Surrogate Model-based Multi-Object Worst Case Analysis for Shunt Capacitive RF MEMS Switch

To analyze the worst case of RF MEMS quickly and efficiently, this paper proposes a surrogate model-based multi-object worst-case analysis framework (SMOW). It contains an adaptive surrogate model and particle swarm optimization (PSO) flow. The Gaussian Process (GP) is applied as a black-box model to approximate the performance of the MEMS device instead of time-consuming EM simulations. Two PSO algorithms with different strategies are adopted to optimize the design variables for the single-objective task and the multi-objective task respectively. Experimental results show that compared with random search (RS), and (PSO) algorithms, our SMOW obtains a speed-up of 5.6-10.8X and improves the quality of solutions in the MEMS shunt switch.