Abstract
In Part 2 of the study, the CFD method to predict the exhaust gas diffusion around a building was proposed. The numerical results were compared with the measured results of one case in Part 1, a case with an emission souce at the downwind corner on the rooftop on the cube.
In practical purpose, minimizing the computational cost is important. Therefore, the k-ε model was choosed as a turbulent model.
The required mesh fineness in the neighbor of walls should be smaller than H/24 where H is the model height. The values of eddy dissipation rate at the inflow boundaries were well determined by the consideration to balance the terms of the turbulent energy generation and dissipation. The turbulent intensity of the exhausted gas had small effects on the results. The log-law was better than the 1/4 power-law for the wall boundary conditions. The turbulent Schmidt Number of 0.6 or 0.7 brought the best predictions of the concentration field.
Through the validation, the following characteristics were clarified; The standard k-ε model can reproduce the recirculation flows neither on the rooftop nor by the sidewalls correctly. The reattachment point was 2.1H behind from the model center, which was longer than the measured value 1.6H. The turbulent energy at the downward region of the rooftop and sidewalls were predicted smaller than the measued values.
Concerning the concentration field, the horizontal width of the plume was narrower and the vertical peak position was higher than the measured value. These discrepancies can be explained by the above mentioned inaccuracy of the predictions of the flow fields.
The characteristics of the horizontal peak position and the concentration decay rate with distance are important but difficult to be predicted by the Gaussian diffusion models. In spite of remaining small defects, the proposed CFD method reproduced them well.
