| Co-Investigator(Kenkyū-buntansha) |
HATANO Itsuo Osaka Univ., Dep. of Precision Eng., Research Associate, 工学部, 助手 (10208548)
YAMAGATA Keiichi Hiroshima Univ., Dep. of Information Sciences, Professor, 総合科学部, 教授 (90029992)
|
|---|
| Research Abstract |
This study deals with on-line and off-line scheduling of a Flexible Manufacturing System (FMS) based on a knowledge engineering approach. To cope with the computational complexity of scheduling problems, we apply a simulation method. When some part types can be processed in a machine, a priority rule is applied to select one part type to be processed. For obtaining an efficient schedule of FMS, we construct a rule base to generate an appropriate priority rule in simulation process. To evaluate the method proposed in our study, we construct three rule-bases whose scheduling objectives are 'Minimizing production leed time', 'Minimizing tardiness time', and 'Just-in-time', respectively. As the result, we confirmed that these three rule bases achieved good performance of each scheduling object.
To cope with the difficulty of correcting a knowledge base of scheduling, First, we develop a self-tuning mechanism to adjust some heuristic parameters in the rule-base automatically analyzing the scheduling results obtained previously. Second, we divide the knowledge base into 3 modules ; schedule evaluation, scheduling policy, and dispatching module, which are represented by fuzzy rules. When the scheduling objective is changed, user can easily correct the knowledge base by revising the membership functions of the fuzzy rules in the schedule evaluation module, because the knowledges of sched-ule evaluation module might be realized intuitively.
To construct a rule base for obtaining an efficient schedule of FMS with on-line, we develop an FMS simulation system that represent uncertain events, such as failure of machine tools, repair time, and processing time. The rule base of on-line scheduling developed in our study, achieved good performance of scheduling objects.
Further research might be developed the method of automated generation of rule bases of off-line and on-line scheduling.
|
