Summary: | 碩士 === 逢甲大學 === 工業工程與系統管理學研究所 === 100 === This study combines dispatching rules and tabu search for flow shop scheduling problems with buffers. Buffers of infinite and finite sizes are considered. An initial solution is constructed by means of dispatching rules. The solution is refined by a tabu search algorithm that uses local search to avoid falling into local optima. Makespan and mean flow time are considered as performance indicators.
This study uses an Exhaustive Attack method verify tabu algorithm to find an approximately optimal solution in a short time. Under the same conditions, the tabu search algorithm requires less CPU time than the Exhaustive Attack method to get a nearly optimal solution. For problems with few jobs, the schedule is quite smooth, and the buffer does not present difficulties. For a case study with an unlimited buffer, the LIFO dispatching rule produces a schedule with the best Makespan. The LPT dispatching rule produces the best schedules for problems with finite buffers that can hold one, two, or four items. The SPT dispatching rule produces schedules with the lowest Makespans when no buffer is available; the LPT dispatching rule produces schedules with the best mean flow time when no buffer is available. LPT is also the best dispatching rule if a buffer is available but can only hold one, two, or four items. The Random dispatching rule performs well for problems with infinite buffers.
The combination of tabu search and dispatching rules outperforms dispatching rules. When the variation in processing time is large, tabu search makes a big improvement; when the variation in processing time is small, tabu search makes a small improvement. The results for problems with infinite buffers showed good Makespan values, and the results for problems with no buffers showed good mean flow time values.
|