【作者】 杨婧；

【导师】 陈英武；

【作者基本信息】 国防科学技术大学 ， 管理科学与工程， 2012， 博士

【摘要】 大型工程项目是一个典型的复杂系统，由成千上万个在时间与空间上相互影响、相互制约的项目任务所构成，同时涉及数量众多且关系复杂的项目参与方。大型工程项目任务及组织规模的增大，以及相互关系复杂性的急剧增加，导致了大型工程项目不确定性因素增多、管理难度加大、管理效率低下等问题。如何描述任务之间、组织之间特别是任务与组织之间的相互作用关系，如何在复杂的关系分析中对大型工程项目进行有效的建模，如何在各种不确定条件下从数量众多且关系繁杂的大型工程项目系统中甄别出关键要素以进行重点管理，进而为优化项目资源配置、提高管理效率提供科学的管理和决策依据，成为大型工程项目管理亟需解决的问题。本文旨在以定性定量相结合的综合集成方法论为指导，通过系统分析方法建立大型工程项目的网络分析模型，并在此模型的基础上，借鉴相关的网络建模理论与方法，从系统、整体的角度辨识和分析对大型工程项目进度具有重要影响的关键组织和任务活动节点，并针对大型工程项目实施过程面临的各种风险因素，考虑在项目完成时间不确定的情况下，依据网络拓扑结构和系统相互作用关系为项目管理者进行关键节点管理提供决策辅助。本文将网络建模与分析方法引入到大型工程项目管理中来，不仅拓展了传统的项目管理理论，并且为管理者能够在项目实践中抓住任务规划与组织管理工作的重点、提升项目管理人员的宏观统筹与总体规划能力、以及提高项目综合管理能力提供了有效的技术方法支撑。论文主要研究工作及创新点如下：1、提出了大型工程项目系统组织——任务相互作用网络模型。把大型工程项目中的各要素集成起来，综合考虑项目任务与组织之间的相互作用关系，构建了大型工程项目组织——任务相互作用网络模型，解决了在工程项目管理中将组织和任务独立开来的问题。基于该模型，可以用社会网络理论中的各种网络指标，对项目组织和任务节点的重要性进行评价，弥补了传统的关键路径法、计划评审技术在项目任务分析中存在的不足；此外，可以将组织与项目结合起来，使得在实施工程项目的动态过程中综合衡量和分析组织节点的重要性成为可能。2、提出了一种基于加权边分解的虚节点算法，解决了网络介数的计算复杂性问题。首先，定义并分析了网络指标在大型工程项目网络模型中的实际意义与作用并给出了计算方法；然后，为解决大型工程项目相互作用网络介数的计算复杂性问题，提出了优化计算网络指标的虚节点算法，改进并优化了介数计算方法。解析并推导了网络平均度、平均边权值与基于虚节点算法的介数优化算法复杂度的关系；通过数值仿真比较在Brandes算法中使用虚节点算法和Dijkstra算法的效率，验证了新算法的时间效率要优于Brandes算法；本算法为后续的网络指标分析奠定了计算基础。3、提出了基于流度的关键任务节点分析方法。为分析项目网络拓扑结构对关键路径的影响，提出了新的更适应描述项目网络任务节点重要度的网络参数——流度。在分析中，采用项目网络生成器RanGen2，生成了1600个不同拓扑结构的项目网络，分析了网络节点的介数、入度、出度、流度等与项目关键路径的相关性，结果表明，流度与项目关键路径存在最大的相关性，当项目完成时间信息不确定时，以度值、介数和流度值为基础计算得到的关键路径与项目实际关键路径拟合度能达到40%-80%以上；而且，通过非关键路径灵敏度分析的仿真结果也表明，流度较大的非关键路径节点对项目完成时间的影响最大。4、提出了组织失效的概念，建立了一套分析组织失效、鉴别关键组织节点的新方法。基于本文提出的大型工程项目组织——任务相互作用网络模型，从解析和仿真的角度分别分析了在随机型组织网络结构、集权型组织网络结构和金字塔型组织网络结构上，组织节点随机失效和优先失效对项目工期的影响，进而为组织节点重要度排序提供了参考依据。研究结果表明，只需一定比例的组织节点优先失效即可导致整个网络完全瘫痪，从而造成工期严重延误。当组织中较少比例节点优先失效时，项目工期将明显多于同样比例节点随机失效的情况。更多还原

【Abstract】 The large-scale engineering project is the typical complex system, whichcomposed of thousands of interdependent tasks and participates. The increasing of sizeand complex relationship among tasks and organization are bringing about the increasedproject uncertainties, management difficulties and low efficiency. How to describe theinteraction relationship between tasks and organizations, how to model the large-scaleengineering project, how to identify and practice key-point management on the criticalelements to optimize project resource, improve the management efficiency and decisionmaking, come into being the urgent problems of large scale project management.Using the system analysis methodology and guiding by the integrationmethodologies of qualitative and quantitative, this paper aims to establish the networkanalysis model of large-scale engineering project. Based on the integrated model, usingfor reference from networks modeling theory and methodology to identify and analysisthe critical organization and tasks nodes that having the important influence on theproject scheduling. In view of various risk factors during project implementationprocess and considering the conditions of the uncertain project completion time, thispaper uses the network topology structure and system interdependent relationship toprovide decision aiding for managers to practice key-point management on the criticalelements. Introducing the networks model and analysis methodology into large scaleproject management, not only expand the traditional project management theory, butalso provide the technical method and support in seizing the focus of the task planningand organization management, improving the macro overall planning capability ofmanagers.The main results and contribution of this dissertation are as follows:1. The organization-task interdependent network model is proposed. Toovercome the shortage of treating the organizations and tasks separately in the researchof project management, the organization-task interdependent network model isestablished. Based on the new model, the importance of project tasks and organizationscan be evaluated using the network index of social network theory, which makes up thelack of traditional theories, such as critical path method and program evaluation reviewtechnique in project tasks analysis; What’s more important is that the new modelintegrates the project participant organizations, making it possible to analyze andevaluate the importance of project tasks in a systematical way.2. The virtual nodes algorithm based on the decomposition of weighted edgesis proposed to solve the calculation complexity of betweenness centrality. At first,the practical meaning of network index in large scale project is defined and analyzed, and the calculation method is provided. Then, in order to solve the calculationcomplexity of betweenness centrality, the virtual nodes algorithm based on thedecomposition of weighted edges is proposed, which improve and optimize thecalculation method of betweenness centrality. The relationship between network meandegree, average edge weight and time complexity based on the virtual nodes algorithmis analyzed and deduced. By numerical simulation, the calculation efficiency ofbetweenness centrality with Brandes’ algorithm with either the traditional Dijkstra’salgorithm or the virtual node algorithm is compared, and the time efficiency of newalgorithm is validated. The new algorithm lay a foundation for subsequent analysis ofnetwork index.3. Analysis method of Critical task nodes based on Flow degree is proposed. Anew network measurement for analyzing the correlation between project networktopology and critical paths is proposed.1600different project networks are generatedby RanGen2and the correlation between nodes’ betweenness centrality, indegree,outdegree, and flow degree are analyzed. Simulation results reveal that when the taskduration is unknown, the match ratio could achieve as high40%-80%for critical pathestimation based on indegree, outdegree and flow degree. Flow degree shows thehighest correlation with critical paths, and it has the most important effect on the projectcompletion time for nodes on the non-critical paths as well.4. The concept of orgazniational malfunction is proposed, a new method toanalyze orgazniational malfunction and indentify critical orgnaization nodes isestablished. By modeling the interacting dependencies between project participantorganizations and project tasks in large-scale engineering project system, wetheoretically analyze the effect of organizational structures, e.g., random structures,centralized structures and hierarchical structures, on project completion time areanalyzed. Results show that the project completion time is much higher when certainproportion of organizations is under priori failure than random failure. What’s more, thewhole network will collapse down with relatively small proportion of nodes under priorifailure. The assumptions and methods used in the cascading failure modeling enable usto provide means for evaluating the criticality of organizational nodes in large-scaleengineering systems.更多还原