An adaptive, data-driven multiscale approach for dense granular flows
B. Siddani, Weiqun Zhang, Andrew Nonaka, John Bell, Ishan Srivastava · 2025
Read An adaptive, data-driven multiscale approach for dense granular flows by B. Siddani, Weiqun Zhang, Andrew Nonaka, John Bell, Ishan Srivastava (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-13458.txt.
Introduction
Bandar Siddani and co-authors present a multiscale method for simulating dense flows of grains, in work from 2025. Coarse models of sand-like flows are fast but rely on closure rules that are often crude. A linked fine model can supply those rules, yet running it everywhere is far too costly.
Their method trains neural networks on fine simulations to stand in for the closure rules. A group of networks also estimates its own uncertainty at each point. Where the estimate looks unreliable, the system runs fresh fine simulations and updates the networks. The networks are shaped to keep the coarse model well behaved. Tests on sloped flows and a collapsing column of grains in three dimensions show the method stays accurate while cutting cost sharply.
Catalog record
| CREATOR | B. Siddani, Weiqun Zhang, Andrew Nonaka, John Bell, Ishan Srivastava |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 69 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-13458
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
An adaptive, data-driven multiscale approach for dense granular flows B. Siddani, Weiqun Zhang, Andrew Nonaka, John Bell, Ishan Srivastava Abstract The accuracy of coarse-grained continuum models of dense granular flows is limited by the lack of high-fidelity closure models for granular rheology. One approach to addressing this issue, referred to as the hierarchical multiscale method, is to use a high-fidelity fine-grained model to compute the closure terms needed by the coarse-grained model. The difficulty with this approach is that the overall model can become computationally intractable due to the high computational cost of the high-fidelity model. In this work, we describe a multiscale modeling approach for dense granular flows that utilizes neural networks trained using high-fidelity discrete element method (DEM) simulations to approximate the constitutive granular rheology for a continuum incompressible flow model. Our approach leverages an ensemble of neural networks to estimate predictive uncertainty that allows us to determine whether the rheology at a given point is accurately represented by the neural network model. Additional DEM simulations are only performed when needed, minimizing the number of additional DEM simulations required when updating the rheology. This adaptive coupling significantly reduces the overall computational cost of the approach while controlling the error. In addition, the neural networks are customized to learn regularized rheological behavior to ensure well-posedness of the continuum solution. We first validate the approach using two-dimensional steady-state and decelerating inclined flows. We then demonstrate the efficiency of our approach by modeling three-dimensional sub-aerial granular column collapse for varying initial column aspect ratios, where our multiscale method compares well with the computationally expensive computational fluid dynamics (CFD)-DEM simulation. 1 Introduction Granular materials have significant applications in industry and nature; yet, a general theory for their dynamics is still an outstanding scientific challenge. Particularly, the flow of granular materials is highly non-Newtonian and can exhibit complex non-equilibrium phenomena such as flow-arrest [ 1 , 2 ] and yield transition [ 3 ] , shear banding [ 4 ] , creep [ 5 , 6 ] , nonlocal dynamics in confined geometries [ 7 ] , normal stress effects [ 8 , 9 ] , and pressure-dependent rheology [ 10 ] . Granular flows can be classified into three distinct regimes, predominantly for convenience: (1) rate-independent quasi-static flows that are modeled using principles of solid mechanics and plasticity [ 11 ] ; (2) rate-dependent dense inertial flows that are modeled using the principles of fluid mechanics [ 12 ] ; and (3) gas-like collisional flows that are well-analyzed using kinetic theories [ 13 ] . However, the boundary between these regimes is not distinct and there are ongoing efforts to extend theories across regimes [ 14 , 15 ] . In this work, we focus our attention on quasi-static and dense granular flows where the granular contact lifetimes are relatively long, inertia is important and the material is predominantly dense. Two classes of numerical methods are particularly well-studied to model granular flows in this regime: (1) the discrete element method (DEM) [ 16 ] , which models individual particles of the material and tracks their trajectory in space and time; and (2) continuum methods such as the mesh-based finite-volume method (FVM) [ 17 , 18 , 19 ] and finite element method (FEM) [ 20 ] , and meshless methods such as the material point method (MPM) [ 21 , 22 , 23 ] and smooth particle hydrodynamics (SPH) [ 24 , 25 ] . DEM is particularly useful for capturing particle-scale physics in granular flows and, in principle, provides the highest fidelity representation of granular dynamics. Furthermore, it can be coupled to computational fluid dynamics (CFD) models such as the finite-volume method (FVM) [ 19 , 26 ] and the lattice Boltzmann method (LBM) [ 27 ] to model the evolution of the solid granular phase in multiphase flows. However, DEM is computationally expensive, particularly for practical applications; for example, just a cubic meter of sand can contain more than 10^{9} particles. On the other hand, continuum methods are easily scalable to practical length and time scales, but require constitutive knowledge to close the set of governing equations such as a rheological relationship between the stress and the strain rate. Significant recent advances have proposed a viscoplastic, scalar \mu(I) inertial rheology for dense granular flows [ 10 , 12 ] , where \mu is a dimensionless stress ratio and I is a dimensionless inertial number. The scalar \mu monotonically increases with I and asymptotes to a constant non-zero value as I\to 0 [ 28 , 8 , 1 ] . The \mu(I) rheology has shown success in predicting velocity profiles in several idealized simple shear configurations [ 29 , 12 ] , and extensions to it have incorporated additional physics such as anisotropic normal stress effects [ 8 ] , Lode-angle dependence [ 30 , 31 ] , nonlocal dynamics [ 7 ] , and transient dilatancy [ 32 , 33 ] . However, the \mu(I) rheology for dense granular flows depends intimately on a multitude of particle-scale properties such as particle shape [ 34 ] , size distribution [ 35 , 36 ] , surface roughness [ 8 , 30 ] , elasticity [ 37 ] and contact restitution [ 38 ] , and a general rheological model that accounts for all these properties is currently unavailable. DEM simulations of a representative volume element (RVE) of a granular material (usually significantly smaller than the scale of the application) can be used to calibrate the rheology for a given set of particle-scale parameters [ 8 , 30 ] . However, even on a reduced RVE scale, these simulations are too expensive to compute the entire rheology, particularly considering the vast design space of particle-scale properties. Furthermore, even in cases where the underlying rheology is sufficiently parameterized, continuum solutions of dense granular flows are complicated because of the rigid plastic response of the material as I\to 0 , which leads to ill-posedness of the standard \mu(I) rheology in continuum solvers [ 39 ] . The presence of a yield stress as I\to 0 requires careful treatment such as using appropriate regularization [ 17 ] , optimization methods [ 40 ] , or treating the material as an elasto-plastic solid [ 41 ] . The ill-posedness of the local \mu(I) rheology was reported in Barker et al. [ 39 ] and confirmed by the appearance of instabilities in continuum numerical solutions with the \mu(I) rheology [ 42 ] . A regularized model of the rheology was proposed that increased the range of inertial numbers where the rheology is well-posed [ 42 ] . The aforementioned evidence suggests that high-fidelity modeling of granular flows necessarily requires multiscale coupling between particle-scale DEM and continuum solvers with appropriately built-in regularization for numerical stability. Hierarchical multiscale methods that couple DEM with a continuum solver have shown promise in accurate and efficient modeling of granular materials [ 43 , 44 , 45 , 23 ] . However, despite being more efficient than a full DEM-based simulation, they are still computationally expensive since they require the evaluation of the constitutive response using DEM at every time step of the continuum solution. There is a growing interest in applying machine