- Physical Review Journals
All JournalsPhysics Magazine

Physical Review E

Reuse & Permissions

It is not necessary to obtain permission to reuse this article or its components as it is available under the terms of the Creative Commons Attribution 4.0 International license. This license permits unrestricted use, distribution, and reproduction in any medium, provided attribution to the author(s) and the published article's title, journal citation, and DOI are maintained. Please note that some figures may have been included with permission from other third parties. It is your responsibility to obtain the proper permission from the rights holder directly for these figures.

Export citation

Export citation

Choose format for download:

Download Citation
  • Open Access

Machine-learning-enhanced collision operator for the lattice Boltzmann method based on invariant networks

Mario Christopher Bedrunka*

Tobias Horstmann

Ben Picard

Dirk Reith

Holger Foysi

  • Department of Engine Acoustics, German Aerospace Center, Bismarckstraße 101, 10625 Berlin, Germany

  • Chair of Fluid Mechanics, University of Siegen, Paul-Bonatz-Straße 9-11, 57076 Siegen-Weidenau, Germany

  • *Contact author: bedrunka.research@gmail.com
PDF

Phys. Rev. E 112, 055308 – Published 12 November, 2025

DOI: https://doi.org/10.1103/rbf2-p8tf

Abstract

Integrating machine learning (ML) techniques in established numerical solvers represents a modern approach to enhance computational fluid dynamics simulations. Within the lattice Boltzmann method (LBM), the collision operator serves as an ideal entry point to incorporate machine learning (ML) techniques to enhance its accuracy and stability. In this work, an invariant neural network is constructed, acting on an equivariant collision operator, optimizing the relaxation rates of nonphysical moments. This optimization enhances robustness to symmetry transformations and ensures consistent behavior across geometric operations. The proposed neural collision operator (NCO) is trained using forced isotropic turbulence simulations driven by spectral forcing, ensuring stable turbulence statistics. The desired performance is achieved by minimizing the energy spectrum discrepancy between direct numerical simulations and underresolved simulations over a specified wave number range. The loss function is further extended to tailor numerical dissipation at high wave numbers, ensuring robustness without compromising accuracy at low and intermediate wave numbers. The NCO's performance is demonstrated using three-dimensional Taylor-Green vortex (TGV) flows, where it accurately predicts the dynamics even in highly underresolved simulations. Compared to other LBM models, such as the BGK and KBC operators, the NCO exhibits improved accuracy while maintaining stability. In addition, the operator shows robust performance in alternative configurations, including turbulent three-dimensional cylinder flow. Finally, an alternative training procedure using time-dependent quantities is introduced. It is based on a reduced TGV model along with newly proposed symmetry boundary conditions. The reduction in memory consumption enables training at significantly higher Reynolds numbers, successfully leading to stable yet accurate simulations.

View figure in article

Physics Subject Headings (PhySH)

Article Text

References (57)

  1. L. S. G. Kovasznay, Turbulence in supersonic flow, J. Aeronautical Sci. 20, 657 (1953).
  2. V. M. Canuto and J. Christensen-Dalsgaard, Turbulence in astrophysics: Stars1, Annu. Rev. Fluid Mech. 30, 167 (1998).
  3. S. B. Pope, Turbulent Flows (Cambridge University Press, Cambridge, 2000).
  4. M. Wang, J. B. Freund, and S. K. Lele, Computational prediction of flow-generated sound, Annu. Rev. Fluid Mech. 38, 483 (2006).
  5. G. Falcucci, G. Amati, P. Fanelli, V. K. Krastev, G. Polverino, M. Porfiri, and S. Succi, Extreme flow simulations reveal skeletal adaptations of deep-sea sponges, Nature (London) 595, 537 (2021).
  6. G. Coleman and R. Sandberg, A primer on direct numerical simulation of turbulence–methods, procedures and guidelines, Technical Report No. A FM-09/01a, University of Southampton, 2010.
  7. S. Succi, R. Benzi, and F. Higuera, The lattice Boltzmann equation: A new tool for computational fluid-dynamics, Physica D 47, 219 (1991).
  8. S. L. Brunton, B. R. Noack, and P. Koumoutsakos, Machine learning for fluid mechanics, Annu. Rev. Fluid Mech. 52, 477 (2020).
  9. J.-X. Wang, J.-L. Wu, and H. Xiao, Physics-informed machine learning approach for reconstructing Reynolds stress modeling discrepancies based on DNS data, Phys. Rev. Fluids 2, 034603 (2017).
  10. R. Maulik and O. San, Adil Rasheed, and Prakash Vedula, Subgrid for two-dimensional turbulence using neural networks, J. Fluid Mech. 858, 122 (2019).
  11. B. Kim, V. C. Azevedo, N. Thuerey, T. Kim, M. Gross, and B. Solenthaler, Deep fluids: A generative network for parameterized fluid simulations, in Computer Graphics Forum (Wiley Online Library, 2019), Vol. 38, pp. 59–70.
  12. S. Chen and G. D. Doolen, Lattice boltzmann method for fluid flows, Annu. Rev. Fluid Mech. 30, 329 (1998).
  13. T. Krüger, H. Kusumaatmaja, A. Kuzmin, O. Shardt, G. Silva, and E. M. Viggen, The Lattice Boltzmann Method (Springer, Heidelberg, 2017).
  14. A. Krämer, K. Küllmer, D. Reith, W. Joppich, and H. Foysi, Semi-Lagrangian off-lattice Boltzmann method for weakly compressible flows, Phys. Rev. E 95, 023305 (2017).
  15. A. Krämer, D. Wilde, K. Küllmer, D. Reith, H. Foysi, and W. Joppich, Lattice boltzmann simulations on irregular grids: Introduction of the natrium library, Comput. Math. Appl. 79, 34 (2020).
  16. D. Wilde, A. Krämer, D. Reith, and H. Foysi, High-order semi-Lagrangian kinetic scheme for compressible turbulence, Phys. Rev. E 104, 025301 (2021).
  17. Z. Guo and K. Xu, Progress of discrete unified gas-kinetic scheme for multiscale flows, Adv. Aerodyn. 3, 6 (2021).
  18. Q. Lou and Z. Guo, Interface-capturing lattice Boltzmann equation model for two-phase flows, Phys. Rev. E 91, 013302 (2015).
  19. D. Ricot, S. Marié, P. Sagaut, and C. Bailly, Lattice Boltzmann method with selective viscosity filter, J. Comput. Phys. 228, 4478 (2009).
  20. I. V. Karlin, F. Bösch, and S. S. Chikatamarla, Gibbs' principle for the lattice-kinetic theory of fluid dynamics, Phys. Rev. E 90, 031302 (2014).
  21. M. Geier, M. Schönherr, A. Pasquali, and M. Krafczyk, The cumulant lattice boltzmann equation in three dimensions: Theory and validation, Comput. Math. Appl. 70, 507 (2015).
  22. A. Krämer, D. Wilde, K. Küllmer, D. Reith, and H. Foysi, Pseudoentropic derivation of the regularized lattice Boltzmann method, Phys. Rev. E 100, 023302 (2019).
  23. S. A. Hosseini, M. Atif, S. Ansumali, and I. V. Karlin, Entropic lattice Boltzmann methods: A review, Comput. Fluids 259, 105884 (2023).
  24. M. C. Bedrunka, D. Wilde, M. Kliemank, D. Reith, H. Foysi, and A. Krämer, Lettuce: Pytorch-based lattice Boltzmann framework, In International Conference on High Performance Computing (Springer, Heidelberg, 2021), pp. 40–55.
  25. A. Corbetta, A. Gabbana, V. Gyrya, D. Livescu, J. Prins, and F. Toschi, Toward learning lattice Boltzmann collision operators, Eur. Phys. E 46, 10 (2023).
  26. J. T. Horstmann, M. C. Bedrunka, and H. Foysi, Lattice Boltzmann method with artificial bulk viscosity using a neural collision operator, Comput. Fluids 272, 106191 (2024).
  27. M. Ataei and H. Salehipour, XLB: A differentiable massively parallel lattice Boltzmann library in Python, Comput. Phys. Commun. 300, 109187 (2024).
  28. D. Marinc and H. Foysi, Investigation of a continuous adjoint-based optimization procedure for aeroacoustic control of plane jets, Int. J. Heat Fluid Flow 38, 200 (2012).
  29. G. Ortali, A. Gabbana, N. Demo, G. Rozza, and F. Toschi, Kinetic data-driven approach to turbulence subgrid modeling, Phys. Rev. Res. 7, 013202 (2025).
  30. P. Lallemand and L.-S. Luo, Theory of the lattice Boltzmann method: Dispersion, dissipation, isotropy, galilean invariance, and stability, Phys. Rev. E 61, 6546 (2000).
  31. P. J. Dellar, Incompressible limits of lattice Boltzmann equations using multiple relaxation times, J. Comput. Phys. 190, 351 (2003).
  32. S. Simonis, M. Haussmann, L. Kronberg, W. Dörfler, and M. J. Krause, Linear and brute force stability of orthogonal moment multiple-relaxation-time lattice boltzmann methods applied to homogeneous isotropic turbulence, Phil. Trans. R. Soc. A. 379, 20200405 (2021).
  33. A. L. Kupershtokh, D. A. Medvedev and D. I. Karpov, On equations of state in a lattice Boltzmann method, Comput. Math. Appl. 58, 965 (2009).
  34. T. Cohen and M. Welling, Group equivariant convolutional networks, in International Conference on Machine Learning (PMLR, 2016), pp. 2990–2999.
  35. T. Cohen, Equivariant convolutional networks, Ph.D. thesis, Taco Cohen, 2021.
  36. F. Zhao, Optimal relaxation collisions for lattice Boltzmann methods, Comput. Math. Appl. 65, 172 (2013).
  37. J. Brasseur, Comments on the Kolmogorov hypothesis of local isotropy in the small scales, in Proceedings of the 29th Aerospace Sciences Meeting, Reno, NV (AIAA, Reston, VA, 1991), AIAA Paper 1991-230.
  38. K. Alvelius, Random forcing of three-dimensional homogeneous turbulence, Phys. Fluids 11, 1880 (1999).
  39. J. Mathew, H. Foysi, and R. Friedrich, A new approach to LES based on explicit filtering, Int. J. Heat Fluid Flow 27, 594 (2006).
  40. B. Fiorina and S. K. Lele, An artificial nonlinear diffusivity method for supersonic reacting flows with shocks, J. Comput. Phys. 222, 246 (2007).
  41. J. Latt and B. Chopard, Lattice Boltzmann method with regularized pre-collision distribution functions, Math. Comput. Simul. 72, 165 (2006).
  42. S. Stolz and N. A. Adams, An approximate deconvolution procedure for large-eddy simulation, Phys. Fluids 11, 1699 (1999).
  43. J. Mathew, R. Lechner, H. Foysi, J. Sesterhenn, and R. Friedrich, An explicit filtering method for large eddy simulation of compressible flows, Phys. Fluids 15, 2279 (2003).
  44. P. Sagaut, Large Eddy Simulation for Incompressible Flows: An Introduction (Springer, Heidelberg, 2006).
  45. M. Meinke, W. Schröder, E. Krause, and T. Rister, A comparison of second- and sixth-order methods for large-eddy simulations, Comput. Fluids 31, 695 (2002).
  46. F. F. Grinstein, L. G. Margolin, and W. J. Rider, Implicit Large Eddy Simulation: Computing Turbulent Fluid Dynamics (Cambridge University Press, Cambridge, 2007), Vol. 10.
  47. N. A. Adams and S. Hickel, Implicit large-eddy simulation: Theory and application, in Advances in Turbulence XII: Proceedings of the 12th EUROMECH European Turbulence Conference, September 7–10, 2009, Marburg, Germany (Springer, Berlin, Heidelberg, 2009), pp. 743–750.
  48. Z. J. Wang, K. Fidkowski, R. Abgrall, F. Bassi, D. Caraeni, A. Cary, H. Deconinck, R. Hartmann, K. Hillewaert, and H. T. Huynh et al., High-order CFD methods: Current status and perspective, Int. J. Numer. Methods Fluids 72, 811 (2013).
  49. M. E. Brachet, D. I. Meiron, S. A. Orszag, B. G. Nickel, R. H. Morf, and U. Frisch, Small-scale structure of the Taylor–Green vortex, J. Fluid Mech. 130, 411 (1983).
  50. J.-I. Choi, R. C. Oberoi, J. R. Edwards, and J. A. Rosati, An immersed boundary method for complex incompressible flows, J. Comput. Phys. 224, 757 (2007).
  51. M. Braza, P. H. H. M Chassaing, and H. H. Minh, Numerical study and physical analysis of the pressure and velocity fields in the near wake of a circular cylinder, J. Fluid Mech. 165, 79 (1986).
  52. D. Russell and Z. J. Wang, A Cartesian grid method for modeling multiple moving objects in 2d incompressible viscous flow, J. Comput. Phys. 191, 177 (2003).
  53. C. Shu, N. Liu, and Y.-T. Chew, A novel immersed boundary velocity correction–lattice Boltzmann method and its application to simulate flow past a circular cylinder, J. Comput. Phys. 226, 1607 (2007).
  54. C. Liu, X. Zheng, and C. H. Sung, Preconditioned multigrid methods for unsteady incompressible flows, J. Comput. Phys. 139, 35 (1998).
  55. J. A. Wright and R. W. Smith, An edge-based method for the incompressible Navier–Stokes equations on polygonal meshes, J. Comput. Phys. 169, 24 (2001).
  56. A. G. Kravchenko and P. Moin, Numerical studies of flow over a circular cylinder at ReD=3900, Phys. Fluids 12, 403 (2000).
  57. https://github.com/lettucecfd/lettuce.

Outline

Information

Phys. Rev. E 112, 055308– Published 12 November, 2025
  • Received 10 December 2024
  • Revised 3 September 2025
  • Accepted 18 September 2025
Crossmark - Check for updates button
Creative Commons Logo

Published by the American Physical Society under the terms of the Creative Commons Attribution 4.0 International license. Further distribution of this work must maintain attribution to the author(s) and the published article's title, journal citation, and DOI.

Published by the American Physical Society

Sign In to Your Journals Account

Filter

Filter

Article Lookup

Enter a citation