RF MEMS switch optimization using ANN and design of antenna with frequency reconfigurability

Abstract

Artificial neural networks (ANN) are becoming highly prominent in the optimization of micro-RF devices, which are very significant in wireless communication applications. In this manuscript, we present the optimization of RF MEMS switches using ANN and the design of an antenna with frequency reconfigurability. A unique procedure is proposed to design reconfigurable antennas with RF MEMS switches using ANN. The novelty of this work lies in the creation of a dedicated dataset for the considered RF MEMS switch with FEM tool simulation and the utilization of cascade feed-forward neural networks for optimization. The design of the dataset and the optimization of RF MEMS switches in different aspects using ANN are the key contributions of this work. Comprehensive analysis was performed using a neural network with the designed dataset. Cascade feed-forward neural networks are highly efficient when compared with other neural networks. The weights and biases of the network were selected using the Xavier approach. The cascade feed-forward neural network is optimized using the LM training algorithm. The optimized cascade feed-forward neural network is further used to predict the optimized RF MEMS switch dimensions for the desired application. The network produces an accuracy of 94.9%. An RF MEMS switch was designed from the dimensions predicted by the cascade feed-forward neural network. The designed switch offers – 55 dB Isolation and – 0.2 dB Insertion. Eventually, an antenna was designed by incorporating identical switches which offer frequency reconfigurability.