Abstract:This paper studies how to model and resolve the dynamic maintenance problem for the continuous degrading system. The main objective is to realize the maintenance cost optimization and determine the optimal maintenance action in the real-time degradation system. First, a deterioration model based on the Wiener process with Brownian motion is described to simulate the deterioration system and estimate the deterioration level. Second, a data-processing method is applied to discretize the continuous degradation into the discrete degradation state. Thus, a discrete-time finite Markov decision process (MDP) model is established to describe the maintenance decision processes. Third, a dynamic maintenance decision framework is constructed by periodic status monitoring and a customized deep reinforcement learning method. The developed decision framework determines the maintenance action by utilizing the specific degradation state at each inspection stage. Fourth, a maintenance decision agent is developed by employing a customized proximal policy optimization with an entropy regularization approach to solve the dynamic maintenance problem, in which the deep actor- critic networks, advantage function estimation method, truncated surrogate approach, and entropy regularization approach are integrated to facilitate the policy optimization processes, and the customized maintenance decision approach is also verified by simulation analysis. To illustrate the effectiveness, comparison studies are conducted and the results demonstrate that: (i) the cumulative maintenance costs are decreased and approaching the approximated equilibrium state by running the proposed maintenance decision algorithm; (ii) the proposed customized deep reinforcement learning approach for the maintenance decision problem can be implemented easily, and the efficiency and stability can be enhanced distinctively; (iii) the proposed maintenance framework can provide a useful decision tool for engineering manager.

Imperfect Maintenance Optimization of Multi-State Rolling Stocks Based on Deep Reinforcement Learning

Multistage Distributionally Robust Optimization for Integrated Production and Maintenance Scheduling

Long-term maintenance optimization for integrated mining operations

Joint optimization of preventive maintenance and production scheduling for multi-state production systems based on reinforcement learning

Maintenance optimisation of multi-unit balanced systems using deep reinforcement learning

Deep Reinforcement Learning for Maintenance Optimization of Multi-Component Production Systems Considering Quality and Production Plan

A Deep Reinforcement Learning Approach to Dynamic Loading Strategy of Repairable Multistate Systems

A Dynamic Maintenance Policy for Degradation System by State Monitoring and Deep Reinforcement Learning

Collaborative optimization of depot location, capacity and rolling stock scheduling considering maintenance requirements

A Deep Reinforcement Learning Approach for Optimal Scheduling of Heavy-haul Railway

Joint optimization of maintenance and quality inspection for manufacturing networks based on deep reinforcement learning

Deep Reinforcement Learning for Dynamic Opportunistic Maintenance of Multi-Component Systems With Load Sharing

Deep Reinforcement Learning for Distributed Flow Shop Scheduling with Flexible Maintenance

Optimization of High-Level Preventive Maintenance Scheduling for High-Speed Trains

Optimal maintenance strategy for multi-state systems with single maintenance capacity and arbitrarily distributed maintenance time

A deep reinforcement learning framework for life-cycle maintenance planning of regional deteriorating bridges using inspection data

Deep reinforcement learning for maintenance optimization of a scrap-based steel production line

Reinforcement learning-driven maintenance strategy: A novel solution for long-term aircraft maintenance decision optimization

Joint Optimization of Condition-Based Maintenance and Performance Control for Linear Multi-State Consecutively Connected Systems

Integrated train timetabling and rolling stock rescheduling for a disturbed metro system: A hybrid deep reinforcement learning and adaptive large neighborhood search approach

Optimal Selective Maintenance Strategy for Multi-State Systems Under Imperfect Maintenance