Remaining Useful Life Prediction for bearing based on Online Oil Parameters and Vibration Signals

Abstract

As the key mechanical component of the power system, remaining useful life (RUL) prediction for bearing has attracted more and more attention. Most of the previous studies are based on the vibration signals. In this paper, RUL prediction is performed based on online oil monitoring parameters and vibration signals. First, several groups of bearing full lifetime test are carried out on the bearing life test bench, and the bearing full lifetime monitoring data such as metal debris, oil viscosity, temperature, density, dielectric constant, water content and vibration signals are collected. After the pretreatment and feature extraction of oil parameters and vibration signals, the support vector machine models for bearing wear status evaluation and RUL prediction are trained and constructed. Secondly, the trigger mechanism of bearing start-up of RUL prediction is analyzed and discussed. The RUL prediction model is verified by full lifetime test data of testing SKF6208 bearings. The results show that the combination characteristics of online oil parameters and vibration signals provides a promising way for accurate evaluation of bearing wear status and effective remaining useful life prediction.