【作者】 杨浩；

【导师】 朱剑英；

【作者基本信息】 南京航空航天大学 ， 机械电子工程， 2005， 博士

【摘要】 智能制造执行系统的研究对于制造系统的信息化和智能化有着重要的意义,已成为国内外制造领域中的热点研究课题。本文以面向敏捷企业的智能制造执行系统为研究对象,采用多 Agent 技术作为系统的实现形式,对多 Agent 系统的建模方法、系统的多 Agent 组织结构、系统中多 Agent 间的通信和协作机制、基于多 Agent 和遗传算法的分布式混合调度等理论问题和关键技术进行了研究,并编制了具有一定实用价值的制造执行系统软件。本文主要创新和成果如下: 1. 提出了基于 UML 和图转换的集成的 MAS 建模方法,详细论述了需求说明、分析和设计各个阶段的模型。它是一种能全面地描述 Agent 的特征,具有模块化、可重用性和很强实用性的系统化建模方法。 2. 系统地分析了敏捷化智能制造系统的资源模型及其 MES 域,规范化了基于多Agent 的 MES 的需求,提出了采用递归方式,对企业的组织结构和控制结构进行物理的和功能的分解的 Agent 映射方法。 3. 提出了敏捷化智能 MES 的混合多 Agent 组织结构,它将层次结构和促进器联邦结构进行了有机结合,并引入了网络域的概念,满足了敏捷化智能 MES 中 Agent的多层次、多类型和多功能的需求。同时,采用了动态逻辑制造单元来实现虚拟企业间的动态联盟以及敏捷化智能 MES 的可重构性。 4. 提出了基于多 Agent 的敏捷化智能 MES 模型,其中的 ERP 接口模块作为敏捷化智能 MES 系统到 ERP 系统的信息代理,是本 MES 系统的核心控制中心。它为 MES和底层控制与 ERP 的信息集成系统的设计提供了很好的参考价值。 5. 提出了敏捷化智能 MES 中 Agent 间的通信方式、基于 KQML 的通信机制和基于 CNP 的协作方法,扩展了系统需要的 KQML 的行为原语,给出了敏捷化智能 MES中 Agent 间的协作关系,实现了敏捷化智能 MES 系统的部分功能。 6. 提出了一种将经典调度算法和分布式调度进行有机结合,来解决敏捷化智能MES 中调度问题的方法。对 CNP 进行了扩展并采用扩展的多层 CNP 作为 Agent 的协商方式,设计了调度遗传算法的各个操作算子,并给出了该算法的步骤、算例和仿真结果,详细地阐述了敏捷化智能 MES 中基于事件驱动和周期的动态调度策略。 7. 建立了敏捷化智能 MES 的实施框架,采用 J2EE 技术和 Grasshopper 平台实现了原型系统,给出了系统的运行环境和示例界面。该系统在江西洪都航空工业集团进行了实际的生产应用,并取得了良好的效果。更多还原

【Abstract】 Intelligent Manufacturing Execution System (MES), the pop research in domestic and overseas manufacturing field, makes much sense to the manufacturing information and intelligence solution. This dissertation focuses on the multi-agent based intelligent MES in agile manufacturing enterprise, its main theoretic issues and key technologies are multi- agent system (MAS) modeling methodology, agents organization structure, agents com- munication and cooperation mechanism, and multi-agent and genetic algorithm based distributed hybrid scheduling. On this basis, the MES software with practicality to some extent is built. The main creativities and achievements of the dissertation are as follows: 1. The unified modeling language (UML) and graph transformation based MAS modeling method is presented, and the model in each requirement specification, analysis and design stage is described in detail. This method can fully abstract the agent characters, and it is a modular, reusable and available systemic modeling method. 2. The resource model and MES domain of agile intelligent manufacturing system are systemically analyzed. The requirements of multi-agent based MES are regularized. An agent encapsulation method is presented. It employs physical and functional approaches to recursively decompose the system according to enterprise organization structure and control strucature. 3. A hybrid agent organization structure of agile intelligent MES is constructed. It combines hierarchical architecture with facilitator architecture, and introduces the domain concept of network, satisfies the needs of different level agents with different styles and functions. Meanwhile, the dynamic logic manufacturing cell concept is adopted to imple- ment the dynamic alliance of virtual enterprise and the reconfiguration of this MES. 4. The multi-agent based agile intelligent MES model is presented, in which the ERP interface module, acting the information agency between this MES and ERP, is the core control center. It provides the design of MES, facility control and ERP information inte- gration system with a good reference. 5. The agent communication method, KQML based communication mechanism and CNP based cooperation method are presented. The KQML performatives are extended according to the needs of this system. And agents’ cooperation relationships and portion of the functions in agile intelligent MES are established. 6. Combining the classical scheduling algorithm with distributed scheduling, a hybrid scheduling method is presented. The CNP is extended and used to express the agent negoti- ation. The operations of the scheduling genetic algorithm are designed, and its steps, example and simulation result are represented. As part of the scheduling, the event-driven and period based dynamic scheduling strategy is described detailedly. 7. The implementation architecture of agile intelligent MES is established. As a frame- work, the prototype of this system is built by adopting J2EE technologies and Grasshopper platform, and its runtime environment and example screens are represented. This system has been applied in the real manufacturing in Hongdu Aviation Industry Group.更多还原