Joint Optimization and Termination of Sequential Prediction Maintenance and Spare Parts Ordering under Dynamic Remaining Useful Time Prediction

In the research of Prognostics and Health Management (PHM), prediction is a dynamic process, and the corresponding maintenance decisions are also dynamic. The problem of updating and termination decisions in dynamic sequential decision-making is a rarely studied but worthy of attention problem. To address these problems, a joint decision-making and termination problem for sequential prediction maintenance and spare parts ordering under dynamic remaining life prediction was studied. The Bayesian update and EM algorithm were used in combination to achieve dynamic remaining life and reliability prediction. An adaptive detection and sequential prediction maintenance strategy based on dynamic prediction results was developed, and a judgment criterion for decision termination was given. At the adaptive detection moment, the system balanced the cost of spare parts storage and stockout, preventive maintenance cost, and post-failure maintenance cost, and determined the optimal maintenance and spare parts ordering time. Based on the termination criterion, the truly executed maintenance and spare parts ordering time were selected from the dynamic redundant decision results. Finally, the effectiveness of the proposed strategy and method were verified through experiments.