Remaining Useful Life Prediction of Rolling Element Bearings using Minimum Redundancy Maximum Relevance and Extreme Learning Machine

Abstract

Estimating the remaining useful life of rolling element bearings is essential to ensure the reliable and efficient operation of rotating machinery, as well as reduce maintenance costs and downtime. In this study, a novel methodology was used to estimate the Remaining Useful Life (RUL) of ball bearings. Data were collected from two platforms: one for testing ball bearings with artificial defects, and another called PRONOSTIA for run-to-failure tests. Noise reduction techniques (Variational Mode Decomposition (VMD), AutoRegression (AR) filtering and Bandpass filtering) were applied to the data to select manualy the features. Using the second platform's data, a Minimum Redundancy Maximum Relevance (MR2) method was used to select automaticaly the features then the extreme Learning Machine (ELM) classification model was constructed. Furthermore, an ELM regression model was developed using the second platform's data to estimate the Remaining Useful Life (RUL). The proposed feature selection method effectively prevents delayed anticipation of failure. The results provide evidence for the effectiveness of the proposed approach in enhancing the accuracy of rolling element bearing Remaining Useful Life (RUL) prediction.