Residual Useful Life Prediction for Rolling Element Bearings Based on Multi-feature Fusion Regression

Abstract

Residual useful life (RUL) prediction is critical in efficient implementation of condition-based maintenances for rolling element bearings (REBs). A multi-feature fusion regression method is reported for predicting the RUL of REBs in this work. In the proposed approach, locally linear embedding technique is employed to fuse original features for erecting a condition indicator of the REBs. The fused condition indicator is then modeled by an adaptive network-based fuzzy inference system for predicting the RUL. The present approach is applied to experimental data collected from an REB. Experimental results show that the proposed approach exhibits better performance than peer approaches. It is capable of accurately predicting the RUL of the REB.