Inverse Design of Ultra-Wideband Transparent Frequency Selective Surface Absorbers Based on Evolutionary Deep Learning

Yaxi Pan,Jian Dong,Meng Wang,Heng Luo,Yadgar Abdulkarim
DOI: https://doi.org/10.1088/1361-6463/ace1fc
2023-01-01
Abstract:Conventional frequency selective surface (FSS) absorbers design is time-consuming, involving multiple electromagnetic (EM) simulations for parameter scanning. A novel reverse design method is proposed utilizing evolutionary deep learning (EDL) based on an improved bacterial foraging optimization (IBFO) algorithm and a deep belief network. It establishes the relationship between the geometric structure and EM response. The combination of IBFO and EDL facilitates an efficient optimization for structural parameters, mitigating the 'one-to-many' problem and accelerating the design process. An optically transparent FSS absorber with an ultra-bandwidth of 8-18 GHz is designed to verify the proposed method's capability. The simulation and experimental results demonstrate that the absorber displays exceptional characteristics such as polarization insensitivity and robustness under a 45 & DEG; oblique incidence angle, making it a suitable candidate for radar stealth and photovoltaic solar energy applications. The proposed method can be applied to the design and optimization of various absorbers and complex EM devices.
What problem does this paper attempt to address?