Abstract
This study applies the shallow water equations to simulate dam-break events and predict both bed topography and Manning's roughness coefficient. A well-balanced finite volume method is used to generate a dataset that captures the complex flow dynamics of dam failures. This dataset is then utilized to train a neural network capable of predicting terrain changes and roughness distribution after the event. The performance of the model is assessed using Root Mean Square Error (RMSE) to ensure accurate predictions. The findings demonstrate that integrating computational hydraulics with neural networks can enhance the prediction of post-event topography and hydraulic roughness. This approach provides valuable insights for flood modeling, hydraulic engineering, and infrastructure planning, contributing to more effective risk assessment and decisionmaking in real-world scenarios.