Structure-informed model reduction of Maxwell-type systems
Please login to view abstract download link
Recent methods for structure-informed model reduction of Hamiltonian systems are applied to problems involving Maxwell's equations in a 3D domain. A variationally consistent method is presented that provably respects the noncanonical Hamiltonian structure of the system, along with a complementary approach that does not require access to derivative information from the high-fidelity discretization. It is demonstrated that these approaches are also effective in the presence of time-dependent forcing, and that its effects can be reliably approximated using non-intrusive, data-driven techniques. This combination is shown to produce a useful reduced-order model that significantly accelerates predictions across the parameter space of the problem.
