Deep Neural Network and Polynomial Chaos Expansion-Based Surrogate Models for Sensitivity and Uncertainty Propagation: An Application to a Rockfill Dam

Computational modeling plays a significant role in the design of rockfill dams. Various constitutive soil parameters are used to such models, which often involve high uncertainties due complex structure dams comprising various zones different parameters. This study performs an uncertainty analysis and global sensitivity assess effect on behavior dam. A Finite Element code (Plaxis) is utilized for analysis. database computed displacements at inclinometers installed dam generated compared situ measurements. Surrogate models tools approximating relationship between input thereby reducing computational costs parametric studies. Polynomial chaos expansion deep neural networks build surrogate compute Sobol indices required identify impact behavior.