Department of Civil and Environmental Engineering, University of Washington, Seattle, WA 98195.
Department of Civil and Environmental Engineering, University of Washington, Seattle, WA 98195
Proc Natl Acad Sci U S A. 2018 Aug 28;115(35):E8125-E8134. doi: 10.1073/pnas.1722313115. Epub 2018 Aug 15.
A generic modeling framework to infer the failure-spreading process based on failure times of individual nodes is proposed and tested in four simulation studies: one for cascading failures in interdependent power and transportation networks, one for influenza epidemics, one benchmark test case for congestion cascade in a transportation network, and one benchmark test case for cascading power outages. Four general failure-spreading mechanisms-external, temporal, spatial, and functional-are quantified to capture what drives the spreading of failures. With the failure time of each node given, the proposed methodology demonstrates remarkable capability of inferring the underlying general failure-spreading mechanisms and accurately reconstructing the failure-spreading process in all four simulation studies. The analysis of the two benchmark test cases also reveals the robustness of the proposed methodology: It is shown that a failure-spreading process embedded by specific failure-spreading mechanisms such as flow redistribution can be captured with low uncertainty by our model. The proposed methodology thereby presents a promising channel for providing a generally applicable framework for modeling, understanding, and controlling failure spreading in a variety of systems.
提出并测试了一种基于个体节点失效时间推断失效传播过程的通用建模框架，该框架在四个模拟研究中得到了验证：一个是关于相互依存的电力和交通网络中的级联失效，一个是关于流感疫情，一个是交通网络中拥塞级联的基准测试案例，另一个是关于级联停电的基准测试案例。量化了四种通用的失效传播机制——外部、时间、空间和功能，以捕捉导致失效传播的因素。对于每个节点的失效时间，所提出的方法展示了推断潜在的一般失效传播机制的显著能力，并在所有四个模拟研究中准确地重建了失效传播过程。对两个基准测试案例的分析也揭示了所提出的方法的稳健性：表明我们的模型可以以较低的不确定性捕捉由特定失效传播机制（例如流量重新分配）嵌入的失效传播过程。因此，该方法为在各种系统中建模、理解和控制失效传播提供了一种通用适用的框架。
