A 3D Machine Learning based Volume Of Fluid scheme without explicit interface reconstruction

Bruno Despr\'es; Moreno Pintore

REVIEW 4 major objections 5 minor 39 references

A 3D Machine Learning based Volume Of Fluid scheme without explicit interface reconstruction

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A trained neural network can replace explicit interface reconstruction in a 3D volume-of-fluid scheme and still converge faster than standard schemes.

desk verdict Solid 3D extension of the ML-VOF idea with honest numerics, but the accuracy claim is only demonstrated on training-family interfaces and there is a small indexing typo in the positivity lemma. read the letter →

arxiv 2507.05218 v1 pith:ECCCLU7Z submitted 2025-07-07 math.NA cs.LGcs.NA

Moreno Pintore , Bruno Després This is my paper · ORCID

classification math.NAcs.LGcs.NA MSC 35Q3568T0776-1076M12

keywords VolumeofFluidMachineLearningCFD3Dinterfacereconstructionneuralnetworkfiniteschemeadvectionequation

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

This paper claims that the hardest step of 3D volume-of-fluid simulation, reconstructing the interface shape to compute fluxes, can be dropped: a fully connected neural network, trained once on synthetic interfaces, maps the local 3×3×3 stencil of volume fractions and a Courant number directly to the flux fraction across a face. The training data are generated purely geometrically from half-space intersections and ellipsoids, so no simulations or experiments feed the network. Embedded in a directional-splitting finite volume advection scheme, the network flux yields numerical convergence as the mesh is refined, with an empirical rate around 1.01 in a constant-velocity test versus 0.16 for upwind and 0.58 for limited downwind, and similar advantages in more general divergence-free velocity fields. This matters because 3D interface reconstruction, especially for irregular or multi-material interfaces, is the main implementation bottleneck of classical VOF schemes; the authors' scheme avoids it and needs no retraining when the equation, domain, or initial condition changes.

What carries the argument

The object carrying the argument is the neural-network flux map $\alpha_{NN}(x,\beta)$: a fully connected feed-forward ReLU network with input dimension $m^3+1$ (here 28, from a 3×3×3 stencil plus the Courant number) and scalar output, trained by minimizing mean squared error between its output and exact geometrical fluxes. Exact fluxes are computed with a convex-hull volume algorithm on synthetic regions defined as cube intersections with one, two, or three half-spaces and with ellipsoids, sampled by Latin Hypercube sampling and enriched by non-flux-preserving permutations. Around this map the paper places three mechanisms: averaging over the eight flux-preserving octahedral symmetries, a complementary-fluid averaging so that $\alpha_{NN}(x)+\alpha_{NN}(Mx)=1$, and a projection onto an admissibility interval that keeps volume fractions between 0 and 1. These mechanisms are what let a purely geometric training set be used as a conservative flux in a finite volume scheme.

What would settle it

Advect an initial condition whose interface is outside the training family, such as a thin cylindrical shell, a sharp wedge, or a three-material junction, on meshes with $N_h = 27, 54, 105$ and compare the final L1 error with limited downwind; if the VOF-ML error plateaus or its convergence rate drops below that of limited downwind, the transferability claim fails. A cheaper check is to evaluate the network on synthetically generated stencils from such shapes and measure whether the flux error remains competitive with the test-set error.

Watch

Extended reading notes

Core claim

The central claim is that the exact transported flux through a cell face can be treated as a regression problem: given the volume fractions in a local cubic stencil and one non-negative Courant number, the flux fraction is approximated by a ReLU network with 9151 trainable weights trained on 144000 synthetic input-output pairs. The paper shows, on advection problems in a periodic cube, that this network flux, when symmetrized over the eight flux-preserving cubic symmetries, made complementary under fluid exchange, and projected onto an admissible interval, produces a scheme whose L1 error decreases toward zero with mesh refinement at a higher empirical rate than the upwind and limited-downwind reference schemes. The same trained network is used across all numerical tests without retraining, and the scheme never reconstructs an interface during inference.

Load-bearing premise

The load-bearing assumption is that the exact flux is determined by a 3×3×3 stencil of volume fractions plus one Courant number, and that interfaces built from half-spaces and ellipsoids are representative enough that the network stays accurate on the curved and multi-scale interfaces that appear in real simulations.

Editorial extensions

If this is right

Interface reconstruction in 3D VOF can be replaced by one network evaluation, so the scheme avoids clipping and plane-positioning algorithms in its implementation.
A network trained once on synthetic planar and ellipsoidal interfaces transfers across equations, domains, and initial conditions without retraining.
On the constant-velocity test the scheme's empirical L1 convergence rate is about 1.01, versus 0.16 for upwind and 0.58 for limited downwind; on rotational and divergence-free shear tests it overtakes limited downwind once the mesh is fine enough.
Mixed-cell counts stay nearly constant when exact interface area should be constant, indicating that the learned flux controls numerical diffusion far better than upwind.
Flux reconstruction errors on the held-out test set are 43.7 times smaller in MSE than upwind and 13.2 times smaller than limited downwind, which matters because flux errors are injected at every time step.

Reading between the lines

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

A testable extension the authors did not run is an initial condition whose interface lies outside the training family, such as a thin cylindrical shell or a sharp wedge; if the L1 error plateaus or the convergence rate drops below limited downwind, the claimed transferability would be falsified.
Because the synthetic dataset is parameterized and cheap to generate, an adaptive strategy could resample stencils that are poorly predicted and retrain, giving the scheme an accuracy guarantee concentrated on the interfaces that actually appear in a run.
The symmetry-averaging trick could plausibly be extended from the eight flux-preserving symmetries to the full 48-element octahedral group, potentially reducing the number of training samples needed and improving accuracy on rotated inputs.
Since the network learns the flux rather than the interface, the same flux map may be reusable inside an unsplit finite volume update or for more than two fluids by componentwise combination; whether such reuse preserves the observed convergence rates is a direct, testable consequence.

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.

Desk Editor's Note

read the letter

Short version: this is a solid extension of the 2D VOF-ML idea to 3D, with clean symmetry handling and honest numerical evidence, but the generalization claim is only supported on the same geometric family used for training, and there's a small notational slip in the positivity lemma.

The genuinely new piece is the 3D flux network: input 3x3x3 volume-fraction stencil plus one Courant number, output the directional flux, no interface reconstruction. The octahedral-symmetry averaging (8 permutations that preserve the flux) and the dataset enrichment by the 24 flux-changing symmetries are well thought out and correctly formalized. The synthetic dataset generation (exact polytope volumes for half-space intersections, Fibonacci-discretized ellipsoids) is reproducible in principle, and the numerical tests on three advection problems show the VOF-ML scheme converging at roughly 1.0 while limited downwind sits around 0.5-0.6. That's a meaningful result for the VOF community.

Soft spots, in order of importance. First, the out-of-distribution question: the tests are all built from spheres and planar cuts, the same primitives as the training data. Nothing tests a genuinely different local topology — two disjoint interface pieces in one stencil, a subgrid thin shell, or repeated crossings. The paper itself claims the training family is 'sufficient' (Remark 2.1) but gives no evidence beyond the held-out split of the same generator. The central advantage could shrink on less friendly interfaces. Second, the error reporting: single deterministic runs, no seeds, no error bars. For a method that starts from a random initialization, that's a real gap. Third, there is a notational inconsistency in Lemma 3.3 and Remark 3.4: the flux bounds are written with the updated value αx_{i,j,k} where the proof needs the old value α_{i,j,k}. It's a typo, not a conceptual flaw, but it should be fixed. Fourth, no code or trained weights are released; given that the whole method is a black-box network, this hampers independent verification.

Overall, the paper is coherent and the math of the symmetry projection is correct. The reader's conditional verdict is about right. I'd engage with it, but I'd want the authors to either narrow the claims or add a genuine out-of-distribution test, and to release the network weights.

Referee Report

4 major / 5 minor

Summary. The paper presents a three-dimensional Volume-of-Fluid (VOF) scheme in which the numerical interface flux is computed directly by a feed-forward neural network from a 3x3x3 stencil of volume fractions and the local Courant number, without any explicit interface reconstruction. The network is trained offline on a synthetic dataset generated from intersections of half-spaces and ellipsoids, with no retraining during the simulations. The authors also propose symmetry-averaging and flux-projection post-processings to enforce physical constraints. They test the scheme on three 3D advection problems (constant, directionally conservative, and general divergence-free velocity fields) and report empirical convergence rates of about 1.01 for VOF-ML, compared with 0.58 for the limited downwind scheme and 0.16 for upwind.

Significance. The conceptual simplification of bypassing interface reconstruction is attractive and could genuinely simplify three-dimensional VOF implementations. The symmetry-averaging construction (Section 3.1.1) and the flux-bounding strategy (Section 3.2) are cleanly motivated and, apart from the index issue discussed below, the algebraic steps are correct. The synthetic dataset generation is reproducible, and the network is used in simulations without any fitting to the advection results, so the scheme is not circularly validated. If the convergence advantage is robust on broader test families, this would be a valuable contribution to the VOF literature. The main caveat is that the numerical evidence so far is confined to interfaces drawn from the same geometric family as the training data, which limits the generality of the central claim.

major comments (4)

[Section 3.2, Eq. (3.7) and Remark 3.4] The upper bounds in (3.7) are written as α_{i+1/2,j,k} ≤ (1/β^x) α^x_{i,j,k} and 1 − α_{i+1/2,j,k} ≤ (1/β^x)(1 − α^x_{i,j,k}), with α^x_{i,j,k} denoting the intermediate x-sweep value. However, the proof of Lemma 3.3 and the compatibility check in Remark 3.4 treat the bounds as involving the pre-update value α_{i,j,k}: the proof requires α_{i,j,k} − β^x α_{i+1/2,j,k} ≥ 0, and Remark 3.4 claims that the choice α_{i+1/2,j,k} = α_{i,j,k} satisfies the inequalities, which is not true for the α^x version in general. Since Definition 3.5 uses these bounds to project the network output, the maximum-principle argument as written is invalid. The authors should correct the index (likely replacing α^x_{i,j,k} with α_{i,j,k} in (3.7)) or revise the projection to be based on an explicitly computable quantity.
[Section 4, Tests 1-3] The initial conditions are all composed of spheres, boxes, and bars, which are the same geometric primitives used to generate the synthetic training data (half-spaces and ellipsoids). No test exercises a genuinely different local interface topology, such as two disconnected interface fragments in one stencil, a thin shell of sub-cell thickness, or an interface crossing a stencil multiple times. The claimed convergence advantage (rates 1.01 vs. 0.58) is therefore demonstrated only within the training distribution. To support the stronger claim that the network approximates the universal stencil-to-flux operator, the authors should include at least one out-of-distribution advection test, e.g., an initial interface generated with a different random process (Fourier modes, random radial functions) or a known challenging benchmark, and report the resulting L1 errors.
[Figures 8, 10, and 12] Each error curve in these figures is a single deterministic simulation per mesh, with no error bars or repeated trials. Since the network weights are initialized randomly, the reported rates (especially 1.01 for Test 1) could be sensitive to the specific training run. The authors should either fix and clearly state the random seed, or better, repeat the training several times (or evaluate with multiple network checkpoints) and show the spread of the L1 errors and of the fitted convergence rates.
[Section 2.2 and Remark 2.1] The mapping from stencil volume fractions to the exact flux is not single-valued: many different sub-cell interfaces can produce the same 3x3x3 vector of volume fractions but different fluxes. The synthetic dataset contains only one interface realization per input vector, so the network is trained to mimic a particular drawing from this conditional distribution rather than the true conditional expectation or worst-case flux. This could explain why accuracy degrades on interface topologies not represented in the training family. The authors should explicitly discuss this non-uniqueness and, ideally, train with multiple interface samples per stencil input (or add a variance penalty) to make the learned flux more robust.

minor comments (5)

[Abstract and Section 1, Eq. (1.1)] The equation is written as ∂tα + ∇·(αu) = ∂tα + u·∇αu = 0; the notation ∇αu is nonstandard and appears to be a typo for u·∇α.
[Section 3.1] 'octohedral' should be 'octahedral'.
[Section 4.1] There is a typo: 'a anti-clockwise rotations' should be 'an anti-clockwise rotation'.
[Tables 3-5] The ratio R_{mix}^{T/0} is reported without explaining how a 'mixed cell' is defined; consider adding a sentence defining the threshold used (ε_mark is defined in Appendix B, but a brief reminder in Section 4 would improve readability).
[Section 4.3] The claim that the normalization procedure 'preserves the mass conservation' is not immediately obvious for a general directional splitting, since normalizing per cell could alter the total mass of each phase; the authors should clarify why the additivity condition (Lemma 3.2) makes the normalization neutral in this respect.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation: the neural network is trained once on synthetic geometric fluxes and then evaluated on advection tests; the only mild concern is train/test shape overlap, which is a generalization issue, not circularity.

full rationale

The paper's derivation chain is non-circular. The VOF-ML flux is defined in Definition 2.3 as alpha_{i+1/2} = alpha_NN(x, beta_x), where alpha_NN is the network trained in (2.4) by minimizing the MSE loss (2.3) on the synthetic dataset D of (2.2), whose outputs are exact geometric fluxes computed from half-space/ellipsoid intersections. The network is trained before any simulation and is then used unchanged ('In all the numerical tests, we use the same neural network (without any retraining or fine-tuning)'). The reported advection errors compare the discrete solution to the exact transported indicator function at final time, so the convergence rates are not fitted outputs. Self-citations to [9,2] are methodological lineage: they motivate the VOF-ML idea and multi-component extension, but the 3D flux definition, dataset generation, symmetry post-processing (3.2)-(3.4), and numerical validations are developed in this manuscript and do not depend on unverified prior claims. The only plausible circularity-like concern is that the test initial conditions (sphere minus rectangular prism, sphere plus bars, sphere) are made of the same geometric primitives used to generate training data, so the tests evaluate in-distribution accuracy. This is a generalization/distribution-shift limitation, explicitly anticipated by the authors in Remark 2.1, not a reduction of the prediction to its training input: the test errors are computed against exact solutions, and no parameter is fit to those errors. Score 1 reflects the minor overlap concern and the presence of non-load-bearing self-citations, with no actual circular step.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The method introduces no new physical entities. Its load-bearing elements are fitted machine-learning parameters (the network weights), hand-chosen hyperparameters, and unproven domain assumptions about locality and training-data coverage.

free parameters (4)

Neural network weights (9151 trainable coefficients) = Trained on synthetic dataset; values not released
The flux approximation αNN is entirely defined by these weights, fitted to minimize MSE over the synthetic dataset. The numerical convergence results depend on this fit, and without the weights the exact model cannot be reproduced.
Architecture and training hyperparameters (L=5, width 50, ReLU, dataset size 144000, 5000 ADAM + 5000 BFGS steps) = Chosen by hand
These choices control the approximation error of the network and are not derived from any theory; they are part of the method design.
Hybridization threshold ε_mark = 0.01
Chosen by hand to define mixed cells where the neural network is used instead of the cheaper limited downwind scheme; affects the accuracy/cost tradeoff reported in Section 4.
Stencil margin m and ellipsoid discretization Npts_el = m=1, Npts_el=10000
The stencil size fixes the network input dimension (3^3+1) and the locality assumption; the ellipsoid vertex count controls the accuracy of the synthetic training data.

assumptions (5)

domain assumption The exact face flux is a function of the volume fractions in a local 3x3x3 stencil and the scalar directional Courant number.
Invoked in Definition 2.3 where α_{i+1/2}=αNN(x,β) with x from stencil S^m. The true flux depends on the sub-cell interface geometry, which is not determined by the volume fractions; this locality ansatz is the foundational modeling assumption of the scheme.
domain assumption The synthetic training family (intersections of 1-3 half-spaces and ellipsoids) spans the interfaces that appear in the numerical tests and in real applications.
Section 2.2 and Remark 2.1. The tests in Section 4 use only spheres and planar cuts, i.e., exactly this family, so generalization outside the family is asserted but not tested.
domain assumption A single Courant number per direction suffices to characterize the velocity in the flux evaluation, even when the velocity varies spatially within the stencil.
Definition 2.3 uses β_x = u_x Δt/Δx, a single scalar; Tests 2 and 3 use spatially varying velocity fields, so the flux model ignores velocity gradients within the stencil.
standard math QuickHull and the convex hull algorithm yield exact (or sufficiently accurate) polytope volumes for the training data.
Section 2.2 relies on [3]; for ellipsoids, the 10000-vertex Fibonacci-sphere approximation is used, and its error is not quantified.
standard math The octahedral symmetry group facts (order 48, subgroups S≠ and S=) are correct.
Used in Section 3.1 to justify dataset augmentation and the 8-fold symmetry averaging in Eq. (3.2).

how reviews work

0 comments

Cite this review

Pith. "Pith review of A 3D Machine Learning based Volume Of Fluid scheme without explicit interface reconstruction." pith.science (2026). https://pith.science/paper/ECCCLU7Z

@misc{pith2026250705218,
  author       = {Pith},
  title        = {Pith review of: A 3D Machine Learning based Volume Of Fluid scheme without explicit interface reconstruction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ECCCLU7Z}},
  note         = {Machine review of arXiv:2507.05218}
}

abstract

We present a machine-learning based Volume Of Fluid method to simulate multi-material flows on three-dimensional domains. One of the novelties of the method is that the flux fraction is computed by evaluating a previously trained neural network and without explicitly reconstructing any local interface approximating the exact one. The network is trained on a purely synthetic dataset generated by randomly sampling numerous local interfaces and which can be adapted to improve the scheme on less regular interfaces when needed. Several strategies to ensure the efficiency of the method and the satisfaction of physical constraints and properties are suggested and formalized. Numerical results on the advection equation are provided to show the performance of the method. We observe numerical convergence as the size of the mesh tends to zero $h=1/N_h\searrow 0$, with a better rate than two reference schemes.

Figures

Figures reproduced from arXiv: 2507.05218 by the authors.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 38 canonical work pages

[1]

Abadi et al

M. Abadi et al. , TensorFlow: Large-scale machine learning on heterogeneous systems , 2015. Software available from tensorflow.org

work page 2015
[2]

Ancellin, B

M. Ancellin, B. Despr ´es, and S. Jaouen , Extension of generic two-component vof interface advection schemes to an arbitrary number of components , Journal of Computational Physics, 473 (2023), p. 111721

work page 2023
[3]

C. B. Barber, D. P. Dobkin, and H. Huhdanpaa , The quickhull algorithm for convex hulls , ACM Transactions on Mathematical Software (TOMS), 22 (1996), pp. 469–483

work page 1996
[4]

Bradbury, R

J. Bradbury, R. Frostig, P. Hawkins, M. J. Johnson, C. Leary, D. Maclaurin, G. Necula, A. Paszke, J. V an- derPlas, S. W anderman-Milne, and Q. Zhang, JAX: composable transformations of Python+NumPy programs , http://github.com/google/jax, (2018)

work page 2018
[5]

Caboussat, M

A. Caboussat, M. Francois, R. Glowinski, D. Kothe, and J. Sicilian , A numerical method for interface reconstruction of triple points within a volume tracking algorithm , Mathematical and Computer Modelling, 48 (2008), pp. 1957–1971

work page 2008
[6]

Cahaly, F

A. Cahaly, F. Evrard, and O. Desjardins , Plic-net: A machine learning approach for 3d interface reconstruction in volume of fluid methods , International Journal of Multiphase Flow, 178 (2024), p. 104888

work page 2024
[7]

Cohen, O

A. Cohen, O. Mula, and A. Somacal , High order recovery of geometric interfaces from cell-average data , ESAIM: Mathematical Modelling and Numerical Analysis, 59 (2025), pp. 693–727

work page 2025
[8]

Delarue and F

F. Delarue and F. Lagouti `ere, Probabilistic analysis of the upwind scheme for transport equations , Archive for rational mechanics and analysis, 199 (2011), pp. 229–268

work page 2011

Show all 39 references

[9]

Despr´es and H

B. Despr´es and H. Jourdren , Machine learning design of volume of fluid schemes for compressible flows , Journal of Computational Physics, 408 (2020), p. 109275

2020
[10]

Despr ´es and F

B. Despr ´es and F. Lagouti `ere, Contact discontinuity capturing schemes for linear advection and compressible gas dynamics, Journal of Scientific Computing, 16 (2001), pp. 479–524

2001
[11]

S. R. Dubey, S. K. Singh, and B. B. Chaudhuri , Activation functions in deep learning: A comprehensive survey and benchmark, Neurocomputing, 503 (2022), pp. 92–108

2022
[12]

Dyadechko and M

V. Dyadechko and M. Shashkov, Reconstruction of multi-material interfaces from moment data, Journal of Computational Physics, 227 (2008), pp. 5361–5384

2008
[13]

F akhreddine, K

A. F akhreddine, K. Alam´e, and K. Mahesh , Directionally-split volume-of-fluid technique for front propagation under curvature flow, International Journal for Numerical Methods in Fluids, 96 (2024), pp. 1517–1554

2024
[14]

Garanger, J

K. Garanger, J. Kraus, and J. J. Rimoli , Symmetry-enforcing neural networks with applications to constitutive modeling , Extreme Mechanics Letters, 71 (2024), p. 102188

2024
[15]

Goodfellow, Y

I. Goodfellow, Y. Bengio, and A. Courville , Deep Learning, MIT Press, 2016. http://www.deeplearningbook.org

2016
[16]

A. Han, R. Chiodi, and O. Desjardins , Capturing thin structures in vof simulations with two-plane reconstruction , Journal of Computational Physics, 519 (2024), p. 113453

2024
[17]

C. W. Hirt and B. D. Nichols, Volume of fluid (VOF) method for the dynamics of free boundaries, Journal of computational physics, 39 (1981), pp. 201–225

1981
[18]

Kawano, A simple volume-of-fluid reconstruction method for three-dimensional two-phase flows , Computers & Fluids, 134-135 (2016), pp

A. Kawano, A simple volume-of-fluid reconstruction method for three-dimensional two-phase flows , Computers & Fluids, 134-135 (2016), pp. 130–145

2016
[19]

Keinert, M

B. Keinert, M. Innmann, M. S ¨anger, and M. Stamminger , Spherical fibonacci mapping, ACM Transactions on Graphics (TOG), 34 (2015), pp. 1–7

2015
[20]

D. P. Kingma and J. Ba , Adam: a method for stochastic optimization , arXiv preprint arXiv:1412.6980, (2014)

2014 arXiv
[21]

Kromer, J

J. Kromer, J. Potyka, K. Schulte, and D. Bothe, Efficient sequential plic interface positioning for enhanced performance of the three-phase vof method , Computers & Fluids, 266 (2023), p. 106051

2023
[22]

R. J. LeVeque , Finite volume methods for hyperbolic problems , vol. 31, Cambridge university press, 2002

2002
[23]

J. S. Lomont , Applications of finite groups , Academic Press, 2014

2014
[24]

Mak and M

A. Mak and M. Raessi , A machine learning approach to volume tracking in multiphase flow simulations , Fluids, 10 (2025)

2025
[25]

W. Meng, C. hao Yu, J. Li, Z. Wu, and R. An , Modeling of interfacial two-phase flows via vof-based ls method with weno scheme in the finite volume method , Computers & Fluids, 291 (2025), p. 106580

2025
[26]

B. K. Miller, M. Geiger, T. E. Smidt, and F. No’e , Relevance of rotationally equivariant convolutions for predicting molecular properties, ArXiv, abs/2008.08461 (2020)

2020 arXiv
[27]

Nakano, M

T. Nakano, M. A. Bucci, J.-M. Gratien, and T. F aney, Machine learning model for gas-liquid interface reconstruction in cfd numerical simulations , Fluids, 10 (2025)

2025
[28]

Paszke et al., Pytorch: An imperative style, high-performance deep learning library , in Advances in Neural Information Processing Systems 32, Curran Associates, Inc., 2019, pp

A. Paszke et al., Pytorch: An imperative style, high-performance deep learning library , in Advances in Neural Information Processing Systems 32, Curran Associates, Inc., 2019, pp. 8024–8035

2019
[29]

Pathak and M

A. Pathak and M. Raessi, A three-dimensional volume-of-fluid method for reconstructing and advecting three-material interfaces forming contact lines , Journal of Computational Physics, 307 (2016), pp. 550–573

2016
[30]

J. E. Pilliod and E. G. Puckett , Second-order accurate volume-of-fluid algorithms for tracking material interfaces , Journal of Computational Physics, 199 (2004), pp. 465–502. 23

2004
[31]

Pitman and N

J. Pitman and N. Ross , Archimedes, Gauss, and Stein , Notices AMS, 59 (2012), pp. 1416–1421

2012
[32]

Y. Qi, J. Lu, R. Scardovelli, S. Zaleski, and G. Tryggvason , Computing curvature for volume of fluid methods using machine learning, Journal of Computational Physics, 377 (2019), pp. 155–161

2019
[33]

M. A. Rushdi, S. Yoshida, C. Hu, T. N. Dief, A. E. Salem, and M. M. K. and , Efficient and accurate machine learning models for volume of fluid (vof ) simulation on uniform cartesian mesh , Engineering Applications of Computational Fluid Mechanics, 19 (2025), p. 2451774

2025
[34]

Tavares, C

M. Tavares, C. Josserand, A. Limare, J. M. Lopez-Herrera, and S. Popinet , A coupled vof/embedded boundary method to model two-phase flows on arbitrary solid surfaces , Computers & Fluids, 278 (2024), p. 106317

2024
[35]

Weymouth and D

G. Weymouth and D. K.-P. Yue , Conservative volume-of-fluid method for free-surface simulations on cartesian-grids , Journal of Computational Physics, 229 (2010), pp. 2853–2865

2010
[36]

Wright, J

S. Wright, J. Nocedal, et al. , Numerical Optimization, vol. 35, Springer, 1999

1999
[37]

D. L. Youngs , Time-dependent multi-material flow with large fluid distortion , Numerical methods for fluid dynamics, (1982)

1982
[38]

J. F. Z. Kraus, F. K ¨ohler and M. Sch ¨afer, A deep learning approach for curvature prediction in algebraic volume of fluid methodsg, International Conference on Computational Methods for Coupled Problems in Science and Engineering, 9 (2021)

2021
[39]

C. Zhao, J. Maarek, S. M. Taleghani, and S. Zaleski , A hybrid continuum surface tension force for the three-phase vof method, Journal of Computational Physics, 504 (2024), p. 112872. Appendix A. Parametrization of the training configurations. The considered training configura...

2024

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

Abadi et al

M. Abadi et al. , TensorFlow: Large-scale machine learning on heterogeneous systems , 2015. Software available from tensorflow.org

work page 2015

[2] [2]

Ancellin, B

M. Ancellin, B. Despr ´es, and S. Jaouen , Extension of generic two-component vof interface advection schemes to an arbitrary number of components , Journal of Computational Physics, 473 (2023), p. 111721

work page 2023

[3] [3]

C. B. Barber, D. P. Dobkin, and H. Huhdanpaa , The quickhull algorithm for convex hulls , ACM Transactions on Mathematical Software (TOMS), 22 (1996), pp. 469–483

work page 1996

[4] [4]

Bradbury, R

J. Bradbury, R. Frostig, P. Hawkins, M. J. Johnson, C. Leary, D. Maclaurin, G. Necula, A. Paszke, J. V an- derPlas, S. W anderman-Milne, and Q. Zhang, JAX: composable transformations of Python+NumPy programs , http://github.com/google/jax, (2018)

work page 2018

[5] [5]

Caboussat, M

A. Caboussat, M. Francois, R. Glowinski, D. Kothe, and J. Sicilian , A numerical method for interface reconstruction of triple points within a volume tracking algorithm , Mathematical and Computer Modelling, 48 (2008), pp. 1957–1971

work page 2008

[6] [6]

Cahaly, F

A. Cahaly, F. Evrard, and O. Desjardins , Plic-net: A machine learning approach for 3d interface reconstruction in volume of fluid methods , International Journal of Multiphase Flow, 178 (2024), p. 104888

work page 2024

[7] [7]

Cohen, O

A. Cohen, O. Mula, and A. Somacal , High order recovery of geometric interfaces from cell-average data , ESAIM: Mathematical Modelling and Numerical Analysis, 59 (2025), pp. 693–727

work page 2025

[8] [8]

Delarue and F

F. Delarue and F. Lagouti `ere, Probabilistic analysis of the upwind scheme for transport equations , Archive for rational mechanics and analysis, 199 (2011), pp. 229–268

work page 2011