## Details

##### Dispersion of fine and ultrafine particles in urban environment. Contribution towards an improved modeling methodology for computational fluid dynamics

by Catherine Gorlé, published February 2010, ISBN 978-2-87516-003-4

PhD thesis the von Karman Institute/Universiteit Antwerpen, Belgium

Numerous studies have demonstrated the existence of a strong link between high concentrations of ne particles (d < 2:5?m) and the prevalence and severity of cardio-respiratory diseases. Recently, research is also focusing on the influence of ultrane particles (d < 100nm). When investigating the possible health impacts of a source of ne or ultrane particles, the dispersion of the particles within the urban canopy is an important aspect, which can be modeled using Computational Fluid Dynamics (CFD). This is, however, a challenging task, mainly because of the highly turbulent nature of urban canopy flows and the resulting importance of turbulent dispersion.

The purpose of this thesis is to improve the understanding of dierent turbulent dispersion modeling methods available for Reynolds-Averaged Navier Stokes Simulations (RANS) and to contribute to the establishment of a validated methodology to model the dispersion of passive, inert particles in urban areas using CFD. To this end, the dispersion in the wake of a single rectangular building immersed in a neutral atmospheric boundary layer without temperature eects is investigated. The correct reproduction of the Atmospheric Boundary Layer (ABL) proles for the velocity and turbulence characteristics in the simulations was addressed rst, secondly the findings from this study were used to model the dispersion in the wake of a rectangular building. Validation was performed using wind tunnel test data.

RANS simulations of an ABL were performed with the standard k-? model to investigate the correct modeling of the velocity and turbulence kinetic energy. A logarithmic velocity prole and a prole for the turbulence kinetic energy as a function of height were used as inlet condition. Equations were derived for the turbulence model constants C? and ?_{?}, which result in a correct modeling of all flow quantities throughout the computational domain.

Large-Eddy Simulations (LES) of the same ABL were performed to verify the specication of an appropriate inlet boundary condition and to quantify the development of the resulting time-averaged and rms velocity inside the computational domain. The inlet condition was derived using the digital filter method proposed by Xie and Castro (2008a). Simulation results showed that 1m downstream of the inlet the time averaged velocity prole increases by 33% close to the ground.

The methodology to correctly model an ABL with RANS was applied for the simulation of the flow and dispersion in the wake of a single rectangular building. First, the focus was on investigating the influence of modifying the turbulence model constants. It was shown that the modication of the constants should be restricted to the region for which it was derived (the

undisturbed ABL). Second, the two most commonly used dispersion models were used to calculate the concentration eld: a particle tracking method and solving the transport equation for a passive scalar. The study confirmed that a satisfactory prediction of the velocity eld is an important, but not sufficient condition for obtaining a satisfactory prediction of the concentration field. The parameters that will govern the turbulent dispersion need to be correctly predicted as well. For the particle tracking model, this requires an accurate prediction of the turbulence kinetic energy and a good estimate of the Lagrangian time scale. For the passive scalar transport model, the value of the turbulent diusion coefficient (usually determined

as the ratio of the turbulent viscosity and the turbulent Schmidt number) is important. It was suggested to calculate these parameters based on the RANS predictions for the turbulence eld, instead of using the default values. Theoretical formulations for estimating the Lagrangian time scale and the turbulent diusion coefficient were proposed and the resulting values were found to be within the range of the values found in literature. The validity of these relations should, however, be veried and the establishment of more accurate estimates would most likely result in improved predictions of the concentration field.

LES were performed to further investigate these aspects. The results show that LES provides a more accurate prediction of both the velocity and the turbulence field. The transport equation for a passive scalar was solved to obtain the concentration eld, which also compared well to the test data. At 4 locations in the flow field the instantaneous velocity and concentration were sampled to enable calculation of the turbulent diusion coefficient. Particles were released from the same locations to obtain an estimate of the Lagrangian time scale. The resulting values for the Lagrangian time scale and the turbulent diusion coefficient have the same order of magnitude as the values obtained from the theoretical formulation, but show a strong dependency on the location within the flow field. Especially at the locations outside of the wake and near the building corner a strong anisotropy of the values is observed, which is obviously not reproduced by the isotropic theoretical formula. When adjusting the turbulent diusion coefficient in the RANS simulation to match the LES values, local improvements in the concentration pattern are obtained.

## Additional Information

Manufacturer | von Karman Institute for Fluid Dynamics |
---|