Electromagnetic Modeling of Numerical Green's Function for Moving Scatterers Using Artificial Neural Networks

This paper presents an electromagnetic modeling method for moving scatterers using artificial neural networks (ANN) in two dimensions. Using the quasi-static approximation method, each time point of a moving object is treated as static, and the NGF is predicted by the ANN for each time point. The training method of ANN is to sample the relationship between the field and source of the moving scatterer at various locations within a finite region as a dataset, and the trained network model can represent the NGF of the moving scatterer at any location. To verify the method, the NGF of a moving scatterer in two-dimensional is obtained by ANNs, and relative errors of the ANN-predicted NGF can reach less than 5%. The electric field obtained based on the ANN-predicted NGF is also shown.