The Research On Multi Component Systems Maintenance Policy Considering Imperfect Maintenance

Multi component systems belong to large and complex systems.With the continuous improvement of modern industrialization and scientific technology,the demand for multi component systems in various fields is gradually increasing.The working environment and structure of a single component system within a multi component system are not completely the same,and given the same maintenance method for different single component systems,there is a tendency for significant deviation from the actual situation;When formulating maintenance strategies for multi component systems as a whole,only considering the failure caused by component aging has certain limitations,and ignoring the correlation between components can lead to unnecessary downtime and maintenance costs.In response to the above-mentioned maintenance decision-making difficulties,this article starts from a single component maintenance strategy considering fixed maintenance effects,establishes an imperfect maintenance mathematical model for complex single component systems,and finally studies the opportunistic maintenance strategy under multi component and multi failure modes.The main work contents are as follows:Firstly,a periodic maintenance strategy with minimum maintenance is developed for the single-component system.Weibull distribution is used to simulate the degradation process of the single-component system.The transfer characteristics between states after maintenance are described by Markov’s "no rear".At each maintenance moment,minimum maintenance and maximum maintenance are provided,two maintenance modes with different maintenance effects.With the aim of minimizing the long-term average maintenance cost of a single component system within a cycle,the optimal preventive maintenance cycle of a given system with given parameters and the choice of maintenance actions under this cycle are solved.Secondly,to solve the problem of ignoring state residence time in Markov model,the optimized version of semi-Markov process is used to describe the transition characteristics between degenerate states of a single component system.On the basis of the traditional imperfect maintenance method,an imperfect maintenance method with random maintenance effect is proposed,which is closer to the real maintenance condition.When formulating maintenance strategy for single-component system,mechanical valve was taken as the verification object,and the optimal preventive maintenance cycle T=44.3 was obtained,and the corresponding optimal long-term average maintenance cost was 0.573.No structure that stipulated maintenance strategy could ensure the optimal solution result.The cost optimality and structural stability of the maintenance strategy are verified by the sensitivity analysis.Finally,a mathematical model of degradation is established for multi-component systems,where the degradation state is composed of the degradation states of multiple individual components.When analyzing the degradation process of a single component,the Gamma distribution is used to describe the system’s degradation process,and the Poisson distribution is used to characterize the external accidental impacts suffered by individual components.Meanwhile,the external impacts on self-degradation and opportunistic maintenance between multiple components are considered.By organically combining deep learning and reinforcement learning,using Deep Q-Learning to develop reasonable maintenance strategies for multi-component systems,and solving the problem of "dimensional explosion" in high-dimensional input states in multi-component systems.This article provides a reasonable method for formulating maintenance strategies for multi-component systems,and provides method support for optimizing maintenance cost and improving safety in the whole life cycle of the system.