Design of Frequency Selective Surface using Residual Network

This letter proposes a frequency selective surface (FSS) inverse design method based on a residual learning framework. Traditional FSS designs that meet industrial demands depend on designers' experience and repeated simulation iterations. The proposed method, however, can automatically design the FSS structure parameters that meet specific frequency response requirements. The residual-learning-based method is not only validated by two passband FSS design examples but has shown advantages over the commonly used convoluted neural network-based methods.