NeuralDEM is a groundbreaking real-time simulation method for industrial particulate flows that replaces traditional Discrete Element Method (DEM) routines with innovative deep learning surrogates. This approach allows for fast and accurate modeling of complex scenarios, such as hopper discharge and fluidized bed reactors, that were previously computationally intensive. NeuralDEM introduces physics representation by modeling DEM dynamics and introducing multi-branch neural operators for real-time modeling. Surprisingly, NeuralDEM can predict macroscopic quantities and global measurements with high accuracy. The system’s ability to capture chaotic fluidized bed trajectories and particle mixing behavior sets it apart as a game-changer in engineering simulations.
https://nx-ai.github.io/NeuralDEM/