Artificial Intelligence Based Deep Learning Surrogate Model for Design Optimization of Microstrip Frequency Selective Surface

This paper centers on the utilization of Artificial Intelligence (AI) based surrogate-model assisted optimization methods for having high performance microwave circuit elements such as Frequency Selective Surfaces (FSS). The fusion of the Modified-Multi Layer Perceptron (M2LP) and a meta-heuristic optimization algorithm Honey-Bee-Mating Optimization (HBMO) algorithm was pursued to improve the design of the FSS unit. The M2LP surrogate model offers a unique strategy for regression problems which makes it a suitable solution tool for optimization of complex microwave designs such as FSS. The outcomes derived from the optimized FSS design parameters obtained from M2LP-HBMO models are compared with the findings obtained from the 3D EM-simulation model. The purpose of this comparison was to verify that the response of the 3D EM simulation model, matches exactly with the proposed approach. As a result, it was seen that the proposed M2LP surrogate effectively optimizes FSS designs for various applications within the given domain.