Implicit space mapping with variable-fidelity EM simulations and substrate partitioning for reliable microwave design optimization

Abstract

Surrogate-assisted techniques enable considerable reduction of the computational cost of EM-driven design optimization processes. Space mapping (SM) is still the most popular method of this kind. Implicit SM (ISM) is particularly attractive because it does not alter the domain of the underlying coarse model and is easy to implement. Yet, ISM is difficult to apply if an equivalent circuit coarse model is unavailable or of poor quality. In this paper, we demonstrate feasible ISM implementation involving variable-fidelity EM simulation models. Preassigned parameters of ISM are introduced as dielectric permittivity values of the substrate sections in the coarse-mesh EM model (utilized as a coarse model for SM). By changing the number of sections, the number of preassigned parameters (and, consequently, the surrogate model flexibility) can be readily controlled. Our approach is demonstrated using a miniaturized rat-race coupler and compared to conventional SM approaches utilizing an equivalent circuit model.