Model Order Reduction Techniques applied to Magnetodynamic Scalar Potential Formulation

摘要

Numerical techniques to extract important information from large systems of equations are in focus of research to cope with the computational effort. This paper discusses the feasibility of model order reduction techniques applied to magnetic scalar potential formulation coupled to the electric vector potential, known as T-$Omega$ formulation. This formulation, compared to the magnetic vector potential formulation, uses a decomposition of the magnetic field into a scalar potential and loop fields to avoid explicit cuts, necessary to not violate Amperés law. The analysis of different techniques, namely the Proper Orthogonal Decomposition (POD) and Proper General Decomposition (PGD), applied to a three dimensional eddy current problem are performed. The presented approach is finally evaluated in terms of convergence and computational effort.