✨ TL;DR
This paper develops a graph neural network approach to classify bridges based on their disaster-resilience roles by analyzing metapaths connecting highways, bridges, and critical buildings. The method helps prioritize bridge maintenance budgets by identifying which bridges are essential for supply chains, medical access, or residential protection during disasters.
Urban infrastructure managers face the challenge of prioritizing bridge maintenance under limited budgets while preparing for disasters. Bridges play multiple critical roles in maintaining urban functions during disasters by providing access to hospitals, commercial facilities, and residences. However, existing approaches rely on single indicators that fail to capture the multi-dimensional importance of bridges in disaster scenarios. There is no systematic way to quantify how different bridges contribute to different aspects of disaster resilience, such as emergency healthcare access, commercial logistics, or preventing residential isolation.
The authors construct a heterogeneous graph with three layers: roads (national highways), bridges, and buildings (hospitals, shops, residences). They define metapaths that connect highways through bridges to specific building types, representing different disaster-resilience functions. A Relation-centric Graph Convolutional Network Variational Autoencoder (R-GCN-VGAE) learns feature representations based on these metapaths. The learned representations enable classification of bridges into three disaster-preparedness categories: Supply Chain (commercial logistics), Medical Access (emergency healthcare), and Residential Protection (preventing isolation). The methodology uses open data sources including OSMnx and applies k-NN tuning strategies. UMAP is used for visualization of the multi-role bridge classifications.