Urban Mobility Institute, Tongji University, Shanghai 200092, China.
School of Transportation Engineering, Tongji University, Shanghai 200092, China.
Int J Environ Res Public Health. 2021 Dec 25;19(1):204. doi: 10.3390/ijerph19010204.
Analysis of the robustness and vulnerability of metro networks has great implications for public transport planning and emergency management, particularly considering passengers' dynamic behaviors. This paper presents an improved coupled map lattices (CMLs) model based on graph attention networks (GAT) to study the cascading failure process of metro networks. The proposed model is applied to the Shanghai metro network using the automated fare collection (AFC) data, and the passengers' dynamic behaviors are simulated by GAT. The quantitative cascading failure analysis shows that Shanghai metro network is robust to random attacks, but fragile to intentional attacks. Moreover, there is an approximately normal distribution between instant cascading failure speed and time step and the perturbation in a station which leads to steady state is approximately a constant. The result shows that a station surrounded by other densely distributed stations can trigger cascading failure faster and the cascading failure triggered by low-level accidents will spread in a short time and disappear quickly. This study provides an effective reference for dynamic safety evaluation and emergency management in metro networks.
地铁网络的健壮性和脆弱性分析对公共交通规划和应急管理具有重要意义，特别是考虑到乘客的动态行为。本文提出了一种基于图注意力网络（GAT）的改进耦合映射格子（CML）模型，用于研究地铁网络的级联失效过程。该模型应用于使用自动售检票（AFC）数据的上海地铁网络，并通过 GAT 模拟乘客的动态行为。定量级联失效分析表明，上海地铁网络对随机攻击具有鲁棒性，但对故意攻击很脆弱。此外，瞬时级联失效速度与时间步长之间存在近似正态分布，而在一个车站的扰动导致稳定状态的扰动大约是一个常数。结果表明，被其他密集分布的车站包围的车站可以更快地引发级联失效，而由低级事故引发的级联失效将在短时间内传播并迅速消失。这项研究为地铁网络的动态安全评估和应急管理提供了有效的参考。
