Neural SPH: Improved Neural Modeling of Lagrangian Fluid Dynamics

Artur Toshev, Jonas A. Erbesdobler, Nikolaus A. Adams, Johannes Brandstetter
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:48428-48452, 2024.

Abstract

Smoothed particle hydrodynamics (SPH) is omnipresent in modern engineering and scientific disciplines. SPH is a class of Lagrangian schemes that discretize fluid dynamics via finite material points that are tracked through the evolving velocity field. Due to the particle-like nature of the simulation, graph neural networks (GNNs) have emerged as appealing and successful surrogates. However, the practical utility of such GNN-based simulators relies on their ability to faithfully model physics, providing accurate and stable predictions over long time horizons - which is a notoriously hard problem. In this work, we identify particle clustering originating from tensile instabilities as one of the primary pitfalls. Based on these insights, we enhance both training and rollout inference of state-of-the-art GNN-based simulators with varying components from standard SPH solvers, including pressure, viscous, and external force components. All Neural SPH-enhanced simulators achieve better performance than the baseline GNNs, often by orders of magnitude in terms of rollout error, allowing for significantly longer rollouts and significantly better physics modeling. Code available under https://github.com/tumaer/neuralsph.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-toshev24a, title = {Neural {SPH}: Improved Neural Modeling of Lagrangian Fluid Dynamics}, author = {Toshev, Artur and Erbesdobler, Jonas A. and Adams, Nikolaus A. and Brandstetter, Johannes}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {48428--48452}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/toshev24a/toshev24a.pdf}, url = {https://proceedings.mlr.press/v235/toshev24a.html}, abstract = {Smoothed particle hydrodynamics (SPH) is omnipresent in modern engineering and scientific disciplines. SPH is a class of Lagrangian schemes that discretize fluid dynamics via finite material points that are tracked through the evolving velocity field. Due to the particle-like nature of the simulation, graph neural networks (GNNs) have emerged as appealing and successful surrogates. However, the practical utility of such GNN-based simulators relies on their ability to faithfully model physics, providing accurate and stable predictions over long time horizons - which is a notoriously hard problem. In this work, we identify particle clustering originating from tensile instabilities as one of the primary pitfalls. Based on these insights, we enhance both training and rollout inference of state-of-the-art GNN-based simulators with varying components from standard SPH solvers, including pressure, viscous, and external force components. All Neural SPH-enhanced simulators achieve better performance than the baseline GNNs, often by orders of magnitude in terms of rollout error, allowing for significantly longer rollouts and significantly better physics modeling. Code available under https://github.com/tumaer/neuralsph.} }
Endnote
%0 Conference Paper %T Neural SPH: Improved Neural Modeling of Lagrangian Fluid Dynamics %A Artur Toshev %A Jonas A. Erbesdobler %A Nikolaus A. Adams %A Johannes Brandstetter %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-toshev24a %I PMLR %P 48428--48452 %U https://proceedings.mlr.press/v235/toshev24a.html %V 235 %X Smoothed particle hydrodynamics (SPH) is omnipresent in modern engineering and scientific disciplines. SPH is a class of Lagrangian schemes that discretize fluid dynamics via finite material points that are tracked through the evolving velocity field. Due to the particle-like nature of the simulation, graph neural networks (GNNs) have emerged as appealing and successful surrogates. However, the practical utility of such GNN-based simulators relies on their ability to faithfully model physics, providing accurate and stable predictions over long time horizons - which is a notoriously hard problem. In this work, we identify particle clustering originating from tensile instabilities as one of the primary pitfalls. Based on these insights, we enhance both training and rollout inference of state-of-the-art GNN-based simulators with varying components from standard SPH solvers, including pressure, viscous, and external force components. All Neural SPH-enhanced simulators achieve better performance than the baseline GNNs, often by orders of magnitude in terms of rollout error, allowing for significantly longer rollouts and significantly better physics modeling. Code available under https://github.com/tumaer/neuralsph.
APA
Toshev, A., Erbesdobler, J.A., Adams, N.A. & Brandstetter, J.. (2024). Neural SPH: Improved Neural Modeling of Lagrangian Fluid Dynamics. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:48428-48452 Available from https://proceedings.mlr.press/v235/toshev24a.html.

Related Material