【作者】 李勇；

【导师】 谭跃进；

【作者基本信息】 国防科学技术大学 ， 管理科学与工程， 2009， 博士

【摘要】 现代战争的演化,尤其是上世纪90年代以后的几场局部战争的形态变化,使得物流保障网络的地位越来越重要,保障的能力、质量己经成为决定战争胜负的关键因素。同时,物流保障网络也是敌对双方在战争中攻防的焦点。因此,物流保障网络的抗毁性研究具有非常重要的理论和实际意义。物流保障网络的抗毁性不仅由网络的结构决定,而且还与网络中的负载有关。当网络中某个节点发生故障时,它上面的流量将转向其他正常节点,而这些接受多余负载的节点又可能由于负载总和大于其容量而发生新的故障。我们将这种相关失效行为称为“级联失效”,考虑网络中“级联失效”的抗毁性称为“级联失效抗毁性”。本文以复杂网络理论为指导,综合运用图论、统计物理、运筹学、概率论、计算机仿真等学科领域知识,系统深入地研究了物流保障网络级联失效抗毁性的建模、分析及其应用。主要研究工作包括:1.建立了物流保障网络级联失效抗毁性模型级联失效的本质是一种基于负载的相关失效,影响级联失效的因素包括:网络结构、负载和级联失效过程。本文通过分析物流保障网络的特点,分别建立了物流保障网络的网络模型、负载模型和级联失效抗毁性模型。从物流保障网络的结构和功能层次上,将其划分为战术保障网络和战略保障网络。用均匀随机网络表示战术保障网络的网络模型,用空间网络表示战略保障网络的网络模型,用逻辑网络表示负载模型。在分析物流保障网络级联失效过程以后,确定了路由机制、节点故障以后负载的重分配原则、网络的失效形式和级联失效抗毁性的度量参数,最后基于这些因素建立了物流保障网络的级联失效抗毁性模型。2.解析分析了典型容量分布下的战术保障网络的级联失效抗毁性容量分布是决定战术保障网络级联失效抗毁性最关键的因素,本文应用概率母函数方法和分支过程方法解析分析了三种典型容量分布下的级联失效抗毁性:相同容量、随机容量以及容量与属性相关。研究结果表明,三种典型容量分布下都存在容量参数的临界值,当容量参数小于临界值时,网络出现大规模的级联故障;当容量参数大于临界值时,网络运行良好。最后,计算机仿真实验验证了上述解析结果的正确性。3.仿真分析了任务约束条件下的战略保障网络的级联失效抗毁性战略保障网络中,既有民用运输负载,又有军事保障负载。并且,同一时刻不同空间位置上的节点负载不同,不同时刻同一节点的负载也是动态变化的。据此,本文将战略保障网络的任务约束条件划分为任务空间约束和任务时间约束,分别分析了任务空间约束和任务时间约束下的战略保障网络级联失效抗毁性。研究结果表明,不同任务约束参数条件下,战略保障网络级联失效抗毁性差异很大。本文还深入分析了产生这种差异的原因。4.实现了用于物流保障网络级联失效抗毁性分析的应用软件系统本文实现了物流保障网络级联失效抗毁性分析软件,该软件作为物流保障网络级联失效抗毁性的平台,可以输入保障网络、编辑保障任务、设置节点的负载和容量。通过该软件,可以运用本文的算法模型来分析各种保障网络实例的级联失效抗毁性。更多还原

【Abstract】 Due to the evolution of modern battle styles, especially the local warfare after 90’s of last century, logistic support networks become more and more important, and the ability and quality of logistic support become key factors that decide the success of modern battles. At the same time, the logistic support networks are the focus of attack and defense. Consequently, study on invulnerability of logistic support networks is of great theoretical and practical significance.The invulnerability of logistic support networks is decided by not only network topology, but also network load. When one vertex fails, its loads will be directed to other normal vertices, which will then fail and cause further load redistributions if the new loads exceed their capacities. This dependent failure behavior is called“cascading failure”and the invulnerability considering cascading failure is called“cascading invulnerability”. Guided by complex network theory, this dissertation studies thoroughly and systematically the modeling, analysis, optimization and application of cascading invulnerability of logistic support networks using methods of graph theory, statistical physics, operations research, probability theory and computer simulation. The main results and contributions of this dissertation are as follows.1) A cascading invulnerability model for logistic support networks is proposed.The essence of cascading failure is a dependent failure based on load. The factors that effect cascading failure include: topology, load and cascading process. By analyzing the characteristics of logistic support networks, a network model, a load model and a cascading invulnerability model for logistic support networks are proposed respectively. Based on structure and function, logistic support networks are divided into the strategic logistic support networks and the tactical logistic support networks. The strategic logistic support networks are modeled as spatial networks, the tactical logistic support networks modeled as uniform random networks, and the loads are modeled as logistic networks. After analyzing the cascading failure process on logistic support networks, the route strategy, load redistribution rule, failure form and measurement of cascading invulnerability are presented. Finally, a cascading invulnerability model for logistic support networks is proposed.2) The cascading invulnerability of tactical logistic support networks with typical distributions of capacity is studied analytically.The distribution of capacity is the most key factor that affects the cascading invulnerability of tactical logistic support networks. The cascading invulnerability with three typical distributions of capacity is studied analytically using the generating function and branching process methods, i.e., uniform capacity, random capacity and property-related capacity. The results show that there is a critical capacity for all three typical distributions of capacity. Above the critical capacity, large scale cascading failures occur; under the critical capacity, large scale cascading failures do not occur. The numerical simulations verify the validity of our analytical results.3) The cascading invulnerability of strategic logistic support networks with task constraints is studied using the numerical simulation methods.There are both civil transport loads and military support loads on strategic logistic support networks. Moreover, the vertex loads at different spatial locations are different at the same time, and the load of same vertex at different time is dynamic. Hereby, the task constraints are divided into the task space constraints and the task time constraints. The cascading invulnerability of strategic logistic support networks with the task space constraints and the task time constraints are studied respectively. The results show that the cascading invulnerability of strategic logistic support networks has remarkable difference under different task constraints. The reasons for the difference are revealed thoroughly.4) A software system for analysis of cascading invulnerability of logistic support networks is implemented.A software system for analysis of cascading invulnerability of logistic support networks is implemented as a platform, on which the logistic support tasks can be imported and edited; the loads and capacities of vertices can be set. Using the software system, the cascading invulnerability of logistic support networks can be analyzed with our models and methods.更多还原