Impact of Mesh Density and Time-Step Variation on Losses and Torque Characteristics in 3D and 2D FEM Models of an Axial Flux Machine

Paper i proceeding, 2026

The Finite Element Method (FEM) is a useful numerical tool for evaluating the electrical machine performance in two and three dimensions (2D and 3D), such as for the Axial Flux Machine (AFM), where the magnetic flux paths are distributed in 3D. However, as is well known, FEM is sensitive to the chosen space and time resolution (choice of mesh density and time step), and the sensitivity is highly dependent on the calculated parameters. This paper aims to quantify the considerations needed specifically when calculating the loss in Permanent Magnet (PM)s, the loss in the stator core, and the characteristics of the torque. The results show the threshold levels of mesh density and time resolution beyond which fur- ther refinement has an insignificant impact. Furthermore, this study demonstrates the selection of mesh density and time step resolution in electromagnetic transient simulations during the early stage modelling of AFMs. To validate these selections, a detailed analysis of losses and torque was conducted for a 2D, extruded 2D, and a 3D model. It was found that at no-load, the (cogging) torque in the 3D model is very sensitive to the mesh density, while at rated conditions, all models show less than 5% deviation of the average torque compared to the results with mesh and time discretization of a chosen trade-off point. Average core loss sensitivity to mesh is less than 1% when improving the mesh from medium to fine, for the 2D and 3D models, and sensitivity to time-steps is below 7% in 3D and 1% in the 2D AFM models. Magnet losses are sensitive to varying mesh density and time steps with a difference of 10% in 2D models and up to 3% for 3D models.