| 英文摘要 | Although manufacturing system engineering have attracted more attentions, it is still in an initial stage comparing with electrical, chemical, and other fields of engineering. The main feature of the manufacturing system engineering is the gap between the theory and the practice. In this thesis, we take the "Advanced Manufacturing Technology Laboratory", which is a sub-project of the Engineering Research Center(ERC) for machinery building automation and is under setting up, as background of study, and take the hedging point policy whose main features are simplicity of the structure and easy to implement as the objective of study. Meanwhile, we do some discussions on the field of job sequence for the reentrant two-machine system. There are three main parts in this thesis. In first part, our research objective is a stochastic production system including two machines and a infinite buffer. The stochastic feature is due to the uncertainties for machine failure and repair. There were some researches to study the tandem machines system, but, they always supposed that the probability of the machine failure is constant in the model established, thus the capacity processes of the machines are homogenous Markov processes and only considered production rate optimization. In this paper, we introduce age, which means the micro time for a machine to being operational, in the model for this king of system and discuss the optimization of the preventive maintenance rate as well as production rate. We establish the dynamic programming model for this problem, then deduce the structure of optimal preventive maintenance rate and production rate. We show that the optimal preventive maintenance rate is the problem to balance between the effectiveness of maintenance activities and the cost expenditure for the activities, and the optimal production rate is determined by hedging surface and critical surface. But, we can not get the analytic solutions of the hedgIng surfaces and critical surfaces. In order to obtain the control policy being able to implement in practice, we attempt to get a heuristic near-optimal control policy by using the structure properties. Some approximated ways are used in discussing the heuristic policy. The continuos age changing processes were replaced nearly by the trapezium changing processes, so that the critical surfaces can transfer to hedging surface by using the consequences obtained in production system models with homogeneous Markov process. Because the hedging surfaces are functions of surplus and age, the heuristic policy, in fact, is a feedback control policy and can easy to implement in practice. Finally, we prove the heuristic policy is near optimal by simulation. In second part, since the analytic solution can not be get in multiple-part system and some approximated methods will face difficulties during dealing with beyond three dimensional system, we apply Genetic Algorithms(GAs) |
修改评论