Neural Network-based Approach to Design FSS Configuration Applicable for Tunable Structures

Frequency selective surfaces (FSSs) have selective transmission/reflection properties and usually consist of an arrangement of periodic unit cells. The design of these complex structures to meet specific properties is time-consuming and usually relies on iterative cycles of simulations. To tackle this demanding process, a machine learning technique with the implementation of a neural network (NN) has proved its effectiveness for designing such structures in recent years. In this work we first design an FSS structure with a new configuration for biasing the lumped elements that are used in tunable structures, and afterwards we use the NN method for predicting the optimal design parameters employed in the constructed FSS shape. This study is conducted by applying the active devices in the FSS structure with cut-slots. The effectiveness of the proposed methodology is verified by employing the trained NN for the constructed structure operating from 2 GHz to 14 GHz.