Automatic fault diagnosis in PMSM using adaptive filtering and wavelet transform

In this paper, a novel fault diagnosis technique is developed for permanent magnet synchronous motors (PMSM) using discrete wavelet transform and support vector machines based on stators current waveform. An adaptive filtering scheme is developed to remove the fundamental component of the current waveform in order to increase the accuracy of fault diagnosis. For this purpose, the instant rotational speed of motor is estimated without use of any external sensor. Our fault diagnosis method achieves accuracy of 97.67% and can be used to detect two types of faults in a PMSM, i.e. broken magnet and eccentric faults. In this method only one stator phase current out of three phases is used. It is shown that the residual waveform after removing the fundamental component by the proposed adaptive filter is a rich source of information for fault diagnosis. This filtering scheme can be generalized to other rotary machine applications in which the current signal can be measured for fault analysis.