Application of power spectral density and the support vector machine to fault diagnosis for permanent magnet synchronous motor

The present study proposes a new technique for diagnosing the Inter-Turn Stator-Winding fault in Permanent Magnet Synchronous Motors. Under faulty operating conditions, the system’s model is built while considering specific parameters linked to fault location and sharpness. The proposed fault-detection method is principally based on the Power Spectral Density estimator’s association with the Support Vector Machine. This classifier is used to separate different regions of system performance. It has been trained to associate the Power Spectral Density current’s magnitude and the stator’s current negative sequence with the fault severity. This method has shown exceptional performance as long as it achieves a fault detection rate of 98.5% for different fault severities. Two Power Spectral Density estimators were compared in terms of their ability to extract fault characteristics, namely Burg’s method and Welch’s method. It was concluded from this study that Welch’s has a higher frequency resolution than Burg’s method.