Maintenance optimization in dynamic environment considering environment-induced maintenance delay

Most maintenance optimization studies for production systems assume stationary environmental states. In practice, however, the operating environments vary over time, changing both degradation dynamics and production reward of the system. Moreover, under certain harsh environmental states, maintenance actions cannot be executed and must be postponed. These factors impact the production reward, degradation penalty, and maintenance cost in a coupled manner, posing challenges for making optimal maintenance decisions. In this paper, we develop a maintenance optimization framework for minimizing the operational cost of a production system operating in dynamic environments with environment-induced maintenance delay. The optimization problem is formulated as a Markov decision process (MDP) over an infinite horizon, including the quantification of the threefold impact of the environment on maintenance. We establish key structural properties and prove that the optimal maintenance policy has a control limit structure. Based on these properties, we develop an improved value iteration algorithm, exploiting the monotone threshold characterization of the optimal policy and avoiding redundant action comparisons, to obtain the optimal solution efficiently. A numerical study of offshore wind turbines illustrates the effectiveness of the proposed framework and algorithm. Managerial insights are provided through sensitivity analysis.