Journal of Wind Engineering and Industrial Aerodynamics 80 (1999) 253262

Wind tunnel velocity proles generated by dierentially-spaced at plates

J.C. Phillips  *, N.H. Thomas , R.J. Perkins , P.C.H. Miller
Department of Applied Mathematics and Theoretical Physics, University of Cambridge, Silver Street, Cambridge CB3 9EW, UK FAST Team, School of Chemical Engineering, University of Birmingham, P.O. Box 363, Birmingham B15 2TT, UK Silsoe Research Institute, Wrest Park, Silsoe, Beds MK45 4HS, UK Received 11 April 1997; received in revised form 30 July 1998; accepted 1 October 1998

Abstract Production of linear shear with low turbulence level in a wind tunnel provides a convenient environment for testing the results of computational uid dynamics simulations and equipment calibration. Boundary layer ow over a at plate at zero incidence provides controlled deceleration of the approach ow according to plate length, with interacting boundary layers between adjoining at plates merging to provide fully developed duct ow. In this way, an array of dierentially-spaced at plates can be used to modify a uniform wind tunnel velocity eld to a specied velocity prole. A one-step iterative scheme is oered to determine plate spacings for simulation of weakly sheared ows, constrained by zero vertical pressure gradient in the downstream ow (representative of boundary layer conditions). The scheme is tested for realisation of uniform shear ow (maximum velocity variation $10% of centreline velocity) by wind tunnel simulation, and produces reasonable results, at least comparable with previous studies. 1999 Published by Elsevier Science Ltd. All rights reserved. Keywords: Wind tunnel; Velocity prole; Flat plates; Iterative scheme; Weak shear

1. Introduction Wind tunnel experiments remain the primary tool for realistic investigation of the interaction of atmospheric boundary layer ows with the built environment [1]. In this context, modication of uniform upstream ow to provide a boundary layer ow which adequately reproduces the full-scale characteristics is usually achieved by
* Corresponding author. Present address: Centre for Environmental and Geophysical Flows, School of Mathematics, University of Bristol, University Walk, Bristol BS8 1TW, UK. E-mail: j.c.phillips@bris.ac.uk.  Present address: Laboratoire de Mecanique des Fluides et dAcoustique, Ecole Centrale de Lyon, 36 avenue Guy de Collongue, BP 163, 69131 Ecully, France. 0167-6105/99/$ see front matter 1999 Published by Elsevier Science Ltd. All rights reserved. PII: S 0 1 6 7 - 6 1 0 5 ( 9 8 ) 0 0 2 0 7 - 4

254

J.C. Phillips et al./J. Wind Eng. Ind. Aerodyn. 80 (1999) 253262

introducing an array of barriers to articially thicken the boundary layer within a reasonable streamwise distance, supplemented by mixing devices to tune the resulting mean and turbulence proles and even turbulence spectra [25]. More exotically, arrays of jets or fans have been used to achieve the same eect [6,7], although necessarily at greater expense, both as equipment and operational costs. However, there are a number of applications which require the production of linear shear ow or weakly sheared ow with low turbulence levels, for which a versatile yet simple experimental method is required. These include linear shear ows for experimental testing of computational uid dynamics predictions and ow measurement equipment calibration, and full-scale investigation of droplet dispersion from agricultural sprays [8,9], where mean shear controls initial droplet removal [10]. We describe here a exible, low cost technique for generating mean velocity proles using an array of parallel at plates to introduce the appropriate momentum decit, and show its application to the benchmark case of weak uniform shear ow. The method described here is a variation on that presented by Lloyd [11], who provided a design procedure based on uniform fully developed channel ow between each pair of plates in the array. His scheme proceeds in a stepwise manner from the wind tunnel oor, specifying the spacing of each pair of plates to produce channel ow with velocity matching the required downstream prole as sole design condition, thus yielding the number of plates required. As a variation on this approach, we oer an alternative design procedure in terms of frictional losses associated with boundary layer development along the surfaces of each plate in the array subject to an overall constraint of zero vertical pressure gradient in the emerging downstream ow. Perhaps the most signicant consideration here is that the plates are thin compared with the boundary layer thickness in order to minimise discrete wake decits in the downstream prole. As a consequence at plates cannot produce high turbulence levels in the downstream ow by vortex shedding, limiting their application to problems with low representative turbulence levels described above. However, the present method is demonstrated to be an adequate approximation for weaker shear ows, requiring a smaller number of plates without risking separation of the upstream approach ow. In Section 2 we outline the design approach and calculations, in Section 3 we report the experimental method and instrumentation, with Section 4 describing the experimental results and nishing with broader discussion and main conclusions in Sections 5 and 6. 2. Design calculations The principle is illustrated in the schematic of Fig. 1, our purpose being to deliver the desired downstream velocity prole using plate spacings within an array of thin plates all of equal length as the design parameter. Neglecting hydrostatic variations, as is usual, and introducing a Fanning friction factor to describe the head loss over each surface, the mechanical energy balance over plate n (see Fig. 1) can be written as follows: p u p u C l((u #u )/2) # " L# L#  L , 2 2 2s L (1)

J.C. Phillips et al./J. Wind Eng. Ind. Aerodyn. 80 (1999) 253262

255

Fig. 1. Schematic diagram of the array of dierentially-spaced at plates in the wind tunnel.

where p is the upstream uniform static pressure, p the downstream static pressure at  L plate n, u the upstream uniform mean velocity, u the prescribed downstream  L velocity at plate n, the uid density, C the friction coecient for ow over plate n, l the plate length and s the spacing between plates n and n#1. L The frictional term in Eq. (1) is based on mean velocity over the plate length ((u #u )/2), assumed representative for velocity modication from u to u . The  L  L friction factor C is assigned according to the Reynolds number at each plate based on this mean ow velocity, following established empirical correlations for laminar and turbulent ows [12]. Re-arranging explicitly in terms of plate spacings s" L p !p u!u \ C l(u #u )  L .  L#  L 8 2 (2)

For present purposes we regard u as prescribed uniform approach ow, u as  L prescribed (uniformly-sheared) exit ow, p as prescribed uniform entry pressure,  p as unknown exit pressure, with s , the plate spacing, as solution variable. Clearly L L we require additional closure constraint to specify p , here being the requirement of L zero vertical pressure gradient in the downstream ow. For weak shear ows considered here, we used the mean downstream pressure dened according to p " p /n as an estimate of constant downstream pressure. N  L An iterative argument was adopted, described in outline as follows. Starting with n plates spaced equally across the control volume (s assigned), the downstream exit L pressure p was calculated for each plate using Eq. (2) and the mean determined. This L was substituted back into Eq. (2) in place of the local value p to specify a new L spacing for each plate in the array based on the prescribed local exit velocity u and L mean downstream pressure. Downstream exit pressures were recalculated from Eq. (2), and a new mean and set of spacings determined. From Eq. (2) it can be seen that plate spacings are simply proportional to the inverse values of pressure and dynamic head dierences, and as the latter term is prescribed, the iteration process simply corresponds to modication of plate spacing by the factor (p !p )/(p !p ), N  L  

256

J.C. Phillips et al./J. Wind Eng. Ind. Aerodyn. 80 (1999) 253262

calculated at each time step. The iteration was repeated until the plate spacings calculated for consecutive steps varied by less than 0.5%. In weak shear simulations, with shear parameter (maximum velocity variation as a proportion of centreline velocity) less than $15% the maximum downstream pressure variation (p / p ) at L  each iterative step was about 10%, determined at the duct walls. The convergence of the scheme is linear, as indicated by the proportionality of plate spacing to pressure head shown by Eq. (2). Convergence of the iteration scheme is characterised by approximately constant static pressure conditions both upstream and downstream of the plate array, with changes in dynamic pressure to match the prescribed shear balanced by changes in drag loss due to variation in plate spacing. In assuming this behaviour, we note the limitation of this technique to simulation of weakly sheared ows where the upstream static pressure remains uniform across the plate array, and dependence on the number of plates specied. In principle, it is possible to increase the range of ow shear simulated by using streamlined plates to minimize blockage of the approach ow, maintaining uniform upstream static pressure, and by using plate length as a variable to allow greater drag loss. The calculation scheme was compared with that due to Lloyd [11] for production of a weak nominally uniform shear, characterised by shear parameter (mean velocity dierence at the tunnel wall expressed as a fraction of the centreline velocity) of $10%. Lloyds scheme calculates plate number explicitly, and so was solved for a range of shear parameters to provide an estimate of the number of plates required for weak shear simulations (Fig. 2). Both schemes were solved numerically for 19 plates 305 mm long and 2 mm wide with upstream approach ow velocity of 10 m s\, yielding plate spacings shown in Fig. 3. The spacings calculated for the present iterative scheme show a uniform increase across the wind tunnel, whereas spacings calculated using Lloyds method increase at an increasing rate near the wind tunnel top. This region is most poorly reproduced in Lloyds original uniform shear simulation [11], with plate spacings showing similar trends at the tunnel top.

3. Experimental method The experiments were conducted in an 18 square section open-circuit NPL-type wind tunnel (windspeeds up to 15 m s\), working section length 3 m [13], in which the plate array was located just downstream of the entry section. Plate arrays were constructed from dural plates 305 mm long and 2 mm thick (Sharpe and Fisher Ltd.) using spacings calculated for the two schemes above (see Fig. 3). A mixing grid (cell dimension 11 mm;11 mm, bar diameter 1 mm) was added immediately downstream of the plate array to allow velocity prole relaxation. Quality of the downstream ows was gauged by measuring the mean velocity proles using a Dantec 55P11 single hot-wire probe in conjunction with 55M01 hot-wire anemometer mounted on a 3-D stepper motor-driven traverse 1.3 tunnel heights downstream of the plate array. The probe was traversed across the tunnel height at a spanwise position corresponding to the tunnel centreline. At each measuring station, 300 voltage readings were taken,

J.C. Phillips et al./J. Wind Eng. Ind. Aerodyn. 80 (1999) 253262

257

Fig. 2. The number of plates required to simulate a uniform ow with specied shear parameter (maximum velocity variation as a proportion of centreline velocity), calculated using Lloyds scheme [11].

Fig. 3. Comparison of plate spacings calculated for simulation of weak shear ow (maximum velocity variation $10% of centreline velocity) using the present iterative scheme and Lloyds scheme [11].

using the ASYST data acquisition package sampling at 500 Hz. These were converted to velocity data within the ASYST package using probe calibrations against pitotstatic measurements taken using a Betz manometer.

258

J.C. Phillips et al./J. Wind Eng. Ind. Aerodyn. 80 (1999) 253262

4. Experimental results Mean streamwise velocity proles are shown for the present iterative scheme (Fig. 4) and Lloyds design calculation (Fig. 5). The eects of plate wakes are manifest as velocity defects in the downstream ow. Fitting trends in these data by eye suggests

Fig. 4. Mean velocity prole produced by at plates specied using the present iterative calculation scheme to simulate weak shear ow (maximum velocity variation $10% of centreline velocity).

Fig. 5. Mean velocity prole produced by at plates specied using Lloyds calculation scheme [11] to simulate weak shear ow (maximum velocity variation $10% of centreline velocity).

J.C. Phillips et al./J. Wind Eng. Ind. Aerodyn. 80 (1999) 253262

259

production of shear ows with parameter about $9% for both schemes. Addition of the mixing grid enabled the downstream velocity prole to satisfactorily relax, although reducing the nal shear parameter simulated. This is shown in Fig. 6 for the present scheme, with reduction in nal shear parameter to $8.6%, determined using N linear regression. Streamwise turbulence intensity ((u/u) was calculated from the velocity data, shown in Figs. 7 and 8 for the present iterative scheme and Lloyds calculation, respectively. For both schemes, the turbulence intensity level produced was around 3% over the tunnel depth, broadly in accordance with results obtained by Lloyd [11] (recall introduction). Peak values of 10% at the walls simply indicate the presence of wall boundary layers in the upstream ow. A smoke tracer (CFT generator, Shell Ondina oil) was used to visualise ow through the plate array, shown in Fig. 9 for the present scheme. For both calculation schemes with and without mixing grid attachment, smoke streams followed approximately horizontal trajectories, indicating approximately zero vertical pressure gradient in the downstream velocity eld.

5. Discussion Use of at plates for mean velocity prole simulation is only suited to applications which require production of weak shear and low turbulence levels, and production of uniform shear still provides a useful reference case in wind tunnel experiments. Within this framework, a simple calculation scheme is attractive. The scheme oered here consists of a single-step iteration which requires the minimum of computation. Given the limited 2-D view of ow dynamics adopted here, there will be a range of convergent solutions for the calculation scheme. For representative weak shear simulation, the

Fig. 6. Mean velocity prole produced by at plates specied using the present iterative calculation scheme to simulate weak shear ow (maximum velocity variation $10% of centreline velocity) with mixing grid added.

260

J.C. Phillips et al./J. Wind Eng. Ind. Aerodyn. 80 (1999) 253262

Fig. 7. Streamwise turbulence intensity prole produced by at plates specied using the present iterative calculation scheme to simulate weak shear ow (maximum velocity variation $10% of centreline velocity).

Fig. 8. Streamwise turbulence intensity prole produced by at plates specied using Lloyds calculation scheme [11] to simulate weak shear ow (maximum velocity variation $10% of centreline velocity).

mean downstream static pressure appears to provide a reasonable estimate of the convergent solution, typically within 10% of the local downstream pressure at any point. Modication of plate spacing is subject to maintaining uniform upstream pressure. This condition will not hold for stronger shear, where small plate spacing will lead to blockage of the upstream approach ow. However, use of at plates as outlined

J.C. Phillips et al./J. Wind Eng. Ind. Aerodyn. 80 (1999) 253262

261

Fig. 9. Smoke visualisation of ow produced by at plates specied using the present iterative calculation scheme. Side view of the at plate array in the wind tunnel with ow direction left to right, showing negligible deection of the smoke tracer indicating zero vertical pressure gradient in the downstream ow.

here may prove a useful platform from which to enhance the range of simulated conditions to higher levels of turbulence and shear. Lloyd [11] added turbulence generating barriers to at plates in an ad hoc manner. We suggest that, with barriers sized to provide desired turbulence characteristics, empirical determination of the Fanning friction coecient for a plate-plus-barrier arrangement will allow direction incorporation into the iterative scheme. Addition of grids to tune turbulence characteristics may prove more attractive than barriers given detailed treatment in the literature, with incorporation of their additional ow resistance into the friction term as before. However, we again caution the requirement of constant upstream pressure for convergence. Simulation of stronger shear may be approached by allowing plate length to vary independently of plate spacing, although this is beyond the scope of our preliminary investigation. Perhaps the biggest weakness of the present iterative scheme is not calculating the plate number required for a given shear, although the graphical correlation presented as Fig. 2 provides a reasonable estimate. Although inferior to Lloyds scheme in this respect, the present iterative scheme does explicitly simulate zero vertical pressure gradient (suitable for boundary layer applications) and appears to be more reliable in uniform weak shear simulation (Fig. 3). The increasing plate spacing near the tunnel top appears to be characteristic for Lloyds design scheme [11] although it is not made clear why this is the case. The eect of plate wakes as velocity defects in the downstream ow is undesirable, although at only 1.3 tunnel heights downstream the measuring station is closer to the array than is standard for wind

262

J.C. Phillips et al./J. Wind Eng. Ind. Aerodyn. 80 (1999) 253262

tunnel ow experiments [1]. As demonstrated here, any uniform grid will allow the velocity defects to relax, with grid dimensions eecting down-stream distance over which this occurs. Streamlined plates would also minimize this eect. 6. Conclusions A calculation scheme has been developed to specify the spacings within an array of at plates to simulate weak shear ows with zero vertical pressure gradient, producing results comparable to the earlier design calculation of Lloyd [11]. An estimate of plate number to create a given shear ow is required, provided here graphically in Fig. 2. The balance of dynamic head to frictional loss in ow between any pair of plates in the array allows this scheme to be extended to at plates with turbulence generators attached via an empirically determined friction coecient. Restriction to weak shear is a consequence of requiring uniform upstream static pressure; this may be maintained for more strongly sheared proles by introduction of plate length as an independent variable. Acknowledgements This work was supported by Agriculture and Foods Research Council (now the Biotechnology and Biological Sciences Research Council) grant LRG 103. We are grateful to Mark Davidson for useful discussion and technical assistance. References
[1] T.V. Lawson, Wind Eects on Buildings. Part 1: Design Applications, Applied Science Publishers, UK, 1980. [2] J. Armitt, J. Counihan, The simulation of the atmospheric boundary layer in a wind tunnel, Atmos. Environ. 2 (1968) 4971. [3] J. Counihan, An improved method of simulating an atmospheric boundary layer in a wind tunnel, Atmos. Environ. 3 (1969) 197214. [4] N.J. Cook, Wind tunnel simulation of the atmospheric boundary layer by roughness, barrier and mixing device methods, J. Wind Eng. Ind. Aerodyn. 3 (1978) 157176. [5] H.P.A.H. Irwin, The design of spires for wind simulation, J. Wind Eng. Ind. Aerodyn. 7 (1981) 361366. [6] H.W. Teunissen, Simulation of the planetary boundary layer in a multiple jet wind tunnel, Atmos. Environ. 9 (1975) 154174. [7] A. Nishi, H. Miyagi, Computer-controlled wind-tunnel for wind-engineering applications, J. Wind Eng. Ind. Aerodyn. 54 (1995) 493504. [8] P.C.H. Miller, C.J. Mawer, C.R. Merritt, Wind tunnel studies of the spray drift from two types of agricultural spray nozzle, Aspects Appl. Biol. 21 (1989) 237238. [9] N.M. Western, E.C. Hislop, P.J. Herrington, E.I. Jones, Comparitive drift measurements for BCPC reference nozzles and for an Airtec twin-uid nozzle under controlled conditions, Proc. British Crop Protection Conf., Weeds, 1989, pp. 641648. [10] J.C. Phillips, P.C.H. Miller, Field and wind tunnel measurements of the airborne spray volume produced by single at fan nozzles, J. Agric. Eng. Res., accepted. [11] A.R.J. Lloyd, The generation of shear ow in a wind tunnel, Q. J. Roy. Met. Soc. 93 (1967) 7996. [12] H. Schlichting, Boundary Layer Theory, 6th ed., McGraw-Hill, New York, 1968. [13] M.J. Davidson, W.H. Snyder, R.E. Lawson, J.C.R. Hunt, Wind tunnel simulation of plume dispersion through groups of obstacles, Atmos. Environ. 30 (1996) 37153731.