Robust optimization of microwave structures using co-simulation-based surrogate models

A robust and computationally efficient technique for design optimization of microwave structures is proposed. We exploit co-simulation-based surrogate models with the essential couplings evaluated using EM solver and most of the designable parameters modelled by distributed circuit elements with physical dimensions corresponding to those of the structure being optimized. Good prediction capability of the surrogate model is ensured by the space-mapping-based alignment process. It is demonstrated that the proposed methodology is able to yield a satisfactory design at the computational cost corresponding to two or three EM simulations of the optimized structure. The design of two microstrip filters is presented for the sake of illustration.