【作者】 江资斌；

【导师】 高阳；

【作者基本信息】 中南大学 ， 管理科学与工程， 2007， 博士

【摘要】 随着虚拟企业理论的普及与其运作经验的不断丰富，加之信息技术的飞速发展，它已成为21世纪企业进行生产经营和市场竞争的主要形式。同时，虚拟企业已跨越了概念形成阶段而步入了实际运作阶段，以至继续需要相应的理论和方法为其提供指导。虚拟企业的生产计划是虚拟企业生产运作的核心内容之一，其关键在于协同。而多智能体技术体现了一种协同作用和优化作用，恰好能满足虚拟企业分布式和异构制造环境的特殊需求。论文从虚拟企业生产计划的特点以及制定和执行虚拟企业生产计划所面临的困难出发，研究并提出了虚拟企业高级计划系统理论框架和基于多智能体的虚拟企业协同生产高级计划系统模型，并以此为基础，研究了虚拟企业任务分解与分配、虚拟企业全局生产计划和成员企业内生产计划与调度等关键问题。最后结合企业的实际运作，开发了原型系统，以验证论文所提出的理论和方法。论文沿着问题—理论—方法—实践这一条主线展开并深入，主要工作和创新如下：(1)通过分析虚拟企业生产计划的主要流程和特点，以及制定和执行虚拟企业生产计划所面临的困难，构建了虚拟企业高级计划系统理论框架，并提出了一个基于多智能体的虚拟企业协同生产高级计划系统模型。虚拟企业高级计划系统理论框架主要由一体化管理和协同优化两个主体组成，其目标是快速响应客户个性化需求，最终实现虚拟企业竞争力的提升。在此理论框架研究的基础上，通过对虚拟企业协同生产高级计划系统的系统需求分析，提出了一种基于多智能体的虚拟企业协同生产高级计划系统模型。该系统模型总体上分为虚拟企业资源规划层、虚拟企业全局生产计划层和企业内详细生产计划三层，它为虚拟企业协同生产高级计划的实施提供了参考。(2)从虚拟企业的总体生产任务出发，重点研究了虚拟企业任务分解方法和任务分配模型。为支持虚拟企业资源规划层的运作，以最小化所有生产任务的制造成本为目标，探讨了制造资源选择问题的数学优化模型，并提出了一个带杂交算子的蚁群优化算法求解该模型。通过对多智能体协商原理和虚拟企业环境下多智能体协商特征的探讨，研究了适用于虚拟企业多智能体协商的形式化模型、协商协议和协商决策模型，以解决用数学规划方法不能解决的虚拟企业任务分配问题。同时，以任务的价格协商为目标，设计了一种新的报价策略，并提出了一个基于改进合同网的多智能体多步协商算法，以实现虚拟企业的“双赢”。(3)通过分析虚拟企业成员企业内部和成员企业之间的协同问题，运用PSL本体论和XML等技术实现了对流程信息语法和语义层次的描述，并规范了对工作流的描述，在此基础上，运用FIPAACL、CORBA和多智能体技术构建了基于多智能体的虚拟企业工作流协同交互模型，以支持虚拟企业的协同生产运作。而后，针对具有浮动开工时间的虚拟企业生产计划问题，提出了一个简易的计划算法。并针对虚拟企业生产计划的特点，建立了其生产任务计划的数学模型，提出了一个基于任务编号编码的混合遗传算法，充分发挥遗传算法所具有的良好全局搜索能力和模拟退火算法能有效避免陷入局部极小的优点，从而提高算法的速度和全局收敛性，最后通过数值仿真，验证了算法的收敛性和有效性。(4)从成员企业生产计划与调度问题的角度，探讨了制造系统中基于多智能体的生产计划与调度、订单能力评价和招投标协商机制。考虑订单的作业批量、作业任务次序约束、制造资源的能力约束和瓶颈约束等，以最小化制造周期为目标，提出了一个实现能力平衡的生产计划模型，设计了基于优先级编码的遗传算法对生产模型进行求解，并通过数值仿真验证了算法的正确性和有效性。同时，针对各种情况的生产任务再调度问题，讨论了相应的再调度策略和基于规则的多智能体再调度流程。(5)采用面向智能体编程(AOP)方法，开发了虚拟企业协同生产高级计划系统原型——VEAPS。通过对原型系统仿真，仿真结果表明，用多智能体技术来研究虚拟企业的协同生产高级计划不仅可行，而且有效。更多还原

【Abstract】 With virtual enterprise’s theory becoming popular, its operation experiencegradually enriching, and information technology speeding up, Virtual Enterprise (VE)has been the main organizational form for enterprises engaging in productionmanagement and participating in market competition in the 21st century. At the sametime, the development of VE has undergone the formation of conception and steppedinto the stage of practical operation. There is a strong need of pertinent theory andmethodology, which can guide the operation of VE. However, production planning isa core issue for the operation of VE in which collaboration is the key. Moreover, dueto multi-agent technology’s function of collaboration and optimization, it rightlyfulfils the special requirement of VE in distributed and heterogeneous manufacturingenvironment. Therefore, after analyzing the characteristics and difficulties ofproduction planning in VE, a theory architecture about advanced planning system(APS) and a multi-agent based system reference model on advanced planning forcollaborative production in VE are presented. Based on the model, such crucial issuesas task decomposition and assignment, global production planning, partners’ detailproduction planning and scheduling in VE are also studied. Finally, combining withenterprises’ practice, a prototype system is developed to verify the theory and themethodology that are proposed in this dissertation systematically.Our research is carried out thorough problem, theory, methodology and practice,the main content or possible creativities of the dissertation are as follows:(1) Based on the analysis of the major process and characteristics of VEproduction planning, and the difficulties in drawing and implementing VE productionplanning, a theory architecture about APS and a system reference model on advancedplanning for collaborative production in VE are presented. The theory frame isconsisted of integrated management and collaborative optimization. Responsingqucikly to the customers’ individualized requirement and promoting VE’scompetitiveness are its ultimate goals. Base on this frame, through analyzing thesystem requirement, a multi-agent oriented reference model on APS is proposed. Thereferential model covers three layers as a whole, namely, the resource planning, globalproduction planning and detailed production planning. It provides an implementalarchitecture for collaborative production planning in VE.(2) Based on the overall task of VE, task decomposition approaches and taskassigning models are studied. To support the operation of resource planning, takenminimizing the tasks’ manufacturing cost as optimization object, a mathematicalmodel on resource selection is supplied and an ant colony optimization algorithm withcross operator is presented. Then, the formal negotiation models, negotiationprotocols and negotiation decision models for multi-agent adapting to VE environment are investigated, which can be employed to solve the problems thatmathematical programming approaches cann’t in VE task assigning. Thereafter, basedon the price negotiation of task, probability function is constructed as offer andcounter offer strategies of agents and the expected price of both sides in the next turnsand their reserved price are also taken into account, an extended contract-net basedprice negotiation algorithm is put forward to make joint gain in VE.(3) Through the analysis of collaboration problems within VE members andbetween them, and the application of PSL ontology, XML and other technologies, thedescription of process information on syntax and semantic level is implemented. Sothe description of workflow is well-formed. Based on that, by employing FIPA ACL,CORBA and multi-agent technologies, a collaboration model for VE workflow isconstructed to support collaborative production operation in VE. Then, an easyplanning algorithm on fluctuated start and end time based planning problem is alsoproposed. Moreover, a mathematical model on tasks planning is presented with regardto the characteristics of VE production planning, and based on the coding of tasknumbers, a hybrid genetic algorithms (HGA) is presented. It combines geneticalgorithm (GA) excelent whole search ability with simulated annealing algorithm (SA)efficiently avoid getting into part minimum, thus the global searching ability of HGAis improved. The numerical simulation shows the good convergence and effectivenessof this algorithm.(4) On the basis of VE members’ production planning and scheduling, themethods of multi-agent based production planning and scheduling, the ability of orderdealing and the bidding mechanism of negotiation are discussed. Taken into accountthe batch of job, task order constraint, the capability constraint and bottleneckconstraint of manufacturing resource, taken minimizing the tasks’ manufacturing costas optimization object, a production planning model achieving ability balance isprsented. A genetic algorithm based on priority coding is designed to solve the modeland it is validated by the numerical simulation example. In addition, based on all sortsof problems of task rescheduling, the relevant strategy and process of reschedulingbased on multi-agent are put forward.(5) By employing agent oriented programming (AOP) methodology, a prototypesystem on advanced planning for collaborative production in virtual enterprise(VEAPS) is developed. The simulation result shows that studing collaborativeproduction planning based on multi-agent technology is both viable and effective.更多还原