【作者】 王伟达；

【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， 2008， 博士

【摘要】 随着经济全球化的深入,面对激烈的市场竞争,许多制造企业采取了基于时间理念的竞争策略,新产品的快速研制和分销配送是该策略的重要核心。因此要求企业能够对动态多变的生产订单做出及时、准确的响应。在制造企业的车间内部建立一套高效的运营管理机制是解决上述问题的有效途径之一。以实现柔性和对多变市场需求作出及时响应为目标的制造执行系统(Manufacturing Execution System, MES)日益为21世纪企业所采用。在MES的研究中,如何使车间制造系统的资源和其控制结构(调度方式)快速、高效和经济地适应市场需求是MES的核心部分。对于我国大部分制造企业而言,尽管已在硬件方面营建了实施MES的条件,但还是普遍缺乏有效和适合的调度等控制技术的支持,因此,从MES的角度研究企业的车间制造系统控制结构及其复杂的优化问题,将对我国制造企业部署MES、提高企业竞争力起到有力的指导和推进作用。本文在分析国内外有关自适应调度理论及相关技术的研究成果的基础上,针对MES系统的特点以及实际需要,对在车间复杂动态环境下建立自适应调度器的关键技术进行了系统深入的研究,全文主要的研究工作如下:建立了自适应调度器的系统框架。自适应调度器由调度知识获取算法、调度知识系统和控制器三个关键部分组成。调度器采取了扩展的自适应调度策略,并使用基于模型参考自适应控制理论设计的体系架构,使之能够在运行策略和体系架构上均能满足复杂动态环境的苛刻要求。调度器还采用了基于工艺过程语言本体论的工艺规划信息表示方案,能够与车间内部其它异构信息系统无缝集成。研究了基于混合归纳学习的调度知识获取算法。将模拟退火算法作为变异算子以串联的形式融入遗传算法,构建成为混合优化方法。再使用Wrapper方式将混合优化方法和决策树学习算法结合。混合优化方法求解不同调度目标下制造系统的最优特征子集,并确定控制决策树规模的最优参数。决策树学习算法用于评价混合优化方法求解过程中染色体编码的适应度。在获取到最优特征子集和最优参数后,决策树学习算法生成调度知识。建立了基于面向对象技术的调度知识系统模型。利用面向对象建模语言描述了调度知识系统的需求模型,设计模型和实现模型,建立了反映知识系统静态结构和动态行为的各种视图,并使用对象约束语言对在设计模型中出现的类对象进行精确语义约束。设计了实施调度策略具有三阶段工作方式的控制器。将完整的调度策略实施过程划分“监测?决策?执行”三个阶段,使之与扩展自适应调度策略的运行机理相匹配。以此为指导思路,构建了具有三阶段工作方式的控制器基本结构,并设计了各功能模块。根据控制器的需要,还定义了仿真模型的组成,给出了生成仿真模型所必需的步骤,分析了仿真时搜索空间的复杂度。这种工作方式大大增强了调度策略的准确性和可实现性。开发了本文研究成果为核心的自适应调度器(AS2-SCHED)。系统支持生产调度人员在车间复杂动态的环境下直观自如地进行调度、介绍了系统的软件结构和工作流程,按照功能的划分描述系统的主要实现技术和主要界面,验证了本文提出的关键技术。更多还原

【Abstract】 In the twenty-first century, the rapid development of new products, coupled with promoted customer delivery, is the foundation of the time-based competition strategies, which have been adopted by many manufacturing enterprises. Accordingly, firms need to react adequately to dynamically changing job orders. The shop floor must posses sophisticated operation management mechanisms for various delivery criteria (e.g. throughput, cycle time, and so forth).Manufacturing Execution System (MES) has been adopted in the 21-century enterprises, which aims at realizing flexibility and response to the changing market requirements as quick as possible. In the researches on MES, how to make the resources and the control structure (scheduling method of resources) of enterprise’s production system adapt to the market requirements quickly, high-efficiently and economically is the core idea. To Chinese manufacturing enterprises, most of them have had the hardware conditions to implement agile manufacturing. However, they still lack effective and proper software technologies to support scheduling and etc. Therefore, to research control structure in the shop floor manufacturing system and their complex optimization problems from MES viewpoint will be a powerful guidance and propulsion.According to features of MES and practical requirements, and based on the research work of the adaptive scheduling techniques done by international and domestic scholars as well. This thesis provides a systematic research on the key techniques of adaptive scheduler under the complex dynamical environment in the shop floor. The research works have been done as follows:The framework of adaptive scheduler, which contains a scheduling knowledge acquisition algorithm, a scheduling knowledge-based system and a controller, was established. The scheduler utilized an Extended Adaptive Scheduling Strategy (EASS) and a Model Reference Adaptive Control (MRAC) theory-based architecture. These technologies make the scheduler adapt harsh requirements of the complex dynamical environment in the respects of strategy and architecture. The process specification language ontology-based representation of process plan information used can enable seamless integration of the scheduler and other heterogeneous information systems in a shop floor.A hybrid inductive learning-based scheduling knowledge acquisition algorithm was proposed. As a mutation operator of GA, SA was in series with GA to develop a hybrid optimization method which called GA&SA. DT and GA&SA were combined in the way of Wrapper. GA&SA was utilized to resolve the optimal subset of manufacturing system attributes and determine the optimal parameters of DT under different scheduling objectives; DT was used to evaluate the fitness of chromosome in the method and generate the scheduling knowledge after obtaining the optimal attributes subset, optimal DT’s parameters.A model for the Scheduling Knowledge-Based System (SKBS), using the object-oriented technology was presented. It employed object-oriented modeling language to describe the requirement model, design model and implementation model of SKBS, so various views were built up to explore the static structure and dynamic behavior of SKBS. The semantics of class in the design model was exactly constrained by using object constraint language.A controller that can deploy the scheduling strategy with a three-phase working approach was proposed. An entire process of the scheduling strategy was divided into three phases which contains monitoring, decision-making and execution. It matched well the mechanism of EASS. It was a guideline to design the basic structure and the function module of the controller. According to the requirement of the controller, the composition of the simulate model was researched and the procedure of the model generation was given. The complexity of the search space was also analyized. This working way strengthens the effectuation and rightness of the control strategy.A prototype system named AS2-SCHED(Adaptive Strategy and Adaptive Structure-based SCHEDuler ) is developed based on the kernel technique mentioned above. AS2-SCHED supports the scheduler in the shop floor in a natural and intuitive manner. The function model and architecture of proposed system are introduced practically. Finally, the application of the system is illustrated to validate these key techniques.更多还原