Fund Project:Project supported by the National Key R&D Program of China (Grant Nos. 2017YFB0803200, 2018YFB0804002), and the National NaturalScience Foundationof China (Grant Nos. 61802429, 61872382).
Received Date:24 December 2018
Accepted Date:31 January 2019
Available Online:23 March 2019
Published Online:05 April 2019
Abstract:Modern systems are always coupled. Previous studies indicate that coupled systems are more fragile than single systems. In a single system, when a fraction of 1-p nodes are removed, the percolation process is often of the second order. In a coupled system, due to the lack of resilience, the phase transition is always of the first order when removing a fraction of nodes. Most of previous studies on coupled systems focus on one-to-one dependency relation. This kind of relationship is called a no-feedback condition. Existing studies suppose that coupled systems are much more fragile without a no-feedback condition. That is to say, if a node depends on more than one node, the coupled system will breakdown even when a small fraction of nodes are removed from the coupled system. By observing the real world system, real nodes are often dependent on a dependency cluster, which consists of more than one other node. For example, in an industry chain, an electronic equipment factory may need several raw material factories to supply production components. Despite part of the raw material factories being bankrupt, the electronic equipment factory can carry out productionnormally because the remaining raw material factories still supply the necessary production components. But theoretical analysis shows that the robustness of such a coupled system is worse than that of one-to-one dependency system. Actually, the coupled system in real world does not usually disintegrate into pieces after some nodes have become invalid. To explain this phenomenon, we model a coupled system as interdependent networks and study, both analytically and numerically, the percolation in interdependent networks with conditional dependency clusters. A node in our model survives until the number of failed nodes in its dependency cluster is greater than a threshold. Our exact solutions of giant component size are in good agreement with the simulation results. Though our model does not have second order phase transition, we still find ways to improve the robustness of interdependent networks. One way is to increase the dependency cluster failure threshold. A higher threshold means that more nodes in the dependency cluster can be removed without breaking down the node depending on the cluster. Other way is to increase the size of dependency clusters, the more the nodes in the dependency cluster, the more the failure combinations are, which increases the survival probability of the node depending on cluster. Our model offers a useful strategy to enhance the robustness of coupled system and makes a good contribution to the study of interdependent networks with dependency clusters. Keywords:interdependent networks/ cascading failures/ percolation/ phase transition
全文HTML
--> --> --> 1.引 言现实世界中的许多复杂系统都可以用复杂网络描述, 例如生物神经系统、社交网络、交通系统、互联网等[1-3]. 随着研究人员对复杂网络认识的逐渐深入, 相关研究成果已成为人们理解现实世界复杂系统的重要工具. 但现实生活中的复杂系统通常并不是孤立的, 往往多个系统之间存在相互依赖关系, 这种依赖性会导致网络鲁棒性的急剧下降. 2003年意大利大停电是典型的相依网络级联失效引起的重大生产事故, 最初电网和通信网中极少的节点失效使意大利接近半数的电力设施瘫痪, 数百万人失去电力供应[4]. 在单个网络中, 随着初始删除节点比例增大, 网络巨分量(giant component)连续减小, 整个解体过程呈现出二阶相变, 但在多个相依网络, 巨分量减小的过程多为一阶相变, 表明相依网络的鲁棒性远不如单个网络. 近些年来, 相依网络的鲁棒性开始引起研究人员的重视[3,5-13], 网络鲁棒性研究主要基于经典统计物理的逾渗模型[14], 该模型能准确描述网络解体的相变过程. 以往关于相依网络鲁棒性的研究主要针对一对一相依节点[12,15-19], 但在现实情况中, 一个节点往往会依赖于多个节点[6], 这多个被依赖节点称为该节点的依赖群. 例如一个通信基站会有多个电厂为其供电, 食物链中一种生物会捕食多种生物. 近年来, 已有****对存在依赖群的网络展开了研究. Shao等[6]提出了一种耦合网络中存在依赖群的模型, 该模型认为只有当某节点依赖群的全部节点失效后该节点才会失效, 这种依赖模式有效改善了相变点. Wang等[20]研究了网络的群渗流现象, 网络节点的存活取决于其所在超级节点是否连向巨分量, 这种网络模型在一定程度上提高了网络的鲁棒性, 但逾渗类型仍然是一阶相变. Parshani等[7]研究了单个网络中存在依赖群的情况, 发现随着依赖群规模增大网络会变得更加脆弱, 极少数节点失效会导致整个网络崩溃. Wang等[9,21]提出了一种单个网络中存在条件依赖群的模型, 认为当某节点依赖的节点失效数量超过一定比例时该节点才失效, 研究发现随着失效比例阈值增大, 一阶相变甚至会变为二阶相变. 本文研究了具有多依赖节点的相依网络的逾渗现象, 提出相依网络的条件依赖群逾渗模型并给出巨分量相变方程. 仿真结果表明, 在确定依赖度分布的情况下, 相依网络逾渗模拟结果与方程数值解相吻合, 并且通过增大失效容忍度$\gamma $或依赖群规模可增强网络鲁棒性. 虽然模型对于任意依赖度分布的相依网络不存在二阶相变, 但本文所述逾渗模型对提高相依网络鲁棒性仍具有一定指导作用. 本文的内容主要包括: 第2部分阐述相依网络的条件依赖群逾渗模型的基本概念; 第3部分通过理论分析给出模型的巨分量相变方程; 第4部分仿真验证理论框架有效性; 第5部分讨论本文研究成果; 第6部分是结论. 2.模 型传统相依网络逾渗模型多为一对一依赖, 一对一依赖可使网络满足无反馈条件, 便于理论分析[3], 但严格的一对一依赖在现实网络中通常是不存在的, 一个网络节点可能会依赖于另一个网络的多个节点, 若节点依赖的所有节点中有任意一个节点失效, 按照传统模型的分析方法, 该节点也会失效, 如图1所示. 图 1 具有多依赖关系的相依网络逾渗示意图, 初始攻击导致红色节点失效, 随后发生级联失效过程, 最终相依网络仅有极少数节点存活 Figure1. The model of percolation of interdependent networks with multiple support-dependence relations. The red node fails after the initial attack. Then the cascading failure process leads to a catastrophiccollapse of the interdependent networks. Finally, only a small fraction of nodes survive.
这种严格的多依赖关系会大幅降低相依网络的鲁棒性, 少部分节点失效就会造成相依网络完全崩溃. 但现实中的网络却没有这么脆弱. 通过观察可以发现, 只要依赖群有部分节点存活则依赖节点仍有效, 例如在某产业链中, 存在生产同类型产品的工厂若干, 其中少部分同类工厂倒闭并不会导致有关产业链的瘫痪[9]. 因此本文提出了一种允许部分相依节点失效的相依网络逾渗模型, 用以描述在部分被依赖节点失效情况下依赖节点仍正常工作的现象, 如图2所示. 图 2 相依网络的条件依赖群逾渗模型 A网络节点随机依赖于B网络的$\tilde k$个节点($\tilde k \geqslant 0$), 反之亦然; A网络的a节点依赖的3个节点有一个失效, 但失效比例未超过容忍度$\gamma = 0.5$, a节点继续工作; B网络的b节点依赖的节点失效比例超过$\gamma = 0.5$, 所以b节点失效 Figure2. The model of percolation in interdependent networks with conditional dependency clusters. Nodes in network A randomly depends on $\tilde k$ ($\tilde k \geqslant 0$) nodes in network B and vice versa. One of the three nodes which node a in network A depends on fails, but the failure proportion does not exceed $\gamma = 0.5$, node a still works. Two of the three nodes which node b in network B depends on fail, the failure proportion exceeds$\gamma = 0.5$, node bfails.
对于任意$\tilde P(\tilde k)$分布求解逾渗方程较为复杂, 本文只考虑依赖群规模固定的情况. 图3是不同$\gamma $取值的RR-RR相依网络逾渗的仿真结果. 图 3 不同$\gamma $取值的RR-RR相依网络逾渗, 每个网络节点数$N = 200000$, 平均度$\left\langle k \right\rangle = 6$, $\tilde P(10) = 1$, 空心标记为仿真值, 实线为逾渗方程数值解 (a) 巨分量${\mu _\infty }$与$p$对应关系; (b) 级联失效迭代次数(number of iterations, NOI) Figure3. The results of the percolation of RR-RR interdependent networks for different $\gamma $, each network has 200000 nodes, average degree is 6, $\tilde P(10) = 1$. The symbols represent the simulation results, and the solid linesshow the corresponding analytical predictions. (a) The size ofthe giant component ${\mu _\infty }$versus $p$; (b) number of iterations.
随着$p$值发生变化, $D(f)$会与横轴相切, 相切时$p$的取值${p_{\rm{c}}}$为相变点. 图4为同分布ER-ER相依网络相变点数值解, 图5为不同$\gamma $取值的ER-ER相依网络逾渗的仿真结果. 图 4 不同$\gamma $值对应的ER-ER相依网络相变点, 网络平均度$\left\langle k \right\rangle = 6$, $\tilde P(10) = 1$ Figure4. Graphical solutions of ER-ER interdependent networks transition for different $\gamma $. The average degree is 6, $\tilde P(10) = 1$.
图 5 不同$\gamma $取值的ER-ER相依网络逾渗, 每个网络节点数$N = 200000$, 平均度$\left\langle k \right\rangle = 6$, $\tilde P(10) = 1$, 空心标记为仿真值, 实线为逾渗方程数值解 (a) 巨分量${\mu _\infty }$与$p$对应关系; (b) 级联失效迭代次数 Figure5. The results of the percolation of ER-ER interdependent networks for different $\gamma $, each network has 200000 nodes, average degree is 6, $\tilde P(10) = 1$. The symbols represent the simulation results, and the solid linesshow the corresponding analytical predictions. (a) The size ofthe giant component ${\mu _\infty }$ versus $p$; (b) number of iterations.
图6、图7分别对比了ER-ER相依网络不同${\tilde P_i}(\tilde k)$分布的逾渗结果和不同$\gamma $取值的相变点. 图 6 不同依赖度分布的ER-ER相依网络逾渗, $\tilde P(10) = 1$, $\tilde P(5) = 1$分别用实心和空心标记表示, 每个网络节点数$N = 200000$, 平均度$\left\langle k \right\rangle = 6$ (a) 巨分量${\mu _\infty }$与$p$对应关系; (b) 级联失效迭代次数 Figure6. The results of the percolation of ER-ER interdependent networks for different dependency degrees, each network has 200000 nodes, average degree is 6. The dependency degrees are $\tilde P(10) = 1$ (solid symbols) and $\tilde P(5) = 1$ (empty symbols). (a) The size ofthe giant component ${\mu _\infty }$ versus $p$; (b) number of iterations.
图 7 不同$\gamma $取值的ER-ER相依网络相变点, 空心标记为仿真结果, 实线是理论值 Figure7. The critical point ${p_{\rm{c}}}$ versus $\gamma $ of ER-ER interdependent networks, The symbols represent the simulation results, and the solid linesshow the corresponding analytical predictions.
将两个生成函数代入(7)式可求出巨分量. 图8给出了不同$\gamma $取值的SF-SF相依网络逾渗仿真结果. 图 8 不同$\gamma $取值的SF-SF相依网络逾渗, 每个网络节点数$N = 200000$, 平均度$\left\langle k \right\rangle = 6$, $\lambda = 2.7$, $\tilde P(10) = 1$, 空心标记为仿真值, 实线为逾渗方程数值解 (a) 巨分量${\mu _\infty }$与$p$对应关系; (b) 级联失效迭代次数 Figure8. The results of the percolation of SF-SF interdependent networks for different $\gamma $, each network has 200000 nodes, average degree is 6, $\lambda = 2.7$, $\tilde P(10) = 1$. The symbols represent the simulation results, and the solid linesshow the corresponding analytical predictions. (a) The size ofthe giant component ${\mu _\infty }$ versus $p$; (b) number of iterations.