An adaptive Newton-Raphson technique for combined vector-scalar potential solutions of large scale 3D magnetic field problems involving anisotropic materials

摘要

An adaptive Newton-Raphson (ANR) technique for handling magnetic material nonlinearities in large-scale 3-D magnetic field problems, for use in conjunction with combined vector-scalar potential 3-D finite element (FE) formulations, is presented. This ANR iterative technique enhances the quality of convergence of the magnetic field solution, as evidenced by a truly 3-D magnetic field case study of a modified Lundell alternator (MLA) performed using CRAY-YMP supercomputers. The method worked very well in the case study of a 14.3-kVA MLA, in which the number of Fe grid tetrahedrons exceeded 113000 with more than 20000 magnetic scalar potential (MSP) unknowns. The results compared very well with corresponding experimental test data.