Applying neural networks to solve the inverse problem of indoor environment

Abstract

The computational fluid dynamics (CFD) analysis and back propagation neural network (BPNN) were employed to solve the inverse problem of indoor environment (IPIE). The approaches to identify the contaminant source locations and inlet boundary conditions by the measurement data at specified observation locations were put forward separately. The inverse design of the indoor boundary conditions by the desired distribution of predicted mean vote (PMV) in the control domain was also presented. The result demonstrated the utility of our proposed method in solving the IPIE.