【作者】 夏文俊；

【导师】 李世其；

【作者基本信息】 华中科技大学 ， 机械设计及理论， 2011， 博士

【摘要】 概念设计作为产品设计的一个重要阶段,在产品全生命周期中具有举足轻重的作用。定性方法和数字样机作为产品设计与分析的有效辅助工具,扩展了设计思维,为产品创新提供了崭新的工具。传统的定量数字样机是在不同时间点变量的数字值集合的层次上进行系统行为推理,对于不精确和非量值的数据缺乏必要的支持。针对产品概念设计阶段存在的大量定性知识和传统定量数字样机在概念设计阶段应用中存在的不足,本文提出了定性数字样机的概念,并研究了定性数字样机一些关键技术问题。在对产品概念设计的特点、传统数字样机系统的建立与应用、定性建模与仿真理论分析与总结的基础上,本文从目标、功能、能力、技术组成四个方面对定性数字样机进行定义,并采用面向对象的体系框架构建策略实现了定性数字样机体系框架的构建。针对概念设计阶段所识别系统组成要素间所具有的有机联系和相互作用关系,以及结构化分析方法在定性知识分析上的优势,本文提出了基于结构模型化的定性数字样机建模方法。通过对系统变量间的相互影响关系进行结构化分析与描述,按照定性仿真模型中的代数约束和定性约束形式在所有具有直接因果影响关系的变量间生成相应的约束形式,并进行约束一致性、行为一致性和量纲一致性过滤,生成可直接用于仿真的定性微分方程。针对Kuipers定性仿真方法实际具备约束满足类问题特性的特点,将其看着是求解一类约束满足问题,从提高仿真效率、方便仿真建模的角度,将Kuipers定性仿真方法根据约束逻辑程序语言进行改进。根据约束求解是基于“先生成后检验”的原则,在逻辑语言的基础上添加了对约束的求解能力,使之更适合于定性知识的应用和程序实现。定性仿真由于其充分不完备性,易导致仿真出现大量冗余结果,当模型量比较大时进行配对组合、全局一致性过滤会非常困难,针对这种情况本文提出了基于强连通图和遗传算法的两种模型分解方法,根据对象系统的实际情况,按照不同的方法进行系统模型分解,并对分解之后的子系统按照其在原系统中的位置进行重新组合,构成新的简化系统。分解后的各个子系统单独进行仿真,其输出结果即为与之相连的子系统的输入。模型分解将完整的配对组合和冗余行为过滤分散在各个子系统中分别进行,所以可以极大提高复杂系统的仿真效率、降低仿真难度。在定性仿真中引入定量知识,能够减少定性仿真中由定性计算导致的模糊性、构造不同层次的系统模型、较为精确的定义系统及其行为,从而减少许多不必要的仿真分支,节省大量无谓的计算,能够很好的提高仿真效率,满足不同的应用需求。本文从扩展系统知识应用的角度,采用模糊数学方法构造定量信息与定性信息相集成的定性仿真算法,实现定性定量知识综合的数字样机。基于模糊数学的定性定量知识结合的仿真方法为有效应用产品概念设计阶段的定量信息提供了有效方法。更多还原

【Abstract】 Product conceptual design activities play an important role in product life cycle. As effective auxiliary product design and analysis tools, qualitative methods and digital prototyping can extend engineers’design mentality and provide new tools for product innovation. Traditional quantitative digital prototyping is lack of the support of non-number and inaccurate data on reason the system behavior. In fact, there is a large number of qualitative knowledge exist in conceptual design stage. Directed towards those reasons a concept of qualitative digital prototyping is proposed in this dissertation, and some key technical problems are researched.After the author analyzed and concluded the product conceptual design features, researched the traditional digital prototyping modeling and simulation theoretical, referenced the qualitative modeling and simulation theoretical, the concept of qualitative digital prototyping that defined from object, function, capability and technology is proposed in this dissertation. An object-oriented framework strategy is used to implement the framework of the qualitative digital prototyping to illustrate the structure of qualitative digital prototyping.The organic connections and interaction relationships exist between the system elements which recognized in product conceptual design stage, and structural analysis method has advantage in relationship analysis between the system elements, so a qualitative digital prototyping modeling method which based on structural modeling method is proposed in this dissertation. In this method, the interaction relationships between system elements are analyzed and described by structural modeling method first. And second, the relevant constraints which in accordance with the form of algebraic constraint and qualitative constraint in qualitative simulation models are generated between the elements which have direct causal relationships. Third, the all constraints generated in second step are filtered by constraint consistency, behavioral consistency and dimensional consistency. Finally, the qualitative differential equations which can directly use for simulation are generated.Because of Kuipers Qualitative Simulation method actually has the feature of the constraint satisfaction problems; it can be looked as solving a class of constraint satisfaction problem. In order to improve simulation efficiency and facilitate system modeling, the author use constraint logic programming language improves Kuipers Qualitative Simulation method. According to the constraint solving is based on "first generate and then to examine" principle, the constraint solving ability is added on the basis of the logical language, and it may be make it more suitable for the application of qualitative knowledge and programming.Because of qualitative simulation method is sufficient and completeness, it easy leads to the simulation contain a large number of redundant result. If the model size relatively large, the matching combination and global consistency filtering will be very difficult. Directed towards those problems two different model decomposition methods which based on strongly connected graph and genetic algorithm are proposed in this dissertation. According to the actual situation of object system, user can select different method to decompose the system model. After accomplish the model decomposition, a new simplified system model which composed by subsystems will be obtained. Each subsystem can simulate separately. Matching combination and redundant behavior filtering are distributed separately in each subsystem, the simulation efficiency will be greatly improved and the difficult of simulation will be reduced.The ambiguity which caused by qualitative computing will be reduced if we Introduce the quantitative knowledge into qualitative simulation. Synthesized the qualitative and quantitative knowledge can construct the system model at different levels, and more precise define the system and its behavior. In order to reduce the simulation branch of unnecessary, save a lot of unnecessary calculations, improve simulation efficiency, and meet different application requirements, fuzzy-based qualitative and quantitative synthesized simulation method is proposed in this dissertation. Fuzzy-based simulation method provides a method for effective application the quantities knowledge in conceptual design stage.更多还原