You are on page 1of 12

Journal of Wind Engineering and Industrial Aerodynamics, 41-44 (1992) 1629-1640

Elsevier

1629

C o m p u t a t i o n o f W i n d P r e s s u r e s o n Low-rise S t r u c t u r e s

D.A. Paterson, and J.D. Holmesb -CSIRO Division of Building, Construction and Engineering, P.O. Box 310, North Ryde, N.S.W. 2113, Australia bCSIRO Division of Building, Construction and Engineering, P.O. Box 56, Highett, Vic. 3190, Australia

Abstract

This report describes the use of some computer programs written by one of the authors (DP) in predicting mean and peak wind pressures on arched-roof buildings. The research was carried out because there is a lack of good windtunnel and full-scale data on arched-roof buildings. The programs have been validated by comparing the results with wind-tunnel and full-scale data associated with the Texas Tech building.

1. INTRODUCTION The related problems of wind-induced pressures on building surfaces and wind velocities at pedestrian level adjacent to buildings are of great importance to engineers and architects. Over the last 20 years, these problems have usually been solved by wind tunnel tests. This is a relatively expensive approach, and requires separate models to be made for each geometrical configuration. Also, for a structure with a curved external surface, such as an arched-roof building, the flow and pressures obtained on a small scale model may differ from full scale due to the difference in Reynolds Numbers in the two cases. The technique of computer modelling of wind flows around structures has been developed by the authors. It seems to be as accurate as wind tunnel modelling in the determination of mean flow velocities and mean surface pressures. It is much less expensive than wind tunnel modelling. Mean surface pressure coefficients, when used with peak gust wind speeds, may be sufficiently accurate for the structural design of many structures for which resonant dynamic response is not expected to be important. This approach is known as the 'quasisteady' method, and has formed the basis of many international codes and
0167-6105/92/$05.00 1992 Elsevier Science Publishers B.V. All rights reserved.

1630 standards for wind loading. The present computer programs have been validated by comparing the results with those of many different wind tunnel and full-scale experiments. One such comparison, with results associated with the Texas Tech building, is reported here. In the first half of 1989, a series of measurements of pressures were made on an instrumented (9.1 x 13.7 x 4.0m) (30 x 45 x 13 ft) experimental building at the Texas Tech University [1]. People working with this building have produced some sets of data of mean, r.m.s, and peak pressures. These are some of the best sets of pressure data on low-rise buildings currently available. They are very useful in determining wind loads on low buildings. These full-scale results have already been compared with some wind tunnel results measured at the University of Western Ontario [2]. For the present paper, wind flows around structures were computed by solving the set of partial differential equations that governs the flow. In this instance, the three-dimensional flow around the Texas Tech Building was calculated and the mean and r.m.s, pressure coefficients were compared with both full-scale and wind tunnel results. The programs tested on the flow around the Texas Tech Building have been used to compute the flows around a large number of arched-roof buildings. Two of these computations are reported here. One of the buildings is a mobile tail enclosure produced by Strarch International Ltd of Melbourne, Australia. The mobile tail enclosure is a structure that allows existing hangars to house bigger aircraft. The front of the aircraft is put in the existing h a n g a r and a mobile enclosure is moved over the tail of the aircraft to keep it out of the weather as shown in Figure 1. The enclosure is supported by semi-circular prestressed arches, erected by the patented Strarch system. The enclosure is much taller and narrower than most arched-roof structures because it has to fit over the tail of an aircraft. The structure on which the pressures were calculated consists of a roof with a radius of 15.5 m supported by vertical walls of 10 m high. The length along the axis is 20 m. The other computation reported here is on an arched-roof building. This building is selected because its dimensions are typical of m a n y arched-roof buildings that have been constructed by Strarch International. It has a rise/span ratio of 0.2, a length/span ratio of 1.0, and an eaves height/rise ratio of 0.45. This is one of many computations involving different arched-roof building shapes, different approach flow angles, and different inlet velocity profiles. Buildings were studied both with and without major openings. The aim of the research was to provide as much information as possible on pressures on as many archedroof buildings as possible under a variety of wind conditions.

1631

i
TYPICALSECTION

I i iI

]1
Ex,

I SIDE ELEVATION

Figure 1. Strarch Mobile Tail Enclosure.

2. MATHEMATICS
Computer modelling of fluid flows starts with the equations of conservation of mass and momentum that are collectively known as the Navier-Stokes equations. These equations are time-dependent. In the present approach they are averaged over time to get the Reynolds equations. The equations for the conservation of momentum are solved for the mean velocities. The conservation of mass equation is combined with the equations for the mean velocities to yield a pressure correction equation. This use of a pressure correction equation is known as the SIMPLE method [3]. The time-averaged equations contain six extra terms that are related to the r.m.s, velocities and are created when the time averaging is done. In the present method, these extra terms are derived from the turbulent kinetic energy (k), which is half the sum of the squares of the r.m.s, velocities, a~d from the rate of dissipation of the turbulent kinetic energy (z). k and ~ are calculated from two further equations [4,5]. No attempt has been made to change the results to

1632 better fit experimental data. Earlier work has indicated that fine tuning of the constants in the mode! is neither necessary nor desirable. These equations are non-linear and so are solved in an iterative manner. They were turned into a tractable form by integrating them over appropriate volumes on a staggered grid. The convection terms in the equations were discretised by using the first-order hybrid upwinding scheme. Computation of the flow over a mobile tail enclosure poses special problems. The roof is semicircular and the walls are vertical so it is necessary to model the transition between the two correctly. Polar coordinates can't easily be used and attempts to use body-fitted coordinate meshes were abandoned because of convergence and accuracy problems. In the end a simple Cartesian mesh was used. Another problem with the mobile tail enclosure is that the most critical load case occurs when the arch is off the aircraft and the wind is blowing into the open end of the enclosure. In this case there are very severe lift and overturning forces. The roof and walls have to be modelled as zero thickness surfaces so derivative boundary conditions have to be set by changing nodal coefficients rather than values over boundaries. Extra care is taken when values over boundaries are used temporarily to increase computational efficiency as in TEACH [6]. This algorithm could not be vectorised and so a new algorithm was developed in which the SOR part of the algorithm was replaced by a Jacobi algorithm. This is fully vectorisable, and does not substantially slow down the overall convergence because of the use of ADI. In addition, significant changes were made to the way in which the boundary conditions are calculated. In the final algorithm, vector lengths in excess of 10,000 are achieved in the setting up of the tridiagonal matrices and in the calculation of the initial approximation. Vector lengths are in excess of 400 when doing the block TDMA solution. Overall, there is a factor of about 30 speed-up from these changes. In the present programs, the locations of solid surfaces (walls and roof) were input as a single 2-D array from a data file. This array gives the height of the roof above ground level. The programs automatically place walls at the edges of the roof. The decision about which walls are solid and which are open is handled by logical variables within the programs. This input method allows geometry data to be created quickly by hand, without the need for a pre-processor. It has been used for other problems involving multiple buildings of irregular shape. A large amount of effort has been expended in optimising the present programs for use on vector processing computers. The computer used for the main calculations is a Cray Y-MP2/216. The original algorithm used the Tri-

1633

Diagonal Matrix Algorithm (TDMA) in one of the three dimensions, coupled with Successive Over-Relaxation (SOR) in the other two dimensions. The direction in which the TDMA was taken was varied in the manner of the Alternating Direction Implicit (ADI) method. But the method differed from the standard ADI method in that no operator splitting was applied; so the number of sweeps for each equation did not have to be a multiple of three. Fastest convergence was found to occur when four sweeps were used ~br the pressure correction equation and one sweep on each of the other equations. Pressure coefficients were calculated in a post-processing program that is quite distinct from the main programs. Mean pressure coefficients were calculated from
m

Cp= 2 P / (p V2) - CP0

(1)

where P is the mean pressure, p is the density, V is the approach velocity at building height and CPo is a reference pressure coefficient. The reference pressure coefficient was obtained by setting the mean pressure coefficient to zero at a point well away from the building. R.m.s. pressure coefficients were calculated from
ci, ffi 2 [ k/3 +

0.s16 I CpI Vo

] / v"

(2)

where V 0 and k 0 are the values of velocity and k in the approach flow at the height at which C~ is required. The expression in Eq. 2 is an approximation consisting of two components. The part proportional to k is associated with small-scale local pressure fluctuations in flow free of mean velocity gradients as discussed by Hinze [7]. The part containing V 0 is a 'quasi-steady' term associated with the amplification of upwind pressure fluctuations by velocity gradients in the flow.

3. RESULTS 3.1 Texas Tech Building


Mean and r.m.s, pressure coefficients, mean velocities and turbulent kinetic energies were calculated for approach flows at angles of 0 (parallel to the long side of the building) to 90 in 10 intervals. A fairly coarse grid of 26 x 29 x 14 with variable grid intervals is used. Velocity vector plots and contour plots of pressure and k have been produced along a large number of slices through and near the building for each approach flow angle. Many have not yet been

1634 published and are available from the first author. Computed mean and r.m.s, pressure coefficients on the line of intersection of the central plane of the building with the building surface have been plotted for each of" the ten approach flow angles. These were compared with results from full scale and wind tunnel experiments for approach flow angles of 0, 60 and 90 . Those from the 90 degree approach angle are included in Figure 2.

Mean Pressure Coefficient, Theta=90


~, Computation x Full Scale Wind Tunnel

~5
I

t"q

R M.S. Pressure Coefficient, Theta=90


~5 '~ Computation x Full Scale

~k

Figure 2. Pressure coefficients for the Texas Tech Building.

1635
The computed mean pressure coefficients agree very well with both full-scale and wind tunnel data, particularly on the front and back faces where the computed pressure coefficients fall between those measured in the wind tunnel and full scale. The agreement at the top of the building is also good. The computed r.m.s, pressure coefficients agree fairly well with the full-scale and wind tunnel results but tend to be slightly too low over the whole of the building surface. A similar level of agreement is found for flows at other approach angles.

3.2 Mobile Tail E n c l o s u r e

In the calculations, an approach flow for an atmospheric boundary layer with a roughness length of 0.01 m was used. The Jensen Number (structure height / roughness length) is a significant scaling number for wind loads on low-rise structures; in the present case the Jensen Number is about 2500. The Reynolds Number (inertial forces in the flow / viscous forces) is also significant for flow around bodies with curved surfaces. For wind flow around full-scale structures with curved surfaces of large diameter, the Reynolds Number is very large (much larger than can be obtained with wind tunnel tests). For the computations of the present report, the Reynolds Number is effectively infinite, since viscous stresses in the flow are neglected in comparison with the turbulent stresses.

-0.61 '
.o.,,

-0.44
-0.32

0,17

CP,e's

+X+0.6 i

~ I

20interval"s

-0.14

Figure 3. External mean pressure coefficients on the Strarch Mobile Tail Enclosure

1636 Calculations of external pressure coefficients were made for a structure with both ends closed and with one end open, based on the mean wind speed at the top of the structure, Those for both ends closed are shown in Figure 3. The values in Figure 3 were computed at points along the centre-line of the structure. The points of computation were at the centres of the wall or angular roof sectors shown in Figure 3.

The m a x i m u m positive pressure coefficient (+0.69) occurs about halfway up the structure of the lower part of the arched roof. The m i n i m u m (maximum negative) pressure coefficient (-0.61) occurs at the crest of the roof. On the leeward side of the structure, the pressure coefficients have fairly uniform negative values, as expected for a fully separated wake region.
There are no full-scale data available to compare the computed pressures with. However, the computational method is currently being applied to other curved roof structures for which some full-scale and wind tunnel measurements are available.

3.3 Typical Arched-roof Building


The geometric variables that are relevant to the wind loading of arched-roof buildings are the span, the length, the rise, and the height of the eaves. Many different geometrical configurations have been examined by the authors. The results from a typical configuration with a rise/span ratio of 0.2, a length,'~pan ratio of 1.0, and an eaves height/rise ratio of 0.45 are presented here. For the computation the Jensen Number is 500. For a structure height of 10 metres, a Jensen Number of 500 corresponds to a surface roughness length of 0.02 metres. This is equivalent to Terrain Category 2 in the Australian Standard [8].

Figures 4a, 4b and 4c show the computed external pressure coefficientson the building for approach wind angles of 45, 0 and 90 degrees respectively from the normal to the axis of the arch. Because of symmetry, values on one half only are shown for approach angles of 0 and 90 .
For the 0 case, positive pressure coefficients occur on the windward wall and the win~:lward edge of the roof, with negative values over the rest of the structure. The highest magnitude negative values occur just upwind of the apex to the rogf.

1637

-0.04 +0.07 -1.02 -0.67 -0.31 -0.65 -0.63 -0.20 -0.28 -0.29 ,~.+0.22
-0.11

,~i-+0-13 +0.01

a)

-0.72
-0.60

-0.57

-0.52

-0.50

0.50
+0.01

,,

-0.65

0.61/ -0.50

-0,35 -0.29 -0.35

b)
Figure 4. Area-averaged external pressure coefficients for the typical arched-roof building. Approach wind angles are: a) 45 , b) 0 , c) 90 .

1638

+0.40
-1.13

-0.42

+0.27

-0.29

-0.26 -0.39 -0.29

-0.29 -0.33 -0.26

c)

Figure 4 (Cont.) For an approach wind angle of 45", positive pressures only occur near the windward corner of the building. The negative pressures on the roof and walls are generally higher than those obtained for the 0 case, with particularly high suctions occurring along the windward end of the arch roof. Except for the windward wall, negative pressures occur everywhere for the approach wind angle of 90 . Again very high suctions occur along the windward edge of the roof, underneath the separating-reattaching flow. In general, for building8 with length/span ratios of unity or less, the computed values of external pressure coefficients are significantly lower in magnitude than those obtained from the current Australian Standard, especially for the negative pressure coefficients. Agreement is much better at very large length/span ratios.

1639 CONCLUSIONS The method of using computer modelling in place of wind tunnel modelling in predicting wind pressures on building surfaces has been shown to be valid. The accuracy of the predictions of pressures on the Texas Tech building is very good. The use of computer modelling in providing pressure coefficients on a variety of arched-roof buildings has led to predictions that can and have been used in the design of such buildings. We have shown that the methods applied in building codes for arched-roof buildings are good when the length/span ratio of the building is large. But they are very conservative when the length/span ratio of the building is small. ACKNOWLEDGEMENT
The authors wish to acknowledge the permission of Strarch International Ltd to publish some results of the computations carried out on behalf of that Company, and the cooperation of Dr Peter Key during the project.

REFERENCES 1. Levitan, M.L., Holmes, J.D., Mehta, K.C. and Vann, W.P. 1989, 'Fieldmeasured pressures on the Texas Tech Building', 8th Colloq. Indust. Aerodyn., Aachen, W.Germany, 4-7 Sept. 2. Surry, D. 1989, 'Pressure measurements on the Texas Tech B u i l d i n g - wind tunnel measurements and comparisons with full scale', 8th Colloq. Indust. Aerodyn., Aachen, W.Germany, 4-7 Sept. 3. Patankar, S.V., 1980, Numerical Heat Transfer and Fluid Flow, Hemisphere, New York. 4. Paterson, D.A. 1986, Computation of wind flows over three.dimensional buildings, PhD Thesis, University of Queensland, Brisbane.
5. Paterson, D.A. and Apelt, C.J. 1989, 'Simulation of wind flows around threedimensional buildings',Bldg & Environ.,24, 39-50.

6. Gosman, A.D. and Ideriah, F.J.K. 1976, 'TEACH-T: A general computer program for two-dimensional, turbulent, recirculating flows', Department of Mechanical Engineering, Imperial College, London.

1640 7. Hinze, J.O., 1959, Turbulence, an Introduction to itr Mechanism and Theory, McGraw-Hill, New York. 8. Standards Australia, 1989, SAA Loading Code, Part 2 Wind loads. Australian Standard ASl170.2 - 1989.

You might also like