Naturally-induced Early Aviation Bearing Fault Test and Early Bearing Fault Detection

Abstract

While the early detection of roller bearing faults has been extensively studied, the research in this area still suffers from the following shortcomings: first, the early bearing faults are artificially implanted, hence not always revealing the true fault mode, morphology, and signal characteristics; second, since the noise reduction & early bearing fault characteristic enhancing algorithms have mainly been developed and validated using data collected under artificially implanted faults, the validity of those diagnosis algorithms is questionable. This paper tries to address those 2 issues. Bearing testing started with brand new and perfectly healthy aero-engine bearings, under multiple times of the typical aero engine load spectrum cycle. Continuously repeating this load spectrum cycle during the test naturally induces early bearing defects, providing the much needed “true failure” test data. The effectiveness of 2 typical modern fault-signal-enhancing algorithms: Maximum Correlated Kurtosis Deconvolution (MCKD) and Fast Spectral Kurtosis (FSK) method is then assessed for early aviation bearing fault, using the artificial implanted fault data and the “true failure” test data collected in this study. Finally, the optimal diagnosis method is proposed. The analysis demonstrates that the aviation bearing early fault progress can be reflected by the change trend of averaging magnitude index at bearing characteristic frequencies.