Application of spectral kurtosis to bearing fault detection in induction motors

This paper deals with the application of the Spectral Kurtosis (SK) to bearing fault detection in asynchronous machines. This one-dimensional spectral measure allows to study the nature of the harmonic components of the stator current of an induction motor running at a constant rotation speed. It provides additional information with respect to second order quantities given by the Power Spectrum Density (PSD). This information can be used to discriminate between constant amplitude harmonics, time-varying amplitude harmonics and noise. Since the harmonic components for a healthy machine can be considered as constant amplitude harmonics, the SK can provide a measure of the distance of the analyzed machine from a healthy one. For example, the amplitude fluctuations of the harmonic components produced by a mechanical fault can be highlighted by the Spectral Kurtosis.