Rolling element bearing fault diagnosis under slow speed operation using wavelet de-noising

Rolling element bearing is widely used in rotating machines and any bearing defect therein can severely affect the overall machine performance. Vibration monitoring is often used to detect bearing faults. However, some rotor dynamic systems operate at very slow speed such as the slewing bearing used in dumpers, rolling mills, and cranes etc. At slow operating speed, the fault related signal features get smeared/masked due to slip, vibration of other machine parts, noise and disturbances. Thus, it becomes very difficult to identify the bearing characteristic frequencies (BCFs) from the vibration signature. In this paper, a novel diagnosis scheme based on envelope analysis and wavelet de-noising with sigmoid function based thresholding is used to extract the fault related symptoms from noisy vibration signatures of defective ball bearings operating at slow speed. The vibration signal is assumed to be composed of a deterministic part representing large scale features, a stochastic part and noise components. After sigmoid function based thresholding of the wavelet coefficients, a Bayesian estimator is used to obtain an approximation of the large scale features in the signal. The uncorrelated noise component in the signal is removed whereas the high-frequency structural ringing of the bearing induced due to impacts with the faults is retained in the large scale features. The envelope spectrum of the large scale features is used for fault diagnosis. The developed diagnosis scheme is tested using the experimental data collected from a machine fault simulator system. (C) 2017 Elsevier Ltd. All rights reserved.