Pith. sign in

REVIEW 4 major objections 6 minor 61 references

Machine Learning Enhanced Collision Operator for the Lattice Boltzmann Method Based on Invariant Networks

T0 review · 4 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper claims that an invariant neural network that sets the relaxation rates of nonphysical moments yields a stable, accurate collision operator for underresolved D3Q27 lattice Boltzmann turbulence simulations.

desk verdict A plausible 3D invariant neural collision operator with real transferability evidence, but the missing constant-rate baseline and reporting gaps keep the central claim from being fully secured. read the letter →

arxiv 2412.08229 v1 pith:2T344IMT submitted 2024-12-11 physics.comp-ph

Mario Christopher Bedrunka , Tobias Horstmann , Ben Picard , Dirk Reith , Holger Foysi This is my paper · ORCID
classification physics.comp-ph PACS 47.11.-j47.27.ep47.27.E
keywords latticeBoltzmannmethodneuralcollisionoperatormultiplerelaxationtimeequivariantnetworkturbulencesimulationlargeeddyTaylor-Greenvortex
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper proposes a neural collision operator (NCO) for the three-dimensional lattice Boltzmann method on the D3Q27 stencil: the standard multiple-relaxation-time structure is kept, but the relaxation rates of nonphysical higher-order moments are computed by a small neural network that is invariant under the octahedral symmetry group of the lattice. Training on forced isotropic turbulence minimizes the energy-spectrum discrepancy against direct numerical simulation and adds dissipation at the cutoff wavenumber, producing an operator that stays accurate and stable in highly underresolved simulations. In Taylor-Green vortex and cylinder-flow benchmarks the NCO matches reference data better than the BGK, KBC, and regularized operators at equal resolution. A second training procedure, based on time-dependent dissipation of a symmetry-reduced Taylor-Green vortex, reduces memory use by a factor of 64 and enables training at higher Reynolds numbers with the same accuracy. The central claim is that learned relaxation rates can put implicit turbulence dissipation inside the collision operator rather than in an explicit subgrid model.

What carries the argument

The load-bearing mechanism is a symmetry-constrained neural collision operator: the collision matrix is diagonal in an Hermite-moment basis, with conserved moments relaxed at rate one, shear moments at the kinematic-viscosity rate, and higher-order moments (grouped into four order groups on D3Q27) relaxed at rates output by a two-layer network with 20 nodes per layer. Equivariance is enforced by relaxing moments of equal order with a single rate; invariance of the network is enforced by averaging its output over all 48 transformations of the octahedral group acting on the moment vector. A sigmoid output layer keeps every learned relaxation time above 0.5, guaranteeing over-relaxation and hence stability. Training minimizes the mean-squared error of the energy spectrum against DNS over wavenumbers $4 \leq \kappa \leq 10$ plus a weighted dissipation term at the cutoff wavenumber, which builds implicit LES-like dissipation into the collision step.

What would settle it

Run the trained NCO on a coarse-grid turbulent channel or boundary-layer flow with walls, never seen in training, and compare mean velocity and Reynolds-stress profiles with DNS; if the profiles deviate beyond the KBC baseline or the simulation becomes unstable, the claimed transferability of learned relaxation rates from forced isotropic turbulence is falsified.

Watch

Extended reading notes

Core claim

The central claim is that a collision operator for three-dimensional lattice Boltzmann simulations can be learned from data while keeping the exact symmetry structure of the lattice. The paper constructs the operator as a multiple-relaxation-time collision step in an Hermite-moment basis, with the relaxation rates of the nonphysical higher-order moments supplied by a two-layer neural network. Keeping one rate per moment order makes the operator equivariant under the full octahedral group, and averaging the network output over all 48 group elements makes the network invariant, so the learned collision behavior does not depend on the orientation of the stencil. Trained on forced isotropic turbulence by matching the DNS energy spectrum over wavenumbers $4 \leq \kappa \leq 10$ with added cutoff dissipation, the NCO tracks reference energy dissipation rates in underresolved Taylor-Green vortex simulations where BGK becomes unstable and KBC over-dissipates, and its cylinder-flow velocity and Reynolds-stress profiles agree with reference data. A second training route using time-dependent dissipation of a symmetry-reduced Taylor-Green vortex achieves the same stability at higher Reynolds numbers and confirms that the learned rates transfer across Mach numbers.

Load-bearing premise

The approach rests on the assumption that relaxation-rate rules learned from forced isotropic turbulence transfer to unseen flow configurations and Reynolds or Mach ranges; the paper demonstrates this for Taylor-Green and cylinder flows but does not derive why the transfer should hold.

Editorial extensions

If this is right

  • Underresolved $32^3$ Taylor-Green simulations with the NCO track reference dissipation rates at Reynolds numbers where BGK becomes unstable, so the learned relaxation rates supply the missing dissipation without an explicit subgrid model.
  • The NCO produces less early-time dissipation than KBC or regularized operators while remaining stable, so accuracy and stability are not forced to trade off in the usual way.
  • The symmetry-reduced training domain cuts memory by a factor of 64 and enables training at higher Reynolds numbers, producing an NCO that remains accurate across Mach numbers from 0.05 to 0.2.
  • At Reynolds number 3900 cylinder flow, the NCO reproduces time-averaged velocity and Reynolds-stress profiles without local grid refinement, indicating that relaxation rates learned from homogeneous turbulence transfer to bluff-body wakes.
  • Training on shear-rate-based dissipation yields very accurate dissipation behavior but unstable underresolved simulations, showing that the training objective itself controls the accuracy-stability balance.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The same invariant-averaging construction should transfer to other stencils such as D2Q9 or D3Q19 by replacing the octahedral group with the stencil's symmetry group, though the paper only demonstrates D3Q27.
  • Because invariance is achieved by averaging over all 48 group elements, inference cost grows by a factor of 48; a cheaper parameterization that builds invariance into the network architecture rather than averaging could make the operator practical for production LES.
  • The cutoff-wavenumber weighting in the loss effectively selects the implicit subgrid dissipation scale; training with different forcing spectra could make the induced dissipation adapt to the grid scale, a connection to classical LES modeling the paper does not explore.
  • If the transferability claim holds broadly, collision operators with learned relaxation rates could be dropped into existing LBM codes with no change to streaming or boundary handling, which is a natural practical route beyond the benchmarks shown.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper introduces a neural collision operator (NCO) for D3Q27 lattice Boltzmann simulations. The collision operator is an MRT model whose relaxation rates for four groups of nonphysical moments are produced by an invariant neural network, constructed by averaging a base network over the full octahedral group of lattice-symmetry transformations applied to the moments. Training is performed on downsampled forced isotropic turbulence by minimizing an energy-spectrum loss with an additional high-wavenumber dissipation penalty. The trained operator is evaluated on a convergence test, Taylor-Green vortex flows, and cylinder flow, and an alternative training scheme based on time-dependent TGV dissipation rates with symmetry-reduced domains is presented. The paper claims that NCO is more accurate than BGK and KBC and remains stable in highly under-resolved regimes.

Significance. If the transferability claims hold, this is a timely and useful contribution: it couples group-equivariant/invariant architectures with the lattice Boltzmann method and provides an implicit-LES-like mechanism by learning nonphysical relaxation rates. The equivariance-preserving MRT construction and the reduced-domain symmetry training procedure are valuable ideas. The paper does not provide code or data, and the reported evidence is single-realization, so the strength of the current verification is limited. The central claim—that an operator trained on forced isotropic turbulence transfers to Taylor-Green vortex and cylinder flows—is plausible but not yet secured by the presented experiments.

major comments (4)
  1. [II.C] The training viscosity is not specified: the text states 'a viscosity of ν = xx' immediately before Section II.D. Without this value the training Reynolds number is unknown, and one cannot tell whether the test cases (TGV Re = 100–1600, cylinder Re = 3900) are interpolation or extrapolation from the training distribution. Since the paper's central claim is transferability, this missing parameter must be supplied together with the corresponding Reynolds number estimate.
  2. [III.A] The loss function in Eq. (29) directly minimizes the energy-spectrum discrepancy on forced isotropic turbulence, so the agreement shown in Figure 5 on the same flow class is a fitting result, not an independent prediction. The independent evidence for transferability is limited to Sections III.C and III.D. In addition, the weight ω = 20 is chosen after comparing only ω = 1 and ω = 100 (Figure 4), with no sensitivity analysis; the claimed accuracy–robustness balance rests on this post hoc choice.
  3. [III.C] All reported comparisons are single trajectories with no ensemble statistics or error bars, despite stochastic forcing (random phases in Eqs. (19)–(27)) and stochastic training. For example, the dissipation curves in Figure 8 and the cylinder profiles in Figure 11 could reflect favorable realizations. At least a few independent forcing and training realizations, or confidence bands, are needed to support the claimed systematic superiority over BGK and KBC.
  4. [II.B] No baseline is provided in which four constant nonphysical relaxation rates are tuned on the same loss function, Eq. (29). The NCO has 1064 network parameters producing four relaxation rates; if a constant-rate MRT with tuned rates matches the dissipation accuracy, then the adaptive and invariant-network components are not responsible for the improvement. Such a baseline is necessary to attribute the observed gains to the machine-learning mechanism.
minor comments (6)
  1. [II.C] The placeholder 'ν = xx' must be replaced with the actual value used in the training simulations.
  2. [III.C, Eq. (33)] The pressure initial condition appears malformed: the standard TGV expression is p = (1/16)(cos 2x + cos 2y)(cos 2z + 2), while the manuscript has cos(2y) cos(2z + 2) inside the parentheses.
  3. [III.D] 'K` arm` an' is a LaTeX artifact and should read 'Kármán'. In addition, the reference name 'Lorenceo & Shih' is inconsistent with 'Lourenco' used in the caption of Figure 11.
  4. [III.B] The convergence study reports only NCO errors; including BGK and KBC curves would make the comparison informative. The text should also state explicitly that the L2 errors are measured after exactly one time step.
  5. [IV.A, Table I] Entries such as '16 3' should be typeset as '16^3' for readability.
  6. [II.B] Equation (15) requires averaging over all 48 elements of the octahedral group, implying 48 forward passes of the network per collision; the computational cost of this averaging is not discussed, which is relevant for practical use.

Circularity Check

2 steps flagged · score 3.0 of 10

No central circularity: the main transfer claim rests on independent TGV and cylinder benchmarks, but the forced-isotropy spectrum and dissipation-rate demonstrations are partly in-sample fits.

  1. fitted input called prediction [Sec. II D, Eq. (29) and Sec. III A, Fig. 5]
    "The loss function compares the shape of the energy spectrum for the wave number range 4 ≤ k ≤ 10 between the results obtained using the nco (see eq. 7) and the ground truth. ... Figure 5: Energy spectrum comparison: analyzing different collision operators relative to dns data with a weight of ω = 20 used in the training procedure of the nco."

    The NCO's relaxation rates are optimized by minimizing MSE(E(κ,θ), E(κ)) on forced isotropic turbulence over exactly the displayed wavenumber range, plus the κmax penalty. The Fig. 5 spectrum match is therefore a training convergence check on the loss function, not an independent prediction of the energy spectrum. The paper labels this section as calibration, and the central transfer claims are separately tested on TGV and cylinder, so this step is a partial, non-central example of fitted input being presented as performance rather than a fully circular derivation.

  2. fitted input called prediction [Sec. IV A, loss definition and Fig. 12]
    "The loss function used is the mean squared error of the difference in the dissipation rates, −dk/dt, between the reference results and the results simulated using the nco. The dissipation is compared across 100 discrete points between t = 0 and t = 15. ... Figure 12 compares the energy dissipation rates obtained using the bgk operator, the kbc operator, the reg operator, and the newly trained nco using time-dependent dissipation data."

    Here the NCO is trained to minimize the MSE of the TGV dissipation curve −dk/dt, and Fig. 12 then reports the same dissipation curves as evidence of accuracy. For configurations contained in Table I, matching the curve is the training objective itself, so the agreement is fitted rather than predicted. Some Fig. 12 rows use resolutions beyond those listed in Table I, which gives partial independent content, but the in-sample portion of the demonstration reduces by construction to the loss function.

full rationale

The paper's headline claim is that an NCO trained on forced isotropic turbulence transfers to unseen three-dimensional Taylor-Green vortex and cylinder flows. Those two benchmarks (Secs. III C and III D) are genuinely external to the training data and provide independent support for the central claim. The visible agreement of the forced-isotropy energy spectrum (Fig. 5) is transparently the quantity minimized in Eq. (29), so it is a fit rather than a prediction; the same holds for the dissipation-rate training in Sec. IV, where the plotted curves are the loss target. These in-sample demonstrations contribute to the paper's overall impression of accuracy but do not force the central transferability conclusion by definition. Self-citations to the Lettuce framework [25] and prior neural-collision-operator work [27] provide implementation and context, but the current derivation does not reduce to those citations, and no load-bearing uniqueness theorem or ansatz-only citation is used. Overall: partial circularity in the in-sample evaluations, but the central derivation is self-contained against external benchmarks.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The NCO is an empirical model whose output rests on trained weights and manually chosen training hyperparameters. It introduces no new physical entity. The main assumptions are the equivariance construction, the coarse-graining of DNS data, and transferability of learned relaxation rates to other flows.

free parameters (3)
  • loss weight omega = 20 (after manual comparison of 1 and 100)
    In Eq. (29) and Section III A, this weight controls the high-wave-number dissipation term and was chosen by hand to balance accuracy and stability.
  • neural network weights = 1064 trainable parameters, optimized with Adam
    These parameters define the NCO and are fitted to training data rather than derived from physics.
  • forcing spectrum parameters kf and c = not reported in text
    In Eq. (24) these set the peak and width of the Gaussian forcing in wave-number space; without numeric values the training data generation is not fully specified.
assumptions (5)
  • domain assumption The MRT collision operator with moments relaxed per order group is equivariant under the full octahedral group.
    Section II B, Eq. (14) states this was checked numerically, not proven, and requires equal relaxation rates for moments of equal order.
  • domain assumption The acoustic-scaling coarse-graining formula in Eq. (28) accurately maps fine-grid distributions to coarse-grid distributions.
    Section II D uses this mapping to create training targets; bias here would propagate into the learned operator.
  • domain assumption Training on forced isotropic turbulence transfers to Taylor-Green vortex and cylinder flow.
    The central benchmarks in Sections III C and III D rely on this transferability, which is tested empirically but not guaranteed.
  • domain assumption Minimizing energy spectrum discrepancy over wave numbers 4 to 10 is sufficient to train relaxation rates for accurate dissipation.
    Eq. (29) defines the training objective; low modes are excluded because of random forcing, and high modes are penalized separately.
  • domain assumption The reduced Taylor-Green vortex symmetry boundary conditions preserve the original dynamics.
    Appendix A derives the boundary conditions; Fig. 14 verifies them for BGK, and the NCO training assumes they hold.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Machine Learning Enhanced Collision Operator for the Lattice Boltzmann Method Based on Invariant Networks." pith.science (2026). https://pith.science/paper/2T344IMT

@misc{pith2026241208229,
  author       = {Pith},
  title        = {Pith review of: Machine Learning Enhanced Collision Operator for the Lattice Boltzmann Method Based on Invariant Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2T344IMT}},
  note         = {Machine review of arXiv:2412.08229}
}
read the original abstract

Integrating machine learning techniques in established numerical solvers represents a modern approach to enhancing computational fluid dynamics simulations. Within the lattice Boltzmann method (LBM), the collision operator serves as an ideal entry point to incorporate machine learning 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 non-physical 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 superior 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 higher Reynolds numbers, successfully leading to stable yet accurate simulations.

Figures

Figures reproduced from arXiv: 2412.08229 by the authors.

Figure 1
Figure 1. FIG. 1: Illustration of group operations applied on the [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2: Illustration of the neural network topology, [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3: 2D representation of the spectral force forcing. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: FIG. 4: Impact of the weight on the last wave number [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5: Energy spectrum comparison: analyzing [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 7
Figure 7. Figure 7: FIG. 7: Volumetric rendering of the vorticity of the 3D [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: FIG. 8: Energy dissipation [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 9
Figure 9. Figure 9: FIG. 9: Volumetric rendering of the vorticity magnitude in three-dimensional cylinder flow, illustrating the K`arm`an [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]
Figure 10
Figure 10. Figure 10: FIG. 10: Domain geometry: The stream-wise domain [PITH_FULL_IMAGE:figures/full_fig_p009_10.png]
Figure 11
Figure 11. Figure 11: FIG. 11: Cylinder flow: Time-averaged profiles along [PITH_FULL_IMAGE:figures/full_fig_p010_11.png]
Figure 12
Figure 12. Figure 12: FIG. 12: Energy dissipation rate, [PITH_FULL_IMAGE:figures/full_fig_p011_12.png]
Figure 13
Figure 13. Figure 13: FIG. 13: Comparison of energy dissipation rates for [PITH_FULL_IMAGE:figures/full_fig_p012_13.png]
Figure 14
Figure 14. Figure 14: FIG. 14: Comparison of the energy dissipation of the [PITH_FULL_IMAGE:figures/full_fig_p015_14.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

61 extracted references · 61 canonical work pages

  1. [1]

    Kovasznay

    L.S.G. Kovasznay. Turbulence in supersonic flow. J. of the Aeronautical Sci., 20(10):657–682, 1953

    work page 1953

  2. [2]

    V. M. Canuto and J. Christensen-Dalsgaard. Turbulence in astrophysics: Stars1. Annual Review of Fluid Mechan- ics, 30(Volume 30, 1998):167–198, 1998

    work page 1998

  3. [3]

    For clarity, it is defined as ∇(n) = ∇(n) α1...αn = ∂ ∂xa1

    (10) The multivariate nabla operator ∇(n) provides a compact way to express n-th order spatial derivatives. For clarity, it is defined as ∇(n) = ∇(n) α1...αn = ∂ ∂xa1 ... ∂ ∂xan . (11) In this study, since we are working in three spatial di- mensions (D = 3), the indices a1 . . . an take values from {x, y, z}. In order to construct orthogonal vectors m, t...

    work page

  4. [4]

    Turbulent flows

    Stephen B Pope. Turbulent flows. Cambridge university press, 2000

    work page 2000

  5. [5]

    M. Wang, J. B. Freund, and S. K. Lele. Computational prediction of flow-generated sound. Annual Review of Fluid Mechanics, 38:483–512, 2006

    work page 2006

  6. [6]

    Extreme flow simulations reveal skeletal adaptations of deep-sea sponges

    Giacomo Falcucci, Giorgio Amati, Pierluigi Fanelli, Ves- selin K Krastev, Giovanni Polverino, Maurizio Porfiri, and Sauro Succi. Extreme flow simulations reveal skeletal adaptations of deep-sea sponges. Nature, 595(7868):537– 541, 2021

    work page 2021

  7. [7]

    Department of Energy

    U.S. Department of Energy. Combustion-Research and Development. Office of Industrial Technologies. , DOE/GO-102001-1213, 2001

    work page 2001

  8. [8]

    The lattice boltzmann equation: A new tool for computa- tional fluid-dynamics

    Sauro Succi, Roberto Benzi, and Francisco Higuera. The lattice boltzmann equation: A new tool for computa- tional fluid-dynamics. Physica D: Nonlinear Phenomena, 47(1):219–230, 1991

    work page 1991

Show all 61 references
  1. [9]

    Machine learning for fluid mechanics

    Steven L Brunton, Bernd R Noack, and Petros Koumout- sakos. Machine learning for fluid mechanics. Annual re- view of fluid mechanics , 52(1):477–508, 2020

    2020

  2. [10]

    Physics- informed machine learning approach for reconstructing reynolds stress modeling discrepancies based on dns data

    Jian-Xun Wang, Jin-Long Wu, and Heng Xiao. Physics- informed machine learning approach for reconstructing reynolds stress modeling discrepancies based on dns data. Physical Review Fluids , 2(3):034603, 2017

    2017

  3. [11]

    A primer on di- rect numerical simulation of turbulence – methods, pro- cedures and guidelines

    Gary Coleman and Richard Sandberg. A primer on di- rect numerical simulation of turbulence – methods, pro- cedures and guidelines. Technical Report AFM-09/01a, University of Southampton , 03 2010

    2010

  4. [12]

    Deep fluids: A generative network for parameterized fluid simulations

    Byungsoo Kim, Vinicius C Azevedo, Nils Thuerey, Theodore Kim, Markus Gross, and Barbara Solenthaler. Deep fluids: A generative network for parameterized fluid simulations. In Computer graphics forum , volume 38, pages 59–70. Wiley Online Library, 2019

    2019

  5. [13]

    Lattice boltzmann method for fluid flows

    Shiyi Chen and Gary D Doolen. Lattice boltzmann method for fluid flows. Annual review of fluid mechanics, 30(1):329–364, 1998

    1998

  6. [14]

    The lattice boltzmann method

    Timm Kr¨ uger, Halim Kusumaatmaja, Alexandr Kuzmin, Orest Shardt, Goncalo Silva, and Erlend Magnus Viggen. The lattice boltzmann method. Springer International Publishing, 10(978-3):4–15, 2017

    2017

  7. [15]

    Subgrid modelling for two-dimensional turbu- lence using neural networks

    Romit Maulik, Omer San, Adil Rasheed, and Prakash Vedula. Subgrid modelling for two-dimensional turbu- lence using neural networks. Journal of Fluid Mechanics , 858:122–144, 2019

    2019

  8. [16]

    Lattice Boltzmann simulations on irregular grids: Introduction of the NATriuM library

    Andreas Kr¨ amer, Dominik Wilde, Knut K¨ ullmer, Dirk Reith, Holger Foysi, and Wolfgang Joppich. Lattice Boltzmann simulations on irregular grids: Introduction of the NATriuM library. Computers & Mathematics with Applications, 79(1):34–54, 2020

    2020

  9. [17]

    High-order semi-lagrangian kinetic scheme for compressible turbulence

    Dominik Wilde, Andreas Kr¨ amer, Dirk Reith, and Hol- ger Foysi. High-order semi-lagrangian kinetic scheme for compressible turbulence. Physical Review E , 104(2):025301, 2021

    2021

  10. [18]

    Progress of discrete unified gas- kinetic scheme for multiscale flows

    Zhaoli Guo and Kun Xu. Progress of discrete unified gas- kinetic scheme for multiscale flows. Advances in Aerody- 16 namics, 3, 12 2021

    2021

  11. [19]

    Semi-lagrangian off-lattice boltzmann method for weakly compressible flows

    Andreas Kr¨ amer, Knut K¨ ullmer, Dirk Reith, Wolfgang Joppich, and Holger Foysi. Semi-lagrangian off-lattice boltzmann method for weakly compressible flows. Phys- ical Review E , 95:023305, Feb 2017

    2017

  12. [20]

    Ricot, S

    D. Ricot, S. Mari´ e, P. Sagaut, and C. Bailly. Lattice boltzmann method with selective viscosity filter. J. Com- put. Physics , 228:4478–4490, 2009

    2009

  13. [21]

    Gibbs’ principle for the lattice-kinetic theory of fluid dy- namics

    Ilya V Karlin, Fabian B¨ osch, and SS Chikatamarla. Gibbs’ principle for the lattice-kinetic theory of fluid dy- namics. Phys. Rev. E - Stat. Nonlinear, Soft Matter Phys., 90(3):1–5, 2014

    2014

  14. [22]

    The cumulant lattice boltzmann equation in three dimensions: Theory and validation

    Martin Geier, Martin Sch¨ onherr, Andrea Pasquali, and Manfred Krafczyk. The cumulant lattice boltzmann equation in three dimensions: Theory and validation. Computers & Mathematics with Applications , 70(4):507– 547, 2015

    2015

  15. [23]

    Interface-capturing lattice boltzmann equation model for two-phase flows

    Qin Lou and Zhaoli Guo. Interface-capturing lattice boltzmann equation model for two-phase flows. Phys. Rev. E, 91:013302, Jan 2015

    2015

  16. [24]

    Hosseini, M

    S.A. Hosseini, M. Atif, S. Ansumali, and I.V. Karlin. En- tropic lattice boltzmann methods: A review. Computers & Fluids , 259:105884, 2023

    2023

  17. [25]

    Lettuce: Pytorch-based lattice boltzmann framework

    Mario Christopher Bedrunka, Dominik Wilde, Mar- tin Kliemank, Dirk Reith, Holger Foysi, and Andreas Kr¨ amer. Lettuce: Pytorch-based lattice boltzmann framework. In International Conference on High Per- formance Computing, pages 40–55. Springer, 2021

    2021

  18. [26]

    Toward learning lattice boltzmann collision operators

    Alessandro Corbetta, Alessandro Gabbana, Vitaliy Gyrya, Daniel Livescu, Joost Prins, and Federico Toschi. Toward learning lattice boltzmann collision operators. The European Physical Journal E , 46(3):10, 2023

    2023

  19. [27]

    Pseudoentropic derivation of the regularized lattice boltzmann method

    Andreas Kr¨ amer, Dominik Wilde, Knut K¨ ullmer, Dirk Reith, and Holger Foysi. Pseudoentropic derivation of the regularized lattice boltzmann method. Phys. Rev. E , 100:023302, Aug 2019

    2019

  20. [28]

    Enhancing lattice kinetic schemes for fluid dynamics with lattice-equivariant neu- ral networks

    Giulio Ortali, Alessandro Gabbana, Imre Atmodimedjo, and Alessandro Corbetta. Enhancing lattice kinetic schemes for fluid dynamics with lattice-equivariant neu- ral networks. arXiv preprint arXiv:2405.13850 , 2024

    2024 arXiv

  21. [29]

    Kinetic data-driven approach to turbulence subgrid modeling

    Alessandro Gabbana, Giulio Ortali, Nicola Demo, Gi- anluigi Rozza, and Federico Toschi. Kinetic data-driven approach to turbulence subgrid modeling. Bulletin of the American Physical Society, 2024

    2024

  22. [30]

    Xlb: A differentiable massively parallel lattice boltzmann li- brary in python

    Mohammamdmehdi Ataei and Hesam Salehipour. Xlb: A differentiable massively parallel lattice boltzmann li- brary in python. Computer Physics Communications , page 109187, 2024

    2024

  23. [31]

    Lattice boltzmann method with ar- tificial bulk viscosity using a neural collision operator

    Jan Tobias Horstmann, Mario Christopher Bedrunka, and Holger Foysi. Lattice boltzmann method with ar- tificial bulk viscosity using a neural collision operator. Computers & Fluids , 272:106191, 2024

    2024

  24. [32]

    Theory of the lattice boltzmann method: Dispersion, dissipation, isotropy, galilean invariance, and stability

    Pierre Lallemand and Li-Shi Luo. Theory of the lattice boltzmann method: Dispersion, dissipation, isotropy, galilean invariance, and stability. Physical review E , 61(6):6546, 2000

    2000

  25. [33]

    Incompressible limits of lattice boltzmann equations using multiple relaxation times

    Paul J Dellar. Incompressible limits of lattice boltzmann equations using multiple relaxation times. Journal of Computational Physics , 190(2):351–370, 2003

    2003

  26. [34]

    However, these sets are chosen as constant values through the simulation and do not adapt to dynamic changes including laminar- turbulent transition

    identified optimal sets by brute-force analysis to con- clude stability criteria in turbulent flows. However, these sets are chosen as constant values through the simulation and do not adapt to dynamic changes including laminar- turbulent transition. In this work, we want to t...

  27. [35]

    Investigation of a con- tinuous adjoint-based optimization procedure for aeroa- coustic control of plane jets

    Daniel Marinc and Holger Foysi. Investigation of a con- tinuous adjoint-based optimization procedure for aeroa- coustic control of plane jets. International Journal of Heat and Fluid Flow , 38(0):200 – 212, 2012

    2012

  28. [36]

    Group equivariant con- volutional networks

    Taco Cohen and Max Welling. Group equivariant con- volutional networks. In International conference on ma- chine learning, pages 2990–2999. PMLR, 2016

    2016

  29. [37]

    Equivariant convolutional networks

    Taco Cohen. Equivariant convolutional networks . PhD thesis, Taco Cohen, 2021

    2021

  30. [38]

    Linear and brute force stability of orthogonal moment multiple-relaxation- time lattice boltzmann methods applied to homogeneous isotropic turbulence

    Stephan Simonis, Marc Haussmann, Louis Kronberg, Willy D¨ orfler, and Mathias J Krause. Linear and brute force stability of orthogonal moment multiple-relaxation- time lattice boltzmann methods applied to homogeneous isotropic turbulence. Philosophical Transactions of the Roya...

    2021

  31. [39]

    This method in- volves a force function ˆFα (k, t) in a Fourier space that distributes the force across a range of small wave num- bers facilitated by randomly generated phases

    developed spectral models that inject a power input P over a discrete time interval ∆t by stimulating random spherical spheres as shown in Figure 3. This method in- volves a force function ˆFα (k, t) in a Fourier space that distributes the force across a range of small wave nu...

  32. [40]

    On equations of state in a lat- tice boltzmann method

    Alexander L Kupershtokh, DA Medvedev, and DI2548181 Karpov. On equations of state in a lat- tice boltzmann method. Computers & Mathematics with Applications, 58(5):965–974, 2009

    2009

  33. [41]

    Fiorina and S

    B. Fiorina and S. K. Lele. An artificial nonlinear diffu- sivity method for supersonic reacting flows with shocks. J. Comp. Phys. , 222(1):246–246, 2007

    2007

  34. [42]

    Comments on the Kolmogorov hypothe- sis of local isotropy in the smallscales

    James Brasseur. Comments on the Kolmogorov hypothe- sis of local isotropy in the smallscales . 1991

    1991

  35. [43]

    Random forcing of three-dimensional homogeneous turbulence

    Krister Alvelius. Random forcing of three-dimensional homogeneous turbulence. Physics of Fluids , 11(7):1880– 1889, 1999

    1999

  36. [44]

    Mathew, H

    J. Mathew, H. Foysi, and R. Friedrich. A new approach to LES based on explicit filtering. Int. J. Heat Fluid Flow, 27(4):594–602, 2006

    2006

  37. [45]

    P. Sagaut. Large Eddy Simulation for Incompressible Flows: An Introduction . Springer Verlag, latest edition, 2006

    2006

  38. [46]

    Lattice Boltzmann method with regularized pre-collision distribution func- tions

    Jonas Latt and Bastien Chopard. Lattice Boltzmann method with regularized pre-collision distribution func- tions. Math. Comput. Simul. , 72(2-6):165–168, 2006

    2006

  39. [47]

    Stolz and N.A

    S. Stolz and N.A. Adams. An approximate deconvolu- tion procedure for large-eddy simulation. Phys. Fluids , 11:1699–1701, 1999

    1999

  40. [48]

    Mathew, R

    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(8):2279–2289, 2003

    2003

  41. [49]

    High-order cfd methods: current status and perspective

    Zhijian J Wang, Krzysztof Fidkowski, R´ emi Abgrall, Francesco Bassi, Doru Caraeni, Andrew Cary, Herman Deconinck, Ralf Hartmann, Koen Hillewaert, Hung T Huynh, et al. High-order cfd methods: current status and perspective. International Journal for Numerical Meth- ods in Flui...

    2013

  42. [50]

    The fundamental bgk operator calculates the predicted profiles with con- siderable accuracy but lacks stability when the flow field becomes turbulent

    at a Reynolds number of 1600. The fundamental bgk operator calculates the predicted profiles with con- siderable accuracy but lacks stability when the flow field becomes turbulent. In contrast, the kbc operator [21] exhibits remarkable stability but fails to accurately cap- tu...

  43. [51]

    A comparison of second- and sixth-order methods for large-eddy simulations

    Matthias Meinke, Wolfgang Schr¨ oder, Egon Krause, and Th Rister. A comparison of second- and sixth-order methods for large-eddy simulations. Computers & Fluids, 31:695–718, 05 2002

    2002

  44. [52]

    Cambridge University Press, 2007

    Implicit Large Eddy Simulation: Computing Turbulent Fluid Dynamics . Cambridge University Press, 2007

    2007

  45. [53]

    Implicit Large-Eddy Simulation: Theory and Application , pages 743–750

    Nikolaus Adams and Stefan Hickel. Implicit Large-Eddy Simulation: Theory and Application , pages 743–750. 10 2009

    2009

  46. [54]

    Small-scale structure of the taylor–green vortex

    Marc E Brachet, Daniel I Meiron, Steven A Orszag, BG Nickel, Rudolf H Morf, and Uriel Frisch. Small-scale structure of the taylor–green vortex. Journal of Fluid Mechanics, 130:411–452, 1983

    1983

  47. [55]

    An immersed boundary method for complex incompressible flows

    Jung-Il Choi, Roshan C Oberoi, Jack R Edwards, and Jacky A Rosati. An immersed boundary method for complex incompressible flows. Journal of Computational Physics, 224(2):757–784, 2007

    2007

  48. [56]

    Numerical study and physical analysis of the pressure and velocity fields in the near wake of a circular cylinder

    M Braza, PHHM Chassaing, and H Ha Minh. Numerical study and physical analysis of the pressure and velocity fields in the near wake of a circular cylinder. Journal of fluid mechanics, 165:79–130, 1986

    1986

  49. [57]

    A cartesian grid method for modeling multiple moving objects in 2d incompress- ible viscous flow

    David Russell and Z Jane Wang. A cartesian grid method for modeling multiple moving objects in 2d incompress- ible viscous flow. Journal of Computational Physics , 17 191(1):177–205, 2003

    2003

  50. [58]

    A novel immersed boundary velocity correction–lattice boltz- mann method and its application to simulate flow past a circular cylinder

    Chang Shu, Ningyu Liu, and Yong-Tian Chew. A novel immersed boundary velocity correction–lattice boltz- mann method and its application to simulate flow past a circular cylinder. Journal of Computational Physics , 226(2):1607–1622, 2007

    2007

  51. [59]

    Preconditioned multigrid methods for unsteady incompressible flows

    C Liu, X Zheng, and CH Sung. Preconditioned multigrid methods for unsteady incompressible flows. Journal of Computational physics , 139(1):35–57, 1998

    1998

  52. [60]

    An edge-based method for the incompressible navier–stokes equations on polygonal meshes

    Jeffrey A Wright and Richard W Smith. An edge-based method for the incompressible navier–stokes equations on polygonal meshes. Journal of Computational Physics , 169(1):24–43, 2001

    2001

  53. [61]

    Characteristics of the plane turbulent near wake of a circular cylinder

    LM Lourenco. Characteristics of the plane turbulent near wake of a circular cylinder. A particle image velocimetry study, 1993

    1993

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.