INVESTIGATION OF FEATURES FOR BALL BEARINGS REMAINING USEFUL LIFE PREDICTION

The ability to accurately predict the remaining useful life (RUL) of rolling bearings plays an important role in the condition monitoring and maintenance of rotating machinery. Some practical challenges are related to the selection of optimal degradation features for effective and accurate RUL prediction. There are few works dedicated to the problem of selecting suitable features for RUL prediction based on physical modelling. This paper proposes a study for predicting RUL of rolling bearing considering various features and prediction methods based on physical fault crack growth modelling. Three feature sets (RMS, level crossing, multiple features (MFs)) are considered as degradation indicators. A nonlinear least squares method is used for initial parameters estimation. Bayesian method and particle filtering are applied for updating the values of the parameters of the physical model. The proposed framework is demonstrated using real test data provided by the FEMTO-ST institute. The results of two methods are compared, considering three different indicators. MFs indicator has the least error in the RUL estimation compared to other indicators. Particle filtering is found to perform more accurately than the Bayesian method when data are collected in real time.