Você está na página 1de 179

NOTE TO USERS

This reproduction is the best copy available.

1*1

OfCrnada

Nationai Library

Bibliothque nationale
du Canada

Acquisitions and
Bibliographie Services

Acquisitions et
services bibliographiques

395 Wellington Street


OttawaON K1A ON4

395. nie Wellington


Otiawa ON K I A O N 4

Canada

Canada

The author has granted a nonexclusive licence allowing the


National Library of Canada to
reproduce, loan, distrihte or sell
copies of this thesis in microfoxq
paper or electronic formats.

L'auteur a accord une licence non


exclusive permettant la
Bibliothque nationale du Canada de
reproduire, prter, distribuer ou
vendre des copies de cette thse sous
la forme de microfiche/^ de
reproduction sur papier ou sur format
lectronique.

The author retains ownership of the


copyright in this thesis. Neither the
thesis nor substantial extracts fiom it
may be printed or othemise
reproduced without the author's
permission.

L'auteur conserve Ia proprit du


droit d'auteur qui protge cette thse.
Ni la thse ni des extraits substantiels
de celle-ci ne doivent tre imprims
ou autrement reproduits sans son
autorisation.

Aspects of Across-Wind Loads and Effects on

Large Reinforced Concrete Chimneys

Jon Kenneth Galsworthy

Graduate Program in Engineering Science


Department of Civil and Environmental Engineering

Subrnitted in partial fulfihent


of the requirements for the degree of
Doctor of Philosophy

Faculty of Graduate Studies


The University of Western On tario
London, Ontario
April3000

O Jon Kenneth Galsworthy 2000

ABSTRACT

Research on the across-wind loads and effects on concrete chimneys has progressed to
the point where reliable predictions for isolated, plain chimneys can now be made. This
is not the case, however, for grouped chimneys and chimneys where the cross-section is
not circufar. In the former case, the wind loads are affected by the presence of one or
more upstream chimneys. In the latter, wind tunnel studies can be used for sharp edged
cross-sections, but when relatively slight modifications are made to the cross-section, the
impact on the aerodynamics is not known. The present work is based on these two areas.

The work begins with a brief review of cylinder aerodynamics and a review of the work
of Vickery and CO-workerson the prediction of the across-wind response for chimneys.

The first part of the original work deals with how the response to across-wind loads is

affected by foundation flexibility. In particular, for chimneys with freestanding liners,


the interaction between the liner and concrete shell increases the relative deflections and
introduces wind-induced stresses in the liner.

A soil-structure interaction model is

introduced that accounts for the flexibility of the foundation and predicts field
observations.

A field and model study of wind loads on pairs of chimneys or cylinders is examined

next.

The field results are summarized.

The mode1 study, conducted at Re = 106

compared to Re = 107 full-scale, covers three diameter ratios and covers strearnwise

separation ratios between three and six diameters and transverse separation ratios of up to
1.5. The effect of the upstream cyiinder on the across-wind loading for the downstream

cylinder is examined.

The Iast part of the work deals with a slight modification from the circular cross-section
and how this affects the aerodynamic Ioads and aeroelastic effects. Results for a wind
tunnel study at ~e=5x10'are used to make post mortem predictions for a chimney that
experienced an increase in the across-wind motions by a factor of ten.

Acknowledgements
1 would like to express my gratitude to my chief advisor, Dr, B.J. Vickery for his

continued support and encouragement throughout the duration of this study and also for
the opportunity to travel to Australia.
1 would also like to thank the members of the Boundary Layer Wind Tunnel Laboratory

for their help with the experiments and full-scale data acquisition; particularly Mr. G.
Dafoe and Mr. D.P. Momsh.
During my time at Monash University, Australia, Dr. J. Holmes provided much
assistance with the experimental program and Dr. W.H. Melbourne was more than
generous in providing access to the facilities. 1 thank both.
The financial support of the Natural Sciences and Engineering Research CounciI and the
Boundary Layer Wind Tunnel Laboratory is gatefully acknowledged.

1 am also thankful to the many friends and colleagues at Western during my tirne as a
graduate student for the enlightening discussions and different perspectivesFinally, words cannot describe the emotional support and encouragement provided by rny
family and my fiance Ingrid. 1 wilI always be grateful.

Table of Contents
Page

TITL.E PAGE
CERTIFICATE OF EXAMINATION
ABSTRACT
ACKNOWDGEMENTS
TABLE OF CONTENTS
LIST OF TAEiLES
LIST OF FIGURES
NOMENCLATURE
CHA.E'TER 1 - INTRODUCTION

1.1 Reinforced Concrete Chimneys


1.2 Motivation and Preview of Thesis

CHAPTER 2 - LI'IEMTURE REVlEW

2.1 Introduction
2.2 Aerodynamics of Circular Cylinders
22.1 Effects of Reynolds Number
2.2.2 Effects of Free Stream Turb~ilence
2.3 Across-Wind Response Model for Chimneys
2.3.1 Vortex Shedding Forces on a Stationary Cylinder
2.3.2 Forces Induced by Motion
2.3.3 Response Predictions
2.4 Surnmary

CHAPTICER 3 - EFFECT OF FOUNDATION ~ X I B I L I T YON T l E ACROSSWl[ND RESPONSE OF R/C CHIMNEYS WITH


FREESTANDING LINERS
29
3.1 Introduction
3.2 Soil-Structure Interaction
3.2.1 Foundation Model
3-3.1.I S hdlow Foundation
3 -2.L .2 Deep Foundation
3.2.2 Structure Model
3.2.3 Foundation Structure Model
3.3 Darnping
3.4 Response of Chimney with Flexible Foundation
3 -5 S ummary and Conclusions

CHAFTER 4 - FULL AND MODEL SCSTUD1.S OF ACROSS WIND


BUFFETING LOADS ON PAIRS OF CIRCULAR CYLINDERS 57
4-1Introduction
4-2 Field Study
4.2-1 Free Vibration Analysis
4 - 2 2 Prediction of Maximum Response
4.2.3 Predictions for an Isolated Chimney
4-3 Model Study
4.3.1 Experiment Setup
4.3.1-1 Flow Propeaies
4.3-2 Results for a Single Cylinder
4.3 -3 Results from Two Cylinder Configurations
4.3.3.1 Case 1: 630mm - 500mm Cylinders
4.3.3.2 Case 2: 5 0 0 m - 5 0 0 m Cylinders
4.3.3.3 Case 3: 400mm - 500mrn Cylinders
4.4 Discussions and Cornpanson of Results
4.5 Summary

CEWPTER 5 - STUDY OF 'IFE EFFECT OF APPURmNANCES ON TEE


ACROSS-WIND RESPONSE OF CNIMNEYS
5.1 Introduction
5.2 Wind Tunnel Siudy
52.1 Flow Properties
5 - 2 2 Surface Pressure and Force Measurements
5 -2.3 Aerodynamic Darnping Measurements
5.3 Prototype Predictions
5.4 Field Study ResuIts
5 -4.1 Modal Properties
5.4.2 Field Observations
5.5 Further Aerodynamic Investigations
5.6 Surnrnary and Conchsions

CHAPTER 6 - S-Y

AND CONCLUSIONS

6.1 S urnrnary of Field Observations


6.2 Effect of Foundation Flexibility
6.3 Summary of Model Results
6.3.1 Pairs of Circular Cylinders
6-3-2Effect of Appurtenances
6.4 Conclusions
6.5 Recornrnendations for Further Study
REFERENCES

vii

List of Tables
Table 3.1 :
Table 3-2:

Foundation and Aerodynamic Damping for Deep Foundation


Foundation and Aerodynamic Damping for S h d o w Foundation

Table 4.1:

Across-wind response mode1 parameters for isolated chimney

viii

Page
45
45

67

List of Figures
Figure 2.1:
Figure 2.2:
Figure 2.3:
Figure 2.4:
Figure 2.5:
Figure 2.6:
Fi,oure 2-7:
Fi,aure 2.8:
Figure 2.9:
Fi,oure 2-10:
Figure 2.1 1:
Figure 2.12:
Figure 2-13:
Figure 2.14:

Disturbed flow field around circular cylinder


Mechanism of separation and vortex formation
Flow field near the separation point
Definition of different flow regrnes in tenns of drag coefficient
Mean drag coefficient as a function of Re for smooth cylinders
Compilation of RMS lift coeffrcient for srnooth cylinders
Compilation of St for smooth cylinders
Spectrum of fifi coefficient for smooth cylinders at high Re
Effect of free stream turbulence on mean drag coefficient
Effect of free stream turbulence on RMS lift coefficient
Spectrum of lift coefficient on mode1 tapered cylinder
Full-scale RMS lift coefficient for concrete chirnneys
Mass-darnping parameter for small amplitudes
Measured and fitted response

Figure 3.1a:
Figure 3.1b:
Figure 3 2 :
Figure 3.3:
Figure 3 -4:
Figure 3.5:
Figure 3.6:
Figure 3.7:
Figure 3.8:
Figure 3.9:
Figure 3.10:
Figure 3.1 1:
Figure 3.12:
Figure 3.13:
Figure 3.14:
Figure 3-15:
Figure 3.16:

End-Bearing Foundation
Shallow Foundation
Model for Soi1 Structure Interaction
Mode Shape for Liner Mode
Mode Shape for Shell Mode
Modal Aerodynamic Damping (Shell Mode) for Deep Foundation
Peak Shell Deflection for Deep Foundation
Peak Liner Deflection for Deep Foundation
Peak Relative (Shell - Liner) Deflection for Deep Foundation
Peak Shell Base Moment for Deep Foundation
Peak Shell Base Shear for Deep Foundation
Peak Liner Base Moment for Deep Foundation
Peak Liner Base Shear for Deep Foundation
Normalized Deflection for Deep Foundation
Normalized Deflection for S hallow Foundation
Normalized Base Forces for Deep Foundation
Normalized Base Forces for Shallow Foundation

Figure 4.1 :
Figure 4.2:
Figure 4.3:
Figure 4.4:
Figure 4.5:
Figure 4.6:
Figure 4.7:
Figure 4.8:
Figure 4.9:
Fi,oure 4-10:

Site plan for the field study


Spectrum of acceleration of the liner at 170m
Spectnim of acceleration of the shell at 170m
Mode shape for the liner mode
Mode shape for the shell mode
Model used for extrapolating to maximum response
X-Y plot of shell acceleration
Variation of peak shell deflection with wind direction
Predictions for isolated stack
Schematic of the 1MW wind tunnel at Monash University

Page
7
8
11
12
13
14
15
15

16
17
19

21
24
25
31
31
38
42
42
44

48
48
49

50
50
51
51
53
53
54

54
59

60
60
61
62
64

65
65
67
69

Figure 4.11:
Fiame 4.12:
Figure 4.13:
Fiowe 4.14:
Figure 4.15:
Fiame 4.16:
Figure 4.17:
Fiwre 4.18:
Figure 4.19:
Figure 4.20:
Figure 4.2 1:
Figure 4.22:
Figure 4.23:
Figure 4.24:
Figure 4.25:
Figure 4-26:
Figure 4.27:
Figure 4.28:
Figure 4.29:
Figure 4.30:
Figure 4.3 1:
Figure 4.32:
Figure 4.33:
Figure 4.34:
Figure 4.35:
Figure 4.36:
Figure 4.37:
Figure 4.38:

Photo of experimentai setup


Experimental setup for mode1 study
Frequency response of tubing
Phase response of tubing
Spectrum of longitudinal turbulence
Horizontal profile of flow properties
Vertical profile of flow properties
Dependence of force coefficients and St on Re
Spectrum of lift coefficient for a single cylinder at Re = 106
Spectrum of downstrearn cylinder lift coefficient for X/Duc = 3.17
and Y/Duc = 0.
Spectrum of differential pressures for upstream cylinder for
X/DUc=3.17 and Y/Duc=O.
= 4.75
Spectrum of downstrearn cylinder lift coefficient for
and Y/Dvc = 0.
Spectrum of differential pressures for upstream cylinder for X/DUc=
4.75 and Y/DUc=O.
S, (downstrearn cyIinder) for al1 630mm-500mm configurations
Mean base pressure coefficient (upstream cylinder) for d l 630mmSOOmrn configurations
RMS lift coefficient (downstream cylinder) for al1 630mm-500mm
configurations
Mean drag coefficient (downstream cylinder) for al1 630rnm-500rnm
configurations
Spectnun of downstream cylinder lift coefficient for X/Duc = 3.0
and Y/Duc = O
Spectrum of differential pressures for upstream cylinder at XIDUC=
3.0 and Y/Duc = 0.
Spectrum of downstream cylinder lift coefficient for X/Duc = 4.5
and Y/Duc = O
Spectrum of differential pressures for upstrearn cylinder at X/Duc =
4.5 and Y/Duc = 0.
Spectrum of downstrearn cylinder lift coefficient for X/Duc = 6.0
and Y/Duc = O
Spectrum of differential pressures for upstrearn cylinder at X/Dvc =
6.0 and Y/DUc = 0.
S, (downstream cylinder) for al1 500mm-500mrn configurations
Mean base pressure coefficient (upstream cylinder) for al1 500mm500mm configurations
RMS lift coefficient (downstrearn cylinder) for al1 500mrn-500rnrn
configurations
Mean drag coefficient (downstream cylinder) for al1 500rnm-500mm
configurations
Spectrum of downstrearn cylinder lift coefficient for X/Duc = 3.75
and Y/Duc = O

Figure 4-39: Spectrum of differential pressures for upstream cylinder at X/Duc =


3.75 and YDuc = O
Figure 4.40: Spectrum of downstrcam cylinder lift coefficient for X/Duc = 5.0
and Y/Duc = 0.
Figure 4.41: Spectrum of downstream cylinder lift coefficient for XDuc = 6-35
and Y/Dvc = O
Figure 4.42: S, for lateral buffeting and vortex shedding for al1 400mm-500mm
configurations
Fiame 4 4 3 : RMS lifi coefficient associated with lateral buffeting
Figure 4.44: Mean drag coefficient (downstream cylinder) for ail 400mm-500mm
confi,wations
Figure 4.45: Amplification of lift forces for al1 630m.m-500mm configurations
Figure 4.46: Amplification of lift forces for al1 500mm-500mm Configurations
Figure 5.1:
Figure 5.2:
Figure 5.3:
Figure 5.4:
Figure 5.5:
Figure 5.6:
Figure 5-7:
Figure 5.8:
Figure 5-9:
Figure 5.10:
Figure 5.1 1:
Figure 5.12:
Figure 5.13:
Figure 5.14:
Figure 5.15:
Figure 5.16:
Figure 5.17:
Figure 5.18:
Figure 5.19:
Figure 5.20:
Figure 5.21:
Figure 5.22:
Figure 5.23:
Figure 5-24:
Figure 5.25:
Figure 5.26:
Figure 5.27:
Figure 5.28:
Figure 5.29:

View of Cod Creek Power Plant


Unit#2chimney
Experiment set up in BLWTL II at UWO.
Distributionof mean wind speed across one ce11 of a uniform grid
Turbulence intensity downstream of a uniform mgid
Vertical profile of turbulence intensity and mean speed 10M
downstrearn of @d
Spectrum of lift coefficient for a = O
Mean pressure distribution for a = O
Spectrum of lift coefficient at a=45
Mean pressure distribution at c(=45
Sketch of flow pattern around cylinder
Spanwise correlation of lift force
Spectrum of lift coefficient for -70
Mean pressure distribution at a=70
Spectrum of lift coefficient at a=90 '
Mean pressure distribution at a=90 "
Mean lift and drag coefficients for the larger enclosure.
Mean lift and ciras coefficients for the smaller enclosure.
RMS lift coefficient for the larger enclosure.
RMS lift coefficient for smaller enclosure.
Strouhal number for large enclosure.
Strouhal number for smdler enclosure.
Mass-damping pararneter for a d 5
Mass-damping pararneter for c(=67.5 "
Mass-darnping pararneter for a=90 "
Peak across-wind response of prototype for a=45
Peak across-wind response of prototype for a=67.5
Peak across-wind response of prototype for a=9O0
Peak across-wind response of prototype without encIosure.

108
109

128
128

139
13 1

Figure 5.30: Peak across-wind response for the prototype with enclosure over
bottom half.
Figure 5.3 1: Mode shape for Iiner mode for Unit #L chimney
Figure 5-32: Mode shape for shell mode for Unit #1 chimney
Fiawe 5.33 : Mode shape for liner mode for Unit #2 chimney
Figure 5.34: Mode shape for sheil mode for Unit #2 chimney
Figure 5.35: Spectrum of shell displacement for peak tip amplitude of 2.5 cm
Figure 5.36: Spectrum of shell displacement for peak tip amplitude of 13 cm
Figure 5.37: Summary of observations from the field study
Figure 5.38: Mean force coefficients for outstand dimension of 0-OSD
Figure 5.39: Mean force coefficients for outstand dimension of 0.05D
Figure 5.40: Mean force coefficients for outstand dimension of 0.01D
Figure 5.41 : RMS lift coefficients for outstand dimension of 0.08D
Figure 5.42: RMS lift coefficients for outstand dimension of O.05D
Figure 5.43 : RMS lift coefficients for outstand dimension of 0.0 1D
Figure 5.44: Strouhal number for outstand dimension of O-OSD
Figure 5.45: Strouhd number for outstand dimension of 0.05D
Figure 5.46: Strouhal number for outstand dimension of 0.0 1D

Nomenclature
Chapter Two

bandwidth parameter

CD

drag coefficient

CL

lift coefficient

CLV

lift coefficient associated with vortex shedding

CO(^,, z2,f ) CO-spectrumof lift force at elevation zl and a


O

cylinder o r chimney diarneter

dp
-

a.

pressure gradient on surface of circular cylinder

f;

natural frequency for mode i

f,

vortex shedding frequency (Hz)

aerodynamic stiffness coefficient

1,

longitudinal turbulence intensity

.=

rnodified turbulence intensity

ratio of mean wind speed to critical speed for vortex shedding

aerodynamic damping coefficient (equivalent viscous damping)

Ka

=rn4
p a ~ ' mas-damping (aerodynamic) parameter
mass-damping parameter for smail amplitudes

Kao

Ks
Ki

=PD'

rnass-damping (structural) pararneter


generalized stiffness in mode i
integral Iength scale of longitudinal turbulence

correlation length in diameters

rn

mass per unit Iength of the structure

Mi
Ma

generalized rnass for mode i


Mach nurnber

where

&=~q-z2l

Reynolds number
spectrum of Iift force per unit length at fiequency f
spectrum of generalized lifi force

Strouhal number
local incident mean wind speed
u

f,D

-2-

cnt

critical wind speed for vortex shedding


incident and local mean flow velocity
Iift force per unit length associated with vortex shedding
motion-induced force per unit length
mode shape
across-wind displacernent
elevation from grade on a chimney
modal coordinate in mode 1
lirniting displacement in aerodynamic damping model
aspect ratio of a chimney (height/diarneter)
kinematic viscosity of a fluid
mass density of a fluid
aerodynamic darnping as a fraction of critical
structural darnping as a fraction of criticai
roo t-mean-square (RMS) quantity
time-averaged (mean) quantity
differentiation with respect to time

xiv

dimensionless frequency
the coefficient in eigenvector

damping constant
normalized damping constant
foundation damping coefficient for rocking
foundation damping coefficient for translation
foundation damping coefficient for cross tems
flexural rigidity of the shell
natural frequency of mode 1
soil shear rnodulus
soil layer thickness
height of the chimney

mass moment inertia of the foundation


m e stiffness constant
normaiized stiffness constant
static stifiess
static horizontal stiffness
static vertical stiffness of the single pile
cornplex stiffness associated with direction i
vertical group stiffness
horizontal group stiffness
rocking group stiffness
coupling group stifmess
complex vertical impedence function of the single pile

element of the shell stiffness matrix


shell stifiess matrix
liner stiffness matrix
m a s of the foundation
mass per unit height of the shell
mass moment of inertia about base
mass of system
mass cross tenn
mass rnatrix cross term for foundation translation and shell flexural
motions
mass matnx cross tenn for foundation translation and liner flexural
motions
element of the mass matrix
mass matrix cross term for foundation rocking and shell flexural motions
mass matrix cross tenn for foundation rocking and liner flexural motions
shell mass matrix
liner mass matrix
radius of circular foundation
pile spacing to diameter ratio
shear wave velocity of the soi1
Lysmer analog velocity
height above datum
dynarnic interaction factor
vertical interaction factor
horizontal interaction factor
complex interaction factors between piles i and j

xvi

complex interaction coefficients for the honzontaf translation and rotation


material damping ratio of the soi1
unit weight of the soil
total damping in each mode
angle between the direction of load action and the plane in which piles lie,
foundation rotation
foundation coefficients for

fh

mode

mass density of soi1


Poisson's ratio of the soil
mode shape

i" shape function


curvature for shape function y i
rigid body shape function corresponding to rotation at the foundation level
rigid body shape function corresponding to translation at the foundation
first vertical natural frequency (rads) of a homogeneous soil layer
first horizontal natural frequency of a homogeneous soi1 Iayer

Chapter Four
bandwidth parameter
constant for a @en chimney

RMS Iift coefficient


drag coefficient at time t
lift coefficient at time r
pressure coefficient with respect to angle and time
corrected mean drag

RMS lift force coefficient

xvii

mean base pressure coefficient


RA& lift coefficient due to lateral buffeting
characteristic diameter

elastic modulus of reinforced concrete


elastic moduIus of brick
central shedding frequency
natural frequency of Liner
natural frequency of liner
longitudinal turbulence intensity
modified intensity
fraction of the total variance under the peak
rotational stiffness of the foundation
aerodynamic damping
correlation length
longitudinal integral lerigth scale
ratio of the wind speed to the critical wind speed
reduced mass
generalized mass
generalized mass of liner
generalized mass of liner
pressure with respect to angle and time
static pressure
solid blockage ratio
Strouhal number

xviii

StrouhaI number
spectnim of longitudinal velocity
mean wind speed
wind speed
critical wind speed
reduced frequency
strearnwise separation
transverse separation
roughness length
damping
structural damping
aspect ratio of the chimney
density of air
liner motions in the liner mode
shell motions in the liner mode
liner motions in the shell mode
shell motions in the shell mode

RMS tip deflection at a particular wind speed


RMS tip deflection at the critical speed
F U S velocity

Chapter Five
b

bar size

bandwidth parameter

c
CL
CD

corrected value
mean lift coefficient
mean drag coefficient

xix

CD,

corrected mean drag

CL=

CP,

RMS lift force coeffkients


rnean pressure coefficient

CL

RMS lift coefficient due to shedding

characteristic diarneter
brick modulus
reinforced concrete elastic moduIus
frequency
central shedding frequency
naturd frequency of liner
naturai frequency of liner
longitudinal turbulence intensity
aerodynamic darnping
rotational stifiess of the foundation
correlation length
mass per unit length
mesh size
generaiized mass of liner
generalized rnass of liner
coherence function
solid blockage ratio
Strouhal number
rnean wind speed
angle of attack

&iP

peak deflection
mean deflection

RMS deflection

6.

aerodpamic damping as a fraction of cntical

density of air

xxi

Chapter One
Introduction
---

1.1 Reinforced Concrete Chimneys


In the 196OYs,governments started to regulate the level of ground concentrations
resulting from emissions from fossil fuel burning power plants and refineries. With this
regulation came the need for the construction of very ta11 chimneys ranging from 300 to
400m in height. More recently, governrnents and replation agencies have acted to limit
the toxicity of the emissions and scnibbers now remove much of the noxious effluent.

Large reinforced concrete chimneys are still required but typical heights are now 150 to
250m.
The need for the 'super tall' chimneys provided a catalyst for research into the
understanding of the wind effects on freestanding slender structures.

Thanks to the

contributions of Davenport (1962) and Vickery (1969) the alonpwind loads and effects
due to atmospheric turbulence were quantified using a rational approach based on random
vibration theory and extrerne value statistics. Another valuable contribution that allowed
Davenport's approach to be used in making predictions and design calculations was made

by Roshko (1961) when he measured the mean drag coefficient on a circular cylinder at
Reynolds nurnbers (-107) typicd of the full-scale chimneys.
Across-wind loads and effects, however, were slower in development. Early approaches
to the vortex-induced response of chimneys were rather simplistic and treated the loading

as a sinusoidal force (Rumrnan, 1970). A breakthrough came when Vickery and Clark
(1972) perfomed a set of wind tunnel experiments on a model of a 366m (1200 ft.)
concrete chimney and developed a model for the forced vibration amplitudes of tapered
stacks. About a decade later (Vickery and Basu, 1983) the same model was extended to
include motion-induced effects, which is the key consideration for steel chimneys, and
the effects of lateral turbulence in the atmosphere. These rnodels are more appropriate in
that the across-wind loading is treated as a narrow banded random process that better
reflects the nature of the tme loading.
The slower development of reliable design guidelines for across-wind loads and effects is
related to the dearth of reliable data. The reason for this is the inability of wind tunnels to
properly scaIe Reynolds numbers and the free Stream flow properties simultaneously.
Thus, data from which the necessary parameters for any response model should be
derived are full-scale measurements on existing chimneys.

The semi-empirical response model of Vickery and Basu and the sources of the full-sale
data used to develop the appropriate parameters are reviewed in Chapter 2.

1.2 Motivation and Preview of Thesis

Sometimes engineers implement designs without properly accounting for the effects of
wind. History is littered with cases where civil engineering structures failed for this
reason. The motivation for this study stems from two such instances for large reinforced
concrete chirnneys. The engineers heeded the s i g s that the structures were in danger of

3
collapse and catastrophic failure was avoided. However, the cases did warrant further
.

study.

In the first instance, two large reinforced concrete chimneys were constructed near a
chimney that was roughly 25% larger in diameter. For wind directions where one of the

new chimneys was in the wake of the larger chirnney, the separation was 5.1 diameters,
based on the upstream chirnney. This caused a large increase in the lateral buffeting
loads on the downstream chimney. In the second instance, an elevator enclosure was
mounted on the surface of two chimneys that had been in service for 20 years. After the
enclosures were instded, the amplitude of the vortex-induced oscillations at the critical
wind speed increased from a value of the order of O.lm for the plain circular cross section
to a value of the order of 1.Om for the modified cross section. This thesis examines the
aerodynamic forces for these two cases in detail.

In both cases, the chimneys were of the 'windshieId7 type. For this form of chirnney, a
reinforced concrete shell shields a liner (or liners) from the wind. The usual method of
analysis examines only the shell resting on a fixed foundation. The effect of foundation
flexibility on the across-wind response is examined in Chapter 3; particular attention is
paid to the influence of foundation flexibility on the linerkhell interactions.
The results of the field study and a model study of across wind loads on pairs of circular
cylinders are presented in Chapter 4. The experimental program was conducted in the
large 1 MW wind tunnel at Monash University, Australia. The maximum Reynolds
number achieved in the experiments was 106 compared to the full scale Re of 107. The
effect of changing upstrearn diameter and the separation between the cylinders on the

across-wind buffeting loads was examined. Cornparisons between the two studies are
also presented.
The results of a wind tunnel study of the effect of surface mounted appurtenances on the
wind effects on chimneys are presented in Chapter 5- The study was conducted in BLWT

II at The University of Western Ontario.

The Reynolds number for this set of

experiments ransed from 3.5 x 10' to 5 x 10'. Small-scale turbulence was added to the
free Stream to promote an earlier transition in the body boundary layer; or sirnulate higher
Reynolds nurnber. The influence of the angle between the wind and the appurtenance on
the aerodynamic forces is examined. The influence of the size of the appurtenance
spanning the height of the cylinder on the wind loads is also exarnined. Predictions of the
across-wind response of the prototype are compared with the reported full scale
observationsTo the writer's knowledge, these aspects of the across wind loads and effects on
reinforced concrete chirnneys have not been observed or studied before. Wind tunnel
studies on pairs of circular cylinders are not new but the focus of this part of the study is
on the across-wind loads whereas most previous studies focused on the mean forces. In
previous instances where lifi forces have been examined, the Reynolds number has been
in the subcritical range.

Chapter Two
Literature Review
2.1 Introduction
The dynarnic response of tall slender structures depends on a number of fluid-structure
interaction phenomena. A convenient way to determine the response is to consider the
dong-wind and across-wind components separately.
The dong-wind response can be again divided into mean and fluctuating components.
The mean response is equivdent to the static response of a structure subject to a drag
force per unit length.

The fluctuating component is primarily the buffeting of the

structure by the along-wind component of turbulence. The latter was the subject of much
research in the 1960's as structures increased in height and flexibility. The work of
Davenport (1961, 1963a, 1963b, 1964) addressed the dong-wind loading and response of
structures in general and later Vickery and Kao (1972) addressed tau chimneys
specifically.
The across-wind response of civil engineering structures can also be divided into a mean
and fluctuating component. However, for structures of circuIar cross-section such as
chimneys, the tirne averaged flow pattern is symrnetric, thus, the mean lift force is zero.
The fluctuating component of the loading depends on three main fluid-structure
interaction phenomena. The most important among these is the shedding of vortices from
the structure in its wake that induces a force across the direction of the wind. When the

6
frequency of the shedding is close to the natural frequency of the structure, resonance
will amplify the structural vibrations. These oscillations cause the second phenornenon,
aeroelastic effects or motion-induced forces, which depend strongly on the ratio of the
frequency of the oscillation and the frequency of the vortex shedding.

The third

important fluid-structure interaction is the buffeting of a structure due to the lateral


component of turbulence.
The modem boundary layer wind tunnel has enabled the accurate prediction of windinduced forces for most buildings, stadiums and bridges. For these structures, sirnilarity
of the key non-dimensional parameters governing the flow properties and the structural
properties are achieved. The parameters can include Froude and Cauchy numbers to
achieve similarity for the main structural actions; and turbulence intensity and mean wind
speed profiles to achieve flow similarity. For most engineering structures, the flow
around a scale model is governed by the geometry of the mode1 if the edges are 'sharp'

and thus define the separation points.


However, the inability of boundary Iayer wind tunnels to properly scale Reynolds number

Iimi ts their applicability for chirnneys and other structures of circular cross-section.
Thus, using aerodynamic data gathered from wind tunnel experiments on circular
cylinders c m be a hazardous exercise if one is not aware how Re affects flow patterns
and the induced loads. This review starts with a brief summary of the flow around
circular cylinders and a discussion on the flow-induced Forces and how these are affected
by Re and turbulence in the incident flow. The across-wind response model for full-scde
structures of circular cross-section from the work of Vickery and CO-workers,employed

in the present work, is then reviewed dong with some of the data gathered fiom full-scale
structures.

2.2 Aerodynamics of the Circular Cylinder


Fluid ffow around circular cylinders has intngued researchers for more than a century.
Despite this passage of time, the complete understanding of the flow field and the
induced forces is still elusive. The complex interactions of local viscous and inertial
forces and the influence of free Stream turbulence or other disturbances have a profound
effect on the Flow patterns and pressure distributions around circuIar cylinders.

In his recent book, Zdravkovich (1997) characterized the disturbed flow around circular
cylinders into four regions (Figure 2.1). The regions are: (i) the stagnation region, (ii) the
body boundary layers, (iii) the region of disturbed irrotational fiow and (iv) the wake.

Figure 2.1: Disturbed flow fieid around circular cylinders (Zdravkovich, 1997)

As a result of intermolecular attractions, there is a no slip condition on the surface of the


cylinder. Thus, a thin boundary layer between the region of disturbed irrotationai flow
and the surface of the cylinder must exist. The thickness of this layer grows with
increasing distance from the stagnation point owing to the equilibrium of viscous and

inertial forces (Schlichting, 1979). The static pressure in the boundary layer for a &en

angular displacement from the stagnation point is constant and equal to the static pressure
impressed by the irrotational flow in the third region.

Over the front portion of the

cylinder this pressure decreases as the disturbed flow is accelerated and the boundary
layer remains atrached. As the outside flow in the third region begns to decelerate, the
pressure gradient becomes positive and opposes the movement of the fluid particles in the
boundary layer. This causes the velocity gradient in the boundary layer to

V ~ N Sat~the

surface of the cylinder as shown in Figure 2.2 (Melbourne et al., 1997) and the boundary
layer separates and is transporteci into the main stream. These separated boundary layers,
or shear layers, bound the fourth region of turbulent rotational flow to the lee of the
cylinder, o r wake. The pressure in the wake of the cylinder is negative and, when
combined with the positive pressure on the front of the cylinder, results in a form drag
force in addition to the viscous drag in the direction of the flow.

decdorating dp/

Boundary layer
vdocity profik
Vdocity det oct finally
r~suitingin r o w r s d

ffow at the surface


and maration of the

boundary Iayur.
Figure 7.2: Mechanism of separation and vortex formation (MeIbourne, 1997)

dx

>O

9
Downstream of the separation point, a reversal of flow occurs and a vortex begins to form
(Figure 2.2)- This vortex continues to grow until it becomes unstable and separates from
the cylinder and is transported downstrearn in the wake. Just prior to this separation of
the vortex from one side of the cylinder, another begins to forrn on the other side. This
process resuits in the farniliar Kirmiin vortex Street (von Kamiiin, 1911). Strouhal(1878)
first observed the phenornenon of vortex shedding when conducting experiments on the
creation of sound- He discovered a non-dimensional relationship between the shedding
frequency, arnbient velocity and the diameter,

The action of vortex shedding alters the instantaneous pressure distribution and results in
fluctuating drag and Iift forces.
The discussion above is a brief overview of the flow patterns and the resulting forces for

a circular cylinder. The rna,@tudes

of these forces, particularly the mean drag force, are

strongly affected by the location of the separation point on the cylinder. For sharp-edged
bluff bodies such as a square or flat plate, the separation points are defined by the
geometry since the boundary Iayer is unabIe to remain attached at the discontinuous
changes in curvature of the body. For the circular cylinder however, this is not the case.

2.2.1 Effect of Reynolds Number

The position of the separation point for a smooth cyIinder in an incident ffow free of
disturbances depends on the ratio of the inertial to viscous forces. Expressed as a nondimensional parameter this ratio forms the Reynolds number,

where V is the velocity of the flow, D is the cyiinder diameter and v is the kinernatic
viscosity of the fluid.
The change in the angular position of the separation point with increasing Re is
surnmarized in Figure 2.3 (Basu, 1982). For Re in the subcritical range, the boundary
layer is laminar at separation and transition occurs in the separated shear layer. As Re
increases, the transition point gradually moves upstream in the shear layer closer to the
cylinder- At a critical Reynolds number the transition occurs imrnediately after laminar
separation and the shear layer sucks itself back ont0 the cylinder surface and causes a
separation bubble to f o m upstream of the final turbulent separation point. At higher Re,
the transition point moves far enough upstream so that the boundary layer is fulIy
turbulent at separation.
The influence of these changes in flow patterns on the drag coefficient of the cylinder is
shown in Figure 2.4 (Eiasu, 1982). The regions of Re are shown at the top of the figure.
For subcritical Re, the flow pattern remains essentially unchanged, thus the drag
coefficient is essentially constant. In the critical range in Re, the wake of the cylinder

(a) Subcritical Re Nos- (5 2x104 )

- - -

entrainment

-y

reversea rlow region

cylinder

(b) T r a n s i t i o n a l Re N o s ,
laminar
separation
transition
/

(cl Transcritical Re N o s .
transition

turbulent

Figure 2-3: Flow field near the separation point. (Basu, 1982)

critical

super
critical

Figure 2.4: Definition of different fiow regimes in te=<

transcritical
(post critical)

of drag coefficient (Basu, 1982)

begins to narrow rapidly thus increasing the base pressure and reducing the drag. After
reaching a minimum, the drap coefficient begins to nse and again reaches a steady
ma=pitude when the separation becomes turbulent. This final regime is referred to as
post critical

'. The region where the 'uphill clirnb' of the drag coefficient occurs is termed

supercritical.
The values of Reynolds number where these changes occur exhibit some scatter. A
compilation of the drag coefficient for a number of expcrimental studies for two-

' Originally termed transcritical after Roshko (1961).

13
dimensionai cylinders is shown in Figure 2.5 (Shih et al., 1993). For very low turbulence
free streams and srnooth cylinders most data suggest a critical Re near 2x10~and the
begnning of the supercritical regime near ~e=5x10'. The start of the post criticai regime

is much harder to define due to the lack of experimental data for hi&

Re. Zdravkovich

to~6x10~.
(1997) suggests a value of 3 . 5 ~ 1 0
The drag coefficient for two-dimensional cylinders in the different Re regirnes are also
shown in Figure 3.5. There is a large amount of scatter in the data but in the upper
subcritical range, the value of the drag coefficient is near 1.2 and the minimum drag
coefficient around ~ e = 5 x 1 0
is~near 0.3. For post critical, the data in Fiapre 2.4 suggest a
range frorn 0.4 to 0.7.

Figure 2.5: Mean drag coefficient as a function of Re for srnooth cylinders (Shih et al., 1992)

14

The fluctuating lift coefficient (RMS) and Strouhd number are also strongly dependent
on Re. Compilations by Fujita (1988) of

CL and S, are shown in Figures 2.6 and Figure

2.7, respectiveIy. In the critical range of Re, the value of the RMS Lift coefficient drops

rapidly and then 'levels off in the supercntical range. There is a large amount of scatter
in the data in this regirne with values ranging from 0.02 to O. 15. The Strouhal number
d s o shows a large arnount of scatter, particularly in the critical and supercritical range of

Re. The scatter in both

CLand S,

is partially explained as the poor organization of the

shedding dong the span of the cylinder. Schewe (1983) notes that the vortex shedding
occurs in different 'cells' and at Re in the upper supercritical and post critical ranges,
these ceils merge and the shedding becomes more organized and the peak in spectra of
the lift force narrows (Figure 2.8, Cincotta et al., 1966).
O. 7
C

0.6

0.6

0-4

0.8

K,

Cheung Ti'=OA%

Ti=9.1%
Savkar Strouhal lift

U
-r(

Schcwe

/ Noudashtr

.r.
U1
U

O
U

2
Y

w
c.

fil

(D

0.2-

.r(
U

"3

0.1-

c.

.c

o . :
10'

IO6

10'

10'

Re

Figure 2.6: Compilation of RMS Iift coefficient For smooth c y h d e r s (Fujita et al., 1988)

Figure 2.7: Compilation of Sr for smooth cylinders (Fujita et d.,1988)

Figure 2.8: Spectnrm of lift coefficient for smooth cyIinders at high Re (Cincotta et al., 1966)

2.2.2 Effects of Free Stream Turbulence


The results and discussion above refer to a two-dimensional smooth cylinder in a uniform

flow with very low turbulence. For 'real' flows in the atmosphere, turbulence intensities
are of the order of 10% to 30%. These velocity fluctuations and associated pressure

fluctuations have a significant impact on the position of the separation point and thus, the
flow-induced forces. This is due to the promotion of transition in the boundary Iayers at

16
Re lower than values observed in smooth flow. This c m be viewed as an increase in the
'effective Re' noted first by Fage and Warsap (1929) when they exarnined the effect on
the drag coefficient when rope netting was placed upstream of the cylinder.
The maximum Reynolds number attainable in wind tunnels such as BLWT II at The
University of Western Ontario and the large 1 MW tunnel at Monash University,
Australia ranges from 5x10' to 106, respectively. Cheung and Melbourne (1983) studied
how the drag and fluctuating lift forces are affected by free stream turbulence in this
range of Re. Their results are shown in Figures 2.9 and 2.10 for the drag and RMS lift
coefficient, respectively. These results suggest an upstream movement of the separation
point and widening of the wake similar to higher Re flows.

10s

106

Re
Figure 2.9: Effect of free stream turbulence on mean cira; coefficient (Cheung and Melbourne, 1983).

Figure 2.10: Effect of free Stream turbuIence on RMS lift coefficient (Cheung and Melbourne, 1983)-

2.3 Across-Wind Response Model for Chimneys


The model employed in the present work was developed by Vickery and CO-workers
(Vickery and Clark, 1972, Vickery and Basu, 1983, Basu and Vickery, 1983). It is
widely employed in various simplified forms as the basis for across-wind response
predictions in the National Building Code of Canada, 1995, the Amencan Concrete
Institute Standard (AC1 307-98), and the model code of the Cornmittee for the Design of
Industrid Chimneys [CICIND].

The predictive mode1 is based on Iinear random

vibration theory and treats the forces associated with vortex shedding in two uncorrelated
parts; those that exist on a stationary body and those induced by motion of the chirnney.

2.3.7 Vortex Shedding Forces on a Stationary Cylinder


As mentioned above, fluid flowing past bluff bodies such as the ckcular cylinder results

in a fluctuating lift force caused by vortex shedding. The centrai frequency of this force
is a v e n by the Strouhai relationship;

where is the mean wind speed, S, is the Strouhal number and D is the diameter. The
RMS lift force per unit lene@ associated with vortex shedding is represented by

where

CL:,,
is the RMS lift coefficient associated with vortex shedding and p is the density

of air. Measurements of the lift force suggest that it is not stnctly periodic. Even in
smooth flow the spectrurn contains a n m o w spread about the central shedding frequency.
The distribution with frequency is suggested by the nature of the free Stream turbulence.
Large scales much greater than the diameter have the effect of slow fluctuations in the
mean speed, thus altenng the central shedding frequency. Smaller scales will cause a
spread in the bandwidth of the force spectrurn about f,. Since the distribution of
fluctuations in velocity is Gaussian, then upon examination of Equation 2.3, one would
expect that the distribution of the shedding frequency would aiso be a Gaussian form.

Vickery and Clark (1972) suggested the following equation;

and compared it with measured spectra of the lift force on a tapered cylinder in turbulent

shear flow (Figure 2.11).

The spectrum only represents those forces associated with

vortex shedding. There is also a lift force associated with the lateral turbulence in the
oncoming flow and c m produce ~i~gnificant
energy at frequencies below f,.
The spectrum is characterized by three parameters, namely the RMS lift coefficient due
to vortex shedding, the Strouhd number, St, and the bandwidth, B. These parameters are
strongly influenced by flow conditions, i.e. turbulence scale and intensity as well as the

Figure 2.1 1: Spectnirn of lift coefficient on mode1 tapered cylinder (Vickery and Clark, 1972)

20
geornetry and three-dimensionality of the stnicture, Le. taper and aspect ratio. For
application to reinforced concrete chimneys where Reynolds Numbers are of the order of
107,it is appropriate to derive the value of the parameters from full-scale rneasurements.
Vickery and Basu (1984) reviewed full-scale data from seven reinforced concrete
structures of circula cross section. Using these data and the results of a study by
Waldeck (1992) and Sanada et al. (1983), Vickery (1995) reported the current best
estimate for the RMS lift coefficient due to vortex shedding for nornindly twodimensional conditions (Figure 2.12). The full-scaIe data suggested that the RMS Iift
coefficient is stron,oly dependent on the turbuIence intensity. He argued that it is only the
scales of turbulence on the order of or less than a diameter that are responsible for the
observed effects. The larger scdes of turbulence would only have the effect of altering
the local mean wind speed. He suggested a modified turbulence intensity that excludes
the very large scales of turbulence. The modified intensity is,

where

4:

is the integral length scale of turbulence. Using ilf as the independent variable,

the relationship shown by the line in Figure 2-13 is a v e n by the relationship,

He also reported the Strouhal number and bandwidth parameter for two-dimensionai
conditions as, S, E 0.235 and B = O. 1+ 2i,

21
As noted above, the parameters

(zLv
and Sr) also depend on aspect ratio, h and the taper.

For values of taper typical of reinforced concrete chimneys, Vickery suggests the
foiiowing relationships,

The correlation of the fluctuating lift force due to vortex shedding at two elevations on a
chimney, zl and

z2,

is given by the coherence function. The relationship suggested by

Basu and Vickery (1983) is

0.04

0.06

0.80

0.10

0.20

Blackburn and Melbourne. 1993


O

0 , f Vickery. 1988
&su. 1982

Figure 2.12: Full-scale RMS Iift coefficient for RIC chirnneys (Vickery, 1995)

where r =

21z, - zzl

bkl)+ 0k311

and !L is a correlation lena& with a value near unity or perhaps

slightly greater. A value of 1 2 is recornmended in AC1 307-98.

2.3.2Forces lnduced by Motion


The primary effect of the motion-induced force is the lock-in phenomenon (Feng, 1968).
This occus for reduced velocities near U S , and is so n m e d because the shedding
frequency Iocks on the frequency of vibration.

Accompanied by this effect are an

increase in the lift force per unit Iength and an increase in the spanwise correlation of the
lift force O\lovak and Tanaka, 1975). Both effects can be modelled by an additional force
per unit Iength centreci on the natural frequency or the frequency of motion. The approach
taken in the response mode1 is to treat the force in two parts, one in phase with the
displacement and one in phase with velocity, after Scruton and Flint (1964). These two

parts are often referred to as the aerodynamic stiffness, ha,and aerodynamic damping, ka.
The lift force per unit lengh associated with the motion of the chimney is then

where y is the across-wind displacement. These magnitudes of the two forces per unit
Iength (Le. h,y and & y ) are of the sarne order but the aerodynamic stiffness is much
smaller in magnitude than the stiffness of the structure and is typically ignored in

23

engineering design. If simple harmonic motion is assumed, a mass-damping parameter


can be developed, Le.

where rn is the mass per unit length and

& is the

aerodynamic damping expressed as a

fraction of critical. An analogous expression, K, can be used to represent the structurd


damping force. The two t e m s ka and Kaare related by,

wheref, is the natural frequency of vibration. The rnaagnitude of Ka is stron& dependent

Gsr -

on--foD

a , Re, turbulence intensity and the amplitude of vibration.


vc,,

The available full-scale data suggest a relationship for Ka,, the rnass-damping parameter

for small amplitudes, as shown Figure 2.13 as a function of

and turbulence

ucrir

intensity. Figure 2.13 can be summarized by the equation,

where k =

U .
-

The magnitude of the aerodynarnic damping increases rapidly in the

&rit

vicinity of the critical wind speed for vortex shedding. This magnitude decreases with
increasing turbulence intensity as shown. A typical value of Ka,,n, for open country

Figure 2.13: Mass-damping parameter for small amplitudes.

exposure is 4 . 6 . K, is also dependent on aspect ratio and is modified using Equation

For larger amplitudes, Vickery and Basu (1983) suggest the following fom,

where

7 is

the RMS displacement of the cylinder and

is a limiting displacement.

Evidence that this form is appropriate is shown in Figure 2.14 (Vickery, 1980) where the
experimental data measured by Wooton (1968) are fitted by an equation of the form,

Experimental
(Wooton, 19681

0.m

0.2

O. 1

0.4

0.6

?.O

2-0

K,

4.0

Figure 2-14: Measured and fitted response (Vickery, 1980)

The fit shown in Figure 2.14 suggests that the response can be separated into two

regirnes: forced vibration and large amplitude- In the forced vibration regirne, Ks> IKaoI
and Equation 2.16 becomes

26

In the large amplitude regirne, Ks< [K,,Iand the amplitude is determined by equating the
energy derived from the fluid and the energy dissipation in the chimney, Le.

Reinforced concrete chimneys are usually precluded from entering the large amplitude

regirne. The reason for this is that the reduced mass is in the range of 100 to 140. For a
typical structural damping of 0.8% to 1.O% of cntical (Basu, 1982), Ksis greater than Ka,.
However, the nonlinear part of the damping cannot always be ignored. A case where this
is tme is the subject of Chapter 5.

2-3.3Response Predicfions
The spectrum of the generalized lift force in a gven mode, i, with shape Hz) is given as,

By substituting the Equation 3.10 into Equation 2.19,

The RMS modal amplitude is then given by,

where Kiis the generaiized stiffness, f;- is the naturai frequency and
inherent and aerodynamic damping.

CE is the sum of

The aerodynamic damping for a a v e n mode is

determined by,

For the purposes of making response predictions in this thesis, a cornputer program was
coded enabling the detailed calculation of the generaiized force and aerodynamic
damping. However, the generd form of the response in a ,aven mode is;

where

- =Cl-,/L/
Grn
M i S:

Cs+ C a ,

In Equations 2.23 and 2.24, the input parameters are evaluated at a height of 5/6 times the
chimney height, which assumes that the loading over the top 1/3 of the chimney is the
most effective for the response. This is generally vaiid for the fundamental mode of a
chimney. In Equation 2-34, the constant C is a function of the mode shape and the taper
of a chimney, but not the stiffness or mass.

2.4 Surnmary
The flow around circular cylinders and a mode1 for the across-wind response of chimneys
are briefly reviewed. The intent of the first part of the review is to highlight the key
findings frorn a very large number of books and papers on the subject. As mentioned in
the introduction, one must take great care when applying aerodynamic data gathered in
wind tunnels to prototype chirnneys. This is, however, what the author intends to do in
Chapters 4 and 5 of this thesis. In both chapters, wind tunnel data are used to perform a
'post mortern' analysis for two events where the wind tunneI is the only tooI available to
further investigate and explain the prototype behaviour.

Chapter Three
Effect of Foundation Flexibility on the Across-Wind
Response of R/C Chimneys with Freestanding Liners
3.1 Introduction
The interaction among the supporting soil, the foundation and the super-structure,
commonly refemed to as soil-structure interaction, alters the dynaxnic characteristics of a
chimney and consequently influences its response to any dynamic Loads. Therefore, the
evaluation of soil-structure interaction is needed for the proper analysis of the chimney7s
response to wind forces.
In areas where seiszllic loads are not significant, ta11 chimneys are often constructed with
a reinforced concrete shell enclosing a freestmding brick liner with both resting on a
cornmon foundation- The traditional method of analysis treats both the shell and liner as
independent fixed base cantilevers with lateral wind forces resisted by the shell. This
analysis ignores the effect of the foundation flexibility and any possible transmission of
base forces to the liner through the cornrnon foundation.
Soil-structure interaction has two main effects on mode shapes. First, the curvature in the
shell is reduced due to rotation of the foundation. This leads to a reduction in structural
darnping which, depending on the underlying soil, may o r may not be compensated by
increased foundation d m p i n g (Novak & El Hifnawy 1983). Second, the liner is excited
by the rotation of the foundation. This leads to an increase in the relative shellniner

30

motions. and ultimately to impacts at the tip of the chimney in the presence of adverse
aerodymamic conditions.
Galsworthy and El Naggar (1997) examined the effect of the foundation flexibility on the
modal properties of a 250m t d i reinforced concrete chirnney.

They found that the

foundation flexibility reduced the natural frequencies by 10% to 20% and increased the
shell deflection relative to that of the liner (represented by mode shapes) by almost 100%
in the first mode (henceforth calIed the liner mode) and 20% in the second mode (shell
mode).

The present chapter examines how the response due to vortex sheddinz is

affected by changes in foundation type and soil properties. Tip deflections and base
forces are examined in detail.

3-2Soil-Structure Interaction
To account for the soil-structure interaction (SSI) efficiently, the substructuring approach
is used in this analysis. The andysis is performed in two steps. In the first step, the
impedance functions (dynamic stiffness) of the foundation system are calculated taking
into consideration the wave propagation away from the foundation system due to the
continuity of the soil medium.

In the second step, the dynamic analysis for the

superstructure is performed using the impedance functions of the foundation system.


3.2.1 Foundation Model

The evduation of accurate wind response of chimneys requires proper values for the
dynamic stiffness and damping of the foundation to be used in the analysis. The variation
of these values with dynamic soil characteristics is usually remarkable and consequently

31
their effect on the superstructure's behaviour is important. In this study, the variation of
the chimney's response with the shear wave velocity of supporting soil for both shallow
and deep foundations is investigated,

The configuration of the deep foundation

considered in this analysis is shown in Fig. 3.la and the configuration of the shalIow
foundation is shown in Fig. 3. lb-

Bedrock
Figure 3. La: End-bearing deep foundation

Firm Stratum
Figure 3.1 b: S hallow Foundation

Several approaches are available for the analysis of foundation systems to account for
dynarnic soil-structure interaction that are mostly based on the continuum approach and
the assumption of elastic or viscoelastic soil.

The analyses used to determine the

impedance functions of shallow and deep foundations are described briefly below.

3.2.1.1 Shallow Foundation


Shallow foundations for chimneys are often idealized by a massless circular disc. The
stiffness and darnping constants of a massless circular disc resting on the surface of a

32
linear viscoelastic halfspace were obtained using either three-dimensional or twodimensional continuum approaches. The analytical solutions include the contribution of
many researchers: Bycroft (1956); Luco and Westmann (1971) and Veletsos and Verbic
(1973). Veletsos and Wei (1971) and Luco and Hadjian (1974) introduced numerical
solutions. For circular bases the cornplex stiffness Kiassociated with direction i are
obtained by the determination of the relationship between harmonic force acting on a
rnassless disc resting on the surface of the halfspace and the resulting displacernent of the
disc. This complex stifiess can be expressed in terms of the true stiffness constant, ki,

and damping constant, Ci, as

in which

k; is static stiffness, a,=-=o R


Y

dimensionless frequency, R is the disc radius, V,

,/C/p = shear wave velocity of the soii, G and p are the soi1 shear rnodulus and mass

density, respectively. The parameters k; and c: are stifiess and damping constants
k.
norrndized as follows: kr=&, c:=&
ki

(Le., see Veletsos & Verbic 1973).


kiR

Embedment is known to increase both stifiess and damping but the increase in
foundation damping is more significant. The response of embedded footings c m be
approximated by assuming that soi1 reactions acting on the base are equal to those of a
surface footing and the teactions acting on the footing sides are equal to those of an
independent layer overlying the halfspace assurning plane strain conditions. Beredugo
and Novak (1972) found that this approximate approach yields reasonable results

33
compared with the finite element predictions- In this study, the plane strain solutions
developed by Novak et al. (1978) for side reactions and the hdfspace solution for base
reactions developed by Veletsos and Verbic (1973) were used to evaluate the stiffness
and damping constants of the shallow foundation.
The impedance functions of a shallow Iayer are different from those of the halfspace. In
the case of a shallow layer, the stiffness increases and geometnc damping (due to wave
propagation away from the foundation) decreases or even vanishes.

Studies of the

behaviour of strata suggest that geometric damping may completely vanish if the
excitation frequency is lower than the first natural frequency of the soi1 Iayer (Kobori et

aI. 1971). For a homogeneous soi1 layer, the first vertical and horizontal natural
frequencies, o,and wu,respectively, are:

where h = Iayer thickness and v = Poisson's ratio of the soi1 Iayer. At frequencies lower
than u,,and ou, the only source of darnping is the material damping.
The static stiffness of a footing embedded in a Iayer of lirnited thickness in the horizontal
and rocking modes was derived by Elsabee and Morray (1977) and in the vertical and
torsional modes by Kausel and Ushijima (1979). The impedance functions are defined
similar to Eq. 3.1. In this case, however,
soi1 Iayer, h.

values are a function of the thickness of the

34
Kausel and Ushijima (1979) recommend taking k' and c' equal to the halfspace functions
except for c' in the low frequency range. For frequencies below the first layer natural
frequencies, it would be safe to ignore geometric darnping completely, and the damping
can be established as a function of stiffness giving c as

in which P is the material damping ratio of the soil.

3.2.7-2Deep Foundation
Stiffness and damping of piles are affected by interaction of the piles with the
surrounding soil. To account for this interaction, a few approaches can be employed: the
continuum approach (Novak & Nogami 1977); the lumped mass mode1 (Matlock et al.
1978); the finite element method (Kuhlemeyer 1979); boundary element solutions
(Kaynia & Kausel 1982); and the simplified approach (Dobry & Gazetas 1988)- In
groups of closely spaced piles, the character of dynamic stiffness and damping is further
complicated by interaction between individual piles.

Therefore, the superposition

approach was used in the analysis. In this approach, the stifmess and damping of single
piles are calculated first, then the group effect is accounted for using the interaction
factors, as discussed here.
The dynamic stiffness (impedance function) of piles can be described as
Ki=ki (a,)+ioci (a,).

35
Calculating the forces needed to produce vibration of the pile head having unit amplitude

in the prescribed direction generates dynamic stifhess. Novak (L974a) described the
stiffness constants, ki,and the constants of equivalent viscous damping, ci,for individual
motions of the pile head as a function of the pile and soi1 properties. Novak and AboulElIa (1978) provided an approach to evaluate the impedance functions of piles in a
Iayered medium,
Dynarnic group effects are quite complex and there is no simple way to alleviate these
complexities. The only simplifications availabIe are the interaction factors approach and
the approximate approach due to Dobry and Gazetas (1988) and Gazetas and Makris
(1991) in which the interaction problem is reduced to the consideration of cylindrical
wave propagation.
A simplified approximate analysis for the dynarnic group effects can be formulated on

the basis of dynarnic interaction factors, a,introduced by Kaynia and Kausel (1982) who
presented charts for dynarnic interaction.

The interaction factors denve from the

deformation of two equally loaded piles and give the increase in deformation of a pile
due to deformation of an equally loaded neighbouring pile.

For a homogeneous

halfspace, the interaction factors between two piles may be aven by @obry & Gazetas
1988; Gazetas & Makris 1991):

-Os

a"-

fi

where

(2)
e
d

-pu- S

-&-S

and g (e0)=cr, ( 0 " ) c o s ' ~ ~(90n)sin'B


~c~

where a, and cc, are vertical and horizontal interaction factors, respectively; S/d = pile
spacing to diarneter ratio; 8 is the angle between the direction of Ioad action and the plane
in which the piles lie; and V,=

the so-called Lysmer analog velocity =

3.4vs
z(1Y )

To analyze a pile g o u p using the interaction factors approach, the impedance functions
of single piles and the interaction factors are calculated first, then the group impedance
functions are computed. The irnpedance functions of a pile group of n piles are then
given by (El Naggar & Novak. 1995):

where K,:

KU, K:

and K:

are the vercical, horizontal, rocking and couplin:: group

stiffness, respectively. In Eq. 3.6,

k;

is the static vertical stifiess of the single pile,

'&[atlwhere ai = cornplex interaction factors between piles i and j, O$ =k; /K",and Kv

37
is the complex vertical impedance function of the single pile . Similady,

kuis the static

horizontal stiffness of the pile [ ~ ~ & [ o rwhereai


~l
= complex interaction coefficients for
the horizontal translation and rotation.

The formulation of the [ a l c a n be found in El

Naggar and Novak (1995).

3.2.2 Structure Mode1


The general arrangement used to develop the mathematical model is shown in Figure 3.2.
The superstructure consists of two thin-walled beams with circular cross-section. The
section is assumed to be stiff in the radiai direction, eliminating ovalling. Torsional
motions are not considered due to the syrnrnetry of the structure, thus, al1 motions
(rotation and translation) are in the vertical plane.

Deformations due to shear and

rotationd inertia are assurned to be negligible.


The Rayleigh-Ritz method is used to establish the stifiess and mass matrices of both the
shell and liner. The mass and flexurd rigidity of the shell and liner are assumed to be
distributed dong the entire lena& of the chimney. A set of shape functions is chosen
such that the mode shape is represented as,

where

yri

is the ilf' shape function and n is the number of shape functions employed.

Thus, the number of degrees of ficedom is equd to the number of shape functions

1,

considered. Hence, the shell and liner stifmess matrices, [ K]~and [K, and their mass

Reinforced

/-

Sheu

Figure 3.2: Mode1 for soil-structure interaction

matrices, [M,] and [ M , ] , have dimensions n x n. In the present analysis, the shape
functions used are the true mode shapes for a uniform cantilever beam for both the shell
and liner. The elements of the shell stiffness and mass matrices are then evaluated by:

and the liner matrices are evaluated similarly.

3.2.3 Foundation Structure Mode1


The foundation flexibility is included in the above andysis through two ri@d body shape
functions,

-- and
H

vu= 1.0

correspondhg to rotation and translation at the

foundation IeveI- The global mass and stiffness matrices are assembled from the
corresponding foundation and structure matrices. The undarnped free vibration equations
of motion of the global system can be written in matrix form as follows:

in which m . and I f are the mass and mass moment inertia of the foundation and
{O)represents the nul1 vector. The other mass matrix terms are:

The other stiffness matrix terms are:

Equation 3.9 defines the eigenvdue problem for the shell-liner-foundation system. Using
matrix iteration (Clough & Penzien 1993) to solve the eigenvalue problem, the undarnped
mode shapes for the liner and shell are obtained. The jth mode is:

where
function,

the coefficient in the eigenvector corresponding to the ith elastic shape

w, in the$

mode; and

Oj, clj = foundation

coefficients forjtlt mode.

41
To dernonstrate the mode shape predictions that result from the model, the chirnney that
will be discussed in greater detail in Chapter 4 is anaiyzed. Exarnple mode shapes are
shown in Figures 3.3 and 3.4 for an end-bearing deep foundation with

V' = 250 mis.

In

the first mode, the liner mode, the liner dominates the motion with both moving in phase.

In the second mode, the shell mode, the motion of the sheIl is 180 out of phase with the
liner. For this case, the amplitude of shell deflection was larger than that of the liner but
this is not always the case, especially for v e r =flexible foundations (Galsworthy & El
Naggar 1997).
For the purpose of cornparison, the mode shape for the shell on a fixed base (dashed line)
is shown in Figures 3.3 and 3.4. Two observations can be made from this comparison:
first, the cwvature of the shell is reduced by the rotation of the foundation; second, the
generalized mass, defined by

shell

liner

that increases as compared to the fixed base case due to the motion of the liner. For the
modes shown, translation of the foundation is not sipificant but may be for higher
modes (Novak 1974b). The computed natural frequencies are 0.203 Hz and 0.256 Hz for
the liner and shell modes respectiveiy. The natural frequency for the fixed base case is
0.282 Hz.

Due to low observed total damping in full scale chimneys, typically c 1% of critical, the
undamped mode shapes are accepted as the true shapes. AIso, the natural frequencies are

Q liner.i (2)
Figure 3.3: Mode shape for Liner Mode

Figure 3.4: Mode Shape for SheH mode.

43
well separated and a proportional damping matrix is therefore a reasonable
approximation. Darnping is discussed in detail below.

3.3 Damping
The total damping in each mode,

is the sum of contributions from foundation,

structurai and aerodynamic damping. To calculate the foundation contribution, Novak


and El Hifnawy (1983) presented a convenient form, based on modal analysis, Le.

where cm,cm,and c,, are the foundation damping coefficients for rocking, translation
and cross terms, respectively andfi is the natural frequency of mode i. H is the height of
the chimney.
The aerodynamic damping was calculated using the approach described by Basu and
Vickery (1983). Figure 3.5 shows the aerodynamic darnping for the deep foundation case
in the shell mode.

It can be noted from Fig. 3.5 that the effect of the foundation

flexibility (lower soi1 shear wave velocity) was to reduce the numerical value of the
aerodynamic darnping.

At lower wind speeds, the positive aerodynamic damping

decreased with the foundation flexibility. As the wind speed increased, the aerodynamic
damping assumed a negative value until it reached a maximum negative value near the
cntical speed. In this region, the effect of the foundation flexibility was to decrease the
negative value of the aerodynamic damping. For higher wind speeds, the aerodynamic

4 - v s = 150 m/s
*vs = 200 m/s
+Vs = 250 d s
-Fixed Base

10

15
20
25
Tip Wind Speed (m/s)

30

35

40

Figure 3.5: Modal aerodynamic damping (SheIl Mode) for deep foundation

damping increased assuming positive values again, and the effect of the foundation
flexibility was to decrease it. Tables 1 and 2 summarize the calculated foundation and
aerodynamic damping as a percentage of criticai for the deep and shdlow foundation,
respectively. The values shown for aerodynamic darnping are at the critical wind speed
for each case.
The larger values of foundation darnping in the Iiner mode can be attributed to the higher
foundation darnping coefficients at the lower frequencies and larger foundation motions
cornpared to the shell mode. The aerodynamic damping showed an opposite trend, i.e.
smaller values in the liner mode. This may be due to the higher values of generalized
mass in the liner mode. In general, the calculated darnping values are quite small which
is not surprising for such a massive superstructure.

Structural damping is a more difficult parameter to calculate theoretically. In practice, a


value is assumed based on Limited full-scale data with most design codes using 1.0% of
critical for concrete structures. In most studies, the contribution of the foundation is
included in the quoted number. Basu and Vickery (1982) suggested that the range of
darnping is 0.4% to 2.0% of critical.

In the results presented below, the structurai

damping was assumed to be constant and equal to 0.7% of criticai. This value was found
to be consistent with the value of structural damping interpreted from field measurements
(Galsworthy & Vickery 1999).
Table 1: Foundation and Aerodynamic Damping Values for Deep Foundation

Foundation

VS

Aerodynamic

Liner Mode

SheIl Mode

Liner Mode

Shell Mode

100

0.35%

0.07%

--07%

-.13%

150

0.3%

0.09%

-.O696

-.14%

200

0.25%

O. 10%

-.04%

-.15%

250

0.20%

0-11%

-.03%

-.17%

Table 2: Foundation and Aerodynamc Damping Values for Shdlow Foundation

VS(&s>

Foundation

Liner Mode

Shell Mode

Aerodynamic
Liner Mode

Shell Mode

3.4 Response of Chimney with Flexible Foundation


The procedure descrbed above was employed to analyze the response of an actual
chimney where an impact occurred between the liner and shell at the tip. The amplitude
required to cause the observed behaviour was of the order of 0.6m. The cause was a
combination of increased lateral force due to the presence of an upstrearn chimney and
the increased relative motions due to foundation flexibility. The results of a field study
and a discussion of the aerodynamic problem cari be found in Chapter 4. The actual
foundation system consisted of end-bearing piles supporting a ngid slab. However, for
the purpose of cornparison, both deep and shallow foundation systems will be considered
here.
The shallow foundation was considered to be a circular reinforced concrete slab and was
assumed to be 30m in diameter and 2m in depth. The foundation was assumed to be rigid
and fully embedded in the soil and the soil was assumed to be homogeneous viscoelastic.
The unit weight of the soil, y, was assurned to be 18 kN/m3 and Poisson's ratio v = 0.3.
The shear wave velocity for the bedrock was assumed to be 500 m/s with y and v being
equal to those of the soil. The deep foundation option consisted of 250 HP360x132 piles
with a rigd cap simila. to the shallow foundation. The piles were arranged in 6 rings
with uniform centre-to-centre spacing of 1.4m. The piles were assumed to be end
bearing.
Figures 3.6-3.12 show the computed response for the chimney supported by the deep
foundation system. The response was calculated assurning an isolated chimney in forest

47

terrain. The values of the parameters used in the vortex shedding mode1 were assumed to
be constant for al1 cases.
The peak shell, liner and relative (shell-liner) deflectiams are shown in Figures 3-6-3.8For lower wind speeds where the shell deflection is dominated by the liner mode, the
effect of the foundation flexibility was to increase the p e a k shell deflection. This may be
attributed to the reduction in the natural frequency in t h e liner mode (up to 18% as shown
by Galsworthy & El Naggar 1997), thus leading to resonance conditions nt: lower wind
speeds. However, for higher wind speeds, the effect O-f the foundation flexibility was to
decrease the peak shell deflection for two reasons: first, the liner deflection increased and
therefore the generalized mass increased; second, t h e negative aerodynamic darnping
decreased, as discussed above, and the peak shell deflection decreased. Liner deflections
were largest for the most flexible foundation and decreased with increasing shear wave
velocity. Relative deflections were significantly larger than in the fixed base case. The
variation was relatively small for the relative deflection in the flexible foundation case
within the range of the shear wave velocity of the soi1 considered.
The results demonstrate that the maximum peak response (at higher wind speeds) was
dominated by the shell mode. This was due primarily a0 the smaller generalized mass in
the shell mode when compared to the liner mode. Alscr, the generalized force in the shell
mode is laser since it is proportional to the fourth power of the critical velocity (which is
higher for the shell mode).

7
= 100 m/s

+Vs

13-vs = 150 m/s

*vs

= 200 m/s
VS = 250 m/s
-Fixed Base
-O-

10

15
20
25
Tip Wind Speed (mk)

30

35

40

35

40

Figure 3.6: Peak shell deflection for deep foundation

70

15
20
25
Tip Wind Speed (m/s)

30

Figure 3.7: Peak liner deflection for deep foundation

u v s = 150 m/s
*vs = 200 d s
-c+ Vs = 250 rnts

10

15
20
25
Tip Wind Speed (m/s)

30

35

40

Figure 3.8: Peak relative (shell - Iiner) deflection for deep foundation

The forces at the base of the chimney (the top of the foundation) are shown in Figures
3.9-3.12.

Figures 3.9 and 3-10show that the effect of the foundation flexibility on the

base forces of the shell was more significant than its effect on the shell deflection. It c m
be noted from Fig. 3.9 that the shell base bending decreased by 30% to 50% of the fixed
base case within the range of soi1 shear wave velocity considered in this study. Figure

3.10 shows that the effect of the foundation flexibility on the shell base shear was sirnilar,
but to a lesser extent. Figures 3.1 1 and 3.12 show the liner base forces. These forces are
unaccounted for in the conventional design approach that assumes a fixed base. Figure
3.11 shows the calculated liner base bending moment with a maximum value of

approximately 100 MN-m that represented more than 30% of the maximum bending
moment of shell base. Figure 3.12 shows that the liner base shear was as high as 0.6 MN,
representing 35% of the shell base shear.

+Vs = 100 m/s


u V s = 150 d s
*vs = 200 m/s
+-Vs = 250 mis
-FTxed
Base

10

15
20
25
Tip Wind Speed (rn/s)

30

35

40

Figure 3.9: Peak shell base moment for deep foundation

10

15

20

25

30

Tip Wind Speed (rn/s)

Figure 3.10: Peak shell base shear for deep foundation

35

40

10

15
20
25
Tip Wind Speed (m/s)

30

35

40

35

40

Figure 3.1 1:Peak liner base moment for deep foundation

10

15
20
25
Tip Wind Speed (m/s)

30

Figure 3.13: Peak liner base shear for deep foundation

52
Figure 3.13 summarizes the maximum tip deflections for the deep foundation case while
Fig. 3.14 sunm.rizes the tip deflections for the shallow foundation case. The deflections
are normalized by the computed maximum shell deflection for the fixed base case. Tt can
be noted from Figs. 13 and 14 that the shell deflections decreased due to the foundation
flexibility and varied from 70% to 90% of the fixed base case whiIe the relative shellliner deflection increased by more than 30%. Figures 3.13 and 3.14 show that the liner
deflection, which cannot be predicted using the conventional approach, was as high as
90% of the shell deflection for the fixed base case.
Figures 3.15 and 3.16 surnmarize the computed maximum base forces for the deep and
shallow foundations, respectively. The forces are normalized by the computed maximum
shell forces for the fixed base case. Figures 3.15 and 3.16 show that both the shell base
bending moment and shear decreased by more than 50% of the value for the fixed base
case due to the foundation flexibility. Figures 3.15 and 3.16 also show that the Iiner base
forces were as high as 35% of the shell base forces. These forces c m not be predicted
using the fixed base analysis.

150%

uLiner Deflection

50

100

1 50

200

250

300

Shear Wave Velocity, Vs ( d s )

Figure 3.13: Nonndized deflection for deep foundation

50

1 O0

150

200

250

Shear Wave Velocity, Vs (m/s)


Figure 3.14: Nomalized deflection for shailow foundation

300

90%

*Liner
+Liner

g 70%

Base Moment
Base Shear

60%

50%

(a 40%

Z 30%

U:
Cu
--%

50

1 O0

150

200

250

300

Shear Wave Velocity, Vs (m/s)


Figure 3.15: Normalized base forces for deep foundation

uSheII Base Shear


+Liner Base Moment
Liner Base Shear

-0-

50

1 O0
150
200
Shear Wave Velocity, Vs (m/s)

250

Figure 3.16: Normdized base forces for shalIow foundation

300

3.5 Summary and Conclusions


Results of the anaiysis showed the si,onificant effect of the foundation flexibility on the
chimney across-wind response and highlighted the importance of including soil-structure
interaction in the analysis of chimneys with freestanding liners.
Changes in the relative deflection are the most drarnatic. The relative deflection was
30% to 35% higher than the fixed base case for the end-bearing foundation and 21% to
32% higher in the shaliow foundation case. This increase in defIection may not cause
collapse of the structure as a whole but rnay exceed design tolerances for a chimney cap.
Also, this increase in the relative deflection may lead to liner andor shell damage if
combined with other adverse conditions such as aerodynarnic interference effects with
Ot her chimneys.

The changes in base forces are rnuch more important for design. As shown above, the
bending moment in the shell decreases si,gGficantly.

For the end-bearing foundation, it

was as low as 53% of the fixed case but ranged between 47% and 79% in the shaIIow
foundation case. The bending moment in the liner ranged between 30% and 35% of sheH
base bending moment for the fixed base case. This vaiue is quite significant given that
chimneys with freestanding liners are ernployed in regions with low seismic activity and
therefore the design of the liner would be based on minimal lateral forces. In these
situations, the liner may be under-designed as the base forces due to wind excitation rnay
exceed the seismic forces used in design.

56
Based on the results shown and the above discussion, the following conclusions are
&am:

1. the effect of the foundation flexibility on the across-wind response of chimneys with

free standing liners is significant and should be included in their analysis;


2. the relative tip deflection of the chimney increases due to the foundation flexibility

and should be accounted for in the design; and


3. the shell base forces decrease due to the foundation flexibility. The liner experiences
base forces as high as 35% of those computed for the shell assurning the fixed base
case. These forces should be considered in the design of the liner.

Chapter Four
Full and Model Scale Studies of Across Wind Buffeting

Loads on Pairs of Circular Cylinders


4.1 Introduction
This chapter deais with the amplification of across-wind forces on circular cylinders due
to the presence of an upstream cylinder. Problems like this are not new. Vickery and
Watkins (1962) studied four inline stacks in both a water and wind tunnel and reported
large amplifications of oscillations in both along and across-wind directions.
Ruscheweyh (1984) reported results from three full-scale studies of inline stacks where
the downstrearn stacks experienced large galloping type motions. Zdravkovich (1988)
reviewed what he termed 'interference-induced' oscillations of two parallel cylinders in
various arrangements and cited exarnples from wind tunnel testing, conductor cables and
chimneys .
The motivation for further work in this area stems from a full-scale study of a group of
chimneys. Unlike previously reported experiences with inline stacks, these were very
large reinforced concrete chimneys. The observed behaviour was not a galloping type
but rather an increase in the across-wind forcing due to the presence of the upstream
chimney. Subsequent to the field study, an experimental program was conducted at
Monash University, Australia. The Reynolds number achieved using the large subsonic
wind tunnel was 106cornpared to full-scale Re of -107.

4.2 Field Study


The three chimneys in the group were of the same type studied in detail in Chapter 3. Of
these chimneys, two were recently constructed (< 1 year old at the time of study) and
replaced the function of the third, which was taken out of service.

The two new

chimneys (Unit #l and Unit #3 in Figure 4.1) were identical and 250 m in height. The
diameter at 35 height

W ~ S
16.5 m-

The older stack was 335 m ta11 and the diarneter was

roughly 25% Iarger at the % height of the newer stacks. The inline spacing ratio for the
old to either of the new chimneys was 5.1 upstrearn diameters and 8.2 between the newer
chimneys.
The study was initiated after a 'clashing' occurred between the liner and shell at the tip of
the Unit #2 stack. The impact was not 'hard' but steelwork and the chimney cap were
both darnaged. While the damage was not catastrophic, it did affect the serviceability of
the chimney and the potentiai for a more severe impact between the liner and shell was a
concern.

Evidence suggested that the behaviour resulted from across-wind motions

associated with a wind direction dong the Iine from the old chimney to the Unit #2
chimney,
,4ccelerometers were installed at a height of 170m on the new chirnneys on the liner and
shell in both the North-South and the East-West axes, eight in total. The purpose of the
instrumentation was twofold; first, to confirrn and adjust the results of the modal analysis
and second, to determine the displacement as a function of wind speed and direction to
enable predictions of the maximum across wind response.

,-Interference Range

True North
Instrument

Location

@
"'O

New Stack

NewStack

Figure 4.1 :Site Plan for the Field S tudy

4.2.1 Free Vibration Analysis

Initial measurements showed the presence of two sipificant modes consistent with the
behaviour described in Chapter 3. Spectra for liner and shell accelerations are shown in
Figures 4.2 and 4.3, respectively. Two modes are apparent from the spectra, the liner
mode with a frequency of 0.205 Hz and the shell mode with a frequency of 0.255 Hz. By
integrating the spectra under the peaks, the reIative magnitude of the shell and liner
displacements were determined for each mode. The ratio of the liner motions in the liner
mode,

GLL,

to those of the shell in the liner mode, os= was 2.95. The ratio of the liner

motions in the shell mode,

GLS,

to those of the shell in the shell mode, oss was -0.48.

These results were then used dong with the soil-structure interaction mode1 outlined in
Chapter 3 to determine the mode shapes.

Frequency (Hz)
Figure 4.2: Specmm of acceleration of the Iiner at 170m.

0.1 O

0-16

0.22

0.28

0.34

Frequency (Hz)
Figure 4.3: Spectnirn of acceleration of the shelI at 170m.

0.40

The variables used to detennine the mode shapes were the reinforced concrete elastic
modulus, Ec, the brick modulus, Eb, and the rotational stiffness of the foundation, Ke. By
fitting the mode shape predictions to the observed results, the final values of the

parameters were; Ec= 34.5 GPa, Eb = 27.6 GPa and Ke = 1 . 9 8 1012


~
N-dradian. The
mode shapes determined using these input parameters are shown in Figures 4.4 and 4.5
for the liner mode and the shell mode respectively.
generalized masses obtained frorn the analysis were; fi,
f-flCll

The natural frequencies and

= 2042 Hz, Mliner= 1 8 . 4 106


~ kg,

= -2565 Hz, Md,[[ = 2.88 x 106 kg with each mode normalized to a unit tip deflection

for the shell.

Figure 4.4: Mode shape for the Liner mode.

Winer

shell

Figure 4.5s Mode shape for the sheli mode.

4.2.2 Prediction of Maximum Response

The results of the free vibration analysis were used in conjunction with the across wind
response mode1 outlined in Chapter 2 to predict the maximum expected tip amplitudes.
Due to the shoa period of study (4 months), the wind speeds associated with the
maximum response were not observed. However, an effort was made to extrapolate to
the maximum response frorn the observed events.
The variation of the RMS tip amplitude in a given mode varies strongly with rnean wind

speed and is similar to the Iift force spectrum if changes in the aerodynarnic damping are
assumed negligible. The following form is suggested.

In Equation 4.1, k is the ratio of the wind speed to the critical wind speed (V/V,d). The
value of the bandwidth parameter, B, was set at 0.35, which was representative of the
forested terrain. Using Equation 4.1, cornparisons could be made between oss,BLL,

GSL,

and Q. As outlined in Chapter 2, the RMS tip deflection at the criticai speed for a aven
mode of vibration is

where A& is the generalised mass and C is a constant for a gven chimney and is a
function of geometry, mode shape and aerodynamic parameters. Strictly speaking the
darnping,

is not constant but any changes between the two modes were assumed to be

minor.
The extrapolation mode1 was normalised to the RMS response of the shell in the shell

mode Le. ass. The response of the shell in the liner mode,

os^,

was then obtained by

multiplying k in Equation 4.1 by the natural frequency ratio, fiinrf"f~r~~


and then
multiplying crvcd by the generalised mass ratio, MfiefflfiRer.The response of the liner
was then given by the ratios discussed in the previous section. Figure 4.6 shows the
predicted behaviour.

0.5

0.6

0.7

0.8

V&,

0.9

1.O

1.1

1 -2

(shell mode)

Figure 4.6: Mode1 used for extrapolating to maximum response

The values of oss,

G L L ~ GSL, and

ou were deterrnined from the measured response

spectra. These were used to define k as this ratio varies strongly with the ratio oss/ou.
This enabled the estimation of the maximum response which occured near k = 1.18.
Over the penod of study, eight usable (0.65 < k c 0.80) records were obtained. The wind
direction was estimated from an on-site anemometer (10 m) and records from a nearby
airport.

The direction was also deduced from the response envelope (Figure 4.7),

assuming the principal axes are aligned with the dong and across wind directions.
The predicted peak shell deflections are shown in Figure 4.8. 'Clear Air1refers to those
observations where no wake interference occurred. The line shown in Figure 4.8 has no
statistical significance and is meant only to show a trend.

The best estimate of the

maximum shell deflection is 0.66 m for the wind direction inline with the old chimney.

-1 O

-5

10

15

East - West
Figure 4.7:X-Y plot of shell acceleration

10

20

30

40

50

Degrees off-line from Old Stack to Stack in Wake


Figure 4.8: Variation of pedc shell deflection with wind direction

60

For wind directions where no significant interference is present, the best estirnate is 0.15

rn to 0.18 m. This suggests that the amplification due to wake effects is equal to 3.7 with
a standard deviation of 0.7 for the case of either of the smaller chimneys directly
downstream of the larger one. For a peak shell displacement of 0.66 rn, the peak relative
motions between the shell and liner would be approxirnately 1 m, which would exceed
the nominal clearance, equal to 0.9 m.
A probability analysis of the records from the nearby airport suggested that wind causing

the most unfavourable effects would be experienced approximately once every five years.
Therefore damage to the steelwork and chimney cap would most likely be much worse
than that experienced and a failure of the liner could be expected.

4.2.3 Predictions for an lsolated Chimney

To aid in interpreting the loads encountered by the chimneys, detailed across wind
response predictions were made using the prediction mode1 outlined in Chapter 2, The

goal of this exercise was to determine the coefficients in the across-wind response mode1
to predict tip deflections consistent with the 'Clear Air' results shown in Figure 4.8.
The predictions of the peak tip amplitudes of the isolated chimney are shown in Figure
4.9.

A roughness length of z, = 0.4 m was chosen as representative of the forested

terrain. Two values of system damping (foundation

+ structural), 0.8%

of critical and

1.0% of critical were used as likely lower and upper bounds of the actual damping. The

approximate upper and lower bounds of the peak deflection (0.18 m and 0.15 m)

67

determined from the fieid results are shown with the dashed lines- The parameters used

in the prediction mode1 are given in Table 4.1.


Table 3.1: Across-wind prediction mode1 parameters for isoIated chirnney

10

20

Mean Wind Speed at Tip (m/s)


Figure 4.9: Predictions for isolated stack

30

68
Equations 4.3 - 4.5 give the values of the Strouhal number, RMS lift coefficient and the
bandwidth parameter. However, to match the field observations, the RMS lift coefficient
was modified to 0.11 from 0.17 predicted by Equation 4.4. The correlation length,t,
used is the value suggested in ACI-307 (1998).
The values of the parameters are suggested by the following equations.

where h is the aspect ratio of the chimney, i, is the longitudinal turbulence intensity and

is a modified intensity, which reflects the fact that only the scales of turbulence of the
order of the diameter have an effect on the RMS lift coefficient.

4.3 Model Study


Subsequent to the field study, a mode1 study was conducted to further investigate acrosswind buffeting loads on the downstrearn cylinder from the wake of the upstream cylinder.
The experimental program was conducted in the 1 M W wind tunnel at Monash

69
University, Austraha.

A schematic of the wind tunnel is shown in Figure 4.10.

The

lower working section was used with the width set at 4 m.


Working Section
lZmxbxIn

TWO F ~ B S1h
Figure 4.10: Schematic of the LMW wind tunnel at Monash University

4.3.1 Experiment Setup

Two circular cylinders spanning the width of the working section were placed in the flow
as shown in Figure 4.11.

The position of the downstream cylinder was fixed and the

upstream cylinder could be moved with the use of a forklift truck. The dimensions of the
setup are shown in Fi-pre 4.12. The longitudinal spacing ranged from 1500 mm to 3000

mm in steps of 250 mm. Vertical spacing r a n p d from O mm to 750 mm, also in steps of
250 mm1. Also shown in Figure 4.12 are the three upstream diameters investigated and
the height of the downstream cylinder. The downstream cylinder was mounted at this
height for ease of installation and not at the ideal mid-height of the test section.

' Sorne possible configurations were not tested due to time constraints and the large expense of operating
the wind tunnel.

Figure 4.1 1 :Photo of expermental setup

Flow Direction

Y=Oto
750 mm

Duc = 400,500
and 630 mm

X = 1500 to 3000 mm

Wind Tunnel Floor

Figure 4-12: Expenmental setup for mode1 smdy

71
The upstream cylinder was instrumented with pressure taps at the stagnation point and at
angles of 90, 180 and 270 from the stagnation point For the downstream cylinder, a
ring of 32 taps at equal angles was instailed on the centreline of the wind tunnel. Surface
pressures were recorded simultaneously at each tap at a sampling frequency of 200 Hz for
120 seconds. Due to the size of the pressure scanning equipment, it was placed outside
the wind tunnel and the required length of the tubing for the downstream cylinder was
3500 mm. The required lena$h of tubing for the upstream cylinder was 6000 mm. These
lengths posed the problem of obtaining a satisfactory frequency response of the tubing
systems.
It is important to note that the upstream cylinder cornes as close as two diameters to the
floor of the wind tunnel. This would have some impact on the fiow pattern around the
upstream cylinder and hence the orientation of the wake behind it. This has not been
examined in detail but may have an impact by changing the 'effective' transverse spacing
ratio but does not affect the principal conclusions from the experirnentsThe tubing systems were optimized using the procedure described in Holmes and Lewis
(1987). Amplitude ratios and phase angles of the proposed systems were rneasured and
compared with theoretical predictions. The predictions were computed using the theory
of Bergh and Tijdeman (1965) implemented in a cornputer program written by Dr. J.
Holmes. Making the task of obtaining a suitable frequency response simpler was the
small range of interest.

For a mean wind tunnel velocity of 30m/s, the shedding

frequency could reach a maximum of Say 30 Hz (Sr = 0.3, V = 30 mis).

Upstream Tubina

: 3500mm- 1.5mm I.D.


Connector
- 2500 mm - 3.0 mm I.D.

2750 mm - 1.5 mm I.D.

Experiment
-Prediction
O

1
Frequency (Hz)

Figure 4.13: Frequency response of tubing

10

20

Frequency (Hz)
Figure 4.14: Phase response of tubing

73

The frequency response and phase angle response of the tubing systems for both of the
cylinders are shown in Figures 4.13 and 4.14, respectively. The lenaas and diameters of
tubing are also noted. In the frequency range of interest the response was acceptable with
a maximum amplification of 8%. The phase angle was linear over the same range for
both systems, indicating only a time shift in the measurements- This presented no
difficultly as correlations between pressure fluctuations on the two cylinders was not
examined. Equipment to measure the phase angle was not available but the accuracy of
the theoreticai predictions of the amplitude ratios lent confidence to the theoretical results
for the phase angle.
4.3.1.1 Flow Properties

The mean wind speed, , was measured using two pitot tubes, one on each side of the
wind tunnel, shown upstream of the cylinders in Figure 4.11. The flow was generated
using two independentIy controlIed fans d s o shown in Figure 4.1 1. Using two pitot tubes
ensured that the mean wind speed was uniform across the width of the wind tunnel.
Profiles of the flow properties were measured using a hotwire and pitot tube
approximately 3m upstrearn of the downstream cylinder location. Mean wind speed and
spectra were estimated for five vertical and five horizontai locations. Time series from
the hotwire signal were recorded at a sampling frequency of 320 Hz for 100 seconds.

The Iongitudinal turbulence intensity and integral length scales were calculated from least
square fits of the velocity spectra. The equation used to fit the spectra was the familiar
von Karman spectmm,

wherex = f - L;'/ is the reduced frequency based on the longitudinal integral length
scale, Lz . The variance of the velocity (and thus intensity of turbulence) and the integral
length scale were estimated from the obtained fit.
An example velocity spectrum is s h o w in Figure 4.15. The abscissa in Figure 4.15 is
the reduced frequency based on the diameter of the test cylinder. It is presented in this
way to draw attention to the peak in the spectrum near a reduced frequency of 0.2 to 0.3.

Figure 4.15: Spectnirn of longitudinal turbulence

75
This peak was most likely due to fhctuations in the wake of the acoustic splitter just
downstream of the fans and upstream of the measunng station. The additional variance
under this peak was estimated graphically and the maximum value was 7% of the
variance obtained from the least square fit. The influence of this additional energy was
assumed to be rninor compared to turbulence generated in the wake of the upstrearn

cylinder.
Horizontal and vertical turbulence and mean velocity profiles are shown in Figures 4.16
The average velocity for both plots is 21.6 m/s,

and 4.17, respectively.

In both

directions, the average turbulence intensity was 5% and the average longitudinal integral
length scale was 0.14m. The standard deviation of the length scale in the horizontai
direction was -01 m and 0.03 m in the vertical direction.

4.3.2 Results for a Single Cylinder

As a baseline cornparison for the two cylinder configurations, a set of wind tunnel runs
for the single cylinder (500 mm diameter) was conducted. The Reynolds number range
was 2.2 x 10' to 10'.

Lift and drag force coefficients were formed by integrating the

pressure distribution around the ring at each time step, Le.

where C , (O,,
t)= ~('i

~ m i c

L /-p
7')-

-1500

-1000

-500

500

1 O00

Horizontal Distance from Centreline (mm)


Figure 4.16: Horizontal profiIe of flow properties

O .2

0-4

0.6

0.8

Normalized Wind Speed,


Local lntegral Length Scale (m),
Local Turbulence Intensity
Figure 4.17: Vertical profile of flow properties

1500

77
is the definition of the pressure coefficient at angle 0; from the stagnation point and at
time r. In Equation 4.5, p,,,

is the static pressure in the free stream.

The mean drag and RMS lift force coefficients, Strouhal numbers and mean base pressure
coefficient were calculated from the resulting time series. The results were corrected for
blockage effects using the ernpirical results of Modi and El-Sherbiny (1971). The
correction formulae are:

where

CD=,
CL<,SICand

are the corrected mean drag and RMS lift force coefficients,

Strouhal number and mean base pressure coefficient, respectively. The subscript c
denotes the corrected value and f is the solid blockage ratio, equal to 10% for al1 test
configurations. All results presented below are the corrected values; the subscript

is

omitted. Also, al1 coefficients are normalized by the mean free stream wind speed.
The coefficients are plotted in Figure 4.18 as a function of Reynolds number, The mean
drag coefficient increases from a minimum value of 0.27 at Re = 2.2 x 10' to a maximum

of 0.46 at Re = 10'. The mean base pressure coefficient follows the same trend, reaching

2x10~

5x10~

Reynolds Number

Figure 4.18: Dependence of force coefficients and S, on Re

a maximum value of 0.625. No discemible Strouhal number was observed below Re =


6.5 x IO', above this value, S, is near constant and equal to 0.26. The maximum value of
the RMS Iift coefficient is 0.30 at Re = 2.2 x 10' and approaches a constant value of
0.130 where strong vortex shedding was observed.
One key observation from Figure 4.18 is the 'leftward shift' of the drag coefficient curve
compared to results for a smooth cylinder in smooth uniform flow. The drag coefficient
appears to be in the classical supercritical regime.
The Strouhd nurnber was determined frorn a fit of the spectra of the lift coefficients. The
spectrum of the lift coefficient for Re = 106 is shown in Figure 4.19. The spectral peak
was fitted to the equation,

Ifs -f

where k is the fraction of the total variance under the peak, B is a bandwidth pararneter
and& is the centrai shedding frequency. The solid line shows the fit with the parameters
shown in the legend. The abscissa is the reduced frequency based on the diameter and
the mean velocity of the free Stream.
The spectnirn reveais a si,gnificant arnount of energy outsidz the peak, which is due to the
lateral component of turbulence in the incident flow. A portion of the energy under the
peak is also due to lateral turbulence but no effort was made to distinguish between the
Iift force due to shedding and that due to turbulence.

0.0

0.1

0.2

0.3

0.4

Reduced Frequency f D N
Figure 4.19: S p e c m of lift coefficient for a single cylinder at Re = 106

0.5

4.3.3 Results from Two Cylinder Configurations

The two cylinder configurations were also tested over a range in Re. However, due to
space limitations, only results for the Re = 106 are presented here. The complete set of
results is reported in Galsworthy (2000).

4.3.3.1Case 1 : 630 mm - 500 mm Cylinders


In this stage of the mode1 studies the ratio of the upstrearn to downstream diameter
matched that of the field study. The diameter of the upstream cylinder was 630mrn (Duc)

and the diameter of the downstream cylinder was 500rnm. The spacing between the
cylinders ranged frorn 1750mrn to 3000rnm in the longitudinal direction (X) and Ornrn to

750mm in the transverse direction (Y).


The vortex shedding from the downstream cylinder was exarnined in detail.

The

spectrum of the lift coefficient for the downstream cylinder for X/Duc=3. 17 and Y/DUc=O
is shown in Figure 4.20. The format of the spectrum is the sarne as Figure 4.19 with the
reduced frequency again based on the mean free Stream veiocity.

Observe that the

Strouhal number is reduced to 0.206 compared to the single cylinder value of 0.260. An
explanation for this is apparent rom the differential pressure spectrum ( 9 0 and 2 7 0 ~from

the stagnation point) for the upstream cylinder, shown in Figure 4.21. The shedding
frequency for the upstream cyiinder matches the downstream cylinder. This coalescence
of the shedding frequencies was present for al1 configurations as shown by the spectra of
the lift coefficient and upstream differential pressure for the largest separation
investigated, Figures 4.22 and 4.23, respectively.

0.0

0.1

0.2

0.3

O .4

0.5

Reduced Frequency f D N
Figure 4.20: Spectnun of downstream cylinder lift coefficient for X/Duc = 3.17 and Y/Duc = 0-

0.O

0.1

0.2

0.3

0.4

0-5

Reduced Frequency f D N
Figure 4.2 1: Specmm of differential pressures on upstream cylinder for X/Due3.17 and Y/DUc=O.

0.0

0.1

0.2

0.3

0.4

0-5

Reduced Frequency f D N
Figure 4.22: Spectnim of downstrearn cylinder Lift coefficient for XIDUC= 4.75 and Y/DuC= 0.

0.0

0.1

02

0.3

0.4

0.5

Reduced Frequency fDN


Figure 4.23: Spectnirn of differentid pressures on upstream cylinder for X/Duc = 4.75 and Y/Duc=O-

A summary of the Strouhal nurnber for the downstream cylinder is shown in Figure 4-24

for al1 configurations. T h e two dashed lines show the value of St for the single cylinder

and 80% of this value, which is equal to the diameter ratio of 0.8. The results confirm
that the strong lateral turbulence generated by the upstrearn cylinder controls the
shedding from the downstream cylinder. However, there is a slight increase in S, with
increasing separation suggesting that the flow around the upstream cylinder is not
independent of the downstream cylinder.

Figure 4.24: S,(downstream cylinder) for al1 630mrn-500mrn configurations

Further evidence of the interaction between the two cylinders is shown by the summary
of the mean base pressure for the upstrearn cylinder in Figure 4.25. One can observe that
the base pressure increases with increasing separation of the two cylinders.

An

explanation for this observation is not readily apparent but it is clear that the flow field

84

around the upstrearn cylinder is different than that of an isolated cylinder, even at the
larger spacing ratios investigated-

Figure 4.25: Mean base pressure coefficient (upstream cylinder) for al1 630mm-500mmconfigurations

The RMS lift coefficient for the downstream cylinder is shown in Figure 4.26 for al1
configurations. The dashed Iine shows the value observed for the single cylinder case.
One can observe Siat the increase in the fluctuating Lift force is quite significant for al1
cases. The increase is most prevalent for the tandem configurations and the maximum
observed value is 0.608. This result makes sense since the lateral intensity of turbulence
is geatest on the centreline of the wake thus one would expect that the increase due to

buffeting would be largest at this location. The magnitude of the RMS lift coefficient
decays with increasing separation of the cylinders, which follows the decay of lateral
turbulence in the wake of the upstream cylinder.

85
The mean drag coeffkient for the dowristrearn cylinder is shown in Figure 4.37 for al1
configurations. The dashed line shows the observed value for the single cyiinder case.
An interesting observation is that for a given Iateral separation, the mean drag coefficient

rernains relatively unchanged for increasing longitudinal separation, For cases where the
downstrearn cylinder is 'shielded' by the upstream cylinder (Y/Duc =O and Y/Duc =0.4),
there is a deficit in the mean drag coefficient. For the other configurations, the mean drag
coefficient is higher.

This is in agreement with observations at subcriticai Re

(Zdravkovich, 1977) and the suggested mechanism is flow speed up through the gap
between the cylinders. M l e this observation is in agreement, the discontinuity in the
mean drag coefficient at a spacing of approximately 3.8 diameters observed by
Zdravkovich and Pndden (1977) is not present.

Figure 4.26: RMS lift coefficient (downstrezun cylinder) for al1 630mm-500mconfigurations

Figure 4.27: Mean drag coefficient (downstreamcylinder) for a11 630rnrn-500mconfigurations

4.3.3.2Case 2:500 mm - 500 mm Cylinders


In this stage of the mode1 studies, the two cylinders were identicai with a diameter of 500

mm. The range in spacing between the cylinders was from 1500 mm to 3000 mm in the
longitudinal direction and O mm to 750 mm in the transverse direction.
The spectra for the lift coefficient for the downstream cylinder and differential pressures
from the upstream cylinder for XIDuc = 3.0 and Y/Duc = O are shown in Figures 4.28 and
4.29, respectively. Again, as for the previous case, the vortex shedding frequency from
both cylinders coalesces as shown by the matching peaks in both spectra. This might be
expected for the small separation but what was somewhat surprising was that S t for both
cylinders essentiaily matched for al1 spacing ratios investigated as shown in Figures 4.30
- 4.33. The 'single peak' behaviour was observed in al1 configurations.

0.0

0.7

02

0-3

0-4

0.5

Reduced Frequency fDN


Figure 4.28: Specaum of downstrciam cylinder lift coefficient for X D u c = 3.0 and Y/Duc = O

100

1 O-'

1 0-2
0-0

0.1

0.2

0.3

0.4

0.5

Reduced Frequency f D N
Figure 4.39: Specmm of differential pressures on upstream cylinder at X/Duc = 3.0 and Y/DUc= 0.

02

0.3

0-4

Reduced Frequency f D N
Figure 4-30: Spectnun of downstream cylinder lift coefficient for X/Duc= 4.5 and Y/Duc = O

1ou

1O-'

10-2

0.0

0.1

02

O -3

0.4

O -5

Reduced Frequency f D N
Figure 4.3 1: Spectrum of differential pressures for upstream cylinder at X/Duc = 4.5 and Y/Duc = 0.

0.0

0.1

0.2

0.3

O -4

0.5

Reduced Frequency f D N
Figure 4-32: Spectrurn of downstream cylinder lifi coefficient for X D u c = 6.0 and Y/Duc = O

0.0

0.1

02

0.3

0.4

0.5

Reduced Frequency f D N
Figure 4.33: Spectrum of differential pressures on upstream cylinder at X/Duc= 6.a and Y/Duc= 0.

The Strouhal number for the downstream cylinder is summarized in Fiboure 4-34.The
dashed line shows the single cyIinder value- One can observe that the Strouhal number
increases for increasing lateral separation of the cylinders.

Also, for a given Iateral

spacing ratio, S t rernains essentially constant for longitudinal separations greater than 3
diameters.
The mean base pressure for the upstream cylinder is shown in Figure 4.35.
behaviour observed for the 630 mm

The

500 mm pair was not observed here. The base

pressure for the upstream cyfinder remained essentially unchanged for ail the
configurations. One observation of interest though is that the base pressure is somewhat
higher than the value observed for the single cylinder case (Figure 4.18).

Figure 4.34: S, (downstream cylinder) for al1 500rnrn-500mconfigurations

Figure 4.35: Mean base pressure coefficient (upstrem cylinder) for al1 500mm-500mmconfigurations

The RMS lift coefficient for the downstream cylinder is sumrnarized in Figure 4.36 with
the observed value for the single cylinder shown by the dashed iine. As for the previous
case, the increase in the rnaa@ude

of the fluctuating lift force is significant in al1

configurations with the maximum observed value being 0.394.

The decay in the

magnitude with increasing separation is again present.


The mean drag coefficient for the downstream c y h d e r is s h o w in Figure 4.37. The
same behaviour as the previous case is observed here with a drag deficit for the Y/DLIC=O
and Y/Duc=O.S configurations and increased drag for the other configurations due to flow
speed up in the gap. For the largest separation of 6 diameters, the mean drag appears as if
it is starting to approach the value for the single cylinder case.

Figure 4.36: RMS lift coefficient (downstream cylinder) for al1 500rnm-500mmconfigurations

Figure 4.37: Mean drag coefficient (downstream cylinder) for al1 500rnm-500mmconfigurations

4.3.3.2Case 3:400 mm - 500 mm Cylinders


In this case, the upstream diameter was 400mrn (Duc) and the downstrearn cylinder was
500mrn. The longitudinal spacing ranged from 1500mm to 2500mm and the transverse
spacing ranged from Omrn to 750mm.
For this case, the behaviour differed from the previous two cases. The spectra for the lift
coefficient of the downstream cylinder and the differentid pressure for the upstrearn
cylinder for X/Duc=3.75 and Y/DUc=O are shown in Figures 4.38 and 4.39, respectively.
Lift coefficient spectra for the downstream cylinder in tandem configurations are also
shown in Figure 4.40 for X/DUc=5 and in Figure 4.41 for X/Dvc=6.25- Here a fit of the
spectra of the lift coefficient was not attempted due to the presence of two peaks.
However, the relative contribution of each peak is shown by the spectral distribution
shown on the second vertical mis. The two peaks are noted by the two values of Strouhal
nurnber2. The higher frequency peak is due to shedding from the upstrearn cylinder as
shown by the differential pressure spectrum in Figure 4.39. As the separation between
the cylinders increases, the peaks separate and the relative contribution to the fluctuating
lift force on the downstream cylinder shifts from being dominated by laterai buffeting
(Figure 4.38) to being dominated by vortex shedding from the downstream cylinder
(Figure 4.41).

' The notion of two Strouhal numbers, while not accurate, is chosen as a convenient way to distinguish
between the two peaks.

0.0

0.1

0.2

0.3

0-4

0.5

Reduced Frequency f D N
Fi,ve 4.38: Spectnun of downstream cylinder lift coefficient for X/Duc = 3.75 and Y/Duc = O

1O-'

O .O

O. 1

0.2

0.3

0 -4

0.5

Reduced Frequency f D N

Figure 4.39: Spectnirn of differential pressures for upstrearn cylinder at X/Duc = 3.75 and Y/Duc = O

0.0

0.1

02

0.3

0.5

Reduced Frequency f D N

Figure 4.40: Specmm of downstream cylinder lit? coefficient for XDuc = 5.0 and YDUc= 0.

0-0

0.1

02

0.3

0.4

0-5

Reduced Frequency f D N
Figure 4.41: Spectrum of downstream cylinder lift coefficient for X/Duc = 6-25 and Y/Duc = O

The two Strouhd numbers are summarized in Figure 4-43 with two dashed lines at
St=0.26 from the single cylinder case and 1.25 x 0.26 = 0.325 representing the shedding

frequency for the upstream cylinder if it was isolated. The frequency of shedding from
the upstream cylinder increases with increased separation of the cylinders, shown by the
upper set of curves.

The shedding frequency for the downstream cylinder remains

approximately constant and slightly below the value observed for the single cylinder for

dl configurations,
Unlike the previous two cases, the RMS lift coefficient due to lateral buffeting could be
quantified using the spectral distribution curves shown above. The portion of the RMS
1st coefficient due to shedding from the downstream cylinder was also quantified and the
value remained near constant at

CL,kums
= 0.11with

a coefficient of variation of less

thanlO% over al1 configurations. The portion due to buffeting is shown in Figure 4.43
for al1 configurations. Observe that

CL,is largest (0.165)

when the cylinders are closest

and decays with increased separation, which agrees with the previous cases. Also, the
magnitudes are largest for the tandem configurations.
The mean drag coefficient for the downstream cylinder is shown in Figure 4.44. As for
the previous two cases, the sarne trends are evident with reduced drag for the shielded
configurations OT/Duc=O and Y/DUc=0.62) and higher drag for YDuc= 1.25.

Buffeting

S hedding

Figure 4.42: S,for Iateral buffeting and vortex shedding for ail 400mm-500mmconfigurations

Figure 4.43: R M S lift coefficient associated with lateral buffeting

Figure 4.44: Mean drag coefficient (downstream cylinder) for al1 400m.m-500mmconfigurations

4.4 Discussion and Comparison of Results


The coalescence of the vortex shedding and buffeting peaks in the lift coefficient spectra
for the first mode1 case provides some insight into the field observations. While it is not
entirely clear what affect the downstrearn cylinder has on the upstream cylinder, the
existence of only one peak in the lift spectra could be expected. If both cylinders were
exposed to the same free Stream wind speed, the central shedding frequency for the larger
cylinder would be 80% of the frequency for the smaller cylinder.

When the larger

cylinder is upstream of the smaller one, then the mean velocity deficit in the wake would
reduce the shedding frequency for the latter. If this deficit is -2096, then the two peaks

99
would rnerge and the vortex shedding response of the smaller diameter cylinder would be
ampiified by the increased lateral turbulence-

For the second mode1 case, the existence of only one peak in the lift spectra is somewhat
surprising. An explmation of the observed behaviour is not readily apparent but it could
be a similar phenornenon to the 'lock-in' of the vortex shedding frequency to the
frequency of vibration, i-e. the vortex shedding frequency of the downstream cylinder
'locks in7to the shedding frequency for the upstream cylinder.

In the third case, the mechanism affecting the frequency of the lift forces are opposite to
the first case. The central shedding frequency for the upstream cylinder would be 35%
higher than the downstream cylinder.

Also, the velocity deficit in the wake of the

upstream cylinder would separate the two frequencies further as the results showed-

For the first two cases, an amplification of the vortex shedding forces can be computed
directly. The across wind response of a chimney near the critical speed is of the form:

where C is a constant for a aven chimney and is a function of mode shape and geometry,

& is the structural damping, m, is the reduced mass, C

is a correlation parameter near 1.2

diarneters and D is a characteristic diameter. The aerodynamic damping term, Ka is


negative near the cntical speed and for large reinforced concrete chimneys is equal to the
value at small amplitudes of roughly -0.6.

100
Due to the lack of information to support otherwise, dl variables in Equation 4.8 are
treated as constants except for the RMS lifi coefficient and the Strouhal number. The
suggested amplification of the lift force on an isolated stationary cylinder is then given by
the ratio,

[z2)
single cylinder

where the lift coefficient for the single cylinder is taken from the mode1 study. The total
value is used as opposed to that predicted by the fitted curve in Figure 4.19.
The results of Equation 4.9 are shown in Figure 4.45 for the case of the larger cylinder
upstrearn. Also included is the predicted ratio from the field study for the inline case.
The predicted amplifications are very large reaching a maximum of close to 8 times the
fluctuang lift force on a stationary isolated cylinder.
amplification decays but is still quite significant.

At larger separations, the

The curves in Figure 4.45 are

extrapolated to a spacing of 5.1 diameters for cornparison with the field result. The
amplification of the lift forces from the inline case is 4.7, which is somewhat higher than
the full scale prediction given in $4.2 (3.7 -+ 0.7) but is not unreasonable given the
uncertainties in both studies.
The results for the second case of the mode1 study are shown in Figure 4.46. Again, the
suggested amplification of the lift force is quite large reaching a maximum of close to 4.

Y/Duc = O
Y/Duc = 0.4
Y/& = 0.8
Y/&,
= 1.2

+
+

Data Points

Full Scale Observation


I

Fi,oure 4.45: AmpIification of lift forces for al1 630mm-500mmconfigurations

Figure 4.46: Amplification of lift forces for al1 500mm-500mmconfigurations

4.5 Summary
The resdts of a field and model study of wind loads on pairs of circular cylinders are
presented. In the field study, it was observed that a chimney in the wake of a chimney
25% larger in diarneter expenenced an increase in the across-wind response by

approximately 3.7 times that it expenenced for wind directions where no interference
occurredFor the model study three diarneter ratios were exarnined.

In the first case, which

con-esponded to the field study, the shedding peak from both the upstream and
downstream cylinders merged which greatly increased the fluctuating lift force compared
to the single cylinder. In the second case, the cylinders were identical and again it was
observed that the two peaks rnerged. For the case where the upstream diarneter was
smaller, the two peaks were separated and the fluctuating lift force associated with
buffeting decayed with increasing separation of the cylinders.

Chapter Five
Study of the Effect of Appurtenances on the Across Wind
Response of Chimneys

This chapter deals with the adverse affects on the aerodynamic loads caused by
modifying the cross section of circular cylinders. The motivation for study in this area
resulted from a field study of two large reinforced concrete chirnneys. Large amplitude
vibrations resulted in a hard impact between the shell and liner and caused a failure in the
liner on one of the stacks.
The chimneys studied were part of Unit#l and Unite

at the Cod Creek Power Station

located near Bismarck, North Dakota. They are shown in Figure 5.1 with Unit #3 on the
ripht side of the photograph. Both stacks were identical with a height of 200 m. The
base diameter of the reinforced concrete shell was 16.75rn and tapered to a diameter of

9.3 m at the tip. The enclosed freestanding brick liner had an average diameter of 8.3 m.
The clearance between the liner and shell at the tip was 0.85 m. The centre-to-centre
spacing of the chimneys was 8.25 diameters, based on the diarneter at % height.
One c m observe frorn Figure 5.2 that a m d f t enclosure is attached to each of the

chimneys up to an elevation of approximately !h of the height. This enclosure was added


to allow year round access to the ernissions ports located at an elevation of 110 m. Prior
to the installation of the enclosure, the manlift rails would ice over in the severe winters

Figure S. 1 :View of C o d Creek Power Plant

Figure 5.2: Unit #2 Chirnney

due to water vapour ernitted from the cooIing towers located to the west of the chimneys.
The enclosures were at an angle of 3iO0, measured clockwise fiom north.

When

originally installed the enclosures spanned the full height of the chimneys to d s o permit
access to the aircraft waming lights near the tip of the chimneys. Prior to this installation,
the chimneys had been in service some 20 years,
Shortly after the enclosures were added to the chimneys, workers at the power plant
noticed that the Unit#2 stack was vibrating with tip amplitudes visible from the ground
on a windy day. Upon inspection, they discovered that the chimney cap at the top of the
liner was severely darnaged and the bnckwork was breaking loose and plurnmeting some
lOOm to the emissions monitoring level.

This obviously rendered the chimney

unserviceable and corrective actions were attempted; the chirnney cap was removed and
the liner was dismantled down to an elevation of approximately 150m, but the manlift
enclosure was left in place.
While this ternporary measure removed the danger of falling bnckwork and allowed
operation of the unit to continue, it did nothing to address the altered aerodynamic forces.
The addition of the enclosure would alter the mean drag force and likely cause a nonzero
mean Iift force on the chimneys. Also, an increase in the strength of the vortex shedding
combined with any changes in aerodynarnic damping would affect the fluctuating lift
response, as evidenced by the observations.

In addition to the still present aerodynarnic problems, removal of the top portion of the
liner worsened the situation in a structural sense. First, the generalized mass of the
chimney was reduced which would result in a higher amplitude response. Second, the

106
nonlinear action of the impact between the liner and shell at the tip, which helped to limit
the amplitude, was removed.
The Boundary Layer Wind Tunnel Laboratory was contacted to investigate the cause of
the observed behaviour and suggest irnmediate corrective measures. Shortly thereafter,
the top portion of the manlift enclosure was removed from each chimney. Subsequent to
this action, wind tunnel studies were c d e d out to define the loads on the chirnney with
the enclosure spanning both the entire height and half the height of the chimneys. It was
also agreed that the chimneys would be monitored for a perod of six months to validate
the predictions based on the wind tunnel experiments.

In this chapter, some of the results from both the field and rnodel studies are presented.
The model study was conducted in two stages in BLWTL II at the University of Western
Ontario. In the first stage, the aim was to estirnate the aerodynamic and aeroelastic forces
acting on the full scde chimneys. This included measurements of surface pressures on a
rgd cyIinder and aerodynamic darnping measurements on an aeroelastic model, which
was the sarne size as the pressure model. The second stage was conducted after the
conclusion of the field study. The purpose of this study was to investigate the influence
of size of appurtenance on the aerodynamic forces (rigd pressure model).

The

motivation for these experiments was the prevalence of ladders, piping etc. on steel
chimneys, which can approach the size (relative to diarneter) of the rnanlift enclosure on
the North Dakota chimneys. The results for the first stage are presented in $5.2 and those
for the second stage are presented in 55.5.

5.2 Wind Tunnel Study


The wind tunnel experiments were conducted in the high speed section of the wind tunnel
at the Boundary Layer Wind Tunnel Laboratory at the University of Western Ontario. A
photograph of the setup is shown in Figure 5.3. The mode1 was 0.324 m (12 % in) in
diarneter and a mock enclosure made of wood spanned the height of the wind tunnel. For
the set of experirnents where the enclosure over the top half of the full-scale chimney was
simuIated, the radiai dimension (outstand) of the enclosure was 14%of the diameter and
the tangential dimension was 28% of the diarneter. For the set of experiments where the
lower half was modelIed, the corresponding dimensions were 10% and 20%. The outside
corners of each mock enclosure were rounded to match the prototype.
The cylinder was instnimented with a ring of 36 pressure taps at the mid-height and 4
rings of 8 taps, 2 above and 2 below, to measure the spanwise correlation of the
fluctuating lift force. An additional 5 taps, three on the surface tangential to the cylinder
and 1 each on the surfaces normal to the surface, were installed on each endosure.

5.2.1 Flow Properties

The testing was conducted near the inlet of the wind tunnel to achieve a nominal twodimensional flow. A grid was designed (shown upstream of the mode1 in Figure 5.3) to
increase the turbulence level in the free Stream and promote an earlier transition in the
body boundary layer or increase the 'effective' Reynolds number. The mesh size, M of
the grid was -254 m (10 in) and the bar size, b was .O318 m (1% in). To determine the

Figure 5.3: Experiment set up in BLWT II at UWO.

Figure 5.4: Distribution of mean wind speed across one ce11 of a uniform grid (Vickery, 1968)

optimal position of the cylinder downstream of the grid, x, the results of Vickery (1968)
were used.
To achieve a uniform mean wind tunnel velocity across the grid, the cylinder was
positioned IOM (2.54 m) downstream of the grid, which corresponds to 80 bar widths.
According to the results shown in Figure 5.4, this would give a uniforrn mean wind speed
across the grid. The longitudinal intensity of turbulence for this position is suggested in
Figure 5.5 at roughly 5%.

The anticipated flow properties were confirmed with

measurements of horizontal and vertical profiles of the turbulence intensity and mean
wind speed. A typical profile is shown in Figure 5.6.
The testing was conducted over a range of Reynolds number from 3.5 x 10' to 5 x IO',
based on the nominal diameter of the cylinder. This was done to examine the sensitivity
of the results.

Figure 5.5:Turbulence intensity downstream of a u ~ i f o r mgrid (Vickery, 1968)

Local lntensity
-0

0.1
1

0-2
1

O -3

0.t

O -5
I

Normalized Mean Speed


Figure 5.6: Vertical profile of turbulence intensity and mean speed 10M downstream of @d

5.2.2 Surface Pressure and Force Measurements


The testing consisted of rotating the cylinder to achieve the desired angle between the
mock enclosure and the oncorning fiow and recording time series at a sarnping frequency
of 400 Hz for 120 seconds at each tap simultaneously. The angle of attack of the wind,

a, ranged from 0' to 180' in steps of 5' or 10'.

Drag and lift force coefficients were

formed from the taps on the centre ring on the mode1 after the data was gathered. The lifi
force component of the pressures from the 4 'side' rings were also formed into separate
coefficients. The resulting time series were analyzed and the time averaged values of the

111

drag and lifi force coeffici. ents and the spectral energy distribution of the lift force were
examined in detail.
The results were correctedl for blockage effects using the empincal results of Modi and
El-Sherbiny (1971). The caorrection formuIae are:
(5.la)

(5.1b)

(5.1~)

(5.ld)

where

Dr,
CLr,S,

and CPc
are the corrected rnean drag and RMS lift force coefficients,

Strouhai number and mean pressure coefficient, respectively. The subscript c denotes the
corrected value and

$ is t h e solid blockage ratio, equd to approximately 10% for al1 test

configurations. Al1 results presented below are the corrected values; the subscript c is
ornitted.
The spectrurn of the lift coefficient for Re = 5 x 10' and the mean pressure distribution
around the cylinder are shown in Figure 5.7 and Figure 5.8, respectively. These results
are for the largest enclosure and an angle of attack of zero degrees. The spectrum is
normalized by the RMS liFt coefficient due to shedding,
fitting the spectral data to t h e equation,

EL,,which was determined by

The solid line shows the obtained fit. The form of the spectrum and the magnitude of the
fluctuating Iift force are comparable to a circular cylinder.

Vortex shedding was

observed at al1 values of Re investigated- The mean pressure distribution shows very
little sensitivity to Re. The largest difference between the distribution for the cylinder
with the attached appurtenance and the plain cylinder is within a

I
30- sector

around the

stagnation point. This is due to the separation of the boundary layer frorn either side of
the enclosure and then the reattachment at approximately 30. Also, the rna,&udes of
both the base pressure and the side pressures are slightly larger than the circular cylinder.

0.0

O. 1

0-2

0.3

Reduced Frequency fDN

Figure 5.7: S p e c t m of lift coefficient for a=OO

0.4

0.5

Shaft Angle = O0
---O---

45

90

135

180

.-.-

R e = 4 . 0 x 10%

225

b!

270

315

360

Angle (degrees)
Figure 5.8: Mean pressure distribution for =O0

The results for the lift spectnirn shown above are indicative of the observations for angles
of attack up to 30. At an angle of attack of 40' to 45' however, a rernarkable transition
occurred as the spectrum of the lift coefficient shown in Figure 5.9 demonstrates. The
strength of the vortex shedding increased drarnaticdly and was accornpanied by a shift in
the Strouhal number to a value of -0.2- The mean pressure distribution shown in Figure
5-10 demonstrates that the boundary layer separates from the back edge of the enclosure

but does not reattach. This creates a wider wake (also suggested by the shift in S,) and
generates a mean lift force. A sketch of the suggested flow pattern is shown in Figure
5.11. The separation points were determined by the intersections shown in Figure 5.10

(Kwok, 1986).

0.0

0-1

0.2

0.3

0.4

Reduced Frequency fDN

Figure 5.9: Spectrurn of lift coefficient at d 5 "

1.5

e R e = 3 . 5 ~10%

0-

--O -Re=4-0~105

0.5

Shaft Angle = 45"

-U-

Re=4.5~10%

- -&- R e = 5 . 0 x l O N i

Se~arationPoints
50' and 248'

Angle (degrees)

Figure 5.10: Mean pressure distribution at a=45"

0.5

Figure 5.1 1: Sketch of flow pattern around cy Iinder

Another change in the behaviour is sugested by the n m o w e d bandwidth in the spectrum


of the lift coefficient is an increase in the spanwise correlation of the shedding. This is
confirmed by the coherence fnction

for n = 0' and a = 45O shown in Figure 5.12.


Similar behaviour was observed at angles of attack, 45' < cc < llOO. Spectra the lift
coefficient for cc = 70 and cc = 90' are shown in Figures 5.13 and 5.15, respectively and
the corresponding mean pressure distributions are shown in Figures 5.14 and 5.16. The
same observations were made for the set of tests where the enclosure over the lower half
of the full-scale chimney was modelled.

0.5

1.5

2.5

0.4

0.5

Az (diameters)
Figure 5-12: Spanwise correlation of lift force

0.0

0.1

02

0.3

Reduced Frequency fDN

Figure 5.13: Spectrurn of lift coefficient for -70"

d R e = 3 - 5 x 1 0 h 5

- -O--Re=4.OxlC%
Shaft Angle = 70

45

90

135

-U-

Re=4.5xlONi

--&--

Re=5.0x10Ni

225

180

270

360

315

Angle (degrees)
Figure 5.14: Mean pressure distribution at u=70

0.0

0.1

0.2

0.3

Reduced Frequency f D N

Figure 5-15:Spectrum of lift coefficient at =9O0

0.4

0.5

45

90

135

225

180

270

315

360

Angle (degrees)

Figure 5.16: Mean pressure distribution at a=90

For angles of attack greater than 70,the flow pattem changes in that the separation point
moves to the leading edge of the appurtenance.

This is evidenced by t h e pressure

distribution in Figure 5.14. The minimum pressure is at the leading edge a n d there is a
very steep positive pressure gradient across the surface of the appurtenance leading to a
separation at the trailing edge. The pressure distribution for an angle of attack of 90"
(Figure 5.16) shows that the separation point has moved to the leading edge of the
appurtenance.
As mentioned above, the change in flow pattern around the cylinder would result in a
nonzero mean lift force. A sumrnary of the observed mean drag and lift force coefficients
for the larger and smaller encIosures are shown in Figures 5.17 and 5.18, respectively.
The change in flow pattem is again evidenced by the change in the meanoforce

coefficients starting at cc = 30' for the larger enclosure and at cc = 40" for the smaller

45

135

90
Angle of Attack,

cc (degrees)

Figure 5.17: Mean lift and drag coefficients for the larger enclosure.

45

90

135

Angle of Attack, a (degrees)


Figure 5.18: Mean lift and drag coefficients for the smaller encIoswe.

120
enclosure. For the larger enclosure, the mean drag coefficient increases to a maximum of
1.33 at a = 90 and the mean lift coefficient reaches a maximum (rna,pitude) of 1.7 at CC

= 100. This gives a resultant mean force coefficient of approximately 2.1.

For the

smdler enclosure, the magnitudes are approximately the same.


Perhaps more interesting than the magnitudes of the mean force coefficients is the
implication for the quasi-steady aerodynamic damping forces.

Den Hartog (1956)

showed that a cross section is prone to a galloping instability if the sum of the slope of
the rnean Iift coefficient and the mean drag coefficient is less than zero, Le.

Figures 5.17 and 5.18 show a strong negative dope in the mean lift coefficient near an
angle of attack of 40 to 45'.

The value of Equation 5.4 in this region is -2.7 for the

larger enclosure and -3.8 for the smaller enclosure.


A surnmary of the observed RMS lift coefficient (total) for both enclosures is shown in

Figures 5.19 and 5.20, respectively. Again, the behaviour noted above is shown clearly
by the changes in magnitude of the flucniating lift force. The magnitude reaches a
maximum of 0.47 near an angle of attack of 50' for the larger enclosure and 0.45 at 55'
for the smaller enclosure. The sensitivity to Re is more apparent here with the maximum
ma-gntude varying by some 15% over the range investigated. The 'clip' at 70' can be
explained by the change in separation points from the back edge of the enclosure to the
leading edge. This change is also present in Figures 5.17 and 5.18.

45

135

90

Angle of Attack, a (degrees)


Figure 5.19: RMS lift coefficient for the larger enclosure.

45

135

90

Angle of Attack,

a (degrees)

Figure 5.20: R M S lift coefficient for smdler enclosure.

1 80

133

The variation in Strouhal number with ande of attack is summarized in Figures 5.21 and
5.22. The drop in the shedding frequency around a = 45' as discussed above, is due to

the widening of the wake because of the separation from the enclosure- The Strouhal
number remains relatively constant up to a = 90' and then starts to increase up to a
maximum near a = 150'. Beyond this angle of attack, no discernible shedding peak was
observed in the lift spectra. A possible explanation for this is that the enclosure behaves
similar to a splitter plate and disrupts the shedding (Roshko, 1955).

52-3Aerodynamic Damping Measurements


After the pressure and force measurements on the

riadcylinder were completed, a set of

tests were conducted on a large aeroelastic model to confinn the implied aerodynamic
instability from the mean force measurernents by estimating the negative aerodynamic
damping near the reduced velocity associated with vortex shedding. For this stage of the
experirnents, only the larger enclosure was tested.
The mode1 consisted of a rigid base pivoted aluminum cylinder constrained to vibrate
only in the across wind direction with the stifmess provided by two springs attached at
the top of the model. For the springs used, the naturd frequency of the model was 10 Hz.
The procedure for the testing was as follows.

The model was held stationary from

outside of the wind tunnel using a long bar. It was then released and the growth of the
oscillation was recorded using an accelerometer mounted on the top of the model. The
time histories were then analyzed using a negative logarithrnic decrement technique to
deduce the aerodynamic darnping. This was repeated severd times at each velocity to
give an average value.

45

90

1 35

Angle of Attack, or (degrees)


Figure 5.2 1: Strouhai number for large enclosure.

45

90

135

Angle of Attack, a (degrees)


Figure 5-22: Strouhal number for smaller enclosure.

Three angles of attack were tested (45O, 67.5' and 90), which were chosen to give a
sufficient coverage of the worst cases for the full scale chimney. The results are shown
in Figures 5.23 - 5.25. A quasi-steady analysis of the fluctuating Lift force suggests that
the mass-damping parameter,

U
rn
where - is the reduced velocity, - is the reduced mass and & is the aerodynamic
foD

PD'

damping as a fraction of critical. The dashed lines shown in Figures 5.23 and 5.24 are the
quasi-steady values suggested by the mean force coefficient measurements. There are no

Quasi-Steady

Figure 5.23: Mass-dampinp parameter for a=45"

5.5

4.5

6.5

7.5

U/f,D
Figure 5.24: Mass-damping parameter for a=67.S0

U / foD
Figure 5.25: Mass-damping parameter for a=90

126

quasi-steady values shown in Fi,gre

5.25 because the mean force measurements

suggested that at this angle of attack the cross section was stable. At an angle of attack of

4s0,the aerodynamic damping is strongly nepative

above a reduced velocity of 5.5 and

agrees well with the quasi-steady values. For an angle of attack of 67S0, the spine of the
mass-damping parameter follows the quasi-steady predictions. At the critical reduced
velocity for vortex shedding (US,), the aerodynamic damping becomes strongly negative
and departs from the quasi-steady spine- The maximum value of the mass-damping
parameter is between 0.75 and 0.80. At 90, the mas-damping parameter again increases
with increasing velocity and reaches a maximum of

- 0.75 at a reduced velocity of 5.75.

5.3 Prototype Predictions


The data gathered from the wind tunnel study was used to make predictions of the
prototype chimneys using the mode1 of Vickery and Basu (1983). One must always
proceed with caution in this direction because of the obvious Re rnisrnatch; Re 5 5 x 105
model cornpared to Re

G 10.'

for the prototype. However, the tests on the rigid model

suggested that the enclosure controlled the separation point on the cylinder. Also, the

main purpose of making the predictions was to confirm that the addition of the enclosure
altered the aerodynarnics sufficiently to cause the observed prototype behaviour, and not
to provide d e s i g data. For the predictions presented below, the RMS lift coefficient,
Strouhal number and mean lift force coefficients input into the across wind response
model were taken directly from the wind tunnel tests. The values used were modified for
the top and bottom halves of the chimney as appropriate. The bandwidth parameter for a
'plain' chirnney,

suggests a value close to 0.3 for the open gassland exposure. However, the specua of
the local lift coefficient (Lift force per unit len,ath) suggest that the shedding becomes
more narrowly banded about the central frequency. It wouid be too conservative though
to use the values from the computed spectra due to the rnismatch of the turbulence
intensity between the mode1 and prototype. The bandwidth, B used in the predictions was
set at 0.16.

Also, to account for the increased spanwise correlation, the correlation

length, k' was increased to 2.0 diarneters as opposed to 1.2 diameters, which is
appropriate for chimneys of circular cross section.
The detailed analysis of the structural properties will be discussed in detail in 55.4.
However, it should be noted here that the naturd frequency of the shell mode was
approximately 0.4 Hz and the generalized mass was 750 000 kg.

These properties

correspond to the undarnaged state of the chimneys.


The predictions for the peak across wind tip deflection for the shell for an angle of attack
of 45" are shown in Figure 5.26. The peak deflection was computed as

where Liis the rnean deflection and Lis the RMS deflection. Two values of systern
darnping, 0.8% and 1.0% of critical were used as bounds and were thought to be
appropriate up to the point where an impact between the liner and shell would occur. The
maximum computed tip deflection for 1.0% darnping was approximately 0.6m at a mean
tip wind speed of 20 m/s. For 0.8% damping, the computed deflection was

Mean Wind Speed at Tip (rn/s)


Figure 5.26: Peak across-wind response of prototype for ~-155'

1O

15

20

25

30

35

Mean Wind Speed at Tip (m/s)


Figure 5.27: Peak across-wind response of prototype for u=67.S0

40

129
approxmately 0.75m at the sarne wind speed However, the increasingly negative
aerodynamic darnping exceeds the structural damping for a mean wind speed of 30 d s at
the tip of the chirnney. For higher wind speeds, the tip deflection would have to be
determined by a force balance between the self-limiting motion-induced force and the
energy dissipation forces in the chimney.
The predictions for angles of attack of 67.5" and 90" are shown in Figures 5.27 and 5.28,
respectively. The peak tip deflections for 67.5" are comparable to the values discussed
above although the critical wind speed is slightly reduced. For an angle of aaack of 90,
the deflection amplitudes are somewhat larger with maximum tip deflections of 0.7m and

0.85mfor system darnping of 1.0% and 0.8% of critical, respectively.

10

15

20

25

30

Mean Wind Speed at Tip (m/s)


Figure 5.28: Peak across-wind response of prototype for cc=90

35

40

130

The worst-case predictions in terms of magnitude (&O0)

d s o happened to ccincide with

the prevailing wind direction, which was close to northwest, Based on these predictions,
the mean wind speed required to cause the observed clashing would be approximately

23rn/s (33 mph) from tfie northwest- This level of wind speed is not uncornmon in the
exposed high plains region of North Dakota.
Across-wind response predictions were aIso computed for two other cases: first, the
enclosure remaining in place over the lower half of the chimneys; and second, the
enclosure completely removed. For the first case, the wind tunnel results applicable to
the lower half of the chimney were applied for an angIe of attack of 90, which, as

mentioned above, corresponded to the prevaihg wind direction. Over the upper haif of
the chirnney, parameters appropriate for a bare isolated chimney were used.
The results for these two cases are shown in Figure 5.29 for the bare chimney and in
Figure 5.30 with the enclosure over the bottom half. For the bare chimney, the peak
across wind tip displacement for the shell wouid be -&lm at a darnping of 0.8% of
critical and a mean wind speed at the tip of 23 m/s. For the case of the enclosure over the
bottom half of the chimney, the peak response is increased to 0-14 m at the sarne damping
and rnean wind speed. An additional effect of the enclosure is that the peak deflection
continues to increase with wind speed up to a maximum of -0.2m at 37rn/s.
The writer would again like ta ernphasize that the presented predictions are intended only
to confirrn that the addition of the enclosure over the full height of the chirnney caused
the observed change in behaviour and resulted in the failure of the liner in the Unit#2
chimney. Difficulties are always present in using wind tunneI data for the prediction of

Mean Wind Speed at Tip (m/s)


Figure 5.29: Peak across-wind response for the prototype wi:!~out enclosure.

10

15

20

25

30

35

40

Mean Wind Speed at Tip (m/s)


Figure 5.30: Peak ricross-wind response for the prototype with enclosure over the bottom half

prototype responses because of the obvious scaling diff~culties. In this case, the
enclosure appears to govern the separation point an one side of the cylinder and the wind
tunnel results appear to be insensitive to Re over the srna11 range exarninzd. However,
the separation of the body boundary layer on the other side of the cylinder, while most
likely closer to the prototype separation point than subcritical Re expenments would
predict, should be viewed as a source of error in the mode1 results.
This always-present difficulty of Reynolds number mismatch justified a further
monitoring of the prototype to confirm that the response of the chimney with the
enclosure over the bottom half of the chimneys was, in fact, safe.

5.4 Field Study Results


The two chimneys were monitored for approxirnately six months following the
completion of the wind tunnel study. For the duration of the study, the liner of the
chimney for Unit #2 stood at a height of 150m. Both chimneys were instnimented with
accelerometers on the shell and liner at a level of lOOm in the North-South and East-West
axes and four accelerometers at the base to measure rocking of the foundation. After
mode shapes and frequencies were determined, two accelerometers were left in place on
each chimney for the duration of the study.
5.4.1 Modal Properties

The displacements of the shell and liner and the rocking of the foundation were
determined using the response spectra. Unlike the chimneys discussed in Chapter 4, the

133

coupling of the Iiner and shell through the rocking of the foundation was not strongHowever, an effort to determine the mode shapes was made using the soil-structure
interaction model outlined in Chapter 3. The variables used to determined the mode
shapes were the reinforced concrete eIastic modulus, Ec7 the brick modulus, Eh and the
rotational stifiess of the foundation, Ke. By trial and error, the final values were; Ec =
28.9 GPa, Eb = 11-4 GPa and Ke = 4.6 x 1012 N-dradian. The mode shapes for the
Unit#l chimney are plotted in Figures 5-31 and 5.32 for the liner and shell mode,
respectivery.

Those for Unit##2 are plotted in Figures 5.33 and 5.34. The naturai

frequencies and generalized masses obtained from the andysis were; fii,,, = -19 Hz,
M i k + 1 6 6 0 x 10' kg, fiteil = -41 H z , MA,[[ = 745 x 103 kg for the UnMI chimney and;

fii,,, = -31 Hz, M r , ~14-5 x loG kg, fi,,[[= -41 Hz, Md,11 = 748 x 103 kg for the UnMZ
chimney with each mode nonnalized to a unit tip deflection for the shell.

Figure 5.3 1:Mode shape for Liner mode for Unit#l chimney

@liner

@ski1

Figure 5.32: Mode shape for shell mode for Unit #l chimney.

Figure 5.33: Mode shape for liner mode for Unit ##2chimney.

Figure 5.34: Mode shape for shell mode for Unit #2chimney

5.4.2 Field Observafions

Over the penod of the monitoring progran, several events were recorded where the mean
wind speed at the tip of the chimney was 10 - 15 m/s and the wind direction was out of
the northwest. A typical spectmm of the shell displacement is shown in Figure 5.35. The
peak tip amplitude of the shell for the event shown was 2.5 c m (1 in.) Two events where
the wind direction was such that the two chimneys were inline were also made. For one
of the events, the wind direction was due West (Unit #2 in the wake of Unit #l) and the
other was due east. The peak amplitudes observed in these cases were 'arnplified' by a
factor of two or more. Also, the Unit #2 chimney exhibited a 'softening' with the
increase in amplitude. This is demonstrated by the spectmm shown in Figure 5.36. The

0.4

0.3

Frequenrcy (Hz)

Figure 5.35: Spectnirn of sheIl displacement for- peak tip amplitude of 2.5 cm (i in)

O -4

Frequenocy (Hz)

Figure 5.36: Spectrum of shell displacement for - peak tip amplitude of 13 cm (5 in).

frequency shifis downward to 0.39 Hz.

This change in frequency corresponds to a

reduction of stifiess of 10% and may indicate that some cracking occurred when the
UnitK chimney experienced the large amplitude oscillations. The observed amplified
oscillations may explain why only the Unit#2 chimney suffered a failure of the liner.
The field observations are sumrnarized in Figure 5.37. Also shown in Figure 5.37 are the
predictions for the case of the enclosure over the lower half of the chimney at an angle of
attack of 90". Unfortunately, there were no events where the wind speed approached that
required for the response to cause a maximum across-wind response. However, the data
gathered did suggest that leaving the enclosure in place on the bottom half of the chimney
was acceptable. In this configuration, the dong wind response would govern the stresses
in the chimney.

---

5 = 1*0%

--------

Mine Observations

Field Observations (Clear Air)

10

15

20

25

30

Mean Wind Speed at Tip (m/s)

Figure 5.37:Surnmary of observations from the field study.

35

40

5.5 Further Aerodynamic Investigations


The rnechanisms responsible for the observed behaviour in the field study were not
surprising. However, what was alarming was the increase in the maagitude of the acrosswind forcing and magnitude of the negative aerodynamic darnping- For this reason, the
effect of the size of appurtenance on the aerodynamic loads was investigated further. The
motivation for this further work is the prevalence of piping, ladders etc. on steel stacks.
The mean forces and vortex shedding were examined for a number of different sizes
rangng from an outstand dimension of 10%of the diameter down to 1%of the diameter.
The surface-mounted appurtenances in these cases, however, were sharp edged.
The mean drag and lift coefficients for outstand dimensions of 8%, 5% and 1% of the
diarneter are shown in Figures 5.38-5-40, respectively. The effect of the appurtenance
continues to be significant even for the srnaIlest size. For the resultant mean force
coefficient per unit length, the maximum magnitude occurs at 90" and is 1.72, 1.61 and
1.06 for the largest to srnaIlest appurtenance, respectively. Also, the sum of the lift slope
and the mean drag coefficient is strongly negative for d l three cases at
-1.6,-2.6

and - 2.6. The angle where the sudden change in forcing occurs does change

with the size of appurtenance. For the largest outstarid dimension, the switch occurs
around 45" to 50, which is simiIar to the larger enclosures exarnined. However, the
angle of attack where the switch occurs in the force coefficients shifts to 65" and 80" for
the 5% and 1%outstand dimensions, respectively.

Angle of Attack, a (degrees)

Figure 5.38: Mean force coefficients for outstand dimension of O.08D

45

90

135

Angle of Attack, a (degrees)

Figure 5.39: Mean force coeff~cienfor outstand dimension of 0.05D

Angle of Attack, a (degrees)

Figure 5.40: Mean force coeff~cientsfor outstand dimension of 0.01D

45

90

135

Angle of Attack, a (degrees)

Figure 5.41: RMS iift coefficient for outstand dimension of 0.08D

45

90

135

Angle of Attack, a (degrees)

Figure 5.42: RMS Iift coefficient for outstand dimension of 0.05D

45

90

135

Angle of Attack, cr (degrees)


Figure 5.43: RMS lift coefficient for outstand dimension of 0.01D

The RMS Lift coefficient for outstand dimensions of 8%, 5% and 1%of the diameter are
shown in Figures 5.41-5.43,respectively. Again, the increase is significant for dl three
cases reaching maadtudes close to 0.5 for the 8% outstand case and 0.3 for the 1%
outstand case. The shift in angle of attack where the switch occurs matches the results
discussed above.
The Strouhal number for outstand dimensions of

8%, 5%

and 1% of the diameter are

shown in Figures 5.44-5.46, respectively. The only sipificant change in behaviour here
over the previous set of experiments is that for the 1% outstand case, the appurtenance
does not interrupt vortex shedding when it is fully in the wake.

45

90

135

Angle of Attack, a (degrees)

Figure 5.43: RMS Iift coefficient for outstand dimension of O.O8D

Angle of Attack, u (degrees)

Figure 5.45: IU'iS lift coefficient for outstand dimension of 0.05D

45

90

135

Angle of Attack, u (degrees)

Figure 5.45: R M S lift coefficient for outstand dimension of 0.0 ID

5.6 Summary and Conclusions


The results of a rnodel study of the effect of appurtenances on the across-wind response
of chirnneys are presented- This work followed reported observations of large acrosswind oscillations of a large concrete chirnney where a manlift enclosure was installed on
the exterior over the height of the chimney. The model study was conducted in two
stages; first, the enclosure on the prototype chimneys was modelled and second, the
effect of the size of an appurtenance on the ma,onitude of the across-wind forces was
examined.
Predictions of the prototype response using the wind tunnel data confirmed that the
addition of the manlift enclosure adversely affected the aerodynamic forces and the
motion-induced forces enough to cause tip amplitudes consistent with the reported field
observations. Subsequent monitoring of the chimneys suggested that the removal of the
enclosure from the top half of the chirnney reduced the across-wind to safe leveIs.
The model study involving much smaller surface mounted appurtenances over the length
of a circular cylinder indicated that any size adversely affects both the aerodynamic
forces and the quasi-steady aerodynamic damping fmces.

Chapter Six
Summary and Conclusions
This study focused on special aspects of the across-wind loads and effects on large
reinforced concrete chirnneys. The design of isolated concrete chimneys of circular
cross-section is based on weil-established codes of practice (AC1 307-98, CICIND) that
rely primarily on the work of Vickery and CO-workers. Sometimes however, this
information is implemented in cases where the current level of understanding is
incomplete. The motivation for the present work stemrned from two such instances that
were the topic of field and mode1 studies.

6.1 Summary of Field Observations


In the first instance, two large reinforced concrete chimneys were constructed near an
existing reinforced concrete chimney that was roughly 25% larger in diameter. For wind
directions where either of the newer chirnneys was directly downstream of the older
chimney, the separation was 5.1 diameters, based on the upstream chimney. The results
of the field study suggested that when either of the smaller chimneys was directly
downstream of the larger one, the across-wind loads were amplified by a factor of
approximately 3.7. This was due to increased lateral buffeting forces resulting from the
wake of the upstrearn chimney.

In the second instance, two large reinforced concrete chimneys that had been in service
some 20 years were fitted with elevator enclosures on the surface of the shell. After the

146
installation, the across-wind oscillations increased by a factor of the order of 10. The
freestanding liner on one of the chimneys suffered major damage and was partially
dismantled-

6.2 Effect of Foundation Flexibility


Both of the cases involved chimneys of the windshield type where the concrete shelI
encloses a freestanding brick liner. The response of this type of structure was exarnined
in detail and it was found that the flexibility of the foundation plays a major role. The

motion of the foundation acts to 'couple' the motions of the liner and shell. It was found
that this coupling reduces the deflections and stresses in the shell, and increases the
relative motions between the liner and shell. It was also found that significant windinduced stresses occur in the liner.
The behaviour is strongly related to the stiffness of the foundation. The chimneys
examined in Chapter 5 did not exhibit strong coupling between the liner and shell and the
motions of the liner were not significant.

6.2 Summary of Model Results


6.2.1 Pairs of Circular Cylinders

The results of a rnodel study of the increased across wind loading due to the presence of
an upstream cylinder were presented in Chapter 4. For the first case, the diameter of the
upstream cylinder was 25% larger than the downstream cylinder. The spectrum of the lift
coefficient for al1 the spacing ratios investigated showed only one peak. Thus, the central

147

frequency of the lateral turbulence due to the shedding from the upstream cylinder
matched the vortex shedding frequency of the downstream cylinder- This was confrrrned
by the spectra of the differentid pressures from the upstream cylinder. The increase in
the mamgnitudeof the RMS lift coefficient over the value for a single cylinder decayed as
the separation between the cylinders increased, but remained ~i~pficant.
For the second case the diameters of the cyiinders were identical. As for the previous
case, the spectra of the lift coefficient showed only one peak.

The increase in the

fluctuating lift force was not as large for this case but was still significant.

In the third case, the upstream diameter was 20% smaller than the downstream cylinder.
The spectra of the Iift coeff~cientfor this set of experiments showed two peaks, the lower
frequency peak due to shedding from the downstrearn cylinder and a higher peak due to
laterd turbulence from the upstream cylinder. The RMS lift coefficient due to buffeting
was quantified and it was dso found to decay with increasing separation between the
cylinders.
For the cases where only one peak was present in the lift coefficient spectrum, the
implied amplification of the vortex shedding loads was evaluated. When the upstream
diarneter is larger than the downstream diarneter by 3596, the amplifications were very
large and approached eight times the single cylinder across-wind loads. As the separation
between the cylinders increased, this amplification reduced to a factor of five times the
single cylinder loads.

Similar results for the identical cylinder case were observed.

Cornparisons with the field study results were encouraging gven the uncertainties in each
study.

6.2.2 Effect of Appurtenances

The results of a wind tunnel study on the impact of surface mounted appurtenances on the
wind loads on chimneys were presented in Chapter 5. The effect of the angle of attack on
the wind Ioads was studied for a range of appurtenance sizes.

For the Iarger

appurtenances, the flow around the cylinder was affected such that a large negative mean
Iift force and increased drag force existed for angles of attack g e a t e r than 45' and up to

150". The change in the mean force coefficients in this range resulted in a strongly
negative lift slope suggesting the cross-section was prone to galloping instability.
Around the same angle of attack, the RMS magnitude of the fluctuating lift coefficient
increased dramatically accompanied by a downward shift in the Strouhal number. Al1 of
these changes in the aerodynamic forces were adverse compared to a plain circular
cylinder.

The same changes in the aerodynamics were observed for much smaller

surface-mounted appurtenances.

While the magnitudes were less severe, the cross-

section was still susceptible to galIoping instability.


Predictions of the across-wind response of the prototype chimneys using the wind tunnel
data were presented. These results confimed that the oscillation amplitudes consistent
with the observed ciashing of the shell and liner were caused by the addition of the
enclosure.

6.3 Conclusions
Based on the results of this work, the following general conclusions are drawn:
1. The effect of foundation flexibility on the response of reinforced concrete

chimneys to wind excitation can 5e significant and should be addressed at the


design stage; particularly for 'windshield' type chimneys where shellkner
interaction is involved.
2. The presence of an upstream chimney, particularly if larger in diameter, greatly
amplifies the across-wind forces through the buffeting action of the increased
level of lateral turbulence.

3. The addition of any surface-mounted appurtenance such as the manlift enclosure


in the field study, but as small as piping on steel chirnneys, adversely affects the
dong-wind and across-wind aerodynamic loads on chimneys of circular crosssection.
4. One further conclusion, while not explicitly dedt with in this thesis, is the
applicability of the across-wind response mode1 of Vickery and CO-workerswhen
the appropriate parameters are available.

6.4 Recommendations for Further Study


Throughout this thesis, an effort was made to emphasize the value of full-scale data.
While the agreement with wind tunnel and field results in this thesis was encouraging, the

150

scaling difficulties should not be overlooked. It is the writer's belief that full-scale data is
required to better assess the increased across-wind loading for groups of chimneys.
Design standards are incomplete in this area.

References

AC1 Comrnittee 307. 1998. Design and Construction of Reinforced Concrete Chirnneys,
Detroit, Michigan.
Basu, KI. and Vickery, B.J. 1983. Across-wind vibrations of structures of circular crosssection - Part II: Development of a mathematical mode for full-scale application,
Journal of Wind Engineering and Indzistnal Aerodynamics, 12(1), 75-98.
Basu, R.I. and Vickery, B.J. 1982. A comparison of mode1 and full-scde behaviour in
wind of towers and chimneys. International Workshop on Wind Tzrnnel Modeling
Critena and Techniques, Gaithersburg, Maryland.
Basu, R.I. 1982. Across-wind response of slender structures of circular cross-section to
atmAosphericturbulence: Volumes 1and II. Ph-D. Thesis, University of Western
Ontario, 1982.
Beredugo, Y. 0. and Novak, M. 1972. Coupled horizontal and rocking vibration of
embedded footings. Canadian Geotechnical Journal, 9(4), 477-497.
Bergh, H. and Tijdeman, H. 1965. Theoretical and experimental results for the dynamic
response of pressure measuring systems, National Aero and Aeronauticd
Research Institute, Rep. NLR-TR F.238.
Blackburn, H-M. and Melbourne, W.H. 1993. Cross flow response of slender circularcylindrical structures: prediction models and recent experimental results. Journal
of Wind Engineering and Industrial Aerodynamics, 49, 167-176.
Bycroft, G. N. 1956. Forced vibrations of a ngid circular plate on a semi-infinite elastic
space and on an elastic stratum. Philosophical Transactions of the Royal S o c i e ~
London, Ser. A., 248,327-368.
Cheung, J.C.K. and Melbourne, W.H. 1983. Turbulence effects on some aerodynarnic
parameters of a circular cylinder at supercritical Reynolds numbers. Jozcrnal of
Wind Engineering and Industrial Aerodynamics, 14, 399-4 10.
CICIND, 1984. Mode1 Code for Concrete Chimneys. Part A: The Shell
Cincotta, T.T., Jones, G.W., and Walker, W.W. 1966. Experimental investigation of wind
induced oscillation effects on cylinders in two dimensional flow at high Reynolds
numbers. National Aeronautics and Space Administration, NASA TMX 57779,
paper 30.1-20.35.
Clough, R-W. and Penzien, J. 1993. Dynamics of Stnrctures. McGraw-Hill, Inc., New
York.

Davenport, AG. 1961. The application of statistical concepts to the wind loading of
structures. Proceeding of the Institzite of Civil Engineers. 19.
Davenport, A.G. 1962. The response of slender, Line-like structures to a gusty wind.
Proceeding of the Instimre of Civil Engineers. 23, pp. 389-408.
Davenport, A.G. 1963a. The relationship of wind structure to wind loading. Proceedings
of the International Conference on Wind Eflects. National Physical Laboratory,
London.
Davenport, A.G. 1963b. The buffeting of structures by gusts. Proceedings of the
Irzternational Conference on Wind Effects. National Physical Laboratory, LondonDavenport, A.G. 1964. Note on the distribution of the largest value of a random function
with appiication to gust loading. Proceeding of the Institute of Civil Engineers.
28, pp. 187-196.
Dobry, R. and Gazetas, G. 1988. Simple method for dynamic stiffness and darnping of
floating pile groups, Geotechnique, 38(4), 557-574.
El Naggar, M. H. and Novak, M. 1995. Nonlinear lateral interaction in pile dynamics.
Jounzal of Soi1 Dynamics and Earthqrrake Engineering, 14(2), 141-157.
Elsabee, F. and Morray, J. P- 1977. Dynamic behavior of embedded foundations.
Research Report R77-33, Civil Engineering Department, Massachusetts Institute
of Tecbnology, Cambridge, Massachusetts,
Fage, A. and Warsap, J.H. 1929. The effects of turbulence and surface roughness on the
drag of a circular cylinder. Aeronarttical Research Coztncil, Rep and Memo 1283.
Feng, C.C. 1968. The measurement of vortex induced effects in flow past stationary and
oscillating circular and D-section cylinders. M.A.Sc. Thesis, University of British
Columbia.
Fujita, K., Ikegami, Y., Kobayashi, K., and Ohashi, M. 1988. Experimental studies on
fluctuating lift force on a single cylinder at high Reynolds numbers. Japan
Journal of Wind Engineering., 37,73-82.
Galsworthy, J.K. and El Naggar, M.H. 1997. Analysis of R/C chimneys with soilstructure interaction. ASCE Geotechnical Special Publication No. 70, 23-35.
Galsworthy, J.K. and Vickery, B.J. 1999. Full scde and mode1 studies of wind loads on a
pair of circular cylinders. Wind Engineering into the 21st Century, Edited by A.
Larsen, G.L. Larose & F.M. Livesey, 3, 1641-1648.
Galsworthy, J.K. 2000. Experimental investigation of wind loads on pairs of circular
cylinders at high Reynolds numbers: Supplement to PhD Thesis. BLWTL-SS62000.

Galsworthy, J.K. 3000. Experimentd investigation of wind loads on circular cylinders


with surface mounted appurtenances: Supplement to PhD Thesis. BLWTL-SS72000.
Gazetas, G. and Makris, M. 1991. Dynamic pile-soil-pile interaction, part 1: Analysis of
axial vibration, Earthquake Engineering and Stmctztral Dynamics, 20, 115-133Holmes, J.D. and Lewis, R.E. 1986. Optimization of dynamic-pressure-measurements
systems. 1. Single point measurements. Journal of Wind Engineering and
Ind~tst?falAerodynamics, 25,249-273.
Kausel, E. and Ushijima, R. 1979. Vertical and torsional stiffness of cylindrical footing.
CiviI Engineering Department Report R79-6, MIT, Cambridge, Massachusetts.
Kaynia, A. M. and Kausel, E. 1982. Dynamic behavior of pile groups. zndInternational
Conference on Numerical Methods in Offshore Piling, Austin, Texas, 509-532.
Kobori, T., Minai, . R. and Baba, K. 1977. Dynamic behaviour of a laterally loaded pile.
9"' International Conference of Soil Mechanics, Tokyo, Session 10 (6), 175-180.
Kobori, T., Minai, R. and Suzuki, T. 1971. The dynamical ground cornpliance of a
rectangular foundation on a viscoelastic straturn- Bullerin Disaster Prevention
Research lizstirute, Kyoto University, 20,289-329.
Kuhlemeyer, R. L-1979. Static and dynamic laterally loaded floating piles. hzrmal of the
Geotechnical Engineering Division, ASCE, 105(GT2), 289-304.
Kwok, K.C.S. 1986. Turbulence effect on flow around circular cylinder. Journal of
Engineehg Mechanics, 112(1l), 1181-1197.
Luco, J. E. and Hadjian, A. H. 1974. Two-dimensiona1 approximations to the threedimensional soil-structure interaction problem. Nuclear Engineering and Design,
31(2), 195-203.
Luco, J. E. and Westmann, R. A. 1971. Dynarnic response of circular footings, Jorrrnal of
the Engineering Mechanics Division, ASCE, 97 (EM6), 138 1-13%.
Matlock, H., Foo, H. C. and Bryant, L. M. 1978. SimuIation of lateral pile behaviour
under earthquake motion. Proceedings of the ASCE Specality Conference on
Ea~hquakeEngineering and Soil Dynamics, Pasadena, California, 2, 1065-1084.
Melbourne, W.H. 1997. B luff-body aerodynamics. In Wind Engineering Course Notes,
Melbozmze, W. H. ed., Department of Mechanical Engineering, Monash,
University, Australia,
Modi, V.J. and El-Sherbiny, S. 197 1. Effect of wall confinement on aerodynamics of
stationary circular cylinders. Proc. of the 3rdInt. Con$ on Wind Efects on
Buildings and Stntciures, Tobo: 365-375.

National Building Code of Canada 1995. Issued by the Canadian Commission on


Building and Fire Codes.
Nogami, T. and Novak, M. 1976. Soil-pile interaction in vertical vibration. International
Journal of Earthquake Engineering and Stncctzcral Dynamics, 4 (3), 277-293.
Novak, M. 1974a. Dynamic stiffhess and damping of piles. Canadian Geotechnical
Journal, 11,574-598-

Novak, M. L974b. Effect of soi1 on structural response to wind and earthquake. Jorrrnal
of Earthqzcake Engineering and Stncctural Dynarnics, 4, 79-96.
Novak, M. and Aboul-Ella, F. 1978. Impedance functions of piles in layered media.
Journal of the Engineering Mechanics Division, ASCE, 104(EM3), 643-66 1.
Novak, M. and El Hifnawy, L. 1983. Effect of soil-structure interaction on damping of
structures. Journal of Earthquake Engineering and Stmctzeral Dynarnics, 11,59562 1.
Novak, M. and Nogami, T. 1977. Soil-piIe interaction in horizontal vibration.
International Jorcrnal of Earthquake Engineering and Stnrchire Dynamics, 5 (3),
263-282.
Novak, M., Nogami, T. and Aboul-Ella, F. 1978. Dynamic soi1 reactions for plane stain
case. Jorernal of the Engineering Mechanics Division, ASCE, 104 (EM4), 953959.
Roshko, A. 196L. Experirnents on the flow past a circular cylinder at very high Reynolds
numbers. Journal of Fluid Mechanics, 10,345-356.
Roshko, A. 1954. On the drag and shedding frequency of two-dimensional bluff bodies.
National Advisory Committeefor Aerortautics, NACA TN 3 169.
Roshko, A. 1955. On the Wake and Drag of Bluff Bodies, Journal of Aeronacctical
Sciences, 22, 134- 132.
Rumman, W S , 1970. Basic structural design of concrete chimneys, Jozcrnal of the Power
Division, ASCE, 96,309-3 18.
Ruscheweyh, H. 1984. Problems with inline stacks: experience with full-scale objectsEngineering Stntctures, 6 , 340-343.
Sanada, S. and Nakamura, 0. 1983, Full-scde measurements of wind forces on a 200m
concrete chirnney. Kajima Institute of Construction Technology, Chofu City,
Tokyo, Japan, 1983.

Schewe, G. 1983. On the forces acting on a circular cylinder in cross flow from
subcritical up to transcritical Reynolds numbea. Journal Fluid Mechanics, 133,
265-285.
Schlichting, H. 1979 Boundary Layer Theory, McGraw-Hill.
Scruton, C. and Flint, A.R. 1964. Wind-excited oscillations of structures, Proceedings of
the Instihrtion of Civil Engineers, 27,673-702.
Shih, W.C.L., Wang C., Coles, D. and Roshko, A. 1993. Experiments on flow past rough
circular cylinders at large Reynolds Numbers, Jarrmal of Wind Engineering and
Indzrstnal Aerodynamics, 49,35 1-368.
Strouhal, V. 1878. ber eine besondere art der tonerregung, Annalen der Physik und
Chernie (Leipzeig), V ,217-251.
Veletsos, A. S. and Verbic, B. 1973. Vibration of viscoelastic foundations, Earthquake
Engineering and Structural Dynarnics, 2. 87-102.
Veietsos, A. S. and Wei, Y. T. 1971. Lateral and rocking vibrations of footings, Journal
of the Soil Mechanics and Foundahons Division, ASCE, 97 ( S m ) , 1227-1248.
Vickery, B.J. 1995. The response of chirnneys and tower-like structures to wind loading.
State of the A n Volume in Wind Engineering, International Association of Wind
Engineering, 205-233.
Vickery, B.J. and Basu, R.I., 1984. The response of reinforced concrete chimneys to
vortex shedding. Journal of Engineehg Stnrctures, 6, 324-333.
Vickery, B.J. and Basu R.I. 1983. Across-wind vibrations of structures of circular crosssection - Part 1: Development of a mathematical model for two dimensional
conditions. Journal of Wind Engineenhg and Indzrsrrial Aerodynamics, 12(1),
49-74.
Vickery, B.J. 1981, Across-wind buffeting in a group of four inline model chimneys,
Journal of Wind Engineering and Indzrstrial Aerodynamics, 8, 177-193.
Vickery, B.J. and Clark, A.W. 1972. Lift or across-wind response of tapered stacks.
ASCE Journal of the Structural Division, ASCE, 98, 1-20.
Vickery, B.J. and Kao, H. 1972. Drag or alongwind response of tapered stacks. ASCE
Journal of the Stntctrrral Division, ASCE, 98,21-40.
Vickery, B.J. 1968 Wind Effects on Stntctzwes, PhD Thesis, University of Sydney,
Sydney, Australia.
Vickery, B.J. 1969 On the reliability of gust loading factors, Civil Engineering
Transaction, Institution of Engineers Azrstralia, 13, 1-9

Vickery, B .J. and Watkins, R.D- 1962. Flow-induced vibrations of cylindricd structures.
Proceedings of the Ist Australasian Conference on Hydraulics and R u i d
Mechanics- 2 13-24 1,
von Kirmiin, Th. 1911. ber den mechanismus des widerstandes, den ein bewegter
korper in einer flssigkeit erzeugt. Nachr. Ges. Wiss. Gottingen, Math. Phys.
Klasse 509-5 17,
Waldeck, J.L. 1992. The measured and predicted response of a 300rn concrete chimney.
Journal of Wind Engineering and Indccstrzhl Aerodynamics, 41,229-240.
Wooton, L.R.1969. The oscillations of large circular stacks in wind. Proceedings of the
Institution of Civil Engineers, 43, 573-598.
Zdravkovich, M.M. 1997 Flow arozrnd circzrlur cylinders Volzrrne 1: Fundamentais.
Oxford University Press.
Zdravkovich, M.M. 1988. Review of interference-induced oscillations in flow past two
paralle1 circular cylinders in various arrangements. Jotrrnal of Wind Engineering
and Indzlstrial Aerodynamics, 28, 183-200.
Zdravkovich, M.M- 1977. Review of flow interference between tow circular cyiinders in
various arrangements. Jozrrnul of Flzrids Engineering. Trans. of ASME. 6 18-633.
Zdravkovich, M.M, and Pridden, D.L. 1977. hterference between two circular cylinders;
series of unexpected discontinuities. Jozrrnal of Indztsrrial Aerodynamics, 2,255270.

Você também pode gostar