Selective Maintenance Decision Optimization Based On Reinforcement Learning

With the development of mechanical and electrical equipment in the direction of large-scale,complexity and precision.It poses a new challenges to the modeling and optimization of complex system maintenance decision-making problems,and it is urgent to carry out research on the optimization of complex system maintenance decision-making.At the same time,in many industrial and military application environments,maintenance resources are often limited,and traditional maintenance decision-making methods are inefficient in solving large-scale complex systems.It is of great practical significance to study the optimization of selective maintenance decisions under limited maintenance resources.Aiming at the problems that need to be solved in the maintenance decision-making of complex systems,this paper focuses on the selective maintenance decision-making method of complex systems.The main contents and innovations points are:(1)A selective maintenance model of complex system considering the uncertainty of break duration is proposed.The model combines the uncertainty of break duration and multiple imperfect maintenance actions to construct a stochastic optimization decision model.The case show that the proposed selective maintenance model in this paper has advantages compared with the selective maintenance strategy that does not consider the uncertainty of break duration.(2)A selective maintenance decision solving method for complex systems based on reinforcement learning algorithm is proposed.The selective maintenance decision-making process is represented by the MDP(Markov Decision Process).The reinforcement learning algorithm is used to solve the optimal maintenance strategy with the decision-making goal as the reward.Finally,simulation case verification shows that the reinforcement learning method proposed in this paper has better computational efficiency than traditional methods,especially for large-scale complex systems.Moreover,it further verifies the superiority of considering the uncertainty of downtime.(3)Aiming at the fault coupling of complex systems,a selective maintenance decision model for load-sharing system considering the failure dependency is proposed.Existing selective maintenance studies mainly focus on component level failure rates,and tend to ignore the influence of working conditions and other components on their degradation.The model presented in this paper considers the effect of load on component failure rate and considers failure dependency between components.The effectiveness and advantages of the proposed model are verified by simulation cases.