Design optimization of microwave structures using low-order local Cauchy-approximation surrogates

A robust and computationally efficient microwave design optimization procedure is presented. This procedure integrates low-order Cauchy-approximation surrogate models with coarse-discretization EM simulations. The optimization engine is space mapping (SM). Instead of setting up a single surrogate model valid for the entire design variable space, a sequence of surrogate models is established in small hyper-cubes containing the optimization path. This allows us to substantially limit the number of training points necessary to create the surrogates and, therefore, reduce the cost of the optimization process. Moreover, our approach eliminates the need for circuit-equivalent coarse models traditionally used by SM algorithms. Our algorithm is successfully illustrated through the efficient design of a number of microwave filters.