A CUSUM-based adaptive bearing fault features tracking method for RUL estimation

This research is goal to develop a non-intrusive motor bearing fault prognostic method for remaining useful life (RUL) estimation. This technique can determine the incipient system fault in the early stage to prevent fatal failure. Considering the difficulty of requiring enough historical data for traditional prognostic methods, this research proposed a novel method with the concept of cumulative summation (CUSUM). It is utilized to select suitable fault characteristics among the sensor data to fuse a health indicator (HI) for RUL prediction. And to track the degradation trend of the HI under high noise, the recursive least square (RLS) method is introduced. Finally, 3 sets of electrical data from a motor with excessively-loaded bearings are used to verify the proposed method.