Functional and structural resilience assessment in urban drainage networks: a physics-guided graph-based surrogate model

The first step in real-time management, optimal design, and retrofitting of urban drainage networks (UDNs) involves developing a model capable of rapidly and accurately assessing resilience under various failure scenarios. This study introduces a novel physics-guided approach based on graph theory as a surrogate model for resilience assessment, offering a computationally highly efficient alternative to traditional hydrodynamic models such as the Storm Water Management Model (SWMM). Given the study's focus on resilience evaluation, only the maximum flow of subcatchments was used as input for the physics-guided model, which employs a nonlinear method derived from SWMM outputs to ensure a fair comparison with SWMM's resilience computations. The graph-based resilience assessment incorporates the weighted shortest path method and introduces modified hydraulically informed graph metrics for flow routing in UDN pipes. This approach enables resilience assessment under structural (i.e., single-pipe failure) and functional failures (i.e., high-intensity rainfalls) significantly faster than the SWMM model. The methodology was validated using two real-world case studies under different rainfall scenarios, demonstrating that the graph-based resilience assessment achieved high consistency with the resilience calculations based on the SWMM hydrodynamic model.