration object
14.2.2 Calibration Procedures
In photogrammetry techniques, calibrations of the cameras are needed prior to the acquisition of the image data. Calibration
procedures are similar for all types of photogrammetric imaging, which identify and correct for camera lens distortion,
establish absolute scaling, and determine the relative positions of each camera in space or in relation to the other cameras.
The calibration procedure also establishes a virtual box, inside of which all measurements acquired are considered
valid. Calibrations for both sets of cameras were performed on the base of Launch Pad 39-A prior to hoisting the stereo
camera rigs to the FSS. The calibration procedure required that each camera pair acquire images of a calibration object
through a sequence of rotations, angles, and, finally, repositions at various distances ranging from 10 ft to the camera pair.
The calibration object was a large 2 m cross covered with patterns capable of being recognized by the image correlation
calibration software. Figure 14.5 shows the calibration object, mounted on a tripod on the ground of Launch Pad 39-A.
Ground calibration was chosen because many of the repositions and rotations needed during the calibration procedure
would be inaccessible or unsafe if attempted at each pair’s final location on the FSS. However, the calibration integrity could
be compromised if either the hoisting process or the thermal expansion of the camera beam caused large relative movement
and/or change in angle within each pair of cameras. These risks were deemed acceptable and outweighed the risks of the
alternative calibration procedure in the camera final mounted position.
14.2.3 Speckle Application Methodology
A dot speckle pattern was stenciled on the ET Panel 2 and Panel 6 locations by orbiter technicians with approved black
epoxy marker paint. The dots were nominal 3/8-in diameter, and they were painted between stringers 5 and 14 on Panel 6 and
over the repaired region for Panel 2. For Panel 6, the pattern extended from just above the LO2 flange, down onto the
intertank structure, and approximated a T shape. For Panel 2, the pattern resembled a square painted over the repaired area
shown in Fig. 14.2. Figure 14.6 shows a close up view of the application of the dot speckle pattern on Panel 6.
14.2.4 Image Acquisition Procedures
Both camera pairs were controlled from the Launch Control Complex (LCC) Firing Room 3, 3.5 miles away, via
underground fiber optic cable. The Panel 2 cameras were connected into and directly controlled by the photogram-
metry rack-mounted computer, which physically stayed on the FSS walkway during the tanking test. The photogrammetry
Fig. 14.6 Speckle pattern on ET Panel 6
Fig. 14.7 Photogrammetry setup schematic
rack-mounted computer was controlled in the LCC by a remote desktop connection. Each individual Panel 6 Phantom
camera was controlled via its own dedicated computer in the LCC. A camera trigger signal originated from the photogram-
metry rack computer and was sent to all cameras, and a common IRIG time code originating from LCC was input into each
camera for synchronization purposes. The results were pairs of synchronized uncompressed black and white TIFF images. A
schematic of the setup is shown in Fig. 14.7.
14.3 Results
There were 1,637 images, spanning the entire duration of the test, successfully acquired from each camera and used for
photogrammetric analysis. The tanking test started at approximately 7:19 am KSC local time and lasted approximately 7.5 h
(from first fill to final drain). Image acquisition ended one-half hour after the 5% tank sensor indicated dry conditions were
present. The extra half hour was included to ensure that the ET was empty of propellant and returned back to its pre test,
empty state. The general acquisition rate was every 10 s during the fill, every 3 s during the pre-pressurization (prepress), and
every 60 s during the drain. A test schedule highlighting significant test events is shown in Table 14.1.
All results were examined post-tanking test. When examining the results post-test, it was noted that Panel 6 data were of
much higher fidelity than Panel 2 results and will therefore be used as the example for the data analysis. Figure 14.8 shows
sample images captured from the left and right camera of Panel 6. The calibrated region is much larger than the patterned
region based on the original goal of measuring a significant portion of the length of the intertank. However, the patterned
region was significantly reduced due to flight safety concerns.
The data were first used in a qualitative surface analysis examination. The surface analysis was a three dimensional
surface visualization feature. It was used as a quick look to determine whether any anomalies (i.e., stringer cracks) occurred
on the ET surface. All nominal conditions were detected from the surface analysis. Several features including the radial
curvature of the ET, the peaks and valleys of the stringers, and the LO2 flange were highly noticeable. Figure 14.9 shows the
surface analysis of the ET. The horizontal lines seen in Fig. 14.9 will be addressed later in this section.
Quantitative measurements were next examined. The data density contained over nine data points per square inch, with
over 20,000 virtual strain gages available from which to make the quantitative measurements. A coordinate system was
created such that the +x direction was defined as sideways to the right in the images, representing the ET hoop direction, and
Table 14.1 Significant events IRIG/EST

noted in the tanking test Image # GMT (local KSC time) Event
0 12:19:38 7:19:38 Test begin
112 12:38:27 7:38:27 LO2 slow fill began
202 12:53:50 7:53:50 LO2 5% sensor wet
313 13:12:34 8:12:34 Equivalent time of stringer crack
900 14:52:25 9:52:25 LO2 fast fill ended
1,245 19:18:35 2:18:35 LO2 pre-pressurization (prepress) began
1,306 19:24:44 2:24:44 LO2 prepress ended
1,465 19:51:49 2:51:49 LO2 drain began
1,607 21:21:05 4:21:05 LO2 5% sensor dry
1,637 21:51:31 4:51:31 End of data acquisition
Fig. 14.8 Panel 6

stereo camera view
Fig. 14.9 Surface analysis visualization
Fig. 14.10 Full field fringe plot (left) extracted points for time history analysis (right)
the +y direction was defined as vertical toward the top of the ET, representing the axial direction. Finally, the +z direction
was defined as outward from the tank, representing the radial direction.
Figure 14.10 shows two ways of extracting the data from the full field measurements. The left image shows a fringe plot
of the unmodified full field radial (z) displacement just at the completion of fill (image 900). The change in radial
displacement is very noticeable and indicates significant movement in the ET, with an inward motion of more than 0.5 in.
at the top of the flange. Gaps or holes in the data indicate portions of the pattern that were not tracked and were mainly due to
the discontinuity of the dot pattern resulting from the abrupt change in direction between the stringer tops and valleys. From
the fringe plots, particular places of interest were identified for further time history analysis. Figure 14.10, right, shows
locations of four discrete points identified for further analysis.
An interrogation of the time history was performed on the point labeled in Fig. 14.10 as “Flange above stringer S6–6”,
which is located near the upper left of the pattern. Figure 14.11 shows the time histories of the three dimensional motion at
this point. Note that “disp e” represents the root-sum-square of the motion in three directions at this location. Significant
event milestones are labeled above the data for reference.
Fig. 14.11 Time history of displacements at a point on the left flange
The time history shows that between the LO2 slow-fill milestone and the completed LO2 fast-fill milestone, the point on
the left flange exhibited almost 900 miles of vertical displacement in the y (downward) direction, which is represented by
the plateau in this line in Fig. 14.11. This displacement disappears at the milestone when the LO2 drain began, and the point
reaches almost the starting position after the LO2 sensor dry milestone. This finding is due to the sag in the ET from the
weight of the propellants. Also, the radial displacement in Fig. 14.11, is negative, indicating contraction due to the super-
cooled liquid inside the ET. Finally, the horizontal displacement in Fig. 14.11, is positive, indicating displacement toward
the right of the image.
An examination into the relative deformation between stringer peaks and valleys was next conducted. The differential
motion function was used, with two endpoints that were located on a stringer peak and a stringer valley. The differen-
tial motion of a valley between stringers 11–12 (S11–12) and stringer 13 (S13) were compared in Fig. 14.12.
The plot in Fig. 14.12, right, shows the vertical difference between the stringer peak and valley. A large abrupt spike
represented by a discontinuity in the data would indicate a possible crack formation, as a crack would show significant
shearing motion between the peak and the valley. In contrast, the gradual changes seen over the course of the first 4,000 s of
the test were determined to be normal ET motion. Similar analyses were conducted for the other stringer locations. The
analysis of this data did not show evidence of any newly developed cracks that might have occurred during the tanking test.
Spatial trends in the data were next extracted using the section line function. A section line represented the data as a
function of position for a discrete time, rather than plotting as a function of time. The section line in Fig. 14.13 extended
vertically between the top and bottom of the speckle region in a valley between stringers 8 and 9.
The data were acquired for significant times of interest that occurred throughout the test, mainly those that are referenced
in Table 14.1. Relative motion between these significant times was determined by the difference in the sequential section
line data. For example, the difference in radial displacement between LO2 Sensor Wet (image 202 data) and LO2 Slow Fill
Begin (image 112 data) was found by subtracting the data set from image 112 from the data set in image 202. The result is
shown as the line labeled “202-112” in Fig. 14.14, below. Rigid body motion, otherwise known as motion that occurred on
the entire ET, resulted in the lines showing a flat, horizontal trend. However, a large slope or a large difference as a function
of position would indicate relative deformation between the two event times. In theory, the “flatness” of the line determined
the type of motion. Figure 14.14 shows a differential section line analysis.
Figure 14.14 suggests that all events up through the filling of the ET (image 900) cause actual deformation, which is noted
by the slope increase in the three data sets leading up to image 900. Data set “202-112” shown with small diamonds in
Fig. 14.14, starts off as a flat line indicating rigid body motion. However data set “313-202” shown with squares in
Fig. 14.14, shows a significant change in radial (z) displacement throughout the section length, indicating relative motion.
Fig. 14.12 Analysis for stringer crack determination
Fig. 14.13 An extracted section line for the valley between stringers 8 and 9
Data set “900-313”, shown with triangles in Fig. 14.14, shows a similar trend. The difference between LO2 fast fill and
LO2 pre-pressurization (images 1,245 and 900, or time 25,559 and 9,590 s) seems to cause rigid body motion only.
Similar phenomena can be seen between the difference of the LO2 drain (image 1,465) and the end of the LO2 prepress
(image 1306). The remaining curves show some form of sloped line and most likely indicate structural deformations of the
ET. The minor plateau around the 5 in. mark in the data represents a false reading in the image 1,465 data due to data
dropout.
A complete analysis which includes further section line extraction and data analysis was performed by NASA Marshall
Space Flight Center (MSFC). Because stringer cracks were not seen in the Panel 6 data, the majority of the data was used to
compare and validate computer modeling efforts on the ET performance. Figure 14.15 shows a comparison of photogram-
metry data, shown in solid lines, to finite element analysis predictions for flange radial deflection and rotation angle. For a
complete analysis on the test-simulation correlation, refer to Ref. [17].
Fig. 14.14 Relative change between significant events for the S8–S9 valley section line
Fig. 14.15 Comparison of photogrammetry data to finite element analysis
14.4 Summary
On December 17, 2010, three dimensional digital image correlation techniques were successfully used as a part of a Space
Shuttle tanking test to help determine root causes for stringer cracks seen on ET-137 in support of the STS-133 mission. The
setup and challenges, along with the final methodology used and some sample results are presented.
Photogrammetric engineers had 10 days from the first definition of the problem to the day of the tanking test. In this
span of time, they had to formulate a plan of action, conduct a site visit to familiarize themselves with the Space Shuttle
and Launch Pad 39-A launch hardware, scout suitable locations for camera placement, order and fabricate materials and
components, ship equipment, organize support, calibrate equipment, mount equipment and acquire test images. All of these
items were completed on a non-interference basis, without compromising the safety of personnel and hardware or the
integrity of the data. It was the first time that stereo photogrammetric imaging equipment had been used on the launch pad
and the first time that measurements had been acquired on the ET during a tanking test.
During the tanking test, photogrammetric engineers were located in the Launch Control Center’s Firing Room 3, 3.5
miles away from Launch Pad 39-A. They were remotely controlling cameras housed in explosion proof housings mounted
on custom fabricated camera bars located on the 215-ft level of FSS at Launch Pad 39-A. Approximately 1,600 images were
successfully acquired during the 8 h test.
From the measurements, the Panel 6 detailed analyses provided significant insight into tank behavior. The images were
computed post-test and the resulting data was in the form of full field fringe plots, data extraction time history analysis,
section line spatial analyses and differential peak-valley motion. The results show clear deformation in the ET during the
test, especially during first fill. The measurements acquired also provided an excellent opportunity to validate computer
modeling efforts. The data was used in a complete data analysis and computer modeling effort by MSFC and yielded
excellent results which validated the stringer finite element analysis.
ET-137 carried Discovery for the STS-133 mission to orbit on February 24, 2011.
Acknowledgments Successful completion of these measurements relied on major contributions from dozens of individuals. The authors were
very fortunate to be assigned Wulf Eckruth of United Space Alliance as our primary liaison, and were pleased that he was awarded a NASA Silver
Snoopy in recognition of his critical efforts. Matthew Van Den Driessche and Gabor Tanacs also provided on-the-fly design engineering
assistance. Mike Payne and Jack Strieter (NASA KSC) provided very strong overall engineering supervision and guidance. Mark Olszewski
and Robert Stratton of Analex gave photographic equipment and advice on numerous occasions.
Patterning the external tank was a major task in and of itself. Eugene Sweet of Lockheed Martin was the team lead for this effort. Kerry Chreist
produced the final design of the dot pattern. JJ Laplante, Tony Marketon, James O’Bryhim and John Hubinger applied the pattern, working on an
elevated platform in cold weather until after midnight. Ed Carillion, Scotty Dornton and Timmy Wirth were there to mix the epoxy paint.
The High Crew and Heavy Equipment teams provided extremely strong support numerous times during this effort. High Crew produced
custom work and support platforms remarkably quickly. The group included James Gibson, David Schuermann, Tony Pope, Thomas Blabey,
Brian Amick, Donald Redfern, Dennis Elmore, Michael Rajnish, Joel Smith, William Evans, Robert Hanover, John Jones, and Mark Kobilis.
Heavy Equipment operated trucks and cranes while also demonstrating their own lift capacity in situations where there was only one chance to do
the job correctly. Members included Paul Kruse, Wally Majid, Susan Thai, Kevin Hunt, Roger Spillers, David Parker, Andrew Shepard, Philip
Reno, James Speigner, David Kobilis, William Mayhew, J. Rhode, Hiawatha Brown, Peter Collins, W. King and Gary Osbourne.
The Launch Equipment Shop was called on for rapid production of many custom components. Gino Digiovanni, Chuck Gardener, Carson
Yates, Oogie Townsend, John Franasiak, and John McCauley got the job done effectively and efficiently over multiple shifts.
Dave Sutton, Mike Beville, Abdoul Alivandi and William Sands quickly designed and installed custom nitrogen purge equipment for the
cameras to meet safety requirements, without which, installation on Launch Pad 39-A’s fixed service structure would not have been permitted.
Tim Wright, Jim Blair and Bruce Hodge were responsible for timing and countdown aspects of the test, including camera communication and
triggering support. Gus Alex provided image engineering support.
Brenda Blackmon and Kim Phillips appeared at the launch pad late one evening to install custom thermal blankets on the camera beams, which
they had designed and fabricated on short notice. It is unlikely that the sensors would have remained calibrated without this insulation.
The entire effort was heavily documented photographically by IMCS photo and media services. Technical photographers included Tim Terry,
Robert Murray, Kenneth Allen, Kevin O’Connell, George Roberts, and Charles Robinson.
References
1. NASA (2012) http://www.nasa.gov/mission_pages/shuttle/shuttlemissions/sts133/main/index.html. Accessed 10 Jan 2012

2. Harwood W (2011)NASA zeros in on root cause of shuttle tank cracks. http://spaceflightnow.com/shuttle/sts133/110111rootcause/. Accessed
9 May 2011
3. GOM Optical Measurement Techniques (2009) ARAMIS user manual – software v 6.1, 20 April 2009
4. Schmidt T et al (2005) Performance verification of 3D image correlation using digital high-speed cameras. In: Proceedings of 2005 SEM
annual conference, Portland, 7–9 June 2005
5. Melis M et al (2009) Impact testing on reinforced carbon-carbon flat panels with BX-265 and PDL 1034 external tank foam for the space
shuttle return to flight program. NASA TM 2009-213642/REV1, 2009
6. Ambur DR et al (2004) Progressive failure studies of stiffened panels subject to shear loading. Compos Struct 65:129–142
7. Littell JD et al (2008) Measurement of epoxy resin tension, compression and shear stress strain curves over a wide range of strain rates using
small test specimens. J Aerosp Eng 21:162–173
8. Amsterdam E et al (2006) Failure mechanisms of closed cell aluminum foam under monotonic and cyclic loading. Acta Mater 54:4465–4472
9. Lesser WP et al (2010) Fatigue crack closure analysis using digital image correlation. NASA TM 2010-216695, 2010
10. Chevalier L et al (2001) Digital image correlation used to analyze multiaxial behavior of rubber-like materials. Eur J Mech A Solids
20:167–187
11. Revilock DM et al (2007) Three-dimensional digital image correlation of a composite overwrap pressure vessel during hydrostatic pressure
tests. NASA TM 2007-214938, 2007
12. Thornburgh RP, Hilburger MW (2010) Longitudinal weld land buckling in compression loaded orthogrid cylinders. NASA TM 2010-216876,
2010
13. Tyson J et al (2002) Biomechanics deformation and strain measurements with 3D image correlation photogrammetry. Exp Tech 26(5):39–42
14. Zhang D et al (2002) Evaluating the material behavior of arterial tissue using digital image correlation. Exp Mech 42(2):409–416
15. Chen J et al (2005) Marking-dots digital image correlation and application to studies of spinal biomechanics. Proc SPIE 5852:455
16. Moerman KM et al (2009) Digital image correlation and finite element modeling as a method to determine mechanical properties of human soft
tissue in vivo. J Biomech 42:1150–1153
17. Oliver ST (2012) STS-133/ET-137 tanking test photogrammetry assessment. AIAA/ASME/ASCE/AHS/ASC structures, structural dynamics,
and materials conference, Honolulu to be presented 23–26 April 2012
Chapter 15
Experimental, Numerical and Analytical Evaluation of Structural
Integrity of Experiment Container Subjected to Reduced Gravity
Flight Testing
Mujahid Umar, Michael G. Jenkins, and Thomas L. Nguyen
Abstract An experiment container was designed, analyzed, fabricated and used for conducting a chemical experiment
(calcium oxalate production) under reduced gravity conditions aboard a NASA-operated aircraft. The container was
constructed of a welded and gusseted aluminum square-tubing frame and walled with polycarbonate panels that were
lined in silicone and bolted to remain in-place while reinforced with square u-shaped side-wall supports. Blast cabinet gloves
were fastened into the appropriate holes using mounting bolts and hardware. The bottom portion of the experiment container
was separated into six compartments in order to house the experiment apparatus that was mounted in the secondary
containment compartment. Eight tie-down straps secured the experiment container to the floor of the aircraft. For evaluation
purposes, separate design loads resulting from 9-g’s forward, 3-g’s aft, 6-g’s down, 2-g’s lateral, and 2-g’s up were imposed
on the container. Analytical models (NASA-required design codes) were used to calculated margins of safety from forces
and stresses. Numerical modeling (FEA) was used to determine stresses overall and at critical hotspots. Experimental
methods were used to validate the analytical and numerical results. In general, the design codes provided factors of safety for
the imposed loading conditions. As further validation of the design, the experiment container successfully flew on a reduced
gravity mission onboard a NASA-operated aircraft.
Keywords Reduced gravity • Experiment container • Structural integrity
15.1 Introduction
Calcium oxalate is a salt crystal found in many biological organisms [1–3]. The production rate of calcium oxalate in micro
and varying gravities is crucial to the survival of plants and astronauts consuming these plants. Knowledge learned from the
study of calcium oxalate production can aid in preventable actions for kidney stones that may possibly develop in astronauts
during prolonged space missions. Furthermore, experimental studies will aid in NASA’s current issues with calcium oxalate
crystal precipitation and filter clogging in the water waste system aboard spacecraft [4].
A microgravity experiment was proposed with the main objective to determine whether gravity affects the quantity and
crystal structure formation of calcium oxalate. The ultimate outcome of the experiment was to establish whether calcium
oxalate-producing plants can produce more calcium oxalate in reduced gravity conditions. Through this research, the aim
was to answer the following questions: Will the change in ionic concentration of secondary metabolite formation interfere
with normal metabolic metabolism? Will a change in the amount of calcium oxalate change the structure of plant cells? Will
the excess or deficiency of calcium oxalate increase or decrease the deterring capabilities of plants? If plants are more
deterrent toward predators, will they ultimately exclude the possibility of sustaining life of herbivores in deep space?
M. Umar • M.G. Jenkins (*)

Mechanical Engineering, M/S EE94, California State University, Fresno, CA 93740, USA
e-mail: jenkinsm@csufresno.edu
T.L. Nguyen
Levitas Consultants, Merced, CA 95344, USA
126 M. Umar et al.
Fig. 15.1 Boeing 727-200 aircraft (a.k.a., “weightless wonder,” “vomit comet”) used for zero gravity experiments [5]
Fig. 15.2 Boeing 727-200 aircraft (a.k.a., “weightless wonder,” “vomit comet”) used for zero gravity experiments [5]
Although the formation of calcium oxalate has been tested and analyzed by other research groups [1–3], all chemical
reactions and formations of calcium oxalate were performed under normal gravitational conditions. The current experiment
was executed and analyzed in reduced gravitational conditions. The hypothesis was that the formation of calcium oxalate
was more abundant in reduced gravity because there are less forces acting upon the ions therefore allowing more freedom for
structure development.
The experiment was designed to be self-contained in an experiment container that was strapped to the cargo floor in a
NASA reduced-gravity aircraft, in this case a Boeing 737–200 (B 727–200), nicknamed the “weightless wonder” or “vomit
comet” (see Fig. 15.1). The aircraft flies prescribed parabolic flight maneuvers so as to produce microgravity, lunar gravity
and Martian gravity conditions (e.g., Fig. 15.2) [5]. The lengths of these reduced gravity periods depend on the “g” level
required for the specific test. Table 15.1 shows typical lengths for various maneuvers. After extensive research and drafts of
proposed user-friendly experiment container devices that would securely hold the hazardous materials, the structural
material and design configuration were finally selected that would not only secure our test specimens but also withstand
the potential experimental stresses and forces as prescribed by NASA [6].
In this paper, details of the experiment container are presented first, followed by details of the analytical, numerical and
experimental methods used to evaluate the structural integrity of the experiment container. Next the results of the methods
are compared and discussed. As further validation of the design, the experiment container successfully flew on a micrograv-
ity mission onboard a NASA-operated aircraft. Finally, some salient conclusive points are made about the structural integrity
of this experiment container under the loading conditions.
15 Experimental, Numerical and Analytical Evaluation of Structural Integrity of Experiment. . . 127
Table 15.1 Maneuver, “g” level Maneuver “g” level Duration (s)
and duration of each parabolic
flight segment Hyper-g Up to 1.8 g 60
Zero-g 0g 17
Lunar-g 1/6 g (0.16 g) 20
Martian-g 1/3 g (0.38 g) 20
Fig. 15.3 Pictorial

representation of experiment
container showing aluminum
frame; polycarbonate sides,
hinged top and bottom; access
ports; storage compartments
15.2 Experiment Container
The design of the experiment container had to meet NASA’s guidelines and specifications [5, 6]. The experiment container
had to have dimensions of 588 mm 1,470 mm 1,470 mm (2400 6000 6000 ) and weigh no more than 135 kg (300 lb.).
Therefore these guidelines led to a container that is strong, resistant to deformation, mobile, multifunction-able, and reusable.
The material chosen for the experiment container frame was 6061-T6 aluminum. This welded frame was reinforced by
doublers and gussets comprised of the same material. The side walls were constructed of polycarbonate polymer. The
combination of these materials proved sufficient. The larger front and rear panels were cut to allow for the attachment and
insertion of blast cabinet gloves. These gloves were fastened into the appropriate holes using mounting bolts and hardware.
All doors were fixed with spring hinges, preventing any pre-mature opening. Sliding doors were held in the closed position
using magnets. Syringe apparatuses were stored and utilized within this container. The configuration of the experiment
container is shown pictorially in Fig. 15.3.
The experiment container was oriented length-wise parallel to the aircraft y-axis [5]. All experimental materials were
secured inside the compartments of the experiment container during takeoff and landing. Two operators were positioned
along the center line of the aircraft and one operator was on the outboard side of the cabin.
15.3 Analytical, Numerical and Experimental Methods
Analytical, numerical and experimental methods were applied to the frame of the experiment container to determine its
structural integrity. Each method included its own simplifications, assumptions and results, which when evaluated separately
provided some information about the structural integrity, but when evaluated collectively provided a complete information
of the structural integrity. Each method is detailed separately in the following subsections.
Analytical method: This method began with a simple static analysis of the experiment container connected to the cargo floor
through cargo floor through tie-down straps only (see Fig. 15.4). NASA requires that structural integrity of all equipment
flown aboard the reduced gravity aircraft must be verified [via accepted method(s), analysis, test, demonstration, similarity,
128 M. Umar et al.
Fig. 15.4 Schematic analysis of experiment box frame and tie-down straps (note: dimensions in inches)
Table 15.2 Summary of analytical method applied to frame of the experiment container
Load condition Type of analysis Maximum calculated Maximum allowable Factor of safety
9-g’s Forward Tension on strap 5,004 N 8,896 N 1.8
Stress in frame (for this case at tie-down strap handle) 4.5 MPa 120 MPa (¼Syp/2) 27
3-g’s Aft Tension on straps 3,473 N 8,896 N 2.6
6-g’s Down Stress in frame (in this case bottom frame) 0.04 MPa 120 MPa (¼Syp/2) 2,725
2-g’s Lateral Tension on straps 1,736 N 8,896 N 5.1
2-g’s Up Tension on straps 422 N 8,896 N 21
Note: Syp ¼ yield strength of aluminum. Factor of 2 is the knock-down factor for welded aluminum
etc.] and documented accordingly. Structures, in particular, are to be verified using accepted practices (e.g., free body
diagrams) with load vectors applied at accurate centers of gravity. As such, factors of safety (FS) of 2.0 or greater shall be
applied to all structural or fracture critical elements. In addition, designers are encouraged to use FS greater than 2.0
whenever practical. The NASA-mandated loading conditions are [6]: forward 9 g, aft 3 g, down 6 g, lateral 2 g and up 2 g.
Normally, material yield strengths are to be used as the maximum allowable throughout all design calculations. Where
appropriate, NASA design codes [6] were applied to determine the factors of safety for the weldments using a knockdown
factor of 2 applied to the yield strengths [7] for the welded 6061-T6 aluminum. Representative mechanical properties
assumed for 6061-T6 aluminum [8] are: ultimate tensile strength, Suts ¼ 289 MPa (42 ksi); yield strength, Sy ¼ 241 MPa
(35 ksi); %elongation (2 in.) ¼ 6.
The summary of the analytical method applied to the frame of the experiment container is shown in Table 15.2.
Numerical method: Finite element analysis (FEA) consisted of a full, three-dimensional analysis using quadrilateral
elements (127,336 nodes and 60,078 elements) under static loading imposed by the forces determined in the straps
determined in the analytical method (e.g., see Figs. 15.5 and 15.6). A commercial solid modeling program suite [9] with
built-in FEA capability was used for the modeling. Typical solution run times on computer workstations per loading
condition were in the 5–8 min range. Subsequent refined analysis of the welded frame corners was conducted to determine
stress “hotspots” for the subsequent experimental method.
Fig. 15.5 Example of FEA modeling showing distorted shape for 9-g loading condition and applied boundary conditions (i.e., frame supported on
the bottom and loaded through the tie-down strap handles)
Fig. 15.6 Example of FEA modeling showing distorted shape and von-Mises stresses for 9-g loading condition
Experimental method: Logistical difficulties prevented completion of the experimental portion of this study before the
publication submittal deadline for this paper. However, the process plan for completion of the experimental method is
detailed as follows.
1. Apply rosette strain gages at critical “hot spots” of the frame experiment box. “Hot spots” are determined from the FEA
model. Up to ten critical “hot spots” are to be instrumented.
2. Position the instrumented experiment container on the T-slot floor of the structural testing high bay.
3. Connect tie-down straps between the handles of the experiment container and eye bolts in T-slots of floor. Use a ratchet
tensioner and electronic force transducer connected to each tie-down strap to apply and monitor tension, respectively.
130 M. Umar et al.
Table 15.3 Comparison of analytical, numerical and experimental results

Load condition Type of analysis Analytical Numerical Experimental
9-g’s Forward Tension on strap 5,004 N N/A –
Stress in frame 4.5 MPa in handle 179 MPa (von Mises) at corner Not completed
3-g’s Aft Tension on straps 3,473 N N/A –
6-g’s Down Stress in frame 0.04 MPa in frame floor 36 MPa (von Mises) at corner Not completed
2-g’s Lateral Tension on straps 1,736 N N/A –
2-g’s Up Tension on straps 422 N N/A –
Fig. 15.7 (a) Fresno state microgravity team, (b) experiment container in cargo hold, (c) microgravity experiment
4. After making the proper electrical connections to the appropriate conditioning and data acquisition equipment, zero all
inputs.
5. Using the ratchet tensioner, apply the appropriate tensile force (measured using the force transducer) to each tie-down
strap as determined from the NASA requirements (see Table 15.1) for each loading condition of 9-g’s Forward; 3-g’s Aft;
6-g’s Down; 2-g’s Lateral; and 2-g’s Up
6. Record the strains and forces for each loading condition.
7. Using appropriate constitutive relations, determine the coordinate, principal and von Mises stresses at each “hot spot” for
comparison to the results from the Analytical and Numerical Methods.
15.4 Comparison of Methods
The goal of this study was to compare the results from the analytical, numerical, and experimental methods for the different
loading conditions in order to assess the structural integrity of the experiment container. These comparisons are shown in
Table 15.3.
15.5 Conclusions
Upon finalizing the design of the experiment container and conducting many analytical calculations, NASA’s engineers
deemed the box structurally sound and ready for flight. The angled aluminum welds and polycarbonate polymer enhanced
the overall strength of the box providing an effective safety barrier between the hazardous materials and the experimenters.
Additional calculations indicated that the welds were capable of withstanding maximum loading of 18 g’s (well beyond the
NASA design requirements). Subsequent numerical analysis and experimental testing of the as-fabricated experiment
container should verify that under the maximum loading of 18 g’s through the tie-down straps, stresses in the frame do
exceed the elastic range. Validation of the design was provided when the experiment container successfully flew on a
microgravity mission onboard a NASA-operated aircraft in July 2011, as shown in Fig. 15.7.
References
1. Finch AM, Kasidas GP, Rose GA (1981) Urine composition in normal subjects after oral ingestion of oxalate-rich foods. Clin Sci 60(4):411–418
2. Chaudhary A, Singla SK, Tandon C (2010) In vitro evaluation of Terminalia arjuna on calcium phosphate and calcium oxalate crystallization.
Indian J of Pharmacology Sci 72(3):340–345
3. Grases F, Millan A, Conte A (1990) Production of calcium oxalate monohydrate, dihydrate, trihydrate: a comparative study. Department of
Chemistry at the University of the Balearic Islands, Palma de Mallorca, and the Urology Service in the Hospital General “Virgen de Lluch”
(Insalud), Palma de Mallorca, Spain. Urological Research 18:17–20
4. Malloy SW, Proposal Number: 2011–25298 Calcium oxalate is a component of waste water on-board spacecraft with humans and is an issue for
precipitation and clogging of filters. Higher Education and Student Programs Branch (AD42)
5. Zero Gravity Corporation (2009) Interface Control Document Boeing 727–200. Document ZGC-ICD Revision A2, Approved by Byron K.
Lichtenberg, Zero Gravity Corporation, Vienna
6. National Aeronautics and Space Administration (2010) Experiment design requirements and guidelines for microgravity research. Document,
AOD 33897 Rev D, Approved by Dominic L. Del Rosso, NASA, Lyndon B. Johnson Space Center, Houston, TX
7. Task Committee on Lightweight Alloys (1962) Suggested specifications for structures of aluminum alloys 6061-T6 and 6062-T6. J Struct Div
88(6):1–46
8. ALCOA Mill Products (2005) 6061-T4 and T6 Sheet: technical data. ALCOA Mill Products, Bettendorf
9. Dassault Systèmes SolidWorks Corp. (2012) SolidWorks 3D CAD and SolidWorks simulation. SolidWorks Corp., Waltham, MA
Chapter 16
Use of Fiber Bragg Grating Strain Gages on a Pipeline Specimen
Repaired with a CFRE Composite System
J.L.F. Freire, V.A. Perrut, A.M.B. Braga, R.D. Vieira, A.S.A. Ribeiro, and M.A.P. Rosas
Abstract Re-establishing the maximum operating pressure of a segment of pipeline with metal loss defects, such as erosion
or corrosion defects, can be accomplished either by replacing the damaged segment altogether, or by applying a localized
repair system. The present paper deals with laboratory tests conducted: (1) to understand and describe how the reinforcement
layers of a carbon fiber epoxy composite material can enable a steel line pipe specimen with a metal loss defect to withstand
pressure loading; (2) to compare the test results with those predicted by Mechanics of Materials and by Finite Element
numerical solutions developed previously. Hydrostatic burst tests were performed on three pipe (API 5L X65 ERW)
specimens: one with metal loss defect, one without metal loss defect, and one with metal loss defect but repaired with a
carbon fiber reinforced epoxy composite system CFRE. Fiber Bragg grating FBG strain gages were used to monitor elastic
and plastic strains during the tests of the repaired specimen. The strain gages were bonded either directly on the surface of
the defect, or were inserted in between some of the composite layers in order to show the reinforcement’s effective
contribution to the strength of the repaired pipes. The analytical and numerical results agreed very satisfactorily with
experimentally determined burst pressures and pressure-strain curves, showing that the behavior of composite reinforced
pipelines can be well predicted by using simple Mechanics of Materials or sophisticated Finite Element solutions.
Keywords FBG strain gages • Pipelines • Composite repair • Carbon fiber composite
16.1 Introduction
Re-establishing the maximum operating pressure of a segment of pipeline with metal loss defects, such as erosion or corrosion
defects, can be accomplished either by replacing the damaged segment altogether, or by applying a localized repair system.
Composite repair systems are seen to be more economical than other repair alternatives; since they are ¼ less expensive than
welded steel sleeve repairs and less expensive than the complete replacement of the damaged segment [1]. The present paper
deals with laboratory tests conducted to understand and describe how the reinforcement layers of a composite material made
from carbon fiber reinforced epoxy (CFRE) can enable a steel line pipe specimen with metal loss to withstand pressure
loading, and to compare the experimental results with results derived from two solutions based on Finite Elements and simple
Mechanics of Materials [2]. Hydrostatic pressure and burst tests were performed on three pipe specimens: one with metal loss
defect, one without metal loss defect, and one with metal loss defect but repaired with a carbon fiber reinforced epoxy
composite system. Fiber Bragg grating FBG strain gages were used to monitor elastic and plastic strains during the tests.
J.L.F. Freire (*) • A.M.B. Braga • R.D. Vieira • A.S.A. Ribeiro

Pontifical Catholic University of Rio de Janeiro, Rio de Janeiro, Brazil
e-mail: jlfreire@puc-rio.br
V.A. Perrut
CENPES, PETROBRAS, Rio de Janeiro, Brazil
M.A.P. Rosas
Prima-7S, Rio de Janeiro, Brazil
134 J.L.F. Freire et al.
16.2 Experimental Methods
The three pipe specimens were made of American Petroleum Institute API 5L X-65 grade steel line pipe [3] produced by the
electric resistance welding process ERW. The nominal specimens’ dimensions were diameter D, thickness t and length A,
equal to 457 mm (18 in.), 8.7 mm and 2.7 103 mm, respectively. Reinforced flathead caps were welded to the pipe ends so
that the hydrostatic internal pressure tests could be performed. Two of these specimens had long, external metal loss defects
produced by sparking erosion in order to simulate corrosion defects. The defects had a uniform profile as shown in Fig. 16.1.
Their nominal dimensions were depth d, length L and width W, equal to 70% of the pipe thickness (6.1 mm), 450 and 85 mm,
respectively. The metal loss area of one of these specimens was reinforced with a carbon fiber reinforced epoxy (CFRE) repair
system in order to restore it to its original pipe strength. The defect area and superposed repairing composites layers of the
CFRE repair system of this specimen were also instrumented with circumferential optical strain gages based on fiber Bragg
grating sensors. The geometric dimensions, mechanical properties and other relevant data regarding the pipeline specimens,
prediction equations, repair system, experimental setup and test sequence are given in Figs. 16.1 and 16.2, and in Table 16.1.
2.7 x 103 mm
B R1=10mm
459mm
B
A A
85mm
R3=4mm
450mm
R2=4mm 6.3mm 8.7mm
A-A
Corte B-B
Fig. 16.1 Geometric dimensions of the test specimens and of the machined spark-erosion defect
Rupture test of
Defect free specimen repaired specimen
Rupture at 26MPa Burst at 27MPa
Hydrostatic test
X X
1 Pressure at 15.4MP during 4 hours
3
3
Test pressure (MPa)
Specimen with defect

Gage bonding and application of 10 cycles test at maximum
CFRP system at pressure of 4.1MPa operating pressure of 12.3MPa
and 24 hours epoxy curing time
3
Specimen with non- 3
repaired defect
Rupture at 7.5MPa
X
2
Time or test event
Fig. 16.2 Sequence of tests carried out on each of the three test specimens
16
Table 16.1 Experimental data

Specimen’s data Parameter Nominal data Specimen 1 defect Specimen 2 with Specimen 3 with defect and
free defect repaired
Specimen dimensions D (mm) 457.2 459 459 459
t (mm) 8.74 9.0 8.5 8.9
Defect dimensions d (mm) 6.1 – 6.2 6.5
L (mm) 450 – 450 450
Steel pipe ERW API 5L X65 mechanical Yield strength (MPa) SMYS ¼ 448 Sy ¼ 512 Sy ¼ 510 Sy ¼ 516
properties Ultimate strength SMUS ¼ 530 Su ¼ 630 Su ¼ 663 Su ¼ 666
(MPa)
Predicted pressure Eq. 1–3 (MPa) Design (1) 12.3 – – –
Hydrostatic test (2) 15.4 – – –
B31-G (3) 4.1 – – –
Burst pressure: standard, numerical, analytical, and present test DNV RP-F101 Eq. 4 7.5 25 (nominal area) 8.2 26 (nominal area)
results (MPa) Finite elements [2] – – – 30
Analytical [2] – – – 29
Test results (present – 26 7.5 27
paper)
Location of rupture (test results) – Nominal area Defect Nominal area
Prediction Design pressure pd ¼ 2:t:SMYS:F
D ; F ¼ 0:72 (1)
equations Hydrostatic test pressure pH ¼ 1:25 pd (2)

ASME B31-G, long defect, safe pressure p31G ¼ 2:t:SMYS:F
D ð1:1Þ 1 dt ; F ¼ 0:72 (3)
to operate with defect [4]
0 1
Use of Fiber Bragg Grating Strain Gages on a Pipeline Specimen. . .
DNV RP-F101 (burst pressure of specimen (4)

2:t:Su : @ 1 dt
with defect) [5] pburst;DNV ¼ A
Dt d
1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
t 1þ0:31L =D:t
Repair system Type of repair system Carbon fiber reinforced polymer (epoxy) applied in continuous layers which adhered to the pipe surface and to each
other by means of using a fast curing liquid epoxy (~3 h application, 24 h full use) adhesive
Number of layers, total thickness and 12 layers; Minimum total repair thickness ¼ 11 mm; Maximum total repair thickness ¼ 19 mm (caused by the
total length of the repair system superposition of adjacent layers). The average thickness of the repair system used in calculations was 15 mm
Total length ¼ 550 mm. The repair system consisted of three adjacent sets of 12 layers that covered and exceeded
each end-side of the defect length by 50 mm and had about 50 mm of partial superposition of their adjacent
boundaries
Mechanical properties of the CFRP (From [2]) Young modulus E (GPa) 49 (c ¼ circumferential); 23 (l ¼ longitudinal); 5.5 (r ¼ radial)
Shear modulus G (GPa) 0.69(r-l); 0.69(l-c); 30(r-c)
Maximum elongation eu (%) 1.6 (circumferential)
Poisson coefficient m 0.43(r-l); 0.0.43(l-c); 0.20(r-c)
Epoxy E ¼ 1.7 MPa; m ¼ 0.45
Experiment Test pressure transducer Gefran model TK
details Volume of water Measurement of volume of water injected during the hydrostatic test employed a resistive level transducer
Pressure device Pneumatic-air driven/water pump Haskel BSS-100
Measurement of the circumferential strains Fiber Bragg grating FBG strain gages with 2 mm gage length. Three sensors were used in the defect surface and were
in the defect and repair areas bonded using cyanoacrylate adhesive. Fifteen sensors were placed in between the various CRFP repair layers
sensors and bonded to the various repair layers using the same epoxy resin employed in the repair process
Data conditioners Pressure data and water volume data: Lynx ADS2000 conditioner. FBG strain gages: conditioner Micron Optics
model sm125-500, data acquisition board NI USB-6008 12 bits (National Instruments) plus optical fibers
Data acquisition rate 1 Hz
135
The three specimens were all tested up to rupture, but before the rupture test, the CFRE repaired specimen was tested
under the following conditions: (1) installing the repair system and bonding the FBG gages while the specimen was
subjected to the maximum allowed operating pressure calculated by the American Society for Mechanical Engineers
ASME B31-G method [4], this calculation taking into consideration the spark-eroded metal loss defect; (2) increasing the
internal pressure to reach the standard hydrostatic pressure test for the repaired specimen, considering in this case that it
recovered the original pipeline strength; (3) decreasing the pressure to zero, followed by ten pressure cycles that varied from
zero pressure to the maximum operating pressure (of the pipeline considered completely restored); (4) increasing the
pressure from zero to the burst pressure. Figure 16.2 shows schematically the sequence of tests performed on each of the
three test specimens.
The specimen that received the repaired system was instrumented with 15 FBG strain gages that had a 2 mm gage length.
The gages were produced in-house and were placed in series of three or six gages along three fibers. In each fiber the gages
were set 500 mm apart from each other. Each fiber, one with three gages and two with six gages, was placed in one of the
three of four channels of the FBG reading device (Micron Optics model sm125-500).
Figure 16.3 shows the gage locations in the repaired test specimen. All the gages were positioned to measure circumfer-
ential strains, and were placed in the following positions: three gages were bonded directly to the defect’s surface (at a
distance of 45 mm from each other, from the center of the defect to the end of the defect area); two gages were positioned
between the first and second layers (one over the center of the defect and another 90 away); two gages were positioned
between the third and fourth repair layers (center and 90 away); two were positioned between the fifth and sixth repair
layers (center and 90 away); two were positioned between the seventh and eighth repair layers (center and 90 away); two
were positioned between the ninth and tenth repair layers (center and 90 away); and lastly, two gages were positioned
between the eleventh and twelfth repair layers (center and 90 away). Figure 16.3 also shows some of the gages bonded
to the defect’s surface, details of the repair system, and the entire specimen showing the three adjacent repair layers used to
cover the full defect length. The fracture area encompassing an originally non-defective area is also shown.
16.3 Pressure Tests and Results
The burst pressures of the defect-free specimen and of the specimen with the unrepaired defect were equal to 26 and
7.5 MPa, respectively. According to the equation of DNV RP-F101 [5], which can be expected to furnish a good
approximation of the burst pressure of the specimen [6, 7], the calculated rupture pressure of the specimen with defect
was 8.2 MPa, which is 9% off from the experimental result.
The test procedure and results are shown and described herein. Figure 16.2 helps to illustrate the sequence of the tests.
The remaining tubular specimen that also contained the defect was pressurized up to 4.1 Mpa (safe operating pressure for the
specimen with an unrepaired defect with depth d equal to 70% of thickness t, according to ASME B31-G [4]). This pressure
level was kept constant during the application of the repair system (layers of bi-directional carbon fiber fabric and liquid
epoxy adhesive) and of the process of bonding the fiber optic strain gages. Subsequently the specimen was maintained at the
same pressure for 24 h, which was the time needed for the epoxy resin to cure completely. The initial measurement readings
(zeroing process) of the fiber optic strain gages were at the 4.1 MPa pressure. Next, the specimen was pressurized at a rate of
0.4 MPa up to 15.4 MPa (hydrostatic test pressure of the pipeline without defect, which is 25% above the design operating
pressure of a API 5L X65 oil or gas pipeline, class 1 division 2, without defect) and it was kept at this pressure for 4 h.
Finally, the pressure was decreased back to zero at a rate of 10 MPa/min. The strain values measured by the gages located on
the defect’s surface and in the third and ninth repair layers are presented in Fig. 16.4.
Circumferential strains measured during the pressure test (4.1–15.4 MPa) by the gages positioned in the third and ninth
layers of the repair system, located immediately above the repair and in a position 90 away from the defect’s center, are
presented in Fig. 16.5. One can see that strains are larger in the locations right above the defect than in the locations outside
(90 away) the defect.
The non-linear distribution of strains along the thickness of the repair system and above the defect’s center area can
be observed in the graph in Fig. 16.6. One can see that the strain’s distribution is not uniform along the repair thickness.
The strains are larger in locations closer to the inner layers, as one would expect if a thick theory of tubular specimens under
internal pressure were applied. One can also see in the graph that strains in the metal surface start reaching elasto-plastic
conditions when total test pressure is equal to 12 MPa (note that the zero measured strain corresponds to the initial test
pressure of 4.1 MPa). This can be stated because the definition of yield strength in API 5L [3] considers a total strain of 0.5%.
After reaching the maximum hydrostatic test pressure of 15.4 MPa, the specimen was submitted to 10 cycles from zero up
to maximum operating pressure. In other words, the internal pressure was increased from zero MPa to 12.3 MPa. The
16 Use of Fiber Bragg Grating Strain Gages on a Pipeline Specimen. . . 137
Fig. 16.3 Fiber Bragg grating FBG strain gage locations and test specimen: (a) general location of the fiber optic strain gages; (b) three strain
gages, belonging to the same fiber, bonded inside the defect on the specimen’s surface; (c) location of the three strain gages bonded on the
specimen’s surface; (d) epoxy filling of the defect’s recess; (e) exit of optical fibers from the repair system; (f) tubular-repaired specimen after burst
test – rupture location outside the defect-repaired area
pressure increase and decrease rates both varied between 10 and 15 MPa/min. The measured strain-pressure cycles presented
very repetitive results, as shown in Fig. 16.7, which shows the superposition of the 10 strains cycles measured with the gage
positioned in the ninth layer above the defect location.
18
Pressure (MPa)
16 9th repair
3rd repair
layer
14
layer
12
defect
10 surface
8
0
0 2000 4000 6000 8000
Srain µe
Fig. 16.4 Circumferential strains measured by the fiber optic strain gages located on the defect’s surface, and in the third and ninth layers of the
repair system, all above the defect’s center position
16
Pressure (MPa)
9th repair
layer, 90 3rd repair
14 apart layer, 90
apart
12
3rd repair
layer, above
10 the defect
9th repair
8 layer, above
the defect
0
0 500 1000 1500 2000 2500 3000
Strain µe
Fig. 16.5 Circumferential strains measured in the third and ninth layers in locations above the defect’s center and 90 away
After the tenth cycle of pressurization, the strain gages were reset to zero and the internal pressure was again increased to
the point that the specimen burst after reaching the rupture pressure of 27 MPa. Figures 16.8 and 16.9 show the increasing
strain gage readings with increasing test pressures for the gages mounted in the third and ninth layers above the defect.
Although the gage mounted in the third layer lasted throughout the test, the gage in the ninth layer failed when the total
measured strain approached 3,200 me and the pressure was around 20 MPa. The gages mounted on the defect’s surface failed
due to strains higher than 5,000 me, which occurred during the 10 test cycles. The graphs in Fig. 16.9 help to illustrate the
behavior sensed by the gages during the first hydrostatic test and its unloading, and during the final burst test. Regarding the
latter, linear behavior was present under loading conditions up to the point that the pressure of 15.4 MPa was reached. This
happened because the 15.4 MPa pressure was the maximum pressure reached during the hydrostatic test. From this point and
higher, the angular slope of the curve depicting the burst test changed. This can be explained by the fact that since the steel
pipe material was under plastic behavior, most of the remaining rigidity presented by the curve was due to the repair system,
which continued to display elastic behavior.
Pressure = 6MPa
Radial position and symmetry line
Strain µε
Strain µε
Epoxy
filled gap
Steel pipe
Radial position mm
CFRE layers
Radial position mm
Fig. 16.6 Circumferential strain distributions across the repair thickness and their variation with test pressure (zero strain corresponds to the
starting test pressure, which is equal to 4.1 MPa)
18
Pressure (MPa)
16
Hydrostatic
14 test
12
10
6
10 cycles at
operating
4
pressure
2
0
0 500 1000 1500 2000 2500 3000 3500
Strain µe
Fig. 16.7 Circumferential strains measured by the fiber optic strain gage located in the ninth layer. The graph shows the hydrostatic test increasing
from 4.1 MPa to maximum pressure of 15.4 MPa and 10 superposed and very repetitive pressure cycles from approximately zero pressure up to the
maximum operating pressure of 12.3 MPa
16.4 Numerical and Analytical Results
The results of the experiment can enable one to understand the structural and reinforcing behavior of the repair system.
It may be considered as a benchmark to be used in comparisons with results generated by a numerical model and by an
analytical model developed to accurately simulate the integrity assessment of pipelines with composite repair systems.
Reference [2] developed the two model solutions for the present case. The solutions used a Finite Element FE model and a
simple Mechanics of Materials model. The FE simulation employed solid elasto-plastic 3D elements and the Ansys 11.0
software. The epoxy resin used to fill the defect gap was also included in both the numerical and the analytical models.
Pressure (MPa)
30
25
Sensor failure
20
9th
15 repair
layer 3rd repair
layer
10
0
0 1000 2000 3000 4000 5000 6000
Strain µe
Fig. 16.8 Circumferential strains in the third and ninth layers of the repair system (above the defect) during the burst test
25 30
Pressure (MPa)
9th repair layer 3rd repair layer

25
20
Hydrostatic 20 Hydrostatic
15
test Burst test test
15 Burst test
10
10
5
10 cycles at operating 5
pressure
0 0
0 500 1000 1500 2000 2500 3000 3500 4000 4500 0 1000 2000 3000 4000 5000 6000 7000
Strain µe Strain µe
Fig. 16.9 History of strain data collected for both gages positioned in the third and ninth layers of the repair system (above the defect) during the
hydrostatic test, 10 test cycles (only gage in the ninth layer) and burst test
The epoxy filling and the composite CFRE material were treated as linear elastic materials. It is beyond the scope of this
article to give more details on both solutions, which are fully described in reference [2]. Two results from these simulations
are presented in Figs. 16.10 and 16.11. Figure 16.10 shows the variation in the circumferential strain caused by increased
pressure at points belonging to the numerical solution and located at depths equivalent to the third and ninth layers of the
repair system. These results are presented together with the strain gage responses given by the gages located in the same
positions of the repair system during the final burst test. One can see that the numerical solution curves agree quite well with
the actual results.
Figure 16.11 shows the comparison between the numerical solution and the simple solution derived from Mechanics of
Materials. In the latter, the structure is composed of three concentric and contacting pipes formed by the steel, epoxy filler
and composite materials. These pipes interfere with and react to the application of internal pressure to the steel pipe walls.
The steel pipe material is modeled by an elasto-plastic bi-linear material, where the yield strength is reached at 0.5% total
strain and the ultimate strength is reached at 9% total strain. The epoxy filler and the composite materials are considered as
presenting a linear behavior, although maximum total (ultimate) strains can be imposed to limit their contribution in adding
strength and/or rigidity to the entire structure response. In the present solution, the maximum pressure to be calculated is the
one that makes one of the three materials reach its ultimate strength first, or, as was the case in the present analysis (see in
Fig. 16.3f that fracture occurred outside the defect area), the ultimate strength of the steel of the pipe without defect. As one
can see from the comparison in Fig. 16.11, both solutions agree satisfactorily, and as a result, the analytical solution agrees
satisfactorily with the actual results of the experiment. This successful comparison is important in terms of opening an
opportunity for predicting the behavior of composite reinforced pipelines using simple Mechanics of Materials solutions.
35
Pressure (MPa)
3rd repair layer
FE numerical solution
30 9th repair layer
FE numerical solution
25
20 9th repair layer

Strain gage 3rd repair layer
Strain gage
15
10
0
0 1000 2000 3000 4000 5000 6000 7000
Strain µε
Fig. 16.10 Comparison between the experimental and the numerical (FE) solutions for point located above the defect in the third and ninth layers
of the repair system
600
von Misesequivalent stress (MPa)
Analytic solution at
defect of steel pipe
500
FE solution for surface defect
400 Analytic solution for

internal layer of repair
FE solution for
300
internal wall of repair
200
FE solution for external
wall of repair
100 Analytic solution for
FE solution for epoxy filler
epoxy filler
0
0 5 10 15 20 25 30 35
Pressure (MPa)
Fig. 16.11 Comparison between the experimental and the numerical (FE) solutions for points located in the steel surface defect, in the epoxy filler
and in the composite repair system. Calculations were stopped when the pressure reached 30 MPa, which is the burst pressure for a nominal section
of a specimen without defect
In the present case, the burst test pressure was 27 MPa (nominal non-defective pipe specimen section), while the FE solution
also displayed fracture at a nominal section for a pressure equal to 30 MPa. The simple analytical solution for the nominal
pipe section is 29 MPa.
16.5 Conclusions
This paper has presented the hydrostatic test results of experiments conducted on three pipeline specimens made of API 5L
X65 ERW steel line pipe to demonstrate how the reinforcement layers of a carbon fiber epoxy composite material can enable
a steel line pipe specimen with a metal loss defect to withstand pressure loading. Fiber Bragg grating strain gages were
successfully used to monitor the elastic and small plastic strains during the tests of the repaired specimen. The strain gages
were either bonded directly to the surface of the defect or were inserted in between some of the composite layers, and they
demonstrated the reinforcement’s effective contribution to the strength of the repaired pipes. The test results compared
satisfactorily to results predicted by a Mechanics of Materials analytic solution and by a Finite Element numerical solution
developed previously, showing that the behavior of composite reinforced pipelines can be satisfactorily predicted by using
simple Mechanics of Materials or sophisticated Finite Element solutions.
References
1. Duel J, Wilson J, Kessler M (2008) Analysis of a carbon composite overwrap pipeline repair system. Int J Press Vessel Pip 85:782–788
2. Rosas MAP (2010) Evaluation of sleeve type repair models for pipelines with thickness loss defects. Doctoral thesis (in Portuguese),
Mechanical Engineering Department, Pontifical Catholic University of Rio de Janeiro, PUC-Rio, September 2010
3. API Specification 5L (2009) Specification for line pipe, 44th edn. American Petroleum Institute, Washington, DC
4. ASME (2009) ASME-B31G manual for determining the remaining strength of corroded pipelines – a supplement to ANSI/ASME B31 code for
pressure piping. The American Society of Mechanical Engineers, New York
5. DNV (1999) Corroded pipelines, recommended practice RP-F101. Det Norske Veritas, Høvik
6. Freire JLF, Vieira RD, Castro JTP, Benjamin AC (2006) Burst tests of pipeline with extensive longitudinal metal loss. Exp Tech 30(6):60–65,
November–December
7. Freire JLF, Vieira RD, Diniz JC (2007) Effectiveness of composite repairs applied to damaged pipeline. Exp Tech 31(5):59–66,
September–October
Chapter 17
Microstructure in the FSW Butt Joint of Aluminum Alloy AA5083
Thaiping Chen, Wei-Bang Lin, and Chang-Mao Hung
Abstract The weight reduction for vehicles is the great expectations for energy saving and environmental preservation.
Aluminum alloys will utilize and be helpful very much in some steel structure of vehicles to reduce their weight. Friction Stir
Welding (FSW) is a novel material joining technique, which can produce superior mechanical properties in the weld zone.
As compared to conventional welding, FSW consumes considerably less energy. No shielding gas or flux is used, thereby
making the process environmentally friendly. The joining does not involve any use of filler metal and therefore any kind of
aluminum alloy can be joined without concern for the compatibility of composition, which is an issue in fusion welding. In
this paper, two plates of aluminum alloy AA5083 were FSW in the butt joint for tensile test and TEM observation on stir
zone. The specimens for TEM are chosen from the cases of the ultimate tensile strength of FSW butt joint over 300 MPa and
larger than 86% of the base material. TEM field images show that the microstructures in the stir zone of FSW joint were
recrystallized grains in micron-meter order and submicron-meter order, and fund amorphous within stir zone of welds.
Boundaries of grains are not definable that is uncommon phenomenon.
Keywords FSW • Aluminum alloys • Microstructure
17.1 Introduction
The weight reduction for vehicles is the great expectations for energy saving and environmental preservation. Aluminum
alloys will utilize and be helpful very much in some steel structure of vehicles to reduce their weight. Friction Stir Welding
(FSW) is a novel material joining technique invented in 1991, which can produce superior mechanical properties in the weld
zone [1]. As compared to conventional welding, FSW consumes considerably less energy. No shielding gas or flux is used,
thereby making the process environmentally friendly. The joining does not involve any use of filler metal and therefore any
kind of aluminum alloy can be joined without concern for the compatibility of composition, which is an issue in fusion
welding.
The FSW process is a complex series of materials processing in the physical properties changed, but it can be studied by
experiment [2, 3]. Taguchi technique for the experimental data analysis was used and found optimal FSW process
parameters [4, 5]. The microstructural evolution and resultant mechanical properties during the FSW of precipitation
strengthened aluminum alloys depend on initial temper as well as FSW process parameters [6]. The mechanical properties
were a function of welding speed in FSW aluminum AA5083 and its fatigue properties [3, 7]. Different welding speeds to
weld AA 5083 were chosen to ensure high welding speeds in friction stir weldability [8]. Furthermore, dissimilar materials
could be joined by the FSW method [4, 5]. The literature about the FSW of dissimilar alloys provided a comprehensive
T. Chen (*)
Department of Electrical Engineering, Fortune Institute of Technology, Kaohsiung 83160, Taiwan
e-mail: taiping@center.fotech.edu.tw
W.-B. Lin
Department of Mechanical Engineering, Chinese Military Academy, Kaohsiung 83059, Taiwan
C.-M. Hung
Department of Vehicle Engineering, Yung-Ta Institute of Technology and Commerce, Pingtung 90942, Taiwan
144 T. Chen et al.
picture of both the current status of the field and the opportunities and challenges for the future [9]. A good understanding of
the thermomechanics in the workpieces can be helpful in evaluating the process as well as the weld quality. The study of the
heat flow into FSW tools is also helpful in evaluating the weld quality [10]. From the perspectives of research and
application, microstructural evolution in FSW welds is the most popular topic [11, 12]. The size of a-Al grains was found
to be increase by increasing both tool rotational and welding speeds. The welded A319 cast Al alloy joints showed better
mechanical properties than the base material [13].
The main purpose of this paper is to observe the microstructure in the stir zone of the aluminum AA5083 FSW joint for its
plastic movement after hot machining. The specimens for Transmission Electron Microscopy (TEM) observation are chosen
from the cases of ultimate tensile strength of FSW butt joint.
17.2 Experimental Procedures
The specimens, namely AA5083 aluminum alloy, used for the friction stir welding (FSW) experiments were purchased from
open market. Two plates of AA5083 aluminum alloys were friction stir welded in the butt configuration by using an adapted
milling machine. The two plates were placed side to side and clamped firmly to prevent the abutting joint faces from being
forced apart. The FSW procedure was based on the procedure described as in the literature [5]. The welding direction of
aluminum alloy was along the line of the joint. The rotation of tool resulted in stirring and mixing of material around the
rotating pin and the translation of tool moved the stirred material from the front to the back of the pin and finished welding
process. The advanced side and retreating side of the welding sheet were defined according to the rotation of tool and the
joint line. The insertion depth of pin into the workpieces was associated with the pin height (length). The tool shoulder
contacting with the workpiece surface depends on the insertion depth of pin, which resulted in generation of welds with inner
channel, surface groove, excessive flash or local thinning of the welded plates and so on. The FSW butt joint of AA5083
aluminum alloy is as shown in Fig. 17.1. The material of FSW tool is AISI 4140.
In this work, there are four major controllable factors, i.e., four-level rotation speed (550, 1,100, 1,250, 1,800 rev min 1),
transverse speed (53, 90, 143, 180 mm min 1), tool tilt angle (1 , 2 , 3 , 4 ) and pin tool length (2.5, 2.7, 2.9, 3.1 mm) were
chosen as shown in Table 17.1. Each FSW butt joint was cut to be five pieces of specimens as shown in Fig. 17.2 for tensile
test based on ASTM standard. The tensile tests were carried out by Instron 8801 Universal Testing Machine, and taken their
Fig. 17.1 FSW butt joint of AA5083
Table 17.1 Process parameters and their levels

Process parameter Unit Level 1 Level 2 Level 3 Level 3
Rotation speed; A rev min 1 550 1,100 1,250 1,800
Transverse speed; B mm min 1 0.9 1.5 2.4 3.0
Tool tilt angle; C Degree 1 2 3 4
Pin tool diameter; D mm 2.5 2.7 2.9 3.1
17 Microstructure in the FSW Butt Joint of Aluminum Alloy AA5083 145
Fig. 17.2 Configuration and 100

size of the tensile specimens
16
50
12
18
Welded
R8 Unit: mm
Table 17.2 Selected data of the Trial no. Ultimate tensile strength (MPa) Elongation rate (%)
ultimate tensile strength over
300 MPa 1 306 10.4
2 319 11.3
3 321 13.4
4 303 9.7
5 320 12.4
loading and elongation record of specimens. Finally, the ultimate tensile strength (UTS) and elongation (EL) can be
calculated on the bases of their fracture loading and elongation of specimens. Five best UTS over 300 MPa and their
elongations were shown in Table 17.2.
The specimens for Transmission Electron Microscopy (TEM) are chosen from the cases of the ultimate tensile strength of
FSW butt joint over 300 MPa, which is about 86% ultimate tensile strength of the base material. The TEM specimens were
taken 1 mm thin piece from the welded zone of FSW butt joint along the transversal direction to the welding line, then
ground by abrasive paper to 0.2 mm thin and thinner by ion miller. The micro-structural observation in the welded zone was
taken by PHILIPS CM200 TEM with 200 kV.
17.3 Results and Discussions
TEM observation in the welded zone shows the structures of AA5083 aluminum alloy which are recrystallized grains after
hot machining. The sizes of the most grains are around 0.5–3 mm and bright field image as shown in Fig. 17.3a. The dark field
image is as shown in Fig. 17.3b. Small grains locate inside surrounding by large grains or beside the edge of the relatively
large grains. Some of the small grains are smaller than 0.1 mm. The most of small grains are about 0.1–0.2 mm. An oval-
shaped zone with its size about large grain is shown in the bottom of Fig. 17.3a. The selected area diffraction image for the
oval-shaped zone is as shown in Fig. 17.3b, which indicates the amorphous zone. The atoms disorder in this zone. Some
broken holes in the amorphous zone after surface tension and electron beam loading were easy and quick to appear. This is a
characteristic as shown in the amorphous zone. Another grain next to the amorphous zone appeared as shown in Fig. 17.3c
from a little oblique specimen. In the dark field image, many bright and dark bands were clear and spread in the large grains.
Another bright field image in the same specimen for the amorphous zone is as shown in Fig. 17.4a. The grains look like
being differentiated by dislocated zone and like relative small sub-grains. The dark field image is as shown in Fig. 17.4b and
<001> zone diffraction image as shown in the left-top. Two bright field images are in the same location as shown in
Figs. 17.4a, c, but another grain in <111> zone. In Fig. 17.4, boundaries of grains are not definable, that is uncommon
phenomenon. This grain in <111> zone has dark field image as shown in Fig. 17.4d and looks like a ripple boundary in the
grain. In the bright field images and in the amorphous zone as shown in Figs. 17.4a, c, a little bit more bright points in
Fig. 17.4c than in Fig. 17.4a can be clearly seen. The selected area diffraction image from an indistinct ring is an obvious
bright circle as shown in the left-top of the Figs. 17.4a, c. After analysis confirmed, this obvious bright circle came from
diffraction of many aluminums metal {111} crystal plane. This indicates that some broken holes in the amorphous zone
under surface tension and electron beam loading are gradually magnifying and recrystallized.
Based on the above TEM analysis; these images indicate that aluminum alloy in the FSW welded zone experienced
severe plastic movement and dynamic recrystallized grains after hot machining. Perhaps low temperature in the welded zone
and good heat conductivity aluminum alloy resulting recrystallized process was frozen, so that two different sizes of grains
produced and incomplete amorphous zone generated. Adhesive strength of the FSW aluminum alloy joint is based on the
Fig. 17.3 TEM micrographs in welded zone: (a) bright field image, (b) dark field image and (c) small polygonal grains surrounding oval-shaped
amorphous in <112> zone axis
Fig. 17.4 Another TEM micrographs of the same specimen using in Fig. 17.3: (a) bright field image for the amorphous zone, (b) dark field image
in <001> zone axis, (c) another grain bright field image at the same location and (d) dark field image in <111> zone axis
17 Microstructure in the FSW Butt Joint of Aluminum Alloy AA5083 147
heat effect and the sizes of grains in welded zone. However, improper FSW operation may cause defects in the weld zone.
In low temperature, the FSW grains tend to be small with relative narrow heat effected zone which is the positive effect to
the FSW joint strength.
17.4 Conclusions
The specimens for TEM are chosen from the cases of the ultimate tensile strength of FSW butt joint over 300 MPa and larger
than 86% of the base material. TEM observation in the welded zone shows the structures of AA5083 aluminum alloy which
are recrystallized grains after hot machining. The sizes of the most grains are around 0.5–3 mm. Small grains locate inside
surrounding by large grains or beside the edge of the relatively large grains. The most of small grains are about 0.1–0.2 mm.
The atoms disorder in the amorphous zone.
Acknowledgement The paper is sponsored by the National Science Council (grant no. 99-2221-E-268-001) and the National Center for High
Performance Computing of the Republic of China.
References
1. Thomas WM, Nicholas ED, Need Ham JC, Murch MG, Templesmith P, Dawes CJ (1995) Friction stir butt welding, GB Patent Application
91259788 (1991). US Patent 5460317 (1995)
2. Chao YJ, Liu S, Chien CH (2008) Friction stir welding of AL 6061-T6 thick plates: part 2 – numerical modeling of the thermal and heat
transfer phenomena. J Chin Inst Eng 31(5):757–767
3. Zhou CZ, Yang XQ, Luan GH (2005) Fatigue properties of friction stir welds in Al 5083 alloy. Scr Mater 53(10):1187–1191
4. Chen TP, Lin WB (2010) Optimal FSW process parameters for interface and welded zone toughness of dissimilar aluminum-steel joint. Sci
Technol Weld Join 15(4):279–285
5. Chen TP (2009) Process parameters study on FSW joint of dissimilar metals for aluminum-steel. J Mater Sci 44(10):2573–2580
6. Dixit V, Mishra RS, Lederich RJ, Talwar R (2007) Effect of initial temper on mechanical properties of friction stir welded Al-2024 alloy. Sci
Technol Weld Join 12(4):334–340
7. Peel M, Steuwer A, Preuss M, Withers PJ (2003) Microstructure, mechanical properties and residual stresses as a function of welding speed in
aluminium AA5083 friction stir welds. Acta Mater 51(16):4791–4801
8. Rodrigues DM, Leitao C, Louro R, Gouveia H, Loureiro A (2010) High speed friction stir welding of aluminium alloys. Sci Technol Weld Join
15(8):676–681
9. DebRoy T, Bhadeshia HKDH (2010) Friction stir welding of dissimilar alloys – a perspective. Sci Technol Weld Join 15(4):266–270
10. Chao YJ, Qi X, Tang W (2003) Heat transfer in friction stir welding – experimental and numerical studies. J Manuf Sci Eng 125(1):138–145
11. Rodrigues DM, Loureiro A, Leitao C, Leal RM, Chaparro BM, Vilaca P (2009) Influence of friction stir welding parameters on the
microstructural and mechanical properties of AA 6016-T4 thin welds. Mater Des 30(6):1913–1921
12. Cui GR, Ma ZY, Li SX (2009) The origin of non-uniform microstructure and its effects on the mechanical properties of a friction stir processed
Al-Mg alloy. Acta Mater 57(19):5718–5729
13. Mahmoud TS, Gaafer AM, Khalifa TA (2008) Effect of tool rotational and welding speeds on microstructural and mechanical characteristics
of friction stir welded A319 cast Al alloy. Mater Sci Technol 24(5):553–559
Chapter 18
Load History Dependency of Plastic Collapse Behavior
of Cracked Piping
R. Suzuki, M. Matsubara, M. Suzuki, T. Shiraishi, S. Yanagihara, K. Sakamoto, and S. Izawa
Abstract The effects of the loading history and the notch interval on the plastic collapse behavior and the plastic collapse
strength were investigated for SUS304 steel pipe with two notches perpendicular to the pipe axis direction. The notched
SUS304 steel pipe was coated with photo-elastic material. The different combined axial tensile and bending loads were
applied to the specimens. The photo-elastic fringes were observed during the loading test. The evolution of the plastic region
in the ligament differed depending on the loading history and the notch interval. On the other hand, the plastic collapse
strength depended on the loading history and did not depend on the notch interval.
Keywords Plastic collapse • Stainless steel pipe • Combined of tension and bending
18.1 Introduction
The stainless steel is widely used for piping as in the light water nuclear reactor. The structural integrity evaluation of the
cracked stainless steel pipe is generally performed using the plastic collapse as a failure criterion, because the significant
plastic deformation occurs before the plastic collapse of the ligament. A single-edge cracked member subjected to combined
tension and bending is a typical structure on the structural integrity evaluation. The complex cracked structure is treated as
the single-edge cracked member in the many cases. The fracture mechanics studies have been performed for the single-edge
cracked member [1–3]. The plastic collapse strength depends on the loading history [4], although the loading history
dependency cannot be considered on general rigid-perfectly or elastic-perfectly plastic analyses. The loading history
dependency of the plastic collapse behavior should be clarified experimentally to evaluate accurately structural integrity.
Recently, the authors developed the statically indeterminate fracture mechanics equipment which can apply arbitrary
combined axial tensile and bending loads to the structural member [5]. In previously, the loading history dependencies of the
plastic collapse behavior and plastic collapse strength were reported for the stainless steel rectangular member and pipe
with single through-wall notch [5–8]. In this study, the plastic collapse behavior is investigated for SUS304 steel pipe with
the through-wall notches perpendicular to the pipe axis direction subjected to combined axial tensile and bending loads using
the photo-elastic coating technique. In addition, the effects of the loading history and the notch interval on the plastic
collapse behavior and the plastic collapse strength are discussed.
18.2 Theoretical Collapse Limit
The stress distribution in the ligament area of the notched pipe at collapsing is shown in Fig. 18.1. The pipe is subjected to
combined axial tensile and bending loads and assumed the elastic-perfectly plastic body [6]. The corrected membrane, sm,
and bending, sb, stresses of the ligament area at collapsing are expressed as
R. Suzuki (*) • M. Matsubara • M. Suzuki • T. Shiraishi • S. Yanagihara • K. Sakamoto

Gunma University, Tenjin-cho 1-5-1, Kiryu, Gunma, Japan
e-mail: r_suzuki@gunma-u.ac.jp
S. Izawa
Oyama National College of Technology, Nakakuki 771, Oyama, Tochigi, Japan
150 R. Suzuki et al.
Fig. 18.1 Schematic

illustration of the stress σy
distribution in the ligament φ
area of the pipe at collapsing θ Rm
O O M
F
YZ
Z β Z
σy
Fig. 18.2 Theoretical 3.5

Theoretical Collapse Limit
Corrected Membrane Stress/Yield Stress, σb/σy

collapse limit curve of a single
Pure Bending
notched pipe with notch angle 3
90 and the loading patterns
ng
2.5 ndi
Be
+
n
Tension Bending io
2 ns
Te
1.5
1 Bending Tension
0.5
Pure Tension
0
0 0.5 1 1.5
Corrected Membrane Stress/Yield Stress, σm/σy
sm p y 2b
¼ ; (18.1)
sy p y
sb 2f2sinb siny Y Z ðp y 2bÞgR2m t

¼ ; (18.2)
sy Ze
where y is the half notch angle, b is the natural axis angle, Rm is the mean radius of the pipe, t is the pipe thickness and sy is
the yield stress of the pipe material. The section modulus of the notch tip, Ze, and the distance between the gravity centers of
cross sections of the pipe with and without notch, YZ, are shown in the next equations, respectively.

R3m t 1 2sin2 y
Ze ¼ p y sin2 y ; (18.3)
Rm cos y Y Z 2 p y
Rm siny
YZ ¼ : (18.4)
p y
The corrected bending stress is plotted as a function of corrected membrane stress in Fig. 18.2. The solid line shows
calculated theoretical collapse limit of the single notched pipe with notch angle 90 .
18.3 Experimental Method
A specimen with 200 mm length was machined from SUS304 steel pipe with 32 mm diameter and 1.5 mm thickness
(Fig. 18.3). Two through-wall notches with notch angle 90 were cut in the specimen perpendicular to the pipe axis direction
at notch interval, L, from 0 to 56 mm by a wire saw. Here, the notch interval, L ¼ 0, shows the single notch. The notch tip
radius was 160 mm and the actual shape of the notch differed with theoretical shape shown in Fig. 18.1.
The notched specimen was coated with a photo-elastic material near the ligaments to observe the progress of the plastic
region of the ligaments. The photo-elastic material was composed of 50 mass% epoxy resin (EPIKOTE), 45 mass%
18 Load History Dependency of Plastic Collapse Behavior of Cracked Piping 151
Fig. 18.3 Schematic A-A Section

illustration of the specimen (Ligament Area)
with two notches of notch A
angle 90 perpendicular to the A
pipe axis direction 90
32
1.5
A
L
200 A
L=0, 1, 10, 12.5, 15, 20, 25, 28 and 56 mm
Fig. 18.4 Schematic SS400 Bar Photoelastic Coating

illustration of the specimen
Joint
attached to the statically
indeterminate fracture Spring
mechanics testing equipment
Tension Tension
4 Point Bending
Specimen
140
384
polysulfide rubber (Thiokol), and 5 mass% polyamine and the coating thickness was 2 mm. The first fringe was set to equal
the yield point (0.2 proof stress) in tension test [6]. The relationship between principal stress distance of the pipe material,
s1–s2, and the maximum share stress, tmax, on the first fringe is expressed as
s1 s2 ¼ 2tmax ¼ 2k; (18.5)
where k is the share yield stress and equals to 0.5sy. Tresca yield criterion is satisfied on the first fringe. Thus, the ligament is
fully yielded when the first fringe occurred at the notch tip (tensile fringe) penetrates to the ligament back or the tensile
fringe connects with compressive fringe occurred at the ligament back. We defined these penetrating and connecting points
of the fringe as the plastic collapse point.
The springs and SS400 bars were placed inside the coated specimen to prevent local collapsing at the transverse loading
points (Fig. 18.4), and the specimen was attached to a statically indeterminate fracture mechanics testing equipment
(Fig. 18.5).
The equipment can apply arbitrary combined axial tensile and bending loads to the specimen. The loading patterns are
shown in Fig. 18.2. The experiments were carried out for five loading patterns as follows:
1. Simultaneous axial tensile and bending loads: Tension + Bending
2. Tension load followed by bending load: Tension ! Bending
3. Bending followed by tensile load: Bending ! Tension
4. Tensile load: Pure Tension
5. Bending load: Pure Bending
Three loads of tension + bending, tension ! bending and bending ! tension applied to pass through the target point
sm/sy ¼ 0.5 on the theoretical collapse limit curve. The change of the photo-elastic fringe pattern was observed during the
loading test.
The photographs of the photo-elastic fringe patterns at collapsing for tension + bending, tension ! bending and bending
! tension loading tests are shown in Fig. 18.6. The compressive fringe was not observed for tension + bending and
bending ! tension loads tests before collapsing and the tensile fringe penetrated to the ligament back. On the other hand,
Fig. 18.5 Photograph of the statically indeterminate fracture mechanics testing equipment and the photo-elastic experimental system
Fig. 18.6 Photographs of the photo-elastic fringe patterns at collapsing for the various loading patterns and the notch intervals
a Green Type b Rice Type c Hundy Type

Notch Notch Notch
Slip Line Slip Line Slip Line

M M M M
F F F F
Fig. 18.7 Schematic illustrations of the slip line fields
Fig. 18.8 Corrected collapse strength against the notch interval
the tensile and compressive fringes were connected for tension ! bending loading test at collapsing. The fringe patterns
outside of the two notches were similar to the single notched pipe at L 28 mm, although fringe patterns inside of the
notches were complex shape and not similar to the single notched pipe. On the other hand, the compressive fringe was not
observed for all loading pattern at L ¼ 56 mm. In addition, the fringes occurred at the two notch tips were not connected
before collapsing only at L ¼ 56 mm.
Green’s slip-line field for pure bending, Rice’s slip-line field for axial force and bending and Hundy’s slip-line field for
axial force are shown schematically in Fig. 18.7a, b, c, respectively [9–11]. These are slip-line fields of the upper bound
solution for single edge cracked or notched member based on rigid-perfectly plastic body. We previously reported that the
trajectories of the top of first fringe were conformed the above slip-line fields geometrically [8].
In the case of 0 L 28 mm, the fringe patterns were close to the shape of slip-line fields intermediate between Green
and Rice types for tension ! bending loading test. This is because that the plastic collapse was occurred under bending
dominant. For tension + bending loading test, the fringe patterns were close to the shape of Rice’s slip-line field. For
bending ! tension loading test, the fringe patterns were close to the shape of the slip-line field intermediate between Rice
and Hundy types.
On the other hand, in the case of L ¼ 56 mm, the fringe patterns were close to the shape of slip-line field intermediate
between Rice and Hundy types for all loading patterns. These results show a two notched pipe with L ¼ 56 mm loaded at
sm/sy ¼ 0.5 is collapsed under dominant tensile load independently of the loading history.
The plastic collapse strength by the photo-elastic coating technique is plotted as a function of notch interval in Fig. 18.8.
The plastic collapse strength in bending ! tension, ●, tension ! bending, ♦, and tension + bending, ■, is higher than that
of theoretical collapse limit and dependent on the loading history. The plastic collapse strength for bending ! tension
Fig. 18.9 Corrected bending

stress against the corrected
membrane stress at collapsing
loading test was lowest than other two loading patterns. On the other hand, the plastic collapse strength was not dependent on
the notch interval.
The corrected bending stress is plotted as a function of corrected membrane stress at collapsing in Fig. 18.9. Here, the
symbols ~ and ▾ show the results of pure bending and pure tensile loading tests, respectively. All experimental results are
higher than the of theoretical collapse limit. This is suggested that the structural integrity of the two notched pipe can
be evaluated simply and safely using the theoretical collapse limit of single notched pipe based on elastic-perfectly
plastic body.
18.5 Conclusions
The effects of the loading history and the notch interval on the plastic collapse behavior and the plastic collapse strength
were investigated for SUS304 steel pipe with two notches perpendicular to the pipe axis direction by the photo-elastic
techniques.
1. The plastic collapse strength was dependent on the loading history and little dependent on the notch interval. The lowest
plastic collapse strength was obtained for bending ! tension loading test.
2. The fringe patterns outside of the two notches were similar to the single notched pipe at L 28 mm, although fringe
patterns inside of the notches were complex shape and not similar to the single notched pipe. The progress of the
plastic region of the two notched pipe were similar to single notched pipe at L 28 mm. However, the progress of
the plastic region at L ¼ 56 mm were different with L 28 mm.
References
1. Matsubara M (2000) A parametric study on J-integral under pressurized thermal shock by using statically indeterminate fracture mechanics.
Nucl Eng Des 196:153–159
2. Okamura H, Takano T, Watanabe K (1972) Application of fracture mechanics to the analysis of statically indeterminate structure. Mech Behav
Mater 1:245–255
3. Okamura H, Watanabe K, Takano T (1975) Deformation and strength of cracked member under bending moment and axial force. Eng Fract
Mech 7:531–539
4. Hasegawa K, Kobayashi H (2004) Failure stresses for pipes with multiple circumferential flaws. ASME PVP2004-2712, 2004
5. Matsubara M, Izawa S, Hirao N, Busujima K, Koyama T, Machida K, Kawada D, Ohta A, Harizuka H (2005) Development of testing
equipment for studying statically indeterminate fracture mechanics. In: Proceedings of 2005 SEM annual conference & exposition on
experimental and applied mechanics, 2005
6. Izawa S, Matsubara M, Nezu K, Sakamoto K (2004) Plastic collapse evaluation on the notched stainless steel piping subjected to combined
tension and bending by photo-elastic coating. Key Eng Mater 270–273:2001–2005
7. Izawa S, Matsubara M, Hirao N, Busujima K, Koyama T, Machida K, Kawada D, Ohta A, Harizuka H (2007) Combined tension–bend system
for large deformation analysis. Exp Tech 31(5):41–45
8. Izawa S, Matsubara M, Hirao N, Busujima K, Koyama T, Machida K, Kawada D, Ohta A, Harizuka H (2004) Plastic collapse behavior of a
SUS316 single-edge notched member subjected to combined tension and bending using photo-elastic coating technique. JSEM 4(1):55–60,
In Japanese
9. Green AP (1953) The plastic yielding of notched bars due to bending. Q J Mech Appl Math 6:223–229
10. Rice JR (1972) The line spring model for surface flows. In: Proceedings of surface flaw symposium, ASME winter annual meeting, pp 71–184
11. Hundy BB (1954) Plane plasticity. Metallurgia 49:109–118
Chapter 19
Production of Metal Matrix Composites Using Thixoinfiltration
Techniques
M.H. Robert, L.M.P. Ferreira, A.F. Jorge, and E. Bayraktar
Abstract The work presents an innovative approach to produce metal matrix composites by using thixoforming techniques.
The proposed procedure is based on the infiltration of thixotropic semisolid metal into an assembly of loose particles of
reinforcing material. Experiments were performed using Al alloys as base metal and SiC and porous ceramic particles
(expanded clay known as cinasite) as reinforcement. Process parameters are investigated to observe their influence in the
quality of products. Results demonstrate that the process is flexible for the manufacturing of different distributions of
reinforcing agents, such as homogeneous or layered composites. Moreover, it is also shown that metallic chips can be used as
raw material, meaning an important possibility for materials recycling. It can be concluded that Thixoinfiltration is a simple
and low cost technique to produce different types of MMCs, with advantages upon techniques involving metal powders,
which usually require long milling and sintering periods. Thixoinfiltration can be also advantageous upon techniques
involving liquid processing, once lower temperatures are involved and so lower energy input, as well as less interface
reactions.
Keywords Metal matrix composite • Thixo-infiltration • Low cost manufacturing • Weight saving • SEM analysis
19.1 Introduction
Among the numerous processes presently available to produce particulate metal matrix composites [1], most is based either
in the powder metallurgy technology or liquid processing of the metal or alloy to which the reinforcing particles are added.
Both routes present inherent difficulties responsible for the high costs of such materials on one hand and for the heterogene-
ity and inconstant quality of the product on the other hand. Powder metallurgy methods involve in some cases highly reactive
powders, in general long mixing (particularly processes requiring high energy milling) and sintering times [2, 3]; above that,
particular attention must be paid to the internal porosity in the product.
The liquid route difficulties are related to the handling of high temperature melts, the dispersion of the reinforcing
particles in the liquid during addition operation and during freezing of the alloy, due to their segregation in the solidification
front. Another inconvenience is acceleration of interface reactions at high temperatures.
Alternative processing has been developed aiming to improve the technological aspects together with the metallurgical
quality of the product. Processes which increase the flexibility of combinations matrix/reinforcement are also welcome,
widening the engineering applications of such special class of materials.
With these perspectives in mind, the utilization of the metallic alloy in the semisolid state (SS), as a thixotropic fluid with
controllable flow behavior, has been investigated in the last two decades with relative success, in the so called vortex method
and its derivations [4, 5]. The thixotropic nature of the SS metal can promote entrapment of added particles preventing their
flotation or sedimentation. However, the operational aspect is not simple due to reactions of the metal with stirrers and the
difficulty to take stirrers out and to transfer the SS material given that its viscosity quickly increases as stirring is stopped.
M.H. Robert • L.M.P. Ferreira • A.F. Jorge

University of Campinas, São Paulo, Brazil
E. Bayraktar (*)
Mechanical and Manufacturing Engineering School, SUPMECA-LISMMA-Paris, Paris, France
e-mail: bayraktar@supmeca.fr
158 M.H. Robert et al.
However, the thixotropic flow behavior of SS metals with particular structural condition has motivated a new approach in the
field of MMCs production: thixoinfiltration – the SS metal is forced to infiltrate within free spaces among an assembly of
reinforcing particles (a preform). All of the results obtained so far show that the infiltration of thixotropic semisolid metal
into preforms of sintered or loose particles can produce composites of reliable quality at low costs for different kinds of
Al alloys and ceramic reinforcements [6–14].
Still another new approach in the manufacturing scenario in general has been strongly brought to light in the last years,
already becoming a mandatory issue: the development of sustainable technologies, as the utilization of recycled material and
techniques with reduced energy input. In this context, the use of SS metals fits perfectly: advantages of using SS instead of
liquid can be related to lower processing temperature, reducing operational costs and energy. Concerning product quality,
can result in better reinforcement distribution once segregation could be reduced as well as time for non desirable interface
reactions. Therefore, using SS a wider range of matrix/reinforcement combination can be investigated.
Advantages of SSM over PM can be related to higher operationally, once PM requires long mixing time, pre-compacting
operations and long sintering times. The present work aims to contribute to the development of new and sustainable
technology to produce MMCs; using SS metal obtained from cast material or recycled machining chips.
Al alloys with two different chemical compositions were used as base metal: Al6.5Zn2.5 Mg1.6Cu and Al2.5Cu0.4 Mg (wt%).
To analyze the thixoforming windows of such alloys, the solidus and liquidus temperatures (Ts and Tl) and the behaviour of
dfl/dT (fl ¼ liquid fraction) in the solidification range were determined using different methods: differential scanning
calorimetry – DSC, simple T t curve during cooling from liquid in practical conditions and thermodynamic simulation
using a commercial software. Results showed Ts and Tl at 476–637 C and 632–622 C respectively for AlZnMgCu and
AlCuMg alloys. AlZnMgCu alloy presents formation of small amount of liquid already at temperatures around 470–480 C
due to the melting of intermetallic phases, mainly the T phase AlCuMgZn and M phase MgZn2.
Variation of fl T for the most complex alloy from thermodynamic simulation shown in Fig. 19.1 indicates high dfl/dT
at higher temperature while a phase is melting; it means that at higher liquid fraction, as for example fl > 0.5, processing
control can be difficult.
For the thixoinfiltration tests, different liquid factions, from 0.2 to 0.5 were used. As reinforcing elements, SiC particles
and foamed ceramic granulates, known as cinasite were used, as loose particles. SiC particles used present dimension around
20–30 mm.
Cinasite is an expanded clay containing basically SiO2, Fe2O3, Al2O3, MgO; granules are approximately spherical with
dimensions around 1–3 mm. Figure 19.2 shows typical cinasite granules and the porous aspect of the interior, responsible for
the low density of the material: 0.85 g/cm3. It is expected, therefore, the resulting product to be a low density MMC.
Fig. 19.1 Variation of fl T of AlZnMgCu alloy according to thermodynamic simulation, using Scheil condition. AlCuMg alloy, on the other
hand, presents a narrow solidification range and more homogeneous variation (almost linear) of fl T within this interval. This small
solidification range can represent difficulties to thixoinfiltration process control
19 Production of Metal Matrix Composites Using Thixoinfiltration Techniques 159
Fig. 19.2 Porous ceramic granulates used in thixoinfiltration experiments to produce low density MMCs: (a) general view; (b) typical internal
aspect
Fig. 19.3 Schematic representation of different assemblies for thixoinfiltration tests: (a) compacted Al ships and SiC particles; (b) multilayer of
Al alloy plates and SiC particles; (c) thick layer of SiC in the bottom and Al alloy plate on top; (d) thin layer of SiC on top of Al alloy plate;
(e) sandwich of Al plates and cinasite particles in the middle
The proposed process of thixoinfiltration consists basically of the following steps: reinforcing particles and metal alloy
are placed in a die and heated to the pre-established temperature to promote partial melting of the alloy, the necessary
transformations to get the thixotropic state and the desired liquid fraction. Pressure is then applied to force the penetration of
the SS metal into the particles assembly. Pressure is kept for a while to improve reinforcement/matrix contact while the
material is cooling.
In this work different assemblies of loose particles/metallic alloy were tested, according to the schematic representation
shown in Fig. 19.3. In case (a) grinded machining chips of AlCuMg alloy were milled together with SiC particles, the
mixture was compacted, the green precursor was heated to work temperature and a pressure was applied with the metal in
the SS state. Cases (b), (c) and (d) are similar using Al alloy plates instead of chips and different layers/relative positioning of
the metal and the SiC particles. In case (e) cinasite was the reinforcement and AlZnMgCu the matrix.
Products were analyzed by optical and electronic microscopy; in the case of the MMCs containing macro particles as the
cinasite, tomography was used to evaluate internal quality.
19.3 Results
19.3.1 Operational Conditions Required
The success of the thixoinfiltration process depends on the ability of the SS metal to flow into the small free interstices
among particles of the reinforcement. Therefore, one important parameter is the viscosity of the SS which depends on the
liquid fraction and size of globular solid in the thixotropic material; this must be compatible with the spaces to be filled.
For the conditions tested in this work, results showed that liquid fractions from 0.2 to 0.5 can be used; size of globular solid
in the SS ranging from 20 to 100 mm where tested and successfully infiltrated. Complete filling of the mould and
Fig. 19.4 AlCuMg-SiC MMCs produced by thixoinfiltration of different alloy/reinforcement configurations: (a) compacted Al ships and SiC
particles; (b) multilayers of Al alloy plates and SiC particles; (c) thick layer of SiC in the bottom and Al alloy plate on top (OM)
reinforcement incorporation could be achieved in most of the cases; as pressure as maintained during solidification of the SS
metal, no defects due to difficulty of infiltration were detected.
Dimensions of globules in the SS alloy could be varied by means of previous deformation in the raw material (cast alloy)
to stimulate the recrystallization route to promote the spheroidization of the structure during heating to SS condition.
Required pressure to promote total infiltration ranged from 2 to maximum 10 MPa, depending on the liquid fraction in the
SS metal, decreasing the pressure as fl increases. These values are low enough to prevent any damage of the reinforcing
particles during processing. In the case of the product containing macro particles of expanded clay, infiltration pressure
required for full infiltration ranged from 7 to 10 MPa, increasing as granules size decreases (small free spaces to be filled than
in the case of coarse granules).
19.3.2 Obtained Products
Figure 19.4 shows AlCuMg-SiC MMCs produced by thixoinfiltration according to different alloy/reinforcement assemblies
as shown previously: (a) compacted Al ships and SiC particles; (b) multilayer of Al alloy plates and SiC particles; (c) thick
layer of SiC in the bottom and Al alloy plate on top.
In the first case, using a mixture of compacted Al chips and SiC powder, an acceptable distribution of the SiC in the
matrix, without any significant agglomeration, is achieved. It can be observed that the matrix is continuous, without evidence
of the initial condition as chips, no interfaces are observed in the matrix. It means that total cohesion of the chips was
provided by the applied pressure in the SS state; oxide layers probably suffered rupture in the process, allowing the distinct
chips to get in contact. Some presence of oxide can be observed, but not as continuous films. Porosity is low, apparently
lower than the usually resulted from typical MP processing; the applied pressure in the SS state was effective to promote
voids collapse. Indeed, some orientation of the structure can be observed, as a consequence of the uniaxial direction of the
applied pressure.
Fig. 19.5 AlCuMg-SiC layered MMC produced by thixoinfiltration; layer of SiC at the top, plate of alloy in the bottom as (d) in Fig. 19.3 (SEM).
(a) General view; (b) and (c) details of the SiC/Al interface
In case (b) when using multi layers of SiC particles and Al plates, a total integration of the constituents is achieved in the
SS state. However, a stronger orientation of the structure can be observed as well as some agglomeration of SiC particles,
resulting in a more heterogeneous distribution of the reinforcing particles. Again low porosity is observed due to the pressure
applied in the SS state.
In case (c) when using a thick layer of SiC in the bottom and only one Al alloy plate on top, resulting product shows also
total integration of the constituents; good distribution and of SiC particles and low porosity level. The microstructure
observed suggests the movements of the SSM flow downwards, promoting the distribution of the reinforcing particles. No
oxide films are observed in the metal flow front; probably the oxide layer present in the surface of the SS metal was broken in
the process due to the pressure applied against the hard SiC particles.
Considering the good results obtained when using high volume fractions of SiC particles to produce bulk composites,
the thixoinfiltration process was also tested to produce a material with specific, localized presence of reinforcements.
In the experiment, a thin layer of SiC was placed on top of an Al alloy plate to promote an upwards flow of the SS metal into
the reinforcing particles layer.
Result is shown in Fig. 19.5; it can be observed reasonable homogeneity of reinforced thickness layer; the unidirectional,
upwards movement of the SS material against the SiC layer in contact with the piston did not allowed the reinforcing
particles to spread throughout the matrix. Porosity can be observed within the composite layer; in this case probably the
applied pressure is less effective to promote SS flow and infiltration than in the previous cases, where the pressure was
applied directly over the SS metal.
Porosity can probably be avoided if better pressing and fl conditions are used, preferably by lowering viscosity of SS
(increasing T, for example) as higher P could cause disturbance in the SiC layer. No oxide films are observed in the SS flow
front; again probably due to rupture during infiltration processing.
Fig. 19.6 (a) General aspect of composite produced by thixoinfiltration of AlZnMgCu alloy into a layer of expanded ceramic granulates,
according to assembly (e) in Fig. 19.3; (b) microstructure of metallic matrix; (d) detail of interfaces metal/reinforcement
The SS metal ability to flow within small spaced among particles, resulting in acceptable composites, can also be used to
produce new materials such as low densities composites (LD-MMCs). In this case, porous reinforcements can be utilized and
the product can present characteristics of both MMC and cellular materials, it meaning good wear properties, for instance,
associated with low weight, high ability to absorb energy in compression, and good thermal and sound insulation.
Figure 19.6 presents typical results of a LD-MMC produced by thixoinfiltration of SS AlZnMgCu into a layer of loose
cinasite granules; assembly used in the experiments as indicate in Fig. 19.3e, with the granules of the porous ceramic
material between two Al plates.
A good and macroscopically homogeneous dispersion of granulates could be achieved; tomography analysis showed
no infiltration defects or excessive agglomeration of granulates in the whole volume of the product, as observed in Fig. 19.7,
which shows examples of images of consecutive and parallel sections of the product (distance between sections: 1.5 mm).
It can be observed in all cases, a homogeneous distribution of the ceramic porous particles throughout the metallic matrix.
The success of the infiltration depends on the proper condition in the SS metal, it meaning appropriate flow behavior resulted
from proper microstructure – size of globules in the slurry compatible with spaces to be filled and fl.
For obtaining more detail on the microstructure, typical microstructure of the metallic matrix is presented in Fig. 19.7a: a
globular, typical thixotropic structure can be observed. Liquid fraction was estimated as 0.2 in this case, and average size of
globular solid (a phase) around 90 mm. Those conditions were adequate to promote full infiltration in the practical conditions
employed.
Interfaces between ceramic particle and metal matrix are not different than usually produced by liquid processing – the
wetting problem between two materials of distinct natures is a major problem. Different kinds of surfaces were observed:
from apparently excellent interaction to a visible detachment of the granules. The solidification under pressure can promote
better physical contact and even promote penetration of liquid in eventual defects in the surface of the particles, as observed
in Fig. 19.7d. This penetration promotes a mechanical attachment of matrix to reinforcing particles, and may be important
for the composite properties.
Relative densities of MMCs with porous ceramic reinforcement produced vary from 0.35 to 0.6, increasing as dimensions
of the ceramic particles increases. The density of the products lies within or close to that considered typical for cellular
materials [8].
As a lightweight porous reinforcement, the expanded ceramic particles contribute to the low density of the produced
material; on the other side, the high stiffness and brittle characteristic of such particles can lead to higher stress levels for
plastic deformation in compressive conditions than those required for purely cellular material.
Fig. 19.7 Sequence of tomographic images of parallel sections of AlZnMgCu alloy reinforced with porous ceramic granulates, produced by
thixoinfiltration
19.4 Conclusions
Results show that thixoinfiltration can be an interesting alternative for the production of different kinds of composites, bulk,
surface layered and even low density composite, when porous ceramic reinforcements are used. Advantages over liquid or
PM routes can be suggested: lower temperatures, pressures and operational time involved; higher flexibility of metal/
reinforcement combinations and higher flexible concerning architectures.
Due to the high fluidity of thixotropic material, near net shape products can be directly manufactured. The process can
also be applied for the manufacture of noble products as composites and low density composites by using directly disposable
material from machining shops – chips – without special pre-treatments. Therefore, the production of MMCs via SS alloys
worth to be developed in all aspects, being the most important feature to be followed investigated the metal/reinforcement.
The complete assessment of the technology can result in new engineering products that fulfill requirements of a sustainable
development.
Acknowledgements The authors want to thank financial support from agencies CAPES and CNPq.
References
1. Chawla N, Chawla KK (2006) Metal matrix composites. Springer Science and Business Media Inc., Berlin
2. Suryanarayana C (2001) Mechanical alloying and milling. Prog Mater Sci 46(1–2):1–184
3. Fogagnolo JB, Velasco V, Robert MH, Torralba JM (2003) Effect of mechanical alloying on the morphology, microstructure and properties of
Al matrix composite powders. Mater Sci Eng 342:131–143
4. Hashim J, Looney L, Hashmi MSJ (1999) Metal matrix composites: production by the stir casting method. J Mater Process Technol 92–93:1–7
5. El-Sayed YK, Mahmoud TS, Ali AA (2011) On the electrical and thermal conductivities of Cast A356/Al2O3 metal matrix nanocomposites.
Mater Sci Appl 2:1180–1187
6. Silva RR, Robert MH (2009) Mechanical and metallurgical characterization of open cell porous Al alloys produced by Thixoforming into
removable space holders Pre-forms. In: Proceedings of METFOAM – 6th international conference on porous metals and metallic foams,
Slovakia, September 2009
7. Robert MH, Jorge AF, Gatamorta F, Silva RR (2010) Thixoinfiltration: a new approach to produce cellular and other low density metallic
materials. J Achiev Mech Eng Manuf 34(2):180–187
8. Ashby MF (2000) Metal foams: a design guide. Butterworth-Heinemann, Boston
9. Fuganti A, Cuginto G (2000) Thixoforming of aluminum alloys for weight saving of a suspension steering knuckle. Metall Sci Technol
18(1):19–23
10. Zoqui EJ, Lourençato LA (2008) Solid state phenomena, thixoforming of aluminum-silicon alloys in a mechanical eccentric press, Brasil,
UNICAMP, Vol. 141–143, pp 517–522
11. Chayong S, Atkinson HV, Kapranos P (2005) Thixoforming 7075 aluminum alloys. Mater Sci Eng A 390(1–2):3–12
12. Abramov VO, Straumal BB, Gust W (1997) Hypereutectic Al-Si alloys with a thixotropic structure produced by ultrasonic treatment. Mater
Des 18:323–326
13. Xia K, Tausig G (1998) Liquidus casting of a wrought aluminium alloy 2618 for thixoforming. Mater Sci Eng A 246:1–10
14. Wan G, Witulski T, Hirt G (1994) Thixoforming of aluminum alloys using modified chemical grain refinement for billet production. La Metall
Ita 86:29–36
Chapter 20
Use of Vision-Based Methods for Measurement of Mechanical
Properties of Exhaust Catalysts
Andrew P. Dickinson, Paul J. Gloeckner, Randy J. Stafford, and Kevin B. Golovin
Abstract Diesel engine exhaust catalyst material properties play an important part in predicting the durability of
aftertreatment systems. In particular, the elastic modulus is critical to the catalyst design. Direct measurement of catalyst
stiffness is difficult due to the low deformations and high stiffness that the catalyst samples experiences before failure.
For this paper, catalyst stiffness is measured using two vision-based methods – Digital Image Correlation and a point
tracking method developed in LabVIEW. Results of the testing show good correlation of the two methods. Additionally, the
testing identified contributions of the loadframe compliance that affected results of previous test methods.
Keywords Digital image correlation • Diesel engine catalyst • Catalyst material modulus
20.1 Introduction
Two vision systems were used to measure displacement of ceramic catalyst flexure specimens, one a commercially available
Digital Image Correlation (DIC) system from Correlated Solutions Inc. and the other was developed with the use of National
Instruments (NI) Vision and LabVIEW software. The DIC system consists of two cameras that when used together are
capable of measuring in-plane and out-of-plane motion, the system also provides whole field deformation information. The
NI system consists of a single camera capable of measuring in-plane motion; the system was developed with the use of NI
Vision software and LabVIEW. The displacement and load data recorded from each system was used to calculate the elastic
modulus (E) for each ceramic catalyst sample; this data was then compared to the calculated values using the load frame
crosshead displacement.
20.2 Background
Diesel engine exhaust catalyst materials are evolving and basic material property measurements are needed to determine if
the catalyst will be mechanically robust for the environment it is operating in. One important material property is the elastic
modulus. The elastic modulus can be calculated using ultrasonic, resonance, and mechanical techniques [1]. Each method
can calculate a different elastic modulus depending on the physical structure of the catalyst. The challenge for the material
researcher is to understand the differences and select the most appropriate technique for future evaluation. The selected
method impacts the development of the catalyst system.
The catalyst samples are ceramic in nature and during mechanical flexure testing fracture after small strains are
introduced; typically less than 1,500 microstrain (me). Due to compliance in the load frame the measured crosshead travel,
A.P. Dickinson (*) • P.J. Gloeckner • R.J. Stafford

Cummins Inc., 1900 McKinley Ave, Columbus, IN 47201, USA
e-mail: hj424@cummins.com, andrew.p.dickinson@cummins.com; paul.j.gloeckner@cummins.com; randy.j.stafford@cummins.com
K.B. Golovin
Cornell University, 1265 Helford Ln, Carmel, IN 46032, USA
e-mail: kbg39@cornell.edu
166 A.P. Dickinson et al.
which is used to calculate the elastic modulus, is large compared to the displacement of the sample at fracture. This
compliance needs to be determined and compensated for to make accurate measurements. Non-contact measurement
systems are ideal for these samples since the motion of the load frame components can be measured simultaneously.
20.3 Ceramic Catalyst Samples
The ceramic catalyst samples being measured are approximately 24 mm in thickness (t), 12 mm in height (w), 90 mm in
length, and have a wall thicknesses of less than 1 mm. Figure 20.1 illustrates a cross section of a catalyst sample. The
samples were cut from a larger block and were individually measured for modulus calculations.
The catalyst samples are designed for maximum flow and reaction surface while having sufficient strength to survive
harsh vehicle environments. Common failure modes of the catalyst are plugging with soot and fracture of the crosshatch
structure; the latter being related to the modulus of the sample.
20.4 Flexure Test Setup
The flexure test setup is illustrated in Fig. 20.2. The outer span (L) for the samples is 80 mm and the distance from the lower
support to the upper loading point (L1) is 20 mm. The fixture is designed to pivot and self-center to accommodate
irregularities in the samples. An MTS test system was used for testing with a data acquisition rate of 5 Hz and a crosshead
speed of 0.50 mm/min.
20.5 NI Vision Measurement Setup
The NI Vision hardware setup consists of an Allied Vision Marlin camera, Xenoplan lens, NI data acquisition hardware, and
computer. The camera was positioned on a tripod at a distance to contain the specimen and test fixture in the frame; filling the
frame with the area of interest will decrease the mm/pixel ratio improving the measurement resolution. The camera is
connected to the computer through a Firewire cable into a 1394b adapter card and configured using NI Measurement and
Automation Explorer. Load is recorded using the NI data acquisition hardware which configured using NI
Measurement and Automation Explorer.
Fig. 20.1 Illustration of

the cross section for the
ceramic catalyst samples t
P P
L1
Fig. 20.2 Illustration

of theflexure testing fixture L
20 Use of Vision-Based Methods for Measurement of Mechanical Properties. . . 167
Fig. 20.3 Catalyst sample in flexure fixture showing tracking locations 1–5
Within Measurement and Automation Explorer the camera is set to a frame size of 800 600 pixels allowing up to
15 frames per second (fps). The camera recording is set to continuous allowing acquisition rate to be controlled by the
LabVIEW program. An external trigger could have been used but acquisition rate timing was not critical.
NI Vision software was used to develop the program to track the desired points on the specimen. The first step in the NI
Vision software was to setup the calibration method; for this testing a calibration grid with 6 mm spacing between points was
used. The second step was to setup the points that need to be tracked. Using the NI Vision software five locations were
tracked on the flexure fixture and catalyst sample as shown in Fig. 20.3 as locations 1–5, in the LabVIEW front panel display.
Each point’s tracking options were set to allow for rotation of the pixel cluster and for subpixel accuracy. The third and final
step is to export the NI Vision program as a LabVIEW script. Once exported the script can be modified to acquire load,
program acquisition rate, configure processing, and save the desired output [2].
20.6 DIC Measurement Setup
The DIC hardware consists of two Allied Vision Dolphin cameras, Xenoplan lenses, LED light array, NI data acquisition
hardware, Correlated Solutions Inc. software, and computer. The camera was positioned on a tripod at a distance to contain
the specimen and test fixture in the frame; filling the frame with the area of interest will decrease the mm/pixel ratio
improving the measurement resolution. The camera is connected to the computer through a Firewire cable and load is
recorded using the NI data acquisition hardware. Triggering of the cameras is controlled by the software.
Before any acquisition can begin the samples and the fixture must be prepared for point tracking; this is done by speckle
painting the fixture and sample. For the fixture white paint is applied to the areas of interest then black paint is speckled on
with a brush. For the catalyst sample only blacks speckles are required as the samples are light colored. Figure 20.4 shows
the speckled pattern on the fixture and sample. The same calibration grid used for the NI setup is used for the DIC setup, but
multiple images of the grid, in multiple positions, are required for the calibration of the out-of-plane measurement.
Vic Snap software from Correlated Solutions Inc. was used to setup the acquisition, acquire images, and acquire load. Vic
Snap will save the load data into a separate data file at corresponds to the captured images; this file will be opened later in the
Vic-3D processing software. The first step is to acquire and process the calibration grid images; the software provides output
indicating the quality of the calibration. For this setup, which has a lot of working room, the calibration process is relatively
easy. With the calibration complete baseline images can be collected; these images are used for establishing the measure-
ment uncertainty and the quality of the setup.
Fig. 20.4 Image of the DIC

sample and fixture setup as
seen by the left camera
With images collected the Vic-3D software is used for processing. The first step is to create the areas of interest for
processing. With the areas selected the software needs inputs for grid size and offset, each which control how the software
looks for pixel matches between images. Smaller grid size and offset can be used with a smaller speckle pattern. After
selecting the areas of interest and setting the grip and step size the processing can start. The Vic-3D software has default
processing options which are sufficient for most acquisitions. If the processing is not adequate with the default settings they
can then be changed; for this work the default settings were adequate. After processing, the software outputs a report that
should be checked for changes in the projection error. Changes in projection error typically indicate changes in the setup
during acquisition, which lead to poor data quality.
With the processing complete the displacement data can be extracted. This process begins by first establishing the
coordinate system for the setup; for these samples the x-y plane is the face of the sample. Next any rigid body motion is
removed by selecting a single point as a reference point; this was done by selecting the lower left support as the zero point.
Now change in location information can be extracted and exported for the points of interest using a node data selection tool
in the software. The DIC processing is not automated but can easily provide whole field deformation information for the
sample and contour plots of the deformation.
20.7 Modulus Calculation
In four point bending the modulus (E) in Pa is calculated from

L1 3L2 4L21 dP
E¼ (20.1)
32tc3 dv
where L1 is the distance between the lower support and the upper loading point in m, L is the span in m, t is the sample
thickness in m, c is half the sample height in m, and dP/dv is the slope of the load deflection curve for the sample in N/m [3].
Figure 20.6 shows the load deflection curves for location five of purple sample one (a) and the center of purple sample 11 (b);
the slope of these curves are used in Eq. 20.1. Data from the NI and DIC systems can be easily curve fit for a calculation of
the slope where the data from the load frame must be corrected for machine compliance (Fig. 20.5).
20.8 Compliance Correction
Another method used to correct for the load frame compliance is to measure the load versus displacement for a given
material with known elastic modulus and subtract from it the theoretical load versus displacement relationship [4].
The difference is the load frame compliance which is then subtracted from the specimen load displacement measurement.
a 160 b 160
140 140
120 120
100 100
Force (N)
Force (N)
80 80
60 60
40 40
20 20
0 0
0 0.00005 0.0001 0.00015 0.0002 0.00025 0.0003 0.00035 0.0004 0 0.00005 0.0001 0.00015 0.0002 0.00025 0.0003 0.00035 0.0004
Displacement (m) Displacement (m)
NI Purple 1 MTS Purple 1 DIC Purple 11 MTS Purple 11
Fig. 20.5 Load deflection curves for: (a) purple one collected with the NI system and (b) purple 11 collected with the DIC system. The dash curve
is the data collected with the MTS system which shows the load frame compliance as the non-linear portion of the base of the curve and as a more
shallow slope, easily seen in (a)
For this testing a copper bar was used for the corrections. Figure 20.6 shows the theoretical, measured, and compliance
curves from the copper bar.
A total of 15 samples were tested, samples 1–10 were measured using the NI system and samples 11–15 were measured
using the DIC system. For each sample the deformation at the center of the specimen relative to the supports was used for the
modulus calculation. Figure 20.7 shows the calculated modulus for the 15 samples for the camera systems and the MTS
system. The average difference between the NI Vision measurements and the MTS system is 11%. The average difference
between the DIC system and the MTS system is 15%.
Each setup has its advantages and disadvantages. For the NI system the advantages are that the script and processing can
be specific for the application and the setup is easier and quicker since no speckle painting is required. Disadvantages for the
NI system are that there is no compensation for out-of-plane motion and only selected points are being tracked. For the DIC
system the advantages are that the software is already written, the system compensates for out-of-plane motion, and
processing allows for whole field deformation information. Disadvantages of the DIC system are that the setup takes
more time and that the software cannot be programmed to be specific for the application.
This evaluation of vision based methods to measure the displacement of the ceramic samples was conducted to determine
if the load frame compliance was impacting the calculation of the modulus and to compare to the compliance correction
method. The noncontact displacement measurements provided by the camera systems is assumed to be more accurate than
using the displacement data from the MTS system since any compliance is removed. From Fig. 20.7 it is easy to see that the
modulus calculation with compliance correction calculates a modulus higher than the vision based methods and higher than
the raw MTS data. With the assumption that the vision based methods are the more accurate method, the compliance
correction method is not a good approach to evaluate the modulus for these samples.
20.10 Future Work
To improve the modulus calculation the dP/dv term in Eq. 20.1 can be more accurately estimated by using the largest vertical
displacement along the entire sample length, which is not always the center of the sample. Figure 20.8 shows the line profile
of vertical displacement from sample purple 12 taken with the DIC system. The largest vertical displacement did not occur at
the center of the sample where the displacement was taken for the dP/dv calculation but to the left of center; the use of the
DIC system helped identify this. Using the NI system the script could be rewritten to track multiple points along the length of
the sample, find the point with the largest vertical displacement, and use that point for the dP/dv calculation.
500
450
400
350
Load (N)
300
250
200
150
100
50
0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
Displacement (mm)
Theoretical Measured Compliance
Fig. 20.6 Copper bar load displacement data used for load frame compliance correction with compliance correction curve
4.00
3.50
3.00
Modulus (GPa)
2.50
2.00
1.50
1.00
0.50
0.00
Purple 1 Purple 2 Purple 3 Purple 4 Purple 5 Purple 6 Purple 7 Purple 8 Purple 9 Purple 10 Purple 11 Purple 12 Purple 13 Purple 14 Purple 15
NI Vision DIC MTS MTS Corrected
Fig. 20.7 Comparison of catalyst sample calculated modulus for both camera systems
0
-10 0 10 20 30 40 50
Change in Vertical Displaecment (mm)
-0.05
-0.1
-0.15
-0.2
-0.25
-0.3
-0.35
Specimen Span (mm)
Fig. 20.8 Change in vertical displacement along the span length for sample purple 12
20.11 Summary
Two different vision based methods were used to measure the displacement of ceramic catalyst samples in four point
bending. The displacements were used to calculate the elastic moduli for the samples, then the moduli were compared to
moduli calculated for the load frame measurements and corrected load frame measurements. The corrected load frame
moduli were larger than the moduli calculated from the raw load frame data and the camera data; this correction approach is
not appropriate for these samples.
The raw load frame data had a discrepancy to the NI camera system by 11% and to the DIC system of 15%; more samples
for each system would need to be collected to draw conclusions on which system calculates the most accurate modulus.
From an application point of view each systems is very different. The NI system can be designed for the specific
measurement and is easier to setup where the DIC system has prepackaged software and the setup is more time consuming.
References
1. ASM Organization (2000) Introduction to mechanical testing and evaluation, vol 08, ASM handbook. ASM International
2. National Instruments Vision Concept Manual, National Instruments
3. Dowling Norman E (1999) Mechanical behavior of materials: engineering methods for deformation fracture and fatigue, 2nd edn. Prentice Hall,
Lebanon
4. Kalidini SR, Abusafieh A, EI-Danaf E (1997) Accurate characterization of machine compliance for simple compression testing. Exp Mech
37(2):210–215
Chapter 21
Review of Methods for Determining Residual Stresses
in Biological Materials
Drew Nelson
Abstract Over the years, a variety of methods for finding residual stresses have been developed in the field of experimental
mechanics. Adaptations of those methods by biomechanical researchers to the determination of residual stresses in various
biological materials are reviewed. Methods considered include deflections and strains from release of residual stresses,
X-ray diffraction, hole drilling, photoelasticity, indentation and membrane deflection.
Keywords Artery • Biological • Bone • Deflections • Hold drilling • Indentation • Membrane • Photoelasticity • Residual
stress • Strains • Virus • X-ray diffraction
21.1 Introduction
Experimental methods for determining residual stresses have been developed largely for application to metals. In recent
years, the existence, formation and role of residual stresses in biological materials has been studied by biomechanical
researchers. The purpose of this paper is to provide an overview of how methods developed for engineering components
have been adapted to study residual stresses in biological materials present in animals. It is hoped that this overview will be
of interest to researchers in experimental mechanics and potentially encourage development of new or improved methods for
determining residual stresses in biological materials. To keep the length of the paper reasonable, the overview will be limited
to naturally occurring residual stresses (as opposed to those that may develop as result of medical implants or dental
restorations).
21.2 Deflections Upon Release of Residual Stresses
A long standing approach for finding residual hoop stresses in thin-walled tubing involves making a longitudinal slit that
causes the tube to spring open upon release of residual stresses, as in Fig. 21.1. The bending moment released can be
estimated from a relation involving cross sectional moment of inertia, modulus of elasticity, Poisson’s ratio and change in
diameter [1]. The variation of hoop residual stresses through the wall can then be computed.
A similar experimental approach has been adopted in the biomechanics of arteries, but with computation of residual
stresses appropriate for biological materials. Fung [2] and Vaishnav and Vossoughi [3] found that when a short length of an
artery is dissected to produce a ring and then slit radially through the wall, it becomes a sector like that depicted in idealized
form in Fig. 21.2. Chuong and Fung [4] obtained a solution for hoop residual stresses based on measurement of “effective”
values of opening angle and associated radii to account approximately for the non-circular cross-sections of actual arteries.
Their approach considered the bending deformation needed to restore the sector to a circular shape and assumed homoge-
neous, cylindrically orthotropic, nonlinear elastic, incompressible material behavior. Compressive residual stresses were
predicted at the inner diameter. Including hoop residual strains in the computation of hoop stresses from simulated blood
D. Nelson (*)
Mechanical Engineering Department, Stanford University, Stanford, CA 94305-4021, USA
e-mail: dnelson@stanford.edu
174 D. Nelson
Fig. 21.1 Change in diameter

upon slitting to release hoop
residual stress
Fig. 21.2 Deflection of

an arterial ring specimen
after slitting
Fig. 21.3 Cross-section intima

of artery
media
adventitia
Fig. 21.4 (a) Two layer

representation of artery
with hoop residual stresses,
(b) bonded layers after slit,
(c) layers separated,
then slit [8]
pressure greatly reduced a sharp peak in tensile stresses at the inner diameter [4, 5], leading to a more uniform distribution of
hoop stress through an arterial wall. Some studies [6, 7] have found that a single radial cut may be sufficient to release
residual stresses, while others [8–11] have suggested otherwise. Additional discussion of the determination of residual
strains by radial slitting is given by Rachev and Greenwald [12]. Computation of residual stresses is more fully discussed by
Humphrey [13].
The actual structure of an artery wall has three layers, as depicted in simplified form in Fig. 21.3, with each layer having
different material properties. Finite element models have been developed to predict residual stresses from ring slitting data,
for example, one that treats each cylindrical layer as a fiber reinforced material [14]. An interesting extension of the radial
slitting approach was developed by Greenwald et al. [8]. As seen in Fig. 21.4a, an artery is considered to have two layers,
such as the media and adventitia. (The innermost layer (intima) in healthy arteries is typically thin and often neglected in
stress analyses.) The usual assumed “zero stress-state” is represented in Fig. 21.4b, while if the layers were “unbonded” prior
to slitting, the result of radial slitting would be different opening angles for each layer, as in Fig. 21.4c. This was investigated
experimentally by freezing a segment of an artery and inserting a rod in it, then removing material from the outer diameter by
turning on a lathe until the inner layer remained, or by supporting a segment within a sleeve and then reaming material from
the inner diameter until the outer layer remained. Radial slitting produced different opening angles for each cylindrical layer
as expected. The opening angle data were not used to predict residual stresses, though. Removal of material in this
experiment brings to mind the Sachs boring technique [15], but without the measurement of changes in length and outer
diameter for use in determining residual stresses.
21 Review of Methods for Determining Residual Stresses in Biological Materials 175
Fig. 21.5 Thin-walled tube with a strip cut to determine longitudinal residual stresses
Fig. 21.6 (a) Longitudinal strip cut from artery, then separated into (b) adventitia, (c) media, (d) intima [17]
Fig. 21.7 (a) Ring cut from artery, (b) slit ring, separated into (c) adventitia, (d) media, and (e) intima [17]
To estimate longitudinal residual stresses in thin-walled tubing, a tongue can be dissected as illustrated in Fig. 21.5 and
the resulting deflection measured, from which residual stresses are computed from the bending moment needed to restore the
strip to its original position [16]. A similar approach has been used for arteries, except that the strip is completely separated
from an artery [17, 18]. For example, as shown in Fig. 21.6a, a strip curls upon removal from an artery and data such as
change in length and curvature can be determined approximately. Similar data can be collected for each of the three layers
seen in Fig. 21.6b–d, which were surgically separated from the initial strip. Likewise, data can be collected for a ring such as
that in Fig. 21.7(a), slit radially as in Fig. 21.7(b), then separated into the three layers as in Fig. 21.7c–e. The information
obtained from multiple slitting has been used to develop an analytical model to determine the three dimensional state of
residual stress in each of three layers [21].
176 D. Nelson
Fig. 21.8 Strains measured

by gages at A, B and C upon
cutting of fibula [34]
Fig. 21.9 (a) Posterior and

(b) cranial views of bovine
vertebra [35]
Residual strains/stresses in arteries influence the size and shape of the opening produced when an artery is cut in coronary
bypass surgery [20]. Compressive residual strains/stresses enhance resistance to rupture [21]. The slitting technique has also
been used to study residual strains/stresses in veins [22], esophagus [23], stomach [24], intestine [25], heart [26], trachea
[27], airways in lungs [28], brain [29] and cartilage [30].
21.3 Strains Upon Release of Residual Stresses
The slitting (crack compliance) method [31, 32] determines the profile of residual stresses vs. depth from strains measured
adjacent to a slit as it is gradually deepened. An early version of this approach was applied to bone by Stanwyck et al. [33] in
the 1980s. A strain gage was applied in the longitudinal direction of a bovine metatarsal bone and a saw cut was made
perpendicular to that axis, 1 mm from the gage and 2 mm deep. The gage registered 180 me. The strain increased to –
280 me when the cut was deepened to 3 mm. No quantitative residual stress analysis was reported. About a decade later,
Tanaka and Adachi [34] conducted an experiment in which strain gages were applied in the longitudinal direction of a rabbit
tibia at locations A, B and C shown in Fig. 21.8. Severing the fibula induced a change in strains from partial release of
residual stresses in the statically indeterminate structure of the tibia and fibula. The authors [34] suggested that the measured
strains indicated release of compressive residual stresses on the lateral side of the tibia and tensile stresses on the medial side.
Residual stresses were not computed from the strains. In another experiment by Adachi et al. [35], biaxial strain ages were
applied to locations L and R in Fig. 21.9a of a bovine vertebra. Strains changes were measured first as the end and growth
plates in Fig. 21.9a were removed with a hand saw, then as regions 1 and 2 in Fig. 21.9b were removed in succession with a
rotary cutter. A simplified model was used to estimate residual stresses. Compressive and tensile residual stresses were
predicted to exist in cortical and cancellous regions, respectively, in both longitudinal and circumferential directions.
21.4 X-Ray Diffraction
The measurement of residual stresses by X-ray diffraction (XRD) is well-established for engineering materials with
crystalline structures [36]. Bone contains crystals of hydroxyapatite, enabling XRD to be applied. In 2000, Todoh et al.
[37] reported tensile residual stresses in the longitudinal direction of specimens cut from a bovine femur, except at a medial
Fig. 21.10 (a) Tibiofibula, (b) tibia with fibula removed, (c) tibia with epiphyses removed, (d) specimens cut from proximal, intermediate and
distal locations of (c), (e) each specimen from (d) cut longitudinally to produce specimens in the anterior, posterior, medial and lateral positions [38]
Fig. 21.11 Measurement locations in specimen taken bovine femur [40]
position, with an unclear trend in circumferential stresses. Tadano and Okoshi [38] measured longitudinal residual stresses in
specimens of rabbit tibiofibula that were cut in the sequence shown in Fig. 21.10a–e. The longitudinal stresses were
relatively small (less than 8 MPa in magnitude) and did not appear to vary much as a function of cutting sequence. On the
other hand, circumferential residual strains measured from (a) to (e) at medial and posterior positions were influenced
noticeably. (Circumferential residual stresses were not found because a specimen suitable for XRD calibration by four
bending could not be obtained in that direction in this study).
Yamada and Tadano [39] found tensile residual stresses (up to 162 MPa) in air dried specimens cut from bovine femurs
that were much higher than those in Todoh et al. [37], a study in which specimens were kept in a saline solution.
Circumferential residual stresses were smaller than longitudinal stresses (approx. 20 to +50 MPa depending on location)
[39]. Using synchrotron X-rays, Yamada et al. [40] measured longitudinal residual stresses in an air dried specimen cut from
a bovine femur. Measurements were made at the locations depicted in Fig. 21.11, resulting in the radial profile shown in
Fig. 21.12. The authors were aware that air drying may have influenced their data.
Using synchrotron X-rays, Almer and Stock [41] found compressive residual stress in the longitudinal direction of a
specimen cut from a canine fibula. They expressed a concern that the stresses may have been associated with dehydration of
the specimen. In a subsequent study [42], they found that when a specimen was transferred from formalin to a saline solution
and kept there for 8 h, residual stresses dropped from about 75 MPa to 10 MPa.
21.5 Hole Drilling
A standard method for determining residual stresses in engineered materials is based on measurement of strains
accompanying the release of stresses via introduction of a hole [43]. In the early 1970s, Wright et al. [44] performed a
pioneering study in which hole drilling was applied to explore residual stresses in bone. Commercially available strain gage
rosettes for holes of 1.6 mm diameter were used. (The depth of holes was not reported.) Transversely isotropic material
178 D. Nelson
Fig. 21.12 Longitudinal residual stresses vs. radial from location at different positions around the circumference of the specimen in Fig. 21.11
[40]
Fig. 21.13 (a) Image before ablation occurred at cross-hair, (b) overlaid images of cell borders before and 10 s. after ablation and (c) 20 s. after
ablation [45]
behavior was assumed and expressions derived relating measured radial strains to residual stresses. Details of the derivation
were not presented owing to length limitations of the paper. Tensile longitudinal residual stresses in the range of 1–16 MPa
were found in bovine femora, tibia and a radius.
Recently, Ma et al. [45] used a laser pulse to ablate a tiny hole (less than 1 mm in diameter) through a layer of cells about
6 mm thick. As illustrated in Fig. 21.13, sudden introduction of a hole caused recoil of adjacent cells. Image processing was
used to capture the displacements during recoil. Subsequently, Hutson et al. [46] presented a model to estimate the stresses
associated with the recoil behavior. In another version of the hole drilling approach, Varner and Taber [47] perforated
epithelia (thin cell sheets) with a specially formed tip of a micropipette, producing holes of about 200 mm diameter.
Observed changes in the shapes of the initially circular holes were used with finite element modeling to estimate stresses.
21.6 Photoelasticity
The use of photoelasticity to measure residual stresses in glass and transparent plastics is well established [48]. Aegeter and
co-workers [49, 50] have applied principles of photoelasticity and measured the retardance through the thickness of wing
imaginal discs of fruit flies. (Those discs are structures in the larvae of certain insects that develop into wings in adults.)
A small loading set-up was used to find the equivalent of stress-optic constants for the disc material. Care was taken to
account for changes in retardance from sources other than stress, such as variations in thickness or density of birefringent
molecules. Knowledge of how built-in stresses change during morphogenesis is of interest in developmental biology [49].
21.7 Indentation and Membrane Deflection Techniques
Matsumoto et al. [11] found that excision of tubular specimens of aortic arteries caused surfaces adjacent to a cut to develop
a surface with microscopic hills and valleys such as those as seen in Fig. 21.14. A scanning micro indentation tester was
developed to measure stiffness at 2 mm intervals across the surface. The stiffness in hills was more than triple that in valleys.
Residual stresses were estimated as the stresses needed to restore the surface to a flat configuration and were obtained from a
finite element model plus measurements of the topography and distribution of stiffness.
Residual stresses in thin membranes of engineered materials have been inferred from deflections induced by a
concentrated forced exerted, for example, by a profilometer [51] or nano-indenter [52].
The micro-indentation set-up depicted in Fig. 21.15 was developed by Zamir and Taber [53, 54] to gather force-
displacement and surface contour data for a two cell thick myocardium layer enveloping cardiac jelly, modeled as a thin
membrane on an elastic foundation. (Cardiac jelly is a gelatinous substance present during development of a heart.)
Microspheres (beads) of 6 mm diameter were used to monitor surface displacements. Residual stresses in the myocardium
layer were estimated from the force-displacement and surface contour data and an inverse finite element method.
Fig. 21.14 Surface

topography of a
100 100 mm section
perpendicular to the
circumferential direction of
a porcine thoracic aorta [51]
Fig. 21.15 Cross-section

of microindentation test set-up
showing myocardium layer
(MY), cardiac jelly (CJ),
piezoelectric transducer (PZT)
and charge-coupled device
(CCD) [53]
180 D. Nelson
Moving down in size, the force-displacement characteristics of a prohead (immature viral shell empty of DNA) have been
studied by Carrasco et al. [55] using an atomic force microscope. As seen in Fig. 21.16, the prohead had a cylindrical body
(radius and length of 20 and 52 nm, respectively) with spherical end caps (radii of 21 nm). The prohead was tested in two
orientations shown in Fig. 21.17 using an indenter with a parabolically shaped tip. Finite element modeling indicated the
highest stiffness for the upright orientation, while measurements showed the highest stiffness in the laid-down orientation.
Additional modeling suggested that the presence of a tensile lateral residual stress about 1.5 times larger in the cylindrical
body than in the end caps could account for observed stiffness behavior. An indentation study by Baclayon et al. [56] also
suggests the existence of residual stresses in another type of viral shell.
21.8 Discussion
The studies summarized here provide a sense for the challenges involved in finding residual stresses in biological materials.
For example, the material behavior is often anisotropic, or viscoelastic, or varying from layer-to-layer, or some combination.
Geometries are typically more complicated than in many engineering components. The residual stresses themselves may
change as a result of material no longer being alive.
There may be opportunities to try new approaches for determining residual stresses in biological materials or to improve
existing ones. For example, it does not appear that the hole drilling method has been applied to investigate residual stresses
in bone since one study in the early 1970s. It may be worth trying the method again given the advances in the computation of
residual stresses (including analysis for orthotropic materials). Slitting, the contour method, etc. might provide other
opportunities. It may also be possible to re-evaluate data from early studies making use of improvements in computational
capabilities that have taken place since the studies.
Fig. 21.16 Prohead [55]
Fig. 21.17 Indentation

orientations [55]
21.9 Summary
Residual stresses exist in blood vessels, organs, thin layers of cells and perhaps even in virus shells. Methods developed
originally for determining residual stresses in engineering materials have been adapted to find residual stresses in biological
materials in ways that account for their more complicated material behavior.
References
1. Standard practice for estimating the approximate residual circumferential stress in straight thin-walled tubing, E1928–07, ASTM, West
Conshohocken
2. Fung Y (1984) Biodynamics: circulation. Springer, New York, pp 54–60
3. Vishnav R, Vossoughi J (1983) Estimation of residual strains in aortic segments. In: Hall C (ed) Biomedical engineering II, recent
developments, Pergamon press, New York, pp 330–333
4. Chuong C, Fung Y (1986) On residual stresses in arteries. J Biomech Eng 108:189–192
5. Chaudhry H et al (1997) Residual stresses in oscillating thoracic arteries reduce circumferential stresses and stress gradients. J Biomech
20:57–62
6. Han H, Fung Y (1996) Direct measurement of transverse residual strains in aorta. Am J Physiol 270:H750–H759
7. Stergiopulos N et al (2001) Assessing the homogeneity of the elastic properties and composition of the pig aortic media. J Vasc Res
38:237–246
8. Greenwald S et al (1997) Experimental investigation of the distribution of residual strains in the artery wall. J Biomech Eng 119:438–444
9. Van Dyke T, Hoger A (2002) A new method for predicting the opening angle for soft tissues. J Biomech Eng 124:347–354
10. Taber L, Humphrey J (2001) Stress-modulated growth, residual stress, and vascular heterogeneity. J Biomech Eng 123:528–535
11. Matsumoto T et al (2004) Residual stress and strain in the lamellar unit of the porcine aorta: experimental analysis. J Biomech 37:807–815
12. Rachev A, Greenwald S (2003) Residual strain in conduit arteries. J Biomech 36:661–670
13. Humphrey J (2002) Cardiovascular solid mechanics: cells tissues and organs. Springer, New York
14. Holzapfel G, Gasser T, Ogden R (2000) A new constitutive framework for arterial wall mechanics and a comparative study of material models.
J Elast 61:1–48
15. Sachs G, Espey G (1941) The measurement of residual stresses in metal. The Iron Age 148, Sept. 18, pp 63–71
16. Sachs G, Espey G (1942) A new method for determination of stress distribution in thin-walled tube. Trans Am Inst Min Metall Eng
147:348–360
17. Holzaplfel G et al (2007) Layer-specific 3D residual deformation of human aortas with non-atherosclerotic intimal thickening. Ann Biomed
Eng 35:530–545
18. Wang L, Gleasson R (2010) A mechanical analysis of conduit arteries accounting for longitudinal residual strains. Ann Biomed Eng
38:1377–1387
19. Holzapfel G, Ogden R (2010) Modelling the layer-specific three-dimensional residual stresses in arteries, with an application to the human
aorta. J R Soc Interface 7:787–799
20. Cacho F, Doblare M, Holzapefl G (2007) A procedure to simulate coronary artery bypass graft surgery. Med Bio Eng Comput 45:8819–8827
21. Volokh K (2008) Prediction of arterial failure based on a microstructural bi-layer fiber-matrix model with softening. J Biomech 41:447–453
22. Xie J et al (1991) The zero-stress state of rat veins. J Biomech Eng 113:36–41
23. Zhao J et al (2007) Opening angle and residual strain in a three-layered model of pig oesophagus. J Biomech 40:3187–3192
24. Zhao J, Liao D, Gregersen H (2005) Tension and stress in the rat and rabbit stomach are location- and direction-dependent. Neurogastroenterol
Motil 17:388–398
25. Dou Y et al (2006) Longitudinal residual strain and stress–strain relationship in rat small intestine. Biomed Eng Online 5:37. doi:10.1186/
1475-925-5-37
26. Omens J, McCulloch A, Crisicone J (2003) Complex distribution of residual stress and strain in the mouse left ventricle: experimental and
theoretical models. Biomech Model Mechanobiol 1:267–277
27. Han H, Fung Y (1991) Residual strains in porcine and canine trachea. J Biomech 24:307–315
28. McKay K et al (2002) Zero-stress state of intra- and extraparenchymal airways from human, pig, rabbit and sheep lungs. J Appl Physiol
92:1261–1266
29. Xu G, Bayly P, Taber L (2009) Residual stress in the adult mouse brain. Biomech Model Mechanobiol 8:253–262
30. Setton L, Tohyama H, Mow V (1998) Swelling and curling behaviors of articular cartilage. J Biomech Eng 120:355–361
31. Prime M (1999) Residual stress measurement by successive extension of a slot: the crack compliance method. Appl Mech Rev 52:75–96
32. Cheng F, Finnie I (2007) Residual stress measurement and the slitting method. Springer, New York
33. Stanwyck T et al (1982) Studies on prestress in bone. Biorheology 19:301–306
34. Tanaka M, Adachi T (1994) Preliminary study on mechanical bone remodeling permitting residual stress. JSME Int J Soc Mech Eng A
37:87–95
35. Adachi T, Tanaka M, Tomita Y (1998) Uniform stress state in bone structure with residual stress. J Biomech Eng 120:342–347
36. Noyan I, Cohen J (1987) Residual stress: measurement by diffraction and interpretation. Springer, New York
37. Todoh M et al (2000) Polychromatic X-ray measurements of anisotropic residual stress in bone femoral bone. JSME Int J C Mech Sys
43:795–801
38. Tadano S, Okoshi T (2006) Residual stress in bone structure and tissue of rabbi’s tibiofibula. Biomed Mater Eng 16:11–21
182 D. Nelson
39. Yamada S, Tadano S (2010) Residual stress around the cortical surface in bovine femoral diaphysis. J Biomech Eng 132:044503-1–044503-4
40. Yamada S et al (2011) Residual stress distribution in the bovine femoral diaphysis measured by synchrotron. J Biomech Sci Eng 6:114–124
41. Almer J, Stock S (2005) Internal strains and stresses measured in cortical bone via high-energy X-ray diffraction. J Struct Biol 152:14–27
42. Almer J, Stock S (2007) Micromechanical response of mineral and collagen phases in bone. J Struct Biol 157:365–370
43. Standard test method for determining residual stresses by the hole-drilling strain-gage method, E837-08e1, ASTM, West Conshohocken
44. Wright T, Barnett D, Hayes W (1977) Residual stress in bone. In: Chang T (ed) Recent advances in engineering science, vol 8, Part III of Proc
Tenth Anniversary Meeting of Soc Eng Sci (held in 1973). Scientific Publishers, Boston, pp 25–32
45. Ma X et al (2009) Probing embryonic tissue mechanics with laser hole drilling. Phys Biol. doi:10.1088/1478-3975/6/3/036004
46. Hutson M et al (2009) Combining laser microsurgery and finite element modeling to assess cell-level epithelial mechanics. Biophys
J 97:3075–3085
47. Varner V, Taber L (2010) On measuring stress distributions in epithelia. In: Garikiptai K, Arruda E (eds) IUTAM symposium on cellular,
molecular and tissue mechanics, IUTAM Bookseries 16, Springer, New York, pp 45–54
48. Standard test method for photoelastic measurements of birefringence and residual strains in transparent or translucent plastic materials,
D 4093–95, ASTM, West Conshohocken
49. Nienhaus U, Aegeter-Wilmsen T, Aegeter C (2009) Determination of mechanical stress distribution in Drosophila wing discs using
photoelasticity. Mech Dev 126:942–949
50. Schluck T, Aegeter C (2010) Photo-elastic properties of the wing imaginal disc of drosophila. Eur Phys J E33:111–115
51. Crowe K, Smith R (1989) A new technique for determination of tensile stress in thin films. J Electrochem Soc 136:1566–1568
52. Hong T et al (1990) Measuring stiffnesses and residual stresses of silicon nitride thin films. J Electron Mater 19:903–909
53. Zamir E, Taber L (2004) On the effects of residual stress in microindentation tests of soft tissue structures. J Biomech Eng 126:276–283
54. Zamir E, Taber L (2004) Material properties and residual stress in the stage 12 chick heart during cardiac looping. J Biomech Eng 126:823–830
55. Carrasco C et al (2011) Built-in mechanical stress in viral shells. Biophys J 100:1100–1108
56. Baclayon M et al (2011) Prestress strengthens the shell of Norwalk virus nanoparticles. Nano Lett 11:4865–4869
Chapter 22
A Comparison of Residual Stress Measurements on a Linear Friction
Weld Using the Contour Method and Neutron Diffraction
Adrian T. DeWald, Daira Legzdina, Bjørn Clausen, Donald W. Brown, Thomas A. Sisneros, and Michael R. Hill
Abstract Linear Friction Welding (LFW) is a solid phase bonding process, which is being used commercially for
fabrication of complex titanium parts. Like other welding processes, LFW joints contain tensile residual stresses that
could negatively impact performance. This paper presents results from recent residual stress measurements on a test
specimen containing a linear friction weld. Residual stress measurements were performed on the test specimen using the
contour method and neutron diffraction. A comparison of the data from the two techniques is provided, which is favorable.
In general, the residual stresses from the LFW process are shown to be high in magnitude and localized near the weld.
Keywords Residual stress measurement • Contour method • Neutron diffraction • Linear friction welding • Titanium
22.1 Introduction
Titanium alloys are widely used in Aerospace applications for their high strength to weight ratio, good corrosion resistance
and metallurgical stability. New joining methods are being implemented that allow for more efficient manufacture of
titanium components. Linear Friction Welding (LFW) is a solid phase bonding process, which is particularly appropriate for
titanium alloys. During LFW, a normal load is applied as two parts are rubbed against one another in a repeating linear
pattern. Due to the titanium’s great affinity for oxygen, nitrogen and hydrogen, protective atmospheres must be used to avoid
contamination of welded material during traditional welding processes. LFW avoids the formation of a liquid phase during
the welding process, and can therefore be carried out in air. The typical defects caused by melting and solidification during
traditional welding process such as pores, pinholes, shrinkage cracks and grain coarsening are therefore avoided. However,
as with all welding and deformation processes, due to the presence of plastic deformation at elevated temperatures the LFW
joint will have residual stresses.
Residual stresses are known to play a significant role in many material failure processes (e.g., fatigue, fracture, and stress
corrosion cracking). For example, tensile residual stresses left behind by LFW are seen as a detriment to performance (e.g.,
could significantly reduce fatigue life). Thus, the ability to accurately quantify residual stresses through measurement is
important from an engineering perspective.
The objectives of this work are to quantify the residual stress in a typical LFW joint and to compare the measured residual
stress from two different techniques (contour method and neutron diffraction). A description of the test specimen is given
followed by descriptions of the measurement methods and results obtained using each.
A.T. DeWald (*)

Hill Engineering, LLC, Rancho Cordova, CA, USA
e-mail: atdewald@hill-engineering.com
D. Legzdina
Honeywell Aerospace, Phoenix, AZ, USA
B. Clausen • D.W. Brown • T.A. Sisneros
Los Alamos National Laboratory, Los Alamos, NM, USA
M.R. Hill
University of California, Davis, CA, USA
184 A.T. DeWald et al.
22.2 Methods
Specimen description. The test specimen used for these measurements was manufactured from Ti-6Al-4 V bar stock. The
starting microstructure is typical bi-modal (duplex) structure consisting of primary alpha particles and lamellae alpha grains
as shown in Fig. 22.1. The volume fraction of the beta phase was not measured in this work, but it is typically between 5%
and 10% for Ti-6Al-4V alloy. Sample blocks machined from bar stock were 38.1 mm tall 50.8 mm wide 12.6 mm
thick. Two of these blocks were joined (at the 50.8 mm 12.6 mm face) using linear friction welding (Fig. 22.2a).
The LFW process produces a narrow bond region and heat affected zone where the microstructure is altered from its
original condition. A photograph of the microstructure from a similar LFW specimen is shown in Fig. 22.3. The weld
centerline has a predominantly acicular microstructure with a small amount of primary alpha particles showing that the
temperature at the weld interfaces was near beta transus temperature. From centerline outward the microstructure is
gradually changing from very fine to coarser structure with progressively higher amounts of primary alpha. The width of
the weld and heat affected zone is about 1.5 mm.
Prior to performing the residual stress measurements, the flash produced by the LFW process was carefully cut off the test
specimen using a wire EDM and the edges of the specimen were trimmed square. In addition, the large side faces were
polished and etched to reveal the location of the LFW joint (the neutron diffraction measurement required visual alignment
with the LFW joint). This resulted in a final specimen size of 71.4 mm tall49.7 mm wide11.7 mm thick. A photograph of
the polished and etched test specimen is shown in Fig. 22.2b.
Residual stress measurements. Residual stresses were measured in the LFW test specimen using the contour method and
neutron diffraction. The measurements were performed on the same test specimen, in sequence. The following is a summary
of the measurement locations and sequence. A brief description of each measurement method is provided below along with
additional measurement details. First, neutron diffraction was used to measure the xx, yy, and zz residual stress components
along the line through the center of the specimen (Fig. 22.4). Following completion of the neutron diffraction measurement,
the contour method was used to measure the yy component of residual stress over the plane shown in Fig. 22.4. This provides
an opportunity to directly compare neutron diffraction and contour method measurements on the same test specimen.
Neutron diffraction. The neutron diffraction measurements were performed using the SMARTS instrument at Los Alamos
Neutron Science Center (LANSCE) at Los Alamos National Laboratory. The SMARTS instrument is described in detail
elsewhere [1] so only a brief description is given here. The SMARTS instrument is dedicated to engineering relevant neutron
diffraction measurements with its two detector banks at 90 to the incident beam that enables measurement of two
orthogonal strain components simultaneously, see Fig. 22.5. Residual lattice strains were determined based on full pattern
Fig. 22.1 Illustration of the starting bi-modal (duplex) microstructure in the Ti-6Al-4V bar stock
22 A Comparison of Residual Stress Measurements on a Linear Friction Weld Using. . . 185
Fig. 22.2 (a) Photograph

of similar test specimen after
completion of LFW and
(b) photograph of test
specimen used for the
measurements performed
here after removal of flash
and polishing and etching
of specimen surface
Fig. 22.3 Photograph of

microstructure near the
LFW joint from a similar
test specimen
Rietveld [2] refinements of the measured diffraction data using the GSAS [3] and SMARTSware software packages [4]. For
materials with hexagonal crystal structure individual lattice strains for the a and c crystal directions, ea and ec , are determined
according to (22.1)
a a0 c c0
ea ¼ and ec ¼ (22.1)
a0 c0

of LFW test specimen
showing dimensions,
reference coordinate frame,
and measurement locations

of neutron diffraction
experimental set-up
The reference lattice parameters, a0 and c0 , were measured on small coupons cut from a similar weldment, using one set
of values for measurements on the weld line and one set of values for measurements in the base material. The average elastic
strain in the given measurement direction was determined using the typical averaging of the a and c strains for untextured
hexagonal materials [5, 6] according to (22.2)
2eai þ eci
ei ¼ (22.2)
3
Measurements were performed with the sample in two orientations, the first measuring eyy and exx and the second
measuring ezz and exx. For each measurement point along the line, a 2 mm 2 mm 2 mm gage volume was used as this
was shown to produce good measurement statistics within a manageable timeframe and was fine enough to resolve the sharp
stress gradient in the weld. The residual stresses are then calculated using Hooke’s law for isotropic materials:
E
si ¼ ð1 nÞei þ n ej þ ek ; i; j; k 2 xx; yy; zz (22.3)
ð1 þ nÞð1 2nÞ
Fig. 22.6 Contour method

principle: (a) a body
containing unknown residual
stress is cut in half, (b) the free
surface deforms as stresses are
released, (c) applying the
opposite of the deformations
back to the part recovers the
initial residual stress
The given error bars for the determined stresses are based on the fitting error bars from the least squares refinement
properly propagated through the calculations.
Contour method. The contour method is a mechanical release technique for measuring two dimensional distributions of
residual stress over a plane through a body [7, 8] (Fig. 22.6). During a contour method measurement the part is cut into two
pieces, resulting in the re-distribution of residual stress, which leads to deformation. The shape (i.e., contour) of the cut
surface is measured and the data is used to compute the initial residual stress state. The basic steps of a contour method
measurement include: (a) cutting the part into two pieces, (b) measuring the deformation of the cut surface, and (c) analyzing
the displacement data to determine residual stress. The cutting step is typically performed using a wire electric discharge
machine (EDM). Surface measurement is typically performed using a coordinate measuring machine (CMM) or laser
scanning device. The computation of residual stress from measured displacement data is achieved through the use of the
finite element method, where measured displacement data is applied as a boundary condition to a finite element model of
the part. Solving for equilibrium in the presence of this boundary condition determines the initial residual stress present
in the body.
Over the past decade, the contour method has been used to measure residual stress from a variety of sources ranging from
various welding processes [7, 9, 10] to mechanical surface treatments [7]. A relatively comprehensive set of data exists
demonstrating the accuracy of contour method measurements versus process models [11], on specimens with well
characterized residual stress [12], and versus other residual stress measurement techniques [13]. The contour method is
an effective tool for generating 2D maps of residual stress normal to a plane in relatively thick components [13].
22.3 Results
Neutron diffraction. A line plot of the measured residual stress from neutron diffraction is shown in Fig. 22.7. The largest
stresses are oriented along the plane of the LFW joint (syy). Large magnitude tensile stresses are present near the LFW joint
with a magnitude of around 800 MPa near the center of the LFW joint. The tensile stresses diminish to near zero over a
distance of 4 or 5 mm from the center of the LFW joint. The other stress components show similar characteristics (peak stress
Fig. 22.7 Line plot of 1000

measured residual stress Neutron diffraction
from the neutron diffraction 800
experiment
Residual Stress (MPa)

600 Szz
Syy
400
Sxx
200
-200
-400
-10 -5 0 5 10 15 20 25 30
Distance from LFW center (mm)
Fig. 22.8 Two-dimensional

map of the measured residual
stress over the contour
measurement plane
Fig. 22.9 Line plot 1000

comparing the measured Neutron Diffraction yy-component
residual stress from 800
the contour method Contour
and neutron diffraction
experiments 600
Residual Stress (MPa)
400
200
-200
-400
-20 -15 -10 -5 0 5 10 15 20
Distance from LFW center (mm)
near the center of the weld, diminishing in magnitude over a short distance). The peak magnitude of the other in-plane
component of stress (sxx) is around 300 MPa and the peak magnitude of the out of plane stress (szz) is around 400 MPa.
Contour method. A plot of the two dimensional residual stress measured using the contour method is shown in Fig. 22.8.
There is a concentrated region of high-magnitude tensile residual stress near the LFW joint. The peak stress magnitude is
around 750 MPa. The tensile stress quickly diminishes to near-zero stress away from the LFW joint.
Comparison. To facilitate comparison, data was extracted from the 2D contour surface along the same line where the neutron
diffraction measurements were performed. The extracted line from the contour method data is plotted along with the
neutron diffraction data in Fig. 22.9. Overall, there is very good correlation between the two measurement techniques.
The neutron diffraction data shows slightly higher magnitude peak stress at the center of the weld (nominally 800 vs.
750 MPa, which is a 6% decrease). The width of the tensile stress region is very similar for both sets of measurements.
There are a few possible explanations for the differences in residual stress measured using the two techniques. It should be
noted that this specimen contains some features that make residual stress measurements challenging. The specimen contains
highly variable microstructure near the LFW joint, which can cause challenges for neutron diffraction (even when the
reference lattice parameters are carefully measured, as was the case here). Furthermore, the residual stress is highly varying
(from near-zero to near-yield) and the tensile stress is contained within a localized region. Large stress gradients can be
challenging to measure (for any technique) and the contour method is prone to smoothing out features associated with high
stress gradients (as part of the filtering process to eliminate surface roughness effects). In addition, the neutron diffraction
measurements neglected the influence of the beta phase on residual stress (due to the small volume fraction it was not
distinguishable in the diffraction data). In light of these challenges, the level of agreement observed here is quite good.
Acknowledgements This work has benefited from the use of the Lujan Neutron Scattering Center at LANSCE, funded by the Department of
Energy’s Office of Basic Energy Sciences. Los Alamos National Laboratory is operated by Los Alamos National Security LLC under DOE
Contract DE-AC52-06NA25396.
References
1. Bourke MAM, Dunand DC, Ust€ € undag E (2002) SMARTS – a spectrometer for strain measurement in engineering materials. Appl Phys A
74(1):S1707–S1709
2. Rietveld HM (1969) A profile refinement method for nuclear and magnetic structures. J Appl Crystallogr 2(2):65–71
3. Von Dreele RB, Jorgensen JD, Windsor CG (1982) Rietveld refinement with spallation neutron powder diffraction data. J Appl Crystallogr
15(6):581–589
4. Clausen B (2004) SMARTSware Manual, LA-UR 04-6581, Los Alamos National Laboratory, ftp.lanl.gov/public/clausen/SMARTSware/,
2004
5. Daymond MR (2004) The determination of a continuum mechanics equivalent elastic strain from the analysis of multiple diffraction peaks.
J Appl Phys 96(8):4263–4272
6. Daymond MR, Bourke MAM, Von Dreele RB (1999) Use of Rietveld refinement to fit a hexagonal crystal structure in the presence of elastic
and plastic anisotropy. J Appl Phys 85(2):739–747
7. DeWald AT, Rankin JE, Hill MR, Lee ML, Chen H-L (2004) Assessment of tensile residual stress mitigation in Alloy 22 welds due to laser
peening. J Eng Mater Technol 126(4):465–473
8. Prime MB (2001) Cross-sectional mapping of residual stresses by measuring the surface contour after a cut. J Eng Mater Technol 123:162–168
9. Prime MB, Sebring RJ, Edwards JM, Hughes DJ, Webster PJ (2004) Laser surface-contouring and spline data-smoothing for residual-stress
measurement. Exp Mech 44(2):176–184
10. Zhang Y, Fitzpatrick ME, Edwards L (2004) Cross-sectional mapping of residual stresses in a VPPA weld using the contour method. Acta
Mater 52:5225–5232
11. Prime MB, Newborn MA, Balog JA (2003) Quenching and cold-work residual stresses in aluminum hand forgings: contour method
measurement and FEM prediction. Mater Sci Forum 426–432:435–440
12. Pagliaro P, Prime MB, Swenson H, Zuccarello B (2010) Measuring multiple residual-stress components using the contour method and multiple
cuts. Exp Mech 50(2):187–194
13. Pagliaro P, Prime MB, Robinson JS, Clausen B, Swenson H, Steinzig M, Zuccarello B (2010) Measuring inaccessible residual stresses using
multiple methods and superposition. Exp Mech 51(7):1123–1134
Chapter 23
Experimental Study of High Speed Ball End Milling
of Titanium Alloy (Ti-6Al-4V)
M.-H. Baccar, E. Bayraktar, T. Rickert, M. Boujelbene, and D. Katundi
Abstract With the evolution of machine tools and the emergence of new cutting tools; and in the framework of the
production of parts with complex geometry, the manufacturers were able to realize more and more parts of complex shape.
The multi-axis machining is the main technique for achieving the free form; in fact the multi-axis milling with ball end tools
attracts the interest of the aerospace industry and the mussel industry which continues to seek ways to improve the surface
quality of finished parts (the shaping of parts in a jet engine, it achieves a material removal process, such as multi-axis
milling). The titanium alloy is widely used in aerospace industry is the subject of this study in fact, the integrity of the
surfaces of parts produced by multi-axis milling is an issue more relevant than ever for the aerospace industry. This paper
aims to study the influence of the position of the tool and the parameters cutting precisely the speed feed Vf, the engagement
of the tool on the roughness 3D, micro-hardness and microstructure alteration created in sub-surface during the milling of
concave surface of the titanium alloy type Ti-6Al-4V. It gives also more details on the different parameters (cutting
conditions) such as cutting forces and also temperature field in cutting, the state of surface and the state of residual stresses.
Observations on the scanning electron microscope (SEM) are carried out before and after machining of the specimens and
the corresponding cutting tools for damage analysis at the surfaces.
Keywords High-speed multi-axis milling • Titanium alloy • Surface integrity • Residual stresses
23.1 Introduction
The titanium alloy Ti-6Al-4V is a material widely used in aerospace, biomedical, and chemical industries thanks to its good
mechanical and thermal properties because of their good strength-to-weight ratio, corrosion resistance and possibility of use
in harsh environments such as high temperature. For example, in the aerospace industry in general the mechanical parts are
extracted by the milling process because of the different forms of parts as the complex forms. However, titanium alloys have
a poor machinability due to its low thermal conductivity, which causes a high chemical reactivity caused by the elevation of
temperature in the field cutting [1], indeed, the quality of a machined surface, is becoming more and more important to
satisfy the increasing demands of sophisticated component performance. However, the surface integrity is one of the most
relevant parameters used for evaluating the quality of machined surface. Indeed, the quality and performance of a product is
directly related to surface integrity achieved by final machining. However, the surface integrity is influenced by a set of
parameter as we include for example the orientation of the tool axis relative to the surface the direction of feed rate in sense
scans. Various studies on the characterization of the influence of the orientation of a tool hemispherical were conducted.
Some have a more experimental [2, 3] and are more attached to observe the tool wear as the final surface finish.
Other studies [4] analyze the trace left by a tool for different inclination angle and propose a model to predict the surface,
but the criteria used to the surface state, are only linear and non-surface. The presence of vibration in the system machine-
tool-piece has a bearing on the quality of finished surface [5]. Other researchers have analyzed the influence of cutting
M.-H. Baccar • E. Bayraktar (*) • M. Boujelbene • D. Katundi

Supmeca-LISMMA-Paris, Mechanical and Manufacturing Engineering School, Paris, France
e-mail: bayraktar@supmeca.fr
T. Rickert
American Stress Technologies, Inc., 840 Watercrest Way, Cheswick, PA 15024, USA
192 M.-H. Baccar et al.
parameters on surface state such [6, 7]. However, the surface integrity including several criteria such as roughness,
microstructural changes also residual stresses, micro-hardness and plastic deformation at the surface. These criteria are
influenced by the same factors mentioned earlier, but the variation of each parameter has a different influence (in the sense
all criteria can be good or bad at a time) on each criterion. However, researchers have studied these criteria a case by
case several researchers found that surface roughness values became larger at high cutting speeds in turning Ti-6Al-4V [8]
and in end milling Ti-6Al-4V using WC-Co and PCD insert [7].
The tool wear has a positive effect on the workpiece indeed, with increasing the tool wear the surface will be smoother
[9], in addition the hardness to the surface layer in turning titanium alloy Ti-6Al-4V is increased about 30% the hardness of
metal located in the heart, this layer is assumed in the order of 100 mm. The same way [10] estimated the same value 30%.
The microstructure at the surface layer can be changed as a result of chemical changes caused by the tool, enlargements and
elongations of the grains for the turning Ti-6Al-4V are observed by Hughes et al. [11]. Moreover, the residual stresses were
studied in terms of cutting parameters for the Ti-6Al-4V by several researchers such as [12]; similar work on the surface
integrity generated by the milling tool hemispherical was carried by Hsi-Yung Feng and Ning Su [13], Boujelbene et al. [14],
Fang and Wu [15], Kudryavtsev et al. [16].
The subject of this paper is to study the topography of a surface machined by the tool hemispherical in three-axis milling
of Ti-6Al-4V with cutting conditions of finishing, we mainly analyze the roughness 3D, the micro-hardness and micro-
structural alterations caused by the effect of tool.
23.2 Experimental Conditions
23.2.1 Workpiece Material
Materials studied here is titanium alloy Ti-6Al-4V, is widely used in aerospace, biomedical, chemical and petroleum
industries because of their good mechanical and thermal properties. Certain properties are listed in Table 23.1. The micro-
structure of Ti-6Al-4V is an alpha-beta alloy (a + b), the alpha phase proportion usually varies from 60% to 90%.
The alpha phase in pure titanium is characterized by a hexagonal close-packed crystalline structure that remains stable
from room temperature; beta phase in pure titanium has a body-centered cubic structure, and is stable from room temperature
to the melting point. Adding alloying elements to titanium provides a wide range of physical and mechanical properties.
Certain alloying additions, notably aluminum, tend to stabilize the alpha phase; that is they raise the temperature at which the
alloy will be transformed completely to the beta phase. This temperature is known as the beta-transus temperature Tb of
about 882 C. Alloying additions such as chromium, copper, iron, manganese, molybdenum, and vanadium stabilize the beta
phase by lowering the temperature of transformation from alpha to beta. The chemical composition is given in Table 23.2.
23.2.2 Machining Conditions
Three axis milling with a hemispherical tool were carried in dry machining out at various cutting conditions on the
Ti-6Al-4V alloy. The machined shape are concave surfaces (see Fig. 23.1), the cutting tests are carried out on a three-
axis high speed milling centre in a ball-end mill with a diameter of 16 mm, The tool used is a hemispherical tool with
removable insert to two teeth, the tool is brand SANDVIK its substrate neat grained, associated with a PVD coating of
TiAlN: Designation of insert: R216F-16 40 E-L P20A.
The cutting style for ball end milling of concave surface used and the cutting conditions are listed in Table 23.3.
Table 23.1 Mechanical Thermal

and thermal properties Density Hardness Modulus E Tensile conductivity Melting point
of Ti-6Al-4V (g/cm3) (HB) (GPa) strength (MPa) (W/m K) ( C)
4.42 345 113.8 995 7.3 1,670
Table 23.2 Chemical Element Al V Fe C N H O Ti

composition (%) of
Ti-6Al-4V alloy % 6 4 0.3 0.08 0.05 0.01 0.2 Balance
23 Experimental Study of High Speed Ball End Milling of Titanium Alloy (Ti-6Al-4V) 193
a b
Vf
θd θup
Downward Upward
milling milling
Workpiece
Vc-eff-mini
Fig. 23.1 (a) Photo of milling of concave surface; (b) tool position for milling of concave surface
Table 23.3 Machining conditions applied in this project

Cutting parameters (Milling with hemispherical tool)
Number Spindle speed Feed speed Axial depth Radial depth Tool position or type Sample
of Sample N (rpm) Vf (mm/min) of cut ap (mm) of cut an (mm) of milling (y) on surface form
1 3,000 600 0.5 0.5 Top of surface(y¼0 ) Convexe
11 3,000 600 0.5 0.5 Top of surface(y¼0 ) Concave
12 3,000 900 0.5 0.5 Top of surface(y¼0 ) Concave
13 3,000 600 0.5 0.5 Upward, yup ¼ 60 Concave
15 3,000 900 0.7 0.5 Downward, yup ¼ 60 Concave
21 3,000 600 0.5 0.5 Upward, yup ¼ 60 Convexe
26 3,000 600 0.2 0.5 Downward, yup ¼ 60 Convexe
23.2.3 Roughness Measurements, Micro-hardness and Microstructure Analysis

by Optical and Scanning Electron Microscopy
General surface characterization of the machined surface has been carried out on the optical surface roughness “Surfascan”
device in 3D. Metallographic analysis and surface evaluation of all of the machined parts in dry machining at various cutting
conditions on the Ti-6Al-4V alloy have been made using optical and scanning electron microscope (SEM) in order to control
the surface topography. Optical microscopy allows us detecting of microstructure alteration of the top machined concave
and convex surfaces. Vickers micro-hardness measurements were made on the instrumented microhardness machine.
These experimental works allowed us optimization of the operational parameters (cutting conditions) and also materials
parameters to design optimal pieces under service conditions. In fact, the machined surfaces were examined in concave
surfaces. Additional measurements have been carried out also for making a comparison of these two types of cutting
strategies.
23.2.4 Measurements of Residual Stresses at the Machined Surfaces
Residual stresses play an important role in operating performance of materials, parts and structural elements. Their effect on
the engineering properties of materials such as fatigue and fracture, corrosion resistance and dimensional stability can be
considerable. The residual stresses, therefore, should be taken into account during design, fatigue assessment and
manufacturing of parts (machining, welding etc.). In the frame of the research project, cutting forces were measured
and X-ray diffraction measurement method was used to obtain the residual stresses. All measurements in “outside” and “out”
were electro-polished using exactly the same time increments: 10 s. The total electro-polishing times are: 0, 10, 20, 30, 50,
70 and 100 s. The depth profiles for the measurements are similar in all cases. It means that for each actual depth
measurement step was 10 mm. Therefore, all of the specimens have the same measurements focusing on the electro-
polishing times (same current) rather than depth measurements.
23.3.1 Surface Topography and Roughness
In order to generate the topography of surfaces in ball-end milling, the study of the surfaces is requested. However, there are
many parameters used in the literature related to surface roughness. Otherwise, for the 2D surface roughness parameters,
the most popular of these parameters is average roughness. It is quoted as Ra symbol. Mathematically, Ra is the arithmetic
value of the departure of the profile from centerline; for the 3D surface roughness parameters, the most used parameters are,
Sa and Sq, respectively, the arithmetical mean of the surface and root mean square roughness. In the present paper,
representative parameters of 3D surface were studied. The Sa and Sq variation of the concave surface machined were
analyzed according to the tool position and the speed feed Vf. The study has shown that the tool position influences
significantly the values of Sa and Sq.
Indeed, when the tool removes the material with a low effective cutting speed Vc-eff-minimale at the top of the concave
surface, y ¼ 0 (see Fig. 23.1b), we recorded high values of Sa and Sq (Sq ¼ 4.6 mm and Sa ¼ 2.8 mm) compared to
downward milling and upward milling (see Fig. 23.2a). Also, the surface topography and SEM observation in this zone
shows traces disturbed of the cutting tool and non-regular ridges (Fig. 23.3a, b) which shows the change of the cutting
direction at the top of the concave surface.
However, the phenomena of vibrations were heard during machining in this area that has been machined with a very
low effective cutting speeds (Vc-eff even zero). By consequence, when the tool removes materials with effective cutting
speeds relatively important (downward milling and upward milling) we recorded a best topography and regular ridges
Fig. 23.2 Variation of Sa and Sq: (a) according to the tool position at N ¼ 3,000 rpm, Vf ¼ 900 mm/min, ae ¼ 0.5 mm, ap ¼ 0.5 mm;
(b) according to feed speed Vf at N ¼ 3,000 rpm, ae ¼ 0.5 mm, ap ¼ 0.5 mm
Fig. 23.3 State of surfaces in the top of concave surface y ¼ 0 , at N ¼ 3,000 rpm, Vf ¼ 900 mm/min, ae ¼ 0.5 mm, ap ¼ 0.5 mm:
(a) topography 3D measured by roughness-tester, (b) SEM observation of a milled surface
Fig. 23.4 State of surfaces in upward milling yup ¼ 60 at N ¼ 3,000 rpm, Vf ¼ 900 mm/min, ae ¼ 0.5 mm, ap ¼ 0.5 mm: (a) topography 3D
measured by roughness-tester, (b) SEM observation of a milled surface
(see Figs. 23.1b, 23.4, and 23.5). Nevertheless the surface state in upward milling yup ¼ 60 , has a slight advantage
compared to downward milling yd ¼ 60 .
This is probably due to the difference of the effective cutting speed and cutting phenomenon (entered the edge of tool in
the material) in the same zone. The result of the study of state surface 3D Sa and Sq at the top of the concave form depending
on the speed feed Vf is show in Fig. 23.2b. This study shows that Sa and Sq increase significantly depending on the speed
feed that is to say according to the feed per tooth fz.
23.3.2 Micro-hardness and Microstructure Analysis by Optical and Scanning

Electron Microscopy
During the machining operations by the hemispherical tool, the workpiece material is exposed to thermal, mechanical and
chemical energy that can change the properties of superficial layer (about 100 mm) of the surface machined. In the case of the
hardness measurement, the hardness profile has been evaluated on the surface as a function of the distance from
the machined surface (Fig. 23.6a). That can give an idea on surface proprieties, the level of the work hardening of the
deformed layer produced at the local surface. However, the hardness measurements were made according to the tool
Fig. 23.5 State of surfaces in downward milling yd ¼ 60 at N ¼ 3,000 rpm, Vf ¼ 900 mm/min, ae ¼ 0.5 mm, ap ¼ 0.5 mm: (a) topography 3D
measured by roughness-tester, (b) SEM observation of a milled surface
Fig. 23.6 (a) Microhardness profile of the distance from the machined surface, (b) microstructure alteration of the top machined concave surface
position; for a depth 120 mm from the machined surface and each 20 mm we measured the hardness two times then we plotted
the average of the two measures for each depth (see Fig. 23.6a).
First when we approaching to the core of work piece, the hardness is decreases for the different positions of tool, the
increase in hardness is more important in the top of concave surface, the value of the hardness is about 375–20 mm then
we approaching at core the hardness can be same as the bulk material (see Fig. 23.6a), this is due to the very low value of the
effective cutting speed (Vc-eff-mini even zero in the centre of the cutter) in the top of concave surface that will cause crushing
of the material, in addition the vibration phenomenon increasing when the tool cut with a low effective diameter (Deff low),
which is unfavourable for obtaining good state of surface. Finally, we observe that hardness in surface layer is increased at
about 8% than the bulk material. This evaluation is observed regularly at the major of the specimens machined under
different cutting conditions.
However, machining by the hemispherical tool can be creates the variation of near surface microstructure, that is caused
by the combined effects of mechanical and thermal loads in milling. Then as in the milling by the hemispherical tool cutting
temperatures are too low to induce any phase transformations such as a white layer on the machined surfaces. It implies that
no thermal damage occurred at the milling conditions, but the mechanical and thermal loads can creates a very thin layer of
hardened material and plastic deformations on top of the machined concave surface. Surface deformation induced strain
hardening was the main factor for the measured surface hardness in previous section.
Initially, one may note that layer alteration of machined surface is according of cutting condition and tool position in end
ball milling. But it is interesting to look at the tool position in the machined surface, it was found that when machining under
dry conditions a thin layer of disturbed or plastically deformed layer was formed immediately beneath the machined surface.
Figure 23.6b, shows a layer that has been disordered as a result of milling in the top of concave surface at y ¼ 0 . This
disturbance in general present a plastic deformation in compression of the surface layer that seems to change the orientation
of grain and grain boundary, under a result of elongation or/and shortening of the grains.
However, it is noted that milling in the top of concave surface disturbs the surface much more than the downward milling
when upward milling, This is probably due to the cutting phenomenon in this area that is more complicated other the two
positions of the tool, because in this area the cutting speed is very low that is can generated a exceptional mechanical and
thermal loading that will influence the microstructure in sub-surface machined. A detail examination of the top surface of all
machined work pieces shows that there is neither heat affected zone (HAZ) nor white interface layer (WIL).
In the same way, Machined surfaces at the different cutting conditions were examined on the Scanning Electron
Microscope (SEM) as indicated in the Fig. 23.7a–c. At the left column, typical structures with very intense regular ridges
are observed. Here the position of the tool (according the milling type on the surface, y ¼ 0 , y ¼ 60 , etc.) play a role on
the form of the ridges. At the top of the surface where y ¼ 0 , more intense ridge are observed regarding to the position of
y ¼ 60 up and downward milling. On the other hand, the parameter, feed speed, Vf can be effective on the ridge form of the
surface that we have explained in Figs. 23.3, 23.4, and 23.5 as the roughness values. The intensity of the ridges is very wavy
and has a large front in case of low feed speeds that is related to the surface roughness evaluation.
At the right column (Fig. 23.7a–c), the detail of ridges and the passage-front of the ridges are presented in higher
magnifications. In literature some of the authors have indicated that the passages of the ridges are the resources of
the residual stresses. In general, this phenomenon is observed as alteration of the grain at the surface of the specimen due
to heavy plastic deformations. Any other damage, like grain growth due to the heat affected zone and/or white layer
formation is observed on the machined specimens.
23.3.3 Measurements of Residual Stresses at the Concave and Convex Surfaces
As well known, residual stress is one of the most relevant practical parameters used for evaluating the quality of the
machined surface, particularly when critical structural components are produced; with the objective to reach the higher
reliability levels. This is the case for the components basically used in aerospace industries. Residual stresses are an effect
from both heats generated and mechanical work going into the surface and subsurface. More precisely, residual stresses in
machining are produced as a consequence of inhomogeneous plastic deformation induced by mechanical and thermal
loadings associated with the chip formation process and the interaction between the tool and the freshly machined surface in
the vicinity of the tool nose [14–18].
Thermal effects tend to give tensile stresses, while mechanical influences contribute to compressive residual stresses,
Phase transformations can support the development of both compressive and tensile residual stresses depending on the
relative volume changes and the accompanying plastic deformations. Residual stresses strongly affect the fatigue life of a
component. The introduction of a compressive mean stress will increase the allowed alternating stress for a given fatigue life
and then is beneficial. Conversely, a tensile mean stress reduces the allowed alternating stress in service [16–23].
In consequence, high tensile stresses generated by the machining of work hardening alloys may be highly deleterious to
fatigue performance. The effect is most significant in the high cycle fatigue regime where the applied stress magnitude is not
sufficient to significantly relax the residual stresses produced during manufacturing.
Measurements of the residual stress distributions (hoop and axial stresses) beneath the machined surface inside of the
convex and concave surfaces are presented in the Fig. 23.8a–d. Cutting test conditions were given in Table 23.3 for different
parts from n 01 up to 28. And also a comparison of residual stress distributions for the hoop and axial stresses were given
here only for the piece n 15.
For the variety of cutting conditions investigated here, residual stresses are always compressive at surface (around
400–500 MPa) and gradually move to low tensile values under the surface before stabilizing at the level corresponding to
that found in the work material before machining, in the order of zero MPa and/or going up to 50–100 MPa.
Regarding to the level of residual stresses measured on the different pieces, one may note the effect of the position of the
tool (according the milling type on the surface, y ¼ 0 , y ¼ 60 , etc.) on the distribution of residual stresses. In case of
the tool position at the top of the surface (y ¼ 0 ), residual stresses arrive at maximum values as compressive stresses values
that higher than other tool position (y ¼ 60 ), in case of downward and/or upward position.
As indicated in Fig. 23.8b, there is not so much difference between hoop and axial stresses compared for the specimen
n 15: at the surface quasi the same compressive values whereas hoop stress values a little bit higher than that of axial stresses
at a certain distance of surface.
Fig. 23.7 (a) Examination of the machined surfaces on the SEM (cutting conditions; milling on the concave surface yd ¼ 60 at N ¼ 3,000 rpm,
Vf ¼ 900 mm/min, ae ¼ 0.5 mm, ap ¼ 0.7 mm). (b) Examination of the machined surfaces on the SEM (cutting conditions; milling on the
concave surface, y ¼ 0 , at N ¼ 3,000 rpm, Vf ¼ 900 mm/min, ae ¼ 0.5 mm, ap ¼ 0.5 mm). (c) Examination of the machined surfaces on
the SEM (cutting conditions; milling on the convex surface yd ¼ 60 at N ¼ 3,000 rpm, Vf ¼ 600 mm/min, ae ¼ 0.5 mm, ap ¼ 0.7 mm)
Fig. 23.8 (a) Residual stress distribution (hoop and axial stresses) beneath the machined surface inside of the concave surface; test conditions
were given in Table 23.3 for different parts from n 11 up to 18. (b) Comparison of residual stress distributions (hoop and axial stresses) beneath the
machined surface inside of the concave surface; test conditions were given in Table 23.3 for the parts n 15. (c) Residual stress distribution (hoop
and axial stresses) beneath the machined surface outside of the convex surface; test conditions were given in Table 23.3 for different parts from
n 21 up to 88. (d) Residual stress distribution (hoop and axial stresses) beneath the machined surface outside of the concave surface; test conditions
were given in Table 23.3 for different parts from n 01 up to 04
Fig. 23.8 (continued)
23.4 Conclusions
The present paper gives a comparative study on the surface integrity in case of ball end milling titanium Ti-6Al-4V in dry
machining of concave and convex surfaces. A very large experimental work has been carried out and experimental results
are the mean values based on the more than 50 specimens.
The following are the specific conclusions can be driven form these experiments for a few criteria related directly to the
surface integrity:
Tool position influences the surface roughness, indeed when tool at upward and downward milling provided the best
finish surface conditions compared to machining in the top of the surface.
The 3D roughness Sa and Sq have influenced by the speed feed Vf, definitely; the values of roughness are increased
according to increasing the speed feed.
The milling by the hemispherical tool produced a thin layer of disturbed or plastically deformed layer formed immediately
beneath the machined surface, and it should be noted that the surface was not heat affected, it means that there are neither heat
affected zone (HAZ) nor white interface layer (WIL) and not other microstructural change such as grain growth etc. The tool
positions obviously influence the microhardness of the machined surface.
For the variety of cutting conditions investigated here, residual stresses are always compressive at surface (around
400–500 MPa) and gradually move to low tensile values under the surface before stabilizing at the level corresponding to
that found in the work material before machining, in the order of zero MPa and/or going up to 50–100 MPa. Tool position
influences also the residual stress evolution in the machined parts.
Acknowledgement Authors of the present paper thank so much to American Stress Technologies Inc, Pittsburgh, USA for their kind help to use
their facilities for residual stress measurements. Personally E. Bayraktar thanks to Dr T. Rickert for his valuable ideas and his personal attempt to
carry out the measurements of residual stresses.
References
1. Sun J, Guo YB (2009) A comprehensive experimental study on surface integrity by end milling Ti–6Al–4V. J Mater Process Technol
209:4036–4042
2. Ko TJ, Kim HS, Lee SS (2001) Selection of the machining inclination angle in high-speed ball end milling. Int J Adv Manuf Technol
17:163–170
3. Kang MC, Kim KK, Lee DW, Kim JS, Kim NK (2001) Characteristics of inclined planes according to the variations of the cutting direction in
high-speed ball-end milling. Int J Adv Manuf Technol 17:323–329
4. Bouzakis K-D, Aichouh P, Efstahiou K (2003) Determination of the chip geometry, cutting force and roughness in free form surfaces finishing
milling, with ball end tools. Int J Mach Tool Manuf 43:499–514
5. Biermann D, Kersting P, Surmann T (2010) A general approach to simulating workpiece vibrations during five-axis milling of turbine blades.
CIRP Ann Manuf Technol 59:125–128
6. Daymi A, Boujelbene M, Ben Amara A, Bayraktar E, Katundi D (2011) Surface integrity in high speed end milling of titanium alloy
Ti–6Al–4V. Mater Sci Technol 27:387–393
7. Amin NAKM, Ismail AF, Khairusshima NMK (2007) Effectiveness of uncoated WC–Co and PCD inserts in end milling of titanium alloy – Ti-
6Al-4V. J Mater Process Technol 192(193):147–158
8. Ribeiro MV, Moreira MRV, Ferreira JR (2003) Optimization of titanium alloy (Ti–6Al–4V) machining. J Mater Process Technol 143
(144):458–463
9. Che-Haron CH, Jawaid A (2005) The effect of machining on surface integrity of titanium alloy Ti–6%Al–4%V. J Mater Process Technol
166:188–192
10. Canteroa JL, Tardiob MM, Cantelia JA, Marcosc M, Miguelez MH (2005) Dry drilling of alloy Ti–6Al–4V. Int J Adv Manuf Technol
45:1246–1255
11. Hughes JI, Sharman ARC, Ridgway K (2004) The effect of tool edge preparation on tool life and workpiece surface integrity. Proc Inst Mech
Eng B: J Eng Manuf 218:1113–1123
12. Ulutan K, Ozel T (2011) Machining induced surface integrity in titanium and nickel alloys: a review. Int J Mach Tool Manuf 51:250–280
13. Feng Hsi-Yung, Ning Su (2000) Integrated tool path and feed rate optimization for the finishing machining of 3D plane surfaces. Int J Mach
Tool Manuf 40:1557–1572
14. Boujelbene M, Moisan A, Bouzid W, Torbaty S (2007) Variation cutting speed on the five axis milling. J Achiev Mater Manuf Eng 21(2):7–14
15. Fang N, Wu Q (2009) A comparative study of the cutting forces in high speed machining of the Ti-6Al-4V and Inconel 718 with a round cutting
edge tool. J Mater Process Technol 209:4385–4389
16. Kudryavtsev Y, Kleiman J, Trufiakov V, Mikheev P (1999) Expert system for fatigue assessment and optimization of welded elements. In:
Proceedings of the third world congress of structural and multidisciplinary optimization, vol 2, Buffalo, 17–21 May, pp 469–471
17. Ezugwu EO, Wang ZM, Machado AR (1999) The machinability of nickel-based alloys: a review. J Mater Process Technol 86:1–16
18. Outeiro JC, Pina JC, M’Saoubi R, Pusavec F, Jawahir IS (2008) Analysis of residual stresses induced by dry turning of difficult-to-machine
materials. CIRP Ann Manuf Technol 57:77–80
19. Arunachalam RM, Mannan MA, Spowage AC (2008) Surface integrity when machining age hardened Inconel 718 with coated cutting tools.
Int J Mach Tool Manuf 44:1481–1491
20. Umbrello D, M’Saoubi R, Outeiro JC (2007) The influence of J-C material constant on finite element simulation of machining of AISI 316L
steel. Int J Mach Tool Manuf 47:462–470
21. Axinte DA, Dewes RC (2002) Surface integrity of hot work tool steel after high speed milling experimental data and empirical models. J Mater
Process Technol 127:325–335
Chapter 24
Very High-Cycle Fatigue Resistance of Shot Peened High-Strength
Aluminium Alloys
M. Benedetti, V. Fontanari, and M. Bandini
Abstract The present paper is aimed at investigating the effect of two shot-peening treatments on the plain and notch
fatigue behavior of the Al-7075-T651 alloy. The W€ohler S-N curves were determined by pulsating bending for the different
experimental conditions exploring fatigue lives comprised between 105 and 108 cycles. It was found that improvement in
plain fatigue resistance due to shot peening steadily declines during fatigue life. Conversely, the reduction in notch fatigue
sensitivity due to shot peening is more pronounced in the very-high cycle fatigue regime. Scanning Electron Microscope
(SEM) fractography revealed a tendency towards sub-surface fatigue crack initiation. The extent of the residual stress
redistribution during loading was investigated by means of X-ray Diffraction measurements.
Keywords Shot peening • Aluminum alloys • Very-high cycle fatigue regime
24.1 Introduction
Aluminium alloys are an attractive class of materials for aircraft and automotive industry because of their high specific static
strength. In aerospace, aluminium alloys face ever stiffer competition from composites. In the automotive context, more and
more engine parts are being made from them. Usually, high static mechanical properties are induced in aluminium alloys by
dispersion hardening through solution and ageing heat treatments. However, aluminium alloys exhibit poor plain fatigue
resistance [1] and high notch fatigue sensitivity [2]. Accordingly, stress raisers, like holes, fillets and grooves, always present
in machine parts, are particularly detrimental to the fatigue response of these alloys, thus limiting their use in highly stressed
mechanical components characterized by complex shapes. For this reason, aluminium alloys are frequently subjected to shot
peening, particularly effective in incrementing the plain and notch fatigue strength of steels and light alloys. The shot
peening mainly results in three fatigue related modifications of the surface layers: roughness, residual stresses and work
hardening. Clearly, the surface roughening after shot peening is detrimental to the fatigue resistance due to the stress
concentration exerted by the surface dimples. In the literature, it is commonly accepted that the improvement of fatigue
strength is mainly induced by the introduction of compressive residual stresses in the surface region, responsible for both
retarded fatigue crack initiation and lower small crack growth rates [3, 4]. The role of work hardening on the fatigue response
is essentially indirect, since it strongly affects the stability of residual stresses by preventing them from relaxing due to
accumulation of plastic deformation [5].
The authors recently studied the effect of three types of shot peening treatments on the plain and notch high-cycle fatigue
response of the Al-7075-T651 alloy under reverse bending loading conditions (R ¼ 1) [6–9]. In particular, it was shown
that (1) residual stress relaxation occurs when the superposition between the compressive residual stress and the compressive
peak stress produced by reverse bending exceeds the material’s cyclic yield stress; (2) the plain fatigue response is directly
correlated with the surface residual stress as successfully predicted using the Sines criterion incorporating the stabilized
M. Benedetti (*) • V. Fontanari

Department of Engineering Materials and Industrial Technologies, University of Trento, via Mesiano 77, 38100 Trento, Italy
e-mail: matteo.benedetti@ing.unitn.it
M. Bandini
Peen Service s.r.l, via Pollastri 7, Bologna 40138, Italy
e-mail: m.bandini@peenservice.it
204 M. Benedetti et al.
residual stress field as mean stresses; (3) gentle peening treatment conducted using small ceramic beads are more effective
than more intense treatments employing larger peening media; (4) shot peening reduces the fatigue notch sensitivity;
(5) the notch fatigue resistance can be satisfactorily predicted by the Sines fatigue criterion incorporating stabilized residual
stresses and combined with a line method based on the critical distance theory to account for the notch sensitivity. However,
the current state-of-the-art-knowledge lacks for investigations proving whether the beneficial effect of shot peening is still
present in the very-high cycle fatigue regime.
The present work is aimed at investigating the very-high cycle fatigue regime of shot peened Al-7075-T651 alloy. For this
purpose, pulsating bending fatigue tests (R ¼ 0.05) have been carried out on smooth and notched samples exploring fatigue
lives comprised between 105 and 108 cycles. Fatigue crack initiation sites have been investigated through scanning electron
microscopy (SEM) fractography and the role of surface roughness on fatigue resistance has been analyzed. Residual stress
profiles were used to discuss the improvement in the fatigue response.
24.2 Materials and Experimental Procedures
The experimentation has been performed on the aluminium alloy Al-7075-T651, widely used for aeronautical applications,
supplied in the form of 4 mm thick rolled plate. The bulk material properties have been determined on five standard
monotonic tensile tests (initial strain rate of 1 103 s1) performed in the longitudinal orientation. The results,
summarized in Table 24.1, show a yield strength higher than 500 MPa, combined with a good material ductility (total
elongation of 16%).
The fatigue characterisation has been carried out on prismatic specimens whose geometry, according to the standard ISO
3928, is illustrated in Fig. 24.1. The microstructure has been tested with the stress axis parallel to the L-direction. The first
type of samples (Fig. 24.1a), termed smooth samples, presents a fillet radius large enough to make any notch fatigue effects
negligible. The samples illustrated in Fig. 24.1b, called notched samples, carry two edge V-notches with 90 opening angle,
whose notch tip fillet radius is 2 mm. This corresponds to a theoretical notch stress concentration factor Kt equal to 1.53.
Part of the specimens has been subjected to controlled shot peening: the parameters of the two peening treatments
considered are summarized in Table 24.2. Each treatment has been performed using small beads, which allow for higher
Table 24.1 Monotonic tensile properties of the Al-7075-T651 alloy

E (GPa) sY0.2 (MPa) UTS (MPa) sF (MPa) T.E. (%) R.A. (%)
73 (1) 510 (5) 580 (5) 780 (10) 16 (2) 22 (2)
E elastic modulus, sY0.2 0.2 % yield stress, UTS ultimate tensile strength, sF true fracture stress, T.E. total elongation, R.A. reduction in area
Fig. 24.1 Geometry of the bending fatigue specimen used in this study: (a) smooth and (b) notched specimens. All dimensions are given in mm
24 Very High-Cycle Fatigue Resistance of Shot Peened High-Strength Aluminium Alloys 205
Table 24.2 Shot peening parameters

Treatment Material Bead size (mm) Bead hardness (HV1) Bead composition Almen intensity Angle of impingement Coverage (%)
C120 Ceramic 60–120 700 ZrO2 67% 4.5 N 90 100
SiO2 31%
V40 Glass 40–70 550 SiO2 31% 4.2 N
Na2O 14%
Fig. 24.2 Four-point bending fatigue test apparatus used in the present study
fatigue performance as compared with larger beads [6]. The treatment named C120 employs small ceramic beads leading to
a gentle and superficial effect, whereas the second one termed V40 has been conducted with very fine glass beads in order to
simulate fine particle shot peening which is known to greatly enhance the fatigue resistance of aluminium alloys [10].
Pulsating (R ¼ 0.05) load-controlled four-point bending fatigue tests have been carried out at room temperature air and at
a nominal frequency of 110 Hz using a resonant testing machine (Fig. 24.2). Different stress levels corresponding to fatigue
lives in the range between nearly 105 and 108 cycles have been considered. Tests were terminated at 108 cycles when no
fracture occurred. The fatigue curves corresponding to 50% of failure probability, represented by the S-N curve:
sP50 ¼ s0P50 N f k
1
(24.1)
were determined by fitting the log(Nf) versus log(s) results. The uncertainty range was assumed to be constant and
approximated by its centroid value. As a representative value of the scatter, the following expression was used:
Ts ¼ 1 : sP90 =sP10 (24.2)
P90, P10 denote the 90% and 10% levels of failure probability, respectively.
The notch sensitivity factor q at a given number of cycles to failure Nf was computed as:

Kf Nf 1 sP50;smooth N f
q Nf ¼ ; Kf Nf ¼ (24.3)
Kt 1 sP50;notched N f
The modifications of the surface layers of the shot peened samples were investigated through microhardness, surface
roughness and residual stress profile measurements. To this regard, microhardness profiles were measured to characterize the
material’s work hardening. An optical profilometer that has plane surface spatial resolution of 0.14 mm, and the z-axial
measurement resolution of 0.01 mm, was used to observe the samples surface. Three dimensional surface topography and
roughness were evaluated.
Table 24.3 Surface roughness Condition Ra (mm) Rq (mm) Rt (mm)

properties
As-received 0.33 0.42 1.34
C120 1.17 1.55 7.12
V40 1.33 1.99 8.45
The analysis of the residual stress field induced by the peening treatments was carried out by measuring the stress profile
by XRD technique. For this purpose, an AST X-Stress 3,000 X-ray diffractometer (radiation Cr Ka, irradiated area 1 mm2,
sin2c method, 11 diffraction angles (2y) scanned between 45 and 45 ) was used. The in-depth measurements were
performed step-by-step removing a very thin layer of material using an electro-polishing device.
Both initial and stabilized residual stress fields were measured. For this purpose, measurements were performed on tested
smooth specimens after failure in a region far enough from the fracture surface (about 2 mm) so that the material rupture was
supposed not to have altered the residual stress field [6]. Measurements were carried out on both specimen sides, subjected to
tensile and compressive bending stresses, respectively.
24.3.1 Surface Characteristics
The effect on the surface roughness exerted by the shot peening treatments is quantified in Table 24.3, where the results of
the optical profilometer measurements made on an assessment area 636 477 mm2 are summarized. It can be noted that the
shot peening treatments introduce a limited increment in the surface roughness with the respect to the as-received condition.
The surface morphology of the as-received, C120 and V40 condition is depicted in Fig. 24.3a–c, respectively. Notably,
the surface morphology of the as-received condition is characterized by striations caused by the rolling process. The surface
of the C120 condition presents small dimples typical of gentle superficial peening treatments. On the contrary, the surface
morphology of V40 is completely different, a rougher surface is evident with respect to the as-received condition, and
however the superficial structure of impact craters is not clearly visible.
A comparison among the microhardness profiles measured for both the peening treatments and the as-received condition
is illustrated in Fig. 24.4a. Both peening treatments are very superficial, the maximum hardness is reached 10–20 mm below
the surface, the depth of the surface layer interested by the plastic deformation is about 60 mm. The C120 treatment causes a
more intense work hardening, and hence higher microhardness values, than the V40 treatment.
XRD measurements were carried out on the fatigue samples in order to characterize the residual stress field prior to
fatigue testing. The obtained stress profiles are illustrated in Fig. 24.4b. Two measurements per peening variant were
performed in order to account for the variability in the residual stress field. Both peening treatments display a sub-superficial
compressive residual stress peak located nearly 15 mm below the surface and a depth of the surface layer interested by
compressive residual stresses equal to about 50 mm. The C120 treatment induces higher residual stresses as compared with
the V40 treatment. The repeatability of both residual stress measurement and peening treatment is very good within a depth
of about 15 mm, where the fatigue response is mostly dictated [6].
24.3.2 Fatigue Behaviour
The results of the pulsating bending fatigue tests as well as the P50 fatigue lines are compared in Fig. 24.5a, b for smooth and
notched specimens, respectively, in the different material variants considered. The parameters representing the fatigue
curves corresponding to 50% of failure probability, according to (24.1) and the results scatter, expressed by (24.2) are listed
in Table 24.4.
All the peening treatments considered were effective in prolonging the fatigue life of the material, even in the presence of
stress concentration, as well as in reducing the large scatter in fatigue results displayed by the virgin material. This
improvement depends on the applied load, being more remarkable for load levels corresponding to shorter fatigue lives,
leading however to higher values of the slope in the P50 fatigue line. Therefore, the increment in fatigue resistance due to
shot peening steadily declines during fatigue life. While the plain fatigue performance at 108 cycles of the peened variants is
comparable with that of the virgin material, an increment of about 20% in notch fatigue resistance is still present in this
Fig. 24.3 Surface morphologies measured through optical profilometer: (a) as-received, (b) C120 peening treatment, (c) V40 peening treatment.
The assessment area is 636 477 mm2
a b
Fig. 24.4 (a) Microhardness profiles of the as-received and peened variants, obtained by averaging the results of three tests. (b) Initial residual
stress profiles of the peened variants measured by XRD technique. Two measurements per peened variant were carried out in order to account for
the variability in the residual stress field
a b
Fig. 24.5 Pulsating bending fatigue curves of the as-received and peened conditions: (a) plain, (b) notched fatigue specimens. Run-out tests are
marked by arrows
Table 24.4 Principal results of fatigue tests

W€
ohler curve
Sample geometry Condition k sP50 (MPa) Ts
Smooth As-received 33 200 1:1.22
C120 12 555 1:1.19
V40 12 590 1:1.10
Notched (R2) As-received 46 120 1:1.21
C120 17 315 1:1.12
V40 18 280 1:1.13
k, sP50 slope and y-intercept of the W€
ohler curve according to (24.1), Ts results scatter
fatigue regime. The lightest V40 peening treatment, despite lower surface residual stresses and slightly higher roughness, is
more effective in improving the plain fatigue resistance with respect to the C120 treatment, which, conversely, performs
better in the notched specimens. This fact might be imputed to the particular surface morphology of the V40 condition,
which presumably exerts a less detrimental effect on the plain fatigue response as compared with that exhibited by C120. On
the contrary, the notch fatigue strength is dictated by the residual stress field rather than surface roughness, since the major
contribution to stress concentration is given by the notch itself.
The comparison among the fatigue parameters listed in Table 24.4 points out that the peened notched variants display a
less steep P50 fatigue line with respect to the corresponding unnotched conditions. This results in a notch sensitivity factor q,
which steadily declines during fatigue life for the peened variants, whilst it remains approximately constant in the virgin
material, as illustrated in Fig. 24.6.
Figures 24.7 and 24.8 and 24.9d illustrate the evolution of the residual stress field at two loading levels for the C120 and
V40 treatment respectively. Figures 24.7a (8a) and 24.7b (8b) refer to the specimen surface subject to compressive and
tensile bending stresses, respectively. It can be noted that some relaxation of the residual stress field occurred on the outer
layer of the specimen side subjected to compressive bending stresses, the more pronounced the higher the load levels,
especially in the C120 condition. Conversely, no appreciable relaxation occurred on the specimen side subjected to tensile
bending stresses. This confirm the observations made in Refs. [6, 7] that relaxation is more like a “quasi-static” effect, due to
the achievement of the material’s plasticization when the compressive bending stresses are superimposed to the compressive
surface residuals stress field. Cyclic relaxation seems to be negligible, since the application of tens of millions fatigue cycles
do not significantly alter the surface residual stress field on the tensioned side of the samples.
SEM observations carried out on the fracture surfaces of the specimens tested in the very high-cycle fatigue regime
revealed sub-superficial fatigue crack initiation in the peened conditions, both in smooth (Fig. 24.9a) and notched
(Fig. 24.9b) samples. On the contrary, crack nucleation was found to occur of the surface of plain (Fig. 24.9c) and notched
(Fig. 24.9d) specimens in the as-received condition. Apparently, shot peening prevents the surface from fatigue crack
initiation, which in turn takes place on the sub-superficial layers, which are not subjected to compressive residual stresses.
Fig. 24.6 Notch sensitivity factor dependence upon fatigue life for the three material conditions investigated
a b
Fig. 24.7 Evolution of the residual stress profile during fatigue life in smooth samples subjected to C120 treatment. (a) Tensioned, (b) compressed
side. Specimen tested at 270 (400) MPa failed after 44.6 106 (530,000) cycles
After nucleation, the fatigue cracks must propagate through compressive residual stresses, which cause it to slow down, thus
prolonging the fatigue life of the component. However, the contribution to the overall fatigue life given by a reduced crack
propagation rate is proportionally smaller in the very high cycle fatigue regime. This might explain the steeper W€ ohler
curves of the peened variants with respect to the virgin material.
24.4 Conclusions
The plain and notch fatigue strength of shot peened aluminium alloy Al-7075-T651 has been experimentally investigated
exploring fatigue lives comprised between 105 and 108 cycles. The following conclusions can be drawn:
1. Shot peening is an effective method to improve the fatigue endurance of Al-alloys even in the presence of geometrical
discontinuities. Notch sensitivity is greatly reduced by shot peening.
a b
Fig. 24.8 Evolution of the residual stress profile during fatigue life in smooth samples subjected to V40 treatment. (a) Tensioned, (b) compressed
side. Specimen tested at 290 (420) MPa failed after 22.8 106 (390,000) cycles
Fig. 24.9 SEM micrographs of the fracture surfaces around the fatigue crack initiation sites: (a) plain C120 (sa ¼ 133 MPa, Nf ¼ 37.7 106),
(b) notched V40 (sa ¼ 105 MPa, Nf ¼ 38.9 106), (c) plain as-received (sa ¼ 133 MPa, Nf ¼ 25.0 106), (d) notched as-received condition
(sa ¼ 95 MPa, Nf ¼ 3.9 106)
2. The beneficial effect of shot peening diminishes with increasing fatigue cycles, presumably due to the occurrence of sub-
superficial crack initiation.
3. Residual stress relaxation occurs only when compressive bending stresses are superimposed to the surface compressive
residual stress field.
References
1. ASM Handbook (1991) Properties and selection: nonferrous alloys and special-purpose materials, Vol 2, 10th edn, American Society for
Metals, Materials Park, Ohio
2. Bian JC, Tokaji K, Ogawa T (1994) Notch sensitivity of aluminium-lithium alloys in fatigue. J Soc Mat Sci Japan 43:840–846
3. Guagliano M, Vergani L (2004) An approach for prediction of fatigue strength of shot peened components. Eng Fract Mech 71:501–512
4. Wagner L (1999) Mechanical surface treatments on titanium, aluminum and magnesium alloy. Mater Sci Eng A263:210–216
5. McClung RC (2007) A literature survey on the stability and significance of residual stresses during fatigue. Fatigue Fract Eng Mater Struct
30:173–205
6. Benedetti M, Fontanari V, Scardi P, Ricardo CLA, Bandini M (2009) Reverse bending fatigue of shot peened 7075-T651 aluminium alloy: the
role of residual stress relaxation. Int J Fatigue 31:1225–1236
7. Benedetti M, Fontanari V, Monelli BD (2010) Numerical simulation of residual stress relaxation in shot peened high-strength aluminium
alloys under reverse bending fatigue. ASME J Eng Mater Technol 132:011012-1-9
8. Benedetti M, Fontanari V, Santus C, Bandini M (2010) Notch fatigue behaviour of shot peened high-strength aluminium alloys: experiments
and predictions using a critical distance method. Int J Fatigue 32:1600–1611
9. Benedetti M, Fontanari V, Bandini M (in press) A simplified and fast method to predict plain and notch fatigue of shot peened high-strength
aluminium alloys under reverse bending. Surf Coat Technol DOI:10.1016/j.surfcoat.2011.12.008
10. Oguri K (2011) Fatigue life enhancement of aluminum alloy for aircraft by fine particle shot peening (FPSP). J Mater Process Technol
211:1395–1399
Chapter 25
The Application of the Combined Creep and Rate-Independent
Plasticity Model to Turbine Components Life Prediction
Alexander Staroselsky and Brice N. Cassenti
Abstract The understanding of micromechanics of the high temperature creep, plasticity, and damage accumulation in
single crystal nickel base superalloys is important for the design of turbine blades and vanes in advanced commercial and
military gas turbines. We developed the crystallographic–based viscoplastic model for non-isothermal high temperature
cyclic deformation and coupled it with damage kinetics. Damage accumulation causes tertiary creep and shear localization
around local concentrators, which is essential for airfoil life prediction. We develop a new robust, computationally efficient
rate-independent crystal plasticity approach and combined it with creep flow rule calibrated for Ni-based superalloys. Engine
operating conditions were assembled into a mission profile, an input to the transient thermal analysis. Non-linear transient
thermal-structural analysis of a turbine airfoil provides the illustration of the life prediction for different engine operating
conditions. We compare the predictions for low cycle and thermal mechanical fatigue for smooth specimen tests and for
notched specimens and demonstrate the transition from safe life to damage tolerant approach. The constitutive model has
been implemented in the commercial finite element software ANSYS as a material user routine. The effects of single crystal
orientation and stress relaxation and redistribution, cycles parameters, strain levels, and dwell time have been numerically
obtained. The model results are used for predicting of thermal cyclic behavior and damage of single crystal airfoil.
Keywords Constitutive model • Crystal plasticity • Airfoil • Life prediction • TMF • Creep
25.1 Introduction
It is well understood that increasing the turbine blade tip speed and inlet temperature offer the greatest improvements to the
fuel efficiency and power output of turbomachinery engines. Turbine blade tip speed determines the maximum pressure
ratios that can be obtained by the turbine, which, in turn, limits the maximum engine power and efficiency. The purpose of
the turbine is to extract work from the high pressure and temperature core flow. As a consequence, turbine airfoils are
subjected to very high temperatures and centrifugal loads, which result in increased deterioration and degradation due to
various forms of damage. The phenomena of oxidation, corrosion, thermo-mechanical fatigue, and creep ultimately set the
limit of the entire engine by limiting the core gas temperature entering the turbine and the engine rotor speed, which
ultimately constrains the time allowed between engine overhauls. In particular, the amount of time an engine can spend in
continuous service depends on the life of the worst airfoil, which is limited by the service-life capability, also referred to as
the durability, of the hot-section airfoils.
Turbine airfoils are made from nickel-based super alloys and can withstand metal temperatures up to of 1,260 C (2,300 F)
[1]. A typical modern single crystal superalloy has the ordered L12 structure, with a matrix based on a g face-centered cubic
structure and regular cubes of g0 phase which occupy from 65% to 70% of the volume. This two-phase structure of a
superalloy provides improved creep strength at high temperatures because the hard phase and the phase boundaries provide
obstacles to dislocation motion. With the increase in temperature and loading, the damage also accumulates more quickly
A. Staroselsky (*)
Pratt and Whitney Co., 400 Main Street, MS 165-16, East Hartford, CT 06108, USA
e-mail: Alexander.staroselsky@pw.utc.com
B.N. Cassenti
Mechanical Engineering Department, University of Connecticut, 191 Auditorium Rd. U-3139, Storrs, CT 06269-3139, USA
214 A. Staroselsky and B.N. Cassenti
Fig. 25.1 Strain rate versus

creep strain for <001> creep
of single crystal PWA 1484
at 850 C and an engineering
stress of 550 MPa
leading to the part degradation. The use of internal cooling and external film cooling for turbine airfoils have allowed the
airfoils to withstand greater heat loads, resulting in increases in inlet gas temperatures. Convection cooling works by passing
cooling air through passages internal to the blade transferring heat by conduction through the blade, and then by convection
into the air flowing inside of the blade. It might reduce the amount of metal creep but increases the thermal gradients, and,
subsequently, increases the probability of thermal mechanical cracking. Thin ceramic thermal barrier coatings (TBC) can
shield the temperature to which the underlying alloy is exposed by hundreds of degrees Celsius. The higher temperature
operation afforded by the ceramic top coat of the thermal barrier coating allows the engine to operate with greater power and
fuel efficiency. External (or so called film) cooling techniques allow another 260 C (500 F) in inlet gas temperatures by
producing a protective layer of cold air on the surface of the airfoils, and it can be targeted upon specific areas of the turbine
blade that absorb the most heat.
In this work we use the so-called a loosely coupled conjugate heat transfer (LC-CHT) analysis where the hot gas
(external), the flow inside the internal coolant flow passages, and the heat conduction in the airfoil solid are coupled only at
their interface boundaries. In order to perform this calculation, the proper treatments of the interface conditions require that
wall temperature and heat flux are balanced. The heat transferred to the solid blade is transmitted to the internal cooling air,
which is injected out of the film cooling holes on the surface of the turbine airfoil. This film injection cools down the external
gas temperature and it disrupts the boundary layer, thereby augmenting the external heat transfer coefficient. In the loosely
coupled approach for transient aero thermal heat conduction in a turbine airfoil, two iteration loops are constructed; the outer
loop is required for transient analyses where the turbine operating conditions, such as rotor speed, engine air flow rate,
turbine inlet temperature, etc., change with time. These thermal results are weakly non-linear because the internal cooling air
temperature is a function of the wall temperature, which must be taken from a heat transfer model of the solid. The inner
iterative loop is required to achieve the heat balance. This is where the loosely coupled conjugate iteration is performed so
that the heat transfer from the hot gas into the turbine blade is balanced with the heat flux through the metal, which is picked
up by the internal cooling air and ejected through film cooling holes and slots. The local metal temperature is the result of the
convergence of the described above iteration routine.
Creep in single crystal superalloys is highly anisotropic. Single crystal crystallography leads to orientation dependent
creep behavior and, hence, it is extremely important to be able to predict the creep rates in different crystallographic
orientations, especially due to non-homogeneous temperature distribution in the airfoil. This anisotropic behavior dictates
the choice of the modeling technique (i.e., a crystal plasticity based model). The proposed state variables approach for
deformation simulations of L12 single crystals gives us a tool to deduce the physics-based behavior of the structure and relate
it to the microstructure evolution. The general form of the incremental governing relation for the visco-plastic shearing may
be written as Dga ¼ g_ a ðta ; sa ; oa ; ra ; TÞDt ¼ g_ acreep Dt þ Dgaplast where T is the temperature, sa is the deformation resistance
and oa is a kinematic hardening of ath slip system; ta is resolved shear stress, and di is the amount of damage associated
with different microstructural damage mechanisms.
It is typical for Ni-based superalloys at operational temperatures to creep without steady state regime. This response is
referred to as “sigmoidal creep” [2]. During the primary creep regime the strain rate decreases with increasing strain until it
reaches a minimal value. After this point, the creep strain rate increases without a long constant strain-rate interval. Also, at
high temperatures and moderate values of the applied load it is observed that creep can occur without any noticeable primary
stage. Figure 25.1 illustrates the sigmoidal character of PWA1484 creep deformation along <001> crystallographic
25 The Application of the Combined Creep and Rate-Independent Plasticity Model. . . 215
direction. The latter tertiary creep interval has a much steeper slope and is associated with active creep cavitations and
deformation localization near the fracture surfaces. Predicting tertiary creep implies that damage mechanisms must be
modeled. Creep deformation mechanisms dominate at high temperature. Time independent plasticity effects become very
important at relatively low temperature (in our case in the range 750–850 C). A combination of these two mechanisms
provides a description of the complete thermal – deformation cycle when extensive deformation takes place at high peak
temperatures, and the residual strain/stress generated on cooling or even at shut down, which can eventually lead to the
failure initiation.
Hot section components in advanced gas turbine engines experience severe cyclic temperature gradients and mechanical
loads, particularly during takeoff and landing operations. As a consequence, thermal mechanical fatigue (TMF) is a major
life-limiting factor for cooled gas turbine blades. TMF cracking is usually taken place when inelastic deformation and
corresponding energy dissipation is suppressed at low temperatures while creep occurs at high temperatures and generates
stress redistribution affecting low temperature state. Critical locations in turbine airfoils are subject to the combined
influences of thermally driven strain transients and creep damage resulting from stresses during operation [3] and might
include pressure and suction sides and both leading and trailing edges [4, 5]. The test data obtained on smooth specimens is
traditionally used for the design purposes. As a result, safety-based design approaches might be employed. However, for the
damage tolerant assessment, it is essential, if not critical, that TMF crack initiation, location, propagation path, and growth
rate information are developed as well. As have been already noted, most TMF cracks in airfoils start from the cooling holes.
Thus, a new thermal fatigue experimental technique and appropriate numerical analysis are needed to measure the
structural life of a specimen (part) containing through holes similar to the ones that drilled in a cooled airfoils.
The plan of the paper is as follows: We present the major constitutive relations in the next section. In Sect. 25.3 we show
the results of the creep-plasticity model calculations to predict the state of the notched specimen and compare the results
with experimental observations. We also show the results of the analysis of a typical aerofoil. We finish the paper with some
concluding remarks.
25.2 Constitutive Relations
The overall plastic response is taken as a sum of responses from small regions of a single crystal playing the role of
representative volume elements (RVE). The deformation of a crystal is taken as the sum of contributions from overall elastic
distortion and generalized plastic deformation. In this formulation creep will be included to the total inelastic strain rate as
will be shown in this section. Our model assumes that standard viscoplastic models can be used to represent the macroscopic
mechanical response of the single crystal material. We use the 12 octahedral slip systems <110> (111). At high
temperatures, the cube slip systems <110> (001) may also contribute to maintaining the plastic flow of superalloys.
Finally the material model parameters are different for the octahedral and cube slip systems. We use Orowan’s assumption
that the creep strain rate is proportional to the density of the mobile dislocations, rm. If we denote an arbitrary reference
dislocation density throughout as r0 , then the non-dimensional parameter rrm serves as a measure of the mobile dislocation
0
density, and can be used to predict tertiary creep. The constitutive law for the secondary-tertiary creep strain rate along ath
a
slip system, fg_ c g is to be written as follows:

ram ta oa n Q
g_ a ¼ g_ 0 sgn ð t a
o a
Þ exp ; (25.1)
r0 sa kY
where g_ 0 ðTÞ is a temperature dependent time parameter, sa is the deformation resistance of ath slip system; ta is the
resolved shear stress. The Arrhenius term with the activation energy Q accounts for the temperature changes and reflects
the accumulated damage along the particular slip system; oa is the slip system back stress, n is the creep exponent assumed
to be equal to 3 in this work. a p P ab b
We have chosen latent hardening evolution to be described by (Asaro 1983) s_a ¼ h0 1 ss h g_ , with
b
hardening matrix hab ¼ q þ ð1 qÞdab for temperature dependent h0 and s . The back stress (or kinematic
hardening) has a limiting saturation value, o1 ¼ cc12 , corresponding to the end of the primary creep stage which evolves
according to the following relationship [6]
o_ a ¼ C1 g_ a C2 jg_ a joa ¼ lðg_ a o1 jg_ a joa Þ (25.2)

The back stress requires two additional experimentally measured coefficients c1 and c2 that are explicit functions of
temperature. It is important to note that hardening terms indirectly account for the microstructure evolution during the creep.
Results reported here have been obtained by using the hardening expressions shown above.
Dislocation generation and motion represents a non-recoverable state for the material. We postulate that dislocation
slip
nP
generation rate is proportional to the rate of entropy production which can be expressed by S_ ta oa _ a
sa g 0, where nslip
a¼1
is the number of active slip systems, which is 18 (12 octahedral and 6 cube) for Ni-base superalloy. We consider two types of
the dislocations: mobile and pinned. Using concepts from chemical kinetics we have chosen to represent the dislocation
density evolution as two body interactions. We assume that dislocation immobilization takes place when two corresponding
dislocation loops interact with each other. Note that since fg_ c ga is already a linear function of the mobile dislocation density,
then for two body interactions the dislocation densities growth rates can be taken as the product of g_ a and linear functions of
the current dislocation densities. Assuming the existence of dislocation densities saturation values rss m for mobile
dislocations and rss
p for pinned, we obtain relations for mobile and pinned dislocations along each slip system as following:
a !
a 2 a
t oa a
e2 rss
m þ rp rp e rm
ss
r_ am¼M g_ in
sa r0
a ss a
t o a r p rp
r_ p ¼ P a
g_ in a (25.3)
s r0
where M and P represents specific time constants, different for octahedral and cube slip systems, e2 is a positive constant.
Equation 25.3 includes the annihilation of mobile dislocations and also includes their conversion to pinned dislocations. The
pinned dislocations grow at a rate that is proportional to the mobile dislocation density because of the presence of the plastic
strain rate term. Dislocation densities evolutions have strong temperature dependence because they are functions of inelastic
strain rates and the expression for g_ a contains Arrhenius term.
In order to predict primary creep in Ni-based superalloys, we have developed a rate-dependent crystal plasticity formalism
with a threshold stress on each slip system. Thus, dislocations cross-slip causes strain hardening which in turn decreases the
rate of creep deformation. Increases in the dislocation density and in the number of dislocation pile-ups makes further
deformation more and more difficult, resulting in the transition of creep from the primary stage to a secondary one. The rate of
the primary creep gradually decreases with deformation or with the development of slip resistance. Since dislocation
strengthening is proportional to the square root of the dislocation density (for example [7, 8]), we introduce the threshold
value to also be proportional to the square root of the total dislocation density. The threshold stress reflects the resistance to
plastic flow arising from hardening associated with an increase of dislocation density and cross-slip. All threshold parameters
as well as hardening parameters depend on temperature.
* pffiffiffiffiffi+n
ta op a k rap ta o p a Q
g_ aprimary ¼ g_ 0 sgn exp (25.4)
sa sa kT

a if a>0
Where hai ¼ and k is temperature dependent fitting parameter. In the formula for the back stress, parameter
0 if a 0
l is a characteristic time parameter describing how fast the back stress converges to its saturation value o1 . With an
increase in primary creep deformations, dislocation density increases thus reducing the applied effective stresses to very
small values. Thus, the primary creep slip rate gradually decreases from a significant value to zero.
The back stress during the primary creep stage is expressed as follows:

o_ ap ¼ lp g_ a o1 jg_ a joap (25.5)
Primary creep is much faster process than secondary and tertiary ones. To match the experimental observations a value of
lp is smaller than the value of l in (25.8) while the saturation value o1 is the same for all creep stages.
The total creep deformation rate for each slip system is the direct sum of tertiary creep Orowan’s type g_ asec expression
(25.1) and the primary one (25.4) as: g_ acreep ¼ g_ aprimary þ g_ asec . Primary creep has been mainly observed in Ni-based superalloy
at relatively low temperatures and high applied nominal stresses. Typical creep curves obtained at 760 C (1,400 F) and the
model predictions are shown in Fig. 25.2.
Fig. 25.2 Model predictions

and creep test results at
temperature 760 C for
different crystallographic
orientations
Due to the creep the stress in the part redistributes and some, mostly non-creeping, sub-volumes of the parts experience
very high stresses leading to plastic deformation. The plasticity causes severe damage significantly reducing service life. The
details of the presented rate-independent formulation can be found in Ref. [9]. Here we briefly summarize the idea how to
combine the computational benefits of rate independent plastic formulation with h the advantages of a power method
affirming the uniqueness of active slip increments calculations.
a a a m
We postulate that a plastic strain rate is proportional to the rate of plastic work, W_ , g_ a ¼ tðso
p a
t o a
W_ p . By dividing
a Þ2 s
dep
both parts on W_ we get ijp ¼ f ðW p Þ, which makes ep an explicit function of W p and subsequently the model can be labeled
p
dW ij
as rate independent. Here h xi is the unit ramp function of xand sa is a temperature dependent latent hardening material
parameter along slip system a. We use the ramp function to satisfy Kuhn-Tucker condition or in other words, there is no
plasticity under unloading, or in other words if the energy is removed from the material (W_ <0) then there is no plastic
p
deformation occur. We use the advantage of the power model to limit activities of the slip systems with small resolved shear
stresses and to assure the unique choice of active slip systems. Values of the exponent m 30 provide good selection of slip
activities and do not cause computational problems specific for stiff rate-dependent power models with very high values of
the exponent values. We generalize the formulation above by substituting the plastic work rate with a weighted sum of the
total and elastic work rates. We replace the plastic work rate in the following way:
m
a ta oa ta oa _
_ _p

g_ ¼ 2 a Wk WW (25.6)
a
ðs Þ s
P
where k is a temperature dependent material parameter. The total work rate is given by W_ ¼ sij e_ ij ¼ sij e_ ij , and the plastic
P i;j
work rate is W_ ¼ ðta oa Þ_ga . After simple algebraic transformations, we get the final form of the time independent
p
plastic strain rate thata can be written in incremental formulation as follows:
*P +
sij Detotal
ta oa m ta oa
ij
ð1 k Þ i;j¼1;3
sa sa sa
a (25.7)
Dg ¼
P ta oa mþ2
1 k
a sa
Thus, we have formulated a rate-independent flow rule which automatically guarantees a unique selection of active slip
systems at each time increment without any additional criteria and constraints. We also introduced one more model “weight”
parameter 0 k<1. If the value of this parameter is close to zero, yield values change gradually as it usually observed
during hardening. In the limiting case of k ! 1 there will be a sudden change in the plastic strain rate when ta oa ¼ sa ,
and, hence, sa can be interpreted as a yield stress. The described model gives good predictions for different crystallo-
graphic orientations.
25.3 Application to the Component Life Prediction
We have completed a life prediction analysis of specimens taken from different sets of creep, TMF and LCF tests at different
peak temperatures and test conditions. The algorithm developed for simple specimens were used for the analysis of a blade.
It is important that the damage
" is the combination of #dissipated energy rate DW per loading cycle (usually generated at high
R P
temperature): DW ¼ ðta oa Þ_ga dt and an effect of maximum principal stresses acting at low tempera-
cycle slipsystemsa¼1;18
ture (if we consider non-isothermal conditions) (sm
max ). High temperature dwell is accounted with the help of the “effective
Rt
hot time” Arrhenius type correction: DT ¼ exp kTQðBÞ dB. A regression analysis of the test data yields for the following
0
empirical formula to predict number of cycles to failure:
" #
1 wDW 1 1
¼e þ k (25.8)
Nf A1 =smmax A2 = DW DT
where smax is the maximum principal stress, A1 ; A2 ; w; m; and k are empirical fitting constants. The developed formula has
been implemented into USERMAT ANSYS in the form of damage rate evolution and the quality of life predictions are
illustrated in Fig. 25.3.
Figure 25.4a illustrates damage distribution in the analyzed blade after a combined mission cycle shown in Fig. 25.4b in
normalized form. The variation in damage can be used to identify locations where airfoil cracking will occur and to allow a
redesign of the airfoil before production starts. It helps to eliminate expensive iterations of blade production and testing to
assess durability. In this special modeling case the potential crack will take place in the pressure side. The estimated life of
the airfoil in this particular model calculation is ~350 actual cycles, which is close to block tests observations of a similar
type blade.
The damage calculations show the peaks at the trailing edge and at the blade tip. Actual tests demonstrated that the blades
were cracked near the predicted zones. Thus, the conceptual process for damage evaluation is established and tested.
4
Log(Predicted Life)
Fig. 25.3 Life prediction

is within 2 interval against 0
measured data for both OP 0 1 2 3 4 5
TMF and isothermal LCF tests Log(Measured Life)
Fig. 25.4 (a) Calculated damage distribution in the blade after 40 h of the block test; (b) schematic picture of the normalized temperature test
mission profile
25.4 Concluding Remarks
The constitutive model has been developed to predict anisotropic creep and plasticity. The model was implemented in the
commercial finite element software ANSYS as a material user routine. The equations governing the mechanical response
have been calibrated using original experimental data. Our damage model bridges the gap between dislocation dynamics and
the continuum mechanics scales and allows prediction of failure initiation as well as creep of single crystal superalloy
components. The analysis shows that the value of dissipated energy DW over the cycle controls the life of the specimens.
Such a damage measure accounts for the combined effect of fatigue and creep/plasticity through the width of the hysteresis
loop and cannot be decomposed into the direct summation of the damages caused by fatigue and inelastic dissipation as
is widely used in engineering practice. From this point of view the creep-fatigue interaction is a crucial factor in decreasing
the structural part life. This fully coupled visco-plastic–elastic model based on crystal morphology was combined with
newly developed transient thermal analysis for the airfoils in order to predict damage evolution. The developed method is
efficient enough to handle transient finite element analysis with 200,000 elements.
References
1. Lakshminarayana B (1996) Fluid dynamics and heat transfer of turbomachinery. Wiley, New York
2. Levitin V (2006) High temperature strain of metals and alloys. Wiley-VCH, Weinheim, p 171
3. Cowles BA (1996) High cycle fatigue failure in aircraft gas turbines: an industry perspective. Int J Fract 80:147–163
4. Heine JE, Warren JR, Cowles BA (1989) Thermal mechanical fatigue of coated blade materials. Wright-Patterson Final Report ERDC-TR-89-
4027, September 1989
5. Krutiy V (2009) Sub-modeling of thermal mechanical fatigue crack propagation. MS thesis. Rensselaer Polytechnic Institute
6. Stouffer DC, Dame LT (1996) Inelastic deformation of metals. Wiley, New York, p 502
7. Kocks UF (1976) Laws for work hardening and low-temperature creep. J Eng Mater Technol 98:76–85
8. Mughrabi H (1975) Description of the dislocation structure after unidirectional deformation at low temperatures. In: Argon A (ed) Constitutive
equations in plasticity. MIT Press, Cambridge, MA, pp 199–251
9. Staroselsky A, Cassenti BN (2010) Combined rate-independent plasticity and creep model for single crystal. Mech Mater 42:945–959
Chapter 26
Modified Constitutive Relation Error Strategy for Elastic
Properties Identification
M. Ben Azzouna, P. Feissel, and P. Villon
Abstract The aim of this paper is to present an inverse approach dedicated to the exploitation of full-field measurements, to
identify elastic properties of heterogeneous materials, such as composites. The method is based on the modified constitutive
relation error principle and could be split in two steps. The first one consists in defining mechanical fields from the available
theoretical and experimental data, for a fixed set of mechanical parameters, by the minimization of a criterion allowing
a compromise between constitutive equation and measurements adequacy. The second step takes the form of minimizing a
cost function defined by using these fields, to identify the sought material properties. Moreover, the robustness of the method
was tested on some numerical examples where white Gaussian perturbations were added to displacement values to simulate
an experimental errors.
Keywords Modified constitutive relation error • Identification • Full-field measurements
26.1 Introduction
The intensive use of composite materials in aeronautic industry requires the knowledge of the parameters governing the
mechanical constitutive material behavior. When focusing on their elastic behavior, composite materials have generally
heterogeneous properties. It is then necessary to characterize them in a reliable way. Full-field measurements are often
needed to exploit heterogeneous tests, hardly possible with classic measure tools like extensometric gages which provide a
limited number of measurement points. It leads to the development of digital images correlation (DIC) techniques [1–3],
which constitute one of the main breakthroughs of the last 30 years. Thus, it is necessary to develop identification strategies
adapted to the fullness of this type of measurements, and based on inverse approaches.
Besides, measurement perturbations, which are inherent in experimental tests, affect the accuracy of identified properties,
especially in elasticity, when the magnitude of strain fields is relatively low. Thus, it is important to take them into account
and use a strategy which is less dependent on perturbations.
Numerous identification methods have been proposed and adapted to the use of full-field measurements. A review of
the most widespread methods (Finite Element Model Updating (FEMU), Virtual Fields Method, Equilibrium Gap Method,
Reciprocity Gap Method and Constitutive Relation Error (CRE)) can be found in [4]. Among the identification
procedures, the so-called modified constitutive relation error (MCRE) turns out as a robust method towards measurements
perturbations [5]. At first, it was used for model updating in vibration dynamics within a validation framework [6, 7]. Then,
the method was adapted to the exploitation of full-field measurements in [8–10].
The present paper presents a formulation of an identification problem based on the MCRE, leading to the minimization of
a cost function composed of two parts. One related to the constitutive equation and another to the distance between the
simulation and the experimental measurements.
An important point to be noted is that the modeling of the boundary conditions represents a shortcoming of some inverse
methods. The present formulation allows identification even if they are not available. No hypothesis on boundary conditions
is thus required, for example, to take into account the distribution of the load on a sample boundary.
M. Ben Azzouna (*) • P. Feissel • P. Villon

Roberval Laboratory, University of Technology of Compiègne, Centre de Recherches de Royallieu, 60205 Compiègne, France
e-mail: mouldi.ben-azzouna@utc.fr
222 M. Ben Azzouna et al.
In the following, we present the identification framework and describe the various steps of the MCRE.
The description of the numerical implementation of the presented formulation and the adopted resolving algorithm will
be discussed.
Identification results will be presented and the robustness of the method will be illustrated on a numerical case.
26.2 The Modified Constitutive Relation Error
26.2.1 Identification Framework
Let us consider a structure subjected to a mechanical load, defined in a domain O and governed by the equations of
continuum mechanics. Here, we are looking for its elastic properties noted Y. We make the hypothesis that boundary
conditions are unknown, and one disposes only of displacement measurements noted u~, available on a given part Om of the
entire domain, as shown in Fig. 26.1. Thus, the mechanical equations we have are:
(26.1a)
(26.1b)
(26.1c)
(26.1d)
with C is the stiffness tensor, u the displacement vector, s the stress tensor and e the strain tensor. Om is a non-zero
measurement subdomain, defined as Om O. Measurements are stored in vector u~, defined as u~ 2 Rn , with n: the number of
measurement points. The formulation of the identification problem, based on the so-called modified constitutive relation
error (MCRE) is a variational inverse approach split in two steps. The first step consists in building admissible stress and
strain fields for a fixed set of mechanical parameters Y, by using the whole experimental and theoretical information in our
disposal. This step defines the reference problem. The second one consists in finding the best model’s parameters Y by
minimizing a cost function derived from the solution fields of the previous step.
26.2.2 The Basic Problem
One of the MCRE’s principles relies on the partitioning of the previous equations into reliable and unreliable equations.
Then, during the identification process, the method consists in the exact verification of the equations which are considered to
be reliable, and the relaxation of uncertain relations, which are verified at best by a weighted minimization of the MCRE
functional. The two relation’s sets are defined as:
1. Reliable equations: mechanical equilibrium (26.1a) and kinematic compatibility (26.1b).
2. Uncertain equations:
• The constitutive equation (26.1c), containing the sought behavior properties.
• Measurements (26.1d), because they are spotted by perturbations. This point represents the difference between MCRE
and CRE methods and allots the adjective modified to the presented procedure.
BC2
Ω
∂2Ω
Ωm
Fig. 26.1 Reference problem ∂1Ω
related to the identification
problem BC1
26 Modified Constitutive Relation Error Strategy for Elastic Properties Identification 223
This separation leads to the following formulation of the identification problem, which defines the reference problem for a
fixed parameters Y:
Find the couple ðu; vÞ that minimizes:
Z Z
1 1
J ðu; vÞ ¼ sðv uÞ : eðv uÞdO þa ðu u~Þ2 dOm
2 O 2 Om (26.2)
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
J1 J2
under reliable equations constraints (26.1a) and (26.1b).with:

• u and v: the sought admissible fields.
The solution of the basic problem is denoted as:
8
< ðuðYÞ; vðYÞÞ ¼ arg min J ðu; vÞ
u;v
: under constraints (26.1a) and (26.1b)
Remark. v is a displacement fields referring to the statically admissible stress. The presented formulation is written in a
displacement way. Actually, if the formulation was expressed in a stress way, one can show that: ∃v such as s ¼ C eðvÞ,
where C is the Hooke’s tensor.
The functional J is composed of two parts. A first component J 1 relative to the specific constitutive equation, and a
second one J 2 expressing the gap between simulated and measured fields. Its minimization leads to the resolution of a linear
system, and then, the achievement of the admissible fields building (see Sect. 26.3.1). a is an hyperparameter used to make
the functional terms have the same unit, and plays also the role of a weighting factor allowing to balance their magnitudes.
Indeed, J 1 and J 2 represent respectively an energy and a square of a distance. It is thus necessary to make them
dimensionless by using their orders of magnitude. For example, in the application which will be shown in Sect. 26.4, J 2
is divided by the square of measurements, when J 1 is divided by an estimation of the elastic energy of the reference
problem. Concretely, the elastic energy of the reference problem can be estimated by approximating material properties in
the case of an homogeneous behavior, or by computing a mean value of each material parameter over the whole structure, if
the behavior is heterogeneous. Thus, a is just the fraction of these two quantities.
26.2.3 The Identification Problem
After solving the reference problem (26.2), admissible fields are obtained and one can now proceed seeking the parameters
Y, thanks to the minimization of a cost function G which has the same form of the functional J but is function of the optimal
fields u and v. Identification problem is then written as:
Y ¼ arg min GðY0 Þ

Y0
with GðY Þ ¼ J ðuðY0 Þ; vðY0 ÞÞ
0
(26.3)
26.3 Numerical Implementation
26.3.1 Discretization of the Basic Problem
To implement the presented method numerically, one should make the discretization of the continuous problem (26.2). The
choice here is to use the finite element method. Thus, after meshing the domain O, solution fields are represented by discrete
vectors containing displacement on the degrees of freedom, denoted U and V. The column vector U~ stores measured
displacements available on the measurement grid and K denotes the discrete rigidity matrix. In order to solve the numerical
problem, discrete formulation will be based on the partitioning of mesh nodes into two groups: internal nodes, denoted by the
index “i”, and border nodes denoted by the index “b”. When dealing with both internal and border nodes, the index “.” is
used. Thus, the problem (26.2) takes the following discrete form:Find (U, V ) that minimize:
(26.4a)
(26.4b)
One can remark that the equilibrium constraint of the reference problem stands only for the internal nodes. Thus, all the
lines of the rigidity matrix are concerned. In order to evaluate the distance-to-measurements term, a discrete operator P is
introduced and plays two roles: extracting an adequate size of the kinematic admissible field U defined on the whole domain,
to fit the subdomain Om if smaller than O, and projecting that extracted part on the measurement grid.To simplify
mathematical writing, we make hypothesis that P is only applicable on the internal nodes of the mesh. Finally, functional
J is minimized under the constraint (26.4b). Thus, a Lagrange multiplier L is introduced:
LðU; V; Li Þ ¼ J ðU; VÞ LTi ðK i: VÞ

@L @L @L
When studying its stationary, we obtain three equations @U ; @V ; @L ¼ ð0; 0; 0Þ. After doing the partitioning on both
internal and border nodes, we obtain the following simultaneous equations:
(26.5a)
(26.5b)
(26.5c)
(26.5d)
(26.5e)
26.3.2 Solving Algorithm
26.3.2.1 Discrete Reference Problem
From Eqs. (26.5c) and (26.5d), one can easily show that (Ub Vb) corresponds to the degrees of freedom of a rigid
body motion on the domain border. Besides, one can show that functional J is insensitive to that rigid body motion, that’s
why it’s chosen equal to zero (Ub ¼ Vb).Thereby, Eqs. (26.5b) and (26.5e) lead to a linear system as a function of U:

K ii þ aPT P K ib Ui aPT U~
¼
K bi K bi K 1
ii K ib Ub 0
By substituting Ui with Ub, Ub is gotten as a solution of the following linear system:

h i
1 1 T ~
K bi ðK ii þ aPT PÞ K 1 ii K ib U b ¼ K bi ðK ii þ aP PÞ aP U
T
(26.6)
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
F
M
Let us note that M becomes a rank-deficient matrix when Om is smaller than O. In that case, we choose a QR decomposition,
where Q is an orthogonal matrix and R an upper triangular matrix, such as:

R R2
M ¼ QR ¼ ½ Q1 Q2 1
0 0
Fig. 26.2 Gradient evaluation x 10−5

3.5
2.5
1.5
G
1
0.5
−0.5
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2
μ/μref
The solution of such problem (26.6) has a particular solution and a second part relative to the kernel of M. Since the
functional’s magnitude at the minimum is insensitive to the kernel choice, it’s chosen equal to zero. After the QR
decomposition, the problem solution is:
Ub ¼ R1 T
1 Q1 F (26.7)
The other quantities (Ui, Vi and Vb) are simply deduced from Ub. Next step is the identification of the material properties
by the minimization of the cost function GðUðY0 Þ; VðY0 ÞÞ.
26.3.2.2 Calculation of the Gradient
A great specificity of the present procedure is the use of the same functional to describe both reference and identification
problems. The interest of that particularity is an immediate access to the expression of the cost function gradient, very
helpful for the minimization step. Generally, when the gradient expression is not directly available, it can be computed by
using adjoint state method. In the present formulation, the reference problem and the adjoint state are coupled and solved
together. Then, let us write the expression of the cost function and its derivative, as function of the Lagrangian:
GðY0 Þ ¼ LðUðY0 Þ; VðY0 Þ; Li ðY0 Þ; Y0 Þ

@G @L @L @U @L @V @L @Li
0 ¼ 0 þ 0 þ 0 þ
@Yk @Yk |{z} @U @Yk |{z} @V @Yk @Li @Y0k
|{z}
¼0 ¼0 ¼0
The partial derivatives of the Lagrangian with respect to U, V and the Lagrange multiplier Li are equal to zero, because
0
solutions of the reference problem (26.4a)–(26.4b). Here, Yk denotes the kth sought material property. Thus, we just need to
work out partial derivatives of the Lagrangian with respect to material properties, that is:
@L 1 @K @K i:
¼ ðU VÞT ðU VÞ LTi V (26.8)
@Y0k 2 @Y0k @Y0k
The Lagrange multiplier is directly given from Eq. (26.5c). One just needs to calculate the derivative of the rigidity
@K
matrix with respect to material properties @Y 0 . Figure 26.2 shows an evaluation of the cost function G for one material
k
property, when fixing others, in the case of a numerical application presented Sect. 26.4 (here m is one of the two Lamé
coefficients). Red line segments represent the gradient at each evaluation point. One can remark that it represents the good
direction, which prove its good estimation.
26.4 Implementation of a Test Case
In this section, we illustrate the method through a numerical example, where data are numerically built. One only has to
solve the direct problem to build the admissible displacement fields, and then use an adequate minimization algorithm,
depending on the number of the sought properties. Random perturbations are mixed with the constructed displacement field,
in order to simulate experimental errors.
The test case being studied (Fig. 26.3) is an elastic square plate, built-in on one side and subjected to uniformly
distributed load on the opposite side. The plate is then meshed by Q4-elements and considered elastically homogeneous.
Measurements are artificially built with given reference values of Lamé coefficients lref and mref . Rather than l, we will
identify m since it is more sensitive to the displacement field here. In order to evaluate the robustness of the method, the
dependence of the cost function on the perturbation magnitude will be studied. Figure 26.4a, b show the trend of the cost
function and its two terms J 1 and J 2 when evaluated on an interval of m values, without perturbations for the first, and with
a perturbation level equal to 5% for the second. du is a random white Gaussian perturbation. Its level is computed as follow:
P2n
~
j¼1 U j
dU ¼ R 0:05 (26.9)
2n
where U ~ j denotes the jth degree of freedom of U,

~ n the number of measurement points and R an array of random numbers
whose elements are normally distributed.
The optimal parameter mopt is identified thanks to a binary search algorithm. Figure 26.4c shows the distribution of the
identification results for perturbation levels varying from 0 to 10%. Here, Monte Carlo method is used and 700 random
samples of perturbation are generated for each level. Figure 26.4d shows the standard deviation of identification smopt as
function of perturbation levels. It reaches 27% for a perturbation level equal to 5%. Figure 26.5 shows a comparison of
identification results obtained by FEMU [4] and MCRE methods. The dashed red curve in Fig. 26.5 shows results of another
MCRE formulation which takes into account a supplementary information concerning the knowledge of the free borders on
boundary conditions, whereas FEMU method is used with displacement information on the entire boundary. Let us underline
that FEMU method always needs the knowledge of the boundary conditions, in the form of either loading or displacement,
contrarily to MCRE approach. Figure 26.4d shows the standard deviation of the presented formulation. For example, for a
3% perturbation level, the result’s recap is given in Table 26.1.
26.5 Conclusion
In this paper, we considered the problem of the identification of material parameters in the framework of static. An
identification strategy based on the use of the modified constitutive relation error was proposed. It has the particularity of
taking into account the whole theoretical and experimental data to build admissible fields, then, the identification step starts
by the minimization of a cost function with regards to those optimal fields. The presented approach is formulated in a
Fig. 26.3 Example being

studied
a x 10−5 x 10−7 b −6
3 1 0.12 2 x 10 39.54
0.0101 39.52
2.5 0.1
0.8 0.0101
39.5
0.0101
2 0.08 1.5 39.48
0.6 0.0101
39.46
J2
J1
J1
J2
1.5 0.06 0.0101
G
G
39.44
0.4 0.0101
1 0.04 1 39.42
0.0101
0.2 0.0101 39.4
0.5 0.02
0.0101 39.38
0 0 0 0.0101 0.5 39.36

0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2
μ/μref μ/μref μ/μref μ/μref μ/μref μ/μref
c d
400 0.35
350 0.3
0%
1%
300
2% 0.25
250 3%
4% 0.2
σμopt
200 5%
6%
0.15
150 7%
8%
9%
0.1
100
10%
50 0.05
0 0
0.5 1 1.5 2 0 1 2 3 4 5
δu
Fig. 26.4 Cost function and identification results (MCRE method). (a) Shape of the cost function without perturbations. (b) Perturbation level
¼ 5% (c) Distribution of identified modulus. (d) Standard deviation for different noise levels
0.4
0.35 MCRE
0.3 FEMU
0.25
σμopt
0.2
0.15
0.1
0.05
0
0 1 2 3 4 5
δu
Fig. 26.5 Standard deviation of identification versus perturbation levels (MCRE and FEMU methods)
Table 26.1 Comparison between FEMU and MCRE methods

dU sFEMU (with BC) sMCRE (without BC) sMCRE (free borders information)
3% 19% 7% 6.5%
displacement way and has the advantage of being robust with respect to measurement uncertainties. When inverting the
linear system in the first step, a QR decomposition was proposed to avoid kernel problems occurring when the size of
measurement zone is smaller than the size of the whole domain zone. The method was applied to heterogeneous materials, in
the case of orthotropic elasticity. Identification of orthotropic parameters is being developed thanks to a coupling with a
minimization procedure using gradient information. Besides, an optimal control problem was studied, taking into account
the only distance-to-measurements term of the MCRE functional. Its results will be discussed during the oral presentation.
The work prospect is to develop the MCRE strategy by taking into account loading information such as the resulting load, in
the unreliable set of equations, by adding a distance-to-load term into the MCRE functional. The next step will focus on the
application of this method to real full-field measurements.
References
1. Chu TC, Ranson WF, Sutton MA, Peters WH (1985) Applications of digital-image-correlation techniques to experimental mechanics. Exp
Mech 25:232–244
2. Hild F, Roux S (2006) Digital image correlation: from displacement measurement to identification of elastic properties – a review. Strain
42:69–80
3. Kobayashi A (1993) Handbook on experimental mechanics, Wiley, New York
4. Avril S, Bonnet M, Bretelle A-S, Grédiac M, Hild F, Ienny P, Latourte F, Lemosse D, Pagano S, Pagnacco E, Pierron F (2008) Identification
from measurements of mechanical fields. Exp Mech 48:381–402
5. Feissel P, Allix O (2007) Modified constitutive relation error identification strategy for transient dynamics with corrupted data: The elastic
case. Comput Meth Appl Mech Eng 196:1968–1983
6. Ladevèze P, Reynier M, Maya N (1994) Error on the constitutive relation in dynamics. In: Bui HD, Tanaka M (ed) Inverse problems in
engineering. pp 251–256
7. Calloch S, Dureisseix D, Hild F (2002) Model validation in the presence of uncertain experimental data. Eng Comput 21(8):808–833
8. Deraemaeker A, Ladevèze P, Romeuf T (2002) Identification de modèles de comportement de matériaux solides: utilisation d’essais et de
calculs. Technologies et Formations 100:36–41
9. Latourte F (2007) Identification Des Paramètres d’une loi élastoplastique de Prager et calcul de champs de contrainte dans des matériaux
hétérogènes. Ph.D. thesis, Université de Montpellier 2
10. Florentin E, Lubineau G (2010) Identification of the parameters of an elastic material model using the constitutive equation gap method.
Comput Mech 46:521–531
11. Claire D, Hild F, Roux S (2002) Identification of damage fields using kinematic measurements. C R Mecanique 330:729–734
12. Ben Azzouna M, Périé J-N, Guimard J-M, Hild F, Roux S (2011) On the identification and validation of an anisotropic damage model by using
full-field measurements. Int J Damage Mech 20(8):1130–1150
13. Ben Azzouna M, Feissel P, Villon P (2011) Identification à partir de mesures de champs – application de l’erreur en relation de comportement
modifiée. In: 10th Colloque national en Calcul des structures – giens
Chapter 27
Identification of the Anisotropic Plastic Behaviour
of Sheet Metals at Large Strains
Marco Rossi, Fabrice Pierron, Michaela Štamborská, and František Šimčák
Abstract The identification of the material properties in sheet metal is usually achieved using uniaxial tests performed
along different texture directions. In this paper an experimental procedure to identify the plastic behaviour of sheet metals up
to large strains using full field measurement is presented. The tests were conducted on notched specimens. That geometry
generates a heterogeneous strain field which has been measured during the test using a digital image correlation system.
The advantage of using an heterogeneous strain field in the identification procedure is that a complex state of stress–strain
can be analyzed at the same time and much more information can be obtained in a single test. On the other hand the stress
field cannot be directly computed from the test and a suitable identification procedure has to be developed. Here, the virtual
fields method (VFM) adapted for large strains and plasticity has been used to identify the hardening behaviour and the
anisotropy of the two materials. The values obtained with the VFM have been compared with the results coming from a
standard identification made with uniaxial tensile tests.
Keywords Anisotropic plasticity • Inverse identification • Virtual Fields Method
27.1 Introduction
In sheet metal forming an anisotropic behaviour of sheet metals is often observed when they undergo plastic deformation.
Such anisotropy is due to the rolling process which introduces an anisotropy in the texture of the material. The correct
identification of the plastic behaviour of anisotropic sheet metal is a very important aspect that has to be taken into account in
order to enhance the results of the numerical models. Traditionally the identification is carried on using uniaxial tensile tests
performed at different directions respect to the rolling one. However in recent years, thanks to the development in the full-
fields measurement techniques, more complex specimens which generate heterogeneous stress–strain fields can be used to
study the behaviour of the material during the plastic flow. The parameters can be then identified using for instance an
inverse approach which includes FE updating [1–3]. Another identification procedure is the Virtual Fields Method [4] which
has been successfully applied in many applications, including elasto-plasticity [5, 6] and large strain plasticity [7].
The use of heterogeneous tests allows to achieve a lot of information from a single test since different stress–strain
conditions are evaluated at the same time.
M. Rossi (*)
Arts et Métiers ParisTech, rue St. Dominique, 51000 Châlons-en-Champagne, France
Università Politecnica delle Marche, via brecce bianche, 60100 Ancona, Italy
e-mail: marco.rossi@ensam.eu, m.rossi@univpm.it
F. Pierron
e-mail: fabrice.pierron@ensam.eu
M. Štamborská • F. Šimčák
Department of Applied Mechanics and Mechatronics, Technical Univerzity of Košice, Letná 9, 042 00 Košice, Slovakia
e-mail: michaela.stamborska@tuke.sk; frantisek.simcak@tuke.sk
230 M. Rossi et al.
In this paper a series of tests was conducted on a notched specimen cut in two different directions. The displacement and
strain field was measured using a stereo-correlation technique and the results have been used to identify the parameter with
VFM. The Hill48 yielding criterion and the Swift hardening law were used as plasticity model and six independent
parameters have been identified at the same time using a minimization algorithm.
27.2 Experiments
The heterogeneous strain field was obtained using a notched specimen, the complete geometry is illustrated in Fig. 27.1.
The specimens were cut from a blank sheet of stainless steel, two different cutting directions have been taken into account,
0 and 90 respectively. The properties measured using standard tensile tests in different direction are listed in Table 27.1.
The value of the Lankford parameter [8], i.e. the ratio of the transverse to the through-thickness strain increment, is measured
at different stages of the test. A random speckle pattern was sprayed using black and white paint in the central part of the
specimen, thus the displacement field was measured using Digital Image Stereo-correlation (DISC) [9, 10]. The VIC-3D
commercial software was adopted. The stereo-correlation allows to measure the out-of-plane movements due to possible
misalignment of the camera respect to the specimen surface or to instabilities of the sheet metal during the process.
The experimental setup is illustrated in Fig. 27.2, an INSTRON 880 tensile machine with hydraulic grips was used to
perform the tests. A couple of JAI Pulnix TM – 4000 CL cameras, able to acquire 2,048 2,048 pixels with 10-bit dynamic
range, were used to record the images for stereo-correlation. The specimens were tested along the longitudinal direction up
to the final break however, because of the occurrence of buckling at large deformations, only the first part of the test was used
in this study, when the displacement can be considered as planar.
In Fig. 27.3 is illustrated the out-of-plane movement measured with DISC at the last step used in the following
computation. The red line shows the zone of the specimen used to perform the identification. In this zone the out-of-plane
movement is moderate and the test can be still considered as planar. However a large deflection is visible at the tip of the
notches showing the onset of plastic instability.
The correlation algorithm was run using a correlation windows of 27 pixels and a step of 5 pixels between two
measurement points. The incremental correlation options was used, i.e. each image is correlated with the previous one
and the measured incremental displacement is added to the one measured in the previous step. A typical displacement map is
shown in Fig. 27.4 (last step of the test).
With the adopted settings 101 216 measurement points have been obtained. The strain maps at each step were then
computed using a polynomial fitting of degree 2 using five measurement points. Because of the large strain, the deformation
was computed as logarithmic strain tensor. An example of the obtained strain maps is illustrated in Fig. 27.5. It is clearly
visible the strain localization due to the notches. The test is rather heterogeneous and looks suitable to be used with VFM in
order to identify the constitutive parameters of the anisotropic model.
Fig. 27.1 Geometry of the specimen
Table 27.1 Mechanical properties of material measured by standard uniaxial tensile tests
Thick. Rp0,2 Rm A80 Ag R 3% R 5% R 10% R 15% R 20%
Dir. mm MPa MPa % %
0 0.75 124 297 44.5 24.8 1.7 1.7 1.7 1.7 1.7
45 0.75 139 305 42 23.5 1.5 1.5 1.5 1.6 1.6
90 0.75 132 290 43.5 24.7 2.2 2.1 2.2 2.3 2.3
27 Identification of the Anisotropic Plastic Behaviour of Sheet Metals at Large Strains 231
Fig. 27.2 Experimental setup
-2
-4
-6
-8
Fig. 27.3 Out-of-plane displacement of the specimen at the final analysed step
Displacement UX Displacement UY
0.5
-3
-20 -20
0 -4
mm
mm
0 0
-5
-0.5
20 20
-6
-1
-20 0 20 -20 0 20
mm mm
Fig. 27.4 Measured displacement fields for the last step

232 M. Rossi et al.
eX eY g XY
0.14
-40 -40 -40
-0.02 0.1
-30 -30 -30
0.12
-20 -20 -20
-0.04 0.05
-10 -10 -10
0.1
mm
mm
mm
0 0 0 0
-0.06
10 10 0.08 10
20 20 20 -0.05
-0.08
30 30 0.06 30
-0.1
40 -0.1 40 40
-20 0 20 -20 0 20 -20 0 20
mm mm mm
Fig. 27.5 Strain maps obtained after the polynomial differentiation
27.3 Identification Using the VFM
The chosen plastic model is the Hill48 yield criterion [11] with isotropic hardening. In the plane stress case the Hill48
criterion can be written as:
seq ¼ ðg þ hÞsx 2 2hsx sy þ ðf þ hÞsy 2 þ 2ntxy 2 (27.1)
where f, g, h and n are parameters which can be computed from the Lankford parameters at different orientations. If we
assume that the equivalent plastic strain seq is equivalent to the yield stress in the rolling direction, i.e. 0 , it can be shown
that [12]:

1 1 R0 1 R0
g¼ ; f ¼ ; h¼ ; n¼ þ R45 1þ g (27.2)
1 þ R0 1 þ R90 1 þ R0 2 R90
Where R0, R90 and R45 are the values of the Lankford parameters measured at orientation 0 , 90 and 45 respectively. The
hardening law is described using a Swift law:
seq ¼ Kðp þ e0 ÞN (27.3)
where seq is the equivalent stress of (27.1), p is the equivalent accumulated plastic strain and K, e0 and N are the parameters
to be identified. Therefore the total number of parameters which have to be identified are six, let us denote with x ¼ {K,e0,N,
R0,R90,R45} the vector of constitutive parameters that has to be identified at the end of the procedure.
The VFM is applied to plasticity at large strains. The procedure is fully described in Rossi et al. [7] where the theoretical
approach was presented for a general three-dimensional state of stress. In the present case the same procedure is applied to
plane stress condition valid for sheet metals. Summarizing, the constitutive parameters are identified by the minimization of
a cost function defined according to the principle of virtual work written for large deformations:
N step Z
N vf X
X Z

CðxÞ ¼
dvi dS
Tj : dFi dv
1PK
T1PK
j n
i¼1 j¼1 V @V (27.4)
T1PK is the 1st Piola-Kirchhoff stress tensor, dv is a virtual displacement field defined by the user, dF• is the
corresponding virtual displacement gradient tensor, V is the volume of the inspected solid, ∂V is the boundary surface
and n the surface normal in the undeformed configuration. The cost function is computed for each measurement step of
the test and for each defined virtual field. Nstep and Nvf are the number of measurement steps and adopted virtual fields
sX sY tXY
150 350
-40 -40 -40 150
-30 -30 300 -30
100
-20 -20 250 -20
100 50
-10 -10 -10
200
mm
mm
mm
0 0 0 0
150
10 10 10
50 -50
20 20 100 20
-100
30 30 50 30
40 40 40 -150
0 0
-20 0 20 -20 0 20 -20 0 20
mm mm mm
Fig. 27.6 Example of stress maps obtained from the strain field using the procedure described in [7]
respectively. Equation 27.3 is an equilibrium equation, the first integral can be viewed as the virtual work of the internal
forces and the second one as the virtual work of the external forces.
The Piola-Kirchhoff stress tensor is obtained from the Cauchy stress tensor T as:
T1PK ¼ detðFÞTFT (27.5)
An example of stress field obtained from this process is illustrated in Fig. 27.6. The noise is due to the fact that the stress
field, because of the plasticity law, is computed starting from the increments of strain at each test [7]. The strain maps
illustrated in Fig. 27.5 are instead related to the total strain measured at the single increment. The stress field is computed
using the constitutive model therefore it depends on the parameters of vector j.
According to the VFM, the parameter vector j is iteratively varied by a minimization algorithm in order to find the best
set of parameters which minimize the cost function of (27.3), or in other words, to find the balance between the virtual work
of the internal and external forces. In this application two sets of virtual fields were used:
( 8
dvx ¼ 0 < dv ¼ x ðj yj LÞ
x
dv1 ¼ dv2 ¼ W L (27.6)
dvy ¼ y=L :
dvy ¼ 0
L is the semi-length of the inspected area and W is the semi-width of the middle section. The coordinate system is chosen
such as the origin is placed at the centre of the specimen, in this way the virtual work of the external forces becomes:
Z
T1PK
j n dv1 dS ¼ 2f j (27.7)
@V
Z
T1PK
j n dv2 dS ¼ 0 (27.8)
@V
where fj is the force measured by the load cell at the time increment j.
The minimization of the cost function of (27.4) was performed using the SQP algorithm implemented in Matlab. The results
are illustrated in Fig. 27.7 and Table 27.2. First the cost function was minimized using only the results of one test in one
direction, then the two tests were introduced in the cost function trying to minimize the global error of the two tests.
234 M. Rossi et al.
Direction 0° Direction 90°

12000 12000
10000 10000
8000 8000
Force
Virtual Work
Virtual Work
VF1 only 0°
6000 6000 VF1 only 90°
VF1 0° + 90°
4000 4000 VF2 only 0°
VF2 only 90°
VF2 0° + 90°
2000 2000
0 0
−2000 −2000
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Test step Test step
Fig. 27.7 Comparison between internal and external virtual work computed with the identified parameters
Table 27.2 Identified parameters

Identified parameters Yielding stress
K e0 N R0 R45 R90 0 90
Test used MPa MPa Mpa
Only 0 981 0.019 0.52 1.36 1.67 0.4 127 90
Only 90 1,032 0.048 0.76 0.65 1.50 1.46 103 126
0 +90 1,061 0.021 0.58 1.56 1.25 1.45 114 112
Reference – – – 1.7 1.5 2.1 124 132
In Fig. 27.6 the virtual work of the internal forces due to the first virtual field (VF1) is compared with the measured force
(see (27.7)) and the virtual work of the internal force due to the second virtual field (VF2) is compared with zero (see (27.8)).
It is clear that one test is not sufficient to correctly identify the material properties since the parameters identified using,
for instance, the specimen cut along the 0 direction give incorrect results in terms of global equilibrium when used with the
strain data relative to 90 , and vice-versa.
Looking at the identified parameter listed in Table 27.2 the same trend is observed. However, even when the two tests are
combined, a difference is noted between the identified parameters and the ones coming from a standard identification.
The reason for this discrepancy has to be better investigated but could be related to theHill48 model, which may not be
appropriate enough to describe the plastic behaviour of the studied material.
27.5 Conclusions
This paper presents an example of identification of an anisotropic model using full-field measurements and the Virtual Fields
Method. The identification is performed using two virtual fields and two different orientations of the specimen. It comes out
that at least two directions are necessary to obtain suitable results and a discrepancy is still observed respect to the values
obtained with the classical identification. In future studies more specimen directions, as 45 , should be taken into account,
moreover a validation of the proposed procedure using simulated data is also required.
References
1. Cooreman S, Lecompte D, Sok H, Vantomme J, Debruyne C (2007) Elasto-plastic material parameter identification by inverse methods:
calculation of the sensitivity matrix. Int J Solids Struct 44:4329–4341
2. Tao H, Zhang N, Tong W (2009) An iterative procedure for determining effective stress–strain curves of sheet metals. Int J Mech Mater Des
5:13–27
3. Kajberg J, Lindkvist G (2004) Characterization of materials subjected to large strains by inverse modeling based on in-plane displacement
fields. Int J Solids Struct 41:3439–3459
4. Grédiac M, Pierron F, Avril S, Toussaint E (2006) The virtual fields method for extracting constitutive parameters from full-field measurement:
a review. Strain 42:233–253
5. Grédiac M, Pierron F (2006) Applying the virtual fields method to the identification of elasto-plastic constitutive parameters. Int J Plast
22:602–627
6. Pierron F, Avril S, Tran V (2010) Extension of the virtual fields method to elasto-plastic material identification with cyclic loads and kinematic
hardening. Int J Solids Struct 47:2993–3010
7. Rossi M, Pierron F (2012) Identification of plastic constitutive parameters at large deformations from three dimensional displacement fields.
Comput Mech 49:53–71
8. Lankford WT, Snyder SC, Bausher JA (1950) New criteria for predicting the press performance of deep drawing sheets. Trans Am Soc Met
42:1197–1232
9. Sutton MA, McNeill SR, Helm JD, Chao YJ (2000) Advances in two dimensional and three-dimensional computer vision. In: Rastogi PK (ed)
Photomechanics, topics applied physics, vol 77. Springer, Berlin, pp 323–372
10. Sutton MA, Orteu JJ, Schreier HW (2009) Image correlation for shape, motion and deformation measurements. Springer, New York
11. Hill R (1948) A theory of yielding and plastic flow of anisotropic metals. Proc R Soc Lond Ser A: Math Phys Sci 193:281–297
12. Rossi M, Broggiato GB, Papalini S (2008) Application of digital image correlation to the study of planar anisotropy of sheet metals at large
strains. Meccanica 43:185–199
Chapter 28
Cohesive Zone Law Extraction from an Experimental Peel Test
Christopher Kovalchick, Shuman Xia, and Guruswami Ravichandran
Abstract The measurement of interface mechanical properties between an adhesive layer and a substrate is significant for
optimization of a high-quality interface. A common method for measuring these properties is the peel test. Although
analytical models exist for the peeling of elastic adhesives from smooth surfaces, there is a need for rigorous experiments
in this area. Tests are conducted using a peel arrangement capable of peel angles from 0 to 180 . Experimental validation
is achieved through a series of displacement-controlled tests using elastic adhesive tapes. The steady-state peel force is
measured at each angle and compared to the predicted values based upon the governing relations of Kendall and Rivlin.
The fibrillation zone of a simple adhesive material is imaged at the contact point between the adhesive and a rigid surface.
By quantifying the size of this process zone, an inverse formulation is developed to extract an angle-dependent cohesive
zone law based upon finite deformation beam theory. The adhesion energy during a peel test is then determined along with
the force distribution in the process zone. This local method of determining the adhesion energy is compared to the global
method of Rivlin in the context of finite deformations, showing good agreement.
Keywords Cohesive zone • Peeling • Adhesion
28.1 Introduction
The measurement of interface mechanical properties between an adhesive layer and a substrate is significant for optimization
of a high-quality interface [1]. A common method for measuring these properties is the peel test, a simple technique
that can be traced back to the results of investigations on the surface energy that exists between two elastic solids in
contact [2]. There are several applications of peeling in the field of biomechanics, particularly in the area of gecko and
cell adhesion [3–6]. In the microelectronics community, the focus of peeling has been on the delamination strength of
thin films to silicon and other substrates [7]. Additionally, environmental conditions such as moisture and thermal
gradients can degrade the strengths of devices due to strain mismatches between the film and the substrate. From an
industrial standpoint, the peel test has been used primarily for comparative strength testing of materials for case-specific
applications. These include quality control devices in which the ultimate peel strength of a material is a critical
parameter.
Consider the peeling of an elastic thin film (tape) from a rigid substrate whose parameters are the film thickness, h, film
width, b, the angle of peel between the film and the substrate, y0, and measured peel force, Pf. Rivlin [8] was the first to
analyze the peel test and did so using a global energy balance in the context of large deformations. He obtained the relation
for the peel force per unit width of the tape in terms of the adhesion energy (G) of peeling,
Pf G
¼ : (28.1)
b 1 cos y0
C. Kovalchick (*) • S. Xia • G. Ravichandran

Graduate Aerospace Laboratories, California Institute of Technology, 1200 E. California Blvd, Pasadena, CA 91125, USA
e-mail: kovalchick@gmail.com
238 C. Kovalchick et al.
Kendall [9] extended Rivlin’s model to include the extensibility of the tape and based on a global energy balance, he
obtained the following expression for the adhesion energy,
P2f Pf
G¼ þ ð1 cos y0 Þ: (28.2)
2Eb2 h b
Note that Kendall’s Equation (28.2) includes the elastic energy of the tape, whereas Rivlin does not consider this term. An
inextensible tape is said to be infinitely stiff. In this case, the elastic modulus is very large and the first term in (28.2)
vanishes. This can also be observed by comparing the quantity of stress Pf/bh to the elastic (Young’s) modulus, E. For a
nearly inextensible material, E >> Pf/bh and the elastic energy term can be ignored. For soft materials, this may not be the
case and the extensibility of the film must be considered, especially at small peel angles (yo close to 0 ) where stretching is
dominant.
The elegance of the energy-balance approach taken by Rivlin and Kendall is in its simplicity. The adhesion energy of the
peel test can be estimated with knowledge of the peel angle y0 and the far-field peel force, Pf. However, these models do not
provide any insight about the forces and displacements in the peel zone, which are governed by the mechanical properties of
the adhesive. In order to obtain such detail, another method of describing this local region during a peel test is needed. One
possible method is to treat this local peel region as a cohesive zone [10, 11] and use a cohesive zone law to obtain information
in this local process zone region. This relates to the idea of approaching the peeling problem from both a macro (global) scale
by using the Rivlin-Kendall models, while considering the micro (local) scale that is governed by the cohesive zone.
This paper examines the mechanics of the peel test through experiments and theoretical analysis with the goal of
extracting a cohesive zone law from the peel test. Displacement-controlled tests are conducted using an experimental peel
test configuration for a range of peel angles using elastic adhesive tapes adhered to a rigid glass substrate, and the experi-
mental results are compared to the classical relations of Rivlin and Kendall ((28.1) and (28.2)). By making the analogy that
the fibrillation zone that develops in these tests is a cohesive zone in a fracture-based study, the displacements in the cohesive
zone are measured experimentally by characterizing its shape using an optical imaging technique. A new theoretical
framework is presented in the context of finite deformation beam theory to extract a cohesive zone law from an experimental
peel test. Through extracting a cohesive zone law, the adhesion energy during a peel test is determined along with the force
distribution in the process zone. This local method of determining the adhesion energy is compared to the global method
used in the Rivlin model.
28.2 Experimental Methods
An experimental configuration is developed for conducting a displacement-controlled peel test of an adhesive tape from a
rigid substrate. Tests are conducted using Scotch Magic TapeTM (3M, Minneapolis, MN), a well-known adhesive system for
comparison to known results in the literature for validation purposes. The backing material of this tape is generally observed
to be elastic, with the modulus obtained independently through a uniaxial tension test. The width and the thickness of the
tape backing are 19 mm and 50 mm, respectively.
The experimental setup is shown in Fig. 28.1. The delrin base can be adjusted to obtain a range of angles from 0 to 180
(0 –90 in the configuration shown here; to obtain 90 –180 , the base is turned around). The peel force, Pf is measured at

the end of the unadhered portion of the tape with a load cell (ALD-MINI-UTC-M, AL Design, Inc., Buffalo, NY). The load
cell is attached to the end of an aluminum peel arm, which is attached to a motorized translation stage (M-410.CG, Physik
Instrumente, Irvine, CA). The stage has a total travel range of 100 mm, a maximum velocity of 1 mm/s, and resolution of
100 nm. The stage is vertically mounted to an optical rail and used to apply a crosshead displacement, which peels the tape
at a constant rate from a glass substrate. The glass substrate used is a generic piece of industrial grade lime glass cut to
size with dimensions 203 127 3.175 mm (800 500 1/800 ) for the length, width, and thickness, respectively.
The glass substrate is held in a notched path on the inside perimeter of the peel base. On the top right hand side of the
stage, two screws are used as stops to keep the glass from moving during a test. Data is acquired for specific sampling rates
and test times from the load cell through a virtual instrument created in Labview (DAQM02, National Instruments,
Austin, TX).
To investigate the fibrillation of the adhesive in the process zone, the geometry of the process zone is measured by
acquiring in-situ images of the fibrils in the deformed configuration using a 2-megapixel monochromatic digital CCD
camera (UP-2000CL B/W, Uniqvision, Santa Clara, CA) attached to a long-distance video lens (K2/S Infinity, Boulder, CO).
28 Cohesive Zone Law Extraction from an Experimental Peel Test 239
Fig. 28.1 Experimental peeling configuration
A sequence of images is acquired every 0.5 s for the duration of a peel test (typically 10 min) at a constant angle. Separate
tests are conducted for initial peel angles (y0) ranging from 30 to 90 and the tests are repeated several (typically 5) times for
a given peel angle. The peel zone mouth opening displacement (d) and length (L) of the zone are measured for each image
using a simple post-processing scheme implemented in Matlab (MathWorks, Natick, MA).
28.3 Local Process Zone Model
The process zone in a peel test is defined as the region from the last detached fibril to the point where the fibrillation starts
(peel front), which is the contact point between the adhesive tape and the glass substrate. The parameters of interest are the
length and opening displacement (height) of the process zone, L and d, respectively, as well as the shape of the backing
material in the process zone. The radius of curvature of the backing can be described analytically. A schematic of the peel
test and process zone is shown in Fig. 28.2 in the deformed configuration.
A closed-form theoretical framework is developed for analyzing the peel test based on the planar beam. This analysis is
used to determine the full-field force distribution in the process zone as well as a cohesive zone law to describe the process
zone. Consider the peeling of an elastic adhesive film from a rigid substrate. The film has an elastic modulus, E with a
rectangular cross-section of width, b and thickness, h. The adhesive backing is treated as a 2D planar beam, while the
deforming adhesive in the process zone itself is considered as a cohesive zone. As the adhesive backing is peeled from
the surface at a fixed peeling angle, y0, it is similar to a bimaterial interfacial crack problem [12]. As the peeling proceeds in
the steady state, the length, L and height, d define the process zone (Fig. 28.2). A base curve s is defined along the thin film
(backing of the adhesive), with y1 and y2 being the coordinate frame in the deformed configuration. P(s) is the axial force along
the tape backing in the beam and far from the process zone (s ! 1), the axial force P(s) is the measured peel force, Pf. Force
equilibrium between two neighboring points on the thin film requires, in component form,
dN
þ kT þ ~
f ~
n ¼ 0;
ds
dT
kN þ ~
f ~
t ¼ 0;
ds
dM
þ N þ m ¼ 0; (28.3a, b, c)
ds
where T(s) and N(s) are the internal axial and transverse forces along s, respectively; ~
t and ~
n are unit vectors in the tangential
and transverse directions, respectively, and ~ f is the cohesive zone force in the undeformed (reference) configuration, as
shown in Fig. 28.3.
Fig. 28.2 Schematic of a

peeling configuration with
the cohesive zone for an
elastic adhesive film from
a rigid substrate in the
steady state deformed
(current) configuration
Fig. 28.3 Force distribution

at a point along the deformed
thin film (adhesive backing)
The thin film is assumed to be inextensible, which is a fairly good assumption at medium to large peel angles considered
in this study (30 –90 ). The curvature of the beam, k, can be related to the shape of the tape backing measured
experimentally. Using the general equation of an ellipse to fit the tape backing,
B 2 1
y2 ðy1 Þ ¼ A y21 2 þ B (28.4)
A
where the constants A and B are determined experimentally for each peel angle, y0. Using this information, the force
distribution in the process zone, T and N, can be obtained by solving the system of equations (28.3a, b, c). A distributed
cohesive zone force, s0 is defined as acting normal to the substrate with respect to the undeformed configuration,

~ 0
f ¼ s0 : (28.5)
1
This requires making the assumption that the deformation of the fibrils in the process zone is normal to the substrate,
which is confirmed experimentally. Implementing the shape and cohesive force description along with the force distribution
solved for in (28.3a, b, c), an expression for the cohesive zone law is derived as a function of y1:

1 dy1 dN dy
s0 ¼ þ T : (28.6)
cos y ds dy1 dy1
This expression is solved by utilizing the boundary condition of the peel force, Pf measured experimentally outside the
process zone, where the cohesive force s0 ¼ 0.
28.4 Results
Peel tests are conducted for angles of 30 –90 at 10 increments. A total crosshead displacement of 6 mm is imposed at a
prescribed velocity of 0.01 mm/s, and the peel force (Pf) is measured over a time period of 10 min. The measured steady-
state peel forces are plotted as a function of the peel angle in Fig. 28.4 and compared to those predicted by the Rivlin and
Kendall models. In the Kendall model (28.2), the elastic modulus of the tape backing material is E ¼ 1.65 GPa, obtained
independently through a uniaxial tensile test. The values of adhesion energy used in both models are determined using
Eq. 28.1 from the experimentally measured 90 peel force, which in this case is G ¼ 34.5 N/m. This value is employed for
calculating the theoretical peel forces at the lower angles. Figure 28.4 shows that the peel forces correlated by the two
models begin to exhibit a difference at small peel angles, due to the important role of the elastic energy term associated with
stretching. However, for the range of peel angles (30 –90 ) investigated in this study, the difference is negligible, which
validates the inextensibility assumption for the present work. Compared to both model evaluations, the measured peel forces
show good agreement within experimental scatter.
Characterization of the process zone is aimed at examining two phenomena: the steady-state behavior of the process zone
geometry, and the dependence of the geometry on the peel angle. The process zone image for three representative tests at
yo ¼ 90 , 70 , and 30 , respectively, after the steady-state has been achieved are shown in Figs. 28.5, 28.6, and 28.7.
From these images, the fibrillation zone can be quantified clearly. The last attached fibril can be identified, which defines
the end of the cohesive zone. While the origin of the zone is more difficult to locate accurately at this resolution, it is
Fig. 28.4 Experimentally

measured steady-state peel
force (Pf) versus peel angle (y)
compared to the predictions
from the Rivlin and Kendall
models, Eqs. 28.1 and 28.2
Fig. 28.5 Image of the

process zone from a 90
peel test of Scotch Magic
Tape adhered to glass


identifiable enough to see a trend in the zone size as a function of peel angle. It is observed that to a good approximation, the
fibrils appear to remain perpendicular to the tape throughout the process zone.
Prior to quantitative analysis of the fibrillation, there are some qualitative trends that can be seen by comparing the images
of the process zones at the three different peel angles. The size of the process zone appears to change as the peel angle
changes. Specifically, the length of the zone (L) appears to increase with decreasing peel angle. To examine how the shape of
this process zone potentially changes during the steady-state, five separate images at equal time increments (100 s) during
the steady-state phase are quantified. The shape of the zone remains self similar during the steady-state period for a single
peel test and the shape of the zone will be treated as constant irrespective of time. The shape is taken to be the average of the
five randomly-selected instances in time during a single test. There is fairly good consistency between the five tests, though
some scatter is present in the specific values of the cohesive zone parameters, L and d. The average values for the entire range
of peel angles (30 –90 ) are given in Table 28.1 using data from the averaging of five separate peel tests at each angle.
As is seen for all of the angles in Table 28.1, the error for the process zone size is typically 20% at a given angle.
A definite increase in the length of the process zone L is observed as the peel angle decreases, while the height of the process
zone d seems to remain relatively constant and appears nearly independent of the peel angle.
Using the experimental parameters obtained for L and d, the shape of the tape backing is fit with the elliptical function in
Eq. 28.4 to determine the constants A and B. These parameters (L, d, A, B) and the steady-state peel force, Pf are input into the
local process zone model in order to extract a relation between the cohesive zone stress, s0 and the height of the process
zone, d using Eq. 28.6. Additionally, the internal transverse and axial forces in the tape backing, N and T, respectively, can be
determined. Figure 28.8 shows the cohesive zone stress as a function of the process zone height, d for a peel angle of 90 .
The cohesive zone force, s0 goes to zero at the end of the process zone, where y2 ¼ d. The general shape of this curve is very
similar to a power-law function. In addition to this case for y0 ¼ 90 , the transverse and axial forces N and T, respectively,
Table 28.1 Process zone length Peel angle, yo (degrees) L (mm) d (mm)
scale parameters for varying peel
angles 90 412 78 226 21
80 505 11 212 51
70 523 22 223 39
60 634 0.7 259 23
50 600 85 207 30
40 621 78 222 21
30 617 32 181 22
Fig. 28.8 Cohesive zone

stress-displacement relation
for a 90 peel test
along with the cohesive zone relation, s0 – d are also determined for the range of peel angles 30 –80 . To determine the
adhesion energy, G of this cohesive zone, the s-d curve can be integrated as
Zd
G¼ sðy2 Þdy2 : (28.7)
0
The adhesion energy can also be determined using the global energy balance approach of Rivlin previously described
(Eq. 28.1). Comparing the values of G obtained through these two different methods, the validity of the local method using
finite deformation beam theory is analyzed. Figure 28.9 shows the adhesion energy, G obtained by both the Rivlin model and
the model presented here as a function of the peel angle, y0.
For a peel angle of 40 , there is a large difference both in the experimental results as well as the correlation between the
two methods of determining G. This angle is the only case in which the cohesive zone model yields a higher value of
adhesion energy than the Rivlin model. For all other angles shown in Figure 28.8, the cohesive zone model predicts a lower
value of the adhesion energy, G. The difference between the two methods is that the Rivlin model determines the global
adhesion energy through the far-field peel force, Pf and peel angle, y0. The new method presented here determines slightly
smaller values of adhesion energy through the far-field peel force, Pf, and the size of the local process zone. The model here
also provides the local force distribution in the process zone, which the Rivlin model does not provide. The present analytical
model is able to quantify the forces in this region given that the size of the process zone can be quantified experimentally.
While the advantage of the model presented here is the ability to describe the behavior in the process zone, the
implementation of this model is more complicated than the Rivlin model. The simplicity of the Rivlin model is that
the adhesion energy can be determined as a function of two far-field variables (Pf and y0) that can easily be measured.
The cohesive zone model here is sensitive to the values of the empirical constants A and B in the curve fit. As previously
mentioned, there are any number of functions that could be used to fit the experimental profile. It is important to note that in the
cohesive zone extraction, an experimentally-obtained data curve is being differentiated four times as it is implemented into
the theoretical framework. While the curve fit itself is a functional form, new error may be introduced as the curve fit is
Fig. 28.9 Adhesion energy G 55

for the range of peel angels GCZ
50
30–90 GRivlin
45
Adhesion Energy G [N/m]

40
35
30
25
20
15
10
5
0
0 10 20 30 40 50 60 70 80 90 100
θ [degrees]
processed several times. As previously mentioned, a different boundary condition could be used in order to produce a force
distribution that is in better agreement with finite element simulations. This may also improve the correlation between the
adhesion energies.
It is also observed that the adhesion energy, G appears to be constant within some quantifiable error, independent of the
peel angle, y0, which is in agreement with previous studies [13]. In the context of the cohesive zone law extraction method
used here, since the adhesion energy is not a function of the peel angle y0, determining the cohesive zone law for a 90 peel
test may be sufficient to describe the cohesive zone law for an entire range of angles.
28.5 Conclusions
This paper has examined the mechanics of the peel test through experiments and theoretical analysis with the aim of
developing governing relations to describe the role of adhesion in the peel test for elastic adhesive tapes and in the fracture
process in general. An experimental peel test configuration has been developed and validated for the study of the peeling of
elastic adhesive tapes. A cohesive zone law is extracted using the experimentally-measured profile of the tape backing in the
framework of finite deformation beam theory for a range of peel angles, providing information about the mechanical
behavior of the adhesive in the process zone including the full-field force distribution and the adhesion energy. The adhesion
energy computed from the cohesive zone law shows good correlation with the value obtained using the Rivlin model, which
utilizes a global energy balance of the system. The new method of using the cohesive zone law to extract the adhesion energy
is unique in that it provides a rigorous method of determining the cohesive zone law based upon the measured shape of the
tape backing during an experiment.
References
1. Hutchinson JW, Suo Z (1992) Mixed mode cracking in layered materials. In: Hutchinson JW, Wu TY (eds) Advances in applied mechanics,
vol 29. Academic, Boston, pp 63–191
2. Kendall K (1971) The adhesion and surface energy of elastic solids. J Phys D: Appl Phys 4:1186–1195
3. Garrivier D, Decave E, Brechet Y, Bruckert F, Fourcade B (2002) Peeling model for cell detachment. Eur Phys J E 8:79–97
4. Autumn K, Sitti M, Peattie A, Hansen W, Sponberg S, Liang YA, Kenny T, Fearing R, Israelachvili J, Full RJ (2002) Evidence for van der
Waals adhesion in gecko setae. Proc Natl Acad Sci USA 99:12252–12256
5. Autumn K, Dittmore A, Santos D, Spenko M, Cutkosky M (2006) Frictional adhesion: a new angle on gecko attachment. J Exp Biol
209:3569–3579
6. Tian Y, Pesika N, Zeng H, Rosenberg K, Zhao B, McGuiggan P, Autumn K, Israelachvili J (2006) Adhesion and friction in gecko toe
attachment and detachment. Proc Natl Acad Sci USA 103(51):19320–19325
7. Harris JH, Rubel E (2008) The role of interfacial compound formation on package reliability. Adv Microelectron 35(4):20–27
8. Rivlin RS (1944) The effective work of adhesion. Paint Technol 9:215
9. Kendall K (1975) Thin-film peeling – the elastic term. J Phys D: Appl Phys 8:1449–1452
10. Dugdale DS (1960) Yielding of steel sheets containing slits. J Mech Phys Solids 8:100–104
11. Barenblatt GI (1962) The mathematical theory of equilibrium cracks in brittle materials. Adv Appl Mech 7:55–129
12. Rice JR, Sih GC (1965) Plane problems of cracks in dissimilar materials. J Appl Mech 32:418–423
13. Thouless MD, Jensen HM (1992) Elastic fracture mechanics of the peel-test geometry. J Adhes 38:185–197
Chapter 29
Application of a New Experimental Method to Determine
Bi-Material Interface Bonding Strength
M.A.K. Chowdhuri and Z. Xia
Abstract The accurate measurement of interface bonding strength is important in order to predict the failure or damage
behavior of structures and materials containing bi-material interfaces. Due to the discontinuity of material properties, stress
singularity may exist at the edges/corners of bi-material interfaces, which causes difficulty in accurate evaluation of interface
bonding strength. In many practical applications, interface failures occur under the actions of combined normal and shear
stresses. Therefore, a biaxial normal-shear bonding strength envelope is preferred. This paper presents a general experimen-
tal method for the determination of interface bonding strength envelopes for different combinations of material types. Finite
element analyses are first carried out to determine the appropriate interface design parameters to eliminate the stress
singularity. Then, tests are carried out with the designed specimens under pure tensile, pure torsional, and various combined
tensile-torsional loadings to obtain the interface failure loads of each loading case. The final interface bonding strength
envelope, which is expressed by a quadratic expression, is determined by an iterative calculation technique integrated with
the finite element analysis. By using the proposed method, the interface bonding strength envelope of a Polyvinylchloride
(PVC)/Epoxy (viscoelastic/viscoelastic) interface is determined.
Keywords Experimental mechanics • Bi-material interface • Interface bonding strength envelope • Finite element analysis
• Iterative algorithm
29.1 Introduction
Most of the polymeric materials possess a capacity to both store and dissipate mechanical energy. Those materials can be
cataloged as viscoelastic materials which are time and temperature dependent. Polymeric materials are increasingly used in
many industries, for examples, in automobile, aerospace, oil and gas, etc. They are often reinforced with other harder
materials such as fibers, ceramic particles, etc. to fabricate the composite materials in order to attain the benefits of both
materials. The interface strength between the polymer and reinforcing agent plays a critical role to the overall behaviors of
these polymeric composite materials since damages or failures are most likely to occur at the bi-material interfaces. So, the
determination of bi-material interface bonding strength, especially interface bonding strength determination of viscoelastic/
viscoelastic or elastic/viscoelastic bonded joints is extremely important. However, the accurate measurement of the interface
bonding strength is still a challenging task. Due to the discontinuity of material properties, there may exist stress singularity
at the interface corner (the point where the interface intersects a stress-free surface) of bi-material interface [1–3]. Stenger
et al. [4], Wu [5] and Reddy [6] mentioned that the stress singularity at the free edge of dissimilar material joint is the main
factor responsible for debonding failure under mechanical and/or thermal loading.
The current methods for the measurement of bi-materials interface strength, for examples, ASTM D3528 [7], ASTM
D2095 [8], single fiber pull-out, push-out tests [9, 10], droplet test [11], single-fiber fragmentation test [12] and peel-off test
[13] have the disadvantages of the presence of non-uniform combined shear-normal interface stress field along the
interface and/or the stress singularities at the interface corners. In order to avoid the stress singularity in bonding strength
M.A.K. Chowdhuri (*) • Z. Xia

Department of Mechanical Engineering, University of Alberta, Edmonton, AB T6G 2G8, Canada
e-mail: khanchow@ualberta.ca
248 M.A.K. Chowdhuri and Z. Xia
measurement, Xu [14] and Wang and Xu [15] suggested convex interfacial joints; Lauke [16], and Schneider et al. [17]
proposed a cylindrical interface between two materials in the flat coupon specimen. Recently, Xia et al. [18] and Lauke [19]
have designed a cylindrical specimen with spherical interface geometry for the determination of the interface bonding
strength between two elastic materials, which can eliminate the stress singularity. Lauke [19] used the maximum normal
stress at the interface pole in his calculation and mentioned that for the cases other than elastic materials, the maximum stress
doesn’t occur at the pole. Again, in many practical applications, the bi-material interface failures occur under the action of
combined normal and shear stress. Therefore, Xia et al. [18] argued that not only the pure tensile strength or the pure shear
strength, a general normal-shear bonding strength criterion (strength envelope) is required to characterize the bi-material
interface strength.
In this paper, a new experimental method is presented to determine the bi-material interface bonding strength by
eliminating the stress singularity from the interface corner. As an example, the interface bonding strength of a viscoelas-
tic/viscoelastic bonded joint is determined using the proposed method.
29.2 Methodology
In order to determine the bi-material interface bonding strength, a new method is developed. This method includes the
design of a specimen with special interface geometry to eliminate the stress singularity, obtaining interface failure loads of
the specimens under different combinations of normal and torsional loadings, and an iterative calculation technique
integrated with the finite element method (FEM) analysis to determine the interface biaxial normal-shear bonding strength
envelope. The complete methodology is described in the following sections through the determination of the interface
bonding strength for a viscoelastic/viscoelastic joint.
29.2.1 Design and Preparation of the Specimen
In order to eliminate the stress singularity, a cylindrical specimen of two bulk materials with a spherical interface is designed
and shown in Fig. 29.1. In this design, the hard material is placed at the concave side of the interface and the soft material is
placed at the convex side of the interface. The bonding angle is denoted by y, and defined as the angle between the tangent of
the interface at the free edge to the generator of the cylindrical surface.
From the geometry of the specimen, the bonding angle can be determined from the ratio of the specimen radius, r to the
interface radius, R.
r
cos y ¼ (29.1)
R
For any given bi-material interface, a critical value of the bonding angle, yc exists. So, the condition for avoiding the
stress singularity at the free edge is y<yc .
Thus, the first task of this method is to determine the critical bonding angle for a given bi-material interface. Let us
consider the interface between PVC and Epoxy. Both materials are considered as linear viscoelastic materials. The elastic
properties of PVC and epoxy are measured through the uniaxial tensile tests with very fast loading rate. The Young’s
modulus and Poisson’s ratio of PVC are obtained as 3.38 GPa and 0.362, respectively. For epoxy, these values are
determined as 1,271 MPa and 0.415, respectively. Next, the viscoelastic properties of the materials are determined through
the uniaxial stress relaxation tests of the materials by an 810 Material Testing System (810 MTS). For the relaxation test,
axial strain is applied and hold for about 1½ h. During this period, load, axial strain, time and transverse strain are recorded
in every second. From the measured data, the uniaxial relaxation modulus and Poisson’s ratio are calculated directly by using
Eqs. 29.2 and 29.3.
The relaxation modulus, G(t) is defined as,
GðtÞ ¼ sðtÞ=e0 (29.2)
where, s(t) is the time dependent stress corresponding to the unit step constant strain, e0.
29 Application of a New Experimental Method to Determine Bi-Material. . . 249
Fig. 29.1 Design of the specimen along with its finite element model
Fig. 29.2 Viscoelastic (a) shear modulus and (b) Poisson’s ratio of materials
The time-dependent Poisson’s ratio, n(t) is defined as the ratio of time-dependent transverse/lateral strain, e2(t), to the
axial constant step strain e0
e2 ðtÞ
uðtÞ ¼ (29.3)
e0
The other material properties for example, the viscoelastic shear modulus can be obtained from the above relaxation
modulus and Poisson’s ratio using Eq. 29.4.
GðtÞ
mðtÞ ¼ (29.4)
2ð1 þ nðtÞÞ
Figure 29.2 shows the viscoelastic shear modulus and Poisson’s ratio for the two materials. The properties equal to their
elastic properties are considered as the viscoelastic properties of the materials at t ¼ 0.
A Prony series with three terms (N ¼ 3) is used to analytically represent the measured properties of the materials shown
in Fig. 29.2. The equations of the representing curves are given by Eqs. 29.5 and 29.6.
Fig. 29.3 von-Mises stress

distribution along the interface
for different bonding angles
For PVC : mðtÞ ¼ 830 þ 305et=12 þ 107et=695 ; nðtÞ ¼ 0:395 0:012et=436 0:021et=76 (29.5)
For Epoxy : mðtÞ ¼ 135 þ 237et=21:25 þ 77et=650 ; nðtÞ ¼ 0:443 0:017et=1055 0:011et=50 (29.6)
The fitting curves represented by Eqs. 29.5 and 29.6 are also shown in Fig. 29.2 along with the experimental curves.
Finite element numerical analysis (using ANSYS) is used to determine the critical bonding angle for the interface. The
axi-symmetric finite element model of the specimen is shown in Fig. 29.1. Eight-nodes PLANE 183 element is used for both
materials. The radius of the cylindrical specimen, r, is 10 mm and a maximum axial tensile stress of 10 MPa is applied at top
end of the specimen keeping the bottom end fixed in the longitudinal direction in 30 equal load steps and the time at the end
of load step is 900 s. After solving the problem, the von-Mises stresses are determined from the stress components along the
interface. If for any bonding angle the stress at the interface corner increases with the refinement of mesh size and doesn’t
converge to a finite value, this indicates the existence of the stress singularity for that bonding angle. On the other hand, for a
non-singular stress case the stress at the interface corner converges to a finite value. Figure 29.3 shows the von-Mises stress
distribution for different bonding angles.
Figure 29.3 demonstrates that the excessive stresses exist near the free edge of the interface for the bonding angle,
y ¼ 60 and the stress value increases without convergent limits with increasing mesh density, indicating the presence of the
stress singularity. Whereas, for the case of bonding angle, y ¼ 40 , converged finite stress value is obtained at the interface
corner. The same analyses are carried out for bonding angles from 40 to 60 by the increment of 1 each time. It is found
that the shape of the von-Mises curve near the free edge starts to change from the upward to the downward direction, when
the bonding angle changes from 48 to 47 . This indicates that the value of critical bonding angle for the tested PVC/Epoxy
interface is in between 47 and 48 .
After determining the critical bonding angle, the test specimens are prepared with the bonding angle of 40 to ensure the
elimination of the stress singularity (y < yc). The specimen is shown in Fig. 29.4. The PVC portion of the specimen is first
fabricated by CNC machining and then epoxy is poured on it in a special designed mold. EPON 8131 resin and EPI-CURE
3072 hardener are mixed at a proportion of 79–21 wt% to make the epoxy portion of the specimen.
29.2.2 Experimental Set up and Conducting Tests
A multi-axial testing machine shown in Fig. 29.5 is used to conduct the tests. This machine has the capacity to apply both
torsional and tensile loading simultaneously (combined loading). An ER-32 grip system is used to hold the specimen in the
machine. Though, this grip is capable of maintaining the perfect alignment of the specimen in the testing machine, it cannot
Fig. 29.4 Specimen for the interface bonding strength determination
Fig. 29.5 Multi-axial testing machine and grips to hold the specimen
hold the viscoelastic materials for a longtime. In order to overcome this problem, tapered aluminum end caps are glued to
two ends of the specimen by using DP 460 adhesive (Fig. 29.4).
The specimens are tested under the various combinations of normal and shear loading. The failure loads and time required
to reach the failure loads are recorded for each test case and are given in Table 29.1.
29.2.3 Interface Bonding Strength Determination
The stress distributions along the interface corresponding to the failure loads are determined by 3D non-linear (material
nonlinearity) finite element analysis using the commercial code ANSYS. As an example, the stress distributions along the
interface of the specimen for a combined loading (tensile load: 2,552.72 N and torsional load: 18.82 N m) and the pure
torsional loading is shown in Fig. 29.6.
It is clear from the above figure that there is no normal interface stress along the interface for the application of pure
torsional loading on the present specially designed specimen. Thus, the maximum interface shear stress developed in this test
case is defined as the interface shear bonding strength, ts, which is equal to 15.14 MPa for the subjected PVC/Epoxy
interface. Since, both the normal and shear stress components exist along the interface for the application of pure normal and
Table 29.1 Maximum failure loads and time required to reach the failure loads
No. Tensile load (N) Torsional load (N m) Time (s)
1 0 33.98 676.67
2 766.80 34.06 358.14
3 1,620.21 28.10 265.95
4 2,552.72 18.82 367.03
5 3,112.75 0 368.92
Fig. 29.6 Stress distribution along the interface for (a) combined loading, (b) pure torsional loading
combined loading, a normal-shear bonding strength criterion (envelope) is required to determine the interface bonding
strength envelope. Let assume a quadratic form for the criterion,
ðsn =ss Þ2 þ ðt=ts Þ2 ¼ 1 (29.7)
where, sn and t are the interface normal and shear stress component, and ss and ts are the interface tensile and shear bonding
strengths, respectively.
Assuming the von-Mises equivalent stress criterion as the first trial criterion as given by Eq. 29.8.
s 2 t 2
n
þ ¼1 (29.8)
26:22 15:14
Now, the location of the maximum von-Mises equivalent stress along the interface is determined and the corresponding
pair of normal and shear stress values for each test case is recorded. These pairs of stresses are considered as the first data set.
The von-Mises stress curve (Eq. 29.8) along with the first data set is shown in Fig. 29.7.
However, it is seen that this stress criterion doesn’t fit the data points well. Next, the second trial criterion is obtained by
applying the best fit curve technique on the first data set as,
s 2 t 2
n
þ ¼1 (29.9)
7:95 15:14
Considering the second trial criterion, a second set of data points is obtained. By fitting the second data set, the third trial
condition is obtained as Eq. 29.10.
s 2 t 2
n
þ ¼1 (29.10)
9:63 15:14
Fig. 29.7 Interface bonding

strength envelope
Following the same procedure, for this trial criterion (Eq. 29.10), the third set of stress data points is obtained. Figure 29.7
shows that the third set stress points are fitted well by the trial criterion expressed in Eq. 29.10. Further iteration with this trial
condition (Eq. 29.10) produces the same data points. This demonstrates that the trial condition expressed by Eq. 29.10 is the
final bonding strength envelope for the tested PVC/Epoxy (viscoelastic/viscoelastic) bi-material interface.
29.3 Conclusions
A new experimental method to determine the bi-material interface bonding strength envelope has been presented through an
example of a PVC/Epoxy (viscoelastic/viscoelastic) bonded joint. The stress singularity at the interface corner of the test
specimen can be avoided by the developed method. This new method can be applied for the other combination of materials
as well. This new methodology can be served as a standard method for the determination of the bi-material interface bonding
strength.
Acknowledgement The research is supported by the Natural Sciences and Engineering Research Council of Canada (NSERC) through a research
grant to Z. Xia.
References
1. Bogy DB (1971) Two edge-bonded elastic wedges of different materials and wedge angles under surface tractions. J Appl Mech 38(2):377–386
2. Hein VL, Erdogan F (1971) Stress singularities in a two-material wedge. Int J Fract Mech 7(3):317–330
3. Williams ML (1952) Stress singularities resulting from various boundary conditions in angular corners of plates in extension. J Appl Mech
19(4):526–528
4. Stenger F, Chaudhuri R, Chiu J (2000) A novel sinc solution of the boundary integral form for two-dimensional bi-material elasticity problems.
Compos Sci Technol 60(12):2197–2211
5. Wu ZX (2004) Design free of stress singularities for bi-material components. Compos Struct 65(3–4):339–345
6. Reedy ED (1993) Asymptotic interface corner solutions for butt tensile joints. Int J Solids Struct 30(6):767–777
7. ASTM D3528 (2008) Standard Test Method for Strength Properties of Double Lap Shear Adhesive Joints by Tension Loading, ASTM
International. ASTM International, West Conshohocken, PA
8. ASTM D2095 (2008) Standard Test Method for Tensile Strength of Adhesives by Means of Bar and Rod Specimens ASTM International.
ASTM International, West Conshohocken, PA
9. Chua S, Dai SR, Piggott MR (1992) Mechanical properties of the glass fibre polyester interphase. Part I: effects due to silanes. J Mater Sci
27(4):913–918
10. Drzal LT, Herrera-Franco P, Ho H (2000) Fiber-matrix interface tests. In: Kelley A, Zweben C (eds) Comprehensive composite materials.
Oxford, Pergamon, pp 71–111
11. Miller B, Muri P, Rebenfeld L (1987) A microbond method for determination of the shear strength of a fiber/resin interface. Compos
Sci Technol 28(1):17–32
12. Zhou XF, Wagner HD, Nutt SR (2001) Interfacial properties of polymer composites measured by push-out and fragmentation tests.
Composites A 32(11):1543–1551
13. Alimuddin MA, Piggott MR (2004) Fracture toughness of fiber-polymer interfaces estimated from single fiber peel tests. Polym Compos
20(5):655–663
14. Xu K (2006) Experimental study on biaxial normal-shear interface bonding strength between E-glass and epoxy polymer. M.Sc thesis,
University of Alberta, Canada, 2006
15. Wang P, Xu LR (2006) Convex interfacial joints with least stress singularities in dissimilar materials. Mech Mater 38(11):1001–1011
16. Lauke B (2007) Stress concentration along curved interfaces as basis for adhesion tests. Compos Interf 14(4):307–320
17. Schneider K, Lauke B, Sch€ uller T (2003) Determination of the interface strength between two polymer materials by a new curved interface
tensile test: preliminary experimental results. Compos Interf 10(6):581–591
18. Xia Z, Chowdhuri MAK, Ju F (2011) A New Test Method for Measurement of Biaxial Normal-Shear Bonding Strength at Bi-material
Interface. Mech Adv Mater Struct doi:10.1080/15376494.2011.643278
19. Lauke B (2011) Doubly-curved interfaces for adhesion strength testing. Compos Interf 18(2):121–133
Chapter 30
Calibration of Barkhausen Noise for Residual Stress Measurement
C. Casavola, C. Pappalettere, and F. Tursi
Abstract Industry has been searching for ways to measure residual stresses accurately, quickly and easily without damage
to the material being tested. Several methods have been developed; however, these are either destructive or of limited
capability. A less conventional approach, the magnetic Barkhausen noise (MBN) method, is of particular interest because of
its potential as a non-destructive industrial tool to measure residual stress and other microstructural parameters. The purpose
of this study is evaluation of the relation between residual stress and magnetic Barkhausen noise in Fe360 Steel specimens.
A MBN-stress calibration set-up was developed. To control the accuracy and the effectiveness of the developed system and
procedure, various MBN measurements were carried out; hardness measurements were also conducted.
Keywords Barkhausen noise • Residual stress
30.1 Introduction
Residual stresses can be measured both by destructive methods and by non-destructive methods. Destructive mechanical
techniques are based on residual stress relaxation by material removal. The residual stresses are calculated from the measured
strains. The mechanical methods are not applicable to quality control due to their destructive nature. The most widely applied
non-destructive residual stress measurement method is X-ray diffraction [1]. It is used for quality control but it has several
drawbacks. Firstly, the measuring is quite time-consuming. Secondly, the method is surface sensitive with measuring depth of
the order of several micrometers in steels. Thirdly, it is sensitive to the surface condition of the part. This is why there is a need
for a simple and fast non-destructive residual stress measuring method for quality control. For ferromagnetic steel, a potential
technique for industrial residual stress measurements is the MBN method [2–4]. It has several advantages in respect to the
others methods: it is simple to use, the equipment is cheap, measuring time is short and information depth is fairly large.
In a ferromagnetic material, MBN signal is generated by the irreversible movement of magnetic domains wall [5]. This
movement can be introduced by applying a time varying magnetic field into the material or by introducing dimensional
change into it. Both will force the domain walls to search for new equilibrium positions and this in turn will generate MBN.
The amplitude of the MBN depends on the stress. In steels with positive magnetostrictive constant, there is a stress range
where elastic tensile stresses in the direction of magnetization increase the MBN amplitude while compressive stresses
decrease it [6]. In addition to stress, the MBN is also sensitive to the microstructure, the hardness and the composition that
determine the domain configuration and the characteristics of the domain wall movement [7]. It is well documented that grain
boundaries, lattice dislocations, second phase materials (e.g., carbides in iron), impurities and high hardness of the
ferromagnetic material act as an obstacle for the movement of domain walls and reduces the amplitude of the MBN. This
effect can be broadly described in terms of hardness [8]: the noise intensity continuously decreases in microstructure
characterized by increasing hardness. In this way, Barkhausen Noise measurements provide information on the micro-
structural condition of the material. As several factors affect the MBN its use in residual stress measurements requires
calibration. This is done by calibration specimens made of similar steel whit similar thermo-mechanical history and by
loading them and registering at the same time the corresponding MBN values.
C. Casavola (*) • C. Pappalettere • F. Tursi

Dipartimento di Ingegneria Meccanica e Gestionale, Politecnico di Bari, Viale Japigia 182, 70126 Bari, Italy
e-mail: casavola@poliba.it
256 C. Casavola et al.
The purpose of this paper is to build a stress-MBN calibration curve for the Fe360 Steel, in order to associate at each value
of the MBN signal a value of stress. I will use three specimens, that will be subjected to a tensile test in which a load known
will be applied on them; in correspondence of each level of load, the MBN will be assessed, thus creating for each load a
point of the calibration curve. During the tensile test, the load applied will not exceed the yield point of the material, in order
to remain in the elastic range and record the MBN both in the loading phase and in the unloading phase of each specimen.
To make sure that the stress applied by the traction machine coincides precisely with the stress felt by the specimen,
preliminarily a heat treatment of annealing will be carried out, in order to cancel the residual stresses induced by the
machining process. Another aspect to be taken into account will be the influence of hardness on the response signal.
Therefore, measures of hardness will be conducted on the test specimens under examination, in order to correlate
the calibration curves obtained for each specimen. Finally, since stress states of compression and high hardness imply, for
the same material, a reduction of the MBN, a experimental validation will be tried comparing the measures relating to
specimens annealed and not.
30.2 Materials and Equipments
The material under study is the Fe360 Steel. It has the elastic modulus of 210 GPa, the yield stress of 240 MPa and the ultimate
strength of 415 MPa. The calibration samples for MBN tests were rectangular with dimensions of 25 mm 210 mm cut from
a 4.5 mm tick plate obtained by lamination. The MBN measurements were performed on three specimens; three points were
analyzed on each specimen, located respectively at middle of the specimen (point 2 in Fig. 30.1), at 25 mm to the left of the
central point along the longitudinal axis of the specimen (point 1 in Fig. 30.1) and at 25 mm to the right of the central point
along the longitudinal axis of the specimen (point 3 in Fig. 30.1).
Values of the MBN were measured with commercial equipment Rollscan 300 and sensor model 6020 (Fig. 30.2). The
equipment Rollscan has been controlled by the software Microscan, that allows to observe the characteristics of the received
wave signal and analyze the parameters of this through the acquisition of several cycles of hysteresis in the point of
investigation. The sensor is not equipped with a system of automated moving and this has led to errors of measurement, since
the inclination of the sensor has been adjusted by means of the sensitivity of the operator, not guaranteeing therefore a
constancy of the contact surface of the sensor with the survey area. This involves a Barkhausen signal not perfectly stable,
but oscillating in the neighborhood of real value; therefore multiple measurements are carried out for each point, and only
the average signal of such measures are considered for the analysis.
A MTS Alliance RT/30 tensile testing machine was used to introduce different amounts of stress into the specimens. This
is an uniaxial electric machine with maximum load of 30 kN. It is equipped with a system of control and acquisition, Test
Works 4, able to set the test parameters (crossbar speed advancement, lower and upper limit on the load, etc.) and record the
instantaneous total load measured from load cell, the position of the crossbar, the time corresponding to the data provided
and other inspection parameters requested. The tests were conducted in such a way that the longitudinal axis of the specimen
Fig. 30.1 Specimen used

in MBN calibration
experiment
Fig. 30.2 Equipment

Rollscan 300 and
sensor model 6020
30 Calibration of Barkhausen Noise for Residual Stress Measurement 257
coincides with the load axis of the machine. In order to derive the longitudinal strain of the specimen and therefore to assess
the stress applied to the specimen an extensometer has been used. The extensometer used, in accordance with the
requirements of standards, has a base length of 50 mm, a weight limited as not to induce bending in the specimen, two
knives of support that will not damage the surface of the specimen and a resistance of 350 Ω.
30.3 Experimental Set-up
In this section are presented the procedures for the determination of optimal measuring parameters and for the realization of
the measurements. As already said, in order to calibrate the Barkhausen signal with known values of stress, corresponding to
that imposed by the traction machine, the specimens have been annealed to eliminate the residual stresses of machining
process. During the heat treatment, the specimens have been brought to 600 C, kept at that temperature for 6 h and
subsequently cooled, in oven off, up to ~60 C. The effectiveness of the annealing process was verified by X-ray diffracto-
meter measurements. The annealing has caused the beginning of oxides of not ferromagnetic nature, since due to the
presence of oils and dirt deposited on the surface of the specimens (note that the temperatures reached in the oven are not
sufficiently high for the formation of ferromagnetic oxides). Given the nature of these oxides, it was preferred not remove
them, in order to avoid the onset of further stress and the increase of surface roughness that would have influenced the
measure of the Barkhausen signal. This way, it is decided to omit the eventual effect of reduction of penetration depth of
the magnetic field caused by the oxide layer. Before performing hardness testing it was necessary to remove the oxide layer
that would hamper the vision of the impression generated by the indenter and influence the hardness value measured.
The MBN measurements were performed along the longitudinal direction of the tester, coinciding with the loading
direction of the tensile test. The load occurs in displacement control at a speed of the crossbar set to 0.3 mm/min as suggested
by standard for materials having Young’s modulus greater than 150 GPa. During this phase, the total instantaneous load P
measured by the load cell, the position of the crossbar, the relative displacements of the two knives of the extensometer and
the time corresponding to the data provided are recorded continuously by the acquisition and control system Test Works 4.
The stress applied to the specimen is obtained by dividing the load P in correspondence of the various load levels for the
cross section of the specimen. The same stress value is obtained by multiplying the deformation measured by extensometer
in correspondence of the various levels of loading for the Young’s modulus of the material. To perform the test, in addition
to speed of the crossbar, the direction of advancement of the crossbar and the lower and upper limits of load are set. Define
the load limit is important because it obliges the machine to interrupt the tensile test in the case in which the threshold values
are mistakenly exceeded; in this case, it tries not to go beyond the elastic limit in order not to plasticizing the specimen, and
not to take it in compression. Therefore, the upper limit was set to 15.5 kN while the lower to 0.1 kN. Since the
measurements must be made on the specimen subjected to a load that must be constantly increased, the load step was set
to 3 kN. The maximum value of applied load was 15 kN, in order to stay in the elastic range. So the specimen was loaded
from 0 kN up to 15 kN by five different phases, and then downloaded from 15 kN up to 0 kN by 3 kN steps. The cycle of
loading and unloading are shown in Fig. 30.3. The traction machine, being a machine with displacement control, has
difficulty in maintaining the imposed load; during the test, in fact, the load became stable at a value of ~0.15 kN in less
Fig. 30.3 The cycle of loading and unloading

Fig. 30.4 Scan curve of the magnetizing frequency in a point
compared to that intended during the loading phase, and of ~0.15 kN in more during the unloading phase. To overcome this
drawback, the wanted load level has been increased/decreased of 0.15 kN respectively during each step of loading/
unloading. In this way, once stabilized the value of the load, the MBN measurements were performed. Using Test Works
4 software, the information about the value of applied instantaneous force, the corresponding time, the position of the cross
bar and the deformation of the extensometer are exported in a text file. To locate the value of the applied stress corresponding
to the MBN detected, the initial and final instants of measuring MBN were recorded in a time reference concordant with the
one used by machine; by means of this indication and using the exported text file, it was possible obtain a load value,
averaged between all those that have fluctuated in the time frame of the MBN measurement, which to associate the MBN
signal in the calibration curve.
30.3.1 Determination of the Optimal Magnetizing Frequency [Hz]
The device Rollscan allows to define the frequency with which to obtain the response signal more sensitive. By means of
function scanning of magnetizing frequency, when positioned stably at a point, the machinery provides a curve of the MBN
signal in that point, by varying the frequency of the field of magnetization in the range between 0 and 1,000 Hz (Fig. 30.4).
It is good practice, to be able to perform measurements with the greater sensitivity, choose the value of excitation frequency
in correspondence of the Barkhausen signal greatest of the scan curve. In the present case, the value chosen is 400 Hz.
Scanning was performed on the three points of the three samples, resulting in nine scans. The curves of scanning have a
equal trend in each point of the specimens, but are shifted between them to a negligible distance, thus presenting the
maximum value of the response signal in the neighborhood of 400 Hz of the frequency of magnetization; this value can
therefore be regarded as optimal for the measures on the all specimens.
30.3.2 Determination of the Optimal Magnetizing Voltage [Vpp]
By means of function scanning of magnetizing voltage, when positioned stably at a point, the machinery provides a curve of
the MBN signal in that point, by varying the voltage of the field of magnetization in the range between 0 and 16 V (Fig. 30.5).
To obtain the maximum sensitivity of measurement it is good to choose the voltage value in correspondence of the 3/4 of bend
of the scan curve. You could also choose the voltage value in correspondence of the maximum response signal, but in most
cases you can reach a value of the MBN close to the full scale: for this reason it is suggested to use the value in correspondence
of the bend. In the present case, the optimal voltage is 6 V. Scanning was performed on the three points of the three samples,
resulting in nine scans. Also in this case, the curves of scanning have a equal trend in each point of the specimens, but are
Fig. 30.5 Scan curve of the magnetizing voltage in a point
2.000 40
1.500 30
1.000 20
Magnetization current (mA)

Barkhausen signal (mV)
500 10
0 0
–500 –10
–1.000 –20
–1.500 –30
–2.000 –40
0 2 4 6 8 10 12 14 16
Time (ms)
Fig. 30.6 Barkhausen signal detected in a generic point
shifted between them to a negligible distance, thus presenting the optimal voltage around 6 V; this value can therefore be
regarded as optimal for the measures on the all specimens.
30.3.3 Determination of the Frequency Range of Response Signal
The frequencies range of the response signal chosen is 70–200 kHz. This range allows to reach a depth of inspection of
0.07 mm. The use of that range is recommended because the excitation frequency is high (400 Hz) and produces a signal
having a penetration depth low, then deciding to use a range of frequencies filter of the response signal more low, that would
filter only the signals more in depth, would be a inefficient contradiction.
30.3.4 MBN Measurements During Tensile Test
During the tensile test, the signal of Barkhausen (Fig. 30.6) has been detected at different levels of load in order to build a
calibration curve of the signal in function of applied stress. In this phase of measurement, Microscan software has been used.
Ten measures for each point of each specimen for each loading phase were carried out. The measurements were performed
on the specimen in horizontal position, clamped just below, completely clamped, at 3 kN of the loading phase, at 6 kN of the
loading phase, at 9 kN of the loading phase, at 12 kN of the loading phase, at 15 kN of the loading phase, at 12 kN of
the unloading phase, at 9 kN of the loading phase, at 6 kN of the loading phase, at 3 kN of the unloading phase, fully
discharged. The measuring plane was isolated from any electronic device that might interfere with the field of magnetization
on the material under examination. The execution of the measurements with the specimen in a vertical position (during the
tensile test) is more difficult and uncomfortable for the operator compared with measurements performed with the specimen
in a horizontal position. It is more complicated, in fact, keep the inclination angle of the sensor respect to the plane of the
specimen constant and keep a continuous contact between the sensor and the specimen surface, when the specimen is in a
vertical position. Therefore, measurements on specimen completely unloaded placed in horizontal and then placed in
vertical without that there is still no force applied have been carried out, in order to test the repeatability of measurement and
to succeed to find the same angle of inclination of the sensor compared to the plane of sample in the two conditions. The
absence of a support of automatic moving has caused an error in the measurements, since it appeared difficult to have an
inclination fixed of the sensor, and then have a contact surface of the ferritic coil of magnetization constant. To obviate this,
10 measurements for each point were performed in order to evaluate the average value of the response signal, the standard
deviation and the oscillation range of made measurements. The measurements incoherent and insignificant, due to errors of
positioning, air gap and overheating of the sensor, were discarded from the statistical study. All tests of measurement were
carried out in acceptable and unchanged external conditions of temperature and humidity.
30.3.5 Hardness Measurements
The layer surface of specimens has been cleaned from oxide formed during the annealing treatment. The oxide removal leads
to a further formation of internal tensions because alter the surface state of the material creating surface tensions, which for
the self-equilibrium of the piece generate subsequent stress that are opposed to those of surface. For this reason, the oxide
removal was executed after finishing the measurement of the Barkhausen signal. The specimens were polished on the side on
which the MBN has been detected in order to realize the hardness measurements in the corresponding points. The hardness
measurements were performed every 2 mm distance along the straight line between point 1 and point 3 (Fig. 30.1). In
the present case, having test specimens with a thickness of about 4.5 mm, tests of Vickers hardness with load of 30 kgf
are performed. 26 points of each specimen were investigated so a profile of hardness along the centerline of the specimen
is built.
30.4 Test Results
This section will show the results obtained from Barkhausen signal measurement and hardness tests. Calibration curve MBN
– applied stress will derive, having preliminarily assessed the effects both of the microstructure (hardness) and of residual
stresses “survived” to the thermal treatment. Regard latter, a dual analysis was carried out:
• Previously considering the “survived” residual stresses average zero along all specimens;
• Subsequently using the measures obtained with diffractometer.
30.4.1 Accuracy Analysis of MBN Measurements
Results on accuracy of the MBN measurements are reported in Table 30.1 and then in the corresponding graph (Fig. 30.7).
Mean, standard deviation, maximum value, minimum value and measuring range (Max-Min) calculated relatively to the 10
measurement carried out for each point of three specimens are indicated. The statistical analysis was performed at 6 kN of
unloading phase. Observing the data, it can be assumed that reliable measurements have been made, with the standard
deviation relatively low and the mean range of measurement of 9.8 mp. The differences between the mean values of MBN in
the three points of the same specimen could be due to initial residual stresses of different entities and/or variations
in hardness from position to position within the same specimen. Same motivation could be given to the differences of the
MBN mean values between the different specimens.
Table 30.1 Statistical analysis Measurement point MBN (mp) Specimen 1 Specimen 2 Specimen 3
of the measurements at 6 kN
of the unloading phase 1 Mean 305.5 285.7 300.5
St. Dev. 3.1 3.9 2.3
Max. 310.3 291.2 305.1
Min. 301.4 279.8 296.8
Range 8.9 11.4 8.3
2 Mean 308.4 289.9 315.9
St. Dev. 3.1 2.6 3.3
Max. 315.8 293.8 320.5
Min. 305.0 285.3 310.5
Range 10.8 8.5 10.0
3 Mean 309.6 283.9 307.3
St. Dev. 3.3 3.7 3.1
Max. 315.3 289.5 312.8
Min. 305.9 279.5 302.0
Range 9.4 10.0 10.8
Fig. 30.7 MBN

measurements made on
the three points of the
five specimens at 6 kN
of the unloading phase
30.4.2 Analysis of MBN Parameters
In this section, the additional parameters supplied by the software Microscan are analyzed, varying the stress applied to the
specimen. The parameters relative to the specimen 1 are reported in Table 30.2 which shows that:
• RMS avg (average of Root Mean Square of all signal amplitudes sampled) grows with increasing load and vice versa
during unloading phase since, as known from literature, tensile stresses favor the movement of magnetic domains causing
therefore a higher output signal; RMS is given in volts and is calculated before smoothing, for the specified analysis
frequency range, as follows:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 Xn1 2
RMS ¼ i¼0 i
x (30.1)
n
• Peak avg (average of the absolute values of the peaks heights of the smoothed envelope curve) evolves similarly to the
RMS, because the traction induces a waveform with highest peaks.
• Peak pos avg (average of the values of the peaks heights of the smoothed envelope curve).
• FWHM avg (average of the Full Width Half Maximum of the smoothed envelope curve) decreases consistently with
increasing load because at the grow of the load the waveform of the signal is more tall and narrow.
• Integral area: (area of the hysteresis loop) increases with increasing load, as it is to be expected because materials under
tensile exhibit hysteresis loops larger. Consistently, all characteristic parameters of the hysteresis loop (coercivity,
remanence and permeability) increases with tension increasing.
Table 30.2 MBN parameters of the specimen 1

Phase RMS avg Peak avg Peak pos avg FWHM avg Coercitivity Remanence Permeability Integral area
Horizontal 274.91 392.58 44.09 62.01 0.32 287.46 1,140.80 585.53
Vertical 274.94 391.58 43.40 63.07 0.31 284.93 1,113.44 574.79
Clamped 289.58 446.70 42.84 51.02 0.32 307.37 1,290.41 599.93
3 kN 300.11 476.24 44.26 46.85 0.34 330.59 1,304.22 644.17
6 kN 308.42 497.00 44.63 44.98 0.35 339.25 1,373.32 663.70
9 kN 318.37 514.91 44.01 44.46 0.35 358.47 1,506.75 702.14
12 kN 326.75 529.91 44.27 43.75 0.35 365.41 1,532.35 713.95
15 kN 332.86 542.60 42.75 42.33 0.35 369.28 1,593.95 717.16
12 kN 326.87 532.68 44.17 43.96 0.35 365.87 1,493.60 715.53
9 kN 318.24 514.07 44.33 44.95 0.35 354.79 1,453.72 695.95
6 kN 307.83 497.54 43.97 45.56 0.34 336.40 1,401.46 655.64
3 kN 299.10 466.27 42.78 50.53 0.33 329.85 1,351.68 637.56
0 kN 275.00 393.11 44.70 63.85 0.32 282.01 1,106.82 578.56
Fig. 30.8 MBN measurements made on three points of a specimen annealed and not
In the subsequent analysis, only RMS avg parameter has been considered; for uniformity of language, the RMS avg
parameter has been indicated by MBN [mp].
30.4.3 Analysis of the MBN Signal Before and After Heat Treatment
In this section, it is analyzed the MBN signal of two specimens characterized respectively by:
• Been subjected to a heat treatment of annealing after the machining process (Annealed Specimen)
• Not been subject to any type of treatment after the process of production (Unannealed Specimen)
The MBN results of the Specimen 1 and the Unannealed Specimen are shown in Fig. 30.8. The MBN signal of Annealed
Specimen is higher than that of the Unannealed Specimen, consistent with what is expected. In the present case, the process
of production has generated compressive residual stress, that is reduced accordingly to the annealing, as known from
literature. Therefore, the results obtained are in accordance to the physical nature of the phenomenon. Hardness
measurements are another factor that validate MBN measurements obtained. The Annealed Specimen, in fact, has an
average value of hardness of 112 HV; the Unannealed Specimen, instead, is characterized by an average value of hardness of
about 122 HV. The reduction of hardness, therefore, is concordant with the increase of the MBN signal.
Table 30.3 Measurements MBN (mp) Specimen 1 Specimen 2 Specimen 3

of hardness
Media 108.3 115.0 112.4
Dv st 2.9 5.1 1.6
Max 115.3 122.1 115.3
Min 105.5 107.2 110.1
Range 9.8 14.9 5.2
Table 30.4 MBN – applied Phase Load (kN) Applied stress (MPa) MBN (mp)
stress of specimen 3
Horizontal 0.0 0.00 273.44
Clamped under 0.0 0.00 273.32
Fully clamped 1.2 11.05 287.94
1, 2–3 3.2 28.83 297.78
3–6 5.9 54.13 307.12
6–9 8.9 81.51 317.71
9–12 12.0 109.03 323.54
12–15 15.0 136.50 327.41
15–12 12.0 109.12 322.99
12–9 8.9 80.85 316.92
9–6 6.1 55.31 307.93
6–3 2.9 26.53 296.93
3–0 0.0 0.00 273.84
Fig. 30.9 MBN – applied

stress calibration curve of
specimen 3
30.4.4 Measurements of Hardness
The measurements of hardness (Table 30.3) performed every 2 mm of distance along the centerline of the specimens
between points 1 and 3 shows that the hardness is almost constant on each specimen. Therefore to consider the average
hardness of each specimen is representative. It can be seen that the average values of hardness varies of few units of HV,
negligible for the purposes of the analysis of the Barkhausen signal. Therefore the subsequent analyzes were conducted
omitting the influence of the microstructure, and taking into account only the residual stresses present before the tensile test
sri, considering at first sri average zero along all specimens and subsequently using the measures obtained by diffractometer.
30.4.5 MBN: Applied Stress Calibration Curve (sri Average Zero)
Table 30.4 and Fig. 30.9 summarize, for specimen 3, the phase under consideration, the applied load, the applied stress and
the relative MBN signal. Each value of the Barkhausen noise is averaged over 30 measurements (10 for point 1, 10 for point
2 and 10 for point 3). The applied stress was either calculated by reference to the force P read by the load cell or to the
deformation read by the extensometer, respectively according to the formulas:
P
s¼ (30.2)
A
Fig. 30.10 MBN – applied

stress calibration curve at
constant hardness 112 HV
s¼Ee (30.3)
where A is the cross-section of the specimen and E indicates the Young’s modulus of the material. It may be noted that the
signal increases with increasing of applied stress, and reaches the maximum peak at 15 kN; consistently, during unloading
phase, the signal decreases until to return, with unloaded sample, at the value that it had in the initial condition of test; this
shows that, during the tensile test, the elastic limit was not exceeded. The curve has a non-linear trend: increasing the load, in
fact, there is an increase D of the signal gradually lower between one step and the next; similarly, in the unloading phase
(from 15 to 3 kN), there is a decrease D of the signal gradually increased. This occurs because the magnetization parameters
used for the measurement have been chosen with the specimen completely unloaded and subjected only to residual stresses
sri; therefore the sensitivity of the instrument, decreases with increasing of the force applied on the piece. This does not
invalidate the efficiency of the study done, but suggests to operator who wants to use the calibration curve obtained in this
work to have more attention and care in carrying out these measures, ensuring perfect adherence between the sensor and
surface to be analyzed, and an inclination angle of sensor respect to the plane of the tester equal to that used in this calibration
procedure. It is well marked, moreover, that the MBN signal at 12 kN is very similar to that at 15 kN. In general, the
technique of Barkhausen is capable of perfectly detecting stresses of magnitude less than 50% of the yield strength of
the material under analysis; for values above or close to this limit the signal is saturated and differences are no longer
detectable, thereby the technique becomes less reliable. So, in the case of steel Fe360 (sy ~ 240 MPa), the MBN signal
detected for stresses greater than 120 MPa (corresponding in this case, with specimens of cross section A ~ 100 mm2, at
loads of 12 kN) is saturated, becoming almost constant.
Similar results were obtained for the other two specimens. Figure 30.10 plots the calibration curves for each specimen.
The MBN measurements were carried out at applied load not perfectly coincident for all three specimens due to the problems
of the load machine in the stabilization to a fixed quantity; so, for each sample the measurements were performed around the
loads established during the test plan. The curves do not intersect and maintain a similar pattern. The shift observed between
the curves may be due to an average value of the sri not zero in each specimen. So in the next paragraph this effect is taken
into account.
30.4.6 MBN: Applied Stress Calibration Curve (sri 6¼ 0)
Table 30.5 shows the values of residual stresses sri measured after heat treatment along the longitudinal direction of the
specimens in the points 1, 2 and 3, using the X-ray diffractometer (XRD). It is noted that for each specimen, the stress value
sri varies from point to point, so in the subsequent analysis, the curves are reconstructed at each point. Taking account of
these stresses, the calibration curves previously obtained, have been corrected by adding to the stress applied by the traction
machine, the sri, according to the equation:
sRa ¼ sa þ sri (30.4)
having denoted by sRa, the real stress applied to the specimen and with sa the stress applied by the testing machine.
Figure 30.11 and Table 30.6 summarize the results of the correction in the case of the samples 3. The curves represent
the trend of the stress at each point of each specimen. It is possible to observe the nearly perfect overlap of these curves in the
case of the samples 3.
Table 30.5 Residual stresses sri measured after heat treatment along the longitudinal
direction of the specimens by XRD
sri (Mpa) Specimen 1 Specimen 2 Specimen 3
Point 1 31.4 20.8 11.2
Point 2 60.8 13.0 34.6
Point 3 22.9 25.3 4.6
Fig. 30.11 MBN – applied real stress calibration curve of three points of specimen 3
Table 30.6 MBN – applied real stress of three points of specimen 3

Phase Load (kN) sa (MPa) sRa _1 (MPa) MBN_1 (mp) sRa _2 (MPa) MBN_2 (mp) sRa _3 (MPa) MBN_3 (mp)
Horizontal 0.0 0.00 11.20 265.72 34.60 282.05 4.60 272.54
Clamped under 0.0 0.00 11.20 264.49 34.60 282.59 4.60 272.89
1, 2–3 3.2 28.83 40.03 290.05 63.43 306.08 33.43 297.21
3–6 5.9 54.13 65.33 300.11 88.73 313.83 58.73 307.44
6–9 8.9 81.51 92.71 311.68 116.11 323.28 86.11 318.18
9–12 12.0 109.03 120.23 316.50 143.63 326.38 113.63 327.74
12–15 15.0 136.50 147.70 323.61 171.10 328.48 141.10 330.16
15–12 12.0 109.12 120.32 314.90 143.72 325.90 113.72 328.17
12–9 8.9 80.85 92.05 308.66 115.45 324.13 85.45 317.97
9–6 6.1 55.31 66.51 300.54 89.91 315.92 59.91 307.34
6–3 2.9 26.53 37.73 289.34 61.13 305.19 31.13 296.25
3–0 0.0 0.00 11.20 265.07 34.60 282.99 4.60 273.45
Fig. 30.12 MBN – applied real stress calibration curve at constant hardness 112 HV
Similar results were obtained for the other two specimens. Figure 30.12 shows the calibration curves MBN – real stress
applied to all specimens (for each sample the stress curve of the point more representative was chosen). The curves of the
samples, thanks to the correction of the applied stress, overlap.
30.5 Conclusions
The aim of the present work was to construct a calibration curve MBN signal – applied stress, in order to associate to a
generic output signal, expressed by the magnetoelastic parameter [mp], a value of stress felt by the material. The curve was
obtained for Fe360 Steel, obtained by rolling. The test conditions were obtained using the traction machine with which a
predetermined load is imposed to the specimens; in particular, they have been loaded second step of 3 kN until a load of
15 kN (value well below the yield strength) is reached. In correspondence of each step of loading the MBN measurements
were carried out and associated with the stress value which the specimen was subjected.
Three measurement points of three samples, were investigated. Average 30 measurements (10 for point) were made on
each specimen for each step of the load. The average of these measurements were associate with the stress. It is preferred to
exceed in the measurements in order to minimize the effect of variance due to errors in the positioning of the probe, since it is
not equipped with a system for automatic moving.
The tests were preceded by a heat treatment for stress annealing in order to cancel the residual stresses due to the
production process.
A dual analysis of the results is carried out:
• Considering annealing totally effective, able to eliminate the internal stress
• Using measurements obtained by means of X-ray diffractometer (XRD) on the investigated points immediately after the
heat treatment
In both cases, curves with a similar trend between them and consistent with literature, were obtained. In fact the MBN
value increases proportionally with increasing of applied stress up to 100 MPa; near this value, the signal has a slower
growth because it is close to 50% of the yield strength (limit at which the magnetic domains become saturated).
The comparison between the curves of the different samples was implemented ignoring the influence of hardness. The
results of the hardness measurements, in fact, have shown insignificant changes of this microstructural characteristic from
sample to sample.
The comparison of the signals arising from specimens annealed is conform to the nature of the phenomenon: after heat
treatment, the response signal of a same material is higher because the annealing annuls the residual stresses of compression
thus facilitates the movement of the magnetic domains.
Finally, the analyze of the further characteristic parameters of the Barkhausen signal and the hysteresis loop, shows trend
consistent to the physical nature of the phenomenon.
References
1. Tanaka K, Akiniwa Y (2004) Diffraction measurements of residual macrostress and microstress using x-rays, synchrotron and neutrons. JSME
Int J :252–263
2. Moorthy V, Shaw BA, Day S (2004) Evaluation of applied and residual stresses in case-carburised En36 steel subjected to bending using the
magnetic Barkhausen emission technique. Acta Mater 52:1927–1936
3. Moorthy V, Shaw BA, Mountford P, Hopkins P (2005) Magnetic Barkhausen emission technique for evaluation of residual stress alteration by
grinding in case-carburised En36 steel. Acta Mater 53:4997–5006
4. Moorthy V, Shaw BA, Hopkins P (2006) Surface and subsurface stress evaluation in case-carburised steel using high and low frequency
magnetic Barkhausen emission measurements. J Magn Magn Mater 299:362–375
5. Jiles DC (2000) Dynamics of domain magnetization and the Barkhausen effect. Czech J Phys 50:893–988
6. Lindgren M, Lepisto T (2001) Effect of prestraining on Barkhausen noise vs. stress relation. NDT&E Int 34:337–344
7. Ju JB, Lee JS, Jang JI, Kim WS, Kwon D (2003) Determination of welding residual stress distribution in api x65 pipeline using a modified
magnetic Barkhausen noise method. J Magn Magn Mater 80:641–646
8. Buttle DJ, Moorthy V, Shaw B (2006) Determination of residual stresses by magnetic methods – measurement good practice guide no. 88, NPL,
2006
Chapter 31
Film Residual Stress Assessment Method via Temporarily
Thermal Relaxation
Xin-Xin Cheng and Chen-Wu Wu
Abstract The concept of temporarily thermal relaxation is extended theoretically to assess the film residual stress. First, the
contribution of the initial stress to the observable displacement field is separated with eliminating the thermal expansion
effect of the material. Then, the residual stress is inversely derived through the displacement increment. Finally, finite
element analysis was carried out to provide numerical description of the partial relieve of the residual stress.
Keywords Film • Residual stress • Assessment • Temporarily thermal relaxation
31.1 Introduction
Almost all of the films used in many products are formed with large residual stress. To describe accurately the film residual
stress is of importance for designing a product or evaluating its reliability [1, 2]. A great many of research have been
attracted to focus on the measurement and assessment of the film residual stress [3, 4].
The method of temporarily thermal relaxation [5] is theoretically extended herein to assess film residual stress. The basic
mechanism on such concept can be reviewed briefly herein. It is well known that non-uniform temperature elevation will
lead to non-uniform distribution of elastic modulus in an initial homogeneous solid of material with temperature-dependent
elasticity [6]. The non-uniform change of modulus should result in the redistribution of the initial stress (herein identical to
the term residual stress), which will contribute to the observable displacement field upon thermal loading experiment.
Furthermore, the contribution is determined by the initial stress, the initial and present elastic properties of the material,
among which the analytical relationship could be established. Therefore, the initial (residual) stress could be inversely
derived after extracting the contribution of the initial stress from the observable displacement field. Theoretically speaking,
such separation can be achieved by solving the heat conduction partial difference equations and thermo-elastic partial
difference equations for an actual experiment.
In this article, the displacement increment accompanied with the thermal softening of modulus in an initially stressed film
subjected to non-uniform heating is calculated based on an axis-symmetry model. For the sake of simplicity, the preliminary
analytical solution is derived for a freestanding film and numerical modeling is utilized to investigate the influences of the
substrate on the film deformation. First, the contribution of the redistribution of the initial (residual) stress to the observable
displacement field is separated by eliminating the thermal expansion effect of the material. Then, the residual stress is
derived inversely from such contribution of the initial stress. Finally, finite element analysis was carried through to provide
numerical description of the partial relief of the initially stressed film.
X.-X. Cheng
Institute of Mechanics, Chinese Academy of Sciences, No. 15 Beisihuanxi Road, Beijing 100190, China
Graduate School of the Chinese Academy of Sciences, No. 19A Yuquanlu, Beijing 100049, China
C.-W. Wu (*)
Institute of Mechanics, Chinese Academy of Sciences, No. 15 Beisihuanxi Road, Beijing 100190, China
e-mail: chenwuwu@imech.ac.cn, wcw_on@yahoo.com.cn (C.W.WU)
268 X.-X. Cheng and C.-W. Wu
31.2 Theoretical Analysis
31.2.1 Formularization
Consider a double layer structure of homogeneous film and homogeneous substrate as shown in Fig. 31.1, in which the film
is initially equiaxial tensioned in plane, i.e., the residual stress can be described via
sr ¼ T rs ; s’ ¼ T rs ; tr’ ¼ 0; (31.1)
where s refers to the normal stress component, t is the shear stress component, r and ’ represents the radial and tangent
direction, respectively. Of course, such stress state shall develop a displacement vector field relative to the stress free state
and can be expressed with the radial component as
1n
ur½rs ¼ T rs r; (31.2)
E
where E and n is the Yang’s modulus and Poisson’s ratio, respectively.

Now, assume the film is subjected to non-uniform temperature elevation, i.e., the temperature of the material within a
circular region of radius r is increased via some heating technique, as shown in Fig. 31.2a, b. One can easily understand that
the initial equilibrium state will be disturbed due to the non-uniform thermal softening of the material. A new equilibrium
state will be developed, which, generally speaking, should be accompanied with the redistribution of the displacement.
Divide the temperature elevated circular region into the center circle and a series of annuluses, therefore the elastic
properties can be assumed to be identical in every annulus as shown in Fig. 31.2c.
The interface traction and displacement should be continuous, that is,
sr½k ¼ sr½k 0 ¼ s½k ; (31.3)
And
ur½k ¼ ur½k 0 ¼ u½k : (31.4)
Moreover, the radial displacement can be related to interface traction [7] as:
1 n½0
ur½1 ¼ s½1 r 1 ; (31.5)
E½0
Fig. 31.1 Sketch

of the initially
equiaxially stressed
film and substrate
31 Film Residual Stress Assessment Method via Temporarily Thermal Relaxation 269
Fig. 31.2 Sketch of the non-uniform temperature elevated film

0 1 s½2 r22 s½1 r12 r22 r 1 s½1 s½2
ur½1 ¼ 1 n½1 r 1 1 þ n½1 ; (31.6)
E½1 r22 r12 r22 r12
across the interface between annulus denoted by k ¼ 1 and that by k ¼ 2.

1 s½k rk2 s½k1 rk1
2 r 2 r k s½k1 s½k
ur½k ¼ 1 n½k1 r k 1 þ n½k1 k1 2 ; (31.7)
E½k1 rk2 rk1
2 rk rk1
2

1 s½kþ1 rkþ1
2
s½k rk2 rkþ1
2
r k s½k s½kþ1
ur½k 0 ¼ 1 n½k r k 1 þ n ½k ; (31.8)
E½k 2
rkþ1 rk2 2
rkþ1 rk2
across the interface between No. k annulus and No. (k + 1) annulus.

1 s½n rn2 s½n1 rn1
2 rn1
2
r n s½n1 s½n
ur½n ¼ 1 n½n1 r n 1 þ n½n1 ; (31.9)
E½n1 rn2 rn1
2 rn2 rn1
2
1
ur½n 0 ¼ 2r n T rs ð1 þ nÞr n s½n ; (31.10)
E
across the interface between the annulus denoted by k ¼ n and the annulus beyond the thermal softening region, in which it
is also implied that E½n ¼ E and n½n ¼ n with E and n being the initial elastic constants of the material.
By utilizing the displacement continuous condition at the interface as indicated in Eq. 31.4 and dividing the radius evenly,
which means that the width of every annulus is d ¼ r=n, one can get the series of equations:
½ Afsg ¼ fFrs g; (31.11)
where the elements of the coefficient matrix are


A11 ¼ 3E½1 1 n½0 þ E½0 5 þ 3n½1
A12 ¼ 8E½0
Ak;k1 ¼ 2E½k ð2k þ 1Þðk 1Þ2
8 h i9
< E½k ð2k þ 1Þ 1 n½k1 k2 þ 1 þ n½k1 ðk 1Þ2 >
> =
Ak;k ¼ h i k ¼ 2; 3 . . . n - 1,
: þE½k1 ð2k 1Þ 1 n½k k2 þ 1 þ n½k ðk þ 1Þ2 >
> ;
Ak;kþ1 ¼ 2E½k1 ð2k 1Þðk þ 1Þ2

An;n1 ¼ 2Eðn 1Þ2

An;n ¼ E 1 n½n1 n2 þ E 1 þ n½n1 ðn 1Þ2 þ E½n1 ð2n 1Þð1 þ nÞ (31.12)
and the elements of the force vector are
Frs½i ¼ 0 i ¼ 1; 2; 3 . . . . . . n - 1
Frs½n ¼ 2E½n1 ð2n 1ÞT rs : (31.13)
Solving the equation, the interface traction can be obtained as
fsg ¼ ½ A1 fFrs g: (31.14)
Thus, the displacement can be obtained at every interface as
1 n½0
u½1 ¼ s½1 d
E½0

s k2 s½k1 ðk1Þ2 ðk1Þ2 kðs½k1 s½k Þ
u½k ¼ E½k1
d
1 n½k1 ½k ð2k1 Þ k 1 þ n ½k1 ð2k1Þ k ¼ 2; 3 . . . n 1 :
d
u½n ¼ 2nT rs ð1 þ nÞns½n (31.15)
E
Again, if the specimen dimension is large enough that boundary effect can be neglected, the displacement outside the
thermal softening disc region is

1 r 2 s½n T rs
ur ¼ ð1 nÞT rs r ð1 þ nÞ : (31.16)
E r
Therefore, the radial displacement increment due to the material thermal softening is

ur ¼ ur ur½rs (31.17)
0 1
B11 . . . B1n
B .. .. .. C
Let ½ B ¼ ½A1 ¼@ . . . A, then Eq. 31.14 can be rewritten as
Bn1 Bnn
fsg ¼ T rs 2E½n1 ð2n 1ÞfB1n Bnn gT : (31.18)
Thus the radial displacement at the location r ¼ r can be obtained as following
2r
u½n ¼ 1 ð2n 1Þð1 þ nÞE½n1 Bnn T rs : (31.19)
E
From Eq. 31.2, the initial radial displacement at the same location can be expressed directly as
1n
u½n½rs ¼ rT rs : (31.20)
E
Substitute the relevant components in Eq. 31.17 by Eqs. 31.18 and 31.19, one can get the equation relating the residual
stress and the displacement increment as
r
U IN½n ¼ 2 2ð2n 1Þð1 þ nÞE½n1 Bnn ð1 nÞ T rs : (31.21)
E
Therefore, the residual stress can be derived through Eq. 31.21 as
E
T rs ¼ U IN½n : (31.22)
r 2 2ð2n 1Þð1 þ nÞE½n1 Bnn ð1 nÞ
As a simple illustration, assuming the film is partially subjected to even temperature elevation, which indicates that the
elastic modulus Eand the Poisson’s rationchange into E½0 and n½0 throughout the whole thermal softening region, respec-
tively. Then one can describe the residual stress as

E E 1 n½0 þ E½0 ð1 þ nÞ
T rs ¼ DUIN : (31.23)
r ð1 þ nÞ E 1 n½0 E½0 ð1 nÞ
To sum up, the radial displacement increment is dependent on the initial residual stress T rs and elastic parameters of the
material. Therefore, the initial stress T rs can be calculated inversely once the displacement increment fur g and temperature
profile of the specimen are obtained by measurement and analysis. To this end, heat transfer analysis and direct thermo-
elastic analysis [5] should also be carried out, of which the former is used to determine the softened elastic properties while
the latter is used to eliminate the thermal expansion effect, which is always inevitably included in the real experimental
process.
31.2.2 Analytical Results
Suppose the radius of the circular film specimen to be R ¼ 0.1 m and the radius of the thermal softening region r ¼ 0.01 m.
Assume the lowest value of the modulus at the center of the thermal softening region is 100, 125 and 150 GPa for three cases
and increase linearly up to the initial value of 200 GPa at the location of r ¼ 0.01 m. Moreover, the Poisson’s ratio n ¼ 0.3 is
assumed to be temperature-independent for simplicity.
The results are shown in Figs. 31.3 and 31.4 with the numerical deviation being lower than 0.1%, which is maintained
after the division number n is larger than 400.
Figure 31.3a shows the radial displacement increment fuIN g induced by the thermal softening of the film stiffness, of
which the lowest elastic modulus is 100, 125 and 150 GPa corresponding to the three cases. It is indicated that the
displacement increment increases with decreasing the lowest elastic modulus. Moreover, the influences of the thermal
softening of the material are neglectable when away from the temperature elevated region, which is seemed to be in
accordance with the Saint-Venant’s Principle.
The radial strain increment er ¼ @UIN =@r is diagramed in Fig. 31.3b, which also shows that the radial strain increment
increases with decreasing the lowest elastic modulus. This is easy to understand because, in the extreme, the initial strain
should be totally released if the material within the heated region were removed just like that appears in hole-drilling
process, for which the lowest elastic modulus is certainly reduced to zero. Generally speaking, the higher the elevated
temperature is the lower the minimum modulus will be maintained. Therefore, it can be expected that the displacement/
strain increments will increase with increasing the elevated temperature.
Moreover, the radial displacement increment and strain increment also increase with increasing the magnitude of the
initial residual stress, as shown in Fig. 31.4a, b. Again, in the extreme, one can understand that if there is no initial residual
stress, there’ll still be no displacement even though the material stiffness is unevenly reduced due to non-uniform thermal
softening. Of course, it should be noted again that no thermal expansion effect need to be taken into account in discussing
this phenomenon.
Fig. 31.3 Radial (a) displacement and (b) strain increment for three cases of different thermal softening conditions
31.3 Numerical Computations
31.3.1 FEA Model
Considering the fact that the substrate has not been involved in the analytical analysis model, the Finite Element Analysis
(FEA) model as shown in Fig. 31.5a is utilized to investigate the influence of the substrate on the deformation of the film.
The dimension of the axis-symmetrical model is 100 mm100 mm and the thickness of the film is tf ¼ 100 mm, the initial
residual stress is realized by stretching the rightmost side of the geometry by 0.15 mm along the radial direction.
As an example, the elastic modulus of the film is linearly correlated to the temperature and that of the substrate is fixed to
be 200 GPa. The film elastic modulus is 200 GPa at room temperature and 100 GPa when temperature reaches 300 C and the
its Poisson’s ratio is always 0.3. The applied temperature profile is shown in Fig. 31.5b, in which the film temperature is
300 C at the center (x ¼ 0) and ascends linearly to 25 C at the location x ¼ 5 mm. In computation, the continuous
conditions of both displacement and traction are adopted across the interface between the substrate and film.
Fig. 31.4 Radial (a) displacement and (b) strain increment for three cases of different initial residual stress levels
Fig. 31.5 FEA model of the film-substrate and temperature elevation profile
Fig. 31.6 Radial stress

(a) contour and (b) mapped
onto film surface path
31.3.2 Numerical Results
The redistributions of the residual stress are shown in Fig. 31.6a, b, and the radial displacement/strain increments are shown
in Fig. 31.7a, b.
In Fig. 31.6a, one can see that the radial stress is partially relaxed around the region of elevated temperature. In
comparison, the initial residual stress is uniform within the film as indicated in the curve portion away from the center
graphed in Fig. 31.6b.
The radial displacement/strain increments as shown in Fig. 31.7a, b indicate that the magnitude of the displacement
increment is nearly one order lower than that of the freestanding film although the tendency of the displacement/strain
increments are similar to that of the freestanding film.
Fig. 31.7 Radial (a)

displacement and (b) strain
increment along film surface
path
31.4 Conclusions
The analytical results show that displacement/strain increments will be developed by non-uniform temperature elevation in
the initially stressed freestanding film. Such displacement/strain increments increase with increasing the magnitude of initial
(residual) stress or elevated temperature. The initial (residual) stress can be derived inversely by the displacement increment
and relevant temperature profile. The numerical results reveal that the constraint effect of the substrate will influence the
magnitude of the displacement/strain increment while not alter the basic characteristics of them.
Acknowledgements This work was supported by the National Natural Science Foundation of China (Grant No. 11002145).
References
1. Tonshoff HK, Seegers H (2000) Influence of residual stress gradients on the adhesion strength of sputtered hard coatings. Thin Solid Film
377–378:340–345
2. Teixeira V (2002) Residual stress and cracking in thin PVD coatings. Vacuum 64:393–399
3. Zhang T-Y, Chen L-Q, Fu R (1999) Measurements of residual stresses in thin films deposited on silicon wafers by indentation fracture. Acta
Mater 47:389–3878
4. Chandra T, Ionescu M, Mantovani D (1999) Residual stress analysis of textured materials by X-ray diffraction method. Mater Sci Forum
706–709:1673–1678
5. Chen-Wu Wu (2012) Temporarily thermal relaxation of residual stress. In: Silva Gomes JF, Mário Vaz AP (eds) Proceedings for 15th
international conference on experimental mechanics, Porto/Portugal, 22–27 July 2012
6. Fukuhara M, Sanpei A (1993) Elastic moduli and internal frictions of Inconel 718 and Ti-6Al-4V as a function of temperature. J Mater Sci Lett
12:1122–1124
7. Timoshenko SP, Goodier JN (2004) Theory of elasticity, 3rd edn. McGraw-Hill Companies, Inc., New York, pp 65–71
Chapter 32
Determining Residual Stresses with the Aid
of Optical Interference Techniques
D. Findeis and J. Gryzagoridis
Abstract Most manufactured components contain residual stresses, introduced during the manufacturing process, which
have the potential to affect the performance and life expectancy of the final product.
This paper presents further findings of a research project which investigates the suitability of using optical interference
techniques, namely Digital Shearography and Electronic Speckle Pattern Interferometry, to non-destructively determine
inherent material properties including residual stresses. The method relies on comparing the deformation characteristics of
samples containing residual stresses to the deformation characteristics of fully annealed samples. Initial test results using
Digital Shearography applied to cantilever samples were encouraging and have been extended to steel and aluminium flat
bars subjected to 3 point bending tests using Digital Shearography and Electronic Speckle Pattern Interferometry. In this
paper the two inspection techniques are described and applied to the prepared samples in order to determine the deformation
curves in response to the applied stresses. From the test results the magnitude of the induced residual stresses for the different
samples are calculated and presented. Comparisons between the information obtained from both techniques are drawn and
discussed and pertinent aspects highlighted.
Keywords Residual stresses • ESPI • Digital Shearography • Simply supported beams
32.1 Introduction
The presence of residual stresses in manufactured components is a reality which cannot easily be avoided. Residual stresses
are produced due to e.g. forming procedures or thermal effects during manufacturing and welding processes and can impact
on the performance and life expectancy of the final product. Residual stresses however are not always detrimental but can
also be beneficial and be used as a technique to reduce crack propagation in areas of high stress concentration [1]. It thus is
important to have techniques at hand to determine the presence and magnitude of residual stresses. These techniques can
take on the form of destructive testing techniques [2, 3], where part of the material of the object is removed, or non
destructive techniques which keep the component intact and unaffected by the test.
Two optical interference techniques, namely Digital Shearography (DS) and Electronic Speckle Pattern Interferometry
(ESPI) have shown promise in determining residual stresses. The later is already used successfully in the destructive Hole
Drilling technique to determine residual stresses [2].
This paper is an extension of work conducted into determining the suitability of using optical interference techniques to
determine residual stresses. The outcomes of DS inspections using cantilevers made of mild steel, aluminium and brass were
presented at the 2011 SEM conference [4]. The results obtained were mixed and although they indicated that a change in the
response to an applied force of annealed cantilevers and identical cantilevers with residual stresses was recorded, the results
were largely inconclusive. Questions were raised as to the effectiveness of securely clamping the sample beam as a
cantilever, as well as the method of applying the transverse load to the cantilever. It was thus decided to conduct further
investigations by testing prepared samples both with and without residual stresses in a three point loading mechanism, using
both DS and ESPI. This paper outlines the methodology and results obtained from the inspections.
D. Findeis (*) • J. Gryzagoridis

Department of Mechanical Engineering, University of Cape Town, Private Bag, Rondebosch 7700, South Africa
e-mail: dirk.findeis@uct.ac.za
278 D. Findeis and J. Gryzagoridis
32.2 Theory
Digital Shearography is a laser based optical interferometric technique [5]. The technique employs an expanded monochro-
matic laser to illuminate the inspection area. The laser light reflecting off the surface of the object is viewed through a CCD
camera connected to a PC. Between the camera and object a purpose built shearing device is placed. For a Michelson
configuration the shearing device first splits the image of the object into two distinct images using a 45 beamsplitter and
then using two mirrors recombines them, allowing the two images to overlap. This overlapping image is then captured by the
camera. By manipulating one of the mirrors in the x or y direction, the magnitude as well as position of image overlap,
referred to as shear, can be controlled. A typical optical set-up is shown in Fig. 32.1 below.
The overlap of the two images produces a unique speckle pattern which is captured and stored in a PC. If the object is
deformed due to an applied stress, and causes a relative movement between the two overlapped images, a change in the
speckle pattern occurs. In addition a controlled phase shift is introduced into the beam path length between image
acquisitions. Comparing the before and after speckle images for areas of correlation and decorrelation produces a saw
tooth fringe pattern. The direction of the fringe intensity gradient provides information on the direction of the displacement
gradient. The mathematical formula for this process is outlined below [6].
I i ðx; yÞ ¼ I B ðx; yÞ þ I MP ðx; yÞ cosðyðx; yÞ þ i p=2Þ (32.1)

I 3 ðx; yÞ I 1 ðx; yÞ
’ðx; yÞ ¼ arctan (32.2)
I 4 ðx; yÞ I 2 ðx; yÞ
bðx; yÞ ¼ fa ðx; yÞ fb ðx; yÞ (32.3)
where i ¼ 1,2,3,4
fa(x,y) ¼ phase distribution after stressing, fb(x,y) ¼ phase distribution before stressing.
In order to calculate the magnitude of the displacement gradient the following equation can be used:
@p lN
¼ (32.4)
@x 2S
where ∂p/∂x ¼ displacement gradient in the x (or y) direction, l ¼ wavelength of the laser light
N ¼ number of fringes counted, S ¼ shear magnitude
As with DS, ESPI also uses a single mode monochromatic laser to produce a speckle interference pattern. Two images are
required to interfere with each other in order to produce the speckle interference pattern. ESPI however differs from DS in
that one of the two images acts as a reference image and is not affected by any object stressing. There are a number of ways
to achieve this, either by using a smooth or specular reference beam. A smooth reference beam is obtained by splitting the
laser beam into 2, one acting as an object beam to illuminate the object and the other acting as the reference beam which
Shearing mirror
Phase
Laser stepping
mirror Camera
Beam
Fig. 32.1 Typical Object

Mirror
Shearography set-up
32 Determining Residual Stresses with the Aid of Optical Interference Techniques 279
Fig. 32.2 Typical

smooth reference
Object
beam ESPI set-up PC
L
a
s
e
r
M2
Reference
Beam
Mirror
Video
Object Beam
Mirror Beamsplitter Mirror
Fig. 32.3 Diagram F

of a centrally loaded
simply supported beam x
t
L/2
F/2 F/2
is expanded and combined with the laser light reflected off the object to form a speckle pattern. The second approach is
to obtain a specular reference beam from the physical reflection off a surface illuminated with the same laser light as used to
illuminate the object. Previous work has shown that by using a variable DS shearing device and setting the shear magnitude
of one of the mirrors to such an extent that the image of the surface placed adjacent to the object being inspected is overlayed
onto the object image, a simple yet effective ESPI system can be created. In order to ensure success, the laser beam has to be
expanded sufficiently to illuminate both the object and reference surfaces. When the object is then stressed only the object
surface deforms, the reference surface remains unaffected. The ESPI image correlation process thus determines the absolute
displacement of the object surface in relation to a stationary reference, whereas DS determines the relative object surface
displacement by considering the relative displacement between two moving object surfaces (Fig. 32.2).
Mathematically the ESPI process can be described as below [6]:
nl
d¼ (32.5)
½cos a þ cos b
where:
d ¼ out of plane displacement of the object due to the applied stress
a ¼ angle between the direction of object displacement and camera viewing angle
b ¼ angle between the direction of object displacement and object beam
l ¼ wavelength of the laser beam
n ¼ no. of fringes counted
As indicated above it was decided to test the prepared samples in a three point bending mechanism. The theory for simply
supported centrally loaded beams is briefly outlined in the following paragraphs. Considering the diagram in Fig. 32.3 above,
the beam’s vertical deflection profile can be obtained and is presented in Eq. 32.6 below:

F Lx2 x3 FL3
y¼ (32.6)
2EI 4 6 48EI
Where E is the Young’s modulus and I is the moment of inertia of the beam. The slope of the beam is given by:

dy F Lx x2
¼ (32.7)
dx 2EI 2 2
The curvature of the beam is obtained by differentiating the expression for the slope:

d2 y F L
¼ x (32.8)
dx2 2EI 2
Since y00 ¼ M=EIand s ¼ Mt/2I the stress on the surface of the beam is given by:

Ft L
s¼ x (32.9)
4I 2
32.3 Results
In order to investigate components with embedded residual stresses two sets of mild steel and aluminium beams were
manufactured. Each set consisted of four machined beams which were made of 6 mm thick flat bar and machined to
approximately 4 mm thickness, two on one side only and two on both sides. The final dimensions of the bars are recorded in
Table 32.1 below and can be seen in Fig. 32.4 further down.
After machining the samples were annealed to relieve any residual stresses that may have been present due to the
manufacturing or machining processes. The mild steel samples were heat treated by placing them into an oven at 600 C for
Table 32.1 Beam dimensions Sample no. Length (mm) Width (mm) Thickness (mm)
of the samples
Mild steel
1 297 30 3.8
2 297 30 4.0
3 297 30 3.9
4 297 30 4.1
Aluminium
1 297 30 3.8
2 297 30 3.9
3 297 30 3.6
4 297 30 3.7
Fig. 32.4 Mild steel

and aluminium samples
Fig. 32.5 Three point beam loading rig with sample
Fig. 32.6 ESPI results of mild steel beams, annealed and shotpeened
5 h and then left to cool down to room temperature. Similarly the aluminium was heat treated in an oven at 400 C for 4 h and
then left to cool down to room temperature.
Following the heat treatment process all beams were inspected using DS and ESPI. A purpose made three point loading rig
was manufactured. The span between the two vertical supports which were made of 8 mm tool steel round bar was 290 mm.
The rig included a loading lever which applied the force onto the beam midway between the two vertical supports. The force
was generated by loading the lever with weights of known mass. The final assembly can be seen in Fig. 32.5 above.
All eight samples were inspected using ESPI and DS. For the aluminium samples masses of 11.28 g for ESPI and 61.31 g
for DS were found to produce suitably dense fringes in the final fringe patterns. For mild steel the equivalent masses were
found to be 33.97 g for ESPI and 192.8 g for DS. Following this set of reference inspections, two samples of each beam set
were selected and shotpeened three times on one side only. In Fig. 32.4 the annealed samples are those painted light grey and
the bare samples are those selected and shotpeened. The shotpeened samples were then inspected again using the same
masses used for the initial inspections.
Figure 32.6 above is the set of ESPI results obtained for the mild steel samples 1 and 2; the first image is the fringe pattern of
the annealed form of the beam, which is then followed with the results for the same sample after it had been shotpeened.
The same scenario also applies to sample 2 which is depicted in fringe patterns 3 and 4. The results identify the midpoint of the
beam on the left hand side which is the dark vertical line. In most of the images the bright spot on the bottom edge of the right
hand side of the beam denote the position of the right hand vertical support. For sample 1 15 fringes are recorded for both the
annealed and shotpeened results indicating no noticeable influence of the residual stresses on the displacement magnitude and
profile. Similarly for image 2, there are 14 fringes recorded for both the annealed and shotpeened fringe pattern results.
Fig. 32.7 DS results of

mild steel beams, annealed
and shotpeened
Fig. 32.8 ESPI results

of aluminium beams,
annealed and shotpeened
The same mild steel beams were also inspected using Digital Shearography. The DS results are reproduced in Fig. 32.7
above. Here again, the top two fringe patterns are associated with the annealed and shotpeened beam inspections of mild
steel beam sample 1 and the lower two fringe patterns with the annealed and shotpeened results of the inspection of beam
sample 2. The fringe patterns for image 1 yielded nine fringes in the annealed state and as seen with the ESPI results, also
yielded nine fringes for the shotpeened sample. For sample 2 there is just under half a fringe difference between the results
obtained for the two different states of sample 2, the shotpeened version being the marginally stiffer of the 2. The results here
too show that there is no clear change in the beam stiffness as a result of the introduced residual stress.
The manufactured aluminium samples were also inspected using DS and ESPI inspection methods. The results of these
inspections are listed in Figs. 32.8 and 32.9 below. In Fig. 32.8 the ESPI results obtained for samples 1 and 2 in both annealed
and shotpeened state show a decrease in the fringe density for the shotpeened beam when compared with the initial
Fig. 32.9 DS results of aluminium beams, annealed and shotpeened
Displacements of Annealed and Shotpeened Aluminium Samples

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
0.00E+00
Vertical Distance (m)
-1.00E-06
Sample1
-2.00E-06
-3.00E-06 Sample 2
-4.00E-06 Shotpeened Sample 1
-5.00E-06
Shotpeened Sample 2
-6.00E-06
-7.00E-06
Distance from Centre of Beam (m)
Fig. 32.10 Graph of displacement of aluminium beam samples
annealed state of the samples. The shotpeened results have two fringes less than the annealed state, indicating that the
samples are stiffer and exhibit a lesser displacement magnitude when exposed to the same load.
Likewise for the results reflected in Fig. 32.9 above which are the shearography inspections of the same samples as used
in Fig. 32.8, the fringes recorded for the shotpeened samples 1 and 2 are 10 and 8 respectively, and 11 and 9 for the annealed
forms of the samples. There thus is a reduction in displacement gradient of one fringe, also indicating that the shotpeened
beams exhibited a stiffer response to the applied loads.
From the above results it was decided to investigate the displacement results for the aluminium beam samples. First the
displacement curvature of the four fringe patterns was determined using Eq. 32.5 above. The angles a and b were determined
to be approximately 10 each. Equation 32.6 was then used to determine the theoretical beam displacements using the
experimentally applied force, beam dimensions, x positions according to the fringe locations and Young’s modulus of
70 GPa. The results of this exercise are shown in Fig. 32.10 above. Using Eq. 32.7 again, a reduced force was determined and
used to produce a theoretical displacement profile that tied up with the experimentally obtained displacement profile for the
shotpeened beam samples 1 and 2. Equation 32.9 was then applied to determine the maximum stress for the applied force
(0.1107 N) and the theoretical force (0.107 N for sample 1 and 0.095 N for sample 2). The difference in stress between the
applied and theoretical forces was then found to be 10.7 and 15.8 kPa for samples 1 and 2 respectively which is ascribed to
the induced residual stresses.
32.4 Conclusions
From the above work it has been shown that both ESPI and Digital Shearography can be used to investigate residual stresses.
The technique produces repeatable results and the centrally loaded beam with uniform end supports has produced experi-
mental displacement results with a 96% or better accuracy when compared with theory.
ESPI has shown to be an easier technique to use than DS. The sensitivity of the technique is greater and there is a direct
relationship between the produced fringe density and magnitude of displacement. In addition the absence of the shear
component as is the case for DS simplifies mapping the location of the fringes onto the inspected surface. In addition the use
of the shearing mechanism to create a specular reference beam has proven to be simple to implement and very effective.
For the mild steel beam samples no noticeable drop in the displacement or rate of displacement profile was noted between
annealed and shotpeened beam samples. Judging by the visual appearance of the mild steel surface after shotpeening the
velocity of the shot was most probably insufficient to induce residual stresses of any significant nature. Unfortunately no
alternative shotpeening equipment was available to repeat the process.
The work conducted has determined that for aluminium samples a reduction in beam deflection and rate of deflection is
observed in shotpeened beam samples when compared with equivalent annealed samples. Using beam theory the surface
stress was computed and found to have reduced by 10.7 kPa between the annealed and shotpeened sample 1 and 15.8 kPa
between the annealed and shotpeened aluminium sample 2. It is suggested that this variation in displacement magnitude is
due to the presence of induced residual stresses.
In order to validate the detection of residual stresses using optical interference techniques, future work should investigate
the above approach for different force magnitudes.
References
1. Duggan TV, Byrne J (1979) Fatigue as a design criterion. The Macmillan Press Ltd, London
2. Balalov VV, Pisarev VS, Moshensky VG (2007) Combined implementing the hole-drilling method and reflection hologram interferometry for
residual stresses determination in cylindrical shells and tubes. Opt Lasers Eng 45:661–676
3. Honner M, Litoš P, Švantner M (2004) Thermography analyses of the hole-drilling stress measuring technique. Infrared Phys Technol
45:131–142
4. Findeis D, Gryzagoridis J (2011) Determining residual stress and young’s modulus – can digital shearography assist. In: Proceedings of the 2011
SEM annual conference and exposition on experimental and applied mechanics, Uncasville, 13–16 June 2011
5. Maas AM, Somers PM (1997) Two-dimensional deconvolution applied to phase-stepped shearography. Opt Lasers Eng 26:351–360
6. Findeis D, Gryzagoridis J, Matlali M (2005) Phase stepping shearography and electronic speckle pattern interferometry. In: Proceedings of the
3rd US-Japan symposium on advancing capabilities and applications in NDE, Maui, 20–24 June 2005
Chapter 33
Interfacial Fracture Toughness and Residual Stress
of Thermally Sprayed Coatings
Yoshifumi Okajima, Toshio Nakamura, and Sanjay Sampath
Abstract Residual stresses generated in coatings during thermal spraying plays an important role on the interfacial
adhesion. Furthermore, their magnitude and distributions are affected by both materials and processing parameters. Thus,
it is of interest to evaluate their evolution and influence on the adhesive behavior of thermally sprayed coatings. Given the
complexity in both the process and dimensions, it is difficult to extract both residual stresses and adhesion strength from a
single specimen. In this study, the residual stresses were determined using an in situ thermo-elastic curvature method which
not only provides a non-destructive method of characterizing stress but also allows for its layer by layer evolution during
deposition. Concurrently, the interfacial indentation test is proposed to characterize the adhesion strength of plasma-sprayed
coatings. Through a combination of these experiments, the interfacial fracture toughness of the plasma-sprayed ZrO2
coatings deposited on Al substrates were evaluated. In order to study the effects of the residual stresses, samples were
sprayed under various processing conditions while the indentation tests on resultant coatings were also compared to the
standard tensile adhesion tests. These test results indicate a certain correlation between the residual stresses and the interfacial
fracture toughness of thermally sprayed coatings.
Keywords Gas turbine • Thermal spray • Thermal barrier coating • Interfacial fracture toughness • Indentation
• Adhesion
33.1 Introduction
The role of gas turbines for electricity supply is still increasing worldwide on account of their high efficiency and low
emission. Increased thermal efficiency demands dictate that the turbine inlet temperature be significantly increased. The latest
engine has achieved 1,600 C of turbine inlet temperature [1]. This means that gas turbine components, which are made of
nickel – and cobalt-base superalloys, will require substantial internal cooling. However, cooling air is only available at the
loss of thermal efficiency. Therefore, thermal barrier coatings (TBCs) are applied on the surfaces in order to reduce the extent
of cooling. TBC should achieve both a high level insulation and high reliability.
Material selection and porosity control have potential to obtain a high-level insulation, but the simplest way is to increase
the coating thickness. However, it is widely known that a thicker coating accelerates TBC spallation [2–6]. This is in part
related to the residual stresses within the coating. However, in order to seek proper correlations between thickness and
adhesion, the knowledge of residual stresses within the coating due to processing and service is required.
Beam curvature during spray deposition was developed by Kuroda et al. [7] following the well known concepts in the semi-
conductor industry. Following this initial work, several studies [8, 9] were conducted examining the relationship between
coating formation and substrate curvature. The Center for Thermal Spray Research (CTSR), Stony Brook [10–12], established
in-situ coating property sensor (ICP) to extract coating modulus and residual stress by resource to measurement of substrate
curvature. Through these developments and associated analysis full field description of residual stresses are now available.
Y. Okajima (*)
Mitsubishi Heavy Industries, Ltd, 2-1-1 Shinhama, Arai-cho, Takasago, Hyogo, Japan
e-mail: yoshifumi_okajima@mhi.co.jp
T. Nakamura • S. Sampath
Center for Thermal Spray Research, State University of New York at Stony Brook, 130 Heavy Engineering, Stony Brook, NY, USA
286 Y. Okajima et al.
Chicot et al. [13] proposed and discussed the evaluation of the interfacial fracture toughness at the metal coating/substrate
interface by means of Vickers indentation test. This methodology has flexibility of sample geometries, but requires crack
length measurements which may be inconsistent and subjective to observer. Watanabe et al. [14] proposed and applied
modified tensile adhesion test for HVOF coatings. This methodology can obtain objective data, but limit sample geometries in
the same way as other works. Okajima et al. [15] was conducted these two interfacial fracture toughness tests and confirmed
good correlation between them. In this study, we examine and discuss residual stress obtained by ICP and interfacial fracture
toughness obtained by the indentation test subsequently for the exact same samples produced by a range of processing passes.
In addition, we compare the interfacial fracture toughness results obtained by indentation and adhesion test.
33.2 Experimental Procedure
33.2.1 In Situ Measurement of Residual Stress and Elastic Modulus
In order to study the effect of coating thickness on residual stress, ICP sensor was used to monitor the curvature and
temperature while spraying of substrate beams under four various passes, that is, 15, 22, 29 and 36 passes. Coatings were
deposited after four passes preheating on grit blasted aluminum beams which length, width and thickness were L ¼ 228.6 mm,
W ¼ 25.4 mm and t ¼ 2.3 mm respectively. Spray gun used here is F4 APS torch with an 8 mm nozzle (Sulzer Metco).
ZrO2-8 wt%Y2O3 (204NS-G from Sulzer Metco, Westbury, NY) was used as feedstock. Current, primary gas, secondary gas
and carrier gas were 550 A, 47.5 SLM Ar, 6.0 SLM H2 and 4.0 SLM Ar respectively. Powder feed rate was calibrated to
30 g/min and spray standoff distance was set 100 mm. As a result, coating thicknesses were varied from 0.18 to 0.37 mm for
various passes.
The stress analysis is developed as follows. Deposition stress (from nature: tensile-quenching) was obtained from the
deposit thickness tc and the curvature changes Dk from the deposition period using the Stoney formula [16]:
DkEs 0 t2s
s¼ (33.1)
6tc
where ts is substrate thickness and Es0 is substrate in-plane elastic modulus calculated from Es0 ¼ Es/(1ns). ns is Poisson’s
ratio of substrate. The properties of materials used in the test are Es ¼ 70 GPa and ns ¼ 0.33. Thermal stress was determined
from the deposit thickness and the curvature changes from the cooling period using Eq. (33.1). Finally, residual stress was
determined by superposition of the deposition and thermal stress. Coating modulus can be also obtained from the cooling
curve of the spraying run using following equation represented by Tsui and Clyne [9]:
Dk 6Ec 0 Es 0 tc ts ðtc þ ts Þðas ac Þ

¼ 02 4 (33.2)
DT Ec tc þ 4Ec 0 Es 0 t3c ts þ 6Ec 0 Es 0 t2c t2s þ 4Ec 0 Es 0 tc t3s þ Es 0 2 t4s
where Dk/DT is an experimental result, and ac ¼ 10 106/ C and as ¼ 23 106/ C are thermal expansion of coating
and substrate. Ec0 can be readily computed.
33.2.2 Interfacial Indentation Test
Five samples were cut from aluminum beams sprayed by the particular processing pass, which had been tested for ICP
measurements. After mounted to the epoxy resin, the sample cross sections were well polished to mirror finished. Vickers
indentation tests were conducted with five loads: 9.8, 19.6, 49.0, 98.1 and 196.1 N for each sample. Loading time was 15 s. One
diagonal of the Vickers indenter has to be in the alignment of the interface. The schematic of the test is shown in Fig. 33.1a.
The diagonal of the indenter 2d1 and the crack length 2a were measured by optical microscopy shown in Fig. 33.1b. They
were plotted as a function of applied load in bi-logarithmic scale. Finally, the critical value (Pc, ac) was determined as the
point of intersections between extrapolated lines ln P–ln d1 and ln P–ln a. Apparent interfacial fracture toughness KCa can be
defined as [13],
33 Interfacial Fracture Toughness and Residual Stress of Thermally Sprayed Coatings 287
Fig. 33.1 (a) Schematic of the interfacial indentation test [13], (b) typical observation of optical microscope
1=2
PC E
K Ca ¼ 0:015 3=2 H
(33.3)
aC i
where,
1=2 ðEs =Hs Þ1=2 ðEc =Hc Þ1=2

ðE=HÞi ¼ 1=2
þ (33.4)
1 þ ðH s =Hc Þ 1 þ ðH c =H s Þ1=2
where H is Vickers hardness, and the subscripts i, c and s denote interface, coating and substrate respectively. For the
coatings, the Young’s modulus is obtained from Eq. 33.2 for each specimen while the Poisson’s ratio is set at nc ¼ 0.2.
33.2.3 Tensile Adhesion Test
Figure 33.2a shows the example of a substrate with a carbon layer. Substrate is the aluminum cylinder of 25.4 mm diameter
and 33.0 mm length. After grit-blasting, a carbon layer was placed along the substrate edge with 2.33 mm width from the
edge in order to introduce a weak interface. Four samples were sprayed simultaneous using the same apray parameters as
ICP and indentation samples. Averaged coating thicknesses were varied from 0.18 to 0.38 mm for various passes. After
deposition of a coating layer, another aluminum cylinder of the same configuration was bonded to the ceramic coating using
an adhesive. The thickness of the adhesive layer was about 0.04 mm. This specimen is loaded to fracture by a tensile loading
machine as shown in Fig. 33.2b.
Interfacial fracture toughness Kc can be given as [14],
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

1=2 PC c c
K c ¼ ð1 aÞ pc 1 G : (33.5)
pc2 R R
Where, PC, R and c are critical tensile load, substrate radius and ligament (¼ R – a) respectively, and a and G(c/R) are
Dundurs’ parameter and geometric factor defined as
mS ðkC þ 1Þ mC ðkS þ 1Þ
a¼ ; (33.6)
mS ðkC þ 1Þ þ mC ðkS þ 1Þ
c 1 1 c 3 c 2 c 3 c 4
G ¼ 1þ þ 0:363 þ 0:731 : (33.7)
R 2 2 R 8 R R R
Here, kj ¼ (3nj)/(1 + nj) and mj ¼ Ej/2(1 + nj).

Fig. 33.2 (a) Substrate b

surface with a carbon layer
before spraying, (b) schematic a
of tensile loading machine
universal
.4
joint
er
25
lay
on
rb
top cylinder
ca
epoxy
3(
coating
2.3
bottom cylinder
(substrate)
universal
joint
Table 33.1 lists the ICP derived stress and modulus results. All the measured deposition stresses are tensile and rise up as
number of pass, while thermal stresses are compressive. In general, deposition stress is generated by the contraction of
the individual splats as they cool from the melting temperature to the substrate temperature, which leads to tensile stress.
On the other hand, thermal stress arises from the misfit strain between coating and substrate, which leads to compressive
stress for a combination of low thermal expansion coating and high thermal expansion substrate such as YSZ coating and Al
substrate. The stresses shown in Table 33.1 agree with these general phenomena. Furthermore, it is obvious that deposition
stress magnitudes are much lower than thermal stress and residual stresses dominated by thermal stresses are compressive,
mainly because the stresses due to CTE mismatch contribute more to the final stress states than that due to the quenching of
splats since various stress-relaxation mechanisms can operate during quenching process. In terms of coating moduli, since
spray parameters, except for passes, were completely same, they should result in same material properties. However they
increased as coating thickness increased. It is interesting observation but needed to be further work.
Figure 33.3 shows the stresses calculated from the curvature data measured by ICP sensor as a function of thickness. It is
obvious that the thicker coating shows the bigger tensile deposition stress and the bigger compressive thermal stress,
consequently, compressive residual stress is rising around 0.3 mm and slightly fall over 0.3 mm.
Figure 33.4 shows the typical observation of the interfacial indentation tests performed on the sample sprayed 22 passes.
For our tests, the cracks were always initiated from the interface and propagating along inside the coatings at 30–40 mm away
from the interface. Nevertheless, the cracks kinked out into the coatings for the higher loads such as 98.1 and 196.1 N. Chicot
et al. [13] observed similar crack propagation: the cracks introduced by lower loads remained in the interface plane (zone I),
whereas the cracks introduced by higher loads were deviated into the coating (zone II). They revealed that Zone II crack
accelerate crack propagation rate, but focused on the analysis of zone I crack. Also, in this paper, zone II crack are not in use
for analysis of the interfacial fracture toughness. The critical value (Pc, ac) was determined as the point of intersections
between extrapolated lines of crack length and indenter size, which exclude kinked out crack results.
Interfacial fracture toughness results obtained by indentation test are shown in Fig. 33.5. The interfacial fracture
toughness decreases as the coating thickness increases to 0.34 mm and slightly increases at 0.37 mm. It is evident that
interfacial fracture toughness has good correlation with the residual stress measured by ICP sensor as shown in Fig. 33.3.
Figure 33.6 shows the typical fracture surfaces of tensile adhesion test. The upper pictures are substrate sides and the ring
of the carbon layer can be still observed as dark area. All samples of failure, except for 15 passes, propagated along
the interface as recognized in Fig. 33.6. Samples of 15 passes partially failed along the epoxy adhesive. In contrast to the
indentation test, the cracks introduced by the tensile test were not propagating through the coating.
Table 33.1 Results of ICP measurement

Number of pass 15 22 29 36
Avg. coating thickness (mm) 0.183 0.272 0.339 0.372
ICP measurement Deposition stress (MPa) 22.7 23.3 22.9 28.1
Thermal stress (MPa) 68.5 77.6 74.9 75.2
Residual stress (MPa) 45.9 54.3 52.0 47.2
Ec (GPa) 21.3 23.5 24.9 25.5
-20 30
-30 25
Stress [MPa]
-40 Thermal Stress (ICP) 20

Residual Stress (ICP)
-50 15
Deposition Stress (ICP)
-60 10
-70 5
-80 0
0.0 0.1 0.2 0.3 0.4 0.5
coating thickness [mm]
Fig. 33.3 Deposition, thermal and residual stresses for various thickness via curvature measurements
(9.8 N, ×500) (98.1 N, x 100) Kinking out Kinking out

Coating
2d1 = 413 µm
2d1 = 96 µm
2a = 1259 µm
2a = 271 µm Substrate
(196.1 N, × 50)
Kinking out Kinking out
(19.6 N, ×200) Coating
2d1 = 708 µm
2d1 = 152 µm
Substrate 2a = 2206 µm
2a = 491 µm
8 Kinking out
7 y = 0.7351x + 3.0992
e=0.272
R2 = 0.9978
ln a (a in micro m)
(49.0 N, ×200) 6
5 Apparent hardness
Coating
4
2d1 = 257µm 3 y = 0.5x + 2.838
ac=9.80um
2 R2 = 0.9642
Substrate 1 KC=0.39MPam0.5
2a = 732 µm
0
-4 -3 -2 -1 0 1 2 3 4 5 6
Pc=0.33N
ln P (P in N)
Fig. 33.4 Example of crack observations and representation of ln a–ln P relation

Int. Fracture Toughness [MPa m0.5]

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0 0.1 0.2 0.3 0.4 0.5
Fig. 33.5 Interfacial fracture toughness obtained by interfacial indentation test
Fig. 33.6 Fracture surfaces of the substrate side (upper) and coated side (lower)
Int. Fracture Toughness [MPa m0.5]
4.0
3.5
3.0 IFT_15pass
2.5 IFT_22pass
2.0 IFT_29pass
1.5 IFT_36pass
1.0 average
0.5
0.0
0.0 0.1 0.2 0.3 0.4 0.5
Fig. 33.7 Interfacial fracture toughness obtained by tensile adhesion test
Interfacial fracture toughness are plotted as a function of coating thickness as shown in Fig. 33.7. It is noted that the
interfacial fracture toughness of 15 passes included cohesive effect because those samples partially failed along the epoxy
layer as mentioned above. From Figs. 33.5 and 33.7, it is obvious that interfacial fracture toughness obtained by two
methodologies have a good correlation, but the values from tensile adhesion test are five to eight times as much as those from
indentation test. Four possible reasons can be considered as below.
Int Fracture Toughness obtained by

Fig. 33.8 Interfacial fracture 1.0
toughness plotted as a 0.9
indentation test [MPa m0.5]

function of residual stress
0.8
level for different passes
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-60 -55 -50 -45 -40
Residual Stress [MPa]
1. Different fracture surface: Failure of adhesion test propagated along the perfect interface, while indentation crack
propagated inside the coating. Since fracture toughness of coating itself is much lower than interfacial fracture toughness
[15], IFT obtained by indentation are affected by the coating toughness.
2. Ideal analytical model: Analytical model for tensile adhesion test is assumed that circumferential crack is propagating to
the center point uniformly. However, actual crack might propagate independently from specific weak points.
3. The effect of mixed-mode: Many interface fracture experiments have shown that the mixed-mode condition greatly
affects the measurements of interfacial toughness. Interfacial indentation test and tensile adhesion test evaluate different
mixed-mode.
4. The effect of adhesive layer: Based on the analysis by Qian et al. [17], even less than 100 mm thickness of adhesive layer
can attribute to interfacial fracture toughness. In this paper, though thickness of adhesive layer is approximately 40 mm,
the effect is not taken into account.
Interfacial fracture toughness obtained by indentation test from the four various passes specimen sets are plotted together
in Fig. 33.8 as a function of residual stress. A least squares fit to all the data is shown in the figure. This result corresponds
with Johnson et al. [18]. They varied compressive residual stress and measured compressive shear failure strain for EB-PVD
TBCs and concluded that TBCs with lower compressive residual stresses were more strain tolerant and required larger
additional applied mechanical strain for failure. The relationship between interfacial fracture toughness and residual stress
shows same tendency with their results.
Evans and Hutchinson [19] analyzed the mechanics of the delamination and spalling of compressed films or coating using
a combination of fracture mechanics and post-buckling theory. They have revealed that the associated stress intensity factor
depends on the magnitude of the pre-stress and the film thickness, which implies that such films are susceptible to
delamination and eventual spalling. The correlation between compressive residual stress and interfacial fracture toughness
studied in this article is compatible with their analysis.
33.4 Conclusion
It is widely believed that residual stresses generated during the deposition of coatings act as pre-existing stresses and
adversely affect the coating durability. Similarly, a thicker coating accelerates TBC spallation life in service. In this paper,
the residual stresses generated during the deposition were measured by ICP sensor and interfacial fracture toughness were
obtained by indentation test after ICP measurement for various coating thicknesses. Interfacial fracture toughness obtained
by indentation test were compared with those obtained by tensile adhesion test. The main conclusions can be summarized as
follows:
1. Compressive residual stresses would increase as coating thickness increase and eventually saturate. Subsequently,
compressive residual stress was decreasing because tensile deposition stress was increasing. Interfacial fracture tough-
ness obtained by both methodologies corresponded with the residual stress. The origin of this correspondence is that
compressive residual stress could enhance the stress intensity factor.
2. Interfacial fracture toughness obtained by tensile adhesion test were much higher than those by indentation. A few
possibilities are considered, but further study will be needed.
Acknowledgements Authors would like to thank Dr. Y. Chen for his assistance in processing some of the plasma-sprayed samples. We also
acknowledge assistance provided by Dr. G. Dwivedi and Mr. V. Viswanathan.
References
1. Hada S, Masada J, Ito E, Tsukagoshi K (2011) Evolution and future trend of large frame gas turbine for power generation – a new 1600 degree
C J class gas turbine. In: International gas turbine congress, Osaka, 13–18 Nov 2011, IGTC2011-189
2. Godoy C, Souza EA, Lima MM, Batista JCA (2002) Correlation between residual stresses and adhesion of plasma sprayed coatings: effects of
a post-annealing treatment. Thin Solid Films 420–421:438–445
3. Bose S, DeMasi-Marcin J (1997) Thermal barrier coating experience in gas turbine engines at Pratt & Whitney. J Thermal Spray Technol
6(1):99–104
4. Steffens H-D, Babiak Z, Gramlich M (1999) Some aspects of thick thermal barrier coating lifetime prolongation. J Thermal Spray Technol
8(4):517–522
5. Choi HM, Kang BS, Choi WK, Choi DG, Choi SK (1998) Effect of the thickness of plasma-sprayed coating on bond strength and thermal
fatigue characteristics. J Mater Sci 33:5895–5899
6. Bossmann H-P, Bachegowda S, Schnell A (2008) Manufacturing optimization for bondcoat/thermal barrier coating system. In: Proceedings of
ASME turbo expo 2008, Berlin, 9–13 June 2008, GT2008-51366
7. Kuroda S, Dendo T, Kitahara S (1995) Quenching stress in plasma sprayed coatings and its correlation with the deposit microstructure.
J Thermal Spray Technol 4:75–84
8. Gill SC, Clyne TW (1990) Stress distributions and material response in thermal spraying of metallic and ceramic deposits. Metall Trans B
21:377–385
9. Tsui YC, Clyne TW (1997) An analytical model for predicting residual stresses in progressively deposited coatings, part 1: planar geometry.
Thin Solid Films 306:23–33
10. Matejicek J, Sampath S (2003) In situ measurement of residual stresses and elastic moduli in thermal sprayed coatings-part 1: apparatus and
analysis. Acta Mater 51(3):863–872
11. Matejicek J, Sampath S, Gilmore D, Neiser R (2003) In situ measurement of residual stresses and elastic moduli in thermal sprayed coatings-
part 2: processing effects on properties of mo coatings. Acta Mater 51(3):873–885
12. Sampath S, Matejicek J (2002) Method and apparatus for determining process-induced stresses and elastic modulus of coatings by in situ
measurement. U.S. Patent 6,478,875
13. Chicot D, Demarecaux P, Lesage J (1996) Apparent interface toughness of substrate and coating couples from indentation tests. Thin Solid
Films 283:151–157
14. Watanabe M, Kuroda S, Yokoyama K, Inoue T, Gotoh Y (2008) Modified tensile adhesion test for evaluation of interfacial toughness of HVOF
sprayed coatings. Surf Coat Technol 202:1746–1752
15. Okajima Y, Nakamura T, Sampath S (2012) Role of process conditions on the interfacial fracture toughness of plasma Sprayed Zirconia. In:
Proceedings of the ITSC 2012, Houston, 21–24 May 2012 p 178–183
16. Stoney G (1909) The tension of metallic films deposited by electrolysis. Proc R Soc A (London) 82:172–175
17. Qian G, Nakamura T, Berndt CC, Leigh SH (1997) Tensile toughness test and high temperature fracture analysis of thermal barrier coatings.
Acta Mater 45(4):1767–1784
18. Johnson CA, Rund JA, Bruce R, Wortman D (1998) Relationship between residual stress, microstructure and mechanical properties of electron
beam-physical vapor deposition thermal barrier coatings. Surf Coat Technol 108–109:80–85
19. Evans AG, Hutchinson JW (1984) On the mechanics of delamination and spalling in compressed films. Int J Solids Struct 20(5):455–466
Chapter 34
Design of Ultrasonic Probe Configuration Using Finite-Difference
Time Domain Simulation
Rahul Sharma, Shyamsunder Baskaran, and H. Murthy
Abstract An analysis tool has been developed for optimizing ultrasonic transmitter-receiver configurations for in-situ
monitoring of crack propagation during cyclic loading in order to develop better crack tip models for fatigue life estimation.
Time-of-flight diffraction technique is simulated using finite-difference time-domain method to study the interaction of
ultrasonic waves from probes with defects/cracks. Governing equations relating velocities to stresses are discretized using
central finite-difference formulation and solved on a staggered grid in an explicit time-marching scheme, with velocity and
stress components offset in time and space. This leads to a leap frog scheme in which the velocity and stress components are
calculated alternately from each other. Grid size is taken as the ratio of minimum wavelength and number of steps per
wavelength ðN ¼ 1520Þ to ensure stability. Time step is obtained from Courant stability criteria. Defect surface is
modelled as traction-free. Perfectly matching layer with small damping factor is applied to minimize the amplitude of
waves reflected from boundaries created at the edges of domain. Simulation matches well with results from finite element
model using ABAQUS / EXPLICIT solver. FDTD method can be used for designing optimum transmitter-receiver configu-
ration since it is computationally less expensive and easy to implement as compared to FEM.
Keywords FDTD • Finite difference time domain • Ultrasonic wave propagation • TOFD • Time of flight diffraction
• Crack • FEM • Staggered grid • Simulation
34.1 Introduction
Engineering components are designed with adequate strength to resist the stresses arising in service conditions. However,
defects and cracks formed during faulty manufacturing processes or due to fatigue, foreign object damage and wear and tear
during service life of the component lead to a significant reduction in strength of the component and its eventual failure. The
detection and sizing of crack is essential to predict life of structural component in service condition. To detect the cracks, one
of the widely used non-destructive testing technique is ultrasonic testing because of its ease of use and repeatability. Further,
it is capable of detecting and sizing cracks in a wide range of locations and orientations and does not require safety
restrictions (like in radiography). To develop better crack growth models under fatigue/ cyclic loading and to investigate the
three dimensional effects in complex geometries, it is useful if we can monitor the propagation of crack front as a function of
time while the experiment is running in a laboratory setup. This is especially important in investigating the crack growth rate
in anisotropic materials like single crystal materials where crack growth is complex in nature. Aim of this work is to develop
a tool to design and optimize an ultrasonic inspection configuration for in-situ monitoring of crack front propagation during
cyclic loading. The transmitter-receiver configuration can be optimized to provide maximum information about the real time
shape and size of the crack front.
R. Sharma • H. Murthy (*)

Department of Aerospace Engineering, Indian Institute of Technology Madras, Chennai 600036, India
e-mail: hsnmurthy@gmail.com
S. Baskaran
IQC Advanced Inspection Solutions Pvt Ltd, 40 ECR, Chennai 600 041, India
294 R. Sharma et al.
Current technology like Time of flight diffraction (TOFD) technique is based on the use of pair of probes and uses the
time of flight of the diffracted waves to estimate the size of cracks in a medium [1, 2]. Using single transmitter-receiver pair
for detection of crack as in case of pitch catch mode does not provide sufficient data to obtain the size and position of the
defect accurately and results in poor resolution. Scanning the whole surface for defect using one pair of transducers is time
consuming. Therefore, ultrasonic probe arrays could be used to obtain comprehensive information about the crack front as a
function of time while the experiment is running. As several sources are in contact with the medium in an array based
ultrasonic imaging, different configurations of transmitter-receiver (T-R) combinations can be employed to achieve more
useful reconstructions. In this study, a tool has been developed to optimize the location for transmitters and receivers in an
arrangement of probes in staggered fashion. The data collected by multiple receiver arrays can be processed to generate a
reconstructed image of the defect.
34.2 Numerical Modelling of Wave Propagation
Conducting experiments over a wide range of defects and specimens is both time consuming and costly. Therefore numerical
methods such as finite-difference time domain (FDTD) technique and finite element analysis are preferred for simulating the
wave propagation phenomenon. Simulation results can be utilized further in the design of experimental setup.
34.2.1 Finite Difference Time Domain (FDTD) Formulation
Using velocity-stress based finite-difference formulation, compression and shear wave propagation in heterogeneous
medium is reported by Virieux [3]. A two-dimensional (2-D) finite-difference model based on velocity-stress formalism
for elastic waves in the ground was reported by Schroder et al. [4]. They used the numerical model to study the interaction of
elastic waves with a buried land mine. Satyanarayanan et al. [5] have reported the simulation of phased array ultrasonic wave
interaction with the defects using the finite-difference time domain (FDTD) method. Based on these previous works, a two-
dimensional (2-D) and three-dimensional (3-D) finite difference model for propagation of elastic waves in homogeneous,
isotropic solids has been implemented in MATLAB. The following governing partial differential equations (in indicial
notation) relating velocity and stress at a point in a solid medium are derived from Newton’s second law of motion and the
time derivative of constitutive relations for isotropic material.
sij;j
v_ ¼ ; s_ij ¼ mðvi; j þ vj;i Þ þ ldij vk;k (34.1)
r
where v_ indicates the derivative of velocity with respect to time and dij is the Kronecker delta. Specimen is assumed to be a
homogenous, elastic and isotropic solid defined by density r and Lame’s constants l and m. Equation (34.1 were re-written
using central finite-difference formulation and solved for velocities and stresses on a staggered grid in an explicit time
marching manner. The finite-difference discretization of the system of first order partial differential equations leads to a
staggered finite-difference grid as shown in Figs. 34.1 and 34.2 for 2D and 3D case respectively.
By introducing the finite differences in space Dx and Dy and in time Dt, the equations for 2D case can be discretized as
follows.
V x kþ0:5 ji0:5; j0:5 V x k0:5 ji0:5; j0:5 sxx k ji; j0:5 sxx k ji1; j0:5 sxz k ji0:5; j sxz k ji0:5; j1
r ¼ þ
Dt Dx Dz
Dt k
) V x kþ0:5 ji0:5; j0:5 ¼ V x k0:5 ji0:5; j0:5 þ sxx ji; j0:5 sxx k ji1; j0:5
rDx
Dt
þ sxz k ji0:5; j sxz k ji0:5; j1
rDz
(34.2)
34 Design of Ultrasonic Probe Configuration Using Finite-Difference Time Domain Simulation 295
Fig.Fig. 34. x
nbsp;34.1 Staggered grid for
i-1 I i+1
two dimensional analysis [5]
j-1
j v
sxz
z
sxx, szz
j+1
Fig. 34.2 Staggered grid for

three dimensional analysis [5] Normal stress sxx syy szz
Shear stress sxy
Shear stress sxz
Shear stress syz
Velocity in X-direction -Vx
Velocity in Y-direction -Vy
Y Velocity in Z-direction -Vz
X
Z
sxx kþ1 ji; j0:5 sxx k ji; j0:5 V x kþ0:5 jiþ0:5; j0:5 V x kþ0:5 ji0:5; j0:5 V z kþ0:5 ji; j V z kþ0:5 ji; j1
¼ ðl þ 2mÞ þl
Dt Dx Dz
Dt
) sxx kþ1 ji; j0:5 ¼ sxx k ji; j0:5 þ ðl þ 2mÞ V x kþ0:5 jiþ0:5; j0:5 V x kþ0:5 ji0:5; j0:5 (34.3)
Dx
Dt
þl V z kþ0:5 ji; j V z kþ0:5 ji; j1
Dz
Here capital letters stand for the numerical value of corresponding field component at a discrete location in space and
time. Superscript represents the time index and subscripts after j character denotes the spatial location in the grid. In the same
manner, discretized equations can be obtained for 3D model for all field components. The velocity and the stress components
in the grid are not known at the same position in time and space and are offset by Dt/2 and Dx/2, Dy/2, Dz/2 as shown in
Fig. 34.2. The above discretization leads to a leap frog scheme in which the velocity components are calculated first, then the
stress components from the velocity components, the velocity components again using the stress components and so on.
Thus, knowing the initial values of the field variables through the entire domain, field components can be determined for all
later times. The grid size in the model is taken as lmin/N, where lmin is the minimum wavelength in a domain and N is the
number of steps per wavelength. It was observed that N value of 15–20 makes the iterative formulation stable. Time step is
obtained from Courant stability criterion, which is a necessary as well as sufficient condition to ensure stability and obtain
numerical convergence in FDTD method [6]. Courant condition relates the space grid size (h), time step (t) and spatial
dimension (N): Dt p1ffiffiffi
h Cmax N . Similar discretization is done for 3D case as well.
Initial condition in terms of displacements is implemented on a finite number of transmitting surface nodes. The
displacement is excited by cosine-modulated Gaussian waveform and is shown in Fig. 34.3. During analysis, signal received
at different receiver locations is averaged and the velocity component normal to the surface is compared with the ABAQUS
results.
Fig. 34.3 Cosine modulated 1

Gaussian pulse with time
shift [7] used as initial
condition at the transmitter
location
0.5
Amplitude
0
−0.5
0 0.2 0.4 0.6
Time (micro−seconds)
Fig. 34.4 Longitudinal a Absorbing Boundary b Reflecting Boundary

and shear waves propagating
in a medium with absorbing
and reflecting boundary
at the edges
Point Point
Source Source
Absorbing Boundary
34.2.2 Boundary Conditions
34.2.2.1 Absorbing Boundary Conditions
While solving the partial differential equations numerically, it is often essential to introduce artificial boundaries to limit the
area of computation. Absorbing boundary conditions are therefore enforced at the edges of the computational domain to
absorb waves and thereby model the unbounded region. Berenger developed a perfectly matching layer (PML) that provides
reflection-less absorption of electromagnetic waves independent of frequency and angle of incidence. The idea behind the
PML condition is to define a non-physical set of equations in the PML region which exhibits large attenuation and does not
produce reflection for a wave propagating into PML region for an adjoining lossless acoustic region. Implementation of
absorbing boundary condition and free surface boundary condition is based on the method followed by Kevin et al. [8].
Reflection of waves from the boundaries has been minimized by implementing the absorbing layers at the edges of the
domain. The damping factor D, is given as D ¼ ð1 a pÞ. Distance a is zero inside the domain and increases as one
moves from inside the absorbing layer to the outside. p is the percentage by which the amplitude of parameters are reduced
as wave propagates through subsequent layers. This technique can be used to reduce the reflections for all acoustics and
elasto-dynamic finite integration methods [8]. An absorbing region of 80 nodes thickness with a damping percentage of
0. 2% (p ¼ 0. 002) worked well for reducing reflections from the artificial ends in this case. Figure 34.4a shows the waves
propagating in media that are absorbed at the boundary using above mentioned condition. It clearly shows that the absorbing
boundary conditions implemented as discussed above eliminates reflections at the boundaries.
34.2.2.2 Free Surface Boundary Condition
Using a staggered grid, free surface boundary can be modelled by explicitly satisfying the zero stress condition at the free
surface. For the given model, all the surfaces exposed to air and the crack surface are modelled as traction free. On traction
free surface, shear stress and the stress normal to the surface should be zero. If the 2D domain is assumed to be X Z plane
then the free surfaces at the top and bottom of the 2D domain (Z ¼ 0 and Z ¼ width) in the X Zplane is given by
szz ¼ 0; sxz ¼ 0. Similarly for free surface at left and right of the 2D domain (X ¼ 0 and X ¼ width) in the X-Z plane has to
Fig. 34.5 Implementation x

of free surface condition
in a 2D grid [11] i-1/2 i i+1/2 i+1 Free Surface
j-1/2
j
u
j+1/2
v
z j+1 sxz
sxx, szz
satisfy sxx ¼ 0; sxz ¼ 0. For achieving the stress free conditions various methodologies have been presented for FDTD
method [8, 9]. Min et al. [10] reported the implementation of free surface condition in which the field variables were defined
at node based grid set whereas the material properties were defined at the cell based grid set. In the cell based grid, the stress
free condition can be obtained by defining the change in material property at the surface without implementation of
additional boundary condition. The FDTD grid was arranged in such a way that the horizontal (x direction) surface always
ends with a (Vx - sxx/szz ) layer and a vertical (z direction) surface always ends with a (Vy- sxx /szz ) layer. The outer most
layers in the leap-frog finite difference grid were considered as dummy layers to implement traction free condition. Shear
stresses in the layer just below the outer dummy layer were set directly to zero. To ensure that the longitudinal stress at the
horizontal surface vanishes, szz(i, j) ¼ szz ði; j þ 1Þ for the upper boundary and szz(i, j) ¼ szz ði; j 1Þ for the lower
boundary as implemented by Rudd et al. [8]. i, j are the indices corresponding to x and z directions corresponding to outer
horizontal dummy layer. Similarly, sxx(i, j) ¼ sxx ði þ 1; jÞ and sxx(i, j) ¼ sxx ði 1; jÞ for left and right traction free
vertical surfaces. The zero stress formulation [11] provides stable and accurate results for vertical and horizontal free surface
boundary. For a given 2D grid represented in a X-Z plane as shown in Fig. 34.5 the traction free surface condition is attained
using anti-symmetry property.
For 3D case, the traction free surface can be modelled by satisfying the condition szz ¼ sxz ¼ syz ¼ 0; . Assuming a free
surface boundary to be at Z¼k, stress free condition for a grid shown in Fig. 34.5 can be achieved using the property of
antisymmetry about the free surface and is given by the Eqs. 34.4–34.6.
szz ¼ 0; szz jj1 ¼ szz jjþ1 (34.4)
sxz jj1 ¼ sxz jjþ1 ; sxz jj3 ¼ sxz jjþ3 (34.5)

2 2 2 2
syz jj1 ¼ syz jjþ1 ; syz jj3 ¼ syz jjþ3 (34.6)

2 2 2 2
For all the FDTD simulations performed in the present work, the above given free surface boundary conditions have been
implemented. If the free surface boundary is irregular or inclined, the zero stress formulation becomes difficult because the
boundary will not align with the layout of finite difference grid. Because of the issues related to numerical stability and
accuracy using finite difference methods, free surface boundary conditions requires very careful handling.
34.2.3 Geometry of the Model Used
The probe array is being designed to inspect crack front in a dog-bone shape specimen used during experiments in a
laboratory. To optimize, transmitters and receivers have to be placed at various angles to determine the position where
relatively better crack tip diffracted signal can be received. Wedges with different angles are required to perform inspection
at appropriate angles. Therefore to avoid use of multiple wedges, a polygonal block has been designed as shown in Fig. 34.6,
to achieve T-R combinations at different angles with respect to crack tip.The designed block can be assumed to be a part of
dog-bone shape specimen, inspected at various angles. Holes can be machined on the top of specimen to hold probes firmly.

The edges are arranged to obtain readings at 13 different locations at an interval of 12 . Simulations were performed on a
20 mm thick aluminium block embedded with 5. 32 mm deep and 0. 3 mm width notch made. The results obtained from
FDTD simulations are compared with finite element analysis based package ABAQUS/EXPLICIT results.
Fig. 34.6 Polygonal block a

can be assumed to be a part
of dog-bone specimen
inspected at various angles.
(a) Dog-bone specimen.
(b) Polygonal block geometry
34.2.4 Finite Element Simulations
The principle of FE method is to solve a large and complex problem in terms of simple and smaller elements. The variation
of the field components in an element is approximated using shape functions. The model can be refined by decreasing the
element size resulting in better mesh resolution and thus tending the approximate solution towards the exact solution at an
expense of computational cost. Depending on the strategy used to convert the differential equations into algebraic equations,
finite element methods can be divided into explicit and implicit types. In explicit types, the value of a variable such as
displacement at a certain time is determined based on its known values at earlier time steps. On the other hand, in implicit
algorithms, values of variables at future time steps are needed to determine their current values. Implicit methods are more
complex to program and require more computational time for each time step. Explicit methods are easier to program,
however they demand smaller time steps compared to implicit approach. For problems involving dynamic phenomenon such
as ultrasonic wave propagation, where instantaneous results at each time step are required, explicit methods are preferred.
ABAQUS/EXPLICIT has been used to model propagation of ultrasonic waves in the notched polygonal block [12].
The polygonal block was discretized using 4-noded square plane strain (CPE4R) elements for 2D and using tetrahedral
element for 3D geometry. Mesh size is chosen such that it provides at least ten elements for the smallest propagating
wavelength present in the ultrasonic pulse. Choosing the correct element size is a crucial issue in studying the wave
propagation by finite element modelling. Correct size of elements reduces the computation time while maintaining the
required accuracy of the results. Trial and error studies showed that in order to have a low computation time and good
convergence, a mesh size of the order of 5 10 5 m should be chosen. This provides 12 elements for the smallest
propagating wavelength present in the ultrasonic pulse. The input ultrasonic pulse in terms of displacements (same as that
used in FDTD simulations as shown in Fig. 34.3) was introduced at finite number of transmitting surface nodes.
Fig. 34.7 Propagation of waves in the 2D FE model of polygonal block with a notch. (a) Waves generated by initial cosine-modulated Gaussian
waveform at the transmitter location. (b) Longitudinal, shear and surface waves generated in the polygonal block due to interaction of the transitted
waves with boundaries. (c) Interaction of waves with notch
a x 10−9 FDTD b −9 FDTD

2 2 x 10
0 0
−2 −2
Amplitude
0 1 2 3 4 5 0 1 2 3 4 5
Amplitude
−9 FEM
x 10−9 FEM
5 x 10
2
0 0
−2 −5
0 1 2 3 4 5 0 1 2 3 4 5
Time (micro−seconds) Time (micro−seconds)
Fig. 34.8 Resultant velocity normal to the surface at receivers 11 and 13. (a) At receiver 11. (b) At receiver 13
34.2.5 Comparison Between FDTD and FE Results
The finite element analysis based on ABAQUS is computationally expensive as compared with the FDTD approach.
Therefore to get qualitative agreement between FDTD and FE results, the simulations are performed on the model scaled
to one-fifth the size of original model. Figure 34.7a shows the waves generated due to the initial cosine-modulated Gaussian
waveform applied at the transmitter location. The Longitudinal, shear and surface waves generated due to interaction of
waves with boundaries can be observed in the FE simulation shown in Fig. 34.7b and the interaction of waves with the notch
is shown in Fig. 34.7c. Similar results were obtained in the FDTD analysis. To further compare the two models, the velocity
component normal to the receiving surface is averaged and compared at different receiver locations (Fig. 34.8). The results
from the two analyses match reasonably well.
The finite element analysis for 3D using ABAQUS is computationally even more expensive as compared to 2D.
Therefore to get qualitative agreement between FDTD and FE results, the simulations are performed on the model scaled
to one-tenth the size of original model. Figure 34.9a shows the waves generated due to the initial cosine-modulated Gaussian
waveform applied at the transmitter location. The Longitudinal, shear and surface waves generated due to interaction of
waves with boundaries can be observed in the FE simulation shown in Fig. 34.9b and the interaction of waves with the notch
is shown in Fig. 34.9c. Again, similar results were obtained in the FDTD analysis. To further compare the two models, the
velocity component normal to the receiving surface is averaged and compared at different receiver locations (Fig. 34.10).
The results from the two analyses match reasonably well and give us confidence in using FDTD approach for designing the
probe configuration.
Fig. 34.9 Propagation of waves in the 3D FE model of polygonal bloack with a notch. (a) Waves generated by initial cosine-modulated Gaussian
waveform at the transmitter location. (b) Longitudinal, shear and surface waves generated in the polygonal block due to interaction of the transitted
waves with boundaries. (c) Interaction of waves with notch
a
x 10−10 FDTD b 2 x 10−10 FDTD
5
0 0
−5 −2
Amplitude
0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
Amplitude
x 10−10 FEM x 10−10 FEM

1
5
0 0
−5
−1
0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
−10
c 5 x 10 FDTD d 2 x 10−10 FDTD
0 0
−5 −2
Amplitude
Amplitude
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5

−9 FEM x 10−9 FEM
1 x 10 1
0 0
−1 −1
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
Fig. 34.10 Resultant velocity normal to the surface at receivers 11a, 13a, 11b and 13b. (a) At receiver 11a. (b) At receiver 13a (c) At receiver 11b
(d) At receiver 13b
34.3 Conclusion
The 2D and 3D model for wave propagation using finite difference time domain (FDTD) technique was successfully
implemented in MATLAB. Absorbing and reflecting boundary conditions were applied appropriately and verified. FDTD
technique is found to be computationally less expensive and easy to implement as compared to ABAQUS. The results from
FDTD analysis match well with results from FE analysis of the scaled models. The model can be further extended for the
design of ultrasonic probe configuration.
References
1. Charlesworth JP, Temple JAG (2001) Engineering applications of ultrasonic time-of-flight diffraction. Research Studies Press Ltd., England
2. Ogilvy JA, Temple JAG (1983) Diffraction of elastic waves by cracks: application to tofd. Ultrasonics 21:259–269
3. Virieux J (1986) P-sv wave propagation in heterogeneous media: velocity-stress finite difference method. Geophysics 51(4):889–901
4. Schroder CT, Scott WR Jr (2000) A finite difference model to study the elastic wave interactions with buried land mines. IEEE Trans Geosci
Remote Sens 38(4):1505–1512
5. Satyanarayanan L, Sridhar C, Krishnamurthy CV, Balasubramaniam K (2007) Simulation of ultrasonic phased array technique for imaging and
sizing of defects using longitudinal waves. Int J Press Vessels Pip 84:716–729
6. Schroder CT, Scott WR Jr (2002) On the stability of the fdtd algorithm for elastic media at a material interface. IEEE Trans Geosci Remote
Sens 40(2):474–481
7. Elsherbeni A, Demir V (2009) The finite-difference time-domain method for electromagnetics with Matlab simulations. Scitech Publishing,
Inc., Raleigh
8. Rudd KE, Leonard KR, Bingham JP, Hinders MK (2007) Simulation of guided waves in complex piping gometries using the elastodynamic
finite integration technique. J Acoust Soc Am 121(3):1449–1458
9. Harker AH (1984) Numerical modelling of the scattering ofelastic waves in plates. J Nondestr Eval 4(2):89–106
10. Min DJ, Shin C, Hai SY (2004) Free surface boundary condition in finite difference elastic wave modelling. Bull Seismol Soc Am
94(1):237–250
11. Graves RW (1996) Simulating seismic wave propagation in 3d elastic media using staggered grid finite differences. Bull Seismol Soc Am
86(4):1091–1106
12. Honarvar F, Khorasani S (2010) Simulation of tofd technique by finite element method. In: Simulation in NDT
Chapter 35
Absorbed Versus Released Energy in the Cracking Process
of Heterogeneous Materials Under Compression
G. Lacidogna, M. Corrado, and A. Carpinteri
Abstract In this investigation, the energy absorbed or released in the post-peak regime of heterogeneous materials
subjected to compression are evaluated and distinguished. To this aim, uniaxial compression tests have been carried out
on different types of rocks, such as limestone, marble and granite. The obtained overall responses, in terms of load versus
displacement curves, are very different and variable from ductile to brittle, with the appearance of catastrophic snap-back
instabilities, even for the same size and slenderness. In the case of snap-back instabilities, in order to obtain the complete
load–displacement response, the tests have been controlled by the circumferential strain with a linked chain placed around
the cylindrical specimen at mid-height. The absorbed energy per unit surface is computed through the overlapping
constitutive law. Such a parameter, that results to be almost constant by varying the size-scale of the specimen, is compared
to the elastic energy accumulated in the body at the point of instability to define a structural brittleness index. Finally, the
released energy is put into relation to the different kinds of energy emissions detected during the loading process, namely,
mechanical, electromagnetic, and nuclear.
Keywords Rocks • Compression test • Brittle fracture • Energy release • Acoustic emission
35.1 Introduction
Damage and fracture characterizing the compressive failure of heterogeneous materials such as rocks and concrete are
complex processes involving wide ranges of time and length scales, from the micro- to the structural-scale. They are governed
by the nucleation, growth and coalescence of microcracks and defects, eventually leading to the final collapse, and to the loss
of the classical mechanical parameters, such as nominal strength, dissipated energy density and deformation at failure, as
material properties. For instance, it is well-established from experiments that a strong localization of deformations occurs in
the post-peak regime, and the energy dissipation is a surface-dominated phenomenon, in close analogy with the behaviour in
tension (see the experimental results by Hudson et al. [1] for rocks, and by Kotsovos [2], van Mier [3] and Jansen and Shah [4]
for concrete). According to these evidences, the Overlapping Crack Model has been proposed by Carpinteri et al. [5] for
modelling the crushing process in concrete-like materials. Such a model, dual to the Cohesive Crack Model routinely adopted
for quasi-brittle materials in tension, assumes a stress–displacement (fictitious interpenetration) law as a material property for
the post-peak behaviour, to which corresponds an energy dissipation over a surface. This simple model has permitted to
explain the well-known size and slenderness effects on the structural ductility, characterizing the mechanical behaviour of
concrete-like materials subjected to uniaxial and eccentric compression tests [5, 6]. Analogously to quasi-brittle materials
subjected to tension, the stability of the overall behaviour of specimens in compression depends on geometrical (size and
slenderness) and mechanical parameters (crushing energy, compressive strength and ultimate strain).
The energy release during experimental tests can be assessed on the basis of the acoustic emission (AE) monitoring
technique, that proves possible to detect the occurrence and evolution of stress-induced cracks. Cracking, in fact, is
accompanied by the emission of elastic waves which propagate within the bulk of the material. These waves can be received
and recorded by transducers applied to the surface of structural elements. This technique, originally used to detect cracks and
G. Lacidogna (*) • M. Corrado • A. Carpinteri

Department of Structural Engineering and Geotechnics, Politecnico di Torino, Corso Duca degli Abruzzi 24, 10129 Torino, Italy
e-mail: giuseppe.lacidogna@polito.it
304 G. Lacidogna et al.
plastic deformations in metals, has been extended to studies in the field of rocks and concrete, and can be used for the
diagnosis of structural damage phenomena [7]. Recently, AE data have been interpreted by means of statistical and fractal
analysis, considering the multi-scale aspect of cracking phenomena [8], showing that the energy release, proportional to the
cumulative number of AE events, is a surface-dominated phenomenon. Analogously, also the localisation of cracks
distribution within the specimen volume by means of the AE technique has physically confirmed the localisation of the
energy dissipation over preferential bands and surfaces during the damage evolution [9–11].
In the present paper, experimental results of compression tests carried out on three cylindrical specimens of different of
rocks, are analyzed from the point of view of the balance between stored, released and absorbed energy during the complete
loading process. The stability of the loading process is also investigated, and described on the basis of a dimensionless
number, function of mechanical and geometrical parameters. Finally, the released energy is put into relation to the different
kinds of energy emissions detected during the loading process, namely, mechanical, electromagnetic, and nuclear.
35.2 Scale Independent Constitutive Laws
Damage localization strongly affects the behaviour of heterogeneous materials in compression, with particular regard to the
post-peak regime [1–4]. According to several experimental evidences, the Overlapping Crack Model proposed by Carpinteri
et al. [5] describes the inelastic deformation due to material damage in the softening regime by means of a fictitious
interpenetration of the material, while the bulk material undergoes an elastic unloading. As a result, it introduces a couple of
constitutive laws in compression, in close analogy with the Cohesive Crack Model: a stress–strain relationship until the
compressive strength, sc, is achieved (Fig. 35.1a), and a stress-displacement (overlapping) relationship describing
the material crushing and expulsion (Fig. 35.1b). The latter law describes how the stress in the damaged material decreases
by increasing the interpenetration displacement, up to a residual value, sr, is reached, to which the critical value for
displacement, wcr, corresponds. The area below the stress-overlapping displacement curve of Fig. 35.1b represents the
crushing energy, GC, which can be assumed as a size-independent material property.
According to the Overlapping Crack Model, the mechanical behaviour of a specimen subjected to uniaxial compression
(see Fig. 35.2) can be described by three schematic stages. A first stage where the behaviour is mainly characterized by the
elastic modulus of the material: a simple linear elastic stress–strain law can be assumed, or even more complicated nonlinear
relationships, taking into account energy dissipation within the volume due to initiation and propagation of microcracks (see
Fig. 35.2b). By approaching the compressive strength, such microcracks interact forming macrocracks, and, eventually,
localizing on a preferential surface. A second stage where, after reaching the ultimate compressive strength, sc, the inelastic
deformations are localised in a crushing band. The behaviour of this zone is described by the softening law shown in
Fig. 35.1b, whereas the outside part of the specimen still behaves elastically (see Fig. 35.2c). The displacement of the upper
side can be computed as the sum of the elastic deformation and the interpenetration displacement w:
d ¼ el þ w; for w<wcr ; (35.1)
where l is the specimen length. Both e and w are functions of the stress level, according to the corresponding constitutive
laws shown in Fig. 35.1. While the crushing zone overlaps, the elastic zone expands at progressively decreasing stresses.
When d wcr, in the third stage, the material in the crushing zone is completely damaged and is able to transfer only a
constant residual stress, sr (see Fig. 35.2d). As a result, very different global responses in the s–d diagram can be obtained
by varying the mechanical and geometrical parameters of the sample. In particular, the softening process is stable under
Fig. 35.1 Overlapping crack

model: (a) pre-peak
stress–strain diagram;
(b) post-peak
stress–overlapping
displacement law
35 Absorbed Versus Released Energy in the Cracking Process of Heterogeneous. . . 305
Fig. 35.2 Subsequent stages in the deformation history of a specimen in compression
Fig. 35.3 Stress–displacement response of a specimen in compression: (a) normal softening, (b) vertical drop, (c) catastrophic softening
(snap-back)
Fig. 35.4 Compression tests on concrete specimens by Ferrara and Gobbi [13] (a); stress versus strain diagrams (b); overlapping law diagrams (c)
displacement control, only when the slope dd/ds in the softening regime is negative, Fig. 35.3a. A sudden drop in the load
bearing capacity under displacement control takes place when the slope is infinite, Fig. 35.3b. Finally, the snap-back
instability is avoided, Fig. 35.3c, if the loading process is controlled by means of the localized interpenetration or the
circumferential strain, the slope dd/ds of the softening branch being positive. In accordance with the study on the stability of
the loading process proposed by the authors in Ref. [8], a catastrophic softening (snap-back) occurs when:
scE 1
B¼ ; (35.2)
ec l 2:3
where l ¼ l/d is the specimen slenderness, ec is the elastic strain recovered during the softening unloading, and
scE ¼ GC =sc d is the energy brittleness number in compression, proposed by Carpinteri et al. [5, 6].
An extended validation of the Overlapping Crack Model for concrete-like materials can be found in Carpinteri et al. [6],
where specimens of different materials as well as different slenderness and/or sizes have been considered. The experimental
stress-overlapping displacement curves obtained for the concrete specimens tested by Ferrara and Gobbi [13] are shown in
Fig. 35.4. As far as the stress–strain constitutive law is considered, the elastic behaviour is independent of any geometrical
Fig. 35.5 Compression tests on rock specimens by Carpinteri et al. [13] (a); stress versus strain diagrams (b); overlapping law diagrams (c)
parameter, whereas the post-peak regime is largely influenced by the slenderness and the scale of the specimen (Fig. 35.4b).
As a consequence, the stress–strain relationship cannot be assumed as a material property. On the contrary, the
stress–displacement curves collapse onto a very narrow band (see Fig. 35.4c), demonstrating that the s-w relationship is
able to provide a slenderness and size-scale independent constitutive law. In general, the post-peak s-w relationships can
be computed from the s–d diagrams by subtracting the elastic elongation, caused by the reduction of the applied stress in the
post-peak regime, del, and the pre-peak plastic deformation, dpl. Similar results are obtained for the limestone specimens
extracted from a pillar of the Syracuse Cathedral [13] (Fig. 35.5). The substantial collapse of the curves onto a narrow band
obtained in comparison to the stress–strain diagrams, extends the applicability of the s–w relationship, originally proposed
for concrete, also to rock materials.
35.3 Experimental Tests
The experimental investigation herein presented focuses on the mechanical response of cylindrical specimens of three
different rock types, namely Luserna stone, Carrara marble and Syracuse limestone, subjected to uniaxial compression tests.
All the three samples have a diameter equal to 52 mm and a length equal to 104 mm. On the basis of previous experiences, a
very brittle behaviour, characterized by a snap-back instability, was expected for the Luserna specimen, an intermediate
response for the Carrara sample, and a ductile behaviour for the limestone. On the other hand, a correct evaluation of the
dissipated and released energies during the complete loading process can be done only if the post-peak branch of
the load–displacement curve is detected. To this purpose, the loading processes of the Luserna stone and Carrara marble
specimens have been controlled by the circumferential strain, that results to be a monotonic increasing function of time, even
in case of brittle behaviour. An extensometer attached to the ends of a linked chain with roller at each hinge was placed
around the cylinders at mid-height, measuring the circumferential expansion. Such a device is particularly effective in
controlling the post-peak phase for specimens with slenderness of about 2.0, characterized by a splitting failure mode.
Different velocities have been imposed during the tests, depending on the brittleness of the considered phase, in order to
better control the loading process: low values for the pre-peak and the first part of the post-peak range (1 105 mm/s
about), and large values for the tail (1 103 mm/s about). Due to the expected ductile behaviour for the limestone
specimen, its loading process was controlled by the longitudinal displacement (piston stroke). In this case, a constant piston
velocity was imposed, equal to 1 103 mm/s. In all the cases, the load was applied by means of rigid platens with friction.
During the tests, each specimen was monitored by the acoustic emission technique. The AE emerging from the
compressed specimens was detected by applying to the sample surface a piezoelectric (PZT) transducer, sensitive in
the frequency range from 50 to 800 kHz for detection of high-frequency AE [8–10, 12].
The load versus circumferential and longitudinal displacement curves obtained from tests are shown in Figs. 35.6a, 35.7a,
and 35.8a. The mechanical response of the Luserna stone is highly instable in terms of load versus axial displacement,
whereas a global softening behaviour with the presence of several peaks and valleys is obtained in terms of load versus
circumferential expansion (Fig. 35.6a). The Carrara marble is characterized by a load versus axial displacement with perfect
vertical drop after the peak load and a smooth softening curve in the load versus circumferential expansion diagram
(Fig. 35.7a). Finally, a ductile response in the load–axial displacement diagram has been obtained for the Syracuse limestone
specimen (Fig. 35.8a).
Fig. 35.6 Compression test on the Luserna stone specimen (d ¼ 52 mm, l ¼ 104 mm): (a) load versus circumferential expansion and longitudinal
displacement diagrams; (b) load versus time curve with cumulated number of AE and, (c) AE count rate
Fig. 35.7 Compression test on the Carrara marble specimen (d ¼ 52 mm, l ¼ 104 mm): (a) load versus circumferential expansion and
longitudinal displacement diagrams; (b) load versus time curve with cumulated number of AE, and (c) AE count rate
Fig. 35.8 Compression test on the Syracuse limestone specimen (d ¼ 52 mm, l ¼ 104 mm): (a) load versus longitudinal displacement diagram,
and (b) load versus time curve with cumulated number of AE and AE count rate
As far as the acoustic emission activity is concerned, in the Luserna stone specimen it is mainly concentrated in the
post-peak stage, as clearly evidenced by the cumulated number of AE and the AE count rate diagrams shown in Fig. 35.6b, c.
The cumulated number of AE events at the end of the test is about 4,000. Analogously, also the Carrara marble sample has
exhibited significant AE events in the post-peak, even if the cumulated number is one order of magnitude lower than that of
the Luserna stone (see Fig. 35.7b, c). On the contrary, almost all of the AE events in the limestone specimen has been
detected in the pre-peak stage, as shown in Fig. 35.8b. The cumulated number at the end of the test is around 2,500.
35.4 Stability of the Loading Process and Energy Release
The mechanical responses obtained for the three considered materials correspond to those shown in Fig. 35.3: normal
softening, vertical drop and snap-back instability. Such a transition can be obtained either by varying the specimens
geometry, the material being the same, or, as in our case, by varying the material properties and keeping the geometric
parameters constant. The procedure described in Sect. 35.2 has been used to derive the overlapping laws for the three
considered materials. They are shown in Fig. 35.9, in terms of normalized stress versus overlapping displacement. The values
of the critical interpenetration displacement, the strength and the crushing energy are reported in Table 35.1. The crushing
energy GC has been evaluated as the area beneath the overlapping law comprised between the peak stress and a residual
stress equal to 0.2sc. On the basis of these mechanical properties, and the specimen diameter, the energy brittleness number
sE has been computed for each specimen. The lowest value is that of the Luserna stone sample, equal to 0.0016, and the
largest one is that of the Syracuse limestone sample, equal to 0.065. The marble specimen presents an intermediate sE, equal
to 0.0065. The values of the parameter B, defined in Eq. 35.2, are reported in Table 35.1. They completely agree with the
critical condition defined by Eq. 35.2: snap-back instability for B < 0.43, and normal softening for B > 0.43. The Carrara
marble specimen, characterized by a vertical drop in the load–displacement diagram, has B ¼ 0.41.
According to the Overlapping Crack Model, the energy absorbed in the post-peak phase by the damaging process is equal
to the crushing energy times the transversal cross-section of the specimen. The values for the three samples are reported in
Table 35.1. On the other hand, the elastic strain energy at the peak load, DE, represents the maximum energy stored in the
specimen, that will be partially absorbed and partially released during the post-peak phase. As can be deduced from the
results reported in Table 35.1, the Luserna specimen releases an energy equal to 156.89 J, and the Carrara sample to 8.18 J.
On the contrary, the limestone specimen requires a supplementary amount of energy, to be provided by the testing machine,
for the advancement of the post-peak damaging process. As regards the Carrara marble specimen, crushing energy is not
exactly equal to DE, even if the load–displacement diagram has a vertical drop, because the crushing energy is computed up
to the residual stress sr ¼ 0.2sc, and not up to a complete failure. The obtained experimental results evidence that the AE
activity can be directly related to the energy release in the post-peak phase. The largest number of AE events, in fact, has
been detected for the Luserna stone specimen, which, according to the previous calculations, is that releasing the largest
amount of energy. Limited AE events have been detected from the Carrara marble specimen, for which the accumulated and
the absorbed energies are almost the same, whereas very low AE activity has been observed for the limestone specimen.
As regards the Luserna stone specimen, it is worth noting that, when the energy release takes place in a catastrophic manner,
Fig. 35.9 Overlapping constitutive laws for Luserna stone, Carrara marble and Syracuse limestone specimens
Table 35.1 Mechanical parameters for the three considered specimens

sc (MPa) ec wcr (mm) GC (N/mm) sE B GC area (J) DE (J)
Luserna stone 168 0.0105 0.12 13.90 0.0016 0.079 29.51 186.40
Carrara marble 75 0.0080 0.60 25.35 0.0065 0.410 49.57 57.75
Syracuse limestone 51 0.0180 6.00 165.00 0.0650 1.889 323.89 85.42
i.e. the compression test is controlled either by the load or by the axial displacement, instead of in a gradual manner, as in the
presented experimental test, the power of the crushing process largely increases, producing different forms of energy
emission. Such phenomenon is evidenced in the next section.
35.5 Additional Forms of Energy Emitted During Brittle Fracture
It is possible to demonstrate experimentally that the failure phenomena, in particular when they occur in a brittle way, emit
additional forms of energy related to the fundamental natural forces. As a matter of fact, in their recent works the authors
have found increasing experimental evidences that energy emission of different forms occurs from solid-state fractures
[14–20]. By subjecting brittle or quasi-brittle materials to compression tests, bursts of neutron emission are observed during
the failure process [14, 15], necessarily involving nuclear reactions, besides the well-known acoustic emission (AE), and the
phenomenon of electromagnetic radiation (EM) [17], which is highly suggestive of charge redistribution during material
failure and at present under investigation [21, 22]. Finding convincing explanations of all this complex phenomenology is
still an open issue. EM emission was initially explained in terms of charge separation occurring across the fractures, even if
no reasonable explanation is found for each crack face to obtain a net charge [23]. More recently, a model preserving the
charge neutrality of crack surfaces, where optical modes get excited and result in oscillating dipoles at each newly formed
crack surface, was proposed [24].
The experimental analysis carried out by the authors may open a new possible scenario, in which the coherent EM
fields associated with the superradiant motions of solids are able to produce neutron bursts in presence of sudden stress-drops
or catastrophic fractures, where charged-particle acceleration may take place. As an example, the results obtained on a
cylindrical specimen drawn from a previous experimental campaign [19] are presented in Fig. 35.10. The material used for the
tests was non-radioactive Luserna stone, a metamorphic rock deriving from a granitoid protolith. The compression test was
performed in axial displacement control, and the specimen was monitored by using AE sensors, EM devices and neutron
devices. The AE sensors were calibrated on frequencies comprised between 50 and 800 kHz, the electromagnetic signals were
detected using a Narda device ELT-400 exposure level tester. The adopted device works in the frequency range between
10 Hz and 400 kHz, the measurement range is between 1 nT and 80 mT. As regards the neutron emissions, an He3 neutron
a b Cps - Luserna stone (D=53mm, H=101mm)

-2
250 1200 250 Average Neutron Background (4,74±0,46)x10 cps 35
Luserna stone (D=53mm, H=101mm)
16 (µT) Load (kN)
Cumulated AE
30
Load (kN) 1000
200 200
25
800
150 150
Load (kN)
Cumulated AE
20
Load (kN)
600
15
100 100
400
10
50 50
200
5
0 0 0 0
0 200 400 600 800 1000 1200 1400 1600 0 250 500 750 1000 1250 1500 1750
Time (s) Time (s)
Fig. 35.10 (a) Load versus time diagram and AE cumulated number (dashed line) for a cylindrical specimen in Luserna stone. A signal of 16 mT is
detected in correspondence to the peak load. (b) Load versus time diagram, and neutron emissions count rate detected by He3 neutron detector
proportional counter [20]
detector proportional counter was used. The compression tests were performed at the Fracture Mechanics Laboratory of the
Politecnico of Torino by means of a servo-hydraulic press, with a maximum capacity of 1,800 kN, working by a digital type
electronic control unit.
Experimental evidence in Luserna stone specimens indicates the presence of AE and EM signals: an increasing AE
activity is always detected as the load increases, while EM emissions are mainly observed when abrupt stress drops occur.
For test specimens with more brittle behaviour, neutron emissions significantly higher than the background level were found.
These emissions are due to piezonuclear reactions, which depend on the different modalities of energy release during the
tests. For specimens with sufficiently large size and slenderness, a relatively high energy release is expected, and hence a
higher probability of neutron emissions at the time of failure.
Accordingly, neutron emissions may be accounted for by the fact that the power threshold for piezonuclear reactions is
exceeded, as well as by the type of catastrophic failure that occurs, which entails a very fast energy release [14]. As a matter of
fact, the experimental results show that a volume approximately exceeding 200 cm3, combined with the extreme brittleness
of the tested material, represents a critical threshold value for a neutron emission of about one order of magnitude higher than
the ordinary background [18, 19]. For the specimen represented in Fig. 35.10, having dimensions of diameter d ¼ 53 mm and
height l ¼ 101 mm, neutron emissions of about five times the background level were observed, concomitant with the sharp
stress drop at the time of failure. Since the analyzed material contains iron, it was found that piezonuclear reactions involving
fission of iron into aluminum, or into magnesium and silicon, should have occurred during compression of the specimens
[14–20]. This conjecture was also confirmed by the Energy Dispersive X-ray Spectroscopy (EDS) tests [25]. This hypothesis
seems to find surprising evidence and confirmation at the Earth crust scale from both the geomechanical and the geochemical
points of view [26].
35.6 Conclusions
In the present paper, the results of compression tests carried out on cylindrical specimens of different rock types have been
analyzed considering the stability of the loading process, and the relationship between absorbed and released energy during
the post-peak phase. The following main conclusions may be drawn:
1. According to the Overlapping Crack Model herein applied, the structural brittleness can be predicted by means of the
energy brittleness number sE. In particular, for the considered specimen dimensions, constant for all the tested materials,
Luserna stone results to be the most brittle, compared to Carrara marble and Syracuse limestone.
2. The results depicted in Figs. 35.6c, 35.7c and 35.8b show how the number of AE events in the post-peak regime is
proportional to the released energy, that is evaluated as the difference between the elastic strain energy at the peak load
and the energy absorbed by the crushing process.
3. It is also evidenced that, when the compression test of a very brittle material – such as Luserna stone – is controlled only
by the axial displacement, the fast energy release at the collapse provokes the concomitant appearance of electromagnetic
radiations and neutron emissions, besides AEs.
Acknowledgements The financial supports provided by the Ministry of University and Scientific Research (MIUR) to the project “Advanced
applications of Fracture Mechanics for the study of integrity and durability of materials and structures”, and by the Regione Piemonte to the RE-
FRESCOS project: “Preservation, safeguard and valorisation of masonry decorations in the architectural historical heritage of Piedmont”, are
gratefully acknowledged. The authors wish also to thank Ing. F. Accornero for his active collaboration in the execution of the compression tests
under circumferential displacement control.
References
1. Hudson JA, Brown ET, Fairhurst C (1972) Shape of the complete stress–strain curve for rock. In: Cording EJ (ed) Stability of rock slopes
(Proceedings of the 13th symposium on rock mechanics). American Society of Civil Engineers, New York, pp 773–795
2. Kotsovos MD (1983) Effect of testing technique on the post-ultimate behaviour of concrete in compression. Mater Struct 16:3–12
3. van Mier JGM (1984) Strain softening of concrete under multiaxial compression. Ph.D. thesis. Eindhoven University of Technology, The
Netherlands
4. Jansen DC, Shah SP (1997) Effect of length on compressive strain softening of concrete. J Eng Mech 123:25–35
5. Carpinteri A, Corrado M, Mancini G, Paggi M (2009) The overlapping crack model for uniaxial and eccentric concrete compression tests. Mag
Concr Res 61:745–757
6. Carpinteri A, Corrado M, Paggi M (2011) An analytical model based on strain localization for the study of size-scale and slenderness effects in
uniaxial compression tests. Strain 47:351–362
7. Ohtsu M (1996) The history and development of acoustic emission in concrete engineering. Mag Concr Res 48:321–330
8. Carpinteri A, Lacidogna G, Pugno N (2007) Structural damage diagnosis and life-time assessment by acoustic emission monitoring. Eng Fract
Mech 74:273–289
9. Carpinteri A, Lacidogna G, Niccolini G, Puzzi S (2008) Critical defect size distributions in concrete structures detected by the acoustic
emission technique. Meccanica 43:349–363
10. Carpinteri A, Lacidogna G, Puzzi S (2008) Prediction of cracking evolution in full scale structures by the b-value analysis and Yule statistics.
Phys Mesomech 11:260–271
11. Weiss J, Marsan D (2003) Three-dimensional mapping of dislocation avalanches: clustering and space/time coupling. Science 299:89–92
12. Carpinteri A, Lacidogna G, Manuello A (2011) The b-value analysis for the stability investigation of the ancient Athena temple in Syracuse.
Strain 47:e243–e253
13. Ferrara G, Gobbi ME (1995) Strain softening of concrete under compression. Report to RILEM Committee 148 SCC, ENEL-CRIS Laboratory,
Milano
14. Carpinteri A, Cardone F, Lacidogna G (2009) Piezonuclear neutrons from brittle fracture: early results of mechanical compression tests. Strain
45:332–339
15. Cardone F, Carpinteri A, Lacidogna G (2009) Piezonuclear neutrons from fracturing of inert solids. Phys Lett A 373:4158–4163
16. Carpinteri A, Cardone F, Lacidogna G (2010) Energy emissions from failure phenomena: mechanical, electromagnetic, nuclear. Exp Mech
50:1235–1243
17. Carpinteri A, Borla O, Lacidogna G, Manuello A (2010) Neutron emissions in brittle rocks during compression tests: monotonic vs cyclic
loading. Phys Mesomech 13:268–274
18. Carpinteri A, Lacidogna G, Manuello A, Borla O (2011) Energy emissions from brittle fracture: neutron measurements and geological
evidences of piezonuclear reactions. Strength Fract Complex 7:13–31
19. Carpinteri A, Lacidogna G, Borla O, Manuello A, Niccolini G (2012) Electromagnetic and neutron emissions from brittle rocks failure:
experimental evidence and geological implications. Sadhana 37:1–20
20. Carpinteri A, Lacidogna G, Manuello A, Borla O (2012) Piezonuclear fission reactions in rocks: evidences from microchemical analysis,
neutron emission, and geological transformation. Rock Mech Rock Eng 45:445–459. doi:10.1007/s00603-011-0217-7
21. Scott DF, Williams TJ, Knoll SJ (2004) Investigation of electromagnetic emissions in a deep underground mine. In: Proceedings of the 23rd
international conference on ground control in mining. Morgantown, 3–5 Aug 2004, pp 125–132
22. Lacidogna G, Carpinteri A, Manuello A, Durin G, Schiavi A, Niccolini G, Agosto A (2011) Acoustic and electromagnetic emissions as
precursor phenomena in failure processes. Strain 47(suppl 2):144–152
23. O’Keefe SG, Thiel DV (1995) A mechanism for the production of electromagnetic radiation during fracture of brittle materials. Phys Earth
Planet Inter 89:127–135
24. Frid V, Rabinovitch A, Bahat D (2003) Fracture induced electromagnetic radiation. J Phys D 36:1620–1628
25. Carpinteri A, Chiodoni A, Manuello A, Sandrone R (2011) Compositional and microchemical evidence of piezonuclear fission reactions in
rock specimens subjected to compression tests. Strain 47(suppl 2):282–292
26. Carpinteri A, Manuello A (2011) Geomechanical and geochemical evidence of piezonuclear fission reactions in the Earth’s crust. Strain
47(suppl2):267–281
Chapter 36
Detailed Measurements of Thread Deformation and Failure
in Thin Walled Aluminum Alloy Joints
Bonnie R. Antoun, Spencer Grange, Gerald W. Wellman, and J. Franklin Dempsey
Abstract This paper describes the development and implementation of the experimental design, apparatus and measure-
ment methods for quantifying the deformation of threads during loading to failure. A linear thread geometry is used to allow
direct optical and contacting measurements of key displacements along the loading axis and across the threaded engagement
section. Full field optical measurements of thread pairs are collected for post-processing using digital image correlation
methods. Thread geometry parameters and material pairings are studied.
Keywords Threads • Fastener • Aluminum • Thin-wall • Thread deformation • Thread failure
Shallow engagement threads in large diameter, thin walled assemblies present failure modes considerably different than
typical fastener threads. Developing an understanding of the damage and failure mechanics of these types of thread is
critically needed for many industries and applications. Observations to date have been limited to post-failure since threads
are not visible once they are assembled or engaged. In order to enable direct observations, a linear thread experimental
capability with supporting diagnostic methods was developed and is used in this study for in-situ measurements of
deformation and failure of individual and multiple threads during mechanical loading. Details of the experimental apparatus
and measurements are described herein.
A photograph of the linear thread experimental setup and apparatus is shown in Fig. 36.1. Fixtures, hardware and
specimens were designed to provide ideal, axial loading of threads. Thread engagement and overall thread stiffness are
completely controllable with features shown in Fig. 36.2 which enable study of the effect of these critical parameters on
thread deformation and failure. Figure 36.3 shows a photograph of linear thread specimen pairs. In this study, ACME type
threads with 10 threads per inch and either 2 mm (0.080 in.) or 1 mm (0.040 in.) thread height were tested. Three material
pairing combinations of 7050-T74 and 7075-T651 aluminum alloys were used: 7075-T74 paired with 7075-T651, 7075-
T651 paired with itself and 7050-T74 paired with itself. Each experiment uses one center specimen and two side specimens.
The side specimens are machined as one double-wide specimen and then divided or split along the length only after the
threads are machined to provide an exact matched set of side specimens. Note that the center specimen has four sides of
threads so it can be used twice, by rotating the center specimen 90 before the next experiment. Both side specimens can be
used twice also, by testing the lower four threads in the first experiment, then machining those four threads away and
engaging the top four threads on the next experiment. This saves material and specimen costs and allows for nearly identical
repeat experiments with one specimen set or pairing, if desired.
B.R. Antoun (*)

Sandia National Laboratories, Livermore, CA 94551-0969, USA
e-mail: brantou@sandia.gov
S. Grange • G.W. Wellman • J.F. Dempsey
Sandia National Laboratories, Albuquerque, NM 87185, USA
314 B.R. Antoun et al.
Fig. 36.1 Experimental

apparatus for linear
thread tests
Fig. 36.2 Internal detail

of fixtures, load cells (top
and bottom) for transverse
load measurement across
threads and a Belleville
washer arrangement for
mechanically imposing
structural compliance
Fig. 36.3 Linear thread

specimens, one center and
two side specimens per test
36 Detailed Measurements of Thread Deformation and Failure in Thin Walled Aluminum. . . 315

linear thread engagement
and transverse extensometry
Applied axial load and total axial displacement are measured directly within the MTS 100 Kip load frame using the
system load cell and LVDT. Relative axial displacement between the thread pairs are measured using a laser extensometer,
which is visible behind the fixtures in Fig. 36.1. Transverse displacements of the threads, at both the top or first thread and the
bottom or fourth thread are measured with contacting extensometers attached to the front face of the fixtures, as shown in
Fig. 36.4. Transverse loads across the threaded pairs are made at the top and bottom of the engaged threads with load washers
internal to the fixtures, shown in Fig. 36.2. All measurements were continually monitored during loading and resulting
deformation to failure. Also during loading, high resolution images of the threaded section were captured for detailed
observation and analysis of damage and failure mechanisms. Two cameras are used in each experiment, one video camera on
the back side of Fig. 36.1 and a Phantom V12.1 high speed camera on the front side. High speed images were post-processed
with DIC methods to obtain full-field measurements of deformation at the individual thread level.
Linear thread specimens were loaded in axial displacement control at a rate of 0.001 in/s to failure. Figure 36.5 shows a
photograph of linear thread failure of the center threads in shear. Figures 36.6 and 36.7 show measurements during a typical
experiment. The threaded geometries and applied mechanical loading and associated boundary conditions were modeled
with finite element methods. Details of the modeling, methods and analyses can be found in Ref. [1]. Additional results for
the aluminum alloy linear thread experiments, including DIC analyses will be published in a subsequent paper.
316 B.R. Antoun et al.

linear thread pair after failure.
In this case, the center threads
failed by shear and remain
engaged with the side threads
Test 9A
7050-T74 Center, 7075-T651 Sides
0.04" Thread Height
20,000 0
Top
18,000
16,000
-500
14,000
Bottom transverse load (lb)
axial load (lb)
12,000 Axial Load

Top Transverse Load
Bottom Transverse Load
10,000 -1000
8,000
6,000 Axial
-1500
4,000
2,000
0 -2000
0.000 0.002 0.004 0.006 0.008 0.010
displacement (in)
Fig. 36.6 Measured axial and transverse loads during typical experiment
36 Detailed Measurements of Thread Deformation and Failure in Thin Walled Aluminum. . . 317
Test 9A
7050-T74 Center, 7075-T651 Sides
0.04" Thread Height
0.010
0.009
Transverse Displacement
Top extensometer
0.008 Bottom extensometer
displacement across threads (in)
0.007
Bottom
0.006
0.005 Top
0.004
0.003
0.002
0.001
0.000
0.000 0.002 0.004 0.006 0.008 0.010
axial displacement (in)
Fig. 36.7 Measured transverse displacements during typical experiment
Acknowledgements Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly
owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy’s National Nuclear Security Administration under contract
DE-AC04-94AL85000.
Reference
1. Dempsey JF, Antoun BR, Romero VJ, Wellman GW, Scherzinger WM, Grange S (2012) Temperature dependent ductile material model failure
constitutive modeling with validation experiments. In: Proceedings of the 2012 SEM annual conference and exposition on experimental and
applied mechanics, Costa Mesa, 11–14 June 2012
Chapter 37
Simulation of Interference Fitted Joint Strength
as Used in Hot Rolling Work Rolls
Martin McMillan, Julian Booker, and David Smith
Abstract Composite work rolls consisting of a hard, wear resistant cast iron hub interference fitted to a tough, fracture
resistant alloy steel shaft are typically used in the hot rolling of sections (e.g. rails, I-beams). The hubs are susceptible to
fracture (due to their brittleness) and radial/axial slipping (due to overloading or improper design) when in service. The shaft
may be reworked several times and new hubs fitted in order to extend the life, although an optimised solution is sought to
identify the interference conditions at the interface that will minimise the probability of failure of both failure modes. This
paper looks at the simulation and optimisation of interference fitted work rolls to try and minimise the torsional failure mode
observed in practice under steady state as well as more extreme working conditions that work rolls are exposed to in service.
Finite element models are validated experimentally using scaled laboratory tests. These models are then used to investigate
the probability of failure during a normal working cycle and during extreme working conditions like cobbles (mill stalls).
A number of key parameters are identified which support the effective design of interference fitted joints generally and under
competing failure conditions.
Keywords Interference fit • Hot rolling • Composite work roll • Engineering design • Coefficient of friction • Finite element
modelling
37.1 Introduction
Hot rolling of steel began towards the end of the seventeenth century, but developed into the huge industry it is today
following the invention of the Bessemer process in 1855 and the need for rolled steel products like rails, rods and wires
during the industrial revolution [1]. Hot rolling can be used to roll many types of steel and with the invention of harder and
tougher roll materials, harder stock materials such as stainless steel can be effectively rolled. Advances in rolling techniques
have led to the demand for better quality products with lower production costs. As a consequence, composite rolls, which
have a hard outer surface and a tough, reusable interior, have been developed. Figure 37.1 shows a typical roll used for
rolling rail sections. The outer hub (mirrored finish) can be removed at the end of its working cycle and replaced up to six
times using the same shaft which reduces running costs.
The hub is fitted to the shaft using an interference fit – a mechanical assembly system that can resist relative movement or
transmit torque through the creation of high radial pressures at the interface of its constituent parts. It provides a low cost
joining method and is widely used in industry with applications to cutting tool holders, wheels and bands for railway stock,
turbine disks, rotors for electric motors and for locating ball and roller bearings to shafts and bores. They are also an effective
way of assembling machine elements such as a gear to a shaft or wear resistant rolls to shafts for hot rolling of steel sections.
The underlying principle involves establishing a pressure between the inside diameter of the hub and the outside diameter of
a shaft through an interference in the dimensions at their radial interface, as shown in Fig. 37.2. Commonly, expansion of the
external part by heating, and/or cooling of the shaft is employed, the part located and then the whole joint returned to
M. McMillan • J. Booker (*) • D. Smith

Department of Mechanical Engineering, University of Bristol, Queen’s Building, BS8 1TR, Bristol, UK
e-mail: martin.mcmillan@bristol.ac.uk
320 M. McMillan et al.
Fig. 37.1 Hot rolling composite roll and common rolling loads
Fig. 37.2 Schematic of the interference fitted joint under torsional loading
ambient temperature whereupon the pressure maintains part location. Dismantling can only occur on a limited number of
occasions because of the resulting damage to the component surfaces on removal of the hub.
There are two competing failure modes of the interference fit: fracture and torsional slip. The pressure at the interface creates
tensile hoop stresses in the hub with the highest stress at the inner radius of the hub. The material is very hard to produce a good
surface finish on the finished stock, however, this means it will also fail under brittle fracture. When combining the stresses due
to thermal and mechanical loading in the mill with any manufacturing residual stresses it is very important to choose an
interference fit that will not create excessive stresses in the hub leading to fracture. Torsional slip is also dependent upon the
interference fit chosen as well as the friction between the two parts. If an insufficient interface pressure is chosen then the hub
is likely to slip causing a bad surface finish on the stock, fracturing the sleeve; and in the worst case scenario, damaging the
shaft, which is much more expensive to replace.
37 Simulation of Interference Fitted Joint Strength as Used in Hot Rolling Work Rolls 321
In order to model the in-service behaviour of composite work rolls, predict failure probability and improve failure
mitigation strategies, it is necessary to establish validated modelling procedures. This paper discusses the torsional aspect of
failure mode modelling using a combination of statistical, numerical and experimental methods.
37.2 Modelling Interference Fitted Joints
37.2.1 Analytical Modelling
The early treatment of interference fits was derived by Gabriel Lamé in 1833 [2] by developing the theory for stresses in gun
barrels (thick walled cylinders). The derivations can be found in many engineering text books [3, 4]. The interface pressure,
p for dissimilar hub and shaft materials, as in the case of work rolls, is given by Eq. 37.1 [3]:
d
p¼ h 2 2 i (37.1)
R Ro þR þ n þ R ð1 n Þ
Eh 2 2 h Es s
R o R
Where R is the shaft radius, Ro is the hub radius, Eh and uh are the Young’s Modulus of Elasticity, and Poisson’s ratio of the
hub respectively, Es and us are the Young’s Modulus and Poisson’s ratio of the shaft respectively and d is the interference
between the hub and shaft.
The holding torque, TH determines the torque required for the interference fit to fail in torsion (due to complete slip along
the contact length, L), and can be found using the Coulomb friction equation by considering the normal force acting as a
moment at the interface as shown in Eq. 37.2 [5]:
TH ¼ 2pR2 Lmp (37.2)
Where m is the coefficient of static friction. An assumption made in Eq. 37.2 is that the interface pressure is constant across
the surfaces and remains constant during sliding, and that the friction coefficient is unidirectional and maintains a constant
value at all times.
Previous work [6] has shown that holding torque variability is mainly a function of the interference and the coefficient of
friction. Figure 37.3 shows the sensitivity analysis for a typical shrink-fitted assembly, where all the parameters in Eqs. 37.1
and 37.2 are statistically characterised using a Normal distribution (mean and standard deviation). It may contribute up to
50% of the variability of failure torque (or holding force), and is the main reason why shrink-fit calculations are better
performed probabilistically rather than deterministically [7].
It is necessary to utilise an accurate value of the static coefficient of friction at the shaft/hub interface in design
formulations, however, this parameter is not a constant, with interface pressure, material hardness/elasticity, surface
50
Percentage Contribution
40
to Variance of TH
30
20
10
0
µ Eh Es L Vh Vs Ro R
Parameter
Fig. 37.3 Sensitivity analysis for a typical interference fitted joint

Fig. 37.4 (a) Boundary conditions for torsional loading (hub resists rotation around z axis), (b) Abaqus FEA model showing von Mises stresses
prior to failure of a hub-shaft in torsion
roughness, surface conditions (dry, lubricated, oxidized, etc.), environment (temperature, humidity) and loading rate
affecting the measured value [8]. The pressure and contact area may themselves be affected by geometrical inaccuracies
caused by the chosen manufacturing route to produce the components. The coefficient of friction should therefore be
statistically characterised under realistic service conditions, i.e. at the calculated interface pressure, for similar surface
conditions, material and surface roughness. It is hypothesised that the tests should also be conducted for the particular loading
regime under consideration for the final joint i.e. axial tension, axial compression or torsion.
37.2.2 Finite Element Modelling
For basic interference fit geometries, the finite element results closely resemble the analytical results, the advantage is that
the models can accommodate edge effects and measure microslip along the contact length before total slip occurs. Finite
element models are most useful during the optimisation stages as they allow specific non-uniform geometries and stress
concentrators to be taken into account and allow fine control over various factors which would be difficult to implement
analytically. Thermal and mechanical loads can also be introduced to simulate the rolling process. For this study, finite
element models were created using Abaqus/Standard to predict the holding torque of a number of experimental tests. 2D
axisymmetric models were used to allow for larger meshes and more efficient computing. Figure 37.4 shows the model
geometry and boundary conditions used for torsional modeling.
37.3 Experiments and Work Roll Properties
Interference fit experiments were conducted to validate the modeling using dimensionally scaled down versions of the work
rolls. Two sizes were used: 16 and 50 mm hub bore diameter. Table 37.1 shows the material properties of the hub (SGACI)
and shaft (alloy steel) materials of a work roll. SGACI material was available for experimental testing, however, the alloy
Table 37.1 Material properties

Spheroidal graphite acicular BS970: 070M55 (EN9)
Property cast iron (SGACI) Alloy steel medium carbon steel
Young’s modulus of elasticity (E) 167 GPa 210 GPa 206 GPa
Poisson’s ratio (n) 0.26 0.29 0.29
Tensile strength (UTS) 570 MPa 650 MPa 700 MPa
Yield strength (sY ) – 540 MPa 560 MPa
Linear coefficient of thermal expansion (a) 13.3 106 C1 12 106 C1 14.8 106 C1
Density (r) 7,200 kg m3 7,800 kg m3 7,800 kg m3
Fig. 37.5 Interference fit 16 mm test setup
Table 37.2 Average coefficient of friction at varying interface pressures for various surface treatments
Interface Pressure Statistical Parameters
Surface treatment 40 MPa 60 MPa 80 MPa 100 MPa Mean Standard deviation
1 0.203 0.153 0.193 0.158 0.18 0.07
2 0.268 0.260 0.227 0.235 0.25 0.10
3 0.271 0.284 0.279 0.233 0.27 0.09
Oxidation (water) 0.287 0.524 0.414 – 0.41 0.17
Oxidation (280 C) 0.239 – 0.263 – 0.25 0.09
steel used in the work rolls was not available so BS970: 070M55 (EN9) steel was chosen due to similar carbon content and
material properties. The material properties are given in Table 37.1.
The experiments were tested under a number of loading conditions and failure cycles, but the focus in this paper is the
holding torque at first slip. The specimens were loaded in torsion by a hydraulic test machine until slip occurred. This was
observed visually on the data-logger and by using a dial gauge controlled by a lever arm attached to the underside of the
interference fit. The experimental setup is shown in Fig. 37.5.
As friction is a dominant factor in the modeling of interference fit joint strength, friction tests were carried out to simulate
a selection of possible manufacturing conditions. The test apparatus consisted of a stick of shaft material onto which two test
cubes made of hub material were compressed with a known force using strain gauged bolts to create an interface pressure.
Using a hydraulic universal testing machine, a load was applied to the test stick to produce a characteristic friction plot.
Asperities were introduced to the test samples by coarse grinding with a surface roughness value between 0.4 and 1.6 mm Ra
to simulate the turned surfaces of the shaft and hub. Samples were also treated to create oxidized surfaces either with water
spray, to simulate oxidation caused by water condensation when assembling the interference fit by immersing the shaft in
liquid nitrogen; or by heating to 280 C for 3 h, to simulate assembly by heating the hub. The results of these tests are given
in Table 37.2 and a schematic explaining the asperity orientations is shown in Fig. 37.6.
Force direction
1 2 3
Fig. 37.6 Key to surface treatments used in Table 37.2. Arrows show asperity alignment. Dry samples (1, 2, 3) – surfaces cleaned with acetone
before test. Oxidation water – SGACI test blocks rusted; alloy steel test stick clean. Oxidation 280 C – SGACI test blocks oxidised in oven; alloy
steel test stick clean
100
90
80
Holding Torque, TH (Nm)
70
60 Prediction
50
40
30
20 Experiment
10
0
40 50 60 70 80 90 100
Interface Pressure, p (MPa)
Fig. 37.7 Comparison of experimental and analytical predictions for holding torque, TH against a range of interface pressures, p (joint size
16 mm)
37.4 Results
Figure 37.7 shows the holding torque as a function of interface pressure for the 16 mm test specimens. The dashed line shows
the simulated values based on the friction values measured experimentally.
Figure 37.8 shows the holding torque as a function of interface pressure for the 50 mm test specimens. The dashed line
shows the simulated values based on the friction values measured experimentally.
For the 16 mm tests the linearly fitted line agrees well with the predicted values, however, there is a large spread in the
data. This is most likely due to the tight tolerances required at small scale and measurement errors as the spread in data for
the 50 mm samples is much smaller, while the predictions are also close to the measured values.
Using the results of the friction measurements and validated models a Monte Carlo simulation was used to generate
statistical data so that a reliability study could be carried out using mill torque data provided by TATA Steel UK (Personal
Communication, TATA Steel UK, 04/11/2009). Data was provided for applied torque during normal operating conditions
and applied torque during a cobble, in which the stock jams in the rollers and creates an unexpected torque spike.
The reliability study was conducted at a range of interface pressures including the maximum and minimum pressure used
by TATA Steel UK – 54 and 48 MPa respectively. The results are given in Table 37.3. The reliability study shows that under
normal operating conditions it is unlikely that a sleeve will slip (less than 2 in 100), but during a cobble the sleeve is twice as
likely to slip. Figure 37.9 shows the Monte Carlo plots and overlap between mill torque and predicted holding torque. It is
important to note that this model does not include any mechanical or thermal loadings which will further reduce the ability to
resist torsional loads. Future work will examine these effects.
5000
4500
4000
Holding Torque, TH (Nm)

3500 Experiment
3000
2500
2000 Prediction
1500
1000
500
0
50 55 60 65
Interface Pressure, p (MPa)
Fig. 37.8 Comparison of experimental and analytical predictions for holding torque, TH against a range of interface pressures, p (joint size
50 mm)
Table 37.3 Reliability of work roll during normal mill operation and cobble
Interface pressure (MPa) R, normal mill operation R, cobble
40 0.9807 0.9564
48 0.9816 0.9613
54 0.9812 0.9619
60 0.9826 0.9650
80 0.9838 0.9683
100 0.9846 0.9707
Probability Density
Torque
Normal Mill Operation Cobble 40MPa 48MPa 54MPa 60MPa 80MPa 100MPa
Fig. 37.9 Reliability of interference fitted work rolls at service torque (sensitised data)
37.5 Conclusions
This paper discussed the torsional failure mode of interference fitted joints as used in hot rolling work rolls. Scaled
experimental measurements validated finite element predictions and provided statistical data to carry out Monte Carlo
simulations for full-size work rolls. A reliability study for a work roll under normal and extreme operating conditions
showed that around 2–4 rolls in 100 are likely to slip during operation. This is a conservative estimate as the effects of
thermal and mechanical loading were not included in this study. Further work will be done to examine the effects of thermal
and mechanical loading using the finite element models and a reliability study will be carried out to examine the fracture
failure mode.
References
1. Roberts WL (1983) Hot rolling of steel. CRC Press, New York

2. Lamé G, Clapeyron E (1833) Mémoire Sur L’Equilibre Intérieur des Corps Solides Homogenes. Mem Divers Savans 4:145–149
3. Timoshenko S (1970) Strength of materials. Part II: advanced theory and problems. Van Nostrand Reinhold Company, New York
4. Faupel JH (1964) Engineering design. Wiley, New York
5. ESDU (2007) 68002 – shafts with interference-fit collars, part 1: some factors affecting the design of an joint. ESDU International, London
6. Booker JD, Truman CE, Wittig S, Mohammed Z (2004) A comparison of shrink-fit holding torque using probabilistic, micro-mechanical
and experimental approaches. Proc Inst Mech Eng Manuf B 218:175–187
7. Furman TT (1981) Approximate methods in engineering design. Academic, London
8. Booker JD, Truman CE (2008) A statistical study of the coefficient of friction under different loading regimes. J Phys D: Appl Phys
41(17):174003
Chapter 38
Modelling of Nosing for the Assembly of Aerospace Bearings
Johnpaul Woodhead and Julian Booker
Abstract Nosing is a cold metal-forming process, used during the manufacture of self-lubricating plain spherical aerospace
bearings. This process ensures the outer bearing ring conforms to the shape of the inner ring, with a central composite liner
in-between. Aerospace bearings must be precision engineered due to the large forces and demanding environments they
operate within, yet many companies are still heavily reliant on empirical data and experimental methods, however; FEA
simulation can be used to predict and characterise complex material behaviour. The nosing process is modelled to predict
load history, contact stresses, stress/strain distribution, visual deformation and temperature variation, in order to provide a
better understanding of process parameters, tooling design and the resultant forces which are needed for processing.
Ultimately, a costly and time-consuming experimental process can be replaced with a virtual rapid one, in order to mitigate
defects, secondary processing and low yield rates in production. From initial assessment, a number of objective functions
have been identified including geometric conformity, frictional moment and die wear. Theoretical and analytical
calculations will be validated against experimental data, both statistically and stochastically, to ensure developed methods
are robust. Future emphasis is placed on non-linear friction modelling and strain-rate dependent forming load predictions.
Keywords Double nosing • Modelling • Forming load • Finite element analysis • Bearings
38.1 Introduction
There are three main components to the self-lubricating, plain spherical bearing; a central inner ball with a hole enables the
bearing to be axially supported; an outer sleeve provides a platform for other components in the assembly; and a composite
liner in-between the inner ball and outer sleeve, affixed to the latter with adhesive, provides lubrication and frictional
properties during operation (Fig. 38.1).
The bearing material is Steel alloy; specifically the outer sleeve is AMS 5643, a martensitic, precipitation-hardening
stainless steel. The material is heat treated by increasing the temperature to 482 C 6 for 1 h then allowing it to cool in the
air, providing high strength, hardness and corrosion resistance. The precise material properties of the composite liner is
commercially sensitive information. Applications within the aerospace market are wide and varied including, but not limited
to, bearings for pitch control, dampers, the main central and tail rotary shafts, flap controls, cargo bay doors and under
carriages.
Nosing, otherwise known as ‘tube-end forming’, is a cold-metal forming process used in the manufacture of plain
spherical bearings. In this process the outer sleeve (or bearing outer race) is placed, together with the inner ball and
composite liner, in between two spherical dies (Fig. 38.1a). The upper die displaces along the bearings axis, and the bearing
is subject to compression at the contact interface between the die and the outer sleeve as the force is translated axially. This
causes the outer sleeve to undergo elastic–plastic deformation, until it geometrically conforms to the shape of the inner ball
(Fig. 38.1b).
Various failures can occur during the nosing process, resulting in either extensive re-working of the bearing or scrapping
of the bearing altogether. Incorrect pressure, displacement (shut height of the machine press), or number of cycles (i.e. the
J. Woodhead (*) • J. Booker

Department of Mechanical Engineering, University of Bristol, Queen’s Building, University Walk, Bristol, Somerset BS8 1TH, UK
e-mail: j.woodhead@bristol.ac.uk
328 J. Woodhead and J. Booker
Fig. 38.1 Schematic of the nosing operation, detailing the assembled components, direction of force and final geometric shape of the bearing
(a) Bearing assembly before nosing; (b) Bearing assembly after nosing
Fig. 38.2 Main failure modes produced during the nosing process, resulting in extensive re-working of the bearing or the scrapping of the bearing
altogether
number of consecutive times the press will lower) are just a few contributing factors, and can produce any combination of the
failure modes below (Fig. 38.2) which were identified with the industrial collaborator.
In this paper, numerical modelling for the prediction of forming load during the nosing process will be explored using
analytical and virtual methods, for a high-yield production bearing. Four analytical and two virtual models will be presented.
Models will be validated against experimental data recorded from a force transducer on the companies’ press machine used to
perform the nosing operation. The results of those models are presented, including load history, contact stresses, stress/strain
distributions, visual deformations and temperature variations. They are all analysed in order to provide a better understanding
38 Modelling of Nosing for the Assembly of Aerospace Bearings 329
of process parameters, tooling design and the resultant forces which are needed for processing. Load history has been selected
for analysis to investigate the performance aspect of nosing, test the capability of FE models and explore its limitations. In
addition, the forming load can be linked to some of the failure modes described in Fig. 38.2, notably ‘open mouth’, ‘church
window’, ‘saturnisation’ and ‘ball crushing’. Ultimately key variables are identified to improve those models.
38.2 Nosing Modelling Approaches for Forming Load
Cold-forming processes are generally difficult to model, with many companies favouring an empirical, trial-and-error
approach. The lack of appreciation of the interaction between machine tooling and components in the design stage can result
in lengthy and costly product introductions, with any failures (such as those mentioned above), being addressed retrospec-
tively, however; this experimental approach can be replaced by a quicker, less costly, virtual prototyping method.
38.2.1 Modelling Data
38.2.1.1 Flow Stress Equation
The mechanical properties of the AMS 5643 alloy Steel used to manufacture the outer sleeve, have been experimentally
determined in previous work [1], using compression testing on standard test specimens [2] to obtain the materials true
stress-true strain behaviour. Equation 1 below states the flow stress rule [3] (otherwise known as Holloman’s equation) to
which the constituent variables were determined, using static strain-rates between 0.1 and 0.5 mm/min.
¼ Ken
s (38.1)
is the true stress, K is the strength coefficient, e is the true strain and n is the strain-hardening exponent. The
Where, s
subsequent flow stress rule for AMS 5643 alloy Steel is as follows.
¼ 1665e 0:118 MPa

s (38.2)
The stiffness of the composite liner compared to that of the outer sleeve (race) is very small, and its contribution to the
prediction of forming load has been assumed as such, therefore it has been neglected from all subsequent analysis.
38.2.1.2 Coefficient of Friction
As the presence of friction impedes free movement at the contact interface between the die and the nosed component,
friction can significantly affect the flow and deformation of the material [3], therefore; interpreting the correct value for the
friction coefficient is paramount. In this model Coulomb’s Law is assumed, whereby friction stress is deduced from the
normal contact pressure by the use of a coefficient [4–6] as follows.
tf ¼ mtn (38.3)
Where, tf is the friction stress, tn is the normal contact pressure, m is the coefficient of friction. A friction coefficient of 0.15
was selected for this modelling, being a typical value for cold-worked alloy Steels [6, 7].
38.2.2 Analytical Methods
The assumptions for the validity of these analytical methods are as follows.
• The stress in the radial direction is small;
• The blank material is rigid and linear hardening;
• Tresca failure criterion applies;

• Coulomb friction model applies;
• The blank conforms fully to the die.
As nosing occurs on both ends of the sleeve during the operation, any value obtained through calculation must be
multiplied by two to give the correct load. In the following equations, Af is the final cross-sectional area after compression, sf
is the flow stress, sx is stress in the meridional direction, sy is the hoop stress, y and dy are the initial and change in angle
respectively, m is the coefficient of friction, r is the die radius, t0 and tD are the initial sleeve thickness and thickness at the
extremity (end) of the sleeve respectively, E is the plastic modulus, s_ is the tool velocity, R and R0 are the sleeves current and
initial mid-thickness radii respectively, and Y is the yield strength.
38.2.2.1 Work Formula
The work formula, albeit a simple expression, provides a good estimation of the order of magnitude of the working load, and
is estimated by considering that plastic flow occurs at yield stress [8]. Neglecting frictional and mechanical constraints, the
working load expression is as follows.
P ¼ A f sf (38.4)
38.2.2.2 Lower and Upper Bound Methods
These methods aim to identify an upper and lower forming limit (a maximum and minimum possible load), within which the
correct experimental load fits between [6].
The lower bound (minimum) expression is as follows.
P ¼ 2pR0 t0 ðsx ÞE (38.5)
The upper bound (maximum) expression is as follows.

" # Z yD
2pR0 t20 Y s_ sy dR
Ps_ ¼ þ 2pr0 t0 s_ dy
4 r2 t0
yE R dy
" Z Z !#
yD yD
s sy cos y
þ 2pr0 t0 mr s_ x t dy þ dy (38.6)
yE r2 yE R
38.2.2.3 Approximate Method
The following method combines elements of Manabe and Nishimura’s [5] work with Reid and Harrigan’s [6]. The
approximate method expression is as follows.

K tD þ t0 1þn
P ¼ 2pR0 t0 ðsx ÞE þ (38.7)
2 4r
Fig. 38.3 Axisymmetric (left) and 3D (right) models of the plain spherical bearing created in Abaqus Explicit, detailing reference points allocated
to the constituent parts of the model assembly
38.2.3 Finite Element Analysis
Abaqus Explicit was used to model both axisymmetric and 3-dimentional models of the bearing (Fig. 38.3). The geometry of
the dies, inner ball, outer sleeve and composite liner were recreated from the original manufacturer’s engineering drawings.
In both models, although the inner ball and composite liner were modelled individually, they have been merged into one
single part. The dies were modelled, in both cases, as rigid parts as is common practice due to the dies being of a (much)
stiffer material than any of the bearings constituent parts.
In order to limit computational time to a reasonable duration, the 3-dimensional model mesh was, generally, not as dense
as the axisymmetric model, however; the outer sleeve was very densely meshed in both cases to mitigate inconsistencies, and
the dies were manually seeded around the contact zones, but coarsely in all other unused areas on the 3D model.
As part of the validation process to verify that both models give similar results, the Von Mises stresses distribution in the
outer sleeve was plotted in the Abaqus results manager module. Figure 38.4 shows that the two models give similar results.
38.2.4 Load History Prediction
The following graph (Fig. 38.5) plots the load history prediction results from all four analytical methods (work formula,
upper bound, lower bound and approximate method), and the two finite element analysis models (axisymmetric and
3-dimentional), together with experimental results provided by the company.
The experimental results, indicated by the solid black line, is an average taken from a small number of trials under
‘steady-state’ forming conditions. All analytical models show a similar trend, with the exception of the lower bound method
which greatly under-estimates the load. The closest fit is the FEA models, which correlate well with each other throughout,
and both predict the forming load accurately at maximum displacement, however; the FEA models deviate from the
experimental data in the latter half of the operation.
Fig. 38.4 Von Mises stresses distribution and centre-line marking on both the axisymmetric (left), and 3D (centre) models, with stress distribution
range in MPa (right)
Fig. 38.5 Load history prediction results for all analytical methods and FEA models, together with the experimental ‘steady-state’ forming data
38.3 Material Non-linearity
38.3.1 Friction Coefficient Calibration
The following graph (Fig. 38.6) plots the load history prediction results for the FEA models, again together with the
experimental data, for different friction coefficients. Upon closer inspection it can be seen that the experimental data
fluctuates between a friction coefficient of 0.15 and 0.10, even drawing closer to 0.05 over the last few millimetres of
displacement before increasing to 0.15 again. Maximum contact pressure at the contact interface between the die and outer
sleeve has been analysed, and shows that the coefficient of friction decreases as pressure increases [9].
Understanding the interaction between the die and the bearing during the forming operation is critical, or more
specifically, understanding the influence of friction, however; friction between the die and work piece is a complex
phenomenon [10]. Variations in geometric tolerance and surface roughness can be particularly influential on frictional
properties, especially at the start of the forming process [11] as indicated by the higher coefficient of friction at the start of
the operation. Thus a non-linear, or variable, friction model is required when modelling this type of process.
Fig. 38.6 Load history prediction results for axisymmetric FEA model for multiple friction coefficients, together with the experimental ‘steady-
state’ forming data
Fig. 38.7 A frequency chart of key variables identified in a recent literature review of over 60 papers in the field of cold metal forming
This is highlighted in a recent literature review, whereby over 60 research papers in the field of cold metal forming
operations was reviewed. Fig. 38.7 shows a frequency chart of the top ten key variables identified by other researchers to
improve FE models used to predict process variables. Accurately calibrating the coefficient of friction can be clearly seen, in
comparison to other parameters.
The coefficient of friction will be calibrated using FE models, as it is not be possible to experimentally determine through
direct measurement due to the high contact pressures (in excess of 1,000 MPa) involved in the nosing process.
Lubrication is the second most frequent, however; the industrial collaborators use only one lubricant consistently
throughout the process.
Thickness distribution is the third most frequent, and can also be analysed using FE models.
Furthermore, the data used to populate the models was experimentally derived from compression testing to ASTM
standards, at strain-rates in the static region of 0.1–0.5 mm/min, however; in reality, the manufacturers perform this
operation at 30 mm/min, which is well within the dynamic region, and hence the material may exhibit other non-linear
Fig. 38.8 An engineering

sketch (left) detailing the
specimen dimensions,
together with a photograph
(right) of a prepared specimen
coated with MolyKote
lubricant and placed centrally
on the lower platen of the
press machine, with the
camera lens in the background
characteristics at higher strain-rates [12]. The importance of taking into account the effects of strain-rate is also high-lighted
in the literature review. The next chapter details experimentation performed to confirm or reject this hypothesis.
38.3.2 Strain-Rate Dependency
Uni-axial compression testing was performed to determine the material characteristics of AMS 5643 and S80 alloy Steels
under static and dynamic strain-rates. ASTM International standards [2] state that compressive material testing must
conform to the following criteria:
• Both ends of specimen shall bear on blocks with flat surfaces and parallel within 0.0002 in./in. (m/m);
• Blocks shall be made of hard material (tungsten carbide recommended for Steel);
• Alignment device used, unless machine is specifically designed for axial alignment;
• Mechanical or electromechanical devices can be used for measuring strain (electrical resistance strain gauges can also be
used);
• Recommended that short specimens of solid cylindrical blocks are used for compression tests of bearing materials;
• The ends of a specimen shall be flat and parallel within 0.0005 in./in (mm/mm) and perpendicular to the lateral surfaces to
within 3’ of arc;
• Specimen strain rate of 0.005 mm/min, or 0.003 mm/min if material is strain-rate sensitive.
The only deviation from ASTM standards was the strain-rates selected, however; as the objective of the experiment was
to determine the effects of strain-rate, this deviation was unavoidable.
ASTM standards recommend using specimens with a length to diameter ratio (L/D) of 0.8, therefore; a specimen of 10
12:5 6 0 mm (Fig. 38.8) was selected. The specimen must be coated with white undercoat, and speckled with black paint to
optimise the tracking of targets by the video extensometry equipment.
MolyKote lubricant was applied to all contact surfaces in order to reduce friction between the platens and extreme
surfaces of the specimens, and promote homogenous deformation throughout the specimen during compression.
An Instron 600kN NX press machine was selected for tests (Fig. 38.9), being the most adequate machine able to deliver a
minimum of two times (security margin) the maximum expected calculated load. BlueHill 2 software was used in
conjunction with the machine, along with a Video Extensometer (comprising of a camera and computer running the required
Video Gauge software).
Several longitudinal and circumferential strains were measured on each specimen to ensure the consistency of
measurements and obtain average values. Constant strain-rates of 0.6, 6, and 60 mm/min were selected, increasing in
factors of 10 to ensure a static to dynamic strain-rate range was tested. A total of 60 specimens were tested, 30 specimens of
each material, 10 specimens at each strain-rate. From the output data, true stress-true strain curves were plotted for both
materials and Poisson’s Ratio, Young’s Modulus, Yield Stress, Strength Coefficient, and Strain-hardening Exponent were
calculated.
Fig. 38.9 A photograph

of the Instron press machine,
with the specimen placed
centrally on the lower platen,
illuminated by a spot-lamp
to ensure no interference
from shadows cast onto
the specimen. The camera,
pointed directly at the
specimen, is fitted with
an adjustable focus lens
38.4 Results
The calculated material properties obtained for each sample have been analysed statistically within each strain-rate range, in
order to present the data stochastically in the following graphs (Figs. 38.10 and 38.11).
38.5 Discussion
A range of statistical tests including chi-squared, f-test and t-tests were performed on the data, but only the most significant
are discussed here. When a test required the comparison between mean values, all strain-rates were compared with one
another (i.e. static to quasi-static, static to dynamic, and quasi-static to dynamic).
Statistical analysis indicated that all tests were fair, and that there is not a large difference between population means and
sample means, therefore; it can be concluded that normal distributions for all strain-rates are not statistically different, and
all give comparable results. F-tests indicated a combination of equal and unequal variances, therefore; the appropriate t-test
(for small samples) was chosen to analyse each individual material property.
The appropriate t-tests indicated that:
– There is no statistical difference between strain-rates for Poisson’s Ratio and Young’s Modulus for either AMS 5643
or S80;
– There is a weak statistical difference between strain-rates for Strength Coefficient (weaker for S80);
Fig. 38.10 Strength 2

coefficient as a function
of strain-rate for AMS 5643
and S80 alloy Steels including
Strength Coeffcient (GPa)

error bars for one standard 1.8
deviation
1.6
1.4
AMS 5643 Strength Coeffcient
S80 Strength Coeffcient
1.2
0.6 6.0 60.0
Strain-rate (mm/min)
Fig. 38.11 A graph to show 0.25

strain-hardening exponent as
a function of strain-rate for
AMS 5643 and S80 alloy
Strain-hardening Exponent
0.2
Steels including error bars
for one standard deviation
0.15
0.1
0.05
AMS 5643 Strain-hardening Exponent
S80 Strain-hardening Exponent
0
0.6 6.0 60.0
Strain-rate (mm/min)
– There is a strong statistical difference between strain-rates for Yield Strength and Strain-hardening Exponent (weaker
for S80).
The Young’s Modulus is largely consistent across all strain-rates, averaging at 200.5 and 158.6 GPa for AMS 5643 and
S80 respectively. Poisson’s Ratio for is very consistent across all strain-rates, averaging at 0.31 and 0.29 for AMS 5643
and S80 respectively. The yield strength and strength coefficient display an increase from static to dynamic strain-rate for
both materials. The strain-hardening exponent displays a decrease from static to dynamic strain-rate for both materials.
38.6 Conclusions and Future Work
Both the AMS 5643 and S80 alloy Steels exhibited similar behaviour in all calculated material properties. This further
validates that tests were fair and comparable between strain-rates, but also between materials. These findings correlate with
work by Bayraktar et al. and Cai et al. [12, 13], in that strain-rate dependency must be taken into account.
Future work will entail incorporating this non-linear, strain-rate dependent material behaviour, into existing FE models to
reduce the percentage error within load history predictions.
In addition, the individual components of the bearing can undergo 28 processing stages prior to nosing, including various
machining operations to their surfaces (i.e. turning, grinding, honing, and polishing). Little is understood as to the stresses
that are potentially induced during these stages, and the effect this has on the bearing components during the nosing process.
As a result, the bearings undergo a stress relief stage after being nosed, however; the extent to which the stress is relieved,
if at all, is unknown. Therefore, future work will also focus on measuring the size and distribution of residual stresses in the
bearing outer sleeve using neutron diffraction, and has been confirmed at the major European research centre, SINQ (The
Swiss Spallation Neutron Source).
References
1. Orsolini A (2010) Major project task 1, Compressive testing, M.Sc. advanced mechanical engineering, in mechanical engineering. University
of Bristol, Bristol
2. American Society for Testing Materials (2009) Committee E-28 on Mechanical Testing. Subcommittee E28.04 on Uniaxial Testing ASTM
International, w.a.o., Standard test methods of compression testing of metallic materials at room temperature. ASTM International
3. Kalpakjian S, Schmid SR (1997) Manufacturing processes for engineering materials Third Ed., Addison-Wesley Longman Inc, California
4. Yuung-Hwa L (2005) Study of preform and loading rate in the tube nosing process by spherical die. Comput Methods Appl Mech Eng 194
(25–26):2839–2858
5. Manabe K-I, Nishimura H (1984) Nosing of thin-walled tubes by circular curved dies. J Mech Work Technol 10(3):287–298
6. Reid SR, Harrigan JJ (1998) Transient effects in the quasi-static and dynamic internal inversion and nosing of metal tubes. Int J Mech Sci
40(2–3):263–280
7. Hensel A (1990) Technologie der Metallformung Eisen- und Nichteisenwerkstoffe. 1990: Deutscher Verlag f€ ur Grundstoffindustrie
8. Rowe GW (1977) Principles of industrial metalworking processes. [New] ed. Edward Arnold, London. xxii, 407 p
9. Booker JD, Truman CE (2008) A statistical study of the coefficient of friction under different loading regimes. J Phys D Appl Phys
41(17):1–12
10. Cora ÖN, Akk€ok M, Darendeliler H (2008) Modelling of variable friction in cold forging. Proc Inst Mech Eng Part J: J Eng Tribol 222
(7):899–908
11. Mahrenholtz O et al (2000) Investigation of the influence of surface roughness on metal forming processes. Mech Res Commun 27(4):393–402
12. Bayraktar E, Levaillant C, Altintas S (1993) Strain rate and temperature effect on the deformation behavior of the original hadfield steel. J Phys
IV France 03(C7):C7-61–C7-66
13. Cai M-C et al (2010) A constitutive description of the strain rate and temperature effects on the mechanical behavior of materials. Mech Mater
42(8):774–781
Chapter 39
Multi-axial Quasi-static Strength of a Clinched Sheet Metal Assembly
S. Coppieters, P. Lava, H. Sol, P. Van Houtte, and D. Debruyne
Abstract Conventionally, the mechanical behavior of clinched connections is investigated by a single shear lap test and/or
a pull-out test. In practical applications, however, there is a strong probability that a combination of shear and pull-out
components is exerted on the clinched joint. This paper deals with the development of an Arcan-like device which enables to
introduce various shear/tensile ratio’s in a clinched sheet metal assembly. Since clinch forming locally results in a complex
region, a good knowledge of the plastic material properties is of the utmost importance to perform a sufficiently accurate
simulation. The identification of post-necking hardening behavior of sheet metal is complex and there is no general
agreement on how to perform this. In this study, different methods are used to identify the post-uniform hardening behavior
of DC05. The impact of these different procedures on the simulation of the multi-axial loading behavior of a so-called non-
cutting single-stroke round clinched connection is investigated. The experimental results are used to check the validity of
numerical models to predict the strength under multi-axial loading.
Keywords Sheet metal • Clinching • Mechanical strength
39.1 Clinch Forming
In industry thin sheet metal is formed into many different semi-finished parts or finished products. Usually, the semi-finished
thin-walled parts need to be assembled or integrated into other products. This urges the need for joining methods suitable for
connecting thin-walled metal products. These joining technologies for sheet metal can be divided into four major categories:
1. Welding
2. Joining with the aid of consumables
3. Adhesive bonding
4. Joining by forming
The first group hosts the well-known and widely spread resistance spot welded joint, see Fig. 39.1a. The latter
technique uses electrodes which exert pressure on the sheets and generate a concentrated welding current necessary to
S. Coppieters (*) • P. Lava

Department of Mechanical Engineering, Catholic University College Ghent, Association K.U. Leuven,
Gebroeders Desmetstraat 1, B-9000 Gent, Belgium
e-mail: sam.coppieters@kahosl.be; pascal.lava@kahosl.be
H. Sol
Department of MeMC, Vrije Universiteit Brussel, Pleinlaan 2, Brussel B-1050, Belgium
P. Van Houtte
Department of MTM, Katholieke Universiteit Leuven, Kasteelpark Arenberg 44, Leuven (Heverlee) B-3001, Belgium
D. Debruyne
Department of Mechanical Engineering, Catholic University College Ghent, Association K.U. Leuven,
Gebroeders Desmetstraat 1, B-9000 Gent, Belgium
Department of MTM, Katholieke Universiteit Leuven, Kasteelpark Arenberg 44, Leuven (Heverlee) B-3001, Belgium
340 S. Coppieters et al.
Fig. 39.1 Three joining categories for sheet metal. (a) Spot welded joint (From [1]). (b) Self-piercing riveted joint (From [2]). (c) Clinched joint
melt the sheet metal and form a welded joint. The second group includes all techniques which need an additional element
to establish the joining. This group can be subdivided in two groups depending on the presence of a forming aspect.
The most obvious example in this group which does not entail any forming of the sheet metal is a bolted connection.
On the contrary, other techniques such as self-piercing riveting (SPR, see Fig. 39.1b) require severe local plastic
deformation of the base material. The third group adhesive bonding covers all joining techniques which connect materials
by an adhesive. The fourth group joining by forming consists of all joining methods which solely rely on forming of the
base material to establish a connection. The focus of this work is on clinch forming (see Fig. 39.1c), also referred to as
press-joining or clinching. This technique relies on joining by forming and has a pronounced local character similar to
spot welded joints and SPR joints. The basic principle of clinch forming processes is to create an interlock between the
combining thin metal parts with the aid of relatively simple tools like a punch, a blank holder and a die. The punch locally
pushes metal into the die and, depending on the shape of these clinching tools, the resulting metal flow targets the creation
of a mechanical interlock. The type of clinched joint under consideration in this paper is shown in Fig. 39.1c. Such a joint
is produced by the so called Non-Cutting Single-Stroke (NCSS) clinch forming process. Non-cutting, highlights the fact
that during forming the sheets are not separated. Single-Stroke indicates that the joint is formed during one flowing punch
movement. The assessment of the strength of clinched connections is usually based on two basic loading cases: a single
shear lap test [3] and a pull-out test [4]. In practice, however, the joint is usually exposed to a combination of shear and
pull-out components. This paper presents an Arcan-like device which enables to exert quasi-static multi-axial loading
conditions on a clinched connection. In addition, the capability of finite element techniques to reproduce these experi-
mental tests is investigated.
39.2 Modified Arcan Device
Figure 39.2a shows the basics of the experimental set up which consists of two separate disk halves which enable clamping
of a specimen. Once the device is provided with a specimen, the two disk halves are connected via this assembled specimen.
The next step is then to mount it into a tensile machine: one of the disk halves is connected to a moving cross-head while the
other is fixed, see Fig. 39.2b. The design shown in Fig. 39.2a allows for the following angular positions: (a ¼ 0∘, 15∘,
30∘, 45∘, 60∘, 75∘, 90∘) of which pull-out (0∘), a ¼ 45∘ and shear loading (a ¼ 90∘) are depicted in the figure. The Arcan
device shown in Fig. 39.2a has a multiple-pin clevis which allows to control the rotation of the disks. To prevent a
potentially damaging side load on the load cell, Pin 3 is omitted in this study. As a consequence, each disk halve can rotate
in its clevis and this behavior cannot be measured with a classical extensometer. Therefore, the setup shown in Fig. 39.2b is
augmented with the Digital Image Correlation (DIC) technique. The Charged-Coupled Device (CCD) cameras (denoted 5, in
Fig. 39.2b) take synchronized images of the Arcan fixtures in the reference and operating state of the Arcan device. The
post-processing of these images is performed in our in-house DIC platform MatchID [5]. As such, the displacement fields at
the surface of the Arcan fixtures are retrieved for different loading steps during testing.
39.3 Material Identification
The mechanical behavior of a clinched connection depends on the final geometry of the joint and the material state after
forming. This can be investigated with the aid of finite element techniques. Clinch forming involves severe plastic
deformation of the sheet metal and a good knowledge of the material behavior is of utmost importance to perform a
39 Multi-axial Quasi-static Strength of a Clinched Sheet Metal Assembly 341
Fig. 39.2 The modified Arcan device. (a) Operating principle. (b) The device mounted in a regular tensile machine. 1. Cross-head 2. Load cell
3. Lamp 4. Upper half disk 5. Camera’s
σ22
MLUT
SSF
σ11
UTT
Fig. 39.3 Identification of the initial yield locus through the MLUT and the UTT
sufficiently accurate simulation. The plastic material behavior of sheet metal is conventionally identified through a standard
tensile test. Unfortunately, this test is limited by the plastic instability which is often countered by extrapolation of the pre-
necking hardening behavior using a predefined hardening law. The latter is in the remainder of this work referred to as
Extrapolation method (EM). The EM procedure may give very different results depending on the hardening law fitted to the
experimental pre-necking data. In addition, in practice it might sometimes be impossible to locally remove a tensile
specimen. To cope with the limitations of the standard tensile test, two alternative material tests are used, namely a
Multi-Layered Upsetting Test (MLUT) and the Post-Necking Tensile Experiment (PNTE).Details on the methods used in
this paper can be found in [6, 7]. The clinched connection used in this study is composed of two DC05 sheet metal sheets.
The plastic material behavior of this material is investigated through the EM, the MLUT and the PNTE. Figure 39.3
schematically shows the initial yielding during the uniaxial tensile test (UTT) and the MLUT. The red circle is obtained from
a uniaxial tensile test and the red triangle stems from the MLUT. Assume that both tests are used to identify the von Mises
criterion. It is clear that then two different initial yield loci are obtained, see Fig. 39.3. Indeed, the true initial yield locus
cannot be described by the von Mises locus and stress state fitting (SSF) of an anisotropic yield locus seems mandatory.
Unfortunately, the red triangle from Fig. 39.3 cannot be accurately determined through a MLUT. On the contrary, the
evolution of this part of the yield surface can be accurately determined. Figure 39.4 shows the results obtained for DC05 and
it can be inferred that the flow stress under biaxial tension is larger than under uniaxial conditions. The clinching tools
typically exert very high pressure on the sheets while sliding lengths remain small. Unfortunately, there is no sheet metal
friction test available which matches the specific conditions as encountered in clinch forming. In this study, the frictional
Fig. 39.4 Identified flow 600

curves for DC05. These data
sets are accompanied with
friction values during clinch
forming: EM-Swift 400
(mt¼0.167), MLUT-Swift
(mt¼0.126), MLUT-Voce
σeq
EM-Swift
(mt¼0.198), PNTE-Swift PNTE-Swift
(mt¼0.137) MLUT-Swift
200 MLUT-Voce
0
0 0.2 0.4 0.6 0.8 1
εeqpl
RP1
KC
R=
11
2m α=45° KC
m
x
Z
Y
KC
KC
RP2
Fig. 39.5 FEA model of the Arcan test
conditions during clinch forming are identified based on the correlation of accurate measurements with numerical responses.
More details on this procedure can be found in [8]. For each identified flow curve shown in Fig. 39.4, the frictional
conditions (mi and mt are the inter-sheet friction and the friction between the tools and the sheets, respectively) during clinch
forming are identified. In the next section the reproduction of the multi-axial loading strength using FEA is under
consideration.
39.4 Results
From here it is assumed that the results of the FE simulation of the clinch forming are available and assembled into an Arcan
specimen which is then used to simulate the mechanical behavior under multi-axial loading. Figure 39.5 shows the
numerical model of the modified Arcan test. The Arcan fixtures are assumed to be infinitely rigid and in the model they
are replaced by kinematic coupling (KC). The principal material directions in the models which take planar anisotropy into
account are oriented with respect to the local coordinate frame of the sheets. The rolling direction in these simulations
corresponds with the Y-direction. Three angular positions (a ¼ 0∘, a ¼ 45∘ and a ¼ 90∘) are considered in this paper.
These test are conducted with a constant cross-head speed of 1 mm/min. The results of these tests are shown in Fig. 39.6.
This figure shows that the shear/tensile ratio influences the maximum strength and the ductility of the connection. The shear/
39 Multi-axial Quasi-static Strength of a Clinched Sheet Metal Assembly 343
a b
1000
α=0 1000 α=45˚

750
Force (N)
Force (N)
500 V DIC V DIC
EM-Swift EM-Swift
MLUT-Swift 500 MLUT-Swift
MLUT-Voce MLUT-Voce
PNTE-Swift PNTE-Swift
250
0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 0.5 1 1.5 2 2.5 3 3.5
V (mm) V (mm)
c 3000
V DIC
EM-Swift
MLUT-Swift
MLUT-Voce
2000 α=90˚ PNTE-Swift
Force (N)
1000
0
0 0.5 1 1.5 2 2.5 3 3.5
V (mm)
Fig. 39.6 Experimental and numerical results of the Arcan test on a DC05-DC05 clinched joint. Three angular positions are shown: (a) a ¼ 0∘
(b) a ¼ 45∘ (c) a ¼ 90∘
tensile ratio has also an impact on the joint’s stiffness. If the contribution of the pull-out component increases, the
surrounding sheet material is deformed rather than the joint itself. There is evidence to suggest that when shear loading
starts to dominate, the plastic deformation is concentrated in the vicinity of the joint resulting in a much stiffer behavior. For
pull-out testing (a ¼ 0∘) the frictional dissipation is larger than the plastic deformation of the joint. As such, this loading
case will allow the assessment of the identified frictional behavior between the sheets. On the contrary, plastic dissipation
predominates in shear loading (a ¼ 90∘) which in turn allows the assessment of the identified material behavior. Fig-
ure 39.6a, b show the results of the numerical models along with the experimental results for (a ¼ 0∘) and (a ¼ 45∘),
respectively. Figure 39.6c shows the results for (a ¼ 90∘).The numerical model shown in Fig. 39.5 is used to predict the
joint strength under multi-axial loading. This numerical model is fed with the flow curves and frictional conditions from
Fig. 39.4. Through the MLUT two commonly used hardening laws are identified, namely the Swift (seq ¼ K(e0 + epleq)n)
and the Voce (seq ¼ C(1 me keeqpl )). For (a ¼ 0∘, 45∘) all the numerical models yield very similar results. Indeed, for
these loading cases the plastic deformation of the joint is relatively weak and instead the surrounding sheet material is
deformed. For the angular position a ¼ 90∘ shown in Fig. 39.6c, the plastic deformation is large and here the material model
gains significance. A clear distinction between the models MLUT-Swift and MLUT-Voce can be made. From Fig. 39.6 it
can be concluded that the Swift hardening model identified through the MLUT yields a better result than the Voce model.
In the case of substantial shear loading, it appears that for a ¼ 90∘ the stiffness of the joint is overestimated. This is caused
by an inaccurate measurement of the relative displacement V between the sheets in shear loading. The measurement of V is
conducted on the rigid disks by DIC, and, as a consequence, these small displacements in the beginning of the test cannot be
accurately measured. This can be remedied by using an extensometer in the close vicinity of the joint. Figure 39.7 shows the
predicted deformation mode of the joint in shear loading along with two experimentally observed failure modes. If failure is
Fig. 39.7 Experimental and numerical deformation mode of the clinched joint after shear loading (a ¼ 90∘)
completely dominated by plastic deformation (left panel of Fig. 39.7), a good agreement between the experiment and the
simulation is found. If ductile fracture in the neck of the joint is observed, however, this cannot be reproduced with
the current model since it does not incorporate a damage model. Figure 39.6c clearly shows that the numerical models
underestimate the ductility of the joint in shear loading. This can be attributed to the absence of a damage model which post-
pones the complete separation of the sheets. Nevertheless, since the crack propagation is initiated beyond the maximum
force, the current model is still capable of predicting the maximum strength of the joint.
39.5 Conclusion
This contribution presents results on the mechanical behavior of clinched connections under multi-axial loading conditions.
In order to exert different pull-out/shear ratios on a clinched assembly, an Arcan-like setup is designed and equipped with the
digital image correlation technique to measure all rigid body motions of the fixtures. Three different multi-axial experiments
on a DC05-DC05 clinched connection are used to assess the validate of FE models and the identified input data of these
models. It is shown that the maximum strength of the connection increases when the magnitude of the shear component
increases. In general, the presented FE model is able to reproduce the experimental observations such as the force-
displacement curve and the deformation mode of the joint. Since the numerical model does not incorporate a damage
model, fracture in the neck could not be reproduced. It is shown that the alternative material tests yield good results.
Moreover, it is expected that these tests yield a hardening behavior which is less dependent on the hardening law which needs
to be chosen a priori. The significance of the material model increases when there is sufficient plastic deformation of the joint.
The current set up allows to investigate crack propagation and to develop failure criteria for clinched joints under multi-
axial loading. In addition, the experiments carry information which can be used to develop simplified joint models. Indeed, it
would be unrealistic to model thousands of clinched connections in a structure because of the exuberant computational cost.
In this case, the complex joint model can be replaced by an equivalent, computationally efficient model [9]. The mechanical
response of clinched joints under multi-axial loading can used to tune and improve the accuracy of such simplified models.
References
1. Martin O, De Tiedra P, Lopez M, San-Juan M, Carcia C, Martin F, Blanco Y (2009) Quality prediction of resistance spot welding of 304
austenitic stainless steel. Mater Des 30:68–77
2. Abe Y, Kato T (2009) Self-piercing rivetting of high tensile strength steel and aluminum alloy sheets using conventional rivet and die. J Mater
Process Technol 209:3914–3922
3. Saberi S, Enzinger N, Vallant R, Cerjak H, Hinterdorfer J, Rauch R (2008) Influence of plastic anisotropy on the mechanical behaviour of
clinched joint of different coated thin steel. Int J Mater Form Suppl 1:273–276
4. Joma M, Billardon R (2007) 10th ESAFORM CONFERENCE on material forming, AIP Conference Proceedings 907:169–174
5. www.matchID.org
6. Coppieters S, Lava P, Sol H, Van Bael A, Van Houtte P, Debruyne D (2010) Determination of the flow stress and contact friction of sheet metal
in a multi-layered upsetting test. J Mater Process Technol 210:1290–1296
7. Coppieters S, Cooreman S, Sol H, Van Houtte P, Debruyne D (2011) Identification of the post-necking hardening behaviour of sheet metal by
comparison of the internal and external work in the necking zone. J Mater Process Technol 211:545–552
8. Coppieters S, Lava P, Sol H, Van Houtte P, Debruyne D (2011) Identification of post-necking hardening behaviour of sheet metal: a practical
application to clinch forming. Key Eng Mater 473:251–258
9. Langrand B, Combescure A (2004) Non-linear and failure behaviour of spotwelds: a “global” finite element and experiments in pure and mixed
modes I/II. Int J Solid Struct 41:6631–6646
Chapter 40
Development of a Test Simulator to Perform Optimized
Experiment Design
Marco Rossi, Fabrice Pierron, Lava Pascal, and Debruyne Dimitri
Abstract The use of full-field measurements in the identification of material properties is currently widespread due to
advancements in the measurement techniques and in the identification methods. In this context, an interesting area which has
not been systematically studied yet is the test design, i.e. the use of advanced design procedures to individuate optimized
specimen geometries which give the best results in terms of quality of identification. Since it is very time and money
consuming performing a test optimization with real specimens, the idea here is to build up a simulator able to correctly
reproduce the images generated during an experimental acquisition starting from numerical models. A full-field measure-
ment technique is then applied to the synthetic images and an identification procedure is performed. The quality of the
identification can be checked looking at the parameters input in the numerical models, thus the geometry of the specimen can
be iteratively varied to improve the effectiveness of the test.
Keywords Experiment design • Synthetic images • Inverse method
40.1 Introduction
The use of simulated experiments is a very useful tool to study and enhance the experimental testing. The advantage is that
the parameters to be identified are known a priori therefore it is possible to quantitatively evaluate the effectiveness of the
studied identification procedure. Another important aspect which influences the identification is the adopted measurement
technique and the error introduced by it, for instance the effect of the spatial resolution when high deformation gradients
have to be measured. Rossi and Pierron [1] described in depth how simulated experiments can be used to study the
identification process, the grid method was used as full-field measurement technique [2] and the virtual fields method
(VFM) [3] as identification procedure. In this work the idea is to develop a similar study using the DIC [4] as full-field
measurement technique. The experimental test which was simulated is the unnotched Iosipescu test [5].
M. Rossi (*)
Università Politecnica delle Marche, via brecce bianche, 60100 Ancona, Italy
e-mail: marco.rossi@ensam.eu; m.rossi@univpm.it
F. Pierron
e-mail: fabrice.pierron@ensam.eu
L. Pascal • D. Dimitri
Department of Mechanical Engineering, Catholic University College Ghent, Association K.U. Leuven, Gebroeders Desmetstraat 1,
B-9000 Ghent, Belgium
e-mail: pascal.lava@kahosl.be; dimitri.debruyne@kahosl.be
346 M. Rossi et al.
40.2 Image Deformation
In the previous paper [1], using the grid method as full-field optical technique, the adopted reference image was a cross-hatch
grid which can be mathematically described. In the present example, instead, the reference image is a random white light
speckle coming from a real image acquisition, see Fig. 40.1. This image has to be deformed according to the displacement
field computed with a FE simulation. Such operation is more complicated because the grey level of the reference image has
to be deformed according to the FE deformation pattern. Such distortion is performed numerically using interpolation
functions, therefore in order to avoid numerical artefacts, the error induced by this deformation process has to be negligible
compared to error due to the noise and the reconstruction error induced by the DIC measurement. This aspect has to be well
investigated in further works.
40.3 Results and Discussions
Starting from the synthetic images, the strain maps are computed using a DIC software [6] and the parameters are identified
using the VFM. A cost function is then defined as:
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u !
N e uX
X ðkÞ 2
1 u Qij
FðL; aÞ ¼ t wij 1 ð0Þ with ij ¼ ½xx; yy; xy; ss (40.1)
N e k¼1 ij Qij
Qij(0) are the reference parameters to be identified, Qij(k) are the parameters identified at the kth simulated test and wij is a
weighting parameter that can be varied to give more or less importance to a particular stiffness component during the
optimization process. L and a are the length of the specimen and the fibre orientation. The used material is a Glass/Epoxy
UD. The geometry and constitutive parameters are the same described in Ref. [1].
In Figs. 40.2 and 40.3 the obtained results are illustrated in terms of cost function maps. In Fig. 40.2 the cost function
includes the identification of the four parameters (wij ¼ [0.25 0.25 0.25 0.25]) while in Fig. 40.3 each map represents the
cost function related to a single stiffness parameter.
The obtained results confirm what already found in Ref. [1]. Using too long or too short specimen lengths deteriorates the
identification because of the low spatial resolution, instead the best fibre angle orientation is around 40 .
Fig. 40.1 Speckle pattern

obtained from a real
acquisition
40 Development of a Test Simulator to Perform Optimized Experiment Design 347
Fig. 40.2 Evaluated cost Cost function

function for different value 90
of length and fibre angle
75 0.48
60 0.3833
Angle
45 0.2867
30 0.19
15 0.0933
0
10 20 30 40 50 60
Length
Fig. 40.3 Influence of Parameter Qxx wij=[1 0 0 0] Parameter Qxx wij=[0 1 0 0]

the different parameters 90 90
in the cost function 75 0.25 75 0.25
60 0.2 60 0.2
Angle
Angle
45 0.15 45 0.15
30 0.1 30 0.1
15 0.05 15 0.05
0 0
10 20 30 40 50 60 10 20 30 40 50 60
Length Length
Parameter Qxx wij=[0 0 1 0] Parameter Qxx wij=[0 0 0 1]
90 90
0.14
0.8
75 75
0.64 0.11
60 60
Angle
Angle
45 0.48 45 0.08
30 0.32 30 0.05
15 0.16 15
0.02
0 0
10 20 30 40 50 60 10 20 30 40 50 60
Length Length
40.4 Conclusions
This paper presents the first results obtained with a test simulator able to reproduce numerically the Iosipescu test. The
procedure is able to create synthetic images which simulate the images acquired during an actual test using DIC. The images
are used to evaluate the best test configuration for an unnotched Iosipescu test. The results are consistent with what already
found in a previous papers however more controls and validations, especially on the generation of the synthetic images has to
be conducted to validate the method.
References
1. Rossi M, Pierron F (2012) On the use of simulated experiments in designing test for material characterization from full-field measurement. Int
J Solids Struct 49:420–435
2. Surrel Y (1996) Design of algorithms for phase measurements by use of phase-stepping. Appl Opt 35:51–60
3. Grédiac M, Pierron F, Avril S, Toussaint E (2006) The virtual fields method for extracting constitutive parameters from full-field measurement:
a review. Strain 42:233–253
4. Sutton MA, Orteu JJ, Schreier HW (2009) Image correlation for shape, motion and deformation measurements. Springer, New York
5. Pierron F, Vert G, Burguete R, Avril S, Rotinat R, Wisnom M (2007) Identification of the orthotropic elastic stiffnesses of composites with the
virtual fields method: sensitivity study and experimental validation. Strain 43:250–259
6. http://www.matchid.org
Chapter 41
Simplified Calculation Method of the Torsion Effect
on the Seismic Behaviour of Timber Building
Thanh Kien Vu, Eric Fournely, Rostand Moutou Pitti, and Abdelhamid Bouchair
Abstract An analytical approach is developed to take into account the vertical axis torsion phenomenon which is more
difficult to evaluate than bending modes for seismic situations. A typological analysis of current buildings is done and an
original classification, based on the distribution of the bracing implantations and the degree of symmetry is proposed.
A parametric study is conducted with the method of multi-2D combination to analyze the influence of different bracing
configurations on the sensitivity of the analyzed structure to the torsion phenomena. The simplified method is checked using
more sophisticated methods according to the approaches proposed in Eurocode 8. The results showed that the proposed
approach gives simple, accurate and safe calculation method to take account of the torsion effect on timber buildings.
All these results lead to the creation of a database that can serve as reference for the analysis of the semi-rigidity diaphragm
influence or the real non-linearity bracings on the load distributions under seismic event.
Keywords Seismic action • Torsion effects • Bracing distribution
41.1 Introduction
Regarding post earthquake observations, a lot of damage on buildings is due to in plan or (and) elevation irregularities. Many
cases can be found for example in Kobe Izmit or Kashiwazaki post earthquake reports from AFPS (French association for
earthquake engineering) mission [1–3]. Usual design methods are able to fittingly describe the transversal behavior of a
building, but the torsion phenomena are much more difficult to be acceptably taken into account [4]. Two origins of torsion
are clearly identified, on one hand, a structural eccentricity due to irregular positions of the bracing or distribution of mass,
and on the other hand, an accidental origin due to the uncertainties on the effective position place and rigidity of the bracings,
the distribution of permanent and quasi-permanent loads and the real seismic action [5, 6].
Many design methods can be used in the actual codes, linear-elastic analysis such as classical modal response spectrum
analysis (including or not a linearization of behavior factor) or simplified method as lateral force method, or nonlinear methods
such as pushover or non-linear time history analysis. Figure 41.1 presents these methods in relation with Eurocode 8
prescriptions (seismic European code [7]). To take into account the torsion, several ways can be used, for example Eurocode
8 proposes a simplified wide-ranging method or a method based on additional moment calculated for each level of mass. These
torsion methods can be used as well as for linear and non-linear analysis. The application of these methods, global analysis and
torsion closely depends on the dimensions of modeling (2 2D or 3 2D, or 3D); Eurocode 8 requirements are reported on
Table 41.1. Grey lines in Table 41.1 correspond to the configurations adapted to the approach developed in this paper.
Accurate methods exist to take into account the influence of torsion effects; Chateauneuf et al. [8] carried study with an
artificial neural network coupled with Monte Carlo simulations. The approach used in this paper is based on a simple
approach with a less wide field than wide-ranging Eurocode proposal; it is based on Priestey and Paulay developments for
T.K. Vu • E. Fournely • R. Moutou Pitti (*) • A. Bouchair

350 T.K. Vu et al.
Fig. 41.1 Possible choices for global analysis in Eurocode 8 requirements
Table 41.1 Models, global analysis and behavior factor compatibilities in Eurocode 8 for linear analysis
All types of buildings Buildings with H<10 m + conditions
Regularity Allowed simplification Behaviour factor Regularity Allowed simplification Behaviour factor
Plan Elevation Model Linear-elastic For linear analysis Plan Elevation Model Linear-elastic For linear analysis
Yes Yes Planar Lateral force Reference value Yes Yes Planar Lateral force Reference value
Yes No Planar Modal Decreased value Yes No Planar Modal Decreased value
No Yes Spatial Lateral force Reference value No Yes Planar Lateral force Reference value
No No Spatial Modal Decreased value No No Spatial Modal Decreased value
reinforced concrete with vertical continuous bracings [7, 9]. This method is adapted to timber structures and an application is
carried out in order to build a data base able to propose a general coefficient of torsion incidence based only on a geometric
frame considering the positions of the bracing.
41.2 Description of the Multi 2D Approach
This approach is based on an initial combination of X and Y direction of seismic action. The structure is considered as a set
of “lateral load flexible elements” without interruption from the foundation to the top of the building and punctually
connected by horizontal diaphragms stiffer and stronger than structural walls.
41.2.1 Continuous Bracing Requirements
The structural walls can be composed by vertical bracings, shear walls, portal frames. . . The vertical continuity is assumed
between each level and an effective anchorage is considered at the base of the structural wall. These assumptions correspond
to design code requirements; adapted detailing must be implemented. The application of this study is based on shear walls,
but the method could be applied to other types of structural walls. In the same way, the study is applied to vertical regular
buildings for which simplified lateral analysis is adequate to predict the lateral deflexion of the bracing elements.
Nevertheless, this approach can be used with a modal spectrum analysis. An example of diaphragm wall composition and
anchorage solutions are presented on Fig. 41.2 [10].
As example, the shear wall can be composed with an OSB panel (thickness of 13 mm) nailed (diameter of 2.8 mm,
fu ¼ 600 N/mm2) on a 147 47 mm2 framing. The basic fastener spacing is s ¼ 75 mm and the framing elements have a
design compressive strength parallel to grain of 24 N/mm2. In internal environment the seismic strength of this shear wall
(height, h ¼ 2,800 mm, length ¼ 1,100 mm) is equal to 10.6 kN. Figure 41.3 shows the evolutions of the strength and the
stiffness of this type of shear walls versus their length according to Eurocode 5 (timber structure European code) [11].
41 Simplified Calculation Method of the Torsion Effect on the Seismic. . . 351
Fig. 41.2 Composition of typical shear wall and detailing on anchorage to avoid overturning and sliding
Fig. 41.3 Mechanical characteristics of typical shear walls used in this study
Fig. 41.4 Example of timber building with typical position of bracing (partition for office buildings) [12]
41.2.2 Combination of 2D Actions and Horizontal Diaphragm Hypothesis
To be able to combine the seismic actions from two distinct directions, these directions must correspond to two effective
direction of bracing, as illustrated in Fig. 41.4. The combination is based on linearized behaviour and classical Newmark
assumptions. In order to justify these hypotheses roof and floor components have to be much stiffer than walls. This
hypothesis can be easily reached with reinforced concrete slabs, but it is more difficult to verify such condition for timber
floors. APA guides [13] give some elements about the evaluation of the horizontal diaphragms behaviour. Fuentes did
recently comparison between such evaluation and finite element modelling results [14]. In a first approach, and regarding
Fuentes results, for dwellings or office buildings the ratio between floor area and opening for vertical circulation is high
352 T.K. Vu et al.
Fig. 41.5 Flowchart of torsion calculation respecting Eurocode 8 requirements
enough to be able to check the stiffness condition. Figure 41.4 gives an example of the possible shear wall implantation on
the directions X and Y. Thus, for each direction five zones are defined: A, BC and D. Zones A are close to the façades. Zone
D is central and zones BC are between the zones A and D. Figure 41.4 also exhibits a “reference rectangle” in which the plan
of the building floor can be inserted. The slenderness of this rectangle is limited to two this study and the ratio between each
lack area and reference rectangle is limited to 10%.
41.2.3 Multi 2D Approach and Accidental Eccentricity Effects
On the base of such plan, structural eccentricity between the centre of mass and the centre of torsion can be calculated.
Combinations of seismic action, of geometrical accidental eccentricities, of actions in seismic situation are synthesized on
the flowchart in Fig. 41.5. This approach was presented by Priestley and Paulay for reinforced concrete [9]. Figure 41.5
presents the adaptation of this approach using the Eurocode requirements.
In this flowchart, k represents the in plan rigidity of a bracing element, K, the rigidity of the whole bracing in x or y
direction and Krz the rigidity of the totality of x and y bracings regarding the behaviour in torsion. In plan position of bracings
are represented by the x and y values of the coordinate of their geometrical centre reported to the centre of torsion (xT, yT) and
the radius of torsion are represented by r. The structural eccentricity is called e0 and the accidental one, ea. The shear force,
the bending moment and the moment of torsion are respectively called V, M and C for each floor and MT on the base of the
building. Vi,x is the shear force at the level i in x direction, Vi,j,x is the shear force for the bracing j at the level i in the x
direction. Vcomb is the combination of the 16 configurations of seismic actions and accidental eccentricities.
41.3 Parametric Position of Bracing
In order to take into account the influence of structural implantation on seismic loadings of the bracing elements, a
parametric study has been carried out. Part of tested configurations is presented in Table 41.2. For each position of bracing
per zone, different distribution of symmetry inside the zone are studied from 1/2–1/2, to 4/5–1/5 in zone A and from 1/2–1/2,
to 1/1–0/1 in other zones; these parameters are only geometrical ones. This way is quite easier to calculate than eurocode
way which can be difficult or impossible to solve. Effectively, torsional radius and radius of gyration can be quite difficult to
determine [15].
In addition to the limits of bracing implantation in one direction (minimum of 20% in zone A), a complementary criterion
is imposed in the other direction (minimum of 100–20% in zone A). Only some of the configurations presented in Table 41.2
are covered by the plan regularity criteria of Eurocode. This last condition is not respected in all cases presented in
Table 41.2. These cases are studied for different slenderness of the “reference rectangle”. Percentages of bracing in Y
direction placed in B position (Table 41.2) are also studied in C and D positions. Thus a total of 32,400 configurations have
been studied.
Table 41.2 Example of parametric position of bracings
Confi X direcon X direcon Confi X direcon X direcon

gura bracing posion bracing posion gura bracing posion bracing posion
in % in % in % in %
ons ons
reference rectangle
100%
Grid to define the zones of 90%
bracing implantaons
70%
55%
45%
30%
A B C D D C B A 10%
0%
0 10% 30% 45% 55% 70% 90% 100%

10% 20% 40% 20% 10%
Zones A BC D BC A
354 T.K. Vu et al.
41.4 Parametric Study Results
For each configuration of bracing, 16 combinations of seismic actions and accidental eccentricities are computed. For each
set of bracing implantation, the maximum value of loading is recorded for each line of bracing. These maxima are plotted on
graphs in Fig. 41.6. The lines of bracing are numbered in x and y direction: 1 and 2 for lines close to façades and 3 and 4 for
BC or D zones. Results are split in regular and irregular cases regarding in plan Eurocode 8 criteria. Even if 1.45 (maximum
value of factor incidence for regular cases) is significantly lower than 1.88 (maximum value for irregular cases), wide
differences are exhibited on regular case graph and on Figs. 41.6 and 41.7. More discriminant criteria must be found.
41.5 Bracing Distribution Classes and Choice of Class Limits
Average values and coefficients of variation are reported in Table 41.3. Slenderness, B–C and D positions of bracing in y
direction, positions for x bracings are single parameters which are not able to find a discriminant criterion even for seven
ranges for positions of X bracing. A combination of parameters as “distribution” is presented in the last line of Table 41.3.
Inside each distribution (I, II, III and IV) the coefficient of variation is lower than the values obtained for single parameters
and the average values obtained for each distribution are sensibly different with only four types of distributions. May be it
could be possible to join III and IV distributions.
The distributions are defined as followed:
– Distribution I, ½–½ perfect symmetry in each zone and each direction
– Distribution II, ⅔–⅓ for one zone in one direction, ½–½ in other zones and direction
– Distribution III, –¼ in one zone except zone A, ½–½ in other zones and direction,
Fig. 41.6 Example of result as torsion incidence for regular and non-regular configurations for X 60–40/Y 80–20 and a rectangle slenderness
of 1.5
Fig. 41.7 Distribution of torsion incidence for: (a) regular configurations and for (b) irregular configurations
Table 41.3 Research of criteria classes to evaluate the influence of torsion

Global average 1,35
Coef. of variation 11,1%
Slenderness 1.0 1.5 2.0
Average per 1,30 1,34 1,38
Coef. of variation 11,0% 11,1% 11,7%
Y positions of bracing B C D
Average per Y position 1,38 1,38 1,37
Coef. of variation 11,7% 10,6% 10,0%
X positions of bracing 60-40 70-30 80-20 50-50 40-60 30-70 20-80
Average per X position 1,32 1,32 1,31 1,34 1,35 1,37 1,38
Coef. of variation 10,6% 10,5% 10,4% 10,8% 10,8% 11,1% 11,5%
Distributions I II III IV
Average per distribution 1,12 1,36 1,44 1,49
Coef. of variation 1,8% 4,5% 3,9% 3,3%
Table 41.4 Impact factor of torsion on the loading of bracing regarding its implantation
Length/width Distributions
Min Max I II III IV
1 2 1.15 1.43 1.56 1.58
Or
Two zones in ⅔–⅓ and two zones in ½–½
– Distribution IV, the four zones in ⅔–⅓
Or
Three zones in ½–½ and 1–0 in zone D.
The ratios, ½–½, ⅔–⅓, –¼ and 1–0 are ratios in symmetry inside a couple of zones; for example, in X direction in the two
zones BC, with 14 m of shear wall in one side and 6 m in the other side, the ratio will be 70–30%, so it could be classed as –¼
in this zone.
In order to take into account the results of this work in a design practice or a verification stage, maximum values of torsion
influence must be applied instead of average value (Table 41.4). These maximum values are presented in Table 41.4.
41.6 Conclusion
This study us to analyse the influence of the in plan distribution of bracing on their loadings in seismic situation. The
calculation is based on a 3D approach as used in New Zealand and also as proposed in one of the global analysis methods of
Eurocode 8. This approach includes the structural and the accidental origin of torsion phenomenon and a data base is build
with more than 30,000 simulations corresponding to various implantations, inside a grid defining the “zones” of the bracing
positions. This study is applied on timber structures with shear walls but the approach can be extended to other types of
bracings and other types of materials. The most influent parameter among the slenderness of the plan, the symmetry of the
bracing, and the implantation zones is a combination of the symmetry and the implantation. This combination is here
illustrated by four “distributions”. The interest of these distributions is their simple geometric character, which can be read
on a plan and the zone grid. With this study, a minimum impact of torsion of 15% is observed even for symmetric
configuration and this torsion influence reach 60% for the lowest symmetry configurations taken into account in the study.
This threshold of 60% can be exceeded for asymmetric configurations out of the limits defined by the distribution IV.
This study has been completed by a finite element study with linear elastic behavior and also with a neural network able to
integrate random values of accidental eccentricity; these numerical approaches are completed by experimental approaches
on reduced models.
356 T.K. Vu et al.
References
1. Mazaheri D, Mouroux P et al (1999) Le séisme du Kocaeli (Izmit, Turquie) 17 août 199, Rapport de la mission AFPS, 132pp
2. Méneroud JP, Betbeder-Matibet J et al (1995) Le séisme de Hyogo-Ken Nambu (Kobé, Japon) 17 janvier 1995. Rapport de la mission AFPS,
173pp
3. Sidaner JF, Fournely E (2007) Le séisme de Chuetsu Oki (Kashiwazaki, Japon) 16 juillet 2007. Rapport de al mission AFPS, 176pp
4. Thanoon WA, Paul DK, Jaafar MS, Trikha DN (2004) Influence of torsion on the inelastic response of three-dimensional r.c; frames. Finite
Elem Anal Des 40(5–6):611–628
5. Stathopoulos KG, Anagnostopoulos SA (2010) Accidental design eccentricity: is it important for the inelastic response of buildings to strong
earthquakes? Soil Dyn Earthq Eng 30(9):782–797
6. Thanh Kien Vu TK (2011) Incidence de la torsion sur la résistance sismique de bâtiments courants avec diaphragmes horizontaux rigides –
Application aux structures en bois. Ph.D. thesis, decembre 2011, 200pp
7. CEN-TC250, Eurocode 8: Calcul des structures pour leur résistance aux séismes – Partie 1: Règles générales, actions sismiques et règles pour
les bâtiments, version EN 1998–1
8. Badaoui M, Chateauneuf A, Fournely E, Bourahla N, Bensaı̈bi M (2012) Evaluation of accidental eccentricity for buildings by artificial neural
networks. Struct Eng Mech 41(4):527–538
9. Paulay T, Priestley MJN (1992) Seismic design of reinforced concrete and masonry buildings – reinforced concrete ductile frames elastic
analysis for lateral forces. New York, pp 168–171
10. Gianquinto M, Fournely E et al (2011) Guide des dispositions constructives parasismiques des ouvrages en acier, béton, bois et maçonnerie
conforme aux Eurocodes, Presse de ponts et chaussées. 385 pp
11. CEN-TC250, Eurocode 5: Conception et calcul des structures en bois – Parie 1.1: généralité – règles communes et règles pour les bâtiments,
2004
12. Xi C (2007) Construction parasismique – règles simples pour les petits bâtiments en ossature bois, Mémoire d’ingénieur Polytech’Clermont-
Ferrand
13. APA (2007) Diaphragms and shear walls – design and construction guide. The Engineered Wood Association, Tacoma, WA
14. Fuentes S, Fournely E, Bouchair A (2011) Rigidité et distribution des efforts dans un plancher-diaphragme en bois, 8ème Colloque national
AFPS, septembre 2011, École des Ponts ParisTech Paris
15. Fardis MN, Tsionis G (2011) Specific rules for design and detailing of concrete buildings-design for DCM and DCH – illustration of elements
design, seismic design of building, Lisbon, February 2011
Chapter 42
Experimental Analysis of Semi-rigidity Joint
in the Standardized Timber Beam
Eric Fournely, Rostand Moutou Pitti, Abdelhamid Bouchair, and Frédéric Dubois
Abstract Truss timber beams, with glued joints and various types of thick glue layers, can exhibit a semi-rigid behaviour in
translation and in rotation. The global behaviour of truss beams depends on the local behaviours of joints. In this study, the
results are obtained from tests on standardized timber beams used in wooden houses. Classical measurement devices, such as
LVDT and load cells, are used in combination with 3D images analysis. A comparison is done between the measurements
given by different experimental techniques. The image analysis results are in good agreement with those of classical
equipments. Thus, global deflexion, axial displacements and relative rotations in joint zones are obtained from two different
techniques. The geometrical configurations of the beams allow reaching the local failure before lateral buckling. The
experimental results are discussed considering load-deflexion curves, and local behaviour of the joints. The behaviour of
the joints is analysed regarding the resistance, the ductility, the failure modes, the cracks evolutions, the rotation and the
axial slip. Physical and mechanical characteristics of used softwood are measured. The mechanical characteristics of
the joints obtained from tests are introduced in a simple beam modelling. The obtained numerical results are discussed
and compared with experimental results.
Keywords Timber structure • Lattice beam • Semi-rigid joints • Experiments • Analytical approach
42.1 Introduction
The standardized lattice timber beams can find many uses in the field of industrial buildings, small and collectives’ houses
[1, 2]. These beams are generally carried out from low-size timber elements, assembled by nails, plates with metal teeth or
adhesives where the efficiency is performed on the basis of the manufacture control and the knowledge of the behaviour of
the joints [3–5]. Thus, the crack failure, the rigidity and the ductility of the connections remain the main parameters
influencing the mechanical behaviour of the beams [6].
The present study is focused on the behaviour of semi-rigid joints of a lattice timber beam used in a standardized
modular system. In the literature, few authors have discussed the influence of the connections in the global behaviour of
the structures and especially timber structures. Among them, Daniunas [7] and Urbonas [8] have studied the behaviour
of the semi-rigid beam-to-beam joints under bending and axial forces. Several methods have been proposed to evaluate
the semi-rigid mechanical performances of moment resisting connections for steel and timber structures with various
E. Fournely • A. Bouchair
CNRS, UMR 6602, Institut Pascal, F-63171 Aubiere, France
R. Moutou Pitti (*)
F. Dubois
358 E. Fournely et al.
types of fasteners [9, 10]. So as so to improve the resistance of semi-rigid connection by nails, Turner et al. [11] have
proposed new tests on adhesives while theoretical approach [12] and finite element analysis have been performed by
Ellegaard [13] and Haller [14]. Also, Stehn and Borjes [15] have studied the influence of the nail ductility on the resistance of a
glulam structure. Silih et al. [16] have examined the influence of the semi-rigid behaviour of the connections in metal-plate-
connected timber trusses. This paper is focused on the semi-rigid behaviour of nailed and glued connections loaded with axial
forces or bending as a part of truss beams global behaviour.
The global behaviour is analyzed according to an experimental study, carried out on a set of beams with various systems
of connections by nails, adhesives or the combination of both. The experimental results illustrate the global behaviour of the
beams and the local behaviour of the assemblies. The glued connection is selected to be able to use timber elements of low
dimensions, which shows a difficulty to be connected by nails [11]. The geometrical characteristics of the specimens of
DOREAN’s beam and the test conditions are detailed. Then the types of wood and nails, the adhesives, the wood moisture
content and wood tests are exposed. Also, the experimental device, the loading history and the instrumentations are
presented. The experimental results in terms of load-deflection curves, resistance/ductility, crack facial aspects and the
different slipping joints are proposed and discussed. The comparisons of the results given by different measurement
techniques used are done. The influence of the semi-rigid character of the axial joints is shown on the deflection of the
central part of the beam and the rotations at its ends. The rotational semi-rigidity is also discussed in order to validate
the manufacturing process as well as the methods of justification with respect to the ultimate limit states and serviceability
limit states.
42.2 Experimental Program
The experimental program concerned full scale tests on eight lattice beams with the same dimensions (Fig. 42.1) where the
variable parameter is the connection technique (nails and various adhesives). The lattice beam specimens represent four
types of joints with two replications for each type. The tests are realized with four-point bending to obtain the global load-
deflection curves of the beam and the local slip in the joint zones. Complementary tests are realized to characterize the basic
wood material (compression tests) and its moisture content.
42.2.1 DOREAN Beam Description and Geometry
The lattice beam (DOREAN®) developed in this study (Fig. 42.1) is an original key of the prefabricated standardized wood
housing system. It leaves large architectural possibilities for the designers. The upper and lower chords are composed of two
wood members and the posts and diagonals are composed of one wood member. The connection between the chords and the
web members is realized by adhesives or nails.
Fig. 42.1 Geometrical description of typical Dorean beam and testing setup
42 Experimental Analysis of Semi-rigidity (GI) Joint in the Standardized Timber Beam 359
42.2.2 Material and Realization
For the glued joint, in addition to the adhesive nature, the determining criteria of the joint performance are the resistance, the
time of catch, the conditioning, the implementation and the fire performance. Thus, eight different adhesives more or less
easy to be included in an industrial process are tested [1]. Many parameters are considered to define the best glue such as
the fire resistance, the mechanical and the ecological characteristics and the manufacturing process. Satisfactory behaviour
of the adhesive-wood interface is obtained with quality control of timber faces, wood moisture content and perpendicular
pressure during polymerisation provided by mechanical fastener such as simple threaded nail. In addition, classical nailed
joints are used as a basis of comparison with the glued joints. The nails (d ¼ 2.7 mm, l ¼ 80 mm) loaded in double shear.
Four sets of two beams were tested; each set corresponds to a specific technique of connection: nails or various adhesives.
For economical reasons, the beams are realized with two types of wood “a” and “b”. “Wood a” is used for vertical bracing
and diagonal struts and “wood b” for the chords. Only “wood a” is of high quality respecting the grade and CE marking
requirements. Moreover during the manufacture of the tested beams, 20 wood samples (30 30 90 mm3) are realized in
order to obtain the axial compression strength of woods.
The average compression strength (MOR) is equal to 48.9 MPa for “wood a” and 38.4 MPa for “wood b”. The coefficient
of variation of the strength is equal to 3.6% for “wood a” and 10.5% for “wood b”. The average moisture content is equal to
11.4% for “wood a” and 12% for “wood b” with very low standard deviation. The apparent Modulus Of Elasticity (MOE) of
these materials show a higher dispersion in comparison with the strengths. The coefficient of variation of this modulus is
equal to 6.2% for “wood a” and 10.9% for “wood b”. The ratio between the average values of MOR of “wood a” and “wood
b” is about 1.25. The same ratio is obtained for the values of MOE. The dispersion of MOE and MOR is quite smaller for
diagonal members “wood a”. Thus, for the calculations, an average value of 12,000 MPa is assumed for the MOE of the
chord and diagonal members (strength graded C30).
42.2.3 Experimental Setup Descriptions
The setup of the tests configurations is presented in Fig. 42.1. The deflection is particularly observed on the nodes C and E to
represent the general behaviour of the beam. The load is applied on the nodes D and F through hydraulic jack and steel
trimmer beam (Fig. 42.2). Kinematic boundary conditions are realized with simple supports in the nodes A and I.
Figures 42.1 and 42.2 show the position of the lateral support system used to prevent the lateral buckling. However, these
lateral supports are not in contact with the specimen and during the test no lateral buckling phenomena was observed. Thus,
the moment of inertia in the transverse direction, given by the double chords is sufficient to guarantee the expected stability
out of plan for this type of beam in bending. Displacements and loads are continuously recorded using LVDT sensors
Fig. 42.2 Experimental setup and some measurement details

(100 mm), extensometers (2 mm), load cells (2 15 kN). The double symmetry of the configuration is controlled
measuring the two loads in the nodes D and F and the displacements on each side of the beam in the nodes C and E.
The selected experimental protocol is based on the standard experiment related to the wood beams deflection tests [3].
During the tests, the loading protocol is applied according to the following steps:
• Preloading of the beam until 4 kN, corresponding to 30% of the estimated failure load Fmax (10 kN for beam 8)
• Constant load during half minute, then unloading to 1 kN (2 kN for beam 8)
• Re-loading until the total collapse of the structure
In order to analyse the kinematic behaviour of lattice beam a non-contact measurements system by video camera is used
(Figs. 42.2 and 42.9). The video targets (called m1, m2,. . ., m11) are glued on the lateral face of the chords and the first
diagonals to evaluate the displacements and rotations on the nodes B and C. These targets present at least two symmetry
axes, and their size is defined to obtain a precision of 1/20,000 of the global length of the analysed picture. Matched cameras
allow determining the 3D displacements of each target. Position of each target is directly measured by the video metric
system at the beginning and during each test. Thus, the relative displacements are calculated on the basis of geometrical
reference (X, Y, Z). These values are used to calculate the axial and the rotational semi-rigid displacement of the diagonal
strut joints.
42.3 Experimental Results
The experimental results are analysed considering the global and local levels. The global measurement concerns the load-
deflection curve of the beam. The local measurement concerns the relative displacement between the members of each
connection. These measurements are performed using the classical transducers (LVDT) and camera exploiting a large
number of targets on the specimen.
42.3.1 Load-Deflection Curves
Based on the complete set of records during the tests, the load–deflection curves (Fig. 42.3) show the global behaviour of the
beams. Until 4 kN, corresponding to a Serviceability Limit State with deflection criteria of l/400, the behaviours of the beams
are quite similar; but larger difference appears for load near the ultimate load. The lower strength is observed for beams with
MSP adhesive. However, this technique of connections performs higher rigidity. The higher resistance is obtained
with SK852 adhesive. The nailed beams show ultimate resistances close to those of the glued beams but without residual
resistance. Besides, the nailed beams exhibit a non linear behaviour from the beginning.
a b
16 16
applied load (kN)
applied load (kN)
14 14
12 12
10 10
8 b1 8 b5
nails ASK
6 b2 6 b6
4 4
b3 b7
SK
2 2 MSP
mid span deflection (mm) b4 852 mid span deflection (mm) b8
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Fig. 42.3 Load-deflection curves. (a) and (b): global curves with different types of glues or nails
Table 42.1 Experimental load-deflection data for the ultimate load

Beam b1 b2 b3 b4 b5 b6 b7 b8
Type of joints Nails Nails SK852 SK852 ASK ASK MSP MSP
E deflection vE (mm) 53.30 73.14 40.07 36.67 52.21 39.65 24.74 19.52
C deflection vC (mm) 34.03 47.41 21.61 24.13 30.09 23.47 13.03 9.89
Jack displacements (mm) 58.86 73.08 38.35 35 50.29 37.16
Left load (kN) 6.36 6.94 7.38 7.61 6.45 5.46 4.56 5.25
Right load (kN) 6.05 7.43 7.66 7.63 6.16 5.02 5.17 5.16
Total load (kN) 12.41 14.37 15.04 15.24 12.61 10.48 9.73 10.41
Fig. 42.4 Local failure in the glued joints
Fig. 42.5 Crack initiated in the nailed joint
Table 42.1 presents a summary of some measured characteristics at the ultimate load corresponding to the maximum load
measured during test. These characteristics are the applied loads on the node F (right load) and node D (left load), the total
load, the displacement of the jack and the displacements of nodes E (mid span) and C. The selected adhesives called ASK,
MSP and SK852 are given in Ref. [1].
42.3.2 Fracture Features in Joint Areas
For the glued beams (Fig. 42.4), a first localized failure occurred in the adhesive of one node (B or C). However, in spite of
this local failure the beam showed a higher resistance before reaching the total collapse. This local failure or yielding in joint
glued area appears in shear in the glued surface without any crack in the chords. The localized failures are observed at the
ends of the first two diagonals BC and CD corresponding to the diagonals undergoing maximum tension and compression.
In the beams connected exclusively by nails (beams 1 and 2), the failure arises in the joint zone of the chords by the brittle
failure of wood in tension perpendicular to grain (Fig. 42.5). Thus, the crack development is quicker than in the glued
adhesive joints. The brittle character of the failure and the non redistribution of the internal forces, after localized damage,
limit the use of this kind of joints. At the end of the test, high value of permanent slip, in translation and rotation between
members, is observed in the joint zone. The final collapses of the beams arises in all cases with high deflection.
a b
16 applied load (kN) 16
applied load (kN)

14 14
12 12
10 10
8 8
beam 1 beam 5
6 6
beam 2 beam 6
4 4
beam 3 beam 7
2 2
tensile joint slip (mm) beam 4 tensile joint slip (mm) beam 8
0 0
0 -1 -2 -3 -4 0 -1 -2 -3 -4
Fig. 42.6 Applied load-axial displacement curves of the joints in tension: whole load–displacement curve
Fig. 42.7 Ratio of the experimental deflections between nodes C and E for a SLS loading
42.3.3 Axial Semi-Rigid Stiffness of the Joints
Figure 42.6 shows the evolution of the applied load versus the axial displacement of the joint between the first diagonal in
tension and the lower chord (BC-ACE). It shows the whole curves of the joint until failure showing the semi-rigid character
of the joint. In fact, the connections represent a large part of the global deflection of the beams. The beams with nailed joints
exhibit flexional rigidity lower than that observed for the glued joints while their resistance is approximately the same. This
lower rigidity appears only for a loading higher than 4 kN, equivalent to a distributed load of 150 daN/m2. The ratio vC/vE of
experimental deflections between node C located at 1/6 of the span (vC) and node E located at the mid span (vE) is presented
in Fig. 42.7. It shows the influence of the rigidity of the connections on the deflection of the beam. It is compared with the
deflection calculated with the hypothesis of pin-ended members.
42.4 Results Analysis
42.4.1 Analytical Approaches
In order to compare the experimental and the theoretical results of lattice beam, a simple beam theory calculation is
conducted. Simple truss hypothesis is assumed, with pined connections between the members of the lattice beam. In order to
describe the axial semi rigidity of the connection at the end of the diagonal struts, short beam elements are inserted at the
extremity of the struts. These elements are characterized by their length, lc, representing the zone of influence of the
connection, the Modulus Of Elasticity of the material E taken equal to that of the diagonal element and the area Ac
representing the variable parameter used to define the value of the semi rigid stiffness of the joint Kc Kc ¼ EA
lc :
c
42.4.2 Comparison of Experimental and Analytical Results
According to the calculations, Fig. 42.8 illustrates the influence of the joint axial semi-rigidity on the global behaviour of the
lattice beam. A ratio of 10 on the joint rigidity can induce a ratio of 7 on the deflection. Figure 42.8 also exhibits the
evolution of the ratio between the deflections of nodes C and E versus the axial semi-rigidity of joints.
Experimental and analytical results are in good agreement when axial stiffness of the connection equal to 6 kN/mm is
considered under the applied load of 4 kN (Fig. 42.8). The analytical model gives a deflection vE ¼ 8.2 mm when the
experimental values of vE vary from 6.6 to 8.8 mm (average ¼ 8.15 mm). The experimental values of the joint stiffness vary
from 5 to 8 kN/mm (average ¼ 6.3 kN/mm).
The largest difference between analytical and experimental results concerns the deflection shape of the beam represented
by the ratio vC/vE. The model predicts a variation of this ratio from 0.505 to 0.525, while the experimental values vary from
0.46 to 0.55. With vC/vE ratio values outside the theoretical interval (0.505–0.525) it is necessary to propose more complete
models taking into account the rotational stiffness of the connections. According to the results provided by simple beam
theory, theoretical rotations of the first and the second lower chord elements (AC and CE) can be obtained by:
vC vE vC
rotAC ¼ and rotCE ¼ (displacements in mm).
630 1120
42.4.3 Video Camera and LVDT Results Comparisons
The comparison between the LVDT and the video measurements is done for the beams b7 and b8 with MSP adhesive.
Figure 42.9b, c shows the positions of video targets and the LVDT sensors. Figure 42.9a shows the evolution of the applied
50 0,53
0,525
40
0,52
measured measured ratio : 0,48
deflection (mm)
joint axial 0,515

30
rigidity
vC/vE
0,51
20 0,505
vE (mm)
v5
vC (mm)
v3 0,5
10 vC// v5
vE
v3
0,495
0 0,49
562,3
1,0
1,8
3,2
5,6
10,0
17,8
31,6
56,2
100,0
177,8
316,2
Fig. 42.8 Sensitivity of Pin ended

the truss beam deflection Measured
members
to joint axial semi-rigidity 4 kN deflection semi-rigidity of connection (kN / mm)
a 12
b
applied load (kN)

10
6
m1-m4 slip on BC direction
4 m6-m10 slip on BC direction c

m7-m10 slip on CD direction
2 LVDT record, equivalent m6-m10
joint slip (mm)

0
0 2 4 6 8 10 12
Fig. 42.9 Load-slip curves at the ends of diagonals BC and CD (beam 7) (a), video targets (b), LVDT equipment (c)
load versus the slip of the joint for the beam 7 as illustration. With video targets, the axial displacements are calculated with
the geometrical displacement, in x and z directions, of targets m1, m4, m6, m10 and m7.
The load-slip curves obtained from video and LVDT measurement in the lower extremity of the diagonal in tension BC
are in the same range. The axial rigidities of the connections, between the diagonals in tension and compression and the
lower chord, are quite close. However, the comparison between the rigidities in the upper and lower extremities of the
diagonal BC exhibits a ratio equal to 2. For the beam 8, with the same type of connection (MSP adhesive) as the beam 7,
the analysis shows a good agreement between the video and LVDT measurements (Fig. 42.9). Video technique gives the
possibility to measure joint displacement on a wider range than LVDT as shown on Fig. 42.9a. The values of joint axial
rigidity for diagonals in tension are quite comparables between the beams 7 and 8. These values are between 4.1 and
10 kN/mm for beam 8 and between 4.8 and 8.3 kN/mm for beam 7. Nevertheless, a larger difference is observed between the
axial rigidities of joints for diagonal in compression 4.3 kN/mm for beam 8 and 1.6 kN/mm for beam 7. Video and LVDT
measurements are in good agreement for these examples, and also on the whole experimentation. However, the values of
axial rigidity between nodes and between compression and tension can be quite different; a ratio higher than 6 can be
observed with the same type of connection.
42.5 Rotational Semi-Rigidity
With three targets of video analysis on each studied member, it is possible to obtain the relative rotation between members.
Figure 42.10 shows the evolution of the applied load versus the rotation of the upper and lower chords and the first diagonal
of the beam 8. The whole curves are presented but the main interesting part in the structural analysis including the rotational
or axial stiffness of the connections is the initial phase. On each graph, the differences between curves indicate the part due to
the bending of the considered member. The curves show that this bending is quite small, especially at the beginning of the
loading. In these curves, “rotat 8-9” represents the rotation of the segment of targets m8–m9.
These experimental rotations can be compared with theoretical values of rotation calculated before. The experimental and
theoretical values of the rotation of the AC element are in good agreement, but the rotation ratios between AC and CE
members are quite different between experimental and theoretical values: 1.08 and 1.92 respectively. This illustrates the fact
that the axial stiffness of connections is not sufficient to model the behaviour of the lattice beams considered in this study.
The local behaviour of the connections, mainly influenced by the relative rotation between members, has an effect on the
bending transfer in the beam nodes. The measured rotations between members are lower than those calculated with the end-
pined model (ratio close to 0.5). Figure 42.11 shows the evolutions of the relative rotations between the diagonal BC and the
upper and lower chords. Although these rotations are small, they can transfer significant bending moment.
a b
12 12
applied load (kN)
applied load (kN)

Rotations of the origin of the lower chord Rotations of the first diagonal
10 10
8 8
rotat 8-9
6 rotat 9-10 6 rotat 4-5
rotat 8-10 rotat 5-6
4 rotat 10-11 4 rotat 4-6
2 2
rotation (rad) rotation (rad)
0 0
0 0,02 0,04 0,06 0,08 0,1 0 0,02 0,04 0,06 0,08 0,1
Fig. 42.10 Load-rotation of the chords and the first diagonal: (a) origin of the lower chord, (b) first diagonal
12
applied load (kN)
y = -2183,1x - 0,0016
R2 = 0,9839
10
y = -2781,9x - 0,1682
8 R2 = 0,9501
6 Relative rotation between

BD chord & BC diagonal
4 AC chord & BC diagonal
Linéaire
linear( regressions
)
2 Linéaire ( )
rotation (rad)
0
0 -0,001 -0,002 -0,003 -0,004 -0,005
Fig. 42.11 Evolution of relative rotation between first diagonal and extremities of chords
42.6 Conclusion
The tests on lattice timber beams (DOREAN®) manufactured with various techniques of connections showed the influence
of the connection on the beam behaviour. In comparison with glued connections, the nailed ones show acceptable strength
and a non linear behaviour limited by the brittle failure of timber in tension perpendicular to grain. The measurement
techniques show the efficiency of the video camera in comparison with classical LVDT. Besides, they allow widest field of
measurements, until the failure, for deflection, joint slips and joint or member rotations. With these techniques the axial
rigidities are measured although a dispersion is observed on the values in different nodes. The study showed the necessity to
take into account the axial rigidity to obtain good accuracy between theory and experimentation. Besides, element rotations
are evaluated for upper and lower chords extremities and also for the first diagonal element. The video data provides end
element rotations, the obtained experimental results are in good agreement with the theoretical results. The results of the
simple simulations illustrate the effects of semi-rigidity even in linear phase of the behaviour. Although this phase represents
a significant part for the behaviour of the beams, it is important to take into account the non linear character in the modelling
as well as the effects of rotation, mainly due to the evolution of the contacts between the connected members. These
preliminary results open the way to develop non linear models completed by tests to provide the data necessary to perform
such sophisticated models.
References
1. Moutou Pitti R, Fournely E, Bouchair H (2010) Comportement d’une poutre treillis en bois avec assemblages semi-rigides. 28ème Rencontres
Universitaires de Génie Civil, AUGC, La Bourboule, 02–04 Juin 2010
2. Komatsu K, Hwang KH, Hosokawa K (2000) Structural performance of glulam semi-rigid portal frames constructed by a hardwood wedge
joint technique I. Lateral shear performance of portal frames subjected to static cyclic loading. Japan Wood Res Soc 46(6):573–580, ISSN:
0021–4795
3. AFNOR, NF EN 408 (2004) Structures en bois, Bois de structure et bois lamellé collé, détermination de certaines propriétés physiques et
mécaniques. p 32
4. Broughton JG, Hutchinson AR (2001) Adhesive systems for structural connections in timber. Int J Adhes Adhes 21:177–186
5. Leichti RJ, Hyde RA, French ML, Camillos SG (2000) The continuum of connection rigidity in timber structures. Wood Fiber Sci 32(1):1–19,
ISSN: 0735–6161
6. AFNOR (2005) Eurocode 5: Conception et clacul des structures en bois – généralités, règles communes et règles pour les bâtiments, NF EN
1995–1.1
7. Daniunas A, Urbonas K (2008) Analysis of the steel frames with the semi-rigid beam-to-beam and beam-to-column knee joints under bending
and axial forces. Eng Struct 30:3114–3118
8. Urbonas K, Dani Unas A (2006) Behaviour of semi-rigid beam-to-beam joints under bending and axial forces. J Construct Steel Res
62:1244–1249
9. Kasai K, Ian Hodgson I, Bieiman D (1998) Rigid-bolted repair method for damaged moment connections. Eng Struct 20:521–532
10. Guan ZW, Rodd PD (2008) Hollow steel dowels – a new application in semi-rigid timber connections. Eng Struct 23:110–119
11. Turner MA, Plaut RH, Dillard DA, Loferski JR, Caudill R (2009) Tests of adhesives to augment nails in wind uplift resistance of roofs. J Struct
Eng 135(1):88–93
12. Kaveh A, Moeaz M (2008) Analysis of frames with semi-rigid joints: a graph-theoretical approach. Eng Struct 28:829–836
13. Ellegaard P (2006) Finite-element modeling of timber joints with punched metal plate fasteners. J Struct Eng 132(3):409–417
14. Haller P (1998) Progress in timber joint development and modelling. In: 5th world conference on timber engineering, Montreux, 17–20 Aug
1998, vol 1, pp 337–344. ISBN: 2-88074-380-X
15. Stehn L, Borjes K (2004) The influence of nail ductility on the load capacity of a glulam truss structure. Eng Struct 26:809–816
16. Silih S, Premrov M, Kravanja S (2005) Optimum design of plane timber trusses considering dual flexibility. Eng Struct 27:145–154
Chapter 43
Nano-Macro Correlation of Nano-Silica Concrete
Joan Schoepfer and Arup Maji
Abstract Concrete mixes using progressively finer size nano-silica particles (7–150 nm) were prepared to study the effect
of nano-size pozzolans (nano-silica). Conventional compression tests demonstrated progressively higher strength with
finer size of the nano-silica down to 12 nm particle size. High-vacuum Scanning Electron Microscope tests revealed that
the nano-silica provide nucleation sites for crystal growth thereby changing the structure of the material at the micron
to sub-micron level. This paper will describe nano-indentation tests that were subsequently performed on the material.
Nano-indentation data was evaluated to determine stiffness, hardness and elastic–plastic work with progressively increasing
indentation depth for each material. Results show variability of nano-scale testing in concrete due to inherent lack of
homogeneity and also demonstrate that nano-scale tests correlate with the macro-scale mechanical response giving
credibility to the hypothesis that alteration of crystal micro-structure (as opposed to void ratio) with nano-silica affects
mechanical response at the micro-scale, which translates to improved macro-scale properties.
Keywords Concrete • Nanoindentation • High-strength
43.1 Introduction
Portland Cement is a complex material comprising of silicate, aluminates, and sulfates. Silicates are generally either
dicalcium silicates, 2CaOSiO2 (usually referred to as C2S) or tricalcium silicates, 3CaOSiO2 (C3S). Addition of water to
cement results in a multitude of reaction products, the most significant of which is Calcium Silicate Hydrate (CSH). CSH
provides strength and bonding. The addition of finer particle sizes of silicon dioxide provide more surface area for reaction
with calcium carbonate producing more CSH in the matrix. This leads to the strength gain observed as silicon dioxide
particle size is reduced in the mixture. Since the resulting CSH is denser than the reactants, a void is left in the space where
the reaction occurred. The finer the silicon dioxide, the smaller and more numerous are the voids created.
Understanding of the natural creation of the nano level void system may lead to the ability to self-assemble concrete
mixtures into a strong, durable matrix. A well disbursed, controlled nano level void system may ultimately lead to a more
ductile concrete as the system allows for more movement and may naturally stop the path of fractures. Small, well
distributed voids allow a path for water and space for water expansion improving the freeze-thaw durability of the concrete.
The addition of silicon dioxide has been shown to increase strength, lower permeability, and improve durability [1–4].
Silicon dioxide reacts with calcium hydroxide, in the presence of heat and water, to form additional CSH. Most silicon
dioxide used in the industry today consists of silicon dioxide particles in the 150 nm range. Material is now available in
particle sizes as low as 7 nm.
J. Schoepfer • A. Maji (*)

MSC01 1070, University of New Mexico, Albuquerque, NM 87131, USA
e-mail: amaji@unm.edu
368 J. Schoepfer and A. Maji
43.2 Testing
Concrete specimens were made with amorphous silica dioxide of particle sizes of 7, 12, 40, 100 and 150 nm. Other
ingredients such as aggregate, water-reducing admixture, fly-ash, cement and water were kept the same for each mix.
ASTM-specified uniaxial compressive tests were conducted for up to 56 days. Figure 43.1 shows the increase in strength
with curing (up to 56 days) and the variation in strength vs. size of nano-particles used. It can be seen that strength increases
as the particle size is reduced to 12 nm, and that trend reverses (strength drops) for the 7 nm size.
SEM Photographs were taken using a JEOL 5800LV scanning electron microscope. The SEM was used in high vacuum
mode at various magnifications up to 60,000. The specimens were approximately 100 ½00 ¼00 slices cut from the
cylinders used for compressive strength testing. The prepared surface for these photos was a cut surface, oriented parallel to
the fracture plane. Specimen prep for this mode involved light sanding of the surface with silica-carbide paper. Grits of 280,
320, 400, and 600 were used sequentially from coarsest to finest to develop a smooth surface. The specimens were then
polished using a rotary lapper to 3 mm. Deionized water was used to facilitate cleaning and remove debris during sanding.
Specimens were then sonically vibrated for 1 min and rinsed again to remove any particles from the preparation. The sample
was then dried under low vacuum pressure for 2 h to remove unbound water. The samples were then coated with 20.3 nm of
gold palladium for protection during high vacuum operations.
In the SEM photo (Fig. 43.2) few if any silicon dioxide particles (spherical) may be found. This suggests that the particles
have chemically reacted and are now part of the matrix. Comparing the two photos in Fig. 43.1 (left and right corresponding
to 12 and 150 n-meter silica particles), it is evident that smaller nano-particles result in more numerous and smaller reaction
sites and end products. Although many researchers have maintained that the increase in strength and lowering of
permeability are a function of pores being filled in by small particles in the mixture [5–7], SEM evidence under this project
demonstrates smaller size of the reaction products as well.
Nanoindentation was performed using a Nanotest 600 machine to assess structural differences in the mixtures at the nano
level. A 10 mm diameter spherical diamond tip was used to make nano sized indentations horizontally into the specimens.
Compressive Strength, MPa

13000 90
Compressive Strength, psi
11500 80
10000 70
8500 60
7000 50
0 50 100 150
Particle Size (nanometers)
Fig. 43.1 Variation of compressive strength with nano-particle size and curing time
Fig. 43.2 SEM photos of hardened concrete (20,000, bar length ¼ 2 mm)
43 Nano-Macro Correlation of Nano-Silica Concrete 369
Fig. 43.3 Load versus time Typical Load vs Time Curve

used for nano-indentation tests 1.8
1.2
Load (mN)
0.6
0
0 200 400 600
Time (sec)
Fig. 43.4 Variation 0.02

of unloading slope from
nano-indentation tests with unloading slope 0.015
particle size (mN/nm)
0.01
0.005
0
0 50 100 150 200
SiO2 Particle size (nm)
Fig. 43.5 Maximum depth max depth

of penetration versus SiO2 500
particle size
max. depth (nm)
400
300
200
100
0
0 20 40 60 80 100 120 140 160
avg SiO2 particle Size (nm)
Between 18 and 24 indentations were made on each specimen. Specimens were loaded at a rate of 0.008 mN/s for 50 s for
each cycle (Fig. 43.3). The load was then held for 20 s and released at a rate of 0.016 mN/s for 7.5 s. The cycle was repeated
for five cycles.
A study comparing nano-indentation curves of non carbonated structures with those of intentionally carbonated structures
[8] did not produce any convincing conclusions.
A comparison was also made of the average slope of the unloading curve for each of the specimens (Fig. 43.4). The
standard deviation of this value was also high, was approximately 42% of the average value for each specimen.
The Unloading slope indicates local stiffness. The stiffness at the nano-scale increases as the particle size decreases from
100 to 12 nm, and then decreases for the 7 nm size samples. This trend is identical to the macro-scale compressive test results
shown in Fig. 43.1.
A comparison was made of the average maximum depth of penetration for each of the specimens (Fig. 43.5). A significant
variation is seen only for the 12 nm SiO2 enhanced mix. While Fig. 43.5 presents only the average values, the scatter in the
370 J. Schoepfer and A. Maji
observed data was very high. The standard deviation was typically higher than 50% of the average value for each specimen.
One of the standard deviations (corresponding to the 12 nm SiO2 enhanced mix) had a standard deviation higher than the
average value.
43.3 Conclusions
High-strength concrete was produced with nano-meter size silica particles added to the concrete.
SEM showed that smaller particle size leads to smaller and more numerous reaction products which alter the micro-
structure of the resulting concrete and the distribution of pore size.
Decreasing size of nano-silica results in higher compressive strength up to 12 nm, and then the strength decreases
corresponding to the 7 nm particle size. A corresponding increase in unloading stiffness (and decrease in penetration depth)
is also observed from the nano-indentation tests. This correlation between macro and nano-scale test results lends credibility
to the premise that nano-scale properties influence macro-scale properties such as compressive strength. However, the large
variation (high standard deviation compared to the average) of nano-indentation tests should be noted as an indicator of the
limited value of nano-indentation tests in such a heterogeneous material.
Acknowledgement This research was sponsored by the Defense Threat Reduction Agency (DTRA).
References
1. Kjellsen KO, Wallevrik OH, Hallgren M (1999) On the compressive strength development of high-performance concrete and paste-effect of
silica fume. Mater Struct 32:63–69
2. Kosmatka SH, Panarese WC (1994) Design and control of concrete mixtures, 13th edn. Portland Cement Association, Skokie
3. Shannag MJ (2000) High strength concrete containing natural pozzolans and silica fume. Cem Concr Compos 22(6):399–406
4. Qing Y, Zenan Z, Deyu K, Rongshen C (2007) Influence of nano-SiO2 addition on properties of hardened cement paste as compared with silica
fume. Constr Build Mater 21(3):539–545
5. Belkowitz J, Armentrout DL (2009) The investigation of nano silica in the cement hydration process. ACI Special Publication SP 267:87–100
6. Bentz DP, Stutzman PE (1994) Evolution of porosity and calcium hydroxide in laboratory concretes containing silica fume. Cem Concr Res 24
(6):1044–1050
7. Papadakis V, Fardis M, Vayenas C (1992) Hydration and carbonation of Pozzolanic cements. ACI Mater J 89(2):119–130
8. Çopuroğlu O, Schlangen E (2008) Modeling of frost salt scaling. Cem Concr Res 38:27–39
Chapter 44
Experimental Results of Green Wood Slice Under Natural Drying
Rostand Moutou Pitti, Frederic Dubois, and Eric Fournely
Abstract The natural or artificial shrinkage due to drying is commonly responsible of crack appears in wood pieces. These
defects are even more marked on the logs pieces cut up and stored in the open air during a long time. In this fact, it is essential
to understand the different processes that lead to these phenomena in the green woods. In this work, the radial and tangential
strains due to the natural drying of a green wood slice are studied by analytical and experimental methods. The analytical
method applied is based on a mathematical model coming from cylindrical strain calculations integrating the orthotropy and
mechano-sorptive effects. The experimental device is based on the mark tracking method composed of an acquisition system
recording the displacement of target posted on the wood slice during the drying process. The sample weight is measured with
an electronic balance during test. As results, the moisture content evolution during drying tests and the determination of the
fibre point saturation are obtained. Also according to the used analytical method, the radial and tangential strain components
are posted versus the distance to the center of slide. Finally, the crack initiation region is indentified.
Keywords Green wood slice • Drying • Images analysis • Strains
44.1 Introduction
Due to the current environmental changes, the scientific community is constantly looking for solutions providing to fix the
gas causing the climatic warming. Among these solutions, the use of the wood materials in engineering development, in
building or in timber structures appears very important. However, the wood behaviour is difficult to understand due to its
heterogeneity, its orthotropic and anisotropic characters, and its time dependent behaviour. In addition, these complexities
are improved when the material is put in contact with water or moisture variations as to reduce the mechanical resistance of
timber structures. Nowadays, it remains necessary to increase the lifetime of timber structures and the materials based on the
wood under sever and fluctuating environmental conditions [1]. Moreover, the initial crack presence after drying process can
decrease the timber structure lifetime by inducing stress concentrations and the crack propagation. In this context, the drying
process and the strain blocking induced by moisture content decreases and the transverse orthotropy is studied in this paper.
The radial and tangential strain evolutions due to the natural drying of a green wood slice are studied by experimental and
analytical approaches. In fact, the natural or artificial shrinkage due to drying is commonly responsible of deformations and
crack appears in wood peaces [2]. These defects are even more marked on the logs pieces cut up and stored in the open air
R. Moutou Pitti (*)

Clermont Université, Université Blaise Pascal, Institut Pascal, EA 3867, BP 10448, F-63000 Clermont-Ferrand, France
F. Dubois
E. Fournely
Clermont Université, Université Blaise Pascal, Institut Pascal, EA 3867, BP 10448, F-63000 Clermont-Ferrand, France
during a long time [3]. In the literature, the deformations and the evolution of drying-induced stresses in wood are studied in
terms of analytical [4–6] and experimental [7, 8] aspects. Among them, Kowalski and Smoczkiewicz-Wojciechowska [9]
have presented a model of drying that permits evaluation of moisture content distribution in dried wood during the drying
rate periods with the acoustic emission method (AE) monitoring the state of stress in dried wood. In the present work, the
image analysis technique [10] is applied in order to measure the mechanical fields in a drying wood slice. A numerical model
is developed for evaluating strain distributions according to transverse orthotropy and a mechano-sorptive approach
inducing strain blockings during drying phase. This model then extrapolates stress distributions until the crack appear.
The first part presents the protocol and the experimental device. The green wood slice is directly placed on an electronically
balance providing the weight measure versus time. In parallel, the mark tracking method composed of an acquisition system
date, records the displacement of target posted on the sample surface during the drying phase. The second section deals with
the analytical approach about the strains calculations. The method is based on a mathematical model coming from cylindrical
strain calculations. The radial and tangential strains are computed in the last part. The model is based on a mechano-sorptive
concept in which a hygro-lock model is activated. A strain validation is proposed.
44.2 Experimental Device
44.2.1 Experimental Protocol
A green wood slice, debited on a peace log with an electric saw is placed on a sensitive scale, Fig. 44.1. The sample is
initially conditioned in water in order to assure its saturation. A flat of 40 mm is machined on the basis of the slice in order to
assure a stability on the side of the scale. Hence, the specimen is positioned on the experimental table, taking care to ensure
its horizontality. The height between the table and axis of the camera and the slice center is hc ¼ 150 mm. Also, the camera
was immediately placed on a slidable support fixed on the table in order to reduce any disturbances due to external shocks.
This ensures that the axis of the camera is absolutely horizontal and the distance between the seat of it and the specimen on
the side of the scale is 100 mm (20 mm). Note that the distance between the axis of the slice and the axis of the support
camera is L ¼ 150 mm.

picture of the experimental
device [11]
44 Experimental Results of Green Wood Slice Under Natural Drying 373
Fig. 44.2 Evolution of

temperature and relative
humidity
Fig. 44.3 (a) Green wood

slice and black marks, (b) real
experimental device
44.2.2 Experimental Materials
The tests were carried out in a room test where the environmental conditions are maintained constant. The temperature and
the relative humidity are measured according to a thermohygrometer. In this case, Fig. 44.2 shows the evolution of the
temperature and the relative humidity versus time. It shown that, during the time test of 120 h, the relative humidity varying
between 33%RH and 40% RH with a mean of 36%; simultaneously, the temperature is around 23 C.
The experimental protocol is based on a green wood slice in Douglas fir with a thickness of 30 mm just peeled, Fig. 44.3a.
In this picture, the left side of the slice corresponding to the hardwood region, and the right side represents the normal wood.
The preparation of the specimen consists on draw black marks according to a polar reference centered at the sample heart.
A marking tracking protocol is implemented in order to record displacements of these marks using the videometric
technology [10], Fig. 44.3b. Note that 23 images are recorded per second, with an automatic record of 5 s every 2 h in
order to reduce storage requirements of images during the test. The specimen is placed on a precision balance in order to
measure the masse evolution during the drying process.
Versus time, sample mass, displacements and the crack evolution are simultaneously recorded until the mass stability
corresponding to the moisture content balance. At the end of the test, we can observe the final crack, Fig. 44.4. Note that the
crack appears when the moisture content of the slice is between 18% and 15%.
Fig. 44.4 Final crack

after drying
Fig. 44.5 Moisture 200

content evolution versus
time and detection of
point saturation fiber 160
Moisture content (%)
120
80
40
0
-10 10 30 50 70 90 110 130 150
Time (h)
44.3 Experimental Results and Analytical Approach
44.3.1 Moisture Content Variation
The first results concern the moisture content evolution during drying tests. The graphic in Fig. 44.5 is characterized by a first
linear evolution corresponding to the free water migration. At the end of test, we can observe the moisture content
equilibrium around 8%. While shrinkage swelling effect start below the fiber point saturation, it is necessary to detect this
point. Assuming that the PSF gives a non linearity in the time moisture content curve, it is focused around 30%.
44.3.2 Strain Analytical Computation

The mark tracking method calculating the displacement field in a Cartesian’s referential U x ; U y , it’s necessary to project
this field in the polar system. In this condition, we note the transition formula
Fig. 44.6 Cylindrical a →

y b →
y
coordinates: (a) radial
representation, (b) tangential
representation
→ →
r r
M1 Δq1−2
M2
M1
Δr1−2
→ M2 →
x x
rd
U r ¼ U x cos y þ U y sin y and U y ¼ Ux sin y Uy cos y (44.1)
U r and Uy are the radial and tangential displacement components, respectively. The radial and the tangential component
of the strain tensor called err and eyy respectively, are defined by

@U r 1 @U y
err ¼ and eyy ¼ þ Ur (44.2)
@r r @y
The marker tracking method given the displacement of particular points, the gradient operator must be discretized.
As shown in Fig. 44.6a, by considering two points M1 and M2 on a common radius and separated by the length Dr 12 , the
radial deformation can be estimated by Eq. 44.3. Also, Ur ðM1 Þ and Ur ðM2 Þ are the radial displacement of M1 and M2 ,
respectively. With the same idea, the tangential component can be evaluated as follows

U r ðM 1 Þ U r ðM 2 Þ U y ðM 1 Þ U y ðM 2 Þ U y ðM1 Þ þ U y ðM2 Þ
er ¼ and eyy ¼ þ (44.3)
Dr 12 Dy12 rd
Dy12 is the distance between two points M1 and M2 that ~ r is their symmetric axis. r d is the distance between the
referential origin and the projection of M1 and M2 on ~
r , as shown in Fig. 44.5b.
44.4 Experimental Analysis
In order to identify each black mark, proposes a notification by the intersection between radiuses and rings. For example, the
marks Mij are the points at the intersection between the radius Ri and the ring Cj , see Fig. 44.7.
Let us note that radial and tangential strains are defined at the middle point comprised between M1 and M2 . According to
this remark, the radial strain is calculated at the point M112 corresponding to a point placed in the radius no. 2 and between
2
rings 1 and 2. With the same way, the tangential strain is calculated at the point M23 placed in the ring no. 2 and
between radiuses 2 and 3.
44.4.1 Radial Strains
Figure 44.8 presents radial strains along radius no. 3. These figures show a non homogeneity of radial shrinkage
deformations with higher values for points of center. It shown that the
In Fig. 44.9 radial shrinkage strains are plotted for each crown. These results illustrates that strain heterogeneity does not
dependent on radius but it is sensible to regions. In the same time, the observation of these graphics shows that radial strains
decreases with the distance of specimen center.
Fig. 44.7 Determination

of radius and crowns
around the slice center
Fig. 44.8 Radial strain

along radius no. 3
44.4.2 Tangential Strains
Experimental tangential strains are measured along crowns C1 to C5 . Firstly, results are presented in terms of tangential
strain evolution versus time for points Clij placed in the intersection of crown Cl and the radius comprised between radius Ri
and Rj . Figures 44.10 and 44.11 show strain evolution along intermediary radius.
Fig. 44.9 Radial strain

along crown no. 3
Fig. 44.10 Tangential strain

for radius between R2 and R3
44.5 Conclusion
The radial and tangential strains due to the natural drying of a green wood slice have been studied by analytical and
experimental methods. In order to obtain the displacements, according to image analysis techniques, the tracks have been
posted on the slice. In the same time, the point saturation fibre around 30% is computed by measuring the mass loss during
the drying process. Also, the radial and tangential strains are computed with the analytical method integrated cylindrical
coordinates. The evolutions of radial strains versus radius no. 3 and crowns no. 3 have been presented. It shown that the
radial strains depend of the region of the slice (hardwood and normal wood). Also, the evolutions of tangential strain versus
Fig. 44.11 Tangential strain for radius between R3 and R4
radius R2 and R4 are posted. It shown also that the tangential strains are very important around the radius R2 in the hardwood.
The apparition of the crack initiation presented in Fig. 44.3 is proved by the combination of both precedent remarks. In the
coming works, the crack growth process will be studied by a new image analysis technique and by a finite element model
taking into account the drying process, the viscoelastic law and the time mechano-sorptive behaviour.
References
1. Moutou Pitti R, Alaa C (2012) Statistical approach and reliability analysis for mixed-mode applied to wood material. Wood Sci Technol.
doi:10.1007/s00226-011-0462-7
2. Schniewind AP, Pozniak RA (1971) On the fracture toughness of Douglas fir wood. Eng Fract Mech 2:223–230
3. Ormarsson S, Dahlblom O, Petersson H (1998) A numerical study of the shape of Sawn timber subjected to moisture variation, Part 1: theory.
4. Cave ID (1972) A theory of the shrinkage of wood. Wood Sci Technol 6:284–292
5. Husson JM, Dubois F, Sauvat N (2010) Elastic response in wood under moisture content variations: analytic development. Mech Time Depend
Mater 14:203–217
6. Kowalski SJ, Molinski W, Musielak G (2004) The identification of fracture in dried wood based on theoretical modeling and acoustic emission.
7. Hansson L, Antti AL (2006) The effect of drying method and temperature level on the hardness of wood. J Mater Process Technol
171:467–470
8. Awadall HSF, El-Dib AF, Mohamad MA, Reuss M, Hussein HMS (2004) Mathematical modelling and experimental verification of wood
drying process. Energy Convers Manage 45:197–207
9. Kowalski SJ, Smoczkiewicz-Wojciechowska A (2007) Stresses in dried wood: modeling and experimentation. Transp Porous Med
66:145–158
10. Peuchot B (1993) Camera virtual equivalent model 0.01 pixel detector. Comput Med Imag Graph 17:289–294
11. Moutou Pitti R (2005) Deformations during the drying green wood slice: modelling and expérimentation. Master of Science, Egletons.
University of Limoges
Chapter 45
Thickness Effect on Cross-Ply GLARE 5 FML Beams
Subjected to Ballistic Impact
A. Seyed Yaghoubi and B. Liaw
Abstract In this study, GLARE 5 fiber-metal laminates (FMLs) of dimensions: 254 mm 25.4 mm (1000 100 ) with
various thicknesses were impacted by a 0.22 caliber bullet shaped projectile using a high-speed gas gun. The specimens
were cut transversely through the impact center using a diamond blade cutting wheel for destructive damage evaluation.
The results showed that failure pattern was changed as the specimen thickness increased. This was very obvious especially
for (5/4) and (6/5) configurations. In addition, a high-speed camera was used to record the damage evolution on the non-
impacted side of the specimens. It was found that when impacted ballistically by the same projectile velocity, debonding
between the bottom (non-impacted side) aluminum layer and its adjacent glass-fiber reinforced epoxy layer was increased by
changing the lay-up configuration from (6/5) to (5/4). This phenomenon was demonstrated experimentally using the video
footage obtained from the high-speed camera.
The 3D dynamic nonlinear finite element (FE) code, LS-DYNA, was used to validate the experimental results.
To validate the reliability of the model, the finite element results were compared to those data obtained experimentally,
namely: the incident versus residual velocity relation, damage patterns and bullet residual lengths. Good agreement between
experimental and FE results was achieved.
Keywords Ballistic impact • Fiber-metal laminates • GLARE 5 • LS-DYNA
45.1 Introduction
GLARE is a member of the family of fiber-metal laminates (FMLs). A patent of GLARE was first filed on the 14th of
October 1987 by AKZO with Roebroeks and Vogelesang named as the inventors [1]. GLARE was considered for aircraft
fuselage skins at a moment in time that the first member of the FMLs family (ARALL) appeared not to be able to cover any
aircraft application. GLARE 5 is a GLARE FML, made of alternating layers of 2024-T3 aluminum alloy sheets and [0 /90 ]s
cross-ply S2-glass/epoxy laminates. The GLARE 5 patent was based on the optimization of a GLARE configuration for
impact performance [1]. Ballistic impact on GLARE FMLs had been considered by Hoo Fatt et al. [2]. They derived
analytical solutions based on test results to predict the ballistic limit and energy absorption of fully clamped GLARE panels
subjected to a ballistic impact by a blunt titanium cylinder. Based on their study, thinner panels absorbed a higher percentage
of deformation energy than thicker panels since it was easier for them to bend and stretch before fracture. Goldsmith et al. [3]
considered quasi-static and ballistic perforation of carbon fiber laminates. They reported that the ballistic limit increased
approximately linearly over the range of thicknesses investigated: 1.3–6.6 mm. Buyuk et al. [4] used LS-DYNA to simulate
the ballistic impact tests on 2024-T3/T351 aluminum material. They showed that mesh refinement did not necessarily
provide better results for ballistic limit simulations without considering and calibrating the interrelated factors, such as
material model parameters. Seyed Yaghoubi and Liaw [5] investigated incident versus residual velocity of GLARE 5 beam
specimens with various thicknesses. They demonstrated that the results obtained from their finite element simulations were
in good agreement with the experimental counterparts. A similar experimental and finite element approach had also been
adopted by Sevkat et al. [6] to study ballistic impact response of glass-fiber reinforced epoxy composite beam.
A. Seyed Yaghoubi (*) • B. Liaw

Department of Mechanical Engineering, The City College of New York, Convent Avenue and 138th Street, New York, NY 10031, USA
e-mail: ASeyed_Yaghoubi@gc.cuny.edu
380 A. Seyed Yaghoubi and B. Liaw
The main objective of this study was to investigate damage evaluation in the GLARE 5 FML beams with various
thicknesses subjected to ballistic impact. The 3D dynamic nonlinear FE code, LS-DYNA, was used to validate the
experimental results namely: the incident versus residual velocity data, post-impact damage patterns and bullet residual
lengths.
All GLARE panels considered in this study consist of 2024-T3 aluminum alloy with a thickness of 0.305 mm (0.01200 ) per
layer and unidirectional S2-glass/epoxy laminated layers, each with a thickness of 0.508 mm (0.02000 ). Each S2-glass/epoxy
layer has a lay-up orientation according to Table 45.1. GLARE 5 panels with various thicknesses were cut into rectangular
beam specimens with dimensions of 254 mm 25.4 mm (1000 100 ). The configuration of GLARE 5 panels with different
thicknesses, MVF and typical mechanical properties of constituents are described in Tables 45.1 and 45.2, respectively
[7–9]. In Table 45.1, the term MVF represents metal volume fraction and is defined as the ratio of the sum of the thicknesses
of all aluminum layers over the total thickness of the fiber-metal laminate [1]:
P
taluminum
MVF ¼ (45.1)
tFML
Table 45.1 GLARE 5 panels tested in this study with different thicknesses
Lay-up configuration (m/n) Prepreg plies and orientation Total thickness MVF
2/1 Cross-ply [0 /90 ]s 1.117 mm 0.546
(0.04400 )
3/2 1.930 mm 0.474
(0.07600 )
4/3 2.743 mm 0.445
(0.10800 )
5/4 3.556 mm 0.429
(0.14000 )
6/5 4.368 mm 0.419
(0.17200 )
The configuration notation m/n means the panel is composed of m aluminum-alloy layers interlaced with n fiber-reinforced
epoxy layers
Table 45.2 Typical values of mechanical properties [7–9]

UD S2 Glass/FM 94
Mechanical property Aluminum 2024-T3 epoxy prepreg
Tensile ultimate strength (MPa) L 455 1,900
T 448 57
Tensile yield strength (MPa) L 359 –
T 324 –
Tensile modulus (GPa) L 72 54
T – 9.4
Ultimate strain (%) L 19 3.5
T – 0.6
Compressive yield strength (MPa) L 303 –
T 345 –
Compressive modulus (GPa) L 74 54.4
T – 11
Density (g/cm3) 2.79 2.00
The symbols, L and T, stand for longitudinal (the rolling direction for the metal) and transverse directions, respectively
45 Thickness Effect on Cross-Ply GLARE 5 FML Beams Subjected to Ballistic Impact 381
Fig. 45.1 (a) The high-speed

gas gun setup for ballistic
impact tests and (b) specimen
container
Fig. 45.2 (a) 0.22 caliber

copper bullet and (b) bullet
mounted in a pair of plastic
sabots
Fig. 45.3 A composite

beam specimen clamped
at the two ends
The ballistic impact tests were performed using an in-house-designed gas gun. The gun consists of a pressurized tank, a
gas storage vessel, a solenoid valve and a stainless-steel barrel. The pictures of the gas gun setup and the specimen container
are shown in Fig. 45.1. Compressed helium/air was used as the propellant. A 0.22 caliber copper bullet was chosen as the
projectile. The bullet mounted inside a pair of plastic sabots and then inserted into the gun barrel. Figure 45.2 illustrates the
0.22 caliber copper bullet and the sabot. For the same type of sabots, the projectile initial velocity can be controlled by
changing its position inside the barrel and/or by varying the input pressure of the gun. The beam specimen was clamped over
a length of 50.8 mm from each side (Fig. 45.3). The specimen fixture was extra supported by four C-clamps inside the
specimen container (Fig. 45.1b). The composite beam specimen was then impacted by the projectile at the center.
The speed of the projectile was measured in three distinct ways. A pair of diode-lasers/amplified-photodiodes, separated by
101.6 mm (400 ), was located near the exit of the gun barrel, forming two optical gates to measure the speed of the projectile at
the gun muzzle. For the second measurement, two pairs of chronographs were used to measure the projectile speed before and
after impact; one pair was located in front of the target, while the second was located behind the target. Thirdly, a high-speed
camera was used to monitor the bullet motion during the test. In this study, the high-speed camera was set perpendicular to the
projectile ballistic trajectory. Using the captured high-speed video, the before and after the impact bullet speeds were then
determined.
Figure 45.4 represents impacted and non-impacted side views of the GLARE 5 FML beams with (2/1) and (5/4)
configurations. The corresponding cross-sectional views are depicted in Fig. 45.5. As is apparent from these figures, aluminum
layers failed through petaling, bulging and plugging. In addition, the fiber breakage/splitting/pullout, matrix cracking,
debonding/delamination and plastic deformations in the specimens were observed. It should be noted that for the (2/1)
configuration the bottom aluminum layer was not torn apart; whereas for the (5/4) configuration the bottom aluminum layer
was split into two. This was due to the debonding growth between the bottom aluminum layer and adjacent prepreg layer.
Fig. 45.4 Impacted (top) and non-impacted (bottom) side views of the GLARE 5 (2/1) and (5/4) FML beam impacted at various incident
velocities
Fig. 45.5 Cross-sectional views of the GLARE 5 (2/1) and (5/4) FML beam impacted at various incident velocities
Fig. 45.6 High-speed camera footage of the GLARE 5 (5/4) FML beam impacted at 342 m/s
Figures 45.6 and 45.7 illustrate damage progression on the non-impacted side of the GLARE 5 (5/4) and (6/5) FML beam
impacted at 342 and 340 m/s. These pictures were obtained using a high-speed camera. Since the difference in the impact
velocities was relatively negligible, it was assumed that both panels impacted at the same incident velocity. For both
specimens bulging initiated during the first 12.6 ms of contact time. A circular-shaped damage contour was induced on the
non-impacted side of the specimens. These damage contours propagated toward the boundaries of the specimens in the form
of an elliptical-shaped damage zone. As time passed by, the damage contours reached the specimen edges along the width.
Then, a crack on the bottom aluminum layers were created and propagated along the aluminum transverse direction causing
the aluminum layers to split into two parts. By further advancing in time, the broken aluminum layers continued to deflect
due to their residual kinetic energy causing the interfacial debonding between the aluminum layer and adjacent prepreg layer
to progress further. Their remaining kinetic energies were dissipated by mean of vibration. This is not shown in the figures.
As these figures suggest, under the same impact velocity, debonding between the bottom aluminum layer and its adjacent
prepreg layer was increased by changing the configuration from (6/5) to (5/4).
The specimens in Figs. 45.6 and 45.7 were removed from the specimen fixture after the impact test for the post-impact
damage analysis. Figure 45.8 illustrates the side view and back side view of these specimens. The cross-sectional views of
the specimens are given in Fig. 45.9. These figures suggest that under the same impact velocity, the (5/4) configuration
showed the highest debonding and delamination when compared to the (6/5) specimen.
45.4 Finite Element Simulation
In this study, ballistic impact on GLARE 5 beam FMLs with various thicknesses were modeled and analyzed using the finite
element software, LS-DYNA. Figure 45.10 exhibits perspective views of the composite beam. Refer to Ref. [5] for more
details on the FE simulation.
Figure 45.11 compares the FE predictions with experimental residual velocity versus incident velocity data for the
GLARE 5 (2/1) and (5/4) FML beam, respectively. The experimental and FE result are shown by a diamond and circular
Fig. 45.7 High-speed camera footage of the GLARE 5 (6/5) FML beam impacted at 340 m/s
Fig. 45.8 Side view (a), and zoomed-in back-side view (b), of the GLARE 5 (5/4) and (6/5) FML beam impacted at 342 and 340 m/s, respectively
Fig. 45.9 Cross-sectional view of the GLARE 5 (5/4) and (6/5) FML beam impacted at 342 and 340 m/s, respectively
Fig. 45.10 Perspective views of the finite element model: (a) far-off view and (b) close-up view
Fig. 45.11 Comparison between FE predictions and experimental residual-incident velocity data for the GLARE 5 FML beam: (a) (2/1) and (b) (5/4)
Fig. 45.12 Comparison of the post-impact damage pattern for the GLARE 5 (2/1) FML beam impacted at 215 m/s. In the figure BRL stands for
the bullet residual length
symbol, respectively. The experimental data was fitted by least-square regression according to the classical Lambert-Jonas
equation [10] for the positive residual velocity values:
VRP ¼ AðVIP V50

P
Þ ¼ A VIP B (45.2)
where A and B are two regression coefficients and P is a power. VR and VI are the residual and incident velocities of the
projectile, respectively, while V50 is the ballistic limit velocity, which is defined as the velocity required for a projectile to
perforate the target 50% of the time.
The Lambert–Jonas curves with different P values are represented by dashed lines [5]. The intersection of the Lambert-
Jonas curve with the incident-velocity axis was considered as the V50 velocity for a given P value. It should be mentioned
that in the figure the symbols below the incident-velocity axes represent rebound velocities of projectiles. As is apparent
from the figure, the FE predictions are in excellent agreement with the experimental results.
Figures 45.12 and 45.13 illustrate the comparisons of the post-impact damage pattern between the experimental and FE
result for GLARE 5 (2/1) and (5/4) specimens impacted at 215 and 321.5 m/s, respectively. The bullet residual lengths were
also compared to the experimental results in the figures. Good agreement between the experiment and FE result was
obtained.
After the FE model was validated based on the abovementioned criteria, it was then used to obtain some insights to the
ballistic impacts. Figure 45.14 depicts the FE predicted contact force histories and bullet residual lengths for the GLARE 5
(2/1) and (5/4) FML beam impacted at 300 m/s. As is clear from the figure, by increasing the specimen thickness, the contact
resistance force also increased; while the bullet residual length decreased. This was because the panel stiffness increased by
an increase in the specimen thickness.
45.5 Conclusions
This study presented experimental and numerical investigations on damage evaluation in the GLARE 5 FML beams with various
thicknesses subjected to ballistic impact at room temperature. The following remarks could be concluded from this study.
• The aluminum layers failed through petaling, bulging and plugging. In addition, the fiber breakage/splitting/pullout,
matrix cracking, debonding/delamination and plastic deformation in the specimens were observed.
• Under the same impact velocity, debonding between bottom aluminum layer and prepreg layer was increased by
changing the configuration from (6/5) to (5/4).
• Good agreement between the FE predictions and experimental results was achieved in terms of the projectile residual
velocity versus incident velocity data, damage patterns and bullet residual lengths.
Fig. 45.13 Comparison of the post-impact damage pattern for the GLARE 5 (5/4) FML beam impacted at 321.5 m/s. In the figure BRL stands for
the bullet residual length
Fig. 45.14 (a) FE predicted contact force histories and (b) bullet residual lengths for the GLARE 5 (2/1) and (5/4) FML beam impacted at 300 m/s
Acknowledgments The authors would like to thank the supports from NASA Faculty Award for Research (FAR) under Grant No. NAG3-2259
and from PSC-CUNY under Grants 63168-00 41 and 64543-00 42. Dr. Kenneth J. Bowles and Dr. John P. Gyekenyesi were the Technical
Monitors of the NASA grant. Part of the equipment was procured through ARO Grant No. DAAD19-03-1-0086, of which Dr. Bruce LaMattina
was the Program Manager.
References
1. Vlot A, Gunnink JW (eds) (2001) Fiber metal laminates, an introduction. Kluwer Academic Publishers, Dordrecht
2. Hoo Fatt MS, Lin C, Revilock DM Jr, Hopkins DA (2003) Ballistic impact of GLARETM fiber-metal laminates. Compos Struct 61:73–88
3. Goldsmith W, Dharan CKH, Chang H (1995) Quasi-static and ballistic perforation of carbon fiber laminates. Int J Solids Struct 32(1):89–103
4. Buyuk M, Kan S, Loikkanen MJ (2009) Explicit finite-element analysis of 2024-T3/T351 aluminum material under impact loading for airplane
engine containment and fragment shielding. J Aero Eng 22:287–295
5. Seyed Yaghoubi A, Liaw B (2012) Experimental and numerical approaches on behavior of GLARE 5 Beams: influences of thickness and
stacking sequence. In: IMAC XXX A conference and exposition on structural dynamics. Hyatt Regency Jacksonville Riverfront Jacksonville,
FL, 30 Jan–2 Feb 2012
6. Sevkat E, Liaw B, Delale F, Raju BB (2009) A combined experimental and numerical approach to study ballistic impact response of S2-glass
fiber/toughened epoxy composite beams. Compos Sci Technol 69(7–8):965–982
7. Data sheets. Aviation Equipment Structures, Inc., Costa Mesa, CA, 1998
8. QA Reports B0319B-2, B1008B-1, B0904A-3. Structural Laminates Company, New Kensington, Pennsylvania, 1994
9. Alloy 7475 Plate and Sheet, ACRP-053-B. Alcoa Mill Products, Bettendorf, Iowa
10. Lambert JP, Jonas GH (1976) Towards standardization of in terminal ballistics testing: velocity representation. Ballistic Research
Laboratories, Aberdeen Proving Ground, Maryland, Report No. BRL-R-1852
Chapter 46
Hole-Drilling Residual Stress Measurement in an Intermediate
Thickness Specimen
Colin Abraham and Gary S. Schajer
Abstract The hole-drilling method for measuring residual stresses is generally applied to two extreme cases. In the “thick”
case, the specimen depth is large compared with the hole depth; its far boundary is sufficiently distant to have negligible
effect on the deformations around the hole. In the “thin” case, the specimen has the form of a thin plate with a through hole
for which the classical Kirsch solution applies. This research focuses on the intermediate case where the specimen has a
thickness a small multiple of the hole depth. The far boundary is then near enough to have significant influence. The
unsymmetrical geometry of a blind hole in a finite thickness plate creates significant bending deformations that are not
present in either extreme case. The intermediate case is investigated here using finite element analysis and also using an
analytical model. Based on the finite-element results, the analytical model is empirically adapted to include the effects of
localized deformations around the hole. This development allows a user to make practical measurements in intermediate
thickness plates without having to do custom finite element analysis or require extensive tables of empirical results.
Keywords Residual stresses • Hole-drilling • Thin-plate stresses
46.1 Introduction
Residual stresses are stresses that exist within a material without any external load. Such stresses are created by most
manufacturing processes and can be beneficial or harmful depending on their location, size and sign. Thus, for effective
design, it is important to be able to quantify these stresses reliably.
The most commonly used method for measuring residual stresses is the “hole drilling” method [1, 2]. This involves
drilling a small hole in the stressed specimen and measuring the resulting deformations around the hole location. The method
is typically applied in two extreme cases. The first is the ‘thick’ case where the specimen is very thick compared with
the hole depth [3]. The second is the ‘thin’ case where the hole is drilled all the way through a thin plate specimen [4, 5].
These two extreme cases are described in the ASTM Standard Test Method E837-08 [1]. The document briefly refers to the
intermediate case, but gives no specific guidance on computational procedure. The work described here addresses this case
and seeks to bridge the gap between the two extremes in a way that is useful for practical measurements.
A significant challenge when doing practical hole-drilling measurements is the numerical evaluation of the calibration
constants required to relate the strains measured during hole-drilling and the stresses originally existing at the hole location.
Analytical formulas are available for the “thin” case, and ASTM E837-08 provides tables of numerical values for the “thick”
case. However, for the intermediate case, the plate thickness/hole diameter ratio introduces an additional variable that makes
necessary the use of a table of tables should a tabular approach be taken. Such an approach is burdensome because the bulk
of the required tables is substantial. In this paper an alternative approach is investigated where analytical equations and an
interpolation function are used to make a smooth transition between “thin” and “thick” cases.
C. Abraham • G.S. Schajer (*)

Department of Mechanical Engineering, University of British Columbia, Vancouver, BC V6T 1Z4, Canada
e-mail: schajer@mech.ubc.ca
390 C. Abraham and G.S. Schajer
46.2 Finite Element Analysis
The strain response measured by a hole-drilling strain gauge rosette depends on the three in-plane stresses sx, sy and txy:
e(yÞ ¼ A ( sx þ sy Þ þ B ( sx sy ) cos2y þ 2B txy sin2y (46.1)
where “A” and “B” are calibration constants that relate the measured strains to the originally existing residual stresses. The
isotropic stress (sx + sy)/2 creates axisymmetric deformations that are described by calibration constant “A”. The two shear
stresses (sx sy)/2 and txy create harmonically varying deformations that are described by calibration constant “B”.
Finite element analysis is used here to investigate the details of the deformations around a drilled hole in a stressed
material [6]. For the case of isotropic stresses and calibration constant “A”, the loading is axisymmetric around the hole. The
effect of residual stresses can be modeled by applying radial stresses of opposite sign to the hole boundary [7]. The outer
radial boundary of the finite element model, comprising >3,200 elements, is chosen to be sufficiently distant (>10 hole
diameters) so that it effectively represents an infinitely wide specimen.
An analytical solution [5] is available for the ‘thin’ case and it can be used to confirm the proper functioning of the
finite element model. A comparison of the finite element and analytical models showed agreement of the indicated
displacements within 0.1%. A further comparison was made for the “thick” case with the numerical calibration constants
“A” provided in E837-08. Here, the computed results agreed to all four significant figures quoted in the original document.
These good agreements give confidence in the reliability of the subsequent finite element computations for the intermediate
thickness case.
The finite element results for the intermediate thickness case show substantial out-of-plane bending caused by asymmetry
through the material thickness. Figure 46.1 illustrates the effect. This behavior does not occur in the “thin” case where a
through hole in a thin plate creates a symmetric structure. Neither does it occur in the “thick” case where the deep material
prevents bending deformations.
Figure 46.2 illustrates the variation of calibration constant “A” for a hole of diameter a and various normalized depths h/a
in plates over a range of normalized thickness W/a. The terminal symbols on the left correspond to the “thin” case, and those
on the right correspond to the “thick” case. The peaks in the curves are created by the bending effect illustrated in Fig. 46.1.
The bending effects grow from the left as the amount of asymmetry increases, and then diminish towards the right as the
thickness and bending stiffness of the plate increases. The bending effects do not exist in either two extreme cases, so
the approach implied in E837-08 of making a simple interpolation of the two extreme values is seen not to be realistic.
Fig. 46.1 Out-of-plane bending caused by through-thickness asymmetry of the relieved residual stresses
Fig. 46.2 Calibration constant “A” versus plate thickness for various normalized hole depths. The terminal symbols show the “thin” and “thick”
cases
46 Hole-Drilling Residual Stress Measurement in an Intermediate Thickness Specimen 391
46.3 Isotropic (Axisymmetric) Model
Observation of the plate deformation illustrated in Fig. 46.1 suggests a practical analytical model of a hole drilled in a
stressed plate. The axi-symmetric case, characterized by the calibration constant “A”, is considered first. Figure 46.3
schematically illustrates the model. It consists of an outer annular ring of thickness W, inner diameter a, and “infinite” outer
diameter. It encloses an offset circular disk of thickness W-h with common lower surface, and a hole of depth h. The residual
stresses relieved by hole cutting act on the hole surface, thereby creating a radial force C and a moment Q on the outer ring.
From consideration of equilibrium of the forces and moments between the inner disk and outer ring, and of compatibility
of displacement and rotation of the boundary between them, it may be shown that [8]:

2hk1 K 2 þ 2K2 ðk1 þ k2 ÞðW hÞ
Q¼ S (46.2)
h2 k1 k2 þ 4ðk1 þ k2 ÞðK1 þ K2 Þ

4k1 ðK 1 þ K 2 Þ hk1 k2 ðW hÞ
C¼ S 1 2 (46.3)
h k1 k2 þ 4ðk1 þ k2 ÞðK1 þ K2 Þ
where Q is the net moment per unit length acting on the outer ring (combining the effects of S, F and M in Fig. 46.3), and C is
the corresponding net radial force per unit length (combining the effects of S and F). The in-plane stiffnesses k1 and k2 and
the bending stiffnesses K1 and K2 are [8]:
F ðW hÞE C WE
k1 ¼ ¼ k2 ¼ ¼ (46.4)
Vr (a) 1n Vr (a) 1þn
M ðW hÞ3 E Q W3 E
K1 ¼ ¼ K2 ¼ ¼ (46.5)
y(a) 12ð1nÞ y(a) 12ð1 þ nÞ
where the subscripts 1 and 2 respectively refer to the inner disk and outer ring. The displacements Ur(r) at radius r on the
surface of the outer ring combine contributions from the centerline radial displacement Vr(r) and the centerline rotation y(r).
Following the method of Vinson [9] the combination of the two contributions is:

Ca W Qa2 12ð1 þ nÞ
U r ðrÞ ¼ ð1 þ nÞ (46.6)
HEr 2 EW 2 r
The analytical model requires that the boundary of the hole remains straight and perpendicular to both disk surfaces.
As the material thickness increases towards the thick case, the plate is no longer sufficiently thin to conform to the Kirchhoff
plate theory and so the model becomes less realistic. Substantial non-perpendicularity then occurs near the hole boundary
and the normals increasingly deviate from straightness. To account for this behavior, an interpolation function is introduced
to make a smooth transition between the “thin” case, as described by the analytical solution above, and the “thick” case, as
described in ASTM E837-08:

W W W
A ¼ f Aanalytic þ 1 - f AASTM (46.7)
h h h
where
2 !
W h W
f ¼ exp 0:2 1 (46.8)
h a h
Fig. 46.3 Analytical model

of a hole in an intermediate
thickness plate
Fig. 46.4 Comparison of

interpolated and finite-
element values of calibration
constant “A” versus plate
thickness for various
normalized hole depths
The form of the interpolation function is chosen to have unit value and zero slope at W/h ¼ 1 so that it returns the “thin”
solution for plates with thickness similar to the hole depth. For thicker plates, W/h increases and the interpolation function
asymptotically approaches zero and returns the ASTM results. The coefficient 0.2*h/a was determined empirically to fit
example finite element results for intermediate plate thicknesses. The response for all relevant geometries can then be found
by using Eq. 46.7 and a single table from E837-08.
Figure 46.4 compares the results of the analytical and finite element models. The finite element results (solid lines) are the
same as shown in Fig. 46.2. The analytical results (dotted lines) generally agree within 7%. This is considered to be an
acceptable agreement here because it is in the same range as the practical accuracy of typical hole drilling residual stress
measurements. The interpolated results are useful because they can be computed straightforwardly by a practical user
without having to complete a time-consuming set of finite element analyses.
46.4 Shear (Anticlastic) Model
The second case needed to analyze hole-drilling data is for shear residual stresses, corresponding to the calibration constant
“B”. Analysis of this case is almost identical to that of the axisymmetric “A” case, but with all terms multiplied by sine or
cosine of 2y. The plate deformations have the same trigonometric dependence, thereby causing the plate to bend into an
anticlastic shape. Following the method of Vinson [9], it may be shown that the stiffness constants for the shear stress case
are [8]:
F ðW hÞE C WE
k1 ¼ ¼ k2 ¼ ¼ (46.9)
Ur (a) 1þn Ur (a) 3n
M ðW hÞ3 E Q W3 E
K1 ¼ ¼ K2 ¼ ¼ (46.10)
y(a) 12ð1 þ nÞ y(a) 4ð1 þ nÞ
Substituting these results into the solution for shear stress loading found following Vinson [9], the deformations in the
outer ring are:
2
Cr a a4 2Qð1 þ nÞ a3
Ur ðr; yÞ ¼ 4 2 ð1 þ nÞ 4 þ cos 2y (46.11)
aEW r r EW2 r3

Cr a2 a4 2Qð1 þ nÞ a3
Uy ðr; yÞ ¼ 2ð1 nÞ 2 þ ð1 þ nÞ 4 þ sin 2y (46.12)
aEW r r EW2 r3
46 Hole-Drilling Residual Stress Measurement in an Intermediate Thickness Specimen 393
Fig. 46.5 Comparison

of interpolated and finite-
element values of calibration
constant “B” versus plate
thickness for various
normalized hole depths
The anticlastic deformation causes the shear stress “B” case to be much stiffer in bending. In addition, the stress
concentration around the hole is double that of the axisymmetric “A” case, 4 compared with 2. Because of this large
stress concentration the boundary at the hole edge between the inner disk and outer ring does not remain straight and
perpendicular, as is assumed by the analytical model. This deviation indicates a need for adjustments to the mathematical
solution to accommodate this behavior. The adjustments were selected so that a minimum change to the equations was made
while still attaining good agreement with the finite element data. These adjustments were found empirically. The surface
deformations over a wide range of thicknesses and hole depths were used as well as the through thickness information.
Equations 46.9, 46.10, 46.11, and 46.12 then adapt to:
F ðW hÞE C WE
k1 ¼ ¼ k2 ¼ ¼ (46.13)
Ur (a) 1þn Ur (a) 2n
M ðW hÞ3 E Q W3 E
K1 ¼ ¼ K2 ¼ ¼ (46.14)
y(a) 12ð1 þ nÞ y(a) 4ð1 þ nÞ
2
Cr a a4 2Qð3 þ nÞð1 þ nÞ a3
U r ðr; yÞ ¼ 4 2 ð1 þ nÞ 4 þ cos 2y (46.15)
aEW r r EW2 r3

Cr a2 a4 2Qð3 þ nÞð1 þ nÞ a3
U y ðr; yÞ ¼ 2ð1 nÞ 2 þ ð1 þ nÞ 4 þ sin 2y (46.16)
aEW r r EW2 r3
Figure 46.5 shows a comparison of the interpolated results using Eqs. 46.13, 46.14, 46.15, and 46.16 and the detailed
results from finite element calculation. Again, there is agreement within about 7%, which should be sufficient for ordinary use.
46.5 Conclusions
An analysis of the surface deformation of an intermediate thickness plate during hole-drilling measurements of residual
stresses shows that the response of the plate is dominated by localized bending around the hole. Such bending does not occur
in either extreme case considered in current hole-drilling practice. At the “thick” plate extreme, the large thickness inhibits
the local bending, and at the “thin” plate extreme, the hole passes through the entire plate thickness, leaving a symmetrical
structure. The bending effect is particularly pronounced when considering the axi-symmetric calibration constant “A”,
causing sharp peaks in the calibration constant versus normalized plate thickness relationship. The presence of these peaks
contradicts the implied suggestion in ASTM Standard Test Method E837-08 that a simple interpolation may be sufficient.
Based on insights gained from initial finite element results, an analytical model of the response of intermediate thickness
plates is presented. The model is found to be most effective near the thin end of the plate thickness range and it successfully
indicates the peaks in the calibration constant “A” observed in Fig. 46.2. At greater plate thicknesses, the analytical model
becomes less effective because of the local distortions created by the stress concentration near the hole. To counteract this
effect, an interpolation scheme is developed that returns the analytical solution for thin plates, the “infinite thickness”
solution for thick plates, and an interpolation of the extreme results in the intermediate range. By this means, realistic results
can be achieved over the entire plate thickness range.
The analytical solution plus interpolation approach is proposed because it allows useful results for the hole-drilling
calibration constants “A” and “B” to be evaluated without having to do custom finite element analyses or to resort to using a
very large set of tabulated numerical values. In the latter case the required tables would become very large because the
variable plate thickness introduces a new variable into the geometry specification and adds another dimension to the required
table size. Either possibility places a heavy burden on the user doing practical residual stress measurements. The focus of
such a user is to make reliable measurements with subsequent numerical calculations being as compact as possible. The
proposed interpolation scheme achieves this objective and returns results that are within about 7% of the values that
are found by detailed finite element analysis. This level of discrepancy can be considered acceptable for the application
because the typical uncertainly in residual stresses evaluated by the hole-drilling method is of similar order.
The results reported in this paper refer to the basic case where the residual stresses within the hole depth are uniform.
Work continues to extend the procedure to the general case where the residual stresses vary within the hole depth.
Acknowledgments The authors sincerely thank American Stress Technologies, Pittsburgh, PA, for their financial support of this work.
References
1. ASTM (2008) Determining residual stresses by the hole-drilling strain-gage method. Standard Test Method E837-08. American Society for
Testing and Materials, West Conshohocken, PA
2. Grant PV, Lord JD, Whitehead PS (2002) The measurement of residual stresses by the incremental hole drilling technique. Measurement Good
Practice Guide No. 53, National Physical Laboratory, Teddington
3. Rendler NJ, Vigness I (1966) Hole-drilling strain-gage method of measuring residual stresses. Exp Mech 6(12):577–586
4. Andersen LF (2002) Experimental method for residual stress evaluation through the thickness of a plate. J Eng Mater Technol 124(4):428–433
5. Nawwar AM, McLachlan K, Shewchuk J (1976) A modified hole-drilling technique for determining residual stresses in thin plates. Exp Mech
16(6):226–232
6. ANSYS (2011) Ansys structural analysis, Version 13.0. Ansys, Inc., Canonsburg
7. Schajer GS (1981) Application of finite element calculations to residual stress measurements. J Eng Mater Technol 103(2):157–163
8. Abraham C (2011) Hole-drilling residual stress measurement in an intermediate thickness specimen. MSc thesis, Department of Mechanical
Engineering, University of British Columbia, Vancouver
9. Vinson JA (2005) Plate and panel structures of isotropic composites and piezoelectric materials including sandwich construction. Springer,
Norwell
Chapter 47
Effects of Strain’s Error on Residual Stresses Calculated by HDM
C. Casavola, G. Pappalettera, C. Pappalettere, and F. Tursi
Abstract The hole-drilling method is an effective and popular semi-destructive technique for residual stress (RS)
measurement. It consists in drilling a very small hole into the specimen. Consequently, RS relaxes in the hole and stresses
in the surrounding region change causing strains also to change. A strain gage rosette, specifically designed and standardized
measures these strains. Using stress–strain relationships, the RS field is calculated from the measured strains. In the case of
not uniform RS, an incremental technique is used in which relieved strains are measured during a series of small hole depth
increments. In contrast to the uniform stress case, not uniform stress calculations are extremely sensitive to errors in the
measured strain: small strain measurement errors can cause significant variations in calculated stresses, particularly for
stresses far from the surface.
This paper presents a detailed analysis of the influence of the strain measurement error on the computed stresses.
Particular emphasis is placed on the analysis of the influence of both the number of total steps and the type of step increment.
Both the integral and power series stress calculation methods are investigated, and their different responses to measurement
errors are described.
Keywords Residual stress • Hole drilling method
47.1 Methodology
The power series method was introduced by Schajer [1] as an approximate, but theoretically acceptable method of
calculating non-uniform stress fields from incremental strain data. It is based upon the assumption that the stress field can
be approximated by terms of the power series, i.e., s0 ðhÞ ¼ 1, s1 ðhÞ ¼ h, s2 ðhÞ ¼ h2 , etc. The corresponding strain
responses on the surface when an hole is drilled into a stress fields is ruled by the coefficients a0 ðhÞ, a1 ðhÞ, a2 ðhÞ and,
b0 ðhÞ, b1 ðhÞ, b2 ðhÞ that can be calculated by finite element analysis. In this way, the measured strains can be divided into
components each one of them corresponding to a term in the power series stress fields. In practical calculation only the first
two terms of the power expansion are considered because the hole drilling method is not reliable for higher order terms. The
maximum depth below the surface is limited to 0.5 rm where rm is the radius of the gage circle. The least-squares analysis
is best done by applying the “normal equations” [2] to each of the transformed strains defined as p(h) ¼ (e3(h) + e1(h))/2;
q(h) ¼ (e3(h) e1(h))/2; t(h) ¼ (e3(h) + e1(h) 2e2(h))/2. Because the use of the transformed variables decouples
the stress/strain equations, it is possible to consider each transformed stress or strain independently of the others. The
transformed stresses P(h) ¼ (sy(h) + sx(h))/2 are calculated from strains p(h) using:
P 0 P 0 P 0
P a1 ðhÞ
a0 ðhÞ P a1 ðhÞ
a1 ðhÞ a ðhÞpðhÞ
P E
0 ¼ P (47.1)
a ðhÞa0 ðhÞ a ðhÞa1 ðhÞ P 1þn a1 ðhÞpðhÞ
C. Casavola (*) • G. Pappalettera • C. Pappalettere • F. Tursi

Dipartimento di Ingegneria Meccanica e Gestionale, Politecnico di Bari, Viale Japigia 182, 70126 Bari, Italy
e-mail: casavola@poliba.it
where P and ’P are the first two power series components of the “P” stress field, and S indicates the summation of the
products of the values corresponding to all the hole depths, h, used for the strain measurements. This calculation is repeated
for transformed stresses Q(h) ¼ (sy(h) sx(h))/2 and T(h) ¼ txy using strains q(h) and t(h) with coefficients b0 ðhÞ and
b1 ðhÞ instead of a0 ðhÞ and a1 ðhÞ, and omitting the factor 1 + v. The Cartesian stress field can be recovered using the
transformed stresses P(h), Q(h) and T(h). An advantage of the Power series method is that the least squares procedure forms
a best fit curve through the measured strain data and this makes the algorithm quite robust. This averaging effect is
particularly effective when strain measurements are made at many hole depth increments. A limitation of the method is that
it is suitable only for smoothly varying stress fields.
The use of finite element calculations as a calibration procedure has also made application of the integral method a
practical possibility. Initial developments in this area were made by Bijak-Zochowski, Niku-Lari et al., Flaman and Manning
[2]. In the integral method, the surface strain relief measured after completing hole depth step j (ej) is supposed to be related
through Eq. 47.1 with RS originally present in the material in all the hole depth steps k with 1 k j:

ð1 þ nÞ X 1 X sx sy 1 X
j j j
sx þ sy
ej ¼ ajk þ bjk cos 2y þ bjk txy k sin 2y (47.2)
E k¼1
2 k E k¼1 2 k E k¼1
where y is the angle of strain gage from the x-axis. The calibration constants ajk and bjk indicate the relieved strains in a hole j
steps deep, due to unit stresses within hole step k. Numerical values of the calibration constants have been determined by finite
element calculations for standard rosette patterns, and are tabulated in Ref. [3] by Schajer and in ASTM E837-08. They are
dimensionless, almost material-independent and hole geometry (depth and diameter) dependent constants. Using the integral
method, stress calculation are effective when few hole depth steps are used [3]. For large number of drilling steps, the
calibration matrices a and b become numerically ill-conditioned: small errors in experimental measurements can cause much
larger errors in calculated residual stresses. Moreover a fundamental physical limitation of the hole-drilling method is due to
the fact that strain are measured on the surface, so the sensitivity of the measurements decreases by increasing the hole depth.
The various mathematical methods use different approaches to minimize the adverse effects of experimental errors. However,
they all have to face the same trade-off between spatial resolution and stress uncertainty. If good spatial resolution of the
variation of residual stresses with depth is required, there will be substantial noise and uncertainty in the individual stress
values. On the other hand, if low sensitivity to experimental errors is required, spatial resolution has to be sacrificed. As with
any other mathematical calculation, the quality of the calculated residual stresses depends directly on the quality of the input
data. The extreme sensitivity to the effects of small experimental errors in strain measurements makes hole-drilling residual
stress calculations particularly dependent on having high quality measured data. Thus, meticulous experimental procedure is
essential, with careful attention given to getting accurate measurements that are as free as possible from noise and other sources
of errors [4]. To reduce this effect, ASTM E837-08 [5] and the H-Drill software [6] adopt the Tikhonov regularization [7, 8].
47.2 Experimental Plan
Several works dealing with the non uniform residual stress measurement by the hole drilling method [9] have emphasized its
high sensitivity to errors in the measured data. The main error sources are related to the measurement of relaxed strains, hole
depth, hole diameter, eccentricity and material properties estimation. In particular, the experimental practice shows that, as is
well known in the literature [10], larger errors in the computed stresses are due to the strain measurement error, since other
errors are generally lower. For this reason, the present study concentrates on the strain measurement error, which is the
dominant error source. In particular we have considered a set of linear deformation (e1L, e2L, e3L), the same set both with
random strain error included between 3 me (e1L3, e2L3, e3L3) and with random strain error included between 15 me
(e1L15, e2L15, e3L15). This last sequence was introduced in order to simulate an effectively different signal from the original
one beyond the measurement error. The problem of optimizing the step distribution was considered by Schajer and Altus [10]
and Stefanescu et al. [11]. In order to provide indications for the optimal step selection, both constant calculation step (20, 10, 4
in a total depth of 1 mm) and incremental calculation step (12, 10, 8, 6, 4 in a total depth of 1 mm) are considered. The errors on
the calculated stresses normalized to the corresponding error on the measured strain are obtained from the equations:

smin s3 e1
E3 ¼
min
3
100

(47.3)
s min e 1L e1L
47 Effects of Strain’s Error on Residual Stresses Calculated by HDM 397

smin s15 e1
E15 ¼
min
15
100

(47.4)
s min e 1L e1L
where s is the stresses calculated considering the linear deformation, s3 and s15 are the stresses calculated considering
the linear deformation with random strain error included between 3 and 15 me respectively.
Figures 47.1 and 47.2 show the maximum value of error E3 for minimum residual stress calculated by means of integral
method and power series method of Restan [12] and H-Drill software [6], both for the case of constant and incremental steps.
Experimental data show that the error obtained by elaborating data by the integral Restan software are higher than those
obtained by the integral H-Drill, both in case of constant and incremental step (Fig. 47.1a and Fig. 47.2). This result can be
partially explained in view of the fact that Tikhonov regularization is implemented in the H-Drill software and not by Restan
software. In fact if the regularization procedure is disabled in the H-Drill software errors grow up even if they still remain
smaller at higher number of drilling steps. It also worth noting that the influence the implementation of the Tikhonov
Fig. 47.1 Plot of the max

percentage error E3 versus
the number of steps used for
the calculation of the residual
stress profile in the case of
constant calculation steps.
Results are reported for the
case of the H-Drill calculation
procedure both integral (with
and without Tikhonov
regularization) and power
series and for the case of the
Restan calculation procedure
both integral and power series:
(a) constant steps, (b) constant
steps – detail
Fig. 47.2 Plot of the max percentage error E3 versus the number of steps used for the calculation of the residual stress profile in the case of
incremental calculation steps. Results are reported for the case of the H-Drill calculation procedure both integral (with and without Tikhonov
regularization) and power series and for the case of the Restan calculation procedure both integral and power series
Fig. 47.3 Plot of the percentage error E3 versus the depth of step used for the calculation of the residual stress profile in the case of 20 constant
calculation steps. Results are reported for the case of the H-Drill calculation procedure both integral (with and without Tikhonov regularization)
and power series and for the Restan calculation procedure both integral and power series: (a) 20 constant steps, (b) 20 constant steps – detail
regularization in H-drill method appears to be very strong in the constant step calculation while in the incremental step case a
reduction of few percent can be achieved. In other word it can be assessed that a proper choice of the calculating steps acts
equivalently to a regularization routine. Incremental stepping technique generally reduces error also in the integral Restan
calculation procedure. Figures 47.1 and 47.2 also show that power series approach appears less sensitive to errors on the
strain measurement.
Moreover, it seems that when strain measurements are made at many hole depth increments, errors obtained by the
H-Drill with the power series method and the integral are comparable, while error in the integral method in Restan is
considerably higher than those obtained in the power series approach in Restan. Using Restan software it seems that errors of
integral method are always larger than errors of power series method.
Figures 47.3, 47.4, and 47.5 show the value of E3 for minimum residual stress calculated by means of integral method
and power series method of Restan [12] and H-Drill software [6] for the case of constant steps.
Experimental data show that, independently of method used, the errors obtained by elaborating data grow up if stress
calculation depth increases. This result can be partially explained in view of the fact that the hole drilling method measures
Fig. 47.4 Plot of the percentage error E3 versus the depth of step used for the calculation of the residual stress profile in the case of 10 constant
and power series and for the Restan calculation procedure both integral and power series: (a) 10 constant steps, (b) 10 constant steps – detail
Fig. 47.5 Plot of the percentage error E3 versus the depth of step used for the calculation of the residual stress profile in the case of four constant
and power series and for the Restan calculation procedure both integral and power series
the strain on the surface. It also worth noting that by reducing the number of steps, at the same stress calculation depth, the
level of error on the stress calculation reduces.
Figures 47.6, 47.7, and 47.8 show the values of E3 for minimum residual stress calculated by means of integral method
and power series method of Restan [12] and H-Drill software [6] for the case of incremental steps.
Experimental data of integral RESTAN show that the incremental steps reduce the error in the last steps. Since the
maximum error occurs generally in the last steps, the incremental steps reduce the amount of the maximum error.
Experimental data of Integral Hdrill show that if Tikhonov regularization is not performed, the incremental steps reduce
errors (compare integral Hdrill no regol, 10 steps constant and incremental); if Tikhonov regularization is performed, the
incremental step does not change much the magnitude of the errors (compare integral Hdrill, 10 steps constant and
incremental).
Figures 47.9 and 47.10 show the maximum value of E15 for minimum residual stress calculated by means of integral
method and power series method of Restan [12] and H-Drill software [6] both for the case of constant and incremental steps.
Fig. 47.6 Plot of the

the depth of step used for the
calculation of the residual
stress profile in the case of 12
incremental calculation steps.
series and for the Restan
calculation procedure both
integral and power series

stress profile in the case of 10
series and for the Restan
integral and power series
By reminding that, in this way, the difference between two effective strain inputs is analyzed it is possible to observe that in
the case of constant steps, the integral H-drill approach appears to be more sensitive. If the incremental steps case is analyzed
instead it appears that sensitivity of the integral H-drill is reduced and becomes comparable with sensitivity of the other
methods the other methods. In other words incremental step algorithm seems to act like a filter on the output signal.
47.4 Conclusions
In this paper the influence of the strain measurement error on the computed stresses has been analyzed. Integral and power
series approaches were compared both for RESTAN and H-drill commercial systems. The effect of regularization was also
investigated. Furthermore the effect of using incremental optimized steps of calculation was explored.

four incremental calculation
steps. Results are reported for
the case of the H-Drill
integral (with and without
Tikhonov regularization) and
power series and for the
both integral and power series

constant calculation steps.
Tikhonov regularization) and
power series and for the
with the power series
approach

series and for the case of the
both integral and power series
It was found that integral RESTAN method and integral H-drill without regularization are strongly sensitive to strain
errors. This sensitivity increases by increasing the number of calculation steps. This situation can be improved in the H-drill
system by properly implementing the Tikhonov regularization.
Also the adoption of incremental steps of calculation appears to be effective in reducing the effects of strain errors. In this
case final results appear to be also not influenced by the number of steps.
Furtherly it was observed that errors on strain measurements introduce larger errors at higher drilling depth. This can be
attributed to the fact that strain gage rosette measures the strain at the surface of the drilled hole so that it becomes less
efficient in detecting strain induced by stress relaxation at higher depth. Moreover with regard to the integral method errors
are also introduced by the mathematical instability of solution especially at higher depth.
Finally, it should be observed that care should be paid to the choice of the elaboration parameters in order to guarantee a
good stability with respect to measurement errors and at the same time to avoid losing of sensitivity which could mask
important features in the stress distribution profile.
References
1. Schajer GS (1981) Application of finite element calculations to residual stress measurements. J Eng Mater Technol 103:157–163
2. Schajer GS (1988) Measurement of non-uniform residual stresses using the hole drilling method. Part I – Stress calculation procedures. J Eng
Mater Technol 110:344–349
3. Schajer GS (1988) Measurement of non-uniform residual stresses using the hole drilling method. Part II – Practical application of the integral
method. J Eng Mater Technol 110:344–349
4. Vishay Micro Measurements (2007) Measurement of residual stresses by the hole drilling strain gage method. Tech Note TN-503-6, 2007
5. ASTM (2008) Determining residual stresses by the hole-drilling strain-gage method. Standard Test Method E837-08. American Society for
Testing and Materials, West Conshohocken, PA
6. Schajer GS (2006) Hole-drilling residual stress calculation program Version 3.01, Vishay, User guide, 2505 West Sixth Avenue Vancouver
7. Schajer GS (2007) Hole-drilling residual stress profiling with automated smoothing. J Eng Mater Technol 129:440–445
8. Schajer GS, Prime MB (2006) Use of inverse solutions for residual stress measurements. J Eng Mater Technol 128:375–382
9. Zuccarello B (1999) Optimal calculation steps for the evaluation of residual stress by the incremental hole-drilling method. Exp Mech
39:117–124
10. Schajer GS, Altus E (1996) Stress calculation error analysis for incremental hole-drilling residual stress measurements. J Eng Mater Technol
118:120–126
11. Stefanescu D, Truman CE, Smith DJ, Whitehead PS (2006) Improvements in residual stress measurement by the incremental centre hole
drilling technique. Exp Mech 46:417–427
12. SINT Technology srl (1995) RESTAN – Sistema per la misura dello stress residuo mediante il metodo del foro. Manuale d’uso e manutenzione,
Calenzano, Firenze
Chapter 48
Hole-Drilling Residual Stress Measurement with Artifact Correction
Using Full-Field DIC
G.S. Schajer, B. Winiarski, and P.J. Withers
Abstract A full-field, multi-axial computation technique is described for determining residual stresses using the hole-
drilling method with DIC. The computational method exploits the large quantity of data available from full-field images to
ameliorate the effect of modest deformation sensitivity of DIC measurements. It also provides uniform residual stress
sensitivity in all in-plane directions and accounts for artifacts that commonly occur within experimental measurements.
These artifacts include image shift, stretch and shear. The calculation method uses a large fraction of the pixels available
within the measured images and requires minimal human guidance in its operation. The method is demonstrated using
measurements where residual stresses are made on a microscopic scale with hole drilling done using a Focused Ion Beam –
Scanning Electron Microscope (FIB-SEM). This is a very challenging application because SEM images are subject to
fluctuations that can introduce large artifacts when using DIC. Several series of measurements are described to illustrate the
operation and effectiveness of the proposed residual stress computation technique.
Keywords Residual stresses • Hole-drilling • Digital image correlation • Scanning electron microscope • Focused ion beam
48.1 Introduction
The hole-drilling method is a well-established and reliable technique for measuring residual stress in a wide range of
materials. It involves drilling a small hole in the specimen, measuring the deformations of the surrounding surface, and
evaluating the local residual stresses from the measured deformations [1, 2]. Traditionally, the surface deformations are
made using strain gauges [3, 4], but in recent years optical methods such as ESPI [5, 6], Moiré [7, 8], and DIC [9–14] have
been applied. Motivations for the use of optical methods include that they avoid the time-consuming attachment and wiring
of strain gauges and that they enable a wide range of hole sizes to be used.
Early residual stress evaluations from optical data used methods based on the techniques used for strain gauge
measurements [5, 10, 11, 15]. However, this approach limits the data usage to a small subset of the total available from
the full-field optical data. Different strategies can be followed to choose the data used, but they too involve only a small
subset of the available data and typically require substantial human guidance to make the data selection.
A significant challenge when using optical measurements is that the deformations to be measured are small relative to the
resolution limit of the measurement methods, and thus the signal-to-noise ratio of the data is often quite modest. However,
this effect can be offset by taking advantage of the large quantity of data available from the full-field optical measurements.
This paper explores the use of full-field data analysis techniques to exploit the great information richness available from
optical images. The objective is to make the calculations as automated as possible and to minimize the need for human
interaction. Techniques of this kind have been developed and successfully applied to ESPI measurements [6, 12, 16]. Here
this approach is extended to DIC measurements and further developed. The DIC method has the advantage over the ESPI
and Moiré methods that it gives displacements in multiple directions, in two dimensions fairly straightforwardly, and in three
G.S. Schajer (*)

Department of Mechanical Engineering, University of British Columbia, Vancouver, Canada
e-mail: schajer@mech.ubc.ca
B. Winiarski • P.J. Withers
School of Materials, University of Manchester, Manchester M1 7HS, UK
404 G.S. Schajer et al.
dimensions with more effort. In addition, the DIC method can be used over a wide range of length scales spanning several
orders of magnitude from microns to meters [9, 17]. ESPI and Moiré are typically useful for holes only in the millimeter
range. In addition, DIC measurements can easily provide displacement estimates in multiple directions, while ESPI and
Moiré most often provide only single directional measurements. This multi-directionality gives isotropic residual stress
measurement sensitivity.
A challenge when using DIC measurements is that the technique has barely sufficient sensitivity to identify the small
deformations that occur during hole drilling. In addition, the measurements are often prone to artifacts that can be much
larger than the deformations of interest. However, the substantial redundancy that exists within the large quantity of data
available from full-field DIC measurements provides opportunities for data averaging, error checking and elimination of
systematic artifacts. The resulting approaches greatly enhance the accuracy and reliability of the measurements and
substantially mitigate measurement concerns. An example application is considered here of Focused Ion Beam (FIB)
hole-drilling residual stress measurements within a Scanning Electron Microscope (SEM). This is a particularly demanding
application because SEM measurements are much less stable than conventional optical measurements, and are prone to
much larger and more serious artifacts. These artifacts include image displacement, image stretching and image skew.
Automatic techniques are developed here to identify and compensate for these artifacts and are shown to be effective.
Because of the scale-independent character of DIC measurements, the methods developed are also useful for conventional
macro-scale measurements and significantly enhance residual stress evaluation accuracy.
48.2 Full-Field Residual Stress Computation
In contrast to the traditional strain gauge style measurements, all optical techniques indicate surface displacements, not
strain. Estimation of surface strains from displacement measurements involves numerical differentiation, which an
inherently noisy process, and so is to be avoided. Thus, it is desirable to work directly with displacement data.
The radial displacements of the surface around a circular hole drilled in a uniformly stressed material with dimensions
much greater than the hole size have a trigonometric form [16]:
dr ðr; yÞ=a ¼ ½Pur ðrÞ þ Q vr ðrÞ cos 2y þ Tvr ðrÞ sin 2y=E (48.1)
where the stress quantities
P ¼ ðsx þ sy Þ=2; Q ¼ ðsx sy Þ=2; T ¼ txy (48.2)
respectively represent the isotropic stress, the 45 shear stress, and the axial shear stress. In Eq. 48.1, ur(r, y) is the radial
profile of the radial displacements caused by a unit isotropic stress P, and vr(r, y) is the radial profile of the radial
displacements caused by unit shear stresses Q or T. The normalizations by hole radius a and Young’s modulus E non-
dimensionalizes the radial displacement profiles ur(r) and vr(r). The resulting numerical values depend on hole depth and can
be computed using finite element analysis [16]. Since DIC measurements are scale independent, it is convenient to measure
the displacements dr(r, y) and hole radius a in units of image pixels.
The corresponding circumferential displacements are:
dy ðr; yÞ=a ¼ ½Q vy ðrÞ cos 2y þ Tvy ðrÞ sin 2y =E (48.3)
A “P” term is absent in Eq. 48.3 because an isotropic stress causes only axi-symmetric displacements, in which case
all circumferential displacements are identically zero. A relationship similar to Eq. 48.1 applies for the axial displacements
dz(r, y), but is not required here because the present example focuses on 2-dimensional measurements. However, if 3-dimensional
measurements were made, the data from the additional dimension could be directly incorporated into the calculations by
extending the procedure described here.
In addition to the surface displacements due to hole drilling described in Eqs. 48.1, 48.2, and 48.3, surface displacements
artifacts are also observed due to relative motion, axial magnification change and shearing of the imaging device. These
artifacts can typically be controlled well when imaging using conventional cameras, but less effectively when imaging using
a device such as a Scanning Electron Microscope [18]. Such artifacts are small and are not of concern in typical microscopy
48 Hole-Drilling Residual Stress Measurement with Artifact Correction Using Full-Field DIC 405
applications. However, even with modern SEM equipment, they become visible and very significant when using DIC to
identify the very small surface displacements from hole drilling.
Relative motion, magnification change and shearing artifacts are systematic in character and thus can be identified and
taken into account within the residual stress calculation. Equations 48.1 and 48.3 can be augmented to include these
quantities. Rearranging the relationships into Cartesian coordinates to fit the axial format typically used for DIC calculations
gives:
1
dx (r,yÞ= a ¼ ður (r) cosy þ vr (r) cos2ycosy vy ðrÞ cos 2y sin yÞw1
2
1
þ ður (r) cosy vr (r) cos2ycosy þ vy ðrÞ cos 2y sin yÞw2
2
þ ðvr (r) sin2ycosy vy ðrÞ sin 2y sin yÞw3
þ w4 þ ðx= HÞw5 þ ðy= HÞw6 (48.4)
1
dy (r,yÞ= a ¼ ður (r) siny þ vr (r) cos2ysiny þ vy ðrÞ cos 2y cos yÞw1
2
1
þ ður (r) siny vr (r) cos2ysiny vy ðrÞ cos 2y cos yÞw2
2
þ ðvr (r) sin2y siny þ vy ðrÞ sin 2y cos yÞw3
þ w7 þ ðy= HÞw8 þ ðx= HÞw9 (48.5)
where the normalized quantities wi are:
w1 ¼ x stress, sx =E w2 ¼ y stress, sy =E w3 ¼ xy shear stress, txy =E

w4 ¼ x image displ.=H w5 ¼ x image stretch =H w6 ¼ x image skew =H
w7 ¼ y image displ. =H w8 ¼ y image stretch=H w9 ¼ y image skew=H
and where x and y are the horizontal and vertical pixel coordinates with origin at the hole center, and H is the image height in
pixels. The numbering system for the quantities to be evaluated is used so that Eqs. 48.4 and 48.5 can be expressed within a
combined matrix equation. The normalizations are done so that the various quantities have similar sizes and so improve
computational stability. Pairs of Eqs. 48.4 and 48.5 exist for each pixel of the image, typically several hundreds of thousands
of pixels. When expressed in matrix format, the resulting equation has the following structure, illustrated for the first six
pixels, with non-zero matrix coefficients indicated by asterisks.
δ x1
δ x2
w1
δ x3
w2
δ x4
w3
δ x5
w4
δ x6
w5 = (48.6)
δ y1
w6
δ y2
w7
δ y3
w8
δ y4
w9
δ y5
δ y6
Gw5d (48.7)
Matrix G has 2 N rows and 9 columns, where N is the number of pixels. Since N typically equals several hundreds of
thousands, the number of rows (¼ number of data) greatly exceeds the number of columns (¼ number of quantities to be
determined), and thus the matrix Eq. 48.7 is highly over-determined. A “best-fit” solution can be obtained by the least-squares
method [19]. This can be conveniently computed by pre-multiplying both sides of Eq. 48.7 by the transpose of matrix G. The
resulting 9 9 matrix equation can be solved routinely.
GT G w 5 GT d (48.8)
By paying attention to the sequence of the required multiplications, it is possible to form the 9 9 GTG matrix and the
1 9 right-side vector GTd by accumulating the various permutations of the products of the matrix coefficients and
displacements at each pixel. This procedure minimizes the required numerical effort by avoiding the explicit creation and
handling of the very large matrix G and right-side vector d.
Theoretically, data from every pixel surrounding the hole could be accumulated into Eq. 48.6. However, data from the
pixels very close to the hole edge are unreliable because the hole cutting process damages the surface imaged in this area.
Data from pixels far away from the hole edge are also not useful because the deformations due to hole drilling become too
small to contribute significantly to the stress calculation. Following previous practice for ESPI calculations, data are taken
from pixels within an annular area surrounding the drilled hole [16]. For the SEM images considered here, an inner radius 1.7
times the hole radius was found to be the minimum sufficient to avoid the faulty data near the hole edge. Choice of outer
radius is less critical; about twice the inner radius appears to provide a reasonable balance between far-field data to identify
measurement artifacts and near-hole data to identify the residual stresses. Inner and outer calculation area radii of 1.7 and 3.4
times the hole radius were used for all calculations reported here. To provide the required image geometry, including
surrounding space for patches and imperfect hole centralization, the SEM magnification was chosen to give a hole diameter
approximately 20% of the image height.
If 3-dimensional DIC were used, Eq. 48.6 would have a similar but larger structure with 3 N rows and 12 columns. Since
the resulting equation would have greater data content, some improvement in residual stress evaluation accuracy can be
anticipated. This improvement is likely to occur from the somewhat superior in-plane displacement evaluations provided by
the 3-D DIC technique. However, the improvement due to the addition of the out-of plane displacements may be modest
because these displacements are small compared with the in-plane displacements and their evaluation accuracy is relatively
poor.
Taking the opposite approach, it is also possible to use one-dimensional data, for example, x-displacements using only the
upper half of the matrix in Eq. 48.6, or y-displacements using only the lower half of the matrix. Such calculations can give
useful results. However, stresses in one direction produce displacements mainly in that direction, with much smaller
displacements in the transverse direction. Thus, calculations using only x-displacements will give less reliable results for
the y-stresses, and vice-versa for calculations using only y-displacements. A calculation using Eq. 48.6 with both
displacement components together is spatially symmetrical and gives superior results for all stress components.
48.3 Digital Image Correlation
Digital Image Correlation is a well-established method for evaluating the surface displacements that have occurred between
successive images of the surface of interest [20, 21]. Typically, DIC is used with optical images measured by digital
cameras. In this study, an SEM is used to record the images. The 2-D DIC method used here involves comparing two
successive images of the target surface, taken before and after hole drilling. The image correlation proceeds by selecting a
local area of pixels, a “patch” within the first image, and then locating the corresponding patch position within the second
image based on the position of maximum correlation. Interpolation techniques allow the relative position of corresponding
patches to be determined to within 0.02 or less of a pixel spacing [22]. To allow flexibility in subsequent residual stress
computations, a custom-written computer code was written for this calculation. For computational compactness and speed
the DIC calculations were implemented using the correlation coefficient curve fitting method [23].
Patch size and shape are major factors that control DIC accuracy and resolution. In general, increasing patch size
improves correlation accuracy. However, since the data from a patch are aggregated to give an overall displacement vector,
it is necessary that the pixels within the patch all have approximately similar displacements. Thus, large patches can only be
used where displacement curvatures are small. Smaller patches must be used for images with sharply varying displacement
fields. In addition, because the DIC evaluated displacements at a given point derive from data from a finite size patch area
around that point, there must be sufficient available image data around that point. Lack of available image data becomes an
issue near to image boundaries and near the hole edge for hole-drilling images. Thus, DIC can only be used for image points
Fig. 48.1 Scanning electron

micrograph of a drilled hole,
showing example correlation
patches of elliptical shape and
variable size
beyond half a patch width from image boundaries. Small patches are therefore desirable when DIC results are needed close
to image boundaries.
The deformations to be identified for hole-drilling measurements are very small and so careful attention to patch size and
shape is needed to maximize the quality of the DIC results. The square patches typically used for general-purpose DIC work
are not ideally suited for use around the boundary of a circular hole because of their protruding. Thus, the width of the
unavailable DIC evaluation area around the hole boundary enlarges in the 45 areas. The use of circular instead of square
patches eliminates this effect.
Since the region adjacent to the hole boundary contains the highest surface deformations, it is desirable to minimize the
radius of the correlation patches used and thus the width of the unavailable boundary margin. Smaller correlation patches are
additionally appropriate near the hole boundary because the deformation gradients are relatively high. However, at locations
further from the hole boundary, larger patches can be used because deformation gradients are lower and boundary proximity
is not an issue. It is proposed here to use variable size patches, where the patch diameter is proportional to radial distance
from the hole center. Thus, smaller patches are used near the hole boundary, and larger patches further away.
A concern with the variable patch diameter strategy is that the patch size can get small in the data-rich region near the hole
boundary. The limitation on patch size applies only in the direction perpendicular to the hole boundary; a patch can be larger
in the parallel direction. Such circumferential enlargement is an acceptable possibility because the associated deformation
gradients are smaller than those in the radial direction. Thus, it is chosen here to use elliptical patches with major axis
oriented in the circumferential direction. Figure 48.1 shows an example SEM micrograph of a drilled hole with a selection of
patches around it chosen with this strategy. The two largest concentric circles indicate the annular area containing the pixels
at which DIC displacements are to be evaluated. The patches at the inner circle have shapes that successfully maintain a
consistent margin from the surface distortions near the edge of the hole. The patch width increases linearly with distance
from the hole, doubling at the outer annular circle. The combination of all these features significantly enhances the quality of
the DIC results.
48.4 FIB-SEM Equipment
A dual beam FEI xT Nova NanoLab 600i FEGSEM/FIB microscope was used to sputter a pattern of nano Pt dots on the
specimen surface. An FEI gas injection system and a standard molecular gas precursor for Pt deposition, (CH3)3 Pt(CpCH3),
in the FIB-assisted deposition mode were used [17]. Before deposition, the sample was allowed to stabilise inside the
vacuum chamber for more than 12 h. A pseudo-random nano Pt dot pattern was mapped onto the carbon-coated surface
using the 1,024 844 pixel bitmap shown in Fig.48.1. This pattern was prepared using Adobe Photoshop 7.0 software
(Adobe Systems, Inc., USA) [17]. Prior the hole-drilling experiments the surface of the beam was carbon coated (50 nm
thick film) using a Gatan PECS 682 precise etching-coating system equipped with a Gatan 681.20000 Film Thickness
Monitor, see [17] for details. This carbon coating eliminated surface charging effects on the amorphous Zr surface.
A series of micro-holes (typically approx. 5.0 mm in diameter and 2.4 mm deep) were FIB irradiated using 30 kV
acceleration voltage and 0.92 nA beam current, taking 93 s to drill each hole. The precise dimensions were individually
measured for each hole.
To achieve consistent and reliable SEM measurements for use with DIC, it is important to choose appropriate imaging
conditions, namely working distance, voltage, current, dwell time, detection of secondary electrons (SE) or back-scattered
electrons (BE), secondary ions (SI+), digital image resolution, etc. These were found following the recommendations in
[17]. Thus each image of 1,024 884 pixels suitable for DIC analysis was integrated from 8 e-beam scans (5 kV, 0.40 nA)
with e-beam dwell time of 3 ms (total image acquisition time ¼ 21.7 s). Such acquired images yielded the lowest standard
deviation of the DIC-indicated displacement within the image area and the effects of image shifts (the step changes in x- and
y-direction) on the resulting displacement/strain field were negligible.
48.5 Experimental Measurements
Several series of measurements were conducted using the FIB-SEM equipment described in the previous section. These
measurements were designed to test and illustrate the features of the computation method and its ability to compensate for
measurement artifacts. In addition, they were designed to investigate effective techniques for using FIB-SEM equipment to
achieve the most reliable and stable measurements.
Hole-drilling residual stress measurements were made on a Zr-based bulk metallic glass specimen (Zr50Cu40Al10 [24]) in
the shape of a rectangular beam 16.4 3.3 1.05 mm. The beam was cut with diamond cutting wheel using Struers
Acutom-5 precision cutting machine, polished with 600-, 1,200- and 2,500-grit paper, and then polished with ¼ micron
diamond suspension. SEM images for DIC analyses were made before and after hole drilling using a FIB, as described in the
previous section.
1. “Baseline” Measurements
The first series of measurements was designed to obtain SEM images with the smallest possible artifacts. The beam
specimen was mounted within the FIB-SEM equipment and SEM images were taken immediately before and after using
FIB to drill a set of five holes equally spaced along the specimen. Changes to equipment setup, e.g., magnification setting
and vacuum maintenance, were minimized between SEM measurements. These measurements gave a baseline case for
“good” measurements.
Figure 48.2a shows an example map of the y-displacements evaluated using DIC from the image shown in Fig. 48.1,
using the variable-size elliptical correlation patches of the type shown in Fig. 48.1. The centers of these patches were
chosen to form a rectangular grid at 15 15 pixel intervals and the displacements at the pixels throughout the image
were linearly interpolated from the DIC estimated displacements at the grid pixels.
Figure 48.2 is presented in “fringe” format. This presentation format mimics the fringe patterns produced by ESPI
measurements, where each fringe corresponds to a displacement of one wavelength. As such, Fig. 48.2 shows contour
plots where a white area corresponds to an integer number of pixel displacement and a gray area corresponds to an
intermediate number of pixels (black at half-pixel displacement). It can be seen that the displacements in Fig. 48.2a are
very small, with only modest values near the hole edge and almost zero elsewhere. The largest displacements close to the
edge of the hole are approximately 0.2 of a pixel, which is a typical size for hole drilling in a high-E material. This small
displacement range places severe challenges on the measurement and computation methods. Great care must be taken to
get as high quality data as possible, to do calculations that make most effective use of the available data and to use
methods that are as immune as possible to measurement artifacts.
Table 48.1 shows the residual stresses computed from the DIC data in Fig. 48.2a using Eq. 48.7. The table also lists the
computed sizes of the computed displacement artifacts w4 to w9. The bulk displacements w4 and w7 evaluate as zero
because the average x and y displacements within the annular computation area were previously subtracted from the DIC
data. This preparatory step was done to centralize the fringe pattern in Fig. 48.2a, but is otherwise not essential. Quantities
w5 and w8 indicate the stretches in the x and y directions respectively, and w6 and w9 indicate the skews parallel to the x
and y directions. In this well-controlled measurement, all these quantities are small.
The third row of Table 48.1 shows the residual stresses calculated using Eq. 48.7 with DIC data for both x and y
displacements. As previously indicated, it is also possible to calculate residual stresses based on x or y displacements
Fig. 48.2 Example “baseline” y-displacement maps in “fringe” format, (a) y-displacements, (b) y-displacements with artifacts subtracted,
(c) ideal data corresponding to (a), (d) residuals ¼ (a–c)
Table 48.1 Residual stresses and measurement artifacts calculated using data from Figs. 48.2 and 48.3.
x-displ. x-stretch x-skew y-displ. y-stretch y-skew
x-stress MPa, y-stress MPa, shear MPa, pix/H pix/H pix/H pix/H pix/H pix/H
Axis from w1 from w2 from w3 from w4 from w5 from w6 from w7 from w8 from w9
Fig. 48.2 x 216 225 6 0.00 0.08 0.04 – – –
Fig. 48.2 y 191 254 23 – – – 0.00 0.12 0.01
F2 x&y 219 255 15 0.00 0.08 0.05 0.00 0.12 0.00
Fig. 48.3 x 226 301 9 0.00 4.65 5.16 – – –
Fig. 48.3 y 223 264 13 – – – 0.00 7.56 3.53
F3 x&y 219 269 11 0.00 4.64 5.16 0.00 7.57 3.53
alone. In such cases, the computed stress that is in the same direction as the chosen displacements is computed more
reliably than the perpendicular stress. The first three rows of Table 48.1 show that the residual stress calculation using
dual axis displacements gives results that combine the single axis results, although not always giving a value between
them. The dual axis results have a bias towards the more reliable of the single axis results, which certainly is a desirable
property.
Returning to Fig. 48.2, map (b) shows the displacement data after the image stretching and skew corresponding to w8
and w9 have been subtracted. Since these artifacts are small in this example measurement, map (b) is very similar to map
(a). Map (c) shows the ideal data that would be expected for the residual stresses and artifacts in map (a), calculated using
the computed w values. This map should be a “clean” version of the actual measurements in map (a). The difference
Fig. 48.3 Example y-displacement maps with artifacts, (a) y-displacements, (b) y-displacements with artifacts subtracted, (c) ideal data
corresponding to (a), (d) residuals ¼ (a–c)
between the measured and ideal data is shown in the residuals map (d). If all goes well the residuals map should be close
to null, with few if any significant features. The root-mean-square (rms) of the residual map (d) is about 0.03 pixels.
2. Measurements Including Artifacts
In a second series of measurements, SEM images were made after the instrument magnification was changed and then
returned to its original setting and the vacuum chamber vented and then re-evacuated. Both of these actions, which occur
when a sample needs to be removed between measurements, are known to make significant changes to the imaging
conditions and thereby introduce measurement artifacts.
Figure 48.3 shows the displacement maps for an example measurement including very severe artifacts. This is for the
same hole as Fig. 48.2, measured after magnification change, venting and re-evacuation. The numerous fringes visible in
map (a) correspond to stretch and skew in the y direction. Each fringe represents one pixel displacement. For the
y-displacements shown in Fig. 48.3, stretch and skew respectively create horizontal and vertical fringes (vice versa for
x-displacements). These artifacts substantially dominate the displacements in Fig. 48.3a. Figure 48.3b shows the
displacement map after subtraction of the stretch and skew artifacts determined from the computed w values. This
process is very effective, and the resulting map is very similar to Fig. 48.2b. Likewise, the residuals map in Fig. 48.3d is
also close to null, similar to Fig. 48.2d. The computed stresses listed in the lower half of Table 48.1, corresponding to
Fig. 48.3 are very similar to those in the upper half, corresponding to Fig. 48.2. These results clearly show the
effectiveness of the artifact modeling in Eq. 48.7. A signal of maximum size 0.2 pixels has been successfully extracted
from among artifacts creating apparent displacements of seven pixels.
After some exploration, it was found that the source of the stretch artifact was hysteresis in the setting of the SEM
magnification. A slightly different image magnification was produced when the magnification setting was made from a
low to a high value than from a high to a low value. The difference is not large, less than 1%, but even this change is
Fig. 48.4 Bending specimen, a

(a) specimen mounted in load
frame, (b) specimen
dimensions
b 16.4
3.3
2.8 0.9typ
7.8 7.8
enough to produce a shift of several pixels in an image that is 1,024 pixels wide. The effect of this artifact was reduced to
less than one pixel by ensuring that the last magnification change was always in the same direction, from low to high.
3. Measurements with Bending Load
In a third series of measurements, the beam specimen was centrally loaded within the three-point bending fixture shown
in Fig. 48.4. SEM images were made on the beam specimen for a series of five equally spaced holes 50 mm cut adjacent to
the holes made during the first two series of measurements. New holes were cut so that the new total stresses (residual +
applied) would be measured. This procedure was used so that the incremental effect of the applied loading could be
identified by subtracting the residual stresses evaluated from the first series measurements from the third series
measurements. This incremental approach was used to investigate hole-drilling residual stress evaluation accuracy
because the initial residual stresses in the sample are not accurately known. Prior attempts to create stress-free samples
were not successful because of the occurrence of recrystallization during the annealing process. The applied stresses
induced in the beam sample were evaluated by measuring its upper surface shape using a Nanofocus mscan SC200 laser
profilometer before and after loading.
Figure 48.5 shows a comparison of the hole-drilling measurements of the applied x-stresses created by the bending
loading (third series measurements minus first series) with the stress values determined from the surface profile
measurements. The equi-spaced arrangement of the holes on a uniform beam specimen gave the linear stress variation
shown in the graph. Good agreement was achieved between the hole drilling measurements and theoretical expectations.
4. Measurements with a Reduced Bending Load
The above (third series) measurements were made under favourable conditions because both pre- and post- hole images
were made with the loaded specimen kept in place within the SEM. A fourth series of measurements was conducted to
evaluate the quality of measurements where the post-hole measurements were made after the specimen was removed, the
bending load changed, and the specimen reloaded within the SEM. The existing holes cut during the third series
measurements were then re-imaged. These measurements therefore included the adverse effects of magnification
adjustment, vacuum chamber venting and specimen movement.
The lower line of Fig. 48.5 shows a comparison of the hole-drilling measurements of the applied x-stresses created by
the bending loading (fourth series measurements minus first series) with the stress values determined from the surface
profile measurements. Again, good agreement was achieved between the hole drilling measurements and theoretical
expectations. The closeness of the agreement is remarkable because the need for very close angular alignment was not
fully appreciated when the specimen was reloaded into the SEM. An angular error of about 0.3 occurred, enough to give
a shift of several pixels across the height of the image, thereby producing a fringe pattern similar to Fig. 48.3a. However,
Eq. 48.7 was effective in compensating for this error, and gave the stress results shown in Fig. 48.5. Subsequently,
Fig. 48.5 Measured and

theoretical stresses for the
bend test specimen. 400
“S3” ¼ series 3 tests,
“S4” ¼ series 4 tests
300
X-Stress, MPa
200
S3 Theoretical
S3 Measured
100
S4 Theoretical
S4 Measured
0
0 2 4 6 8
Distance from Support, mm
Table 48.2 Computed residual stresses and artifacts for a specimen rotated by 0.1
Angle x-stress y-stress shear x-stretch y-stretch x-skew y-skew Dx-skew Dy-skew
degrees MPa MPa MPa pix/H pix/H pix/H pix/H pix/H pix/H
0.1 49 196 10 7.68 7.05 6.20 0.47 – –
0.0 19 196 3 7.66 7.03 4.59 2.12 1.61 1.65
0.1 21 193 8 7.69 6.99 2.98 3.76 1.61 1.64
measurement procedures were adjusted to minimize the rotation error when reloading specimens. The adverse effects of
the venting, specimen removal, replacement and the SEM re-evacuation can be seen by the increased scatter of the S4
points around the theoretical line in Fig. 48.5 compared with that of the S3 points.
5. Specimen Rotation Measurements
The occurrence of specimen rotation in the fourth series measurements prompted the addition of a short fifth series of
measurements to investigate residual stress evaluation stability and angle estimation in the presence of known specimen
rotations. SEM images were made on a single hole from the fourth series where the beam specimen was rotated 0.1
using a precision rotary stage mounted within the SEM.
Table 48.2 lists the residual stresses and the artifacts computed from the images taken at the three angular positions.
The computed residual stresses remain very stable and the skew values follow the expected trends very closely. A rigid-
body rotation is represented mathematically as the combination of equal and opposite axial skews, 0.1 corresponding to
943 p 0.1/180 ¼ 1.65 pixels in images 943 pixels high. The skew differences evaluated from the images closely
reproduced this value. All these results demonstrated the effective functioning of the residual stress computation and
artifact correction method.
48.6 Discussion
Residual stress evaluations from DIC data are challenging because the displacements created by hole drilling are small in the
high-E materials of common interest, typically 0.1–0.3 pixels. This modest displacement sensitivity of DIC is just sufficient
under ideal conditions, but causes a tendency to serious artifacts in non-ideal cases. The effectiveness of the proposed full-
field residual stresses calculation method is demonstrated using measurements with a FIB-SEM. This is a very challenging
application because images produced by a SEM are subject to small but very significant fluctuations. The fact that useful
DIC-based residual stress evaluations are even possible is a testament to the high quality and stability of modern SEM
imaging.
The local fluctuations in the DIC images can be seen by comparing the measured and idea fringes shown Figs. 48.2 and
48.3, panels (b) and (d). The noise, with rms about 0.12 pixels, is large compared with the stress-induced displacement of up
to 0.3 pixels. This low signal-to-noise ratio significantly impairs residual stress evaluation accuracy, even when using the
proposed full-field method with artifact correction. Based on observation of the variation of the computed residual stresses
for various cases, the standard error is estimated to be about 50 MPa for FIB-SEM measurements. Errors of this size can be
seen in Fig. 48.5. An additional source of error specific to FIB-SEM measurements is the possible introduction of localized
residual stresses due to the FIB hole drilling. The size of this possible stress introduction could not be evaluated here because
the measurements reported in Fig. 48.5 are all differential in character.
Conventional hole drilling residual stress measurements on the macro-scale using optical measurements of holes mm or
larger in diameter are much less prone to measurement artifacts. However, they retain the same low displacement sensitivity
as FIB-SEM measurements and thus can benefit from the use of the full-field computation method presented here. Useful
residual stress evaluation results were achieved by Nelson [10] when using 16 selected points within the measured images;
certainly much greater computational stability can be expected when using data from several hundred thousand pixels with
automated artifact correction. The full-image data usage also removes any need for human selection of the particular data to
be used for the residual stress calculation. The time required to complete the full-field calculations is small compared with
the measurement time, about 20 s on an ordinary laptop (Intel Atom 1 GHz, 1 GB of RAM) computer for the DIC, and about
3 s for the stress evaluation.
48.7 Conclusions
A full-field computation technique is described and successfully demonstrated for determining residual stresses using the
hole-drilling method with DIC. The computational method exploits the large quantity of data available from optical images
to ameliorate the effect of modest deformation sensitivity of DIC measurements and requires minimal human guidance in its
operation. In addition, the substantial data redundancy allows for commonly occurring measurement artifacts such as image
shift, stretch and shear to be accounted for and their effects mathematically eliminated. Other systematic artifacts could
similarly be accounted for and eliminated. The multi-axial deformation measurement capability of DIC is also exploited to
provide further data and also to enable uniform residual stress sensitivity in all in-plane directions.
The proposed method is demonstrated using measurements where residual stresses are made on a microscopic scale
within a Scanning Electron Microscope with hole drilling done using a Focused Ion Beam. This is a very demanding
application because SEM images are subject to fluctuations that can introduce large artifacts when using DIC. Several series
of measurements are described to illustrate the operation and effectiveness of the proposed residual stress computation
technique. Even though the deformations due to hole drilling are small, about 0.2–0.3 pixels, and the rms noise is relatively
large, about 0.1 pixels, the averaging of the large available quantity of data and the artifact compensation allowed useful
residual stress measurements to be made with estimated uncertainty of about 50 MPa.
The described mathematical technique is generally applicable to full-field hole-drilling image measurements, including
the ESPI, Moiré and optical DIC techniques. More accurate results are expected from optical data because of their greater
measurement stability compared with SEM images. By formulating the required least-squares calculations in an efficient
sequence, computation time is modest, just a fraction of a minute on a personal computer. The mathematical procedure is
generally automatic with minimal human guidance required.
Acknowledgments Author GSS was supported by a grant from the Natural Sciences and Engineering Research Council of Canada (NSERC).
The measurements were made within the Stress and Damage Characterization Unit at the University of Manchester, U.K., supported by the Light
Alloys Towards Environmentally Sustainable Transport (LATEST) EPSRC Portfolio Project. We are grateful to P. Liaw (the University
of Tennessee, U.S.A.) and Y. Yokoyama (Himeji Institute of Technology, Japan) for provision of the sample and A. Gholinia (the University
of Manchester, U.K.) for technical and scientific suggestions during the experiment.
References
1. Lu J (ed) (1996) Handbook of Measurement of Residual Stresses. Chapter 2: Hole-drilling and ring core methods. Fairmont Press, Lilburn
2. Schajer GS (2010) Advances in hole-drilling residual stress measurements. Exp Mech 50(2):159–168
3. ASTM (2008) Determining residual stresses by the hole-drilling strain-gage method. ASTM standard test method E837-08. American Society
for Testing and Materials, West Conshohocken
4. Grant PV, Lord JD, Whitehead PS (2002) The measurement of residual stresses by the incremental hole drilling technique. Measurement good
practice guide No.53. National Physical Laboratory, Teddington
5. Makino A, Nelson D (1994) Residual stress determination by single-axis holographic interferometry and hole drilling. Part I: theory. Exp Mech
34(1):66–78
6. Steinzig M, Ponslet E (2003) Residual stress measurement using the hole drilling method and laser speckle interferometry: part I. Exp Mech
27(3):43–46
7. Nicoletto G (1991) Moiré interferometry determination of residual stresses in the presence of gradients. Exp Mech 31(3):252–256
8. Wu Z, Lu J, Han B (1998) Study of residual stress distribution by a combined method of Moiré interferometry and incremental hole drilling.
J Appl Mech 65(4): Part I: 837–843, Part II: 844–850
9. McGinnis MJ, Pessiki S, Turker H (2005) Application of three-dimensional digital image correlation to the core-drilling method. Exp Mech
45(4):359–367
10. Nelson DV, Makino A, Schmidt T (2006) Residual stress determination using hole drilling and 3D image correlation. Exp Mech 46(1):31–38
11. Lord JD, Penn D, Whitehead P (2008) The application of digital image correlation for measuring residual stress by incremental hole drilling.
Appl Mech Mater 13–14:65–73
12. Winiarski B, Withers PJ (2010) Mapping residual stress profiles at the micron scale using FIB microhole drilling. Appl Mech Mater
24–25:267–272
13. Winiarski B, Withers PJ (2012) Micron-scale residual stress measurement by micro-hole drilling and digital image correlation. Exp Mech
52(4):417–428
14. Focht G, Schiffner K (2003) Determination of residual stresses by an optical correlative hole drilling method. Exp Mech 43(1):97–104
15. Baldi A, Bertolino F (2007) Sensitivity analysis of full field methods for residual stress measurement. Opt Lasers Eng 45(5):651–660
16. Schajer GS, Steinzig M (2005) Full-field calculation of hole-drilling residual stresses from ESPI data. Exp Mech 45(6):526–532
17. Winiarski B, Schajer GS, Withers PJ. Surface decoration for improving the accuracy of displacement measurements by digital image
correlation in SEM. Exp Mech
measurements. Part I: SEM imaging at magnifications from 200 to 10,000. Exp Mech 6(47):775–787
19. Dahlquist G, Bj€ork Å, Anderson N (1974) Numerical methods. Prentice-Hall, Englewood Cliffs, Chapter 4
20. Schajer GS, Steinzig M (2010) Dual-axis hole-drilling ESPI residual stress measurements. ASME J Eng Mater Technol 132(1):71–75
21. Sutton MA, McNeill SR, Helm JD, Chao YJ (2000) Advances in two-dimensional and three-dimensional computer vision. In: Rastogi PK (ed)
Photomechanics. Springer, Berlin/Heidelberg
22. Sutton MA (2008) Digital image correlation for shape and deformation measurements. In: Sharpe WN (ed) Springer handbook of experimental
solid mechanics. Springer, Berlin/Heidelberg
23. Pan B, Xie H-M, Xu B-Q, Dai F-L (2006) Performance of sub-pixel registration algorithms in digital image correlation. Meas Sci Technol
17(6):1615–1621
24. Tian JW, Shaw LL, Wang YD, Yokoyama Y, Liaw PK (2009) A study of the surface severe plastic deformation behaviour of a Zr-based bulk
metallic glass (BMG). Intermetallics 17(11):951–957
Chapter 49
Anisotropic and Pressure-Dependent Plasticity Modeling
for Residual Stress Prediction
Michael B. Prime
Abstract In order to provide well characterized residual stress specimens, 10-mm thick, 60-mm diameter disks of
2024-T351 Aluminum were plastically indented by opposing 15-mm diameter indenters of hardened steel. The residual
stresses in the disk specimen were measured using neutron diffraction, the contour method, and the slitting method. A finite
element model of the indentation process was constructed, but matching the measured residual stresses proved unexpectedly
challenging. An attempt was made to improve the agreement honestly (without any unjustified parameter changes) by
improving the constitutive model for 2024. Cyclic stress–strain curves were measured in multiple directions in the source
plate of 2024-T351 which showed plastic anisotropy on the order of 15% in the flow strength. Recent literature has also
shown a pressure dependence in 2024-T351 which would increase the flow strength in the triaxial stress region under the
indenter. Combining anisotropy, cyclic loading, and pressure dependence effects in Abaqus has significantly improved
agreement with the data, but a completely accurate prediction remains elusive.
Keywords Residual stress • Plastic anisotropy • Neutron diffraction • Bauschinger effect • Pressure-dependent yielding
49.1 Introduction
Residual stresses play a significant role in many material failure processes like fatigue, fracture, and stress corrosion
cracking [1, 2]. Residual stresses are the stresses present in a part free from external load, and they are generated by virtually
any manufacturing process. The subject of this study is indented-disk test specimens that were designed to provide a
controlled distribution of residual stress [3] in order to develop and test methods for measuring residual stresses.
Accurate finite element modeling of the disk specimens has proven somewhat challenging. In previous work, accurate
residual stress modeling for disks of 316L stainless steel was achieved only after cyclic testing and subsequent calibration of
a combined hardening model to capture the Bauschinger effect [3]. In this paper, we explore accurate modeling for indented
disks made of 2024-T351 Aluminum, which have proven even more challenging to model accurately.
49.2 Experimental
Disks were plastically compressed through the thickness by two cylindrical indenters of smaller diameter, see Fig. 49.1.
The disks of 2024-T351 aluminum were 60 mm in diameter and 10 mm thick. The indenter material used was A2 tool steel
characterized by a high hardness (58 HRC) and a high yield stress (about 1,300 MPa). The indentation was performed quasi-
statically to a maximum load 99.6 kN and then unloaded. The residual stresses in the specimens were measured using
neutron diffraction [4, 5], the contour method [6–8], and the slitting method [9, 10], with the measurement details on the
2024 specimen discussed in this paper reported elsewhere [3, 11].
M.B. Prime (*)

Los Alamos National Laboratory, Los Alamos, NM 87545, USA
e-mail: prime@lanl.gov
416 M.B. Prime
Fig. 49.1 Indented disk Loading direction

residual stress specimens
were made using 2024-T351
Aluminum
Indenters (A2 steel)
z Disk
Loading direction
49.3 Modeling
The disk indentation process was modeled using ABAQUS version 6.11. For initial runs with isotropic plasticity models, a
half-symmetry axi-symmetric model of the specimen, see Fig. 49.2, was built using 15,000 four-node quadrilateral elements
with reduced integration (CAX4R). Square elements 0.1 mm on a side gave a 50 300 mesh in the disk. The contact
behavior between the indenter (master surface) and the disk (slave surface) was assumed frictionless because lubricant was
used during the experimental test, and a surface-to-surface contact algorithm was used. A pressure load was applied to the
top surface of the indenter to match the experimentally applied load. Figure 49.3 shows the 1/8th symmetry, 3D mesh used
for later simulations with anisotropic material models. Twenty node brick elements with reduced integration (C3D20R) were
used with elements approximately 0.5 mm on a side.
The indenter was modeled as elastic with Young’s modulus of 204 GPa and Poisson’s ratio of 0.3. For the Al 2024, the
Young’s modulus was 73.2 GPa and Poisson’s ratio was 0.33. The plasticity model for the Al 2024 was varied during
the study and is described below.
49.3.1 Model Attempt Using Isotropic Plasticity with Combined Hardening
Small compression samples were taken from the 12.7 mm thick Al 2024-T351 plate. The results showed modest plastic
anisotropy, with the flow stress approximate same in the rolling and through-thickness directions of the Al 2024 plate, and
about 10–15% lower in the transverse direction. Subsequently, specimens were extracted in the rolling and transverse
directions of the plate and tested in compression-tension cyclic tests. The plate was too thick to extract such samples in the
through-thickness direction. Figure 49.4, shows the data from the cyclic tests. Based on the nearly identical behavior in
the loading-only compression tests, the through-thickness direction was assumed to have the same cyclic behavior as rolling
direction.
Based on the experience with 316L specimens [3], a cyclic combined hardening model was carefully calibrated for the
2024 Al, see Fig. 49.4. The resulting model prediction did not match the data very well; see Figs. 49.5 and 49.6. This
anisotropy was ignored in this initial modeling because 15% errors in residual stress or strain would have been acceptable for
this application. Note, however, that the ~15% plastic anisotropy in the flow stress has produced more than a factor of two
difference in the residual radial strains, see Fig. 49.5.
49 Anisotropic and Pressure-Dependent Plasticity Modeling for Residual. . . 417
Fig. 49.2 Axisymmetric z

finite element model Displacement
12.7 mm
Axisymmetry Indenter
axis
Contact surfaces
Disk
7.5 mm
Symmetry
plane
5 mm
r
30 mm
Fig. 49.3 1/8th symmetry

3D finite element mesh for
simulations with anisotropic
material models
49.3.2 Reverse Yielding
In ABAQUS, a combined hardening model cannot be used simultaneously with anisotropic plasticity, and clearly anisotropic
plasticity is necessary to accurate model the stresses. A short study was performed to estimate the magnitude and importance
of reverse yielding during unloading of the indenters on the disk. The constitutive data was fit using a kinematic hardening
model. The model fit is shown in Fig. 49.7. Such a model only allows for linear strain hardening, but does a reasonable job of
fitting the loading portion of the data and conservatively estimates the reverse yielding.
418 M.B. Prime
Fig. 49.4 Isotropic plasticity,

combined hardening cyclic
stress–strain curves calibrated
for 2024-T351 aluminum
alloy. The model fits the cyclic
behavior but does not fit the
anisotropy
Fig. 49.5 Residual elastic 1.0E-3

radial strains in 2024-T351
disk, model versus data. Neutron, along transv.
Taken at mid-thickness along 5.0E-4
diameters of disk Neutron, along rolling
corresponding to rolling and
transverse directions of plate FEM, isotropic
Residual radial strain
0.0E+0
material
-5.0E-4
-1.0E-3
-1.5E-3
-2.0E-3
-30 -20 -10 0 10 20 30
r (mm)
A simulation of the indentation process with the kinematic hardening model showed no reverse plasticity during the
unloading. Therefore, it should not be necessary to accurately model the cyclic behavior of the 2024 Al so long as premature
reverse yielding is not predicted.
This simple model did not give a very good prediction of the residual stresses because of the inability of the model to fit
the finer details in the loading portion of stress–strain curve, as shown in Fig. 49.7. Although such details may seem minor,
most of the disk only sees plastic strains of under 1%, which is where the fit is least accurate. Figure 49.8 shows the plastic
strain in the disk after an indentation simulation. Except for a small stress concentration region near the edge of the indenter,
the disk only sees plastic strains of less than 2%.
Fig. 49.6 Residual hoop 100

stress in 2024-T351 disk,
models versus data. Taken
50
at mid-thickness along
diameters of disk
corresponding to rolling 0
and transverse directions
of plate material
Hoop stress (MPa)

-50
-100
-150
-200
Neutron, along transv. Neutron, along rolling
-250 Contour, along rolling Slitting
FEM, isotropic
-300
-30 -20 -10 0 10 20 30
r (mm)
Fig. 49.7 In order to check 500

for reverse yielding, an
anisotropic model with
kinematic hardening was
fit to the data
True stress (MPa)
rolling and TT data
transverse data
0 roll/TT fit
transv. fit
-500
-0.025 0.000 0.025
True strain
49.3.3 Anisotropic Plasticity with Isotropic Hardening
Since it was no longer necessary to model the reverse plasticity, an anisotropic fit using isotropic hardening was used. Plastic
anisotropy was modeled using Hill’s potential function [12] in ABAQUS. Because the conversion of uniaxial stress–strain
curves to effective-stress–effective-strain curves depends on the R ratios, an iterative process was used to fit the R ratios to
the data. Figure 49.9 shows the fit for optimizing for the range of strains up to 1%, which gave an R of 1.16, or 16%
anisotropy. This fit was applied in ABAQUS using the *POTENTIAL function to define the anisotropy and a tabular
description of the stress–strain curve. Simple one element problems were used to verify that the implementation matched the
calibration data of Fig. 49.9.
420 M.B. Prime
Fig. 49.8 Equivalent plastic

strains in the disk after the
indentation simulation are
mostly under 2%
(Deformation exaggerated by
a factor of 10)
Fig. 49.9 Anisotropic 500

plasticity fit to data, optimized
over range 0–1.0% plastic
strain 400
Stress (MPa)
300
Transverse data
Rolling data (=TT)
200 Transverse fit
Rolling/TT fit
100
0
0 0.005 0.01 0.015 0.02
Plastic strain
The anisotropic model was used on the indentation simulation. The resulting residual strain and residual stress predictions
are compared with data in Figs. 49.10 and 49.11. The model now shows the observed anisotropy in the results, but the overall
magnitude is still incorrect.
49.3.4 Pressure Dependent Plasticity
In addition to the residual stresses, the simulations were also compared with load–displacement data taken during the
indentation process [13]. The displacement portion of the load–displacement data has large uncertainty because of the
need to correct for machine compliance and the presence of lubricant during the indentation, but is informative nonetheless.
It was hoped that the data was accurate enough to see if the model had larger inaccuracies during the loading or unloading
portions of the simulation. Comparing the model prediction to the load–displacement data indicates that the modeled yield
strength during loading needs to be about 10% stronger than what was observed in the uniaxial compression testing in order
to match the data. The residual stress results are also consistent with that observation.
Fig. 49.10 Residual elastic 1.0E-3

radial strains predicted with
the anisotropic plasticity Neutron, along transv.
model, versus data Neutron, along rolling
5.0E-4 FEM, anisotropic, along R
FEM, anisotropic, along T
FEM, isotropic

0.0E+0
-5.0E-4
-1.0E-3
-1.5E-3
-2.0E-3
-30 -20 -10 0 10 20 30
r (mm)
Fig. 49.11 Residual hoop 100

stresses predicted with the
anisotropic plasticity model,
versus data 50
0
Hoop stress (MPa)
-50
-100
-150
-200
Neutron, along transv. Neutron, along rolling
-250 Contour, along rolling Slitting
FEM, anis., along rolling FEM, anis., along transv.
FEM, isotropic
-300
-30 -20 -10 0 10 20 30
r (mm)
Some recent studies have indicated that under conditions of high triaxial compressive stress, such as under the indenters
in this experiment, the yield strength in 2024 Al increases [14]. The triaxiality is defined as the ratio of mean stress to
effective stress:
sm
¼ (49.1)
s
422 M.B. Prime
Fig. 49.12 Triaxiality at

peak load. The scale zoomed
in to show the compressive
triaxiality under the indenter
Figure 49.12 shows the triaxiality from Eq. 49.1 plotted at peak load during the simulation. Under the indenter the
specimen sees a triaxiality of about 0.8 through the thickness, with peak triaxialities even higher.
Bai and Wierzbicki [14] give a relation between the yield stress and the triaxiality as

syld ¼ s ep 1 c ð 0 Þ (49.2)

Where s ep is the stress–strain curve from the reference test taken at a triaxiality of 0, which was 1/3 in our uniaxial
compression tests. They found a value of c of 0.09 for 2024 Al. This would give a strength increase of 4% at a triaxiality of
0.8 and more where the triaxiality is more negative.
The pressure dependent yield strength was implemented into the ABAQUS model by using a user defined field
to calculate triaxiality and then augmenting the tabular stress–strain data to provide stresses as a function of the triaxiality
to match Eq. 49.2. Test problems were used to verify that the implementation gave the expected behavior. The Appendix
gives the Abaqus input for the constitutive model including the user routine for calculating triaxiality.
The resulting residual stress predictions move the predictions in Figs. 49.10 and 49.11 in the right direction, but only by
about 6 MPa, compared to the 50 MPa that would be needed to match the data.
49.4 Conclusion
The models do not predict the residual stresses in the indented disks as well as hoped. Further studies have shown that the
predictions are insensitive to the friction coefficient between the indenter and disk and to other parameters. The most likely
explanation remain that the constitutive model is inaccurate in some regard. The next step would be to obtain some in situ
data during the indentation process. It is hoped that such data would be able to identify if the inaccuracy occurs during the
loading or unloading of the indenter, which would narrow down what aspect of the model to improve. The original in situ
data was load–displacement data on the indenter, but it lacks the necessary sensitivity. The addition of strain gauges to take
load-strain data might prove informative (Figs. 49.13 and 49.14).
Fig. 49.13 Effect of pressure 1.0E-3

dependent plasticity model on Neutron, along transv.
residual elastic radial strains Neutron, along rolling
versus data 5.0E-4 FEM, anisotropic, along R
FEM, anisotropic, along T
FEM, P dependent, along R

0.0E+0
FEM, P dependent, along T
-5.0E-4
-1.0E-3
-1.5E-3
-2.0E-3
-30 -20 -10 0 10 20 30
r (mm)
Fig. 49.14 Effect of pressure 100

dependent plasticity model on
residual hoop stress versus
data 50
0
Hoop stress (MPa)
-50
-100
Neutron, along transv.
Neutron, along rolling
-150
Contour, along rolling
Slitting
-200
FEM, anis., along rolling
FEM, anis., along transv.
-250 FEM, P dependent, along R
FEM, P dependent, along T
-300
-30 -20 -10 0 10 20 30
r (mm)
Acknowledgements This work was performed at Los Alamos National Laboratory, operated by the Los Alamos National Security, LLC for the
National Nuclear Security Administration of the U.S. Department of Energy under contract DE-AC52-06NA25396. By acceptance of this article,
the publisher recognizes that the U.S. Government retains a nonexclusive, royalty-free license to publish or reproduce the published form of this
contribution, or to allow others to do so, for U.S. Government purposes.
Appendix
For reference, the information to reproduce the constitutive behaviour of this paper is given below in two parts: (1) the
information for the input file and (2) the user subroutine for the triaxiality user defined field.
424 M.B. Prime
Input File Text
*USER DEFINED FIELD

*DEPVAR
1
*INITIAL CONDITIONS, TYPE¼FIELD, VARIABLE¼1
,0
*Material, name¼Al2024_cyclic
*Density
2.7e-09,
*Elastic
73200., 0.33
*Plastic, hardening¼isotropic, dependencies¼1
80.55, 0.0000e+000, 0.0, -3.000
131.58, 2.5000e-004, 0.0, -3.000
207.90, 5.0000e-004, 0.0, -3.000
252.85, 7.5000e-004, 0.0, -3.000
294.90, 1.0000e-003, 0.0, -3.000
322.78, 1.2500e-003, 0.0, -3.000
341.79, 1.5000e-003, 0.0, -3.000
358.45, 1.7500e-003, 0.0, -3.000
369.07, 2.0000e-003, 0.0, -3.000
377.69, 2.2500e-003, 0.0, -3.000
384.92, 2.5000e-003, 0.0, -3.000
391.61, 2.7500e-003, 0.0, -3.000
397.88, 3.0000e-003, 0.0, -3.000
402.24, 3.2500e-003, 0.0, -3.000
406.36, 3.5000e-003, 0.0, -3.000
410.22, 3.7500e-003, 0.0, -3.000
413.93, 4.0000e-003, 0.0, -3.000
416.46, 4.2500e-003, 0.0, -3.000
419.62, 4.5000e-003, 0.0, -3.000
422.66, 4.7500e-003, 0.0, -3.000
425.34, 5.0000e-003, 0.0, -3.000
427.84, 5.2500e-003, 0.0, -3.000
429.73, 5.5000e-003, 0.0, -3.000
432.32, 5.7500e-003, 0.0, -3.000
434.48, 6.0000e-003, 0.0, -3.000
436.50, 6.2500e-003, 0.0, -3.000
438.19, 6.5000e-003, 0.0, -3.000
440.25, 6.7500e-003, 0.0, -3.000
442.01, 7.0000e-003, 0.0, -3.000
443.77, 7.2500e-003, 0.0, -3.000
445.70, 7.5000e-003, 0.0, -3.000
447.06, 7.7500e-003, 0.0, -3.000
448.54, 8.0000e-003, 0.0, -3.000
450.01, 8.2500e-003, 0.0, -3.000
451.31, 8.5000e-003, 0.0, -3.000
452.60, 8.7500e-003, 0.0, -3.000
454.08, 9.0000e-003, 0.0, -3.000
455.52, 9.2500e-003, 0.0, -3.000
457.04, 9.5000e-003, 0.0, -3.000
458.35, 9.7500e-003, 0.0, -3.000
459.51, 1.0000e-002, 0.0, -3.000
461.23, 1.0375e-002, 0.0, -3.000

463.95, 1.0938e-002, 0.0, -3.000
467.77, 1.1781e-002, 0.0, -3.000
492.94, 1.3047e-002, 0.0, -3.000
480.53, 1.4945e-002, 0.0, -3.000
490.26, 1.7793e-002, 0.0, -3.000
503.44, 2.2064e-002, 0.0, -3.000
520.75, 2.8472e-002, 0.0, -3.000
542.84, 3.8083e-002, 0.0, -3.000
562.77, 4.8083e-002, 0.0, -3.000
580.89, 5.8083e-002, 0.0, -3.000
597.08, 6.8083e-002, 0.0, -3.000
611.45, 7.8083e-002, 0.0, -3.000
624.45, 8.8083e-002, 0.0, -3.000
636.16, 9.8083e-002, 0.0, -3.000
646.78, 1.0808e-001, 0.0, -3.000
656.68, 1.1808e-001, 0.0, -3.000
665.60, 1.2808e-001, 0.0, -3.000
673.75, 1.3808e-001, 0.0, -3.000
681.30, 1.4808e-001, 0.0, -3.000
688.23, 1.5808e-001, 0.0, -3.000
694.57, 1.6808e-001, 0.0, -3.000
700.46, 1.7808e-001, 0.0, -3.000
705.87, 1.8808e-001, 0.0, -3.000
710.90, 1.9808e-001, 0.0, -3.000
715.59, 2.0808e-001, 0.0, -3.000
51.32, 0.0000e+000, 0.0, 2.000
83.83, 2.5000e-004, 0.0, 2.000
132.45, 5.0000e-004, 0.0, 2.000
161.09, 7.5000e-004, 0.0, 2.000
187.88, 1.0000e-003, 0.0, 2.000
205.64, 1.2500e-003, 0.0, 2.000
217.75, 1.5000e-003, 0.0, 2.000
228.37, 1.7500e-003, 0.0, 2.000
235.14, 2.0000e-003, 0.0, 2.000
240.62, 2.2500e-003, 0.0, 2.000
245.23, 2.5000e-003, 0.0, 2.000
249.50, 2.7500e-003, 0.0, 2.000
253.49, 3.0000e-003, 0.0, 2.000
256.26, 3.2500e-003, 0.0, 2.000
258.89, 3.5000e-003, 0.0, 2.000
261.35, 3.7500e-003, 0.0, 2.000
263.71, 4.0000e-003, 0.0, 2.000
265.33, 4.2500e-003, 0.0, 2.000
267.34, 4.5000e-003, 0.0, 2.000
269.28, 4.7500e-003, 0.0, 2.000
270.99, 5.0000e-003, 0.0, 2.000
272.57, 5.2500e-003, 0.0, 2.000
273.78, 5.5000e-003, 0.0, 2.000
275.43, 5.7500e-003, 0.0, 2.000
276.80, 6.0000e-003, 0.0, 2.000
278.09, 6.2500e-003, 0.0, 2.000
279.17, 6.5000e-003, 0.0, 2.000
280.48, 6.7500e-003, 0.0, 2.000
281.60, 7.0000e-003, 0.0, 2.000
426 M.B. Prime
282.72, 7.2500e-003, 0.0, 2.000

283.95, 7.5000e-003, 0.0, 2.000
284.82, 7.7500e-003, 0.0, 2.000
285.77, 8.0000e-003, 0.0, 2.000
286.70, 8.2500e-003, 0.0, 2.000
287.53, 8.5000e-003, 0.0, 2.000
288.35, 8.7500e-003, 0.0, 2.000
289.29, 9.0000e-003, 0.0, 2.000
290.21, 9.2500e-003, 0.0, 2.000
291.18, 9.5000e-003, 0.0, 2.000
292.01, 9.7500e-003, 0.0, 2.000
292.75, 1.0000e-002, 0.0, 2.000
293.85, 1.0375e-002, 0.0, 2.000
295.58, 1.0938e-002, 0.0, 2.000
298.02, 1.1781e-002, 0.0, 2.000
301.31, 1.3047e-002, 0.0, 2.000
306.15, 1.4945e-002, 0.0, 2.000
312.34, 1.7793e-002, 0.0, 2.000
320.74, 2.2064e-002, 0.0, 2.000
331.77, 2.8472e-002, 0.0, 2.000
345.84, 3.8083e-002, 0.0, 2.000
358.54, 4.8083e-002, 0.0, 2.000
370.08, 5.8083e-002, 0.0, 2.000
380.40, 6.8083e-002, 0.0, 2.000
389.55, 7.8083e-002, 0.0, 2.000
397.83, 8.8083e-002, 0.0, 2.000
405.30, 9.8083e-002, 0.0, 2.000
412.06, 1.0808e-001, 0.0, 2.000
418.37, 1.1808e-001, 0.0, 2.000
424.05, 1.2808e-001, 0.0, 2.000
429.24, 1.3808e-001, 0.0, 2.000
434.05, 1.4808e-001, 0.0, 2.000
438.47, 1.5808e-001, 0.0, 2.000
442.51, 1.6808e-001, 0.0, 2.000
446.26, 1.7808e-001, 0.0, 2.000
449.71, 1.8808e-001, 0.0, 2.000
452.91, 1.9808e-001, 0.0, 2.000
455.90, 2.0808e-001, 0.0, 2.000
**
*Potential
1.162, 1.162, 1., 1.2364, 1.1, 1.1
**
User Subroutine to Give Triaxiality as User Defined Field
SUBROUTINE USDFLD(FIELD,STATEV,PNEWDT,DIRECT,T,CELENT,
1 TIME,DTIME,CMNAME,ORNAME,NFIELD,NSTATV,NOEL,NPT,LAYER,
2 KSPT,KSTEP,KINC,NDI,NSHR,COORD,JMAC,JMATYP,MATLAYO,
3 LACCFLA)
C
INCLUDE ’ABA_PARAM.INC’
C
CHARACTER*80 CMNAME,ORNAME
CHARACTER*3 FLGRAY(15)
DIMENSION FIELD(NFIELD),STATEV(NSTATV),DIRECT(3,3),
1 T(3,3),TIME(2)
DIMENSION ARRAY(15),JARRAY(15),JMAC(*),JMATYP(*),
1 COORD(*)
C
C Cet Pressure from stress invariants output
CALL GETVRM(’SINV’,ARRAY,JARRAY,FLGRAY,JRCD,JMAC,JMATYP,MATLAYO,LACCFLA)
PRESS ¼ ARRAY(3)
SMISES ¼ ARRAY(1)
C Calculate triaxiality
IF (SMISES .ne. 0) THEN
TRIAX ¼ -PRESS / SMISES
ELSE
TRIAX ¼ 0
END IF
FIELD(1) ¼ TRIAX
STATEV(1) ¼ FIELD(1)
C If error, write comment to .DAT file:
IF(JRCD.NE.0)THEN
WRITE(6,*) ’REQUEST ERROR IN USDFLD FOR ELEMENT NUMBER ’,
1 NOEL,’INTEGRATION POINT NUMBER ’,NPT
ENDIF
C
RETURN
END
References
1. Withers PJ (2007) Residual stress and its role in failure. Rep Prog Phys 70(12):2211–2264
2. James MN (2011) Residual stress influences on structural reliability. Eng Fail Anal 18(8):1909–1920
3. Pagliaro P, Prime MB, Clausen B, Lovato ML, Zuccarello B (2009) Known residual stress specimens using opposed indentation. J Eng Mater
Technol 131:031002
4. Hutchings MT, Withers PJ, Holden TM, Lorentzen T (2005) Introduction to the characterization of residual stress by neutron diffraction.
Routledge, USA
5. Woo W, Choo H, Brown DW, Feng Z, Liaw PK (2006) Angular distortion and through-thickness residual stress distribution in the friction-stir
processed 6061-T6 aluminum alloy. Mater Sci Eng A 437(1):64–69. doi:10.1016/j.msea.2006.04.066
6. Prime MB, Sebring RJ, Edwards JM, Hughes DJ, Webster PJ (2004) Laser surface-contouring and spline data-smoothing for residual stress
measurement. Exp Mech 44(2):176–184
7. Wilson GS, Grandt AF Jr, Bucci RJ, Schultz RW (2009) Exploiting bulk residual stresses to improve fatigue crack growth performance of
structures. Int J Fatigue 31(8–9):1286–1299
8. DeWald AT, Hill MR (2009) Eigenstrain based model for prediction of laser peening residual stresses in arbitrary 3D bodies. Part 1: model
description. J Strain Anal Eng Des 44(1):1–11
9. Cheng W, Finnie I (2007) Residual stress measurement and the slitting method. Mechanical Engineering Series Springer Science + Business
Media, LLC, New York
10. Jones KW, Dunn ML (2008) Fatigue crack growth through a residual stress field introduced by plastic beam bending. Fatigue Fract Eng Mater
Struct 31(10):863–875
11. Pagliaro P, Prime MB, Robinson JS, Clausen B, Swenson H, Steinzig M, Zuccarello B (2011) Measuring inaccessible residual stresses using
multiple methods and superposition. Exp Mech 51(7):1123–1134. doi:10.1007/s11340-010-9424-5
12. Hill R (1948) A theory of the yielding and plastic flow of anisotropic metals. Proc R Soc London Math Phys Sci 193(1033):281–297.
doi:10.1098/rspa.1948.0045
13. Pagliaro P (2008) Mapping multiple residual stress components using the contour method and superposition. Ph.D. dissertation, Universitá
degli Studi di Palermo, Palermo
14. Bai Y, Wierzbicki T (2008) A new model of metal plasticity and fracture with pressure and lode dependence. Int J Plast 24(6):1071–1096
Chapter 50
Simulation of Triaxial Residual Stress Mapping for a Hollow Cylinder
Mitchell D. Olson, Wilson Wong, and Michael R. Hill
Abstract This paper describes a novel method to determine a two-dimensional map of the triaxial residual stress on a
radial-axial plane of interest in a hollow cylindrical body. With the description in hand, we present a simulation to validate
the steps of the method. The simulation subject is a welded cylindrical nozzle typical of a nuclear power pressurized water
reactor pressurizer; in the weld region, the nozzle inner diameter is roughly 132 mm (5.2 in.) and the wall thickness is
roughly 35 mm (1.4 in.). The pressure vessel side of the nozzle is carbon steel (with a thin stainless steel lining), the piping
side is austenitic stainless steel, and between the two are weld and buttering deposits of nickel alloy. Weld residual stresses in
such nozzles have important effects on crack growth rates in fatigue and stress corrosion cracking, therefore measurements
of weld residual stress can help provide inputs for managing aging reactor fleets. Nuclear power plant welds often have large
and complex geometry, which has made residual stress measurements difficult, and this work validates a new experimental
technique for measurements on welded nozzles.
Keywords Residual stress • Weld simulation • Dissimilar metal weld • Pressurizer safety/relief nozzle • Finite element
50.1 Introduction
Residual stresses in welds have an important effect on crack growth rates in fatigue and stress corrosion cracking. Residual
stress measurements can provide a distribution of residual stresses to be used as input for predicting crack growth rates as
well as for validation of weld simulation models. Components of interest are often large and have complex geometry that
makes measurement difficult.
Here we describe a method for determining the triaxial state of stress on a radial-axial plane of a cylindrical body, present
a simulation that validates the method, and define a set of experimental methods for follow-on experiments. Prime recently
introduced the contour method [1], which determines a two-dimensional distribution, or “map”, of residual stress normal to a
plane through a part. DeWald et al. [2] have applied the contour method to map hoop residual stress in a cylindrical nozzle as
a function of position in a radial-axial plane. Pagliaro et al. subsequently presented an extension of the contour method to
determine the triaxial stress on the contour plane [3], and the present work draws from that approach to develop triaxial stress
mapping for a cylindrical body.
50.2 Proposed Triaxial Stress Measurement Method
The proposed method consists of a series of section cuts and measurements to find the original triaxial state of stress on a
single radial-axial plane of interest in a cylindrical body. An outline of the section cuts and measurement steps appears in
Fig. 50.1. For each section cut there is an associated measurement of released stress and then a final measurement of the
M.D. Olson (*) • W. Wong • M.R. Hill

Department of Mechanical and Aerospace Engineering, University of California, One Shields Avenue, Davis, CA 95616, USA
e-mail: mitchelldolson@gmail.com
430 M.D. Olson et al.
Slice
Removal
D
Contour
Cut
C
Opening
Cut Released
Stress iii
B
Initial
Configuration
A Released
Stress ii
-
Plane of Released
Interest Stress i
Fig. 50.1 Measurement outline. Original stress ¼ released stress from the section cut (i) + remaining stress in configuration B. Remaining stress
in configuration B ¼ released stress from the contour cut (ii) + remaining stress in configuration C. Remaining stress in configuration C ¼
released stress from slice removal (iii) + remaining stress in configuration D. It follows that the original stress ¼ released stress from the section
cut (i) + released stress from the contour cut (ii) + released stress from slice removal (iii) + stress in slice (D)
remaining residual stress in a thin slice. The section cuts for the cylindrical body are: a section cut at 180 from the plane of
interest, which will be called the opening cut, a section cut at the plane of interest, called the contour cut, and a section cut to
remove a thin slice from the cylindrical body. Measurements would determine the released stress due to each section cut, and
a final measurement would determine the remaining residual stress in the slice.
After all measurement steps have been performed, the stress components from each step are superposed to find the
original triaxial stress state on the plane of interest in the cylinder. A theoretical justification of a similar technique has been
shown by Pagliaro et al. [3] and relies on Bueckner’s superposition principle [4]. According to superposition, the original
residual stress in a body is the sum of the released stress due to a cut and the stress remaining in the body after the cut, at
every point within the body. The outline of Bueckner’s superposition principle for this experiment is shown in Fig. 50.1,
which shows that the original residual stress in the nozzle at the plane of interest is the sum of the following: released stress
due to the opening cut, released stress due to the contour cut, released stress due to slice removal, and the remaining stress in
the nozzle slice. Defining the released stress in terms of the stress in each configuration of Fig. 50.1, we have
sri ¼ sA sB
srii ¼ sB sC
sriii ¼ sC sD (50.1)
where the sA, sB, sC, and sD are the stresses in the cylindrical body in configurations A, B, C, and D of Fig. 50.1, and sir,
siir, siiir are the released stresses at each step. It remains to determine whether measurements of sir, siir, siiir, and sD are
feasible.
50.3 Method Validation by Finite Element Modeling
In order to validate the method just described, we employ a weld model developed earlier by Fredette et al. [5]. That model is
a complete numerical simulation of a welding process for the safety/relief nozzle in a pressurized water reactor that connects
stainless steel piping to a carbon steel pressurizer. The model was implemented in commercial finite element software [6]
and provides a complete residual stress state for use in validating the proposed stress measurement method. The residual
stress from the weld simulation was used to create a series of three-dimensional welded nozzle models that match the
configurations A, B, C, and D of Fig. 50.1. The three-dimensional finite element models were created by revolving
50 Simulation of Triaxial Residual Stress Mapping for a Hollow Cylinder 431
Fig. 50.2 Dimensioned physical nozzle similar to the one used in this study
Fig. 50.3 Nozzle model diagrams (a) showing material boundaries and (b) angle of rotation, y
the original two-dimensional, axisymmetric weld-simulation model about an axial line of symmetry, replicating the original
stresses along specified radial-hoop planes, and then allowing the model to come to equilibrium. This work was enabled
using an available capability in the commercial software (called symmetric model generation) [6].
The cylindrical nozzle is approximately 362 mm (14.25 in.) long with a 132 mm (5.2 in.) inner diameter with a 35 mm
(1.378 in.) wall thickness (Fig. 50.2). The pressure vessel side of the nozzle is carbon steel (SA-508 Gr2), while the piping
side of the nozzle is stainless steel (316 SS). The weld was laid down in two phases, the first phase was a weld butter
consisting of 49 passes that was post stress relieved, and the second is a girth weld in a single V configuration having 9
passes. Both the weld and weld butter are made of nickel based weld filler metal (Alloy 182) (Fig. 50.3a). Further details
detailed regarding geometry, materials, and welding can be found in Ref. [5].
A series of models is used to simulate the measurement method of Fig. 50.1. All models use elastic material properties,
with elastic constants defined in the weld model and any effects of plasticity that may arise during sectioning are therefore
ignored. To simulate configuration A of Fig. 50.1, we use a half nozzle model with symmetry boundary conditions are
applied on two faces, y ¼ 0 and 180 (Fig. 50.3b), which replicates a complete nozzle. To simulate configuration B,
the boundary condition at y ¼ 180 is removed, and to simulate configuration C, the boundary condition at y ¼ 0 is also
removed. Configuration D is simulated by removing nearly all the elements in the half model, except those in a thin slice of
3.7 included angle (slice thickness 4.26 mm at the inner diameter). For each of the above configurations, the stresses at the
plane of interest were extracted for further analysis.
Once the stresses at each configuration were determined at the plane of interest, the released stresses at each step were
computed by subtracting the stresses in successive configurations, i.e., following equation Error! Reference source not
found.
50.4 Results
Figure 50.4 shows the released stress from the opening cut. Both axial and hoop released stresses are in the range of
10 MPa and the radial released stress is negligible (less than 0.2 MPa). Line plots of the data show that the axial released
stress is distributed linearly through the thickness at a given axial position and that the hoop released stress has a hyperbolic
distribution that is offset by a constant at a given axial position (i.e., both axial and hoop released stresses are a combination
of uniform and bending fields, as may be expected given the large distance from the plane of interest to the opening cut).
The released stress from the contour cut is shown in Fig. 50.5, and the stress release is triaxial, as noted earlier by Pagliaro
et al. [3]. Both the axial and hoop released stresses are in the range of 400 MPa, and the radial released stress is in the range
of 250 MPa. The hoop stress release is complete, so that no further hoop stress release will occur in subsequent steps.
Figure 50.6 shows that the in-plane released stresses due to slice removal are a significant portion of the superposed total,
in the range of 400 MPa for axial, and 200 MPa for radial released stress (the hoop released stress is zero because the
plane of interest is normal to the hoop direction).
The remaining in-plane stresses in the slice are shown in Fig. 50.7, which shows that axial stress is in the range of
200 MPa and radial stress is significantly smaller, in the range of 100 MPa.
Figure 50.8 shows contour plots of the original weld-simulation stresses compared with superposition totals. The contour
plots match very well, and confirm the validity of the proposed measurement method.
a 30 b 30
20 20
(mm)
(mm)
10 10
0 0
−30 −20 −10 0 10 20 30 40 −30 −20 −10 0 10 20 30 40
(mm) (mm)
−10 0 10 −10 0 10
Stress (MPa) Stress (MPa)
Fig. 50.4 Released stress from first section cut at the plane of interest: (a) axial stress, (b) hoop stress; the radial released stress, not shown, is
negligible
a 30 b 30
20 20
(mm)
(mm)
10 10
0 0
−30 −20 −10 0 10 20 30 40 −30 −20 −10 0 10 20 30 40
(mm) (mm)
−400 −300 −200 −100 0 100 200 300 400 −200 −150 −100 −50 0 50 100 150 200
c 30
20
(mm)
10
0
−30 −20 −10 0 10 20 30 40
(mm)
−400 −300 −200 −100 0 100 200 300 400

Stress (MPa)
Fig. 50.5 Released stress due to the contour cut: (a) axial, (b) radial, and (c) hoop
a b
30 30
20 20
(mm)
(mm)
10 10
0 0
−30 −20 −10 0 10 20 30 40 −30 −20 −10 0 10 20 30 40
(mm) (mm)
−400 −300 −200 −100 0 100 200 300 400 −200 −150 −100 −50 0 50 100 150 200
Fig. 50.6 Released stress due to slice removal at the plane of interest: (a) axial and (b) radial
a b
30 30
20 20
(mm)
(mm)
10 10
0 0
−30 −20 −10 0 10 20 30 40 −30 −20 −10 0 10 20 30 40
(mm) (mm)
−200 −150 −100 −50 0 50 100 150 200 −200 −150 −100 −50 0 50 100 150 200
Fig. 50.7 Stresses remaining in the slice: (a) axial and (b) radial
a
30 30
20 20
(mm)
(mm)
10 10
0 0
−30 −20 −10 0 10 20 30 40 −30 −20 −10 0 10 20 30 40
(mm) (mm)
−400 −300 −200 −100 0 100 200 300 400

Stress (MPa)
b 30 30
20 20
(mm)
(mm)
10 10
0 0
−30 −20 −10 0 10 20 30 40 −30 −20 −10 0 10 20 30 40
(mm) (mm)
−200 −150 −100 −50 0 50 100 150 200

Stress (MPa)
c
30 30
20 20
(mm)
(mm)
10 10
0 0
−30 −20 −10 0 10 20 30 40 −30 −20 −10 0 10 20 30 40
(mm) (mm)
−400 −300 −200 −100 0 100 200 300 400

Stress (MPa)
Fig. 50.8 Triaxial residual stress at the plane of interest from weld process model (left) and proposed measurement (right), (a) axial, (b) radial,
and (c) hoop
50.5 Recommended Experimental Approach
The results in Figs. 50.4, 50.5, 50.6, and 50.7 are useful in supporting selection of experimental methods for
measurements of sir, siir, siiir, and sD. For the released stress due to the opening cut, it would be possible to measure
strain on the inner and outer surfaces at the plane of interest using metallic foil strain gages. Given strain change at the
inner and outer surfaces at a set of axial positions, corresponding axial and hoop stress fields through the thickness can
be computed using uniform and bending distributions. The weld simulation used here suggests that sir, is small, less
than 2.5 % of the total stress, but that may not be the case for a different weld procedure or configuration. The proposed
method was developed with the contour method in mind for measuring the released stress due to the contour cut, siir,
and the contour method is probably the simplest technique for this step. The released stresses due to slice removal could
be measured using either an array of strain gages or digital image correlation [7] and elastic stress–strain relations.
Remaining stresses in the slice could be measured using a variety of measurement techniques, including hole drilling
[8], slitting [9], neutron diffraction [10], or x-ray diffraction [11]. In general, each measurement technique will be able
to resolve the expected stress magnitude (200 MPa), but each has practicalities to consider. For example, the welded
nozzle considered in this study will have significant grain size, composition, and lattice parameter variations, which may
cause problems for diffraction techniques. Neutron diffraction will average stresses over the size of a gage volume,
typically 2–5 mm along one dimension, with smaller gage volumes having smaller averaging but requiring increased
measurement time. Hole drilling will average the stress over the hole depth (typically 1 mm), whereas slitting will
average stress over the entire thickness. Slitting would also provide only the axial component of stress, thus rendering
the method a biaxial stress measurement technique (axial and hoop). One note of caution is the reliance of the present
technique on superposition and therefore also upon elastic material behavior; plasticity occurring during any step in the
sequence of Fig. 50.1 will result in systematic measurement error.
Two additional investigations were performed to determine the effects arising from practicalities of the above measure-
ment techniques. The first was a simulation of the contour method measurement. A contour measurement comprises
measurement of out-of-plane displacements at the cut plane, imposing those out-of-plane displacements as a boundary
condition on an elastic (stress free) model of the sectioned part (here, the half-nozzle), and determining the triaxial stress at
the plane of interest. This is different than the process used above, where we took the difference between a model with
(configuration B) and without (configuration C) a boundary condition on the plane of interest. To determine the effect
of using only the out-of-plane displacements, the contour simulation took the out-of-plane displacements from the model of
configuration C on the plane of interest and applied them as a boundary condition on a stress-free half model. The difference
between the stress due to the imposed displacements and the released stress of Fig. 50.5 are less than 1% of the peak stresses,
so that there is no systematic error in determining siir using the contour method.
A second investigation was performed to assess to effect of using a thickness-average of residual stress in the slice for sD,
rather than the surface stress. A through-thickness average of stress in the three-dimensional slice model was computed and
compared with the surface stresses reported in Fig. 50.7. The difference between the average and surface stress is shown in
Fig. 50.9. While differences are quite small in most locations, there are significant differences localized near material
interfaces with values as large as 7% of peak axial stress and 20% of peak radial stress.
a b
30 30
20 20
(mm)
(mm)
10 10
0 0
−30 −20 −10 0 10 20 30 40 −30 −20 −10 0 10 20 30 40
(mm) (mm)
−40 −30 −20 −10 0 10 20 30 40 −40 −30 −20 −10 0 10 20 30 40

Fig. 50.9 The difference between the through-thickness averaged stress in the slice and the surface stress at the plane of interest: (a) axial and
(b) radial. Note: the color scale is 10% of the full range of the superposed axial stress and 20% of the full range of the superposed radial stress
50 Simulation of Triaxial Residual Stress Mapping for a Hollow Cylinder 435
50.6 Conclusions
The present simulation shows that the proposed triaxial measurement sequence of Fig. 50.1 has no inherent systematic error.
However, in applying the method, one needs to be aware that for each measurement step there will be uncertainty, and it
follows that the superposed total will suffer from the combined uncertainties of each measurement step. Thus, the
practitioner should make efforts to minimize uncertainties at each of the steps.
Acknowledgements The Electric Power Research Institute, Materials Reliability Program (Paul Crooker, Senior Program Manager) provided
financial support for this work. During the course of the work the first author (Olson) was supported by the Los Alamos National Laboratory
Summer Program (Mike Prime, mentor). The authors also acknowledge: Mike Prime (Los Alamos National Lab) for helpful discussions; Matthew
Kerr (Nuclear Regulatory Commission) for guidance in making use of the weld simulation model; Adrian DeWald (Hill Engineering, LLC) for
information related to the contour method and how it was implemented on the physical nozzle of Fig. 50.2.
References
1. Prime MB (2001) Cross-sectional mapping of residual stresses by measuring the surface contour after a cut. J Eng Mater Technol 123:162–168
2. DeWald AT, Hill MR, Willis E (2011) Measurement of welding residual stress in dissimilar metal welds using the contour method. In:
Proceedings of the ASME 2011 pressure vessels & piping division conference, Baltimore. Paper 57720
3. Pagliaro P, Prime MB, Robinson JS, Clausen B, Swenson H, Steinzig M, Zuccarello B (2010) Measuring inaccessible residual stress using
multiple methods and superposition. Exp Mech 51(7):1123–1134
4. Bueckner HF (1973) Field singularities and related integral representations. In: Sih GC (ed) Mechanics of fracture. Noordhoff International
Publishing, Leiden, pp 239–314
5. Fredette LF, Broussard JE, Kerr M, Rathbun HJ (2011) NRC/EPRI welding residual stress validation program – phase III details and findings.
In: Proceedings of the ASME 2011 pressure vessels & piping division conference, Baltimore. Paper 57645
6. Abaqus/Standard, version 6.10, Simulia, Inc., Providence, 2010
7. Pan B, Qian K, Xiw H, Asundi A (2009) Two-dimensional digital image correlation for in-plane displacement and strain measurement: a
review. Meas Sci Technol 20(6):1–17
8. Schajer GS, Stenzig M (2005) Full-field calculation of hole drilling residual stresses from electronic speckle pattern interferometry data. Exp
Mech 45(6):526–532
9. Prime MB, Hill MR (2006) Uncertainty analysis, model error, and order selection for series-expanded, residual stress inverse solutions. J Eng
Mater Technol 128(2):175–185
10. International Organization for Standardization – Technical Specification (2005) Non-destructive testing – standard test method for determining
residual stresses by neutron diffraction. ISO/TS 24432
11. European Committee for Standardization – Technical Specification (2002) Non-destructive testing – x-ray diffraction for polycrystalline and
amorphous material – part 1: general principles. EN 13925–1

Experimental and Applied Mechanics Volume 4 PDF

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

Experimental and Applied Mechanics Volume 4 PDF

Enviado por

Direitos autorais:

Formatos disponíveis

Conference Proceedings of the Society for Experimental Mechanics Series

For further volumes:

Experimental and Applied Mechanics,

ISSN 2191-5644 ISSN 2191-5652 (electronic)

# The Society for Experimental Mechanics, Inc. 2013

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

Vancouver, BC, Canada Carlos E. Ventura

16 Use of Fiber Bragg Grating Strain Gages on a Pipeline Specimen Repaired

35 Absorbed Versus Released Energy in the Cracking Process

Trenton Carpenter, Cody Ray, and Roberto Albertani

T. Carpenter (*) • C. Ray • R. Albertani

1.2 Experimental Setup

Table 1.1 Wind tunnel test matrix

1.3 Force Estimation from Vorticity

1.3.1 Surface Reconstruction

1.3.2 Numerical Fluid Model

1.3.3 Lift from Circulation

Fig. 1.5 Vorticity contours of

Fig. 1.6 Contour map

1.4 Force Estimation from Poisson Model

1.5 Results and Discussion

1.5.1 Aerodynamic Loads

1.5.2 Surface Reconstruction

Table 1.2 Strain sensor configuration

1.5.3 Material Characterization

1.5.4 Lift Calculation

1.7 Future Work

Adam Kammers and Samantha Daly

Keywords Deformation mapping • Digital image correlation • Scanning electron microscopy

2.2 Small Scale Patterning

A. Kammers (*) • S. Daly

2.3 Image Distortion Correction

2.4 Stress Relaxation

Fig. 2.6 Surface showing

Fig. 2.7 The u displacement

Martha E. Grady, Mark D. Losego, Paul V. Braun, and Nancy R. Sottos

Keywords Self-assembled monolayers • Laser spallation • Interfacial adhesion

M.E. Grady (*)

Table 3.1 List of chemicals used to prepare self-assembled monolayers

3.2 Specimen Preparation

3.2.1 Self-Assembled Monolayers

3.2.2 Transfer Printing

3.3 Experimental Protocol

V max þ V min V max V min

Table 3.2 Material properties used to calculate substrate stress

3.4 Results and Discussion

1. Cahill DG et al (2003) Nanoscale thermal transport. J Appl Phys 93:793–818

J.P.M. Hoefnagels, C.C. Tasan, F.J. Peters, and M.G.D. Geers

Keywords DP steel • Ductile damage • Damage characterization • Strain path dependence

4.2 Experimental Procedure, Results, and Discussion

4.2.1 Model Microstructures

J.P.M. Hoefnagels (*) • F.J. Peters • M.G.D. Geers

4.2.2 In-situ Marciniak Experiments

4.2.3 Relevant Damage Mechanisms

4.2.4 Quantitative Analysis of Damage Mechanisms

4.2.5 Quantitative Evolutions of the Damage Mechanisms; Preliminary Results

L.A. Liew, D.T. Read, and N. Barbosa III

5.2.1 Specimen Design and Gripping

Fig. 5.1 Essential design

Fig. 5.2 (a) and (b): Details

Fig. 5.3 Schematic of the

5.2.2 Fatigue Test Protocol