A hybrid approach estimates the real-time health state of a bearing by accelerated degradation tests, Machine learning

Abstract

For Remaining useful life (RUL) prediction, this article presents a paradigm that separates the whole bearing life into many health states and then builds unique local regression models for each of those states, rather than searching for an overall regression model with multiple health state assessments. A method that utilised both unsupervised learnings and supervised learning to estimate a bearing’s real-time health status is presented without previous information. The primary technology used to perform health status assessment and RUL prediction is the support vector machine. The efficacy of the suggested framework has been shown via experiments, including accelerated deterioration testing on rolling element bearings.