Summary: | 碩士 === 國立交通大學 === 電機與控制工程系 === 88 === Scheduling is one of the most basic but the most difficult problem to be solved in the manufacturing system. The difficulty is that the most scheduling problems belongs to the NP-Complete combinatorial problems, for which the development of efficient optimum-producing polynomial algorithm is unlikely. Therefore, practical schedules are commonly generated by simple heuristic algorithm or experienced schedulers. The interaction of jobs, as they compete for limits resources, is not visible, and overall shop goal such as on-time delivery of jobs are beyond control. Thus, there is a press need for improving the scheduling operation in complex manufacturing environment.
Lagrangian relaxation technique has been used to solve scheduling problems. After obtaining the dual solution, list scheduling method is applied to produce feasible schedule for minor scheduling problems. However, if we use the linear incremental cost function of list scheduling to evaluate nonlinear scheduling problems for deciding the prior orders during the operation procedure, we may distort the real situation, due to the sensitive-response feature in linear analytic method. For improving this distortion, we can apply the compound incremental cost function when we use the linear method to evaluate the nonlinear problems. Although it is difficult to find out a quasi-optimal solution for the long-horizon job shop flow problems within a short time, section processing algorithm is able to both substantially reduce the CPU time and improve its cost function to some extent.
In this thesis, we are going to transform the scheduling problems into the optimal ones first, and then apply Lagrangian relaxation technique and the list-scheduling alogrithm to produce the feasible solution. As for the long-horizon job shop flows problems, we will combine section processing and compound incremental cost function to solve them and to greatly improve both the CPU time and the cost function.
|