Fitting EM Scattering of 3D Rough Surface using Deep Neural Networks

摘要

This paper explores the potential of deep neural networks to fit electromagnetic scattering problems. We use numerically calculated EM scattering data to train deep neural network so that it can be used to predict scattering of a new problem. First, we calculate the induced current of a randomly rough surface incident by tapered waves using three-dimensional method of moment (MoM) based on beam simulation method^[1]. Then, we use the data to train pix2pix network^[2] where the rough surface is used as input and the induced current is used as the desired output. The results show the feasibility of the basic idea of this paper. It provides a new way to solve the calculation of electromagnetic forward problems and paves the way for the corresponding calculation of electromagnetic inverse problems.