A Neural Network-Empowered Inverse FSS Design and Synthesis Approach for 5G Shielding Applications

In this paper, an intelligent technique for Frequency Selective Surfaces (FSS) synthesis for 5G shielding applications is presented. Our approach applies a Neural Network (NN) architecture trained on a dataset of FSS electromagnetic (EM) responses to directly predict the optimal geometrical configurations from the desired shielding specifications. The proposed solution is a very useful tool for offering radiofrequency (RF) designers a smart service to optimize RF system design considering the harmful EM environment of the device under test. The developed solution is mainly based on multivalued inverse machine learning (MIML) method. It was optimized so that it takes account of the effect of polarization and incidence angles of incident EM waves and makes the proper arrangement of the shielding FSS screen, relying on the desired frequency behavior. We demonstrate the efficiency of our proposed approach through comprehensive simulations and experimental validations. Results indicate that the proposed NN-empowered inverse FSS design proposed method achieves superior performance in terms of shielding effectiveness (SE) and design efficiency compared to traditional direct machine learning (ML) modeling approaches.