Diagnostic and Prognostic Health Management of Electric Vehicle Powertrains: An Empirical Methodology for Induction Motor Analysis

Abstract

The growing interest in electric vehicles has led to an increased focus on the development of efficient and reliable electric motors. To ensure reliable operation, it is essential to incorporate on-board diagnostic and prognostic tools that can detect and predict potential failures. This paper proposes an approach to diagnose and predict the health condition of induction motors used in electric vehicle powertrain applications using machine learning techniques. The proposed approach utilizes vibration signals collected from accelerometers attached to the motor and employs decision forest and decision tree algorithms to classify the health condition of the motor. The study aims to identify the most significant features of the vibration signals and evaluate the effectiveness of the proposed approach in diagnosing and predicting the health of the motor. The models are trained on full extracted features and selected features using Principal Component Analysis (PCA) and Correlation Analysis (CA) to improve the classification performance. The experimental results demonstrate that the combination of selected features using PCA with the Decision Forest (DF) algorithm achieves the best classification performance for the simulated motor fault conditions. This suggests that machine learning techniques can be effective in diagnosing and predicting the health of induction motors used in electric vehicle powertrain applications.