Quantitative Fault Severity Degree Assessment for Rolling Bearings Based on Multivariate Regression Model optimization

Abstract

The relationships among fault signal resources, condition indicators and assessment models play an important role in quantitative fault severity degree assessment for rolling bearings. This paper aims to assess fault severity degree by using optimal condition indicators and regression models. Original signals of rolling bearings in various fault severity degrees are collected by available test points, then thirty condition indicators of original signals for each fault severity are calculated, which can describe the bearing fault severity; Those complex relationships among multiple test points, condition indicators and fault severity degrees are built by multiple regression models, then the assessment results with various test points and condition indicators are studied. The minimum Root Mean Square Error (RMSE) of assessment outcomes is used to select optimal condition indicators and assessment models. Finally, rolling bearing acceleration vibration signals provided by Case Western Reserve University were used to verify the effectiveness of proposed methods in this paper.