Investigation of prognostics and health monitoring for bearings under overload conditions

Rolling ball bearings are essential in the function of rotating machinery. Taking advantage of prognostics and health management techniques in the maintenance scheme of bearings can translate to increased safety, increased efficiency, and decreased costs of repairs of a system. We aim to expand on the capabilities of current bearing diagnostic and prognostic techniques, which typically rely on vibration data alone, by introducing data from multiple other sensors in our analysis. Our accelerated degradation test bed incites degradation and ultimately failure in 3/4" open steel rolling ball bearings through an external force applied in the radial direction. Temperature, shaft rotation speed, vibration, sound pressure, motor current, and external radial load are collected for multiple loading schemes to build a database of run-to-failure data for developing and testing prognostic models. After bearing failure, the failure mode is determined through post-mortem analysis of the bearing, looking for discoloration, brinelling, fatigue area, etc. Data are evaluated through time- and frequency-domain statistics as well as empirical modeling methods to identify signs of early degradation or changes in performance. Relationships among the signals allow us to create a more sophisticated analysis for detecting bearing degradation, predicting bearing failure, and estimating remaining useful life.