Sei sulla pagina 1di 13

International Journal of Dynamics of Fluids ISSN 0973-1784 Volume 4, Number 1 (2008), pp. 1 12 Research India Publications http://www.ripublication.com/ijdf.

htm

CFD Supported Examination of Scaling Issues for Open Wind Tunnel Model Testing
Nadir Yilmaz1,* and A. Burl Donaldson2 Department of Mechanical Engineering, New Mexico Institute of Mining and Technology, Socorro, NM 87801, USA. * Corresponding Author Email: yilmaznadir@yahoo.com 2 Department of Mechanical Engineering, New Mexico State University, Las Cruces, NM 88003, USA. E-mail: bdonalds@nmsu.edu Abstract This paper presents an examination of the velocity scaling question for model testing in an open wind tunnel and utilizes CFD calculations to support conclusions which are drawn. The application considered presently is the inner courtyard, or patio area of an actual structure which is subject to free stream wind, and the parameter of interest is the resulting air velocity and pattern at various locations in the courtyard area. To support this study, data for the full scale was collected. Similarly, wind tunnel testing was conducted at free stream velocities which bracketed a typical wind velocity range and the velocities at the comparable 5 locations in the recirculation region were measured. A 1:93 scale model was used for the wind-tunnel model experiments. The CFD code Star-CD was used to predict air velocities at each of these 5 locations for both full scale and model scale subject to their respective free stream velocities. Conclusions reached are that over a representative range of velocities, the model measurements provide a meaningful reflection of the full scale behavior, so that wind tunnel testing can be performed at the same free stream velocity as the full scale case. Results also show that the CFD code Star-CD provides a high quality predictive capability for both scales. Keywords: CFD Modeling, Velocity Field Measurement, Scaling, WindTunnel Simulation.
1

Introduction
Open, low speed wind tunnels, just as the name implies, typically draw ambient air into a test section by a suction fan at the tunnel exhaust. Wind speeds are generally sufficiently low that compressible effects are negligible (M < 0.3), but they are useful

Nadir Yilmaz and A. Burl Donaldson

in the study of incompressible flow around scale models, and to determine drag coefficients and flow streamlines. Because the wind tunnels are open, then the properties of the air are fixed, and the test section pressure is not much lower than ambient. Hence, there is little opportunity to match Reynolds numbers between model and full size structures because the only available parameters are velocity multiplied by a dimension. So, if the model is dimensionally scaled from full size, say 50:1, then to match Reynolds number, the velocity would need to be increased by a factor of 50. This velocity multiple generally would not be available in an open wind tunnel, unless the full scale velocity of interest is order of 1-2 ft/s. Model testing is performed to indicate how the full scale case will respond to a stimulus without undergoing the expense and time which would be required for large scale parametric testing. For the case of open wind tunnel testing of a building design, the issue of concern is how the free steam velocity should be scaled in order that measurements made with a scale model are representative and useful. This issue was encountered in consideration of the recirculation zone in an open courtyard area of a structure (see Figure 1) which is subject to prevailing winds. The ultimate application of this study is the consideration of wind barriers which might be installed on the actual structure for the purpose of reducing wind in the courtyard area.

Figure 1: A Model of the structure under study

Flay and Teunissen [1] compared wind tunnel measurements with full-scale measurements taken around a small suburban airport. The wind-tunnel measurements were carried out in a boundary-layer wind tunnel with a 1:1920 scale model of the airport. For most of the dependent variables measured, the agreement was good but at very low pressure, the mean velocity measurement in the simulation was less than for full scale. Another comparison between wind-tunnel and full-scale was made by Robertson and Moran [2]. Full-scale measurements were done on a canopy roof structure. Wind-tunnel tests were carried out using two wind tunnels. There were discrepancies not only between full-scale and wind-tunnel data but also between different wind tunnels in which the same model was used and between the same wind

CFD Supported Examination of Scaling Issues

tunnel in which models with different scales were used. Utsunomiya et al. [3], however, made a set of experiments on a full-scale structure and in a wind-tunnel, which showed good agreement overall. It was concluded that it was necessary to determine the gradient-wind direction in the full-scale situation to make more accurate comparisons. Richards and Hoxey [4] compared results from wind-tunnel experiments and CFD modeling of a specific type of structure with full-scale data. In their work, computational predictions obtained by a CFD package showed reasonable agreement with the full-scale measurements but wind-tunnel models showed some scatter for low pressure regions. Oliveira and Younis [5] studied turbulent flows around full scale structures. They used data to validate the uncertainties associated with the use of CFD. This information was used to predict wind loads for full-scale buildings. A CFD analysis was used by Thiis [6] to predict snowdrifts around buildings and it was recommended that computational models should be used in planning building design. In work by Hoxey et al. [7], pressure comparisons were made between full scale measurements and wind tunnel measurements for a 6 m cube. The results were found to be in good agreement with one exception which was attributed to location of flow reattachment. For a building which was instrumented for pressure measurement, Sterling et al. [8] noted that the actual air flow was highly unsteady. Wind tunnel studies at a 1:50 scale were used by Bienkiewicz and Hamm [9] to study the wind induced point pressure on a building. A mismatch between the wind tunnel and actual data was attributed to the fluctuating pressure in the actual case. In a paper by Yilmaz and Donaldson [10], the accuracy and validity of the CFD modeling of a structure with various wind screens was established based on the comparison of the point-wise hot-wire anemometer measurements in the wind-tunnel. One of the main purposes was to evaluate the effectiveness of the wind screens experimentally and computationally and to find the best screen with respect to the economical, aesthetic and engineering considerations. The current work presents velocity field measurements in the inner courtyard of Faywood Hot Springs Resort for the full scale case and an open wind tunnel model; corresponding CFD predictions were made for a given free stream wind velocity for both cases. The CFD calculations were carried out using the Star-CD code. The concern deals with selection of an appropriate free stream velocity for wind tunnel model testing which will provide a useful reflection of velocity for the full scale case. In the comparison of experimental data for the full scale and model, it becomes readily apparent that with prevailing wind, the speed and direction are highly variable; for the wind tunnel, these conditions are constant. Therefore, it should not be anticipated that the conditions for the full scale case and model case can be made the same because the forcing function for the two cases is different.

Full Scale Results


Portable hot wire anemometers were utilized during a windy day to measure both free stream velocity and velocity in the inner courtyard of the structure at various times. Schematic representation of locations of measurements and the experimental data for these 5 locations are shown in Figure 2. (After the data was collected and compared

Nadir Yilmaz and A. Burl Donaldson

with comparable case CFD calculations discussed in the next paragraph, it was determined that the hot wire anemometers were out of calibration. So, each of the anemometers had to be calibrated after-the-fact, and the data shown in this figure includes the indicated corrections.) The free stream wind velocity varied up to more than 7 m/s, but because only 3 hot wire anemometers were available, not all of the free stream velocities show a corresponding data point at the individual interior locations.
2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 0 2 4 6 Free Stream Velocity [m/s] 8 Velocity at Center [m/s]
0.5 Velocity at North [m/s] 0.4 0.3 0.2 0.1 0 0 1 2 3 4 5 Free Stream Velocity [m/s]

(a)
Free Stream
0.7 Velocity at West [m/s] 0.6 0.5 0.4 0.3 0.2 0.1 0 0

(b)

North West East

Center South

2 4 6 Free Stream Velocity [m/s]

(d)
1.6 Velocity at South [m/s]
Velocity at East [m/s] 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0 1 2 3 4 Free Stream Velocity [m/s] 5

1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 0 1 2 3 4 Free Stream Velocity [m/s] 5

(c)

(d)

Figure 2: Schematic representation of locations of full scale experimental data for (a) Center, (b) North, (c) South, (d) West, (e) East

The computational study was carried out using Star-CD, a commercial computational fluid dynamics software package. Experimental evidence indicated that velocity fluctuates over time at all of the measured locations. Therefore, the k-

CFD Supported Examination of Scaling Issues

turbulence model was used for the numerical modeling. Relative mesh size was changed based on the locations of interest between 0.5 % and 10 % with tetrahedral elements in order to achieve an adaptive grid. The CFD generated velocity contour plot was made for a free stream velocity of 3 m/s, which was representative (approximate average) of the collected data, and is shown in Figure 3. The velocity distribution was taken at an elevation of 1.5 m above ground level for the purpose of comparing it to velocity field measurements for which hot-wire anemometer probes were also located at 1.5 m above ground level.

Figure 3: Velocity contour plot in courtyard for free stream velocity of 3 m/s

Based on the line fitting of Figure 2, CFD predictions and the experimental data were found to be in fairly good agreement. The velocity ratio (the ratio of the velocity in the courtyard to the mean free stream velocity) can also be found from the plot. The slope of the line fitted to the data is 0.18 for the center location. The slope of the fitted line can be used to calculate the average velocity at a specific point in the courtyard corresponding to a free stream velocity which is in the range of the experimental data. In other words, if the free stream velocity is 3 m/s, then the velocity in the center location of the courtyard is expected to be around 0.54 m/s. The CFD result for the center point also shows that the velocity in the center location of the courtyard is about 0.52 m/s for the free stream velocity of 3 m/s. In the north, there are less data points than the center position, but the results are similar. The majority of the data points are located around the fitted line while a few points stray from the line. From this plot, the velocity ratio is 0.05. In the south, the collected data fit well to the line. The velocity ratio for the south location is 0.17. In the east, the majority of the points fall near the fitted line with some departures noted. The velocity ratio at the east location is 0.079. Finally, in the west region, not as many data points fall along the fitted line. A larger number of data points deviate

Nadir Yilmaz and A. Burl Donaldson

from the line in comparison to the other locations. Even with this observation, the velocity ratio is still close to the other values. The velocity ratio for the west location is 0.057. In addition to the line fitting approach, it is important to make a comparison based on experimental uncertainties due to some scatter in measurements. Figure 4 shows a comparison of the CFD calculations to the experimental data with uncertainty analysis. Free stream and court yard velocities were averaged over time and standard deviations for each measurement location was found. Because the number of hot-wire anemometers was limited and the measurements were collected at different times for certain locations, the averaged free stream velocities are not the same. However, in order to make a valid comparison, CFD analyses were done for each of the averaged free stream velocities based on the experimental data. There are a few data points which compare poorly, but overall the agreement between CFD prediction and the experimental data is fairly good and CFD results stay within the deviation limits.
1.0 Courtyard Velocity [m/s]
Center Experimental Data CFD

0.8 0.6 0.4 0.2


South East North West

0.0 1.8 2.0 2.2 2.4 2.6 2.8 Free Stream Velocity [m/s]

Figure 4: Comparisons between CFD predictions and averaged experimental data points. Bar limits indicate one standard deviation about the mean.

In comparison of CFD predictions to full-scale measurements, overall results show good agreement. Although there are some regions in which CFD predictions and average experimental data are not close, explanations will be given later to elaborate this issue.

Wind Tunnel Model Results


The issue of velocity scaling was considered by comparing wind tunnel measurements to CFD predictions. Wind tunnel testing provides for a steady free stream so that transients in velocity and flow direction are not experienced. Therefore, it was possible to select two velocities for testing and compare results on the basis of

CFD Supported Examination of Scaling Issues

percentage reduction in the courtyard area velocity as compared to the free stream for each case. For the wind-tunnel model experiments, a 1:93 scale model was used. The experimental model was run with a free stream velocity of 1.22 m/s (Re =15,000 based on the inner diameter of the courtyard) and again at a free stream velocity of 12.2 m/s (Re =150,000 based on the inner diameter of the courtyard); an increase in free stream velocity by a factor of 10. The results from the test with 1.22 m/s are summarized in Table 1 and comparisons can be made to the results shown in Table 2 for a wind speed of 12.2 m/s.

Table 1: Experiment-CFD comparison for no wind screen, 4 ft/s (1.22 m/s) wind speed Court yard Position Center North East South West Minimum Maximum Velocity Velocity (m/s) (m/s) 0.24 0.43 0.015 0.13 0.10 0.24 0.08 0.40 0.04 0.32 Average Experimental Velocity (m/s) 0.32 0.06 0.17 0.21 0.15 CFD CFD Predicted Percent Velocity (m/s) Variation 0.28 12.34 0.17 206.01 0.21 24.17 0.23 10.43 0.21 35.26

Table 2: Experiment-CFD comparison for no wind screen, 40 ft/s (12.2 m/s) wind speed Court yard Position Center North East South West Minimu m Velocity (m/s) 2.40 1.10 2.00 2.40 1.00 Maximum Velocity (m/s) 3.30 2.00 3.10 3.30 1.90 Average Experimental Velocity (m/s) 2.84 1.40 2.55 2.80 1.36 CFD Predicted Velocity (m/s) 2.80 1.69 2.10 2.30 2.07 CFD Percent Variation 1.41 20.71 17.65 17.86 52.21

In comparison of the experimental values in Table 2 for the east and west case, symmetry is not found. Without resorting to a significant investigation of this observation, it is tentatively concluded that flow patterns within the wind tunnel during this test were influenced by vagaries in the entering room air due to room HVAC system circulation. Also to aid with the comparison of full scale results to wind tunnel results, a CFDgenerated contour plot for the case of 1.22 m/s is shown in Figure 5 (a), and the contour plot for the case with 12.2 m/s, shown in Figure 5 (b).

Nadir Yilmaz and A. Burl Donaldson

Figure 5: CFD results for no wind screen with free stream velocity of (a) 1.22 m/s (b) 12.2 m/s

For the two wind speeds considered experimentally in the wind tunnel (which differed by a factor of 10, i.e., 1.22 m/s and 12.2 m/s), the inner courtyard average (at 5 locations) velocity at the slower speed was 8.1% of that for the higher speed. In other words, the velocity scaling would have been perfect at 10%; the value of 8.1% indicates the scaling is quite good. The reason the lower velocity case not being precisely 10% of the higher velocity case might be due to the error in the both experiments and the poor reading capability of one of the hot-wire anemometers at low wind speeds. Furthermore, in the area where there is circulation, the hot-wire anemometer readings may have been influenced by directional effects. However, according to CFD-generated contour plots, the slower wind speed case corresponded to precisely 10% of the higher wind speed case for the courtyard average velocity; this is apparent in both contour plots. Since the free stream velocity was reduced by a factor of 10 and this reduction resulted in a decrease in the average court yard velocity by a factor of ten (approximately), it can be concluded that the courtyard velocity is directly proportional to the mean stream velocity for this flow regime. In other words the courtyard velocity is a fixed percentage of the free stream velocity.

Comparisons between Full Scale, Wind Tunnel Model, and CFD Predictions
As a final comparison, Figure 6 shows wind-tunnel results, full-scale measurements and CFD predictions. To put all of the velocities on the same basis, the inner courtyard velocities were normalized by the free stream velocity. As seen in the figure, CFD, wind-tunnel and full-scale results show excellent agreement for center and south locations. The greatest inconsistency occurs at the west region. Overall, CFD and wind-tunnel results are in better agreement than CFD and full-scale results. The reason for this is that wind-tunnel experiments are better controlled than field measurements, as expected.

CFD Supported Examination of Scaling Issues


0.6 0.5 Velocity [m/sec] 0.4 0.3 0.2 0.1 0 North East Center South West Full Scale CFD Wind Tunnel

Figure 6: Comparison of average full-scale and wind-tunnel measurements to CFD values for a free stream velocity of 3 m/s

The percent variation of CFD predicted values with respect to wind-tunnel and full-scale measurements for center, north, south, east and west are shown in Figure 7.
70 Percent Variation 60 50 40 30 20 10 0 North East Center South West

CFD vs Full-Scale Measurements CFD vs Wind-Tunnel Data

Figure 7: CFD percent variation with respect to full-scale measurements and windtunnel data

For center and south, the percent variation for CFD vs. full-scale and CFD vs. wind-tunnel is quite small. Overall, the smallest variation occurs between CFD and the wind-tunnel results because of controlled conditions in the wind-tunnel as compared to open air with a full-scale structure. Moreover, hot-wire anemometers are direction dependent devices and if the anemometer thermal element is not directly perpendicular to the flow stream, an error can easily occur. Also, wind-tunnel tests can be performed several times to test repeatability of the measurements. However,

10

Nadir Yilmaz and A. Burl Donaldson

full-scale measurements depend on the weather conditions and occurrence of high wind speeds in the southwest are not highly predictable except during the spring season, and even then, there is significant fluctuation in wind speed and direction. The greatest discrepancy occurs in the west region for both CFD vs. wind-tunnel and CFD vs. full-scale measurements and between CFD and full-scale measurements at north and east regions. This discrepancy can be explained with the assistance of the CFD-generated velocity vector plot shown in Figure 8.

(a)

(b)

Figure 8: Velocity vector plots inside courtyard for 3 m/s free stream velocity: a) Plan view, b) Side view from 45 degree angle with free stream vector approaching from right

As seen in the figure, there is high circulation in west and east sides and the vectors have a vertical component. And because the anemometers are directiondependent devices, velocity measurements can be expected to have higher uncertainty when approach direction is uncertain, as in these regions compared to the other regions. In the north, the reason for the higher deviation of full-scale measurements from CFD predictions is that the lowest velocity occurs in this region so uncertainty in the hot-wire anemometer measurements may be greater. To emphasize the problem encountered in comparison of CFD results to the full scale data, consider again the variable direction of the free stream wind. The CFD calculation presumes that the direction is constant; in an actual wind, the direction is variable, just as the speed is variable. This observation will have an obvious impact when comparisons are made.

Summary and Conclusions


The current work presents velocity field measurements in the inner courtyard of Faywood Hot Springs and the validation of the CFD model for both a full-scale case and a wind tunnel model case. The CFD simulation was carried out under relevant experimental conditions for each case. The velocity field measurements are taken at five locations (Center, North, South, West and East).

CFD Supported Examination of Scaling Issues

11

For the full scale case, in center and south positions, velocity field measurements show fair agreement with the CFD predicted velocity values. The greatest discrepancy between experimental data and numerical prediction is observed in the west position. This discrepancy was attributed to the accuracy of the hot-wire anemometers (directional sensitivity was not recognized during measurements). In fact, there is a high circulation area around the west region. Since hot-wire anemometers are direction-dependent devices, unless the wind direction is steady and known, it is not easy to maintain proper orientation of the probe. For the wind-tunnel model experiments made with a 1:93 scale model, the measurements showed good agreement for center and south regions with CFD. However, unlike the full-scale data, good agreement with the CFD results was found in north and east regions. This was due partly to the fact that the wind-tunnel experiments can be controlled and repeated as contrasted to natural events. The finding that velocity for wind tunnel testing did not need to be scaled with dimension in order to simulate full scale behavior has obvious limitations and cannot necessarily be extrapolated to the general case. However, in considering various flow phenomena, modest changes in Reynolds number do not necessarily have a profound effect on flow patterns. So for this case, the velocity in recirculation regions is found to be primarily a function of the free stream velocity, the shape of the structure, and the location within the structure. Another observation which can be made from this study is that CFD software packages are highly useful and although they may not totally replace model testing, can be viewed as at least an important supplement in considering outcomes of large scale events without the cost and time associated with large scale testing.

References
[1] Flay, R. G. J., and Teunissen, H. W., 1983, Comparison of simulated and full-scale wind structure over a small suburban airport, J. Wind Eng. Ind. Aerodyn., 15(1-3), pp. 65-76. Robertson, A. P., and Moran, P., 1985, Comparisons of full-scale and windtunnel measurements of wind loads on a free-standing canopy roof structure, J. Wind Eng. Ind. Aerodyn., 23(1-3), pp. 113-125. Utsunomiyai, H., Nagao, F., and Yoshimura, S., 1989, Wind-tunnel and fullscale comparisons on the change of local wind characteristics due to an open cut, J. Wind Eng. Ind. Aerodyn., 32(1-2), pp. 51-61. Richards, P. J., and Hoxey, R. P., 1992, Computational and wind tunnel modeling of mean wind loads on the Silsoe structures building, J. Wind Eng. Ind. Aerodyn., 43(3), pp. 1641-1652. Oliveira, P. J., and Younis, B. A., 2000, On the prediction of turbulent flows around full-scale buildings, J. Wind Eng. Ind. Aerodyn., 86(2), pp. 203-220. Thiis, T. K., 2000, A comparison of numerical simulations and full-scale measurements of snowdrifts around buildings, Wind Struct. Int. J., 3(2), pp. 73-81.

[2]

[3]

[4]

[5] [6]

12 [7]

Nadir Yilmaz and A. Burl Donaldson Hoxey, R. P., Richards, P. J., and Short, J. L., 2002, A 6 m cube in an atmospheric boundary layer flow, Part 1, Wind Struct. Int. J., 5(2-4), pp. 165176. Sterling, M., Baker, C. J., and Hoxey, R. P., 2003, Short term unsteady loading on a low-rise building, Wind Struct. Int. J., 6(5), pp. 403-418. Bienkiewicz, B., and Ham, H. J., 2003, Wind tunnel modeling of roof pressure and turbulence effects on the TTU test building, Wind Struct. Int. J., 6(2), pp. 91-106. Yilmaz, N., and Donaldson, A. B., 2006, "Consideration of wind barriers for an inner courtyard," Architectural Engineering and Design Management, 1(4), pp. 281-293.

[8] [9]

[10]

Potrebbero piacerti anche