On the use of Spectral Kurtosis for diagnosis of electrical machines

This paper explores the efficiency of Spectral Kurtosis (SK) in the area of electrical machines diagnosis. In the literature, Spectral Kurtosis is mainly presented as a tool used to detect non-stationary components in a signal. However, classical use of SK is unsuitable for detection of new stationary components or slow evolutions in a spectrum. In order to detect different types of faults, three indicators are designed from the original definition of the Spectral Kurtosis. These indicators are first tested and compared on synthetic signals. Then, their performance are demonstrated for unbalance detection in a Induction Machine (IM) using current signal.