Académique Documents
Professionnel Documents
Culture Documents
ON LOW-RISE BUILDINGS
by
A DISSERTATION
IN
CIVIL ENGINEERING
DOCTOR OF PHILOSOPHY
Approved
Chair~ommittee
hool
December, 1997
ACKNOWLEDGEMENTS
ii
TABLE OF CONTENTS
ACKNOWLEDGMENTS ii
ABSTRACf vii
UST OFTABLES ix
UST OF FIGURES xi
CHAPTER
1. INTRODUCITON 1
1.1 General Remarks l
1.2 Background Information 2
1.2.1 Introduction 2
1.2.2 Wind Damage 2
1.2.3 Wind Load Codification 3
1.3 Objectives and Scope 4
1.3.1 A Brief Background Review 4
1.3.2 The CUSfiTU Cooperative Program 6
1.3.3 Research Objectives and Scope 6
2. LITERATURE REVIEW 9
2.1 Introduction 9
2.2 Wind-tunnel Simulation of Wind and Loads 10
2.2.1 Leading Edge Phenomenon 10
2.2.2 Wind Turbulence 11
2.2.3 Effects of Non-stationarity of Natural Wind 13
2.2.4 Characteristic Length-scale 14
2.2.5 Velocity Protile and Scale Effect 15
2.2.6 Reynolds Number and Blockage Effect 16
2.3 Separation Bubble and Conical Vortices 17
2.3.1 Separation Bubble 18
2.3.2 Conical Vortices 20
iii
2.4 Summary 23
3. WIND LOADS AND BUILDING PERFORMANCE 25
3.1 Introduction 25
3.2 Atmospheric Boundary Layer 25
3.3 Wind and Turbulence 27
3.4 Wind Effects on Buildings 29
3.4.1 Observations of Wind-induced Damage 30
3.4.2 Air-flow and Building Failure 32
3.4.3 Engineering Attention and Building Performance 35
3.4.4 Improvement of Building Practice 38
4. THE WERFL FACILITY 40
4.1 Introduction 40
4.2 Test Building and Meteorological Tower 41
4.3 Anemometers and Pressure Measurement System 44
4.3.1 Anemometers 44
4.3.2 Pressure Measurement System 44
4.4 Data Acquisition System and Data Validation 46
4.4.1 Data Acquisition System 46
4.4.2 Data Analysis and Validation 47
5. DESIGN OF EXPERIMENTS 48
5.1 Introduction 48
5.2 Incident Wind and Pressures 48
5.3 Wind-flow Measurement 49
5.4 Flow Visualization 51
5.5 Pressure Measurement 52
5.6 Comparison of Velocity Measured with Different Anemometers 53
iv
6. SEPARATION BUBBLE AND ITS EEFECTS ON PRESSURE 55
6.1 Introduction 55
6.2 Structure of Separation Bubble 56
6.2.1 Flow Measurement 56
6.2.2 Velocity Vectors 56
6.2.3 Boundary of Separation Bubble 60
6.3 Wind Flow Characteristics 61
6.4 Pressure Distribution 61
6.5 Mechanism of Peak-pressure Generation 65
6.5.1 Non-conventional Pressure Coefficient 66
6.5.21ncident Wind and Pressure Coefficient 66
6.5.2.1 Incident Wind 67
6.5.2.2 Conventional Pressure Coefficient 68
6.5.2.3 Wind Angle and Pressure Coefficient 69
6.5.2.4 Non-conventional Pressure Coefficient 71
6.6 Spectral Analysis of Pressure 74
6. 7 Summary 77
7. CONICAL VORTICES AND THEIR EFFECTS ON PRESSURE 79
7.1 Introduction 79
7.2 Theoretical Background of Vortex and Pressure 79
7.3 Characteristics of Pressure on Leading Roof-comer 81
7.3.1 Pressure Coefficient for Tap 50101 82
7.3.2 Pressure Coefficient for Tap 50501 89
7.3.3 Predicting Pressure Coefficient from Incident Wind 94
7.4 Pressure Zones and Critical Wind Angles 99
7.5 Features of Conical Vortices 101
7.5.1 Position of Vortex Core 101
7.5.2 Forms and Patterns of Vortex Pair 103
v
7.6 Comparison of Wind Velocity Measured with Different Anemometers 111
7.7 Summary 113
8. CONCLUSIONS AND SUGGESTIONS 117
8.1 About Separation Bubble 117
8.2 About Conical Vortices 118
8.3 Suggestions for Further Study 119
REFERENCES 120
APPENDICES
A. TECHNICAL SPECIFICATION
OF THE SONIC ANEMOMETER 126
B. DATA ACQUISffiON LOG 127
C. PRESSURE COEFFICIENT DATA
FOR SELECTED TAPS FROM MODE-15 145
D. DERIVATION OF EQUATION
FOR PRESSURE PREDICTION 158
vi
ABSTRACf
Two fundamental flow phenomena, the separation bubble (SB) and conical vortex~
over the roof (flat and rectangular) of the Texas Tech test building are studied in terms of
flow characteristics and pressure-generating mechanisms. Major fmdings contribute to
understanding the mechanisms of pressure generation and the roles of turbulence and
other properties of the incident wind in loading effects.
Much of the debate on wind-tunnel simulation priority in the wind engineering
community has failed to distinguish the different effects the incident wind has on two
types of quantities: single events, e.g., minimum (peak) pressures, and statistics~ such as
the mean and the rms pressures. The turbulence intensities, reflecting the gust structure
and the directional fluctuations of the free-stream wind, might bear significant influences
collectively on the pressure statistics. It is the intensity of individual longitudinal gusts
(as well as the incident wind angles) that decides the data-run-wise peak-pressure
coefficients on the roof corner~ while the peak-pressure coefficients associated with the
SB are governed by the lateral directional fluctuation in the incident wind. Proper
simulation of the incident wind profile (boundary layer type) is probably the most
important single input in the wind- structure interaction process.
The separation bubble, having a mean reattachment-point approximately 10 ft from
the leading edge, is oblong and elongated in the horizontal direction. Pressure distribution
on the roof surface is intimately related to the structure of the SB. Conditional sampling
technique indicates that the wind flow associated with strong suctions is highly three-
dimensionaL Introduction of the non-conventional pressure coefficient makes it possible
to isolate the effect of wind speed from that of the wind direction on pressure generation.
The mechanism of peak-pressure generation associated with the SB is governed by
directional fluctuations of incident wind. Primary peaks of pressure coefficient
(conventional) are often an outcome of combined wind gust and fast direction fluctuation.
Non-Gaussian-pressure zone along the short roof-axis is found to extend about l5 ft from
vii
the leading edge. This dimension is comparable to the stream-wise overall size of the
vortex circulation and to the height of the test building (13 ft).
Low pressures (high suctions) on a leading roof comer are always related to strong
vortices. The wind angles favoring vortex formation over a roof comer are bounded in a
50 range symmetric about the diagonaL Within this range, the form and dimension of a
single vortex, and the combination (pattern) of the vonex pair vary significantly with the
incident wind angle. Instantaneous pressure coefficients for taps underneath either of the
vortices can be satisfactorily predicted from the mean pressure coefficient and the wind
direction and speed of the incident wind. Peak pressures on the roof corner are produced
by wind gusts approaching at wind angles conducive to strong vortex formation. This
mechanism is different from that of the SB in that the fluctuations of wind direction (not
the direction itselt) are important in the case of the SB.
A practical means of wind load prediction is established. It is a uhybrid" approach
that combines the mean pressure coefficient.. obtainable from either full-scale experiment
or model test in properly simulated boundary layer wind flow, with easily available site-
specific wind data measured at an appropriate point.
viii
LIST OF TABLES
ix
B.21 Sonic data acquisition log (10/12/1997) 142
B.22 Sonic data acquisition log (10/12/1997) 143
B.23 Sonic data acquisition log (10/12/1997) 144
X
LIST OF FIGURES
xi
6. 7 Pressure coefficient versus wind angle 71
6.8 Wind-structure interaction and loading effect 72
6.9 Simultaneous traces of incident wind and pressures 73
6.10 Spectra of pressure measured at six taps along short roof-axis 76
6.11 Pressure zones on flat roofs: Gaussian and non-Gaussian 76
6.12 Typical trace of pressure coefficient: Gaussian 78
6.13 Typical trace of pressure coefficient: non-Gaussian 78
7.1 Pressure distribution under a free point-vortex 80
7.2 Instrumented roof comer and taps 81
7.3 Tap 50101 pressure coefficients versus wind angle 84
7.4 Schematic view of vortices as they affect pressure tap 50101 85
7.5 Straight flow and vortices formed at incident wind around 225 85
7.6 Time series of pressure coefficient and incident wind angle (M49nl60) 87
7.7 Time series of pressure coefficient and incident wind angle (M49n163) 88
7.8 Tap 50501 pressure coefficients versus wind angle 90
7.9 Location of tap 5050 1 with respect to vortices 91
7.10 Simultaneous series (M49n 161) 92
7.11 A close-up look of incident wind and pressure (M49nl6l, tap50501) 93
7.12 Mean Cp versus wind angle (tap 50501, Mode-15) 95
7.13 Traces of variables used in pressure prediction 96
7.14 Comparison of pressure coefficients 97
7.15 Schematic view of variability of upstream wind 98
7.16 Division of pressure zones on tlat roof 100
7.17 Schematic view of flow visualization setup 101
7.18 Position of vortex core derived from flow visualization 102
7.19 Schematics of vortices and incident wind angle 103
7.20 Vortex at the left edge (x =l0'-06n) 105
7.21 Vortex at the left edge (x =6ft) 105
7.22 Effect of wind angle on vortex shape and core position 106
xii
7.23 A pair of near-circular vortices 106
7.24 A big oblong vortex forming at the right edge 107
7.25 A big oblong vortex forming at the left edge 108
7.26 Vortex and its relationship to pressure-tap location 108
7.27 Wind direction and pressure coefficient 110
7.28 Comparison of longitudinal velocity spectra 112
7.29 Comparison of lateral velocity spectra 112
7.30 Comparison of horizontal velocity components 114
7.31 Comparison of vertical velocity component and wind direction 115
B.l Coordinate system 128
B.2 Definition of instrument u- v- w-components 129
B.3 Locations of sonic anemometer and grid 136
B.4 Schematics of experimental setup 138
B.5 Sketch of experimental setup for velocity measurement comparison 139
B.6 Definition of instrumental u- v- w-components (M49n252-259) 140
B.7 Defmition of instrumental u- v- w-components (M49n26l-265) 141
B.8 Defmition of instrumental u- v- w-components (M49n267) 142
B.9 Defmition of instrumental u- v- w-components (M49n269-272) 142
B.10 Defmition of instrumental u- v- w-components (M49n274-283) 143
B.ll Definition of instrumental u- v- w-components (M49n284-287) 144
C.l Tap 50901 pressure coefficient data 146
C.2 Tap 50205 pressure coefficient data 147
C.3 Tap 50505 pressure coefficient data 148
C.4 Tap 50905 pressure coefficient data 149
C.5 Tap 50209 pressure coefficient data 150
C.6 Tap 50509 pressure coefficient data 151
C.7 Tap 50909 pressure coefficient data 152
C.8 Tap 50123 pressure coefficient data 153
C.9 Tap 50523 pressure coefficient data 154
xiii
C.lO Tap 50823 pressure coefficient data 155
C.11 Tap 51123 pressure coefficient data 156
C.12 Tap 51423 pressure coefficient data 157
0.1 Definition sketch 158
xiv
CHAPfERl
INTRODUCTION
rigid structures, and can be dealt with by equivalent static loads. Standards and codes
often specify non-dimensional pressure coefficients for different structural elements for
limited variations of buildings, and specify force coefficients for other structures. These
coefficients are used in conjunction with site specific wind data and terrain characteristics
to detennine design wind loads. In most cases, the coefficients are established from wind-
tunnel tests conducted in simulated atmospheric boundary layers. The other approach is
applicable whenever the buildings and other structures are likely to be susceptible to wind-
induced vibrations. For these buildings and structures, customized wind-tunnel tests or
other experimental methods~ or a detailed procedure is to be followed to determine wind
loads. Wind-tunnel tests are more appropriate when more exact definition of dynamic
response is needed, and for determining exterior pressure coefficients of buildings with
complicated geometry.
It is thus obvious that adequate and efficient structural design against wind induced
damage relies heavily on precise specification of wind loads. Wind-tunnel testing
techniques have made significant contributions in the load codification. Engineered
structures such as the main-wind-force-resistance frames of many high-rise buildings have
performed well in severe storms. However, there is still a need to improve the
performance of residential buildings and other types of low-rise structures. Peak pressures
of short duration are important in the case of low-rise buildings, where the damage of
components and cladding often leads to structural destruction. Roof edge and roof comers
sustain the largest peak pressures of all the surfaces of a building. It is important to
understand the mechanism of pressure generation for appropriate design and developing
strategies to reduce the intensity of pressures in those regions.
2.1 Introduction
Historically~ wind engineering has focused on preventing failure of the main-wind-
force-resistance-system (MWFRS) of structures. However, direct losses from the breach
of the building envelope and the consequential water damage to structural elements and
building contents have been tremendous when severe storms struck the increasingly
constructed and densely populated coastal regions. Recent years have seen a shift of focus
on improving the performance of building envelopes for mitigation of property losses.
The regions of worst suction~ where initial damage starts, on the roofs of low-rise
buildings, comprise only a small fraction of the total roof area and its influence on
structural loads is small; it is, however, critically important for roof-cladding or roof-
covering systems (Kind~ 1986). Failure of roof systems tends to begin at the upwind
corner of the rooftop, and an initial failure could precipitate failures over large areas (Kind
and Wardlaw, 1982). Our ability to design against such component failures underlies
considerable economic significance, and this, to a large degree, depends on how well we
know the complicated wind loading effects on structures.
As aforementioned, the technique we have relied on heavily for determining wind
loads has room for improvement. The staggering amount of property-loss our society has
suffered recently testifies to this. The unpleasant reality, from the technical viewpoint, is
that it is precisely with those critical regions in a building envelope that we have been least
successful in load predicting. Significant discrepancies of wind-induced pressures were
found in modeVfull scale comparisons for those regions.
Two flow phenomena associated with the critical loading can be identified for tlat
roofs of low-rise buildings: ( l) flow separation and possible reattachment, at normal-to-
wall wind; and (2) conical vortices formed over roof corners at oblique wind. In the
situation when the wind approaches normal to one of the walls., the flow above the
9
10
stagnation point on the windward wall will go over the roof. The up-wash flow separates
at the roof edge to form a separated shear layer.. which may reattach to the roof at some
distance downstream if the stream-wise dimension of the roof is large enough, to fonn a
separation bubble (SB). When wind hits a building at skewed angles, a conical vortex or a
pair of vortices may form after flow separation at the two oblique leading roof edges.
Both flow phenomena involve separation and vortex formation, which is closely related to
the pressures developed on the roof. The magnitude of the outward-acting pressure is
strongly dependent on the condition of the vortices, which are themselves very sensitive to
the incident wind, specifically, to the fluctuations of the wind speed and direction.
This chapter intends to present a thorough literature review of the two flow
phenomena and their loading effects on flat roofs of low-rise buildings, and on the state-
of-the-art of wind-tunnel simulation technique.
which typically realize only a couple of times in 15 minutes. Relevant works are scarce in
the literature. Only a few authors have made brief statements about the non-stationarity of
wind possibly being the reason for full-scale- model-scale mismatch.
Surry ( 1989) compared the mean, rms and peak pressure distributions from the full-
scale data of the WERFL and that from the wind-tunnel results. He used the wind-tunnel
data for an oblique wind angle of 60 and the full-scale data with wind angles in the range
of 50 - 70. Reasonable agreement was found in the mean and rms pressure data, but the
peak pressure data in the full-scale were much larger than the model predictions. The
scatter bands, indicating the span of the results from 8 full-scale runs, showed that the
results for this wind angle were highly variable; significant differences extended to both the
front walls and the wake, in the regions where quasi-steady type of response was
expected. He concluded that this observation was strongly suggestive of significant
differences in the gust structures of natural wind and that of the simulated winds. Lin et at
14
( 1995) aJso pointed out that, among other factors, non-stationary effects in the full-scale
might be the cause of discrepancies.
Figure 2.1 General flow pattern at nonnal-to-wall wind (hlb<1J.5, after Cook~ 1990)
For the conical vortices developed over the upwind roof-comer, Lin et al. (1995)
argued that the length from the stagnation point on the two windward walls, where a
boundary layer began, to the roof edge, where the boundary layer separated, was an
appropriate characteristic length. Since the length over which the wall boundary developed
was a portion of the wall height (h), h could be taken as a convenient scale length. They
went on to suggest that similar wall-boundary-layers would develop for similar geometry.
In the same paper, Lin et al. used pressure models having plan dimensions 3 times the
length of the long side of a 1:50 model of the WERFL building, with varied heights of 1, 2
and 3 times of the exactly scaled model They found that although the mean, rms and peak
15
values of Cp at the same location, on a rays or at (x, y). changed from model to model. the
pressure coefficients seemed to be independent of the plan dimensions, especially for the
mean pressure coefficient. However, the curves of Cp versus s/h or (xlh, ylh) essentially
collapsed for different model heights (in the same flow) and thus tended to be independent
of the building dimensions. They suggested adopting the height rather than one of the plan
dimensions as the characteristic scale, provided the height to plan dimensions ratio fell in a
certain range (Lin et al, 1995).
large values of h/8, the rooftop was below the logarithmic portion of the velocity profile;
16
that is., it was in a non-universal portion of the flow where conditions depend on the
specific characteristics of the underlying surface roughness. In such cases.. the roughness
parameter Zo or the ratio hiZD would have little direct physical significance. The non-
dimensional velocity gradient at roof level appeared to result in lower pressures. Because
increased velocity gradient could be expected to result in reduced flow up-wash and thus
reduced suctions.. in a similar way as reduced angle-of-attack of a delta wing leading to
smaller lift on the wing.
L'1 an extensive study by Lin et al. (1995) .. it was observed that the largest magnitude
of Cp generally increased with model height and changed with flow characteristics. It was
also found that., near the roof corner, the mean pressures were much more sensitive to the
change of flow characteristics than the peak or rms pressures. Away from the corner, the
peak and rms pressures became more sensitive.
Although the difference of pressures induced on models immersed in the uniform
flows (Je =ex:) and in the atmospheric flows, typically Je =50 - 500, was remarkable, it
was usually sufficient to ensure that the Jensen number was in the correct range (Cook,
1990).
separation. In their view, almost all the factors other than the Re were Correctly' scaled
4
and simulated. They claimed that the model test results were consistent with the full-scale
data on the basis of the Re, and the differences could be corrected by extrapolating the
full-scale data to the model results in tenns of the Re. In another work., Savory and Toy
(1996) compared the pressure distributions of a 1:43 Silsoe Structures Building model
tested at the Re ranging between 2.35xl04 to 7.14xl04 , based on the ridge height. They
observed significant reduction of suctions at the leading eaves of the windward slope with
decreasing Re; the greatest dependence of pressure coefficients on the Re was with
regions having surface discontinuities, that is at the windward eaves and the ridge.
Unlike the situation of flow-structure interaction in the open atmosphere, the
streamlines around the model structures in the wind-tunnel tests are not completely free to
extend laterally; they are confined to some degree by the presence of the two side walls
and the ceiling of the wind tunneL Flow is accelerated in the constricted segment between
the model and the walls, increasing the loading of the structure. This is called blockage
effect and needs be minimized by choosing a low ratio of the projected area of the model
to that of the wind tunnel cross section. While approximate corrections to some extent are
possible, it is always a good practice to work with low blockage ratio. Studies at the
Bristol University in UK indicated that ground-mounted models were less susceptible to
blockage effect and 10% blockage might be acceptable without correction (Kirrane and
Steward, 1978), although most workers regarded 5% as the 'safe' limit (Cook, 1990).
Momep.\um from
V Reynolds stress
~
.,.__-w~--
Figure 2.2 Flow over flat roofs (reproduced after Cook, 1985)
19
~ Cp=-1.2
. Q.65
...
(a) Uniform wind flow (b) Boundary-layer wind flow
Figure 2.3 Pressure distributions for flat roofs (reproduced after Cook, 1985)
As a direct result of the different flow characteristics over the roof, the pressure
distributions on the roof were very dissimilar (see Figure 2.3). The pressure distribution
changed from nearly unifonn in the unifonn incident flow to increasingly negative toward
the upwind roof edge.
The flow inside the separation bubble at the central cross-section was basically a two-
dimensional flow where the reattachment process depends on the mass balance within the
separation bubble. The turbulence structure of the incident wind affected the entrainment
into the reattaching shear layer, and thus had a strong and direct influence on the mass
balance inside the bubble, and hence on the reattachment process and the pressure
fluctuations (Kind, 1986).
The size of the separation bubble was expected to scale to the across-wind dimension
of the front wall or two times the building height, whichever is smaller. The dynamics of
the vortex governing the loading characteristics on the roof depend on the wind angle, the
slenderness ratio, the roof pitch and so on; it was also strongly influenced by the incident
turbulence (Cook, 1990).
High suctions of short duration were found in regions of the roof close to the leading
edge both in full-scale measurement and in model test with properly simulated boundary-
layer flows (Cook, 1985). Melbourne (1975) speculated that the observed high suctions
were associated with instability of the shear layer triggered by incident turbulence.
20
Wagaman (1993) and Letchford (1995) are among the early researchers in full-scale
investigation of the separation bubble (SB). Wagaman (1993) established the length and
height of the SB over the WERFL test building through flow visualization; Letchford
(1995) made some observations about the dynamics of the SB and hypothesized the
mechanism of peak-pressure generation. Their work will be referred to later for
comparisons with this study.
As sketched in Figure 2.2(a), the flow separating along the edge entrains into a
circulation which increases as it moves downstream to create a pair of conical vortices.
These two vortices would have different strength unless they are formed at quartering
wind (Cook, 1985).
21
The conical vortices forming over the upwind roof corner of a low-rise building are
quite similar to the vortex over a delta wing at a moderate angle of attack. The vortices
break down at a certain point downstream. The center of each vonex is a region of high
negative pressure. The two vortices in Figure 2.4(a) produce two lobes of negative
pressure on the underlying roof comer~ as shown in Figure 2.4(b). These vortices~ acting
directly or interacting with the incident turbulence, are the principal cause of high uplift on
roofs (Cook, 1985).
Unlike the vortex of the SB, conical vortices are three-dimensional and highly stable
(Kind, 1986). Fluid particles in the vortices follow helical paths. The three-dimensional
reattachment process is governed by the laws of vortex dynamics and is essentially
independent of the viscous effects or the turbulence structures of the shear layer. This
statement is supported by the fact that inviscid analyses have been highly successful in
predicting this type of flow. Consequently, the turbulence intensity and scale is not likely
to have any significant effect on the three-dimensional conical vortex flow by virtue of
their effects on turbulence entrainment processes (Smith., 1984). In fact Stathopoulos et al.
(1978) found the occurrence of similar worst mean-suctions for very different turbulence
intensities of 35% and 15%. Direction fluctuation in the approaching wind due to the
turbulence causes a ''wandering"' of flow pattern and thus a 'smearing' of the mean
suction-peaks (Kind, 1986).
For the conical vortex phenomenon, the important length scale is likely to be the
distance over which the boundary layer on the windward walls of the building grows. That
is the length from the wall stagnation-point, where a new boundary layer embedding
vorticity begins, to the roof comer or any point on the roof edge, where the boundary
layer separates and the wall boundary-layer vorticity rolls into vortices developing over
the roof. This wall boundary layer would be similar for those models that have similar
frontal walls and roof-comer features, regardless of the plan dimensions, if the models are
exposed to similar flows and reattachment takes place on the roof. That is, the wake
pressure cannot have any substantial upstream influence. Hence~ the wall boundary layer
22
development (WBLD) length appears to be the non-dimensionalizing length of choice (Lin
et al.~ 1995). Since, in similar flows~ the WBLD length is likely to be a fraction of the
building height (h), h itself is an acceptable substitute. However, the WBLD length is
expected to be affected by the approaching boundary-layer flow conditions, which would
influence the entrainment rate in the wall boundary layer and the separating shear layers
over the roof, and hence the reattachment distance on the roof and the pressure.
Therefore, h is not expected to be a global collapsing parameter if the approaching flow
condition changes. A fixed power law velocity profile of the fonn, VIVh= {z/h)a., inherently
implies self-similarity, and the source of the most important residual differences is likely to
be the variation of the incident turbulence characteristics eaves height or along the
stagnation streamline (Lin et al., 1995).
The magnitude of high suctions at the upwind roof comer, and equivalently the
strength of the vortex, changes significantly with the wind angle. However, the location
does not seem to change very much. Each high suction lobe shifts only slightly from b=20o
to b=l0 or symmetrically b=80 to /r-70, when wind angle varies from about 15 to 75,
where b is the angle of the lobe axis with the roof edge (Lin et al., 1995). Within the two
lobes near the roof comer, the worst wind angles were around 30 and 60
correspondingly. For other locations, however, the wind direction corresponding to the
worst suctions was location-specific. Peak pressure coefficients beyond -18 for rough flow
and mean pressure coefficients beyond -12 for smoother flow were observed, at x/h=0.026
and ylh=<l0065 (or at a ray of an angle /r-14 at s/h=0.0268). The roof-level turbulence
intensity was 17%. It was unknown whether the suctions would continue to increase
toward the roof corner (Lin et al, 1995). However, Kind ( 1986) pointed out that the true
worst suctions on the upwind roof corner might have been missed in some wind-tunnel
simulations due to a lack of pressure taps sufficiently close to the roof edge. The suctions
seemed to increase monotonically toward the apex of a flat roof, as close to the apex as
installation of pressure taps allows (Kind, 1986; Jamieson and Carpenter, 1993). The true
worst suction remains to be identified in full-scale studies.
23
In an effon to better understand the conical vortex phenomenon and the pressures
developed on the outside surface of the roof~ the works of Kawaii and Nishimura ( 1996),
Marwood (1996} and Banks et al (1997) are representative. Kawaii and Nishimura (1996)
took detailed measurements of the incident wind and the pressure on the square roof of a
model (120 mm x 120 mm x 60 mm). They calculated the correlation between a reference
tap and the rest of the taps, corning up with a theory to separate the pressure fluctuations
into a low-frequency pan and a high-frequency part caused by different mechanisms.
Marwood (1996) measured the detailed flow field inside the conical vonex on a model
using a Laser Doppler Anemometer (LDA). By developing a simultaneous pressure-
velocity measuring technique and a conditional statistical technique, he could relate the
surface pressure to the velocity events within the vortices. Banks et aL (1997) performed
simultaneous flow visualization and pressure measurement on a 450 mm cuboidal modeL
Among their findings were: ( 1) the relationship between the wind angle and the angular
vortex-core location, (2) a linear change of the vonex-core displacement with the distance
from the roof apex, and (3) a good correlation between the pressure profiles and the
vortical circulation.
2.4 Summary
To conclude this chapter, a brief summary might be usefuL The main points discussed
in this chapter were as follows:
1. With its reasonable agreement with the full-scale measurements of building
pressures, wind-tunnel simulation will continue to be a valuable tool for research
and applications;
2. The turbulence intensity might be a more relevant factor than the integral scale for
wind-loading simulation on low-rise structures; the lateral turbulence seemed to be
equally, if not more, important than the longitudinal turbulence. Usually only the
longitudinal turbulence was reported, but the longitudinal turbulence and the
lateral turbulence are likely to be correlated, due to the inherent three-
24
dimensionality of the turbulence. More realistic simulation of wind~directional
3.1 Introduction
An overall picture of the wind loading consists of three fundamental elements: the
climate., the near-surface wind., and the structure. Wind is generated at a considerable
distance above the earth's surface, in the first instance., by weather systems, which are
themselves initiated by pressure gradients resulting from differential heating of the
atmosphere. As it is with the flow of any fluid over a surface, a boundary layer, called the
atmospheric boundary layer (ABL), is formed over the earth's surface. All ground-based
structures are immersed in the ABL.
Structures are usually built to stand various environmental influences. The
performance of the existing buildings and other structures in severe winds is the only
realistic indicator of the success or failure of our design and construction practice against
wind loads. Buildings come in all sizes and shapes, and they are complex entities whose
performance would fare anywhere between a complete success and a total failure in the
event of strong winds. In general, the wind-loading effect on a building is governed by a
few factors: the wind characteristics, the terrain or topographical features, the building
geometry and its orientation with respect to the wind direction. Different parts of a
building envelope receive varying amount of wind loading, and the existence of adjacent
buildings might affect a building unfavorably. The wind damage a particular building
might sustain is the outcome of a competition between the loading effect and the
resistance capacity of the building components.
25
26
transparent to the solar radiation incident on the earth; but the earth is capable of
absorbing the solar energy. The heated surface emits energy in the fonn of terrestrial
radiation to warm up the air close to the ground. Differential wann-up of the air over the
water bodies and the landscapes gives rise to an uneven spatial temperature distribution
and creates pressure systems initiating the large-scale air motions. The earth~ s surface
poses a boundary for the atmosphere to draw momentum from and to move over.
The retarding frictional effect exerted by the surface of the earth on the moving air
slows down the air motion near the surface; this effect is diffused by the shear forces
(Reynolds stresses) and the turbulent mixing throughout the ABL. The depth of the ABL
ranges from a few hundred meters to several kilometers in the case of neutrally stratified
flows~ depending on the wind intensity, the terrain roughness and the angle of latitude.
Within the boundary layer, the wind speed increases with the elevation; it reaches the
gradient speed at the top of the boundary layer. Beyond the boundary layer, in the free
atmosphere where the surface effect has diminished, the wind flows approximately at the
gradient speed along the isobars (Simiu and Scanlan, 1996). The large-scale air motions
are significantly modified by an apparent force termed the Coriolis force caused by the
rotation of the earth. It is within the lowest portion of the ABL that most of human
being'"s engineering activities take place. A sketch of the ABL in the suburban areas is
shown in Figure 3. I.
Characteristics of the ABL are closely related to the terrain features over which the
air moves. The underlying terrain roughness governs the extension and profile of the near-
surface portion of the atmosphere. The terminology exposure category is used in ASCE 7-
95 to classify the roughness characteristics. In the classification, account is taken of the
variation in the surface roughness that arises from the natural topography and vegetation
as well as from the constructed features. The exposure in which a specific building or
other structure is sited is assessed as category A, B, CorD, with corresponding gradient
height of 1500, 1200, 900 or 700 ft, respectively, representing the urban terrain, the
suburban terrain, the flat open terrain and the open waters. In addition, an exponent is
assigned to each category to facilitate a convenient power-law expression of the mean
wind-speed profile of the ABL.
7
6
.~
en Structure-generated j
c 5
-8 turbulence
g
8.
II)
4
3
12 hours 1 hour 1 min t I<E-<___,\.____--310>.
b
~ 2
cf
l
0
0.001 0.01 0.1 l 10 100 1000
Frequency, cycle/hour
10.0
.!!
E
5.0
run mean= 7.58 m/s
0.0
0 10 20 30 40 50 Seconds 60
60.0
Wind direction series M49n072
~------~~~~==~====~~~~------------~
40.0
tn
1 20.0
g 0.0 ~....1'-1.....---v-.::.w-+-t-~~~~~~~~'Vl+...___-+-=1-___---'I:..L-~__,.,,......
20.0
-40.0 "------------------------~
0 10 20 30 40 50seconds60
Figure 3.3 Time series of wind speed and direction (WERFL, T1U)
In this section.. some damage patterns of buildings are first established, followed by a
description of the wind-building interactive process; finally, the issue of engineering
attention and performance improvement is discussed. The materials covered in this section
closely follow the works of Minor (1979, 1984) and Mehta (1984).
Figure 3.6 Local airflow and loading effects (after Minor., 1979)
34
With flat or gently-sloped roofs" the normal-to-ridge wind generates high suctions at
the windward eaves and the ridge. Oblique wind direction usually produces worst-case
uplifting forces at the leading roof comer. Parallel-to-ridge wind causes significant loads
on the roof verge and the gable wall. For most wind directions., generally two of the four
wall-comers are under severe loading effects (see Figure 3.6).
Other than the external pressures acting on the outside surfaces of a building
enclosure, the internal pressure is in most situations different from the static pressure, due
to air penneability present by either design or component failure. The internal pressure
may aggravate the loading effects on the building enclosure. Windward wall openings
usually cause an increase in the pressure within the building. This pressure increase
combines with the outward-acting pressures already acting on the roof, the leeward wall..
and the sidewalls to intensify the forces on these components. This is demonstrated in
Figure 3.7.
Conversely. openings in the sidewalls or the leeward wall cause a decrease in the
internal pressure. Such a pressure drop collaborates with the inward-acting pressures on
the windward wall However.. the decrease in pressure partly counteracts the outward-
35
acting forces on the external surfaces of the roof.. and on the side and leeward walls to
reduce the effects of pulling apart the building envelope (Minor7 1981 ). Local component
failure could lead to a total disaster when the building loses its integrity (see Figure 3. 7).
The presence of adjacent buildings of comparable height might magnify the suctions
on the side walls7 but pressures on the windward wally on the roof and leeward wall are
basically not changed by the proximity of other buildings (Sachs .. 1978). Building damage
by the impact of flying debris in severe storms is also an issue of serious concern.
4.1 Introduction
It is perhaps no overstatement that wind engineering at Texas Tech University staned
with the 1970 Lubbock Tornado. In the decade after the tornado, about 50 extreme
windstorms in the United States were documented by the researchers from Texas Tech.
During the same period, other investigators representing the civil defense interests, the
architecture and engineering protessions, the nuclear industry, and the academia were also
involved in the conduct of technical post-storm damage surveys (Minor, 1979). The
decade of 1980, prior to the establishment of the Wind Engineering Research Field
Laboratory (WERFL) at Texas Tech University, saw extensive research in the area of
wind-induced loads on low-rise buildings, primarily using the wind-tunnel testing
technique. The resul[S of the research have had significant impact on wind-loading
codification. Nevertheless.. problems remained, and there was a need to fully validate the
model-scale results for low-rise buildings. A few full-scale experiments and some
comparative studies revealed problems and difficulties. Both the wind-tunnel testing
technique and the data quality of the full-scale measurements were matters of concern
(Surry, 1979)
The WERFL located on the campus of Texas Tech University in Lubbock, Texas is a
unique facility for full-scale studies of wind loading effects on low-rise buildings. Since its
initial operation in 1988, the WERFL has represented one of the best-instrumented full-
scale installations to provide data of the highest quality for comparison with and
verification of model-scale experiments. Located in Lubbock, a city on the West Texas
High Plains, the site of the WERFL experiences persistent wind throughout the year. The
prevailing winds are from the south and southwest. Sustained winds of 20-35 mph are
frequent occurrences in the windy seasons of spring and late fall. Figure 4.1 shows trees in
front of the Texas Tech Museum (about one mile east of the field site) shaped mainly by
40
41
unrelenting winds in the springs. The generally flat and open landscape presents an ideal
terrain characteristics for the wind-tunnel simulatio~ by minimizing the number of
uncertainties to make the simulation of wind flow a less daunting task. All these factors
make the site an ideal location for a full-scale wind engineering research facility.
The test building is a prefabricated rectangular metal building with a nearly flat
duopitch roof. It is constructed on a rigid undercarriage that rides on a circular steel track
embedded in a concrete slab. When jacked up and supported by the four wheels installed
at the comers of the undercarriage, the building can be rotated to the desired orientation
with respect to approaching wind direction. Temporary anchorage of the building is
provided by bolts embedded at 15 intervals in the concrete foundation slab. A schematic
indicating the dimensions of the test building is shown in Figure 4.3.
Also shown in Figure 4.3 are the definitions of the wind azimuth (a), the building
position (~) and the angle of attack (9). The azimuth angle is measured clockwise from the
true north to the direction of the wind. The building position angle is the angle measured
43
clockwise from the true north to the assigned building north., i.e. ., the outward pointing
nonnal of the wall with the door. The angle of attack is a relative wind direction with
respect to the building orientation; it is measured clockwise from the building north to the
wind direction, i.e. e =a - p, if a > p, or a=360 + a - p, if a < p.
(a)
_L 0.25'
Il2.83'
30.25'
(b)
a- wind azimuth
P- building position
9- angle of attack
Figure 4.3 Schematics of test building and definitions of position and angle
(a) Dimensions of test building, (b) Building position and wind angle
The meteorological tower is a guyed latticed metal tower. It is located about 150 ft
(49 m) to the west (280 azimuth) of the test building, so that it is upstream of the test
44
building at the prevailing wind directions from the south and southwest. The anemometers
are installed on the tower in such a way as to minimize the interference of the wind flow
by the tower.. Instrumentation is selected at six levels on the tower: 3, 8, 13, 33, 70 and
160ft (l, 2.5, 4, 10., 21., and 49 m) above the ground.
4.3.1 Anemometers
The anemometer placement on the meteorological tower for Mode-49, a data
acquisition mode incorporating a 3-component sonic anemometer created specifically for
this study, is as follows. One 3-cup anemometer is mounted at 3 ft; five UVW
anemometers of propeller type are installed at 8, 13, 33, 70, and 160ft levels. Two 3-cup
anemometers are mounted, one at 13 ft and the other at 19 ft, on two pules located
between the test building and the meteorological tower to provide additional wind speed
measurements.
Sensors .
I / Multiplexor boards
AID
5.1 Introduction
The wind-tunnel simulation technique has long been the main means of determining
wind-loading effects on structures. However, researchers were always aware of the
limitations of model test results for prototype applications. After full-scale wind pressure
data became available from field measurements conducted in laboratories like the
WERFL at Texas Tech University, comparative studies ensued, and discrepancies
between the wind-tunnel predictions and the full-scale data were soon noted. There was
quite a debate on the simulation priorities., and this is the setting of the current study.
In the course of time, exploration of flow mechanisms associated with the peak
pressures took place. Trying to explain the causes of high suctions related to the
separation bubble (SB) flow situation, Melbourne (1977) hypothesized that the high
suctions were due to an instability of the bubble triggered by the incident turbulence.
Research by Letchford (1995), Kawaii and Nishimura {1996), Marwood (1996) and
Banks et al. (1997) on the conical vortices is representative of the renewed effort. Their
works went one step further by looking into the details of the incident wind, the near-
surface wind flow and the pressure on scaled models tested in the wind tunneL
48
49
Wind measured in the far field" on the meteorological tower might be too dissimilar to
the near-field' wind to permit synchronization with the pressures.
Figure 5.2 Set-up of sonic anemometer for wind flow measurement on roof
51
5.4 Flow Visualization
As an important element in the experimental research, flow visualization has played
a broad role in improving our physical understanding of complicated flow phenomena.
Effort was made in this study to seek feasible means to visualize the wind-flows over the
roof of the WERFL test building. Three techniques were explored: tuft-grid, smoke
injection and airfoil-grid, each with its own advantages and shortcomings. The tuft-grid
technique was used for most of the flow visualization. These methods were inexpensive,
suitable for full-scale applications and yet effective to provide insightful information
about the wind flow of interest.
The tuft-grid method has been used for flow visualization in wind tunnel tests. A 6.4
m (L) x 2.1 m (H) (21 ft x 7 ft) metallic frame with a square-grid system of 15.2 em (6
in.) was built. Bright-orange colored yarn segments were tied to the grids. The yarns
align with the local wind flow to reveal the location, the overall shape and the dimensions
of the SB or the conical vortices. The grid consists of three detachable panels, each has
dimensions of 7 ft x 7 ft. Figure 5.3 shows the set-up of a grid of 14 ft x 7 ft on the roof
of the test building.
In the current study, a direct time series comparison of the incident wind velocities is
sought. The velocity measurements were simultaneous, with closely spaced anemometers
of different types. Specifically, the comparisons are between measurements by the three-
component sonic anemometer and the UVW anemometer at 13ft leveL The experimental
set-up is shown in Figure 5.4. The two anemometers are about 6.5 ft apan in the east-
west direction.
CHAPTER6
SEPARATION BUBBLE AND ITS
EFFECfS ON PRESSURE
6.1 Introduction
Field surveys of building damage inflicted by severe winds indicate that roof eaves
are one of the most vulnerable portions of the cladding. When wind approaches a
building normal to one of its walls~ the obstructed flow is forced to either go over the
building, or to descend along the wall and detour around the wall comers. A conceptual
model of the flow around a cubic obstacle was given by Woo et al. (1977). The up-wash
flow cannot negotiate the sharp eaves to remain attached because of the viscosity of the
air; it separates to form a shear-layer, which may reattach to the roof surface to generate a
vortex called the separation bubble (SB). The SB and the pressures induced on the
underlying roof surface are worth investigating, from the viewpoint of both academic and
practical imponance.
The WERFL at Texas Tech University has documented peak-pressure coefficients
up to -10 at a pressure tap (50123) located one foot from the leading roof edge. The
strongest suction ever documented for the building cladding was from tap 50501, located
on one of the four roof-comers. Tap 50501 recorded peak-pressure coefficients up to -13.
However, worst suctions for the eaves and the roof corners do not occur at the same or
even close wind angles. The SB is not a special case of the conical vortex.
In this chapter, a time-averaged structure of the SB is fust established from flow
measurements. Mean pressures developed on the roof underneath the SB are then
examined in relationship to the flow structure of the SB. A concept termed non-
conventional pressure coefficient is introduced to make possible the separation of the
effect of direction fluctuation from that of the speed variation on the pressure-generating
mechanism. Spectral analysis is performed to explore the characteristics of the pressures
associated with the SB.
55
56
6.2 Structure of Separation Bubble
Systematic measurement of wind flow velocity within the SB was conducted using
the 3-component sonic anemometer. Based on the measurement~ the mean structure of the
SB is established in the form of normalized velocity vectors.
-
\C)
58
Table 6.1 Statistics of velocity vectors
h=0'-06'' h=1 '-04''
Run# X ex s Run X ex s
9 0'-05" 139.8 - 145 0' 146.2 1.078
49 1' 166.5 0.087 146 1' 155 1.204
48 2' -5.8 0.146 147 2' 164.5 0.495
47 3' -4.5 0.187 148 3' 167.1 0.258
46 4' -1.4 0.200 150 4' 171.9 0.180
45 5' -0.5 0.210 151 5' 179.4 0.116
44 6' -5.9 0.143 152 6' -172.7 0.091
43 7' -6.6 0.119 153 7' -165 0.138
38 8' -8.4 0.099 154 8' -161.6 0.136
37 9' -63 0.030 155 9" -164.45 0.180
36 10" -89.5 0.018 156 10' -166.05 0.227
35 11' -153.3 0.038 197 11' -168.6 0.275
31 12' -171.6 0.108 199 12' -172.0 0.301
30 13' -174.8 0.148 200 13' -171.6 0.267
29 14' -176.8 0.203 201 14' -173.8 0.325
26 15' -175.9 0.214 202 15' -175.6 0.340
h=1 '-00" h=1 '-08,..
13 0'-05" 142.4 1.182 50 0' 144 1.057
14 1' 155.8 0.787 51 1 150.5 1.192
15 2' 154.7 0.125 54 2 157.6 0.902
16 3' -31.3 0.063 55 3 165.9 0.551
18 4' -11.5 0.052 58 4 169.3 0.330
19 5' -56.5 0.018 59 5 176.3 0.295
20 6' -96.7 0.029 60 6 -179.2 0.262
22 7' -96.6 0.030 63 7 -173.8 0.276
23 8' -118.9 0.045 64 8 -172.7 0.279
24 9' -143.8 0.077 65 9 -172.4 0.282
25 10' -158.2 0.119 66 10 -172.4 0.298
204 11, -162.8 0.110 67 11 -174.1 0.325
205 12' -168.5 0.155 68 12 -175.6 0.366
206 13' -170.0 0.183 69 13 -176.3 0.396
207 14' -175.2 0.277 70 14 -177.6 0.408
208 15' -175.4 0.258 71 15 -178.6 0.436
59
Table 6.1 Continued
h =2\t-03" h = 3'-00"
Run# X. a s Run X a s
72 0 152.4 1.139 105 0 158.6 0.962
73 0'-06" 151.42 1.234 106 0'-06" 160 1.080
74 1' 154.04 1.312 107 1' 160.6 1.080
81 2' 157.94 1.078 108 2' 162.4 1.127
82 3' 164.63 1.113 109 3' 165.9 1.052
84 4' 169.93 0.816 112 4' 171.8 1.204
88 5' 173.66 0.672 113 5' 175.3 1.094
92 6' 177.31 0.571 115 6' 178.6 1.017
94 7' -178.23 0.601 116 7' -178.8 0.916
95 8' -175.7 0.488 117 8' -176.1 0.833
97 9' -174.96 0.508 119 9' -174.7 0.769
98 10' -175.38 0.606 120 10' -172.6 0.741
99 11' -175.25 0.440 121 11' -172.5 0.633
100 12' -175.51 0.546 122 12' -171.9 0.645
101 13' -175.28 0.542 126 13' -173.7 0.703
103 14' -174.77 0.573 127 14' -172.9 0.712
104 15' -177.03 0.486 128 15' -174.1 0.709
h = 4'-02"
213 0' 163.63 1.038
215 1' 164.34 1.116
216 3' 167.87 1.074
217 5' 172.49 1.128
218 7' 176.71 1.159
219 9' 179.56 1.059
The velocity magnitudes (s) in Table 6.1 are normalized by the corresponding mean
wind speeds, measured at roof-height level on the meteorological tower. For plotting,
sizes of the velocity vectors are scaled by the vector with the largest magnitude, which is
assigned a convenient size.
A few observations can be made about the time-averaged SB:
1. The vortex is oblong and elongated in the horizontal dimension;
2. The mean reattachment point (RP), as identified by a nearly downward-pointing
velocity vector, is approximately 10 ft from the leading edge; the mean height of
the SB is about 4.5 ft.
60
3. The region with predominant flow reversal is confmed close to the roof surface. In
this region, time series records indicate frequent flow reversal and low mean
speed;
4. In the immediate proximity of the roof, there exist four distinct zones of wind
flow. They are the leading edge zone (ZONE-0, where the separated shear layer
meets the reverse flow to create a wedged region, the flow reversal zone (ZONE-
m, the reattachment zone (ZONE-Ill) and the forward flow zone (ZONE-IV);
5. Despite the conceivably random and turbulent nature of the flow inside the SB,
the averaged structure of the SB is amazingly organized and regular.
Table 6.3 Coord"mates and mean C;p of seIected taps along the short roo f-axtS
.
location
tap X (ft) y (ft) Cp (mean) Remarks
50123 1.00 23.17 -1.18 Mode-15
50223 1.67 23.17 -1.05 Mode-49
50523 4.67 23.17 -0.73
50923 8.67 23.17 -0.72 reattachment
51423 14.18 23.17 -0.43 occurs at l 0 ft
52323 22.58 23.17 -0.25 (3m)
52923 29.25 23.17 -0.21
64
Cp= -1.18
~ -1.05
i i""-..~: 0.73 -0.72
.'
II
I
I
I
I
I
I
I
I
.......- - - -
'I
I
l
l
-0.25 -021
I l I
I I I
Since the mean reattachment point (RP) is at about 10 ft (3 m) from the leading
edge, taps 50123, 50223., 50523 and 50923 are located upstream of the reattachment
point; taps 51423, 52323 and 52923 are downstream of the reattachment point.
A closer look at the velocity vectors near the leading edge of the roof reveals that
taps 50123 and 50223 are in the wedge-shaped region where the separated shear layer
meets the reverse flow (Figure 6.4). These two taps experience the strongest suction, this
could possibly be explained by a vacuum-like effect: the air in that wedge-shaped parcel
is drafted upward to create a zone of extremely low pressure (high suction).
Taps 50523 and 50923 are located within the zone of flow reversal and their mean
pressures differ only slightly. Taps 52323 and 52923 are in the forward flow zone and the
difference in the mean pressures is small. Tap 51423 assumes a mean pressure coefficient
falling sontewhere in between the values for the two zones before and after.
65
Wind---..
wedge-shaped region
100
ISO
Eo
-50
Seconds
-100
840 650 660 670 680 690 700 710 720
0.0 r---------__,.....,..-----.------------.
-0.5
-t.O 1----------~-:.------.........::;r.-~~~~~~:i-H
a..-1.5
o.2.o run mean= -1.04
2.5
-3.0 run minimum= -4.79
-3.5 ~...- _ __.__ __.__ __,___---'_ ___,__ _ _.___--..~.__ _ ___,
640 650 660 670 680 690 700 710 720
2
Seconds !
0 ~--~----~----~--~----~----~-~----~
640 650 660 670 680 690 700 710 720
IS
..r -2
.............................._........ --
0+----+----+----+----+----+----~---r----+----r--~
.......... -.-.,........... .
c
~ -4
.!~ .......~~~-'-.!.!~--.-----
..6
II
i -8 i
l -10~----------------------------------------------~
J
~ ~ ~ ~ ~ m ~ ~ ~ ~ ~
wind Wigle
Figure 6.7 Pressure coefficient versus wind angle (tap 50223., Mode-49 data)
Loading ( Cp)
Some cause-effect relationship between the incident wind and the pressure
coefficients is of interest. It is assumed that the direction fluctuations (not the direction
itself) are the most important single input; other contributing factors such as the wind
profile are put aside to simplify the problem. It is from this basic assumption that the non-
conventional pressure coefficient was introduced.
Figure 6.9 displays four synchronized segments of 80-second duration for the
incident wind direction, the conventional pressure coefficient., the non-conventional
pressure coefficient and the wind speed. A close comparison of the direction fluctuations
(Figure 6.9 (a)) with the trace of the non-conventional pressure coefficient (Figure 6.9
(c)) reveals that the primary peaks of Cp (t) coincide with large-and-fast fluctuations of
the wind direction. It is difficult, though, at this stage to quantify the phenomenon.
Qualitatively, the rate of a direction fluctuation, the magnitude of the fluctuation, the
immediate history of wind direction tluctuation., including the amount of deviation of the
direction from the windward normal, are among those factors affecting the Cp (t).
The SB is usually considered as a two-dimensional phenomenon; fluctuations of the
incident wind direction disturb a certain delicate balance within the SB. It takes some
time for the balance to establish itself. A fast-and-large wind direction fluctuation about
270 preceding a more or less constant wind direction seems to break this balance and
initiate low pressures (high suctions). This is not the case with the conical vortices, where
the vortices are three-dimensional and stable; the peak pressures are results of combined
gusts and incident wind directions (not the fluctuations) conducive to vortex formation.
73
Wind c:iraction series. M49n072
~~----------------------------------------~
60
40
.20
I o~~~--~~~yq~-w~H+~~~~~~~~
l-20
-40
-80
~ ~----------------------------------------~
560 570 580 590 600 610 640
-3.0
560 570 580 590 600 610 630 640
10.0
.~
8.0 .' '>~,_ I
Observations indicate that the primary peaks of the conventional pressure coefficient in
the SB were often an outcome of combined fast-and-large fluctuations of incident wind
direction and wind gust, with the former playing a more important role (Figure 6.9 (b)
and (c)). Slow (but not necessarily small) fluctuations of the incident wind direction only
induces weak suctions ( Cp (t) of small magnitudes), see segments from 564 second - 580
second in Figures 6.9 (a) ;ud 6.~(c).
It is not clear at this point what was happening in terms of the wind flow inside the
SB at instants of high suction. Simultaneous, multiple-point velocity measurements are
desirable for an affumative answer. Letchford (1995) claimed that some shear-layer
instability, which was strongly influenced by the incident turbulence, was an important
parameter in this respect. He postulated that instability caused by incident turbulence
forces the separated shear layer to prematurely reattach onto the roof, creating peak
suctions.
Effort was made in this study to establish certain flow patterns associated with high
suction, using the conditional sampling technique adopted by Marwood ( 1996) for the
conical vortex situation. No simple and clear pattern emerged, except the observation that
flow inside the SB at moments of high suction was quite three-dimensional Thus the
flow defies a straightforward two-dimensional characterization. Correlation analysis
between the incident wind and the pressure at tap 50223 failed to show good correlation.
More work is needed for quantification of the wind-pressure relationship. However, the
current study suggests an answer to the question posed by Letchford (1995) as to the
cause of shear-layer instability, i.e., fast-and-large directional fluctuation in the incident
wind might be the initiator of the shear-layer instability.
0.12
~0.08
0.04
0.00
0.001 0.01 0.1 f, Hz 1
Figure 6.10 Spectra of pressure measured at six taps along short roof-axis
non-Gaussian
Wind Gaussian
z
Figure 6.11 Pressure zones on flat roofs: Gaussian and non-Gaussian
(after Kumar and Stathopoulos, 1997)
77
Typical time series of Gaussian and non-Gaussian pressure distributions are
displayed in Figures 6.12 and 6.13, respectively. Wind pressure spectra, representing
energy content in the pressure fluctuation, receive contributions not only from the
mechanical turbulence but also from the building generated turbulence., both varying with
the building configuration, the exposure condition and architectural details (Kumar and
Stathopoulos, 1997).
Recalling the structure of the SB discussed in Section 6.2, the non-Gaussian pressure
zone as represented by taps 50223, 50523, 50923 and 51423 could be related to the wind
flow in the SB. These taps are underneath the overall flow circulation of the SB, and the
pressure fluctuation is non-Gaussian. Of course, the pressure becomes progressively less
non-Gaussian as the location moves away from the leading roof-edge.
6.7 Summary
The main points discussed in this chapter can be summarized as follows.
l. The overall SB is oblong and elongated in the upstream wind direction. Four flow
zones near the roof surface can be identified: (a) the leading edge zone, where the
separated shear-layer meets the reverse flow to create a wedge-shaped region; (b)
the reverse tlow zone; {c) the reattachment zone; and {d) the forwarding flow
zone;
2. The mean point of reattachment (PR) is approximately 10 ft from the leading roof
edge; the mean height of the SB is about 4.5 ft;
3. The pressure distribution on the roof surface underneath the SB is directly related
to the structure of the SB. The uplifting pressure on the roof generally decreases
with distance from the leading edge;
4. The introduction of the non-conventional pressure coefficient made it possible to
separate the effect of wind speed from that of the wind direction on pressure
generation;
5. Time series analysis indicates that fast-and-large fluctuation of the incident wind
direction governs the mechanism of peak-pressure generation;
78
6. The primary peaks of conventional pressure coefficient are often an outcome of
combined wind direction fluctuation and wind gust.
7. The non-Gaussian pressure zone is found to extend at least 15ft from the leading
edge~ and thls dimension is comparable to the height of the test building.
0.0
-o.2
c..
0
-o.4
-o.6
seconds
-o.a
0.00 60.00 120.00 180.00
-1.5
c..
0
-2.0
-2.5
-3.0
seconds
-3.5
0.00 60.00 120.00 180.00 240.00 300.00 360.00
7.1 Introduction
Post-disaster surveys have repeatedly revealed the wlnerability of low-rise building
roofs to severe wind. It is now well-established knowledge that, at oblique wind, the
leading roof comers of rectangular low-rise buildings sustain outward acting pressures
higher than any other part of the building envelope. Yet, it is precisely with the roof-
corner region that our efforts at load prediction were the least successful. Despite the
progress made over recent years, simulation techniques and results by different
researchers were not always consistent; model-scale predictions of wind loading in
simulated wind flow still have room for improvement. Among others, the works of
Tieleman et al (1993), Kawaii and Nishimura (1996), Marwood (1996) and Banks et aL
(1997) are representative of an effort to understand the conical vortex phenomenon over
the roof and the pressure induced on the underlying roof surface.
Full-scale studies of the incident wind and its relationships with the conical vortex
and the pressure could uniquely contribute to a better understanding of the wind
simulation requirements, and eventually provide guidance for improving wind-loading
simulation.
This chapter intends to discuss the fundamentals of the conical vortices above a flat-
roof corner and their relationship with the surface pressures.
79
80
The assumption of minimum pressure occurring below the axis of a vortex core is
partly supported by the potential flow theory. In a free vortex, the theory states., the
streamlines are in the form of a nest of concentric circles, and the tangential velocity is
inversely proportional to the distance from the vortex center. The circulation along any
circumferential path in the vortex is independent of its radius. The pressure distribution
on an underlying surface can be estimated from the velocity field using
in which pis the surface pressure at a point, p min is the peak suction, y is the distance of
a point on the surface from the vortex axis, parallel to the roof surface, zh = 1.55 Yuz., is
the height of the vortex core above the surface, y 1n. is the half-width of the vortex, ie.
the value of y at p/p min= 0.5 (Figure 7.1). More detailed derivation can be found in
Marwood (1992). The above formula is only approximate, given the fact that the vortices
above a roof corner are usually oblong (elongated in the horizontal direction) and far
from the ideal circular free-point-vortex. The pressure distribution on an underlying
surface is shown in Figure 7.l.
P min
Figure 7.1. Pressure distribution under a free point-vortex (after Marwood, 1996)
81
In their work of simultaneous pressure measurement and flow visualization on
cuboidal models tested in wind tunneL Banks et aL ( 1997) found that the pressure
distribution under roof-comer vortices was asymmetric with respect to the vortex axis.
Pressures for locations having angular position (from the roof edge) greater than the
vortex axis exceeded the symmetric curve-fit values. However, their fmdings indicated
that the angular position of highest (mean and peak) suctions did follow the vortex core
position closely.
(y)
---------------~
0 0 0
50209 50509 50909
50205 50505 50907
0 0 0
Some statistics of the pressure coefficients for these nine taps are listed in Table 7 .2.
A general trend of the pressure distribution can be identified: as the location moves away
from the two roof edges and from the apex, the intensity of suction drops rapidly. ln other
words., only a small portion of the upwind roof comer sustains severe wind loading. It is
of vital importance, however, that this portion be adequately designed and constructed
against intense local uplifting force to prevent initial damage to the building envelope.
Mode-15 data of pressure coefficient versus the wind angle are reproduced here for
taps 5010 I and 50501 to give a general idea about the characteristics of pressure on flat
roof corners of low-rise buildings. Similar graphs for the other seven taps are included in
the Appendices. Synchronized traces of pressure coefficient and incident wind are also
provided to emphasize the points to be made.
Tap 50101 is either beneath the vortex generated at the left roof-edge with incident
wind angles of around 260~ or under the vortex formed at the right roof-edge with wind
angles around 190 (Figure 7.4). At around 225~ two small vortices are generated but
neither of them is big enough to reach tap 50101, and straight flow passes over the tap.
Figure 7.5 gives a schematic view of the phenomenon. Synchronized traces of the
pressure coefficient and the incident wind direction in Figures 7.6 and 7.7 support this
claim.
Pressure coefficient traces of tap 50101 are characterized by small fluctuations
around a nearly constant value of about - 0.5, which is sporadically interrupted by brief
moments of large fluctuation (Figure 7.6(a)). The same figure shows a generally flat
pressure coefficient, except a brief moment of strong suction lasting about 5 seconds
(290s - 295s). This eventful 5 seconds for the pressure corresponds to incident wind
angles close to 190 (Figure 7.6(b)), Le., the angles at which the vortex formed from the
flow separation at the right edge is big enough to encompass tap 50101. A close look at
the fluctuation of the wind speed within these 5 seconds reveals that the pressure
fluctuation closely follows the wind speed variation.
84
AOAva.U.an
o.s
0--~~-+~~~~~---+------+------r------+-----~----~
~r----,~~-r~~~~~j---~~r-t--=~~---t
c -1~-----+------+-----~~
i -15~-----+------+------+-----
2~-----+------+------+------+------+------+--.---+----~
-25~-----+------+------+------+------+------+-r----+----~
~~----~------~----_.------~-------~---------~-------~-------~
0 90 135 180 270 315
AOA
AOA.va. RMS
1.4
12+------+------+------+------~--~~--~~1-----~-------
~ OB+------+------+-----~-----,~~~~----~~~--~------
:1
~ OB+------+------+-----~--~~--~~~~~--~~
~4+-----~------r---~~
02~~--~~~~~~~~
0+---~-+------+-----~------~----~------~----~------~
0 90 135 180 225 270 315
AOA
A.OAva. Min
10+------+------+-----~------~~--~K-~~~~--~~----~
c
2
15~-----+------+-----~--------~----~------~--------~----~
25
A.OAva. Max
i ~~i'l~t.#~~f\4~
0 90 135 180
AOA
225 270 315
Figure 7.3 Tap 50101 pressure coefficients versus wind angle (Mode-15)
85
Figure 7.4 Schematic view of vortices as they affect pressure tap 50101
(Roof corner)
--tap50101
t
225
Figure 7.5 Straight flow and vortices formed at incident wind around 225
A similar argument accompanies Figure 7.7, except here., instead of one, there are
several short moments of high suction, and the pressure tap was under the influence of
the vortex from flow separation at the left roof-edge, when the wind angles were close to
260. However, a pressure-coefficient peak did occur at 27-second, at wind angles not so
much off225, due to higher-than-the-mean wind speed at the moment.
At this point., it is worthwhile to go back and look at the wind-angle/pressure-
coefficient relationship shown in Figure 7.3. The distribution of the minimum pressure
coefficient has significant scatter and no distinct peak exits, unlike the distribution of the
86
mean pressure coefficient. This is presumably caused by what can be called the "direction
smear'' effect't ie., the minimum pressure coefficient is an individual event, it took on its
values at wind angles that were probably much off the run-averaged wind direction, a
convenient wind direction commonly used for documentation. Therefore, if the
instantaneous wind angle at the instant of pressure peak occurring had been used for
plotting, it might have been a very different picture: with much-reduced scatter and
possibly two distinct peaks, like the case for the mean pressure coefficient. The scatter
with the rms distribution was likely caused by the non-stationary nature of the
atmospheric wind.
To end this subsection, a brief summary is helpful:
1. The pressure coefficient traces at tap 50101 were characterized by small
fluctuations around a nearly constant value of about -0.5, which was sporadically
interrupted by brief moments of large fluctuations.
2. The pattern of pressure fluctuation observed at tap 50101 was presumably
associated with the straight flow passing over the tap, and occasional large
vortices formed from flow separation at one of the roof edges.
87
Cp
4.0 ~----------------------------~~+-----------~
~.0 ~----------------------------~~+-----------~
seconds
4.0 ~--~--~------~--~--~--------~--~--~--~
255 265 275 285 295 305
240
230
~220
Q)
~10
"'0
200
190
seconds
180
255 265 275 285 295 305
2.0
1.0
0.0
~ -1.0
-2.0
-3.0
-4.0
0 10 20 30 40 50 60
AOAva.Man
0.5
-0.5
0
-- -.... -~-..-. ..... ~-
~. ....... 'f'
I;~ ,...4...:" .
c 1
:1 1.5
2
~~
-2.5
-3
_.._
"--:
-3.5 ~-
AOAvs. RMS
1.8
1.6 ...
1.4
4~ I
~r
....
.... ...
.... ~ ......
....
,..
I
I
0 J -~
0 45 90 135 180 225 270 315 360
A.OA.
AOAvs. Min
o~w.~~~~~----~----~----r-----r-----r---~
-2~----~~~~~~~~~---+----~~--~~-
-4+----+---+---
c -6+------+------+------+~~~
~ -8~----~-----+------~----~~--~~---4------~----~
-10t-------+------t-------+------4~~~~L-~~+------+---------
12+------+------+------+------+---~-+------+------+---------
-14
0 45 90 135 180 225 270 315
AOA
1.5
= 05 lii~Sil~dt!~+~~~d~--~
~ o~~~~---~--.-~~
~+-----~----~------+-----~
-1
0 45 90 100 270 315
NJA
Figure 7.8 Tap 50501 pressure coefficients versus wind angle (Mode-15)
91
In the same figure, the wind gust at 404 second induces only a weak suction due to
wind angles not favoring vortex formation. The wind at 400 second seems to be in the
right angle range but it fails to create a high suction because of much lower wind speed at
that instant.
t"" Wind
The above argument can be generalized to taps 50901, 50205 and 50209., except that
the range of wind angle for taps 50205 and 50209 should be approximate 240 - 250 out
of symmetry of pressure tap location. All these taps are close to the apex and also close to
the roof edges.
Similar to the case with tap 50101, the minimum pressure coefficient of tap 50501
has more scatter than that of the mean pressure coefficient. This can again be attributed to
the udirection smear," ie., the erroneous assignment of the angle-of-attack to some of the
data points. The scatter would have been much less if the actual instantaneous angle,
instead of the run-mean angle, had been used for plotting.
The wind-speed-dependent minimum-pressure coefficient (conventional) is not a
fundamental quantity. It takes on values that fall in a wide range, depending on the
incident wind. The conventional minimum pressure coefficient would reduce to the mean
pressure coefficient, if its wind-speed dependence could be removed by normalization
with respect to the instantaneous dynamic pressure. This is verified through a successful
construction of pressure coefficient series.
(a) Pressure Coefficient
e~ 10 ~------------~~~~-r~------+--------------
~ : t----IIFUU.-.:,.-:::----1-f/'-L--r:
6 ~--------~----------~----~----~----~----~
390 400 410 420 seconds 430
Figure 7.11 A close-up look of incident wind and pressure (M49n 161, tap 50501)
94
It seems appropriate at this point to establish the flow pattern of the vortices above a
flat roof comer at oblique wind angles:
1. A pair of vortices form over the roof comer at wind angles close to 225, but they
are relatively small;
2. Only one large vortex would form at a time from flow separation on either side,
depending on the incident wind angle, since the two ranges of wind angle
favoring the formation of large vortices at the two roof edges do not overlap;
3. Pressures on the underlying roof-corner reflect the flow field over the roof-corner.
In conclusion, strong vortical circulation over flat roof-corners and the associated
high suctions on the underlying surface are conditional on favoring incident wind angles
and wind gusts. This viewpoint will be strengthened in the following subsection, where
the pressure is successfully predicted from incident wind, and will be revisited later in the
context of flow visualization.
Traces of the variables used in the above computation are demonstrated in Figures
7.13 and 7.14. Figure 7.l3(a) is the incident wind direction series {<Xi), which has
considerable fluctuation with a range of about 87. Figure 7.13(b) shows the wind speed,
u(tJ, with a run-mean U = 10.33 rn/s. Figure 7.13(c) gives the instantaneous pressure
coefficient, Cp(a;), corresponding to a; at instant ti. The factor [(u(ti)/U)2 ] that accounts
for the wind-speed dependence of the instantaneous pressure coefficient (Cp(ti)) is
presented in Figure 7.13(d). Figure 7.14 compares the measured pressure coefficient with
the predicted one (Cp(ti)). The similarity is amazing, given the simplicity of the model
used.
WtM4
-o.5
-1.0
-1. ~
-2. 0
-2. ~
....0
0'\
(a) Measured pressure coefficient (M49n 158)
~ M49N158.\4
0.0
-2.0
--4. 0
-6.0
-8.0
In spite of the complexity of the actual situation and the simplicity of the model, the
general agreement of the pressure coefficient is satisfactory. Of course, while the wind
speed and the wind direction are the most important parameters, other factors might also
be important. For instance, the venical profile of wind speed is important in determining
how much airflow would go over the roof when the flow approaches the building.
99
Perceivably, the fundamental features of the profile are preserved when the wind flow
actually fluctuates, and it is probably this fact that renders it possible to predict the
pressure from the upstream wind measured at only one point. The fmdings made here are
important from the viewpoint of practical applications, since a point-measurement of the
approaching wind should not pose a problem in many practical situations, and the wind
tunnel technique can predict the mean pressure coefficients fairly accurately.
45
CD
30
@Y G) 0 90
,.
270
-(
13 ft (roof height)
15
@)
13 ft (roof height)
0
0 10 20 30
I 180
3.0 , - - - - - - - - - - - - -
x, z (feet)
offset\ ..... _ .........o .... .o
2.0
X =0.159y + 0.0353
.,....c.- ...
_... ....-ot-
"" x = -o.0064yz + o.2n4y - o.4841
0
1.0 _o..----~~
------o- 0
___ 0~
----o-
~ z = 0.0689y + 0.0937
I
a----
displacement
0.0 .___ _ ___.__.....___... _ _,____..._ _._____.__ ___._ _.._____...___...____.
5 10 15 y (feet) 20
/ Rightedge
It!' (short wall)
180
--
~ 210
t"
Left edge (long wall) j )"
225"
Figure 7.19 Schematics of vortices and incident wind angle
104
In general, vortices forming over the roof-corner take various sizes and forms, and
there are a variety of combinations (patterns) of the vortex pair, depending on the
incident wind angles. Here., the wind angles of concern are in the range of 180-270.
First., a single vortex is examined. The grid is set perpendicular to the long wall to
visualize the vortex developing from flow separation at the long roof edge. Presumably,
the vortex at the short roof edge is symmetrically similar. Figure 7.20 shows the vortex at
a wind angle of about 220, and Figure 7.21 at a wind angle of 210. The wind was
blowing from right to left in the figures.
A comparison of the two vortices indicates that at the larger wind angle (220,
Figure 7.20) the vortex is more oblong, and its core has a larger offset from the roof edge
and a smaller displacement from the roof surface. This is conceptually generalized and
depicted in Figure 7.22 based on extensive observations.
At a certain range of wind angle around 225 (roof comer diagonal), a pair of
vortices would form. To visualize the vortex pair, a grid of 14 ft long was orientated
nearly perpendicular to the diagonal of roof comer; the camera was set downstream of the
grid and the wind was blowing toward the viewer. This setup remains true for the pictures
presented later, unless otherwise specified. Figure 7.23 captures one of those instants.
Unfortunately., the wind velocity i:aata from the sonic anemometer were bad for that
data run, due to a wind speed overshoot. Wind gusts exceeding 30 m/s (60 mph) were
recorded on the meteorological tower at 13 ft, while the speed range for the sonic
anemometer was set at 30 rn/s. The wind angle was presumably about 230- 235 at that
particular instant of the picture, since the vortex on the right-hand side of Figure 7.23
(forming at the left roof edge, as defmed in Figure 7 .19) is slightly oblong shaped.
105
Core position
Roof comer
2<Xt
Figure 7.22 Effect of wind angle on vortex shape and core position
Figure 7.24 A big oblong vortex forming at the right edge (M49n220, 200~
108
Figure 7.25 A big oblong vortex forming at the left edge (M49n220, 250,
9.5 ft
9ft
1 50501
Right edge
(short wall)
0 'II _ J i . J II I ~ . I_ 1. : I a . : J. ,1_ I
.
{5-2 ...__....
fs:i3:ss 1 --------...
-4 ~~~----~------------------------+---~~~~--~--~
I
I
I
6 I I
I
t f t
-----------------
I
I
60 I II w
1\ \iW I "' II .. I ... , - I
~
e:
0>
Q) I - I I
0 30 ''' -- ._flfll '\.E r I II
seconds
0 --------~--------~-----+~--------~------~--------
360 370 380 390 400 410 420
Flow visualization supports the above presumption. In Figures 7.24 and 7.25, two
big oblong vortices were observed. Figure 7.26 shows that the vortex (at wind angle of
200, was large enough to encompass tap 5010 L Figure 7.27 gives the pressures
coinciding with the vortex. Referring to Figure 7.27, at 8:13:12, the wind angle was
around 30 (equivalent to 240.. an angle favoring vortex formation at the left edge), a
(non-conventional) pressure coefficient peak occurred at tap 50101. Similar phenomenon
happened at 8:13:58 with a wind angle of 70 (equivalent to 200, an angle favoring
vortex formation at the right edge), except it was with the right edge.
0.1
.L
N
<:1
0.01
~:1
:1
rn
L
0..001
0..0001
0.00001 0.000] o.cxn 0.01 0.1 10
rzii.J (cycles]
_ Wind Tunnel J :100 Rll _ Field Sonic _rteldUVW
_ Field 3-Cup _ Univcnal
fia frequency [Hz). z [4m] is roof height.
U and Su "'2 arc mcau and variance.. respec::tively. of longitudinal velocity at z.
-
..!...
N
(
0.1
~ 0.01
r::;
J
L
o.cxn
0.0001
0.00001 0.0001 0.001 0.01 0.1 1 10
r'z/U (cycles)
_ WindTunnell:lOORII _Field Sonic _FaeldUVW
_ Field ~-Cup - Universal
(is frequency [Hz], z [ 4mJ is roof beighL
U is mean longitudinal velocity at z, and Sv""" 2 is variance of lateral velocity at z.
In the experimental setup., the instrumental u-, v- and w-components of the UVW
anemometer and the sonic anemometer are consistent. Figures 7.30 compares the
instrumental u- v-components of the wind velocity. No attempt was made to fmd a mean
wind direction to decompose the velocity components accordingly to obtain the two
conventional velocity components in the horizontal plane. Comparisons of the w-
component of the wind velocity and the wind direction are shown in Figure 7 .31. The
most outstanding contrast is that high frequency fluctuations in all three velocity-
components and in the wind direction are missing from the UVW anemometer signals.
Peaks of short duration are left undetected by the UVW anemometer. The fact that the
two anemometers are set to sample at different rates, 10Hz for the UVW anemometer
and 30 Hz for the sonic anemometer, are not be responsible for the observed differences.
The fundamental cause is the retarded frequency response of the moving blades of the
UVW anemometer. Because of the inertia., the blades simply could not speed up or slow
down fast enough to keep up with rapid variations of wind speed. Occasionally.
overshoots of velocity peak would occur, due to abrupt changes of the wind direction.
Finally, it needs to be stressed that the differences observed represent the turbulence
part of natural wind fluctuations, and turbulence fluctuations are indeed important from
the viewpoint of wind loading effects on structures.
7.7 Summary
The major points discussed in this chapter can be summarized as follows.
1. In tenns of local flow-pressure relationship, a fundamental concept is that low
pressures (high suctions) on the roof comers are always related to strong
cornering vortices;
2. Wind angles favoring vortex-formation are bounded by 200 and 250;
3. Instantaneous pressure coefficients for tap 50501 could be satisfactorily predicted
from the mean pressure coefficient data and the incident wind properties. Pressure
prediction for tap 50101 was less successful, presumably because there are two
flow phenomena (straight flow and vortical flow) involved, and as a result, the
u-oornponent (sonic anemometer} v-component (sonic anemometer)
20 6
3
j 15
~..
ll:
0
gb 10 l:'
'Q -3
~ ~ -6
seconds
-9 ~~--~------~----~------~--~~
0 20 40 60 80 100 0 20 40 60 80 100
20 6
8... 15 II) 3
0 ~ 0
- 10 'Q -3 I
b -- - 'f ~.,.. I
. 5
II t '"
, ~onds I
11.)
> 0 seconds l :: I . ' ' ' ',
0 20 40 60 80 100 0 20 40 60 80 100
-
~
w-component (sonic anemometer) Wind direction (sonic anemometer)
6
30
20.. J Ill
.., I . j
t. lnu
i 3 a
tn
10
.e.' 0
g
!' r
l -3
-6
seconds
l :~~ I .: ' ."
?> o
-30
' ' ' JC<Ondi
I
0 20 40 60 80 100 0 20 40 60 80 100
ii~. 03 n IV'~
rI\ ~~ n.r .~l".1 uLr\u:vvu "ldl>-c .. ,. . n"A.
l.J, -6 I
1 ' ' ' , ' , , scoonds I -30
0 20 40 60 80 100 0 20 40 60 80 100
-
U\
116
This study set out to clarify the mechanisms of pressure generation for two important
flow phenomena significantly affecting the wind loading on low-rise building roofs: the
separation bubble and the conical vortices. Findings in this study have partially clarified the
mechanisms of pressure generation for these two flow situations.
In a retrospective view, much of the debate on wind-tunnel simulation priority has
failed to distinguish the effects incident wind has on two different types of quantities:
single events, e.g., minimum (peak) pressures, and statistics, such as the mean and the rms
pressure coefficients. The turbulence intensities, reflecting the gust structure and the
directional fluctuations of the free-stream wind, might bear significant influences
collectively on the pressure statistics. It is the intensity of individual longitudinal gusts that
decides the data-run-wise peak-pressure coefficients on the roof corner, while the peak-
pressure coefficients associated with the SB are governed by the lateral directional
fluctuations in the incident wind. Proper simulation of the incident wind profile (boundary-
layer type) is probably the most important single input in the wind- structure interaction
process.
The major contributions from this study to the understanding of the wind flow and
the related loading effects on low-rise buildings are summarized as follows.
117
118
3. The pressure distribution on the roof surface underneath the SB is directly related
to the structure of the SB, with the uplifting pressure on the roof generally
decreasing with distance from the leading edge.
4. The introduction of the non-conventional pressure coefficient makes it possible to
separate the effect of wind speed from that of the wind direction on pressure
generation.
5. Time series analysis indicates that fast-and-large fluctuations of the incident wind
direction govern the mechanism of peak-pressure generation.
6. The primary peaks of the conventional pressure coefficient are often outcomes of
combined wind direction fluctuations and wind gusts.
7. The non-Gaussian pressure zone is found to extend approximately 15 ft from the
leading edge, and this dimension is comparable to the overall longitudinal size of
vortical circulation and the height of the test building.
8. It is not very clear as to what was happening with the flow inside the SB at those
moments of pressure peaks, except that the flow was quite three-dimensional
Predictions underestimate the pressure peaks while over estimating the mean
pressure.
5. The form and dimension of a single vortex and the pattern of combination of the
vortex pair varied considerably with the incident wind angle.
A practical means of wind load prediction is established. It is a "hybrid" approach that
combines wind-tunnel simulated mean pressure coefficients with easily available
information of site-specific wind data.
ASCE., 1995: Minimum Design Loads for Buildings and Other Structures., ASCE7-95,
American Society of Civil Engineers.
Akins, R.E., Peterka, J.A., and Cermak, I.E., 1977; "Mean Force and Moment
Coefficients for Buildings in Turbulent Boundary Layers/' Journal of Wind
Engineering and Industrial Aerodynamics, Vol 2'1 pp195-209.
Atkins, R.E., 1992: "Effects of Turbulence on Mean Force and Momentum Coefficients,"
Journal ofWind Engineering and Industrial Aerodynamics, Vol 41, pp701-712.
Bachlin, W . ., Plate, E.J.'I and Kamarga., A.., 1983: "Influence of the Ratio of Building
Height to Boundary Layer Thickness and of the Approach Flow Velocity Profile
on the Roof Pressure Distribution of Cubical Buildings," Journal of Wind
Engineering and Industrial Aerodynamics, VoL 11, pp63-74.
Bailey, A., 1933: "Wind Pressures on Buildings.,n Inst. Civil Eng., selected papers, No.
139.
Bailey, A.., and Vincent., N.D.C., 1943: 'Wind Pressure Experiments at the Severn
Bridge," Journal of Institution of Civil Engineers, VoL 11'1 pp363-.
Banks, D., Meroney, R.N., Sarkar, P.P., and Zhao, Z.'l 1997: ''Visualization of Flow and
Separation about a Low Rise Building: Comer Vortex Zones," submitted to the
Journal of Wind Engineering and Industrial Aerodynamics.
Chiu, G.L.F., Perry'l D. C., and Chiu, A.N.L., 1993: "Structural Performance in Hurricane
Iniki.,, Proceedings of the 7th US National Conference on Wind Engineering,
Vol!, pp155-l61, University of California., Los Angeles.
Cochran, L. S., and Cermak, J.E .., 1992: uFull- and Model-scale Cladding Pressure on the
Texas Tech University Experimental Building," Journal of Wind Engineering and
Industrial Aerodynamics~ VoL 41-44, pp 1589-1600.
120
121
Cook., N. J .., 1985: The Designer"s Guide to Wind Loading of Building Structures. Part I:
Background. Damage Survey, Wind Data and Structural Classification,
Butterworths., London, UK.
Cook., N. J., 1990: The Designer's Guide to Wind Loading of Building Structures. Part II:
Static Structures, Butterworths, London., UK.
Dalley, S., 1993: "''Investigation into the Simulation Priorities for Wind Tunnel Tests on
Low-rise Buildings," Proceedings of First International Conference on Wind
Engineering. European and African Regional Conference, pp95-104.
Gartshore, I.S., 1973: "'The Effects of Free Stream Turbulence on the Drag of Rectangular
Two-dimensional Prism," University ofWestern Ontario, Canada BLWT-4-73.
Hiller, R., and Cherry, N. J., 1981: 'The Effects of Stream Turbulence on Separation
Bubbles,, Journal of Wind Engineering and Industrial Aerodynamics, Vol 14,
pp375-386.
Hoxey, R., Robertson, A., Richardson G., and Short, L., 1996: "Correction of Wind-
tunnel Pressure Coefficients for Reynolds Number Effects,u Abstract, the Third
International Colloguium on Bluff Body Aerodynamics and Applications," ppA-
VI-1, Virginia Tech and State University.
Jamieson, N.J., and Carpenter, P., 1993: "Wind Tunnel Pressure Measurements on the
Texas Tech Building at a Scale of 1:25 and Comparison with Full-scale,n
Proceedings of the 7th US National Conference on Wind Engineering, Vol 1,
pp303-312, University of California, Los Angeles.
Jenson, M., 1958: 'The Model Law for Phenomena in Natural Wind," Ingenicpren
(International Edition), Vol 2, ppl21-128.
Kawaii, H., and Nishimura., G., 1996: "Characteristics of Fluctuating Suction and Conical
Vortices on a Flat Roof in Oblique Flow," Journal of Wind Engineering and
Industrial Aerodynamics. Vol60, pp211-225.
Kind, R.J., 1986: 'Worst Suctions Near Edges of Flat Rooftops on Low-rise Buildings,"
Journal of Wind Engineering and Industrial Aerodynamics, VoL 25, pp31-47.
122
Kind7 R.J., and Wardlaw, R. L., 1982: uFailure Mechanism of Loose-Laid Roof-Insulation
Systems," Journal of Wind Engineering and Industrial Aerodynamics, VoL 9,
pp325-34L
Kirrane, P.P., and Steward S.J., 1978: "The Effect of Blockage on Shear Flow in the
Wind Tunnel,"' Department of Aeronautical Engineering7 Thesis 221, University of
Bristol
Kumar, K.S., and Stathopoulos T., 1997: 'tpower Spectra of Wind Pressures on Low
Building Roofs," Second Europe and Africa Conference on Wind Engineering,
ppl069-1076, July 1997, Genoa, Italy.
Levitan.. M.L., Holmes J. D., Mehta K.C., and Vann W. P., 1989: "Field Measured
Pressures on the Texas Tech Building,'' Proceedings of the 8th Colloquium on
Industrial Aerodynamics, ppl3-24, Aachen., West Germany, September 4-7. 1989.
Levitan, M.L., and Mehta, K.C., 1992a: "Texas Tech Experiments for Wind Loads Part 1:
Building and Pressure Measuring System," Journal of Wind Engineering and
Industrial Aerodynamics, Vol 41-44, ppl565-1576.
Levitan, M.L., and Mehta, K.C., 1992b: ~'Texas Tech Experiments for Wind Loads Part
II: Meteorological Instrumentation and Terrain Parameters," Journal of Wind
Engineering and Industrial Aerodynamics, VoL 41-44, pp1577-1588.
Levitan, M.L., Cochran, L., and Norville, H.S., 1993: ~'Damage Assessment of Hurricane
Andrew in Louisiana. " Proceedings of the 7th US National Conference on Wind
Engineering, Vol 1, pp393-402 . University of California, Los Angeles.
Lin, J.X., Surry, D. and Tieleman, H.W., 1995: 'The Distribution of Pressure Near Roof
Corners of Flat Roof Low Buildings," Journal of Wind Engineering and Industrial
Aerodynamics., VoL56, pp235-265.
Marwood., R., 1996: "An Investigation of Conical Roof Edge Vortices," Ph.D.
dissertation, Lincoln College, University of Oxford.
Mehta, K. C., 1984: "Wind Induced Damage Observations and Their Implications for
Design Practice," Journal of Engineering Structures., ASCE, Vol 6, 1984, pp242-
247.
123
Melbourne, W.H.~ 1992: 'Turbulence and the Leading Edge Phenomenon," Proceedings
of the 2nd International ColloQuium on Bluff Body Aerodynamics and
Applications, Melbourne, Australia.
Minor, J.E., and Mehta K.C., 1979: "Wind damage Observations and Implications,"
Journal of the Structural Division, ASCE .. November, 1979., pp2279-229l.
Minor, J.E., 1981: "Wind Effects on Buildings.,H The Thunderstorm in Human Affairs, E.
Kessler (Ed.), pp89-109, University of Oklahoma Press, Norman.
Mochida, A, Murakami, S., Shoji, M., and Ishida, Y., 1993: 'Numerical Simulation of
Flow-field around Texas Tech Building by Large-Eddy Simulation/' Journal of
Wind Engineering and Industrial Aerodynamics, VoL 46-47, pp455-460.
NBCC, 1995: "User's Guide - NBC 1995, Structural Commentaries (Part 4). National
Research Council Canada, Ottawa, Canada.
Panofsky, H.A., and Dutton J. A, 1984: Atmospheric Turbulence- Models and Methods
for Engineering Anplications, John Wiley & Sons, New York. NY.
Saathoff, P.J., Melbourne, W.H .., 1987: "Free-stream Turbulence and Wind Tunnel
Blockage Effects on Stream-wise Surface Pressures~" Journal of Wind Engineering
and Industrial Aerodynamics, VoL 26, pp353-370.
Sachs, P., 1978: Wind Forces in Engineering, 2nd Ed., Pergamon Press, Elmsford, NY.
Sandri, P., 1992: ''Frequency Response Requirements for Measuring Fluctuating Wind
Pressure on Buildings," Master's thesis, Department of Civil Engineering, Texas
Tech University.
124
Sarkar, P.P., and Zhao .., Z., 1997: ~ow Visualization and Measurement on the Roof of
the Texas Tech Building.," to be published in the Journal of Wind Engineering and
Industrial Aerodynamics.
Savory, E., and Toy., N.., 1996: ''Reynolds Number Effects on Wind Loads on Low Rise
Buildings," Abstract.. the Third International Colloguium on Bluff Body
Aerodynamics and Applications.,"' ppA-VI-5, Virginia Tech and State University.
Selvam, P.P .., and Konduru., P.B., 1993: "'Computational and Experimental Roof Comer
Pressures on the Texas Tech Building," Journal of Wind Engineering and
Industrial Aerodynamics, Vol 46-47., pp449-454.
Sheffield, J.W., 1993: ''A Survey of Building Performance in Hurricane Iniki and Typhoon
Omar.," Proceedings of the 7th US National Conference on Wind Engineering,
Vol 2, pp653-662, University of California, Los Angeles.
Simiu, E., and Scanlan, R. H., 1996: Wind Effects on Structures: Fundamentals and
Applications to Design, 3rd Ed., John Wiley & Sons Inc., New York, NY.
Smith, T.L., and McDonald, J.R., 1990: "Roof Wind Damage Mitigation: Lessons from
Hugo.,"' Department of Civil Engineering, Texas Tech University.
Stathopoulos, T., Davenport, A.G., and Surry, D., 1978: "The Assessment of Effective
Wind Loads Acting on Flat Roofs.," Proceedings of the 3rd Colloguium on
Industrial Aerodynamics, Part 1, pp225-239.
Stathopoulos, T., and Surry, D., 1983: "Scale Effects in Wind Tunnel Testing of Low
Buildings," Journal of Wind Engineering and Industrial Aerodynamics, Vol.l3,
pp313-326.
Surry, D., 1989: "Pressure Measurements on the Texas Tech Building-II: Wind Tunnel
Measurements and Comparison with Full Scale," Proceedings of the 8th
Colloquium on Industrial Aerodynamics, Aachen, West Germany, September 4-7,
1989, pp25-35.
Tieleman~ H.W., 1992: uProblems Associated with Flow Modeling Procedures tor Low-
rise Structures~" Journal of Wind Engineering and Industrial Aerodynamics, Vol
41-44, pp923-934.
4
Tieleman, H ..W., Surry, D., and Lin, J.X., 1994: 'Characteristics of Mean and Fluctuating
Pressure Coefficients under Comer (Delta Wing) Vortices," Journal of Wind
Engineerin& and Industrial Aerodynamics, Vol 52, pp263-275.
Tieleman, H.W., 1995: "Simulation of Surface Winds for Assessment of Extreme Wind
Loads on Roofs," Proceedin&s of the 9th International Conference On Wind
Engineering, pp 1162-1169, New Delll India.
Van der Hoven, 1957: "'Power Spectrum of Horizontal Wind Speed in the Frequency
Range from 0.0007 to 900 Cycles per Hour.," Journal of Meteorology, Vol 14.
ppl60-164.
Vognild., R.A, 1993: "Are Building Codes Adequate for Hurricane Protection?"
Proceedings of the 7th US National Conference on Wind Engineering, Vol. 2,
pp833-838., University of California, Los Angeles.
4
Wagaman, S. A.., 1993: 'Full-scale Flow Visualization over a Low-rise Building,"
Master's thesis, Department of Civil Engineering., Texas Tech University,
Lubbock, TX.
Woo., H.G.G., Peterka, J. A., and Cennak J.E., 1977: "Wind-tunnel Measurements in the
Wakes of Structures,'' NASA Contractor Report NASA CR-2806, National
Aeronautics and Space Administration.
Yin., J.M., 1996: uProbability Study of Wind Pressures on Buildings," Ph.D. dissertation.
Department of Civil Engineering, Texas Tech University. Lubbock, TX.
APPENDIX A
A.2 Environmental
Operating temperature: -40C to +60C
Storage temperature: -50C to +75oC
Relative humidity: 5 % to 100 %
Precipitation: up to 300 mmlhr
Altitude: 0 to 3000 m
Moisture ingress: IP65
126
APPENDIXB
DATA ACQUISffiON LOO
127
128
Anemometers on the Meteorological Tower. The sampling frequencies were 7.5
Hz and 10Hz for runs M49n001- 125 and runs M49n126-287~ respectively.
Pressures. The sampling frequencies for pressure data were 22.5 Hz and 30 Hz for runs
M49n001-l25 and runs M49nl26-287~ respectively.
v
Figure B.l Coordinate system
u
129
In the following tables7 BLK is the number of firings averaged by the anemometer.
The basic sampling rate for wind velocity is 10ms. The desired output (sampling)
frequency is determined by a user-specified integer value between 1 and 250. BLK=5
indicates an output frequency of 1000 ms I (5x 10 ms) =20 Hz. 'Range' represents the
maximum axial wind speed corresponding to 2.5 Volts analog output. Here a linear
relationship applies. 'x' is the horizontal distance of the sonic anemometer from the
leading roof edge; .. h' is the height of the sonic anemometer from the building roof. 'H'
and 'V" in the column 'sonic' denote the anemometer orientation being horizontal (H) or
vertical (V). 'Bldg' specifies the orientation of the test building; 'U' is the approximate
run-averaged upstream wind speed at 13ft on the meteorological tower.
The positive wind velocities of the instrumental u- v- w-components are indicated in
the following graph for runs M49n00l-156, M49nl96-219.
The following definition of u- v- and w-components are good for runs
M49n158-169, M49n220-226 (Figure B.2).
U+ ~V+ Wmd
~
For other data runs, the definitions are given wherever convenient along with other
information. The sonic anemometer was in operation for automatic data runs
M49nl70-195.
131
Tables B.l - B 11 list details of experimental setup for data runs M49n008-156.
.. ..
>Y
T a bleB .2 S oruc
. d ata acquJStt.J.on o~ {2116/1997)
Run# BLK Range X b sonic Bldg. start U,mpb
M49n022 5 30mls 7'..00.. 1' .()()" H 285 16:36:09 13
M49n023 .. .. 8'..00" .. .. .. 16:59:32 12
M49n024
M49n025
..
..
.... 9'-00,.
10'..()*' ..
'tt ..
..
..
..
17:23:00
17:49:08
12
12
Sonic data were sampled at 22.5 Hz for Runs M49n058 - M49n 125; T=20 minutes, 27000 data
points
Pressure data were sampled at 22.5 Hz for Runs M49n058 - M49n 125; T=20 minutes, 27000 data
points
M49n058 ., 4' .. .. .. 16:37:29 14
..
u
M49n059 .. s .. .. .. 16:59:56 12
T ableB .5 Sorne
. d ata aCQUJStUon 01 (2123/ 1997)
Run# BLK Ran2e X h Sonic Bldg. start u.mph
M49n060 5 30m/s 6' 1'-08" H 240 12:56:29 18
M49n061 .. .. T .. .. 13:21:07 20
M49n062 . .. . .." . 13:43:57 17
M49n063 .. .. . .." .. .. 14:06:49 17
M49n064 .. .. s . .. . 14:29:42 18
M49n065 .. .. 9' . .. .. 14:52:34 19
M49n066 .. .. 10" .. .. . 15:15:27 18
M49n067 ..
..
..
..
11' .. ..
..
.. 15:38:19 19
M49n068 12. .. .. 16:01:12 14
M49n069 .. .. 13' .. .. .. 16:24:29 ..
M49n070 .. .. 14' .. .. . 16:47:21 ..
M49n071 . .. 15' .. .. .. 17:10:13 15
M49n072 .. .. oo 2'-03" .. .. 17:47:32 17
133
Table B.6 Sonic data acquisition lo! (2/26/1997)
Run# BLK Range X h Sonic Bldg. start U~mph
M49n073 5 30m/s o~ -06'' 2"-03 .. H 15 14:20:07 19
M49n074 .. . 1' .,. .. .. 14:42:58 16
M49n077 .. ... 2' ., . .. 16:18:55 15
M49n078 .. .. .. .. .. 16:41:37 17
. .. .
19
..
n
Runs M49n126 and after were sampled at 30 Hz for both the pressure and the sonic
anemometer. With total data points of 27()(X), the run duration was 15 minutes.
T ableB.10 Somc
. data acqwstnon 1og (3/06/1997)
Run# BLK. Range X h Sonic Bldg. start U,mph
M49n126 5 30mls 13. 3"-00'" v 270 15:50:50 15
M49n127 .. 14' ... . .. 16:09:16 16
M49nl28
"
.. 15 .. .. 16:27:45 16
M49n130 .." . o 1'-04"
"
H .. 17:55:31 15
Runs M49n 131 - 143 are for flow visualization of conical vorte~ but the video turned out not clear
3/13/1997
Runs M49nl58 - 169
The 3-D sonic was placed above the leading roof corner to measure the direction I
magnitude of the upstream wind. Conical vortex was visualized at different location by
moving the grid tied with yarns. Details of experimental setup are given in Table B.l2.
Figure B.3 is a schematics view of the locations of the sonic anemometer and the grid.
Wind~
Grid location
Runs M49nl70-195 were automatic data runs with bldg. position of 165.
137
Tables B.l3- 816 list details of experimental setup for data runs M49nl96-219.
..
n
Run#
M49n203
Date: 3/24/1997
Flow Visualization of conical vortices (M49n220-224); SB (M49n226)
Figure B.4 and Table B.17 give the experimental setup and details.
M49n224 ,, .. - - .. 21:07:49 18
Sonic
(ROOF)
L Sonic anemometer was placed about 2 ft from each edge, 4' -02" above the roof;
2. At the end of Run 223, the camera was knocked over by a gust over 60 mph. The sonic
was blown off its position by 22 to the right, which was found after tests were over, thus
sonic data must be corrected by this amount for runs 224 and 226, and probably for a
small portion of 223 toward the data run end.
3. The sonic data were bad for runs m49n222, 223, 224, with voltages being close to zero.
probable because the 30 m/s range was exceeded and the anemometer started to
malfunction.
139
Freguency-Res.ponse Comparison of Anemometers
Earlier, large-and-fast direction fluctuations were observed using the three-
component sonic anemometer, and it was suspected that the propeller type UVW
anemometers might not respond adequately to swift direction changes and wind gusts of
short duration. For comparison, the three-component sonic anemometer was setup side by
side with the UVW anemometer on the meteorological tower at 13ft. Figure B.5 shows
the experimental setup.
6' -04"'
t< >f
09/28/1997
Outstanding measurement of wind flow inside the SB
Tables B.l8- B23 list details of experimental setup for data runs M49n252-287.
Figure B.6 indicates the definition of instrumental u- v- w-components for data runs
M49n252-259.
D
Figure 8.6 Definition of instrumental u- v- w-components (M49n252-259}
By mistake, the sonic anemometer was not oriented as desired. The actual orientation is
shown in the above sketch. Again, the right-hand-rule applies.
141
10/05/1997
Data runs intended to find the upper boundary of the separation bubble.
W+
D
Figure B.7 Definition of instrumental u- v- w-components (M49n261-265)
142
10/08/1997
Run#
M49n267
Figure B.8 indicates the definition of instrumental u- v- w-components for data run
M49n267.
W+
U+._l
D
Figure B.8 Definition of instrumental u- v- w-components (M49n267)
10/12/1997
Runs M49n269 - 272
D
Figure B.9 Definition of instrumental u- v- w-components (M49n269-272)
143
10/12/1997
Runs M49n274- 283
U+ ~ U+ ---.,.
.-v+ V+ t
f ol f ol
(M49n274 - 275) (M49n276 - 283)
W+
"\+
+-+U+ V+ .J
ol
(M49n284-286)
DI
(M49n287)
The configuration of Mode-15 is identical with that of Mode-49. In these two modes,
the test building was in its basic form and without any appurtenances attached.
The selected taps for which their pressure coefficients are provided are taps from the
heavily instrumented roof corner, and taps along the short roof-axis. They are taps 50901,
50205, 50505, 50905, 50209, 50509, 50909, 50123, 50523, 50823, 51123 and 51423.
The mean, rms, min. and max. pressure coefficients are given versus the wind angle of
attack. Similar figures for taps 50101 and 50501 were provide in the text.
145
146
Ol 1 +---+-----f--+----+---,..,---r:""~~-~-----+----i
il 0.5 +---+-----f-~....
0~~~~~~-L~~~~
0 45 90 135 180 225 270 315 360
AOA
a -5+-----+---~~
.....
~ -10+---+----~--~--~~~~--~~--+---~
-15~--------~--~----~--~--~---------
AOAvs. Mean
~~~~--~~-iM..~==~--~~~ 1~1
0 45 90 135 180 225 270 315 360
AOA
AOAvs. RMS
AOAvs. Min
AOAvs. Max
::E 0.5
=
0 + - -..... :;._=-.....;_+---Z....t
AOAvs. Mean
c: .0.5 +---+---------i---=-::-11~
I
~ -1+----~-~-----
AOAvs.RMS
0.8+-----+-----+-----+-----~----r-~~~----r---~
~ 0.6
a: 0.4 -t----+----+-----::":11111
02.,~~~~~~~~
AOAvs.. Min
o~~rr~~~,-----~----~----~----~----~----~
-2 -t-=-~~~~~..__,
.5-4+-----+----t---~
~
~+-----+-----~--~~----~----~----~----~----~
~~----~----~----~----~----~----~----~----~
0 45 90 135 180 225 270 315 360
AOA
AOAvs. Mean
o . .~~~~~~t----r--~~.-~~~~~
c: ...().5 ~;::;..._....::111111. . . .' - =...........~
I -1+-----r-----~--~
:E 1.5 +---+----+------1~~~~-__,._ _-+__,;:::..---+-----t
-2+-----+-----r----1~------~------~-----+~---+------t
-~5._--------------~~--~----~----~----------~
0 90 135 180 225 270 315 360
AOA
AOAvs. RMS
1.2
AOAvs. Min
OT-~r-~----~----~----~----~----~----~----~
-2 ..pr::----'11111~~~~..... -::;:IE"%=::--+----""'".:'::-
.s -4 +---+----+---'~
::s
~+-----+----+----+----4~~~-+-----~~--~---~
-8~----~----~----~----~---~----~----~-----
0 45 90 180 225 270 315 360
AOA
AOAvs. Mean
0~~--~--~~----~----~----~----~----~--~
..0.5 +--.-;:::~~IIIFZ~~+--~
i -1
:I -1.5 +---+---+---+---+--.::IIWII!i---r4!JIK-+----+------I
~+-----+----+-----+-----+---.::11~~-+------+------1
-~5L-----L-----L-----~----~----~----~----~--~
0 45 90 135 180 225 270 315 360
AOA
AOAvs.RMS
0
o.s +----+-----4-----....._---+-~
:E 0.6 +----+----4------+----~~--~----'--:-t-----+------t
a: 0.4 +------+------+---
0.2 ~~--=-~::"'"""":1:=----t---::E:~illlll
o~~~~----~~--~---~---~---~---~----~
0 45 90 135 180 225 270 315 360
AOA
AOAvs. Min
0~--~~~--~----~----~----~-----r----~----~
-2~~-w~~~...._~-.
c: -4+----+------+--___,.,.~--r.-;..;:;;
2 ~T-----T-----;----;-----~~~~~~~-----r----~
~+----+-----+------+-----+---~~----~-----+---~
-10~----~----~----~----~----~----~----~----~
0 45 90 135 180 225 270 315 360
AOA
AOAvs. Max
AOAvs. Mean
0~----~----~----------~--~----~----~----~
..0.2
..0.4.f---~
I -o.s + - - - + - - - + - - -
:.:~ ..0.8 -1----+----1----~
1+-----+-----+-----~~ ~----~~~~
-1.2 +----f----..J-----J.---Z.--ii-----4-._..Za-J-.:.......--J.--~
-1.4 ..L-----'-----'----"----1----___...IL____;;____._ _---~._ _---J
0 45 90 135 180 225 270 315 360
AOA
AOAvs.RMS
1~----~---~----~----~----~----~----~----~
0.8+-----4-----~----~----~----~~-.--+-----+----~
0 ~6+-----4-----~----~----~----~~~~----+-------
:E
a: 0.4 - 1 - - - - - t - - - - - + - - -
0.2 ..f.IIIE-----.,~::--=~-+-~~
AOAvs. Min
AOAvs.Max
M
0.5
0+--~~~--+-----~
-o.5~-----'-------'-----~----~----"------L-----~----
o 45 90 135 180 .225 270 315 360
AOA
=-o.5
0
.... ~~.
AOAvs. Mean
... ~~ ~
I
:IE -1
IMeanl
~
-1.5
0 45 90 135 180 225 270 315 360
AOA
AOAvs.RMS
3
2.5 .. !
2 I
UJ !
..
:1 1.5
~
0.5
0
1
0 45
.. 90
....
135
....180
~ 225 270 315
j
!
I
i
360
I+RMSI
AOA
.5-4
:1
0
-2
-6
&.. I .~. ~. 41 ~
~ .
~
AOAvs. Min
..
I
I+Minl
-a
~
l
0 45 90 135 180 225 270 315 360
AOA
AOAvs.Max
5
~ ;
4
..
)( 3
.
j
-
!
I+Maxj
.-~
:12
i
1
0
0 45
.....
90
~..-.....
135
Jf!_
180
......
AOAvs. Mean
o~--~-=~~~--~--~~----~--~----~--~
-o.5
i -1+---~--------+----~----+-~~
:1 -1.5 +---~--------+--~---+-----t---z;:::;:p~..;.._-+-------t
~+---~--------+-------~----------+--------t----+~~-+-------t
_,5----~----~----~----~---------------------
0 45 90 135 180 225 270 315 360
AOA
AOAvs.RMS
AOAvs. Min
~~~~jili~----~:::r===r=::J==~
:ic ~t=====t====j=====t===~=r~~~~~~~~~!===:j
-6
~+----~--~--------+---+----t-~~~~-~----~
-10------~--~----~----~----~----~----~--~
0 45 90 135 180 225 270 315 360
AOA
AOAvs.Mean
o~~~~~r2~~--~~~~~r-~----~----~--~
-o.5 ~~-~~U~IQ!!!!!!!
I _,
:::E -1.5 +----+----+---+---+-------+---+--~+------!
~+-----+-----+-----+-----+-----+-----~~--+---~
-~5._---~----~----~----~----~----~----~--~
0 45 90 135 180 225 270 315 360
AOA
AOAvs. RMS
1~----~----~----~----~-------~-----r-------~----~
0.8 + - - - - - + - - - - + - - - - + - - - - + - - - - - + - - -............- - + - - - - - i
= 0.6
a: 0.4 +----+-::----+---------+-----+-~:4
-.r.p
,.
02,.~~~. .~~.a~ ....~~----~---1~-~
0+-------+--------+---~-+-~---+-----+-----~----+-------~
AOAvs. Min
c -~ +--~~~~~~~~~~~~~!~~~~i~aii~l=~~
-4
5 ~+-----~-----+-----+----~----+-~~~.-~----~
~+-----+--------~----~--~-----+--~----~----~
-10~------~-------~-------~----~----~----~----~---~
0 45 90 135 180 225 270 315 360
AOA
AOAvs. Max
1.5 .,.----~----..,---~--...,.---...,.---..,..-----,------,
1~~-+-----~----~-~~~--~----~~--~---~
= 0.50+--...;.
:::E
..0.5+-----+-----~--~----~----~----~~L-~---~
-1------------------------~----~-------~----~---~
0 90 135 180 225 315
AOA
-2+----+---+----+--+---+---+-~--+------1
-2.5 ...___ __.__ _.....___ __.__ __.___ __.__ _ _ _ ___.__ ___,
AOAvs. RMS
1 ----~--~~--~--~--~~--~--~--~
0.8 +----+-----it----+----+--t---~~--+-----1
I'll 0.6 +---+---1------+---+-----:~lr--:::-~~......___---1
r! 0.4 +----1-----+--+-----+----z~
6._--~--------~----~--~--~~~~--~
0 45 90 135 180 225 270 315 360
AOA
AOAvs. Mean
0
i
~
~.5+---~~~~~~~~----~~--~
-1.5
1+--------+-----+-----+-----+-----+-----+-----+---~
+---------+---+----+----+---+-----+----+----i
~~----~----~----~----~----~----~----._--~
..
.
0 45 90 135 180 270 315 360
AOA
AOAvs.RMS
0.8 ..,...----..,.......--..,.......---,.-----,----..,......----T----T-----,
0.6 +-----+-------+-----t-----t----t----t-..-...:.r---t-----1
tn
~ 0.2
o. t;~:t~L~:~::l:::J~M~II~;t=~J
4
.~.~._""':16::~........~=--=--+--:---:
0+-----t----~----~~--~----~------~----~----~
0 45 90 135 180 225 270 315 360
AOA
AOAvs. Min
o~----~----~----~~~~-----~~-----~----~--~
-2
=~+------r----~--~----~----~~--~~
~
~+------+------t------+----~---~------r-~--+---~
-8~----~----~----~----~----~~----~----~--~
0 45 90 135 180 225 270 315 360
AOA
AOAvs.Max
2~-----------r------r-----~-----r----r----~---~---~
1.5 +-----t-----+----+------+-----+-------+-----+----1
0.5
0+---__...:ll"f-----~--.::;.._,.;;,~
AOAvs. Mean
AOA
AOAvs.RMS
0.8
Cl)
0.6
:1 0.4
a:
0.2 IRMsj
0
0 45 90 135 180 225 270 315 360
AOA
AOAvs. Min
AOA
AOAvs. Max
2
1.5
)( 1
:1 0.5 IMaxl
0
-o.5
0 45 90 135 180 225 270 315 360
AOA
The Cp(a) in the text was based on wind measurement on the meteorological tower.
In the following derivation for the equation used to predict Cp(t) from wind measured at a
point close to the test building~ reference is made to the definition shown in the following
sketch. The parameters with a bar above are averaged values.
LlP(t) = Cp *Qoc(t).
To avoid the need to compensate for time lag between measured qoc:(t) and corresponding
pressure fluctuations, use the approximation,
Qoc{t) = {q Jq)*q(t),
158