Review of signal-based methods for the detection of bearing damages
In this work, the stator current of a permanent-magnet excited synchronous machine was analyzed using signal-based methods, such as the short-time Fourier transform, the wavelet packet decomposition and the envelope detection. Thereby, the ability of these methods for the detection of bearing damages were evaluated. The data used for the analysis were generated by means of a simulation model developed at IEM. This model is able to reproduce the effect of bearing damage. The signal-based methods were designed and applied to the stator current data. For the evaluation, statistical characteristics are calculated from the extracted frequency bands. The evaluation criteria were derived based on the linear discriminant analysis.