Enhancing Electric Vehicle Reliability: Advanced Fault Detection for Demagnetization Faults in PMSM Motor Using Artificial Neural Network

Authors- Suprotip Ghosh Hazra, Professor Rekha Chaudhari, Omkar R Kale, Niteen J. Kharade, Abhishek M Pandey

Abstract-Electric Vehicles (EVs) have emerged as a promising solution to address environmental challenges and reduce the dependence on fossil fuels. PMSMs have a critical function in the propulsion setups of EVs, delivering superior efficiency and power density. Nevertheless, demagnetisation faults in PMSM motors present notable challenges to the reliability and maintenance of EVs. In this investigation, we suggest an Artificial Neural Network (ANN)-based system for detecting faults to tackle demagnetisation issues in PMSM motors of EVs. Our research aims to develop a robust fault detection system to enhance the reliability and safety of EV propulsion systems. By employing advanced machine learning techniques, specifically artificial neural networks (ANNs), we conducted comprehensive experiments and analyses to assess the efficacy of our proposed approach. The ANN model was trained with motor performance data, encompassing motor current, torque, and rotor speed, to pinpoint demagnetisation faults precisely. The outcomes of our analysis reveal the efficacy of the ANN-centric fault detection strategy in accurately recognising demagnetisation faults in PMSM motors. The ANN model exhibits remarkable accuracy, precision, recall, and F1-score, exceeding conventional fault detection techniques. Visual depictions of motor performance data and the progress of neural network training further confirm the resilience and dependability of our suggested approach. The pragmatic implications of our research bear significance for the automotive sector, especially in refining EV reliability and maintenance protocols. The adoption of our fault detection system has the potential to curtail downtimes, reduce maintenance expenditures, and prolong the lifespan of EV propulsion systems, ultimately fostering consumer trust in electric vehicles. Furthermore, our analysis suggests potential avenues for future research, exploring advanced machine learning techniques such as deep learning and reinforcement learning. The examination of real-time integration of fault detection systems and the inclusion of sensor fusion methodologies are also pivotal in bolstering the reliability and resilience of fault detection algorithms.

DOI: /10.61463/ijset.vol.12.issue2.151