Convolutional causal learning for aerodynamic flows

Kai Fukami; Qiong Liu; Ryo Araki; Ryo Koshikawa

arxiv: 2601.19104 · v2 · pith:2U6XBJHQnew · submitted 2026-01-27 · ⚛️ physics.flu-dyn · physics.comp-ph

Convolutional causal learning for aerodynamic flows

Ryo Koshikawa , Ryo Araki , Qiong Liu , Kai Fukami This is my paper

Pith reviewed 2026-05-21 15:03 UTC · model grok-4.3

classification ⚛️ physics.flu-dyn physics.comp-ph

keywords causal learningaerodynamic flowsvortical structuresmode decompositionconvolutional neural networkunsteady flowsdata-driven modelingsnapshot data

0 comments

The pith

A convolutional neural network with information-theoretic decomposition extracts time-dependent vortical structures that causally influence future aerodynamic forces from snapshot data.

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

The paper presents a technique that combines information-theoretic machine learning for time-varying mode decomposition with convolutional neural networks to identify causal links in unsteady aerodynamic flows directly from data snapshots. It isolates informative vortical structures whose time evolution contributes to changes in coefficients such as lift, and demonstrates this on gust-airfoil encounters, jet-wing interactions, and turbulent wakes at varying Reynolds numbers. In the wake case, large-scale motions are tied to force changes without any supplied spatial scale information. If the extraction is causal rather than correlative, the method supplies a pathway to build predictive models and control strategies for flows where traditional equations are hard to close or solve in real time.

Core claim

The central claim is that aerodynamic causality can be captured from snapshot data with a time-varying mode decomposition technique referred to as information-theoretic machine learning. When this decomposition is paired with a convolutional neural network, it yields spatial continuous modes and identifies time-dependent informative vortical structures that contribute to the future evolution of aerodynamic coefficients. A low-order representation of these structures and coefficients is obtained via autoencoder compression. The method applied to vortex-gust airfoil interactions extracts the time-varying gust effect on lift response, while in the turbulent separated wake it reveals the link to

What carries the argument

The information-theoretic machine learning time-varying mode decomposition integrated with a convolutional neural network, which produces spatial continuous modes and isolates vortical structures that drive future aerodynamic coefficients.

If this is right

In gust-airfoil cases the time-varying influence of the gust on the lift response is isolated in an interpretable form.
In turbulent wakes the relationship between large-scale vortical motion and lift force is recovered without any supplied spatial length-scale information.
Autoencoder compression yields a low-order representation that characterizes both the informative vortical structures and the associated aerodynamic coefficients.
The overall framework supplies a basis for data-driven causal modeling and control across a range of unsteady aerodynamic flows.

Where Pith is reading between the lines

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

The same snapshot-based causal extraction could be tested on three-dimensional or compressible flows to check whether the identified structures remain predictive when additional physics enter.
If the low-order representations prove stable across Reynolds numbers, they might be inserted into reduced-order controllers for real-time gust alleviation on aircraft.
The technique might be applied to other snapshot-rich domains such as ocean current data to locate causal drivers of transport without assuming fixed length scales.

Load-bearing premise

The information-theoretic decomposition applied to snapshot data extracts genuine causal contributions to future aerodynamic coefficients rather than statistical correlations.

What would settle it

Introduce a controlled perturbation to one of the vortical structures identified as causal and measure whether the aerodynamic coefficients change in the predicted way; absence of the expected change would falsify the causal extraction.

Figures

Figures reproduced from arXiv: 2601.19104 by Kai Fukami, Qiong Liu, Ryo Araki, Ryo Koshikawa.

**Figure 1.** Figure 1: An example of the given state q and the informative component qI decomposed by a data-driven technique. This study proposes information-theoretic convolutional learning that achieves causality-based mode decomposition while identifying reduced-order representations of causally important vortical structures. While existing causal mode decompositions operate in a point-wise manner, which yields spatial disco… view at source ↗

**Figure 2.** Figure 2: Informative mode extractor F based on (a) convolutional autoencoder and (b) convolutional neural network. To extract the informative vortical structure qI from the given vorticity field q with respect to a future target of lift, i.e., λ = CL, we construct an informative mode extractor F, qI (t) = F(CL(t + ∆t), q(t); w), (5) where w is the weight parameter of the mode extractor. This extractor is implemente… view at source ↗

**Figure 3.** Figure 3: Informative mode decomposition of extreme vortex-airfoil interaction. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

**Figure 4.** Figure 4: Informative mode decomposition for experimental transverse gust encounter at [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗

**Figure 5.** Figure 5: Informative modal structure of spanwise-averaged separated flow over wing section at [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗

**Figure 6.** Figure 6: (a) Informative modal structure of three-dimensional separated flow over wing section at Re = 20, 000 visualized with the iso-surface (Qth = 100) colored by streamwise velocity u. (b) Scale-decomposed fields with two cut of length-scales. (c) Dependence of the streamwise variation of the magnitude of Q-criterion on time window ∆t. addition to providing time-varying modal structures based on causality. This… view at source ↗

**Figure 5.** Figure 5: figure 5. The balancing parameter [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗

read the original abstract

This study aims to capture aerodynamic causality from snapshot data with a time-varying mode decomposition technique referred to as information-theoretic machine learning. The current approach extracts time-dependent informative vortical structures, contributing to the future evolution of the aerodynamic coefficients. The present decomposition is employed with a convolutional neural network, enabling the identification of the spatial continuous mode. In addition, a low-order representation, characterizing the informative vortical structures and their corresponding aerodynamic coefficients, can also be identified by considering autoencoder-based data compression. The present technique is applied to a range of aerodynamic examples, including extreme vortex-gust airfoil interactions, experimentally measured transverse jet-wing interaction, and a turbulent separated wake across different Reynolds numbers. For the cases of gust-wing interaction, the time-varying gust effect on the lift response is extracted in an interpretable manner. With the example of a turbulent wake, the relationship between large-scale vortical motion and lift force is identified without any spatial length-scale information. The proposed approach could serve as a foundation for data-driven causal modeling and control for a range of unsteady flows.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This paper combines info-theoretic time-varying decomposition with CNNs and autoencoders to pull vortical structures from flow snapshots that appear tied to lift changes, but the causal framing rests on correlations without interventional checks.

read the letter

Colleague, the main point is that this work takes standard information-theoretic decomposition, makes it time-varying, and layers on convolutional networks plus autoencoder compression to identify vortical structures in aerodynamic snapshot data that influence future force coefficients. It applies the pipeline to gust-wing interactions, experimental jet-wing cases, and turbulent wakes across Reynolds numbers. In the wake example it reports a link to lift without feeding in any spatial length-scale data upfront, which is a concrete result if the extraction holds. The gust case also shows time-dependent effects on lift in a way that looks interpretable from the description. Those applications are the part that feels useful for unsteady aerodynamics work. The soft spot is the causality claim itself. Snapshot statistics plus mode decomposition can surface strong predictors, but the abstract gives no sign of ablation on shuffled time series, surrogate data, or actual interventions that would change the modes and check whether the coefficient response moves as predicted. Without those steps the distinction from correlation stays thin, especially in the turbulent case where the claim is strongest. This is aimed at the data-driven fluid mechanics crowd who already use modal methods and want something framed around future prediction for control ideas. A reader who cares about causal inference in physics would see the setup but would probably want the missing validation tests before treating the structures as causal drivers. I would send it for peer review. The cases are relevant and the combination is specific enough to discuss, even if the full manuscript needs to add quantitative checks and clearer separation of correlation from directed influence.

Referee Report

1 major / 2 minor

Summary. The manuscript introduces a convolutional causal learning framework that applies an information-theoretic machine learning time-varying mode decomposition to snapshot data in order to extract time-dependent vortical structures that causally influence the future evolution of aerodynamic coefficients. The decomposition is combined with a convolutional neural network to obtain spatially continuous modes and with an autoencoder to produce low-order representations of the informative structures and coefficients. The method is demonstrated on three cases: extreme vortex-gust airfoil interactions, experimentally measured transverse jet-wing interactions, and turbulent separated wakes at multiple Reynolds numbers. The authors claim that the approach identifies interpretable causal links, such as the time-varying gust effect on lift and the relationship between large-scale vortical motion and lift force in the wake without requiring explicit spatial length-scale information, and that it can serve as a foundation for data-driven causal modeling and control of unsteady flows.

Significance. If the extracted modes can be shown to isolate genuine causal contributions rather than predictive correlations, the work would offer a useful data-driven route to interpretable causal modeling in unsteady aerodynamics. The integration of information-theoretic decomposition with CNNs and autoencoders for both spatial continuity and dimensionality reduction is a constructive combination that could support downstream control applications across gust, jet, and wake problems.

major comments (1)

[Abstract and §4.3] The central causality claim for the turbulent-wake example (abstract and §4.3) rests on the assertion that the time-varying mode decomposition identifies structures whose influence on lift is causal. Snapshot data alone yields joint statistics; without reported interventional tests (e.g., targeted mode perturbation, time-shuffling, or surrogate-data controls that preserve correlations while destroying directed temporal influence), the extracted relationship remains consistent with non-causal predictive modeling. This issue is load-bearing for the claim that the method extracts causality rather than correlation.

minor comments (2)

[§3] The method section would benefit from explicit equations or pseudocode showing how the information-theoretic objective is optimized and how the time-varying modes are updated at each snapshot.
[§4] Quantitative error metrics, ablation studies, or baseline comparisons (e.g., against standard DMD or POD) should be added to the result figures for the gust-wing and jet-wing cases to allow assessment of improvement over existing decompositions.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive review and positive assessment of the work's potential. We address the single major comment below, proposing targeted revisions to strengthen the presentation of the causality claims.

read point-by-point responses

Referee: [Abstract and §4.3] The central causality claim for the turbulent-wake example (abstract and §4.3) rests on the assertion that the time-varying mode decomposition identifies structures whose influence on lift is causal. Snapshot data alone yields joint statistics; without reported interventional tests (e.g., targeted mode perturbation, time-shuffling, or surrogate-data controls that preserve correlations while destroying directed temporal influence), the extracted relationship remains consistent with non-causal predictive modeling. This issue is load-bearing for the claim that the method extracts causality rather than correlation.

Authors: We agree that interventional or surrogate tests provide stronger validation for directed causality. Our information-theoretic time-varying mode decomposition quantifies directed information transfer (via measures such as transfer entropy) from vortical structures to future aerodynamic coefficients, which is a standard approach for inferring causal influence in observational dynamical systems. However, we acknowledge that the current results are most accurately described as identifying structures with significant predictive influence on lift that is consistent with causal mechanisms in the flow. In the revised version we will (i) moderate the language in the abstract and §4.3 to emphasize information-theoretic directed influence rather than interventional causality, (ii) add surrogate-data controls that preserve spatial correlations while destroying temporal ordering, and (iii) include a brief discussion of the distinction between observational and interventional causality. These additions directly address the concern while preserving the method's core contribution. revision: yes

Circularity Check

0 steps flagged

No circularity: method applies standard information-theoretic decomposition plus CNN/autoencoder to snapshots without reducing predictions to fitted inputs or self-citations

full rationale

The paper presents a time-varying mode decomposition based on information-theoretic machine learning, combined with convolutional networks and autoencoders, to extract vortical structures from aerodynamic snapshot data. No equations or procedures in the provided abstract or description reduce a claimed prediction or causal extraction to a fitted parameter or self-referential definition by construction. The approach is applied to example flows (gust interactions, jet-wing, turbulent wake) as an empirical technique rather than a closed derivation. Central claims rest on the interpretability of extracted modes for future aerodynamic coefficients, but these are not shown to be statistically forced by the input data fitting itself. Self-contained against external benchmarks with no load-bearing self-citation chains or ansatz smuggling visible.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that snapshot data of aerodynamic flows contains extractable causal information via information-theoretic decomposition; no free parameters or invented entities are identifiable from the abstract alone.

axioms (1)

domain assumption Snapshot data of unsteady flows contains sufficient information to extract time-dependent causal vortical structures via information-theoretic decomposition
Invoked in the description of the time-varying mode decomposition technique applied to aerodynamic examples

pith-pipeline@v0.9.0 · 5717 in / 1233 out tokens · 83872 ms · 2026-05-21T15:03:54.001650+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith.Foundation.RealityFromDistinction reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

The informative component is defined as the state that maximizes mutual information with λ at a future time stamp... H(λ|qI)=0... I(qR;qI)=0 (eqs. 2-4, §2)
IndisputableMonolith.Cost.FunctionalEquation washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

w∗=argminw ||q−qI||² + β||I(qR;qI)||² (eq. 6); convolutional autoencoder / CNN extractor

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

27 extracted references · 27 canonical work pages

[1]

Potential flow and forces for incompressible viscous flow.Proc

Chien-Cheng Chang. Potential flow and forces for incompressible viscous flow.Proc. Roy. Soc. A, 437(1901):517– 525, 1992

work page 1901
[2]

¯Otomo, P

S. ¯Otomo, P. Gehlert, H. Babinsky, and J. Li. V ortex force map method to estimate unsteady forces from snapshot flowfield measurements.Exp. Fluids, 66(3):1–13, 2025

work page 2025
[3]

J. L. Lumley. The structure of inhomogeneous turbulent flows.Atmos. Turbul. Radio Wave Propag., pages 166–178, 1967

work page 1967
[4]

P. J. Schmid. Dynamic mode decomposition of numerical and experimental data.J. Fluid Mech., 656:5–28, 2010

work page 2010
[5]

S. L. Brunton, B. R. Noack, and P. Koumoutsakos. Machine learning for fluid mechanics.Annu. Rev. Fluid Mech., 52(1):477–508, 2020

work page 2020
[6]

Taira, G

K. Taira, G. Rigas, and K. Fukami. Machine learning in fluid dynamics: A critical assessment.Phys. Rev. Fluids, 10(9):090701, 2025

work page 2025
[7]

Cremades, S

A. Cremades, S. Hoyas, and R. Vinuesa. Classically studied coherent structures only paint a partial picture of wall-bounded turbulence.Nat. Commun., 16(1):10189, 2025

work page 2025
[8]

M. S.-A. A. Cremades, A. Martinez-Sanchez, A. Lozano-Duran, and R. Vinuesa. X-CAL: Explaining latent causality in physical space for fluid mechanics. arXiv:2601.03311, 2026

work page arXiv 2026
[9]

Arranz and A

G. Arranz and A. Lozano-Durán. Informative and non-informative decomposition of turbulent flow fields.J. Fluid Mech., 1000:A95, 2024

work page 2024
[10]

Fukami and R

K. Fukami and R. Araki. Information-theoretic machine learning for time-varying mode decomposition of separated aerodynamic flows.AIAA J., pages 1–9, 2025. 8 APREPRINT- JANUARY28, 2026

work page 2025
[11]

Linot, B

A. Linot, B. Lopez-Doriga, Y . Zhong, and K. Taira. Extracting dominant dynamics about unsteady base flows. Fluid Dyn. Res., 57(3):031401, 2025

work page 2025
[12]

LeCun, L

Y . LeCun, L. Bottou, Y . Bengio, and P. Haffner. Gradient-based learning applied to document recognition.Proc. IEEE, 86(11):2278–2324, 1998

work page 1998
[13]

C. E. Shannon. A mathematical theory of communication.Bell Syst. Tech. J., 27(3):379–423, 1948

work page 1948
[14]

Martínez-Sánchez, G

Á. Martínez-Sánchez, G. Arranz, and A. Lozano-Durán. Decomposing causality into its synergistic, unique, and redundant components.Nat. Commun., 15(1):9296, 2024

work page 2024
[15]

Huang, D

C.-W. Huang, D. Krueger, A. Lacoste, and A. Courville. Neural autoregressive flows. InInternational conference on machine learning, pages 2078–2087. PMLR, 2018

work page 2078
[16]

G. E. Hinton and R. R. Salakhutdino. Reducing the dimensionality of data with neural networks.Science, 313(5786):504–507, 2006

work page 2006
[17]

Fukami and K

K. Fukami and K. Taira. Grasping extreme aerodynamics on a low-dimensional manifold.Nat. Commun., 14(1):6480, 2023

work page 2023
[18]

Towne, S

A. Towne, S. T. M. Dawson, G. A. Brès, A. Lozano-Durán, T. Saxton-Fox, A. Parthasarathy, A. R. Jones, H. Biler, C.-A. Yeh, H. D. Patel, and Taira K. A database for reduced-complexity modeling of fluid flows.AIAA J., 61(7):2867–2892, 2023

work page 2023
[19]

Q. Liu, L. C. Trujillo Corona, D. Espinoza, F. Shu, and A. Gross. Design and dimensional transfer of reinforcement learning-based closed-loop airfoil flow control.Theor . Comput. Fluid Dyn., 39:49, 2025

work page 2025
[20]

Fukami, H

K. Fukami, H. Nakao, and K. Taira. Data-driven transient lift attenuation for extreme vortex gust–airfoil interactions.J. Fluid Mech., 992:A17, 2024

work page 2024
[21]

G. I. Taylor. On the dissipation of eddies.Meteorology, Oceanography and Turbulent Flow, pages 96–101, 1918

work page 1918
[22]

Biler, G

H. Biler, G. Sedky, A. R. Jones, M. Saritas, and O. Cetiner. Experimental investigation of transverse and vortex gust encounters at low Reynolds numbers.AIAA J., 59(3):786–799, 2021

work page 2021
[23]

Andreu-Angulo, H

I. Andreu-Angulo, H. abinsky, H. Biler, G. Sedky, and A. R. Jones. Effect of transverse gust velocity profiles. AIAA J., 58(12):5123–5133, 2020

work page 2020
[24]

Fukagata and K

K. Fukagata and K. Fukami. Compressing fluid flows with nonlinear machine learning: mode decomposition, latent modeling, and flow control.Fluid Dyn. Res., 57(4):041401, 2025

work page 2025
[25]

Smith, K

L. Smith, K. Fukami, G. Sedky, A. Jones, and K. Taira. A cyclic perspective on transient gust encounters through the lens of persistent homology.J. Fluid Mech., 980:A18, 2024

work page 2024
[26]

Fukami, L

K. Fukami, L. Smith, and K. Taira. Extreme vortex-gust airfoil interactions at reynolds number 5000.Phys. Rev. Fluids, 10(8):084703, 2025

work page 2025
[27]

Fujino, Y

J. Fujino, Y . Motoori, and S. Goto. Hierarchy of coherent vortices in turbulence behind a cylinder.J. Fluid Mech., 975:A13, 2023. 9

work page 2023

[1] [1]

Potential flow and forces for incompressible viscous flow.Proc

Chien-Cheng Chang. Potential flow and forces for incompressible viscous flow.Proc. Roy. Soc. A, 437(1901):517– 525, 1992

work page 1901

[2] [2]

¯Otomo, P

S. ¯Otomo, P. Gehlert, H. Babinsky, and J. Li. V ortex force map method to estimate unsteady forces from snapshot flowfield measurements.Exp. Fluids, 66(3):1–13, 2025

work page 2025

[3] [3]

J. L. Lumley. The structure of inhomogeneous turbulent flows.Atmos. Turbul. Radio Wave Propag., pages 166–178, 1967

work page 1967

[4] [4]

P. J. Schmid. Dynamic mode decomposition of numerical and experimental data.J. Fluid Mech., 656:5–28, 2010

work page 2010

[5] [5]

S. L. Brunton, B. R. Noack, and P. Koumoutsakos. Machine learning for fluid mechanics.Annu. Rev. Fluid Mech., 52(1):477–508, 2020

work page 2020

[6] [6]

Taira, G

K. Taira, G. Rigas, and K. Fukami. Machine learning in fluid dynamics: A critical assessment.Phys. Rev. Fluids, 10(9):090701, 2025

work page 2025

[7] [7]

Cremades, S

A. Cremades, S. Hoyas, and R. Vinuesa. Classically studied coherent structures only paint a partial picture of wall-bounded turbulence.Nat. Commun., 16(1):10189, 2025

work page 2025

[8] [8]

M. S.-A. A. Cremades, A. Martinez-Sanchez, A. Lozano-Duran, and R. Vinuesa. X-CAL: Explaining latent causality in physical space for fluid mechanics. arXiv:2601.03311, 2026

work page arXiv 2026

[9] [9]

Arranz and A

G. Arranz and A. Lozano-Durán. Informative and non-informative decomposition of turbulent flow fields.J. Fluid Mech., 1000:A95, 2024

work page 2024

[10] [10]

Fukami and R

K. Fukami and R. Araki. Information-theoretic machine learning for time-varying mode decomposition of separated aerodynamic flows.AIAA J., pages 1–9, 2025. 8 APREPRINT- JANUARY28, 2026

work page 2025

[11] [11]

Linot, B

A. Linot, B. Lopez-Doriga, Y . Zhong, and K. Taira. Extracting dominant dynamics about unsteady base flows. Fluid Dyn. Res., 57(3):031401, 2025

work page 2025

[12] [12]

LeCun, L

Y . LeCun, L. Bottou, Y . Bengio, and P. Haffner. Gradient-based learning applied to document recognition.Proc. IEEE, 86(11):2278–2324, 1998

work page 1998

[13] [13]

C. E. Shannon. A mathematical theory of communication.Bell Syst. Tech. J., 27(3):379–423, 1948

work page 1948

[14] [14]

Martínez-Sánchez, G

Á. Martínez-Sánchez, G. Arranz, and A. Lozano-Durán. Decomposing causality into its synergistic, unique, and redundant components.Nat. Commun., 15(1):9296, 2024

work page 2024

[15] [15]

Huang, D

C.-W. Huang, D. Krueger, A. Lacoste, and A. Courville. Neural autoregressive flows. InInternational conference on machine learning, pages 2078–2087. PMLR, 2018

work page 2078

[16] [16]

G. E. Hinton and R. R. Salakhutdino. Reducing the dimensionality of data with neural networks.Science, 313(5786):504–507, 2006

work page 2006

[17] [17]

Fukami and K

K. Fukami and K. Taira. Grasping extreme aerodynamics on a low-dimensional manifold.Nat. Commun., 14(1):6480, 2023

work page 2023

[18] [18]

Towne, S

A. Towne, S. T. M. Dawson, G. A. Brès, A. Lozano-Durán, T. Saxton-Fox, A. Parthasarathy, A. R. Jones, H. Biler, C.-A. Yeh, H. D. Patel, and Taira K. A database for reduced-complexity modeling of fluid flows.AIAA J., 61(7):2867–2892, 2023

work page 2023

[19] [19]

Q. Liu, L. C. Trujillo Corona, D. Espinoza, F. Shu, and A. Gross. Design and dimensional transfer of reinforcement learning-based closed-loop airfoil flow control.Theor . Comput. Fluid Dyn., 39:49, 2025

work page 2025

[20] [20]

Fukami, H

K. Fukami, H. Nakao, and K. Taira. Data-driven transient lift attenuation for extreme vortex gust–airfoil interactions.J. Fluid Mech., 992:A17, 2024

work page 2024

[21] [21]

G. I. Taylor. On the dissipation of eddies.Meteorology, Oceanography and Turbulent Flow, pages 96–101, 1918

work page 1918

[22] [22]

Biler, G

H. Biler, G. Sedky, A. R. Jones, M. Saritas, and O. Cetiner. Experimental investigation of transverse and vortex gust encounters at low Reynolds numbers.AIAA J., 59(3):786–799, 2021

work page 2021

[23] [23]

Andreu-Angulo, H

I. Andreu-Angulo, H. abinsky, H. Biler, G. Sedky, and A. R. Jones. Effect of transverse gust velocity profiles. AIAA J., 58(12):5123–5133, 2020

work page 2020

[24] [24]

Fukagata and K

K. Fukagata and K. Fukami. Compressing fluid flows with nonlinear machine learning: mode decomposition, latent modeling, and flow control.Fluid Dyn. Res., 57(4):041401, 2025

work page 2025

[25] [25]

Smith, K

L. Smith, K. Fukami, G. Sedky, A. Jones, and K. Taira. A cyclic perspective on transient gust encounters through the lens of persistent homology.J. Fluid Mech., 980:A18, 2024

work page 2024

[26] [26]

Fukami, L

K. Fukami, L. Smith, and K. Taira. Extreme vortex-gust airfoil interactions at reynolds number 5000.Phys. Rev. Fluids, 10(8):084703, 2025

work page 2025

[27] [27]

Fujino, Y

J. Fujino, Y . Motoori, and S. Goto. Hierarchy of coherent vortices in turbulence behind a cylinder.J. Fluid Mech., 975:A13, 2023. 9

work page 2023