Design of a Reconfigurable 5-Fingers Shaped Microstrip Patch Antenna by Artificial Neural Networks

This paper designs a reconfigurable 5-fingers microstrip patch antenna using artificial neural networks (ANNs). The particular neural networks selected for this task are the source difference method (SD), multilayer perceptron (MLP), prior knowledge input method (PKI) and difference prior knowledge input method (PKID). Employing fine and coarse models for training the networks enables to develop fast and accurate EM-ANN models. The proposed antenna has four modes of operation, which are controlled by two PIN diode switches with ON/OFF states, andit resonates at multiple frequencies between 2-6 GHz. Simulations demonstrate considerable savings in computational costs as compated to the EM simulation software while maintaining the same level of accuracyas the fine models. Keywords—Artificial neural network,source difference method, multilayer preceptrons, prior knowledege method, difference prior method, reconfigurable microstrip antenna, PIN diodes