Dynamic Scheduling Using the Mixture of Experiential Intelligence and Emergent Technology, and its Performance Verification Using a MES Model
| Project/Area Number |
15560098
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Production engineering/Processing studies
|
|---|
| Research Institution | HIROSHIMA UNIVERSITY |
|---|
Principal Investigator |
OBA Fuminori Hiroshima University, Graduate School of Engineering, Professor, 大学院・工学研究科, 教授 (10081267)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
EGUCHI Toru Hiroshima University, Graduate School of Engineering, Research Associate, 大学院・工学研究科, 助手 (80253566)
橋本 雅文 広島大学, 大学院・工学研究科, 助教授 (10145815)
|
|---|
| Project Period (FY) |
2003 – 2005
|
| Project Status |
Completed (Fiscal Year 2005)
|
| Budget Amount *help |
¥3,500,000 (Direct Cost: ¥3,500,000)
Fiscal Year 2005: ¥900,000 (Direct Cost: ¥900,000)
Fiscal Year 2004: ¥1,000,000 (Direct Cost: ¥1,000,000)
Fiscal Year 2003: ¥1,600,000 (Direct Cost: ¥1,600,000)
|
|---|
| Keywords | Dynamic Scheduling / Real-time Scheduling / Reactive Scheduling / Proirity Rule / Genetic Algorithm / Neural Network / MES / AGV / ニューラルネットワーク / 分散オブジェクト / 衝突回避 / デッドロック / アジャイル生産システム / ジョブショップ |
|---|
| Research Abstract |
1.A robust real-time scheduling rule for dynamic job shop has developed using a neural network. The neural network is trained efficiently by two training stages. In the first stage, the solution space is divided into several spaces and multiple networks are trained separately using a simulated annealing. In the second stage, those multiple networks are integrated into a network using a back-propagation algorithm. The trained neural network has proved to be very effective and robust by the numerical experiments.
2.Dynamic scheduling method using the mixture of a genetic algorithm and a priority rules has been proposed. In real-world dynamic scheduling environments, it is very difficult to search optimal schedules in real-time. The proposed method can effectively search good schedules based on the priority rule having high performance in dynamic environments. The method can also tune the weight between the priority rule and the genetic algorithm.
3.In the scheduling method described in (2), it was found that the priority rule plays a very important role. The dynamic scheduling method can have high performance by using the characteristics of the good priority rules that make schedules having appropriate due-date allowances. Bi-criteria scheduling for meeting due-dates and minimizing setup times can also be efficiently generated by changing the characteristics of the priority rule.
4.A MES (Manufacturing Execution System) has developed for the verification of the dynamic scheduling method. The system consists of a production system and a transportation system. The production system consists of multiple computers that can communicate through network, and emulates the parts flow in a factory. The transportation system consists of a host computer and multiple AGVs that can process the transport requests generated by the production system. The developed MES was shown to be able to test the effectiveness of the various scheduling method.
|
Report
(4 results)
Research Products
(54 results)
