【作者】 胡海波；

【导师】 杨丹； 傅鹂；

【作者基本信息】 重庆大学 ， 计算机应用技术， 2012， 博士

【摘要】 非功能需求是说明软件系统或组件必须要达到的质量、运行能力等方面的需求，是软件开发过程中需要关注的关键要素。由于具有主观性、相对性、多样性和交互性等特性，非功能需求的研究已成为需求工程业界关注的热点。本课题源于国家863项目和重庆市自然基金项目，重点研究非功能需求之间的交互问题，该问题也可表述为非功能属性之间的相容和失配。在方法论层面，非功能需求框架中的柔目标依存图是专门用于处理非功能需求交互的结构化、图形化建模语言，可对非功能需求的复杂关系进行可视化建模和过程化支持，为设计决策提供支持。但它在形式化、自动化方面的能力有限，实例数据模型不具有语法和语义互操作性，难以用于机器理解和自动化推理。在问题域方面，非功能需求交互问题是系统复杂性和涌现性的体现。须在全局范围内来考察非功能需求，了解它们之间产生的相容和失配关系和机理，进而检测出失配关系，为需求权衡和设计决策提供依据。非功能需求的实现将会横切多个基本功能组件，使得非功能需求和功能需求之间产生复杂的依赖关系。功能需求之间产生的交互也会为非功能需求的可满足性和关联带来错综复杂的影响。分析非功能需求的交互和作用机理，还需要结合对功能需求交互的机理分析来展开。本文通过本体论和语义万维网技术扩展非功能需求框架和柔目标依存图在非功能需求建模和分析方面的能力，探寻本体模型中的公理化和规则化相协同的语义建模方法；实现非功能需求知识表示、语义互操作、实例模型的共享与重用。采用本体化的方法对非功能需求交互关系、相容失配的机理进行建模和规约，通过规则的推理实现非功能需求的自动化分析和验证。主要研究内容和创新性成果包括以下五个方面：①研究非功能需求交互的形式化本体建模方法，构建出使用语义万维网本体语言(OWL)描述的非功能需求形式化本体。主要工作包括：以非功能需求框架和柔目标依存图为需求建模的参考模型进行形式化建模，论述了用于非功能需求交互的概念层次模型、属性关系、依存关系建模模式、本体公理化等本体建模的技术和方法，构建出非功能需求的核心本体模型。为非功能需求概念建模和语义表示提供完整的方法论支持，奠定了形式化规约的基础。②研究柔目标依存图的语义表示方法、结构化语法特性和语义互操作性，提出了语法一致性检验规则和度量指标。主要工作是论述了基于知识管理的非功能需求语义表示框架；提出柔目标依存图符合严格无回路有向图的约束条件，依次对柔目标依存图中非功能需求的分解、关联等依存关系的结构化特性进行分析，建立了语法一致性检验规则；对实例模型的度量指标进行了分析，并定义了判定规则；通过实验研究验证了语义规则的正确性。③研究柔目标依存图中因非功能需求之间的依存关系而产生相容、失配的交互机理，提出形式化和规则化的表示方法。主要工作是对非功能需求可满足性评估过程进行分析，对评估准则进行语义建模；进而提出基于领域知识的隐式关联规则表示方法、基于波动影响评估的间接语义路径关联规则和失配模式，实现了形式化表示方法和自动化检测规则，可用于分析和检测柔目标依存图实例数据模型中潜在的失配关系，并通过实验案例研究验证了语义规则的正确性和方法的有效性。④研究功能需求与非功能需求的依存关系和形式化语义表示方法，提出用于分析功能需求的交互所引发非功能需求隐式关联和失配关系的分析方法。主要工作包括：基于特征-柔目标依存图的语法，建立了功能需求和非功能需求依存关系表示的语义表示模型；对功能需求交互的模式和机理进行了形式化的分析；提出了基于语义万维网技术的功能需求交互的自动化检测方法，分析并定义了由功能需求交互引发的非功能需求交互的隐式关联规则；最后通过案例研究验证了规则的正确性和方法的有效性。⑤研究非功能需求知识知识管理方法，将模式化、关注点分离、可扩展性、可重用等特性，融入到上述研究过程中，为方法论和技术体系在解决非功能需求分析建模问题的多样性、有效性、实用性等方面提供充分的理据。更多还原

【Abstract】 Non-Functional Requirements (NFRs) address important issues such as qualityproperties and performance for software systems, products or services. NFRs are vitalto the success of software system. Duo to the distinct nature from FunctionalRequirements (FRs), as subjectivity, relativity, diversity and interactivity, NFRsproblems are challengeable and currently acting as one of the hottest spot issues inrequirements engineering society. The works of this thesis are derived from projectsgranted by the National High Tech Program (863Program) and Nature ScienceFoundation of Chongqing. The thesis focuses on interaction problems of NFRs, a.k.a,harmony and conflict relationships among nonfunctional attributes of software system.As a methodology, Softgoal Interdependency Graph (SIG) from the well knownNon-Functional Requirement Framework is a graphical and structural modelinglanguage which is specialized to deal with NFRs interdependencies and interactions.SIG is an operative tool to model and visualize complex correlations among NFRs bypresenting a process-oriented support for NFRs refinement, trade-off and deign decision.Nevertheless, the disadvantages of SIG lie in the limitation of formalize and automation,lacing of syntactic and semantic interoperability for instance data models, as well asdifficulties of carrying out machine readability and automated reasoning.For the nature of NFRs problems, it is the reflection of system complexity andemergency. It is better to investigate NFRs with a systematic view to know the truth ofmechanism for harmony and conflict, to detect conflict as early as possible. Carryingout this task can help the developer to make decision with abundant facts for rationale.On the other hand, NFRs will be developed and implement as cross-cutting concernsamong multiple components, results in complex dependencies between NFRs and FRs.Consequently, it is essence to address the implicit correlations of NFRs which arecaucused by FRs interaction.The thesis aims to extent the capability of NFR Framework and SIG for modelingand analyzing NFRs, with the aid of Ontology and Semantic Web technologies. Thework also explores approaches to modeling NFRs in semantic level with the synergy ofontological axioms and semantic rules, to facilitate knowledge representation, semanticinteroperability, sharing and reusing of instance data models of NFRs. Furthermore, thethesis proposes formalized and ontological methods for modeling and representing NFRs interaction, as well as mechanism of harmony and conflicts. Thus leads us to a setof well defined semantic rules for automated reasoning with NFRs interaction. Thethesis contributes NFRs modeling and reasoning with five aspects, as follows.Firstly, the author proposes a formal and ontological approach to model NFRsinteraction, develops an NFRs core ontology represented with Web Ontology Language(OWL). The work starts with formalizing NFR framework and SIG which are acting asreference models for representing NFRs interaction. An Ontology Engineering processis performed then, in detailed means of concept hierarchy, attitudes derivation, patternsof modeling relations between/among NFRs, and axiomatization as well. AnOWL-based core ontology is developed to support conceptual modeling and knowledgerepresentation for NFRs.Secondly, the thesis investigates methods to represent an instance model of SIG,with analyzing its structural properties and syntactic interoperability, and exploresmeans of syntactic consistency checking as well as instance model metrics. The workbegins with an ontological analysis for knowledge-based framework of SIG to representNFRs interaction. By showing the axioms and restrictions of SIG as a strictlycircuit-less directed graph, the author discusses structural properties of softgoaldecomposition and correlation as interdependencies. Moreover, a set of structuralcharacteristics derived from Graph Theory is selected as metric factors for instancemodels. A group of semantic rules are defined then for automated reasoning, andvalidated by testing them with seven instance SIG models from literature.Thirdly, the thesis investigates and formalizes the mechanisms of NFRsinteractions as harmony and conflicts in SIG models, which are caused byinterdependencies of softgoals in the graph. The work begins with analysis of NFRssatisficing evaluation process, with semantic modeling for evaluation criteria. Therewith,the thesis proposes approach to correlation rules for implicit softgoal interdependencieswith domain knowledge, as well as rules for indirectly semantic correlation byextending so called delta evaluating method. A group of reasonable rules in SemanticWeb environment is defined as well, to facilitate analysis and detection of conflictsbetween/among NFRs in SIG instance models. An experimental study is performed then,to validate correctness of rules, and feasibility proposed methods.Fourthly, the thesis takes FRs into account, by extending currently method torepresent dependencies of NFRs and FRs in an incorporative model. We propose meanswith a set of correlation rules to detect NFRs conflict which are derived from FRs interaction. The start point of the work is to represent interdependencies of FRs andNFRs with a modeling language named F-SIG. Thus the author investigates andformalizes mechanisms of FRs interaction, proposes an ontological approach to detectFRs interaction with semantic rules. The rules for implicit correlations of NFRs causedby FRs interaction are presented then to detect potential conflicts of NFRs. A case studyon smart home system is performed, to validate the feasibility proposed methodsFinally, we put a knowledge based method for NFRs interaction into above works,with regarding to its properties as modularity, cross-cutting concerns, extendibility andreusability. The scattered works can also act as supplementary contributions for thethesis, to make the above works more feasible in real practice on NFRs problems.更多还原