Multimodal Fusion-Based Fault Diagnosis of Electric Vehicle Motor for Sustainable Transportation

Abstract

Electric vehicles (EVs) are essential for sustainable transportation, and various ecofriendly vehicles are being manufactured. In EVs, the traction motor is a crucial prime mover for propelling the vehicle forward. However, traction motors are susceptible to faults like any other motors which can compromise their performance, safety, and longevity. This study proposes a reliable fault diagnosis strategy by using information fusion of vibration and current sensor data. Initially, vibration and current signals fusion-based diagnostic methods have been developed in the laboratory environment for induction motors (IMs) having seven fault conditions. This developed method involved wavelet synchrosqueezing transform (WSST) for the decomposition of the acquired vibration and current signature and further converted into a time-frequency spectrum. Thereafter, a multi-input fusion network (MiFN) has been designed for the fusion of vibration and current information. Finally, the developed fault diagnosis method has been extended and validated on an electric two-wheeler for diagnosing the faults in the brushless direct current motor (BLDC) hub motor. The suggested approach demonstrated significantly better classification accuracy than the signature of each sensor across a range of different speed situations. The achieved accuracies are in the range of 97.50%–98.35% in the laboratory environment and 90%–95% in the electric two-wheeler. The experimental results demonstrate that the suggested diagnosis methodology is highly accurate and remarkably reliable for pragmatic working conditions of EVs.