Accurate modeling of microwave structures using variable-fidelity response features

This paper describes a novel approach to low-cost, accurate modeling of microwave structures using response features. Computational efficiency of the proposed method originates from the less nonlinear dependence of the feature points on the designable parameters of the structure of interest than conventionally used responses (e.g., S-parameters vs. frequency). Further decrease of the surrogate model setup cost is obtained by exploiting variable-fidelity EM simulations that are blended together using co-kriging interpolation. Comprehensive verification and comparisons with benchmark techniques is provided.