【作者】 黄毅；

【导师】 郑力；

【作者基本信息】 清华大学 ， 管理科学与工程， 2011， 博士

【摘要】 随着经济全球化进程的全面推进，以机械化和自动化为特征，以规模经济为战略的传统制造企业正逐步转型为以信息技术和先进制造技术为依托，能快速响应市场波动和技术创新的现代制造企业。制造执行系统（MES）是连接制造企业上层管理和底层生产的“信息枢纽”，必须具备强大的实时监控能力，快速响应各类“意料中”的生产状态和生产异常；必须具备强大的快速重构能力，快速响应各类“意料外”的系统业务逻辑需求变化。因此，本文深入分析MES重构需求和监控需求，提出以模块粒度维和信息粒度维为主线的可重构制造执行系统体系结构（Reconfigurable Manufacturing Execution System Architecture，RMESA），系统研究了MES实现快速重构和实时监控的理论和方法。模块粒度维的核心是跨粒度模块体系结构，将MES解构为代表数据的实体模块、代表业务逻辑的服务模块、代表人员参与的人机交互模块、代表流程逻辑的业务流程模块、代表功能划分的领域模块、以及代表外部系统通讯通道的接口模块，通过规范各类模块的组织、耦合与设计模式，通过使用元语言以及代码生成、代码复用、代码模板等方法，通过持续提炼模块通用元素，通过使用重构需求分解方法，提升MES构建与重构的效率和质量，减少其工作量和复杂度。信息粒度维的核心是基于复杂事件处理（CEP）和模型驱动诊断（MBD）的事件处理框架，近年来由于RFID技术趋于成熟并在实时监控领域表现出极大潜力，因此本文专注于研究基于RFID技术及其事件处理的生产实时监控。事件处理框架按信息聚合量（粒度）从小到大定义4类事件：原始事件、简单事件、复杂事件和状态事件，基于ALE的简单事件处理模块将设备产生的RFID标签读取事件转化为代表对象时空状态的简单事件，基于CEP的复杂事件处理模块将简单事件转化为代表逻辑现象的复杂事件，基于MBD的系统状态诊断模块则综合系统组件模型、组件逻辑约束、复杂事件、简单事件等条件，实时推理系统组件状态，产生代表生产异常或关键状态的状态事件。同时，为了优化系统状态诊断模块，提出基于树分解的改进投射质蕴含项产生算法，极大提升了系统状态逻辑模型的知识编译性能。RMESA及其RMES原型系统已应用于多个实例，仍持续研发并产业化中。更多还原

【Abstract】 With the trend of economic globalization, traditional manufacturers characterizedby mechanization, automation and economies of scale are transformed into modernmanufacturers characterized by information technology, advanced manufacturing andadaptability to changes. To bridge the critical information gap between EnterpriseResource Planning (ERP) systems and device control systems, a ManufacturingExecution System (MES) requires the capability of real-time monitoring to response tothe “expected” production situations and exceptions, as well as requires the capability ofquickly reconfiguring to response to the “unexpected” requirement change of businesslogic. Therefore, a Reconfigurable Manufacturing Execution System Architecture(RMESA), composed of the dimension of module granularity and the dimension ofinformation granularity, is proposed to meet the reconfiguration requirement and themonitoring requirement of MESs.In the dimension of module granularity, a cross-grained module framework isproposed to build and reconfigure MES with higher efficiency, higher quality and lowercomplexity. Accordingly, a MES can be composed of six types of module: entitymodules that abstract data, service modules that abstract business logic, human-machineinteraction modules that abstract the participation of humans, business process modulesthat abstract process logic, domain modules that abstract functional organization andinterface modules that abstract communication with external systems. In addition, theorganization, the coupling and the design patterns of these modules are normalized;meta-language technology such as code generating, code reusing and code templatingare deployed; common module elements are being collected; and a reconfigurationrequirement analyzing method is also proposed.In the dimension of information granularity, an event processing framework, whichis based on complex event processing (CEP) technology and model-based diagnosis(MBD) technology is proposed to support RFID-based real-time monitoring.Accordingly, there are four types of event defined: tag-read event, simple event,complex event and state event, from the minimal granularity to the maximal granularity.The simple event processing module, based on the Application Level Events (ALE) standard, is designed to transform machine-generating tag-read events into simpleevents that represent the temporal and spatial status of objects. The complex eventprocessing module, based on the CEP technology, is designed to transform simpleevents into complex events that represent meaningful occurrences. The system statediagnosis module, based on the MBD technology, is designed to reason aboutcomponent states in real time by considering the component model, the internel logicalconstraints, complex events and simple events. State events that represent productionexceptions or important situation are finally generated. As well, a tree-decompositionbased, projected prime implicate generation algorithm is proposed to greatly improvethe reasoning performance by speeding up the knowledge compilation on thecomponent state logical model.RMESA and the implemented system RMES have been applied successfully tosome cases. They are being researched, developed and industrialized.更多还原