Compact dual-band metamaterial antenna using deep neural network for next-generation wireless communication

Abstract

In this research paper, a novel deep neural network (DNN) methodology is used to accurately predict the resonant frequency of electric-inductive-interdigital capacitive (EL-IDC) metamaterial unit cells. The approach employs an equivalent circuit model (ECM) to identify critical design parameters that influence inductance and capacitance, facilitating an accurate analysis of electromagnetic behavior. DNN is implemented with multiple hidden layers, utilizing a comprehensive dataset generated from variations in geometric and material properties. Comparative evaluations of Bayesian optimization, Levenberg–Marquardt, and scaled conjugate gradient training algorithms with DNN reveal that Bayesian optimization achieves superior predictive accuracy while maintaining computational efficiency. The proposed methodology is further validated by designing an asymmetric coplanar stripline (ACS)-fed compact metamaterial antenna, achieving dual frequency bands 3.1 to 3.9 GHz and 7.0 to 8.7 GHz. Experimental and simulation results demonstrate the antenna’s excellent performance, including low reflection coefficients, effective impedance matching, and radiation characteristics. This study emphasizes the transformative potential of DNN, which is used to enhance the design functionality of metamaterials for next-generation wireless communication applications.