INTELLIGENT FAULT DIAGNOSIS OF ELECTRIC VEHICLE POWERTRAINS USING MACHINE LEARNING TECHNIQUES

Abstract

Electric vehicles (EVs) are widely acknowledged as eco-friendly means of transport. They operate by transforming electrical energy into mechanical energy through various types of motors, which aligns with the sustainable ideals of smart cities. The motors in EVs draw and utilize electrical power from renewable energy (RE) sources via interfacing connections, employing power electronics technology to generate mechanical power through rotation. The dependable functioning of an EV heavily depends on the state of these connections, particularly between the output of the 3-phase inverter and the brushless DC (BLDC) motor. This paper utilizes machine learning (ML) tools to identify and classify faults in the connection lines from the 3-phase inverter output to the BLDC motor during the operational phase on the EV platform, focusing on double-line and three-phase faults. Several ML-based tools for fault detection and classification, including Decision Tree, Logistic Regression, Stochastic Gradient Descent, AdaBoost, XG Boost, K-Nearest Neighbour, and Voting Classifier, were optimized to enhance robustness and reliability in identifying and categorizing faults. The ML classifications were developed using datasets that represent both healthy and faulty conditions, taking into account a combination of six essential parameters crucial for the reliable operation of EVs. These parameters include the current supplied to the BLDC motor from the inverter, the modulated DC voltage, output speed, measured speed, and readings from the Hall-effect sensor. Furthermore, the effectiveness of the proposed fault detection and classification methods utilizing ML tools was evaluated by comparing their detection and classification efficiencies through various statistical performance metrics across the classifiers.