EmmiAI/NeuralDEM

Model Details

NeuralDEM models for simulating particle systems and particle-fluid coupled systems.

NeuralDEM trains two models separately for these two systems where datasets are generated for the respective use-case.

• Particle simulations: A hopper with an outlet at the bottom is initially full of particles where particles exit the hopper through the outlet.
• Particle-fluid multi-physics simulations: A fluidized bed reactor where air comes in from the bottom and pushes around the particles within the reactor.

How to use

We provide a standalone GitHub repository with relevant tutorial code on how execute inference. The corresponding repository can be found here: https://github.com/Emmi-AI/NeuralDEM

Dataset

The datasets for NeuralDEM are available on Hugging Face Datasets at EmmiAI/NeuralDEM. Additionally, this repository contains a single simulation sample from the test set to demonstrate inference.

License

NeuralDEM is shared under the CC-BY-NC 4.0 license.

Citation instructions

@article{alkin2024neuraldem,
  title={{NeuralDEM} for real time simulations of industrial particular flows},
  author={Benedikt Alkin and Tobias Kronlachner and Samuele Papa and Stefan Pirker and Thomas Lichtenegger and Johannes Brandstetter},
  journal={Nature Communications Physics},
  year={2025}
  doi={10.1038/s42005-025-02342-4},
}