The Use of Artificial Neural Networks for Predicting Response of Frequency Selective Surfaces

Abstract

In the real world problems, mostly there is difference between dimensions of simulated and corresponding fabricated structures. In case of Frequency Selective Surface (FSS), this variation can be even larger as the frequency gets into terahertz range. One of the approach to study such variations is metamodeling. In this work, we address the problem of modeling such structures with statistical variations in their geometries by using artificial neural network (ANN). We train and test this metamodel and finally show its performance statistics.