Optimizing Seismic Retrofit of Bridges: Integrating Efficient Graph Neural Network Surrogates and Transportation Equity

Natural hazards can substantially damage bridge networks. Therefore to protect these lifelines, a bridge maintenance and retrofit strategy is imperative. To prepare for earthquakes in particular, conventional methods for seismic risk analysis needs to calculate network response which is a computationally intensive process. This prevents the optimal bridge maintenance for seismic events from being widely done for large networks. Also, it is critical to integrate measures of equity in these optimal asset management solutions to ensure equitable post-disaster access to mobility across all income levels. In this study, we propose a graph neural network (GNN) surrogate model for rapid estimation of transportation performance and a bridge maintenance optimization framework that considers network connectivity and equity measures. The proposed multi-objective optimization, via genetic algorithm with graph neural network surrogates, identifies the optimal bridge retrofit strategies constrained by a retrofit budget. The efficacy and accuracy of the proposed algorithm are demonstrated using a road network in the California Bay Area as a case study. Also, it will be shown that integrating income disparities in the region into the optimization framework can lead to improved transportation equity measures.