Comprehensive Study of Wind Loads on Parapets

Rania Bedair

A Thesis

in

The Department

of

Building, Civil & Environmental Engineering

Presented in Partial Fulfillment of the Requirements

For the Degree of Doctor of Philosophy (Building Engineering) at

Concordia University

Montreal, Quebec, Canada

April 2009

© Rania Bedair, 2009

1*1 Library and Archives
Canada

Published Heritage
Bibliothgque et
Archives Canada

Direction du
Branch Patrimoine de l'6dition

395 Wellington Street 395, rue Wellington

Ottawa ON K1A 0N4 Ottawa ON K1A 0N4
Canada Canada
Your file Votre reference
ISBN: 978-0-494-63348-9
Our file Notre r6f6rence
ISBN: 978-0-494-63348-9

NOTICE: AVIS:

The author has granted a non- L'auteur a accorde une licence non exclusive
exclusive license allowing Library and permettant a la Bibliotheque et Archives
Archives Canada to reproduce, Canada de reproduce, publier, archiver,
publish, archive, preserve, conserve, sauvegarder, conserver, transmettre au public
communicate to the public by partelecommunicationou par I'lnternet, preter,
telecommunication or on the Internet, distribuer et vendre des theses partout dans le
loan, distribute and sell theses monde, a des fins commerciales ou autres, sur
worldwide, for commercial or non- support microforme, papier, electronique et/ou
commercial purposes, in microform, autres formats.
paper, electronic and/or any other
formats.

The author retains copyright L'auteur conserve la propriete du droit d'auteur

ownership and moral rights in this et des droits moraux qui protege cette these. Ni
thesis. Neither the thesis nor la these ni des extraits substantiels de celle-ci
substantial extracts from it may be ne doivent etre imprimes ou autrement
printed or otherwise reproduced reproduits sans son autorisation.
without the author's permission.

In compliance with the Canadian Conformement a la loi canadienne sur la

Privacy Act some supporting forms protection de la vie privee, quelques
may have been removed from this formulaires secondaires ont ete enleves de
thesis. cette these.

While these forms may be included Bien que ces formulaires aient inclus dans
in the document page count, their la pagination, il n'y aura aucun contenu
removal does not represent any loss manquant.
of content from the thesis.

Canada
 COMPREHENSIVE STUDY OF WIND LOADS ON PARAPETS

Rania Bedair, Ph. D.

Concordia University, 2009

The current thesis aims at defining and evaluating the local (components and cladding)

wind loads on parapets. For the first time, it was attempted to measure such loads in full-

scale, in order to address the issues encountered in previous wind tunnel studies. Field

testing was carried out using the full-scale experimental building (3.97 m long, 3.22 m

wide and 3.1 m high) of Concordia University (located near the soccer field at the Loyola

Campus). In order to define individual surface pressures as well as their combined effect

from both parapet surfaces, simultaneous peak and mean wind-induced pressures were

measured on both exterior and interior surfaces of a uniform perimeter parapet with a

height of 0.5 m. Roof edge and corner pressures were also recorded. In addition, a

complete wind flow simulation was performed in the Boundary Layer Wind Tunnel

(BLWT) of Concordia University using a 1/50 scale model of the experimental building

with two different parapet heights, equivalent to 0.5 and 1 m. The choice of geometric

scale based on correctly modeling the turbulence intensity at the roof height. The wind

tunnel results were compared with the field data for validation purposes. In general, the

comparison shows good agreement, although some discrepancies were identified for

critical wind directions.

In the past, it was difficult to directly model and record the parapet surface pressures, due

to modeling limitations. Therefore, wind loads on parapets were mainly estimated from

pressures measured on the wall and the roof of the building in the vicinity of a parapet.

The current results demonstrate, in general, that the design method provided in the ASCE

7-05 overestimates the total load on the parapet. In addition, design recommendations

are provided and can be considered by the standards.

Numerical simulation of the wind flow over the test building model with the parapet was

also performed by using the CFD code Fluent 6.1.22. The steady-state RANS equations

were solved with two modified k-s turbulence models, namely the RNG model and

the RLZ k-s model. Considering the current state-of-the-art, peak pressures are not

predicted reliably by computational approaches. Therefore, in the present study only

mean wind-induced pressures on the roof and on parapet surfaces were computed. The

computational results show that parapets act to reduce high negative pressures on the

leading edge and to make the distribution of mean pressures on the roof more uniform.

The simulated pressures are generally in good agreement with the corresponding wind

tunnel data.

IV
 ACKNOWLEDGMENTS

I wish to express my sincere gratitude to Dr. Ted Stathopoulos for his supervision,

guidance, encouraging and support during the course of this work.

Acknowledgments are due to all the members of thesis committee for their constructive

suggestions and to Dr. Patrick Saathoff for his assistance during the early stage of the

experimental work

I would like to express my sincere gratitude to my parents, sisters, brother and my lovely

sons. Special worm appreciation is given to my husband for his assistance,

understanding, encouraging, support and love.

v
 TABLE OF CONTENTS

LIST OF TABLES xi

LIST OF FIGURES xii

NOMENCLATURE xvi

ABBREVIATIONS xix

CHAPTER 1 INTRODUCTION

1.1 Testing of Low-rise Building 1

1.2 Flat Roofs with Parapets 3

1.3 Significance of the Study 3

1.4 Research Objectives 4

1.5 Work Methodology 5

1.6 Flow Patterns on a Low Building 6

1.7 Outline of the Thesis 7

CHAPTER 2 LITERATURE REVIEW

2.1 Full-scale Investigations 10

2.1.1 Building roofs with parapets 15

2.2 Wind Tunnel Investigations 16

2.2.1 Wind loads on roofs with parapets 20

2.2.2 Wind loads on parapets themselves 26

vi
 2.3 Numerical Computations 27

2.4 Current Wind Loading Standards Regarding Parapets 32

2.5 Summary 33

CHAPTER 3 NUMERICAL SIMULATION

3.1 Introduction to FLUENT 42

3.2 Governing Equations 43

3.3 Turbulence Models 44

3.3.1 Standard k-smodel 45

3.3.2 Realizable ^-fmodel 45

3.3.3 RNG k-emodel 46

3.4 Numerical Simulation Procedure 47

3.4.1 Computational domain 48

3.4.2 Grid generation 49

3.4.3 Boundary conditions 50

3.5 Sensitivity study of the numerical solution 53

CHAPTER 4 EXPERIMENTAL WORK

4.1 Full-scale Testing 62

4.1.1 Experimental building 63

4.1.2 Measurement of wind speed and direction 64

4.1.3 Upstream wind flow parameters 65

4.1.4 Reference static pressure 66

vii
 4.1.5 Pressure measurements 67

4.1.6 Verification of the Collected Data 69

4.2 Wind Tunnel Modeling 70

4.2.1 Modeling of the experimental Building 71

4.2.2 Boundary Layer Simulation 72

4.2.3 Wind-induced pressure 75

4.2.3.1 Pressure Measurements 76

4.2.3.2 Extreme value analysis 77

CHAPTER 5 EXPERIMENTAL RESULTS AND COMPARISONS

5.1 Local Loading Coefficients 95

5.1.1 Correlation of pressures on the roof

and the interior parapet surface 98

5.1.2 Correlation of pressures on the exterior

And the interior parapet surface 100

5.1.3 Total local loading coefficients on parapets 101

5.1.4 Parapet pressures: full-scale versus wind tunnel 102

5.1.5 Comparisons with other studies 103

5.1.6 Comparison with NBCC (2005) 105

5.2 Area-Averaged Loading Coefficients 106

5.2.1 Comparisons with the ASCE 7-05 108

viii
CHAPTER 6 NUMERICAL RESULTS AND DISCUSSION

6.1 Flow Patterns around the Building 126

6.2 Mean pressure distribution on the roof 127

6.3 Effect of parapet height on roof pressures 129

6.4 Mean pressures on the parapet itself 129

CHAPTER 7 CONCLUSIONS, RECOMMENDATIONS AND FUTURE

WORK

7.1 Summary 139

7.2 Concluded Remarks 141

7.3 Recommendation Based on the Current Study 145

7.4 Limitations of the Present Study 146

7.5 Future Work 147

REFERENCES 149

APPENDIX A

A. 1 Velocity and Turbulence Parameters 170

A.2 Anemometer Calibration 175

A.3 DSM System Setting and Operating Principle 176

A. 5 Gumbel Plot 179

ix
 APPENDIX B

B. 1 Wind Speed and Turbulence Intensity Profiles

(ESDU, 1983) 180

B.2 Minimum Design Loads for Buildings and Other

Structures, Sec. 6.5 Analysis Procedures (ASCE 7-05) 181

B.2.1 Main wind force-resisting system 181

B.2.2 Components and cladding 182

x
 LIST OF TABLES

Table 3.1 Specification of computational domain sets used in the

current analysis 55

Table 4.1 Stability classification (Sedefian and Bennet, 1980) 79

Table 4.2 Full-scale testing parameters 80

Table 4.3 Wind tunnel modeling parameters for different simulated

exposures 81

Table 4.4 Wind tunnel modeling parameters used in the present study 82

Table 5.1 Characteristics of different models used in each study 112

Table 5.2 Comparison of parapet load coefficients derived from

NBCC (2005) with the current measured values 113

Table 6.1 Numerical and experimental pressure coefficients 131

Table B.l Internal pressure coefficient, GCpi 183

xi
 LIST OF FIGURES

Figure 1.1 An aerial view of low buildings in a city center 8

Figure 1.2 Vortex formation on roof edges (Cook, 1985) 9

Figure 2.1 Corner mean pressure coefficients in terms of hp/L

(Stathopoulos et al., 1992) 36

Figure 2.2 Parapet wall at the WERFL of TTU(parapet under

construction, 2004) 37

Figure 2.3 Roof corner pressure coefficients for cut-parapet

configurations (Stathopoulos and Baskaran, 1988-b) 38

Figure 2.4 Turbulent air flow field around a cube (Murakami, 1990) 39

Figure 2.5 Mean stream lines expected around the center line of

the current building 40

Figure 2.6 Assumed load acting on the parapet (ASCE-7, 2005) 41

Figure 3.1 Flow chart of numerical computation method 56

Figure 3.2 Building configurations and computational domain (not to scale) 57

Figure 3.3 Grid arrangements for the computational domain:

(a)
(b) Near thewhole
For the building model
domain. 58

Figure 3.4 Inlet and incident vertical profiles of

(a) Turbulence intensity, (b) Mean velocity,
(c) Turbulent kinetic energy, (d) Turbulent dissipation rate 59

Figure 3.5 Computational domain used in sensitivity analysis 60

Figure 3.6 Experimental and computational pressure coefficients

with different computational domains 61

Figure 4.1 Basic exposure of the experimental station (South-West) 83

Figure 4.2 Details of the test building 84

xii
Figure 4.3 Details of parapet section and tap locations 85

Figure 4.4 Flow patterns around a rectangular building (ASHRAE, 1999) 86

Figure 4.5 Mean wind speed and direction measured during 24 hours
(a) Wind speed
(b) Wind direction 87

Figure 4.6 Typical time histories for wind speed and wind direction

measured at the anemometer height 88

Figure 4.7 Typical time histories for parapet surface pressure coefficients 89

Figure 4.8 The wind tunnel in simulated terrain

(a) Wind tunnel inlet

(b) Inside view of the tunnel 90

Figure 4.9 Building models with different parapet heights 91

Figure 4.10 Details of pressure tap layout of the building model 92

Figure 4.11 Profiles of turbulence intensity and mean velocity, and the
spectrum at the roof height:
(a) Longitudinal turbulence intensity profile and
(b) Mean wind velocity profile
(c) Wind spectrum at the roof height 93

Figure 4.12 Extreme value behavior for a segment of the Cp time history
of Tap Bli for perpendicular wind direction:
(a) Time history (data point 1-600)
(b) Gumbel plot 94

Figure 5.1 Surface parapet pressure coefficients recorded near the corner
(Tap Bl)
(a) Exterior surface parapet pressure coefficients
(b) Interior surface parapet pressure coefficients 114

Figure 5.2 Surface parapet pressure coefficients recorded at the mid-span

(Tap B3)
(a) Exterior surface parapet pressure coefficients
(b) Interior surface parapet pressure coefficients 115

Figure 5.3 Comparison of peak pressure coefficients on the roof and the
inside parapet surface measured near corner (Tap Bl)
(a) Full-scale pressure coefficients
(b) Wind tunnel pressure coefficients 116

xiii
Figure 5.4 Comparison of peak pressure coefficients on the roof and
the inside parapet surface measured at mid-span (Tap B3)
(a) Full-scale pressure coefficients
(b) Wind tunnel pressure coefficients 117

Figure 5.5 Effect of wind direction on the correlation coefficients of

Cpr and Cpi 118

Figure 5.6 Comparison between Pressure coefficents measured in

Full-scale and Wind tunnel 119

Figure 5.7 Effect of wind direction on the correlation Coefficients of

CPe and CPi 120

Figure 5.8 Peak local load coefficients on the parapet measured

(a) Pressure coefficients at the corner (Tap Bl)
(b) Pressure coefficients at mid-span (Tap B3) 121

Comparison of peak pressure coefficients from different studies:

(c) Exterior surface parapet pressure coefficients
Figure 5.9
(d) Interior surface parapet pressure coefficients
(c) Simultaneous combined pressure coefficients
from both surfaces 122

Variation of peak loading coefficients (obtained from the

most critical azimuth) with different tributary areas:
Figure 5.10
(a) Exterior parapet loading coefficients
(b) Interior parapet loading coefficients
(c) Net loading from both surfaces 123

Comparison of edge region loading coefficients with

ASCE 7-05
Figure 5.11
(a) Exterior parapet loading coefficients
(b) Interior parapet loading coefficients
(c) Net loading from both surfaces 124

Comparison of interior region loading coefficients with

Figure 5.12 ASCE 7-05 :
(a) Exterior parapet loading coefficients
(b) Interior parapet loading coefficients
(c) Net loading from both surfaces 125

Mean velocity vector field around the building

Figure 6.1 (Vertical cross-section)
(a) Upstream standing vortex
(b) Near wake recirculation

xiv
 (c) Separation region 132

Figure 6.2 Mean velocity vector field around the building (Plan-view)

Figure 6.3 Contours of pressure coefficient on roof surface from

different studies (hp = 0 m):
(a) RNG k- s (present study)
(b) N-R two layer method (Zhou, 1995)
(c) Experimental (Hunt, 1982) 133

Figure 6.4 Contours of pressure coefficients on roof surface (current study):

(a) Wind tunnel, hp = 0.1 m
(b) RNG k-s, hp = 0.1 m 134

Figure 6.5 Pressure coefficient along the center line of building roof:
(a) hp = 0.1 m
(b) hp = 0.2 m 135

Figure 6.6 Effect of parapet height on roof pressure 136

Figure 6.7 Change of parapet surface pressure coefficients with

parapet height
(a) Individual mean pressure coefficients
(b) Combined mean pressure coefficients 137

Figure A. 1 Calibration of the anemometer used in field testing 170

Figure A.2 Mean velocity and turbulence intensity profiles for

different exposures
(a) Terrain-1
(b) Terrain-2
(c) Terrain-3

(d) Terrain-5 175

Figure A.3 DSM-3000 system diagram (Wang, 2005) 177

Figure A.4 ZOC-33 system diagram (Wang, 2005) 178

Figure B. 1 External pressure coefficients, GCpf (Walls) 184

Figure B.2 External pressure coefficients, GCP (Roofs) 185

xv
 NOMENCLATURE

B Building width (m)

cp Pressure coefficient

Cpe Exterior parapet surface pressure coefficient

Cpi Interior parapet surface pressure coefficient

Cpr Pressure coefficient on the roof

A Cp Net pressure coefficient on the parapet

Cpe-max Maximum instantaneous pressure coefficient on parapet exterior surface

Cpe-min Minimum instantaneous pressure coefficient on parapet exterior surface

Cpi-max Maximum instantaneous pressure coefficient on parapet interior surface

Cpi-min Minimum instantaneous pressure coefficient on parapet interior surface

dp Distance between first grid line and solid boundary (m)

f Frequency (Hz)

Gu Gust factor

GCpe Exterior peak parapet surface pressure coefficient

GCpi Internal peak parapet surface pressure coefficient

H Building height (m)

hp Parapet height (m)

I Turbulence Intensity

Je Jensen number

k Turbulence kinetic energy (m 2 /s 2 )

ks equivalent sand-grain roughness height (m)

L Building length (m)

Lx Longitudinal scale of turbulence (m)

m Mode

p Static pressure (N/m 2 )

p0 Reference pressure (N/m 2 )

qp Velocity pressure evaluated at the top of the parapet (N/m )

R Scaling length

Re Reynolds number

s Dispersion

T Time (s)

U, V, W Velocity components along x, y, z directions

Ug Gradient velocity (m/s)

u* Friction velocity (m/s)

u' Velocity fluctuation

x Distance from the leading corner (m)

Z+ Dimensionless distance

z Height above mean ground level (m)

zg Gradient height (m)

z0 Aerodynamic roughness height parameter (m)

a Power law exponent

e Turbulence dissipation rate (m /s )

^ Vertical distance from the base of the parapet (m)

xvii
0 Wind azimuth (°)

K Von Karman constant

jj. Dynamic viscosity of the air (Pa . s)

v Kinematic viscosity of the air (m 2 /s)

p Air density (kg/m )

<Je Standard deviation of wind direction (°)

<ju Standard deviation of wind speed (m/s)

xviii
 ABBREVIATIONS

2D Two Dimensional

3D Three Dimensional

ABL Atmospheric Boundary Layer

AIAA American Institute of Aeronautics and Astronautics

ASM Algebraic Stress Model

ASCE American Society of Civil Engineers

ASHRAE American Society of Heating, Refrigerating and Air-Conditioning

Engineers

BAL Building Aerodynamic Laboratory

BLWT Boundary Layer Wind Tunnel

BRE Building Research Establishment

CFD Computational Fluid Dynamics

CSCE Canadian Society of Civil Engineers

CVM Control Volume Method

CWE Computational Wind Engineering

DSD Down-Stream Distance

DSM Digital Service Module

ESDU Engineering Sciences Data Unit

FS Full Scale

LES Large Eddy Simulation

MBMA Metal Building Manufactures Association

xix
MSC Meteorological Service of Canada

NBCC National Building Code of Canada

NIST National Institute of Standards and Technology

NRC National Research Center

PVC Polyvinyl Chloride

RANS Reynolds-Average Navier-Stocks

RE Reynolds Equation

RLZ Realizable

RNG Renormalization Group

RSM Reynolds Stress Model

RVM Random Vortex Method

rms root mean square

SIMPLE Semi-Implicit Method for Pressure-Linked Equations

TTU Texas Tech University

UWO University of Western Ontario

USD Up-Stream Distance

USNRC United States Nuclear Regulatory Commission

WERFL Wind Engineering Research Field Laboratory

WT Wind Tunnel

ZOC Zero, Operate, Calibrate

xx
 Chapter One

INTRODUCTION

The majority of structures built all over the world can be categorized as low-rise

buildings. An industrial or residential low-rise building can be defined as a structure that

has a height to width ratio (H/B) of less than unity. Wind loads on low-rise buildings are

highly fluctuating and difficult to be determined since such buildings are located in the

lower region of the atmosphere where atmospheric turbulence and speed gradients are

stronger. The lateral strength of a low-rise building is mainly governed by wind loads

rather than the high seismic zones, particularly in zones of severe wind. This type of

buildings is more prone to wind damage than other structures (NBCC, 2005). Large part

of wind damage to a low-rise building is restricted to the envelope of the building, in

particular to roof sheathing.

1.1 Testing of Low-rise Buildings

Boundary layer wind tunnel experimentation is considered the basic tool of almost all

wind engineering studies of wind loads on structures. Wind tunnel enables modeling the

complex nature of the wind itself and its interaction with an object. Several wind-tunnel

examinations were focused on the evaluation of wind-induced pressures on flat roofs of

low-rise buildings, with and without parapets. Current technology shows that most of

1
these experiments can provide highly reliable results. However, model testing of low-rise

buildings has limitations of simulation. Natural wind conditions are simulated by

adopting the turbulent boundary layer in wind tunnel; however, the Reynolds numbers in

wind tunnel tests are much smaller than those for real buildings.

Recently, Computational Fluid Dynamics (CFD) has the potential to be used for

optimization analysis of different building shapes and arrangements. The boundary

effects can be avoided with the numerical simulation. However, since wind flow is

turbulent and its interaction with buildings is characterized by high Reynolds number of

the order of 106 - 107, numerical calculation for such cases in 3-D is very complicated.

Therefore, the application of CFD method is limited and based on many assumptions.

Full-scale testing is the most reliable to obtain wind-measured data, which represents the

real life wind loads on buildings and structures. Consequently, in wind engineering

research, it is extremely important to conduct field measurements of wind flow on full-

scale buildings in order to reveal the real characteristics of both atmospheric surface layer

and wind pressures. Such measurements provide data to validate wind tunnel results, to

overcome the deficiencies of model-scale testing and to improve the understanding of

wind loads acting on structural systems (Mehta, 2004). Reliable full-scale results are also

essential for verification and developments of numerical simulation of atmospheric flow

around buildings. However, this type of testing is expensive, time consuming and has

manpower intensive nature. As a result, limited number of full-scale wind studies was

reported, compared to the wind tunnel and numerical ones.

2
1.2 Flat Roofs with Parapets

Figure 1.1 shows an aerial photograph of buildings in a city center. It is quite clear that

the majority of flat-roofed buildings have parapets around them. A parapet is a

significant element for any low or high-rise building. Parapets are used for joining of

wall and roof membrane in addition to their safety purposes. Many flat or shallow pitch

roofs fail near the windward edge in the region of severe local suction. Various

solutions, ranging from specialized aerodynamic devices to traditional architectural

features, were used in an attempt to reduce these high local roof loads. Parapets are

probably the most widely used, they are expected to modify the wind flow over the roof

of a building and, thus, change the effective wind loads acting on the roof. Therefore,

they are considered the best-documented architectural device for reducing high roof

corner and edge loads. Parapets work by lifting the separated shear layer clear off the

roof surface, so dissipating the high local edge suctions over a much larger area

(Blackmore, 1988). However, the benefits gained by reducing edge loads may be offset

by increasing loads on interior regions (Stathopoulos, 1982).

1.3 Significance of the Study

Understanding and defining the wind loads on roof parapets is very important since these

loads must be included in the design of the main wind force resisting systems of

buildings, in addition to being essential for the design of parapets themselves.

3
Several previous investigations were focused on how the parapet modifies wind loads on

building roofs. However, no study was carried out in an attempt to measure the wind

loads on parapets themselves until recently (Stathopoulos et al., 2002-a and 2002-b) due

to modeling limitations, as it was very difficult to directly model and record the parapet

surface pressures in the wind tunnel. Accordingly, in the past practice, empirical values

were suggested based on a rational approach rather than from experimental

investigations, due to insufficient research. The ASCE-7 (2005) standards, estimate the

wind loads on parapets from pressures measured on the wind-ward wall and the roof of

the building in the vicinity of parapets. On the other hand, the NBCC (2005) does not

include any recommendations regarding the design wind loads of parapets. The

importance of the current study is evolved owing to the lack of guidance provided by

code provisions.

1.4 Research Objectives

The current thesis aims at understanding the flow-pressure mechanism on and around

roof parapets and measuring surface pressures on both faces of the parapet in order to

determine the design coefficients related to wind loading on parapets. Specific objectives

include:

• Evaluate mean and simultaneous peak pressures on both parapet surfaces

experimentally.

• Estimate the appropriate loading coefficients used for parapet design based on a

comprehensive experimental study.

4
 • Inspect the ability of a CFD code in predicting the wind pressures on building roofs

with parapets and on the parapets themselves.

• Provide code provisions with more information regarding the real wind loads on

parapets.

1.5 Work Methodology

In order to achieve a precise assessment of the previous objectives, the following path of

action is followed:

• Perform a complete field testing using a full-scale low-rise building located in an

open area and utilize the data to evaluate wind tunnel results. The experimental

building is located at one corner of the soccer field at Loyola Campus, Concordia

University, Montreal, Qc. A wooden perimeter parapet is fixed on the building roof

with tapping on all its surfaces. The field testing has dealt with extensive

measurements of the wind-induced pressures on the roof and the parapet.

• Carry out a detailed wind tunnel investigation, using a scaled model of the

experimental building, and investigate the surface parapet pressures as well as roof

pressures in the presence of parapet. Using 1:50 length scale, the experimental

building model is tested in the wind tunnel of Building Aerodynamics Laboratory

of Concordia University.

• Perform a three-dimensional numerical computation, using the CFD code, Fluent

6.1.22, with the RNG k-emodel and the RLZ k-s model and compare the results

with the experimental data.

5
 • Compare the overall results with the current codes of practice, namely NBCC

(2005) and ASCE 7 (2005), and develop a set of design recommendations regarding

wind loads on parapets.

1.6 Flow Patterns on a Low-rise Building

At any location on the exterior of a structure, the wind-induced pressures are likely to be

highly unsteady and to vary significantly from point to point. This is due to turbulence in

flow and turbulence caused by flow separation from the sharp edges of the building

(Stathopoulos and Surry, 1981). The scale of the resulting pressure fluctuations must

then depend on both the building size and the size of eddies in the oncoming wind.

High suctions on roofs are caused by the deflection and acceleration of the wind flow

after its separation at the roof edge (Stathopoulos, 1981). Corner areas are particularly

susceptible to wind damage and pressures on roofs are a function of wind direction and

the highest, i.e. worst, suction (Kind, 1988). At the oblique wind directions, near 45°

wind azimuth, strong vortices occur at the upwind roof corners.

Cook, 1985, explained the mechanism of roof edge vortex. Figure 1.2 shows the vortex

formation at roof corner and pressure distribution on a building roof. The flow marked 1

separates on the windward edge and tends to be displaced under the flow marked 2 as the

flow 2 separates immediately downwind of flow 1. The vorticity of shear layers from 1

adds to that of 2 and this process continues along the roof edges (windward), resulting in

a strong conical vortex called the Delta Wing Vortex. These vortices generally occur in

6
pairs, one on each windward edge of the roof. The center of each vortex is a region of

high negative pressure. The pair of vortices produces negative pressure behind each

windward edge of the roof.

1.7 Outline of the Thesis

The current thesis is organized in seven chapters as follows:

• Chapter two presents review for the available previous works about wind loading

on low-rise building roofs with parapets, in both wind tunnel and full-scale, in

addition to the research studies that have simulated wind flow using CFD

algorithms.

• Chapter three introduces theoretical background of the numerical solution and

describes the simulation procedure that has been performed using Fluent code.

• Chapter four presents full-scale and wind tunnel experimental work.

• Chapter five compares and evaluates the experimental results.

• Chapter six evaluates the numerical results

• Chapter seven includes conclusions, recommendations and future work.

7
An aerial view of low buildings in a city center, showing a plethora of
parapets

8
 Wt05« ol
aniefed How

Flow structure Pressure distribution

Figure 1.2: Vortex formation on roof edges (Cook, 1985)

9
 Chapter Two

LITERATURE REVIEW

Several previous studies focused on the investigations of wind loads on building roofs

and the effect of different parapet configurations on these loads. However, due to

modeling restrictions, it was very difficult to directly model and record the parapet

surface pressures. Therefore, wind loads on parapets themselves had not been examined

until recently (Stathopoulos et al, 2002-a and 2002-b). This chapter discusses mainly the

experimental studies that dealt with the wind loads on low-rise building roofs with and

without parapets, in both full-scale and wind tunnel. Recent investigations that attempt to

examine the wind loads on parapets themselves are also discussed. Some previous

numerical studies, which utilized CFD to simulate wind flow on and around low-rise

buildings, are presented.

2.1 Full-scale Investigations

Full-scale testing is important to achieve reliable data, which represent the real life wind

loads on buildings and structures, and to provide verifications of the wind tunnel results

(Mehta, 2004). Over the last four decades, a limited number of full-scale tests are

pursued for wind effects on low-rise buildings. Most of full-scale buildings are designed

to hold out wind loads using data and principles outlined in national codes of practice,
which almost completely relay upon data from wind-tunnel experiments. Eaton and

Mayne (1974) conducted a wind pressure experiment on a two-story building located in

Aylesbury, England. The building was 7 m x 13.3 m with a variable roof pitch of 5° to

45 0 from the horizontal. 72 pressure measuring transducers were used. Wind speed and

direction were measured with a 10 m mast at 3, 5 and 10 m height. Wind profile

upstream and wind flow in the downstream urban environment were investigated using

an anemometer at 20 m mast. Wind tunnel experiments were followed by Sill et al.

(1992) at 17 laboratories worldwide using a 1:100 model of the Aylesbury building.

Comparison between full-scale and wind tunnel measurements indicated that the

traditionally used similarity parameter, Jensen number, is not sufficient to ensure

similarity when significant isolated local roughness are presented. Moreover, the

variation in pressure coefficients in different experiments is due to the differences in the

method of data acquisition and in the measuring point of reference static and dynamic

pressures.

Marshall (1975) conducted field experiments on a single-family dwelling in Montana.

The dwelling was 6.8 m x 23 m with a wing of 5 m x 5.8 m and roof pitch of 11.5°. The

pressure transducers were mounted on the roof surface to avoid penetrating the roof

membrane of the house. Wind speed and direction were measured by a prop-vane

anemometer at a height of 6.1m. Wind tunnel testing was completed using a model of the

same house with scale of 1:50 in order to compare the model results with the full-scale

data. The comparison showed that the discrepancy in the results between model and full-

scale was perhaps due to inadequate simulation of wind characteristics in wind tunnel.

11
Marshall (1977) performed another field testing on a mobile home at National Bureau of

Standards in Gaithersburg (Maryland). The full-scale dimensions of that home were 3.7

m x 18.3 m and it could be rotated to obtain different wind angles of attack. The

anemometers were mounted at a five levels ranging from 1.5 m to 18 m. The surface

pressures and the total drag and lift force were measured. The results indicated that

negative pressure fluctuations occurred on the end walls and along the perimeter of the

roof.

Kim and Mehta (1979) studied the roof uplift loads on a full-scale building at Texas Tech

University (TTU). The analysis of spectra of wind speed and wind loads showed peaks

and lows at same frequencies. A statistical model was developed from this study which

concluded that the fluctuating component of roof uplift is best represented by a Gamma

probability density function. Levitan et al. (1986) studied pressure coefficients on the

roof of another building at TTU. The study indicated that mean and peak pressure

coefficients on the wind-ward roof area are greater for wind azimuth 6 = 60° than those

for 9 = 90°. Subsequently, the researchers at TTU constructed the Wind Engineering

Research Field Laboratory (WERFL) which is a permanent laboratory to study wind

effects on low-rise buildings in the field. The building is 9.1 m wide, 13.7 m long and 4

m high with an almost flat roof and it can be rotated in order to allow control over wind

angle of attack. The meteorological tower supports wind instruments at six levels: 1, 2.5,

4, 10, 21 and 49 m above the ground. A significant amount of data on winds and surface

pressures was collected (Levitan and Mehta, 1992-a and 1992-b).

12
Mehta et al. (1992) studied the roof corner pressure on the building to obtain baseline

data in the field for stationary winds. The largest pressures act along the roof edges while

significantly lower pressures occur at the interior taps and at the corner tap. The trends

shown in this investigation regarding the roof pressure data are reasonable indicating a

degree of validation of the field data. Moreover, Wagaman et al. (2002) studied the

separation bubble formed by flow perpendicular to the full-scale TTU building and

provided useful information for flow visualization researchers.

The Silsoe Research Institute in England provided useful full-scale measurements on the

Silsoe Structures Building. The building was 12.9 m wide, 24.1 m long and 4.1 m high

and had a gable roof angle of 10°. The structure consists of seven cold-formed steel

portal frames. Hoxey and Moran, 1983, focused on the geometric parameters that affect

wind loads on the building. Curved eaves and conventional sharp eaves were tested and

the study showed some inadequacy of many national standards for the prediction of loads

on low-rise buildings. Robertson (1992) studied the wind-induced response of the Silsoe

building in the field. One-hour recordings of free-stream wind pressure, wind direction,

external and internal pressures and structural strain were measured under strong wind

conditions (8 to 14 m/s). This work provided the details of pressure coefficient

information necessary to obtain a specific level of accuracy in predicting wind-induced

internal structural loadings (stresses).

Hoxey and Robertson (1994) completed full-scale measurements of surface pressures on

the Silsoe Building. They examined methods for analyzing full-scale data to obtain

13
pressure coefficients. The study shows that the quasi-steady method predicts the

maximum pressure exerted on the cladding and these pressures can be averaged over an

appropriate area to obtain design cladding loads. Moreover, approximating the non-

uniform wind load over a roof slope by an area-averaged pressure coefficient led to

acceptable predictions of stresses and deflections.

An experimental low-rise building was built, in 1985, in Concordia University for

research dealing with the control of rain penetration through pressurized cavity walls.

The building was a small one-room house 3.3 m high, 3.7 m long and 3.2 m wide with

sloped roof, which was modified to a flat roof in 1990. It is located in an open area,

beside the soccer field of Loyola Campus. Stathopoulos and Baskaran (1990-a)

examined the wind pressures on roof corners of that building in the field. A three-cup

anemometer and wind vane were mounted on a tower 20 m away from the building and at

a height of 4.7 m. The results of the study indicated that very high suctions occur indeed

on points very close to the roof corner for oblique wind directions and the comparison

with wind tunnel results showed good agreement in terms of mean pressure coefficients.

Maruyama et al. (2004) measured the pressure and flow using a full-scale 2.4 m cube

located in an open field. A number of anemometers were arranged on a 10 m high tower

around the cube. Wind and pressure data were measured simultaneously. The surface

boundary layer develops up to the height of the tower and the cube is well immersed in

the turbulent layer. The correlation between the upcoming flow velocities at an upwind

14
measuring point and the pressures on the cube surface was relatively strong: positive

correlation on the wall and negative correlation on the roof.

2.1.1 Building roofs with parapets

Stathopoulos et al. (1999) carried out a field testing on the full-scale experimental

building of Concordia University in order to investigate the effect of parapet on roof

pressures; which may be considered the only full-scale study regarding roof parapets. The

anemometer and vane were located on a tower mounted on the building roof. Field data

were compared with wind-tunnel experimentation of the building model with different

length scales (Marath, 1992) and with the results of other studies (Figure 2.1).

The full-scale investigation showed that lower parapets which have hp /L < 0.02 increase

the corner suctions and higher parapets (hp/L > 0.02) reduced these suctions, where hp is

parapet height and L is the building length. Considering very low parapets, hp/L = 0.008

and 0.016, roof pressures were found to be more depended on the ratio of parapet height

to building height (h p /H). Such study in full-scale provided useful clarification for the

previous wind tunnel results regarding the effect of parapet on roof pressures. However,

only mean pressures were examined at that stage. It should be noted that the current

study is conducted using the same full-scale building after it was relocated to the corner

of the soccer field in 2003.

15
Recently, the researchers at TTU recognized the importance of studying parapet loads in

full-scale. Therefore, as part of the NIST project at Texas Tech, a parapet was

constructed at the WERFL building (see Figure 2.2). The experiment was intended to

examine wind-induced pressures on the inside, outside and top surfaces of the parapet, in

addition to the internal pressures inside the parapet section. However, no complete study

can be shown until now.

2.2 Wind Tunnel Investigations

The first comprehensive study aimed at defining wind loads on low-rise buildings was

performed by Stathopoulos (1979) in the boundary layer wind tunnel of the University of

Western Ontario. The parameters examined included building geometry, surrounding

terrain, wind direction and building attachments. One of the significant developments in

this study was the introduction of a pneumatic averaging system that allowed for the

spatial and time averaging of the surface pressures recorded over the surface of the

building. This enabled the pressures recorded at each individual pressure tap to be

combined at each instant in time to provide a time series of the combined pressures. By

doing so, area-averaged loads for areas of various sizes could be accurately determined.

The analysis methodology provided in this study is applied to the majority of wind tunnel

studies.

The full-scale measurements on TTU building provided a criterion data for verifying

wind tunnel experiments and numerical simulations. Surry (1986) compared pressure

16
measurement results on the full-scale building at TTU with those obtained from a 1:50

scaled model tested in wind tunnel. The study also examined the effect of terrain

roughness on mean and peak pressure coefficients. The comparison showed good

agreement, however for oblique winds the data indicated significant differences in peak

coefficients.

Okada and Ha (1992) tested three scale models of the TTU building, 1:65, 1:100 and

1:150. The comparison shows good agreement in terms of mean wind-pressure

coefficients but large differences of rms and peak pressures. The authors concluded that

the most important factor regarding these differences was related to the frequency-

response characteristics of the pressure-measurement system used in the study.

Cheung et al. (1997) attained good agreement of a 1/10 scale model of TTU building with

full-scale results, in which no artificial increase in the longitudinal turbulence intensity

was made. The authors suggested that the increased Reynolds number was believed to

play a significant part in their achievement.

Ham and Bienkiewicz (1998) performed a study of approach wind flow and wind-

induced pressure on a 1:50 geometrical scale model of TTU building at Colorado State

University. The physical simulation technique developed using conventional wind tunnel

devices led to a very good representation of the TTU nominal flow. This agreement was

attributed to improved modeling of approach flow and relatively high frequency response

of the pressure measuring system.

17
Tieleman et al. (1998) interested in the effect of the different flow parameters on the

observed pressure coefficients, they compared the full-scale data collected from TTU

laboratory with results of mean, rms, and peak pressure on the roof of a l:50-scaled

model of the same building. The authors stated that the reproduction of the horizontal

turbulence intensities and their small-scale turbulence content in the wind tunnel is very

important for the possibility of agreement between model and field roof pressure

coefficients.

Tieleman et al. (2008) represented the distributions of peak suction forces in separation

regions on a surface-mounted prism by Gumbel distributions, using the field and

laboratory data of the WERFL of TTU. The authors stated that wind engineers must

realize that wind tunnel experiments of the atmospheric flows never duplicate the non-

stationary conditions that occur in the atmosphere. Consequently, the peak distributions

obtained over a long period with the method of moments exhibit much greater dispersion

of the peaks than the distributions obtained over a much shorter period with the Sadek-

Simiu (2002) procedure.

Full-scale testing of the Silsoe Building provides an opportunity to undertake detailed

full-scale/model scale wind pressure comparisons. Richardson and Surry (1991 and

1992) performed wind tunnel model of scale 1:100 for the building. Collaborative

research with the Boundary Layer Wind Tunnel Laboratory of University of Western

Ontario was conducted. The comparisons of wind tunnel results with the obtained full-

scale data showed that very good agreement is possible between full and model-scale

18
data. However, the model-scale underestimates the suctions significantly where the

separation occur on the wind-ward roof slope.

Richards et al. (2007) reported a wind-tunnel modeling of the Silsoe 6 m Cube at the

University of Auckland. The approach taken in this study was to match the velocity

profile and the high-frequency turbulence as closely as possible. Similar mean pressure

distributions were obtained as a result. In addition, reasonable agreement is obtained by

expressing the peak pressure coefficient as the ratio of the extreme surface pressures to

the peak dynamic pressure observed during the run.

Lin et al. (1995) proved experimentally that corner vortices for both smooth and

boundary layer flow dominate the absolute suction pressures coefficient, Cp, while their

effects are dramatically reduced with increasing distance from the leading corner and

with increasing tributary area. The relationship between the effective load and tributary

area is insensitive to the building dimensions if area is normalized by H2 where H is the

building height (Lin and Surry, 1998) and it is only slightly dependent on the shape of the

areas that include the corner point. In this region, loads reduce rapidly with increasing

tributary area. The authors used simultaneous time series of pressures measured at

locations within the corner region to form new time series of uplift loads by instantaneous

spatial averaging, for various building heights and plan dimensions.

Kawai (1997) also investigated the structure of conical vortices on a flat roof in oblique

flow by hot-wire velocity measurements in smooth and turbulent one. The author stated

19
that the strength of the conical vortices was larger in the smooth flow than in the

turbulent flow. Consequently, the larger mean suction acted on the roof in smooth flow,

and each conical vortex grew and decayed alternatively to produce imbalance in the

suction distribution.

Uematsu and Isyumov (1998) provided an empirical formula for estimating the minimum

pressure coefficient in the leading edge and corner regions which could be applied to

buildings with roof pitches larger than approximately 4:12. In the leading edge and

corner regions, where severe suctions occurred, the effect of time average on the peak

pressure was related to that of spatial average through the equivalent length and the

square root of the area. The analysis of wind effects on design (Kasperski, 1996), taking

into account a characteristic load combination of at least dead load, showed that final

design was based on a negative wind-induced bending moment in the downwind frame

corner for many practical cases.

2.2.1 Wind loads on roofs with parapets

Leutheusser (1964) carried out the first detailed study regarding the effect of parapets on

the wind loads of flat roofs under uniform conditions. The author believed that, for

oblique wind directions, the parapets cause an extreme reduction of the high suctions

particularly at the corners and edges of the roof. Furthermore, the presence of parapets

made the roof pressure distribution more uniform but it did not affect the magnitude of

the average roof pressure coefficient and the pressure distribution over the building

20
sidewalls. However, the study was carried out in uniform flow conditions and the mean

velocity and turbulence intensity profiles found in natural wind were disregarded, thus

the above results may not be representative of turbulent flow conditions. Subsequently,

Columbus (1972) studied the effect of parapets considering turbulent, in addition to

uniform flow conditions. However, the study shows that parapets do not cause any

reduction on local mean pressures in turbulent flow in contrast to the case of uniform

flow.

Kind (1974) studied the gravels blown off rooftops and found that the critical wind

speeds increase with increasing parapet height. This indicates that parapets reduce the

pressure coefficients on the roof. On the contrary, Davenport and Surry (1974) indicated

that local mean suctions are increased when the parapet is added, particularly for oblique

wind direction. Kramer (1978) suggested that parapet height significantly affect the

corner roof pressures. For the case of h/B > 0.04, where hp is parapet height and B is

building width, corner pressures were reduced by 70 %. Consistent with that results,

Sockel and Taucher (1980) concluded that when h/H = 0.2, where H is building height,

the mean suction pressures at the roof corner were reduced by 50 %. Also, parapets act

to reduce the fluctuating pressure components.

A systematic approach was followed by Stathopoulos (1982) to clarify the effect of

parapets on flat roof pressures of low-rise buildings as well as to inspect the discrepancies

that were found between some authors. The study concluded that the addition of parapets

around the roof causes a reduction of the roof-edge local suction by 30%, these suctions

21
slightly increase on the interior roof. On the other hand, parapets tend to increase the

mean local suction and positive peak and mean pressure coefficients as well, particularly

at the corner areas. Root mean square (rms) pressure coefficients show little changes in

the building with parapets. The study has recommended specifications to the NBCC

regarding the wind loads on low-rise buildings with parapets.

Lythe and Surry (1983) performed a comprehensive research examining a number of

parapet and building heights in order to study the effects of parapets on wind load

distribution on roofs. The study showed that:

• The distribution of loads on a roof consists of a very highly loaded corner region, a

highly loaded edge region and a more moderately loaded interior region.

• Edge region width is increased with building height and by the addition of

parapets.

• Corner region width remains nearly constant with building height except for cases

of very high parapets.

• Corner region mean and peak pressures increase on low buildings with low parapets

but decrease otherwise.

An investigation of wind-induced failure of various roofing systems was carried out by

testing different building shapes, each with a number of different parapet heights (Kind,

1986); and followed by pressure measurements on two small low-rise building models in

a relatively thick boundary layer, with pressure taps very close to the roof edges (Kind,

1988). The author found that increasing parapet height leads to a decrease of worst

22
suction coefficients. Moreover, worst suctions on flat-roofed low-rise buildings were only

mildly sensitive to the characteristics of the approach-flow boundary layer.

Stathopoulos and Baskaran (1988-a, 1988-b and 1988-c) produced a series of articles

clarifying the effect of parapets on local surface pressures (see Figure 2.3). Such study

examined the effects of building height, ranging from 12 to 145 m, and large range of

parapet heights including low parapets (0-3 m) on both local and area-averaged roof

pressures for a variety of wind directions. The effect of one-side parapet, such as

billboard, on the roof corner pressures was also examined. However, the building tested

in this case is a high-rise building, with height H = 96 m. The conclusions of these

studies could be summarized as follows:

• Parapets do not affect roof interior wind loads.

• Low parapets (<1.0 m) reduce local high suctions on roof edges by 30 %, increase

them on roof corners by 100 % and increase area-averaged loads on roof corners by

20 %.

• High parapets (>1.0 m): reduce local high suctions on roof edges by 15 % and

increase these suctions on roof corners by 50 %.

• One-side parapets induce higher mean and peak local suctions at roof corners, even

for high parapets, in comparison with perimeter parapets.

Kareem and Lu (1992) studied the mean and fluctuating pressure distributions on the roof

of a square cross-section building with adjustable height. Measurements were done in

two terrains representing open country and urban flow conditions. Perimeter parapet

23
walls of two different heights were introduced. Parapets were found to reduces peak

suction pressure. In agreement with Kind (1988) the effectiveness of perimeter parapets

in reducing suction pressure coefficients increased by increasing their height.

Badian (1992) carried out wind tunnel experiments in order to evaluate the wind loads on

flat roof edges and corners on buildings with parapets, the models were tested with high

number of taps distributed on one corner and along its adjacent edge. The study

confirmed that parapets generally reduced the high suctions on the roof edges.

Furthermore, parapets with a height of 1.0 m or lower increased peak and mean pressures

for low buildings.

Bienkiewicz and Sun (1992) performed fundamental studies of wind loading on the roof

of a 1/50-scaled model of the TTU test building with some parapet heights. The presence

of low parapets (hp/H < 0.04) results in an increase in the overall maximum mean and

peak suction. Moreover, high parapets (h p /H> 0.08) reduce both values. The researchers

believed that parapets are very useful in reducing high pressures created at roof corners

and edges. Nevertheless, the height and shape of parapets significantly affect this

reduction. Therefore, some studies were interested in examining different parapet

configurations.

Surry and Lin (1995) investigated the sensitivity of high suctions near corners to the use

of relatively minor alterations to the roof corner geometry. Different parapet

configurations were used as a modification to the roof corners. The study shows that:

24
 • Porous parapets leading to a reduction of the high suctions of up to 70 % near the

corner.

• For the saw-tooth partial parapets, the high suctions near the corner observed are

reduced by up to 40 %.

• Reductions of about 60 % in the high suction magnitude are also reached near the

corner, for the rooftop splitter configurations with the porous splitters being slightly

better than the solid ones.

Mans et al. (2003) have also studied the effect of a single parapet (above only one wall)

on local roof pressures near the leading corner of a low-rise building. The study focused

on a comparison of the results with those previously obtained for a continuous parapet.

The results of this study agreed with the conclusion of Baskaran and Stathopoulos (1988-

b) and indicated that isolated parapets, regardless of height, generate larger suction

pressures on the roof surface in comparison to a continuous parapet and to the case of no

parapet, with regards to low-rise buildings. In contrast, tall continuous parapets (hp/H >

0.4) may reduce the corner suction pressures, which agreed with Stathopoulos and

Baskaran, 1988-c.

Pindado and Meseguer (2003) examined different parapet configurations, including non-

standard configuration (cantilevered parapets), to reduce the wind suctions generated on

the roofs by conical vortices. This study demonstrated that low-height parapets (hp/H <

0.05), where hp is parapet height and H is building height, with medium porosity are more

25
efficient than solid parapets. Also, low-height cantilevered parapets (h/H < 0.031)

produced a very effective reduction more than vertical parapets, either solid or porous.

2.2.2 Wind loads on parapets themselves

The first investigation for the wind loads on parapets themselves was performed by

Stathopoulos et al. (2002-a). Due to modeling limitations, no study previously attempted

to directly model and record the parapet surface pressures. Instead, in past practice the

wind loads on the parapet were estimated from pressures measured on the wall and roof

of the building in the vicinity of the parapet. In this case, it was assumed the pressures

on the front wall surface are equivalent to the external parapet surface pressures and the

interior parapet surface pressures are identical to the roof edge pressures. This

assumption does not take into consideration the differences between roof edge/corner

suctions and those on the inside surface of the parapet. Therefore, the primary focus of

the study was to determine the validity of this estimation method. The study concluded

that parapet design loads obtained by applying the National Building Code of Canada

(NBCC, 2002) for windward wall and roof region were significantly higher than actual

loads on the tested parapet. However, the dimensions of the modeled building (L/H = 1

and L/H = 2, where L is building length and H is building height) classify it more as an

intermediate-rise building rather than a low-rise building.

Subsequently, Stathopoulos et al. (2002-b), performed an experimental study to

determine wind pressure coefficient appropriate for the design of parapets for low-rise
buildings (L/H = 3). The study focuses on measuring the surface parapet pressures

simultaneously in order to investigate the combination effect of parapet surface loads.

The comparison of the results with the ASCE 7-02, provisions shows that the latter to be

on the conservative side for all cases.

A study was completed by Mans et al. (2005) in order to analyze wind-induced pressures

on parapets of low-rise buildings. In agreement with Stathopoulos et al. (2002-a) the

authors found that evaluation of parapet loads, using pressures recorded on the wall and

roof in the vicinity of a parapet; overestimates the measured parapet loading by

approximately 10 %.

2.3 Numerical Computations

Many studies previously attempted to simulate wind flow on and around structures by

means of numerical algorithms (Computational Fluid Dynamics, CFD) in order to predict

different wind parameters. Hanson et al. (1984) examined the possibilities of developing

a numerical model to simulate wind flow around buildings. The study used both the

Random Vortex Method (RVM) and the Control Volume Method (CVM). The results

were not compared with experimental data. However, significant contribution was made

to demonstrate the advantages of CVM over RVM. The latter was found not compatible

with turbulence models and inefficient in predicting wind flow around three-dimensional

sharp edge buildings.

27
Paterson (1986) attempted to solve the Reynolds Equation (RE) using the standard

k-s turbulence model developed by Launder and Spalding (1974). The author employed

the CVM to discretize the RE. The computed results were compared with different wind

tunnel and full-scale measurements. However, the recirculation zone and the pressure on

the building sides were not acceptable for all cases. It was included that the predicted

parameters need improvements particularly at separation zone and in the wake.

Murakami and Mochida (1988) computed the steady wind conditions around a cubic

model using the standard k-s model. The study revealed that the numerical evaluation

of wind pressures on flat roofs of rectangular buildings is very complex. Murakami

(1990) attempted the numerical simulation of the airflow around a cube for unsteady flow

using the standard k-s model for turbulence. The author concluded that the k-s model

with a fine mesh can reproduce mean velocity field and mean pressure field more

accurately than a coarse mesh compared with wind tunnel results. However, significant

differences were observed in the distribution of the turbulent energy around the windward

corner and in the wake (see Figure 2.4).

Stathopoulos and Baskaran (1990-b) evaluated wind effects around buildings through an

individually developed code. In order to improve the results of standard k-s model in

predicting recirculation and separation regions, two simple modifications were applied to

the calculations, namely, a streamline-curvature correction and a preferential-dissipation

correction. A new zonal-treatment procedure was developed to link the solid boundary

nodes with the computational domain for the turbulence variables (k and s).

28
In 1992, Murakami analyzed velocity-pressure fields and wind-induced forces on and

around a building model, and compared the results with those from the wind tunnel. The

study confirmed that the results of the 3D computation match well to the experimental

data, while 2D results include significant discrepancies. More detailed time-averaged

flow fields around a cube within a surface boundary layer using three types of well

known turbulence models, namely the k—e eddy viscosity model, the Algebraic Stress

Model (ASM) and the Large Eddy Simulation (LES) model, were given by Murakami et

al. (1992). The model equations used for ASM were based on the methods of Rodi,

1976, and Gibson and Launder, 1978, except for the treatment of the wall reflection term

(Murakami et al. 1990). The accuracy of these simulations was assessed by comparison

with results from wind tunnel tests. LES results show the best agreement with the

experimental data. Standard k—e model overestimates the kinetic energy around the

leading edge of the building and thus it underestimates both the length and height of the

recirculation zone on the top of the building. However, some modified k—e models can

overcome this drawback.

Selvam (1992) simulated the three-dimensional wind flow around the TTU building

model using k - s model. The computational data with staggered grid arrangement agreed

well with the experimental results. However, the numerical results from the non-

staggered grid failed to predict the roof pressures. In addition, Selvam (1997) used LES

to solve the Navier-Stokes equations for wind flow around the TTU building. The

computed mean pressures were in good agreement with field measurement results.

However, the peak pressures were much higher than the field data.

29
Stathopoulos and Zhou (1993) proposed a two-layer methodology combining the

&-£-model in the external flow region with either a one-equation model (Norris and

Reynolds, 1975), or a modified k-s model, in the near wall area. This two-layer method

based on the one equation model was found effective in predicting the separation above

the roof surface and near the side walls of a cubic building.

He and Song (1997) also, simulated the wind flow around the TTU building and roof

corner vortex using LES method. The authors stated that the three-dimensional roof

vortex model was in good agreement with wind-tunnel and full-scale results.

A systematic investigation of numerical effects on the computation of turbulent flows

over a square cylinder has been made by Lee (1997). The author found that some

conventional k-s models may give reasonable prediction when proper numerical

parameters are included.

Meroney et al. (1999) compared recirculation zones around several building shapes using

standard k-s model, RNG k-s model and Reynolds Stress Model (RSM) using Fluent

4.2.8. All constants in these turbulence models are default values. The authors stated

that the RSM provides a more accurate flow field than both types of k-s models. Also,

the results showed that pressure coefficients on the front, back and rooftop can be

predicted very well when compared with experimental data, although recirculation zones

around the building were not reproduced accurately. However, no explanation was

30
provided regarding this good agreement of pressure coefficients and bad agreement of

flow field with the experimental data.

Chang and Meroney (2003) investigate the sensitivity of high roof suctions to the

presence of surroundings using Fluent code with k-s model. The overall numerical

results appear similar to the experimental data of a 1/50 scale model of TTU building.

However, the numerical results of the front edge region at flow separation indicate higher

suction.

In order to improve the overestimated kinetic energy around the leading edge of the

building by the standard k-s model, Gao and Chow (2005) proposed a method to change

the speed distributions around the sharp corner of the building. By limiting the

longitudinal velocities in the first cell adjacent to the sharp edge of the cube and making

good use of the wall functions at the intersection cells of the velocity components. In this

case, the positions of maximum turbulent kinetic energy and the flow separation and

reattachment could be predicted by the standard k-s model.

Zang and Gu (2008) used a revised k-s model by Re-Normalizing the Group of

equations (RNG k-s model) to solve the RANS equations, in order to investigate the

wind-induced interference effects on the pressure distributions on a building adjacent to

another one in staggered arrangement. The numerical results were in qualitative

agreement with the experimental data. The author also confirmed that the RNG k-s

model seems to be a useful tool for predictions of wind pressures.

31
Based on the previous survey, the expected flow stream lines around the current test

building with the parapet are presented in Figure 2.5.

2.4 Current Wind Loading Standards Regarding Parapets

Wind standards and codes provide very little guidance regarding the wind loading on

parapets, or their influence on the local (cladding and component) loads. Moreover, their

recommendations are based on a rational approach and do not come from experimental

research. Defining a wind loading standard to be used in the design of low-rise buildings

requires a consideration of many parameters, including building geometry, surrounding

terrain and wind direction. Thus, experimental studies for evaluating parapet loads

become a primary task for the researchers in this area. Although, the standards do

consider how the loading on the parapet itself may affect the roof surface pressure

coefficients, they do not consider wind loads on parapets themselves.

The ASCE 7-02 considered the importance of the design wind loads of parapets. In the

absence of research, a design method in the wind load section uses wall pressures for the

outside surface of the parapet and roof edge/corner suctions for the inside surface in order

to estimate the total drag force on the parapet, as shown in Figure 2.6. This method may

be conservative since it does not consider the lack of simultaneous occurrence of

maximum pressure value on the outside surface and the maximum suction on the inside

surface of the parapet on the windward side of a roof (Stathopoulos et al., 2002-b).

32
For the design of the main structural system the ASCE 7-02 (2002) standard recommends

net coefficients (GCP) of + 1.8 and - 1.1 for the windward and leeward parapets,

respectively. However, after completion of preliminary work on which the present study

is based (Stathopoulos et al., 2002-a and 2002-b), these values were revised in the ASCE

7-05 version to + 1.5 and - 1.0. With regards to the loads on the component and

cladding, ASCE 7-05 recommends GCP values of + 3.8 and - 2.4 for windward and

leeward parapets, respectively, at corner region and + 2.8 and - 2.2 for mid-span location.

Appendix B includes the ASCE 7-05 recommendations, as written in the standard.

NBCC (2005) does not provide explicit recommendation regarding wind loads on

parapets. However, a similar procedure with that of the ASCE 7-05 can be used.

2.5 Summary

The main points concluded in this chapter can be summarized as follows:

• Peak suction pressures occur at the leading corner of the building, where the flow

separates from the building roof creating two vortices. These pressures reduce

significantly with increasing distance from that corner. The worst angle of attack is

generally between 30° and 60° from the normal extending of one face of the

building.

• Large differences have been recorded for root mean square (rms) and peak pressure

values particularly near roof corners. These differences were due to inequality in

tap diameter, edge geometry, turbulence scale or Reynolds number, in addition to

33
 the corner vortex development or the presence of convective and non-stationary

effects in full-scale.

• Regarding the effect of parapet on roof pressures:

- The addition of parapets around the building roofs reduces the high suction

pressures on roof corners (Leutheusser, 1964; Kind, 1974; Stathopoulos, 1982).

- The presence of parapets made the roof pressure distribution more uniform

(Leutheusser, 1964; Lythe and Surry, 1983; Baskaran and Stathopoulos, 1988).

Parapet heights significantly affect roof corner pressures: the addition of

relatively high parapets (h p > 1 m) reduces the high suction pressures on roof

edges and corners, on the other hand, low parapets (h p < 1 m) may increase

these pressures (Kramer 1978; Sockel and Taucher, 1980; Kind, 1988; Kareem

and Lu, 1992; Stathopoulos and Baskaran, 1987; Badian, 1992; Bienkiewicz

and Sun; Mans, 2003).

Various parapet configurations were examined as in order to effectively reduce

the corner suction pressures (Stathopoulos and Baskaran, 1988; Surry and Lin,

1995; Pindado and Meseguer, 2003; Mans et al., 2003)

Some discrepancies exist between authors as they considered that parapets do

not cause any reduction in roof mean pressures (Columbus, 1972) or that the

addition of parapets increases the local mean suctions on building roofs

(Davenport and Surry, 1974).

• The few studies that dealt with the wind loads on the parapet itself provided very

useful information; however, only limited wind tunnel investigations were

completed. These studies concluded that code provisions may be conservative in

34
 predicting parapet design loads (Stathopoulos et al., 2002-a and 2002-b; Mans et al.,

2005).

• No previous full-scale study was attempted to measure the real wind-induced

pressures on parapets.

• With regard to the numerical computational studies:

- The standard k - £ model overestimates the kinetic energy around the leading

edge of the building and thus it underestimates both the length and height of the

recirculation zone on the top of the building.

Several modified k-e models have been proposed to overcome significant errors

such as overestimation of turbulence kinetic energy k, results in the standard k-

smodel. These modified models accurately estimate the flow separation and

reattachment to the roof

Theoretically, non-linear models such as LES and RSM are more accurate than

linear models ( k - e models); therefore such models are recommended, although

they still require some improvements. However, these methods apply to

unsteady-state large scale motion of turbulent flow and need finer grid

arrangement, i.e. more computational resources, compared to k-smethod.

Some attempts were made in order to compute peak pressures on different

building by means of non-linear models. However, these pressures were not

predicted reliably.

35
 -0.6

-3.5 -

0.02 0.04 0.06 0.08 0.1 0.12 0.14

h/L

Figure 2.1: Corner mean pressure coefficients in terms of h/L

(Stathopoulos et al., 1999)

36
Figure 2.2: Parapet wall at the WERFL of TTU

(parapet under construction, 2004)

37
Figure 2.3: Roof corner pressure coefficients for cut-parapet configurations

(Stathopoulos and Baskaran, 1988-b)

38
 turbulent

Figure 2.4: Turbulent air flow field around a cube (Murakami, 1990)

39
Turbulent
boundary
layer Parapet Recirculation

Figure 2.5: Mean stream lines expected around the center line of the current

building

40
TO
S
•t
«

•CA
•JS
C
3
«
a
2T
6!
a
65
O
ft-

B"
TO
s
O
ST
«
•a
to
•1
Si
TS
re

•
C/3
0
w1

©
©
Ul
Positive wall pressure
 Chapter Three

NUMERICAL SIMULATION

Accurate computer predictions are very useful since numerical methods can be less

expensive and less time consuming than comparable field or wind tunnel testing.

Numerical simulation of wind flow over the test building model with parapets is

performed using the commercial Computational Fluid Dynamics (CFD) code Fluent

6.1.22. The steady-state Reynolds-Averaged Navier-Stokes (RANS) equations are solved

using modified k-s turbulence models. This chapter presents the numerical simulation

procedure and explains the revised turbulence models that were used. The computation

has been conducted for wind flow perpendicular to the building face (i.e. zero wind

azimuth). Three-dimensional calculations of air flow are essential to obtain accurate

results for flow simulation around buildings. The numerical simulation was adopted for

the same configurations of the building model; thus wind flow characteristics for the

wind tunnel are used.

3.1 Introduction to Fluent

Fluent is based on a finite-volume discretization of the equations of motion and is utilized

for modeling fluid flow and heat transfer in complex geometries. Fluent can identify key

areas of the flow such as building geometries, stack configurations and incident flow

42
combinations which create areas of flow (Flowe and Kumar, 2000). A complete

description of the code, many validation examples, and a number of published papers and

reports may be found on the web at http://www.fluent.com. The program provides complete

mesh flexibility, solving flow problems with unstructured meshes. Fluent package

consists of many tools for defining a separate flow problem, setting boundary and initial

conditions and solving a set of complex equations for conservation of mass, momentum

and energy. A steady-state Reynolds averaged turbulence model, k-s, has been used.

Therefore, the output of the computation consists only of mean values.

3.2 Governing Equations

The governing equations of incompressible turbulent wind flow around a bluff body are

the Continuity equation and the steady-state Reynolds-averaged Navier-Stokes (RANS)

equations:

3U;
-^ = 0 3.1
8X:

dUj'
J
' +
pUjU j 3.2
dx dx, dx, v dx,J dx 1 y

where: i,j= 1, 2, 3, U and u' are mean and fluctuating velocity, respectively, p is air

density, JLI is the dynamic viscosity and p is the mean pressure. The conservative general

form of all flow equations under steady state conditions, including equations for the

scalar quantities such as k and s, when derived for a flow property, <f>, which represents

the unknown variables (u, p, etc.), can be expressed in the following form:

43
 d(pUO) | d(pVO) = d dO ao
+- s
V
3.3
dx dy dx dx dv dy source term
Convection terms diffusion terms

Equation 3.3 is known as Transport equation and is used as the starting point for

computational procedures in finite volume method for developing CFD codes. The key

step is integrating Transport equation over a three-dimensional control volume.

For the continuity equation (3.1): O = 1, = 0 and =0

For the momentum equation (3.2): <t> = U„ + jut) and = — ( - P)

dxj

3.3 Turbulence Models

There are many types of turbulent models utilized in CFD codes. Fluent provide

different turbulent models, which explained in detail in Fluent 6.1.22 User's Guide

(2005), Volume-2. Three types of kinetic energy-dissipation rate (k-s) models are

described below namely: the Standard k-s model, the Realizable (RLZ) k-s model and the

Renormalization Group Theory (RNG) k-s model. A major difference between these

models is the method of calculating turbulent viscosity. The current simulation uses both

RLZ k-s and RNG k-s models since the standard k-s model was found not effective in

predicting the recirculation zones accurately.

44
3.3.1 Standard k-s model

The standard k-s model includes some simplifying assumptions, such as a turbulent

equilibrium assumption near the wall and the use of various constants in the turbulent

transport equations. Due to these assumptions, the model needs the least computer

resources comparing with the other models. However the validity of such method is

limited. The turbulent kinetic energy (k) and its dissipation rate (e) are obtained from the

following transport equations (Launder & Spalding, 1972):

Dk _d_ Mt dk
+ Gk-ps 3.4
' Dt 8X: k y dx-

Ds _d_
M+
EL ds + Cu-Gk ClEp- 3.5
Dt dx, as dX: k

k2 — d U ,
where: /ut = pC — , Gk = -pu^j representing the generation of turbulent

kinetic energy due to the mean velocity gradients.

3.3.2 Realizable k-e model (RLZ)

To improve the simulation results of the standard k-s model for cases in which the flow

includes strong streamline curvature, vortices and rotation, Shih et al. (1995) applied

some modifications, which include making CM variable and adding a new equation for s

based on the dynamic equation of the mean square velocity fluctuation. This modified

45
model called Realizable k-s. The dissipation equation in this model represents, very well,

the spectral energy transfer. Meanwhile, the destruction term does not have any

singularity even if k vanishes or becomes negative. Shih et al. (1995) found that the RLZ

k-s model performs well, comparing to the standard k-s model, for a variety of flows

including the recirculation in the wake. The transport equations for this model can be

obtained as follows:

Dk Mt dk
M+ + Gk-ps 3.6
Dt dx,. ' k j dxj

Ds ds
JU +ti- + pCxSs - pC2p 3.7
Dt dxj ers jdxj k + *Jvs

where: Cx = max 0.43, 7 , T]=S~, S = ^ISijSy ,

77 + 5

C2=\.9,ak=\,(7e=\.2.

3.3.3 RNG k-s model

The RNG k-s model used in Fluent is based on a mathematical technique called

Renormalization Group. Choudhury, 1993, provided more detailed description of RNG

theory and its application to turbulent flow. The transport equations have a similar form

as those of the standard k-s model and can be written as follows:

46
 f
p
Dk
—
d
= —
3.8
Dt dx: \

2
Ds d ( s
3.9
dxj k

where: C/ £ = 1.42, C ^ = 1.68, /ueff is the effective viscosity, a.k and ae are the inverse

effective Prandtl numbers for k and s respectively.

3.4 Numerical Simulation Procedure

The main steps during a numerical simulation consist of: pre-processing, solving and

post-processing.

Pre-processing is to prepare the work for numerical calculations including: selection of

the related physical phenomena that need to be numerically solved (i.e. wind-induced

pressures on the building roof and parapet), setting up the geometry of the region of

interest (i.e. computational domain), creating proper mesh generations for that domain

(i.e. grid formation) and setting up the boundary conditions for the governing equations.

Solving contains the procedures to get the solution of the equations. Fluent provides two

solvers: a segregated solver and a coupled solver. The current study uses the segregated

solver, which is appropriate for incompressible flow. The Semi-Implicit Method for

Pressure-Linked Equations (SIMPLE) scheme (Van Doormal and Raithby, 1984) is used

for introducing pressure into the continuity equation. The numerical simulation of wind

47
flow conditions around a low rectangular building is carried out until the residues of all

equations reach the round-off limit of machine accuracy. The residue is the difference

between the results of two successive steps and the relative residue is defined as the ratio

of the current residue to the residue after the first iteration. The relative residue is used as

the convergence criterion and is set to 10"6 for all simulations. The pressure and velocity

distribution are obtained when the residues reach the convergence limit.

Post-processing implies retrieving graphical and numerical information about the

variables from the solution. Wind-induced pressures on the building roof and parapet are

expressed in the form of pressure coefficient (Cp) as follows:

C ={£j!22LZlA 3.10
' Yipul

where p0 is the static pressure at free stream, x/2 pU 2

H is the dynamic pressure in which

UH is the mean wind speed at roof height. Figure 3.1 presents a flow chart of the

numerical solution method.

3.4.1 Computational domain

The wind flow is simulated in three dimensions over the building at 1/50 scale model.

The dimensions of the building model are L x B x H = 0.08 x 0.065 x 0.06 m 3 . The

building model with roof overhang was also tested. The computational domain was

selected to be large enough compared with the building model, so the small building set

in a large open field does not affect the mean wind flow conditions at the boundary of the

domain. The length of upstream field is 6H, where H is the building height. The
downstream field length must be long enough to ensure that the flow at the outlet

boundary is not affected by the building; 30H down-stream from the leeward face of the

building model is selected. The upper boundary is about 8H from roof surface and the

domain width is extended about 8H from each sideward face. The computational domain

has dimensions 2.24 x 1 x 0.5 m 3 . Figure 3.2 shows the details of computational domain

and building configuration. The figure also shows the configurations of the building

model with overhang

3.4.2 Grid generation

Some of previous numerical studies have attributed poor prediction of wind pressure of

flat roofs because of coarse grid arrangement (Murakami, 1990). of a non-staggered grid

system is easier than that of a staggered system; however the pressures near solid surfaces

had to be obtained by extrapolation. Also, a non-staggered grid system is not

recommended to predict the roof pressures of low-rise buildings (Selvam, 1992). The

grid scheme used in the current computation is the rectangular staggered grid system,

since this is considered as the most 'body-fitted' system for buildings of rectangular

shapes. Grid arrangements are generated with dense grid lines near solid boundaries and

relatively coarse grid lines far from solid boundaries. The flow properties significantly

change near the windward corners of building roof, because of the high vorticity behind

them. The distances between the first grid line to the roof surface and to the windward

wall are very important parameters for the numerical evaluation of roof pressures, in

order to achieve good prediction (Stathopoulos and Zhou, 1993). Very small distances

49
from the first grid line to the roof, upstream wall and parapet have been chosen (0.0007

m). For down stream and side walls of the building and the parapet 0.002 m is set, since

surface pressures are not sensitive to the mesh at these locations the (Stathopoulos and

Baskaran, 1990-b; Stathopoulos and Zhou, 1993). Figure 3.3 presents the current grid

arrangement for the computational domain.

3.4.3 Boundary conditions

Adequate boundary conditions for turbulent flow should be applied in order to obtain

good estimation of wind pressures on buildings. 'No-slip walV boundary conditions are

set for ground surface (i.e. bottom of the computational domain). The standard wall

function approach (Launder and Spalding, 1974) with sand-grain-based roughness

modification (Cebeci and Bradshaw, 1977) is used to model the near-wall region for

bottom boundary. In Fluent 6.1 the equivalent sand-grain roughness height, ks, can be

obtained as follows (Blocken et al., 2007):

where z„ = 0.0016 m is the aerodynamic roughness length, obtained from wind tunnel

data, and Cs is the roughness constant. Considering the default Cs = 0.5, ks should be

0.031 m. However, Fluent Inc. (2005) warns the users to obey the requirement zp > ks,

where zp is the distance from the bottom boundary to the center point of the wall adjacent

cell. The value of zp = 0.002 m is selected since accurate solutions for near-ground flow

can only be obtained with high grid resolution near the bottom of the domain. Therefore,

ks will be limited to 0.002 m at the bottom of the computational domain. In this case, Cs

50
should be 7.83. However, Fluent Inc. (2005) recommends considering a value between

0.5 and 1.0 for Cs\ therefore the default Cs = 0.5 is used. It should be noted that if the

upstream length in the computational domain is about the same as that between the start

of the turntable and the building model in the wind tunnel, which is the case of this study,

then taking a lower Cs value, in comparison with that satisfying equation 3.11, is allowed.

This can be related to the very low roughness of the turntable, where no roughness

elements are present during the tests.

Note that, using incorrect roughness height can lead to stream-wise changes (horizontal

inhomogeneity) in the vertical profiles of wind speed and turbulent parameters as the

flow travels through the computational domain. Blocken et al. (2007) explain the

inconsistency in the requirement for the Atmospheric Boundary Layer (ABL) flow

simulation and provide different solutions such as using a small, as possible, upstream

distance (which is limited to 0.36 m in this study).

The value of zp can be obtained from the following equation:

z +
U* ZD 3 1 2

where: Z + is normalized wall distance, u* is friction velocity, vis the kinematic viscosity

(v=ju/p) and p is the center point of the wall-adjacent cell. The logarithmic law for

mean velocity is known to be valid for 30 < Z + < 60. In Fluent, the log law is employed

when Z + >11.225. When the mesh is such that Z + <11.225 at the wall adjacent cells,

Fluent applies the laminar stress-strain relationship. The velocity profile of the turbulent

flow in the recirculation region behind a backward facing step was measured by Adams

51
and Johnston (1998). Zhou (1995) compared the velocity profile measured by Adams

with the log-law velocity profile and found that the measured velocity distribution

deviates significantly from the logarithmic part of the wall function when: Z + > 7.

The inlet profiles of mean velocity (U) and turbulence parameters (k and s) were

interpreted in Fluent as User Defined Functions (UDF) and the wind tunnel measured

data was used. The ''Inlet-velocity'' boundary condition was taken at the inlet of the

computational domain. The mean wind speed profile was modeled by the power low

with exponent a = 0.29 and maximum Ug = 12 m/s at gradient height zg = 0.6 m;

turbulence quantities are given by Richards and Hoxey (1993) as follows:

{ \a
U_ z
3.13
u„ KZtJ

u*2
k = - = 3.14

£= 3.15

KZ

where u* is the friction velocity (0.56 m/s) for an aerodynamic roughness length z0 =

0.0016 m, K is von Karman constant (0.42). 'No-slip wall' boundary conditions are

assumed for building and parapet surfaces with Cs = 0.5 and ks = 0. 'Symmetry' boundary

conditions (i.e. with zero shear slip) are applied to the sides of the computational domain;

while ' o u t f l o w ' boundary condition, assumes no stream-wise gradients at this location, is

set to the outlet of the computational domain. To avoid additional stream-wise gradients,

at the top of the domain, the values of the mean velocity and turbulent parameters are

52
fixed by using the inlet profiles values at this height. These values are: U= 7.95 m/s, k =

1.044 m 2 /s 2 and £ = 0 . 1 0 4 m 2 /s 3 (Blocken et al. 2007).

A 3-D simulation is made in an empty domain, with grid distribution similar to the

domain with the building model, to evaluate the extent of stream-wise changes.

Comparisons between the inlet (x = 0) and incident (building position, x = 0.36) profiles

of mean velocity and turbulent quantities is presented in Figure 3.4. The first 0.2 m from

ground surface is considered. Figure 3-c shows the vertical profile of k obtained from the

wind tunnel testing. The numerical profile agrees with the experimental one. In general,

minor differences occurring between the two positions for all parameters; which means

that the stream-wise changes are successfully limited.

3.5 Sensitivity Study of the Numerical Solution

Any numerical solution is considered to be converged when the difference between the

final solution and the successive approximation tends to approach zero as the number of

iterations increased. Sensitivity is defined from the behavior of the system output for

variations in the system input. Sensitivity is one of the factors that characterize the

performance of the iterative procedure for a steady-state condition. This issue is studied

by considering two main factors: extent of computational domain and utilizing different

number of nodes. Figure 3.5 shows the computational domain cluster for the considered

study in plan (x-y) and sectional (x-z) views. An Upstream Distance (USD) from the

windward wall and a Downstream Distance (DSD) from the leeward wall define the

53
boundaries of the computational domain along x-direction. Distances DS and DT are

used for y- and z-directions respectively. Four domain sets are considered in the present

analysis. The extent of the computational domain along x-, y- and z-direction has been

chosen based on the study of Paterson, 1986, and Murakami and Mochida, 1988. Table

3.1 lists the specifications of all domain sets. Note that, in this analysis the total number

of nodes (195,000) is kept constant. Considering D1 as the base, the effects on computed

results by increasing or decreasing the domain size are analyzed.

The induced pressure values are evaluated for the variation in the domain distance as

shown in Figure 3.6. The induced suctions along the center line of the roof are presented.

Increasing the size of the domain by keeping the number of grid points constant,

increases the grid spacing. When comparing the observations of the figure including the

experimental results, D1 domain specifications appears as a good choice. This domain

was used in the present simulation. The analysis shows that an increase in the size of the

domain without increasing the number of nodes may reduce the accuracy of the

computed results.

54
Table 3.1: Specification of computational domain sets used in the current analysis

x-direction y-direction z-direction

SET USD DSD DS DT

~D1 6H 30H 8H 8H

D2 13H 26H 10H 8H

D3 17H 34H 13H 12H

55
Figure 3.1: Flow chart of numerical computation method

56
Figure 3.2: Building configurations and computational domain
(not to scale)

57
 (b)

3.3: Grid
atr
»»&<neatiQ
the
m C mput
°dcl anri.
and
° »tional
(b)f0t
ung

S8
 0.2
x = 0.0
1
»
3C = 0.36m
1
11
0 15 tt
t1
11
I\
\

1t
I
0.1 »1
•a i%
•s
x w
\\
u
0.05 \1
\t
\\

x
-J'S-tfr.i im/m . . .

0.1 0.2 0.3 0.4 0.5 0.6 0 2 4 6 8

Turbulence Intensify, c„ u Mean velocity, u (m/s)

(a) (b)

0.2
\i x = 0.0
\t x = 0.36
\ o WT(x = 0.36)
i
0.15 i
O',
1»
%
»
1
^1 \
M 0.1
i
O \1
I. O 1
'5
X O 1
<1J
0.05 A

f1 O
/
0O
J* O
/

Turbulentkiuetic energy, k (in2 s2) Turbulent dissipation rate, e (m /s )

(C) (d)

Figure 3.4: Inlet and incident vertical profiles of (a) Turbulence intensity, (b) Mean
velocity, (c) Turbulent kinetic energy, (d) Turbulent dissipation rate

59
 USD DSD
-H-

Figure 3.5: Computational domain used in sensitivity analysis

60
 x/L
wind

/», = 0.01 m

SET-D1

SET-D2

SET-D3

• Exprimental, WT

Figure 3.6: Experimental and computational pressure coefficients with different

computational domains

61
 Chapter Four

EXPERIMENTAL WORK

This chapter presents the details of the full-scale experimental building and field

measurements. In addition, a complete wind tunnel simulation of wind flow over a

model of that building is presented

4.1 Full-Scale Testing

Full-scale testing has been completed to examine surface parapet pressures. The test site

should be subjected to adequately strong winds for most part of the year to facilitate

having maximum number of observations and good response of the instruments. The city

of Montreal experiences strong winds in the winter season. However, because of severe

winter conditions, it is difficult to carry out field experiments between December and

April. The measurements of this work had been carried out between May and November

of three consecutive years.

4.1.1 Experimental building

The ideal type of terrain for testing low-rise buildings is the open country i.e. with no tall

trees, buildings and other obstructions in the surrounding area. A small low-rise building

62
with a perimeter parapet was used to measure wind-induced pressures on parapet

surfaces. The building is located at the corner of the soccer field of the Loyola Campus

(about 8 km west from downtown Montreal) of Concordia University and housed the

instruments required for the experimentation. The exposure of the test site can be

classified into two categories, open country and suburban, depending on the wind

direction. Two sides of the building are exposed to a suburban type of terrain, with small

buildings and high trees, for winds from North-East and South-East. The other two sides

are exposed to open country type of terrain for winds from North-West and South-West.

The basic direction for the strongest winds in the Montreal area is South-West, thus the

basic exposure of the experimental building is considered as open country terrain.

Figure 4.1 shows the upstream of the building, facing South-West. The building is a

made of brick and is 3.10 m high, 4.00 m long and 3.25 m wide. The roof is larger than

the base, 4.45 m long and 3.70 m wide (Figure 4.2). More details about this experimental

station can be found in Stathopoulos et al. (1999). A perimeter parapet with a height of

0.5 m is attached to the roof. The parapet is made out of wood, 11 mm (7/16") aspenite

board, separated by a 13 mm stud. The parapet section is 0.16 m thick and is fixed to the

roof by plate connection fastened with screws to hold them together. The interior,

exterior and top parapet surfaces are drilled for tapping. Each pressure tap had 3.2 mm

inside diameter and is connected by a steel pipe to a plastic tube. Figure 4.3 shows the

locations of all tapping and the definition of wind angles. The subscripts e, i and t refer

to the exterior, interior and top surface tapping respectively, while the subscript r refers to

the roof tapping.

63
4.1.2 Measurement of wind speed and direction

For measuring wind speed, a three-cup anemometer, from Teledyne Geotech, is used.

This anemometer provides an output voltage proportional to wind velocity. A wind vane

is attached to the same tower with the anemometer to indicate the azimuth of the incident

winds (See Figure 4.2). In the current study, the acquisition of accurate records of wind

speed and direction is very important. The anemometer has to be placed in a spot near

the location of measurement that is not affected by the wind flow around the building.

Therefore, it is mounted on a metal tower on the top of the building. According to the

ASHRAE (1999) recommendations the anemometer should be placed at a height of \ .5R

above the building, in which R is scaling length and is calculated as follows:

R = B ™ B ™ 4 1

where: Bs = smaller of H and B, Bi = larger of H and B. At this height the building does

not influence the flow (Figure 4.4). Considering Bs = 3.6 m (H + hp) and Bi = 4.5 then,

1.5R = 5.8 m and the minimum height for placing the three cups is equal to 9.4 m from

the ground surface. The height of the tower is 6.7 m and the height of the anemometer

from the base of the building is about 10 m. The output of the anemometer is given in

volts and is directly proportional to the wind speed. The anemometer was calibrated in

the Boundary Layer Wind Tunnel (BLWT) of Concordia University. Calibration details

are presented at Appendix A. Wind vane is attached on the same tower with the

anemometer for the purpose of azimuth measurements for the incident winds. The wind

speed and direction are indicated on a compact box. For calibration purposes, the wind

vane is pointed to a particular direction, e.g. north, and the output in volts is measured.

64
The present calibration was checked repeatedly at regular intervals during the

experimentation, to confirm its steadiness.

4.1.3 Upstream wind flow parameters

According to the NBCC (2005), the exposure of the test site can be classified as category

C (open country terrain). The standard recommends roughness length z0 = 0.02 m and

power low exponent a = 0.16 for such terrain. Other standards recommended even

higher values; for example, ESDU (1983) considers z0 = 0.03 m for open country terrain.

Although, field observations of the complete mean wind profile were not available, the

upwind flow parameters could be estimated based on the recorded turbulence intensity in

the site. The current field observations showed that the average turbulence intensity at

the anemometer height was 22 %, during the entire period of test records. Consequently,

the average turbulence intensity at the roof height was 28 % and the average roughness

length was 0.11 m. These values were estimated according to the charts provided by

Tieleman, 2003, who proposed a method derived from basic flow relations to estimate the

upwind roughness parameters. Considering a = 0.16, the velocity at the roof height was

found about 17 % lower than that measured at the anemometer height. Another

commonly used measure of the variability of the wind is the gust factor, Gu, which had an

average value of 1.64 during the measurement period. A complete description for

calculating the wind flow parameters is presented in Appendix A.

65
As already mentioned, due to severe winter conditions in Montreal area, all field

measurements were carried out during the summer and fall seasons, when the strongest

winds blow from West and South-West. Therefore, limited data were recorded for wind

azimuth 90° - 1 8 0 ° . The randomness of the natural wind leads to a significant variability

in wind direction, thus higher variation of wind direction occurs on the site comparing

with the wind tunnel conditions. The US Nuclear Regulatory Commission (USNRC)

recommended 'the standard deviation of the wind direction' (erg) method to simply

classify the atmospheric stability (Mohan and Siddiqui, 1998). Stability is an expression

applied qualitatively to the property of the atmosphere, which governs the acceleration of

the vertical motion of air (Snaebjornsson, 2002). Table 4.1 presents the classification of

atmospheric stability as used by Sedefian and Bennet (1980). In the present study, data

was only considered when oe < 7.5°, which classified the atmosphere as ''moderately

stable'' (Group E). In order to obtain reliable response of the instruments, data was

collected at windy days and was only considered when the wind velocity is higher than

5.6 m/s (20 km/hr) at the anemometer height. Table 4.2 summarizes the field testing

parameters of upstream wind flow during the entire period of the experimentation.

4.1.4 Reference static pressure

In full-scale testing special consideration is required to obtain adequate reference

pressure. Atmospheric static pressure is used as reference pressure, as it provides the

most useful data for experiments on building aerodynamics (Levitan, 1993). Placement

of the static pressure point is critical; it should be placed at a sufficient distance away

66
from the building so that it is not affected by the static pressure field of the building. The

reference pressure has to be steady in order to obtain reliable field pressure data (Levitan

and Mehta, 1992). At the initial stages of measurements, the reference pressure was

measured at various locations around the building to ensure stability and negligible

effects on the measured pressures. Reference pressure was measured at the ground level

at a distance of 22 m west from the test building. The pressure at this location was

transmitted to the test building using PVC flexible tube of 4 mm inner and 11.2 mm outer

diameter. A common way of checking the reference pressure fluctuations is to connect

the reference pressure to both ends of a transducer and measure the difference to ensure

that reference pressure value is not fluctuating.

4.1.5 Pressure measurements

Pressures were measured with 10 differential pressure transducers, 160 PC,1 psi (6895

Pa) range, manufactured by Micro-Switch (a Honeywell Division) operated from a single

positive supply voltage ranging from 6 to 12 volts DC. The instruments were calibrated,

periodically, in the wind tunnel. For collecting mean wind pressures, the transducers

were connected to pressure tapping with 4.5 m long PVC flexible. All tubes were

collected in a 0.1 m diameter hole through the building roof and thus, the tubes were

easily connected to the transducers. For measuring instantaneous peak pressures, four

transducers were mounted inside the parapet section next to the tap being monitored, and

could easily be moved from one tap to another (see Figure 4.3). Each tapping point was

connected to a transducer by a 0.15 m long PVC flexible tube with 3.2 mm internal

67
diameter. The length and diameter of the tubes control the frequency response of the

pressure system (Holmes, 1984). The tube diameter should not be less than 3 mm to

avoid condensation problems (Levitan, 1993). The length of the tube should be as short

as possible to maximize the frequency resolution of the system (Gumley, 1983).

Therefore, 0.15 m was decided to be a reasonable compromise between placing the

pressure taps on all parapet surfaces, while obtaining satisfactory frequency resolution.

In this case, only one side (side B) of the perimeter parapet was examined, due to

technical restrictions. When no measurements are taken, all pressure taps have been

sealed to avoid concentration of moisture inside the tubes. Using a compressor, air was

blown inside each tube before starting measurements to ensure that all tubes are

completely dry. A Hydra Data Logger unit (Model 2635-A) from Fluke and Philips was

used for data acquisition during the experimentation. The instrument is a 21-channel data

acquisition unit with maximum sampling rate of 1 Hz. Data is subsequently transferred

to a portable PC unit. Two channels are connected to the anemometer and the vane for

collecting wind speed and direction through a data acquisition unit. The other channels

are connected to the pressure transducers for collecting pressure difference records.

Although it is recognized that the sampling frequency of 1 Hz may be on the low side,

this was the only instrumentation available at the time of field measurements.

Wind-induced pressure acting on the building surfaces is a function of wind speed and

direction, air density (/?), barometric pressure, building geometry and surrounding terrain.

While wind speed and air density can vary significantly between experiments, it is not

possible to compare wind-induced pressure directly from one experiment to another.

68
Therefore, a dimensionless form of pressure measurement, pressure coefficient (Simiu

and Scanlan, 1986) is commonly used:

C 4.2
x
' ApU2

The denominator in this equation is equal to the mean dynamic pressure. In the current

study pressure coefficients were referenced to the mean dynamic pressure at the building

height (roof level).

4.1.6 Verification of the collected data

The first step in full-scale experimental analysis is verification of the collected data. The

wind speed and direction, measured by the anemometer, had been compared to the data

provided by the Meteorological Service of Canada (MSC), which is available on the web

at www.ec.ge.ca. The wind parameters are observed at Trudeau Airport (10 m above the

ground) where the exposure is considered as flat open terrain. Figure 4.5 shows a sample

of the data collected in two different days in order to check the correctness of readings for

both exposures considered. The difference in wind speed could be attributed to the

difference of geographical features for each site. Power law is used for converting the

data obtained from MSC to match the results measured at the experimental site (Loyola)

in terms of height and roughness:

f \
>
U
z, z (Zg GI
'
ZGr u y
^ /

69
where: Uz is the velocity measured at a height Z, ZQ is the gradient height and a is power

law exponent. Note that, the subscript 1 refers at data of Dorval location, while the

subscript 2 refers at current data of Loyola site.

Verification of the field measurements required a visual inspection of time history plots.

Checks are performed to ensure that the time series data have been recorded properly.

For each record time history plot is carefully examined to see if there are any immediate

problems. The time histories are also compared for similar wind speeds and directions to

see if the observations are consistent. Figure 4.6 shows the wind speed and wind

direction for a 10-minute sampling period, while Figure 4.7 shows the time history for the

exterior and interior surface parapet pressure at a corner tap (Bl) for the same record.

Visual inspections of wind stationary were also completed for all records.

4.2 Wind Tunnel Modeling

The experimental building has been modeled to be tested in the BLWT of Concordia

University. The wind tunnel has a 12.2 m long working section 1.8 m wide with variable

height (1.6 m high at the measurements location). All information about the design and

fabrication of this wind tunnel is given by Stathopoulos (1984). The boundary layer

develops naturally over a rough floor with the assistance of some passive devices.

Roughness blocks with different heights and arrangements were used. This implies that

different wind tunnel floor roughness develop conditions representative of different

terrain exposures. The heights and arrangements of these blocks were adjusted to create

70
simulated wind velocity and turbulence profiles comparable with those of the ESDU

standards. The inlet screen (Figure 4.8-a) was installed behind 6 horizontal rods (metal

pipes) of different diameters and a mounted board with boxes was placed next to the inlet

screen, which produced the inlet setting. Figure 4.8-b shows an inside view of the wind

tunnel with the current roughness configuration.

4.2.1 Modeling the experimental building

The modeling in wind tunnel is extremely complex particularly when dealing with small

structural elements, such as parapets. The accuracy of the results and the capability of

comparison with full-scale data depend mainly on the correctness of the modeling of

building geometry and the simulation of the atmospheric boundary layer characteristics.

Extensive measurements were carried out by Stathopoulos, 1984, to evaluate the

simulation characteristics of the BLWT of Concordia University including velocity

profile, intensity, scale and spectrum of the longitudinal turbulence component. The

study indicated that a 1:400 geometric scale is most appropriate. However, the

experimental building is a small low-rise building; its eave height is 3.1 m, which would

be modeled to 8 mm in 1:400 length scale. This small model is not of sufficient size to

develop parapet details to study the aerodynamic forces. Moreover, it is too small to

allow adequate level of small-scale turbulence in the incident flow at roof level and leads

to relatively low Reynolds number (Tieleman, 2003).

71
To overcome the problems of modeling small details, a geometric model scale of 1:50 is

utilized in the present study. The choice of modeling scale was based on the importance

of correctly modeling the turbulence intensity at the roof height, which was indeed

similar with that measured in the field. The model is made out of plexi-glass with two

sets of perimeter parapets: 0.01 m and 0.02 m, of an equivalent full-scale height of 0.5 m

and 1 m respectively, attached to the roof (see Figure 4.9). Pressure taps were drilled on

the exterior and the interior as well as the top surface of the parapet. The parapet

thickness is 3 mm and the tapping has an inside diameter of 1mm. Figure 4.10 illustrates

the pressure tapping layout of the model. The definition of wind approach angle is also

shown in the figure. The number of pressure taps located on the parapet itself varied with

parapet height. A total of 48 taps were located on the 0.5 m parapet and 96 taps on the 1

m parapet. In addition, 42 pressure taps were placed on the roof surface.

4.2.2 Boundary layer simulation

In order to correctly model the wind flow, both velocity and turbulence characteristics of

the natural wind should be accurately simulated in the wind tunnel. Wang et al, 1996,

suggested that correct modeling of roughness length leads to correct pressure coefficients

even with a mismatch of the scale ratio. However, the wind engineering vision now

moves toward considering correct simulation of the turbulence characteristics in testing

low-rise buildings with larger length scale, while the mean flow profile parameters may

not require exact simulation (Tieleman, 1998). Increasing the upstream ground

roughness acts to retard the wind flow close to the ground and hence reduces mean wind

72
speed at the roof level. On the other hand, higher turbulence will be obtained. In this

case, it will be possible to examine larger building models. However, this is restricted by

the wind tunnel blockage, which should be less than 5% (Laneville, 1990). In order to

examine low-rise buildings with larger length scale, researchers in the Boundary Layer

Wind Tunnel II of the UWO used additional turbulence generating devices, including

spires and a tall barrier, which were added across the wind tunnel at the entrance (Ho et

al., 2005 and St. Pierre et al., 2005). These devices were used to create turbulence

characteristics consistent with larger length scales, 1:100 and 1:50.

Following the same concept, the current study applies a different approach to attain

correct matching of the turbulence characteristics as well as roughness length with those

in the field. In this case a length scale of 1/50 was used and the wind tunnel blockage

was 0.2%. To achieve higher turbulence intensity at roof height, which is the most

important to be matched since the pressure coefficients are normalized by this height (Ho,

1992); turbulent shear flows are generated over different upstream roughness

configurations. For each arrangement, turbulence intensity and mean velocity profiles

are obtained. Essentially, the required roughness blocks are higher than in full-scale, but

they are placed 1.2 m upstream of the center of the turntable where the model is located

see Figure 4.8-b). Table 4.3 shows the experimental parameters measured in the wind

tunnel using different arrangements of the upstream roughness elements. By comparing

each parameter with the corresponding full-scale value, it was found that Terrain-4 is the

most appropriate. Note that, the selected simulated exposure shows reasonable match of

73
Jensen number (J e = H / z0), which is one of the important dimensionless parameters in

similarity requirements.

Figure 4.11-a depicts the turbulence intensity profile for the simulated exposure

compared with the ESDU (1983) profile, which is described in Appendix B. The

turbulence intensity measured at the test site is also shown at the anemometer and the

roof heights. Clearly, the measured profile agrees with the ESDU profile below 5 m

height, while a clear difference occurs above this height. Since the main goal of the

current wind tunnel simulation is to match the turbulence intensity at the roof height, this

discrepancy may be acceptable. The full-scale turbulence values agree well with ESDU

profile at both roof and anemometer heights, considering their variability resulting from

natural wind. Mean wind velocity profiles are shown in Figure 4.11 -b, where the

velocity was normalized by the reference velocity at the wind tunnel (12 m/s). The wind

tunnel profile agrees well with that of ESDU. By fitting power low profile, an exponent

a = 0.29 was found. The vertical profiles of mean wind speed and turbulence intensity

for all other tested exposures are presented in Appendix A.

Spectral density function has been determined at roof height using the velocity time

histories (Figure 4.11-c). The spectra are normalized by the square of velocity

fluctuation at roof height and the frequency is normalized by the mean velocity at roof

height. By fitting the Von Karman equation (Simiu and Scanlan, 1996) to the spectra, it

was found that the length scale of turbulence, Lx = 0.22 m at the roof height and 0.28 m at

10 m level (see Appendix A for more details).

74
Wind tunnel measurements are made at a gradient speed of 12 m/s. The velocity profile

is recorded at the centre of the turntable without the model in place. A general-purpose

program was developed (Wang, 2005) in a host PC computer that controlled the scan

frequency and duration of a hot-wire device via a multiplexer. This program reported

statistics data on-line and recorded data for off-line process. The scan frequency is set at

1000 Hz and the scan duration is set routinely at 32,768 seconds. The hot-wire probe is

attached to a traversing gear system, which could be moved up and down mechanically.

At each measurement height, the mean and standard deviation of the tunnel speed is

recorded, providing the mean velocity and turbulence intensity profiles. Roughness

length, z 0 = 0.08 m, was obtained by fitting the logarithmic law. Velocity time histories

were recorded for roof height (z = 0.062 m) and for 10 m elevation in full-scale or

anemometer height (z = 0.2 m).

4.2.3 Wind-induced pressure

The accuracy of the measured pressures depends on the achieved dynamic similarity.

Correct dynamic similarity requires matching of the Reynolds number in full-scale and

wind tunnel. However, this is not possible in conventional wind tunnels. The Reynolds

number is defined as:

where U is the oncoming velocity, D is the characteristic length and v is the kinematic

viscosity of the air. A previously held view considered that bluff-body aerodynamics is

Reynolds number insensitive for sharp edge elements (Simiu and Scanlan, 1996), where

75
points of flow separation are more or less fixed. However, recent investigations

demonstrate that relaxing the Reynolds number in wind engineering might cause minor

systematic errors (Larose and D'Auteuil, 2006). The Reynolds number for the current

full-scale building is 2.2 x 106 and for the wind tunnel model is 5 x 104 with a length

scale of 1/50 and 6.2 x 103 with a length scale of 1/400, where D is the building length

and U is the mean hourly wind speed at 10 m height. It can be noted that using larger

model act to minimize the difference in Reynolds number between full and model scale.

4.2.3.1 Pressure measurements

Pressure fluctuations were measured with a Scanivalve's Digital Service Module (DSM-

3000 system) that was connected to two Zero-Calibrate-Operate (ZOC-33) pressure

transducers, each of which was factory-designed to scan 64 pressure taps simultaneously.

The DSM system was controlled by a host computer. The installation and operation of

this measuring system, as well as the calculations of scanning time and frequency, are

presented in Appendix A. To correspond to acceptable wind tunnel testing practices on

the bases of the assumed time scale, the scanning frequency is 250 Hz and the scanning

duration is 24 sec (details are presented in Appendix A). The scanning frequency in the

wind tunnel is higher than the corresponding full-scale frequency. Therefore, an attempt

was made to compare the measured peak pressures with those obtained using 50 Hz

scanning frequency. The comparison reveals an average difference not more than 10%

for all cases.

76
Tubing system of 645.4 mm length leads from the surface tap to the scanivalve. This

system consists of 10 mm length steel tube of 0.8 mm internal diameter, drilled into the

model; connected to 365 mm long and 1.6 mm internal diameter flexible PVC tube; leads

to 25.4 mm long steel restrictor, of 0.16 mm internal diameter to keep the frequency

response flat; leads to 245 mm long tube connected to the scanivalve. The resulting

system responds to pressure fluctuations on the model up to about 100 Hz with negligible

attenuation or distortion (Stathopoulos and Saathoff, 1992).

Pressure time histories were recorded simultaneously at all tap locations and statistics

pressure coefficients (maximum, minimum, mean and rms) were determined. Pressure

coefficients (Cp) were referenced to the mean dynamic pressure (1/2pU2) at the building

height. Measurements were repeated for 13 wind angles, ranging from 0° to 180°.

Therefore, due to building symmetry, a complete data set was obtained.

4.2.3.2 Extreme value analysis

The peak point pressure and area-averaged load coefficients presented in this study are

not the absolute worst coefficients recorded within the sample time. Extreme values of

time series data are significant for selection of the pressure peaks. As an application of

the Gumbel distribution (Gumbel, 1958), extreme value ' T y p e V is used for peak

selection (also applied by Suresh Kumar, 1998; Holmes and Moriarty, 1999). In this

case, the total time history of 6000 data-point is broken down into ten segments each of

600 data-point then ten segmental peaks are selected and fitted by Gumbel Plot (see

77
Appendix A for more details) with a cumulative probability of 0.975 (Uematsu and

Isyumov 1998). The resulting mode (m) and dispersion (5) are used to determine the

expected peak value for each pressure coefficient. Such peaks are believed to be more

statistically stable quantities in comparison to the absolute worst recorded peaks. A

sample for one case of Tap Ble and for wind direction 0 = 0° is presented in Figure 4.12

showing the time history for the first segment, i.e. 600 data-point, (Figure 4.12-a) and the

Gumbel plot for this segment (Figure 4.12-b).

78
 Table 4.1: Stability classification (Sedefian and Bennet, 1980)

Stability classes Group oo 0

Highly unstable or convective A 22.5 < ere

Moderately unstable B 17.5 < c r 0 < 22.5

Slightly unstable C 12.5 < <j0 < 17.5

Neutral D 7.5 < c j e < 12.5

Moderately stable E 3.75 < <jg < 7.5

Extremely stable F 2.0 <C70< 3.75

Low wind night time stable G <70 < 2.0

79
 Table 4.2: Full-scale testing parameters

Parameters Data range (average)

Upstream exposure (NBCC, 2005) C, open country

Eave height, H (m) 3.1

Parapet height, hp (m) 0.5

Anemometer height, za (m) 10

Roof dimensions, LxB (m) 4.45 x 3.7

Power exponent parameter, a 0.16

Roughness length, z0 (m) 0.035 - 0.21 (0.11)

Mean wind velocity at za, Ua (m/s) 5.6 - 11.4 (9.5)

Mean wind velocity at H, Uroof (m/s) 4.6 - 9.5 (8.1)

Turbulence intensity at za, Ia (%) 18 • 29 (22)

Turbulence intensity at H, Iroof (%) 23 36 (28)

Length scale of turbulence at za, LXiia (m) 36 • 83 (72)

Gust factor, Gu 1.55 - 1.95(1.64)

Mean wind direction, 9 (°) 5 177

Standard deviation of wind direction, creC) 3.6 - 7.5 (6.3)

80
 Table 4.3: Wind tunnel modeling parameters for different simulated exposures

Simulated wind tunnel exposures'" (scale 1:50)

Parameter Full-scale
Terrain-1 Terrain-2 Terrain-3 Terrain-4 Terrain-5

a 0.16 e 0.15 0.24 0.28 0.29 0.34

za (m) 10 0.2 0.2 0.2 0.2 0.2

z0 (m) 0 . 0 3 5 - 0.21 e 0.00/ 0.04 7 0.01f 0.08 7 1.05 7

u* (m/s) 0.50 0.63 0.7 0.73 0.8

HI z0 14.76 - 88.5e 620 77 44 39 3

Ua (m/s) 5.6 -11.4m 11 10.6 10.4 10.2 10

Uroof (m/s) 4.6 - 9.5e 9 7 6.7 6.3 6.1

Ia (%) 18 - 29 e 10 12 15 19 20

I roof (%) 23 - 36e 14 28 25 28 30

Lx, a (m) 36 - 83e 0.40 0.35 0.33 0.28 0.26

Lx,roof (m) 0.32 0.28 0.25 0.22 0.2

/ full-scale equivalent
e estimated (Simiu and Scanlan 1996)
m measured

81
 Table 4.4: Wind tunnel modeling parameters used in the current study

Parameters Measured values

Model scale 1-50

Parapet height, hp (m) 0. 01, 0. 02

Reference height, z r e /(m) 0.2

Power exponent parameter, a 0.29

Roughness length, z 0 (m) 0.08^

Gradient wind speed, UG (m/s) 12

Wind speed at roof height, Uroof (m/s) 6.3

Turbulence intensity at zref, Iref (%) 19

Turbulence intensity at roof height Iroof (%) 28

Length scale of turbulence at zref, LXt ref (m) 0.28

Length scale of turbulence at roof height, LXi roof (m) 0.22

/ : full-scale equivalent

82
Figure 4.1: Basic exposure of the experimental station (South-West)

83
 (Dimensions in cm.)

•MBBBMMWBMBI

Anemometer and wind vane

Figure 4.2: Details of the test building.

84
 B5e B4e B3e B2e Ble
o St
B5i
o o

B4i
o

B3i
o

B2i
o

Bli 10°
90°
B5t B4t B3t B2t Bit
O Alt Ali O Ale
B5r B4r B3r B2r Blr

O A2t A2i o A2e

0 A3t A3i o A3e

N
O A4t A4i o A4e

o A5t A5i o A5e

18^1 102 ,, 102 |, 102 ,, 102 ,, ,, 18

I 445 L

Tap locations

4
sr<iL
Pressure transducers Parapet section

(Dimensions in cm)

Figure 4.3: Details of parapet section and tap locations.

85
 UPWIND VORT EX !
L zoN E OF
(RCCIRCULATING FlOW. L,

Figure 4.4: Flow patterns around a rectangular building (ASHRAE, 1999)

86
 - e — Loyola (16 Jul, 2 0 0 4 )

— Dorval (16 Jul, 2 0 0 4 )

-i 1 1 1 1 1 r
10 12 14 16 18 20 22 24
Time (lirs)

(a) Wind speed

360

300

® T
- B — Loyola (16Jul, 2 0 0 4 )

2 ISO - * — Dorval (16Jul, 2004)

Loyola (18Jul, 2004)

cs
c - • — Dorval (18Jul, 2 0 0 4 )

(b) Wind direction

Figure 4.5: Mean wind speed and direction at the anemometer height,

measured during 24 hours

87
 100 200 300 400 500 600
Time (sec)

(a) Wind speed

120

100

80

S
© 60 14Nov., 2005
'•46>
mean = 43°
'•= 40 - mis = 6.9°

2 0 -----

100 200 300 400 500 600

Time (sec)

(b) Wind direction

Figure 4.6: Typical time histories for wind speed and wind direction measured at

the anemometer height

88
 B l e ( 1 4 N o v . , 2005)

m e a n = 0.53
n n s = 0.28

100 200 300 400 500 600

T i m e (sec)

(a) Exterior parapet surface pressure coefficient

Bli (14Nov..2005)

mean = 0.68
mis =0.21

100 200 300 400 500 600

T i m e (sec)

(b) Interior parapet surface pressure coefficient

Figure 4.7: Typical time histories for parapet surface pressure coefficients

89
 (a) Wind tunnel inlet

(b) Inside view of the tunnel

Figure 4.8: The wind tunnel in simulated terrain

90
1 m high equivalent parapet 2 m high equivalent parapet

(Dimensions in cm)

Figure 4.9: Building models with different parapet heights

91
 I
0°

90°

o
Tal 351 31 3H

, t).3

351 3SI ZaU

0.6 j k 1.28 j, 1.28 ^ 1.28 j 1.28 j 0.28 j 0.28 j , 0.6

8.9

(Dimensions in cm)

Figure 4.10: Details of pressure tap layout of the building model

92
 40
» W i n d tunnel

35 ESDU, zo = 0.08 m
a = 0.29

30

B25
a
•a
« 20 -I
>
u
W 15 -
anemometer ht.

10

V
r o o f ht.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.0 0.2 0.4 0.6 0.8 1.0
<Tu/U U/Uref.

(a) Turbulence intensity (b) Mean wind velocity

ILL1.'

Wind-tunnel H 3 1m
Von-Karman, Lx=0.22 z0 = 0 . 0 8 m

0.01 0.1 1 10 100

(c) Wind spectrum at the roof height

Figure 4.11: Profiles of turbulence intensity and mean velocity, and the spectrum at

the roof height for Terrain-4

93
 Wind
y 0.18

iS^N
r*-
+ "Ls 1

Tap BI e

0 = 0°

(a) Time history (600 data points)

t - y Wind
j / 0.18«" Ble
3

2.5

2 4-

1.5

! •

' 0 Tap Ble

-0.5
0=0°

0 0.5 1 1.5 2.5

Cp-min

(b) Gumbel plot

Figure 4.12: Extreme value behavior for a segment of the Cp time history of Tap Bli

for perpendicular wind direction.

94
 Chapter Five

EXPERIMENTAL RESULTS AND COMPARISONS

This chapter includes an analysis for all experimental results obtained from the wind

tunnel (WT) and the full-scale (FS). Note that, all pressure data are presented in the

form of pressure coefficients, normalized by using the dynamic pressure at roof height.

Local surface pressures are measured on both exterior and interior parapet faces as well

as on the roof edge surface. Area-average loading is recorded for different tributary

effective areas. The wind tunnel results have been systematically compared with the full-

scale collected data throughout the analysis of parapet surface pressure coefficients. Full-

scale records for both positive and negative instantaneous peak pressures are displayed in

a form of data ranges (error bars) and their averages. These divergences are due to

unsteadiness of the direction of natural wind and the repeatability of the collected data.

Note that, the lines connecting the data points for different wind angles have only been

added for clarify purposes.

5.1 Local Loading Coefficients

The design of a parapet may require three feasible load cases: wind loads on the exterior

parapet surface, the interior parapet surface and the net load on the parapet. The local

fluctuating wind pressures on different parapet surfaces may be either towards the surface

95
(positive pressures) or away from the surface (negative pressures or suctions). The net

load on the parapet is defined as positive when it acts to push the parapet toward the roof

surface and negative for the reverse.

Examining the individual surface pressures, Figure 5.1 presents instantaneous maximum

and minimum pressure coefficients, as well as mean values, on both exterior and interior

surfaces. Data are presented for exterior and interior taps, Ble and 57, respectively,

which are located at the mid-height of the parapet and at a distance of 0.18 m from the

leading corner.

Figure 5.1-a shows the effect of wind direction on local pressure coefficients on the

exterior surface (Cp e ). For both wind tunnel and full-scale results, the maximum positive

peak pressure occurred for winds approaching perpendicular to the parapet, while the

minimum negative peak pressure occurred for winds approaching parallel to the parapet.

The maximum full-scale positive peak coefficient of 2.4 occurred near 8 = 15°, which is

about 10 % higher than the wind tunnel pressure coefficient at the same wind angle. On

the other hand, the maximum full-scale negative peak coefficient o f - 2 . 7 9 occurred near

8 = 90° in the separation zone down stream of the leading corner of the building. This

value is about 15 % higher than the wind tunnel value.

Figure 5.1 -b shows the effect of wind direction on local pressure coefficients on the

interior parapet surface (Cp,). The maximum negative peak pressure coefficient was

found to occur at wind angles near 8 = 45° to the face of the parapet. A full-scale

96
pressure coefficient of - 3 is recorded near 6 = 45°, which is about 10 % more than the

wind tunnel value. The high suction recorded on the interior parapet surface was due to

the formation of sturdy vortices at the leading corner of the building, which is expected

for this cornering wind. The maximum positive peak pressure generally occurred for

wind perpendicular to the interior surface of the parapet (i.e. 0 = 180°). A full-scale

positive peak Cpt of 2 was recorded in near <9= 180°.

Pressures were also measured at mid-span location on the exterior and interior parapet

surfaces, tap B3e and B3t respectively. Figure 5.2-a shows that the maximum positive

peak pressure on the exterior surface occurred, also, for winds approaching perpendicular

to the parapet. A full-scale peak positive coefficient of 2.28 occurred near 0 = 0°, which

is about 5 % lower than the wind tunnel pressure coefficient. The maximum negative

peak pressure occurred for winds approaching parallel to the parapet. A peak negative

coefficient o f - 2 . 4 3 occurred near 0 = 90°. This value is approximately equal to the wind

tunnel result.

Figure 5.2-b shows that the maximum negative peak pressure coefficient on the interior

surface occurred for winds parallel to the face of the parapet. A full-scale value of -2.3

was recorded near 0 = 90°, which is matching the wind tunnel results. A maximum

positive peak pressure of 2.2 occurred near 0=135°, approximately 10 % higher than the

wind tunnel value. Generally, mean pressure coefficients measured in the wind tunnel

show very good agreement with the full-scale results except some discrepancies at wind

angle of 180°. Considering peak pressure coefficients, a difference of 10 - 15% is

97
observed particularly at critical wind directions. However, the overall data shows good

agreement between wind tunnel and full-scale.

As mentioned in Chapter two, the method used by some standards for estimating parapet

loads is to assume the inside local pressure on the parapet is equivalent to the local roof

pressure obtained at a tap next to the parapet edge. In the present field testing, roof

tapings are installed along side B at a distance of 0.025 m from the edge of the parapet

section. Mean and peak pressure coefficients on the roof (Cpr) and the inside surface of

the parapet (Cp,) are compared. Figure 5.3 shows the results at the corner, while Figure

5.4 shows the results at mid-span. For both locations, the mean pressures on the roof and

inside parapet tapping are almost equal for all wind azimuths, while some discrepancy is

found in both positive and negative peak pressures for specific wind directions. The

negative peak pressures at the corner measured on the roof are approximately 10 - 15 %

higher than those on the inside parapet surface. At mid-span location, negative peak

pressures on the roof were also 10 -15 % higher than those on the inside parapet surface.

Positive peak pressures are almost very close, for both locations, although a discrepancy

up to 50 % was found for wind angles of 135° and 180°, while the pressures on the inside

surface of the parapet are relatively high.

5.1.1 Correlation of pressures on the roof and the interior parapet surface

In the absence of design loads for parapets, a designer might use roof edge or corner

design loads to estimate the suction on the inside surface of the parapet. However, this

98
may significantly overestimate the peak suction since the roof design point loads are the

maximum loads in a corner or edge region and thus may not correspond to the load very

close to the edge. Another factor that may be important when using roof edge pressure to

evaluate the wind load on a parapet, is the correlation of the roof pressure fluctuations

with those on the inside surface of the parapet. The total fluctuating load on the parapet

is obtained from the pressure difference between the outside and inside faces of the

parapet, measured simultaneously. Perfect correlation between pressures on the inside

surface of the parapet and the roof edge pressures would be necessary to replace the one

with the other and keep equivalence.

Figure 5.5 shows cross-correlation coefficient, R, between pressures measured on the roof

edge and interior parapet surface at corner and mid-span locations, respectively for wind

directions of 0° - 180°. The present data (equivalent full-scale hp = 0.5 m; hpIH = 0 . 1 6 ;

x/L = 0.04) are compared with those from Stathopoulos et al., 2002-a (equivalent full-

scale hp = 1 m; hp/H= 0.13; x/L = 0.08), see table 5.1. The present data indicates that the

correlation is relatively high, approximately 0.9 for both locations, at 90°. However, the

correlation decreases to a minimum of 0.75 and 0.63 at 60° for corner mid-span locations,

respectively. It is likely that the correlation between roof and interior parapet surface

pressures depends somewhat on the location of the roof tapping. The vortices that

produce the significant pressure fluctuations travel downstream from the separation point

at some fraction of the free stream velocity. Thus, there will be a time lag between a

pressure drops near the edge compared to that occurring some distance away from the

edge. This will result in some loss of correlation between parapet and roof pressures.

99
Furthermore, for oblique wind angles (30°-60°) a time lag is expected between parapet

and roof instantaneous pressures due to lateral movement of the delta-wing vortices. This

time lag will reduce the correlation of the pressure signals. At the corner location, the

lateral movement is small and thus the correlation is relatively high. On the other hand,

at the mid-span location, the vortices have large lateral movement and the correlation is

relatively low. It should be noted that, similar performance was found by Stathopoulos et

al, 2002-a as shown in Figure 5.5. However, for oblique wind angle the correlation

reduces to 0.7 and 0.57 for corner and mid-span locations, respectively. As the

correlation values depend on the tap location, it is expected that higher correlation can be

obtained for roof tapping closer to the edge, i.e. parapet, which is the case of the present

testing. This may explain the difference in correlation between the two studies

5.1.2 Correlation of pressures on the exterior and interior parapet surface

The local fluctuating wind load on a parapet may be either toward the roof (exterior

parapet surface pressure) or away from the roof (interior parapet surface pressure). As a

first approximation, the peak local load on a parapet A Cp may be estimated by taking the

difference between peak pressure values on both surfaces. However, this requires that

these peak values occur simultaneously, i.e. very good correlation between the measured

pressure on the inside and outside parapet surfaces is assumed. Therefore, the correlation

coefficients of Cpe and Cp, for all wind directions, at five tap locations along the parapet,

are evaluated below.

100
Figure 5.6-a shows wind tunnel results while Figure 5.6-b shows full-scale data. The

overall correlation between Cpe and O , is relatively low. Maximum positive value of 0.5

was found for wind direction perpendicular to the parapet surface ( 6 = 0°) for all tap

locations; i.e. the absolute maximum peak pressures do not occur simultaneously on the

exterior and interior parapet surfaces. Therefore, net peak pressure coefficients may be

calculated only by taking the simultaneous difference between Cpe and C/>,. Generally,

although there are some differences in the individual values, the correlation coefficients

between exterior and interior pressures measured in wind tunnel shows good agreement

with the respective set of full-scale data.

5.1.3 Total local loading coefficients on parapets

The net peak pressure coefficients (ACp) on a parapet were estimated by taking the

difference between peak pressure values on the outside and inside surfaces of the parapet.

For positive and negative load respectively: A C p ~ (Cpe_max - Cpi_min) and (Cpe_min -

Cpi_max)• This approximate value of A Cp is expected to be larger than the actual value

since the maximum and minimum pressure coefficients do not occur simultaneously.

The most representative evaluation of A Cp is obtained by taking the simultaneous peak

pressure difference between the exterior and the interior surface of a parapet as follows:

A Cp — (Cpgjnax-inst.— CPi_min-inst.)simultaneous (5.1)

Figures 5.8-a and 5.8-b show the variation of both inward (positive) and outward

(negative) A Cp with wind direction at the corner and mid-span locations, respectively.

Two curves are shown for each peak pressures, one was obtained with the simultaneous

101
peak pressure difference and the other was obtained with the non-simultaneous peak

pressure difference calculated from the data shown in Figures 5.1 and 5.2. As expected,

data shows significant differences between simultaneous and non-simultaneous A Cp

values. Considering the simultaneous peak pressures at the corner (Figure 5.7-a), the

maximum positive (inward) pressure coefficient occurred near 6 = 45° while the

maximum nigative (outward) pressure coefficient occurred near 0 = 180°. Clearly, the

maximum negative and positive coefficients occur at the same azimuth as the coefficients

of interior parapet surface. This indicates that the pressure on the interior surface

dominates the net loading at this point of the parapet. However, the positive pressures on

the exterior surface make an important contribution, since the net windward coefficients

are 35 % higher than those on the interior surface. Figure 5.7-b shows that the maximum

positive peak coefficients (ACp, simultaneous) occurs near 6 = 30°, while the maximum

negative peak coefficient occurred near 0 = 135°.

5.1.4 Parapet pressures: full-scale versus wind tunnel

A comparison between the pressure coefficients measured in full-scale and the

corresponding data obtained from the wind tunnel is shown in Figure 5.8. Mean and

peak pressure coefficients for all taps and critical wind directions are plotted in the figure,

which represents 265 date points all together. The solid line represents the linear

regression line for Cp, FS and WT, (y = 1.012 x + 0.019), while the dotted line represents

the 45° line (y = x). The figure shows the agreement between full-scale and wind tunnel

results. Although, there is some dispersion near 45° and 135° wind directions, the overall

102
agreement seems to be excellent taking into consideration the unsteadiness of the

direction of natural wind.

5.1.5 Comparisons with other studies

Figure 5.9 presents instantaneous maximum and minimum peak pressures measured at a

single corner point. The current full-scale results were compared with three different

studies. Table 5.1 presents the main characteristics of different models used in each

study. Note that although the pressure coefficients in the study of Mans were referenced

to the mean dynamic wind speed at a height of H + hp, the data presented here were

referenced to the mean dynamic wind speed at roof height to match that in the present

study. Regarding tap location, for the present study x/L and £/hp were 0.04 and 0.5

respectively, where x is the distance from the leading corner and ^ i s the vertical distance

from the base of the parapet. The tap locations presented in Stathopoulos, 2001 and

2002, are x/L = 0.08 and tyhp = 0.4, while those presented in Mans et al, 2005, are x/L -

0.06 and &hp = 0.83.

Figure 5.9-a shows the coefficients on the exterior surface. A clear difference is shown

for Mans' results (particularly for wind azimuth near 45°); where positive peak

coefficients are lower than those of the other three studies. This difference may be

attributed to the difference in tap location (the distance from the base of the parapet). In

the experiments of Mans et al. (2005) the tap is located relatively close to the top of the

parapet {^/hp = 0.83), where the pressure on the exterior surface is expected to mitigate;

103
while in the current study and Stathopoulos et al. (2002-a) the tap is located at the mid

height of the parapet. Note that, a similar explanation was suggested in Mans et al, 2005.

In contrast, negative peak coefficients of Mans are higher than the other studies. Figure

5.9-b shows peak pressure coefficients for the interior surface of the parapet. The most

evident discrepancy between all studies is for the worst suction. It is clear that the

current results match the Stathopoulos et al. (2002-a) study, which has also equal

building length and height (L/H = 1) and relatively high parapet with respect to building

height {hp/H= 0.13). Moreover, Mans et al. (2005) and Stathopoulos et al. (2002-b, L/H

= 3) data show reasonable agreement for the most critical suction, although, higher

coefficients than those of the current study, by a factor of 2, were found. This difference

may be due to the fact that suctions on the interior surface of the parapet are dependent

on the minimum distance required for flow reattachment to occur, after separation at the

leading edge of the parapet. The building dimensions of Mans et al. (2005) study allow

for reattachment to occur on roof surface. However, flow reattachment may not occur on

the roof of a building with a reduced length/width ratio and high parapet, as is the case of

the present study. Consequently, most of wind flow passes above the parapet with small

circulation down stream (at the interior face). This will significantly reduce the suctions

on the interior surface of the parapet.

The comparison, also, suggests that worst suctions on the interior surface depend on the

ratio of building length to its height (L/H). Figure 5.9-c compares the net peak pressures

on the parapet measured simultaneously. The worst net pressure occurs at wind azimuth

45° for all cases; that indicates that suctions on the interior surface dominate the net

104
loading on the parapet, as mentioned previously. Very good agreement is shown between

the current study and Stathopoulos et al. (2002-a, L/H = 1). However, for building with

(L/H= 3) and the study of Mans et al. (2005) pressure coefficients are about 40 % and 20

%, respectively, higher than those of the current study. It is clear that Mans et al. (2005)

data shows that the interior suctions are controlling the net loading with almost no

contribution of the external pressure, which is not the case of Stathopoulos et al. (2002-b,

L/H = 3) and the present study, as already discussed in section 5.1.2. Although,

comparison results are generally good, there are some cases with some differences.

5.1.6 Comparison with NBCC (2005)

Currently, no provision exists in the National Building Code of Canada (NBCC) for the

design wind loads of parapets. In this case, it can be recommended that the parapet load

be determined using the design wind loads for the windward wall and for the roof. Table

5.2 shows the current coefficients, measured in full-scale and wind tunnel, and those

derived from the NBCC provisions using this method. The design force coefficient ( C »

for a particular region (corner or edge) of a parapet is given by:

(Cp)parapet ~~ (CpCg)waH - (CpCg)roof (5.2)

where CpCg is the design pressure coefficient. (CF) p a r apet for the present study is

determined by multiplying the measured values of (Cpejnax.inst. - CPi_min-in.,L)sitnultaneous by

0.8 (directionality factor) as it occurs with all CpCg values in the NBCC. The value of

(Cp)parapet for the corner region, based on the NBCC provisions, is significantly greater

than the measured force coefficient (about 45%), however; less difference (about 20 -

105
30%) is found between those values in mid-span location. The overestimation of

(Cp)parapei by NBCC method is due to the fact that the values of (CpCg)wau and (C p C^) wa ii

are both greater than the maximum CpCg on either side of the parapet. This can mostly be

as a result of the fluctuating pressures on the exterior and interior surfaces of the parapet

are uncorrelated, as explained in section 5.1.1. Therefore, it is not practical to use the

critical wall and roof values to estimate the total peak load. Based on a wind tunnel

testing of different building and parapet configurations, Stathopoulos et al. (2002-a)

shows similar trend.

5.2 Area-averaged Pressure Coefficients

Design of cladding and other structures and non-structured elements requires spatially-

averaged coefficients over a variety of areas. Area-averaged coefficients were derived by

simultaneously integrating the recorded pressures for a group of taps according to:

" I h-

where: n is the number of taps, Cpj is the pressure coefficient on the individual tap, lj is

the tributary length of the tap and hj is the tributary height of the tap. For each loaded

area coefficients were developed for all suitable tap combinations. Subsequently, the

largest maximum and minimum coefficients were obtained. Coefficients were calculated

separately for both exterior and interior parapet surfaces. The analysis was repeated for

the combined action from both surfaces, which was calculated instantaneously using the

same loaded area and location.

106
Figure 5.10 depicts the worst negative and positive peak coefficients for three load cases.

Full-scale data were compared with the wind tunnel results of two different parapet

heights. As expected, for all cases both positive and negative load coefficients tend to

decrease with increasing load area. Examining area loading on the exterior surface,

Figure 5.10-a shows that positive peak coefficients tend to increase with parapet height as

the 1 m parapet loads are about 10 % higher than those recorded on the 0.5 m parapet. In

contrast, negative peak loading coefficients on the 1 m parapet are about 10-15% less

than those on the 0.5 m parapet. The positive and negative full-scale area loads were

found to be approximately 18% less than the wind tunnel loads of the parapet with the

same height (0.5 m).

Figure 5.10-b compares peak coefficients on the interior parapet surface as function of

tributary (effective) area. For relatively small areas, less than 1 m , the positive

coefficients tend to increase with parapet height, while the negative coefficients tend to

somewhat decrease with increasing parapet height. This is consistent with the results

from the literature, which found that taller parapets reduce the largest suction pressure on

the building roof (Kind, 1988; Kareem and Lu, 1992; Stathopoulos and Baskaran, 1987;

Mans, 2003). A difference of about 10% was found. The positive and negative full-scale

area loading coefficients are in good agreement with the wind tunnel coefficients.

The combined action from both parapet surfaces is presented in Figure 5.10-c. The worst

peak positive loading coefficient occurs on the 1 m parapet, while the worst negative

107
peak occurs on the 0.5 m parapet, particularly for areas greater than 1 m 2 . The full-scale

positive peak coefficients match well the wind tunnel results for the same parapet height,

while the negative peaks are approximately 30% greater than those recorded in the wind

tunnel.

5.2.1 Comparisons with the ASCE 7-05

In the absence of research, the ASCE 7-05 recommends the following equation for the

design wind pressure on the component and cladding elements of parapets:

p = qP (GCPe ~ GCPi) 5.4

where: qp is the velocity pressure evaluated at the top of the parapet, GCpe is the external

pressure coefficient and GCpi is the internal pressure coefficient (see Appendix B for

details). Two load cases should be considered. In Load Case A (windward parapet), the

positive wall pressures are applied to the exterior surface while negative roof edge or

corner pressures are applied to the interior surface. In Load Case B (leeward parapet),

the positive wall pressure is applied to the interior parapet surface while the negative wall

pressure is applied to the exterior surface. The measured data are referenced to the 3-

sec. velocity pressure at 10 m height to be consistent with the ASCE 7-05 format

according to the following form:

{GCP\ 5.5

where (GCp)eq is the equivalent measured pressure coefficient, Cp is the peak coefficient

(maxima or minima) based on the mean wind pressure at the roof height (qa) and qiom,3s is

108
the basic wind pressure in the ASCE 7-05. As the code provides different factors for the

edge and interior regions, the comparisons are presented in two different figures. Figure

7.10 compares the ASCE 7-05 GCP for wall edge and roof corner regions with those on

the exterior and interior surfaces of the parapet edge, respectively.

In Figure 5.11-a the exterior surface loads on parapet edge are compared with the GCP

values on 'wall edge zone' obtained from ASCE 7-05 (Appendix B, Figure BI). The

negative 'wall edge zone' coefficient suggested by the standard covers the negative

surface pressure coefficients on parapet edge. However, the positive 'wall edge zone'

coefficient underestimates the positive surface pressure coefficients by 20 %.

Figure 5.11-b shows the (GCp)eq values for the interior surface of parapet edge and those

for the roof corner zone obtained from ASCE 7-05 (Appendix B, Figure B2). The

negative 'roof corner zone' coefficient overestimates the negative surface pressure

coefficients on parapet edge by about 40%. On the other hand, the positive 'wall edge

zone' coefficient adequately envelops the positive surface pressure coefficients.

The net loading from both parapet surfaces is presented in Figure 7.11-c. The 'edge

zone' coefficient for the Load Case A overestimates the net positive pressure coefficients

on the edge of windward parapet by about 50%. The 'edge zone' coefficient for Load

Case B overestimates the net negative pressure coefficients on the edge of leeward

parapet by about 30%.

109
A similar comparison was made in Figure 5.12 between wall interior and roof edge

regions and the parapet mid-span. In Figure 5.12-a the exterior surface loads on the

parapet mid-span are compared with the GCP values for 'wall interior zone' obtained

from ASCE 7-05 (Appendix B, Figure BI). The negative 'wall interior zone' coefficient

found to overestimate the negative surface pressure coefficients on parapet mid-span by

10 - 35%. However, the positive 'wall interior zone' coefficient underestimates the

positive surface pressure coefficients by 10 - 30%.

Figure 5.12-b shows the equivalent GCP values for the interior surface of parapet interior

region compared with those for roof edge zone obtained from ASCE 7-05 (Appendix B,

Figure B2). The negative 'roof edge zone' coefficient overestimates the negative surface

pressure coefficients on parapet mid-span by 35%. In addition, the positive 'wall interior

zone' coefficient overestimates the positive surface pressures by 25%.

The net loading from both parapet surfaces is presented in Figure 5.12-c. For Load Case

A, the 'interior zone' coefficient overestimates the net positive pressure coefficients on

the mid-span of wind-ward parapet by 25% and 40% for the higher and lower parapet,

respectively. For Load Case B, the 'interior zone' coefficient overestimates the net

negative pressure coefficients on the mid-span by 30% and 20% for the higher and lower

parapet, respectively.

Generally, the comparisons of the results of the present study with the ASCE 7-05

provisions shows the latter to be on the conservative side for most cases. However, the

110
provisions underestimate the positive pressure coefficients on the exterior parapet

surface. The overall approach adopted within the standards, which utilizes the area-

averaged pressure coefficients from the wall and roof of the building, was found to

significantly overestimate the combined effect from both surfaces. This is attributed to

poor correlation between the exterior and interior surface pressures, i.e. the peak

pressures on both surfaces do not occur simultaneously, as shown before.

Based on the results of the present study, the following recommendations can be

proposed for wind standards and codes of practice, particularly for the ASCE 7-05 (the

design wind pressure on the components and claddings of parapets):

• Wall pressure coefficients can be used as a good estimate for those on the exterior

parapet surface.

• Pressure coefficients on the corner of the interior parapet surface can be reduced

from + 3.8 and - 2.4 for windward and leeward parapets, respectively, to + 2.4 and

-2.0.

• Pressure coefficients on the mid-span of the interior parapet surface can be from

+ 2.8 and - 2.2 for windward and leeward parapets, respectively, to + 2.0 and - 1.8.

Ill
 Table 5.1: Characteristics of different models used in each study

Study L/H h/H x/L £/hp

a
Present study 1.2 0.16 0.04 0.50

Stathopoulos et al, 2002-a 1.0 0.13 0.08 0.45

Stathopoulos et al, 2002-b 3.4 0.10 0.08 0.45

4
.
wind
A

Mans et al, 2005 3.0 0.10 0.06 0.83 ^ L 4

112
Table 5.2: Comparison of parapet load coefficients derived from NBCC (2005) with

the current measured values

Study Corner Mid-span

(CPCG)WAU (NBCC, 2005) 1.8 1.8

(CPCG)ROOF (NBCC, 2005) -5.4 -2.5

('CF)parape, (NBCC, 2005) 7.2 4.3

FS 3.72 2.88
(CF)parapet (present study)
WT 3.68 3.36
Note: these values represent the largest peak values from all wind directions

113
 t-yWind
Y O.lSw^Bi,,

yo
4.45 m
-i -
• FS ( C p , m e a n ) •WT(Cp,mean)

* FS (Cp,max.) -WT(Cp,max.)

A FS (Cp,min.) -WT(Cp,min)
30 60 90 120 150 180
w i n d direction (degree)

(a) Exterior surface parapet pressure coefficients

^—Bli

• FS (Cp.mean) - W T (Cp.mean)

• FS (Cp.max.) -WT(Cp,max.)

-4
a FS (Cp,min.) - W T (Cp.min.)
30 60 90 120 150 180

w i n d dii-ection ( d e g r e e )

(b) Interior surface parapet pressure coefficients

Figure 5.1: Surface parapet pressure coefficients recorded near the corner (Tap BI)

114
 Wind

Kvn
r*'

• FS (Cp,mean) -WT(Cp,mean

* FS (Cp, max.) -0— WT, (Cp, max)

A FS (Cp, min.) -A—WT(Cp, min)

30 60 90 120 150 180
wind direction (degree)

(a) Exterior surface parapet pressure coefficients

1— s . Wind

• FS (Cp,mean) -B— WT(Cp,mean)

• FS (Cp.max.) -o— WT (Cp^nax.)

A FS (Cp,min.) WT'(Cp/nin.)
0 30 60 90 120 150 180
wind direction (degree)

(b) Interior surface parapet pressure coefficients

Figure 5.2: Surface parapet pressure coefficients recorded at the mid-span (Tap B3)

115
 " FS (Cp i,mean) • FS (Cpr,mean)

-3
> FS (Cpi,max.) <>FS (Cpr,max.)

-4

30 60 90 120 150 180 i FS (Cpi.min.) £ F S (Cpr.min.)

w i n d direction (degree)

(a) Full-scale

h-^Wind
-Bli
z .

- - WT (Cpi,mean) - WT (Cpr,mean)

• - - WT (Cpijnax.) 6 WT (Cpr,max.)

0 30 60 90 120 150 180 A WT (Cpi,min.) A WT(Cpr,min.)

wind direction (degree)

(b) Wind tunnel

Figure 5.3: Comparison of peak pressure coefficients on the roof and the inside

surface of the parapet near the corner (Tap BI)

116
 2 -- J. ..

J <> T
<>
,> i i >j j I
UO U o if ' L 1
" a a
• a 0 « 9 fl fl a
-1
• FS (Cpynean) • FS (Cpiynean)
IV l r t^j r I 's J
-2
ii-li' i
t • FS (Cpynax.) OFS(Cpr,max.)

-3
A FS (Cpynin.) A F S (Cpiynin.)
0 30 60 90 120 150 180
wind direction (degree)

(a) Full-scale

z
Wind

B3i

B3r

* *

• - - WT (Cpi,mean) -WT(Cpr^iean)

• WT (Cpi,max.) -0 WT (Cpiynax.)

30 60 90 120 150 180 A WT (Cpi,min.) -A WT (Cpr^nin.)

wind direction (degree)

(b) Wind tunnel

Figure 5.4: Comparison of peak pressure coefficients on the roof and the inside

surface of the parapet at mid-span (Tap B3)

117
 z
Wind

B3i -Bli

B3r —A
A s
-Blr

4.45 m

• WT (Blr - Bli)

— * — WT (B3r - B3i)

- - - « — Stathopoulos et a I , 2002-a (corner)

0 30 (SO 90 120 150 180 - " A - - - .Stathopouloset al., 2002-a (mid-span)

wind direction (degrees)

Figure 5.5: Effect of wind direction on the correlation coefficients of Cp, and C«

118
 Wind

R(Ble-Bli)

R(B2e-B2i)

R(B3e-B3i)

R(B4e-B4i)

30 60 90 120 150 180 R(B5e-B5i)

wind direction (degrees)

(a) Wind tunnel data

B5e_
/
B4e Bjje B2e Ble

R(Ble-Bli)

• R(B2e-B2i)

R(B3e-B3i)

• R(B4e-B4i)

30 60 90 120 150 180 R(B5e-B5i)

wind direction (degrees)

(b) Full-scale data

Figure 5.6: Effect of wind direction on the correlation coefficients of Cpe and Cp;

119
 e Wind

• FS (Cp, max-s imult) - WT (Cp, max-s imult)

O FS (Cp,min-simult) - WT (Cp,min-s imult)

o FS ( C p , m a x - n o n s imult) - WT(Cp,max-nonsimult)

30 60 90 120 150 180 • F S ( C p , m i n - n o n s imult) - WT (Cp, min-no n s imult)

wind direction (degree)

(a) Pressure coefficients at the corner (Tap BI)

F S (Cp, max-s imult) - WT (Cp, max-s imult)

F S (Cp, min-s imult) - WT (Cp, min-s imult)

F S (Cp, m a x - n o n s i m u h ) • WT (Cp, ma x-no n s imuh)

30 60 90 120 150 180 F S (Cp, m i n - n o n s imult) - WT (Cp, min-no n s imult)

wind direction (degree)

(b) Pressure coefficients at mid-span (Tap B3)

Figure 5.7: Peak local load coefficients on the parapet measured

120
 ---•'• ^
y=x .

o -P
o Wind angle = 0
J...
i j / p
CP ° Wind angle = 4 5
' ' ' ^Jjkr^
1 2 3 -
• Wind angle = 9 0
i -3 -2 -ijsHr t

D
* Wind angle = 135
| « | . . t
+ Wind angle = 1 8 0
v D
- « ,

y = 1.012X + 0 . 0 1 9

C P (FS)

Figure 5.8: Comparison between pressure coefficents measured in full-scale and

wind tunnel

121
 8
FS, current study
6
L / H = 1.2,hp/H = 0.16
4
WT, Stathopoulos e t a l (2002-a)
2 L / H = 1.0,hp/H = 0.13
0 I.
- a — WT, Stathopoulos e t a l (2002-b)
-2 L/H = 3.0, h p / H = 0.10

-4 - e — WT, M a n s et al (2005)
L/H = 3.4, h / H = 0.10
-(S

-8

30 60 90 120 150 180

w i n d direction (degree)

(a) Exterior surface parapet pressure coefficients

6 - - a FS, Current study

L / H = 1.2,hp/H = 0.16

WT, Stathopoulos e t a l (2002-a)

L/H = 1.0,hp/H = 0.13

- a — WT, Stathopoulos e t a l (2002-b)

x — L / H = 3.0, h p / H = 0.10

-4 - e — WT, M a n s et al (2005)
L/H = 3 . 4 , h / H = 0.10

——
i
30 60 90 120 150 180
w i n d direction (degree)

(b) Interior surface parapet pressure coefficients

8
FS, Current study
L / H = 1.2, h p / H = 0.16

WT, Stathopoulos e t a l (2002-a)

L/H = 1 . 0 , h p / H = 0.13

- a — WT, Stathopoulos et al (2002-b)

L/H = 3 . 0 , h p / H = 0.10

- a — WT, M a n s et al (2005)
L/H = 3 . 4 , h p / H = 0.10

30 60 90 120 150 180

Mind direction (degree)

(c) Simultaneous combined pressure coefficients from both surfaces

Figure 5.9: Comparison of peak pressure coefficients from different studies

122
 FS (hp = 0.5m)

ft. - W T (hp = 0.5m)

u
- W T (hp = 1m)

0.1 10
area (in2)

(a) Exterior parapet loading coefficients

FS (hp = 0.5m)

- W T (hp = 0.5m)

- W T (hp = l m )

0.1 10
area (m2)

(b) Interior parapet loading coefficients

FS (hp = 0.5m)

% 0 - W T (hp = 0.5m)

- W T (hp = l m )

0.1 10
area (in2)

(c) Net loading from both surfaces

Figure 5.10: Variation of peak loading coefficients (obtained from the most
critical azimuth) with different tributary areas

123
 • FS (hp = 0.5 m)

O WT (hp = 0.5 m)
0
a WT (hp =1.0 m)

ASCE 7-05 (wall edge)

0.1 10
area (m2)

(a) Exterior parapet loading coefficients

-1
3

2 • FS (hp = 0 5 m)

1 o WT (hp = 0.5m)

0 A WT (hp = 1 m)

-1 ASCE 7-05 (wall edge)

-2 ASCE 7-05 (roof comer)

-3
0.1 1 10
urea (m2)

(b) Interior parapet loading coefficients

• FS (hp = 0.5)

<> WT (hp = 0.5m)

w A WT (hp = 1 m)
O
ASCE 7-05 (Load Case A)

ASCE 7-05 (Load Case B)

0.1 10
area (in2)

(c) Net loading from both surfaces

Figure 5.11: Comparison of edge region loading coefficients with ASCE 7-05

124
 • FS (hp = 0 .5 m)
1 -
- « — W T (hp = 0.5m)
0
-A—WT (hp = lm)
0 -1
ASCE 7-05 (wall interior)
-2
-3
O.I 1 10
area (in2)

(a) Exterior parapet loading coefficients

2
FS (hp = 0 .5 m)
1 -
- WT (hp = 0 .5m)
0
- WT (hp = lm)

-1
- ASCE 7-05 (wall interior)

•ASCE 7-05 (roof edge)

0.1 1 10
area (m2)

(b) Interior parapet loading coefficients

• FS (hp = 0 5)

-e WT (hp = 0.5m)

—h— W T ( h p = lm)

ASCE 7-05 (Load Case A)

ASCE 7-05 (Load Case B)

0.1 1 , 10
area (in2)

(c) Net loading from both surfaces

Figure 5.12: Comparison of interior region loading coefficients with ASCE 7-05

125
 Chapter Six

NUMERICAL RESULTS AND DISCUSSION

Numerical simulation of wind flow over the test building model with parapets is

performed with the commercial CFD code Fluent 6.1.22. Simulation details were

explained in Chapter three. This chapter presents an analysis and comparisons of

numerical outcomes. The computation has been conducted for wind flow perpendicular

on the building face (i.e. zero wind azimuth). Only mean wind-induced pressures on the

roof and on parapet surfaces are computed, since peak pressures are not predicted reliably

by current computational approaches. Two different turbulence models, namely: RNG

and RLZ k-s are used. The computational results are compared with the present

experimental results as well as with those from previous numerical studies. Note that, all

pressure data are presented in the form of pressure coefficients, normalized by using the

dynamic pressure at roof height.

6.1 Flow Patterns around the Building

The wind flow conditions around the building with parapet have been numerically

simulated using RNG k-s model. Data is presented for the building with 0.01 m parapet

height which corresponds to a full-scale parapet height of 0.5 m. The mean velocity

vector fields in cross-section through the center of the building are presented in Figure

126
6.1. The reversed flows in front of the building near the ground, the upstream standing

vortices (Figure 6.1-a); and the circulation flow region behind the building (the near wake

recirculation region) are fully predicted (Figure 6.1-b). Very small arrows indicate very

low mean velocity, which is the characteristic of the flow in such regions. Mean stream

lines in the approach flow are coming closer together toward the top edge of the

windward parapet and very high velocity at this edge is clearly shown (Figure 6.1-c).

The presence of the parapet makes the separated flow to be raised from the roof edge and

a separation region is formed behind the windward parapet. Smaller vectors in this

region indicate lower velocity and hence lower pressure, comparing with the case of no

parapet where the flow separates at the roof edge. Flow reattachment does not occur on

the roof surface, as expected for buildings with reduced length/width ratio. Therefore, the

bulk of the wind flow passes above the parapet height with a small amount of

recirculation above the roof. This significantly changes the flow dynamics over the

building by reducing the pressure coefficients on the roof surface. The separated flow

over the leeward parapet edge continues circulating and merged with the near wake

recirculation flow. The wind conditions around a plan view, at roof height, are presented

in Figure 6.2. The separation regions, found in the experiments by Murakami, 1990,

beside the side walls have been successfully predicted. Also, the recirculation behind the

building is clearly shown.

6.2 Mean Pressure Distribution on the Roof

In order to monitor the general distribution of wind pressures on the building roof without

parapet, the current numerically predicted pressures in the form of pressure coefficient

127
contour-lines are presented in Figure 6.3-a and compared with some previous numerical

and experimental results. The agreement between the current study and the numerical

results of Zhou (1995), for the two-layer method, is reasonable (Figure 6.3-b),

considering the differences in building geometries. Furthermore, the current study agrees

with the experimental results of Hunt, 1982, (Figure 6.3-c).

Figure 6.4-a shows the results of the roof with 0.01 m parapet. It can be clearly seen that

the high negative pressures (suctions) near the windward edge have been reduced by

adding the parapet, which has been experimentally verified by many previous studies

(Kind, 1988; Baskaran and Stathopoulos, 1988-a). In addition, the presence of parapet,

also, acts to modify the pressure distribution on the roof to be more uniform, as indicated

by Kareem and Lu (1992). Figure 6.4-b shows the current wind tunnel experimental

results for building roof with 0.01 m parapet. Numerical results show good performance

compared to those from the wind tunnel.

The computed pressure coefficients along the center line of the roof, for both RNG and

RLZ k-s models, are compared with the wind tunnel experimental results (Figure 6.5).

The results of the building with two different parapet heights: hp = 0.01 m (Figure 6.5-a)

and hp = 0.02 m (Figure 6.5-b), are presented. RLZ k-s model shows significant

discrepancy in the pressure coefficients comparing with the measured data. In contrast,

the general agreement between numerical results of RNG k-s model and experimental

data is reasonable. However, the locations of maximum mean pressure coefficient along

the building centerline are different. The numerical maximum pressure coefficient occurs

128
at a distance of x/L of 0.25 for lower parapet and of 0.35 for higher parapet. On the

contrary, the experimental maximum pressure coefficient occurred closer to the roof

edge, x/L of 0.15 for both parapet heights.

6.3 Effect of parapet height on roof pressures

In an attempt to examine the effect of parapet height on roof pressures, the negative

pressure coefficients at a point near the leading edge along the center line of the roof are

plotted for different parapet heights (Figure 6.6). Numerical results are compared with

the wind tunnel experimental data of Stathopoulos et al. (2002-b). The model for that

study was a cube of 30 m long and has been tested for different parapet heights.

Therefore, the ratio of parapet height to the building height (hp/H) has been used for

comparison. As shown, the computed pressure values for low parapet height (hp/H =

0.03) tend to show higher suction near the windward edge of the roof, comparing with the

case of no parapet. Increasing parapet height (hp/H = 0.67, 0.13 and 0.2) causes an

evident reduction in the negative pressures near the windward edge of the roof, which

had been concluded previously in many experimental studies (Bienkiewicz and Sun,

1992; Baskaran and Stathopoulos, 1988-a).

6.4 Mean pressures on the parapet itself

In an attempt to study the wind-induced pressures on parapets themselves the pressure

coefficients on the exterior and interior parapet surfaces, along a vertical line at centerline

129
of the parapet, are presented in Figure 6.7-a. The computational results on the exterior

surface are compared with wind tunnel data as well as those from Stathopoulos et al.

(2002-b), at the same locations. Numerical and experimental data shows that exterior

mean pressures reach their maximum values at the base of the parapet and gradually

reduce with parapet height. With regard to the interior pressures, numerical computed

coefficients seem to reach their maximum values within the mid-height of the parapet.

The total pressure coefficients resulting from combined both surface coefficients are

shown in Figure 6.7-b and are compared with the wind tunnel data. Generally, good

agreement between the numerical and wind tunnel measurements can be shown.

Table 6.1 compares the mean pressure coefficients on the roof (C pr ) and on the inside

surface of the wind-ward parapet (C pi ). The full-scale parapet height is 1 m. Data were

obtained at the corner region and at the mid-span location. Cp values are presented at

points very close to the roof edge (x/l = 0.05). The computed values of Cpr and Cpi at the

mid-span location are approximately the same. On the other hand, at the corner region

Cpr is approximately 10% larger than Cp,. The comparison with the wind tunnel

experimental results, at the same points, shows that the numerical predictions fall on the

lower side, except for the corner roof pressures.

130
Table 6.1: Numerical and Experimental pressure coefficients

Corner region Edge mid-span

Test type
Cpr Cpi Cpr Cpi

Numerical -1.01 -0.80 -0.91 -0.76

Experimental -0.73 -0.71 -0.69 -0.68

131
(a) Upstream standing vortex (b) Near wake recirculation

r
j, / /'yy/fUt
// i
' / / / f l!\
-"VV///,

(c) Separation region

9.50e»00

9.03e«00

8.56e«00

8.08e»00

7.61e*00

7.13e+00

6.66c»00

6.19e+00

5.71c«00

5.24e*00
ill 4.77e«00

4.29e»00

3.82e+00

3.35e.OO

2.87e*00

3.40c»00

1.92c.00

1.45«*00

9.76e-01

5.03e-01

2.90e-02

Velocity V e c t o r s Colored By Velocity Magnitude (m/s) F i b 07. 2 0 0 9

FLUENT 6.1 (2d. segregated, rngVc)

Figure 6.1: Mean velocity vector field around the building with 0.01 m parapet

(Vertical cross-section)

132
Figure 6.2: Mean velocity vector field around the building (Plan-view)

133
 (a) RNG k-s, (present study)

wind

• • I I I ! I I L

(a) N-R two layer method (Zhou, 1995)

wind

(b) Experimental (Hunt, 1982)

Figure 6.3: Contours of pressure coefficient on roof surface from different studies
(h p = » m )

134
 wind

(a) Wind tunnel, hp = 0.01 m

wind
-0.6
n
-0.81 -0.72 -0.61 -0.43 -0.3

-0.18

i 'i

(b) RNG k-e, hp = 0.01 m

Figure 6.4: Contours of pressure coefficients on roof surface (current study)

135
 hp= 0.01 m

Exprimental, WT

• Numerical, RNG k-e

• Numerical, RLZ k-e

(a) hp = 0.01 m

x/1

h p = 0.02 m

Exprimental, WT

• Numerical, RNG k-e

• Numerical, RLZ k-e

(b) hp = 0.02 m

Figure 6.5: Pressure coefficient along the center line of building roof

136
137
 wind

h p = 0.02 m

« Cpe - WT

Cpe - RNG k-e

- O — Cpe - Stathopo lous etal.,2002-b

• Cpi - WT

Cpi- RNG k-e

(a) Individual mean pressure coefficients

Figure 6.7: Change of parapet surface pressure with parapet height

138
 Chapter Seven

CONCLUSIONS, RECOMMENDATIONS AND FUTURE WORK

In the past, only a limited number of studies were attempted to directly measure parapet

surface pressures in the wind tunnel, due to modeling limitations. However, no full-scale

study was reported. Consequently, wind loading standards and codes of practice provide

very little guidance regarding the design wind loads of parapets. Recently, the ASCE-7

(2005) considered the importance of defining parapet loads. However, due to insufficient

research, the proposed recommendations are based on a rational technique rather than on

experimental investigations. The present study has attempted to reduce this lack of

information regarding parapet design loads based on comprehensive experimental and

numerical investigations.

7.1 Summary

In wind engineering research, it is extremely important to carry out field measurements

for wind flow on real buildings. Full-scale testing is suitable to obtain reliable wind-

measured data and to understand the flow mechanism around buildings. Consequently,

this type of testing is the only reliable method to validate wind tunnel results as well as to

verify and develop numerical simulations. Comprehensive field measurements of parapet

surface pressures have been completed using a small low-rise building located in an open

139
area. A perimeter parapet of a height of 0.5 m is attached to the building roof and is

drilled for tapping on all parapet surfaces. Mean and simultaneous peak pressures, on the

exterior and interior parapet surfaces as well as the roof edge, were recorded.

The experimental building is modeled to be tested in the BLWT of Concordia University.

Based on the importance of correctly modeling the turbulence intensity at the roof height,

a geometric model scale of 1:50 is used. Two sets of perimeter parapets: 0.01 and 0.02

m, equivalent to 0.5 and 1 m height in full-scale respectively, have been attached to the

building model roof. Both local and area-average pressure coefficients are recorded on

the exterior and the interior parapet surfaces.

Accurate computer predictions are very useful since this method is less expensive and

time consuming than comparable field or wind tunnel testing. Numerical simulation of

wind flow over the building model with the parapet was performed with the commercial

CFD code Fluent 6.1.22. The steady-state RANS equations were solved using two

different modified k-£ turbulence models: the RNG k-e model and the RLZ k-£ model.

The computations were performed for flow perpendicular to the building face ( 0 = 0°).

The experimental building with different parapet heights was simulated in order to

examine the effect of parapet on roof pressures. Considering the current state-of-the-art

peak pressures are not predicted reliably by computational approaches. Therefore, in the

present study only mean wind-induced pressures on the roof and on parapet surfaces were

computed and compared with the current experimental results as well as with previous

numerical studies.

140
7.2 Concluding Remarks

The wind tunnel results have been systematically compared with the full-scale collected

data through the evaluation of the descriptive parameters of parapet surface pressures.

Generally, the comparisons reveal reasonable agreement between the two different

techniques, confirming successful wind tunnel simulation. Mean pressure coefficients

measured in the wind tunnel show very good agreement with the full-scale results.

However, the analyses demonstrate some differences in the maximum peak pressure

coefficients particularly for critical wind directions, where the highest pressures usually

recorded. These differences are largely attributed to the unsteadiness of the direction of

natural wind. Moreover, locations and density of the pressure taps on the model may be

considered the most important parameter in affecting the experimental results. Specific

conclusions regarding the measurements of local parapet surface pressures can be

illustrated as follows:

• For exterior parapet surface, the maximum positive peak pressure coefficients

occur for winds approaching perpendicular to the parapet face, while the

maximum negative peak pressure coefficients occur for winds approaching

parallel to the parapet section.

• For interior parapet surface, the maximum negative peak pressure coefficients

occur at the leading corner for wind angle near 45°. On the other hand, the

141
 maximum positive peak pressure coefficients occur for winds perpendicular to the

interior parapet surface (9= 180°).

• The maximum negative peak pressure coefficients measured on the roof edge are

approximately 15 % and 10 % higher than those on the interior parapet surface,

for corner and mid-span locations, respectively.

• Peak drag coefficients, which include the combination of pressures from both

parapet surfaces, were found to occur during cornering winds. Negative pressure

coefficients on the interior parapet surface are found to dominate the net loading

on the parapet. However, the positive pressures on the exterior surface make an

important contribution.

• The local inward and outward loads on the parapet measured near the leading

corner were approximately 25% and 30%, respectively, higher than those at the

mid-span. Simultaneous pressures are found to be lower than non-simultaneous

pressures.

• The correlation between roof edge and interior parapet surface pressures depend

on the wind direction. High correlation obtained for 9 = 0°, while, for oblique

wind directions (30° - 60°) lowest correlation was found. This indicates that roof

edge pressures may not be an adequate estimation for the interior parapet surface

pressures, in the absence of experiments.

142
 • The overall correlation between the pressures on the exterior and the interior

parapet surfaces is considerably low, which indicates that the absolute maximum

peak pressures do not occur simultaneously on both surfaces. Therefore, the total

load coefficients on a parapet may only be obtained by simultaneously combined

the pressure coefficients on both surfaces. The correlation between the exterior

and interior parapet surface pressures, measured in the wind tunnel, shows

reasonable agreement with the respective full-scale correlation.

• The parapet design loads obtained by combining the NBCC (2005) design load

coefficients for windward wall and roof were significantly higher than actual

loads on the parapet.

Area averaged pressure coefficients show that the net loads on the 1 m height parapet are

higher than those on the 0.5 m height parapet by about 15 %. The area-averaged loads on

the exterior and the interior surfaces of the parapet, in addition to their simultaneous

combination, were compared with the ASCE 7-05 provisions. Specific conclusions are

listed as follows:

• For the exterior parapet surface:

The 'wall edge zone' coefficient suggested by the standard underestimates the

surface pressure coefficients on parapet edge up to 20 %.

The negative 'wall interior zone' coefficient overestimates the negative surface

pressure coefficients on parapet mid-span by 10 - 35%. While the positive 'wall

143
 interior zone' coefficient underestimates the negative surface pressure coefficients

on parapet mid-span by 10 - 30%.

• For the interior parapet surface:

The 'roof corner zone' coefficient overestimates the negative surface pressure

coefficients on parapet edge by about 40 %.

The 'roof edge zone' coefficient overestimates the surface pressure coefficients

on parapet interior locations by 25 - 35 %.

• For combined action from both surfaces:

The coefficient for the Load Case A overestimates the net positive pressure

coefficients on the edge of wind-ward parapet by 25 - 50%.

The coefficient for Load Case B overestimates the net pressure coefficients on the

edge of lee-ward parapet by 20 - 30 %.

Wind tunnel experimentations, with their reasonable agreement with full-scale data, will

continue to be beneficial for research and applications. On the other hand, full-scale

investigation is essential to validate the wind tunnel testing results. Properly conducted

field testing can also overcome the difficulties of laboratory test deficiencies (Mehta,

2004).

Numerical simulation results of wind flow around a low-rise building with parapet, using

Fluent 6.1.22, show that the computed mean pressure coefficients on parapets and on the

144
roof with parapet agree with the wind tunnel data, for wind perpendicular to the building

face. The RNG k-e turbulence model shows better performance in predicting mean

pressures when compared with the RLZ k-smodel.

It can be realized that it is possible to achieve similar trends, of numerical simulations

and experimental measurements, for mean wind pressures over a low-rise building.

However, equivalence between results of the two techniques may be achieved with care

taken to provide accurate inlet flow profiles, adequate grid resolution, proper boundary

conditions and appropriate turbulent model.

As also concluded by Stathopoulos (2002-c), the numerical wind tunnel is still virtual

rather than real and further experimental studies will be required to increase the level of

confidence in the computational results, in spite of some interesting and visually

impressive results produced with Computational Wind Engineering (CWE).

7.3 Recommendations Based on the Current Study

For the design of the main wind force-resisting system the ASCE 7-02 standard

recommends net coefficients (GCP) of + 1.8 and - 1.1 for the windward and leeward

parapets, respectively. However, after completion of preliminary work on which the

present study is based (Stathopoulos et al., 2002-a and 2002-b), these values were revised

in the ASCE 7-05 version to + 1.5 and - 1.0, respectively.

145
With regards to the loads on the components and cladding, ASCE 7-05 recommends GCP

values of + 3.8 and - 2.4 for windward and leeward parapets, respectively, at edge region

and + 2.8 and - 2.2 for mid-span location. Based on the results of the present study, the

following recommendations can be proposed for wind standards and codes of practice,

particularly for the ASCE 7-05 (the design wind pressure on the components and

claddings of parapets):

• Wall pressure coefficients can be used as a good estimate for those on the exterior

parapet surface.

• Pressure coefficients on the corner of the interior parapet surface can be reduced

from + 3.8 and - 2.4 for windward and leeward parapets, respectively, to + 2.4 and

-2.0.

• Pressure coefficients on the mid-span of the interior parapet surface can be from

+ 2.8 and - 2.2 for windward and leeward parapets, respectively, to + 2.0 and - 1.8.

7.4 Limitations of the Presents Study

Although every effort has been made to produce the best possible results in terms of

wind-induced pressures on parapets, there are some limitations to be mentioned. These

include but are not limited to the following:

• Using specific building configuration and exposure in the full-scale testing, which

may not be representative of other cases

146
 • Using sampling rate of 1 Hz in field measurements may not be adequate to capture

the most critical peak values. However, comparisons with the results obtained by

using higher frequencies are encouraging.

7.5 Future Work

The current thesis has addressed some outstanding concerns regarding to the design wind

loads of parapets on low-rise buildings. By providing a complete data set, further studies

may assist in expanding the understanding of the results presented in this study. Future

research will also help in answering questions raised from the present results. The

proposed experimental and numerical investigations are listed as follows:

• Examining different building and parapet heights:

The current study deals with specific building and parapet height. Only tow parapet

heights were tested in the wind tunnel. It is recommended to study the effect of

parapet height on parapet loads by examining different parapet heights. It is also

recommended to examine parapet loads for different building heights.

• Measuring surface pressures on a single parapet:

The experimentations conducted in the current study considered a parapet extending

around the entire perimeter of the building. Some previous studies experienced

significant changes in roof corner pressures in the case of an isolated parapet

(Stathopoulos and Baskaran, 1988; Mans et al., 2003). A single parapet located at

147
 the leading corner of the building roof may significantly change the flow dynamics

around the building, as the corner vortex from the separation at the roof will interact

with a second vortex associated with separation from the end of the parapet (Mans

et al., 2003). It is suggested that a full-scale study may be performed using a

parapet located along a single building edge. Pressures may be recorded on the

exterior and interior surfaces of such parapet and compared with the present results

of the perimeter parapet.

• Visualizing the flow field around the parapets:

A flow visualization study may be useful in understanding the physical nature of

wind flow above the parapet and clarifying the existing pressure data. The study

can include different building and parapet heights in addition to a single parapet.

Such study may also help in verifying whether the current results are appropriate for

different building configurations.

• Numerical computation of peak pressures:

The current numerical simulation is performed in steady-state conditions and thus

only mean pressures are provides. In order to compute wind-induced peak

pressures on parapets, and hence calculate the design wind loads, the time

dependant form of RANS equations needs to be solved. Either Large Eddy

Simulation (LES) or the Algebraic Stress Model can be used. In this case different

wind directions may be examined as it was concluded that the largest peak

pressures on a parapet occur at an oblique wind angle, i.e. near 45°.

148
 REFERENCES

Adams, E., Johnston, J., 1998. 'Flow Structure in the Near Wall Zone of a Turbulent

Separated Flow'. Journal of American Institute of Aeronautics and Astronautics, AIAA.

26: 932-939.

ANSI A.58.1. 'Minimum Design Loads for Building and Other Structures'. American

National Standards Institute, New York.

ASCE-7, 2005. 'Minimum Design Loads on Buildings and Other Structures'. American

Society of Civil Engineers.

ASHRAE, 1999. 'Hand book Applications'. American Society of Heating, Refrigerating

and Air-Conditioning Engineers. Millstar Electronic Publishing Group, Inc.

Baskaran, A., Stathopoulos, T., 1988. 'Roof Corner Wind Loads and Parapet

Configuration'. Journal of Wind Engineering and Industrial Aerodynamics, 29: 79-88.

Bienkiewicz, B. and Sun, Y., 1992. 'Local Wind Loading on the Roof of a Low-rise

Building'. Journal of Wind Engineering and Industrial Aerodynamics, 45: 11-24.

Blackmore, P., 1988. 'Load Reduction on Flat Roofs - The Effect of Edge Profile'.

Journal of Wind Engineering and Industrial Aerodynamics, 29: 89-98.

149
Blocken, B., Carmeliet, J., Stathopoulos, T., 2007-a. 'CFD Evaluation of the Wind Speed

Conditions in Passages Between Buildings—Effect of Wall-Function Roughness

Modifications on the Atmospheric Boundary Layer Flow'. Journal of Wind Engineering

and Industrial Aerodynamics. 95: 941-962.

Blocken, B., Stathopoulos, T., Carmeliet, J., 2007-b. 'CFD Simulation of the

Atmospheric Boundary Layer-Wall Function Problems'. Atmospheric Environment. 41:

238-252.

Blocken, B., Stathopoulos, T., Saathoff, P., Wang, X., 2008. 'Numerical Evaluation of

Pollutant Dispersion in the Buit Environment: Comparisons between Models and

Experiments'. Journal of Wind Engineering and Industrial Aerodynamics, 96: 1817-1831.

Cebeci, T., Bradshaw, P., 1977. 'Momentum Transfer in Boundary Layers'. Hemisphere

Publishing Corporation, New York.

Chang, C., Meroney, R., 2003-a. 'The Effect of Surroundings with Different Separation

Distances on Surface Pressures on Low-Rise Buildings'. Journal of Wind Engineering

and Industrial Aerodynamics, 91: 1039-1050.

Chang, C., Meroney, R., 2003-b. 'Concentration and Flow Distributions in Urban Street

Canyons: Wind Tunnel and Computational Data'. Journal of Wind Engineering and

Industrial Aerodynamics, 91: 1141-1154.

150
Cheng, Y., Lien, F., Yee, E., Sinclair, R., 2003. 'A Comparison of Large Eddy

Simulations with a Standard k-s Reynolds-Averaged Navier-Stokes Model for the

Prediction of a Fully Developed Turbulent Flow over a Matrix of Cubes'. Journal of

Wind Engineering and Industrial Aerodynamics, 91: 1301-1328.

Cheung, J., Holmes, J., Melbourne, W., Lakshmanan, N., Browditch, P., 1997. 'Pressure

on a 1/10 Scale Model of the Texas Tech Building'. Journal of Wind Engineering and

Industrial Aerodynamics, 69-71: 529-538.

Columbus, J., 1972. 'The Study of Pressure Coefficients on Large Flat Roofs and Effects

of Parapets on These Coefficients'. Engineering Science 400 report, University of

Western Ontario, London, Canada.

Cook, N., 1985. 'The Designers Guide to Wind Loading of Building Structures'.

Butterworth.

Daugherty, R., et al., 1983. 'Fluid Mechanics with Engineering Applications'. 8th

edition, McGraw-Hill, New York.

Davenport, A., 1961. 'The Spectrum of Horizontal Gustiness near the Ground in High

Winds'. Quarterly Journal of the Royal Metrological Society, 87: 372.

151
Davenport, A., Surry, D., 1974. 'The Pressure on Low-Rise Structures in Turbulent

Wind'. Proceed of the Canadian Structural Engineering conference, the Canadian Steel

Industrial construction Council, 1-39.

Eaton, K., Mayne, J., 1975. 'The Measurement of Wind Pressures on Two-Story Houses

at Alesbury'. Journal of Wind Engineering and Industrial Aerodynamics, 1: 67-109.

ESDU, 1983. 'Strong Winds in the Atmospheric Boundary Layer. Part 1: Mean-Hourly

Wind Speeds'. Engineering Sciences Data Unit, Data Item 82026.

Ferziger, J., Peric, M., 1999. 'Computational Methods for Fluid Dynamics'. 2ed edition.

Springer-Verlag Berlin Heidelberg.

Flowe, A., Kumar, A., 2000. 'Analysis of Velocity Fields and Dispersive Cavity

Parameters as A Function of Building Width to Building Height Ratio Using a 3-D

Computer Model for Squat Buildings'. Journal of Wind Engineering and Industrial

Aerodynamics, 86: 87-122.

Fluent 6.1.22, 2005. 'User's Guide'. Fluent Incorporated.

Gao, Y., Chow, W., 2005. 'Numerical Studies on Air Flow Around a Cube'. Journal of

Wind Engineering and Industrial Aerodynamics, 93: 115-135.

Gumble, E., 1958. 'Statistics of Extremes'. Colombia University Press, New York, USA.

152
Gumley, S., 1983. 'Tubing Systems for Pneumatic Averaging of Fluctuating Pressures'.

Journal of Wind Engineering and Industrial Aerodynamics, 12: 189-228.

Hall, D., Spanton, A., Macdonald, R., Walker, S., 1996. 'A Review of Requirements for

Simple Urban Dispersion Models'. Building Research Establishment, BRE Client Report

CR 77/96.

Ham, H., Bienkiewicz, B., 1992. 'Local Wind Loading on the Roof of a Low-Rise

Building'. Journal of Wind Engineering and Industrial Aerodynamics, 45: 11-24.

Ham, H., Bienkiewicz, B., 1998. 'Wind Tunnel Simulation of TTU Flow and Building

Roof Pressure'. Journal of Wind Engineering and Industrial Aerodynamics, 77-78: 119-

133.

Hanson, T., Summers, D., Wilson, C., 1984. 'Numerical Modelling of Wind Flow over

Buildings in Two Dimensions'. International Journal for Numerical Fluids, 4:25-41.

He, J., Song, S., 1997.'A Numerical Study of Wind Flow around the TTU Building and

the Roof Corner Vortex'. Journal of Wind Engineering and Industrial Aerodynamics, 67-

68: 547-558.

Ho, T., 1992. 'Variability of Building Wind Loads'. Ph.D. Thesis. University of Western

Ontario, London, Ontario, Canada.

153
Ho, T., Surry, D., Davenport, A., 1991. 'Variability of Low Building Wind Loads Due to

Surroundings'. Journal of Wind Engineering and Industrial Aerodynamics, 38: 297-310.

Ho, T. Surry, D., Morrish, D., Kopp, G., 2005. 'The UWO Contribution to the NIST

Aerodynamic Database for Wind Loads on Low Buildings: Part 1. Archiving Format and

Basic Aerodynamic Data'. Journal of Wind Engineering and Industrial Aerodynamics,

93: 1-30.

Holems, J., 1984. 'Effect of Frequency Response on Peak Pressure Measurements'.

Journal of Wind Engineering and Industrial Aerodynamics, 17: 1-9.

Holmes, J., Moriarty, W., 1999. 'Application of the generalized Pareto Distribution to

Extreme Value Analysis in Wind Engineering'. Journal of Wind Engineering and

Industrial Aerodynamics, 83: 1-10.

Hoxey, R., Moran, P., 1983. 'A Full-Scale Study of Geometric Parameters that Influence

Wind Load on Low- Rise Buildings'. Journal of Wind Engineering and Industrial

Aerodynamics, 13: 277-288.

Hoxey, R., Robertson, A., 1986. 'Pressure Coefficients for Low-Rise Building Envelopes

Derived from Full-Scale Experiments'. Journal of Wind Engineering and Industrial

Aerodynamics, 53: 283-297.

154
Hoxy, R., Robertson, A., Richardson, G., Short, J., 1997. 'Correction of Wind-Tunnel

Pressure Coefficients for Reynolds Number Effect'. Journal of Wind Engineering and

Industrial Aerodynamics. 69: 547-555.

Hunt, A. 1982. 'Wind-Tunnel Measurements of Surface Pressures on Cubic Building

Models at Several Scales', Journal of Wind Engineering and Industrial Aerodynamics,

10: 137-163.

Kareem, A., Lu, P., 1992. 'Pressure Fluctuations on Flat Roofs with Parapets'. Journal of

Wind Engineering and Industrial Aerodynamics, 41: 1775-1786.

Kasperski, M., 1996. 'Design Wind Loads for Low-Rise Buildings: A Critical Review of

Wind Load Specifications for Industrial Buildings'. Journal of Wind Engineering and

Industrial Aerodynamics, 61: 169-179.

Kawai, H., 1997. 'Structure of Conical Vortices Related with Suction Fluctuation on a

Flat Roof in Oblique Smooth and Turbulent Flows '. Journal of Wind Engineering and

Industrial Aerodynamics, 69: 579-588.

Kind, R., 1974. 'Wind Tunnel Tests on Building Models to Measure Wind Speeds at

Which Gravel is Blown off Roof-Tops'. National Research Council of Canada, Ottawa,

Ontario.

155
Kind, R., 1986. 'Worst Suctions near Edges of Flat Rooftops on Low-Rise Buildings'.

Journal of Wind Engineering and Industrial Aerodynamics, 25: 31-47.

Kind, R., 1988. 'Worst Suctions Near Edges of Flat Rooftops With parapets'. Journal of

Wind Engineering and Industrial Aerodynamics, 31: 251-264.

Kleiser, L., Zang, T., 1991. 'Numerical Simulation of Transition in Wall-Bounded Shear

Flows'. Annul Review of Fluid Mechanics, 23: 495-537.

Kopp, G., Surry D., Mans, C., 2005. 'Wind Effects of Parapets on Low Buildings: Part 1.

Basic Aerodynamics and Local Loads'. Journal of Wind Engineering and Industrial

Aerodynamics, 93: 817-841.

Kopp, G., Mans, C., Surry D., 2005. 'Wind Effects of Parapets on Low Buildings: Part 2.

Structural Loads'. Journal of Wind Engineering and Industrial Aerodynamics, 93: 843-

855.

Kopp , G., Mans, C., Surry, D., 2005. 'Wind Effects of Parapets on Low Buildings: Part

4. Mitigation of Corner Loads with Alternative Geometries'. Journal of Wind

Engineering and Industrial Aerodynamics, 93: 873-888.

Kramer, C., et al., 1978. 'Wind Pressure on Block-Type Buildings'. Proceed of 3 ^ coll.

on Industrial Aerodynamics, W. Germany, 241- 254.

156
Landsea, C., Pielke, Jr., R., 1998. 'Normal Hurricane Damages in the United States:

1925-1995'. Weather and Forecasting, 13: 621-631.

Laneville, A., 1990. 'Turbulence and Blockage Effects on Two-Dimensional Rectangular

Cylinders'. Journal of Wind Engineering and Industrial Aerodynamics, 33: 11-20.

Launder, B., Spalding, D., 1974. 'The Numerical Computation of Turbulent Flows'.

Computer Methods in Applied Mechanics and Engineering, 3: 269-289.

Larose, G., D'Auteuil, A., 2006. On the Reynolds Number Sensitivity of the

Aerodynamics of Bluff Bodies with Sharp Edges. Journal of Wind Engineering and

Industrial Aerodynamics, 94: 365-376.

Lee, S., 1997. 'Unsteady Aerodynamic Force Prediction on a Square Cylinder using k-s

Turbulence Models'. Journal of Wind Engineering and Industrial Aerodynamics, 67: 79-

90.

Leutheusser, H., 1964-a. 'The Effects of Wall Parapets on the Roof Pressure Coefficients

of Plock-Type and Cylindrical Structures'. Technical Publication Series, No. TP6404,

Department of Mechanical Engineering, University of Toronto, Toronto, Ont.

Leutheusser, H., 1964-b. 'Parapets Improve Roof Wind Loading'. The Consulting

Engineer: 94-96.

157
Levitan, M., 1993. 'Analysis of Reference Pressure Systems used in Field Measurements

of Wind Loads", Ph. D. Thesis, Texas Tech University, Texas, USA.

Levitan, M., Mehta, K., 1992-a. 'Texas Tech Field Experiments for Wind Loads part I:

Building and Pressure Measuring System'. Journal of Wind Engineering and Industrial

Aerodynamics, 41: 1565-1576.

Levitan, M., Mehta, K. 1992-b. 'Texas Tech Field Experiments for Wind Loads Part II:

Meteorological Instrumentation and Terrain Parameters'. Journal of Wind Engineering

and Industrial Aerodynamics, 41-44: 1577-1588.

Ligrani, P., Moffat, R., 1986. 'Structure of Transitionally Rough and Fully Rough

Turbulent Boundary Layers'. Journal of Fluid Mechanics, 162:69-98.

Lin, J., Surry, D., 1998. 'The Variation of Peak Loads with Tributary Area near Corners

on Flat Low Building roofs'. Journal of Wind Engineering and Industrial Aerodynamics,

77: 185-196.

Lin, J., Surry, D., Tieleman, H., 1995. 'The Distribution of Pressure near Roof Corners of

Flat Roof Low Buildings'. Journal of Wind Engineering and Industrial Aerodynamics,

56: 235-265.

158
Lythe, G., Surry, D., 1983. 'Wind Loading of Flat Roof with and without Parapets'.

Journal of wind Engineering and Industrial Aerodynamics, 11: 75-94.

Mans, C. 2001. 'Wind Loads on Parapets'. M.Sc., University of Western Ontario,

London, Ontario, Canada.

Mans, C. Kopp, G. and Surry, D. 2003. 'Wind Loads on Low Buildings With Various

Parapet Configurations'. Eleventh international conference on wind engineering.

Mans, C., Kopp, G., Surry, D., 2005. 'Wind Effects of Parapets on Low Buildings: Part 3.

Parapet Loads'. Journal of Wind Engineering and Industrial Aerodynamics, 93: 857-872.

Marathe, R., 1992. 'Full Scale Study of Wind Loads on Flat Roofs with Parapets'. Master

Thesis, Concordia University, Montreal, Canada.

Maruyama, T., et al. 2004. 'Field Experiment Measuring the Approaching Flows and

Pressures on a 2.4 Cube'. 5 th International Colloquium on Bluff Body Aerodynamics and

Applications, Ottawa, Canada.

Mehta, K., Levitan, M., Iverson, R., McDonald, J., 1992. 'Roof Pressures Measured in

the Field on a Low Building'. Journal of Wind Engineering and Industrial Aerodynamics,

41: 181-192.

159
Mehta, K., 2004. 'Importance of Field Testing in Wind Engineering'. Proceeding of the

National Conference on Wind Engineering, Nagpur, Phoenix Publication, New Delhi,

India.

Melaragno, M., 1996. 'Severe Storm Engineering for Structural Design'. Gordon and

Breach Science Publishers, Amsterdam.

Meroney, R., Leitl, B., Rafailidis, S., Schatzmann, M., 1999. 'Wind Tunnel and

Numerical Modeling of Flow and Dispersion about Several Building Shapes'. Journal of

Wind Engineering and Industrial Aerodynamics, 81: 333- 345.

Mikkelsen, A., Livesey, F., 1995. 'Evaluation of The Use of The Numerical K-s

Chameleon II, For Predicting Wind Pressures On Building Surfaces'. Journal of Wind

Engineering and Industrial Aerodynamics, 57: 375-389.

Mohan, M., Siddiqui, T., 1998. 'Analysis of Various Schemes for the Estimation of

Atmospheric Stability Classification'. Journal of Atmospheric Environment, 32: 3775-

3781.

Murakami, S., 1990. 'Computational Wind engineering'. Journal of Wind Engineering

and Industrial Aerodynamics, 36: 517-538.

160
Murakami, S., Moshihiko, A., Hayashi, Y., 1990. 'Examining the k-s Model by Means

of a Wind Tunnel Test and Large Eddy Simulation of the Turbulence Structure around a

Cube'. Journal of Wind Engineering and Industrial Aerodynamics, 35: 87-100.

Murakami, S., et al., 1992. 'Numerical Study on Velocity-Pressure Field and Wind

Forces for Bluff Bodies by k-e, ASM and LES'. Journal of Wind Engineering and

Industrial Aerodynamics, 44: 2841-2852.

Murakami, S., Mochida, A., 1988. '3-D Numerical Simulation of Air Flow around a

Cubic Model by Means of the k-s Model'. Journal of Wind Engineering and Industrial

Aerodynamics, 31: 283-303.

NBCC, 2005. 'Supplement to the National Building Code of Canada'. Association

Committee on the National Building Code, National Research Council of Canada,

Ottawa, Ont.

Norris, L. and Reynolds, W., 1975. 'Turbulent Channel Flow with a Moving Wavy

Boundary'. Report No. FM-10, Department of Mechanical Engineering, Stanford

University.

Okada, H., Ha, Y., 1992. 'Comparison of Wind-Tunnel and Full-Scale Pressure

Measurement Tests on Texas Tech Building'. Journal of Wind Engineering and Industrial

Aerodynamics, 41: 1601-1612.

161
Patankar, S., 1980. 'Numerical Heat Transfer and Fluid Flow'. McGraw-Hill, New York.

Paterson, D., 1986. 'Computation of Wind Flows over Three-Dimensional Buildings'.

Ph.D. Thesis, Department of Civil Engineering, Queensland, Astralia.

Pindado, S., and Meseguer, J. 2003. 'Wind-Tunnel Study on the Influence of Different

Parapets on the Roof Pressure Distribution of Low-Rise Buildings'. Journal of Wind

Engineering and Industrial Aerodynamics, 91: 1133-1139.

Richards, P., Hoxey, R., Connell, B., Lander, D., 2007. 'Wind-Tunnel Modeling of the

Silsoe Cube'. Journal of Wind Engineering and Industrial Aerodynamics, 95: 1384-1399.

Richardson, G., Surry, D., 1991. 'Comparison of Wind-Tunnel and Full-Scale Surface

Pressure Measurements on Low-Rise Pitched-Roof Buildings'. Journal of Wind

Engineering and Industrial Aerodynamics, 38: 249-256.

Rhie, C., Chow, W., 1983. 'Numerical Study of the Turbulent Flow Past an Airfoil with

Trailing Edge Separation'. Journal of American Institute of Aeronautics and

Astronautics, 21: 1525-1532.

Robertson, A., 1992. 'The Wind Induced Response of Full-Scale Portal Framed

Building'. Journal of Wind Engineering and Industrial Aerodynamics, 41:1677-1688.

162
Sadek, F., Simiu, E., 2002. 'Peak Non-Gaussian Wind Effects for Data-Base-Assisted

Low-Rise Building Design'. Journal of Engineering Mechanics, 5: 530-539.

Sedefian, L., Bennet, E., 1980. 'A Comparison of Turbulence Classification Schemes'.

Journal of Atmospheric Environment, 14: 741-750.

Selvam, R., 1992. 'Computation of Pressures on Texas Tech Building'. Journal of Wind

Engineering and Industrial Aerodynamics, 41: 1619-1627.

Selvam, R., 1997. 'Computation of Pressures on Texas Tech University Building Using

Large Eddy Simulation'. Journal of Wind Engineering and Industrial Aerodynamics, 67:

647-657.

Shih, T., Liou, W., Shabbir, A., Yang, Z., Zhu, J., 1995. 'A New k-s Eddy Viscosity

Model for High Reynolds Number Turbulent Flows- Model Development and

Validation'. Computers Fluids, 24: 227-238.

Sill, B., Cook, N., Fang, C., 1992. 'The Aylesbury Comparative Experiment: A Final

Report'. Journal of Wind Engineering and Industrial Aerodynamics, 41: 1553-1564.

Simiu, E., Scanlan, R., 1986. 'Wind Effects on Structures: An Introduction to Wind

Engineering'. Wiley-Interscience, New York, 3rd edition.

163
Snaebjornsson, J., 2002. 'Full- and Model Scale Study of Wind Effects on a Medium-

Rise Building in a Built Up Area'. PhD theses, Norwegian University of Science and

Technology, Norway.

Sockel, H., Taucher, R., 1980. 'The Influence of A Parapet on Local Pressure
th
Fluctuations'. Proceed of 4 coll. on Industrial Aerodynamics, W. Germany, 107-118.

Sparks, P., Schiff, S., Reinhold, J., 1994, 'Wind Damage to Envelops of Houses and

Consequent Insurance Losses'. Journal of Wind Engineering and Industrial

Aerodynamics, 53: 145-155.

Stathopoulos, T., 1979. 'Turbulent Wind Action on Low-Rise Buildings'. Ph. D. Thesis,

University of Western Ontario, Canada.

Stathopoulos, T., Surry, D., Davenport, A., 1981. 'Effective Wind Loads on Flat Roofs'.

Journal of Structural Division, ASCE, 107: 281-297.

Stathopoulos, T., 1981. 'Wind Pressure Functions for Flat Roofs'. Journal of Engineering

Mechanics Division, ASCE, 107: 889-905.

Stathopoulos, T., 1982. 'Wind Pressures on Low Buildings with Parapets'. Journal of the

Structural Division, ASCE, 108: 2723-2736.

164
Stathopoulos, T„ Surry, D., 1983. 'Scale Effects in Wind Tunnel Testing of Low-

Building'. Journal of Wind Engineering and Industrial Aerodynamics, 13: 313-326.

Stathopoulos, T., 1984. 'Design and Fabrication of a Wind Tunnel for Building

Aerodynamics'. Journal of Wind Engineering and Industrial Aerodynamics, 16: 361-376.

Stathopoulos, T., Baskaran, A., 1988-a. 'Wind Pressure on Flat Roofs with Parapets'.

Journal of Structural Engineering, ASCE, 113:2166-2180.

Stathopoulos, T., Baskaran, A., 1988-b. 'Turbulent Wind Loading of Roofs with Parapet

Configuration'. Canadian Journal of Civil Engineering, 15: 570, 578.

Stathopoulos, T., Baskaran, A., 1988-c. 'Roof Corner Wind Loads and Parapet

Configurations'. Journal of Wind Engineering and Industrial Aerodynamics, 29: 79-88.

Stathopoulos, T., Baskaran, A., 1990-a, 'Full Scale Measurements of Wind Pressures on

Flat Roof Corners'. Journal of Wind Engineering and Industrial Aerodynamics, 39: 1063-

1072.

Stathopoulos, T., Baskaran, A., 1990-b. 'Boundary Treatment for the Computational

Three-Dimensional Wind Flow Conditions Around a Building'. Journal of Wind

Engineering and Industrial Aerodynamics, 35: 177-200.

165
Stathopoulos, T., Zhou, Y., 1993. 'Numerical Simulation of Wind-Induced Pressures on

Buildings of Various Geometries'. Journal of Wind Engineering and Industrial

Aerodynamics, 46-47: 419-430.

Stathopoulos, T., Marathe, R., Wu, H., 1999. 'Mean Wind Pressures on Flat Roof

Corners Affected by Parapets: Field and Wind Tunnel Studies'. Journal of Wind

Engineering and Industrial Aerodynamics, 21: 629-638.

Stathopoulos, T., Saathoff, P., 1991. 'Wind Pressure on Roofs of Various Geometries'.

Journal of Wind Engineering and Industrial Aerodynamics, 38: 273-284.

Stathopoulos, T., Saathoff, P., Du, X. 2002-a. 'Wind Loads on Parapet'. Journal of Wind

Engineering and Industrial Aerodynamics, 90: 503-514.

Stathopoulos, T., Saathoff, P., Bedair, R. 2002-b. 'Wind Pressure on Parapets of Flat

Roofs'. Journal of Architectural Engineering, ASCE, 8: 49-54.

Stathopoulos, T., 2002-c. 'The Numerical Wind Tunnel for Industrial Aerodynamics:

Real or Virtual in the new Millennium?'. Wind and Structures, 5:193-208.

St. Pierre, L., Kopp, G., Surry, D., Ho, T., 2005. 'The UWO Contribution to the NIST

Aerodynamic Database for Wind Loads on Low Buildings: Part 2. Comparison of Data

with Provisions'. Journal of Wind Engineering and Industrial Aerodynamics. 93: 31-59.

166
Streeter, V., 1971. 'Fluid Mechanics'. 5th edition, McGraw-Hill, New York.

Suresh Kumar, K., 1998. 'Simulation of Fluctuating Wind Pressures on Low Building

Roofs'. Ph.D. thesis. Concordia University, Montreal, Canada.

Surry, D., 1989. 'Pressure Measurements on Texas Tech Building-II: Wind Tunnel

Measurements and Comparison with Full-Scale'. Proceed of the eighth coll. on Industrial

Aerodynamics and Wind Engineering, W. Germany.

Surry, D., Lin, J., 1995. 'The Effect of Surroundings and Roof Corner Geometric

Modifications on Roof Pressures on Low-Rise Buildings'. Journal of Wind Engineering

and Industrial Aerodynamics, 58: 113-138.

Tamura, T., et al., 1997. 'Numerical Prediction of Wind Loading on Buildings and

Structures—Activities of AIJ Cooperative Project on CFD'. Journal of Wind Engineering

and Industrial Aerodynamics, 67-68: 671-685.

Tieleman, H., Hajj, M., Reinhold, T., 1998. 'Wind Tunnel Simulation Requirements to

Assess Wind Loads on Low Rise Building'. Journal of Wind Engineering and Industrial

Aerodynamics, 74: 675-685.

Tieleman, H., 2003. 'Roughness Estimation for Wind-Load Simulation Experiments'.

Journal of Wind Engineering and Industrial Aerodynamics, 91: 1163-1173.

167
Tieleman, H., 2003. 'Wind Tunnel Simulation of Wind Loading on Low-Rise Structures:

A Review'. Journal of Wind Engineering and Industrial Aerodynamics, 91: 1627-1649.

Uematsu, Y., Isyumov, N., 1998. 'Peak Gust Pressures Acting on the Roof and Wall

Edges of a Low-Rise Building'. Journal of Wind Engineering and Industrial

Aerodynamics, 77-78: 217-231.

Uematsu, Y., Isyumov, N., 1999. 'Review: Wind Pressures Acting on Low-Rise

Building'. Journal of Wind Engineering and Industrial Aerodynamics, 82: 1-25.

Van Dormal, J., Raithby, G., 1984. 'Enhancement of the SIMPLE Method for Predicting

Incompressible Fluid Flow'. Journal of Heat transfer, 7: 147-163.

Versteeg, H., Malalasekera, W., 1995. 'An Introduction to Computational Fluid

Dynamics - The Finite Volume Method'. John Wiley and sons, New York.

Von Karman, T., 1948. 'Progress in the Statistical Theory of Turbulence. Proceeding of

National Academic Society, 34: 530-539.

Wagaman, S., Rainwater, K., Mehta, K., Ramsey, R., 2002. 'Full-Scale Flow

Visualization over a Low-Rise Building'. Journal of Wind Engineering and Industrial

Aerodynamics, 90:1-8.

168
Wang, Z., Plate, E., Rau, M., Keiser, R., 1996. 'Scale Effects in Wind Tunnel Modeling'.

Journal of Wind Engineering and Industrial Aerodynamics, 61: 113-130.

White, F., 1991. 'Viscous Fluid Flow'. 2 nd edition, McGraw-Hill, New York.

Wieringa, J., 1993. 'Representative Roughness Parameters for Homogeneous Terrain'.

Boundary-Layer Meteorol. 63:323-363.

Wu, F., Sarkar, P., Mehta, K., Zhoa, Z., 2001. 'Influence of Incident Wind Turbulence on

Pressure Fluctuations near Flat-Roof Corners'. Journal of Wind Engineering and

Industrial Aerodynamics, 89: 403-420.

Yakhot, V., et al., 1992. 'Development of Turbulence Models for Shear Flows by a

Double Expansion Technique'. Journal of physics Fluid, A4: 1510-1520.

Zhang, A., Gu, M., 2008. 'Wind Tunnel Tests and Numerical Simulations of Wind

Pressures on Buildings in Staggered Arrangement'. Journal of Wind Engineering and

Industrial Aerodynamics, 96: 2067-2079.

Zhou, Y., 1995. 'Numerical Evaluation of Wind Effects on Buildings'. Ph.D. Thesis,

Concordia University, Montreal, Canada.

169
 APPENDIXA

A.l Velocity and Turbulence Parameters

Mean wind speed ( U ) decreases with a decreasing distance from the ground, because of

frictional effect. A common representation of the wind profile above the earth's surface

is the 'power law\ which is employed as an empirical model for mean wind speed

profiles and is given by:

— r \ a

U z
A.l
u

where: z is the height above the mean ground level, zg is the gradient height, which is

equivalent to the top of the boundary layer and where the velocity is constant (gradient

velocity, Ug) and a is the power exponent parameter.

The boundary layer profile approaches zero speed at the ground surface and its shape

may be sensitive to the local roughness effect and the ''logarithmic law' may be applied:

( z \
U = — Cg In A.2
K

where: K is Von Karman's constant (0.4 - 0.42), Cg is the geostrophic drag coefficient

and z0 is the roughness length parameter.

170
The ' turbulence intensity' is a simple measure of the gustiness (turbulence) wind within a

specific time period, when the mean wind velocity was defined. The longitudinal

component can be defined as:

where: au is the standard deviation of a fluctuating wind velocity and U is the mean

along wind velocity. Another commonly used measure of the variability of the wind is

the 'gust factor' which is defined as:

where: is the largest maximum recorded wind velocity during a specific time

period.

A more detailed approach for the analysis of turbulence phenomena is by means of

'power spectral density function', which define the process in time rather than amplitude

domain. Spectrum of longitudinal turbulence (Sn) represents the distribution of turbulence

energy over a frequency range in which:

oo
a2 = |S(n)dn A.5

Spectra measurements are performed for the current flow conditions. The experimental

data has been compared with some empirical and analytical representations such as

Davenport's empirical equation (Davenport, 1961) and Von-Karman's analytical

expression (Von-Karman, 1948), respectively:

171
 nS<») _ 2
x = 1200— A.6
10
. 0 + x T

nS 4x _ nL,
A.7
\5/6 x
v 2
(l + 70.8X ) 2 : ~v z

where n is the frequency and x is a non-dimensional frequency, used to normalize the

standard spectra by the appropriate variance, a , with

0.35
25 ( z - d )
Lr = .0.063
(m) A.8

172
 0U/U U/Ug

(a) Terrain-1

30 r?
30 r O 1
o Experimental © experimental
/ i
0 1 -FSDU
25 ESDU
o 1 /°
20
o 1
20 1
/°
o \ Field T.I. /°
© 1 (0.17- Jo
' 15 /o
/©
012 o I
f
anemometer ht.
/©o
N
KT10 +
0.18 % \ (0.22-

/ roof ht. j y

0
0 0.1 0.2 0.3 0.4 0.! 0.0 0.2 0.4 0.6 0.8 1.0
Ou/U u/u„

(b) Terrain-2

Figure A.2: Mean velocity and turbulence intensity profiles for different exposures

173
 (c) Terrain-3

a u /U U/Ug

(d) Terrain-5

Figure A.2: Continued

174
A.2 Anemometer Calibration

The calibration consists of placing the anemometer in the wind tunnel and recording the

corresponding output in terms of volts for various speeds. Simultaneously, the wind

speed is measured using a hot-wire probe. Subsequently, a graph is plotted of volts vs.

wind speed in m/sec and the points are fitted by a straight line (Figure A.l).

14

y = 9.087x
R2 = 0.9999

0 0.2 0.4 0.6 0.8 1.2 1.4 1.6

volts

Figure A.l: Calibration of the anemometer used in Held testing

175
A.3 DSM System Setting and Operating Principle

Figure A.3 shows a diagram for installing the Digital Surface Module (DSM-3000)

system as presented by Wang, 2005. Two transducers are used, each of 64 scanning

channels. A host computer can control the DSM computer via Internet, while the DSM

computer connected to the Zero-Operate-Calibrate (ZOC-33) pressure scanner through

local cable connection. At the same time, the DSM computer controls a device named

DSM-CPM that organizes the working mode of the ZOC-33 pressure scanner. Figure

A.4 shows the ZOC-33 valve system and operating modes diagram. ZOC-33

incorporates integrated pneumatic valves that allow four working modes, namely, Purge,

Calibrate, Operate and Leak-Test. These modes are controlled by two pistons

respectively.

By setting the period in microseconds (|as) to 62 (as:

The scanning frequency = 1 / 6 4 channels x (62 x 10"6) sec ~ 250 Hz.

In wind tunnel testing strong wind is simulated, therefore a velocity scale of 1/2 is

assumed. For 1/50 length scale, the time scale is obtained as follows:

Tmod el ( Lmod el }( vmod el \

Tfiill-scale ^ ^ full-scale y y
vfitll-scale J

Corresponding to 10 minuets in full-scale, scanning time in wind tunnel = 24 sec.

Therefore, 6000 frame / scan are obtained (250 Hz x 24 sec).

176
 Static P

Valve

Figure A.3: DSM-3000 system diagram (Wang, 2005)

177
 CONTROL AIR SUPPLY
-TEST PRESSURE INPUTS- -Px VALVE TYP.
EXHAUST
P2 P3 P4
Px CONTROL
SOLENOID VALVE

CAL CONTROL
SOLENOID VALVE

PISTON TYP.

VOLT SIGNAL OUTPUT FIELD REPLACEABLE

PRESSURE SENSOR

Figure A.4: ZOC-33 system diagram (Wang, 2005)

178
A.5 Gumbel Plot

Gumbel's plot (Gumbel, 1958) is used to determine the probability term P as follows:

ln[-ln(JP)]=-t—^ (A. 10)

which can be written as:

x = m + sy (A.ll)

where, y = - ln[- ln(P)] is the reduced y-variant, m is the mode and s is the dispersion.

The observed extreme values are sorted into ascending order of magnitude, after which is

assigned a rank, r, where r - 1 for the smallest and r = n for the largest of n values. An

estimate of P corresponding to each extreme value can be calculated from their ranks as

follows:

P = -L7 (A. 12)

n +1

The mode and dispersion can be determined from plotting equation A . l l . Mode

corresponds to the x value when the reduced ^-variant is zero and the dispersion is the

slope of the line fitting the data (Suresh Kumar, 1998).

179
 APPENDIX B

B.l Wind Speed and Turbulence Intensity Profiles (ESDU, 1983)

The ESDU (82026) equilibrium mean wind speed and turbulence intensity model has

three scaling parameters: roughness height z0, friction velocity, u* and the gradient height

zg. The velocity profile can be calculated as follows:

/ \ r \2 r \3 f
23_z__15 1
i{z) = 2.5u * In +- + -3 (B.1)
\ Z
0 J 4 zg 8 J vz? J \zs J

This equation is valid up to zg in which:

u*
z
s = (B.2)
6/c

where fc is the Coriolis parameter, which depends on earth's self rotation and latitude.

Eq. B.l can be simplified for speed profile up to 300 m as follows:

—
/ \
z 34.5 fcz
u (z)= 2.5m* In +- (B.3)
U J M*

The turbulence intensity can be obtained as follows:

f u* \
[u'(z
/ = (B.4)
u(z) { u* J Ku(z))

where: u ' (z) is fluctuation component of the wind velocity and:

u'(z) _ 7.57[0.538 + 0.091n(z/z o )]' ? " i

(B.5)
u* l + 0.1561n(u*/(/ c zJ)

6/cz
where: rj = 1
u *

180
B.2 Minimum Design Loads for Buildings and Other Structures,

[This Section has been taken verbatim from ASCE 7-05]

Parapets: (Analysis Procedures)

B.2.1 Main wind force-resisting system

The design wind pressure for the effect of parapets on main wind force-resisting systems

of rigid, low-rise or flexible buildings with flat, gable, or hip roofs shall be determined by

the following equation:

pP = qP GCpn (Ib/sj) (B.6)

where

pp = combined net pressure on the parapet due to the combination of the net

pressures from the front and back parapet surfaces. Plus (and minus) signs

signify net pressure acting toward (and away from) the front (exterior)side

of the parapet.

qp = velocity pressure evaluated at the top of the parapet.

GCpn = combined net pressure coefficient;

= + 1.5 for windward parapet

= - 1.0 for leeward parapet.

181
B.2.2 Components and cladding

The design wind pressure on the components and claddings elements of parapets shall be

designed by the following equation:

p = qp{GCp-GCpi) (B.7)

where

qp — velocity pressure evaluated at the top of the parapet.

GCP = external pressure coefficient from Figure B.l.

GCPi = internal pressure coefficient from Table B.l, based on the porosity of the

parapet envelop.

Two load cases shall be considered:

Load Case A shall consist of applying the applicable positive wall pressure from Figure

B.l to the front surface of the parapet while applying the applicable negative edge or

corner zone roof pressure from Figure B.2 to the back surface.

Load Case B shall consist of applying the applicable positive wall pressure from Figure

B.l to the back of the parapet surface, and applying the applicable negative wall pressure

from Figure B.l to the front surface.

Edge and corner zones shall be arranged as shown in Figures B.l and B.2. GCP shall be

determined for appropriate roof angle and effective wind areas from Figures B.l and B.2.

if internal pressure is present, both load cases should be evaluated under positive and

negative internal pressure.

182
 Table B.l Infernal Pressure coefficient, GCP-,

Enclosure Classification GC„,

Open Buildings 0.00

Partially Enclosed Buildings + 0.55 or - 0.55

Enclosed Buildings +0.18 or -0.18

Notes:

1. Plus and minus signs signify pressures acting toward and

away from the internal surfaces, respectively.
2. Tow cases shall be considered to determine the critical
load requirements for the appropriate condition:
(i) a positive value of GCPi applied to all internal surfaces
(ii) a negative value of GCPi applied to all internal surfaces

183
 10 500
a. •u
O 1 *
a (5) 1 i
•1.4 •1.4
42
<•1.1
o -1,0
•0.6 -0,S

o
s •0.6
•JX4
•0.2
<s 0
<0
« +0,2
.0.4
<© -0.6
c +0.7
fO.B
Q>
"S2 +1.0 +1.0
in
1 10 20 50 100 200 5001000
(0.1) (0.9) (1.9) (4.S) (9.3} (18.6)
Effective Wind Area, ft 2 ( r f )

Notes:

1. Vertical scale denotes GCP to be used with

2. Horizontal scale denotes effective wind area, in square feet (square meters).
3. Plus and minus signs signify pressures acting toward and away from the surfaces,
respectively.
4. Each component shall be designed for maximum positive and negative pressures.
5. Values of GCP for walls shall be reduced by 10% when 0< 10°.
6. Notation:
a : 10% of least horizontal dimension or 0.4/?,whichever is smaller, but not less
than either 4% of least horizontal dimension or 3 ft (0.9 m).
h : Mean roof height, in feet (meters), except that eave height shall be used for
6 < 10°.
6 : Angle of plane of roof from horizontal, in degrees.

Figure B.l External pressure coefficients, GCpf (Walls)

184
 & 10 108
i i ! 11 <5 s 1
•3.0
! L ; Root i •n
•3.0
€6 to Overhang iI
g. ! i •26
to
-J6
i\ !
•Zt
i ^ j ; \ «

if !
-"V— .......
•2.0
57 1 r — -2,0
1.6 rs. j •1.8
1
-17
•1.6
i I Sj"
o -l.fl i I •

"Xtr
•1.2
"1 - .. .. .....
•1.2
•1.0
-1.1
-1.0
OH •ftj -98
i i •at
\
•ftO i ;
i
1 10 20 50 100 200 5G01020
42 „ „.„ f HI) [yf (IS! <«J {}.3| (tsj) i«5){SS)
0 i i
-<SXI)*£i> ' L i Effective Wind Area, ft2 [ d )

*M i
1 10 20 50 100 200 500 1 000
pi) i9) (U| «6) rn psi) P5HKJ]
Effective Wind Area, It 1 {nr?)

Notes:

1. Vertical scale denotes GCP to be used with qh-

2. Horizontal scale denotes effective wind area, in square feet (square meters).
3. Plus and minus signs signify pressures acting toward and away from the surfaces,
respectively.
4. Each component shall be designed for maximum positive and negative pressures.
5. If a parapet equal to or higher than 3 ft (0.9 m) is provided around the perimeter of
the roof with 9 < 7°, Zone 3 shall be treated as Zone 2.
6. Values of GCP for roof overhangs include pressure contributions from both upper
and lower surfaces.
7. Notation:
a : 10 percent of least horizontal dimension or OAh, whichever is smaller, but not
less than either 4% of least horizontal dimension or 3 ft (0.9 m).
h : Eave height shall be used for 9 < 10°.
e : Angle of plane of roof from horizontal, in degrees.

Figure B.2 : External pressure coefficients, GCP (Roofs)

185