Computer-Aided Civil and Infrastructure Engineering, Год журнала: 2025, Номер unknown

Опубликована: Апрель 14, 2025

Abstract Climate change has intensified extreme weather events, with floods causing significant socioeconomic and environmental damage. Accurate flood forecasting is crucial for disaster preparedness risk mitigation, yet traditional hydrodynamic models, while precise, are computationally prohibitive real‐time applications. Machine learning surrogates, such as graph neural networks (GNNs), improve efficiency but often lack physical consistency interpretability. This paper introduces HydroGraphNet, a novel physics‐informed GNN framework that, the first time, integrates Kolmogorov–Arnold Network (KAN) to enhance model interpretability in unstructured mesh‐based forecasting. The embeds mass conservation laws into loss function, ensuring physically consistent predictions. Additionally, it employs an autoregressive encoder–processor–decoder architecture that captures spatiotemporal dynamics mitigating error accumulation over long horizons. Validation on data from White River near Muncie, Indiana, demonstrates 67% reduction prediction error, near‐zero balance 58% improvement critical success index major events compared baseline model. These results highlight potential of proposed advance improved

Язык: Английский