Evaluation and classification of stator turn-to-turn faults using electrical equivalent circuits for surface permanent magnet brushless direct current motors

Stator turn-to-turn faults occur due to improper loading, eccentricity in the rotor, and increases in the operating temperature. During the occurrence of a stator turn-to-turn fault, an abnormal temperature increase occurs, and if this state is left unattended for a long duration, it can lead to degradation of the permanent magnet. This paper presents an analytical modeling scheme for surface permanent magnet brushless DC motors for diagnosing and classifying stator turn-to-turn faults using SIMULINK® during non-stationary operating conditions. A significant increase in the stator current, back EMF, torque, and speed is observed. A current signature analysis is performed during non-stationary operating conditions using a fast Fourier transform method to identify the severity of the fault. Furthermore, a simple and efficient classification model is developed by selecting the best classifier among the decision trees, neural network, support vector machine, discriminant analysis, and ensemble classifier. A statistical evaluation of the current signal for fault feature extraction and ranking is performed based on minimum redundancy and maximum relevance, Chi-square test, Relief F, analysis of variance, and Kruskal–Wallis test. The dataset for classification is extracted from a Simulink analytical model. Neural network-based classifiers can classify faults precisely and rapidly with a minimum number of features.