Application and numerical error analysis of multiscale method for air flow, heat and pollutant transfer through different scale urban areas

作     者：Cui, Peng-Yi Zhang, Yan Zhang, Jin-Hao Huang, Yuan-Dong Tao, Wen-Quan 

作者机构：Univ Shanghai Sci & Technol Sch Environm & Architecture Shanghai Peoples R China Xi An Jiao Tong Univ Sch Energy & Power Engn Key Lab Thermofluid Sci & Engn MOE Xian Shaanxi Peoples R China 

出 版 物：《BUILDING AND ENVIRONMENT》 (建筑与环境)

年 卷 期：2019年第149卷

页      面：349-365页

核心收录：

学科分类：0830[工学-环境科学与工程（可授工学、理学、农学学位）] 08[工学] 0813[工学-建筑学] 0814[工学-土木工程] 

基　　金：Foundation for Innovative Research Groups of the National Natural Science Foundation of China Shanghai Sailing Program [18YF1417600] 

主　　题：Multiscale method Wind-tunnel measurement Buoyancy flow Pollutant dispersion Numerical error analysis 

摘      要：There is an increasing concern on the effect of the outdoor environment on the indoor air quality (IAQ) through ventilation. In urban areas, the physical phenomenon that the air pollutants pass over the neighborhood scale-street scale-indoor scale, and affect the health of the residents belongs to multiscale problems. The adopted multiscale method models the problem in nesting grid system approach by solving from top (entire domain with relative coarse grids) to down (subsequent sub-domains with finer grid). In this paper, a 3-D multiscale model with Richardson number (Ri) equal to 0.85 was firstly established to investigate the above multiscale phenomenon, then evaluated by comparing with the full-scale intensive simulation, since our previous study showed that when Ri = 0.85 the standard k-epsilon turbulence model can better predict the flow and temperature fields. The numerical error analysis of the multiscale method was conducted by considering the effects of two aspects (boundary condition interpolation schemes and sub-domain partitioning) on the regularity of error production and transfer. Results show that the applied multiscale method can save 83.8% computing time. Furthermore, it can be concluded that among the three interpolation schemes for the boundary reconstruction the linear and spline interpolation methods are appropriate while the nearest method should not be applied because of its 1st-order accuracy. If insensitive interfaces are chosen as the partitioning interfaces of the lower-scale domain the computational accuracy of the multiscale method can be greatly improved.