【作者】 吴克寿；

【导师】 肖人彬；

【作者基本信息】 华中科技大学 ， 工业工程， 2011， 博士

【摘要】 本论文受国家自然科学基金“公理设计的扩展研究及其稳健优化设计方法”(项目批准号：50575083)支持。针对制造业信息系统软件体系结构(Software Architecture,SA)的设计、优化与评价问题,从理论和应用两方面对公理设计(Axiomatic Design,AD)进行扩展性研究,提出一种基于AD的SA构建方法,以提高SA构建质量与效率。首先,提出了融入独立公理的用例驱动SA设计方法。该方法以AD理论概念框架为指导,按照面向对象的SA设计过程,提出一种功能需求(Functional Requirements,FRs)与设计参数(Design Parameters, DPs)抽取、分解和映射的新规则,从纵向和横向两方面并行对SA进行设计。在纵向设计上,从系统的用户需求出发,将FRs和DPs分别进行用例级、活动级、交互级和类级的4级Z字形分解与映射,完成用例图、活动图、交互图和类图的SA动态和静态建模过程。同时,从横向把FRs和DPs的关系抽象成对应的数学模型,并用设计矩阵(Design Matrix, DM)来描述。在此基础上,应用独立公理对DM进行耦合性分析和处理,从功能和结构上保证各级建模的独立性。以某银行系统的登录模块为例,说明和验证了该方法的可行性和有效性。其次,提出了一种针对耦合设计的SA优化方法。在研究设计结构矩阵(Design Structure Matrix, DSM)及其优化方法的基础上,利用DM与DSM之间存在互补性的特点,对二者有效集成,把DM的解耦问题转化为DSM的优化问题；同时引入结构二叉树及遗传算法(Genetic Algorithms, GA)理论,综合应用割裂、聚类、划分3种方法对DSM进行优化；提出位置反馈量的概念,对其进行定量分析,以此作为评价DSM优化效果的度量指标。以无人驾驶汽车控制系统(Unmanned Vehicle Control System, UVCS)在信息融合和行车决策模块为例,说明和验证了该方法的可行性和有效性。再次,提出了一种基于信息度量的SA评价方法。针对SA设计方案的多选性问题,研究设计方案评价优选的处理策略,将面向对象软件度量指标与信息度量理论相结合,提出一种新的适用于软件设计迭代过程的SA设计方案评价模型。该模型考虑了与SA复杂性相关的规模、耦合性和内聚性等质量属性,结合信息公理构建多目标综合评价模型,以信息量作为SA评价的统一测度,消除了量纲和权重的影响,为SA设计方案的评价选优提供一种具有理论基础的、更为客观的综合评价方法。以UVCS视频监控控制子系统的设计方案优选评价为实例,说明和验证了该方法的可行性和有效性。最后,在本文研究成果基础上对UVCS进行了体系结构的构建,对前文所提出的方法进行实例验证,通过逐层递进的方式对本文提出的3个方法的集成效果进行了验证,同时也为本方法在实际SA构建过程中的应用提供了示范。首先在构建用例模型的基础上从纵向和横向完成用户需求到SA的映射,在此基础上设计UVCS的体系结构；然后对设计方案中存在的耦合设计进行解耦优化,得出控制系统设计的一个或多个解耦设计方案；最后针对满足功能需求的多个解耦设计方案,从规模、耦合度和内聚度3个方面对备选解耦设计方案进行度量,计算每个方案的总信息量,并进行排序,将信息量最小的设计方案确定为UVCS案例的SA最终设计结果。通过对UVCS体系结构的构建,进一步证明了本文所提出方法的正确性、有效性和实用性。更多还原

【Abstract】 This dissertation is supported by the Natural Science Foundation of China (Grant No.50575083). Aiming at the problems of design, optimization and evaluation in software architecture (SA) construction processes, an extending research is conducted on Axiomatic Design (AD) to improve the construction quality and efficiency of SA.Firstly, a SA design approach is proposed by utilizing the use case driven SA design method and independence axiom. Based on the analysis on independence axiom and use case driven design method, a new use case driven SA design approach is proposed, which is under the guidance of independence axiom and utilizes UML as the SA modeling tool. A set of rules are defined to extract, decompose and map functional requirements (FRs) and design parameters (DPs) and SA design is processed from horizontal and vertical directions, respectively. This approach starts from use case model, and then FRs and DPs are iteratively decomposed into four levels, they are, use case, activity, interaction and class levels, guided by zigzagging mapping. At each iteration, the relation between FRs and DPs are described by design matrix (DM), and coupling analysis and process are applied on DM to assure the function and structure independence of modules. A practical example, the design of the login module in the bank system, is provided to illustrate this approach.Secondly, a SA optimization approach is proposed by utilizing DM and DSM. Aiming at the deficiency that AD theory does not provide a specific decoupling method, DM and DSM are simultaneously used to optimize SA design. Through effectively integration, the problem of SA optimization is converted into the problem of DSM optimization.In this approach, a complexity metric, that is, position feedback metric is devised for DSM optimization. In addition, structure binary tree and GA are introduced, and three mechanisms, tearing, clustering and partitioning, are comprehensively devised in this approach. A practical example, the design of the information fusion module and operation decision making module in unmanned vehicle controlling system (UVCS), is provided to illustrate the efficiency of the proposed approach.Thirdly, a SA evaluation approach is proposed based on information measurement. Aiming at the feature of multiple selections on SA design solutions, a multiple attributes evaluation approach is proposed, which considers both quality features of SA, such as scale, cohesion and coupling, and information axiom. It is not necessary for the designers to normalize the criteria, but just to calculate the information content for each design solution, and then the optimal solution can be decided. In order to meet the need of information content calculation and design solution evaluation, a suit of SA metrics and the information content formulation are devised. A case study on the evaluation of the design solutions of video controlling module in the UVCS is presented to illustrate the effectiveness of the proposed approach.Finally, the construction of UVCS architecture is progressive conducted based on the above design, optimization and evaluation approaches, which can also verify the integrated effect of above approaches and supplies a demonstration of how above approaches applied to practical SA construction as well. The use case model is firstly developed based on the AD conceptual framework to reduce the design couple and the design is verified by the independence axiom. In the following design iterations, the coupling design is optimized by using the proposed optimization approach, and one or more decoupled design solutions can be obtained. The proposed evaluation approach is utilized to deal with the selection of multiple decoupled design solutions to get the optimal design. Lastly, the SA construction of UVCS is successfully accomplished.更多还原