Beruflich Dokumente
Kultur Dokumente
OF
n~B
Journal of Wind Engineering
and Industrial Aerodynamics 57 (1995) 191-201
ELSEVIER
Abstract
Model-scale data for the Silsoe Structures Building, from the BRE Boundary Layer Wind
Tunnel number 3, is compared to recent full-scale data. Assessment of boundary layer simulation precedes a comparison of surface pressure measurements. For transverse winds both sets of
data compare well. Model-scale measurements underestimate the mean pressure coefficient for
cornering winds. This is exemplified by data for one tapping point at the gable end of the roof.
For this tap, the quasi-steady assumption is shown to apply to the reference dynamic pressure
spectrum.
1. Introduction
The Silsoe Structures Building, erected at the Silsoe Research Institute, Wrest Park,
Silsoe in 1987, was designed specifically as a full-scale test facility for the measurement
of wind pressures and their effects on the structure [1]. The building can be configured with either a curved or a sharp eaves detail. Wind-tunnel tests on scale models
of the building have been made in several laboratories to evaluate the modelling
process.
A series of full-scale experiments utilizing a set of equipment from the Building
Research Establishment (BRE), Garston, was started in N o v e m b e r 1990. These were
complemented with wind-tunnel measurements, using the recently completed blowdown boundary layer wind tunnel [2] at BRE, at a 1 : 100 linear scale. In this paper the
coordinate system is defined with x in the streamwise direction, y in the lateral (or
transverse) direction and z normal to the floor with velocity components, u, v and w,
respectively.
* Corresponding author.
0167-6105/95/$09.50 1995 Elsevier Science B.V. All rights reserved
SSDI 0 1 6 7 - 6 1 0 5 ( 9 4 ) 0 0 1 14-6
192
G.M. Richardson, P.A. Blackmore/J. Wind Eng. lnd. Aerodyn. 57 (1995) 191-201
!~
56.0
!~
t----0-.
46.0
-0 ~---
"i
-0-----(~-
.+
I
-0
.....
-O------O-
oE!
-~0
.1[
129.3
l, I
Fig. 1. Model-scale tapping plan for the curved eaves Silsoe Structures Building (all dimensions are in mm).
2. Experimental details
The 1 : 100 scale models of the Silsoe Structures Building, manufactured by BRE,
have been used on a n u m b e r of occasions [ 3 - 9 ] to determine pressures at a total of 81
points for the curved eaves and 74 points for the sharp eaves model. The c o m p a r i s o n
with full-scale data reported here is for a set of 32 taps, illustrated in Fig. 1, c o m m o n
193
to both models and the full-scale building. The model tappings are 1 mm diameter
holes, connected to a pair of scanivalve pressure scanning switches via 400 mm long,
1.6 mm internal diameter PVC tubes. These tubes were restricted at their mid-length,
by brass restrictor tubes, to give a fiat frequency response to 100 Hz. The differential
pressure transducers were Scanivalve model type PDCR24D.
The full-scale surface pressures were sensed by 9 mm diameter tapping points in the
centre of 450 mm square aluminium sheets. These spanned the corrugations with
tapered infills both to prevent flow beneath and to avoid a sudden change in the
surface profile. The pressure was conveyed from the central tapping to a local
transducer via 1 m long, 6.4 mm internal diameter plastic tubing. The transducer
assemblies included two solenoid valves which allowed automatic zero and calibration during the operation cycle. Solid state Honeywell pressure transducers, reference 163PC01D36 were used. These were all simultaneously logged on a 32 channel
CED 1401 data acquisition system controlled by a C O M P A Q 386/20e computer.
The ridge height reference wind dynamic pressure was computed from instantaneous velocity measurements at both model and full-scale, although a direct
measurement via pitot-static probe pressures was used for the model-scale mean
pressure coefficient (Cp) data. At model-scale, a Dantec type 51 cross-wire probe was
used. The reference static pressure was obtained from a pitot-static measurement,
corrected for underestimation of static pressure in the turbulent boundary layer. At
full-scale, all three components of wind speed were sensed simultaneously using
a Solent ultrasonic anemometer (model 1012R2) which was mounted upstream of the
building at ridge height, together with the static pressure probe used for the static
reference pressure.
Model-scale time history data were recorded at 320 Hz and low pass filtered at
160 Hz. Full-scale data were unfiltered and recorded at 5 Hz. The model-scale ridge
height velocity was approximately 9 m/s and is typical of full-scale.
The pressure coefficient was determined from the time history record at full-scale.
The 60 min of unbroken full-scale record was divided into consecutive 10 min periods.
Model-scale Cp were taken from a sample of 16 consecutive 6 s mean values. Assuming a longitudinal length-scale factor of I : 100, 6 s is equivalent to 10 min at full-scale.
Thus the model-scale Cp is averaged over a time interval 16 times the equivalent
length of the full-scale Cp data point (10 rain mean). The rms of wind direction, 8 at
model-scale, is similar to full-scale measurements (for an equivalent period) when
conditions are near stationary.
3. Comparison of results
3.1. Scale factors
There are three interdependent scaling factors; length, velocity and time. The length
scaling was dictated by the 1:100 scale model and the bottom 500mm of the
wind-tunnel velocity profile was matched to produce a 1 : 100 scaling of Zo. The mean
longitudinal velocity at ridge height was typically between 8 and 13 m/s during
194
G.M. Richardson, P.A. Blackmore/J. Wind Eng. Ind. Aerodyn. 57 (1995) 191-201
Height
z(m)
,oo ~z
101 ~1o=
.1
of."
~glZtZo)
~
i~ ~
z
5lo
Velocity ratio
Height
20-
z(m)
15
10
eo
o
o
oe
0-m-0.1
o12
ola
o14
RMSz
-fiz
ks
full-scale measurement. Thus a model-scale mean velocity of 9 m/s gave an approximate velocity scaling of unity. The timescale was determined from a comparison of the
ridge height velocity spectra.
3.2. Profiles
G.M. Richardson, P.A. Blackmore/J. Wind Eng. Ind. Aerodyn. 57 (1995) 191-201
195
10 2
I ....... s=~(?)uoaet-ste
Sww(r~)Model-sccLle
. . . . .
.-~
~j 1 0 - '
10 -~
10-'
10 '
10 "
10 ~
10 ~
10-'
10 0
10'
frequency (n)
ation of U , with increasing height reflected the increasing roughness of the upstream
fetch. The surface of the wind tunnel varied from the smooth turntable centre via
250 m m spaced 10 m m high permeable barriers of interlinked chain to systematically
placed 10 m m high solid barriers. At the inlet the air was blown through a 0.5 m
square grid over a sharp toothed barrier of 484 m m total height 14 m upstream of the
test section. Between heights of 80 and 500 m m U , was 0.72 m/s implying Zo of
0.26 m m for a velocity of 9.9 m/s at 10 m (equivalent) height. The transition to this
value was gradual. At model heights, Zo was about 0.10 mm, consistent with full-scale.
A comparison of the local turbulence intensity is made in Fig. 3. The model-scale
intensity, like full-scale, shows increasing longitudinal turbulence in the b o t t o m 5 m of
the profile. The bottom model-scale measurement reflects the effect of the smooth
turntable centre where the rms is normalised by a larger ~ than would be expected for
the upstream roughness.
Comparisons of the reference longitudinal velocity and vertical velocity spectra at
building/model height are given in Fig 4. The model-scale frequency data has been
reduced by a factor of 100.
3.3. Static references
Model and full-scale static reference measurements were made in the freestream
flow at ridge height to windward of the building. Both measurements were calibrated
versus the static pressure measured at the ground surface. At model-scale, the model
was rotated whilst monitoring the static pressure: no effect due to the model's static
pressure field was observed. The static pressure was measured 365 m m to windward
and 170 m m to one side of the turntable centre. At full-scale, for 0 = 270 , the
reference position was 18 m out from the most instrumented side wall and 3 m to
windward. For the curved eaves building this position was not changed for changing
wind direction. This does result in static pressure errors for some wind directions.
G.M. Richardson, P.A. Blackmore/,L Wind Eng. Ind. Aerodyn. 57 (1995) 191 201
196
1.0
0.75
0.5
!i
i
,
iI
Ii
I
ii
i)
0.25
C~
0.0
0o
-0.75
I?
FS
24C
WT
24C
mean
mean
I
-1.0
45
90
135
180
225
i
270
315
360
of mean
('p
for w a l l - t a p 24.
197
0.0
-0.2~
-0.~
-0.7~
O*
Cp -1.o
-1.25
-1.5
I FS
-1.75
-2.0
45
90
135
180
225
270
315
WT
26C m e a n
26C m e a n
360
from Figs. 2-6 and for wind directions close to 180 it can be concluded that the 1 100
scaling of the atmospheric boundary layer has resulted in accurate determination of
the Cp for tranverse winds.
3.4.3. Gable end roof(O = 240).
The variation of mean Cp with wind direction is given in Fig. 7 for tap 5 on the
gable end roof. The model-scale underestimates the minimum pressure recorded at
full-scale. The wind angles at which the minima occur differ by 5 . A comparison of Cp
for angles between 180 and 360 suggests a different process occurring at model-scale
which cannot be accounted for by static pressure error. These differences may reflect
differences in the effect of vortices at full and model scales as such flow structures are
likely to be important at these angles. The model underestimates Cp at 240 and,
consistent with this observation, the low frequency content of the model spectra for
tap 5 is also underestimated. At full-scale the increased power at low frequency is due,
in part, to non-stationary properties of the atmosphere which are not modelled in the
wind-tunnel boundary layer. However, the significance of this power in terms of the
overall variance of the measurement is small.
Almost all the power present in the tap spectra can be accounted for by multiplying
the reference q spectra by the predicted Cp [10], including the power contributed by
the effect of the changing wind direction upon the tapping pressure (Fig. 8). This
indicates that for this tapping and for 0 = 240 that the spectra is not attenuated by
building effects within the bandwidth shown here.
For this tapping, at least, there is a clear relationship between the simultaneously
recorded pressures when the direction of the freestream wind is considered in relation
G.M. Richardson, P.A. Blackmore/J. Wind Eng. Ind. Aerodyn. 57 (1995) 191-201
198
0.0
--0.25
~r
***a
I
--0.5
--0.75
0o
Cp -lo
-1.25
-1.E
FS
WT
-1.7E
-2.[
45
90
135
180
225
270
315
5C
5C
mean
mean
360
10~
10~
10~
.-
~-F-scc~e Top 5
- - - M-scole Top 5
-- M-scole Mod q
.a,
~,~ " ' ~I ~ .
'
D- 100
,~c
u~ lo,
1o ~
lO~
lO"
10 '
10 '
10
frequency
(n)
100
10'
Fig. 8. C o m p a r i s o n of spectra: Full-scale tap 5 with model-scale of both t a p 5 and its modified q, wind
angle of 240:.
to its effect on the local pressure coefficient. This linkage is established for low
frequency mean characteristics,
Cp(O)=p/O.
(1)
In practice Cp(O) is affected by the rate of change of C o with 0 since p and {/ are
"assumed" to be related to the mean direction 0 alone. The large scatter occurring at
the minimum in Fig. 7 is in part due to this effect. The individual 10 min records at
G.M. Richardson, P.A. Blackmore/J. Wind Eng. Ind. Aerodyn. 57 (1995) 191-201
199
full-scale are sensitive to the spread of wind directions either side of the mean. If 0 is
recorded at a sharp minimum, as in Fig. 7, the true value of Cp is under-predicted.
Without removing lateral turbulence from the flow this true Cp cannot be measured.
However, this has been taken account of in the calculation of the modified q in Fig. 8
[10]. If short duration gusts are considered in relation to Fig. 8 then the following
equation for the peak pressure coefficient (t~p) is suggested to hold true; where 0 is the
direction of the gust which may be unknown,
C'p (0)
ff(O)/q(O).
(2)
Full-scale data, for a tapping point in the middle of the windward roof experiencing
attached flow, show little difference between the magnitude and change of Cp (0) and
(~p(0) when plotted versus the known wind direction 0. For a tap in rapidly changing
flow field (e.g. tap 5), the fluctuation in 0 causes both a horizontal spread of the
minimum in Fig. 7 as well as more negative values as indicated above in reference to
Fig. 7.
The significance of Eq. (2) is that t~p should be determined from a simultaneous
measurement of q and p. The normal form of a peak coefficient appearing in
wind-tunnel data is as follows,
~'p* (#)
=/~/~,
(3)
which is the basic form used in the Cook-Mayne method to determine the value of the
loading coefficient that results in a design load of the desired design risk given a wind
speed of the same risk.
The preferred method in windloading codes today is to generate the pseudo-steady
loading coefficient ~p, where
(4)
Eq. (2) relates closely to (4). The difference is in the record of p and q. In Eq. (2)/3 and
occur in the same sample period. Further accuracy in coded wind-load data would
result if the common wind-tunnel practice of combining the measurement of q with
a calibration period for all data channels preceding the measurement of surface
pressures was supplemented with an additional channel measuring q. If q is determined from a pitot-static probe, a correction for the static probe underread needs to
be made together with a matching of the pneumatic response to that of the tapping
point tubing. Alternatively a hot wire anemometer can be used. In both cases, sample
spectra should be plotted to ensure that the characteristics of the tapping and dynamic
pressure signals are both free from extraneous features such as fan crossing frequency
and aliasing.
Fig. 9 presents a comparison of full-scale 10 min peak pressure coefficients for tap
26, calculated according to Eq. (2), with model-scale data calculated according to Eq.
(3) but with/) and ~ originating from the same sample period. The model-scale data
have also been modified by a reduction factor of 0.85 accounting for the underestimation of ~ due to the long tube lengths used. The factor was determined from
200
G.M. Richardson, P.A. Blackmore/J. Wind Eng. Ind. Aerodyn. 57 (1995) 191 201
0.0
-0.25
-0.5
-0.75
Cp - i . o
-1.25
!
-l.Tf ....... ~
-2.0 '
0
]
45
90
135
180
225
270
315
zx FS
26C m a x i m uLx[a
[] WT
26C
maxim
360
WIND DIRECTION ( 0 )
Fig. 9. C o m p a r i s o n of full-scale p e a k with m o d e l - s c a l e e x t r e m e p r e s s u r e coefficients for t a p 26.
a simultaneous comparison of time history data for the pitot and cross-wire specifically recorded for the purpose. The key difference between these data sets is the lack of
scatter at model-scale, where extreme value analysis to 16 data points has been
applied. Most of the 33 model-scale data points are discernable from the mass of
full-scale data. However, the "once in 50 year windstorm" extreme value analysis is
applied to both top and bottom of Eq. (2). It is the significantly longer equivalent
record length at model-scale that is more important in explaining these more negative
and more consistent coefficients. There is a degree of similarity between these peak
pressure coefficients and the mean data in Fig. 6. As expected the mean data shows
less scatter.
4. Conclusion
For transverse flow (0 = 180"), there is close agreement between model and fullscale data. For the curved eaves model Reynolds number effects appear to be absent.
The model-scale data for a corner tap and a cornering wind (0 = 240 ) indicates
that the mean pressure can be seriously underestimated in the wind-tunnel for this
vortex-loaded region of the roof near the gable end. The reason for this has not been
identified.
The quasi-steady assumption has been applied successfully to the tap 5 reference
dynamic pressure spectrum indicating absence of building induced turbulence effects
in this region for 0 = 240 .
The comparison of peak pressures, in both cases relying on simultaneous measurements of p and q, show reasonable agreement.
G.M. Richardson, P.A. Blackmore/J. Wind Eng. Ind. Aerodyn. 57 (1995) 191-201
201
Acknowledgement
The c o n t r i b u t i o n of the D e p a r t m e n t of the E n v i r o n m e n t in f u n d i n g this collaboration between the U K Building Research E s t a b l i s h m e n t (BRE) a n d Silsoe Research
Institute is gratefully acknowledged, as is the financial s u p p o r t provided by the
M i n i s t r y of Agriculture, Fisheries a n d F o o d for E n g l a n d a n d Wales ( M A F F ) for the
o n g o i n g research p r o g r a m m e at Silsoe.
References
[1] A.P. Robertson and A.G. Glass, The Silsoe Structures Building - its design, instrumentation and
research facilites, Divisional Note, DN 1482, AFRC Engineering, Silsoe, 1988.
[2] G.V. Parkinson and N.J. Cook, Blockage tolerance of a boundary-layer wind-tunnel, J. Wind Eng.
Ind. Aerodyn. 42 (1992) 873-884.
[3] G.M. Richardson and D. Surry, Comparisons of wind-tunnel and full-scalesurface pressure measurements on low-rise pitched-roof buildings, J. Wind Eng. Ind. Aerodyn. 38 (1991) 249-256.
[-4] G.M. Richardson and D. Surry, The Silsoe Building:a comparison of pressure coefficientsand spectra
at model and full-scale,J. Wind Eng. Ind. Aerodyn. 43 (1992) 1653-1664.
[5] G.M. Richardson and D. Surry, The Silsoe Structures Building:Comparison between full-scaleand
wind-tunnel data, J. Wind Eng. Ind. Aerodyn. 51 (1994) 157-176.
[-6] S. Dalley, Wind tunnel measurements on a low-rise building and comparison with full-scale, PhD
Thesis, University of Surrey, 1993.
[7] S. Dalley and G.M. Richardson, Reference static pressure measurements in wind-tunnels, J. Wind
Eng. Ind. Aerodyn. 42 (1992) 909-920.
[8] P.A. Blackmore, Personal communiqu6 No. 1 (1991).
[9] P.A. Blackmore, Personal communiqu6 No. 2 (1991).
[10] R.P. Hoxey and P.J. Richards, Full-scalewind load measurements point the way forward, in: Proc. 1st
European and African Regional Conference,Guernsey, September 1993, J. Wind Eng. Ind. Aerodyn.
57 (1995) 215-224.