Surrogate Model Assisted Frequency Selective Surfaces Multifunctional Optimization

In this paper, we presented a novel differential evolution (DE)-based optimization approach for accelerating the iterative design process of frequency selective surfaces (FSS). By integrating a Bayesian neural network (BNN) and lower confidence bound (LCB) prescreening methods, our approach constructs a simulation surrogate model that effectively leverages a priori knowledge from equivalent circuits. This method is well-suited for optimizing multifunctional filtering FSS designs using the same design prototype. The proposed algorithm’s performance was validated through a typical case study, demonstrating its effectiveness in enhancing the optimization process. The realization details and code are available for further exploration.https://github.com/LeeHongji/FSS_auto_optimization.git