NeuralDEM, Now Open Source: Real-Time Deep Learning for Industrial Particulate Flows
NeuralDEM is now open source. Following its initial introduction in November 2024 as a first end-to-end deep learning alternative to computationally intensive CFD–DEM multiphysics simulations, we are releasing both the dataset and the model so the community can explore the data, evaluate NeuralDEM, and build on it.
What NeuralDEM is
NeuralDEM presents a field-based physics representation and a multi-branch neural operator that together replace slow particle-resolved routines with fast, adaptable deep learning surrogates. The method treats the Lagrangian discretization of DEM as an underlying continuous field, while directly modeling macroscopic quantities (e.g., occupancy, transport, residence time, mixing) as auxiliary fields. A multi-branch Transformer couples phases (particles and fluid) and supports conditioning on macroscopic parameters (e.g., internal friction angle, flow function coefficient), removing the need for microscopic parameter calibration.
Scale and fidelity
NeuralDEM is built for industrially relevant regimes - from slow, pseudo steady hopper flows to fast, transient fluidized bed reactors. In our numerical experiments, NeuralDEM faithfully models systems with 500k particles and 160k CFD cells, maintaining stability and mass conservation over long rollouts (e.g., 28 s, or 2,800 ML timesteps) while running in real time on a single GPU. It recovers macroscopic observables such as outflow rate, drainage time, residual material, residence time, and mixing indices, and it supports direct macroscopic conditioning - a practical path for engineering workflows.
Real-world applications
- Energy & process engineering: particle transport optimization and process intensification
- Pharma & chemicals: powder flow, coating, mixing
- Mining & construction: granular handling and flow prediction
- Agriculture & food: grain processing and storage dynamics
Resources
- Model & Dataset (Hugging Face): https://huggingface.co/EmmiAI/NeuralDEM
- Code (GitHub): https://github.com/Emmi-AI/NeuralDEM
- Paper (arXiv): https://arxiv.org/abs/2411.09678
- Demo: https://emmi-ai.github.io/NeuralDEM
Ready to accelerate your development?
Explore how AI-powered models are transforming physics simulations and learn how to build or integrate yours.
Latest updates
.webp)
.webp)
.webp)
Emmi AI raises EUR 15M to bring AI to the heart of industrial engineering
Funding from 3VC, Speedinvest, Serena, and PUSH marks the largest seed round ever raised by an Austrian startup and accelerates Emmi AI’s vision for AI-enabled engineering.
Emmi AI releases AB-UPT: Scaling Neural Surrogates to 100M+ Mesh Cells
Emmi AI presents AB-UPT, a novel architecture that scales neural surrogates for computational fluid dynamics (CFD) to industrial-scale problems with over 100 million mesh cells - achieving state of the art accuracy, mesh free inference, and physics consistent predictions.
NeuralDEM, Now Open Source: Real-Time Deep Learning for Industrial Particulate Flows
NeuralDEM introduced in November 2024 as the first end to end deep learning alternative to CFD–DEM multiphysics simulations, is now open source. The dataset and model enable real-time, physically consistent simulation of industrial particulate flows at production scale.
.svg)
