Gwak Sang-Hwan, Goh K-I
Department of Physics, Korea University, Seoul, 02841 Korea.
J Korean Phys Soc. 2022;81(7):680-687. doi: 10.1007/s40042-022-00549-0. Epub 2022 Jul 25.
Network robustness has been a pivotal issue in the study of system failure in network science since its inception. To shed light on this subject, we introduce and study a new percolation process based on a new cluster called an 'exclave' cluster. The entities comprising exclave clusters in a network are the sets of connected unfailed nodes that are completely surrounded by the failed (i.e., nonfunctional) nodes. The exclave clusters are thus detached from other unfailed parts of the network, thereby becoming effectively nonfunctional. This process defines a new class of clusters of nonfunctional nodes. We call it the no-exclave percolation cluster (NExP cluster), formed by the connected union of failed clusters and the exclave clusters they enclose. Here we showcase the effect of NExP cluster, suggesting a wide and disruptive collapse in two empirical infrastructure networks. We also study on two-dimensional Euclidean lattice to analyze the phase transition behavior using finite-size scaling. The NExP model considering the collective failure clusters uncovers new aspects of network collapse as a percolation process, such as quantitative change of transition point and qualitative change of transition type. Our study discloses hidden indirect damage added to the damage directly from attacks, and thus suggests a new useful way for finding nonfunctioning areas in complex systems under external perturbations as well as internal partial closures.
自网络科学诞生以来，网络鲁棒性一直是网络科学中系统故障研究的关键问题。为了阐明这一主题，我们引入并研究了一种基于名为“飞地”簇的新簇的新型渗流过程。网络中构成飞地簇的实体是被故障（即无功能）节点完全包围的相连未故障节点集。因此，飞地簇与网络的其他未故障部分分离，从而实际上变得无功能。这个过程定义了一类新的无功能节点簇。我们将其称为无飞地渗流簇（NExP簇），它由故障簇及其包围的飞地簇的连通并集形成。在此，我们展示了NExP簇的影响，表明在两个实证基础设施网络中出现了广泛且具有破坏性的崩溃。我们还在二维欧几里得晶格上进行研究，以使用有限尺寸标度分析相变行为。考虑集体故障簇的NExP模型揭示了网络崩溃作为一种渗流过程的新方面，例如转变点的定量变化和转变类型的定性变化。我们的研究揭示了除直接来自攻击的损害之外新增的隐藏间接损害，因此为在外部扰动以及内部局部关闭情况下寻找复杂系统中的无功能区域提供了一种新的有用方法。
