【作者】 李长仪；

【导师】 包振强；

【作者基本信息】 扬州大学 ， 计算机应用技术， 2006， 硕士

【摘要】 当前,急剧变化的市场环境使得制造企业必须以高度敏捷性来赢得竞争优势。由此,动态调度技术成为调度领域的研究热点,其理论价值和现实意义日益显著。本文以基于知识的动态调度决策机制为研究对象,构建了基于多Agent的调度实现框架,对动态调度策略、动态调度决策模型,以及与之相关的基于粗模糊集的Agent调度知识获取、融合和更新等理论问题和关键技术进行了系统的研究。1.将调度涉及的对象抽象为Agent实体,基于合同网协议构建动态调度框架;提出基于模糊Petri网的动态调度决策机制,运用规格化的条件属性知识和调度规则知识,以模糊Petri网作为推理机制,进行模糊推理决策,最终为各任务确定执行资源,实现任务的动态调度。基于模糊Petri网的知识决策模型,在性能评价的基础上,综合影响决策结果的诸因素:规则可信度、条件属性可信度和条件属性权值,通过先定性再定量的决策过程,使调度执行过程更趋合理、有效。2.通过“资源可用性”和“上相似任务”的概念定义了任务的相似性知识,根据资源对任务的可用性与任务间对资源的需求相似性,提出一种动态再调度策略,并构建了多阶段动态生产调度模型,从而避免任务的无谓中断与释放,达到减少受影响任务数量、降低任务调整成本的目的。3.基于任务和资源完成任务的成本、质量、负荷和时间等属性,通过一种从专家的行为实例数据集中学习和归纳决策知识的粗模糊方法,将模糊集中的隶属度函数作为粗糙集的属性,以模糊截集获取系统的分类知识。在此基础上构造决策表,以粗糙集理论对决策表属性进行分析;针对不同专家获取的决策表冲突问题,提出基于投票表决的知识融合策略,实现了以多位专家为对象所获取知识的整合,得到公共调度决策表。4.在分析制造系统所承受的内外更新动力的基础上,提出基于任务执行参数的反馈更新的方法,形成闭环的知识更新模型,从而实现对资源可信度评价的积累;提出基于人机交互的模糊控制更新方法,以完成决策属性、规则可信度,以及条件属性重要度的更新,使系统感知并适应制造外部环境的发展;在以上两种更新方法的基础上,构建调度知识更新器,并对其体系结构和更新过程进行了分析、描述。5.对基于知识的动态调度决策系统的设计与实现进行了探讨,建立了基于Jini技术的动态调度系统构架。结合本文的理论研究,初步实现了基于知识的动态调度决策原型系统。更多还原

【Abstract】 As essential sector of production running in manufacturing system, the study on scheduling technology holds the balance in innovation and development of manufacturing enterprise. This dissertation focuses on the knowledge-based decision mechanism for dynamic scheduling. Its response strategy to uncertain incidents and knowledge-based decision model are studied, and with Contract Net Protocol as the coordinating platform, an implementation framework is built upon the multi-agent theory. As extensive research, the dissertation also involves the acquisition, fusion and update of scheduling knowledge.1. The entities, which are involved in scheduling system, are portrayed by all kinds of agents in the abstract. Integrating these agents, the structure of dynamic scheduling is constructed with Contract Net Protocol as the coordinating platform. This dissertation presents a dynamic scheduling decision mechanism based on the Fuzzy Petri Net theory. The Knowledge-based Decision Model, which belongs to the decision mechanism, takes standardized attributes knowledge and knowledge of scheduling rules to do fuzzy reasoning. As a result, the right resource is selected for each task to be performed. Upon that, dynamic scheduling is actualized. On the basis of capability evaluation, the Knowledge-based Decision Model integrates the degrees of attributes’credibility, the weight of each attribute and the degree of scheduling rule’s credibility to make decision. In virtue of the decision-making which is carried out by qualitative analysis earlier and quantificational later, the dynamic scheduling leads to rational and effective.2. Through introducing the availability of resource and Prevailing Approximate Task, the approximation knowledge among tasks is defined. According to the availability of resources and the approximation of requirement among tasks, a new dynamic scheduling strategy is presented and a Multi-Stage Dynamic Scheduling Model is established. They avoid the needless interruption and abandonment of task, as a result, manufacturing system achieves decreasing the cost of adjusting tasks and the quantity of influenced tasks.3. Based upon the attributes of tasks and resources to accomplish these tasks: priority, cost, quality, load, time and so on, rough fuzzy methods, which extracts knowledge from the data set recording experts’action instances, are applied. In this way, the attributes defined in Rough Sets are generated from Membership Functions, and then, the grade knowledge is acquired in Fuzzy Cut-Set method. With the knowledge extracted from the data set recording experts’action instances, the decision table is built.更多还原