Data-Driven Surrogate-Assisted Optimization of Metamaterial-Based Filtenna Using Deep Learning

In this work, a computationally efficient method based on data-driven surrogate models is proposed for the design optimization procedure of Frequency Selective Surface (FSS)-based filtering antenna (Filtenna). A Filtenna acts as module that simultaneously pre-filters unwanted signals, and enhances desired signals at operating frequency. However, due to typically large number variables FSS unit elements, their complex interrelations affecting scattering response, challenging task. Herein, deep-learning-based algorithm, Modified-Multi-Layer-Perceptron (M2LP), developed render an accurate behavioral model cell. Subsequently, M2LP applied optimize elements being parts under design. The exemplary device operates 5 GHz 7 band. numerical results demonstrate presented approach allows almost 90% reduction computational cost process compared direct EM-driven At same time, physical measurements fabricated prototype corroborate relevance methodology. One important advantages our technique cell can be re-used various bands without incurring any extra expenses.