【作者】 张晶；

【导师】 张云生；

【作者基本信息】 昆明理工大学 ， 冶金工程控制， 2009， 博士

【摘要】 分布式实时软件强调信息流的实时交互和并行处理,支持实时并行的质量过程,适用于钢铁行业的生产过程质量控制。软件主要特征是完成不同功能的软件构件分布在网络的不同节点上,节点之间包含实时交互的并行信号流,设计时不但要求系统的功能正确性,而且要求计算这些功能结果的时间要及时。现代钢铁企业质量过程控制系统以质量信息流的跟踪控制为中心,具有实时性、分布性、并行性等复杂行为特征,需要分布式实时软件设计理论和方法的支持。在影响分布式实时软件的诸多因素中,基于实时交互规约的框架模型是软件运行过程执行效能的关键。传统软件框架模型多侧重于利用面向对象技术的对象调用和状态转换传送控制作用,不能实现从系统建模到软件设计的一致性实时并行约束。而面向执行体模型通过变换数据相互作用,更适应于框架模型的实时并行处理。本论文基于云南省自然科学基金项目,针对钢铁质量过程控制领域,在模型架构定义中,由面向对象转向面向执行体；通过分离以数据为中心的功能构件和以时序为中心的交互机制,建立面向执行体结构的框架模型来支持分布式实时软件的设计和开发。面向执行体结构的设计机制强调构件间的并行处理和时序调度,是一种适应实时模型关系的可行技术。在分布式实时软件的框架模型中,引入这种机制,将其具体应用到框架模型研究涉及的各项关键技术中,能够较好地规约软件所要求的各类行为特征,尤其是对功能性业务逻辑和时间约束特征的关注分离,并最终实现可构造的、易于动态适应的分布式实时软件框架模型。面向钢铁质量过程控制的分布式实时软件框架模型研究,主要围绕实时交互框架、构件设计模型和构件组装规约三个研究点展开,具体为面向执行体的框架模型建立、执行体结构的构件语义设计、执行体构件的组装规约分析以及应用实例验证：1.面向执行体的框架模型研究主要建立一类面向钢铁质量过程控制,适应分布式实时应用环境的执行体交互机制,其能分离执行体之间的逻辑交互关系,支持实时并行的质量过程控制。在框架模型中,借鉴构件和连接子思想,将执行体设计为参数化行为封装,提高执行体的重用性；独立的监控器机制控制执行体之间的逻辑交互,支持并行信号流的跟踪控制；执行体端口行为优先级可配置,提供对实时性的灵活支持。这种框架模型使得以数据流为中心的执行体功能逻辑和以时序调度为中心的交互逻辑相分离,增强框架的灵活性与适应性。2.执行体结构的构件语义设计主要研究基于谓词逻辑的执行体语义模型,其能规约执行体的静态结构和动态行为；封装执行体内部逻辑,抽象质量过程的操作规范和质量要求；建立执行体端口间行为连接,保障执行体的动态特性；提供执行体包含机制,支持执行体的任意嵌套。该设计模型不仅能够描述执行体质量过程参数的属性规约,还能满足实时并行的端口行为建模。3.执行体构件的组装规约分析主要研究小粒度、低层次内部执行体分层递阶组合为大粒度、高层次复合执行体的组装理论,其具体分为复合执行体属性规约推导、复合执行体端口规约推导、和复合执行体行为规约推导三个部分,用于执行体模型组装时的语义行为描述和层次化抽象。同时从规约层面支持执行体的演化,降低执行体组装的不稳定性。4.基于上面的相关研究成果,面向昆明钢铁集团有限公司质量过程控制实际,构建了执行体结构的分布式实时软件框架模型,实现对复杂钢材形成过程的有效质量控制。基于该模型开发的软件系统配置灵活、易于组装,通过实践证明了面向执行体的软件框架建模方法可以为钢铁企业质量过程控制的应用起到很好的促进作用。本文的关键问题包括：1.构建可构造的、易于动态适应的执行体框架模型。解决构件间的并行处理和时序调度问题,动态灵活地表示交互逻辑,支持实时并行的质量过程；并要求框架可定制,能动态优化、插入和升级,不损害平台独立。2.设计完备的执行体语义模型。重点解决实时执行体的静态属性和动态行为建模；进而定义端口行为连接,描述执行体间的通信过程,确定复杂实时交互行为,适应更广泛意义上的时间并行模型。3.分析执行体的组装规约推导规则。重点解决大粒度复合执行体的属性、端口和行为规约推导,从语义层面减少实时并行的质量过程建模复杂性。更多还原

【Abstract】 In order to manage quality information better in steel manufacturing process, distributed real-time software is aimed at providing a fundamentally support of real-time and concurrent quality process behavior by the exchange of data flow to remarry time, concurrency and controlled complexity. Software focuses on developing functionality processing and timing constraints of complex multi-node on a distributed platform that has real time, interactive, and concurrent quality data flows. Quality process control system in steel industry, by the track and control of quality data, must be distributed, real time, and concurrent. It is inevitable to need the support of software design methodology. Key to distributed real-time software is research in framework model based on real-time interaction specification.Traditional object-oriented software model focuses on the transfer of control flow through method call and state transition, and the tight timing and concurrent correspondence between system modeling and software design is often lost. Effective framework modeling must be based on actor-oriented design that is interacted by the exchange of data flow to remarry time, concurrency and controlled complexity. With the support of Nature Sceince Foundation in Yunnan Province, P.R.C, the main effort of this dissertation here, aimed at quality process control domain in steel industry, is to separate timing attributes from functional properties, based on actor-oriented model rather on object-oriented model, and introduce actor-oriented.framework model to promote distributed real-time software design and development.Key to actor design is research in concurrent models and real time scheduling among actors. It is a feasible technology of adapting to real time model. Introducing this mechanism to the framework model of distributed real-time software and specifically applying it to each essential technology, can well specify and realize all kinds of behavior characteristics required by the software, especially give more cocncern of separation on functional logic and timing constraint characteristics in distributed real-time software, finally making realization of a flexible and adaptable distributed real-time software framework model. The research of framewok model in distributed real-time software based on steel quality process control mainly revolves three points:real-time framework, actor model and composition specification. And the research includes four parts:the establishment of actor-oriented framework model, the design of actor semantics, the analysis of actor composition specification and the verification of application in steel industry. Some contributions are shown as follows. Firstly, actor-oriented software framework model mainly studies a kind of actor interaction mechanism adaptable for distributed real-time application in steel quality process control domain. It can separate the interactive logic among actors, and support real-time and concurrent quality process. The framework merges advantages of ideas of component and connector, regards actors as an encapsulation of parameterized actions, and imporves the reusability of actor to a great extent. Independent monitor mechanism can control the interactive behaviors among actors and support the track of concurrent signal flow. Software framework can support real-time behavior flexibly by configuring port behavior priority. The model separates interactive attributes from functionality properties, and enhance the flexibility and adaptability of the framework.Secondly, actor semantics design mainly studies actor semantics model based on predicate logic. The model can specify static properties and dynamic behaviors of actors. Encapsulation of actor’s functional logic is a particular kind of abstraction that emphasizes specification and requirement of quality process. Port behavior correlation ensures the dynamic characteristics of actors. Composite actor can be contained by another composite actors, so the hierarchy can be arbitrarily nested. Our actor model not only specifies properties of quality process parameters, but also supports port behavior modeling to remarry time and concurrency.Thirdly, one fundamental problem of actor composition specification analysis is how to decompose a coarse-grained composite actor at higher level into more manageable fine-grained subactors at lower level from a view of specification. It is divided into three parts:properties specification, port specification and behavior specification, which are used to semantic analysis and hierarchy deduction when implementing actor composition. Meanwhile it supports the actor evolution from semantic level, and reduces the instability of actor composition.Lastly, based on the work above, a framework model of actor-oriented distributed real-time software in Kunming Iron&Steel Co.Ltd.(KISC) quality process control, which effectively supports quality process control of complex steel manufacturing, is designed at the end of the dissertation. The resulting control software based on framework model is configurable, liable to compose on a different platform. Practice of control software proves that actor-based software modeling methodology, which is explored in this dissertation, can bring about the gains of productivity of steel industry quality process control.The key aspects of our work are as follows:At first, we establish actor-oriented software framework model which is adaptable in the dynamic environment, liable to compose and reuse. Software framework can manipulate concurrent processing and real time scheduling among actors, denote interactive logic flexibly, and support quality process to remarry time and concurrency. In addition, framework model can be customized, liable to optimize, plug and upgrade, independent from the platform.Moreover, we design formal actor semantic model which focuses on static properties and dynamic behavior of real-time actor modeling. Port behavior correlation defines communication among actors, ensures real-time interactive behavior, and adapts universal timing and concurrent model. Finally, we analyze actor composition specification at a theoretical level which emphasizes on deducing properties specification, ports specification and behavior specification of composite actor. Actor composition specification reduces the complexity of timing and concurrent quality process modeling from a view of semantics.更多还原