Artificial Neural Network Model for Predicting the Design of Cyclopean Concrete Dams in Water Harvesting and Planting Projects

Lizarbe-Alarcón H.; Palomino-Pariona D.; Estrada-Cárdenas J.; Tacas-Evanan J.; Ircañaupa-Huamani A.; Tacuri-Gamboa A.; Ayala-Bizarro R.; Bravo-Anaya R. · Mathematical Modelling of Engineering Problems (2025) · DOI: 10.18280/mmep.121133

Resumen

This study proposes a model based on artificial neural networks (ANNs) for predicting preliminary design parameters of cyclopean concrete dams in high-Andean microwatersheds, used in water harvesting and storage projects. A database was developed using digital elevation models (DEMs) and technical records from projects executed by the Sierra Azul Fund Executing Unit (UEFSA) between 2019 and 2024. Geospatial processing of the DEMs yielded morphometric variables and storage volumes, which were used as model inputs, while the dam axis location, length, and height were the outputs. The selected architecture was a multilayer perceptron (MLP) with two hidden layers, trained via backpropagation with the Adam optimizer. Results show R² values greater than 0.90 and acceptable average errors for preliminary design use. It is concluded that the model reduces variability inherent in empirical methods and enhances standardization in formulating water harvesting projects in high-Andean regions. This approach improves the efficiency, accuracy, and sustainability of hydraulic infrastructure development in rural contexts with high water vulnerability. Copyright:

Instituciones: Universidad Nacional de San Cristóbal de Huamanga