Rapid seismic resilience assessment of bridge networks using Stacking-GAN integration

Abstract

Comprehensive evaluation of the seismic resilience of bridge networks is crucial for building resilient cities. This paper proposes a rapid assessment method for seismic resilience of bridge networks using Stacking-GAN integration. The method considers both the structural and traffic characteristics of the bridge network and introduces a comprehensive resilience index that integrates structural and functional resilience to comprehensively assess the seismic resilience of the bridge network. To address the need for rapid assessment, a Stacking model-based seismic resilience evaluation model for bridge networks is developed. The integration of Stacking and GAN models addresses the issue of insufficient training data while simultaneously improving the model’s prediction accuracy and generalization capability. Additionally, using a parametric modeling approach for bridges and based on 14 urban road networks, a sample library containing 5000 bridge networks was established using a GAN model. Finally, using a bridge network from a region in Anaheim as an example, the structural and functional resilience of the network was assessed. Results show that under various PGA levels, the maximum relative error between calculated and predicted comprehensive resilience is 8.74 %, with a maximum absolute error of 0.025, validating the model’s effectiveness.