Artificial Neural Network based Design of RF MEMS Capacitive Shunt Switches

摘要

Artificial neural networks (ANNs) have appeared as a very efficient alternative to time consuming full-wave simulations of electrical characteristics of RF MEMS. In this paper, a new ANN based method to be used in the design of RF MEMS devices is proposed. ANNs are trained to model dependence of the scattering parameters and the resonant frequency of an RF MEMS switch on the switch geometrical parameters, as well as to perform the opposite procedure, i.e., to determine values of the geometrical parameters to achieve the desired electrical resonant frequency. The developed models can be used for fast simulation and optimization of the switch characteristics replacing time consuming procedures in full-wave simulators, which leads to a significant reduction of time needed for the device design.