Experimental Bearing Fault Detection, Identification, and Prognosis through Spectral Kurtosis and Envelope Spectral Analysis

Abstract Rolling element bearings are machinery elements that are widely used in aerospace and industrial applications to support rotating shafts, thereby reducing mechanical friction and heating. However, these components are subject to several kinds of faults which, in general, can be divided into single-point or localized and distributed faults. A common way to detect ball bearing localized or raceway faults is by identifying and analyzing the so-called characteristic “bearing frequencies.” This article presents an approach to bearing condition-based maintenance which comprises fault detection and identification and condition prognosis, based on spectral kurtosis and squared envelope spectral analysis of stator current. In this work, envelope spectrum analysis, based on spectral kurtosis, is applied to identify these frequencies and to evaluate its applicability as a severity estimation index. To evaluate the performance of this index, experimental tests were carried out considering bearings with outer raceway defects at different stages, simulating different fault severities. The numerical results confirm the applicability of the described methodology for detecting and identifying faults, as well as estimating the generated fault severity levels, suggesting that the proposed fault severity index can be applied for bearing monitoring in real world industrial applications.