No abstract

The main idea of this article is to model and analyze the short circuit fault between the turns of the stator windings of a Permanent Magnet Synchronous Motor (PMSM). To accomplish this objective, a numerical model describing both the healthy and defective state of the PMSM is developed. Besides, this dynamic model is simulated and tested to study motor behavior under different fault conditions. Also, the frequency domain analysis based on the famous fast Fourier transform (FFT) as well as the time-frequency analysis using discrete wavelet transform (DWT) is established. This allowed extracting signatures related to the presence of an inter-turn short-circuit (ITSC). In the proposed method, ITSC detection is based on the decomposition of stator currents and electromagnetic torque. DWT and spectral analysis show that the low-frequency wavelet details as well as the total harmonic distortion (THD) can be easily used as a good short-circuit indicator. The simulation results of a healthy and faulty motor show the effectiveness of these two approaches but with a significant superiority of the DWT over the FFT.

In a sensorless control of PMSM based on Extended Kalman Filter (EKF), the correct selection of system and measurement noise covariance has a great influence on the estimation performances of the filter. In fact, it is extremely difficult to find their optimal values by trial and error method. Therefore, the main contribution of this work is to prove the efficiency of Biogeography-Based-Optimization (BBO) technique to obtain the optimal noise covariance matrices Q and R. The BBO and EKF combination gives a BBO-EKF algorithm, which allows to estimate all the state variables of PMSM drive particularly, the rotor position and speed. In this paper, three evolutionary algorithms namely Particle Swarm Optimization (PSO), genetic algorithms (GAs) and BBO are used to get the best Q and R of EKF. Simulations tests performed in Matlab Simulink environment show excellent performance of BBO-EKF compared to GAs-EKF and PSO-EKF approaches either in resolution or in convergence speed.