Model-based Prognostic of the Remaining Useful Life of Bearings Considering Model Parameter Uncertainty

Bearings are one of the most used components in rotating machinery. They are also the components which suffer from the damage most frequently. For this reason, accurate prediction of the remaining useful life (RUL) of this equipment is very important. In this paper, a Bayesian approach is presented for the RUL prediction of rolling bearings, based on a two-stage processing of the degradation feature of vibration data collected experimentally. Threshold level crossing of the vibration signal in time is as a defect detection indicator. Then, a modified Paris crack growth model is developed for the bearing's defect propagation on slow-time and fast-time scales. Probability distributions of the uncertain parameters in the model are introduced and the Metropolis-Hasting algorithm is applied to the Markov chain Monte Carlo method for generating samples from the posterior distribution, which are used to estimate the RUL distribution and to describe different types of bearing degradation and fault growth processes.