mlsanityData to 14 Oct 2026

OmniFluids: Physics Pre-trained Modeling of Fluid Dynamics

arXiv 2506.10862 ↗First posted 12 Jun 2025physics.flu-dyn cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Computational fluid dynamics (CFD) drives progress in numerous scientific and engineering fields, yet high-fidelity simulations remain computationally prohibitive. While machine learning approaches offer computing acceleration, they typically specialize in single physical systems or require extensive training data, hindering their applicability in highly nonlinear and 3D flow scenarios. To overcome these limitations, we propose OmniFluids, a pure physics pre-trained model that captures fundamental fluid dynamics laws and adapts efficiently to diverse downstream tasks with minimal data. We develop a training framework combining physics-only pre-training, coarse-grid operator distillation, and few-shot fine-tuning. This enables OmniFluids to retain broad physics knowledge while delivering fast and accurate predictions. Architecturally, OmniFluids integrates a mixture of operators, a multi-frame decoder, and factorized Fourier layers, seamlessly incorporating physics-based supervision while allowing efficient and scalable modeling of diverse tasks. Extensive tests on a broad range of 2D and 3D benchmarks show that OmniFluids outperforms state-of-the-art AI-driven methods in terms of flow field prediction and turbulence statistics. It delivers 10--100× speedups over traditional solvers while maintaining a comparable accuracy and accurately identifies unknown physical parameters from sparse, noisy data. This work demonstrates the potential of training a unified CFD solver exclusively from physics knowledge, offering a new approach for efficient and generalizable modeling across complex fluid systems.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 34 of its 63 references that are in the map. First posted on arXiv 12 Jun 2025.

Builds on

34 of 63 references.

References · 34

  1. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
  2. 2015Distilling the Knowledge in a Neural Networkcited by 10,324
  3. 2016Layer Normalizationcited by 5,669
  4. 2020Fourier Neural Operator for Parametric Partial Differential Equationscited by 2,908
  5. 2019DeepONet: Learning nonlinear operators for identifying differential equations based on the universal approximation theorem of operatorscited by 2,018
  6. 2021Characterizing possible failure modes in physics-informed neural networkscited by 643
  7. 2021Learning the solution operator of parametric partial differential equations with physics-informed DeepOnetscited by 540
  8. 2021Physics-Informed Neural Operator for Learning Partial Differential Equationscited by 527
  9. 2021Machine learning accelerated computational fluid dynamicscited by 471
  10. 2022Message Passing Neural PDE Solverscited by 386
  11. 2021Factorized Fourier Neural Operatorscited by 249
  12. 2023Neural Operators for Accelerating Scientific Simulations and Designcited by 221
22 more
  1. 2023Convolutional Neural Operators for robust and accurate learning of PDEscited by 214
  2. 2018PDE-Net 2.0: Learning PDEs from Data with A Numeric-Symbolic Hybrid Deep Networkcited by 210
  3. 2020Physics-informed learning of governing equations from scarce datacited by 167
  4. 2024DPOT: Auto-Regressive Denoising Operator Transformer for Large-Scale PDE Pre-Trainingcited by 151
  5. 2023Scalable Transformer for PDE Surrogate Modelingcited by 128
  6. 2023In-Context Operator Learning with Data Prompts for Differential Equation Problemscited by 101
  7. 2024Weak baselines and reporting biases lead to overoptimism in machine learning for fluid-related partial differential equationscited by 91
  8. 2023PINNacle: A Comprehensive Benchmark of Physics-Informed Neural Networks for Solving PDEscited by 87
  9. 2022Deep transfer operator learning for partial differential equations under conditional shiftcited by 83
  10. 2023Can Physics-Informed Neural Networks beat the Finite Element Method?cited by 68
  11. 2021Encoding physics to learn reaction-diffusion processescited by 61
  12. 2023From Zero to Turbulence: Generative Modeling for 3D Flow Simulationcited by 49
  13. 2022Multi-resolution partial differential equations preserved learning framework for spatiotemporal dynamicscited by 44
  14. 2023A Neural PDE Solver with Temporal Stencil Modelingcited by 22
  15. 2024PhyMPGN: Physics-encoded Message Passing Graph Network for spatiotemporal PDE systemscited by 18
  16. 2024Graph Neural PDE Solvers with Conservation and Similarity-Equivariancecited by 17
  17. 2022Discovery of partial differential equations from highly noisy and sparse data with physics-informed information criterioncited by 16
  18. 2024P2C2Net: PDE-Preserved Coarse Correction Network for efficient prediction of spatiotemporal dynamicscited by 13
  19. 2023Monte Carlo Neural PDE Solver for Learning PDEs via Probabilistic Representationcited by 11
  20. 2022Deep Random Vortex Method for Simulation and Inference of Navier-Stokes Equationscited by 5
  21. 2024Conservation-informed Graph Learning for Spatiotemporal Dynamics Predictioncited by 4
  22. 2025MultiPDENet: PDE-embedded Learning with Multi-time-stepping for Accelerated Flow Simulationcited by 3

Built on by

Not cited yet.