Detection and Prediction of Demagnetization fault for PMSMs Based on 3D-FEA and AI Technology

This study introduces a combined diagnostic approach that merges three-dimensional finite element analysis (3D-FEA) and artificial intelligence techniques for the detection and prediction of demagnetization faults in permanent magnet synchronous motors (PMSMs). The proposed approach analyzes the behavior of the magnetic flux density distribution near the stator windings using 3D finite element analysis (3D-FEA). The 3D-FEA model was developed in EMWorks EMAG, providing a visual representation of magnetic field variations as a function of fault type and severity. The second method is based on Extreme Gradient Boosting (XGBoost), an artificial intelligence algorithm. In the XGBoost-based approach, overall fault classification is performed, including the identification of unknown faults. By examining the magnetic flux density and magnetic field intensity in corresponding regions of the machine, a diagnostic index is developed to quantify the extent of the fault. The results demonstrate that the proposed methods can accurately detect and predict demagnetization faults in PMSMs with a high degree of reliability.