Tracking the severity of naturally developed spalls in rolling element bearings

Abstract

Condition monitoring of rolling element bearing is vital for condition-based maintenance (CBM) in many industries. A key obstacle at present is the ability to accurately quantify the severity of the bearing faults, which is commonly measured in terms of the bearing defect size. Limitations of previous studies in the area include: (i) most accelerometer-based approaches were developed for artificial bearing faults instead of naturally developed spalls, and (ii) a systematic comparison between accelerometers and alternative measurements is not available. Therefore, this thesis aims at obtaining effective methods to estimate and track the growth of bearing spalls. This has been achieved by both advancing the processing of accelerometer signals and exploiting the capabilities of alternative measurements. Firstly, a novel approach based on accelerometers is proposed, which utilises natural frequency perturbations to estimate spall size. By comparing it with the well-established existing methods, it was found that all methods are effective for artificial spalls, but only the newly proposed approach is successful for naturally developed faults. Then, three alternative measurements (acoustic emission, instantaneous angular speed, and radial load) are investigated and benchmarked against acceleration on UNSW’s bearing test rig. It was found that radial load was far superior in fault-size estimation comparing to all other sensors, and achieved more precise results than accelerometers with less complex processing. This was justified considering radial load as a proxy for radial displacement, whose potential was recently suggested by theoretical studies. To confirm this, in the last part of this work, actual displacement sensors (proximity probes) were installed on the bearing test rig and a larger gearbox facility. Both experiments demonstrated that the proposed displacement approach can effectively estimate the size of natural spalls, with very limited signal processing required. This thesis has therefore provided three significant novel contributions to the field of bearing fault severity assessment: (i) the development of a new acceleration-based approach, effective on natural spalls for the first time, (ii) the collection and analysis of a new and comprehensive database of alternative measurements, obtained on naturally developed spalls, (iii) the discovery of the superior effectiveness of direct displacement measurements.