Machine Learning-Based-Inverse Topological Design Method for Active Frequency-Selective Surface (ITDM-AFSS)

To design active frequency-selective surface (AFSS) with a large number of degrees of freedom (DoFs), in this communication, a machine learning-based inverse topological design method (ITDM) is proposed for AFSS design, termed ITDM-AFSS. Different from the conventional frequency-selective surface (FSS) design, the AFSS electromagnetic (EM) responses under on/off states of p-i-n diode simultaneously participate in the training in the proposed ITDM-AFSS and are simultaneously considered in the design process. In contrast to parametric modeling, the topological design domain can be viewed as a combination of discrete binary pixels. This allows for a significantly larger number of DoFs, enabling the exploration of more optimal structures. The effectiveness of ITDM-AFSS is demonstrated through a numerical example, and two other popular models and an evolutionary algorithm-based design method are employed to compare with the proposed ITDM-AFSS. Additionally, the measurements of the fabricated topologies are carried out to validate the performance of the AFSS design derived from ITDM-AFSS.