An effective method to reduce the computing time of nonlinear time-stepping finite-element magnetic field computation

Abstract

Time-stepping finite-element methods have been widely used to compute the magnetic field of electrical machines. Because the reluctivities of magnetic materials are nonlinear, the finite-element equations have to be solved iteratively. In this paper, an effective method for reducing the computing time of the Newton-Raphson method coupled with the incomplete Cholesky-conjugate gradient algorithm for solving time-stepping finite-element problems is presented. The proposed method is based on a proper prediction of some predefined error tolerances in the iteration processes at each time-stepping finite-element computation. The computational analysis on an induction motor shows that the proposed strategy can reduce the nominal computing time by as much as 50%.