NeuralPFEM - Graph Neural Network (GNN) version

NeuralPFEM is a mesh-based neural surrogate model designed for free-surface fluid flows. It combines a Lagrangian neural network model with the efficient remeshing strategy of the Particle Finite Element Method (PFEM).

In this version, the core of the architecture is a GNN module.

During training, the model uses the mesh connectivity provided by the simulations to construct the graph. At inference time, the PFEM mesh generation algorithm is employed to build the graph dynamically.

Gmsh alpha-shape is employed as mesher

Cite

For more details:

@article{lanteri2025,
         title = {A mesh-based Graph Neural Network approach for surrogate modeling of Lagrangian free surface fluid flows},
         journal = {Computers & Fluids},
         volume = {301},
         pages = {106773},
         year = {2025},
         issn = {0045-7930},
         doi = {https://doi.org/10.1016/j.compfluid.2025.106773},
         url = {https://www.sciencedirect.com/science/article/pii/S0045793025002336},
         author = {Federico Lanteri and Massimiliano Cremonesi},
        }