Reflectarray antenna direct optimization using surrogate models with several geometrical degrees of freedom per polarization

Abstract

In this work, surrogate models based on support vector regression (SVR) of a multi-resonant unit cell with several degrees of freedom (DoF) per polarization are trained and used in a reflectarray antenna design and optimization. Since the unit cell has multiple sharp resonances when considering several DoF, the training process is carried out in a hyper-rectangle around a plane of stability. Results of SVR models with four geometrical DoF are shown to provide highly accurate results for the design and analysis of a very large contoured-beam reflectarray for space applications. The direct optimization layout with the surrogate models allows to improve the cross-polarization figures of merit several dB.