Quasi-Newton model-trust region approach to surrogate-based optimisation of planar metamaterial structures

Abstract

A novel implementation of aggressive space mapping (ASM) for the automatic layout synthesis of planar metamaterial structures is outlined in this article. Specifically, we employ a model-trust region optimisation approach to significantly reduce the computational burden associated with the direct optimisation of high-fidelity models. A Visual Basic for application (VBA) link to a commercial full-wave electromagnetic (EM) solver is created, to ensure that the automated Matlab-based platform has complete control of the design and analysis of the entire ASM process. The validity and e±ciency of our approach is demonstrated with examples of complementary split-ring resonator (CSRR)-loaded transmission lines, comparing both modified and unmodified version of the quasi-Newton iteration within the ASM framework.