Fast fluid dynamics simulation of the airflow distributions in urban residential areas

Abstract

Conventional CFD models can resolve complex physical fields around single or multiple buildings with high spatial resolution, but they are unable to meet the demand for fast simulations with meter-level spatial resolution and minute-level temporal resolution due to the huge computational domain with numerous grids for urban residential areas. In this paper, three fast fluid dynamics (FFD) models with different pressure-correction schemes (i.e., SIPC, NIPC and RIPC) for solving Navier-Stokes (N-S) equations are implemented in OpenFOAM. The computational accuracy and speed of these FFD models is validated and analyzed through three cases. For the prediction of airflow distribution in urban residential areas, the average relative error between the simulation results of the FFD models and the wind tunnel experimental data is less than 15 %. The speeds of these FFD models show SIPC > NIPC > RIPC from fast to slow under the same number of grids, turbulence model and time step size, and these speeds are about 15 times faster than the commercial CFD code Ansys Fluent. The computing time of the three FFD models approximately shows a linear increase with the number of grids, and the difference in computing time between the other two FFD schemes and the SIPC scheme becomes larger and larger as the number of grids increases from 0.3 million to 3.0 million. There is almost no difference in the number of grids required to achieve grid independence for the three FFD models. On the premise of ensuring accuracy, the SIPC scheme has the fastest computational speed, and it could be preferred to quickly evaluate the airflow distribution of different residential areas in urban planning stage.