A Novel Optimized CNN-SVM Approach for Inter-Turn Stator Fault Detection and Localization in Permanent Magnet Synchronous Motors

Identification and localization of Inter-Turn Faults (ITF) in permanent magnet synchronous motors (PMSM) are critical aspects of motor health monitoring. It can effectively span motor life, reducing the time of silence and preventing sudden failures. In this respect, the study has proposed an optimized hybrid Machine Learning (ML) method to improve accuracy and predictive performance concerning ITF identification. In this method, the combination of Convolutional Neural Networks (CNN) and Support Vector Machines (SVM) is utilized to determine the fault location, in which the PMSM motor data are collected under different faulty conditions, and CNN enters the features extracted from the data into SVM to locate the fault. The model was tuned to make it more accurate and performing, which is done through Bayesian optimization. With this method, the parameters are automatically selected to obtain the best ones. The data used incorporates various PMSM phase currents under healthy conditions and faulty conditions to locate ITFs correctly in all three phases from 10% to 90%, thus proving superiority over other methods. Root Mean Square Error (RMSE), Normalized Root Mean Square Error (NRMSE), Coefficient of Determination (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>), and Mean Absolute Percentage Error (MAPE) are the different criteria used in evaluating the performance of this model and other models considered for comparison. The results of this research prove that the proposed approach will perfectly locate ITFs, increasing the motor's reliability and performance.