Induction Motor Stator inter-turn Fault Diagnosis Approach based on Machine Learning Models

Induction motors (IMs) are essential components in industrial applications due to their widespread use and critical importance. However, stator inter-turn faults (ITFs) can significantly impede the operation of IMs, leading to potential disruptions in industrial processes. Extensive research has been conducted to develop methods for detecting ITFs and evaluating their severity. In this paper, a machine-learning (ML) based approach is utilized for the detection and severity assessment of ITFs in IMs. The training data are generated through simulation tests based on an accurate mathematical model of the IM under ITFs. When ITFs occur, they directly affect the motor's balance and consequently modify the motor impedance matrix. This serves as the key feature for detecting such faults. By using the impedance matrix, the voltage measurements are incorporated into the ITF detecting feature, which helps to distinguish between ITF-induced imbalances and those caused by voltage unbalances. After feature extraction, different ML models are trained and validated using simulation measurements, and their performance is further tested on a separate dataset to evaluate their generalization capabilities. The results demonstrated that k-Nearest Neighbors (kNN) and Gradient Boosting models achieved high accuracy in fault detection, with estimation errors remaining low across various fault ratios and voltage conditions. Both models effectively distinguish between ITFs and SVU-induced anomalies, maintaining consistent detection accuracy.