Ultra-Low-Sidelobe-Level Concentric-Ring-Array Pattern Synthesis Using Bessel Neural Networks [Antenna Designer´s Notebook]

The present paper proposes a new approach to constrained optimal pattern synthesis for a circular ring array with ultra-Iow-sidelobe levels. The approach uses an artificial neural network (ANN) with hidden Bessel activation functions, trained using the Levenberg-Marquardt (LM) algorithm. The proposed approach is capable of producing minimal sidelobe levels for an arbitrary pattern shape using only a few iterations, and without high memory requirements as compared to existing alternatives.