ANN element characterization for reflectarray antenna optimization

Abstract

In this paper, artificial neural networks (ANNs) for modelling reflectarray periodic element is evaluated. A reflectarray antenna based on a 3-layer stacked patch element is chosen. Every element in the reflectarray must shift the phase of the reflection coefficient a given amount to obtain the prescribed radiation diagram. Different shifts are obtained from different geometrical configuration of the reflectarray element. Then, optimizing a whole reflectarray involves a large number of full wave electromagnetic (EM) computations. ANNs are found to represent the complex reflection coefficient of the reflectarray element as a function of the geometrical parameter, the incident angle and the frequency. A good agreement is achieved between the ANN outputs and the EM solver solutions by Method of Moment (MoM). Using ANNs in place of full wave EM simulation is proposed for reducing the time in optimization purposes.