Degradation Monitoring and RUL Prediction of Rolling Element Bearing Using Proposed C-MMPE Feature

Abstract

Effective feature extraction and selection from vibration signals is a challenging task in bearing fault diagnosis and prognosis, as it directly reflects the health of rolling bearings. In this paper, Cumulative Modified Multiscale Permutation Entropy (C-MMPE) feature is proposed to reflect the effective degradation trend of bearing. Modified Multiscale Permutation Entropy (MMPE) indicates the effective fluctuation in the signals signifying the presence of an incipient fault. Then, the cumulative effect of MMPE is carried out to show the increasing monotonic trend of bearing health indicators (HI). Finally, the exponential degradation model is performed on cumulative MMPE to predict the remaining useful life (RUL) of the bearing. The proposed feature results on bearing reveal an early indication of fault and effectively predict the RUL compared to other features such as root mean square (RMS), kurtosis, skewness, permutation entropy (PE), and multiscale permutation entropy (MSPE).