Machine Learning-Based Estimation of Seepage and Health Assessment of Rockfill Dams

Puntos clave

• The aim is to accurately estimate seepage behaviors in rockfill dams using observational data and machine learning techniques.
• Developed machine learning models using observational data.
• Applied Random Forest, Gradient Boosting Decision Tree, and Neural Network techniques.
• Optimized hyperparameters for each model.
• Neural Network achieved the highest precision for seepage estimation.
• Estimated error range was confirmed to be ±171L/min in healthy conditions.
• Proposed a health assessment metric based on the estimation errors.