Multi-Fidelity Surrogate Modeling based on Numerical Eigenfunction Expansions

In this work, a novel physics-based method for supplying low-fidelity models is studied. Specifically, numerical eigenfunction expansions are used to efficiently provide approximate electromagnetic solution data for a multi-fidelity (MF) surrogate optimization process. Based on our results, the proposed method can generate highly accurate MF models 3 times faster compared to state-of-the-art methods and with errors less than 3% when compared with full-wave solutions.