Convergence of Core Losses in a Permanent Magnet Machine, as Function of Mesh Density Distribution, a Case-Study Using Finite-Element Analysis

To limit numerical inaccuracies in electromagnetic finite element analysis, the mesh resolution needs to be set with care. Within electric vehicle propulsion, core losses are highly important while excessive simulation times should be avoided, due to the wide operating range. This paper investigates the procedure of achieving core loss convergence, in all parts of a permanent magnet synchronous machine, by varying mesh density. The finite element analysis is done in Ansys Maxwell and the core loss error is quantified by an estimated fully converged result. Two approaches to mesh tuning are compared. It is found that interdependence, between the mesh density in one part and the core loss in another, should be considered. The required mesh resolution is found to differ among operating points. Finally, the flexibility offered by this convergence method is shown. Once the convergence behaviour has been identified, a mesh can be found for any desired compromise of error and simulation time. For the investigated case, the estimated error in total core loss is below 1.2% when an angular machine segment covering one pole, is modelled with 15500 mesh elements. Reducing the number of mesh elements to 7200 gives an error below 3.3%.