Enhanced Metro Resilience Assessment Combining Dynamic Cascade-Based Simulation and Deep Learning-Based Attributed Graph Clustering

Resilience assessment is crucial for urban metro systems, enabling risk identification and guiding recovery strategies during extreme events. This study develops an Enhanced Metro Resilience Assessment (EMRA) method integrating dynamic cascade-based simulation and deep learning-based attributed graph clustering. Specifically, EMRA incorporates four human-centered resilience indicators, including network efficiency, economic benefit, passenger load capacity, and point of interest density, into the metro network’s topology. A passenger flow-based load capacity model is employed to simulate the impacts of disasters, incorporating dynamic cascade effects. A deep learning-based graph clustering algorithm is then used to assess the heterogeneous recovery capabilities of metro stations, distinguishing recovery times to optimize the overall resilience evaluation. Validated on the Shanghai metro system, EMRA provides insights into network vulnerabilities, revealing that stations with cut-vertex characteristics, like Shanghai South Railway Station, significantly disrupt performance (9.72% decline). The study also highlights the benefits of increasing transfer stations for improving network stability and resilience. The findings underscore the value of cascade-based disaster simulations and clustering-based recovery modeling in conducting robust resilience assessments, offering valuable insights for improving the resilience of urban metro systems.