Design of Miniaturized-Element Frequency Selective Surface Using Neural Networks

Abstract

Recently, the research on the design of frequency selective surface (FSS) has developed rapidly, and new achievements have been continuously obtained. Compared with the frequency selective surface in the traditional sense, the miniaturized-elements frequency selective surface (MEFSS) has now become an important direction of FSS development. In this paper, a new equivalent circuit model is proposed to explain the MEFSS interlayer coupling based on neural network. Compared with the equivalent circuit formed by cascading each layer of MEFSS by transmission lines in the past, the circuit models has higher rationality and accuracy. Here by developing a Back-Propagation neural network based machine learning tool, the relationship between coupling elements and structural parameters can be obtained, the transmission coefficients of the MEFSS structure can be obtained directly without full-wave simulations.