Mixed Proper Orthogonal Decomposition with Harmonic Approximation for Parameterized Order Reduction of Electromagnetic Models

Abstract

This paper presents some preliminary investigations on a hybrid Model Order Reduction approach for parameter-dependent electromagnetic systems. Starting from an integral equation formulation of the field problem, we introduce a first level of compression based on the well-established Proper Orthogonal Decomposition (POD). The result is a small-scale approximation of the full-order discrete field formulation, which retains an explicit dependence on the set of free parameters defining the geometry. The evaluation of the reduced model for arbitrary parameter configurations remains very expensive, as it requires the construction of the full system equations before its projection onto a lower-dimensional space. This problem is solved by constructing a surrogate macromodel of the parameterized reduced-order system through a multivariate Fourier approximation. Numerical results applied to a moving coil over a finite ground plane show model compression above 99% while preserving accuracy on currents and fields within 1%.