You are on page 1of 9

Author version of the paper published in: Int. J. Environment and Pollution, Vol. 44, Nos.

1/2/3/4, 2011, pp. 359-367 359

Simulation of flow in an idealised city using various CFD codes


Istvn Goricsn, Mrton Balcz, Mikls Balogh, Kroly Czder, Anik Rkai, Csilla Tonk
Department of Fluid Mechanics, Budapest University of Technology and Economics (BME), 1111, Bertalan L. 4-6., Budapest, Hungary * Corresponding author, balczo@ara.bme.hu

Abstract: In the framework of the COST 732 action several teams run models to verify the proposed Best Practice Guideline (Franke et al. 2007). The data set concerns the MUST experiment, where wind field and dispersion measurements were conducted for an array of obstacles as a full scale experiment, as well as in a wind tunnel. The pollutant transport research group at the Department of Fluid Mechanics of the BME started simulations of flow within this idealised city using two commercial CFD codes, FLUENT and MISKAM. The paper presents results of the flow simulations in the 0 degree MUST case through velocity profiles, contour plots and validation metrics.
Keywords: atmospheric dispersion, CFD, MISKAM, FLUENT, validation Reference to this paper should be made as follows Goricsan et al. (2011) Simulation of flow in an idealised city using various CFD codes, Int. J. Environment and Pollution, Vol. 44, No. 1/2/3/4, pp. 359-367. Biographical notes: Istvn Goricsn received his PhD in Mechanical Engineering at the Budapest University of Technology and Economics (BME). His research interests include issues related to dispersion in atmospheric boundary layers, wind tunnel measurement of wind loads acting on buildings and structures, as well as pollutant transport related CFD. Mrton Balcz graduated as mechanical engineer 2001 from the BME and the University of Karlsruhe and is research assistant at the Department of Fluid Mechanics. His research area is pollutant transport in urban environment. Mikls Balogh received his diploma degree in meteorology at Etvs Lornd University, Faculty of Science. Since 2006 he is doctoral candidate at the Department of Fluid Mechanics, focussing on CFD code development. Kroly Czder, Anik Rkai and Csilla Tonk are undergraduates at the BME. They study Environmental Engineering and made a contribution to the simulations performed in the framework of COST 732.

Introduction

Urban air pollution is becoming one of the most important issues in environmental problems in cities. To determine the wind field and the concentration distribution in urban micro-scale areas (length scale up to 5 km) it is necessary to use obstacle resolving methods. Traditional solution is the physical modelling in

Author version of the paper published in: Int. J. Environment and Pollution, Vol. 44, Nos. 1/2/3/4, 2011, pp. 359-367 360 wind tunnels, but numerical simulation (CFD) can also be used for this purpose. Different requirements at the environmental regulatory side (e.g. large amount of modelled buildings, fast model preparation, limited simulation time and computer capacity) make simpler methods necessary, at the expense of more limited accuracy. The micrometeorological flow and dispersion models fulfil the mentioned requirements. The COST Action 732 is addressed to the improvement and quality assurance of micro scale obstacle accommodating meteorological models and their application to the prediction of flow and transport processes in urban or industrial environment. Within the framework of this COST action an exercise is performed where several groups run models applying the Best Practice Guideline for the CFD Simulation of Flows in the Urban Environment (Franke et al., 2007) to evaluate the models against the measurements. The experimental data set comes from the Mock Urban Setting Test (MUST) where wind field and dispersion experiments were conducted for several wind directions in a 10x12 array of 120 standard 40-feet containers (LWH = 12.2x2.4x2.54m) settled in flat terrain, thus simulating an idealised city (see Figure 1). A detailed description of the test is given by Yee and Biltoft (2004). The experiment was also measured in a wind tunnel and an extensive validation dataset was produced by Leitl et al. (2007). The pollutant transport research group at the BME performed simulations of flow within this idealised city using two commercial CFD codes. MISKAM (Eichhorn, 2003) is a special code for simulating flow and dispersion processes in urban environment, widely used for regulatory purposes in Europe, while FLUENT is a dominant, general purposed CFD code. This paper presents the results of flow simulations in the 0 degree MUST case, which means that the flow is perpendicular to the longer side of the containers (see Figure 1). Results are compared to the wind tunnel measurements and several metrics are determined for validation.

Used models, grids and simulation parameters

Both FLUENT and MISKAM solve the Reynolds-averaged Navier-Stokes equation using turbulence models to take the unsteady effect of turbulence into account. Table 1 summarizes all important parameters of the codes and of the simulations. The FLUENT simulations ran on the same grid in full scale (F-FS) and wind tunnel scale (F-WTS) and with different reference velocities. Because MISKAM uses a non-equidistant Cartesian grid, from which buildings are simply blocked out, in case of the coarse MISKAM grid (M-1) both the position and the size of the containers had to be adapted. In case of two finer MISKAM grids (M-08, M-05) only the position had to be adapted due to the Cartesian grid. The inlet boundary conditions of FLUENT were taken from wind tunnel measurements. MISKAM generated its own inlet profile by the assumption of an equilibrium boundary layer for a given reference velocity Uref at a given height Zref, resulting in a logarithmic velocity profile.

Goricsan et al. (2011): Simulation of flow in an idealised city using various CFD codes
Table 1 Model and grid description, boundary conditions and simulation parameters
F-WTS-1 F-WTS-10 F-FS-1 F-FS-10 M-1 M-08 M-05

361

Simulation name Model description numerical code used turbulence model wall treatment velocity-pressure correlation advection scheme Computational domain and grid model scale grid type

FLUENT 6.3.26 realizable k- (Shih et al., 1995) non-equilibrium wall function (Kim, 1995) SIMPLE method 2nd order upwind

MISKAM 5.01 modified k- (Kato and Launder, 1993 and Lpez, 2002) logarithmic wall function splitting method by Patrinos and Kistler (1977) 1st order upwind

number of grid cells [million] domain area domain height distance of buildings from boundary as a multiple of container height (in/out/side/top) average grid resolution at the containers 8x17.3x3.6 mm 0.6x1.3x0.27m as a multiple of container height (approx.) 0.24x0.51x0.1 1.2 max. cell growth rate Boundary conditions Inlet velocity profile obtained from wind tunnel measurement reference height zref reference velocity at zref inlet turbulence profile ground surface and wall boundaries surface roughness building roughness top boundary outlet boundary 97.2mm 1 m/s 10 m/s 7.29m 1 m/s

full scale wind tunnel scale (1:75) full scale block structured, body fitted with hexahedral cells Arakawa-C non-equidistant Cartesian grid, with buildings blocked out from the grid 1.5 1.3 2 4.8 4x4.19m 300X314m 300x314m 0.28m 21m 130m 21m 23/33/20/7 23/33/20/50 23/33/20/7 1x1x0.5m 0.4x0.4x0.2 1.2 0.8x0.8x0.5m 0.5x0.5x0.5m 0.3x0.3x0.2 0.2x0.2x0.2

logarithmic profile with roughness length z0=0.02m 7.29m 1 m/s generated from the assumption of equilibrium boundary layer no-slip 0.02m 0.01m constant values taken from the top of the inlet profile no-flux with pressure correction

10 m/s

obtained from wind tunnel measurement no-slip 0.017m 0.017m symmetry outflow

0.017m 0.017m

Results

The results are presented here through velocity profiles, contour plots and validation metrics. The place of velocity profiles and contour plots are adjusted to the wind tunnel measurements, therefore the values are compared with values obtained from wind tunnel measurements normalized to the reference velocity Uref . 3.1 Vertical velocity profiles Vertical profiles of U and W velocity components at 21 positions among the array of obstacles were compared to the measurement. Here only a few of them can be shown in Figure 1 and 2. Note that container height corresponds to Z/Zref = 0.35. The following observations can be made: the full scale FLUENT simulations (F-FS) provide closer horizontal velocity component values than the wind tunnel scale (F-WTS); reference velocity has no effect (F-FS-1 and F-FS-10 are almost similar); the fine grid MISKAM run (M-05) slightly underpredicts the longitudinal velocity component U, but gives in most profiles a better agreement than the coarse one, showing that grid resolution has considerable effect in the MISKAM simulations;

Author version of the paper published in: Int. J. Environment and Pollution, Vol. 44, Nos. 1/2/3/4, 2011, pp. 359-367 362
Figure 1 Vertical velocity profiles for selected cases with profile positions

Goricsan et al. (2011): Simulation of flow in an idealised city using various CFD codes

363

the negative vertical velocity W, observed in the wake of obstacles, which is a consequence of the top separation bubble above the obstacles, can be seen clearly in wind tunnel profiles (3, 9, 18) but is only partially shown by FLUENT and is not resolved by MISKAM. This was also observed in MISKAM simulations of the flow above an array of buildings by Eichhorn (2004), and can be explained by the suppression or the absence of the top separation bubble which is a known problem of k- turbulence models.

Figure 2 Velocity profiles in the middle of the array of obstacles (For position see Fig. 1)

3.2 Velocity field in horizontal plane Wind tunnel and simulation results in a horizontal cross section are shown in Figure 3. Points and vectors show the wind tunnel data, the contour and the stream-traces represents the MISKAM (M-1) and FLUENT (F-FS1) values at about 1/3 of the container height. The flow can be clearly separated in two regions: (a) unhindered flow in the longitudinal street; (b) separation area behind the containers. It can be observed that: both models resolve the two regions well as it can be seen from the stream-traces; the vectors are mostly parallel to the stream-traces, fore-spoken that beside the absolute values, the direction of simulated velocity vectors are correct. Comparison of the contour colour and the colour of the measurement dots shows similarity of the longitudinal U and cross-flow V velocity components. FLUENT seems to agree better, while MISKAM is underestimating the U component in the wake and gives longer separation bubble than the FLUENT simulation (Figure 3, top left and right). Both models predict cross-flow V velocities on lower magnitude, especially at container corners. The observations made above and at the vertical profile diagrams suggest that simulated flow velocities in the wake of obstacles are smaller in magnitude and air exchange between the separation and

Author version of the paper published in: Int. J. Environment and Pollution, Vol. 44, Nos. 1/2/3/4, 2011, pp. 359-367 364 the free flow is especially in MISKAM simulations weaker. This would mean e. g. that in later dispersion simulations based on these wind fields an unrealistic high accumulation of pollutants can be expected if sources are located near the surface between buildings.
Figure 3 Velocity in Z=0.9 m plane, comparison of wind tunnel and MISKAM/FLUENT data

3.3 Validation metrics Flow variables obtained from MISKAM and FLUENT runs were validated against the wind tunnel experiments in 566 points of vertical profiles using the BOOT software (Chang and Hanna, 2004), which calculates the metrics shown in Table 2.

Goricsan et al. (2011): Simulation of flow in an idealised city using various CFD codes

365

Table 2 Determining the membership function for various validation metrics Validation metrics (ideal value) FB, Fractional Bias (0) NMSE, Normalized Mean Square Error (0) R, Correlation Coefficient (1) FAC2, within a factor of two (1) Range -0.3<FB<0.3 1<FB<1.2 -1.2<FB<-1 FB<-1.33 or FB>1.33 NMSE<4 9<NMSE<16 25<NMSE R>0.8 0.8>R>0.5 R<0.5 0.5<FAC2 0.3<FAC2<0.4 FAC2<0.2 Membership function Good Fair, Underestimation Fair, Overestimation Poor Good Fair Poor Good Fair Poor Good Fair Poor

Typical magnitudes of these performance measures and estimates of model acceptance criteria have been summarized by many investigators including Chang and Hanna (2004), Ziomas, Tzoumaka & Balis (1998), Zawar-Reza (2005) and Park and Seok (2007). Based on validation metrics shown in Table 3, the full scale FLUENT simulations (F-FS-1 and F-FS10) delivered the best results, followed by the finest grid MISKAM simulation (M-05). Underprediction of W (hit rate below the limit of 0.66 given e.g. in VDI, 2005) is a common problem of both models (see also Figure 4), as it was explained in section 3.1. Opposed to this the metrics of mean flow velocity U fulfil all acceptance criteria for all simulations, while turbulent kinetic energy TKE is underpredicted.
Table 3 Collected validation metrics calculated for vertical profile measurements
U/Uref F-WTS-1 F-WTS-10 F-FS-1 F-FS-10 M-1 M-08 M-05 |R| 0.92 0.94 0.94 0.95 0.97 0.96 0.95 FAC2 0.86 0.86 0.88 0.90 0.89 0.90 0.91 FB NMSE Hit rate 0.72 0.72 0.64 n.a. 0.66 0.81 0.79 0.75 W/Uref |R| 0.90 0.89 0.89 0.87 0.83 0.83 0.74 FAC2 0.25 0.35 0.41 0.39 0.12 0.12 0.31 FB NMSE Hit rate 0.20 0.27 0.31 n.a. 0.25 0.15 0.14 0.20 TKE/Uref2 |R| FAC2 0.75 0.50 0.75 0.42 0.80 0.78 0.79 0.80 0.66 0.44 0.65 0.47 0.69 0.67

n.a.

n.a.

FB 0.64 0.74 0.40 0.34 0.78 0.74 0.54

NMSE 0.67 0.88 0.28 0.23 1.05 0.95 0.52

Hit rate 0.16 0.13 0.41 0.48 0.02 0.04 0.24

Author version of the paper published in: Int. J. Environment and Pollution, Vol. 44, Nos. 1/2/3/4, 2011, pp. 359-367 366

Figure 4 Simulated versus measured data in case of the F-FS-1 and the M-05 simulations

Summary

This paper presents results of FLUENT and MISKAM numerical simulations of flow in an idealised city (MUST experiment). Both codes are able to resolve main features of the flow as it was proven by data visualization of the results and also by validation metrics. On the other hand, not all details of the flow measured in wind tunnel could be recognized in the simulations. While FLUENT with realizable k- closure performed slightly better, both models need further improvement. Application of higher order schemes for advection (MISKAM), changes in the turbulence closure and the use of dense computational grids can help to overcome these weaknesses. Until then, wind tunnel measurements are also necessary to give fully trusted answers for micro scale flow problems in urban environment.

Acknowledgements
The authors would like to thank for the support of the NKFP 3A/088/2004 project. The MUST data set has been provided by the Defence Threat Reduction Agency (DTRA) / Dugway Proving Ground (DPG) for use in COST732.

References
Britter, R., Schatzmann, M. (eds.) (2007) Model Evaluation Guidance and Protocol Document, COST Action 732, COST Office: Brussels, ISBN 3-00-018312-4 Britter, R., Schatzmann, M. (eds.) (2007) Background and Justification Document to Support the Model Evaluation Guidance and Protocol, COST Action 732, COST Office: Brussels.

Goricsan et al. (2011): Simulation of flow in an idealised city using various CFD codes

367

Chang, J.C. and Hanna, S.R. (2004) Air quality model performance evaluation, Meteo. Atmos. Phys. 87, 167-196. Eichhorn, J. (2003) Numerical Modeling of Urban Air Quality: an Extension to the Flow And Dispersal Model MISCAM, Proceedings of the Fifth International Conference on Urban Climate (ICUC-5) Lodz, Poland, 1-5 September 2003. P.3.9. Eichhorn, J. (2004) Application of a new evaluation guideline for microscale flow models, 9th International Conference on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes, Garmisch-Partenkirchen, Germany, June 1-4, 2004, 44-48. Franke, J., Hellsten, A., Schlnzen, H. and Carissimo, B. (eds.) (2007) Best Practice Guideline for the CFD Simulation of Flows in the Urban Environment, COST Action 732, COST Office: Brussels. ISBN 3-00-018312-4 Kato, M. and Launder, B.E. (1993) The modelling of turbulent flow around stationary and vibrating square cylinders, Ninth Symposium on Turbulent Shear Flows, Kyoto, Japan, August 1993, 10.4.1-10.4.6. Kim, S. E. and Choudhury, D. (1995) A near-wall treatment using wall functions sensitized to pressure gradient, Symposium on Separated and Complex Flows, August 13-16, 1995, Hilton Head, South Carolina ASME FED 217. Leitl, B., Bezpalcova, K. and Harms, F. (2007) Wind tunnel modelling of the MUST experiment, 11th International Conference on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes, Cambridge, United Kingdom, July 2-5, 2007. Lpez, S.D. (2002) Numerische Modellierung turbulenter Umstrmungen von Gebuden. PhD thesis, University of Bremen: Germany. Park, O.-H., Seok, M.-G. (2007) Selection of an appropriate model to predict plume dispersion in coastal areas Atmospheric Environment 41, 6095-6101. Patrinos, A.N.A. and Kistler, A.L. (1977) A numerical study of the Chicago lake breeze, Boundary-Layer Meteorol. 12, 93-123. Shih, T.-H., Liou, W.W., Shabbir, A., Yang, Z., and Zhu. J. (1995) A new k-epsilon eddyviscosity model for high Reynolds number turbulent flows Computers Fluids 24 (3), 227238. VDI (2005) VDI 3783, Part 9: Environmental Meteorology. Prognostic Microscale Windfield Models: Evaluation for Flow around Buildings and Obstacles. Beuth-Verlag, Berlin. Yee, E. and Biltoft, C.A. (2004) Concentration fluctuation measurements in a plume dispersing through a regular array of obstacles, Boundary Layer Meteorology 111, 363-415. Zawar-Reza, P., Kingham, S. and Pearce, J. (2005) Evaluation of a year-long dispersion modeling of PM10 using the mesoscale model TAPM for Christchurch, New Zealand, Science of the Total Environment 349, 249-259. Ziomas, I.C., Tzoumaka, P. and Balis, D. (1998) Ozone episodes in Athens, Greece. A modelling approach using data from the medcaphot-trace, Atmospheric Environment 32, 2313-2321.

You might also like