You are on page 1of 181

CRANFIELD UNIVERSITY

Taofiq AMOLOYE

Wind Engineering: A Review of the Eurocode provisions for the


Wind Loading on Low-rise Buildings

SCHOOL OF ENGINEERING
MSc Aerospace Dynamics

MSc THESIS
Academic Year: 2011 - 2012

Supervisors: J. C. Holt
K. P. Gary
September 2012
CRANFIELD UNIVERSITY

SCHOOL OF ENGINEERING
MSc Aerospace Dynamics

MSc THESIS

Academic Year 2011 - 2012

Taofiq AMOLOYE

Wind Engineering: A Review of the Eurocode provisions for the


Wind Loading on Low-rise Buildings

Supervisors: J. C. Holt
K. P. Gary
September 2012

This thesis is submitted in partial fulfilment of the requirements for


the degree of Master of Science

© Cranfield University 2012. All rights reserved. No part of this


publication may be reproduced without the written permission of the
copyright owner.
ABSTRACT
Building codes such as the Eurocode have usually been used as a cheaper
alternative to wind tunnel studies in the consideration of wind loading on a
structure. It is often the case that very tall buildings and large structures have
enough economic justification for expensive wind tunnel studies in their design
stage. Such wind tunnel studies, as per state-of-the-art, feature simultaneous
scanning of and acquisition of loading data from hundreds of pressure tappings
with subsequent high-speed computer data processing and analysis. This is not
the case for low-rise buildings which do not find their way into the wind tunnel
except in the case where they are unusual edifices. Low-rise buildings,
however, are the most damaged in wind storms. In addition, in the present
times, their shapes are increasingly losing touch with the traditional and generic
forms dealt with in the Eurocode. Therefore, the question is: How well does the
Eurocode, which was put together with information from wind tunnel studies
performed in the and using currently outdated data acquisition
techniques, deal with present building shapes? The study was based on models
of a simple cuboidal building; a quasi-rectangular building with inset faces in its
plan; and a building plan featuring a re-entrant corner possessing curved
surfaces at the internal and external junctions of its wings. It was concluded
from the results of the study that adapting the Eurocode wind loading provisions
to irregular building plans characteristic of modern times gives very unsafe
solutions. The variations of pressure with wind direction on the internal walls of
the wings of and the curved surface at the internal junction of the re-entrant
corner were observed to follow coherent wave forms which are mutually similar.
These call for further research.

Keywords:

Cuboidal Irregular Pressure Tapping Tributary

I
II
DEDICATION
This work is dedicated, wholeheartedly, to God, The Almighty Lord and Master
of the wind Who says:

“And among His Signs is this, that He sends the winds as glad tidings, giving
you a taste of His Mercy, and that the ships may sail at His Command, and that
you may seek of His Bounty, in order that you may be thankful.” [Translation
from the Arabic text]

Al-Qur’an, the Romans (Chapter 30): 46

III
ACKNOWLEDGEMENTS
I express much gratitude to my immediate supervisor, Jenny Holt. I thank her
not only for conceiving the idea of the subject of this thesis which I have found
to be very interesting but also for her time, the right pieces of advice and help
which she offered me on its completion. The contribution of my co-supervisor,
Professor K. P. Gary, is also worth mentioning and I thank him a lot as well.

I always admired the works of Paul Dancer, the model maker. I am grateful to
him for taking the time to make my models accurately (as they can get in the
School of Engineering workshop) at such small scales and with such nice finish.

I have to be grateful to the staff of the Estate services of Cranfield University


who provided me with the geometrical information on the buildings studied in
this work.

Many thanks and acknowledgements are also in order to the editors of and
contributors to the Journal of Wind Engineering and Industrial Aerodynamics
without which I would have found much difficulty in developing my thoughts and
ideas on the project.

I greatly appreciate the efforts which Mateuz Pintar must have put into getting
the Computational Fluid Dynamics results used in this study. I have to thank
him not only for his direct and indirect inputs into this thesis (which came as
soon as I asked) but also for being a peaceful and very nice flatmate. The rest
of my flatmates also fall in this rank and I really thank them a lot.

I would be lacking hindsight if I did not acknowledge the contributions of my


teachers and lecturers, both formal and informal, so far at Cranfield and
elsewhere who have contributed immensely to my general knowledge and, in
particular, to my knowledge of Aerodynamics. Of notable mention are Dr K.
Bofah of Kwame Nkrumah University of Science and Technology in Kumasi,
Ghana and John D. Anderson, Jr. of the Smithsonian Institution, Washington,
DC, United states. The former introduced me thoroughly, in the pursuit of my
Bachelor’s degree, to the ‘Fundamentals of Aerodynamics’ using one of the

V
numerous well-known texts written by the latter. I have since admired the writing
style of John D. Anderson and it informed the style I adopted for this thesis.

I must appreciate the insight my internship with the Aerodynamics group at


Volvo Car Corporation in Sweden has afforded me into the industrial practice of
the discipline.

I have found that some of the best informal teachers are one’s classmates and I
appreciate well whatever contributions my classmates have made to my
educational development.

I have to acknowledge the contributions of the general staff of my Department,


the Library and the IT department. In particular, I found the thesis template
provided by the IT department to be highly beneficial to this write-up. In a similar
vein, the ease of referencing with Refworks is worthy of mention.

I would be too ungrateful not to acknowledge my friend(s) who may have been
on the lookout for this paragraph. I have your name(s) ingrained in my heart and
I thank you for the friendship we share which I hold dearly.

I am indebted to my sister, Dr. Mrs. Azizat Amoloye-Adebayo, who showed me


the dividing line between ‘story-telling’ and academic report writing with her
comments on this write-up. I am even much more indebted to her as well as to
my other siblings and family (biological and adopted), too numerous to mention
here, for their unconditional and unending love and prayers. It would amount to
another thesis write-up to effectively express my thanks to everyone. However,
naturally, Taofiqah Amoloye, Hikmah Amoloye and Mubarak Amoloye also
deserve unconditional mention as much as my love and appreciation. Also, it
behoves me to mention my fraternal love and admiration for all the members of
the Cranfield Islamic Society who made Cranfield University a home-away-from-
home for me especially in the month of Ramadan.

Words cannot begin to express my gratitude, appreciation, love and adoration


for my mother, Dr. Mrs Bilqis Amoloye, and my late father, Dr. Abduljelil
Amoloye. I had a dream when I was very young and I am living it now thanks to
your belief and support. In a similar vein, I remain forever indebted to my

VI
godparents in their numbers who, indeed, have chosen to remain friends of the
family as much in my father’s demise as in his lifetime.

At the completion of my programme at Cranfield, I have found that I am better


poised to look, critically, into engineering problems and suggest solutions even
when they are not immediately relevant to my field. I recognise this as a favour
and blessing of God, who, I believe, was not only my general Supervisor for this
work but also, the Only Supervisor I have known for all my affairs. I thank Him a
lot and I submit that to Him belongs the Ultimate acknowledgement and glory.

Alhamdulillah!

VII
TABLE OF CONTENTS
ABSTRACT ......................................................................................................... i
DEDICATION ..................................................................................................... iii
ACKNOWLEDGEMENTS................................................................................... v
LIST OF FIGURES ............................................................................................. xi
LIST OF TABLES .............................................................................................. xv
LIST OF EQUATIONS ..................................................................................... xvii
NOTATIONS .................................................................................................. xviii
1 Introduction...................................................................................................... 1
1.1 Wind Engineering in Historical Perspective .............................................. 1
1.2 Motivation for the study ............................................................................. 4
1.3 Aim and Objectives ................................................................................... 7
1.4 Scope and Limitations............................................................................... 8
2 Background Studies and Review of Literature ............................................... 11
2.1 Atmospheric Wind ................................................................................... 12
2.1.1 The Atmospheric Boundary Layer .................................................... 13
2.1.2 Terrain Roughness Categorization ................................................... 16
2.1.3 Turbulence ....................................................................................... 17
2.1.4 Mean Wind Velocity profile ............................................................... 20
2.1.5 Wind Data and Wind Rose ............................................................... 21
2.2 Building Aerodynamics ........................................................................... 23
2.2.1 Tall Buildings .................................................................................... 31
2.2.2 Low-rise Buildings ............................................................................ 34
2.3 Wind loading ........................................................................................... 39
2.3.1 Loading Coefficients ......................................................................... 41
2.4 Building Codes: The Eurocode ............................................................... 45
2.4.1 Scope of the Eurocode ..................................................................... 49
2.5 Principles of Wind-tunnel Experiments ................................................... 50
2.5.1 Field of Application of Wind tunnel Experiments .............................. 50
2.5.2 Guidelines to Wind tunnel Experiments............................................ 51
2.6 Computational Wind Engineering ........................................................... 54
2.7 Relevant past works on the present study .............................................. 55
2.8 Chapter Summary ................................................................................... 59
3 Research Methodology ................................................................................. 63
3.1 Case Studies .......................................................................................... 63
3.2 Models construction and Pressure tapping locations .............................. 70
3.3 Research methodology ........................................................................... 73
3.4 Chapter Summary ................................................................................... 74
4 Experiment .................................................................................................... 75
4.1 The Cranfield ABL Wind Tunnel ............................................................. 75
4.2 Simulation of the Cranfield Atmospheric Boundary Layer on the models 76

IX
4.3 The Procedure ........................................................................................ 79
4.4 Error Analysis ......................................................................................... 81
4.5 Chapter Summary ................................................................................... 85
5 Results and Discussion ................................................................................. 87
5.1 The Sports Hall ....................................................................................... 89
5.2 Building 83, Hangar Building ................................................................... 98
5.3 The Trafficmaster Building .................................................................... 103
5.4 Effects of Tributary Area variation ......................................................... 111
5.5 Summary of Observations .................................................................... 116
6 Conclusion................................................................................................... 119
6.1 Project Summary .................................................................................. 119
6.2 Concluding Remarks............................................................................. 120
6.3 Recommendations ................................................................................ 122
REFERENCES ............................................................................................... 123
Appendix A Principles of Wind tunnel Experiments in detail ....................... 129
Appendix B Building Models Dimensions and Pressure tapping location ... 146

X
LIST OF FIGURES
Figure 1-1 The Gherkin in London [5] ................................................................. 7
Figure 2-1The mean ABL wind velocity profile over different terrains [25] ....... 14
Figure 2-2 Wind speeds at 3 heights during a gale [40] ................................... 15
Figure 2-3 Wieringa terrain roughness classification [54] ................................. 17
Figure 2-4Spectrum of Wind speed variation within the Atmospheric boundary
layer [48] ................................................................................................... 20
Figure 2-5 Cranfield Wind rose [12] .................................................................. 23
Figure 2-6 Flow around a rectangular body with a. a short along-wind length
[14] b. with a long along-wind length [40] ................................................ 24
Figure 2-7 a. Industrial chimney [6] b. Tree Canopy at Milton Keynes Coachway
(picture taken by author) c. Rendering of the Millenium tower in Abuja
(under construction) [2] ............................................................................. 25
Figure 2-8 Non-dimensional power spectrum of longitudinal turbulence at model
and full scale [46] ...................................................................................... 28
Figure 2-9 a. Computer rendered image of Burj dubai and b. A real picture of
the building during construction (adapted from [17]) ................................. 32
Figure 2-10 Wind Behaviour of Burj Dubai [17] ................................................ 32
Figure 2-11 Flow around a tall building [40]...................................................... 33
Figure 2-12 Flow around a Low-rise building [40]............................................. 35
Figure 2-13 Flow around buildings with Roof pitches of 20-30 degrees [14] .... 36
Figure 2-14 Corner vortices in an obliquely approaching wind [40] .................. 37
Figure 2-15 The "Alan G. Davenport's Wind loading chain" [48] ...................... 40
Figure 2-16 Regions of validity of Bernoulli Equation in the flow around a
rectangular bluff body [14] ......................................................................... 41
Figure 2-17 Cp (t) on low-rise buildings [40] ..................................................... 43
Figure 2-18 Resolution of forces on a building [27] .......................................... 44
Figure 2-19 Division of walls and roof surfaces into zones in the Eurocode
based on the building dimensions and expected nature of flow around the
building (adapted from [15]) ...................................................................... 47
Figure 2-20 Pressure tappings with tributary areas shown a. Uniform distribution
b. Bias distribution [27] .............................................................................. 53
Figure 2-21 Outline of the test building used by Geurts [35] ............................ 55

XI
Figure 2-22 Aylanc's and Kasperki's building model [13]. The dash lines on the
model signify the tapping positions. .......................................................... 57
Figure 2-23 Lin's and Surry's building models’ dimensions [52] ....................... 58
Figure 2-24 Simulation of a Westerly wind over the Cranfield University Sports
Hall (signified by S H here) (adapted from [53]) ........................................ 59
Figure 3-1 An aerial view of a section of the London city centre -The Gherkin is
marked 'A' [1] ............................................................................................ 64
Figure 3-2 Irregular building plans a. Re-entrant corners b. Recessed bays c.
Central wells d. Inset storeys (adapted from [27]) ..................................... 65
Figure 3-3 Building form selection from Cranfield University Campus [7] ......... 67
Figure 3-4The Sports Hall viewed from the South-West (picture taken by the
current author)........................................................................................... 68
Figure 3-5The Hangar building (in the fore-ground) viewed from the South-West
(picture taken by the current author) ......................................................... 69
Figure 3-6The Trafficmaster building viewed from the South-West (picture taken
by the current author during a flight test) ................................................... 69
Figure 3-7 Wind tunnel models of a. Hangar building and b. Trafficmaster
building ...................................................................................................... 70
Figure 3-8 Constructed 1:200 scale Aerodynamic models of the case study
buildings fitted with pressure tappings ...................................................... 70
Figure 3-9 An elevation view of the Hangar building model focusing on one of its
inset faces (the approaching wind is assumed to be flowing into the present
page. Its directions of flow at this corner of interest are signified by the
green arrows) ............................................................................................ 73
Figure 4-1 Schematic representation of the ABL wind tunnel facility at CU [54]75
Figure 4-2 Isometric view of the test section [12] ............................................. 76
Figure 4-3 Set up of the wind tunnel for simulating the Cranfield ABL ............. 77
Figure 4-4 Comparison between the wind tunnel data and ESDU predictions for
wind velocity profile (V P) and turbulence profile (T P) .............................. 77
Figure 4-5 The Setup of the Sports hall and inset a. the Trafficmaster building b.
the Hangar building showing the Aluminium tape and the Pitot-static tube78
Figure 4-6 Tubes from pressure taps fitted on to the transducers located under
the turntable .............................................................................................. 78
Figure 4-7 Orientation of the buildings relative to North (adapted from [7]) ...... 80
Figure 4-8 Repeatability analysis of the transducers ........................................ 84

XII
Figure 5-1 Matlab generated contours of surface Cp for a westerly wind on the
West face of the SH model based on a. mean wind velocity at different
heights and b. mean wind velocity at 50mm.............................................. 87
Figure 5-2A westerly (270 degrees) wind flow over the Sports Hall model ...... 90
Figure 5-3 A SW (225 degrees) wind flow over the Sports Hall model ............. 91
Figure 5-4 Comparison between EN and EMV on the roof of the Sports Hall
model for W and SW wind ......................................................................... 92
Figure 5-5 Comparison between EN and EMV on the South face of the Sports
Hall model for W and SW wind.................................................................. 93
Figure 5-6 Comparison between EN and EMV on the West face of the Sports
Hall model for W and SW wind.................................................................. 94
Figure 5-7 Variation of Cp with wind direction for pressure taps 1 and 6 on the
West face of the SH model........................................................................ 95
Figure 5-8 CFD prediction of surface pressure coefficient on the Sports Hall .. 97
Figure 5-9 Comparison between a. experiment surface pressures and b. CFD
predictions of the surface pressures on the roof of the Sports Hall ........... 98
Figure 5-10 Pictorial representation of a Westerly and a South-westerly wind
flow over the Hangar building .................................................................... 99
Figure 5-11 A South-Westerly flow over the HB model .................................... 99
Figure 5-12 A Westerly flow over the HB model ............................................. 100
Figure 5-13 The HB model roof divided into Zones using the Eurocode ........ 101
Figure 5-14Key figure for the dimensions and pressure tapping locations on the
Trafficmaster (TM) building model........................................................... 103
Figure 5-15 Flow around the TM model for a a. Westerly wind and b. a South-
Westerly wind .......................................................................................... 104
Figure 5-16 Countour of Cp for a Westerly wind flow over the TM model
(Faces1,6 and 7) ..................................................................................... 104
Figure 5-17 Countour of Cp for a Westerly and South-Westerly wind flow over
the TM model (Faces 4and 5) ................................................................. 105
Figure 5-18 Graph of Cp variation with wind direction for pressure tappings 20,
83 and 89 on the TM model .................................................................... 106
Figure 5-19 Graph of Cp variation with wind direction for pressure tappings 14,
83 and 95 on the TM model .................................................................... 107
Figure 5-20 Graph of Cp variation with wind direction for pressure tappings 35
and 47 on the roof of the TM model ........................................................ 108

XIII
Figure 5-21 Countour of Cp for a South-Westerly wind flow over the TM model
(Faces 3 – the roof) ................................................................................. 109
Figure 5-22 Tributary area variation on the roof of the SH model for 8 different
configurations .......................................................................................... 111
Figure 5-23 Effect of tributary area variation on the wind loading on the roof of
the Sports Hall model .............................................................................. 112
Figure 5-24 Tributary Area distribution over Face 5 of TM building for
Configurations 1-4 ................................................................................... 113
Figure 5-25 Effect of tributary area variation on the wind loading on Face5 of the
TM model ................................................................................................ 115

Figure A-1 Roughness, Barrier and Mixing devices method of simulating the
ABL [28] .................................................................................................. 133
Figure A-2 Hot-wire probe [24] ....................................................................... 141
Figure A-3 Multi-hole probe [24] ..................................................................... 141
Figure A-4 Schematic of a 4-hole cobra probe [24] ........................................ 142
Figure A-5 Single point pressure measurements a. Short tube b. Restricted tube
c. Leaked tube [40] .................................................................................. 143
Figure B-1 Key figure for the dimensions and pressure tapping locations on the
Sports Hall (SH) model ........................................................................... 146
Figure B-2 The West face of the SH model .................................................... 147
Figure B-3 The South face of the SH model ................................................... 147
Figure B-4 The Roof of the SH model ............................................................ 148
Figure B-5 Face 1 of TM model ...................................................................... 148
Figure B-6 Face 2 of TM model ...................................................................... 149
Figure B-7 Face 3 (the roof) of TM model ...................................................... 150
Figure B-8 Face 4 of TM model ...................................................................... 151
Figure B-9 Face 5 of TM model ...................................................................... 151
Figure B-10 Face 6 of TM model .................................................................... 152
Figure B-11 Face 7 of TM model .................................................................... 152
Figure B-12 Face 8 of TM model .................................................................... 153
Figure B-13 The Hangar building (HB) model plan dimension and tapping
location .................................................................................................... 154

XIV
LIST OF TABLES
Table 4-1 Important parameters at wind tunnel scale and full scale ................. 80
Table 5-1 Comparison between EN and EPV for the Sports Hall ..................... 96
Table 5-2 Comparison between EN and EPV on the Hangar building ........... 102
Table 5-3 Comparison between EN and EPV for Face4 of the TM model in a
westerly wind and south-westerly flow .................................................... 105
Table 5-4Comparison between EN and EPV on the roof of the west wing of the
TM model for a south-westerly wind ........................................................ 110
Table 5-5Number of tappings in each of the configurations used for tributary
area variation assessment ...................................................................... 115

XV
LIST OF EQUATIONS
(2-1) .................................................................................................................. 18
(2-2) .................................................................................................................. 18
(2-3) .................................................................................................................. 18
(2-4) .................................................................................................................. 18
(2-5) .................................................................................................................. 18
(2-6) .................................................................................................................. 20
(2-7) .................................................................................................................. 21
(2-8) .................................................................................................................. 21
(2-9) .................................................................................................................. 26
(2-10) ................................................................................................................ 42
(2-11) ................................................................................................................ 43
(4-1) .................................................................................................................. 84
(5-1) .................................................................................................................. 88
(5-2) .................................................................................................................. 88
(A-1) ............................................................................................................... 136
(A-2) ............................................................................................................... 137
(A-3) ............................................................................................................... 137
(A-4) ............................................................................................................... 137
(A-5) ............................................................................................................... 137
(A-6) ............................................................................................................... 138
(A-7) ............................................................................................................... 138
(A-8) ............................................................................................................... 138

XVII
NOTATIONS
Area

Reference Area

Width of the structure (the length of the surface perpendicular to the wind

direction if not otherwise specified)

Force coefficient

Pressure coefficient

External pressure coefficient

Depth of structure (the height of the surface parallel to the wind direction

if not otherwise specified)

Zero plane displacement

Young’s modulus for a structural material

Force

Frequency in the wind tunnel

Frequency at full scale

Shear modulus for a structural material

Acceleration due to gravity

Height of structure

Reference height

Longitudinal turbulence intensity

Lateral turbulence intensity

XVIII
Vertical turbulence intensity

Turbulence intensity in the wind tunnel

Turbulence intensity at full scale

Von karman constant

Characteristic length of a structure

Characteristic length in the wind tunnel

Characteristic length at full scale

Length scale of longitudinal turbulence

Length scale of lateral turbulence

Length scale of vertical turbulence

Natural frequency

Free stream static pressure

Stagnation pressure

Static pressure at a point

Velocity pressure at height above ground

Velocity pressure at a height

Power spectrum of longitudinal turbulence velocity

Period of time

Time in the wind tunnel

Time at full scale

Velocity

Velocity at height above ground

XIX
Shear velocity or friction velocity

̅ Mean wind speed

Velocity in the wind tunnel

Velocity at full scale

Viscosity of air

Height above ground

Roughness length

Roughness length in the wind tunnel

Roughness length at full scale

Reference height above ground at which the power and log laws are

Matched

*
Longitudinal integral scale

Exponent used to match the power and log laws

Boundary layer height

Density of a structure

Density of air

Structural damping ratio

Geometric scale

Velocity scale

Time scale

Frequency scale

XX
Standard deviation of longitudinal turbulence

Variance of longitudinal turbulence

Standard deviation of lateral turbulence

Standard deviation of vertical turbulence

Jensen number

Reynolds number

Strouhal number

Strouhal number in the wind tunnel

Strouhal number at full scale

XXI
1 Introduction

“This bountiful activity, the evolution of wind culture, and the capability of mankind to exploit
beneficial wind aspects and protect itself from the prejudicial ones, may be framed in four
periods (Cermak, 1975)”[57 p814]

The provisions for the design of buildings against the wind as contained in the
Eurocode as well as other building design codes have often been the only basis
by which designers account for wind effects on their structures. With an ever-
dynamic wind and an equally changing world, it is apposite to review the codes
whenever it is suspected that history may have taken its toll on them.

The motivation for this study is discussed in Section 1.2. It is better appreciated
with a brief history of wind engineering which is presented in Section1.1.
Subsequently, the aim and objectives of the study as well as the scope and
limitations are discussed.

The background studies including the review of relevant past works is


presented in Section 2 followed by the Research Methodology adopted (Section
3), the Experiment performed (Section 4), the Analysis of the Results (section5)
and Conclusion (Section 6) in that respect.

1.1 Wind Engineering in Historical Perspective


Solari [57 p814] could not be more right when he notes that out of all natural
phenomena known to man, the wind has attracted much more than a fair share
of “speculation, observation, experience, research, applications and
technologies” in the long course of human recorded history. His paper offers a
good insight into the creation of the International Association of Wind Engineers
(IAWE); an association borne out of earlier international and national

1
conferences on industrial aerodynamics and wind engineering [25][48]. At one
of these conferences chaired by Professor Jack Cermak, the christening and
official labelling of the then new discipline as wind engineering was adopted
[48]. Stathopoulos [58 p1] describes wind engineering citing Cermak ( ) “as
the rational treatment of interaction between wind in the atmospheric boundary
layer and man and his works on the surface of earth.” He notes that it requires
the synthesis of a far reaching number of other earlier disciplines ranging from
fluid mechanics, meteorology, structural mechanics, and physiology to
aerodynamics.

The history of wind engineering can be generally grouped (for discussion


purposes only) into five periods of time [18] or four periods [57] or to cover just
two periods: past and present [25]. Baker [18] analysed the development of
wind engineering in the context of the prevailing socio-political (which may be
viewed from economic perspectives) and intellectual conditions of specific times
and places. On this basis, he also makes a theoretical enquiry into the future of
the discipline.

As such, he highlighted the “traditional” period (up to AD) in which people


held very crude superstitious views of the wind; the “empirical” period
( – ) in which the greater awareness about science and the industrial
revolution was born; the “establishment” period ( – ) which saw the
two World Wars and, subsequently, experiments that provided information
contained in building codes; “the period of growth” ( – ) which saw the
High-rise structure construction boom and the development of new building
codes as well as the revision of existing ones; and “the modern period” (
onwards), now experiencing burgeoning computer technology growth.

Baker [18] who presented his paper in maintains that the future of wind
engineering can also be foretold by a cue from the pre-dominant socio-political
and intellectual conditions of the world as we know it now. He expects these
contexts to be different for different parts of the world, naturally, with the focus
of wind engineering likely to be on sustaining and maintaining the already built
environment and the development of sustainable energy sources in the

2
developed world. Nonetheless, since the phenomenon of climate change is
global, he posits that it will be of importance on the frequency and intensity of
wind storms for all parts of the world.

The importance of computing power is not going to wane down and in time, time
series analysis of wind loading may be able to displace some of the canonical
statistical and frequency domain methods. It can be expected that
Computational Fluid Dynamics (CFD) techniques will ever be improving and
that there may be a decline in the use of Atmospheric Boundary Layer (ABL)
wind tunnels. Applications of CFD, however, may only be wide spread in
assessing pedestrian comfort rather than in obtaining surface pressure
distributions on structures. There is still the need of further dedicated full scale
studies to validate loading data in codes, wind tunnel data and CFD results [18].

Baker [18] has suggested that the highlighted areas of possible development
will lead to another round of code revision in the next couple of decades or so.
Again, he is optimistic that different types of codes might be developed that will
facilitate load calculations in time domain using standard wind and loading
datasets. He contends that the education sector will have to play quite a pivotal
role in seeking out new ways such as the development of multi-disciplinary
research teams to obtain research funding needed to facilitate and sustain the
above developments and the survival of wind engineering within universities.

He concludes by noting the active and vibrant nature of wind engineering which
needs much further research and development. He also recalls the basic
justification of the discipline which is “to help societies to cope with one of the
most destructive forces of nature” [18 p867]. At the time of writing, however, it is
observable that the echoes of austerity measures in national macro-economic
policies are resonating all across Europe and much of the rest of the world
amidst the crippling financial crisis [44]. It remains to be seen what this holds for
the future of wind engineering.

3
1.2 Motivation for the study
A number of prominent practitioners, researchers, consultants and experts in
the wind engineering discipline such as Peter Irwin [46] , Leighton Cochran [25],
Theodore Stathopoulos [58], John Holmes [39], N J Cook [29], D Boggs [22]
and C J Baker [18] have all expressed concerns about the adequacy of the
building design codes to cope with predicting the wind loading on buildings to
an acceptable match with reality.

It is customary in the design stage to have wind tunnel studies on models of


large structures such as tall buildings, stadia, arenas and long span bridges and
results from these project-specific studies invariably contribute to the database
of the knowledge of wind loading of structures [46]. The vast majority of
structures are low-rise buildings which often do not have any economic
justification of being designed from ‘project-specific-wind-tunnel-studied’ models
unless they are of unusual shapes representative of the ever imaginative minds
of the modern architects [25];[46]. Designers of this category of buildings
usually, then, have to resort to the crude wind loading information in codes
which are in much need of concerted reviewing and updating [46];[58];[22].
Holmes [39 p59], after studying the considerable variations in the instantaneous
pressure distributions on a model of a low-rise building in a wind tunnel, argues
“that if a single distribution is [to be] used for design, it must be a conservative
one for most effects”. This is invariably the approach taken in the codes.

It is an integral approach of the codes to define the approaching wind as a


function of height for a number of generic exposure conditions or categories
which the designer will have to struggle to choose from. Making this choice, as
if not hard enough, is compounded by the fact that in actual cases, a site is
almost always a combination of a number of categories; these categories may
even be varied for different wind directions of interest [22]; [58]. Another real
concern for low-rise buildings is the fact that they are usually susceptible to
speed up effects on the wind caused by terrain effects [25]. As such, in Irwin’s
words, “the choice of exposure category is one of the major sources of
uncertainty in predicting wind loads from code provisions.” [45 p2]

4
Rarely are buildings built to stand alone in the middle of nowhere. Rather, they
are usually in the vicinity of other buildings of similar height range (or indeed
mixed with a variety of heights) such as is seen in the middle of a city
[22];[29];[45]. This means in some cases, the building might be shielded from
the full force of the approaching wind (shelter effect) and so the actual loading
on it will be lower than predicted by the codes. Yet in some other quite possible
cases, just the right arrangement of surrounding buildings can increase the load
a building in question experiences by “channelling” and further accelerating the
oncoming wind into a narrow gap [22]. These are difficult situations for the
codes to handle [29];[45 p2] and Irwin has suggested a research team be
assembled with expertise not only in “wind loading and wind tunnel
methodology but also in structural liability and risk” [45 p2]. Such a team, he
continues, is to be tasked with studying the errors amounting from the stated
variability and proffer ways in which to better deal with them in the codes.

In the present days, the architects come up with very interesting building forms
and shapes. The wind loading on a building, however, is greatly dependent on
its shape. The loading information well documented in codes is mostly for
rudimentary quasi-rectangular shapes which are rapidly losing the favour of
designers. Apperley et al. [10] relay the dilemma of wind loading codification
when they surmise that the numerous geometrical possibilities of buildings in
addition to the complex nature of interactions of buildings with their environment
preclude any precision in the definition of the appropriate wind loads in the short
term or even at all. Irwin [45] and Boggs [22] also agree to the insufficiency in
the geometrical variations documented in the codes. Boggs went further to note
that even for the quasi-rectangular building shapes, the critical wind direction
may actually be different from the square-on situation documented in the codes
owing to possibilities of directional influences by exposure condition or the wind
climate itself. Additionally, the -dimensionality of the flow around the building
only adds to the already difficult situation. Holmes [39] agrees with this where
he notes that studies which were carried out on low-rise buildings in the
emphasised the importance of natural wind turbulence and local vortex
shedding in generating the fluctuations in the wind loading.

5
Compilers of codes are not unaware of these limitations and their knowledge of
these facts is expressed in the defined scope and areas of application of the
codes where designers who have projects outside of the scope of the building
codes are directed to design them using relevant recognized literature or resort
to the use of the wind tunnel procedure [22];[4]. The latter direction seems fair.
After all, the codes, in the first place, were generated from wind tunnel
experiments [26]; [45]. This is where Irwin [45], who has been widely
acknowledged as an authority in wind engineering because of his wide ranging
experience in international projects such as the World Trade Centre
Replacement Towers in New York and the Burj Dubai in Dubai amongst others
[55], has found another deficiency in the codes. He notes that the wind tunnel
testing and technology have come a long way from what they were in the
when most information used in the codes today were gathered. For one thing,
many more pressure taps can be fitted and scanned simultaneously on wind
tunnel models now and for the other, the data can be gathered and processed
with much more precision than was ever possible in the given the
computing prowess of present times. It will be interesting to see what bearing
this will have on the adequacy of the codes.

Finally, with a motivation to improve the codes, physical modelling in


atmospheric boundary layer wind tunnels has evolved and come of age to be
sanctioned with confidence as a reliable way to get loading information
superseding that provided by codes in accuracy [26]; [45];[24];[22]. Yet, the
motivation behind the advancement of boundary layer wind tunnels goes
beyond just the improvement of the codes for the generic geometries they
cover. It may be quite clear now that unconventional low-rise building designs
even of modest value may now find their ways into the boundary layer wind
tunnel as much as their high-rise counterparts. This is because it is fast
becoming common to design buildings with curved and smooth shapes such as
is the case of the Gherkin in Figure 1-1. These offer possibilities of structural
savings from the general reduction in the imposed aerodynamic forces on them
(refer to the literature for details) [25];[46].

6
Figure 1-1 The Gherkin in London [5]

This trend in building design has provided the avenue to challenge the
underpinning assumption behind the use of wind tunnel studies for design - the
issue of insensitivity of most building shapes to Reynolds number effects. Irwin
[46] suggests, in line with Baker’s recommendation[18], giving further
considerations to this established assumption and to look for ways to interpret
wind tunnel tests on Reynolds number sensitive shapes.

1.3 Aim and Objectives


The aim of this study is to review the Eurocode in the light of some of the
highlighted deficiencies with an extended view to assess the quality assurance
of using more pressure tappings in obtaining wind loading information from wind
tunnel experiments and thereby, populating the database of the Department of
Aerospace Dynamics with the processed information.

To achieve this aim, the following objectives were set:

1. To gain relevant background information on the subject of Building


aerodynamics needed to understand and work with the Eurocode.
2. To explore the principles of wind tunnel experiments for studies of
surface pressure distributions on buildings and also, briefly, to explore
computational approaches to the said matter.

7
3. To analyse the strengths and limitations of earlier similar studies on the
same subject as contained in existing literature in order to better situate
the present work.
4. To identify three building geometries, as case studies, based on the
perception of the level of difficulty the Eurocode would experience in
order to predict the wind loading on them from the simple to the
intermediate and difficult levels on the Cranfield Univeristy (CU) campus.
5. To have models of these buildings constructed and fitted with as many
pressure tappings as possible so as to be able to determine with some
measure of accuracy any gained precision in the wind loading estimate
with number of tappings.
6. To conduct sufficient wind tunnel tests on these models by varying the
azimuth of the approaching wind and possibly varying the wind speed.
7. To analyse the data and compare relevant results with corresponding
information from the Eurocode and also with available computational
result.
8. To conclude on the review of the Eurocode, the effect that more tappings
may have on the measured wind loading and the Reynolds number
sensitivity issue.
9. To highlight areas of further research.

1.4 Scope and Limitations


The discussion so far seemingly demonstrates the vastness and diversified
nature of the discipline of wind engineering. In fact, Cook [27] who has written a
voluminous text on the guide to the wind loading of building structures writes in
its second part:

“The development, testing and application of the engineering models used in


this Guide are the result of more than two decades of research in the field
now known as wind engineering. This field is so large that it cannot all be
covered adequately by any one individual or even any one research
organisation.” [27 p4-5]

8
The relevant part of the Eurocode to this study is
[15] and its annex in current use in the United Kingdom. It contains
information not only about the wind loading of low-rise structures but also about
medium height buildings, large ware-houses, bridges, towers, chimneys e.t.c.
There is no doubt there may be equal needs for further research on each and
every kind of structure documented in the code but the scope of the present
study is limited to the provisions of the code for low-rise buildings for reasons
already discussed. As such, no attempt was made to quantify the possible
dynamic responses of the case studies nor was there any attempt to review the
provisions of the Eurocode for dynamic responses.

In addition, the wind loading on a structure calculated from a building code is a


translation of the multiplication of the dynamic pressure with a pressure or force
and/or friction coefficient and several modification factors to take into account
pressure fluctuation effects for instance. Also, for cladding designs especially,
external as well as internal coefficients are reported. In a similar vein, net
pressure coefficients are specified for some specific kind of structures. This
study only considers the effectiveness of the code in predicting the effect of
wind on some chosen external surfaces of the buildings tested that is to say the
external pressure coefficients specified.

More so, it has been discussed that wind loading of low-rise structures is liable
to shelter effects [22], terrain effects [25] and heterogeneity of roughness
exposure [22]; [58]. These are beyond the scope of this study as the building
models used were tested in a lone environment with a homogeneous roughness
exposure from all wind directions neglecting any possible terrain effects that
may be present in the real situation. These should be taken into consideration
should the result of this study find its way into comparison with any full scale
corroborative measurements on the actual buildings.

Moreover, the models of the buildings were constructed with bare surfaces. As
such, effects of the roof design and furniture effects are out of the scope of the
present study.

A number of limitations to this study were identified as follows:

9
1. The number of possible pressure tappings to be used was limited to a
hundred.
2. The simulation of the atmospheric boundary layer in the wind tunnel was
achieved by Roughness, barrier and mixing devices method.
3. The visualisation of the results with Matlab contour plots is tied to
limitation 1.

10
2 Background Studies and Review of Literature

“Sometimes gentle, sometimes capricious, sometimes awful, never the same for two moments
together; almost human in its passions, almost spiritual in its tenderness, almost divine in its
infinity” John Ruskin, The Sky [9 p110]

The use of the Building codes to assess the wind loading on a building naturally
requires an understanding of the nature of the wind and the knowledge of
Building Aerodynamics. Some emphasis need to be placed on ‘Building
Aerodynamics’ because while aerodynamics is fairly general in such concepts
of fluid flow as flow attachment and separation, its application to specific
disciplines require ‘specific understanding’ of or ‘fine-tuning of the mind’ to the
discipline in question. In this line of thought, it is also interesting to note that
‘Building’ in this regard does not necessarily have its limited definition as in
everyday usage; ‘Building’ in this context usually means a building or any part
thereof so, sometimes, distinctive terms such as ‘member’ or ‘structure’ may be
used to avoid confusion [51]. Moreover, the validation of the building codes with
‘carefully’ performed wind tunnel experiments demands the grasp of the
appropriate experimental techniques and methods of data analysis. Indeed
Kasperski [49 p527] says it all when he notes that “an inappropriate
experimental technique may lead to complete useless results.”

Some of the basic concepts already discussed have been explored and
documented along with the appraisal of some relevant past works in this
Chapter. A summary is presented in Section 2.8. An attempt is also made in
that sub-section to assess the stated objectives of the current study in the light
of the gathered information. This approach is also used in subsequent Chapters
of this thesis.

11
2.1 Atmospheric Wind
The excerpt at the beginning of this chapter gives an interesting personification
of the sky to which the atmospheric wind may, at first, be synonymous. In a
better perspective, the atmospheric wind is an integral constituent of the sky.
The excerpt encapsulates a poetical as well as technical representation of the
nature of the atmospheric wind.

The wind which receives its energy from the sun [58] is “a dynamic and random
phenomenon in both time and space” [21 p16]. The differential solar heating of
different parts of the earth’s surface leaves the equatorial regions of the earth
as the hottest. The temperature differences thus achieved between the poles
and the equator, give rise to pressure differences resulting in what is called
atmospheric circulation. This simply means the response of nature to balance
out the pressure differences. The atmospheric circulation is susceptible to the
annual variations of the sun’s position in relation to the north and south of the
equator (seasonal effects), unequal distribution of water and land (geographical
effects) and variations in the speed of the rotation of the earth from the greatest
at the equator to the least at the poles (Coriolis effects) [58];[40].

In effect, the wind speed can be seen as possessing an average or mean value
on which random variations or “gusts” or “turbulence”, as they are more readily
called, are superimposed. With no clear correlation between the random
variations at different heights, the mean wind speed increases with height.
There is no repetition in the waveforms in time (see Figure 2-2). The wind
speed at any particular time of interest can only be described statistically as it
does not lend itself to exact prediction. The facts discussed so far are true for
the lowest portion of the atmosphere adjacent to the surface and it is referred to
as the Atmospheric Boundary Layer (ABL) [21].

Depending on geographical location, the various wind types have earned


different names even when they are the same. At the low latitudes for instance,
very strong winds which derive their energy from latent heat of condensation of
water vapour are called hurricanes in the West, typhoons in the Far East and
Cyclones in Australia and the Indian Ocean. This kind of local winds may have

12
not so much influence on global atmospheric circulations but nonetheless, they
may possess high intensity [58].

In the mid-latitudes from about to degrees within which Europe lies, the
strongest winds are ‘gales’. They can have large horizontal dimensions
extending more than sometimes. As such they usually span so many
countries in Europe’s case. They are generated by large and deep depressions
or extra-tropical cyclones of synoptic scale [40]. They have a return period of
years. These winds are used to specify design wind speeds in the Eurocode.

2.1.1 The Atmospheric Boundary Layer


D’Alembert ( ) is famous for the problem he encountered in trying to
reconcile the odds between theories of classical fluid mechanics and practical
experience. “A singular paradox which I leave to geometricians to explain”, he
concludes exuding much frustration [8 p791]. It might be of interest to
D’Alembert to learn that Prandtl who came up with the explanation and solution
to the seeming paradox was not a geometrician.

Prandtl ( ) exposed the existence of a boundary layer between a moving


fluid and a solid in which viscous effects are most prevalent and outside of
which the classical theories of inviscid flow would still hold [25]. In this thin layer,
there is a sharp velocity gradient with no fluid motion directly at the surface. The
fluid motion at the edge of this layer possesses the speed of flow experienced
beyond this layer into the fluid. Prandtl came up with his hypothesis when
working with flat plates but a boundary layer indeed exists between any moving
fluid and a solid surface no matter the dimension [25].

The ABL is large in scale in human terms but it is still thin when compared to
the size of the atmosphere [25]. It also varies in its depth. In very large storms
such as extra-tropical depressions, this layer extends up to to . In
thunderstorms, it is around . In so many ways, the ABL is similar to a
turbulent boundary layer on flat plates or on airfoils at high wind speeds [40].

13
Figure 2-1The mean ABL wind velocity profile over different terrains [25]

Frictional effects near the ground surface exhibit such an important role as to
retard the wind speed close to the surface as seen in Figure 2-1 [58];[40]. This
causes a gradient in the wind speed with height above ground. The height at
which the wind is no longer under the influence of surface roughness is called
the gradient height . This height varies with varying surface roughness . The
wind at this height which is free from surface frictional effects is called gradient
wind speed . It is taken to remain constant above the gradient height [58].

In strong winds with speed of about or more, turbulence generation is


predominantly mechanical and thermal effects on the generation of turbulence
are negligible. The ABL stratification based on temperature is disturbed by the
turbulent mixing ensued. The ABL then is considered to be ‘neutrally’ stable
[42]. For practical purposes of simulating this kind of wind in wind tunnels to
assess wind loading and environmental safety, the boundary layer is taken to
be adiabatic [28]. This would not be the kind of boundary layer simulation to be
used in a study on effluent dispersal which is affected by thermal effects [28].

14
A fully developed ABL flow has the following characteristics as depicted in
Figure 2-2 [40]:

 the increase of the average wind speed with height;


 the turbulence in the wind speed at all heights;
 the vast range of frequencies in the turbulence or gust in the air flow;
and
 the apparent similarity in the patterns of gusts at all heights, most
noticeable in lower frequency gusts or turbulence.

Figure 2-2 Wind speeds at 3 heights during a gale [40]

The Coriolis effects on the wind in the ABL produces a mean velocity vector
which is not constant in direction with height. This effect reduces with
decreasing height above the earth’s surface leading to a geostrophic imbalance.
The mean wind velocity vector which is parallel to the isobars at the gradient
height gradually rotates towards and possesses a component in the low
pressure side as the ground surface is approached. In essence, the mean wind
speed may possess slight variations in its direction as well as in its magnitude
with height. This phenomenon is denoted as the Ekman spiral and changes in
mean wind direction of up to about degrees have been measured in full-
scale [40]. However, this effect is normally neglected in wind engineering since
the directional change due to it is minimal over the height range of normal

15
structures [40]. The region of the ABL closest to the surface and normally
regarded as being out of the influence of this effect is known as the inner or
surface layer with a depth of about [40]. This is the region of interest in
this study.

2.1.2 Terrain Roughness Categorization


With the mean wind speed over a terrain being a function of the surface
roughness of such terrain, it is natural that wind speeds may be lower in a city
or urban area than in an open terrain. However, the turbulence which is also
dependent on the surface roughness is much higher in the city [58].

The Davenport ( ) roughness classification is arguably the most instructive


reference for practical purposes. Davenport classified the different terrains into
eight categories. Wieringa ( ) extended this classification to include terrains
with low roughness, as presented in Figure 2-3, such as the ocean for instance
[54].

The Engineering Sciences Data Unit (ESDU) classification is another


type of classification which finds common usage. Although the terms used by
these terrain roughness categorizations are different sometimes, they are all
identical to some greater or less extent [54].

16
Figure 2-3 Wieringa terrain roughness classification [54]

2.1.3 Turbulence
From earlier discussion about the atmospheric wind, it is evident that the wind
speed can be regarded as having a fluctuating component superimposed on the
mean wind speed. The fluctuating component which has been identified as
turbulence or gustiness is caused by meteorological (convective movement)
and/or mechanical (ground roughness) means. In very strong winds, the
mechanically generated turbulence is assumed to dominate. As such, it
becomes of utmost interest to engineers. Naturally, mechanical turbulence is a
function of height above ground and terrain roughness. Therefore, it decreases
with increasing height and increases the rougher the terrain [58].

The level of turbulence in the wind speed can be quantified by its standard
deviation or root-mean-square as presented in Equation (2-1) for the
longitudinal turbulence component [40].

17
(2-1)
∫ ̅

2.1.3.1 Turbulence Intensity and Length Scales

Another measure of turbulence is the turbulence intensity which is the ratio of


the standard deviation of each fluctuating component to the mean value of the
wind speed [40].

⁄̅ (2-2)

⁄̅ (2-3)

⁄̅ (2-4)

Measurements which were taken close to the ground in gales produced by


synoptic depression systems show that the standard deviation of longitudinal
wind speed, , is equal to to a very good estimation. Then, the
longitudinal turbulence intensity can be expressed as in Equation (2-5).

(2-5)
⁄ ⁄ ⁄

As such, the turbulence intensity is a function of the roughness length . This


means it also decreases with increasing height above the ground. The
longitudinal component of turbulence intensity is invariably higher in magnitude
compared to the corresponding lateral and vertical components. Therefore, it
receives much more attention [40].

For an ‘ideal’ neutral ABL, there are two characteristic length scales. The first
which is on a velocity scale is for the outer region of the flow which is dependent
on the rate of rotation of the earth and the latitude and the second is for the flow
in the inner layer of the ABL and is dependent on the size and density of the
roughness on the surface [40].

18
2.1.3.2 Wind Structure and Spectral Density Function

The non-deterministic characteristics of the natural boundary layer wind in both


time and space domains have been established thus far. The sporadic nature of
the ABL wind speed is caused by eddies and vortices within the flow moving
along at the mean speed. These eddies and vortices are never identical for any
arbitrarily taken points in time. As a consequence of this, only statistical and
probability theories can be used to describe them. It is common in many wind
loading codes to use a peak gust wind speed for design purposes. However,
the dynamic nature of wind also means that that the peak gust within an
averaging period of for instance is also a random phenomenon [40].
Nonetheless, assuming stationarity (a term used to define the tendency of the
mean wind speed to remain relatively steady for a period of to an hour),
an expected or average value for this can be defined [58];[40]. This is also a
very basic idea behind wind tunnel testing; the justification for it is the fact that
processes giving rise to the mean flow have time scales much greater than an
hour. In addition, usually, large variations of the mean speed with time lasts a
period of several days [58].

The spectral density has been used to describe the distribution of turbulence
with frequency [40]. The full wind spectrum put up by Van der Hoven ( ) and
which was presented by Davenport at the International Conferences on Wind
Engineering (ICWE) - in [48] is presented in Figure 2-4 . Van der Hoven
(1957) identified several distinctive features and time scales of the wind speed
spectrum from full-scale measurements he made of the horizontal wind speed
at Brookhaven, New York in the ; “Wind speeds are driven by weather
fluctuations which vary on a macro-meteorological scale with periods in the
order of hours and days”[48 p3]; Details of the wind structure in a strong wind is
predominantly dependent on mechanically generated turbulence which are
generated at a micro-meteorological scale with periods of seconds and
minutes. Consequently, spectral peaks appear in two major regions, one at a
period of days associated with synoptic storm systems and a period of day
corresponding to diurnal frequency; the other at a period of is associated

19
with the locally generated turbulence within the ABL the fluctuations of which
are separated by a spectral gap at periods of about to [48]; [58].

Figure 2-4Spectrum of Wind speed variation within the Atmospheric boundary


layer [48]

2.1.4 Mean Wind Velocity profile


2.1.4.1 Logarithmic Law

The log law which Prandtl (1904) originally derived for the turbulent boundary
layer on a flat plate has been found in the unmodified form to be most adequate
in modelling strong wind conditions in the ABL near the surface. It is usually
derived in a number of different ways, the simplest of which is presented in
Equation (2-6) [40]:

̅ (2-6)
[ ]

Even though the log law is based on the physics of the boundary layer and very
valid in the bottom to of the layer [58] (at least for a fully developed
wind flow over a uniform terrain) it may not be accurate sometimes since these
ideal conditions may rarely be met [40]. In addition, the log law is unsolvable for
heights below the zero-plane displacement since the logarithm of negative

20
numbers do not exist. More so, it can be more difficult to integrate over a height
range. All of these factors make it a less desirable option a number of times to
wind engineers who are in favour of the power law [40].

2.1.4.2 Power-law

The power law, presented in Equation (2-7), has no theoretical basis as it was
derived solely with empirical insights [58]; [40]. It is quite easily integrated over
a height range. This makes it convenient to use when the objective is to
determine the bending moment at the base of a tall structure for instance [40].

̅ ̅ (2-7)

(2-8)
( )

The exponent, , in the power law changes with terrain roughness as well as
with height range when compared and matched to the log law. Equation (2-8)
relates the exponent to the roughness length, where is a reference
height at which a match is made between the laws. It may be taken as the
mean height in the range in which matching is required or half of the maximum
height over which the matching is required to be made [40].

Taking to be and matching the laws over a height range of with


Equation (2-8), Holmes reports that the two ‘laws’ are “extremely close”.
Additionally, “the power law is quite adequate for engineering purposes.” [40
p49]

2.1.5 Wind Data and Wind Rose


Guilbauld [12] who used the Cranfield campus as a case study in his research
into urban micro energy recovery systems has made a comprehensive review of
the wind data and the wind rose for the campus.

Wind data are usually collected by weather stations and airports. Due to the
dynamic characteristics of the wind in time and space already discussed, the

21
recorded data for the wind speed and direction are usually average values with
period of measurement commonly few minutes in the range of [12].

Wind roses are built from wind data recorded for several years usually a
decade, at least, to be reliable [12]. The Cranfield wind rose presented in Figure
2-5 was made with data from to . The length of a sector in the rose
represents the percentage of that direction as a proportion of the overall values.
Also, the sectors are divided into divisions to give the proportions of the
different wind speed ranges. It should be noted from the figure that strong winds
can come from any direction. With more than of the data in the
direction ( degrees), the dominant wind direction for Cranfield is established.
For this reason, it is the primary wind direction of concern in this study.

After much statistical analysis to take into account the seasonal variations in the
wind speed and the fact that the wind data used for the Cranfield wind rose
were recorded majorly in the day time, Guilbaud [12] concludes that the long-
term average wind speed of is suitable for such studies as he made on
the Cranfield campus bearing in mind that the value was optimistic from
comparisons he made to some earlier studies he reported. This was also
considered suitable for the current study.

22
Figure 2-5 Cranfield Wind rose [12]

2.2 Building Aerodynamics


Buildings belong to the category referred to as ‘bluff bodies’ in aerodynamics.
This is in relation to the nature of flow around them. As opposed to the flow
around streamlined bodies such as an aircraft wing where the flow streamlines
follow the ‘body contour’ (strictly speaking, the displaced body surface due to
the presence of the thin boundary layer) closely from the leading edge usually
towards the trailing edge before separation, the flow around bluff bodies from
the wind ward extremity separates before getting to the leeward extremity
[14];[40]. The separation occurs at the sharp leading edge corners of the bluff
body, as depicted in Figure 2-6a. This occurs as the momentum of the fluid at
these corners “overcomes the weaker cohesive viscous forces of attachment”
[14 p69-70]. A thin region of high shear and vorticity similar to a detached

23
boundary layer on an airfoil called the ‘free shear layer’ separates this region of
separated flow from the outer flow. The free shear layers are inherently
unstable in a sheet form. Therefore, they roll up and become sources of
downwind vortex shedding towards the turbulent wake which forms between
them on the leeward side of the bluff body [14];[40];[47]. For a bluff body with a
long ‘after body’ depicted in Figure 2-6b, the separated region ‘re-attaches’ on
to the surface at some distance downwind. However, it may not be fully
stabilized and may still lead to vortices being formed and rolled up along the
surface [40].

Figure 2-6 Flow around a rectangular body with a. a short along-wind length [14]
b. with a long along-wind length [40]

Vortex shedding excitation is usually a source of concern for wind engineers;


the across-wind excitation of very tall buildings, chimneys and trees by this
phenomenon being an example [47];[51]. It is also common knowledge that “the
highest peak local loads on buildings are often caused by the flow separation
and generation of vortices near building corners” [47 p709]. While the turbulent
intensity in the approaching wind can reduce the strength of vortex excitation
[47], engineers have come up with practical means of dealing with it. If vortex
shedding is to be mitigated or assuaged, the interaction of the two shear layers
must be prevented. For circular cylindrical shapes, a way to this is to vary the
diameter in a helical manner along the height by wrapping ‘strakes’ around
them as illustrated in Figure 2-7a. This essentially serves to disturb the
correlation of the vortex sheets along the heights. Another way is by wrapping a

24
porous shroud of slightly larger diameter around the cylindrical shape as
depicted in Figure 2-7b in the case of a tree canopy. This serves to bleed air
into the wake between the shear layers and prevent their interaction [51]. This
may have informed the design of the decorative porous shroud proposed to be
around the Millennium towers under construction in Abuja, Nigeria in Figure
2-7c. As for very tall buildings, a practical way into ameliorating the vortex
shedding effects is discussed in Section 2.2.1.

Figure 2-7 a. Industrial chimney [6] b. Tree Canopy at Milton Keynes Coachway
(picture taken by author) c. Rendering of the Millenium tower in Abuja (under
construction) [2]

The location of the separation points and the consequent geometry of the wake
have a significant bearing on the surface pressure distribution and total forces
on a bluff body. For a rectangular prism, the separation points are fixed by the
geometry as the boundary layer flow from the windward face fails to remain
attached around the sharp corners resulting is flow separation. In the case of

25
other bluff shapes especially those with curved surfaces, the separation points
are not readily predictable. For instance, separation takes place at different
positions in the flow around a circular cylinder depending on the relative
magnitude of the forces which dominate the boundary layer flow [58]. The
relative magnitude of the inertial forces exerted by the fluids momentum to the
cohesive viscous forces in the boundary layer is expressed in the form of a
dimensionless parameter, in Equation (2-9), referred to as the Reynolds
number, [58]. of airflow around buildings are usually large given the large
length factor (characteristic length) and the low viscosity of air [14].

(2-9)

In the case of bodies with curved surfaces, the drag coefficient (a term adopted
from aeronautics to describe the coefficient of the along-wind force) is greatly
dependent on the [58]. These bodies are shaped so that their surface
curvatures match the path of adjacent streamlines in the moving fluid around
them for a particular flow direction and speed. This means that relatively small
changes in the on-coming flow direction or significant changes in the speed will
most likely lead to substantial changes in the magnitude and distribution of
surface pressure distribution on the streamlined body. To mitigate this, buildings
with curved surfaces are designed with surface roughness such as protruding
mullions or indentations to trip the laminar flow across such surfaces into a
turbulent one. This effectively collapses the varying flow patterns which would
have been the case with changing wind speeds for a given wind direction into a
single pattern [14]. Bodies with sharp corners, on the other hand, have drag
coefficients which are almost independent. Also, the along-wind length of the
body has an important effect on the drag force. More so, the existence and
location of re-attachment points have significant bearing on the pressure-
induced drag force. The size of the wake influences the drag force so that for
two bodies with the same frontal area, the one with the longer along-wind length
and, consequently, narrower wake will experience less drag [58].

26
Nevertheless, the general effects of are usually ignored in bluff body
aerodynamics on the grounds that a multitude of relevant shapes in this field
are sharp edged bodies for which flow patterns are almost predictable since the
sharp edges fix the flow separation. The reliable flow separation at these sharp
edges such as wall-roof junctions ensures a relatively constant flow pattern and
characteristics for a given wind direction and over a wide range of expected
wind speed on these bodies [46]; [14]; [14]; [40]. The reliance on the certainty of
flow separation at these sharp edges thus permits reliable wind pressure
distribution studies on models of large sharp-edged buildings to be made at
relatively low tunnel wind speeds neglecting the dissimilitude of without
much of significant errors in the surface pressure distribution patterns [14]; [14];
[25].

Yet, words of caution have been given in regards to this basic assumption of
insensitivity of sharp edged bodies’ surface flow patterns. Hoxey et al. [43 p554-
555] who made detailed measurements on a low-rise building at full scale and
model scale in a wind tunnel found “significant effect associated with the
separated flow region on the windward roof slope of the building”. They noticed
that while it appeared from their full scale measurements that the effects are
limited to the region of flow separation at the windward eave, it is possible that
the effects extend to the stagnation region on the windward wall. They found it
evident that the change in flow pattern in the separated flow region modifies the
flow downstream of it. In addition, it affects the pressure distribution outside the
separated flow region [43]; [46]. They reinforce their point by comparing the
consistency of their observations with those of other researchers who have
noticed under-estimation of full-scale measured suctions with model-scale
studies (the work and conclusion of Vickery et al. [62]; [61] are also noteworthy
in this regard ) and also with a cited observation that the re-attachment length in
a flow over a backward-facing step is sensitive especially at model-scale
[43]. Therefore, they went on in their paper to call for urgent further investigation
into this issue geared towards the quantification of these effects for a wide
range of structures and the development of practical engineering approaches to
take them into account. Irwin [46] supports this call and also cited the research

27
carried out by Long et al. in comparing wind tunnel results with full scale
measurements on the Texas Tech experimental building to support the findings
of Hoxey et al.: Long et al. found that the significant mismatch present at low
turbulence levels were reduced substantially at high turbulence levels [46].

Irwin [46] went further to explain the effect on the approaching flow
turbulence with the power spectrum of turbulence (which is a basic parameter
simulated in boundary layer wind tunnels) using that of the longitudinal
turbulence in non-dimensional form depicted in Figure 2-8.

Figure 2-8 Non-dimensional power spectrum of longitudinal turbulence at model


and full scale [46]

It is evident from the figure that the curve consists of a part which is proportional
to at low frequencies. This goes through a peaky transition to the inertial sub-

range region characterised by behaviour. Viscous forces dissipate the
turbulence energy into heat at substantially high frequencies. The spectral curve
declines rapidly in the ‘viscous cut-off’ region as it is called. Theoretical analysis
of this region suggests that the spectrum varies as . It is noticeable from the
figure that the range of frequencies where the inertial sub-range applies
covering the range of significance for wind loading in the model scale is
relatively small compared to the full scale: For typical model scales and
velocities, the viscous effects are likely to be prominent for non-dimensional

28
frequencies beyond about 30 to 300 while at full scale, the viscous effects

do not begin to have prominence on the turbulence until about . The

good thing is the frequency range where the non-dimensional model and full
scale spectra remain similar is that of most interest. However, when unusually
small models are tested with low velocities such as is the trend in the design of
larger structures like very tall towers, caution will be required [46].

The discussion so far has established the issue of sensitivity of wind flow
around sharp edged bluff bodies to beyond that of conjecture. While it is
apparent that turbulence and surface roughness have the tendency to suppress
the effects noticeable in the flow around streamlined bodies, an in-depth
quantification of how much confidence can be put into these has not been
achieved for most building forms. Similarly, the interactions of these with
effects on buildings with rounded corners of various radii have not been fully
established [46].

As important as the has proven to be in the accuracy of wind tunnel model


studies of buildings, another non-dimensionless parameter called the Jensen
number, is probably of more significance at least where mean and fluctuating
pressures are concerned. In the case of a building in a turbulent boundary layer
flow, it is the ratio of a characteristic body dimension, usually the height, to the
characteristic boundary layer length represented by the roughness length . Its
effect was shown to be greatest on the roof and side walls of buildings where
increased turbulence in the flow over the rougher ground surfaces supports
shorter flow reattachment lengths. Jensen ( ) proved this in a series of
experiment he carried out by matching wind tunnel measurements on models of
small buildings with those in full scale. Thus, he established the requirement of
equality of ⁄ to match wind tunnel results with full scale [40]; [27]; [40].

The general effect of wind on buildings is three-fold. The building must have
sufficient strength to withstand the wind-induced forces; it must have sufficient
stiffness to satisfy occupant comfort and serviceability criteria; and the wind
may likely cause a dynamic response of the building (along-wind, cross-wind

29
and torsional modes [21];[27; 40; 51]. Lawson [51] notes that all buildings can
move. Thus, it is necessary to consider the “stability” of the building. A stable
building self-recovers its equilibrium position after it has been distorted from it.
Therefore, a building can either be “Statically” or “Dynamically” stable, the
difference being that the velocity of the building induces additional forces or
moment on the dynamically stable building. It is however not necessary for the
whole building to exhibit a dynamic response; a part, the roof for instance, may
exhibit a dynamic response on a building which is statically stable. In such a
context, it is pertinent to refine the vocabulary; “building” means a building or
any part thereof and sometimes the term “member” or “structure” may be used
instead [51]. In this regard, the dynamic response of a low-rise structure may
be insignificant provided it would have been appropriately designed for both
strength and stiffness requirements [21].

Stathopoulos [58] highlights the general characteristics of wind-induced


pressure distributions on a building model in a turbulent shear flow condition as
follows:

 Pressures on the wind-ward face are positive. They diminish away from
the stagnation region as the flow accelerates around the sides and upper
edge of the face;
 Pressures reduce downwards along the face centre due to decreasing
velocity with decreasing height in the boundary layer. This, however,
translates to increased velocity and decreased pedestrian comfort at
street level;
 Pressures on the leeward face are negative (suction) with their absolute
value appearing to decrease downwards the face;
 Roof and side pressures are invariably negative with very large localized
suctions at the corners and eaves; and
 Differences in between the wake pressure and base pressure of
windward faces are responsible for horizontal flows through arcades or
around corners. These are not too easy to manage especially for very tall
buildings [58].

30
These form the theoretical basis with which the results of the experiments were
viewed in Chapter5.

2.2.1 Tall Buildings


Buildings in this category have medium to high aspect ratios (where aspect ratio
is the ratio of the height to the breadth of the building [40]). In the present day,
they may have heights well in excess of . Burj Dubai in Figure 2-9 is
notable in this range [17]. A building of such height extends well into the upper
regions of the ABL where its upper levels may be subjected to the highest winds
of a synoptic windstorm such as a tropical cyclone or a winter gale in the
temperate regions. In such a case, its resonant dynamic responses in the
along-wind, cross-wind and torsional modes will be excited [40];[51];[27].

The dynamic response of a tall building in the along-wind direction is


predominantly produced by the turbulence in the natural wind [40]. It is, usually,
common that the overall wind loading on a tall slender building originates from
the across-wind vortex shedding excitation which induces a significant dynamic
response [47]. The contribution to this from cross-wind turbulence is usually
small [40]. A practical solution to this for tall buildings has been employed in the
engineering of Burj dubai (in Figure 2-9) where the cross-section of the building
is varied with height while introducing setbacks into the tower at different levels
as shown in Figure 2-10. This, essentially, ensures that the Strouhal number,
which relates to the shedding frequency, varies with height. Thus, the vortex
excitation loses its coherence due to this ‘confusion’ [47]; [17].

31
Figure 2-9 a. Computer rendered image of Burj dubai and b. A real picture of the
building during construction (adapted from [17])

Figure 2-10 Wind Behaviour of Burj Dubai [17]

The general flow characteristics around a tall building with a rectangular cross-
section are presented in Figure 2-11. The downward deflection of high velocity
air from upper levels results into a very strong flow around street level below the

32
stagnation point on the windward face (which occurs at a height of of
the total building height). This usually causes pedestrian discomfort. Flow
separation and re-attachment at the side walls results in extreme local cladding
pressures. At the rear face, the pressure is negative and is of lower mean
values [40].

Figure 2-11 Flow around a tall building [40]

It is not surprising then that tall buildings are some of the more wind-sensitive of
structures. Their responses to wind are of utmost concern to structural
engineers. They command the attention of experimenters both in full-scale and
in the wind tunnel [40]. Holmes [40] identifies the two main areas of concern for
tall buildings as follows:

 The vulnerability of their glazed cladding to the direct wind pressure as


well as flying debris in windstorms; and
 The serviceability issues which arise from extreme motion towards the
top of the buildings.

Of significant interest to structural engineers are also the aerodynamic


interference effects which other neighbouring tall buildings may have on the one
of interest in a cluster of tall buildings. This is very usual in a city-centre

33
business district. In the design of a tall building to be situated in a complex of
tall buildings, wind tunnel investigations are usually carried out to reveal any
important interference effects which neighbouring buildings may have on the
new one. It is also quite possible that existing buildings will experience
unexpected higher loads resulting from new buildings of similar size in the
immediate surroundings of these buildings. This means that this would have to
be considered when considering load factors [40].

2.2.2 Low-rise Buildings


Holmes [40 p163] defines low-rise buildings as roofed low-rise structures of less
than height. This definition is also adopted for the purposes of analysis in
this thesis.

A number of factors make wind loading assessment for low-rise buildings not
easier when compared to that of tall buildings. These are:

 The fact that low-rise buildings are usually completely situated within the
lowest region of the ABL where the turbulence intensities are high due to
the earth’s surface roughness. Also, in this region, interference and
shelter effects are of paramount importance. All of these are usually
difficult to quantify [40];
 The most significant wind loads on a low-rise structure are those of the
suction on the roof with many structural failures initiating there.
However, roof loadings vary with variations in building geometry [40];
and
 It is often the case for low-rise structures to have a single internal space.
The internal pressures on this internal space can become very
significant particularly when a dominant opening happens to be in a
windward wall. The assessment of the magnitude of the internal
pressure peaks and their correlation with external pressure peaks are
very important [40].

34
Figure 2-12 Flow around a Low-rise building [40]

Illustrated in Figure 2-12 is the flow pattern around a low-rise building with a
gable roof. There is flow separation at the eave of the windward wall. At some
distance downwind of this separation point on the roof, the flow re-attaches
resulting in a separation zone or ‘bubble’ the extent of which is time-variant. The
separation zone, which is separated by a high turbulence and high velocity
gradient free shear layer (as in earlier discussion) from the outer flow, rolls up
sporadically shedding downwind vortices. This process may result in the
production of high suction peaks on the roof surface [40].

A high turbulence approaching wind with a longitudinal turbulence intensity at


typical roof heights of low-rise buildings of or greater shortens the extent of
the separation zone towards the eave. These small separation zones are
representative of high initial negative pressures with fast pressure recovery
downwind. In aerodynamic terms, roofs with pitches up to around degrees
are flat for approaching wind normal to the ridge or gable end. A similar case to
this is when the mean wind direction is parallel to a ridge line. Then, the roof is
also aerodynamically flat no matter what roof pitch [40].

For winds perpendicular to the ridge line and, also, roofs with pitches between
and degrees, a second flow separation appears at the ridge as shown in
Figure 2-13. This results in high suction (negative pressure) regions on both
roof slopes. A second flow re-attachment is experienced with consequent
pressure recovery downwind of the ridge [40]. With roof pitches greater than

35
about degrees, the wind pressure on the upwind roof slope changes to a
positive pressure and acts towards the roof surface. This leaves a fully
separated flow without re-attachment downwind of the ridge. Thus, it results to
fairly uniform negative mean pressures on the downwind roof slope [40]; [14].

Figure 2-13 Flow around buildings with Roof pitches of 20-30 degrees [14]

The flow features discussed above apply only to low-rise buildings with
(slenderness) ratios less than around . In the case where this ratio is higher,
the roof generally experiences more suction [40]. The significance of this is that
flow patterns around buildings are dependent on what characteristic dimension
of the building the wind ‘sees’ as the scaling length. Usually, this length is either
(twice the height of the structure) or the cross-wind breadth whichever is
smaller [27].

The flow pattern on the side walls of low-rise buildings is similar to that which
has been described for the roof only with generally lower mean pressure
coefficients. The windward walls experience positive pressures. The leeward
walls experience negative pressures of lower magnitudes due to the re-
circulating turbulent wake of the building. The magnitudes of these negative
pressures are dependent on building dimensions inclusive of the roof pitch
angle [40].

36
Figure 2-14 Corner vortices in an obliquely approaching wind [40]

Corner vortices sharing great similarities to those found on a delta wing in


aeronautics ensue at the roof corner for quartering approaching winds on the
roof corner in question as illustrated in Figure 2-14 [40].The axes of these
vortices are slightly inclined to the adjacent walls forming the corner. On a low
pitched roof with square or rectangular plan form, the pressures beneath these
vortices are highest. However, these pressures act over small areas. Hence,
they are more significant for cladding considerations rather than the design of
major structural members [40].

A number of Full-scale experiments which have been carried out to investigate


wind pressure distributions on low-rise buildings have shown the highly dynamic
nature of the wind pressures, area-averaged wind loads, and load effects on
this category of structures [40]. Holmes [40] relates the time-variation of these
quantities to the following factors:

 Induced fluctuations in pressure due to approaching wind turbulence


which may arise from the wakes of upwind buildings for instance (see [40
p178-179] for more details on aerodynamic interference and shelter
effects on low-rise buildings); and

37
 Local unsteadiness in the surface flow which may result from vortex
shedding or other unsteady flow phenomena associated with the region
of flow separation around sharp corners, roof eaves and ridges.

In reality, these two identified factors may interact to create a more complicated
flow pattern. In addition, the varying quantities are not only time-variant but also
space-variant. This means that the same pressure or response variation with
time may not be experienced simultaneously at different points separated from
each other on a building [40]. However, resonant dynamic response of a low-
rise building can normally be ignored [40]. It is interesting to note that it is
usually the case that the majority of structural damage in windstorms is incurred
by low-rise buildings (family dwellings) naturally because they are most often
non-engineered and ill-maintained [40];[25];[11].

The rationale behind and the caution that goes with the usefulness of wind
tunnel studies to predict flow patterns on sharp edged buildings at full scale
despite the dissimilitude in have been noted earlier in this study. The need to
scale the model studies with the appropriate has also been presented.
directly projects its influence on the mean pressure distributions on a low-rise
building through the variation of mean velocity profile with height [40]; [40].

However, after series of international comparative wind tunnel studies on


models of the full scale measurements taken in the on a low-rise gable
roofed building in England (the Aylesbury Experiment [32]), the need for scaling
such quantities as the turbulence intensity, scale and spectra density became
apparent. Even though Eaton and Mayne [32] consider it unlikely that local
details in the surface roughness (trees and hedges) upstream of the Aylesbury
experimental house had effects on the measurements taken, later comparative
wind tunnel studies on this compelled other researchers to conclude otherwise
[37]; [41];[61];[10]; [41]. Vickery and Surry [61] who explored the sensitivity of
the measured pressures to the level of details modelled in the wind tunnel using
eight different boundary layer simulations of the same experiment, found a
generally increased correlation between the model scale and full scale when
more details were modelled. They subsequently conclude that unsteady

38
pressures which are dependent on the turbulence from local details present the
primary modelling challenge. Apperley et al. [10 p220] conclude much earlier
that “comparisons between model scale and full scale [experiments] on low-rise
structures will be inherently more limited than for tall structures due to this
sensitivity to local details.” They explain further that while , , relates
directly to the mean velocity profile and the corresponding turbulent intensities
for homogeneous terrains, isolated local roughness elements can have
significant alterations on the turbulence intensities and surface pressure without
corresponding alterations to the apparent roughness height. Holmes and Best
[41];[40] note that in a fully developed turbulent boundary layer flow over a
rough ground surface, the turbulence intensity and spectra need to also scale
with the ratio near the ground. These essentially mean that in a wind
tunnel study, the turbulence intensity similarity will only be attained with
equality if the turbulent inner surface layer of the ABL has been correctly
modelled.

It would be easier to work with turbulence parameters above as if they were


non-dimensional quantities. Regrettably, it is not easy to independently vary
them in a wind tunnel test. It has been pointed out that fluctuating and peak
external surface pressures are dependent on the turbulence properties of the
approaching wind. As a result, peak load effects are also functions of the
upwind turbulence. However, for adequately simulated boundary layers having
turbulence quantities near the ground scaling as , “peak load effects can
be reduced to a variation with .” [40 p168 -169]

2.3 Wind loading


Wind loading is the prevalent horizontal force on buildings. Consequently, it is
the dominant loading for cladding considerations. In building codes, the wind
loading derivation is based on the wind climate characteristics measured at
meteorological stations. The dynamic pressure calculated with knowledge of the

39
local wind climate characteristics, is multiplied by a pressure or force coefficient
and a number of other modification factors [35].

In the field of wind engineering, “wind loading” is considered synonymous to


“Alan G. Davenport”. In fact, the “wind loading chain” (presented here in Figure
2-15) which he developed and presented in his doctoral research in has
been officially named the “Alan G. Davenport Wind Loading Chain” by the IAWE
[48 p2].

Figure 2-15 The "Alan G. Davenport's Wind loading chain" [48]

His hypothesis was based on the structure of wind discussed earlier. It


recognizes that the wind loading a particular building or structure is subjected to
can be determined by the aggregate effects of the local wind climate, which
must be described in statistical terms; the local wind exposure, which is affected
by terrain roughness and topography; the aerodynamic characteristics of the
building shape and the potential for load increases due to possible wind-
induced resonant dynamic responses. In addition, clear criteria must be in place
to assess the significance of the effects of the predicted wind action. Such
criteria include the effects of wind on the structural integrity of the building or
structure and its exterior envelope and numerous serviceability considerations
ranging from the control of the wind-induced drift, the effects of building motion
on occupant comfort to the pedestrian comfort level around the building or
structure [48].

The evaluation of the wind loading and its effects, thus, relies on many mutually
connective considerations with each one requiring scrutiny and systematic
assessment. The analogy of this hypothesis or process to a physical chain is in
recognition of the fact that the outcome of the process is determined by the

40
weakest link or component. While much is gained only by improving the weak
links, little is lost by not paying much attention to the already strong ones. This
approach has found worldwide usage in building codes and standards [48].

Davenport ( ) also demonstrated the gains which can result from improved
description of the local wind climate and how the reliability of wind loading
prediction is improved with the use of wind tunnel aerodynamic data over the
loading information in building codes [48].

2.3.1 Loading Coefficients


2.3.1.1 Pressure Coefficient

Bernoulli ( ) presented the equation that has, perhaps, the most significant
usage in incompressible fluid mechanics and wind engineering. The equation
relates the fluid pressure energy, potential energy and kinetic energy along a
streamline in the steady flow of an incompressible and inviscid fluid [14].

Figure 2-16 Regions of validity of Bernoulli Equation in the flow around a


rectangular bluff body [14]

The Bernoulli equation with its overwhelming significance in building


aerodynamics is only valid, in strict terms, in an inviscid flow [14]. This is a very
important piece of information to the current study.

41
Using the Bernoulli equation, surface wind pressures on buildings or structures
are often expressed as dimensionless pressure coefficients, , obtained
relative to the dynamic pressure of the approaching wind so as to be
independent of the velocity [14];[58]; [40]. A general expression for the time-
varying pressure coefficient, , for buildings in stationary or large-scale
windstorms is given in Equation(2-10):

(2-10)

As a matter of convention, “positive pressure coefficients indicate pressures


acting toward a surface while negative pressure coefficients indicate suctions or
pressures acting away from a surface” [14 p72]. The magnitude of positive wind
pressures are limited by the dynamic pressure of the approaching wind. This
means at complete stagnation, on a windward face. In reality, the
highest mean pressure coefficients on a windward face have magnitudes less
than this theoretical value [40];[58];[14]. Negative values experienced near
flow separations on roofs and sides of buildings do not have clear limits to their
magnitudes [14];[58]) and can get as low as -12 in full scale [14].

The wake region is characterized by some pressure gradient with height. The
Bernoulli equation is of limited application in this turbulent and chaotic region
where streamlines are non-existent (see Figure 2-16). This notwithstanding, the
mean pressure in the wake is influenced by the mean velocity “along the line of
flow separation around the windward face of a building” [14 p71];[58]. It explains
why surface pressures in this region can still be expressed in dimensionless
forms with respect to the approaching wind dynamic pressure. These pressure
coefficients are invariably negative [58].

42
Figure 2-17 Cp (t) on low-rise buildings [40]

Figure 2-17 illustrates a typical variation of on a low-rise building where ̅


is the mean or time-averaged pressure co-efficient; , the root-mean-

square fluctuating value, or standard deviation from the mean; ̂ , the maximum
value of the pressure coefficient in a given time period; ̌ , the minimum value
of the pressure coefficient in a given time period [40].

2.3.1.2 Force Coefficient

Similarly to the pressures, forces on buildings can also be defined in a non-


dimensional way as in Equation (2-11).

(2-11)

By convention similar to that of aeronautics, these forces can be resolved into


two axes system as illustrated in Figure 2-18. In the wind axis system, the
forces are parallel and perpendicular to the mean wind direction. In the body
axis system, they are parallel and perpendicular to the geometrical axis of the
building. The cross-wind force and the along-wind force components are
referred to as the lift and the drag respectively [40].

43
Naturally, buildings are attached to the ground in the vertical plane. This fact
explains the high lift force on the roof as there is no chance for a flow under the
building to develop a counteractive downward acting lift force [14].

The drag force is the integral sum of the viscous skin friction force and the
pressure drag resulting from the pressure difference between the windward and
leeward sides of the building. Skin friction forces are proportional to the flow
velocity but inversely proportional to the size of the body in the flow. Skin friction
drag on a building is usually small compared to the pressure drag. Even so, it
can be significant on buildings with very long dimensions in the airflow direction
when there is the tendency for separated flow re-attachment to roof and walls.
Pressure drag is a function of the square of the flow velocity, the frontal area of
the building and the cross-sectional area of the building’s wake [14].

Figure 2-18 Resolution of forces on a building [27]

44
2.4 Building Codes: The Eurocode
Building codes and standards have often been the only contact point for
structural engineers (especially those designing low-rise buildings) and
estimations of wind loading on structures since they (building codes) started
coming into prominence in the years after World War II [40]. These codes of
practice (which are almost always legally binding) are often based on extensive
research. But for reasons which have been made clear earlier, they are usually
simplified models of wind loading based on quasi-steady assumption and using
mean or steady flow coefficients of pressure [39];[40]. This limits the level of
accuracy that can be expected from their loading predictions.

In most cases, building codes have been developed on a national basis.


However, with the growth of continental and regional integration fostered by
increasing world trade, the Eurocode has been developed and adopted by
member countries of the European Union;

“In , the Commission of the European Community decided on an action


programme in the field of construction, based on article of the Treaty. The
objective of the programme was the elimination of technical obstacles to
trade and the harmonisation of technical specifications.” [15 p5]

The Eurocode replaced all individual national building codes of member


countries of the Comité Européen de Normalisation (European Committee for
Standardization), (CEN) in . It, however, includes national annexes to
specify values for which it ‘inevitably’ leaves national choice open. Without the
national annex and corresponding translation of the Eurocode in the official
language of a member country, it cannot be used. The Eurocode (including
those parts not concerned with wind loading) consists of documents which
are in use in at least 28 European countries in which national annexes are
required. The end result is a confusing number of regulations [36]. It should be
noted once again that the relevant annex to the present study is the one issued
in the United Kingdom (UK) (see [16]).

45
Nonetheless, Holmes [40 p288] expresses optimism in the prospects of
internationally harmonised wind loading codes when he notes that “the growth
of world trade is expected to reduce the number of loading standards in use and
gradually force more consistency in their format and ‘content’ [emphasis added
on ‘content’ by the current author]”. Beal [20 p27] does not share this view by
his witty observation that:

“To use a Eurocode [inclusive of parts not concerned with wind loading], an
engineer must first read what it says, then read what the national annex
says, then work out what the result of combining them would be, then
memorise this (because it is not written down anywhere) and then remember
what it was that he or she was trying to design.”

Aylanc and Kasperski [13] also note the unfortunate loss of harmony desired in
the wind loading provisions of the Eurocode for member countries of the CEN in
the bid to provide transparency and ease of comparison between national
versions.

All the same, the Eurocode as well as other advanced wind loading codes and
standards invariably contain the following features to varying levels of details
[40]:

 Some basic or reference wind speed specification for various locations or


zones within the territory which the code applies at a height of usually
in open country terrain;
 Some modification factors for the effects of height and terrain types and
sometimes for change of terrain or wind direction or topography or for
shelter effects;
 Shape factors (pressure or force coefficients) for structures of a number
of shapes; and
 Some form of estimations of possible wind-induced resonant dynamic
effects on flexible structures.

In the Eurocode, there is distinction between ‘Principles’ (represented with the


letter P) which comprise of general statements, definitions, requirements and

46
analytical models having no alternatives and ‘Application Rules’ for which it is
right to make use of alternatives so far as they comply with relevant Principles.

The pressure coefficients in the Eurocode are specified by zones. These are
obtained with the use of a subdividing length (the scaling length) which is the
smaller value from times the height or the across wind dimension of the
building. The centre bay of the roof is subdivided by and into three
zones starting from the leading edge [13].

Figure 2-19 Division of walls and roof surfaces into zones in the Eurocode based
on the building dimensions and expected nature of flow around the building
(adapted from [15])

The zones represent regions of expected flow features such as stagnation


(zone D), separation (zone A, F and G), separation/attachment (zones B and
H), unstable re-attachment (zone C and I) and wake (zone E).

47
The specified values of the pressure coefficients are for the orthogonal wind
directions , and . The values represent the most unfavourable values
obtained in a range of and can be used for wind directions either side
of the relevant orthogonal direction [15].

These pressure coefficients are specified for loaded areas of and


as , for local coefficients, and for overall coefficients, respectively [15;
36]. The overall pressure coefficients on the vertical walls are primarily required
for the overall structural design of the building. The local coefficients are
significant for cladding design. The relation between these values has also been
specified. However, sometimes some national annexes such as the UK NA
specify same values for both [16].

The basic wind speed specified (on which the coefficients are based) is a
mean wind speed at height in open country terrain with an
annual probability of exceedence of ( -year return period) [40].

The service level wind loads specified with this basic wind speed are combined
with other loads on a building. Load factors (partial safety factors) to be applied
to achieve an acceptable small risk of failure are specified [36]; [46]. These are
not given in the but they are described in ‘Basis for
Design’. The values to be used for these are specified in National Annexes of
. So, they may have varying values across countries [36].

Davenport [31] who presented a paper on the relationship of reliability principles


to wind loading gave a good insight into how the load factor can be influenced
by the various uncertainties as expressed by the coefficients of variations of the
variables that are inputs to the wind loading estimation [46]; [31]. He notes that
the unreliability of wind loading estimations from codes stem from three
groupings of uncertainty: uncertainties from the inherently dynamic nature of the
wind and the consequent loading; uncertainties in the available knowledge of
the precise modelling of the wind loading process; uncertainties introduced by
the deliberately simplified wind loading models in the codes to make them
amenable to engineering use [31]. He asserts further that though these

48
uncertainties will vary with the extents of simplification, they can be ameliorated
with skilful development of standards which give maximum information with
minimum complexity. He concludes that these uncertainties, which map out the
extent of the unreliability of any set of design regulations and also where the
balance will lie between safety and economy with adequate considerations to
the costs of insurance and strength, are presently dealt with by implicit
conservatism in the choice of design values of load and strength and by explicit
safety factors or margins. Thus, he observes that the currently available building
codes could do more to promote close examination of these uncertainties taking
an extensive advantage of a systematic evaluation process.

2.4.1 Scope of the Eurocode


The Eurocode “gives guidance on the determination of natural wind actions
for the structural design of building and civil engineering works for each of
the loaded areas under consideration. This includes the whole structure or
parts of the structure or elements attached to the structure, e.g. components,
cladding units and their fixings, safety and noise barriers.” [15 p9]

It is only applicable to buildings and civil engineering works with heights up to


and to bridges having no span greater than so far as they satisfy
the criteria for dynamic response. It is concerned majorly with the wind loading
on land based structures and their components as well as appendages.

The writers of the Eurocode submit that

“…the Eurocode standards provide common structural design rules for


everyday use for the design of whole structures and component products of
both a traditional and an innovative nature. Unusual forms of construction or
design conditions are not specifically covered and additional expert
consideration will be required by the designer in such cases.” [15 p6]

The building shapes, roof forms and structural elements dealt with in the
Eurocode are, even though wide-ranging, generic. They do not feature ‘special’
or irregular forms (or combinations of these forms and the effects these

49
combinations might have on the wind loading) some of which will be discussed
in Section 3.

Geurts [36] informs that multitudes of the loading coefficients in the Eurocode
have been obtained with wind tunnel studies dating back to the .

2.5 Principles of Wind-tunnel Experiments


Reports of primitive wind tunnel studies on bluff bodies representative of
buildings such as that of Kernot ( ) and Irminger ( ) predate the advent
of the two World Wars which ushered in the rapid development of the wind
tunnels for aeronautical applications[18; 25; 40].

Now for over half a century ago, ABL wind tunnel experiments have been
primarily used to derive the information in the building codes. They have also
served as the ‘lender of last resort’ when it became necessary to have more
precise wind loading estimates than obtainable from the codes or where the
scopes of these codes fall short. After pointing out the dilemma of the wind
loading codification for low-rise buildings (see Section 1.2), Apperley et al. [10
p208] suggest that “the best that can be done is to develop some simplified
loading models which provide reasonable bounds on the likely loads, while still
incorporating the major parameters found to be important”. They contend that
“wind tunnel testing is the only method available to provide the mass of data
from which to derive such simple models.” [10 p208]

2.5.1 Field of Application of Wind tunnel Experiments


In accordance with the defined scope of the Eurocode and of other notable
building codes such as the ASCE-7(the building code produced by the
American Society of Civil Engineers) in the prediction of wind loading, wind
tunnel experiments are directed to be undertaken when buildings:

 possess shapes basically not of the traditional or simplified generic forms


covered in the codes and will not be amenable to innovative use of these

50
codes. This may be with regards to the shape of the plan but also, to the
height of the structure; and
 are located in such complex environments as to warrant a detailed
evaluation of the interaction effects which may lead to reduced or
increasing wind loads.

Wind tunnel experiments are also used in the review of existing codes, in the
search for data to be incorporated in future versions of these codes and in the
validation of methods such as computational wind engineering and other
numerical methods [34-36].

Boundless as they may seem, wind tunnel experiments are not used to estimate
[36]:

 internal pressure coefficients;


 friction coefficients;
 effects of pressure equalisation; and
 dynamic forces on slender structures with limited stiffness, such as
cables, bridge decks and flexible roof coverings.

2.5.2 Guidelines to Wind tunnel Experiments


Where it has been considered necessary to carry out a wind tunnel experiment,
most building codes including the Eurocode and ASCE – 7 give guidelines and
demands on how to go about it. These also cover how best to analyse the data
obtained from it. These are discussed substantially in Appendix A: Principles of
Wind tunnel Experiments in detail. However, a general demand is that the
report of a wind tunnel experiment carried out should contain ways in which the
researcher(s) was (were) able to meet the demands in the following guidelines
[36].

The UK national annex (UK NA) to the Eurocode demands that:

“Tests for the determination of wind loads on static structures should not be
considered to have been properly conducted unless:

a) the natural wind has been modelled to account for:

51
 the variation of mean wind speed with height above ground
appropriate to the terrain of the site; and
 the intensity and scale of the turbulence appropriate to the
terrain of the site at a determined geometric scale;
b) the building has been modelled at a geometric scale not more than
the following multiples of the geometric scale of the simulated
natural wind, with appropriate corrections applied to account for any
geometric scale discrepancies within this range:
 for overall loads; and
 for cladding loads;
c) the response characteristics of the wind tunnel instrumentation are
consistent with the measurements to be made;
d) the tests enable the peak wind loads with the required annual risk of
being exceeded to be predicted.” [16 p2-3]

The measurements of pressure are effected as pressure differences between


the building surface and a reference pressure [36]. Guerts and van Bentum [36]
who give a good insight into the general demands of a well performed ABL wind
tunnel experiment in their paper note the minimum demands that:

 The position of the reference pressure is chosen so, that this pressure is
independent of wind direction or changes made to the model; and
 The frequency-response characteristics of the pressure measurement
equipment need to be available on request.

Geurts and van Bentum [36] reports further on the demands for local loads:

 that a selection of locations of expected increased local loads be


made; and
 that measurements of pressure at sufficiently high frequency be
made such that extreme loads which corresponds to loaded areas
designated in the codes can be determined.

In addition, when the measured pressure will be used to obtain the overall wind
loads on the whole structure then:

52
 All surfaces should be fitted with large number of pressure tappings;
 The contribution of friction should be provided for with the use of the
appropriate sections of the building code; and
 Pressure measurement at all tappings should be simultaneous to keep
the time information.

The wind induced forces and moments on buildings and most bluff bodies can
be obtained from the integration of surface normal pressures measured since
the contribution of these far exceeds that of the shear stresses [27]. In such a
case, Cook [27] explains that the pressure tappings should be spaced
sufficiently close together so that each measured pressure is representative of
the ‘tributary area’ around the tapping. The tappings and their tributary areas
illustrated in Figure 2-20 can be uniformly distributed over the surface to obtain
the forces or biased distributed by varying the number and spacing of the taps,
thereby varying the relative tributary areas, to take into account other wind
effects such as moments [27].

Figure 2-20 Pressure tappings with tributary areas shown a. Uniform


distribution b. Bias distribution [27]

53
2.6 Computational Wind Engineering
Computational Wind Engineering (CWE) entails the application of CFD methods
to wind engineering. CFD is a numerical and computer-based method of
getting solutions to numerous types of fluid flow problems for engineering and
scientific applications [25].

The complexity of wind flows around buildings have been made clear and the
dependence of the fluctuating surface pressures of such buildings on turbulence
characteristics of the likes of intensity and integral scale have been explained
(Section 2.2.2). The varying length scale of turbulence has also been talked
about (Section 2.1.3.1). As such, present CFD methods have only been found
to produce acceptable results in predicting mean surface pressures on
buildings. These methods which incorporate simplifications of turbulence in the
fluid flow equation have not been able to predict the fluctuating and peak
surface pressure values and distributions to any reasonable accuracy. This is
because in order to be able to do that, CFD must adequately simulate the inflow
turbulence scales and structure, appropriately resolve these scales with a
detailed numerical mesh and correctly resolve the turbulence closure problem
(accounting for turbulence modes too small to be captured by the mesh).
Attempts to do these are limited by computer memory and speed [40];[33]; [25].

Nevertheless, Large Eddy Simulation (LES) is one CFD method that has given
better correlations with carefully obtained wind tunnel data. It is noteworthy that
it requires more computation than other CFD methods [14];[33]. CFD
techniques have also given acceptable accuracy where insights into the nature
of flow around a building are sought for pedestrian comfort analysis and
environmental considerations [40];[33]. The interested reader is referred to the
references cited for more details.

54
2.7 Relevant past works on the present study
Given that the Eurocode is a fairly recent adopted document in , there
appears to be very limited review of its provisions for wind loading in the
relevant literature. Even in the few reviews perused, researchers have found it
less burdening to focus on the annexes of the Eurocode relevant to them for
reasons of nationality and/or domicile. This is the approach taken by the current
author as well. However, it is worthy of note that some reviews have predated
the official adoption of the Eurocode.

Some of such reviews are the works of Geurts [35] and of Aylanc and Kasperski
[13].

Geurts [35] conducted his review based on the Dutch annex (In reality, his work
was based on the Dutch code with references to the Eurocode. However, it is
expected that the Dutch code would be the source of information contained in
the Dutch Annex to the Eurocode) and with both full scale and wind tunnel
model experiments on the building in Figure 2-21. The full scale experiment he
undertook presented him a chance to assess the code provisions for building
cladding. That is to say, he was able to review the internal pressure coefficients
as well as the external ones.

Figure 2-21 Outline of the test building used by Geurts [35]

55
He used the mean values of the pressure coefficients he obtained from his
experiment since these are the only ones used in the codes. He concludes that
the wind loading provisions of the codes are generally not valid for buildings.
This is due to the assumptions of quasi-steady theory which are necessary to
provide the simplified models contained in them. Davenport, in , already
noted that the penalties incurred by the adoption of these necessary
simplifications in building codes are related to more variability in the ratio of
actual load to code specified load [31]. Howbeit, Geurts [35] went on to
conclude that the errors are generally small but can work, usually uncertainly,
against safety or economy.

Aylanc and Kasperski [13] based their work on the German annex. Their
analysis was based on the extreme value method. They assessed the code
specified values for the local external pressure coefficients. They found
mismatches between wind tunnel peak pressures, measured on the building
model illustrated in Figure 2-22, and code predicted values with the former
larger than the latter. This was especially pronounced at corners of roofs.
Nonetheless, they observed the “astonishingly good agreement” between these
values at the mid-region of the roof (zone ) [13 p10].

They also found faults in the specification of uniform values of pressure


coefficients for the designated zones in the Eurocode as they noted that peak
pressures in a zone may, actually, be considerably different from these uniform
values; this was especially the case in the intermediate zone on the side wall.
Such deviations were observed on both the uneconomic and unsafe sides.

Probably faced with a similar limitation on the number of pressure tappings to


be used as that of the present study, Aylanc and Kasperski [13] took advantage
of expected symmetry of flow around their block model and placed the pressure
taps on just one half of the model. This means that the pressures they
measured and analysed for a complete face for a particular wind direction were
not from the same time-series.

56
Figure 2-22 Aylanc's and Kasperki's building model [13]. The dash lines on the
model signify the tapping positions.

Even more limited are the relevant past works to the assessment of the effect of
varying tributary areas of pressure tappings on the wind loading obtained by the
integration of surface pressures. That of Lin and Surry [52] is noteworthy.

Lin and Surry [52] explored the variation of peak loads with tributary area near
roof corners of flat low building roofs for a number of building heights and roof
dimensions depicted in Figure 2-23. They observed that in these areas of the
roof, including the corner point, “the load reduces rapidly with increasing
tributary area from the very high local peak value at the corner”. They conclude
that their “results demonstrate that the relationship between the effective load
and tributary area is insensitive to the building dimensions if area is normalized
by where is the building height, and is only slightly dependent on the
shape of the areas that include the corner point.” [52 p185]

57
Figure 2-23 Lin's and Surry's building models’ dimensions [52]

Pintar [53] simulated the wind flow over the CU Sports Hall and its immediate
neighbouring buildings in the west for a westerly, depicted in Figure 2-24, and a
south-westerly wind. The aim of his study was to “investigate the feasibility of
utilising wind energy for Cranfield University” [53 p3]. While deliberating on how
best to simulate the turbulent flow, he notes that “turbulence model is a set of
equations which removes the need to resolve all scales of turbulent eddies” [53
p19]

The simulation was carried out with STAR CCM+ using a “k-ε” turbulence model
with a high “Y+” wall treatment method [53]. In the validation of his method with
experimental results of fluid flow available for a different block model, he notes
that “the results almost match in regions above [the] building where the flow is
simple. In regions of high turbulence [that is closer to the ground surface] there
is a large difference between results.” [53 p21] The simulation used

58
cells and converged at about iterations after hours for each of the wind
directions he studied!

Figure 2-24 Simulation of a Westerly wind over the Cranfield University Sports
Hall (signified by S H here) (adapted from [53])

2.8 Chapter Summary


The ABL wind is a “dynamic and random phenomenon in both time and space”
[21 p16]. As such, exact predictions of the wind speed at any specific time of
interest is not possible. This can only be satisfied by statistical means. When
strong wind speeds are about or exceed , mechanically generated
turbulence (by surface roughness) dominates in the ABL. The mean wind
velocity profile can be represented by the power or log law. The dominant wind
direction is and an average wind speed of is found to be
appropriate for Cranfield (Section 2.1).

Buildings are considered bluff bodies in aerodynamics. The predictable flow


separation on bluff bodies with sharp edges is fixed by body geometry and
not . This is not the case for circular bluff shapes or those with curved
surfaces. In any case, effects are generally ignored in bluff body
aerodynamics on grounds that most of the relevant shapes in the field are sharp
edged. However, prominent and well-informed researchers including Hoxey and

59
Irwin have concluded on possible sensitivity of flow around sharp edged bodies
and buildings to .They have called for more research geared towards
quantifying this effect for wide-ranging building shapes (Section 2.2 ).

Buildings can be classified tall or low-rise with the distinction usually set at
height. Low-rise buildings are almost always completely within the ABL.
Appropriate modelling of surface pressure distributions over these buildings
depend on how well , turbulence intensity and turbulence scale are scaled
(Section 2.2).

Wind loading is the dominant horizontal force on buildings. It can be estimated


from pressure coefficients or force coefficients. By convention, positive pressure
coefficients indicate pressures acting towards a surface (attached flow) and
negative pressure coefficients indicate pressures acting away from a surface or
suction (detached flow) (Section 2.3).

Building codes including the Eurocode contain specified values of these


coefficients for wind loading estimation on a number of generic shapes of
buildings and structures. The Eurocode is in use in at least European
countries. Each country has a national annex to the Eurocode. The variations of
building shapes covered in the Eurocode do not begin to spell out the present
complex reality of building (external) shapes and the pieces of information from
which they were made were acquired from wind tunnel studies carried out in the
to the (Section 2.4).

Wind tunnel studies have improved since these times. They offer the better and
more reliable option to wind loading estimates from codes. Where the codes
including the Eurocode have recommended wind tunnel experiments, guidelines
which should be followed for such experiments have been provided. Wind
loading estimates on buildings can be obtained from wind tunnel studies by
integration of measured surface normal pressures (Section 2.5).

CFD methods applied to wind engineering have not been particularly successful
where prediction of fluctuating and peak surface pressures is concerned. There
has been some reasonable success in the prediction of surface mean flows and

60
pedestrian wind comfort around buildings. LES is one such CFD method that
has proven to be promising with the results obtained from its application but ,
generally, the wind engineering community has legitimate reasons to be
cautious about CWE (Section 2.6).

Past reviews of the Eurocode with wind tunnel and/or full scale studies
examined have predated the official adoption of the Eurocode in . They
have also been based on the annexes in use where the researchers are based.
General mismatches, albeit small, were found between the code reported
coefficients and experiment values. These were especially true of corners of
roofs. The sign of the errors were found to be uncertain. Researchers also
found faults in the sub-division of side walls into zones in the Eurocode while
noting that peak pressures in a zone may be significantly different from those
specified uniform values in the code. This was especially true of the
intermediate zone . However, some researchers found very good agreement
between the Eurocode and wind tunnel experiments at the mid-region (zone )
of the roof. Researchers who assessed effects of tributary area of tappings on
the wind loading estimate observed the rapid reduction of the load with
increasing tributary area from the very high local peak value at the roof corners
they examined towards the lower values in the inner regions of the roof. Results
from CWE can take a very long time and very huge computational space to be
obtained (Section 2.7).

With the completion of this present section objectives 1, 2 and 3 have been
fulfilled. It is considered that the research has been positioned to fulfil objectives
4 and 5 in the following chapter.

61
3 Research Methodology
“Theoretical fluid dynamics, being a difficult subject, is for convenience, commonly divided into
two branches, one treating of frictionless or perfect fluids, the other treating of viscous or
imperfect fluids. The frictionless fluid has no existence in nature, but is hypothesized by
mathematicians in order to facilitate the investigation of important laws and principles that may
be approximately true of viscous or natural fluids.” Albert F. Zahm, 1912 (Professor of
aeronautics and developer of the first aeronautical laboratory in a U. S. university, The Catholic
University of America) [8 p189]

3.1 Case Studies


In a much similar way to the subject of the above excerpt, a cuboidal building
having a rectangular plan (sometimes with a duo-pitch roof in the case of a low-
rise or flat roof in the case of a tall one) such as have been discussed in Section
2.2.2 always serves its purpose of simplifying the subject of building
aerodynamics. While structures with such simplicity in shape have been built, in
contrast to the non-existence of the frictionless fluid, a quick survey of the
buildings wherever the reader is (or perhaps more conveniently from the
vantage point view presented of some locations in the London city centre within
the red rectangles in Figure 3-1) would reveal most likely the complex reality of
building (external) shapes. A non-expert may find it difficult and confusing to
begin to classify the variety of building shapes and forms existing in this reality.
Interestingly, however, experts in wind engineering and building aerodynamics
also experience difficulty in making this classification especially where
presentation of loading data in a logical fashion in building codes is concerned
[27].

63
Figure 3-1 An aerial view of a section of the London city centre -The Gherkin is
marked 'A' [1]

Nonetheless, some form of division has been attempted. This stemmed from
the fact that some specific building shapes and form stand out clearly and aid
the process of classifying them and intermediate forms into major classes [27].
Cook [27] presents two main classes thus:

 Curved structures: This constitutes all structures that are mainly curved
even though they may possess some flat sections; and
 Flat-faced structures: This constitutes all structures that are mainly flat
faced even though they may possess some curved sections.

64
He went on to present some further sub-divisions of these classes appropriately
into ‘roofs’ and ‘walls’ (which may be vertical or inclined).

Figure 3-2 Irregular building plans a. Re-entrant corners b. Recessed bays c.


Central wells d. Inset storeys (adapted from [27])

The external walls of a building may be combined to form regular or irregular


polygons of different numbers of sides; to form re-entrant corners translating
into L-, T-, V-, X-, Y-plan shapes (depicted in Figure 3-2a for an X-plan); to
form recessed bays having H-plan shapes (illustrated in Figure 3-2b) or more
complex plan shapes; to form internal corners opening at the roof to create
central wells (as seen in Figure 3-2c); inset faces and storeys of varying levels
of complexity to the one presented in Figure 3-2d.

It is also a possibility to have combinations of two or more of these highlighted


shapes and forms as a second look to Figure 3-1 would reveal.

Detailed analysis is presented in reference [27] which gives some possible


variations of these forms. In this text, Cook also gives descriptions of the nature
of flow that may be expected around these building forms. These do not fall
immediately in the scope of this study. However, it is interesting to apply the

65
“Principle of Minimum Energy” to the wind which can be considered to be
general for wind flows [51 p30 -31]; [27]. Lawson [51 p30-31] puts it in simple
terms as: “Wind is lazy, and will always take the easiest path”. He notes that it is
never the case that “all the wind will go this way and none the other”, but, given
two possible paths with varied ratio of difficulty (pressure or friction losses), the
proportion of wind going one way will be inversely proportional to the ratio of
difficulty of following that path as opposed to the other.

In fulfilment of objective 4 (see Section1.3), buildings were selected on the


Cranfield University campus as illustrated in Figure 3-3; -The Sports Hall; -
Building 83, Hangar Building; and – The Trafficmaster Building.

This choice of buildings and the Cranfield terrain were considered to be


adequate for the present research. The Cranfield Campus site can be correctly
categorized as an ‘open’ terrain (see Figure 2-3). The Eurocode has been
reported to specify reference wind speeds for ‘open’ country terrain (see
Section2.4). It contains far-reaching information on simple cuboidal building
shapes such as that of the Sports Hall in Figure 3-4. More so, it would be
interesting to compare experiment results on the Sports Hall model with Pintar’s
[53] CFD results.

The Hanger building, in Figure 3-5, was observed to possess a good enough
intermediate level of difficulty since it is just a combination of vertical walls. The
faces outlined in red ink, in Figure 3-7a, are inset faces to their adjoining lower
roof surfaces outlined in blue ink in the same figure. Cook [27] explains that an
adjoining lower roof serves as the ground plane for an inset face. The flow on
this face interacts with that on the lower roof such that when the face is
windward, positive pressures are experienced on the lower roof. The extent of
this region of interaction is defined by some fraction of the scaling length of the
inset face. However, the Eurocode specifies loading data for roofs having no
such flow interactions.

66
Figure 3-3 Building form selection from Cranfield University Campus [7]

The idea was that the Trafficmaster Building, in Figure 3-6, provides the
combination of forms which may just pose some more difficulty to the Eurocode.
It is a re-entrant corner with a V-plan and with circular faces at the internal and
external junctions of the two wings of the building. When the wind flow over the
building is in the direction of the black arrow in Figure 3-7, the circular face at
the internal corner outlined in red ink is essentially a windward face with positive
pressures. The positive pressures on this face act on the downwind end of the
face adjoining it from the upwind direction which is supposed to be a side face
experiencing suctions [27]. More so, in the same figure, local high-suctions do
not occur on the face outlined in blue ink for winds in the direction of the green

67
arrow over the building. In general, Cook [27] points out that surface pressure
distribution on faces of re-entrant corners is dependent on the whole plan
shape. This shape is not dealt with in the Eurocode.

Figure 3-4The Sports Hall viewed from the South-West (picture taken by the
current author)

There are numerous roof forms including the pitched (duo-pitch, mono-pitch,
mansard and multi-pitch, hipped, skew hipped), the hyperbloid, barrel-vault and
flat roofs. A flat roof may have sharp eaves with or without parapets; or curved
or mansard eaves [27]. Flat roofs with sharp eaves are of significance to the
present study (because the Sports Hall is flat roofed).

68
Figure 3-5The Hangar building (in the fore-ground) viewed from the South-West
(picture taken by the current author)

Figure 3-6The Trafficmaster building viewed from the South-West (picture taken
by the current author during a flight test)

69
Figure 3-7 Wind tunnel models of a. Hangar building and b. Trafficmaster
building

3.2 Models construction and Pressure tapping locations


With the exception of the model of the Hangar building which had been
constructed prior to this study, information regarding the geometric dimensions
of the buildings was sought and gathered from the Estates services of Cranfield
University.

Figure 3-8 Constructed 1:200 scale Aerodynamic models of the case study
buildings fitted with pressure tappings

70
For reasons discussed in details in Appendix A.1 , the aerodynamic model of
the Hangar building, illustrated in Figure 3-8, was constructed at a scale
of the real building. Giulbauld [12] who also gathered information regarding the
geometric dimensions of this building from the Estates services of Cranfield
University, has included the information in his report [12]. This model appears to
be a good representation of the Hangar building at such a scale. Aerodynamic
models of the Sports hall and the Trafficmaster building were also constructed
at the Cranfield workshop at this scale (refer to Appendix A.4 for a brief
discussion on model construction). More detailed information about the
dimensions as well as pressure tappings positions in non-dimensional forms is
included in Appendix B.

The locations of the tappings on the models were arrived at after a number of
considerations. On the one hand, the demands on an acceptable wind tunnel
experiment presented in Section 2.5.2 require an adequate number of tappings
on ALL external surfaces of the building considered for wind loading estimation.
Additionally, the tappings should be present in designated loading areas defined
in the codes. Fulfilling these would require a large number of pressure tappings
considering that the designated loaded regions on the buildings vary with wind
direction. On the other hand, the number of tappings which can be used for this
study is limited to (see Limitation1). By contrast, Aylanc and Kasperski [13]
used tappings concentrated on a half side of their simple block model in
Figure 2-22. Similarly, Lin and Surry [52] used a total of tappings
concentrated just in ‘one’ roof corner region of their block models (see Figure
2-23). On a different level, the guidelines in Section 2.5.2 demand that the
pressure measurements to be used for overall wind loading estimation should
be taken simultaneously so as to ensure that measured pressures on all
surfaces are of the same time series and the problem of non-correlation of
fluctuating pressures can be taken out. More so, in tune with the guidelines,
regions of expected increase in magnitudes of local loads (such as roof corners
and upwind corners of side walls as Section 2.7 reveals) and also other regions
of interest like the intermediate loaded zone designated in the codes were
desired to be fitted with adequate tappings.

71
In view of the aforementioned, it was the view that the present study must
inevitably be carried out with a close scrutiny of possible symmetry of flow
around the models and the prevailing wind direction in Cranfield (which is ).
It is worthy of note that all the models were fitted with tappings except the
Trafficmaster building model which was fitted with one less.

The Sports Hall model being a cuboid was expected to have a symmetrical flow
nature around it. Accordingly, only the roof and the west and south faces were
tapped (see Figure 4-7 for cardinal point orientation). The Trafficmaster building
is also symmetrical about an imaginary centreline drawn at the centre of the
curved surfaces. Thus, the tappings were concentrated on the roof and walls of
the western half of the building (see Figure 4-7 for cardinal point orientation).
However, some tappings were put in limited areas of the other half adjoining the
curved parts to check for symmetry assumed in the nature of flow around the
building. The Hangar building is not symmetrical at all. A very complex flow
nature was expected over the roof and the entire building model roof as
explained earlier. For instance, Cook [27] explains further that the interaction of
two crossing flows at their point of convergence close to the internal corner of
an inset face which is flushed to the side of the building, such as depicted in
Figure 3-9, causes very high suctions on the roof of the low part and the side
wall of the higher inset part [27]. The suction peaks are not correlated across
the entire roof surface and the whole building. As a result, it was considered
better to obtain ‘complete’ information (of the same time series) of the complex
flow per wind direction at least in one of the directions relative to the wind.
Hence, the distribution of the tappings only on the roof (to obtain ‘complete’
across-wind information in the x-y plane) as opposed to any other possible
distribution of tappings given the constraints already highlighted.

72
Figure 3-9 An elevation view of the Hangar building model focusing on one of its
inset faces (the approaching wind is assumed to be flowing into the present
page. Its directions of flow at this corner of interest are signified by the green
arrows)

These choices made have not been without consequences; relatively large
areas on the models highlighted with the red ellipses in Figure 3-8 as well as in
other parts not highlighted have been left without tappings. These may appear
unsatisfactory to a well-informed wind engineer on how suitable the pressures
measured therewith would be for this study. Be that as it may, it was conceived
that a suitable method of looking at and analysing the data obtained with the
choice of the tapping locations adopted would satisfy the aim of this study.

3.3 Research methodology


As shall be evident in the next chapter, the data available from the wind tunnel
experiments is time-averaged. Consequently, the quasi-steady method of data
analysis was adopted (refer to Appendix A.5 for a brief discussion on this).

The Sports Hall model, being the simplest, was used as a bench mark test for
this research. It was expected that if the performance of the Eurocode
prediction of wind loading on this simple building shape was less than desirable,
the observation would cut across the other buildings. However, in the case that
it was otherwise, it might not be expected to be so for the other chosen building

73
shapes. In such a case, it was then important to carry on further assessment of
the other buildings.

3.4 Chapter Summary


Modern building shapes and forms are combinations of different forms. They
may not be exactly amenable to any form of classification from the first sight.
Naturally, the vast possibilities of these building shapes and forms are
significantly different from a simple cuboidal building (or other generic forms in
the Eurocode for that matter) in geometry as much as in the nature of flow
around them. In fulfilment of objective 4, 3 buildings were selected on the
Cranfield University campus: -The Sports Hall which is a simple cuboidal
building; –Building 83, Hangar building which incorporates inset faces in its
plan and elevations; and –The Trafficmaster building which is a re-entrant
corner with a V-plan having curved surfaces at the internal and external
junctions of its wings (Section 3.1).

Aerodynamic models of these buildings were constructed at scale and


fitted with tappings each as a fulfilment of objective 5 (except for the traffic
master building which was fitted with ). The locations of the tappings on the
models were arrived at after such considerations as the demands of an
acceptable wind tunnel experiment, the limitation on the number of tappings to
be used, close scrutiny of expected flow symmetry around the models and the
prevailing wind direction in Cranfield (Section 3.2).

The quasi-steady method of data analysis was adopted. The Sports Hall model
would be used as a benchmark (Section 3.3).

The research is better positioned to fulfil objective 6 in the following chapter

74
4 Experiment
“I do not see then, I admit, how one can explain the resistance of fluids by the theory in a
satisfactory manner. It seems to me on the contrary that this theory dealt with and studied with
profound attention gives, at least in most cases, resistance absolutely zero: a singular paradox
which I leave to geometricians to explain.” Jean LeRond d’Alembert, 1768 [8 p791]

Experiments are usually undertaken to clear doubts and provide answers to


theoretical questions. Many a times, however, experimental results may serve
to confuse the inadequately informed researcher the more. The experiments
undertaken in this study were carried out following the guidelines already
presented in Section 2.5.2 and discussed in more details in Appendix A. Where
relevant and applicable, references are made to the stated areas of this thesis
either for more details on facilities and equipment used or on decisions taken or
for reasons for taking these decisions.

4.1 The Cranfield ABL Wind Tunnel


All the experiments were carried out in the Cranfield ABL wind tunnel illustrated
here in Figure 4-1.

Figure 4-1 Schematic representation of the ABL wind tunnel facility at CU [54]

It is an open return circuit wind tunnel comprising of an axial three-bladed fan


powered by a DC electric motor; a settling chamber with a
contraction ratio; and a long flow fetch of upstream of a closed test section
(depicted in Figure 4-2) [12; 54]. The cross-sectional area of the test section
is . The roof of the tunnel is not adjustable nor is it with a constant
slope. Also, the test section does not feature slotted walls with a plenum
chamber (refer to Appendices A.1 and A.2 for the significance of these).

75
Figure 4-2 Isometric view of the test section [12]

The circle in the middle of Figure 4-2 represents the automatically operated
diameter turn table in the test section capable of rotating a full degrees [12;
54]. The maximum achievable wind speed in the tunnel is for the set up
used.

The control room, where the monitoring system that has the Labview software
installed is situated, is next to the working section.

4.2 Simulation of the Cranfield Atmospheric Boundary Layer on


the models
Wind tunnel simulations of the ABL with the use of Roughness, barrier and
mixing devices methods are usually obtained by trials and errors (refer to
Appendix A.1 for more details). Therefore, once simulations of the boundary
layer over specific locations of interest having good agreement with ESDU data
have been achieved, wind tunnel owners and operators usually keep the
simulation technique and data. These are usually available on request.

The School of Engineering which has operated the wind tunnel for many years
has such ready-made information for Cranfield and other choice locations.

76
Figure 4-3 Set up of the wind tunnel for simulating the Cranfield ABL

The set-up in Figure 4-3 uses uniformly distributed lego blocks on the fetch floor
to represent the surface roughness. The elliptic wedge vortex generators which
produce the initial momentum deficit can be seen far upstream. The barrier
which induces turbulence is not visible in this figure.

120

100
Height FUll SCale m

80
Wind tunnel V P
60
ESDU V P
40 Wind tunnel T P
ESDU T P
20

0
0.000 0.500 1.000 1.500
U/U10

Figure 4-4 Comparison between the wind tunnel data and ESDU predictions for
wind velocity profile (V P) and turbulence profile (T P)

The wind velocity and turbulence profiles of the generated boundary layer are
presented in comparison to ESDU estimates in Figure 4-4. The boundary layer

77
was modelled at a geometric scale of to full scale and with
(full scale).

Figure 4-5 The Setup of the Sports hall and inset a. the Trafficmaster building b.
the Hangar building showing the Aluminium tape and the Pitot-static tube

The models were attached to the tunnel floor with strips of aluminium tape
which has a very strong grip. In addition, the Hangar building model was
screwed to the tunnel floor through two screw holes on its roof.

Figure 4-6 Tubes from pressure taps fitted on to the transducers located under
the turntable

The dynamic pressure and reference wind speed were measured


simultaneously as the model surface pressures with a pitot-static tube

78
positioned at no less than away to the side of the models at its closest to
any part of the biggest model (the Hangar building model). This is depicted in
Figure 4-5. Its vertical position at to the tunnel floor corresponds to
in full scale.

The tubes from the pressure tappings on the models were taken through a hole
at the centre of the turntable under the models. They were fitted on to the
transducers as illustrated in Figure 4-6. The tubes had been blown through with
a hand pump and numbered before-hand to ensure they were not
blocked and that they were not mixed up. They were also made long enough to
avoid them crippling when the turn table rotates through large angles. This is
not ideal as the transient response of the transducers is dependent on the
length of the tubes (see Appendix A.3). However, it was very necessary due to
the limitation of the setup used.

The pressure transducers were built in-house at the CU school of engineering.


They were calibrated prior to the undertaken of the experiment. Five sets of
these transducers which are of similar kind were used. Each transducer takes
tubes. The first transducer was connected to tappings , the second to
tappings and so on to tapping . The extra ports on each transducer
were connected to the reference static pressure sensed by the pitot-static
probe. The pitot-static probe was also calibrated prior to the experiment.

4.3 The Procedure


The models were orientated, with the aid of Figure 4-7, to the approaching wind
from the tunnel fetch with North at zero degrees. After set-up is complete, the
test section was closed and locked. A measurement was taken at zero wind
(with the fans not in operation). Then the fan was put on and ramped up till the
mean wind speed is .

79
Figure 4-7 Orientation of the buildings relative to North (adapted from [7])

The atmospheric temperature and pressure were regularly updated in the


Labview software used for acquiring the data. Once the test was started, a set
of measurements was taken for wind directions starting from zero and
rotating left in angular progressions of degrees. The rotation was done
automatically.

For each building, runs returning mean values of samples each were
taken per wind direction and at over periods in order to accurately
capture the mean value. Each reported value is the average of the runs.
With a geometric scale of , the corresponding calculated values of the
frequency and period for the full-scale are presented in Table 4-1 (refer to
Appendix A.2 for details of the equations used).

Table 4-1 Important parameters at wind tunnel scale and full scale

Parameter Wind tunnel scale Full-scale

Geometric dimension

Rounghness length,

Velocity

80
Sample frequency

Sample period

Number of Samples

Although the reported value of the average wind speed over Cranfield is
(see Section 2.1.5), it is clear that the wind speed used for the tests
was . This was a necessary piece of advice given by the supervisor of
this study to up the measured pressures and aid the response of the
measurement system. More so, the same flow pattern was expected on the
models within this range of speeds. In effect, the calculated velocity scale
is . There was not enough justification for any meaningful variation of wind
speed (and ) since the maximum wind speed achievable with the set up used
in the tunnel was .

4.4 Error Analysis


The Aylesbury comparative experiment (ACE) where full scale data and data
from a number of different wind tunnel studies on the same model of the
Aylesbury House were compared revealed that careful attention needs to be
paid to reference static pressure measurements. While fluctuating pressures
were unaffected, wind tunnel to wind tunnel comparison of the mean pressures
showed significant differences in the reference static pressure measurement
[56];[30];[62]. Dalley and Richardson report that it was suggested by Sill, Cook
and Blackmore that these static errors were likely due to the variations in the
reference positions used [30].

This was a major motivation behind the work of [30] who explored different
reference static pressure measurement locations in the wind tunnel to find the
most suitable. They found no best reference position for static pressure
measurement in the wind tunnel even though they would prefer to measure it at
the middle of the turntable in the working section when no model is present.
However, this is not too practical as in the ideal case; the dynamic pressure

81
measurement and the reference static pressure measurement should take
place concurrently with the model surface tap pressure measurement [30].
Finally, they recommend that when measuring the static pressures, the dynamic
pressures should be closely monitored and that large samples be taken to stem
out low frequency fluctuations.

Sill, Cook and Fang [56] conclude that their analysis of the results of the ACE
was suggestive that errors in reference static pressure are greater than those
measured at the surface taps on the model. Dalley and Richardson [30] note
the interesting observation that an error in the static pressure measurement will
carry through in the pressure coefficient appearing in both slope and offset. This
is due to the fact that both the numerator and denominator will be influenced by
the static error. They also cite the work of others like Greenway and Wood
( ), Goldstein ( ), Fage ( ), Wood( ), Tommree( ) and Shaw
in their discussion about what effects the tunnel gradient, static pressure field
around the model and turbulence can have on the pressure measurements.

A detailed survey of the static pressure variation in a boundary layer wind tunnel
flow would be required to reveal such effects as the inherent longitudinal
gradient of pressure in the free-stream (unless the tunnel has been specifically
designed to eliminate it (see Appendix A.2)), longitudinal changes in the
measured static due to boundary layer devices (this is irrespective of the wind
tunnel design) and cross tunnel variations of the static pressure possibly from
irregularities in the wind tunnel structure. In addition, a pressure reference
probe placed in the vicinity of the model may well be responding to the local
static pressure field around the model and not exactly that of the free stream
even at model heights away to the side of the model [30]. More so, the results
obtained by Dalley and Richardson[30] on the vertical variation of the static
pressure in the wind tunnel are suggestive of a correlation between turbulence
intensity and the static pressure errors within the boundary layer.

This effect of turbulence on the static error is the most difficult to quantify. It has
been shown, however, that the sign of the error is dependent on the curvature
of the surface in which the static hole is bored relative to the turbulence scale so

82
that a tap in a flat surface over-reads whereas that in the side of a pitot tube
under-reads. It is the conception that the magnitude of this error is dependent
on the turbulence intensity as well as the ratio of the turbulence scale to
curvature of the surface [30]. Moreover, the magnitude of this error has also
been shown to be dependent on the spacing as well as the size of taps
particularly when the flow turbulence scale is comparable in size to the pressure
tap [25; 30].

Given the importance of the Eurocode, it is most presumable that the wind
loading information contained therein has been corrected for most of these error
sources as well as for those of blockage effects and those inherent in the
measurement devices.

For simplicity sake, only the latter sources of error were assessed in this study.
The blockage ratio calculated for the Hangar building model which is the biggest
is less than (and less than which is the threshold above which
corrections are needed. See Appendix A.2)

The repeatability of the pressure transducers was analysed and assessed for
sensitivity to wind direction in Figure 4-8. To do this, pressure readings at
tappings and on the Sports Hall model representing the five
identical transducers respectively have been followed for the runs of the
experiment. Notwithstanding the high turbulence in the simulated ABL, it can be
seen that the transducers possess high precision which was not affected by
wind direction and/or position of tapping relative to the oncoming flow. As such,
the pressure values measured being the averages of the mean values of the
runs were expected to be very close to their true values detectable by the
transducers.

83
Figure 4-8 Repeatability analysis of the transducers

However, the accuracy of the transducers needed to be taken into consideration


as well to assess the inherent errors in measurement. The transducers are fitted
with HCLA0075...U pressure sensors with typical accuracy of (Full
Scale Span; see [3] for data sheet)

(4-1)
√ √

The dynamic pressure and the surface pressures were expected to have the
same amount of measurement error quantified in Equation (4-1) (where
standard deviation in the measured mean; published standard
deviation from data sheet ; and number of samples) in them. The maximum
error in the values analysed and discussed in this study (that is for the
and wind directions) was quantified to be using the error analysis method
contained in [50]. It should be noted that this is for the worst case scenario
when the accuracy of the transducers have been given as stated earlier.
However, it was gathered from the supervisor of this study that the accuracy of

84
the pressure sensors has been found to be within . In such case,
the maximum expected error in the values would be less than .

It is notable that the pressures measured are pressure differentials and not
absolute pressure values. Thus, what may be of equal importance is the
resolution of the transducers which is (analogue to digital) [3].

4.5 Chapter Summary


The ABL over Cranfield was simulated at the scale of with the use of
Roughness, barrier and mixing devices method in the Cranfield ABL wind
tunnel. The tunnel which does not feature any adjustable or constant slope
roofs is an open return circuit one with a fetch of . The walls are not slotted
and there is no plenum chamber incorporated. The reference dynamic pressure
and wind speed were measured at a vertical height of (equivalent
full scale) and no less than away to the left side of any part of the biggest
model (Hangar building model).

Five identical transducers were used. The models were set with North at zero
degree to the approaching wind at the start of any run. Runs of the
experiment were performed for each of the models. In each run, the mean
values of samples taken at over per wind direction were
recorded. A total of wind directions with a progression of starting from
zero degree were used. The mean wind speed used for the tests was .

The transducers have very high precision which is not affected by wind direction
and turbulence. There are many possible error sources identified. The errors in
measurement affect both the numerator and the denominator of the pressure
coefficients. It was estimated the largest error on either side of a reported
pressure coefficient in this study would be based on the worst case scenario
when the transducers were reported to have accuracies within .

It is the view that enough data was gathered to aid the fulfilment of the rest of
the objective.

85
5 Results and Discussion

“If you can’t explain it simply, you do not understand it well enough.” – Albert Einstein

A number of limitations were encountered in processing and visualising the


results. For instance, since the pressure tappings could only be located at some
distance away from the model surfaces’ corners and edges, measurements
from tappings closest to corners had to be extrapolated to the corners as
depicted in Figure 5-1 (The locations of the pressure tappings are signified by
the cross-hairs in this and every other relevant figure to this effect). In addition,
the Matlab code used to generate the plots works its way around the input data
such that it tries to present a logical variation of the pressures in between two
pressure tappings. Where the input data is limited, the result of this may go
against theoretical and practical reasoning. These mean that the contour plots
presented cannot be taken as absolute representations of the real flow
especially at the corners. Nonetheless, they were assessed to be good for the
purposes of this chapter so far as the attention is focused on the locations of the
pressure tappings.

Figure 5-1 Matlab generated contours of surface Cp for a westerly wind on the
West face of the SH model based on a. mean wind velocity at different heights
and b. mean wind velocity at 50mm

87
It is interesting to observe that the colour bar in Figure 5-1a extends to include
values beyond which is practically and theoretically incorrect. This is so
because the plot was made with pressure coefficients based on the calculated
mean wind velocity at the different heights of the tappings.

The dynamic pressure was only measured at the height of equivalent to


full scale. Consequently, the dynamic pressures at the different heights
of the tappings were calculated with the knowledge of the mean wind velocity
profile. These were estimated with Equation (5-2) using the rearranged log law
in Equation (5-1):

⁄ (5-1)

(5-2)

Apperley et al. [10] simulated the ABL up to the gradient height and reported
pressure coefficients on their model of the Aylesbury house estimated with the
dynamic pressure measured at this height. Using the square of the mean wind
velocity ratio as in Equation (5-1), they expected inaccuracies of the order of
in the mean pressure coefficients estimated therefore for the appropriate
comparison to full scale coefficients (which are based on dynamic
pressure). However, Holmes [38] appears to suggest that when only the surface
layer of the ABL is simulated and reference dynamic pressure (and mean wind
speed) taken at equivalent height of full scale (which is much closer to the
model) as done in this study, correction errors to the different heights on the
building may be avoided. In any case, the values specified in the Eurocode
are based on the mean wind velocity. Thus, where comparisons are made
to the Eurocode values, the values reported are based on the measured
dynamic pressure in the wind tunnel. The contour of such values illustrated in

88
Figure 5-1b can be seen to give more reliable agreement to theory. Thus, the
deficiency of the log law discussed earlier (in sub-section 2.1.4.1) is highlighted.

Where possible, a single colour bar is maintained for the contour plots.
However, there were cases where the use of different colour bars was
inevitable. This is to say attention should be paid to the colour bars in the plots.

In such places where comparisons are made, the Eurocode specified values
are designated . The ‘experiment mean values’ and the ‘experiment peak
values’ are designated and respectively. These are also comparable
to specified values for and respectively. The s are usually
the measured values with the highest magnitude for (westerly) and
(south-westerly) degrees wind directions considered for this discussion. The
summary of the findings are discussed in Section 5.5.

5.1 The Sports Hall


In Figure 5-2, it can be seen that the results of the experiment follow theoretical
expectations. However, the at stagnation on the windward face is less than
the theoretical value of . This is so because the theoretical definition of the
pressure coefficient is based on Bernoulli equation which assumes an inviscid
flow. The real flow is viscous and some viscous dissipations of the total
pressure abound. The positive pressures can be seen to diminish away and
downwards from the stagnation region. This is caused by the acceleration of the
flow away from this region and the decrease in flow velocity with decreasing
height in the boundary layer respectively.

89
Figure 5-2A westerly (270 degrees) wind flow over the Sports Hall model

The roof and the side walls can be seen to experience negative pressures with
localised high suctions at the eaves and the corner where the flow separates. At
the upwind eave of the roof, a cylindrical vortex exists [27] giving rise to the
moderately negative pressures of . This is very similar to what
happens at the upwind corner of the south face (the side wall). Just downwind
of these regions of high suctions, the flow experiences some pressure recovery
towards zero but the sense of the pressure is still negative. This is caused by
the action of the cylindrical vortices re-energising the flow downwind of the
separation and causing a tendency towards re-attachment [27].

This tendency is aided by the long dimension of the building in the along-wind
direction. However, the re-attachment is not fully stabilised and so the
pressures on these regions of the roof and side walls are characterised by
sporadic changes in sense.

90
Figure 5-3 A SW (225 degrees) wind flow over the Sports Hall model

For a quartering ( ) wind on the building illustrated in Figure 5-3, the trails of
the corner vortices on the roof characterised by high suctions are noticeable.
The suctions can be seen to be highest along the inclined axis of the vortices
and not exactly at their origin. Cook [27] reports of some debate on how close to
a roof corner a representative pressure can be obtained. Lin’s and Surry’s [52]
results support a side of this debate predicting unlimited negative values of the
pressure towards the roof corner where they observed the highest suction on
their models at radial distance to the corner (where is the height of
the building). The result of the current study (where the suction at is
higher than at radial distance from the corner) supports the other side of
the debate where the theory of vortices [27] sides. The theory suggests that the
smaller the radius of the vortices get, the more viscosity will impose limits on the
negative pressures. It is noteworthy that these localised high suctions are only
significant for cladding design.

The stagnation region moved to the junction of the west and the south faces of
the building. The variation of the pressure with height on the west face appears

91
not to follow theoretical expectations. This is different than the south face on
which the pressure can be seen to decrease expectedly with height. This
anomaly is suggestive that the plot may have been ‘flipped’ mistakenly when
the figure was created. Time resource was limited to find this out. However, it
was conceived that it would have no negative bearing on the subsequent
comparisons made.

Figure 5-4 Comparison between EN and EMV on the roof of the Sports Hall model
for W and SW wind

Figure 5-4, Figure 5-5 and Figure 5-6 present an overview of the comparison
between and for the roof, south face and west face of the building
respectively for a westerly and south-westerly wind flow.

The Eurocode can be seen to generally over-predict the suctions over the roof
by about . This is so because the UK NA uses the most unfavourable
values of localised suctions in each of the defined zones to specify the
corresponding overall loading coefficients. Doing this ignores the fact that the
suctions vary in strength across the area of the zones. Despite this, the striking
agreement between the and in zone on the roof for the normal wind

92
to the eave (westerly wind) is noticeable. This is consistent with Aylanc’s and
Kasperski’s findings [13].

Again in Figure 5-5, the can be seen to largely over-predict the suctions in
each of the three zones of the side wall for the wind normal to the west face. A
particular problem arises in Figure 5-5 and Figure 5-6 for the case of the
quartering wind which is the predominant wind direction in Cranfield.

Figure 5-5 Comparison between EN and EMV on the South face of the Sports Hall
model for W and SW wind

It is important to recall here that the s are specified to be used for a range of
angles around the relevant orthogonal direction. This, essentially,
defines a cone in the range of which the variations of the surface pressures
have been ignored. However, it is evident from Figure 5-7 which presents the
pressures measured by pressure taps and on the west face of the building

93
model that the surface pressures are indeed functions of the wind direction. In
addition, the spatial variation of the pressures on the face is obvious. While
pressures measured by tap can be observed to have a polynomial
relationship with the wind direction, it is difficult to define a specific relationship
of pressures from tap with the wind direction. The spatial correlation of the
pressures on the face are noticed only for the and wind directions
when the face is leeward and windward respectively.

So, the questions arise: How is the case of the quartering wind to be treated
effectively with the Eurocode for the building walls? Which of the west face and
the south face is to be taken as the windward face or the side face with the
present provisions? What will be the consequences of whichever choice is
made?

Figure 5-6 Comparison between EN and EMV on the West face of the Sports Hall
model for W and SW wind

These questions highlight Boggs’ [22] concerns about the building design codes
(refer back to Section 1.2). If the south face was taken to be a windward face
(as Zone in the Eurocode) in Figure 5-5, the would be over-predicting the
pressure on it by about . On the other hand, it would be dangerously unsafe
to treat either of the faces as a side face since the predictions of the Eurocode
would then be in the wrong sense.

94
It is suspected that in this case, a second look at the wind rose (in Section
2.1.5) might render the west face the design wind ward face since the next
predominant wind direction after is the in Cranfield. For a normal wind on
the west face, the over-predicts the pressure by .

0.8

0.6

0.4

0.2

0 Cp tap1
Cp

0 45 90 135 180 225 270 315 360


-0.2 Cp tap6
Poly. (Cp tap1)
-0.4

-0.6

-0.8

-1
Wind direction (degrees)

Figure 5-7 Variation of Cp with wind direction for pressure taps 1 and 6 on the
West face of the SH model

It is worth to mention, again, the dynamic nature of the wind and that strong
winds can come from any direction. As a consequence of this, it may be viewed
a one-sided discussion to have based the current review on the analysis of just
wind directions. This point of view is considered to be well-founded especially
since the Eurocode states that the coefficients specified are “the most
unfavourable values obtained in a range of wind directions either side
of the relevant orthogonal direction” [15 p33] . Similarly, it may be a point of
argument that making the tubes which connected the tappings to the
transducers as long as was necessary for the set-up used (see Section 4.3)
could have affected the frequency response of the measurement system and
attenuated the measured peaks. However, these appear to be refuted with the

95
observation that some excellent agreements were found between the and
on the wind ward face (zone ) and some zones of the Roof ( , and )
as is evident in Table 5-1. In fact, both signs of the coefficient and the exact
magnitude were observed in zone . The for zone E was obtained for a
and wind direction.

Table 5-1 Comparison between EN and EPV for the Sports Hall

Zone EN EPV
A -1.2 -0.76
B -0.8 -0.38
C -0.5 -0.25
D 0.7 0.67
E -0.3 -1.3
F -2 -1.9
G -1.4 -0.97
H -0.7 -0.8
0.2 0.2
I -0.2 -0.2

Pintar’s [53] result of the CFD simulation of a wind flow over the Sports Hall
is presented in Figure 5-8. It should be noted from Figure 2-24 that the
simulation domain also included other buildings west of the Sports Hall which
have not been shown here. This explains the wide range of the colour bar. For
the wind, it was expected that the Sports Hall would not be in the wake of
these buildings and the comparisons of the surface pressures predicted with the
results of the experiment have been made based on this wind direction in
Figure 5-9. It was necessary to compare the surface pressures and not the
surface pressure coefficients because as it can be seen in Figure 5-8, the CFD
prediction of the pressure coefficients is disturbing. Nevertheless, the general
features of the pressure distribution (such as the high suctions along the axis of
the corner vortices highlighted in black ink in the figure) already discussed have
been captured.

96
Figure 5-8 CFD prediction of surface pressure coefficient on the Sports Hall

It is interesting to observe the laudable agreement between the experiment


results and CFD in the region of the roof bounded by the black ellipses in Figure
5-9a and Figure 5-9b respectively. This is credited to the fact that the flow over
this region is less turbulent than towards and around the upwind roof corner
where the agreement can be seen to be less than desirable. The agreeable
result of this comparison between the surface pressures is suggestive that the
CFD predicted pressure coefficients in Figure 5-8 might have been based on a
wrong reference wind velocity. A quantitative comparison between the and
the CFD results would have been desirable but for the limited time to correct
these coefficients.

97
Figure 5-9 Comparison between a. experiment surface pressures and b. CFD
predictions of the surface pressures on the roof of the Sports Hall

5.2 Building 83, Hangar Building


Reference is made to the discussion in Section 3.1 on the expected nature of
flow on the building. It should be noted that due to the inclined orientation of the
building to the north (see Figure 4-7), both the as well as the wind flow
obliquely on the building as shown in Figure 5-10. It is important to keep in mind
the limitations of the contour plots highlighted at the beginning of this chapter.
The complex flow over the building is characterised, for instance, by interactions
of positive pressures on the walls of the inset faces with the suctions on the
adjoining lower roof such as in the regions bounded by the white boxes in
Figure 5-11 and Figure 5-12 annotated . The extent of these regions may
cover the whole spread of the lower roof or some part of it depending on the
scaling length of the inset face as discussed earlier.

98
Figure 5-10 Pictorial representation of a Westerly and a South-westerly wind flow
over the Hangar building

Figure 5-11 A South-Westerly flow over the HB model

99
This mean that suctions may or may not be present at the eaves of the lower
roof. Also, the white boxes annotated signify regions of interaction of cross
flows where the more negative pressures can be observed. Most of the
tappings on the roof of this building were located in these regions of interaction.
This is good for the current review because it was possible to check if the
provisions of the Eurocode could be adapted to these regions. However,
extrapolating measurements from some of these tappings (in the region
bounded by the black boxes in Figure 5-11 for instance) to the corners next to
them as was necessary to generate the figures may or may not be
representative of the true complex flow.

Figure 5-12 A Westerly flow over the HB model

100
Also, as a consequence of the dis-uniformity in the distribution of the pressure
tappings over the roof surface, defining s for the identified zones in Figure
5-13 was considered to be erroneous. As such, the review of the were
based on the s in the zones. This is summarised in Table 5-2

Figure 5-13 The HB model roof divided into Zones using the Eurocode

It can be seen from this table that adaptation of the to predict the wind
loading on the roof of this building results generally into very unsafe conditions.
Zone , from the table where the code under-predicts the suction by
about , is very noteworthy in this regard. The thought must be entertained
whether tap should have been interpreted to be located in this zone and not
zone or in which case, the solution would not be of much concern. In this

101
regard, it is interesting to observe that tap which was interpreted to be in the
same zone ( ) returned .

Table 5-2 Comparison between EN and EPV on the Hangar building

EPV
Zone EN Cp Tap no Wind direction
A -1.2 -0.74 45 270
B -0.8 -1.01 67 270
C -0.5 -0.92 69 225
0.36 96 225
D 0.7 -0.54 77 225
E -0.3 -0.56 87 225
F -2 -2.1 38 270
G -1.4 -1.31 32 270
H -0.7 -1.55 90 225
0.2 0.04 36 225
I -0.2 -1.6 82 225

The zones were arrived at using the guidelines provided by Cook in his text
[27]. They are based on the assumption that the wind loading on an irregular
plan can be adapted from the data on a regular plan. Cook notes that “the
division [of an irregular plan] is made into logical parts according to the flow
characteristics and the particular interactions…must be separately accounted
for” [27 p272]. Hence, the view is held that the ‘trick’ to getting a good
representation of the current complex flow with adaptation from the codes (if it is
ever possible) can only come with experience. Nonetheless, in Figure 5-12, the
agreement is observed in zones and where the local flow effectively sees a
simple cuboidal building.

102
5.3 The Trafficmaster Building
Figure 5-14 is the key figure to the nomenclature adopted for this building.

Figure 5-14Key figure for the dimensions and pressure tapping locations on the
Trafficmaster (TM) building model

Figure 5-16 presents a description of the nature of flow around the building for a
westerly wind and south-westerly wind. In the westerly wind flow on the
building, the west wing of the building behaves like a simple cuboidal building.
The flow sees the cross-wind dimension of Face as the scaling length of the
wing. It stagnates on Face . The under-predicts the localised pressure on
this face but its agreement with the is commendable ( ;
; and ).

The flow accelerates away from the stagnation region until it separates at the
sharp edges of Face . At the upwind edges of Faces and , negative
pressures are recorded due to the flow separation. On Face , the flow
separation with the consequent suctions at the upwind corner is affected by the
interaction of the positive pressures from Face which the flow ‘sees’ as a
setback wind ward face.

103
Figure 5-15 Flow around the TM model for a a. Westerly wind and b. a South-
Westerly wind

Figure 5-16 Countour of Cp for a Westerly wind flow over the TM model (Faces1,6
and 7)

This region of interaction can be seen to be wide ranging in Figure 5-16 where
the Eurocode expected suctions in zones and have been damped out into
positive pressures and the is significantly less (in magnitude) than the
in zone . The situation is different for Face which is a long face relative to
the scaling length. The flow tries to re-attach but in an unstable manner towards
Face as it deflects around the external corner in to the wake.

104
Figure 5-17 Countour of Cp for a Westerly and South-Westerly wind flow over the
TM model (Faces 4and 5)

The results presented in Figure 5-17 and Table 5-3 suggest that if the
adaptation of the Eurocode to Face is to be valid, there will be no need for
zone on this face (or probably, its extent will be reduced). Otherwise, the
over-predicts the suction in zones and .

Table 5-3 Comparison between EN and EPV for Face4 of the TM model in a
westerly wind and south-westerly flow

Zone EN EPV
A -1.2 -0.83
B -0.8 -0.83
C -0.5 -0.27
E -0.3 -1.14

In a south-westerly flow on the building illustrated in Figure 5-15b, the flow


enters the re-entrant corner and gets trapped into a region of stagnated flow
defined by the wedge in the figure before it rises over the building into the wake
region [27]. The can be seen in Figure 5-17 and Table 5-3 to under-predict
the suctions (both and ) by very substantial amounts on Face which
falls in the wake.

105
The outward bulge of the circular Face extends the effect of the positive
pressures from the stagnation region beyond the extent of the wedge usually
defined by the scaling length of the building. This can be observed in Figure
5-18 and Figure 5-19 where the pressures as well as the suctions on the
internal and (towards) the external corners of Faces (taps and
respectively) and (taps and respectively) follow the variation of pressure
with wind direction on (circular) Face (tap ). (Refer to Appendix B for clarity
on the locations of these tappings and tappings and in Figure 5-20. It
should be noted that the in these graphs were based on the calculated wind
pressure at the various heights of the tappings.)

0.8

0.6
Cp tap20
0.4
Cp tap89
Cp tap83
Cp

0.2
Poly. (Cp tap20)
0
Poly. (Cp tap89)
0 45 90 135 180 225 270 315 360
-0.2 Poly. (Cp tap83)

-0.4

-0.6
Wind direction (degrees)

Figure 5-18 Graph of Cp variation with wind direction for pressure tappings 20,
83 and 89 on the TM model

The variation of with wind direction in these figures can be observed to


follow some coherent ‘wave’ forms. Time resource was limited to effectively
study, identify and/or parameterize these wave forms. However, the graphs are
suggestive that the pattern of flow on the curved surfaces of the building
dominates the whole flow. This inference was taken because the ‘wave’ forms
in Figure 5-18 and Figure 5-19 do not share a similar form with the trend in

106
Figure 5-7 which represents the variation of pressure with wind direction on a
wall of a cuboidal building. A consequence of this is that the building is
sensitive.

It can also be deduced from these figures that the flow around the building is
very complex. It does not appear to be symmetrical except for a wind.
Therefore, the loading data on a wing of the building may not be appropriate to
be used on the other.

1.4

1.2

0.8 Cp tap14
0.6 Cp tap95
Cp tap83
Cp

0.4

0.2 Poly. (Cp tap14)


Poly. (Cp tap95)
0
0 45 90 135 180 225 270 315 360 Poly. (Cp tap83)
-0.2

-0.4

-0.6
Wind direction(degrees)

Figure 5-19 Graph of Cp variation with wind direction for pressure tappings 14,
83 and 95 on the TM model

107
0.2

0
0 45 90 135 180 225 270 315 360
-0.2
Cp tap35
-0.4
Cp tap47
Cp

-0.6 Poly. (Cp tap35)

-0.8 Poly. (Cp tap47)

-1

-1.2
Wind direction (degrees)

Figure 5-20 Graph of Cp variation with wind direction for pressure tappings 35
and 47 on the roof of the TM model

Figure 5-21and Table 5-4 summarise the assessment of the on the roof of
the west wing of the building. The over-predicts both and in zone
; agrees with and under-predicts the in zone ; over-predicts
and under-predicts in zone ; and under-predicts the suctions (both
and ) in zone . These disagreements are likely due to the fact that the
adaptation of to the building does not take into account the -dimensional
nature of the complex flow over the building which may stem from building-
induced turbulence and/or oncoming wind turbulence.

108
Figure 5-21 Countour of Cp for a South-Westerly wind flow over the TM model
(Faces 3 – the roof)

The Eurocode provides no data on flat roofs of circular cylinders. Hence, no


comparison could be made on this part of the roof of the building. However it is
suspected that interactions between the individual building shapes (cuboid and
cylinder) will make the flow on that part of the roof to be different from that on
the roof of a lone cylinder.

109
Table 5-4Comparison between EN and EPV on the roof of the west wing of the
TM model for a south-westerly wind

Zone EN EPV
F -2 -0.94
G -1.4 -1.9
H -0.7 -0.89
0.2 0.12
I -0.2 -0.47

Also, it may have been observed that analysis have not been made on Faces
and . This is because they are curved surfaces and it was not very straight-
forward as it was with the cuboidal sections to get the provisions of the codes
for the wind loading on them. That is to say it would have required more time
which is limited for the completion of this work.

110
5.4 Effects of Tributary Area variation

Figure 5-22 Tributary area variation on the roof of the SH model for 8 different
configurations

Figure 5-22 and Figure 5-24 present the configurations used to assess the
effect of tributary area variation on a flat and a curved surface respectively. The

111
roof of the Sports Hall model and the curved surface Face of the Trafficmaster
model were used respectively. The assessment was based on these surfaces
given the relatively uniform distribution of the tappings on them. The force
coefficients compared for the different configurations in the bar charts in Figure
5-23 and Figure 5-25 were based on the dynamic pressure measured at
in the wind tunnel and a reference area (where is the height of the
building).

Where it is possible, in the configurations, a tapping closest to the centre of a


defined tributary area was used to estimate the wind loading on that area.
Where not, such as in some tributary areas in Config , and in Figure 5-22,
the tappings used have been identified with the blue circles.

-3.5

-3.0

-2.5

-2.0 180
Cfz

225
-1.5 270

-1.0

-0.5

1 2 3 4 5 6 7 8
0.0 Configuration

Figure 5-23 Effect of tributary area variation on the wind loading on the roof of
the Sports Hall model

This means there are other variations of these configurations which could have
been assessed as well. However, the variations used were conceived to be
adequate enough for the purpose of the discussion.

112
It should be noted that the entire wind loading estimates in this sub-section
have the negative sense. For convenience, words like ‘higher’ and ‘lower’ or
‘less’ are used in the comparison. These are only with regards to the magnitude
of the quantities and not their actual sense.

Figure 5-24 Tributary Area distribution over Face 5 of TM building for


Configurations 1-4

Table 5-5a presents the number of tappings in each of the configurations


assessed on the roof of the Sports Hall model. Config contains tappings. It
is the crudest of the configurations. Large portions of the surface like the centre
region have been left untapped. Figure 5-4, however, reveals that this region
spans an area where the sense of the pressure changed downwind for a
wind due to the long span of the roof in this direction resulting in an unstable
flow re-attachment. In a wind, the suction at the roof corner becomes
higher. This is evident in the upshot in compared to the wind. For a

113
wind, the dimension of the roof in the along-wind direction is short. There
is no possibility of flow re-attachment downwind the upwind eave. Even though
the suction at the roof corners becomes less compared to the wind, is
higher because of the relatively higher suctions which cover the whole centre
region. This trend can be observed for all the configurations except in Config
where for wind is higher than for wind. This is due to the fact that
the suctions along the axis of the corner vortices which are the highest (for all
wind directions) have been factored in for the estimation. Refer back to the
tributary areas bounded by the red boxes in Figure 5-22 to observe this.

Factoring in the individual pressures (from the tappings) in the centre region in
Config reduces when compared to Config because the changes in the
sense of the pressures over this region are taken account of. The reduction in
load is particularly observable for wind when the roof experiences positive
pressures in the downwind end of this region.

For normal winds ( and ), the highest suctions are experienced in the
middle of the eave and verge [27] (the regions bounded by the yellow boxes in
Figure 5-22 for Config and ). Accordingly, using more tappings in these
regions of the roof increases the load in Config when compared to Config for
a wind. This is not observed for the other wind directions because
increasing the number of tappings all over the roof gives the opportunity to
capture regions where less suctions are experienced for these wind directions
and so the load reduces.

Again, increases in Config when compared to Config for a wind


because positive pressures downwind in the centre region have not been
accounted for. This is rectified in Config where more tappings are used in this
region compared to Config .

Configs with tappings do not present much of variations in the obtained


when compared to to Config with tappings. However, when a comparison
is made between Config and Config , it can be observed that the load
reduces on the roof with increasing number of tappings. This trend is expected

114
to extend beyond Config . Additionally, a general observation is that the load is
a function of wind direction on a flat roof of a cuboidal building.

-1.4

-1.2

-1.0

-0.8
180
Cf

-0.6 225
270
-0.4

-0.2

1 2 3 4
0.0

Configuration

Figure 5-25 Effect of tributary area variation on the wind loading on Face5 of the
TM model

Table 5-5Number of tappings in each of the configurations used for tributary area
variation assessment

115
Accordingly, it can be deduced from the discussion so far that the knowledge of
the expected flow features and their expected variations with wind direction on a
building geometry is key to getting reliable wind loading estimates from wind
tunnel studies when there is a limitation on the number of tappings to be used.

As it can be seen in Figure 5-25, all the configurations assessed on the curved
Face of the Trafficmaster building presented about the same magnitude of
load. This is suggestive that the load on a curved wall is fairly constant with
tributary area. Expectedly, the load varies with wind direction.

5.5 Summary of Observations


At the completion of the analysis of the results, the following observations were
made:

1. Surface pressure coefficients based on the wind pressure measured at


the equivalence of in the wind tunnel are more consistent with
theoretical and practical expectations than those based on calculated
wind speeds at different heights on the buildings tested.
2. The Eurocode, generally, over-predicts the in designated zones of
the roof and side walls of the Sports Hall. However, a striking agreement
was noticed in zone of the roof between the ( ) and the result
of the experiment for a normal wind to the eave. In addition, in zone
(on a windward wall) as well as in zones , and of the roof were
accurately predicted by the . Where disagreements were noticed, they
were attributed to the fact that the UK NA to the Eurocode specifies the
same values of localized pressure coefficients for and .
3. The surface on a building is a function of the wind direction. The
specified in Eurocode are to be used for of a relevant
orthogonal direction. Doing this neglects the actual variation of pressure
on the building with wind direction. A particular problem arose for the
quartering wind on the walls. In such a case, the Eurocode resulted in
both unsafe and uneconomic solutions on the walls.

116
4. The available CFD predictions of the surface pressures on the roof of the
Sports Hall considered agreed with experiment results in the region
designated as zone in the Eurocode. This was attributed to the fact that
the turbulence in the flow over that region of the roof is lower than other
parts where disagreement was noticed.
5. The adaptation of the Eurocode provisions on a simple cuboidal building
to predict the on the roof of the Hangar building results, generally, into
very unsafe conditions. The flow over the building is very complex having
interactions of positive pressures from its windward inset faces with
suctions on their adjoining lower roofs. In addition, the interaction of
cross-flows close to the junction of its inset faces flushed with the
building geometry on all sides, and a lower roof results into high suctions
on the lower roof. However, some agreement between the and
experiment result was noticed at some roof corner of the building (that
bear resemblance to roof corners of simple cuboidal buildings) where
complex cross-flow interactions were absent on the building (zone and
).
6. In a similar vein, the adaptation of the Eurocode to predict the loading on
the Trafficmaster building results, generally, into very unsafe conditions.
Where agreement was observed between the and (or where the
over-predicts the ), unsafe disagreement existed between the
and . This was also attributed to the inability of the Eurocode to
predict the complex flow around the building. Cook [27] elucidates that
peak loads are effected by turbulence from the approaching wind and
building-generated turbulence. More so, “…the pressure distribution on
the faces either side of an internal corner [of a re-entrant
corner]…depends on the whole plan shape…” [27 p288].
7. The variation of with wind direction on the internal side walls of both
wings of the Trafficmaster building follow some ‘wave’ form similar to that
followed by the curved surface present at the internal junction of the
wings. The variation of with wind direction on the roof of the building
also follows a wavy trend.

117
8. On the roof of the Sports Hall model, the overall load reduces with
increasing number of tappings i.e. decreasing size of tributary areas of
the tappings. However, on a curved wall (Face of the Trafficmaster
building), the overall load remains fairly constant with tributary area size.
In addition, the magnitude of the load depends on wind direction.

118
6 Conclusion
“For the stronger we our houses do build,

The less chance we have of being killed”

The Tay Bridge Disaster, William McGonnegal [18 p867]

6.1 Project Summary


Building codes such as the Eurocode have usually been used as a cheaper
alternative to wind tunnel studies in the consideration of wind loading on a
structure. It is often the case that very tall buildings and large structures have
enough economic justification for expensive wind tunnel studies in their design
stage. Such wind tunnel studies, as per state-of-the-art, feature simultaneous
scanning of and acquisition of loading data from hundreds of pressure tappings
with subsequent high-speed computer data processing and analysis. This is not
the case for low-rise buildings which do not find their way into the wind tunnel
except in the case where they are unusual edifices. Low-rise buildings,
however, are the most damaged in wind storms. In addition, in the present
times, their shapes are increasingly losing touch with the traditional and generic
forms dealt with in the Eurocode. Therefore, the question was: How well does
the Eurocode, which was put together with information from wind tunnel studies
carried out in the and using currently outdated data acquisition
techniques, deal with present building shapes?

This defined the major aim of the study and its focus on low-rise buildings. In
addition to this, it was of interest to know what the limit might be on the required
number of pressure tapping to obtain adequate and reliable wind loading
information from wind tunnel studies. However, this was already limited by the
number of pressure tapping available for the study which was a hundred.

To achieve these, it was necessary to explore the nature of the atmospheric


wind; to understand the aerodynamics of buildings; to understand the concept
of wind loading; to assimilate the wind loading provisions of the Eurocode; to

119
dissect the principles of wind tunnel experiments; to explore computational
means which may be available on the same matter; to peruse past relevant
works available in the literature; to identify suitable cases on which to base the
study; to conduct the wind tunnel experiments undertaken in line with
recognized guidelines; and to analyze and discuss the results. These were in
fulfilment of the set objectives1-7 of the study.

The Sports Hall which is a simple cuboidal building with a flat roof; Building 83,
Hangar Building which features inset faces in its plan and elevation; and the
Trafficmaster Building which is a re-entrant corner with a V-plan having curved
surfaces at the internal and external junctions of its wings were selected on the
Cranfield Campus as case studies. The conclusions of the study are presented
in the next sub-section.

6.2 Concluding Remarks


In fulfilment of objective 8, the following conclusions were drawn from the
observations made:

A. The Eurocode in use in the UK over-predicts the load on the structure of


a simple cuboidal building with a flat roof. However, it is adequate for the
design of cladding elements on such a building. This notwithstanding, it
is in need of some minimal updates in that regard. In any case, adapting
its provisions to buildings of irregular plans particularly those featuring
inset faces and re-entrant corners, generally, results into very unsafe
conditions. This is because the complex flows around these irregular
buildings depend on their whole planforms.
B. The wind loading on a building is a function of the wind direction. The
definition of uniform loading coefficients to be used for of a
relevant orthogonal wind direction as adopted in the Eurocode ignores
this fact. It was demonstrated that when the predominant wind direction
is skewed on a simple cuboidal building, the Eurocode provisions are not
adequate for the safe design of its walls. Therefore, the Eurocode will not
be adequate for the design of buildings with irregular plans in skewed
approaching winds.

120
C. It can be expected that the quality of the wind loading estimate from a
wind tunnel study is better with more pressure tappings. However, where
there is a limitation on this, it was demonstrated that the knowledge of
the expected flow features on the building geometry in question in the
wind directions of interest can inform the distribution of these tappings
which will give very reliable loading estimates.
D. In the course of the literature review, it was gathered that concerns have
been raised about the real possibility of sensitivity of the flow pattern
around sharp edged buildings to by well known researchers and
practitioners in the wind engineering community. This was not
independently assessed since the speed range possible in the CU ABL
wind tunnel for the setup used does not give enough justification for such
tests. However, it is concluded that a building with sharp edges and
curved surfaces will be sensitive. This is from the observation that the
pressure variations on the internal side walls of the re-entrant corner
studied follow the same trend as that on the curved surface at the
internal junction of the wings of the corner.
E. Dedicated CFD predictions of surface pressures on a building with a
simple plan can provide predictions which have a reasonable agreement
with results of experiments in regions of the building where the wind flow
is less turbulent.

The reliability of the load factors specified in the Eurocode could not be
reviewed due to the fact that these are only applied after the combination of the
wind loads and other loads on the structure. It should be pointed out once more
that the wind loading is the predominant horizontal loading on a building.
Hence, it can be assumed that it would be of much significance in the
evaluation of the overall loads on a structure. Therefore, the view is held, in
consistence with Davenport’s ( cited in [48]; see Section 2.3), that the
reliability of wind loading prediction can be improved with the use of wind tunnel
aerodynamic data over the loading information in building codes. It is expressed
further, also in agreement with Davenport’s observation about building codes
[31], that the use of localised pressure or suction peaks to specify loads for

121
overall structural design of simple cuboidal buildings in the UK NA may be
overly conservative especially since load factors are also specified. It is
absolutely important not to compromise safety but a balance must also be
struck between safety and economy especially at such times as this of global
financial crises [44].

6.3 Recommendations
The data acquired from this study is extensive. Time was only available to
analyse some portion of it. Similar analysis to those carried out in this study can
be made on this set of data for the rest of the wind directions.

More so, time was limited to review the Eurocode provision for the walls of
circular cylinders. This can be done with the loading information obtained on the
Trafficmaster building assuming the presence of the wings of the building will
not compromise the effectiveness of doing it. In any case, it may serve to
highlight the effectiveness or not of the Eurocode provisions to predict the wind
loading on circular cylinders incorporated in this kind of building geometries.

In addition and in fulfilment of objective 9, the variations of with wind direction


on the internal side walls and roof of both wings of the Trafficmaster building are
identified for further research. These follow some trends. These trends can be
studied better, identified and parameterized. Doing this may be of benefit to the
knowledge of the complex flows around similar building shapes to this and
subsequent codification of such knowledge.

122
REFERENCES
[1] , "The Gherkin" (Swiss Re Tower) - Google Maps , available at:
http://maps.google.co.uk/maps?q=domicilirity&oe=utf-
8&aq=t&rls=org.mozilla:en-GB:official&client=firefox-a&um=1&ie=UTF-
8&hl=en&sa=N&tab=wl (accessed 7/16/2012).

[2] , The Abuja Millenium TowerThe Abuja Millenium Tower | D-Review.net ,


available at: http://www.d-review.net/?p=396 (accessed 6/27/2012).

[3] , Amplified Pressure Sensors from Sensortechnics , available at:


http://www.sensortechnics.com/hcla (accessed 7/26/2012).

[4] , ASCE 7-10 , available at:


http://www.asce.org/PPLContent.aspx?id=2147486026 (accessed 5/9/2012).

[5] , Building the Gherkin , available at: http://www.buildingthegherkin.com/


(accessed 8/15/2012).

[6] , Industrial chimney. Sariñena, Huesca province. Aragon. Spain.


[WE047202] > Stock Photos | Royalty Free | Royalty Free Photos >
Visualphotos.com , available at:
http://www.visualphotos.com/image/2x3120165/industrial_chimney_sariena_huesc
a_province_aragon (accessed 6/27/2012).

[7] Cranfield University Campus( 2012), , Google Earth.

[8] Anderson, J. D. (2007), Fundamentals of Aerodynamics, 4th ed, McGraw-


Hill Higher Education, Boston.

[9] Anderson, J. D. (2011), Introduction to Flight, 7th ed, McGraw-Hill


College.

[10] Apperley, L., Surry, D., Stathopoulos, T. and Davenport, A. G. (1979),


"Comparative measurements of wind pressure on a model of the full-scale
experimental house at Aylesbury, England", Journal of Wind Engineering
and Industrial Aerodynamics, vol. 4, no. 3–4, pp. 207-228.

[11] Association of British Insurers (2003), The Vulnerability of UK Property to


Windstorm Damage, , Association of British Insurers.

[12] Aurelien, G. (2010), A 'LIVING LABORATORY' FOR URBAN


MICRO ENERGY RECOVERY SYSTEMS (unpublished Master of Science
thesis), Cranfield University, MK43 0AL, Bedfordshire, United Kingdom,
Cranfield, United Kingdom.

123
[13] Aylanc, N. and Kasperski, M. (2009), "Design wind loads for cladding
elements", EACWE 5, 19-23 July, pp. 1.

[14] Aynsley, R. M. (1999), "Shape and Flow: The Essence of Architectural


Aerodynamics", Architectural Science Review, vol. 42, no. 2, pp. 69-74.

[15] B/525/1, ( 2005), BS EN 1991-1-4:2005+A1:2010


Eurocode 1. Actions on Structures. General actions. Wind actions, ICS
91.010.30 (Technical aspects) ed., European Union.

[16] B/525/1, ( 2008), NA to BS EN 1991-1-4:2005+A1:2010


UK National Annex to Eurocode 1. Actions on Structures. General Actions.
Wind Actions, ICS 91.010.30 (Technical aspects) ed., European Union.

[17] Baker, F. W., Korista, D. S. and NOVAK, C. L. (2007), "BURJ DUBAI:


ENGINEERING THE WORLD’S
TALLEST BUILDING", THE STRUCTURAL DESIGN OF TALL AND
SPECIAL BUILDINGS, vol. 16, pp. 361-365.

[18] Baker, C. J. (2007), "Wind engineering—Past, present and future",


Journal of Wind Engineering and Industrial Aerodynamics, vol. 95, no. 9–
11, pp. 843-870.

[19] Baskaran, A. and Stathopoulos, T. (1988), "Roof corner wind loads and
parapet configurations", Journal of Wind Engineering and Industrial
Aerodynamics, vol. 29, no. 1–3, pp. 79-88.

[20] Beal, A. (2010), "Eurocodes in Britain: the questions that still need
answering", Proceedings of ICE, February, pp. 27-34.

[21] Boggs, D. and Dragovich, J. , The Nature of Wind Loads and


Dynamic Response, available at:
http://www.cppwind.com/support/papers/papers/structural/240-2.pdf (accessed
5/9/2012).

[22] Boggs, D. and Lepage, A. , Wind Tunnel Methods, available at:


http://www.cppwind.com/support/papers/papers/structural/240-6.pdf (accessed
5/9/2012).

[23] Boggs, D. W. (1992), "Validation of the aerodynamic model method",


Journal of Wind Engineering and Industrial Aerodynamics, vol. 42, no. 1–3,
pp. 1011-1022.

[24] Cochran, L. (2006), "State of the Art Review of Wind Tunnels and
Physical Modelling to Obtain Structural Loads and Cladding Pressures",
Architectural Science Review, vol. 49, no. 1, pp. 7-16.

124
[25] Cochran, L. and Derickson, R. (2005), "Low-rise Buildings and
Architectural Aerodynamics", Architectural Science Review, vol. 48, no. 3,
pp. 265-276.

[26] Cochran, L. and Derickson, R. (2011), "A physical modeler's view of


Computational Wind Engineering", Journal of Wind Engineering and
Industrial Aerodynamics, vol. 99, no. 4, pp. 139-153.

[27] Cook, N. J. (1990), The Designer's guide to wind loading of building


structures
Part 2 Static structures, [Garston, Watford] : Building Research
Establishment, Dept. of the Environment, Butterworths.

[28] Cook, N. J. (1978), "Wind-tunnel simulation of the adiabatic atmospheric


boundary layer by roughness, barrier and mixing-device methods", Journal
of Wind Engineering and Industrial Aerodynamics, vol. 3, no. 2–3, pp. 157-
176.

[29] Cook, N. J. (1988), "Discussion of session 5: Low-rise structures - local


wind loads", Journal of Wind Engineering and Industrial Aerodynamics, vol.
29, no. 1–3, pp. 119-134.

[30] Dalley, S. and Richardson, G. (1992), "Reference static pressure


measurements in wind tunnels", Journal of Wind Engineering and Industrial
Aerodynamics, vol. 42, no. 1–3, pp. 909-920.

[31] Davenport, A. G. (1983), "The relationship of reliability to wind loading",


Journal of Wind Engineering and Industrial Aerodynamics, vol. 13, no. 1–3,
pp. 3-27.

[32] Eaton, K. J. and Mayne, J. R. (1975), "The measurement of wind


pressures on two-storey houses at Aylesbury", Journal of Industrial
Aerodynamics, vol. 1, no. C, pp. 67-109.

[33] Franke, J., Stathopoulos, T., Baniotopoulos, C. C., Pfeiffer, F.,


Rammerstorfer, F. G., Salençon, J., Schrefler, B. and Serafini, P. (2007),
"Introduction to the Prediction of Wind Loads on Buildings by Computational
Wind Engineering (CWE)", in Wind Effects on Buildings and Design of
Wind-Sensitive Structures
CISM Courses and Lectures, Springer, Vienna, pp. 67-103.

[34] Geurts, C. , The Use of Wind tunnel experiments for wind loading on
structures, available at: http://www.ruhr-uni-
bochum.de/imperia/md/content/asib/ws08_09/boundary_layer_windtunnel.pdf
(accessed 7/5/2012).

[35] Geurts, C. (1998), "Validation of Wind loading codes by experiment ",


HERON, vol. 43, no. 0046-7316, pp. 85-99.

125
[36] Geurts, C., Bentum, C., Stathopoulos, T., Baniotopoulos, C. C., Pfeiffer,
F., Rammerstorfer, F. G., Salençon, J., Schrefler, B. and Serafini, P.
(2007), "Wind Loading on Buildings: Eurocode and Experimental
Approach", in Wind Effects on Buildings and Design of Wind-Sensitive
Structures
CISM Courses and Lectures, Springer, Vienna, pp. 31-65.

[37] Hansen, S. O. and Sørensen, E. G. (1986), "The Aylesbury experiment.


Comparison of model and full-scale tests", Journal of Wind Engineering and
Industrial Aerodynamics, vol. 22, no. 1, pp. 1-22.

[38] Holmes, J. D. (1980), "Comparative measurements of wind pressure on


a model of the full-scale experimental house at Aylesbury, England",
Journal of Wind Engineering and Industrial Aerodynamics, vol. 6, no. 1–2,
pp. 181-182.

[39] Holmes, J. D. (1988), "Distribution of peak wind loads on a low-rise


building", Journal of Wind Engineering and Industrial Aerodynamics, vol.
29, no. 1–3, pp. 59-67.

[40] Holmes, J. D. (2001), Wind Loading of Structures, First ed, Spon Press,
London; New York.

[41] Holmes, J. D. and Best, R. J. (1977), "Wind Tunnel Measurements of


Mean Pressures on House Models and Comparison with Full Scale Data",
Vol. 6th Australian Hydraulics and Fluid Mechanics Conference, 5-6
December,1997, Adelaide, Australia, pp. 26-30.

[42] Houghton, E. L. and Carruthers, N. B. (1976), Wind Forces on Buildings


and Structures: an introduction
, Edward Arnold, London.

[43] Hoxey, R. P., Robertson, A. P., Richardson, G. M. and Short, J. L.


(1997), "Correction of wind-tunnel pressure coefficients for Reynolds
number effect", Journal of Wind Engineering and Industrial Aerodynamics,
vol. 69–71, no. 0, pp. 547-555.

[44] International Monetary Fund (2012), Global Financial Stability Report


The Quest for Lasting Stability, ISSN 1729-701X, International Monetary
Fund, Publications Services, Washington, DC.

[45] Irwin, P. A. , WIND ENGINEERING RESEARCH NEEDS, available at:


http://www.iawe.org/about/TCWE.pdf (accessed 5/9/2012).

[46] Irwin, P. A. , Wind Engineering Research Needs,


Building Codes and Project Specific Studies, available at:
http://www.iawe.org/Proceedings/11ACWE/11ACWE-Irwin.pdf (accessed
5/9/2012).

126
[47] Irwin, P. A. (2008), "Bluff body aerodynamics in wind engineering",
Journal of Wind Engineering and Industrial Aerodynamics, vol. 96, no. 6–7,
pp. 701-712.

[48] Isyumov, N. (2012), "Alan G. Davenport's mark on wind engineering",


Journal of Wind Engineering and Industrial Aerodynamics, , no. 0.

[49] Kasperski, M. (2003), "Specification of the design wind load based on


wind tunnel experiments", Journal of Wind Engineering and Industrial
Aerodynamics, vol. 91, no. 4, pp. 527-541.

[50] Lawson, N.J., ( 2006), Advanced Measurement Techniques: Part 1 -


Errors
(Aerospace Dynamics MSc Lecture notes), School of Engineering,
Cranfield University.

[51] Lawson, T. (2001), Building Aerodynamics, Imperial College Press,


London.

[52] Lin, J. X. and 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, vol. 77–78, no. 0, pp. 185-196.

[53] MATEVŽ, P. (2012), RENEWABLE ENERGY FOR CRANFIELD


UNIVERSITY
(Group design project) (unpublished MSc Advanced Mechanical
Engineering thesis), Cranfield University, .

[54] Merlot, D. (2010), Effect of distributed roughness on surface drag


(unpublished Master of Science thesis), Cranfield University, MK43 0AL,
Bedfordshire, United Kingdom, Cranfield, United Kingdom.

[55] Rowan Williams Davies & Irwin Inc., Peter A. Irwin, Ph.D., P.Eng.

[56] Sill, B. L., Cook, N. J. and Fang, C. (1992), "The Aylesbury Comparative
Experiment: A final report", Journal of Wind Engineering and Industrial
Aerodynamics, vol. 43, no. 1–3, pp. 1553-1564.

[57] Solari, G. (2007), "The International Association for Wind Engineering


(IAWE): Progress and prospects", Journal of Wind Engineering and
Industrial Aerodynamics, vol. 95, no. 9–11, pp. 813-842.

[58] Stathopoulos, T., Stathopoulos, T., Baniotopoulos, C. C., Pfeiffer, F.,


Rammerstorfer, F. G., Salençon, J., Schrefler, B. and Serafini, P. (2007),
"Introduction to Wind Engineering, Wind Structure, Wind-Building
Interaction", in Wind Effects on Buildings and Design of Wind-Sensitive
Structures
CISM Courses and Lectures, Springer, Vienna, pp. 1-30.

127
[59] Surry, D. (1991), "Pressure measurements on the Texas tech building:
Wind tunnel measurements and comparisons with full scale", Journal of
Wind Engineering and Industrial Aerodynamics, vol. 38, no. 2–3, pp. 235-
247.

[60] Surry, D. (1992), "Wind tunnel simulation of the Texas tech building",
Journal of Wind Engineering and Industrial Aerodynamics, vol. 43, no. 1–3,
pp. 1613-1614.

[61] Vickery, P. J. and Surry, D. (1983), "The Aylesbury experiments revisited


— Further wind tunnel tests and comparisons", Journal of Wind
Engineering and Industrial Aerodynamics, vol. 11, no. 1–3, pp. 39-62.

[62] Vickery, P. J., Surry, D. and Davenport, A. G. (1986), "Aylesbury and


ACE: Some interesting findings", Journal of Wind Engineering and
Industrial Aerodynamics, vol. 23, no. 0, pp. 1-17.

128
Appendix A Principles of Wind tunnel Experiments in
detail
A.1 Atmospheric Boundary Layer Simulation in Wind tunnels
Comparative studies between full-scale and model-scale experiments have
highlighted the problems of simulating the turbulence in the ABL adequately.
This is usually the case because known methods of simulating the boundary
layer assume the real boundary layer to be fully developed over a
homogeneous fetch of surface roughness. However, it is often not the case in
reality even for synoptic windstorms over a long homogeneous fetch. It may be
the case that this occurs over open seas with uniform wave heights and coming
after large fetches of flat open country or desert terrain. Unfortunately, relatively
few buildings or structures are subject to these conditions. Possibly fewer urban
centres will have flat homogeneous upwind roughness of enough stretch to
result into a fully developed boundary layer flow over them. Nonetheless, these
methods and the assumption of an adiabatic boundary layer in strong winds
associated with large scale extra-tropical depressions with speeds of or
more have been the benchmark for wind tunnel flow assessment [40]; [28].

The UK NA to the Eurocode demands that:

“Tests for the determination of wind loads on static structures should not be
considered to have been properly conducted unless:

a) the natural wind has been modelled to account for:


 the variation of mean wind speed with height above ground
appropriate to the terrain of the site; and
 the intensity and scale of the turbulence appropriate to the
terrain of the site at a determined geometric scale;”[16 p2]

There are two ways of doing this: a wind tunnel simulated boundary layer flow
can either be allowed to grow ‘naturally’ over a long fetch of rough wind tunnel

129
floor representative of the modelled terrain or ‘artificially’ kick-started to produce
an immediate velocity gradient downstream of the fetch [40];[22];[10];[28].

A ‘naturally’ grown wind tunnel boundary layer flow requires large wind tunnel
dimensions. Ideally, it will keep growing until it catches up with the boundary
layer on the roof of the tunnel but in most practical cases, this equilibrium
condition is not achieved. As a result, tests of tall structures are carried out in
boundary layers that are not fully developed but are deep enough to envelop
the complete model. Although the full height of the ABL has been found to be
dependent on the wind speed and the latitude through dimensional analysis, the
typical height is around . Taking a geometric scale of , it is
expected that a wind tunnel of a height of will be required to simulate the full
ABL [40]. Even if this were achieved, its significance would be for models of
very tall buildings with actual dimensions large enough to allow for their
geometrical details to be represented adequately at this scale. For low-rise
buildings with typical heights as defined in Section 2.2.2, this scale would result
in extremely small models which would prohibit reproduction of geometrical
details of the buildings [10];[40]. In addition, the consequent huge differences in
the between model and full scale might then render the wind tunnel test data
unreliable [10];[40];[38]. Thus, for such low-rise buildings, resources are only
concentrated on modelling the inner surface layer of the atmosphere (see
section 2.1.1) which is approximately deep in full scale. When that is
done, larger and more practical models of the building with scales in the range
to can be used [40]. Then, the reference mean wind speed is taken
from a height (usually corresponding to as used in the full-scale Aylesbury
tests) closer to that of the building [38];[32].

It is obvious that the simulation of the boundary layer is limited by


considerations such as the available wind tunnel dimensions and geometrical
scaling of the models as well as the cost. In the early days of the ABL
simulations when existing aeronautical wind tunnels were of low aspect ratio
(that is short in relation to their height and width), it became pertinent to

130
investigate ways of using these tunnels for the simulations to avoid the huge
costs of building new more suitable ones [40].

These ways usually incorporated the use of tapered fins or spires to ‘artificially’
produce an immediate velocity gradient upstream of the roughened short tunnel
floor. The desired turbulent nature of this flow was achieved by the introduction
of upstream grids or barriers along with the turbulence generated by the
roughened floor [40];[28]. It is most likely that the flows produced by these
methods would still be in a process of development at the end of the short test
section. Also, the vortex structures shed in the wakes of the various upstream
devices would result in undesirable turbulence characteristics at positions of
measurement which might never be detected unless detailed fluctuating velocity
measurements were carried out including spatial correlations [40]. Moreover,
the desired turbulence characteristics might decay downstream if they were
simulated adequately at all [28]. It is a relief that wind pressures and forces on
structures are, apparently, dependent only majorly on turbulence characteristics
as discussed in Sections 2.1.3.1 and 2.2.2 in the along-wind direction and not
also on the detailed structures of the eddies within the turbulence [40].

Cook [28] cites in his paper the work carried out by Davenport and Isyumov
( ) and by Cermak and Arya ( ) on the review of the problems of
simulating the ABL in wind tunnels. He reports that both sources conclude that
the naturally grown wind tunnel boundary layer is a good model of the adiabatic
ABL. They argue that it is much more preferable to that achieved by artificial
methods as regards both the mean and the turbulent velocity characteristics.
The “naturally” grown boundary layer, while it may be restricted to research
institutions having economic justifications for a very long wind tunnel, gives
excellent agreement with the atmospheric data. The purely ‘artificial’ methods
(in which only the mean velocity profile is reproduced) still find areas of
application especially where the mean velocity in the near field of buildings is
the parameter of concern. The multitude of present simulation methods are
hybrids of these two techniques in which a natural boundary layer is allowed to
grow over an intermediate fetch of wall roughness after an initial artificial start.

131
These methods, depicted in Figure A-1, are generally classified as the
‘Roughness, barrier and mixing-device methods’ [28].

These methods were developed largely based on intuition and trial-and error.
Accordingly, the various hardware used differ across the methods including
those developed by Counihan, in 1969, and Cook, in 1973 [40];[28]. However,
these varying devices perform similar functions. The surface roughness acts as
a momentum sink which establishes the Reynolds stress profile in the layer. It,
therefore, controls the mean velocity profile and turbulence characteristics. It is
perhaps the most important component as it gives values to the three ‘law-of-
the-wall parameters , and (see [27] for more details). The barrier gives an
initial momentum deficit and depth to the layer which is mixed by turbulence
generated by the mixing device into the developing layer. Theoretically, all
these should result into a flow just downstream of the roughness section with
characteristics of a naturally grown turbulent boundary layer over a much longer
fetch of the same roughness without any undesirable consequences of the
presence of the barrier or mixing-device. This may not be the case as discussed
earlier [28];[40].

132
Figure A-1 Roughness, Barrier and Mixing devices method of simulating the ABL
[28]

The desired role of the barrier is, thus, to create the boundary-layer height [28].
The advantage of this is that simultaneous control of the turbulence intensities
and length scales to match the model scaling ratios can be achieved by the
adjustment of the height of the barrier [40];[28].

The simulation described so far can be achieved with more ‘active’ methods
than the ‘passive’ ones discussed already. The physical barrier can be replaced
by air-jets whence the mixing-device may be omitted completely. A major
advantage claimed for using this is that the momentum deficit can be adjusted
independently of the wind speed [28]. However, Cook [28] is of the opinion that
the expense of using these methods may only be justified for large wind tunnels
where using the ‘passive’ methods for differing applications (and/or differing
wind directions) may amount to major construction exercise as opposed to how
easy it may be for smaller wind tunnels in which allowances have been made
for the easy incorporation of the flow devices for the ‘passive’ methods.

133
The possible variation of upstream surface roughness and exposure conditions
in full scale with wind direction has to be taken into account when wind tunnel
model studies are commissioned. In addition, the model studies must
incorporate the necessary requirements to transform the site climate data to the
reference velocity used in the wind tunnel with considerations to the type of gust
measurements obtained as well as the height of the measurement used [22].

A.2 Modelling techniques


The justification for and the caution in the use of atmospheric wind tunnel
studies on surface pressures on buildings have been elaborated in Section
2.2.2. It may be argued that, in the present times, it is often the case to
commission wind tunnel studies without much consideration to the concept of
‘Dimensional Analysis’. This may appear to be right because ABL wind tunnel
techniques have already been well established. However, it may also appear to
be suggestive that all dimensionless parameters that may be of interest in every
study have been figured out. This cannot be taken to be true. While
commenting on the work of Baskaran and Stathopoulos [19], Cook [29 p130 -
131] writes:

“Surely one must always be able to express the problem in terms of non-
dimensional parameters. The art is finding the ‘right’ normalising parameters
to obtain a universal result.”

Thus, the knowledge of Dimensional analysis cannot be over-emphasised in the


modelling of structures for wind effects.

In the Dimensional analysis of building structures, the response of such


buildings to wind loading, including resonant dynamic effects, may be
considered as a function of such basic variables as follows: ; (Section
2.1.2); (Section 2.1.3.1), , ; (Section 2.1.3.1), , ; ; ; ; ; ; ; ;
[40].

“The Buckingham Pi theorem states that, for dynamic similarity to exist


between two systems, if the two systems depend upon independent
variables, then all dependent variables, made “non-dimensional” using some

134
of the independent variables, shall be functions of non-dimensional
groups of the independent variables; all the independent variables being
used” [51 p31].

Satisfying the three dimensions of ‘Mass’, ‘Length’ and ‘Time’ eliminates


variables and accounts for the . Hence, the stated variables can be
reduced to independent dimensionless groups inclusive of some of the
following [40] :

 Reynolds number, , ;

 Jensen number, , ;

 Strouhal number, , ̅
;

 Turbulence intensities, ; ; ;

 Length ratios, ; ; ;

 Density ratio, ;

 Froude number, √ (inertial forces(air)/ gravity forces(structure));

 Cauchy number, (internal shear forces in structure/inertial forces

(air)); and
 Critical damping ratio,

For dynamic similarity or correct scaling between the model and full-scale
building/structure, the values of the non-dimensional groups should be the
same [40]; [51]. As earlier stated, the groups are not a unique set. Therefore,
other non-dimensional groups can be formulated from the sixteen basic
variables even though there are only thirteen independent groups. However, it
will be the case that the additional groups can be formulated by taking products
of the stated groups or their powers [40]. It is noteworthy that equality in all of
these parameters will be required to guarantee ‘strict similarities’ in flow
patterns and pressure distribution between model and full scale. Unfortunately,
it is never possible to satisfy each and every one of them simultaneously. In this

135
case of conflict, the important ones in the conditions of interest must be satisfied
(see [51] for more details).

The UK NA to the Eurocode goes further to demand that:

b) “the building has been modelled at a geometric scale not more than
the following multiples of the geometric scale of the simulated
natural wind, with appropriate corrections applied to account for any
geometric scale discrepancies within this range:
 for overall loads; and
 for cladding loads;” [16 p2]

The geometric scale is defined in Equation (A-1) [36 p50]:

(A-1)

The walls and roof of a closed test section wind tunnel constrain the flow around
the modelled building or complex of buildings. This is dependent on the
blockage ratio which is the maximum cross-sectional area of the model at any
cross-section divided by the area of the wind tunnel cross-section [40]. Holmes
[40] explains that depending on how high the blockage ratio is, significant
increases in the flow velocities around, and consequent pressure distribution,
on the model can be expected. However, the reverse is the case for an open
test section; the velocities around the model are reduced. He identifies the
following ways of mitigating the blockage effect:

 It can be ensured that the blockage ratio is small enough to be equal to


or less than the maximum of below which rules usually absolves the
experiment of any corrections;
 Higher blockage ratios and corresponding corrections can be used (it is
usually not certain what the appropriate corrections are as only very little
information about this is available for building models in a turbulent wind
tunnel boundary layer flow); and

136
 The design of the walls and/or roof of the test section can be made in a
way as to limit the blockage errors. Slotted walls and roof with
symmetrical aerofoil slats backed with a plenum chamber are
incorporated in the slotted wall concept of achieving this. With this, it has
been claimed blockage area ratios as high as can be used without
correction. This has been used at the BRE in the UK [36].

The upstream roughness length is scaled from . For the determination of


wind loads on buildings, it is usually the case to use a lower value of as a
conservative choice. This results in the minimum demand in Equation (A-2) that
[36 p50]:

(A-2)

The profile of the mean wind velocity has been discussed to be modelled by the
appropriate application of the Jensen law. This can also be done by application
of the appropriate exponent in a power law profile. Modern ABL wind tunnels
possess a number of profile characteristics obtainable on demand. As much as
the geometric scale, the wind velocity scale, ; and time scale, are also
relevant [36 p51]. They have the following definitions in Equations (A-3) and
(A-4):

(A-3)

(A-4)

The frequency , being the inverse of time , scales as presented in Equation


(A-5).

(A-5)

137
The wind velocity used in the wind tunnel has to achieve the minimum
requirements specified by and . A wind velocity scaling in the range
is commonly used for the West European wind climate [36 p52].

The importance of scaling the sample frequency and time appropriately in the
wind tunnel cannot be over-emphasised [10]. This can be achieved by ensuring
equality in the both at full and model scales as in Equations (A-6) and (A-7).

(A-6)

(A-7)

At the onset of the deliberate use of the boundary layer wind tunnel for studies
on building models, the tests were performed in a uniform flow with velocity not
varying with height [25]. This produced inaccurate results. Not until the
understanding of the nature of wind (such as discussed in section 2.1) was
gained was there any progress made [25]. The reference wind velocity, ,
which would provide reliable estimates of the wind loading is taken at some
reference height, . This reference height may be taken to be equal to the
height of the building when the building is lower than twice the average height of
the surrounding buildings [36].

The importance of the equality of the turbulence characteristics at full and


model scale have been elucidated earlier (Section 2.2.2). These have also been
said to be unobtainable simultaneously. However, for wind loading studies,
lower levels of turbulence than is the real case, which result into higher loads,
are taken in compliance with “conservatism” [36 p52]. The requirement is given
in Equation (A-8).

(A-8)

In the comparison with full-scale results, wind tunnel experiments may correctly
be seen as a matching process in which the simulations might just be varied
until an acceptable agreement is achieved instead of being independent studies

138
[25]. The bid to defy this was the motivation behind the work of Surry [60];[59]
who gathered wind tunnel data on the model of the Texas Tech building in
advance of the full-scale studies. He found very good agreement between the
two experiments particularly when the approaching wind was nearly normal to
the ridge and the differing frequency responses in the measuring instruments
have been accounted for on the magnitudes of the peaks. However, for
quartering winds on the ridge, some significant differences in the peak
coefficients were observed. He suggests that this may be attributed to obvious
difference in the “gust-to-mean ratios in these storms [turbulence intensities] or
to non-stationarity” [59 p246].

The need to keep a non-varying pressure gradient in the along-wind direction


informed the practice of installing roofs of adjustable heights in the early
boundary layer wind tunnels. With this, the increasing velocity gradient in the
direction of flow was guaranteed and the free-stream velocity outside of the
boundary layer almost kept constant. This also allows the reduction in the
blockage errors for large models. In the case of smaller models with lower
blockage ratios, it has been found to be unnecessary to adjust the roof since the
measurement errors when the height of the roof is kept constant or with a
constant slope are quite negligible [40].

The Ekman spiral (see Section 2.1.1) cannot be simulated in conventional ABL
wind tunnels. In any case, the change in wind direction due to it is usually
ignored for the height range of most structures [40].

A.3 Measurements
The UK NA to the Eurocode finally demands that:

c) “the response characteristics of the wind tunnel instrumentation are


consistent with the measurements to be made;
d) the tests enable the peak wind loads with the required annual risk of
being exceeded to be predicted.” [16 p3]

139
Holmes [39] relays that a measurement system of an inadequate frequency
response can attenuate peaks in measured pressures. In addition, a general
demand is that all instrumentations used in the wind tunnel test should have
known calibration results [36].

Measuring the simulated boundary layer

Pitot tubes and hot wire (or film: refer to [27] for further details) anemometry
have traditionally been considered appropriate for measurements of velocity
and pressure in aerodynamics practice. Thermal anemometry (see Figure A-2),
being the more responsive, has been used to measure the mean and peak
properties of the wind profile [24];[36]. However, Cochran [24] who performed a
state-of-the-art review of wind tunnel instrumentation for wind engineering
practices reports that multi-hole pressure probes, such as the cobra probe in
Figure A-3, have been designed to be even more responsive than the thermal
anemometry system. They incorporate typically between and (as seen in
Figure A-4) holes in the probe head alongside miniature pressure transducers;
this gives them the good frequency response. Therefore, in recent years, many
commercial wind-engineering laboratories have used them to collect time-series
pressure data that can be converted through analytical software to speed and
direction over wide ranging angles. These can include reverse flow in separated
regions even in highly turbulent flows; this cannot be achieved by a lone hot
wire or film [24]. Nonetheless, this promising technology is still under evaluation
by the wider wind engineering community.

It should be noted that the techniques discussed above are of the intrusive type.
Non-intrusive techniques like the Laser Doppler Anemometry (LDA) and Particle
Image Velocimetry (PIV), although more expensive and more difficult to setup,
can also be deployed for the same purpose yielding possibly better results [24].

140
Figure A-2 Hot-wire probe [24]

Figure A-3 Multi-hole probe [24]

141
Figure A-4 Schematic of a 4-hole cobra probe [24]

Measurement of pressures

The measurements of pressure are taken as pressure differences between the


building surface and a reference pressure [36]. Guerts and van Bentum [36]
who give a good insight into the general demands of a well performed ABL wind
tunnel experiment in their paper note the minimum demands that:

 The position of the reference pressure is chosen so, that this pressure is
independent of wind direction or changes made to the model; and
 The frequency-response characteristics of the pressure measurement
equipment need to be available on request.

It is usually the case that the pressure sensor unit is remotely located from the
point of pressure measurement for possible geometrical and/or cost reasons.
The adequate transmission of the fluctuating pressures such as the peak
pressures or suctions between these ends depends greatly on the length of the
tubes transmitting the pressures and the frequency response of the whole
system. Consequently, the frequency response of the pressure measurement
system is of importance [40].

142
For single point measurements, three systems as illustrated in Figure A-5 are in
use. These have varying degrees of frequency response. Therefore, they are
suitable for different purposes [40].

Figure A-5 Single point pressure measurements a. Short tube b. Restricted tube
c. Leaked tube [40]

Geurts and van Bentum reports further on the demands for local loads:

 that a selection of locations of expected increased local loads be


made; and
 that measurements of pressure at sufficiently high frequency be
made such that extreme loads which corresponds to loaded areas
designated in the codes can be determined.

In addition, when the measured pressure will be used to obtain the overall wind
loads on the whole structure then:

 All surfaces should be fitted with large numbers of pressure tappings;


 The contribution of friction should be provided for with the use of the
appropriate sections of the building code; and
 Pressure measurement at all tappings should be simultaneous to keep
the time information.

143
A.4 Model Construction
There are two ways of modelling the structure of a building in question in a wind
tunnel. The model can either be constructed as an Aerodynamic model or as an
Aeroelastic model.

The Aerodynamic model, in which the structure is modelled as a rigid non-


flexible body (and which is the simpler and cheaper) is used when resonant
dynamic actions of a structure can be ignored. With the availability of cheap
pressure transducers and the geometrical possibility of fitting numerous
pressure taps on an aerodynamic model, it can also be used to assess resonant
dynamic structural loads on a building through the acquisition and analysis of
simultaneously obtained pressure time-series data on the model. This
notwithstanding, it may be required to gain a more accurate description of the
building’s dynamic response to the approaching wind. In such a case,
Aeroselastic models, in which the flexibility of the structure as well as such
parameters as the variation of mass with height and the damping are modelled,
are used [23];[40];[24].

A.5 Analysis of Measured data


Methods of Analysis

Depending on the data available from the wind tunnel and what information is
desired from the experiment, a number of different approaches to the analysis
of the data are available. Two of such, with far reaching definitions and usage,
are briefly discussed here.

The first is the quasi-steady method which forms the basis for many wind
loading codes and is used when only time-averaged information is available. It
assumes that the fluctuations in the surface pressures on the model adequately
follow the variations in the oncoming flow. Thus, extreme wind loads can be
predicted or estimated using mean pressure or force coefficients with a peak

144
gust wind speed. It ignores building-induced pressure fluctuations. Also, it
assumes full correlation of pressure peaks when used over a large area [40].

The second is the extreme value method. This is used when a sufficient number
of time series with adequate length is available. The peaks in the values of the
parameters of interest are analysed based on the extreme value analysis
available in the literature. The values obtained with this method have been
applied in the Eurocode [36].

Sample length, number of samples and sample frequency

In the determination of statistical values of the (see Section 2.3.1.1), one or


more time series are required to be measured for every wind direction and for
every measurement channel, per configuration investigated. However, when the
mean values are of interest, a single time series is required per wind condition
with a length chosen so, that a longer time series will not present a different
mean value. When using the extreme value analysis, more than one time series
is required with a fixed length . All the time series are measured with the same
sample frequency determined using the [36]. Geurts and van Bentum relay
the following minimum demands for an extreme value analysis:

 Same length for all the time series;


 Time series are independent without overlaps;
 The time series each represent full scale duration of at least . This is
determined using the ; and
 For every wind direction, at least 24 time-series are required.

The sample frequency is, at the minimum, twice the frequency of interest for the
wind effect explored.

145
Appendix B Building Models Dimensions and Pressure
tapping location
Figure B-1 is the key figure for the Sports Hall (SH) model dimension and
pressure tapping locations presented in Figure B-2, Figure B-3 and Figure B-3.

Figure B-1 Key figure for the dimensions and pressure tapping locations on the
Sports Hall (SH) model

The normalizing parameter which is the height of the SH model.

146
Figure B-2 The West face of the SH model

Figure B-3 The South face of the SH model

Figure 5-14is the key figure for the dimensions and pressure tapping locations
on the Trafficmaster (TM) model presented in Figure B-5 all through to Figure
B-12. The normalising parameter ; this is the height of the TM
model.

147
Figure B-4 The Roof of the SH model

Figure B-5 Face 1 of TM model

148
The vertical axis in Figure B-5 adjoins Face .

Figure B-6 Face 2 of TM model

The vertical axis in Figure B-6 adjoins Face .

149
Figure B-7 Face 3 (the roof) of TM model

150
Figure B-8 Face 4 of TM model

The vertical axis in Figure B-8 adjoins Face .

Figure B-9 Face 5 of TM model

The vertical axis in Figure B-9 adjoins Face .

151
Figure B-10 Face 6 of TM model

The vertical axis in Figure B-10 adjoins Face .

Figure B-11 Face 7 of TM model

The vertical axis in Figure B-11 adjoins Face .

152
Figure B-12 Face 8 of TM model

The vertical axis in Figure B-12 adjoins Face .

Figure B-13 presents the plan dimensions of the Hangar building (HB) model
and pressure tapping locations. The part where tapping is located is the
highest part of the model with . This is the normalising parameter for
the dimensions of this model.

153
Figure B-13 The Hangar building (HB) model plan dimension and tapping
location

154
155

You might also like