【作者】 姜洋；

【导师】 刘文剑；

【作者基本信息】 哈尔滨工业大学 ， 航空宇航制造工程， 2010， 博士

【摘要】 制造执行系统(MES)是连接企业上层ERP系统和生产现场PCS系统的桥梁,它实现了制造企业由生产管理到生产控制的转换。随着MES技术的不断发展,以单件/小批量为特点的型号生产企业也应用了一些具有针对性的生产调度与管理系统。近年来,市场竞争日趋激烈,驱使着型号生产企业做到研制和批产并行,企业进而对MES的可集成性、协同性、敏捷性提出了更高的要求。因此,如何有效地构建适应型号生产需求的MES系统,是重要并且有待深入研究的课题。本文在分析了MES系统的发展趋势,以及国内外关于制造执行平台各项关键技术发展现状的基础上,对型号生产的特点和实际需求进行了探索,利用计算机集成制造技术和人工智能技术提高型号生产各过程的工作效率,进而从总体上缩短生产周期,降低废品率和生产成本,提高企业研制和批产的并行能力。全文主要研究内容如下:建立了基于多Agent的协同制造执行平台体系结构。为实现型号生产过程的集成与协同,引入了多Agent技术。面向制造系统设计了一种多Agent的分布自治控制模式,进而给出了制造系统下Agent的结构、通讯方式和语言,以及多Agent间的协同机制。在以上技术基础上,构建了协同制造执行平台体系结构,并设计了基于虚拟制造单元与Web服务技术的制造资源组织模式,以及基于PSL和XML的制造信息表达模式。研究了基于本体技术的型号生产领域知识集成方法。为满足知识优化集成和重用的需求,建立了型号生产中的领域知识模型和基于本体理论的领域知识形式化表达方法,进而提出了一种基于本体的领域知识集成器。该集成器划分为本体构建、本体映射和本体合并的三个模块,规定了各模块的运行机理,并对本体映射过程给出了量化的本体相似度计算方法,以一个实际的知识集成和检索的例子,探讨了该领域知识集成器的应用性。研究了型号生产中的业务过程协调和冲突消解方法。针对型号生产的前期任务协调,给出了基于事例推理的任务分解方法和基于合同网协议的任务分配方法;针对任务执行过程中,由信息资源共享不协调造成的业务冲突,给出了冲突消解的原则与策略,进而建立了一种混合访问控制模型,该模型将PBAC与RBAC相结合,把协同制造环境中对共享资源的操作限制在具体的任务之中。给出了任务执行主体间信任度的定义、分类和计算方法,从而提出了一种授权评价方法。该方法为调解解信息共享不协调造成的业务冲突,提供了一种量化的解决方案。仿真结果表明,所提模型和方法能够制定出符合业务实际的执行序列,从而验证了对消解业务冲突的有效性。研究了面向动态车间调度的混合变临域搜索算法。将粒子群优化算法作为局域搜索引擎融入到变临域搜索算法中,构成适宜单件/小批量生产调度的改进型优化方法。在新的调度决策点,应用经训练的B-P人工神经网络,来确定改进变临域搜索算法的关键参数,使优化算法适应车间执行环境的变化,并对原有计划进行有效的重调度。通过对比和应用实例,验证了这种混合变临域搜索算法对动态生产调度的有效性。基于以上关键技术,结合实际生产对制造执行平台的需求,开发了平台原型系统,并验证了该协同制造执行平台在实际应用中的有效性与可行性。更多还原

【Abstract】 Manufacturing Execution System (MES) is considered as the bridge between ERP system of enterprise level and PCS system of job shop, which has achieved the switch from production management to production control in a manufacturing enterprise. With the development of MES techniques, project production enterprises, which with characteristic of unitary and little batch, have applied domain production management and scheduling system as well as. Recently, as drastic market competition drives project production enterprises to combine the develop mode and batch one, enterprises get higher requirement on integration, collaboration and agility of MES. Thus, it is an important problem to how to effectively construct the MES, which meet requirements of project production.According to trends of MES, features and requirements of project production was researched based on the home and aboard research work of manufacturing execution platform and and its key technologies. Computer Integration Manifacturing technology and artificial intelligence technology are applied in project production process to resolve the problems of long production period, high rejection rate and cost, and improve the parallelism of development and batch production. The main points are summarized as follows:The architecture of collaborative manufacturing execution platform based on multi-agent was established. In order to realize integration and cooperation of project production process, the multi-agent technology was introduced. A multi-agent distributed self-control mode was designed for manufacturing system. Moreover, the agent structure, communication mode and cooperation model were proposed. To support the architecture of collaborative manufacturing execution platform, the organizational pattern of manufacturing resources on VMC and Web Services and the expression pattern of manufacturing informantion on PSL and XML were designed.The ontology-based domain knowledge integration of project production was reaserched. To meet the requirement of optimization integration and reuse of knowledge, the model and ontology-based formal expression of domain knowledge were introduced. Moreover, a knowledge integration device based on ontology of design knowledge, which consists of ontology construction, ontology mapping and ontology merging at the core, was proposed. To achieve the ontology integration process, an algorithm of ontology mapping was designed. Finally, an application case was presented to illustrate the validity of the integration mechanism.The Business coordination and conflict resolution methods for project production were studied. In order to resolve the problem of pre-project coordination, a case-based reasoning task decomposition method and contract net-based task allocation protocol were introduced. Futhermore, a hybrid access control model was presented to resolve business conflicts caused by incompatible information resource sharing. The HAC model, which combines the PBAC model with the role-based one, limits the operation to public resource in a concrete task, to guarantee the confidentiality and the integrality of the collaborative environment. An authorized evaluation method was introduced for the resource access of across administration domains in collaborative environment, in order to judge, by the confidence value, whether to the dynamic authorization to users or not. Finally, the simulation experiment to the algorithm was carried on. The result of the simulation experiment meets requirements of the application so that the validity of the conflict resolution mechanism was proved.To deal with the problem of dynamic production scheduling in project production, a hybrid variable neighborhood search (H-VNS) algorithm was presented by combining dynamic characters of shop floor. Through apply the particle swarm optimization (PSO) to VNS, the low efficiency of local search was controlled and the local optimization was assured in neighborhood structure sets. In the scheduling period, the global optimization of solution was assured too by applying the improved VNS. In the rescheduling point, the analytic value of some important parameters in the impoved VNS was solved by imposing the trained B-P neural network. Finally, experimental results illustrate the effectiveness of the proposed hybrid algorithm in a variety of shop floor conditions.The prototype system is developed based on the above-mentioned key technologies and actual production requirements for the manufacturing execution platform, and the collaborative manufacturing execution platform is proved feasible and efficient.更多还原