LSTM-based deep learning approach for remaining useful life prediction of rolling bearing using proposed C-MMPE feature

Prognostic health management (PHM) is essential for the predictive maintenance of industrial systems, aiming to predict the remaining useful life (RUL) of system to ensure safe, reliable, and cost-effective operation of the machinery. This work proposes an innovative method for RUL prediction of bearings, by combining a health indicator (HI) proposed from the absolute cumulative modified multiscale permutation entropy (C-MMPE) feature with a deep learning long short-term memory (LSTM) model. The work also introduces a virtual health degree for bearings, using an exponential degradation pattern as the target function for the LSTM model output. Experimental validation showcases the effectiveness of proposed approach, achieving a high score value of 0.81 and demonstrating a lower mean absolute error value of 7.38 in RUL prediction for test bearings compared to conventional features and regression labeling functions. This highlights the superior RUL prediction capability of the proposed methodology.