ANN-based physio-chemical prediction of the photochemical cycle and the reactive air pollutant dispersion in an urban environment

This paper investigated the mechanism of the air pollutant’s reactive dispersion in the ideal urban environment via a hybrid numerical intelligence model consisting of Architectural Institute of Japan (AIJ) wind tunnel data, k-ε model Computational fluid dynamics (CFD) simulation, and Artificial Neural Network (ANN) machine learning algorithms. The results showed that the normalized velocity distribution stay unchanged under different inflow speeds. The phenomenon of pollution accumulation on the rear of the building resulting from the turbulent kinetic energy (TKE) distribution was discovered. It was determined that the pollutant spreading region enlarges proportionally with the increase in Damköhler number of ozone (DaO3) number when it is close to 1. In contrast, the pollutant spreading region was unaffected by the Damköhler number of nitrogen oxide (DaNO) number when it is far less than 1. Moreover, the ANN model showcased the strong advantage of characterizing the sophisticated nonlinear spatial diffusion and reaction of the air species by generating acceptably accurate predictions with significantly lower computational and time costs compared with CFD simulation.