Pith. sign in

REVIEW 66 references

Learnable-Differentiable Finite Volume Solver for Accelerated Simulation of Flows

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2507.01975 v1 pith:RMONZYON submitted 2025-06-23 cs.LG cs.AIphysics.flu-dyn

classification cs.LGcs.AIphysics.flu-dyn
keywords flowssimulationfinitegridsldsolversolvervolumecoarse
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Simulation of fluid flows is crucial for modeling physical phenomena like meteorology, aerodynamics, and biomedicine. Classical numerical solvers often require fine spatiotemporal grids to satisfy stability, consistency, and convergence conditions, leading to substantial computational costs. Although machine learning has demonstrated better efficiency, they typically suffer from issues of interpretability, generalizability, and data dependency. Hence, we propose a learnable and differentiable finite volume solver, called LDSolver, designed for efficient and accurate simulation of fluid flows on spatiotemporal coarse grids. LDSolver comprises two key components: (1) a differentiable finite volume solver, and (2) an learnable module providing equivalent approximation for fluxes (derivatives and interpolations), and temporal error correction on coarse grids. Even with limited training data (e.g., only a few trajectories), our model could accelerate the simulation while maintaining a high accuracy with superior generalizability. Experiments on different flow systems (e.g., Burgers, decaying, forced and shear flows) show that LDSolver achieves state-of-the-art performance, surpassing baseline models with notable margins.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

66 extracted references · 51 canonical work pages

  1. [1]

    1995.Computational Fluid Dynamics

    John David Anderson and John Wendt. 1995.Computational Fluid Dynamics. Vol. 206. Springer

  2. [2]

    Yohai Bar-Sinai, Stephan Hoyer, Jason Hickey, and Michael P Brenner. 2019. Learning data-driven discretizations for partial differential equations.Proceedings of the National Academy of Sciences116, 31 (2019), 201814058

  3. [3]

    Ernnie Illyani Basri, Adi Azriff Basri, Vizy Nazira Riazuddin, S F Shahwir, Z Mohammad, and K A Ahmad. 2016. Computational fluid dynamics study in biomedical applications: a review.International Journal of Fluids and Heat Transfer 1, 2 (2016), 2–14

  4. [4]

    Peter Bauer, Alan Thorpe, and Gilbert Brunet. 2015. The quiet revolution of numerical weather prediction.Nature525, 7567 (2015), 47–55

  5. [5]

    Federico Berto, Stefano Massaroli, Michael Poli, and Jinkyoo Park. 2022. Neural solvers for fast and accurate numerical optimal control.Advances in Neural Information Processing Systems(2022)

  6. [6]

    Steven L Brunton, Bernd R Noack, and Petros Koumoutsakos. 2020. Machine learning for fluid mechanics.Annual review of fluid mechanics52, 1 (2020), 477–508

  7. [7]

    Keaton J Burns, Geoffrey M Vasil, Jeffrey S Oishi, Daniel Lecoanet, and Benjamin P Brown. 2020. Dedalus: A flexible framework for numerical simulations with spectral methods.Physical Review Research2, 2 (2020), 023068

  8. [8]

    Zhao Chen, Yang Liu, and Hao Sun. 2021. Physics-Informed Learning of Govern- ing Equations from Scarce Data.Nature Communications12, 1 (2021), 6136

Show all 66 references
  1. [9]

    Gideon Dresdner, Dmitrii Kochkov, Peter Christian Norgaard, Leonardo Zepeda- Nunez, Jamie Smith, Michael Brenner, and Stephan Hoyer. 2023. Learning to correct spectral methods for simulating turbulent flows.Transactions on Machine Learning Research(2023)

  2. [11]

    Han Gao, Luning Sun, and Jian-Xun Wang. 2021. PhyGeoNet: Physics-informed geometry-adaptive convolutional neural networks for solving parameterized steady-state PDEs on irregular domain.J. Comput. Phys.428 (2021), 110079

  3. [12]

    Somdatta Goswami, Katiana Kontolati, Michael D Shields, and George Em Karni- adakis. 2022. Deep transfer operator learning for partial differential equations under conditional shift.Nature Machine Intelligence4, 12 (2022), 1155–1164

  4. [13]

    Gaurav Gupta, Xiongye Xiao, and Paul Bogdan. 2021. Multiwavelet-based opera- tor learning for differential equations.Advances in Neural Information Processing Systems34 (2021), 24048–24062

  5. [14]

    Jayesh K Gupta and Johannes Brandstetter. 2023. Towards Multi-spatiotemporal- scale Generalized PDE Modeling.Transactions on Machine Learning Research (2023)

  6. [15]

    Quercus Hernandez, Alberto Badias, Francisco Chinesta, and Elías Cueto. 2023. Thermodynamics-informed neural networks for physically realistic mixed reality. Computer Methods in Applied Mechanics and Engineering407 (2023), 115912

  7. [16]

    Philipp Holl and Nils Thuerey. 2024. PhiFlow: Differentiable Simulations for PyTorch, TensorFlow and Jax. InInternational Conference on Machine Learning

  8. [17]

    Masanobu Horie and Naoto Mitsume. 2024. Graph Neural PDE Solvers with Conservation and Similarity-Equivariance.International Conference on Machine Learning(2024)

  9. [18]

    Steeven Janny, Aurélien Beneteau, Madiha Nadri, Julie Digne, Nicolas Thome, and Christian Wolf. 2023. Eagle: Large-scale learning of turbulent fluid dynamics with mesh Transformers. InInternational Conference on Learning Representations

  10. [19]

    Hrvoje Jasak, Aleksandar Jemcov, Zeljko Tukovic, et al. 2007. OpenFOAM: A C++ library for complex physics simulations. InInternational workshop on coupled methods in numerical dynamics, Vol. 1000. Dubrovnik, Croatia, 1–20

  11. [20]

    2005.Spectral/hp Element Methods for Computational Fluid Dynamics

    George Karniadakis and Spencer J Sherwin. 2005.Spectral/hp Element Methods for Computational Fluid Dynamics. Oxford University Press, USA

  12. [21]

    T Kobayashi and M Tsubokura. 2009. CFD application in automotive industry. In 100 Volumes of ‘Notes on Numerical Fluid Mechanics’ 40 Years of Numerical Fluid Mechanics and Aerodynamics in Retrospect. Springer, 285–295

  13. [22]

    Dmitrii Kochkov, Jamie A Smith, Ayya Alieva, Qing Wang, Michael P Bren- ner, and Stephan Hoyer. 2021. Machine learning–accelerated computational fluid dynamics.Proceedings of the National Academy of Sciences118, 21 (2021), e2101784118

  14. [23]

    Zongyi Li, Daniel Zhengyu Huang, Burigede Liu, and Anima Anandkumar. 2023. Fourier neural operator with learned deformations for pdes on general geometries. Journal of Machine Learning Research24, 388 (2023), 1–26

  15. [24]

    Zongyi Li, Nikola Kovachki, Kamyar Azizzadenesheli, Burigede Liu, Kaushik Bhattacharya, Andrew Stuart, and Anima Anandkumar. 2021. Fourier Neural Operator for Parametric Partial Differential Equations. InInternational Conference on Learning Representations

  16. [25]

    Zongyi Li, Nikola Kovachki, Chris Choy, Boyi Li, Jean Kossaifi, Shourya Otta, Mohammad Amin Nabian, Maximilian Stadler, Christian Hundt, Kamyar Aziz- zadenesheli, et al. 2024. Geometry-informed neural operator for large-scale 3D PDEs.Advances in Neural Information Processing S...

  17. [26]

    Zongyi Li, Nikola Borislavov Kovachki, Kamyar Azizzadenesheli, Burigede Liu, Kaushik Bhattacharya, Andrew Stuart, and Anima Anandkumar. 2021. Fourier Neural Operator for Parametric Partial Differential Equations. InInternational Conference on Learning Representations

  18. [27]

    Zijie Li, Dule Shu, and Amir Barati Farimani. 2024. Scalable transformer for PDE surrogate modeling.Advances in Neural Information Processing Systems36 (2024)

  19. [28]

    Xin-Yang Liu, Min Zhu, Lu Lu, Hao Sun, and Jian-Xun Wang. 2024. Multi- resolution partial differential equations preserved learning framework for spa- tiotemporal dynamics.Communications Physics7, 1 (2024), 31

  20. [29]

    Zichao Long, Yiping Lu, and Bin Dong. 2019. PDE-net 2.0: Learning PDEs from data with a numericsymbolic hybrid deep network.J. Comput. Phys.399 (2019), KDD ’25, August 3–7, 2025, Toronto, ON, Canada Mengtao Yan et al. 108925

  21. [30]

    Zichao Long, Yiping Lu, Xianzhong Ma, and Bin Dong. 2018. PDE-net: Learning PDEs from data. InInternational Conference on Machine Learning. 3208–3216

  22. [31]

    Lu Lu, Pengzhan Jin, Guofei Pang, Zhongqiang Zhang, and George Em Karni- adakis. 2021. Learning Nonlinear Operators via DeepONet Based on the Universal Approximation Theorem of Operators.Nature Machine Intelligence3, 3 (2021), 218–229

  23. [32]

    Yiping Lu, Aoxiao Zhong, Quanzheng Li, and Bin Dong. 2018. Beyond finite layer neural networks: Bridging deep architectures and numerical differential equations. InInternational Conference on Machine Learning. 3276–3285

  24. [33]

    2023.An introduction to ansys fluent 2023

    John E Matsson. 2023.An introduction to ansys fluent 2023. Sdc Publications

  25. [34]

    2016.The Finite Volume Method

    Fadl Moukalled, Luca Mangani, and Marwan Darwish. 2016.The Finite Volume Method. Springer

  26. [35]

    Tobias Pfaff, Meire Fortunato, Alvaro Sanchez-Gonzalez, and Peter Battaglia

  27. [36]

    Michael Poli, Stefano Massaroli, Atsushi Yamashita, Hajime Asama, and Jinkyoo Park. 2020. Hypersolvers: Toward fast continuous-depth models.Advances in Neural Information Processing Systems33 (2020), 21105–21117

  28. [37]

    Maziar Raissi, Paris Perdikaris, and George E Karniadakis. 2019. Physics-Informed Neural Networks: A Deep Learning Framework for Solving Forward and Inverse Problems Involving Nonlinear Partial Differential Equations.J. Comput. Phys. 378 (2019), 686–707

  29. [38]

    Maziar Raissi, Alireza Yazdani, and George Em Karniadakis. 2020. Hidden Fluid Mechanics: Learning Velocity and Pressure Fields from Flow Visualizations. Science367, 6481 (2020), 1026–1030

  30. [39]

    Chengping Rao, Pu Ren, Yang Liu, and Hao Sun. 2022. Discovering Nonlin- ear PDEs from Scarce Data with Physics-Encoded Learning. InInternational Conference on Learning Representations

  31. [40]

    Chengping Rao, Pu Ren, Qi Wang, Oral Buyukozturk, Hao Sun, and Yang Liu

  32. [41]

    Chengping Rao, Hao Sun, and Yang Liu. 2021. Physics-Informed Deep Learning for Computational Elastodynamics Without Labeled Data.Journal of Engineering Mechanics147, 8 (2021), 04021043

  33. [42]

    Pu Ren, Chengping Rao, Yang Liu, Zihan Ma, Qi Wang, Jian-Xun Wang, and Hao Sun. 2023. Physr: Physics-informed deep super-resolution for spatiotemporal data.J. Comput. Phys.492 (2023), 112438

  34. [43]

    Pu Ren, Chengping Rao, Yang Liu, Jian-Xun Wang, and Hao Sun. 2022. PhyCRNet: Physics-informed convolutional-recurrent network for solving spatiotemporal PDEs.Computer Methods in Applied Mechanics and Engineering389 (2022), 114399

  35. [44]

    Alvaro Sanchez-Gonzalez, Jonathan Godwin, Tobias Pfaff, Rex Ying, Jure Leskovec, and Peter Battaglia. 2020. Learning to simulate complex physics with graph networks. InInternational Conference on Machine Learning. 8459–8468

  36. [45]

    Tapio Schneider, João Teixeira, Christopher S Bretherton, Florent Brient, Kyle G Pressel, Christoph Schär, and A Pier Siebesma. 2017. Climate goals and computing the future of clouds.Nature Climate Change7, 1 (2017), 3–5

  37. [46]

    J S Shang. 2009. Computational fluid dynamics application to aerospace science. The Aeronautical Journal113, 1148 (2009), 619–632

  38. [47]

    2014.CFD vision 2030 study: a path to revolutionary computational aerosciences

    Jeffrey P Slotnick, Abdollah Khodadoust, Juan Alonso, David Darmofal, William Gropp, Elizabeth Lurie, and Dimitri J Mavriplis. 2014.CFD vision 2030 study: a path to revolutionary computational aerosciences. Technical Report

  39. [48]

    Zhiqing Sun, Yiming Yang, and Shinjae Yoo. 2023. A neural pde solver with temporal stencil modeling. InInternational Conference on Machine Learning. PMLR, 33135–33155

  40. [49]

    Zhiqing Sun, Yiming Yang, and Shinjae Yoo. 2023. A neural PDE solver with temporal stencil modeling. InInternational Conference on Machine Learning. 33135–33155

  41. [50]

    Kejun Tang, Jiayu Zhai, Xiaoliang Wan, and Chao Yang. 2024. Adversarial Adaptive Sampling: Unify PINN and Optimal Transport for the Approximation of PDEs. InInternational Conference on Learning Representations

  42. [51]

    Alasdair Tran, Alexander Mathews, Lexing Xie, and Cheng Soon Ong. 2022. Factorized Fourier neural operators. InInternational Conference on Learning Representations

  43. [52]

    Kiwon Um, Robert Brand, Yun Raymond Fei, Philipp Holl, and Nils Thuerey

  44. [53]

    Qi Wang, Pu Ren, Hao Zhou, Xin-Yang Liu, Zhiwen Deng, Yi Zhang, Ruizhi Chengze, Hongsheng Liu, Zidong Wang, Jian-Xun Wang, et al. 2024. P 2C2 Net: PDE-Preserved Coarse Correction Network for Efficient Prediction of Spatiotem- poral Dynamics.Advances in Neural Information Proce...

  45. [54]

    Rui Wang, Karthik Kashinath, Mustafa Mustafa, Adrian Albert, and Rose Yu

  46. [55]

    Rui Wang, Robin Walters, and Rose Yu. 2021. Incorporating symmetry into deep dynamics models for improved generalization. InInternational Conference on Learning Representations

  47. [56]

    Haixu Wu, Huakun Luo, Haowen Wang, Jianmin Wang, and Mingsheng Long

  48. [57]

    Mengtao Yan, Zhiming Zhang, Shangce Gao, and Shuyang Cao. 2022. Predicting aerodynamic pressure on a square cylinder from wake velocity field by masked gated recurrent unit model.Physics of Fluids34, 11 (2022)

  49. [58]

    InProceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining

    Towards Physics-Informed Deep Learning for Turbulent Flow Prediction. InProceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. 1457–1466

  50. [59]

    Ruiyang Zhang, Yang Liu, and Hao Sun. 2020. Physics-Informed Multi-LSTM Networks for Metamodeling of Nonlinear Structures.Computer Methods in Applied Mechanics and Engineering369 (2020), 113226

  51. [60]

    Jiawei Zhuang, Dmitrii Kochkov, Yohai Bar-Sinai, Michael P Brenner, and Stephan Hoyer. 2021. Learned discretizations for passive scalar advection in a two- dimensional turbulent flow.Physical Review Fluids6, 6 (2021), 064605

  52. [61]

    2005.The Finite Element Method: Its Basis and Fundamentals

    Olek C Zienkiewicz, Robert L Taylor, and Jian Z Zhu. 2005.The Finite Element Method: Its Basis and Fundamentals. Elsevier. Learnable-Differentiable Finite Volume Solver for Accelerated Simulation of Flows KDD ’25, August 3–7, 2025, Toronto, ON, Canada Appendix This supplementa...

  53. [63]

    Bocheng Zeng, Qi Wang, Mengtao Yan, Yang Liu, Ruizhi Chengze, Yi Zhang, Hongsheng Liu, Zidong Wang, and Hao Sun. 2025. PhyMPGN: Physics-encoded Message Passing Graph Network for spatiotemporal PDE systems.International Conference on Learning Representations(2025)

  54. [67]

    The learning rate is set to5× 10−4, with a decay factor of 0.9 applied every 5000 steps

    and the Adam optimizer. The learning rate is set to5× 10−4, with a decay factor of 0.9 applied every 5000 steps. The model is trained for 5000 epochs with a batch size of 20. PeRCNN: The PeRCNN model [40] is used with the standard settings. The optimization is performed using ...

  55. [2020]

    Solver-in-the-loop: Learning from differentiable physics to interact with iterative pde-solvers.Advances in Neural Information Processing Systems33 (2020), 6111–6122

  56. [2021]

    InInternational Conference on Learning Representations

    Learning meshbased simulation with graph networks. InInternational Conference on Learning Representations

  57. [2023]

    Encoding physics to learn reaction–diffusion processes.Nature Machine Intelligence5, 7 (2023), 765–779

  58. [2024]

    International Conference on Machine Learning(2024)

    Transolver: A fast transformer solver for PDEs on general geometries. International Conference on Machine Learning(2024)

Pith tools