A Fuzzy Logic Approach to Remaining Useful Life Estimation of Ball Bearings

Abstract

The paper deals with the development of a modelling and prediction scheme capable of estimating a remaining useful life of ball bearings. In particular, a multiple model-based Takagi-Sugeno scheme is developed, which is able to follow the system degradation over the time and predict it in the future. Contrarily to the typical framework, multiple models designed with historical data are used to support diagnostic decisions. In particular, health status determination of the currently operating bearing is supported by the knowledge gathered from the preceding bearings, which went through the run-to-failure process. In both historical and actual bearing cases an efficient modelling scheme with low computational burden is proposed. It is also shown how to exploit it for predicting the bearings remaining useful life. Finally, the proposed approach is applied to data gathered from the PRONOSTIA Platform, designed for the purpose of IEEE Data Challenge pertaining remaining useful life estimation of ball bearings.