Physics-based vs data-driven constitutive modeling of granular media down an inclined plane

¹ Institute of Data Science and Artificial Intelligence (DATAI), University of Navarra, Pamplona, Spain
² Department of Physics and Applied Mathematics, University of Navarra, Pamplona, Spain

^* e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.
^** e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.

Published online: 1 December 2025

Abstract

We present a numerical study of 3D granular flow down an inclined plane, using Discrete Element Method (DEM). The data of individual grains are used to compute the macroscopic density, velocity, and stress fields through a coarse-graining technique (CG). We begin by analyzing granular flows with the analytical rheology model μ(I), which has proven to be effective in describing dense, quasistatic, and inertial flow regimes. We also present a data-driven approach that utilizes machine learning methods to build constitutive models. This approach does not rely on predetermined balance equations; instead, the resulting constitutive model is trained directly on DEM-CG data to learn patterns and relationships. In general, our results suggest the potential of ML approaches in predicting stress distributions in dense granular flows. As expected, random forest and neural network analysis are more effective compared to simple linear regression. In particular, neural networks appear as a promising avenue for advancing predictive accuracy in future studies.