⚡
Industrial Engineering Just Got a Turbo Boost. Emmi AI is Live!

NeuralDEM

NeuralDEM introduces the first end-to-end deep learning surrogate for the discrete element method (DEM) used in modeling particulate systems. By replacing computationally intensive DEM routines with fast, multi-branch neural operators, NeuralDEM enables real-time simulation of industrial processes.

This breakthrough captures long-term transport phenomena and models complex multiphysics interactions, such as those in fluidized bed reactors, without extensive microscopic calibration.
Real-time Deep Learning Surrogates
NeuralDEM replaces traditional, slow DEM routines with fast deep learning models. This breakthrough enables immediate simulation feedback, transforming how industrial particulate flows are modeled.
Scalable Multi-branch Neural Operators
The architecture leverages multi-branch neural operators to manage industrial-scale simulations efficiently. It handles a variety of flow regimes—from slow pseudo-steady processes to rapid transient dynamics—ensuring robust performance across scenarios.
Seamless Coupled CFD-DEM Integration
NeuralDEM integrates CFD and DEM simulation components within one unified framework. This integration accurately captures complex multiphysics interactions, including those in fluidized bed reactors, with real-time efficiency.
Direct Macroscopic Process Modeling
The model directly predicts macroscopic observables without relying on detailed microscopic calibration. This approach streamlines the simulation process while preserving essential transport characteristics.

Paper

ARXIV

Authors

Benedikt Alkin
Tobias Kronlachner
Samuele Papa
Stefan Pirker
Thomas Lichtenegger
Johannes Brandstetter

contact
© 2025 Emmi AI GmbH — Made with ❤️ because smashing particles without AI is so last century