Development of Autonomous Decentralized Scheduling System with Optimization Function

点击翻译按钮获取中文摘要

In most plants, production instructions are given to each production section consisting of one more batch units. It is often the case that each production section has accumulated enough information about its operation and has the potential for creating a schedule of its own. If there exists an algorithm which enables a production section to create a schedule while taking into account the relationships with other sections, it may lead to the speedy development of a scheduling system with no supervisory system controlling the scheduling for the entire plant. Such a scheduling system should also be able to easily and swiftly adjust itself to various requirement changes.From the above viewpoint, we have developed an autonomous decentralized scheduling system, in which an individual scheduling system belonging to each production section generates a schedule in the following steps : 1) Generate an initial schedule, ii) Exchange the data among production sections, iii) Optimize the schedule b … More y incorporating information from other production sections. By repeating steps ii) and iii), a plausible schedule for the entire plant is gradually generated. As the schedule is independently generated at each production section, each production section can use any optimization method which is suitable for that section at step iii). It is also possible to introduce some constraints which are valid to a specific production section without changing the scheduling systems of the other production sections.In order to verify the effectiveness of the proposed system, it is applied to a large scale scheduling problem of a batch plant which has a complicated flow structure and some intermediate storage tanks. The result of the example problem shows that the proposed system has an ability to derive a feasible schedule which satisfies every type of constraints.A new algorithm of calculating the starting times of jobs is proposed for the autonomous decentralized scheduling system. In the proposed algorithm, the latter stages can generate schedules more freely from the schedules of the former stages, and the resulting schedules of the latter stages affect the scheduling of the former stages. Therefore, the proposed algorithem is more effective when the scheduling of the latter stages is important. The result of the example problems shows that the proposed scheduling system has the ability to generate a schedule by focusing on the stage critical for obtaining the good schedule of the entire process. Less