Application of Machine Learning for Analyzing Demagnetization Fault in IPMSM using Finite Element Method

Interior permanent magnet synchronous motors (IPMSM) are preferably used in electric vehicles, ship propulsion, and industries because of their high torque and excellent dynamic performance. Nearly 10% of faults in IPMSM are due to demagnetization, which must be detected early to avoid machine damage. This work considers a 550 W, 220 V IPMSM finite element model in ANSYS Maxwell for analysis. Demagnetization severities such as 5%, 10%, and 25% are investigated and compared with healthy IPMSM. As the demagnetization severity increases, average torque reduces, and asymmetry in the flux distribution over IPMSM increases. The fundamental peak amplitude of radial air gap flux density is 0.4585 T for healthy IPMSM, which reduces to 0.4237 T for a 5% demagnetized machine. Stator current data obtained under healthy and demagnetized conditions from ANSYS Maxwell was analyzed using machine learning toolbox in MATLAB, and the Fine Tree algorithm provides an accuracy of 73.9%.