Maintenance Optimization for Multi-Indenture Systems Based on Reinforcement Learning  
Author Qian Qian Zhao


Co-Author(s) Xiaojian Zhang


Abstract This paper studies a maintenance strategy of multiindenture systems with a modular redundant structure. In each failure path of the hierarchical system, only one unit (component, module, or subsystem) could be selected as a maintenance unit. A reinforcement learning (RL) algorithm is developed to determine maintenance actions for each decision epoch that minimizes the average maintenance cost over a finite time horizon. Therefore, based on the consideration of the maintenance units, the minimal repair, corrective replacement of faulty components, and preventive maintenance decisions are considered dynamically, and an effective solution method for dynamic maintenance strategy is investigated. The numerical experiments are conducted to illustrate the maintenance model.


Keywords multi-indenture systems; modular-level redundancy; preventive maintenance (PM); reinforcement learning
    Article #:  RQD28-327

Proceedings of 28th ISSAT International Conference on Reliability & Quality in Design
August 3-5, 2023