Long-term RUL Prediction of Bearings with Signal Amplitude Regulation and Accumulative Feature

Remaining Useful Life (RUL) prediction is essential for the running bearings. An accurate RUL estimation can help the maintenance decision reliable. The accuracy of RUL prediction is greatly influenced by the health index feature. Traditional health index feature suffers from the long constant process and the vertical change degradation. In this paper, a regulated amplitude signal and an accumulative feature is proposed. The regulated signal reduces the vertical change degradation, which makes the prediction easily. And the accumulative feature changes the long constant process to an increase process, which makes the prediction possible at the early life time. The Support Vector Regression (SVR) model is adopt for the direct RUL prediction. In order to verify the effective of our method, the PRONOSTIA platform is used. The result shows that our method behavior better than the traditional feature method.