Rapid post-earthquake evacuation time prediction in urban bridge networks using a machine-learning-based surrogate model

As the frequency and intensity of earthquakes continue to increase, accurately assessing post-earthquake evacuation performance in urban transportation systems has become a critical challenge. Seismic damage to bridges, which serve as critical links within road networks, can significantly degrade overall network functionality and severely impede vehicular evacuation processes. However, conventional traffic simulation-based approaches require substantial computational resources and time, limiting their applicability for large-scale scenario exploration and real-time decision-making. To address these limitations, this study proposes a machine-learning-based surrogate modeling framework to efficiently predict post-earthquake evacuation performance. The Evacuation Time Estimate (ETE) is adopted as a quantitative indicator of network performance. A large-scale dataset is generated through simulation-based traffic analyses conducted on the real-world road network of Pohang, South Korea, which includes 48 bridges. Bridge damage scenarios are systematically defined based on variations in epicentral locations within road-traffic-dense areas of Pohang City and earthquake magnitudes. In total, 100,000 simulation samples are constructed to train and validate the surrogate model. The trained model is then applied to predict ETE for various earthquake scenarios, achieving a substantial reduction in computational time compared to conventional simulations while maintaining high predictive accuracy. Furthermore, numerical analysis results indicate that evacuation time variability increases with earthquake magnitude, particularly when epicenters are located closer to the road network. These findings demonstrate that the proposed surrogate model effectively captures seismic impacts on transportation systems and provides a practical and efficient tool for rapid post-earthquake evacuation assessment and informed decision-making.