Health Status-Based Predictive Maintenance Decision-Making via LSTM and Markov Decision Process

Abstract

Maintenance decision-making is essential to achieve safe and reliable operation with high performance for equipment. To avoid unexpected shutdown and increase machine life as well as system efficiency, it is fundamental to design an effective maintenance decision-making scheme for equipment. In this paper, we propose a novel maintenance decision-making method for equipment based on Long Short-Term Memory (LSTM) and Markov decision process, which can provide specific maintenance strategies in different degradation stages of the system. Specifically, the LSTM model is firstly applied to predict the remaining service life of equipment to distinguish its health state quantitatively. Then, based on the bearing residual life prediction curve, the degradation process model is constructed, and the corresponding parameters of the model are identified. Finally, the bearing degradation curve is obtained by the degradation process model, based on which the Markov decision process model is constructed to provide accurate maintenance strategies for different health conditions of system. To demonstrate the effectiveness of the proposed method, an experimental study with the full life cycle data set of rolling bearings is carried out. The experimental results show that the proposed method can achieve efficient maintenance decisions for bearings under different health states, which provides a feasible solution for the maintenance of bearing systems.