Current Signal-Based Fault Classification Using MFCC-DWT Feature Fusion and AI Techniques in IPMSM used in Electrical Vehicle

Permanent magnet synchronous motor (PMSM) is known as one of the most promis- ing machines for electric vehicle (EV) propulsion due to its high torque density, efficiency and excellent speed regulation. However, motor faults may seriously affect the performance, safety, and reliability of the system. Traditional methods of fault detection cannot provide satisfactory performance in the aspects of accuracy, adap- tation to dynamic working condition and real-time performance. To address these issues, in this paper, we propose a hybrid fault classification framework by combining features from MFCC and wavelet transform. Comprehensive information from the time and frequency domains is combined together in the proposed method, which improves the discriminator for distinguishing the fault types better. The tests were conducted on dataset in two cases: with MFCC only and the combined version using two classifiers catboost and random forest. Finally, the obtained results are very encouraging with 97.10 % in the first case and 86.00 % in the second case.