Stator winding fault diagnosis using flux signal in permanent magnet synchronous motor

Electric motors are devices that convert electrical energy into mechanical energy and are used in many applications such as electric vehicles, airline transportation and robotics. Failure of electric motors can cause increased maintenance costs, breakdowns in production lines. For these and similar reasons, fault diagnosis is of great importance. Permanent Magnet Synchronous Motors (PMSM) are widely preferred due to their advantages such as simple structure, high efficiency and easy control. Therefore, fault diagnosis and detection of PMSM is important. In this thesis, a decomposition and machine learning based fault diagnosis and detection method is proposed for the detection of stator winding faults in PMSM. For stable and dynamic operating conditions, the motor has been operated at no load and at full load. The current, speed and flux signals of the healthy and faulty motor are recorded and the flux signal is used for fault diagnosis and detection. The collected flux signals are decomposed into Intrinsic Mode Functions using the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise method. By using Permutation Entropy and Hausdorff Distance methods, high frequency components of the signal are filtered and low frequency components carrying fault information are used. The signal is divided into windows by applying windowing to the filtered signal. By applying statistical analysis methods in time and frequency dimension to each signal window, the signal features were extracted and the feature matrix was created. This feature matrix has been transferred to the Support Vector Machine and fault classification has been performed. With the results obtained, it has been proven that the proposed method gives successful results for fault diagnosis and detection.