Optimal maintenance strategy for large-scale production systems under maintenance time uncertainty

Many complex industrial systems have numerous operational states and oftentimes suffer from a variety of uncertainties. Determining how to address the uncertainty and efficiently reduce the number of system states has practical significance. This work is dedicated to developing a maintenance strategy based on the approximate dynamic programming (ADP) method for large-scale maintainable systems suffering from maintenance time uncertainties. In this work, an optimal schedule algorithm is proposed to address the optimal assignment problem of how to assign a certain number of components to several technicians, with the goal of minimizing the total maintenance time subject to maintenance time uncertainty. Thereafter, an optimal maintenance strategy for a system with a large number of states over a finite time horizon is developed based on the Markov decision process combined with ADP. To solve such a maintenance strategy, a solution algorithm is proposed where the system state reached in the next decision period is estimated by a simulation technique, and the post-decision state method is utilized to achieve the reduction of the number of system ergodic states. The results of numerical examples and a practical application demonstrate the effectiveness, high performance and applicability of the developed strategy.