Surrogate-based decision-making for post-earthquake recovery scheduling and resilience assessment of subway systems considering the effect of infrastructure interdependency

This study proposes a method for post-earthquake recovery and seismic resilience assessment of subway systems considering the effect of infrastructure interdependency. A two-stage recovery decision-making model is established. In the pre-earthquake stage, reserved restoration resources are determined by balancing recovery speed and recovery efficiency. In the post-earthquake stage, the optimal restoration schedules are obtained by solving the established functional recovery optimization model using the NSGA-II. A deep learning-based surrogate model for functionality computation is developed to significantly reduce the computational costs in the recovery decision-making and optimization process. A seismic resilience assessment and recovery decisionmaking process is conducted for a real, large-scale subway system. The results indicate that the subway system exhibits high recovery efficiency in its early-stage recovery and a high level of seismic resilience. The study highlights the significant impact of the municipal power supply system (MPSS) as a critical infrastructure system. The recovery time affected by the MPSS accounts for more than 33% of the total recovery time of the subway system. Ignoring the impact of the MPSS led to an average 12.13% underestimation of the center of resilience. These results underscore the substantial influence of the MPSS on the recovery and seismic resilience of subway systems.