Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Predicting Onflow Parameters Using Transfer Learning for Domain and Task Adaptation

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Transfer learning lets a convolutional network trained on CFD surface-pressure data predict angle of attack and onflow speed in a new flow domain by retraining only the final layers.

desk verdict A competent, honest engineering demonstration of transfer learning for pressure-based onflow estimation; the main gap is the missing from-scratch baseline, but the modest claims mostly hold up. read the letter →

arxiv 2506.14784 v1 pith:TIPPBBXP submitted 2025-05-26 cs.LG physics.flu-dyn

Emre Yilmaz , Philipp Bekemeyer This is my paper · ORCID
classification cs.LGphysics.flu-dyn
keywords transferlearningconvolutionalneuralnetworksurfacepressuredataangleofattackpredictiononflowspeeddomainadaptationtaskRANSCFD
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper proposes a transfer-learning recipe for predicting angle of attack and onflow speed from sparse surface-pressure measurements: train a convolutional neural network once on simulated pressure distributions, freeze its feature-extracting layers, and retrain only the final predictor layers on data from a new flow regime, a new task, or an extended envelope. The goal is to make data-driven onflow-parameter predictors adaptable with little new training, which matters for wind tunnel monitoring and for flight and wind turbine operations where sensor faults or changing conditions would break a fixed model. On CFD data for an NLR 7301 airfoil, the recipe recovers offline-level prediction accuracy when the turbulence model changes or the domain is extended, while changing the task from angle of attack to speed works only when the final two layers are retrained and adaptation to noisy pressure data is reported as not effective. If the recipe holds on real measurements, a reusable predictor could follow a changing data distribution at a fraction of the retraining cost.

What carries the argument

The central mechanism is weight freezing with selective retraining. A ConvNet (the AlexNet-derived 1-D 'ConvNet-D' and the shallower 'ConvNet-S') is first trained offline on source-domain surface-pressure arrays; for transfer, a network with the same architecture is initialized from those weights and only the final fully connected layer or layers before the output are unfrozen, while all convolutional filters stay fixed. This operationalizes the assumption, taken from prior ConvNet transfer work, that the convolutional features are generic and reusable, so the regression head alone must adapt to the target domain or task. The trade study retrains either one layer (0.086% of ConvNet-S weights, 0.017% of ConvNet-D) or two layers (98.697% of ConvNet-S, 70.359% of ConvNet-D), and the outcome of task adaptation depends on retraining enough of that head.

What would settle it

Run the same transfer-learning pipeline with target-domain data from a physical wind-tunnel experiment instead of a second CFD turbulence model: if the retrained network's test-set mean absolute error on the experimental pressure data does not approach the offline-trained baseline within the same order of magnitude, the claim that frozen convolutional features transfer to physically different pressure distributions is refuted.

Watch

Extended reading notes

Core claim

The paper's central claim is that a ConvNet trained to map 1-D arrays of airfoil surface pressure to onflow parameters can be adapted to a target domain with a different data distribution, to an extended domain, or to a new prediction task by freezing the weights of the convolutional layers and retraining only the fully connected layers immediately preceding the output node. The main evidence is that after transfer learning to a target domain generated with a different RANS (Reynolds-averaged Navier-Stokes) turbulence model (RSM instead of Spalart-Allmaras), the retrained network's test-set mean absolute error on the target domain is similar to what the offline-trained network achieved on the source domain, for both angle of attack and onflow speed. When the domain is extended beyond the offline-training envelope, transfer learning reduces the error in the extrapolated region, in the shallow architecture by up to an order of magnitude. Task adaptation from predicting angle of attack to predicting onflow speed succeeds for the shallow network only when two fully connected layers are retrained (98.697% of its weights, still with frozen convolutions), while the deeper network remains an order of magnitude worse than a dedicated offline-trained speed predictor. The paper also states that transfer learning to a noisy-pressure domain is not effective and that, although transfer learning shortens training time for larger datasets and denser architectures, real-time learning has not been achieved.

Load-bearing premise

The approach rests on the premise that the convolutional features learned from source-domain pressure distributions are generic enough to be reused for target domains and tasks, so freezing all but the final layers costs little accuracy.

Editorial extensions

If this is right

  • A ConvNet trained on pressure data from one RANS turbulence model can be reused for a different turbulence model by retraining only the final predictor layers, reaching source-domain-level accuracy on the target domain.
  • An offline model trained on only part of the angle-of-attack and speed envelope can be extended to the full envelope by transfer learning, with errors in the extrapolated region falling by up to an order of magnitude for the shallow architecture.
  • Transferring the angle-of-attack predictor to the speed-prediction task works for the shallow network only when two fully connected layers are retrained, which unfreezes most of its weights while keeping the convolutions fixed.
  • Adapting to pressure data corrupted by Gaussian noise does not transfer well: the retrained network's errors stay much higher than in the other demonstration cases, so noise robustness requires a different approach.
  • Transfer learning reduces training time relative to offline retraining, particularly for larger datasets and denser networks, but the resulting training times still exceed the limits for real-time learning.

Reading between the lines

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

  • Editorial inference: if the frozen-feature assumption is as general as the paper suggests, the same recipe should transfer across airfoil geometries, not just across turbulence models; this can be tested by pretraining on one airfoil's pressure arrays and fine-tuning on another geometry's arrays.
  • Editorial inference: the paper's failure on noisy data implies the learned convolutional features are not robust to input corruption, so adding a denoising preprocessing step or unfreezing convolutional layers in the noise case is a natural next experiment the authors did not run.
  • Editorial inference: the real-world motivation depends on steady RANS CFD standing in for quasi-steady wind tunnel, flight, and wind turbine conditions, so a direct validation on experimental surface-pressure data is the critical step that would make the method operational.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes a transfer-learning framework for predicting two onflow parameters, angle of attack and freestream speed, from sparse surface-pressure data on an airfoil. A ConvNet is trained offline on a source domain, then its convolutional weights are frozen and selected fully connected layers are retrained on a target domain or for a changed prediction task. Demonstrations are carried out on steady RANS CFD data for the NLR 7301 airfoil, covering adaptation to a change in turbulence model (SA vs. RSM), extension of the operating domain, addition of sensor noise, and task transfer from angle-of-attack prediction to speed prediction. The paper reports that transfer learning recovers source-level accuracy in the domain-change case, improves extrapolation in the domain-extension case, and works for task adaptation only when two layers are retrained, while the noisy-data application is acknowledged to be ineffective.

Significance. If the transferability premise were established, the framework would be practically useful for updating onflow-parameter predictors when the data distribution shifts, with modest training-time savings relative to offline retraining (Table II). The paper has several strengths: it uses a held-out test-set protocol, reports training times with mean and standard deviation, clearly states the failure of the noise-adaptation scenario, and provides a useful negative result for one-layer task transfer. The central claim, however, is currently underdetermined because the key comparisons lack a from-scratch target-domain baseline, so the observed improvements could be attributable to ordinary retraining on the target data rather than to reuse of the pretrained convolutional features.

major comments (3)
  1. [Section IV.C, Fig. 12] The central quantitative claim lacks the necessary control. The figure compares N_TL on the target domain D_R with N_OL on the source domain D_S, but the only target-domain comparator, 'N_OL on D_R', is the source-trained network evaluated on target data, not a network trained from scratch on D_R. Since Section II justifies freezing the convolutional layers by the premise that their features are 'generic and can be reused for target tasks', the paper must report a from-scratch target model trained with the same architecture, data split, and hyperparameters. Without such a baseline, the recovery of source-level MAE on D_R could be explained by the target training set being sufficient to train the small unfrozen head, or even the whole network, without any benefit from the pretrained features. Figure 20 shows that in the one scenario where a true from-scratch baseline is provided, the one-layer transfer fails and only after unfreezing 98.697% of the shallow network's weights does it approach the baseline; this makes the missing control in Fig. 12 consequential rather than pedantic.
  2. [Section IV.D, Fig. 15] The domain-extension claim has the same control gap. The label 'N_OL on D_e' denotes the source/initial-domain network evaluated on the extended domain, which is a zero-shot extrapolation result, not a from-scratch model trained on D_e. The improvement of N_TL on D_e over N_OL on D_e is expected simply because N_TL has been trained on D_e training data. To demonstrate that transfer from the initial domain is beneficial, the paper needs a from-scratch network trained on the merged D_i plus D_e data under the same protocol. Without this baseline, the reported improvement in MAE on the extended domain does not support the conclusion that transfer learning, rather than additional training data, is responsible.
  3. [Section IV.F, Fig. 20 and Section IV.A] The task-adaptation experiment itself undercuts the generality of the transferability premise. One-layer transfer produces no improvement over the source network, and the only successful configuration retrains 98.697% of the weights of the shallow architecture (Section IV.A). The statement in Section IV.F that 'transfer learning is not the same as nearly fully retraining the entire network' is technically true only because the convolutional layers remain frozen, but those layers contain a small fraction of the total weights; the model is in effect almost completely retrained. This should be framed as fine-tuning with most weights unfrozen, and the generic-feature claim should be tested by freezing the convolutional layers while retraining only the final layer(s) and comparing against a from-scratch target model.
minor comments (5)
  1. [Section IV.B, Figs. 7-12] The MAE values in the bar charts are reported as single point estimates without training-seed variation; only Table II reports mean and standard deviation. Since the differences in Figs. 12 and 15 can be small on a log scale, it would be helpful to report repeated-run statistics or at least state how many seeds were used.
  2. [Section IV.A, Table III] In the ConvNet-D parameter table, the row 'Conv4 (256, 256)' appears twice; if the second occurrence is meant to be Conv5, the table should be corrected.
  3. [Appendix A.1] The appendix header says 'the parameters of the ConvNet-D, ConvNet-D, and FCNN architectures', where the second 'ConvNet-D' should presumably be 'ConvNet-S'.
  4. [Sections IV.C-IV.F] All transfer-learning MAE demonstrations fix the dataset size at nd = 1024. Since the introduction motivates transfer learning for settings with limited or streaming data, the paper would be strengthened by varying the target-domain training set size in the transfer-learning phase.
  5. [Overall] No code or data availability statement is included; providing the CFD datasets or trained models would improve reproducibility, given that all demonstrations are in-house generated.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's transfer-learning claims rest on held-out test-set measurements, not on fit identities or self-citation chains.

full rationale

The central claims are empirical. In Section II, the transferability premise ('the features extracted on source targets using convolutional operations are generic and can be reused for target tasks') is stated as an assumption and cited to external work [38], then tested by held-out MAE comparisons in Figs. 12, 15, 18-20. The offline networks are trained on source-domain training splits; the transfer networks are initialized from them and retrained on target-domain training splits; the reported accuracies are computed on test splits that were not used to set any weights. No parameter is fitted to the quantity being predicted, and no quoted result is an identity or a renamed fit. Self-citations [28, 36, 63] are background or tooling references (SMARTy for standard Halton sampling, also cited to [64]) and are not load-bearing for the transfer claim. The skeptical observation that a true from-scratch-on-target baseline is missing in Figs. 12 and 15 is a valid experimental-design limitation, but it does not make the outcome equivalent to the inputs: the absence of a control is not circularity. The paper even reports negative results (one-layer task adaptation fails; noisy-domain transfer is ineffective), which is inconsistent with a claim forced by construction.

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

The paper introduces no new physics, particles, or forces; it combines standard components. The entries above capture the hand-chosen experimental settings that the central claims depend on and the domain assumptions, stated or implicit, that carry the demonstration. The main result is an empirical measurement on held-out CFD test sets, so the circularity burden is low.

free parameters (3)
  • Number of retrained layers (1 vs 2) = 2 layers for task adaptation; 1 layer for other cases
    Chosen per scenario; the central task-adaptation result (Fig. 20) depends on this choice. Retraining one layer fails, retraining two layers unfreezes 98.697% of ConvNet-S weights and succeeds.
  • Noise variance levels = 0.01, 0.001, 0.0001
    Hand-selected in Section IV.E; the noise study is reported only at these three levels with 10 noisy replicates per sample.
  • ns and nd for demonstration cases = ns = 75 surface points, nd = 1024 runs
    Selected in Section IV.B after observing that MAE around 10^-1 is reached at nd = 1024 and that ns has no consistent trend; all transfer learning demos use these values.
assumptions (5)
  • domain assumption Features extracted by convolutional layers on the source domain/task are generic and reusable for the target domain/task
    Invoked in Section II as the justification for freezing all layers except those preceding the output; the paper's own task-adaptation and noise results show the assumption has limits.
  • domain assumption The SA-based and RSM-based datasets are related but different distributions, making transfer sensible
    Section IV.C asserts this prerequisite: distributions differ strongly near separation but are similar at low angles of attack. Relatedness is asserted qualitatively, not measured quantitatively.
  • domain assumption Steady RANS CFD solutions represent quasi-steady wind tunnel, flight, and wind turbine conditions
    Section III states this assumption explicitly; all demonstrations are in-silico, and the paper lists CFD-to-experiment transfer as future work.
  • domain assumption Surface pressure sampled at equidistant points carries enough information to identify alpha and V_infinity
    This inverse-map identifiability underpins the whole predictor; error maps (Figs. 10-11) show the assumption degrades near separation regions.
  • standard math Halton sequences provide low-discrepancy coverage of the design space
    Used in Section III via the SMARTy toolbox; standard quasi-Monte Carlo property.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Predicting Onflow Parameters Using Transfer Learning for Domain and Task Adaptation." pith.science (2026). https://pith.science/paper/TIPPBBXP

@misc{pith2026250614784,
  author       = {Pith},
  title        = {Pith review of: Predicting Onflow Parameters Using Transfer Learning for Domain and Task Adaptation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TIPPBBXP}},
  note         = {Machine review of arXiv:2506.14784}
}
read the original abstract

Determining onflow parameters is crucial from the perspectives of wind tunnel testing and regular flight and wind turbine operations. These parameters have traditionally been predicted via direct measurements which might lead to challenges in case of sensor faults. Alternatively, a data-driven prediction model based on surface pressure data can be used to determine these parameters. It is essential that such predictors achieve close to real-time learning as dictated by practical applications such as monitoring wind tunnel operations or learning the variations in aerodynamic performance of aerospace and wind energy systems. To overcome the challenges caused by changes in the data distribution as well as in adapting to a new prediction task, we propose a transfer learning methodology to predict the onflow parameters, specifically angle of attack and onflow speed. It requires first training a convolutional neural network (ConvNet) model offline for the core prediction task, then freezing the weights of this model except the selected layers preceding the output node, and finally executing transfer learning by retraining these layers. A demonstration of this approach is provided using steady CFD analysis data for an airfoil for i) domain adaptation where transfer learning is performed with data from a target domain having different data distribution than the source domain and ii) task adaptation where the prediction task is changed. Further exploration on the influence of noisy data, performance on an extended domain, and trade studies varying sampling sizes and architectures are provided. Results successfully demonstrate the potential of the approach for adaptation to changing data distribution, domain extension, and task update while the application for noisy data is concluded to be not as effective.

Figures

Figures reproduced from arXiv: 2506.14784 by the authors.

Figure 1
Figure 1. FIG. 1: Categorization of Transfer Learning [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2: Transfer Learning Framework [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3: ConvNet-D (C: convolutional, R: ReLU, P: pooling, m: max, aa: adaptive [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (19 more)
Figure 4
Figure 4. Figure 4: FIG. 4: ConvNet-S (C: convolutional, R: ReLU, P: pooling, m: max, aa: adaptive [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5: NLR 7301 [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6: Design of experiments for 128 CFD runs [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7: The variation of the test set MAE values with the architecture type and dataset [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]
Figure 8
Figure 8. Figure 8: FIG. 8: The variation of the test set MAE values with the number of surface data points [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 9
Figure 9. Figure 9: FIG. 9: The variation of the test set MAE values with the numbers of surface data points [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]
Figure 10
Figure 10. Figure 10: FIG. 10: Test set absolute errors for offline learning and the task of predicting [PITH_FULL_IMAGE:figures/full_fig_p016_10.png]
Figure 11
Figure 11. Figure 11: FIG. 11: Test set absolute errors for offline learning and the task of predicting [PITH_FULL_IMAGE:figures/full_fig_p016_11.png]
Figure 12
Figure 12. Figure 12: FIG. 12: Test set MAE values for transfer learning between the datasets generated using [PITH_FULL_IMAGE:figures/full_fig_p018_12.png]
Figure 13
Figure 13. Figure 13: FIG. 13: Test set absolute errors obtained with [PITH_FULL_IMAGE:figures/full_fig_p019_13.png]
Figure 14
Figure 14. Figure 14: FIG. 14: Test set absolute errors obtained with [PITH_FULL_IMAGE:figures/full_fig_p020_14.png]
Figure 15
Figure 15. Figure 15: FIG. 15: Test set MAE values for transfer learning beyond the boundaries of the source [PITH_FULL_IMAGE:figures/full_fig_p022_15.png]
Figure 16
Figure 16. Figure 16: FIG. 16: Test set absolute errors for transfer learning beyond the boundaries of the source [PITH_FULL_IMAGE:figures/full_fig_p023_16.png]
Figure 17
Figure 17. Figure 17: FIG. 17: Test set absolute errors for transfer learning beyond the boundaries of the source [PITH_FULL_IMAGE:figures/full_fig_p024_17.png]
Figure 18
Figure 18. Figure 18: FIG. 18: The variation of the test set MAEs with noise levels for the task of predicting [PITH_FULL_IMAGE:figures/full_fig_p025_18.png]
Figure 19
Figure 19. Figure 19: FIG. 19: The variation of the test set MAEs with noise levels for the task of predicting [PITH_FULL_IMAGE:figures/full_fig_p026_19.png]
Figure 20
Figure 20. Figure 20: FIG. 20: Test set MAE values for the case of task adaptation to predicting [PITH_FULL_IMAGE:figures/full_fig_p027_20.png]
Figure 21
Figure 21. Figure 21: FIG. 21: Test set absolute errors obtained with [PITH_FULL_IMAGE:figures/full_fig_p033_21.png]
Figure 22
Figure 22. Figure 22: FIG. 22: Test set absolute errors obtained with [PITH_FULL_IMAGE:figures/full_fig_p034_22.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Fusing CFD and measurement data using transfer learning

    cs.LG 2025-07 conditional novelty 6.0 of 10

    Transfer learning fuses dense CFD with sparse wind-tunnel data into a parametric MLP surrogate that outperforms gappy POD on transonic pressure prediction.

Reference graph

Works this paper leans on

64 extracted references · 63 canonical work pages · cited by 1 Pith paper

  1. [1]

    Finley and P

    T. Finley and P. Tcheng, Model attitude measurements at NASA Langley Research Center, inProceedings of the 30th Aerospace Sciences Meeting and Exhibit(AIAA, 1992) pp. 1–7

    work page 1992

  2. [2]

    Lee, Study of optical techniques for the Ames Unitary Wind Tunnels Part 3

    G. Lee, Study of optical techniques for the Ames Unitary Wind Tunnels Part 3. Angle of attack (1992), nASA-CR-190541, Progress Report by MCAT Institute

    work page 1992

  3. [3]

    R. K. Bogue and H. W. Jentink, Optical air flow measurements in flight (2004), NASA/TP- 2004-210735

    work page 2004

  4. [4]

    Crawford, Angle measurement system (AMS) for establishing model pitch and roll zero, and performing single axis angle comparisons, inProc

    B. Crawford, Angle measurement system (AMS) for establishing model pitch and roll zero, and performing single axis angle comparisons, inProc. of the 45th AIAA Aerospace Sciences Meeting and Exhibit(2007) pp. 1–10

    work page 2007

  5. [5]

    K. G. Toro, Technology review of wind-tunnel angle measurement, inProc. of the 11th Inter- national Symposium on Strain-Gauge Balances(2018) pp. 1–10

    work page 2018

  6. [6]

    Valasek, J

    J. Valasek, J. Harris, S. Pruchnicki, M. McCrink, J. Gregory, and D. G. Sizoo, Derived angle of attack and sideslip angle characterization for general aviation, Journal of Guidance, Control, and Dynamics43, 1039 (2020)

    work page 2020

  7. [7]

    Lerro, A

    A. Lerro, A. Brandl, and P. Gili, Model-free scheme for angle-of-attack and angle-of-sideslip estimation, Journal of Guidance, Control, and Dynamics44, 595 (2021)

    work page 2021

  8. [8]

    J. E. Zeis Jr.,Angle of attack and sideslip estimation using an inertial reference platform, Master’s thesis, Air Force Institute of Technology, Wright-Patterson Air Force Base, OH (1988), ADA194876, Available athttps://apps.dtic.mil/sti/tr/pdf/ADA194876.pdf

    work page 1988

Show all 64 references
  1. [9]

    Gracey, Summary of methods of measuring angle of attack on aircraft (1958), National Advisory Committee for Aeronautics, Technical Note 4351

    W. Gracey, Summary of methods of measuring angle of attack on aircraft (1958), National Advisory Committee for Aeronautics, Technical Note 4351

    1958

  2. [10]

    M. S. Selig and B. D. McGranahan,Wind Tunnel Aerodynamic Tests of Six Airfoils for Use on Small Wind Turbines: Period of Performance: October 31, 2002-January 31, 2003, Tech. Rep. NREL/SR-500-34515 (National Renewable Energy Laboratory, 2004)

    2002

  3. [11]

    Bridges and M

    J. Bridges and M. P. Wernet, The NASA subsonic jet particle image velocimetry (PIV) dataset (2011), Report/Patent Number E-17439, NASA/TM-2011-216807

    2011

  4. [12]

    Narbuntas, S

    T. Narbuntas, S. Conley, K. Kallstrom, and A. M. Abrego, Wind tunnel performance test of a ducted fan system (2020), NASA/TM-2020-220484. 35

    2020

  5. [13]

    N. T. Smith, J. T. Heineck, and E. T. Schairer, Optical flow for flight and wind tunnel background oriented Schlieren imaging, inProc. of the 55th AIAA Aerospace Sciences Meeting (2017) pp. 1–18

    2017

  6. [14]

    T. J. Rohloff, S. A. Whitmore, and I. Catton, Fault-tolerant neural network algorithm for flush air data sensing, Journal of Aircraft36, 541 (1999)

    1999

  7. [15]

    Quindlen and J

    J. Quindlen and J. Langelaan, Flush air data sensing for soaring-capable UA Vs, inProc. of the 51st AIAA Aerospace Sciences Meeting including the New Horizons Forum and Aerospace Exposition(2013) pp. 1–17

    2013

  8. [16]

    O. T. Topac, S. Y. Sara Ha, X. Chen, L. Gamble, D. Inman, and F.-K. Chang, Hybrid models for situational awareness of an aerial vehicle from multimodal sensing, AIAA Journal61, 305 (2023)

    2023

  9. [17]

    Goupil, Airbus state of the art and practices on FDI and FTC in flight control system, Control Engineering Practice19, 524 (2011)

    P. Goupil, Airbus state of the art and practices on FDI and FTC in flight control system, Control Engineering Practice19, 524 (2011)

    2011

  10. [18]

    Ossmann, H

    D. Ossmann, H. D. Joos, and P. Goupil, Enhanced sensor monitoring to maintain optimal aircraft handling in case of faults, Journal of Guidance, Control, and Dynamics40, 3127 (2017)

    2017

  11. [19]

    Vitale, F

    A. Vitale, F. Corraro, N. Genito, L. Garbarino, and L. Verde, An innovative angle of attack virtual sensor for physical-analytical redundant measurement system applicable to commercial aircraft, Advances in Science, Technology and Engineering Systems Journal6, 698 (2021)

    2021

  12. [20]

    Mersha, D

    B. Mersha, D. Jansen, and H. Ma, Angle of attack prediction using recurrent neural networks in flight conditions with faulty sensors in the case of F-16 fighter jet, Complex Intelligent Systems9, 2599 (2023)

    2023

  13. [21]

    Schrittwieser, I

    J. Schrittwieser, I. Antonoglou, T. Hubert, K. Simonyan, L. Sifre, S. Schmitt, A. Guez, E. Lockhart, D. Hassabis, T. Graepel, T. Lillicrap, and D. Silver, Mastering Atari, Go, chess and shogi by planning with a learned model, Nature588, 604 (2020)

    2020

  14. [22]

    Brown, B

    T. Brown, B. Mann, N. Ryder, andet al., Language models are few-shot learners, inAdvances in Neural Information Processing Systems, Vol. 33 (Curran Associates, Inc., 2020) pp. 1877– 1901

    2020

  15. [23]

    Jumper, R

    J. Jumper, R. Evans, A. Pritzel, and et al., Highly accurate protein structure prediction with AlphaFold, Nature596, 583 (2021). 36

    2021

  16. [24]

    J. C. Ross, C. C. Jorgenson, and M. Norgaard, Reducing wind tunnel data requirements using neural networks (1997), technical Memorandum, NASA TM-112193

    1997

  17. [25]

    C. F. Lo, J. L. Zhao, and R. DeLoach, Application of neural networks to wind tunnel data response surface methods, inProc. of the 21st AIAA Aerodynamic Measurement Technology and Ground Testing Conf.(2000) pp. 1–13

    2000

  18. [26]

    Rajkumar and J

    T. Rajkumar and J. Bardina, Prediction of aerodynamic coefficients using neural networks for sparse data, inProc. of the AAAI FLAIRS Conf.(2002) pp. 242–246

    2002

  19. [27]

    X. Guo, W. Li, and F. Iorio, Convolutional neural networks for steady flow approximation, in KDD’16(Association for Computing Machinery, New York, NY, USA, 2016) pp. 1–10

    2016

  20. [28]

    Yilmaz and B

    E. Yilmaz and B. German, A convolutional neural network approach to training predictors for airfoil performance, inProc. of the 18th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference(2017) pp. 1–19

    2017

  21. [29]

    Lee and D

    S. Lee and D. You, Prediction of laminar vortex shedding over a cylinder using deep learning (2017), arXiv:1712.07854 [physics.flu-dyn]

    2017 arXiv

  22. [30]

    Nagawkar and L

    J. Nagawkar and L. Leifsson, Multifidelity aerodynamic flow field prediction using random forest-based machine learning, Aerospace Science and Technology123, 107449 (2022)

    2022

  23. [31]

    Morimoto, K

    M. Morimoto, K. Fukami, K. Zhang, and K. Fukagata, Generalization techniques of neural networks for fluid flow estimation, Neural Computing and Applications34, 3647 (2022)

    2022

  24. [32]

    Hines and P

    D. Hines and P. Bekemeyer, Graph neural networks for the prediction of aircraft surface pressure distributions, Aerospace Science and Technology137, 108268 (2023)

    2023

  25. [33]

    Wassing, S

    S. Wassing, S. Langer, and P. Bekemeyer, Physics-informed neural networks for parametric compressible Euler equations, Computers and Fluids270, 106164 (2024)

    2024

  26. [34]

    X. Yan, J. Zhu, M. Kuang, and X. Wang, Aerodynamic shape optimization using a novel optimizer based on machine learning techniques, Aerospace Science and Technology86, 826 (2019)

    2019

  27. [35]

    W. Chen, K. Chiu, and M. D. Fuge, Airfoil design parameterization and optimization using b´ ezier generative adversarial networks, AIAA Journal58, 4723 (2020)

    2020

  28. [36]

    Yilmaz and B

    E. Yilmaz and B. German, Conditional generative adversarial network framework for airfoil inverse design, inProc. of the AIAA A VIATION 2020 FORUM(2020) pp. 1–18

    2020

  29. [37]

    J. Ling, A. Kurzawski, and J. Templeton, Reynolds averaged turbulence modelling using deep neural networks with embedded invariance, Journal of Fluid Mechanics807, 155 (2016). 37

    2016

  30. [38]

    Oquab, L

    M. Oquab, L. Bottou, I. Laptev, and J. Sivic, Learning and transferring mid-level image representations using convolutional neural networks, inProc. of the 2014 IEEE Conference on Computer Vision and Pattern Recognition(2014) pp. 1717–1724

    2014

  31. [39]

    Y. Sun, S. Li, H. Gao, X. Zhang, J. Lv, W. Liu, and Y. Wu, Transfer learning: A new aerodynamic force identification network based on adaptive emd and soft thresholding in hypersonic wind tunnel, Chinese Journal of Aeronautics36, 351 (2023)

    2023

  32. [40]

    Krishnan, S

    S. Krishnan, S. Ghosh, S. Atkinson, V. Andreoli, T. Vandeputte, and L. Wang, Transfer learning based modeling of industrial turbine airfoil characteristics, inProc. of the AIAA SCITECH 2022 Forum(2022)

    2022

  33. [41]

    Z. Wang, X. Liu, J. Yu, H. Wu, and H. Lyu, A general deep transfer learning framework for predicting the flow field of airfoils with small data, Computers and Fluids251, 105738 (2023)

    2023

  34. [42]

    R. Li, Y. Zhang, and H. Chen, Transfer learning from two-dimensional supercritical airfoils to three-dimensional transonic swept wings, Chinese Journal of Aeronautics36, 96 (2023)

    2023

  35. [43]

    Cappugi, A

    L. Cappugi, A. Castorrini, A. Bonfiglioli, E. Minisci, and M. S. Campobasso, Machine learning- enabled prediction of wind turbine energy yield losses due to general blade leading edge erosion, Energy Conversion and Management245, 114567 (2021)

    2021

  36. [44]

    Vaiuso, G

    A. Vaiuso, G. Immordino, M. Righi, and A. Da Ronch, Multi-fidelity transonic aerodynamic loads estimation using bayesian neural networks with transfer learning, Aerospace Science and Technology163, 110301 (2025)

    2025

  37. [45]

    C. M. Roelofs, C. G¨ uck, and S. Faulstich, Transfer learning applications for autoencoder-based anomaly detection in wind turbines, Energy and AI17, 100373 (2024)

    2024

  38. [46]

    L. Y. Pratt, J. Mostow, and C. A. Kamm, Direct transfer of learned information among neural networks, inProc. of the Ninth National Conference on Artificial Intelligence - Volume 2, AAAI’91 (AAAI Press, 1991) pp. 584–589

    1991

  39. [47]

    Naik and R

    D. Naik and R. Mammone,Learning by learning in neural networks, Vol. 24 (Chapman and Hall, London, 1993)

    1993

  40. [48]

    T. M. Mitchell and S. Thrun, Learning one more thing (1994), technical Report, CMU-CS- 94-184, CMU

    1994

  41. [49]

    S. J. Pan and Q. Yang, A survey of transfer learning, IEEE Transactions on Knowledge and Data Engineering22, 1345 (2010). 38

    2010

  42. [50]

    Devlin, M.-W

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, BERT: Pre-training of deep bidirectional transformers for language understanding, inProc. of the North American Chapter of the As- sociation for Computational Linguistics: Human Language Technologies (NAACL-HLT 2019) (2019) p. ...

    2019

  43. [51]

    M. D. Zeiler and R. Fergus, Visualizing and understanding convolutional networks, inProc. of the European Conference on Computer Vision – ECCV 2014(Springer International Pub- lishing, 2014) pp. 818–833

    2014

  44. [52]

    Lecun, L

    Y. Lecun, L. Bottou, Y. Bengio, and P. Haffner, Gradient-based learning applied to document recognition, Proc. of the IEEE86, 2278 (1998)

    1998

  45. [53]

    V. Mnih, K. Kavukcuoglu, D. Silver, and et al., Human-level control through deep reinforce- ment learning, Nature518, 529 (2015)

    2015

  46. [54]

    Ronneberger, P

    O. Ronneberger, P. Fischer, and T. Brox, U-Net: Convolutional networks for biomedical im- age segmentation, inMedical Image Computing and Computer-Assisted Intervention – MIC- CAI’2015(Springer International Publishing, Cham, 2015) pp. 234–241

    2015

  47. [55]

    Redmon, S

    J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, You Only Look Once: Unified, real- time object detection, inProc. of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)(2016) pp. 779–788

    2016

  48. [56]

    Goodfellow, Y

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

    2016

  49. [57]

    Krizhevsky, I

    A. Krizhevsky, I. Sutskever, and G. E. Hinton, Imagenet classification with deep convolutional neural networks, inProc. of the Advances in Neural Information Processing Systems, Vol. 25 (Curran Associates, Inc., 2012) pp. 1–9

    2012

  50. [58]

    TorchVision maintainers and contributors, Torchvision: PyTorch’s computer vision library, https://github.com/pytorch/vision(2016)

    2016

  51. [59]

    R. J. Zwaan, Data Set 4: NLR 7301 Supercritical airfoil oscillatory pitching and oscillating flap (1982), NATO AGARD Report 702: Compendium of unsteady aerodynamic measurements

    1982

  52. [60]

    Schwamborn, T

    D. Schwamborn, T. Gerhold, and R. Heinrich, The DLR TAU-Code: Recent applications in research and industry, inProc. of the ECCOMAS CFD 2006 Conference(2006)

    2006

  53. [61]

    S. R. Allmaras, F. T. Johnson, and P. R. Spalart, Modifications and clarifications for the implementation of the Spalart-Allmaras turbulence model, inProc. of the 7th International Conference on Computational Fluid Dynamics, ICCFD7-1902(Springer International Pub- 39 lishing, ...

    1902

  54. [62]

    Eisfeld, C

    B. Eisfeld, C. Rumsey, and V. Togiti, Verification and validation of a second-moment-closure model, AIAA Journal54, 1524 (2016)

    2016

  55. [63]

    Bekemeyer, A

    P. Bekemeyer, A. Bertram, D. A. H. Chaves, M. D. Ribeiro, A. Garbo, A. Kiener, C. Sabater, M. Stradtner, S. Wassing, M. Widhalm, S. Goertz, F. Jaeckel, R. Hoppe, and N. Hoffmann, Data-driven aerodynamic modeling using the DLR SMARTy Toolbox, inProc. of the AIAA A VIATION 2022 ...

    2022

  56. [64]

    J. H. Halton, Algorithm 247: Radical-inverse quasi-random point sequence, Communications of the ACM7, 701 (1964). 40

    1964

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.