Professional Documents
Culture Documents
Wind Load Coefficient
Wind Load Coefficient
Wind Load Coefficient
www.elsevier.com/locate/engstruct
Received 20 March 2001; received in revised form 22 August 2001; accepted 22 August 2001
Abstract
In an earlier paper a parametric study of wind loads on cantilevered grandstand roofs was reported. The effects of grandstand
geometry, sub-roof venting, leading edge fascias and upstream structures were investigated in a comprehensive wind tunnel study.
This paper takes this data and derives equivalent static wind load distributions for the design of simple cantilevered grandstand
roofs. A trapezoidal load distribution is recommended over the current triangular distribution. Correlations measurements are reported
which may be used to obtain equivalent static wind load distributions for other grandstand roof structural systems. 2002 Elsevier
Science Ltd. All rights reserved.
0141-0296/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 1 - 0 2 9 6 ( 0 1 ) 0 0 0 8 5 - 2
208 C.W. Letchford, G.P. Killen / Engineering Structures 24 (2002) 207217
Fig. 1. Parameters investigated and sign convention for lift and moment coefficients.
is a trussed cantilever type structure supported by an The tests were conducted in the University of Queens-
over roof tie and backstay system. The roof is rectangu- land, Department of Civil Engineerings Boundary Layer
lar in plan with a width of 140 m and a span of 30 m. Wind Tunnel, which is 3 m wide by 2 m high and has
The roof is 35 m above ground level with an upward some 12 m of upstream fetch for boundary layer develop-
pitch of 3. The seating is in two tiers. Full details of ment. Full details of the boundary layer simulation, a sub-
the experimental apparatus and procedure are described urban terrain at 1:200, are described in Ref. [2]. Good
in Ref. [2], and only a brief summary is given here. agreement with mean velocity and turbulence intensity pro-
A 1:200 scale model was constructed with the roof files as well as longitudinal component velocity spectrum
manufactured from a sandwich of perspex sheets was obtained. The 0 wind direction (f) was defined as
machined so that pneumatically averaged panels on both normal to the leading edge of the grandstand roof.
the top and bottom could be analogue differenced to Net pressures were sampled at 400 Hz for 15 s and
obtain the net pressure acting on the roof. The model repeated 10 times. The 15-s sampling time corresponded
was rigid and only net pressures were measured. Each to ca 15 min full scale. A FisherTippett type 1 extreme
panel contained 12 pressure taps (six above and six value distribution was fitted to the peak pressures. Mean,
below). Four panels made up one bay of the structure, rms and hourly mean extreme (maxima and minima)
which consisted of 14 bays on a 10 m grid. Panel 1 was were obtained and non-dimensionalized by the mean
at the leading edge, Panel 4 at the rear. Pressures were dynamic pressure at roof height. Cross correlation coef-
area-averaged within each panel. Fig. 2 shows the tap- ficients were also obtained between various panels. The
ping and panel configuration. The sandwich construction
peak bending moment coefficient per unit width of roof
of the roof allowed pressure tubes to be contained
about the cantilever support is defined in Eq. (1). Posi-
within, but resulted in a 9 mm thick roof, somewhat dis-
tive is defined as corresponding to uplift (refer to Fig.
torted from the full scale. The frequency response of the
1). The earlier paper [2] reported extensively on this
pressure measurement system was good (half power) to
moment coefficient as a function of the parameters
150 Hz which according to Letchford et al. [6] would
lead to a maximum attenuation of mean extreme area- investigated.
averaged pressure coefficients of 5%.
The effect of the distorted roof thickness was dealt with
M
in the earlier paper [2] and the conclusion from surface oil CM . (1)
flow visualization was that the area-averaged pressures 1/2rV2hL2
would not be significantly affected by the over thick model.
Fig. 2. Cross-section of model with full-scale dimensions and roof pressure tapping arrangement.
C.W. Letchford, G.P. Killen / Engineering Structures 24 (2002) 207217 209
3. Analysis techniques approach is presented in Ref. [13]. For uplift, each panel
contributes equally to the uplift and the peak factors
3.1. As1170.2 provisions were weighted accordingly.
A simple equivalent static pressure distribution is one
The Australian wind load code [1] presents a model that causes simultaneous peaks in both uplift and bend-
for the equivalent static pressure distribution for the root ing moment. Killen [5] proposed a trapezoidal equivalent
bending moment on a cantilever grandstand roof. The static peak pressure distribution, which led to the simul-
distribution is triangular in form as shown in Fig. 3 with taneous maxima in uplift and root bending moment. This
a leading edge peak pressure coefficient of Cp,c=5 for is shown schematically in Fig. 4. The relationship
relatively stiff cantilevers. For more flexible structures, between peak moment and uplift coefficients and the
this is modified by the natural frequency of the cantilever assumed trapezoidal pressure distribution is shown in
as illustrated in Eq. (2). Pressures are obtained by multi- Eq. (3).
plying by the mean dynamic pressure at roof height.
Melbourne [7,8] pioneered this approach with measure- C12(3CMCL) (3a)
ments on aeroelastic models of several grandstand roofs. C22(2CL3CM) (3b)
V h where CM and CL are the mean extreme (maxima and
Cp,c5.0 for 0.4 (2a)
Lnc minima) moment and uplift coefficients obtained from
Cp,c5.02.5 V h
Lnc
0.4 for
V h
Lnc
0.4 (2b)
covariance integration. Thus both upwards and down-
wards equivalent static peak pressure distributions are
obtained while pressures are calculated by multiplying
where, Vh is the design mean hourly wind speed, and L by the mean dynamic pressure at roof height.
the length and nc the first mode cross wind natural fre-
quency of the cantilever, respectively. 3.3. LRC technique
Although Ref. [1] only gives recommendations for
upwards loads, Melbournes later studies [8] of actual The LRC technique proposed by Kasperski and Niem-
grandstand configurations incorporating upstream struc- ann [4] was used to estimate equivalent static loads for
tures indicate downwards loads of approximately half individual load effects. The technique employs the same
the magnitude of the uplift are possible. data as covariance integration, mean and standard devi-
ation coefficients, cross correlations, peak factors and
3.2. Covariance integration structural influence coefficients but extracts the expected
pressure distribution that leads to the mean extreme
The covariance integration method proposed by (maxima and minima) load effect coefficient. Thus sep-
Holmes and Best [3] was used to estimate both peak roof arate equivalent static peak pressure distributions for
uplift and bending moment coefficients in Ref. [2]. This bending moment and uplift for both positive and nega-
method employs mean and standard deviation pressure tive going loads are obtained and again pressures are
coefficients, cross correlations, peak factors (typically 6 calculated by multiplying by the mean dynamic pressure
for upward loads and 3.5 for downward loads) and struc- at roof height.
tural influence coefficients for the load effect. Root Appendix A gives an example calculation for each
bending moment peak factors were obtained by weight- method and fluctuating pressure coefficient data for other
ing individual peak panel pressures by their influence structural load effects to be examined.
coefficient, thus the leading panel was weighted more
highly than the trailing edge panel. An alternative
Fig. 3. Assumed triangular equivalent static pressure distribution Fig. 4. Simplified trapezoidal equivalent static pressure distribution
from AS1170.2 [1]. after Killen [5].
210 C.W. Letchford, G.P. Killen / Engineering Structures 24 (2002) 207217
4. Results and discussion between these pressures both the LRC and the trap-
ezoidal distributions produce lower values. It is seen that
4.1. Roof pitch and wind direction the trapezoidal distribution is a good approximation to
the LRC expected peak panel pressures and represents
It was shown in the earlier paper [2] that there was a simple model for implementation in wind load codes.
little relationship between peak roof moment coefficient Both these equivalent static pressure distributions are
and roof pitch (7) on a centre bay. This result was considerably larger than the triangular distribution
supported by an aeroelastic study of grandstand roofs by assumed in the Australian Wind Load Code AS1170.2
Kawai et al. [9]. The maximum upward moment [1]. For the centre bay in an isolated grandstand there is
occurred for a roof pitch of +3 and wind direction per- no downwards load due to wind loading.
pendicular to the leading edge of the cantilever. It was Fig. 6 shows the corresponding results for the end bay
also noted that for low reduced velocities (Vh/Lnc0.4) at a wind direction of 30. Once again the trapezoidal
the Australian wind loading code [1] was unconservative distribution represents a good approximation to the LRC
for most angles of attack. peak panel pressures and the actual peak panel pressures
The variation of peak bending moments on an end bay measured on the model. Also the AS1170.2 result is
of the roof with wind direction and as a function of roof included for comparison and indicates an unconservative
pitch revealed that it was slightly more sensitive to pitch loading distribution. For this wind direction there is also
than the central bay, but still not as sensitive as a free no downward load apart from on the leading panel.
roof [10]. The uplift increases as the wind direction is
increased to about 30 and then decreases slightly. This 4.2. Leading edge fascias
variation is due to conical vortices forming along the
edges of the roof. Surface oil flow visualisations con- Fascias are becoming common features on grandstand
firmed this interpretation. The maximum end bay roofs because they provide highly visible signage. In the
moments occurred for the higher positive pitches and for earlier paper [2], various fascia scenarios were tested and
a wind direction of 30. peak moment coefficients reported. Two fascia depths,
For all subsequent tests the roof pitch was +3 and F/L=0.1 and 0.2, were studied and the effects of position,
centre bay results were analysed for 0 wind direction either fully above, fully below the roof level or in-
while end bays were analysed for 30. between, the effects of gaps behind the fascia and wind
Fig. 5 shows the results for the centre bay at 0 wind direction were examined. Fig. 1 shows the geometry
direction. The figure shows the mean and extreme associated with a leading edge fascia. In this study the
(maxima and minima) panel pressure coefficients as well fascia extended the full length of the grandstand roof
as the LRC distributions for peak upwards and down- leading edge. Here the peak pressure distributions in the
wards moment and uplift, and the simplified trapezoidal presence of the larger fascia (F/L=0.2) on the centre bay
distribution. They are plotted as a proportion of the can- for 0 wind direction and on the end bay for 30 are
tilever span with x=0 at the root/base of the cantilever. presented. Both fascias above, and below the roof have
It is clear that the peak upwards panel pressure coef- been analysed while the contribution of the fascia load-
ficients represent the upper envelop of panel pressures ing to the root bending moment has been ignored
and that by incorporating the reduced correlation because of the very small lever arm.
Fig. 5. Centre bay equivalent static peak pressure distributions, iso- Fig. 6. End bay equivalent static peak pressure distributions, isolated
lated grandstand H/L=1.1, wind direction=0. grandstand H/L=1.1, wind direction=30.
C.W. Letchford, G.P. Killen / Engineering Structures 24 (2002) 207217 211
Fig. 7. Centre bay equivalent static peak pressure distributions, iso- Fig. 9. End bay equivalent static peak pressure distributions, isolated
lated grandstand H/L=1.1, wind direction=0, leading edge fascia grandstand H/L=1.1, wind direction=30, leading edge fascia mounted
mounted above the roof. above the roof.
Table 1
Worst case upward peak pressure coefficients for roof internal bays, excluding dynamic effects
BASE: Isolated grandstand pitch, a7 (maximum at f=0, H/L1.4) 3.6 6.0 100 100 156 192
For H/L1.4 4.6 4.5 87 95 136 182
Any fascia on leading edge F/L=0.1c 3.8 4.2 78 83 122 160
Maximum effects on shielded grandstands at f=0 and s/L=6, only
Maximum uplift with half height grandstand upstream H/h=2 2.6 5.8 91 88 142 168
Maximum uplift with same height grandstand upstream H/h=1 1.7 2.9 48 48 75 92
Maximum uplift with approx. double height grandstand upstream H/h=0.58 1.0 2.6 40 38 62 72
a
For a grandstand with a sub-roof vent of 1% of the total elevation area of the structures, a further reduction of load resulting in a 10%
reduction of the total roof moment at the support is allowed [2].
b
C1 is at the leading edge and C2 the trailing edge of the cantilever, respectively.
c
The loads from the fascia itself must be added to the roof system to establish the total wind load effect.
Table 2
Worst case upward peak pressure coefficients for roof end bays, excluding dynamic effects
BASE: Isolated grandstand pitch, a7 (maximum at f=30, H/L1.4) 3.4 4.4 100 100 122 156
For H/L1.4 3.7 6.3 134 128 163 200
Any fascia on leading edge F/L=0.1 and m/L=0.03b 3.5 5.7 122 118 149 184
Maximum effects on shielded grandstands at =0 and s/L=6, only
Maximum uplift with half height grandstand upstream H/h=2 1.5 3.7 73 67 89 104
Maximum uplift with same height grandstand upstream H/h=1 1.2 4.5 84 73 102 114
Maximum uplift with approx. double height grandstand upstream H/h=0.58 1.9 2.9 63 62 77 96
a
No reduction in load for a grandstand with sub-roof venting is permitted for end bays [2].
b
The loads from the fascia itself must be added to the roof system to establish the total wind load effect.
214 C.W. Letchford, G.P. Killen / Engineering Structures 24 (2002) 207217
Table 3
Worst case downward peak pressure coefficients for roof internal bays, excluding dynamic effects
BASE: isolated grandstand pitch, a7 (maximum at f=90) 0.5 0.8 100 100 13 14
Maximum download with half height grandstand upstream H/h=2, s/L=6, f=0 0.5 0.6 33 8 5 2
Maximum download with same height grandstand upstream H/h=1, s/L=6, f=0 0.0 1.4 133 108 37 56
Maximum download with approx. double height grandstand upstream H/h=0.58,
0 2.6 248 200 84 126
s/L=6, f=0
Table 4
Worst case downward peak pressure coefficients for roof end bays, excluding dynamic effects
BASE: isolated grandstand pitch, a7 (maximum at f=90) 2.2 1.8 100 100 48 51
Maximum download with half height grandstand upstream H/h=2, s/L=6, f=0 0.1 0.7 22 15 15 20
Maximum download with same height grandstand upstream H/h=1, s/L=6, f=0 0.5 1.8 53 33 30 38
Maximum download with approx. double height grandstand upstream H/h=0.58,
0 2.4 83 60 62 94
s/L=6, f=0
1.75 0.591 CMmean(0.884)
1.85 0.618
Cpmean:= Cpsd:= Cpmax: CMsigma:=(psiTACpsigmCpsigmapsi)0.5
1.74 0.572
CMsigma(0.263)
1.57 0.5
5.59 0.18
6.19 0.33 A.1.2.4. Covariance integration gives CMmin as
Cpmin:=
5.23 0.37 CMmin:=CMmeangminCMsigma
4.63 0.4 CMmin(0.219)
0.25 1 0.8 0.61 0.42 0.875
0.25 0.8 1 0.74 0.42 0.625
Area:= rho:= psi: A.1.2.5. Covariance integration gives CMmax as
0.25 0.61 0.74 1 0.7 0.375
CMmax:=CMmeangmaxCMsigma
0.25 0.42 0.42 0.7 1 0.125
CMmax(2.613)
0.524 0.45
A.1.2.9. LRC effective static pressure distribution 0.334 0.641
Cpmin
0.579 0.364 0.598 0.417
LRCp:=psiTACpsigmarhoCpsigma EigvecCM
0.52 0.38 0.601 0.473
LRCp(0.145 0.152 0.123 0.074) 0.345 0.722 0.41 0.438
gmin
CpLRCmin:=CpmeanLRCTp
CMsigma
A.1.3. Roof shear analysis
0.357
0391
CpLRCmin 1
0.561
1
0.86 psi:=
1
CMminck:=psiTACpLRCmin 1
CMminck(0.219)
Which is the same as CMmin.
A.1.3.1. Calculation of weighted gust factor for shears
A.1.2.10. LRC effective static pressure distribution
Cpmax (g1psi1+g2psi2+g3psi3+g4psi4)
gsmax:= gmax6.435
(psi1+psi2+psi3+psi4)
gmax
CpLRCmaxCpmeanLRCTp (gb1psi1+gb2psi2+gb3psi3+gb4psi4)
CMsigma gsmin:= gmin
(psi1+psi2+psi3+psi4)
5.37
2.463
5.642
CpLRCmax
4.803
3.414 A.1.3.2. Calculation of mean and standard deviation
shear force coefficients
CMmaxck:=psiTACpLRCmax
CSmean:=psiTACpmean
CMmaxck(2.613)
CSmean(1.728)
Which is the same as CMmax.
CSsigma:=(psiTACpsigmarhoCpsigmaApsi)0.5
A.1.2.11. Eigenvalue analysis of the covariance matrix CSsigma(0.483)
CovCM:=CpsigmarhoCpsigma
A.1.3.3. Covariance integration gives CSmin as
0.349 0.292 0.206 0.124
CSmin:=CSmeangsminCSsigma
0.292 0.382 0.262 0.13
CovCM CSmin(0.537)
0.206 0.262 0.327 0.2
0.124 0.13 0.2 0.25
0.959
0.213 CSmax(4.837)
EigvalCM
0.045
0.092 A.1.3.5. Simultaneous peaks gives CSmin as
EigvecCM:=eigenvecs(CovCM) Cssimmin:=psiTACpmin
C.W. Letchford, G.P. Killen / Engineering Structures 24 (2002) 207217 217
CSsimmin(0.32) CSmaxck:=psiTACpLRCmax
CSmaxck(4.837)
A.1.3.6. Simultaneous peaks gives CSmax as Which is the same as CSmax
T
CSsimmax:psi ACpmax
A.2. Data for end bay at a wind direction of 30
CSsimmax(5.41)
G:=1.74
1.81 0.533 5.23 0.28
1.62
Cpsd:=
0.457
0.507
0.555
Cpmax:=
3.62
4.03
4.06
Cpmin:=
0.275
0.07
0.02
Area:=
0.25
0.25
0.25
rho:=
0.71
0.48
0.4
1
0.88
0.79
0.88
0.91
0.79
0.91
1
psi:=
0.625
0.375
0.125
A.1.3.8. LRC effective static pressure distribution
Cpmin
LRCp:=psiTACpsigmarhoCpsigma
LRCp(0.243 0.266 0.249 0.176)
gsmin References
CpLRCmin:=CpmeanLRCpT
CSsigma
[1] Standards Australia, AS 1170.2, SAA loading code Part 2: Wind
0.512 Loads; 1989.
[2] Killen GP, Letchford CW. A parametric study of wind loads on
0.493 grandstand roofs. Engineering Structures 2001;23:72535.
CpLRCmin [3] Holmes JD, Best RJ. An approach to the determination of wind
0.472
load effects on low rise buildings. JWEIA 1983;7:27387.
0.673 [4] Kasperski M, Niemann HJ. The LRC (loadresponse-correlation)
method a general method for estimating unfavorable wind load
CSminck:=psiTACpLRCmin distributions for linear and non-linear structural behavior. JWEIA
1992;4144:175363.
CSminck(0.537) [5] Killen GP. A parametric study of wind loads on grandstand roofs.
MEngSt thesis, Department of Civil Engineering, The University
Which is the same as CSmin. of Queensland, 1997.
c1min:=2(3CMminCSmin) c1min(0.237) [6] Letchford CW, Sandri P, Levitan ML, Mehta KC. Frequency
response requirements for fluctuating pressure measurements.
c2min:=2(2CSmin3CMmin) c2min(0.838) JWEIA 1992;40:26376.
[7] Melbourne WH, Cheung JCK. Reducing the wind load on large
c1max:=2(3CMmaxCSmax) c1max(6.004) cantilevered roofs. JWEIA 1988;28:40110.
[8] Melbourne WH. The response of large roofs to wind action.
c2max:=2(2CSmax3CMmax) c2max(3.671)
JWEIA 1995;54/55:32535.
c1 is at the free end of the cantilever; c2 is at the support [9] Kawai H, Shimura M, Yoshie R, Wei R. Wind induced response
end of cantilever. of a large cantilevered roof. In: Proceedings 4th Asia-Pacific
Symposium on Wind Engineering, 1416 July, 1997, Gold Coast,
University of Queensland; 1997:1914.
A.1.3.9. LRC effective static pressure distribution [10] Ginger JD, Letchford CW. Wind loads on planar canopy roofs,
Cpmax Part 2: fluctuating pressure distributions and correlations. JWEIA
1994;51:35370.
gsmax [11] Mallam A. Structural dynamics of cantilevered grandstand roofs.
CpLRCmax:=CpmeanLRCpT
CSsigma Unpublished BE thesis, University of Queensland; 1998.
[12] Letchford CW, Denoon RO, Johnson G, Mallam A. Dynamic
4.985 characteristics of cantilever grandstand roofs. Engineering Struc-
5.397 tures (accepted).
CpLRCmax [13] Holmes JD, Rains GJ. Wind loads on flat and curved roof low-
5.053 rise buildings application of the covariance integration
method. In: Proc. Colloque Construire avec le vent, Nantes,
3.914 France; 1981:V.1.118.