Pith. sign in

REVIEW 3 major objections 5 minor 26 references

A CNN-LSTM surrogate trained on just four sinusoidal pulsating pipe flows can predict drag reduction for arbitrary non-sinusoidal flows, so long as the training data cover the local flow states involved.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-03 13:14 UTC pith:2CEIIRR3

load-bearing objection A credible extension of the CNN-LSTM surrogate to arbitrary non-sinusoidal pulsations with a useful but imperfect generalization diagnostic; the main held-out result stands, though the PTD mechanism claim needs tempering. the 3 major comments →

arxiv 2512.24757 v2 pith:2CEIIRR3 submitted 2025-12-31 physics.flu-dyn

Generalization Capability of Deep Learning for Predicting Drag Reduction in Pulsating Turbulent Pipe Flow with Arbitrary Acceleration and Deceleration

Sota Kumazawa , Yasuhiro Yoshida , Tomohiro Nimura , Akira Murata , Kaoru Iwamoto This is my paper
classification physics.flu-dyn
keywords pulsating pipe flowdrag reductionCNN-LSTM surrogategeneralizationlocal temporal similarityrelaminarizationdirect numerical simulationflow-state space
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Pulsating pipe flow can cut friction drag, but searching waveforms by direct numerical simulation is expensive. This paper claims that a CNN-LSTM surrogate, trained only on a handful of sinusoidal pulsating flows, predicts arbitrary non-sinusoidal pulsating flows accurately because it learns local short-time evolution rather than the global waveform. The model reproduced 33 unseen non-sinusoidal flows with a mean absolute error of 6.6 in phase-averaged drag reduction rate, comparable to the 6.7 error for sinusoids, and once intermittent and relaminarizing cases were added to training it covered drag reduction up to 86% with MAE 9.2. The paper's quantitative guide is the pulsating trajectory difference (PTD): distance in the normalized C_f–Re_b plane between test and training trajectories, which correlates with prediction error (C=0.62). If true, the useful rule is to choose training data that tile the local flow-state space, not the waveform parameter space.

Core claim

The central discovery is that a recursively applied CNN-LSTM sequence-to-sequence model, trained only on sinusoidal pulsating pipe flows, generalizes to arbitrary non-sinusoidal pulsating flows because it predicts local temporal evolution (a short input-output window, Δt*_seq = 0.5) rather than the global waveform shape. With four sinusoidal training flows, the model achieves a phase-averaged drag-reduction MAE of 6.6 over 33 unseen non-sinusoidal flows with drag reduction from −1% to 23%; the time-averaged MAE is 3.0. When representative intermittent-transition and relaminarizing flows are added to training (total 84,000 snapshots over six flows), the model covers drag reduction up to 86% w

What carries the argument

The central objects are (i) the CNN-LSTM Seq2Seq-with-TDNN architecture that compresses r–z velocity fields into latent vectors and recursively advances them in time, augmented by a physics-informed loss term that penalizes wall-shear-stress error (the term that matters; a continuity term added little), and (ii) the pulsating trajectory difference (PTD), a phase-averaged Euclidean distance from each point of a test flow's trajectory to the nearest point among all training trajectories in the normalized C_f–Re_b plane. PTD does the argumentative work: it converts the abstract idea of 'local temporal similarity' into a computable number that predicts when the surrogate will fail, giving C=0.62

Load-bearing premise

The argument rests on treating proximity in the normalized C_f–Re_b plane as a sufficient proxy for similarity of the full flow state, even though the paper itself notes that PTD 'does not fully represent high-dimensional turbulent structures'; if two flows share a trajectory but differ in near-wall vortical structure, the learned local operator may not transfer.

What would settle it

Construct two non-sinusoidal waveforms whose phase-averaged C_f–Re_b trajectories coincide (PTD_DNS ≈ 0) but whose near-wall turbulence is phase-shifted or structurally different, e.g., one with a double peak in Reynolds shear stress and one without. If the surrogate's MAE differs markedly between them, or if a flow with large PTD_DNS is predicted accurately because its unseen structures are simple, then PTD is not a sufficient state descriptor and the training-data rule based on it would fail.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Training data for flow-surrogate models should be selected to tile the local flow-state space (C_f–Re_b trajectory) rather than to match the global pressure-gradient waveform.
  • Pulsation period matters more than amplitude for training coverage, because period controls the wall-shear-stress phase lag relative to bulk velocity.
  • Qualitatively distinct regimes—intermittent laminar–turbulent transition and relaminarization—must appear in training data; without them the model fails (MAE up to 78 for relaminarizing cases), and with one representative example per regime it recovers (MAE 5.6).
  • The recursive local-evolution design lets a model trained on simple sinusoids predict complex non-sinusoidal waveforms, so waveform complexity itself is not the obstacle; coverage of local states is.
  • Drag reduction rate, a global scalar, can be predicted accurately (time-averaged MAE 3.0) as a by-product of predicting the full spatiotemporal field, suggesting the surrogate is a viable low-cost replacement for DNS in this parameter space.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • One testable extension is to use PTD as an a priori data-acquisition criterion: before running expensive DNS for a new waveform, estimate its C_f–Re_b trajectory from a cheap low-fidelity model and add the waveform to training only if its PTD exceeds a threshold; the paper's C=0.62 correlation suggests this would control worst-case error.
  • If the local-similarity hypothesis holds beyond pipe flow, the same PTD-style metric could be applied to channel flow, boundary layers, or other drag-reduction contexts where a flux–response phase plane exists—offering a general rule for 'which training data do I need?'
  • The paper's observation that PTD_ML is systematically smaller than PTD_DNS implies the model acts as a contraction toward training-data states; an inference is that uncertainty estimates could be built by tracking PTD growth during recursive prediction, flagging prediction trustworthiness in real time without DNS.
  • A stronger test of the central mechanism would be to train the same architecture on Fourier-random waveforms and test on sinusoids: if local similarity is the operative principle, error should again be predicted by PTD rather than by waveform family.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper describes a physics-informed CNN-LSTM model that predicts the spatiotemporal evolution of pulsating turbulent pipe flow from DNS data. The model is trained on sinusoidal pulsation cases and is evaluated on unseen sinusoidal and arbitrary non-sinusoidal waveforms. In §3.2 a model trained on four sinusoidal flows achieves a phase-averaged MAE of 6.6 for the drag-reduction rate over 33 unseen non-sinusoidal cases with RD in [-1%, 23%]; in §3.3, after augmenting the training set with sinusoidal flows from intermittent and relaminarizing regimes, the model reports MAE 9.2 over 36 cases with RD up to 86%. The paper introduces the pulsating trajectory difference (PTD, Eq. 14) in the Cf–Reb plane and reports a correlation C=0.62 with prediction MAE, which it interprets as evidence that generalization is governed by local temporal similarity between test and training flows.

Significance. If the claimed predictions are robust, this is a practically valuable result: a recursive CNN-LSTM trained on a modest number of sinusoidal DNS cases can predict drag reduction for a wide range of non-sinusoidal waveforms, and the proposed training-data-coverage guideline is directly useful for surrogate-model construction. The paper contains genuine held-out tests on non-sinusoidal flows, an explicit ablation of the physics-informed loss terms, and reproducible implementation details (fixed random seeds, data ranges, loss definitions). The 6.6/9.2 MAE results are explicit and falsifiable. The main issue is that the PTD-based mechanism is not as strongly established as claimed.

major comments (3)
  1. [§3.2, Eq. (14)] PTD as defined is a static Euclidean distance in the normalized (Reb, Cf) plane, with no information about the local time derivative or the preceding short trajectory. The Cf–Reb relation in pulsating flow is hysteretic: a test point on the deceleration branch can be close to a training point on the acceleration branch with opposite evolution direction, even though the CNN-LSTM learns a 10-step temporal operator. Thus small PTD does not establish 'local temporal similarity' in the sense needed to transfer the learned operator. The C=0.62 in Fig. 16 is consistent with coarse phase-plane coverage and with the mechanical dependence of MAE on deviations in Cf. Please either augment PTD with directional/trajectory information (e.g., local dReb/dt and dCf/dt, or distance between short sequences) or present only the weaker claim that phase-space coverage correlates with accuracy.
  2. [Abstract and §3.3] The abstract states that the model, trained exclusively on a limited set of sinusoidal flows, predicted 36 arbitrary non-sinusoidal flows with MAE 9.2. Reading §3.3, the 36-case result is obtained after increasing the total training data to 84,000 snapshots and adding representative sinusoidal flows from the intermittent and relaminarizing regimes; without these additions the relaminarizing case has MAE 78. This makes the abstract's attribution of the 9.2 result misleading. Please separate the 33-case/6.6-MAE experiment from the 36-case/9.2-MAE augmented experiment, both in the abstract and in §3.3.
  3. [§3.2/Appendix A] The four-waveform training configuration was selected by screening 48 combinations on a 16-case subset of the target test data. This is a valid model-selection step only if those 16 cases are excluded from all reported test statistics; the manuscript implies this, but the process should be stated explicitly, and the chosen split should be justified as independent. As written, the 'arbitrary' generalization claim is partly conditioned on information from the target distribution (only the remaining 33 cases are truly blind). Please clarify the split and, ideally, report the dependence of the MAE on the screening subset.
minor comments (5)
  1. [§2.2.2] The text says prediction starts from a single DNS sequence whose initial distribution is a steady flow without pulsation, but the test ranges described in §3.1–3.2 are 80<t*<150. Clarify whether the initial input is at t*=0 or t*=80; this affects reproducibility.
  2. [§2.1, Eq. (5)] The notation '−⟨dp*/dz*⟩>2' is hard to parse; define the acceleration phase and T*_acc more explicitly.
  3. [§3.1.1, Eq. (8)] Clarify that uθ from DNS is used only in training for the continuity residual, and state whether this imposes any restriction on the application of the model when uθ is not available.
  4. [§3.3, Fig. 18] The high-RD and relaminarizing cases are assessed only via MAE of the drag-reduction rate. Since these regimes are qualitatively different, at least one representative comparison of phase-averaged velocity/Reynolds stress (as in Fig. 13) would substantially strengthen the claim that the model reconstructs the flow field, not just the scalar.
  5. [Fig. 7] The statement that the marginal MAE increase in the large-data regime is 'a result of variability' needs quantitative support (error bars or multiple seeds).

Circularity Check

0 steps flagged

No significant circularity: the generalization results are genuine held-out predictions, and the PTD correlation is an independent diagnostic rather than a fitted input.

full rationale

The paper's central claim is that a CNN-LSTM trained only on sinusoidal pulsating DNS flows can predict arbitrary non-sinusoidal pulsating flows. The reported errors (MAE 6.6 for 33 non-sinusoidal cases with RD -1% to 23%, and MAE 9.2 for 36 cases including relaminarization) are computed on test flows not used in weight fitting, so they are genuine held-out predictions rather than fitted inputs renamed as predictions. The PTD metric in Eq. (14) is an independently defined Euclidean distance between test and training trajectories in the normalized Cf-Reb plane; it is not derived from the prediction error, and the correlation C=0.62 with MAE is not an identity. Although both PTD and the drag-reduction error involve Cf, the paper partially addresses this by substituting tau_w for Cf and obtaining a similar correlation (0.50), and the moderate correlation coefficient shows the relationship is statistical rather than constructionally forced. The training-data combination was selected using a separate 16-case benchmark subset, with the main results reported on the remaining independent 33 cases, explicitly guarding against leakage. Self-citations to Matsubara et al. [11] supply the architecture, hyperparameters, and prior DNS validation; these are not invoked as an external uniqueness theorem, and the central generalization claim is independently tested on held-out data. The paper's own limitation statement that PTD 'does not fully represent high-dimensional turbulent structures' is a validity caveat about the similarity metric, not evidence that the prediction reduces to its inputs. No load-bearing step reduces by construction to a fitted parameter or to a self-citation chain.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 0 invented entities

The central claim rests on DNS training data (validated only for sinusoids by prior work), the data-driven choice of which sinusoidal waveforms to train on, and the sufficiency of C_f-Re_b local proximity as a state descriptor. The PTD metric is a new diagnostic but not an independently verified physical entity.

free parameters (4)
  • Adaptive loss weights λ1 and λ2 = Not reported
    Weights balancing L_pixel, L_continuity, and L_tauw in Eq. (6); said to be dynamically updated during training, but neither the values nor the update rule is given.
  • Training waveform combination for Section 3.2 = (T*, A*) = (6,7), (8,5), (4,7), (6,3)
    Selected as the best of 48 combinations using a 16-case benchmark subset of the test data (Appendix A). This is a data-driven selection, not a first-principles choice.
  • PTD normalization denominators = Mean values of the C_f and Re_b axes
    Eq. (14) normalizes each coordinate by its mean; this choice affects PTD magnitudes and the reported C=0.62 correlation with MAE.
  • Training snapshot count = 14,000 (Section 3.2); 84,000 (Section 3.3)
    Chosen using the error-plateau result in Fig. 7; a data-diet tuning choice that affects model performance.
axioms (4)
  • domain assumption Incompressible Navier-Stokes DNS at Re_tau=180 with the stated grid and domain faithfully represents pulsating turbulent pipe flow.
    The governing equations and grid are given in Section 2.1, but validation is inherited from Matsubara et al. [11] for sinusoidal flows; the arbitrary spline-driven waveforms are not separately validated against experiment or grid-convergence data.
  • ad hoc to paper Local temporal similarity in the C_f-Re_b plane is a sufficient proxy for flow-state similarity.
    This is the load-bearing assumption behind PTD and the generalization interpretation; the paper acknowledges it does not fully represent high-dimensional turbulent structures (Section 3.2).
  • domain assumption Phase-averaged statistics are sufficient for evaluating prediction quality.
    MAE and correlation are computed from phase-averaged quantities (Eqs. 10-13); the authors admit instantaneous fields diverge chaotically, so only statistical agreement is claimed.
  • domain assumption Training on a single r-z cross-section and testing on four cross-sections is sufficient for statistical representativeness.
    Section 2.2.2 states training uses one r-z cross-section while test uses four; this presumes the learned dynamics are homogeneous across circumferential positions.

pith-pipeline@v1.3.0-alltime-deepseek · 17544 in / 12198 out tokens · 136782 ms · 2026-08-03T13:14:55.124006+00:00 · methodology

0 comments
read the original abstract

The spatiotemporal evolution of pulsating turbulent pipe flow was predicted by deep learning. A convolutional neural network (CNN) and long short-term memory (LSTM) were employed for long-term prediction by recursively predicting the local temporal evolution. To enhance prediction, physical components such as wall shear stress were informed into the training process. The datasets were obtained from direct numerical simulation (DNS). The model was trained exclusively on a limited set of sinusoidal pulsating flows driven by pressure gradients defined by their period and amplitude. Subsequently, 36 pulsating flows with arbitrary non-sinusoidal acceleration and deceleration were predicted to evaluate the generalization capability, defined as the predictive performance on unseen data during training. The model successfully predicted drag reduction rates ranging from $-1\%$ to $86\%$, with a mean absolute error of 9.2. This predictive performance for unseen pulsations indicates that local temporal prediction plays a central role, rather than learning the global profile of the pulsating waveforms. This implication was quantitatively verified by analyzing the differences in periodic $C_f$--$Re_b$ trajectories between the training and test datasets, demonstrating that flows exhibiting local similarity to the training data are more predictable. Furthermore, it was demonstrated that flows exhibiting intermittent laminar--turbulent transition and relaminarization become predictable when such regimes are incorporated into the training data. The results indicate that accurate prediction is achievable provided that the training data sufficiently cover the local flow-state space, highlighting the importance of appropriate training data selection for generalized flow prediction.

Figures

Figures reproduced from arXiv: 2512.24757 by Akira Murata, Kaoru Iwamoto, Sota Kumazawa, Tomohiro Nimura, Yasuhiro Yoshida.

Figure 1
Figure 1. Figure 1: Schematic of the computational domain and coordinate system. [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Example of generating an arbitrary non-sinusoidal pressure gradient waveform [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Map of drag reduction rates (RD) for sinusoidal pulsating flows in the (T ∗ , A∗ ) parameter space obtained from DNS, where T ∗ and A∗ are defined in Eq. (2). Points (a), (b), and (c) correspond to (T ∗ , A∗ ) = (5, 5), (9, 5), and (5, 9), respectively, and are utilized as training data candidates in Section 3.1. The symbols indicate flow regimes: +, turbulent flow; ■, relaminarizing flow (RD = 86%) [PITH… view at source ↗
Figure 4
Figure 4. Figure 4: Map of drag reduction rates (RD) for arbitrary non-sinusoidal pulsating flows in the (T ∗ , Aacc) parameter space obtained from DNS. Here, Aacc denotes the effective amplitude for arbitrary non-sinusoidal waveforms, as defined in Eq. (5). The symbols indicate flow regimes: +, turbulent flow; ■, relaminarizing flow (RD = 86%). 11 [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Architecture of the physics-informed CNN LSTM Seq2Seq with TDNN model. [PITH_FULL_IMAGE:figures/full_fig_p014_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Conceptual illustration of pressure gradient waveforms and the proposed gener [PITH_FULL_IMAGE:figures/full_fig_p017_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Dependence of the mean absolute error (MAE) of the drag reduction rate on the [PITH_FULL_IMAGE:figures/full_fig_p018_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Mean absolute error (MAE) of the drag reduction rate ( [PITH_FULL_IMAGE:figures/full_fig_p021_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Mean absolute error (MAE) of the drag reduction rate ( [PITH_FULL_IMAGE:figures/full_fig_p022_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Response of bulk velocity and wall shear stress to variations in the pulsation [PITH_FULL_IMAGE:figures/full_fig_p023_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Visualization of instantaneous streamwise velocity fields. Left panels: DNS [PITH_FULL_IMAGE:figures/full_fig_p024_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Phase-averaged bulk velocity and wall shear stress. [PITH_FULL_IMAGE:figures/full_fig_p026_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Radial profiles of: (a) phase-averaged streamwise velocity; (b) root-mean-square [PITH_FULL_IMAGE:figures/full_fig_p028_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Mean absolute error (MAE) of the drag reduction rate ( [PITH_FULL_IMAGE:figures/full_fig_p030_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Trajectories of four sinusoidal pulsating flows (training data) and a represen [PITH_FULL_IMAGE:figures/full_fig_p032_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Relationship between the MAE values of the drag reduction rate ( [PITH_FULL_IMAGE:figures/full_fig_p034_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Visualization of instantaneous streamwise velocity fields characterizing high [PITH_FULL_IMAGE:figures/full_fig_p036_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: Mean absolute error (MAE) of the drag reduction rate ( [PITH_FULL_IMAGE:figures/full_fig_p038_18.png] view at source ↗
Figure 7
Figure 7. Figure 7: Table A.2: Screening results of training data combinations. The table lists the sinusoidal parameter sets (T ∗ , A∗ ) and the resulting mean absolute error (MAE) of RgD. The best￾performing configuration is highlighted in bold. The screening process was conducted using a subset of 16 benchmark data selected from the 49 test data. No. Training waveforms (T ∗ , A∗ ) MAE Remarks Combination of 3 waveforms 1 (… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

26 extracted references · 2 linked inside Pith

  1. [1]

    Pulsatile pipe flow transition: Flow waveform effects

    Brindise, M.C., and Vlachos, P.P., 2018, “Pulsatile pipe flow transition: Flow waveform effects”, Physics of Fluids, Vol. 30, 015111

    2018

  2. [2]

    Highly Energy-Conservative Finite Difference Method for the Cylindrical Coordinate System

    Fukagata, K. and Kasagi, N., 2002, “Highly Energy-Conservative Finite Difference Method for the Cylindrical Coordinate System”, Journal of Computational Physics, Vol. 181, Issue 2, pp. 478–498

    2002

  3. [3]

    Synthetic Turbulent Inflow Generator Using Machine Learning

    Fukami, K., Nabae, Y., Kawai, K., and Fukagata, K., 2019, “Synthetic Turbulent Inflow Generator Using Machine Learning”, Physical Review Fluids, Vol. 4, 064603

    2019

  4. [4]

    Direct Numerical Sim- ulation for Drag Reduction by Pulsating Turbulent Pipe Flow

    Iwamoto, K., Morino, Y., and Murata, A., 2009, “Direct Numerical Sim- ulation for Drag Reduction by Pulsating Turbulent Pipe Flow”, Pro- 43 ceedings of ASCHT09 2nd Asian Symposium on Computational Heat Transfer and Fluid Flow

    2009

  5. [5]

    Adam: A method for Stochastic Optimization

    Kingma, D. P. and Ba, J., 2014, “Adam: A method for Stochastic Optimization”, arXiv preprint arXiv:1412.6980

    Pith/arXiv arXiv 2014

  6. [6]

    Prediction of the Drag Reduction Effect of Pulsating Pipe Flow Based on Machine Learning

    Kobayashi, W., Shimura, T., Mitsuishi, A., Iwamoto, K., and Murata, A., 2021, “Prediction of the Drag Reduction Effect of Pulsating Pipe Flow Based on Machine Learning”, International Journal of Heat and Fluid Flow, Vol. 88, 108783

    2021

  7. [7]

    Turbulent Com- bined Oscillatory Flow and Current in a Pipe

    Lodahl, C. R., Sumer, B. M., and Fredsøe, J., 1998, “Turbulent Com- bined Oscillatory Flow and Current in a Pipe”, Journal of Fluid Me- chanics, Vol. 373, pp. 313–348

    1998

  8. [8]

    Spectral Dynamics of Pulsating Tur- bulent Pipe Flow

    Manna, M. and Vacca, A., 2008, “Spectral Dynamics of Pulsating Tur- bulent Pipe Flow”, Computers & Fluids, Vol. 37, Issue 7, pp. 825–835

    2008

  9. [9]

    Studies of the wall shear stress in a turbulent pulsating pipe flow

    Mao, Z., and Hanratty, T. J., 1986, “Studies of the wall shear stress in a turbulent pulsating pipe flow”, Journal of Fluid Mechanics, Vol. 170, pp. 545–564

    1986

  10. [10]

    Influence of Large-Amplitude Os- cillations on Turbulent Drag

    Mao, Z., and Hanratty, T. J., 1994, “Influence of Large-Amplitude Os- cillations on Turbulent Drag”, AIChE Journal, Vol. 40, pp. 1601–1610

    1994

  11. [11]

    Pre- diction of pulsating turbulent pipe flow by deep learning with general- ization capability

    Matsubara, K., Mitsuishi, A., Iwamoto, K., and Murata, A., 2023, “Pre- diction of pulsating turbulent pipe flow by deep learning with general- ization capability”, International Journal of Heat and Fluid Flow, Vol. 104, 109214

    2023

  12. [12]

    Compressed Convolutional LSTM: An Efficient Deep Learning Framework to Model High Fidelity 3D Turbulence

    Mohan, A. T., Don, D., Michael, C., and Daniel L., 2019, “Compressed Convolutional LSTM: An Efficient Deep Learning Framework to Model High Fidelity 3D Turbulence”, arXiv preprint arXiv:1903.00033

    Pith/arXiv arXiv 2019

  13. [13]

    Gener- alization techniques of neural networks for fluid flow estimation

    Morimoto, M., Fukami, K., Zhang, K., and Fukagata, K., 2022, “Gener- alization techniques of neural networks for fluid flow estimation”, Neural Computing and Applications, Vol. 34, pp. 3647–3669

    2022

  14. [14]

    Effect of waveform on turbulence transition in pulsatile pipe flow

    Mor´ on, D., Feldmann, D., and Avila, M., 2022, “Effect of waveform on turbulence transition in pulsatile pipe flow”, Journal of Fluid Mechanics, Vol. 948, A20. 44

    2022

  15. [15]

    Critical assessment of turbulent drag reduction through spanwise wall oscillations

    Quadrio, M., and Ricco, P., 2004, “Critical assessment of turbulent drag reduction through spanwise wall oscillations”, Journal of Fluid Mechan- ics, Vol. 521, pp. 251–271

    2004

  16. [16]

    Physics- informed neural networks: A deep learning framework for solving for- ward and inverse problems involving nonlinear partial differential equa- tions

    Raissi, M., Perdikaris, P., and Karniadakis, G.E., 2019, “Physics- informed neural networks: A deep learning framework for solving for- ward and inverse problems involving nonlinear partial differential equa- tions”, Journal of Computational Physics, Vol. 378, pp. 686–707

    2019

  17. [17]

    Ex- perimental study on drag-reduction effect due to sinusoidal riblets in turbulent channel flow

    Sasamori, M., Mamori, H., Iwamoto, K., and Murata, A., 2014, “Ex- perimental study on drag-reduction effect due to sinusoidal riblets in turbulent channel flow”, Experiments in Fluids, Vol. 55, 1828

    2014

  18. [18]

    Boundary-Layer Theory

    Schlichting, H. and Gersten, K., 2017, “Boundary-Layer Theory”, Springer, 9th ed

    2017

  19. [19]

    Numerical Simulation of Pulsating Turbulent Channel Flow

    Scotti, A. and Piomelli, U., 2001, “Numerical Simulation of Pulsating Turbulent Channel Flow”, Physics of Fluids, Vol. 13, pp. 1367–1384

    2001

  20. [20]

    Experimental Inves- tigation of Pump Control for Drag Reduction in Pulsating Turbulent Pipe Flow

    Souma, A., Iwamoto, K., and Murata, A., 2009, “Experimental Inves- tigation of Pump Control for Drag Reduction in Pulsating Turbulent Pipe Flow”, Proceedings of Sixth International Symposium on Turbu- lence and Shear Flow Phenomena, Vol. 2, pp. 761–765

    2009

  21. [21]

    Machine learning in fluid dynamics: A critical assessment

    Taira, K., Rigas, G., and Fukami, K., 2025, “Machine learning in fluid dynamics: A critical assessment”, Physical Review Fluids, Vol. 10, 090701

    2025

  22. [22]

    Turbulent-turbulent transient concept in pulsating flows

    Taylor, P.S., and Seddighi, M., 2024, “Turbulent-turbulent transient concept in pulsating flows”, Journal of Fluid Mechanics, Vol. 982, A20

    2024

  23. [23]

    Towards physics-informed deep learning for turbulent flow prediction

    Wang, R., Kashinath, K., Mustafa, M., Albert, A., and Yu, R., 2020, “Towards physics-informed deep learning for turbulent flow prediction”, KDD ’20: Proceedings of the 26th ACM SIGKDD International Con- ference on Knowledge Discovery & Data Mining, pp. 1457–1466

    2020

  24. [24]

    Physics-guided deep learning for generating turbulent inflow conditions

    Yousif, M.Z., Yu, L., and Lim, H., 2022, “Physics-guided deep learning for generating turbulent inflow conditions”, Journal of Fluid Mechanics, Vol. 936, A21. 45

    2022

  25. [25]

    A transformer-based synthetic-inflow generator for spatially develop- ing turbulent boundary layers

    Yousif, M.Z., Zhang, M., Yu, L., Vinuesa, R., and Lim, H., 2023, “A transformer-based synthetic-inflow generator for spatially develop- ing turbulent boundary layers”, Journal of Fluid Mechanics, Vol. 957, A6

    2023

  26. [26]

    Data-centric artificial intelligence: A survey

    Zha, D., Bhat, Z. P., Lai, K. H., Yang, F., Jiang, Z., Zhong, S., and Hu, X., 2025, “Data-centric artificial intelligence: A survey”, ACM Comput- ing Surveys, 57(5), 129. 46

    2025