Research On Condition-based Maintenance Decision Of Multi-unit System

Maintenance is a key factor to influence the development of industry, military and aviation fields, which is closely bound up with the productivity, quality, reliability and resources expenses of equipments. The growing importance of maintenance has generated an increasing interest in the development and implementation of optimal maintenance strategies for improving system reliability, preventing the occurrence of system failures and reducing maintenance costs of deteriorating system. Therefore, in the past decades, maintenance and replacement decision analysis becomes one of hot topics discussed by many scholars and practitioners.In practice, the equipments are mostly integrated systems which consist of several functional parts such as mechanical units, electronic units and hydraulic units. The integration characteristic brings on the interaction between units namely economic, structural and/or stochastic dependencies. Then the optimal decision on one unit is not necessarily optimal for the whole system. In order to improve the practicability of maintenance models, the new policy and analytical approach should be studied to solve the optimization problem of multi-unit system. This paper focuses on condition-based maintenance for multi-unit systems in the context of economical dependences. The major contributions of this dissertation include:Firstly, to investigate the problem of optimization for the maintenance of multi-unit system based on the continuous-state deterioration model, a kind of new opportunistic policy is proposed to combine the corrective and on-condition maintenance. Then the maintenance model based on this policy is built and the iteration method to get the solution is analyzed according to the characteristic of deterioration model. This method can solve the problem of determining the optimal inspection interval of system and the opportunistic threshold of deteriorating unit, which combines the condition-based maintenance and corrective maintenance to achieve the optimization of the whole system.Secondly, since the continuous-state deterioration model result in the complex decision model and most of them has to resort to simulation tools, the discrete-state deterioration model is introduced to analyze the maintenance decision under the more complex conditions. The optimal model of inspection and maintenance for the deteriorating unit is presented with the semi-Markov decision process to solve the problem of joint optimization of sequential inspection schedule and preventive maintenance threshold. When the sojourn time of deteriorating state, inspection interval and maintenance time are the stochastic variables which follow the general distribution, the phase-type approach is proposed to preserve the analytical tractability of the transition probability matrix. To solve the problem that the PH distribution changes the state space of system, the value iteration algorithm for the semi-Markov decision process is improved to get the optimal inspection and maintenance policy. The algorithm is proved to be convergent theoretically and implemented for numerical calculation.Thirdly, two kind problems of multi-unit system maintenance based on the discrete-state deterioration model are analyzed:To achieve the optimization aim of series system consist of two types of units with different failure mode, an opportunistic policy is presented to combine the corrective and condition-based maintenance. The optimal model based on this policy is built and an iteration algorithm based on sequential decision is proposed according to the characteristic of discrete-state deterioration model. This policy and model can solve the problem of combing the corrective and condition-based maintenance in the system level when the different inspection style is taken.To analyze the optimal maintenance policy of multi-unit system consist of different deteriorating units which all take the condition-based maintenance, the multidimensional decision model based on the opportunistic policy is presented. This model realizes the combination of the inspection interval and preventive maintenance threshold of different units to reach the optimal aim of the whole system. To solve the problem of state space overextended in the multidimensional decision process, a heuristic approach is given to decompose of the multi-unit problem into several single-unit Markov decision problems.The examples demonstrate: when the economical dependence between the units of system is stronger, the opportunistic policy given in this paper is more effective in the decreasing the long-run cost rate of system.The maintenance models and analytical methods to deal with the problem of multi-unit system maintenance decision are complicated and require new technologies and methodologies. There is more work to do to make it practical. The research work of this dissertation only focus on the small parts of the multi-unit system maintenance and can provide some reference for the future work.