Novel method for rolling element bearing health assessment—A tachometer-less synchronously averaged envelope feature extraction technique

Abstract

The assessment and diagnosis of bearing health using vibration data has been a research topic of interest for many years and includes developments in an assortment of signal processing methods and classification algorithms. This paper investigates detecting bearing degradation at different levels of damage, in that estimating the bearing health at the various stages of degradation is important for predicting failure as well as making maintenance decisions. The proposed technique does not require a measure of the rotational shaft speed or bearing cage speed, which makes it very suitable in certain applications in which it is very difficult or not cost effective to measure the rotational speed. To effectively estimate the bearing health state, a novel tachometer-less synchronously averaged envelope (TLSAE) signal processing and feature extraction technique for rolling element bearing is proposed. The Tachometer-Less Synchronous Averaged Envelope (TLSAE) method consists of first using a narrow band pass filter around a calculated bearing fault frequency of interest and using the derivative of the phase of the Hilbert Transform of this narrow band signal to generate a synthesized tachometer signal that is representative of the impact due to a bearing defect. This synthesized tachometer signal is combined with the high frequency envelope method to perform synchronous averaging on the envelope signal, resulting in a defect synchronous envelope spectrum in which the frequency content is in terms of the fault frequency orders. The proposed method is further compared and evaluated with other existing methods, in particular to the traditional Fourier Transform technique, the bearing envelope analysis technique, and the empirical mode decomposition signal processing methods on the basis of whether each method provides an enhanced level of indication that can determine the health of rolling element bearings. Data from a bearing test-rig is used to facilitate the comparison and evaluation of the signal processing methods. Vibration data was collected from the test-rig for bearings with different levels of degradation. The calculated vibration features from the tachometer-less synchronously averaged envelope (TLSAE) technique are compared to the other feature extraction techniques; with the synchronous average method providing a set of bearing vibration features that can distinguish all three levels of damage on the outer race of the rolling element bearing. Future work looks to further investigate this proposed technique for data collected during a run to failure test in order to consider its merits for early detection of incipient bearing damage and whether it provides a consistent monotonic trend from spall initiation to bearing failure.