Advanced Signal Processing Techniques for Demagnetization Detection in PM Generators at Variable Speed

In this paper, demagnetization detection in permanent magnet generators is analyzed using advanced signal processing techniques. The t-f decomposition of the current signal is calculated using the min-norm method. Three signal processing techniques are investigated which are the adaptive demodulation, the time-varying resampling and the envelope analysis. The merits of these methods are thoroughly compared against the commonly adopted short-time Fourier transform analysis. Moreover, traditional signal processing techniques may not be suitable for fault detection when the machine operates at variable speed. Under such condition, interference between the fundamental frequency of the machine and the fault harmonic can lead to false alarms. Simulation test and experimental results demonstrate effective performance and feasibility of the proposed methods for demagnetization detection of permanent generators operating under non-stationary conditions.