Predicting Dam-Break Discharges in Anura3D using Artificial Neural Networks

Abstract Submission: Dam-break events pose significant risks to downstream communities and ecosystems, necessitating accurate prediction of discharge hydrographs to improve flood risk management and mitigation strategies. Traditional hydraulic models, often require extensive computational resources and detailed input data. This study proposes the use of artificial neural networks (ANNs) to predict breach hydrographs during idealized dam-break events. Two-dimensional dam-break scenarios were implemented by Anura3D, open-source geotechnical software based on material point method. Different length and heights in reservoir were used to simulate dam-break scenarios using single-point method of Anura3D. Discharge hydrographs were obtained at three observation points on the downstream floodplain of the numerical domain. Some of the cases were compared with the laboratory experiments that were carried out in a dam-break flow testing facility at the USDA-ARS National Sedimentation Laboratory for validation. Numerical measurements were used to train an ANN model. The data was divided into training (50%), testing (25%), and validation (25%) sets, iterating over 20,000 timesteps. The optimal model was identified through a trial-and-error method to determine the best configuration of hidden nodes. The evaluation metrics—Mean Absolute Error (MAE), R-squared, and Average Square Error (ASE)—demonstrated predictive accuracy. Excel GUI is developed for demonstration and making it more friendly for users. The presentation will also include potential improvements and broader applications of this modeling approach.

Learning Objectives/Expected Outcome (Optional) : Attending international conferences offers a wealth of benefits that can significantly enhance both academic and professional development. These conferences serve as platforms for exchanging knowledge, fostering collaboration, and advancing one’s career, all while providing exposure to diverse perspectives and cultures.