【作者】 赵强；

【导师】 肖人彬；

【作者基本信息】 华中科技大学 ， 工业工程， 2008， 博士

【摘要】 本文依托国家自然科学基金项目(NSFC-60474077)和教育部新世纪优秀人才支持计划项目(NCET-05-0653),以任务驱动为主线,综合运用多智能体理论、博弈理论及粒子群优化算法,对虚拟企业生命周期管理涉及的合作伙伴选择、生产任务调度及信息共享三个关键问题进行研究,其主要内容如下:首先,分析了任务驱动虚拟企业生命周期管理过程及关键问题。在对任务驱动的虚拟企业生命周期管理内涵进行说明的基础上,建立了任务各阶段与虚拟企业生命周期各阶段的对应关系,并对任务驱动的虚拟企业生命周期管理过程进行了论述。此后,重点分析了虚拟企业合作伙伴选择、生产任务调度及信息共享三个关键问题的求解难点。分析指出合作伙伴选择问题的难点主要包括优化选择、谈判与签约及协调管理;生产任务调度问题的难点主要包括任务调度优化及协调管理;信息共享问题的难点主要包括信息共享平台构建、信息共享的机制设计及策略选择。其次,给出了任务驱动的虚拟企业合作伙伴选择结构模型及优选算法。利用多智能体技术,建立了任务驱动的虚拟企业合作伙伴选择结构模型。针对合作伙伴优选问题,建立了基于工期约束成本最小化的优化模型,并采用限制法证明了该问题是NP完全问题。针对问题求解,提出改进粒子群优化求解算法。改进算法新增初始扩大、变异及局部更新三个过程,上述过程的综合运用提高了粒子群算法的优化求解效果。以敏捷制造实例验证了合作伙伴选择模型及优选算法的有效性。第三,给出了虚拟企业生产任务调度的层次结构模型及调度优化算法。建立了虚拟企业及合作伙伴两个层次的生产任务调度结构模型。针对调度优化问题,以最小化生产延续时间为优化目标,建立了调度优化模型,其中虚拟企业调度层优化主要考虑生产任务之间的时序逻辑关系、作业时间及伙伴自身已确定的生产任务等影响因素,而合作伙伴调度层优化则主要考虑资源占用、转运时间等影响因素,并给出了粒子群优化求解算法。以大型设备制造实例验证了生产任务调度模型及优化算法的有效性。第四,给出了虚拟企业信息共享博弈模型及促进高层次信息共享的激励方法。以盟主及一个合作伙伴组成的虚拟企业为研究对象,建立了虚拟企业信息共享博弈模型,指出了信息共享博弈的“囚徒困境”,即盟主及合作伙伴都试图通过减少共享的信息量来增加自身利润,结果导致虚拟企业高层次信息共享无法实现,并且使得虚拟企业、盟主及合作伙伴利润降低。为实现虚拟企业高层次信息共享,从单次博弈及重复博弈两个方面提出了解决方法。针对单次博弈,提出了竞拍与惩罚机制相结合的信息共享激励机制设计方法,而针对重复博弈,提出了基于博弈策略及学习调整、增加博弈次数的方法,通过仿真验证了方法有效性。最后,基于上述研究成果,建立了原型系统,并通过300吨汽车起重机研发及制造实例验证了其可行性、有效性和实用性。更多还原

【Abstract】 This work is supported by Natural Science Foundation of China (NSFC-60474077) and Program for New Century Excellent Talents in University (NCET-05-0653). Taking task-driven as a. clue, adopting multi-agent theory, game theory and particle swarm optimization (PSO) algorithm integratedly, this dissertation focuses on the study of three key issues of life cycle management of virtual enterprise, which are partner selection, production task scheduling and information sharing. The main contents are as follows.First, this dissertation analyzes the process of life cycle management of task-driven virtual enterprises and its key problems. On the basis of expaining the meanings of life cycle management of task-driven virtual enterprise, a corresponding relationship figure between each task phase and each life cycle phase is established, and then the life cycle management process of task-driven virtual enterprise is briefly described. Thereafter, the focuses are paid on the difficulty analysis on problems’ solution of partner selection, production task scheduling and information sharing. The analysis points out that the difficulty of the partner selection mainly embodied in selection decision-making, in negotiating and contracting, and in coordination management; the difficulty of production task scheduling mainly reflected in tasks scheduling optimization and in corresponding cooperation process; the difficulty of information sharing mainly embodied in establishment of the information-sharing platform, in the design of mechanisms and in the choice of strategies when dealing with information sharing.Second, this dissertation presents a partner selection structure model for task-driven virtual enterprises and proposes a corresponding partner selection optimization algorithm. The partner selection structure model of task-driven virtual enterprise is proposed based on multi-agent technology. As for optimization problem on partner selection, based on minimizing costs within the time constraints, the study proposes a partner selection model to solve optimization problem of partner selection. Using restriction method, this optimization problem for partner selection is proved to be a NP-complete problem. To find the solution of this optimization problem, an improved PSO algorithm is proposed which includes initial enlargement, variance and local updating process. Utilizing three processes synthetically improves the optimization effect of PSO algorithm. Two examples of partner selection for agile manufacturing are presented to illustrate and prove the applicability of the presented model on partner selection in virtual enterprise and the validity of the proposed optimization algorithm.Third, a model of production task scheduling in virtual enterprises is presented and a corresponding scheduling optimization algorithm is proposed. The hierarchy structure model on task scheduling includes virtual enterprise layer and partner layer. As for scheduling optimization problem, a model for minimizing manufacturing cycle is built, in which some key factors in virtual enterprise layer are considered, such as time sequence and logic relationship among production tasks, operation timing and partner’s own production tasks, however, some key factors in partner layer are considered, such as resource possessing, transportation time and so on, and the PSO algorithm is proposed. A large-scale equipment manufacturing case is presented to illustrate and prove validity of the presented production task scheduling model and the proposed optimization algorithm.Fourth, a game model of information sharing of virtual enterprises is proposed and strategies and methods for improving high-level information sharing are presented. Take a virtual enterprise which is made up of a leader enterprise and a partner enterprise as research object, the game model of information sharing is established and "the prisoner’s dilemma" of information sharing is put forward. In the prisoner’s dilemma of information sharing, both the leader and the partner try to increase their own profit through sharing lower information with its counterpart, so the result comes out that the profit of the virtual enterprise and of each partner’s decrease and high-level information sharing is impracticable. To achieve high-level information sharing, solutions are presented both for single and repeated game. For single game, an incentive mechanism design method is presented which combines the auction and punishment mechanism, while for repeated game, a method based on game learning, strategy adjustment and increasing game times is presented. The following simulation proves that the methods are effective.Finally, based on the mentioned research achievements, a prototype system is established, what’s more, a case of the R&D and manufacture of 300-tons automobile crane is presented to prove the feasibility, validity and applicability of the mentioned research conclusions.更多还原