Artificial neural networks assisted analysis of an RF-MEMS switch for W-band applications

This paper presents an efficient approach for the analysis of a W-band RF-MEMS switch based on an artificial neural network (ANN) model of the switch return and insertion losses. The analysis is performed over the 92–95 GHz frequency range, with particular emphasis on the influence of the switch recess dimension. In the proposed ANN modeling approach, the recess dimension is used as the primary input parameter. Two separate ANN models are employed to capture the frequency dependence of the return and insertion losses. Moreover, a power flow analysis is carried out by converting the modeled return and insertion losses into reflected, transmitted, and absorbed power components. The developed ANN model enables fast and accurate evaluation of loss characteristics and power distribution, facilitating efficient parametric studies and design optimization of W-band RF-MEMS switches.