Envelope Spectral Kurtosis and SVD Analysis of Stray Flux Signals for the Diagnosis of Field Winding Faults in Synchronous Motors

Abstract

In this work, the application of an unsupervised algorithm based on Singular Values Decomposition (SVD) and on the Envelope Spectral Kurtosis of stray flux signals is proposed as a way to detect field winding faults in synchronous motors. These signals are analyzed both under starting and under steady-state regimes for two different types of flux sensors. Furthermore, a new fault indicator based on the SVD of the envelope of the spectral kurtosis is proposed, yielding satisfactory results for fault classification, observing a common trend in both cases at starting and in steady-state independent of the flux sensor type and position. A comparative analysis between the two motor operating regimes is carried out, obtaining several conclusions about the suitability of each regime for the application of the methodology in real applications.