Pith. sign in

REVIEW 3 major objections 5 minor 72 references

Physics-Informed Neural Networks for microflows: Rarefied Gas Dynamics in Cylinder Arrays

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read One physics-informed neural network trained on sparse DSMC data reproduces rarefied microflow fields in a cylinder array with under 2% residual error.

desk verdict Useful first demonstration of PINNs for rarefied flow in a non-convex geometry, but the noise-filtering and accuracy claims are under-validated because all comparisons are against the same noisy DSMC data. read the letter →

arxiv 2501.13108 v1 pith:FUFSGXDP submitted 2025-01-07 physics.comp-ph physics.flu-dyn

Jean-Michel Tucny , Marco Lauricella , Mihir Durve , Gianmarco Guglielmo , Andrea Montessori , Sauro Succi This is my paper · ORCID
classification physics.comp-phphysics.flu-dyn PACS 47.45.-n47.11.-j
keywords physics-informedneuralnetworksrarefiedgasdynamicsdirectsimulationMonteCarloKnudsennumbercylinderarraydeviatoricstressmicroflowsperiodicboundaryconditions
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 aims to show that a single physics-informed neural network can replace many direct-simulation Monte Carlo runs for rarefied gas microflows in a non-convex periodic cylinder array. Across Knudsen numbers from 0.1 to 3, the network trained on a limited set of DSMC-generated fields keeps continuity and momentum residuals below 2 percent, produces smooth velocity, pressure, and deviatoric stress fields, and filters DSMC's statistical noise. It also predicts the withheld Knudsen number 0.7 flow well, while extrapolation to Knudsen number 5 remains physically consistent but locally overshoots by about 20 percent. If correct, this makes parametric studies of MEMS, air filtration, and shale-gas flows much cheaper, since each DSMC field costs about 20 hours on four GPUs while the PINN trains in under two hours on one GPU and evaluates in under two seconds.

What carries the argument

The central object is a feed-forward physics-informed neural network with a truncated Fourier feature layer that enforces periodic boundary conditions exactly, mapping (x, y, ln Kn) to the six output fields. The loss function combines data-matching terms on the DSMC fields with normalized residuals of the continuity equation and the Cauchy momentum equation, with extra residual-evaluation points concentrated near the cylinder surface where Knudsen layer and S-layer effects are strongest. Z-score normalization, with pressure centered per Knudsen number, keeps the different variables and loss components comparable in magnitude.

What would settle it

Run a deterministic kinetic solver, such as the discrete velocity method or the discrete unified gas kinetic scheme, at Knudsen number 0.7 in the same periodic cylinder array and compare its velocity and deviatoric stress fields with the PINN prediction; if local differences exceed the DSMC statistical noise, the claim that the PINN captures the Knudsen-layer physics would be wrong.

Watch

Extended reading notes

Core claim

The central discovery is that encoding only the macroscopic conservation laws, the incompressible continuity equation and the Cauchy momentum equation, as residuals in the loss is enough for a PINN to approximate the six macroscopic fields of a steady, laminar, weakly compressible rarefied gas in a cylinder array, provided the network is trained on DSMC data covering the transition regime. Using the Knudsen number as a third input, the network interpolates between the simulated Knudsen numbers and filters the statistical noise that dominates second-order moments at low Knudsen number. Extrapolation to Knudsen number 5 remains physically consistent but locally overshoots by about 20 percent.

Load-bearing premise

The claim stands on the assumption that the time-averaged DSMC fields, especially the deviatoric stress tensor, are accurate and unbiased representations of the true rarefied flow in the non-convex geometry; if the DSMC noise is biased or under-resolved near the cylinder's S-layer, the PINN inherits that error.

Editorial extensions

If this is right

  • Under 2 percent continuity and momentum residuals mean the PINN satisfies the macroscopic conservation laws almost as well as a converged solver across the whole training range.
  • Interpolation in the Knudsen number works: the withheld Knudsen number 0.7 field is reproduced well, so the approach can bridge gaps between discrete DSMC simulations.
  • The smoothness of the PINN outputs turns it into a denoiser for DSMC statistical fluctuations, especially for pressure and deviatoric stress components.
  • The trained network yields local effective-viscosity maps that deviate from Newtonian behavior, indicating that a single effective viscosity cannot capture rarefied flow in the array.
  • The same setup extends naturally to three dimensions with ten outputs, and the computational cost stays orders of magnitude below DSMC.

Reading between the lines

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

  • A natural next step would be to use the trained network as a surrogate for querying local quantities, such as the S-layer stress discontinuity, at any Knudsen number inside the training range without running new DSMC simulations.
  • Because the physics loss only constrains divergence and momentum balance, the reliability of the stress outputs is ultimately limited by the DSMC data; comparing the trained PINN against a deterministic kinetic solver would separate data noise from network approximation error.
  • The hard periodic-boundary Fourier layer is reusable for other periodic porous-media geometries, but transferring to non-periodic or three-dimensional domains will require new boundary-condition treatments.
  • One could test whether the network's denoising ability allows accurate training on shorter DSMC averaging times, potentially cutting the roughly 20 hours per field data generation cost further.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper presents a physics-informed neural network (PINN) that approximates the steady, weakly compressible rarefied gas flow through a periodic cylinder array as a function of the Knudsen number (Kn). The network is trained on DSMC (SPARTA) data for Kn in 0.1–3, with Kn=0.7 and Kn=5 held out, and is regularized by residual terms for the continuity and Cauchy momentum equations. The authors claim less than 2% error on these residuals, effective filtering of DSMC statistical noise, good interpolation at Kn=0.7, and limited extrapolation at Kn=5 with a local overshoot of about 20%, while maintaining physical consistency and large computational savings.

Significance. If the central claims are adequately supported, this would be a useful demonstration of PINNs for rarefied gas flows in a non-convex geometry, an extension beyond prior convex-domain studies. The hard periodic boundary condition via a Fourier feature layer, the systematic treatment of data normalization, and the honest reporting of extrapolation failure are strengths. The paper also quantifies the computational cost advantage over DSMC. However, the quantitative validation currently falls short of supporting the accuracy and denoising claims, so the contribution is more incremental than definitive at this stage.

major comments (3)
  1. [Section 5b, Eq. (23), Fig. 12] The abstract and conclusion claim 'less than 2% error,' but this refers to the continuity and momentum residual losses (Eqs. 19–21), which are optimized during training with explicit weights (Table 1), not to an independent accuracy measure. No numerical value of the global mean distance E̅ for the held-out Kn=0.7 case is reported, and Fig. 12 lacks quantitative axis labels. The central claim that the PINN 'accurately approximates' the true flow fields therefore needs a direct error metric against held-out DSMC data, with the residual-loss distinction made explicit.
  2. [Figs. 7 and 10 captions; Section 4b, Eqs. (16)–(17)] The captions of Figs. 7 and 10 state that normalization was carried out on the whole dataset, which includes the held-out Kn=0.7 and Kn=5 fields. Since the Z-score statistics μ and σ in Eq. (16) and the pressure normalization in Eq. (17) are computed from the entire dataset, information from the test fields influences the preprocessed inputs and outputs of the training data. This compromises the independence of the held-out evaluation; normalization statistics should be computed from the training set only, or the sensitivity to this leakage should be quantified.
  3. [Section 5a, Table 2; Section 5b and Fig. 8] The DSMC reference fields carry large statistical noise, especially for second-order moments at low Kn (Table 2 reports momentum residual standard deviations of about 19 at Kn=0.1), and the PINN predictions are compared exclusively against these same noisy time-averages used for training. Low residual losses and smooth outputs therefore do not establish that the PINN recovers the true kinetic solution; the 'effective filtering of DSMC intrinsic statistical noise' claim requires validation against a higher-fidelity reference (e.g., longer DSMC sampling or an independent kinetic solver) at least at one Knudsen number. As written, the observed smoothness could equally result from over-smoothing real physical structure near the cylinder such as the S-layer discussed in Sections 1 and 5b.
minor comments (5)
  1. [Eq. (30)] In the definition of Ψ_xy, the numerator uses |τ_yy| rather than |τ_xy|; this appears to be a typo and should be corrected.
  2. [Eq. (20)] Inside the norm, the terms involving the stress divergence and the convective term appear to be missing a '+' operator between them; please check the formula for clarity.
  3. [Fig. 12] The y-axis of Fig. 12 has no numerical scale; please provide the actual values of E̅, in particular for Kn=0.7 and Kn=5, so readers can judge the magnitude of the interpolation error.
  4. [Section 3, paragraph on DSMC parameters] The sentence 'To represent, the gas flow, 6.6 million particles were simulated, which at least 100 particles in every cell' is grammatically incomplete; rephrase to state that at least 100 particles per cell are ensured.
  5. [Section 6, Conclusion] The conclusion repeats 'less than 2% error on the loss function'; please clarify in both the abstract and conclusion that this is a PDE residual error, not a prediction error against independent data.

Circularity Check

1 steps flagged · score 3.0 of 10

Headline residual-error claim is the training objective itself; the derivation otherwise remains self-contained.

  1. fitted input called prediction [Abstract; Section 4a with Table 1; Section 5b]
    "The PINN achieved under 2 percent error on these residuals and effectively filtered DSMC intrinsic statistical noise."

    Equations (19)-(21) define the continuity and momentum residuals that are minimized during training, with weights w_cont = w_mom = 0.1 (Table 1). Section 5b further states that "from the training onset, the continuity and Cauchy momentum residuals very quickly become lower than the ones computed from the second-order difference scheme on the DSMC fields." The reported "under 2 percent error" is therefore the value of the training objective itself, and the comparison baseline is finite-difference noise derived from the same noisy DSMC data used for training. The residual magnitude is minimized by construction, so it is not an independent accuracy or denoising test against the true kinetic solution.

full rationale

Most of the derivation is self-contained: the continuity and Cauchy momentum residuals (Eqs. 7-8, 19-21) are independent governing equations, not derived from the DSMC data; the Kn=0.7 field is genuinely withheld from the weight updates; and no load-bearing argument rests on a self-citation or an imported uniqueness theorem. The main circular element is the headline "under 2 percent error on these residuals": those residuals are exactly the loss terms being minimized (Table 1 sets their weights to 0.1), so their low value is a property of the training objective, not an independent accuracy certificate. The noise-filtering claim is also validated only against the same time-averaged SPARTA fields used for training, so smoothness and low residuals do not by themselves establish fidelity to the unknown kinetic solution; the paper's own caption that "Normalization was carried out on the whole dataset" further means the Kn=0.7 held-out evaluation is not fully out-of-sample. These are validation weaknesses rather than a collapse of the derivation, so the score is moderate.

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

The core of the paper is a data-driven surrogate; the only free parameters are hyperparameters and the a posteriori acceleration constant. The key axioms are the accuracy of the DSMC ground truth and the adequacy of a smooth network to represent flow fields that have known kinetic discontinuities (S-layer) in a non-convex domain.

free parameters (5)
  • Acceleration scaling constant 0.00766 = 0.00766
    Chosen a posteriori in Eq. (13) to keep the flow laminar (Re<1) and weakly compressible (Ma<0.1) across the Kn range; it sets the volumetric force for all DSMC runs.
  • PDE loss weights w_cont and w_mom = 0.1
    Manual hyperparameters in Table 1 balancing data and physics loss terms; not optimized or justified beyond preliminary studies.
  • Number of Fourier modes n = 3
    Truncated Fourier layer for hard periodic BCs; the sufficiency of n=3 for all Kn is not tested.
  • Weight decay w_reg = 0.001
    Regularization hyperparameter chosen by hand (Table 1).
  • Neural network size = 10 layers x 100 neurons
    Architecture choice from Table 1, not systematically searched (paper acknowledges lack of hyperparameter search).
assumptions (4)
  • domain assumption DSMC with the variable hard sphere model and the specified grid/particle count provides an unbiased approximation to the Boltzmann equation for these flows.
    Section 3 argues criteria meet or exceed established standards, but no convergence study vs. particle count or grid is shown; the PINN inherits any DSMC bias.
  • domain assumption The flow is steady, laminar, weakly compressible, and isothermal, allowing the simplified continuity and Cauchy momentum equations (7)-(8).
    Section 2 states the assumptions; Re~0.13 and Ma~0.072 are verified a posteriori, but isothermality is only asserted.
  • ad hoc to paper A single smooth neural network with tanh activation and 3 Fourier modes can represent the macroscopic fields, including the stress tensor, across the full Kn range.
    This is the representational assumption of the architecture; no universal approximation guarantee for the specific normalization and the S-layer discontinuity is provided.
  • ad hoc to paper Z-score normalization statistics computed over the dataset (which appears to include the held-out Kn=0.7 and 5 fields) do not compromise the testing procedure.
    Section 4b says means and stds are computed on the data set; if test fields are included, the held-out evaluation uses information from those fields, which is a mild leakage.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Physics-Informed Neural Networks for microflows: Rarefied Gas Dynamics in Cylinder Arrays." pith.science (2026). https://pith.science/paper/FUFSGXDP

@misc{pith2026250113108,
  author       = {Pith},
  title        = {Pith review of: Physics-Informed Neural Networks for microflows: Rarefied Gas Dynamics in Cylinder Arrays},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FUFSGXDP}},
  note         = {Machine review of arXiv:2501.13108}
}
read the original abstract

Accurate prediction of rarefied gas dynamics is crucial for optimizing flows through microelectromechanical systems, air filtration devices, and shale gas extraction. Traditional methods, such as discrete velocity and direct simulation Monte Carlo (DSMC), demand intensive memory and computation, especially for microflows in non-convex domains. Recently, physics-informed neural networks emerged as a meshless and adaptable alternative for solving non-linear partial differential equations. We trained a PINN using a limited number of DSMC-generated rarefied gas microflows in the transition regime with Knudsen number from 0.1 to 3, incorporating continuity and Cauchy momentum exchange equations in the loss function. The PINN achieved under 2 percent error on these residuals and effectively filtered DSMC intrinsic statistical noise. Predictions remained strong for a tested flow field with Kn equal to 0.7, and showed limited extrapolation performance on a flow field with Kn equal to 5 with a local overshoot of about 20 percent, while maintaining physical consistency. Notably, each DSMC field required about 20 hours on 4 graphics processing units (GPU), while the PINN training took less than 2 hours on one GPU, with evaluations under 2 seconds.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

72 extracted references · 69 canonical work pages

  1. [1]

    Molecular momentum transport at fluid-solid interfaces in MEMS/NEMS: a review,

    B.-Y . Cao, J. Sun, M. Chen, and Z.-Y . Guo, "Molecular momentum transport at fluid-solid interfaces in MEMS/NEMS: a review," International journal of molecular sciences, vol. 10, pp. 4638-4706, 2009

    work page 2009

  2. [2]

    A review of rarefied gas flow in irregular micro/nanochannels,

    A. Taassob, A. Bordbar, S. Kheirandish, A. Zarnaghsh, R. Kamali, and A. S. Rana, "A review of rarefied gas flow in irregular micro/nanochannels," Journal of Micromechanics and Microengineering, vol. 31, p. 113002, 2021/10/08 2021

    work page 2021

  3. [3]

    Realistic air filter media performance simulation. Part II: Beyond finite -volume computational fluid dynamics procedures,

    B. Zhou, P . Tronville, and R. Rivers, "Realistic air filter media performance simulation. Part II: Beyond finite -volume computational fluid dynamics procedures," HVAC&R Research, vol. 19, pp. 503-512, 2013/07/04 2013

    work page 2013

  4. [4]

    Critical Review of Fluid Flow Physics at Micro - to Nano-scale Porous Media Applications in the Energy Sector,

    H. Singh and R. S. Myong, "Critical Review of Fluid Flow Physics at Micro - to Nano-scale Porous Media Applications in the Energy Sector," Advances in Materials Science and Engineering, vol. 2018, p. 31, 2018

    work page 2018

  5. [5]

    G. M. Kremer, An Introduction to the Boltzmann Equation and Transport Processes in Gases: Springer Berlin Heidelberg, 2010

    work page 2010

  6. [6]

    Cercignani, The Boltzmann Equation and Its Applications: Springer New York, 1987

    C. Cercignani, The Boltzmann Equation and Its Applications: Springer New York, 1987

    work page 1987

  7. [7]

    Succi, The Lattice Boltzmann Equation: For Complex States of Flowing Matter : Oxford University Press, 2018

    S. Succi, The Lattice Boltzmann Equation: For Complex States of Flowing Matter : Oxford University Press, 2018

    work page 2018

  8. [8]

    Devienne, Frottement et Échanges Thermiques dans les Gaz Raréfiés

    M. Devienne, Frottement et Échanges Thermiques dans les Gaz Raréfiés. Paris: Gauthier- Villars, 1958

    work page 1958

Show all 72 references
  1. [9]

    Discontinuity of the velocity distribution function in a rarefied gas around a convex body and the S layer at the bottom of the Knudsen layer,

    Y . Sone and S. Takata, "Discontinuity of the velocity distribution function in a rarefied gas around a convex body and the S layer at the bottom of the Knudsen layer," Transport Theory and Statistical Physics, vol. 21, pp. 501-530, 1992/08/01 1992

    1992

  2. [10]

    Rarefied gas flow over an in-line array of circular cylinders,

    S. Taguchi and P . Charrier, "Rarefied gas flow over an in-line array of circular cylinders," Physics of Fluids, vol. 20, p. 067103, 2008

    2008

  3. [11]

    Numerical study of the generalized cylindrical Couette flow of rarefied gas,

    E. M. Shakhov and V. A. Titarev, "Numerical study of the generalized cylindrical Couette flow of rarefied gas," European Journal of Mechanics - B/Fluids, vol. 28, pp. 152 -169, 2009/01/01/ 2009

    2009

  4. [12]

    Gatignol, Théorie cinétique des gaz à répartition discrète de vitesses : Springer-Verlag, 1975

    R. Gatignol, Théorie cinétique des gaz à répartition discrète de vitesses : Springer-Verlag, 1975

    1975

  5. [13]

    Efficient prediction of gas permeability by hybrid DSBGK-LBM simulations,

    J. Li, "Efficient prediction of gas permeability by hybrid DSBGK-LBM simulations," Fuel, vol. 250, pp. 154-159, 2019. 24

    2019

  6. [14]

    Pore-scale gas flow simulations by the DSBGK and DVM methods,

    J. Li, M. T. Ho, M. K. Borg, C. Cai, Z.-H. Li, and Y . Zhang, "Pore-scale gas flow simulations by the DSBGK and DVM methods," Computers & Fluids, vol. 226, p. 105017, 2021/08/15/ 2021

    2021

  7. [15]

    Subsonic Rarefied Gas Flow over a Rectangular Cylinder,

    O. I. Rovenskaya, "Subsonic Rarefied Gas Flow over a Rectangular Cylinder," Computational Mathematics and Mathematical Physics, vol. 62, pp. 1928 -1941, 2022/11/01 2022

    1928

  8. [16]

    Discrete unified gas kinetic scheme for all Knudsen number flows: Low-speed isothermal case,

    Z. Guo, K. Xu, and R. Wang, "Discrete unified gas kinetic scheme for all Knudsen number flows: Low-speed isothermal case," Physical Review E, vol. 88, p. 033305, 09/27/ 2013

    2013

  9. [17]

    Discrete unified gas kinetic scheme on unstructured meshes,

    L. Zhu, Z. Guo, and K. Xu, "Discrete unified gas kinetic scheme on unstructured meshes," Computers & Fluids, vol. 127, pp. 211-225, 3/20/ 2016

    2016

  10. [18]

    A comparative study of discrete velocity methods for low -speed rarefied gas flows,

    P . Wang, M. T. Ho, L. Wu, Z. Guo, and Y . Zhang, "A comparative study of discrete velocity methods for low -speed rarefied gas flows," Computers & Fluids, vol. 161, pp. 33 -46, 2018/01/15/ 2018

    2018

  11. [19]

    Progress of discrete unified gas -kinetic scheme for multiscale flows,

    Z. Guo and K. Xu, "Progress of discrete unified gas -kinetic scheme for multiscale flows," Advances in Aerodynamics, vol. 3, p. 6, 2021/02/06 2021

    2021

  12. [20]

    Discrete unified gas kinetic scheme simulation of microflows with complex geometries in Cartesian grid,

    Q. He, S. Tao, X. Yang, W. Lu, and Z. He, "Discrete unified gas kinetic scheme simulation of microflows with complex geometries in Cartesian grid," Physics of Fluids, vol. 33, p. 042005, 2021

    2021

  13. [21]

    G. A. Bird. (1994). Molecular Gas Dynamics and the Direct Simulation of Gas Flows . Available: https://app.knovel.com/hotlink/toc/id:kpMGDDSGF3/molecular-gas- dynamics/molecular-gas-dynamics

    1994

  14. [22]

    An open source, parallel DSMC code for rarefied gas flows in arbitrary geometries,

    T. J. Scanlon, E. Roohi, C. White, M. Darbandi, and J. M. Reese, "An open source, parallel DSMC code for rarefied gas flows in arbitrary geometries," Computers & Fluids, vol. 39, pp. 2078-2089, 2010/12/01/ 2010

    2010

  15. [23]

    Rarefied gas effect in hypersonic shear flows,

    J. Chen and H. Zhou, "Rarefied gas effect in hypersonic shear flows," Acta Mechanica Sinica, vol. 37, pp. 2-17, 2021/01/01 2021

    2021

  16. [24]

    Monte Carlo analysis of macroscopic fluctuations in a rarefied hypersonic flow around a cylinder,

    S. K. Stefanov, I. D. Boyd, and C.-P . Cai, "Monte Carlo analysis of macroscopic fluctuations in a rarefied hypersonic flow around a cylinder," Physics of Fluids, vol. 12, pp. 1226-1239, 2000

    2000

  17. [25]

    DSMC Analysis of Rarefied Gas Flow Over a Rectangular Cylinder at All Knudsen Numbers,

    C.-H. Cheng and F. -L. Liao, "DSMC Analysis of Rarefied Gas Flow Over a Rectangular Cylinder at All Knudsen Numbers," Journal of Fluids Engineering, vol. 122, pp. 720 -729, 2000

    2000

  18. [26]

    A. A. Abramov, A. V. Butkovskii, and O. G. Buzykin, "Rarefied cylindrical Couette problem: Effect of a very sharp transition from weakly rarefied flow near the inner cylinder to free molecular flow at very low average Knudsen numbers," Physics of Fluids, vol. 36, 2024

    2024

  19. [27]

    Study of Rarefied Flow Around Rectangular Cylinder Using DSMC,

    P . S. Vignesh Ram, H. D. Kim, and M. Yaga, "Study of Rarefied Flow Around Rectangular Cylinder Using DSMC," in 31st International Symposium on Shock Waves 1 , Cham, 2019, pp. 937-943

    2019

  20. [28]

    High -Speed Rarefied Flow Past a Rotating Cylinder: The Inverse Magnus Effect,

    B. John, X. -J. Gu, R. W. Barber, and D. R. Emerson, "High -Speed Rarefied Flow Past a Rotating Cylinder: The Inverse Magnus Effect," AIAA Journal, vol. 54, pp. 1670-1681, 2016

    2016

  21. [29]

    Aerodynamics of a Spinning Cylinder in Rarefied Gas Flows,

    V. V. Riabov, "Aerodynamics of a Spinning Cylinder in Rarefied Gas Flows," Journal of Spacecraft and Rockets, vol. 36, pp. 486-488, 1999

    1999

  22. [30]

    The effect of Knudsen layers on rarefied cylindrical Couette gas flows,

    N. Dongari, R. W. Barber, D. R. Emerson, S. K. Stefanov, Y . H. Zhang, and J. M. Reese, "The effect of Knudsen layers on rarefied cylindrical Couette gas flows," Microfluidics and Nanofluidics, vol. 14, pp. 31-43, Jan 2013

    2013

  23. [31]

    Continuum and kinetic simulations of heat transfer trough rarefied gas in annular and planar geometries in the slip regime,

    M. Hadj-Nacer, D. Maharjan, M.-T. Ho, S. K. Stefanov, I. Graur, and M. Greiner, "Continuum and kinetic simulations of heat transfer trough rarefied gas in annular and planar geometries in the slip regime," Journal of Heat Transfer, vol. 139, p. 042002, 2017. 25

    2017

  24. [32]

    A two -dimensional computational study of gas flow regimes past of square cylinder confined in a long microchannel,

    K. S. Shterev and S. K. Stefanov, "A two -dimensional computational study of gas flow regimes past of square cylinder confined in a long microchannel," European Journal of Mechanics - B/Fluids, vol. 64, pp. 47-54, 2017/07/01/ 2017

    2017

  25. [33]

    Hidden physics models: Machine learning of nonlinear partial differential equations,

    M. Raissi and G. E. Karniadakis, "Hidden physics models: Machine learning of nonlinear partial differential equations," Journal of Computational Physics, vol. 357, pp. 125 -141, 2018/03/15/ 2018

    2018

  26. [34]

    Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,

    M. Raissi, P . Perdikaris, and G. E. Karniadakis, "Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations," Journal of Computational Physics, vol. 378, pp. 686 -707, 2019/02/01/ 2019

    2019

  27. [35]

    Physics - informed machine learning,

    G. E. Karniadakis, I. G. Kevrekidis, L. Lu, P . Perdikaris, S. Wang, and L. Yang, "Physics - informed machine learning," Nature Reviews Physics, vol. 3, pp. 422 -440, 2021/06/01 2021

    2021

  28. [36]

    On the convergence of physics informed neural networks for linear second -order elliptic and parabolic type PDEs,

    Y . Shin, J. Darbon, and G. E. Karniadakis, "On the convergence of physics informed neural networks for linear second -order elliptic and parabolic type PDEs," arXiv preprint arXiv:2004.01806, 2020

    2004 arXiv

  29. [37]

    Physics-informed neural networks (PINNs) for fluid mechanics: a review,

    S. Cai, Z. Mao, Z. Wang, M. Yin, and G. E. Karniadakis, "Physics-informed neural networks (PINNs) for fluid mechanics: a review," Acta Mechanica Sinica, vol. 37, pp. 1727 -1738, 2021/12/01 2021

    2021

  30. [38]

    Physics-Informed Neural Networks for Low Reynolds Number Flows over Cylinder,

    E. H. W. Ang, G. Wang, and B. F. Ng, "Physics-Informed Neural Networks for Low Reynolds Number Flows over Cylinder," Energies, vol. 16, p. 4558, 2023

    2023

  31. [39]

    NSFnets (Navier -Stokes flow nets): Physics - informed neural networks for the incompressible Navier -Stokes equations,

    X. Jin, S. Cai, H. Li, and G. E. Karniadakis, "NSFnets (Navier -Stokes flow nets): Physics - informed neural networks for the incompressible Navier -Stokes equations," Journal of Computational Physics, vol. 426, p. 109951, 2021/02/01/ 2021

    2021

  32. [40]

    Physics-informed neural networks for rarefied -gas dynamics: Thermal creep flow in the Bhatnagar –Gross–Krook approximation,

    M. De Florio, E. Schiassi, B. D. Ganapol, and R. Furfaro, "Physics-informed neural networks for rarefied -gas dynamics: Thermal creep flow in the Bhatnagar –Gross–Krook approximation," Physics of Fluids, vol. 33, 2021

    2021

  33. [41]

    Physics-Informed Neural Networks for rarefied -gas dynamics: Poiseuille flow in the BGK approximation,

    M. De Florio, E. Schiassi, B. D. Ganapol, and R. Furfaro, "Physics-Informed Neural Networks for rarefied -gas dynamics: Poiseuille flow in the BGK approximation," Zeitschrift für angewandte Mathematik und Physik, vol. 73, p. 126, 2022/05/26 2022

    2022

  34. [42]

    Simulation of rarefied gas flows using physics - informed neural network combined with discrete velocity method,

    L. Zhang, W. Ma, Q. Lou, and J. Zhang, "Simulation of rarefied gas flows using physics - informed neural network combined with discrete velocity method," Physics of Fluids, vol. 35, 2023

    2023

  35. [43]

    Physics -informed neural networks for solving forward and inverse flow problems via the Boltzmann -BGK formulation,

    Q. Lou, X. Meng, and G. E. Karniadakis, "Physics -informed neural networks for solving forward and inverse flow problems via the Boltzmann -BGK formulation," Journal of Computational Physics, vol. 447, p. 110676, 2021/12/15/ 2021

    2021

  36. [44]

    Uniformly accurate machine learning-based hydrodynamic models for kinetic equations,

    J. Han, C. Ma, Z. Ma, and W. E, "Uniformly accurate machine learning-based hydrodynamic models for kinetic equations," Proceedings of the National Academy of Sciences, vol. 116, pp. 21983-21991, 2019

    2019

  37. [45]

    Learning Invariance Preserving Moment Closure Model for Boltzmann–BGK Equation,

    Z. Li, B. Dong, and Y . Wang, "Learning Invariance Preserving Moment Closure Model for Boltzmann–BGK Equation," Communications in Mathematics and Statistics, pp. 1 -43, 2023

    2023

  38. [46]

    Data -driven discovery of governing equations for fluid dynamics based on molecular simulation,

    J. Zhang and W. Ma, "Data -driven discovery of governing equations for fluid dynamics based on molecular simulation," Journal of Fluid Mechanics, vol. 892, p. A5, 2020

    2020

  39. [47]

    Learning of viscosity functions in rarefied gas flows with physics-informed neural networks,

    J.-M. Tucny, M. Durve, A. Montessori, and S. Succi, "Learning of viscosity functions in rarefied gas flows with physics-informed neural networks," Computers & Fluids, vol. 269, p. 106114, 2024/01/30/ 2024. 26

    2024

  40. [48]

    Deep Learning Closure of the Navier– Stokes Equations for Transition-Continuum Flows,

    A. S. Nair, J. Sirignano, M. Panesi, and J. F. MacArt, "Deep Learning Closure of the Navier– Stokes Equations for Transition-Continuum Flows," AIAA Journal, vol. 0, pp. 1-14, 2023

    2023

  41. [49]

    Neural network -based finite volume method and direct simulation Monte Carlo solutions of non -equilibrium shock flow guided by nonlinear coupled constitutive relations,

    G. Garg, T. K. Mankodi, E. Esmaeilifar, and R. S. Myong, "Neural network -based finite volume method and direct simulation Monte Carlo solutions of non -equilibrium shock flow guided by nonlinear coupled constitutive relations," Physics of Fluids, vol. 36, 2024

    2024

  42. [50]

    Physics-informed machine learning: A survey on problems, methods and applications,

    Z. Hao, S. Liu, Y . Zhang, C. Ying, Y . Feng, H. Su, et al., "Physics-informed machine learning: A survey on problems, methods and applications," arXiv preprint arXiv:2211.08064, 2022

    2022 arXiv

  43. [51]

    A comprehensive and fair comparison of two neural operators (with practical extensions) based on FAIR data,

    L. Lu, X. Meng, S. Cai, Z. Mao, S. Goswami, Z. Zhang , et al. , "A comprehensive and fair comparison of two neural operators (with practical extensions) based on FAIR data," Computer Methods in Applied Mechanics and Engineering, vol. 393, p. 114778, 2022/04/01/ 2022

    2022

  44. [52]

    Understanding and mitigating gradient flow pathologies in physics-informed neural networks,

    S. Wang, Y . Teng, and P . Perdikaris, "Understanding and mitigating gradient flow pathologies in physics-informed neural networks," SIAM Journal on Scientific Computing, vol. 43, pp. A3055-A3081, 2021

    2021

  45. [53]

    When and why PINNs fail to train: A neural tangent kernel perspective,

    S. Wang, X. Yu, and P . Perdikaris, "When and why PINNs fail to train: A neural tangent kernel perspective," Journal of Computational Physics, vol. 449, p. 110768, 2022/01/15/ 2022

    2022

  46. [54]

    On stresses in rarified gases arising from inequalities of temperature,

    J. C. Maxwell, "On stresses in rarified gases arising from inequalities of temperature," Philosophical Transactions of the royal society of London, vol. 170, pp. 231-256, 1879

  47. [55]

    A review on slip models for gas microflows,

    W. M. Zhang, G. Meng, and X. Y . Wei, "A review on slip models for gas microflows," Microfluidics and Nanofluidics, vol. 13, pp. 845-882, Dec 2012

    2012

  48. [56]

    Direct Monte Carlo Simulation of Transverse Supersonic Rarefied Gas Flow around a Cylinder,

    M. Y . Plotnikov, "Direct Monte Carlo Simulation of Transverse Supersonic Rarefied Gas Flow around a Cylinder," Fluid Dynamics, vol. 39, pp. 495-502, 2004/05/01 2004

    2004

  49. [57]

    Direct simulation Monte Carlo on petaflop supercomputers and beyond,

    S. Plimpton, S. Moore, A. Borner, A. Stagg, T. Koehler, J. Torczynski, et al., "Direct simulation Monte Carlo on petaflop supercomputers and beyond," Physics of Fluids, vol. 31, 2019

    2019

  50. [58]

    Non-equilibrium effects on flow past a circular cylinder in the slip and early transition regime,

    X.-J. Gu, R. W. Barber, B. John, and D. R. Emerson, "Non-equilibrium effects on flow past a circular cylinder in the slip and early transition regime," Journal of Fluid Mechanics, vol. 860, pp. 654-681, 2018

    2018

  51. [59]

    Molecular-gas-dynamics simulations of turbulent Couette flow over a mean-free-path-scale permeable substrate,

    R. M. McMullen, M. C. Krygier, J. R. Torczynski, and M. A. Gallis, "Molecular-gas-dynamics simulations of turbulent Couette flow over a mean-free-path-scale permeable substrate," Physical Review Fluids, vol. 8, p. 083401, 08/28/ 2023

    2023

  52. [60]

    Kokkos: Enabling manycore performance portability through polymorphic memory access patterns,

    H. C. Edwards, C. R. Trott, and D. Sunderland, "Kokkos: Enabling manycore performance portability through polymorphic memory access patterns," Journal of parallel and distributed computing, vol. 74, pp. 3202-3216, 2014

    2014

  53. [61]

    How important are activation functions in regression and classification? A survey, performance comparison, and future directions,

    A. D. Jagtap and G. E. Karniadakis, "How important are activation functions in regression and classification? A survey, performance comparison, and future directions," Journal of Machine Learning for Modeling and Computing, vol. 4, 2023

    2023

  54. [62]

    Adaptive activation functions accelerate convergence in deep and physics -informed neural networks,

    A. D. Jagtap, K. Kawaguchi, and G. E. Karniadakis, "Adaptive activation functions accelerate convergence in deep and physics -informed neural networks," Journal of Computational Physics, vol. 404, p. 109136, 2020/03/01/ 2020

    2020

  55. [63]

    Adam: A method for stochastic optimization,

    D. P . Kingma and J. Ba, "Adam: A method for stochastic optimization," arXiv preprint arXiv:1412.6980, 2014

    2014 arXiv

  56. [64]

    Physics -Informed Neural Networks with Hard Constraints for Inverse Design,

    L. Lu, R. Pestourie, W. Yao, Z. Wang, F. Verdugo, and S. G. Johnson, "Physics -Informed Neural Networks with Hard Constraints for Inverse Design," SIAM Journal on Scientific Computing, vol. 43, pp. B1105-B1132, 2021

    2021

  57. [65]

    A method for representing periodic functions and enforcing exactly periodic boundary conditions with deep neural networks,

    S. Dong and N. Ni, "A method for representing periodic functions and enforcing exactly periodic boundary conditions with deep neural networks," Journal of Computational Physics, vol. 435, p. 110242, 2021/06/15/ 2021. 27

    2021

  58. [66]

    Data normalization and standardization: a technical report,

    P . J. M. Ali, R. H. Faraj, E. Koya, P . J. M. Ali, and R. H. Faraj, "Data normalization and standardization: a technical report," Mach Learn Tech Rep, vol. 1, pp. 1-6, 2014

    2014

  59. [67]

    DeepXDE: A deep learning library for solving differential equations,

    L. Lu, X. Meng, Z. Mao, and G. E. Karniadakis, "DeepXDE: A deep learning library for solving differential equations," SIAM Review, vol. 63, pp. 208-228, 2021

    2021

  60. [68]

    A. A. Martín Abadi, Paul Barham, Eugene Brevdo,, C. C. Zhifeng Chen, Greg S. Corrado, Andy Davis,, M. D. Jeffrey Dean, Sanjay Ghemawat, Ian Goodfellow,, G. I. Andrew Harp, Michael Isard, Rafal Jozefowicz, Yangqing Jia,, M. K. Lukasz Kaiser, Josh Leven berg, Dan Mané, Mike Schu...

    2015

  61. [69]

    CUDA, release: 11.8,

    NVIDIA, "CUDA, release: 11.8," ed, 2022

    2022

  62. [70]

    An extended Navier-Stokes formulation for gas flows in the Knudsen layer near a wall,

    Z. L. Guo, B. C. Shi, and C. G. Zheng, "An extended Navier-Stokes formulation for gas flows in the Knudsen layer near a wall," Epl, vol. 80, p. 6, 2007

    2007

  63. [71]

    Computation of Effective Viscosities for Rarefied Gas Flows Using Ray -Tracing,

    J.-M. Tucny, S. Leclaire, D. Vidal, and F. Bertrand, "Computation of Effective Viscosities for Rarefied Gas Flows Using Ray -Tracing," International Journal of Applied and Computational Mathematics, vol. 9, p. 110, 2023/09/21 2023

    2023

  64. [72]

    On the invalidity of the extended Navier- Stokes equations to compute rarefied gas flows in a cylinder array (in revision),

    J.-M. Tucny, S. Leclaire, D. Vidal, and F. Bertrand, "On the invalidity of the extended Navier- Stokes equations to compute rarefied gas flows in a cylinder array (in revision)," 2024

    2024

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.