Condition-based maintenance strategy for redundant systems with arbitrary structures using improved reinforcement learning

The condition-based maintenance (CBM) decision-making for redundant systems has attracted increasing attention. Most existing studies are dedicated to k-out-of-n redundant systems and the search of the optimal maintenance policy is efficient for low-dimensional CBM. In practical applications, complex system structures and failure criteria are commonly observed, posing challenges for searching the optimal CBM policy. This paper studies the optimal CBM strategy for redundant systems with arbitrary system structures using improved reinforcement learning, considering failure and economic dependences. The decisions of imperfect repair and replacement of failed components are considered dynamically, and an efficient solution method of dynamic maintenance strategy is investigated via improved reinforcement learning incorporating re-learning and pre-learning processes. Numerical studies are conducted and the results indicate that the proposed method is effective in reducing the maintenance cost and efficient in searching the optimal CBM strategy for redundant systems.