NeuralDEM presents an end-to-end approach to replace Discrete Element Method (DEM) routines and coupled multiphysics simulations with deep learning surrogates.
Benedikt Alkin †, *, 1, 2, Tobias Kronlachner †, *, 1, 3, Samuele Papa †, 1, 4, 5,
Stefan Pirker 3, Thomas Lichtenegger 1, 3, Johannes Brandstetter @, 1, 2
† core contributor, * equal contribution
1 NXAI GmbH, Linz, Austria
2 ELLIS Unit Linz, Institute for Machine Learning, JKU Linz, Austria
3 Department of Particulate Flow Modelling, JKU Linz, Austria
4 University of Amsterdam, Amsterdam, Netherlands
5 The Netherlands Cancer Institute, Amsterdam, Netherlands
@ Correspondence to: johannes.brandstetter@nx-ai.com
Advancements in computing power have made it possible to numerically simulate large-scale fluid-mechanical and/or particulate systems, many of which are integral to core industrial processes. Among the different numerical methods available, the discrete element method (DEM) provides one of the most accurate representations of a wide range of physical systems involving granular and discontinuous materials. Consequently, DEM has become a widely accepted approach for tackling engineering problems connected to granular flows and powder mechanics. Additionally, DEM can be integrated with grid-based computational fluid dynamics (CFD) methods, enabling the simulation of chemical processes taking place, e.g., in fluidized beds. However, DEM is computationally intensive because of the intrinsic multiscale nature of particulate systems, restricting either the duration of simulations or the number of particles that can be simulated. Moreover, the non-trivial relationship between microscopic DEM and macroscopic material parameters necessitates extensive calibration procedures. Towards this end, NeuralDEM presents a first end-to-end approach to replace slow and computationally demanding numerical DEM routines with fast, adaptable deep learning surrogates. NeuralDEM is capable of picturing long-term transport processes across different regimes using macroscopic observables without any reference to microscopic model parameters. First, NeuralDEM treats the Lagrangian discretization of DEM as an underlying continuous field, while simultaneously modeling macroscopic behavior directly as additional auxiliary fields. Second, NeuralDEM introduces multi-branch neural operators scalable to real-time modeling of industrially-sized scenarios — from slow and pseudo-steady to fast and transient. Such scenarios have previously posed insurmountable challenges for deep learning models. Notably, our largest NeuralDEM model is able to faithfully model coupled CFD-DEM fluidized bed reactors of 160k CFD cells and 500k DEM particles for trajectories of 28s, which amounts to 2800 machine learning timesteps. NeuralDEM will open many new doors to advanced engineering and much faster process cycles.
NeuralDEM presents the first end-to-end solution for replacing computationally intensive numerical DEM routines and coupled DEM-CFD simulations with fast and flexible deep learning surrogates. NeuralDEM introduces two conceptually novel modeling paradigms:
Hoppers are industrially used for short as well as long term storage of particulate material, showcasing slow and pseudo-steady macroscopic behavior. DEM is the preferred method since the air around the particles can usually be neglected due to the slow velocities in the system. In our experiments, the hopper geometry is filled with 250k particles, which gradually exit the domain over the simulation duration when the hopper empties. Based on the material properties and the geometry properties, different flow regimes are obtained in a hopper.
Most macroscopic quantities of interest to engineering applications emerge from the microscopic particle-particle interactions modeled by DEM. To re-generate these emerging phenomena with NeuralDEM, we train our model to predict auxiliary fields at each timestep of the simulation.
Fluidized bed reactors are characterized by fast and transient phenomena and are widely used in industry for a variety of processes. Fluidized bed reactors showcase strong interactions of the particles with the surrounding fluid, necessitating an accurate modeling of particles, the gas phase, as well as particle-gas interactions. Thus, modeling approaches need to combine DEM parts with simulations of the surrounding fluid. Simulation are performed with a coupled CFD-DEM method, using 500k particles and the fluid, i.e., air, that is uniformly pushed into the reactor from the bottom is modeled on a grid of 160k hexahedral cells.
Particles
Solid fraction
Fluid velocity
Iso-surface
Solid fraction
Fluid velocity
Slow inlet velocity
Fast inlet velocity
The fast and chaotic nature of fluidized bed also means that numerical differences will result in sharp changes in the simulation. Both different initial conditions and slight differences in the numberical choices made during the simulation will lead, in the long term, to different rollouts. NeuralDEM captures these physics very well. Would you be able to distinguish between the CFD-DEM simulation and the NeuralDEM simulation?
@article{alkin2024neuraldem,
title={{NeuralDEM} - Real-time Simulation of Industrial Particulate Flows},
author={Benedikt Alkin and Tobias Kronlachner and Samuele Papa and Stefan Pirker and Thomas Lichtenegger and Johannes Brandstetter},
journal={arXiv preprint arXiv:2411.09678},
year={2024}
}