Transfer learning-enhanced deep reinforcement learning for aerodynamic airfoil optimisation subject to structural constraints

David Ramos; Eusebio Valero; Gonzalo Rubio; Lucas Lacasa

REVIEW 4 major objections 6 minor 51 references

Transfer learning-enhanced deep reinforcement learning for aerodynamic airfoil optimisation subject to structural constraints

T0 review · 4 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read A reinforcement-learning airfoil designer pre-trained on a fast neural surrogate and fine-tuned on the accurate XFoil solver reaches nearly the same lift-to-drag improvement as full-XFoil training while cutting estimated solver time by…

desk verdict A credible, practical DRL airfoil-optimization paper with a systematic comparison of transfer-learning strategies, but the hybrid-case claim of 'comparable performance' overgeneralizes and needs qualification. read the letter →

arxiv 2505.02634 v2 pith:6R3PRWST submitted 2025-05-05 cs.LG physics.comp-ph

David Ramos , Lucas Lacasa , Eusebio Valero , Gonzalo Rubio This is my paper · ORCID

classification cs.LGphysics.comp-ph

keywords transferlearningdeepreinforcementairfoilshapeoptimisationlift-to-dragratiostructuralconstraintProximalPolicysurrogateaerodynamicmodelParticleSwarm

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that a deep reinforcement learning (DRL) agent that reshapes airfoils can be trained for a fraction of the usual cost by first pre-training the policy with a fast neural-network surrogate for the aerodynamics and then fine-tuning it with the slower, accurate panel-method solver XFoil. The optimisation target is the lift-to-drag ratio $C_L/C_D$, optionally combined with a structural requirement — preserving the airfoil's maximum thickness — that is folded into the reward through a Gaussian penalty. On the full UIUC evaluation set, the best transfer-learning agent reaches a mean lift-to-drag improvement of $136 \pm 44$, very close to the $140 \pm 49$ of the agent trained entirely with XFoil, while the estimated solver time drops by about 86% in the purely aerodynamic case and up to about 94% in the hybrid case. The paper also argues that a single trained agent outperforms Particle Swarm Optimisation in aerodynamic improvement and is orders of magnitude faster per airfoil at evaluation time. If these results hold, surrogate-pretrained DRL turns airfoil optimisation into a one-time training cost and makes room for expensive high-fidelity solvers inside the loop.

What carries the argument

The load-bearing machinery is a two-solver training pipeline with a scalarised reward. The reward at step $i$, $R_i = \lambda_i\kappa_i(C_L/C_D)_i - \lambda_{i-1}\kappa_{i-1}(C_L/C_D)_{i-1}$, combines the differential gain in lift-to-drag with a Gaussian structural regulariser $\lambda_i = $e^{{-\sigma(x_i-1)^2}}$$ acting on the maximum-thickness ratio $x_i$, so one scalar drives both aerodynamic gain and thickness preservation. The transfer mechanism is the weight handover: the PPO actor-critic policy pre-trained on the fast surrogate NeuralFoil (4 ms per call) is re-initialised on XFoil (73 ms per call) with all weights shared and training continued (strategy #1), plus an added entropy term that pushes the fine-tuned agent to re-explore the design space. The geometry space is bounded by the 18-parameter CST representation, and each episode starts from a randomly chosen member of a 20-airfoil NACA set, so the learned policy applies to many shapes rather than a single one. The whole argument works by shifting most of the learning onto the cheap solver and keeping only a short corrective phase on the accurate one.

What would settle it

Time the full training runs end-to-end on identical hardware instead of estimating solver cost from step counts times nominal per-call times, and re-run the PSO baseline with hyperparameters tuned for the same XFoil budget on the same UIUC airfoils. The quantitative claims stand or fall on two numbers: the transfer-learning saving staying near 86% of real wall-clock time, and tuned PSO failing to close the gap to the DRL improvement of $141\pm48$ against PSO's $105\pm50$.

Watch

Extended reading notes

Core claim

The central discovery, stated on the paper's own terms, is that transfer learning from a cheap surrogate removes most of the expensive-solver burden of DRL airfoil optimisation without sacrificing measured performance. The agent works in an 18-parameter Class-Shape Transformation (CST) representation and receives, at each step, the differential reward $R_i = \lambda_i\kappa_i(C_L/C_D)_i - \lambda_{i-1}\kappa_{i-1}(C_L/C_D)_{i-1}$, where $\kappa$ is the NeuralFoil confidence (set to 1 when XFoil is the solver) and $\lambda_i = $e^{{-\sigma(x_i-1)^2}}$$ penalises departures of the maximum-thickness ratio $x_i = MT_i/MT_0$ from 1, scalarising aerodynamics and structural preservation into one objective. An agent pre-trained for 26,312 steps with NeuralFoil and fine-tuned for 10,240 steps with XFoil (transfer strategy #1, which copies all weights and continues training) scores an improvement of $136 \pm 44$ on the UIUC evaluation set versus $140 \pm 49$ for the agent trained from scratch with 81,920 XFoil steps, while the estimated solver time falls from 5,980 s to 853 s, an 86% reduction. The same pipeline in the hybrid case cuts estimated time by up to about 94% at the price of roughly 10% lower aerodynamic improvement, and the paper reports that DRL dominates PSO in per-airfoil lift-to-drag improvement while being about $1.65 \times $10^{5}$$ times faster at inference, with the caveat that PSO preserves maximum thickness almost exactly whereas DRL does so only approximately.

Load-bearing premise

The comparison against classical optimisation hinges on Xoptfoil2's default, untuned Particle Swarm settings being a reasonably strong baseline; the paper itself states that thorough PSO hyperparameter tuning could shift those results.

Editorial extensions

If this is right

A transfer-learning-trained agent performs very close to the fully XFoil-trained agent on the whole UIUC set (improvement $136\pm44$ vs $140\pm49$; best $236(28)$ vs $241(37)$), so surrogate pre-training is nearly a free speedup rather than a performance tax.
Because the trained policy optimises a new airfoil without any solver call (about 0.015 s per airfoil on one CPU thread), DRL is roughly $1.65\times10^5$ times faster than single-process PSO per airfoil, and the advantage grows with the number of airfoils to be shaped.
The structural regulariser $\sigma$ is a tunable trade-off knob: raising it from 0 to 1000 lowers mean improvement from $140\pm49$ to $54\pm35$ while cutting mean maximum-thickness deviation from $64\pm22\%$ to $7\pm8\%$, so practitioners can set the aerodynamics-versus-structure balance without re-engineering the reward.
The paper's own conclusion extends the surrogate-first, solver-second recipe to more expensive settings — CFD-based optimisation and three-dimensional wing design — where the relative saving from pre-training would be even larger.

Reading between the lines

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

The 86–94% saving is computed by multiplying training step counts by nominal per-call solver times (73 ms for XFoil, 4 ms for NeuralFoil), not by measuring end-to-end wall-clock training; a direct timing study is the natural next check, since policy updates and evaluation overhead are excluded from the figure.
The hybrid comparison is asymmetric: PSO enforces the thickness constraint almost exactly (mean deviation $\Delta MT = 0.67\%$) while DRL enforces it softly ($\Delta MT = 11.5\%$ at $\sigma=15$); adding an exact thickness constraint to the DRL reward, or a soft penalty to PSO, would put the two on equal footing and isolate the algorithm comparison from the constraint-enforcement comparison.
All training and evaluation happen at one operating point ($AoA=2^\circ$, $Ma=0.5$, $Re=10^6$); since the reward already downweights low-confidence NeuralFoil predictions via $\kappa$, a testable prediction is that the transfer-learning time saving shrinks at off-design conditions where the surrogate is less certain and the fine-tuning phase has to work harder.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

The headline: this is the most systematic comparison of transfer-learning strategies for DRL airfoil optimization I have seen, and the empirical core—measured improvements, statistical tests, evaluation on a large airfoil set—is largely credible. The hybrid-case claim of 'comparable performance' is where it overreaches.

What is actually new: the paper compares four weight-transfer strategies (full copy, partial copy, frozen/fine-tuned splits) for pretraining on NeuralFoil and fine-tuning on XFoil, and measures time reduction against a no-TL baseline. That direct comparison is not in the prior DRL airfoil work they cite. The Gaussian max-thickness penalty is a simple but sensible way to scalarize the aerodynamic/structural trade-off, and the sensitivity sweep over sigma is useful. Credit also for releasing code in pyLOM, for using Mann-Whitney U and paired sign tests rather than just eyeballing means, and for honest discussion of PSO's untuned defaults.

The soft spots, in proportion:

1. Equation (2) has a sign error: STATE <- -STATE + alpha*ACTION flips the state, which would make the reported learning impossible. This is almost certainly a typo, but as written it is wrong and will confuse anyone implementing the method.

2. The evaluated-airfoil counts do not reconcile. Table II reports 1456 airfoils for the pure-aerodynamic DRL agent; Table IV and Table V report 1982 for the same setup. The difference is too large to wave off as XFoil non-convergence. The authors need to explain which number is correct.

3. The abstract's unqualified claim that TL achieves 'comparable performance' breaks down in the hybrid case at sigma >= 30. The stress-test note holds up: comparing Table IV with Table VI, improvement drops by 16% at sigma=30, 26% at sigma=100, and 52% at sigma=1000, with best-median dropping 8-20%. The text says 'only slightly lower... around a 10%,' which the paper's own tables contradict. The 'up to 94% time reduction' is largest precisely where the performance loss is largest. This is not a minor quibble; it is an internal inconsistency between the reported tables and the qualitative summary.

4. The PSO baseline uses Xoptfoil2 defaults with no tuning. The authors acknowledge this. It weakens the 'DRL outperforms PSO' headline only modestly because the speed advantage is enormous, but the aerodynamic-improvement comparison might shift with tuned PSO.

5. The solver-time comparison multiplies step counts by assumed per-call times (73 ms XFoil, 4 ms NeuralFoil) rather than measuring wall-clock. For a training-efficiency claim, that is a reasonable proxy but should be labeled as such more prominently.

The paper deserves a serious referee. The central empirical claim—that transfer learning from a surrogate cuts training cost by roughly 86% in the pure-aerodynamic case with near-equal performance—is credible and useful. The hybrid overgeneralization is fixable by qualifying the claim to low sigma and pure-aerodynamic settings, and by correcting the internal numbers. My recommendation: send it to peer review, but the referee should demand those corrections before acceptance.

Referee Report

4 major / 6 minor

Summary. The paper proposes a transfer learning (TL) enhanced deep reinforcement learning (DRL) framework for two-dimensional airfoil shape optimisation, using Proximal Policy Optimization (PPO) over an 18-parameter CST representation. The reward scalarises a multi-objective problem: it combines the lift-to-drag ratio with a Gaussian penalty that preserves the airfoil's maximum thickness as a structural-integrity proxy. The agent is trained either directly with XFoil or pre-trained with the NeuralFoil surrogate and then fine-tuned with XFoil under four transfer strategies. The authors evaluate the resulting policies on the UIUC/Aerosandbox airfoil dataset and compare them with Particle Swarm Optimization (PSO) as implemented in Xoptfoil2. The central empirical claim is that TL-enhanced DRL achieves performance close to full XFoil training while reducing estimated solver time by about 86% in the purely aerodynamic case and up to about 94% in the hybrid aerodynamic/structural case.

Significance. If the central claims hold after revision, the paper makes a useful contribution to data-driven aerodynamic shape optimisation. It demonstrates a concrete multi-fidelity DRL pipeline, evaluates it on a large airfoil dataset, includes several transfer-learning strategies including a non-converging control, and reports statistical comparisons. The paper also provides detailed PPO hyperparameters, solver settings, CST bounds, and states that the framework is available through the open-source pyLOM library, which supports reproducibility. The main qualitative result—that a cheap surrogate can pre-train a policy that is then fine-tuned with a high-fidelity panel method—is plausible and of broad interest. However, the headline comparability claim in the hybrid case is not supported by the paper's own tables, one methodological equation appears to contain a sign error, and the PSO baseline is explicitly untuned; these issues must be addressed before the empirical conclusions can be accepted.

major comments (4)

[Section II.B.1, Eq. (2)] The state update is written as STATE ← −STATE + α ⊙ ACTION. As written, this replaces the current state by its negative plus a bounded action increment, which is not an incremental geometry modification and contradicts the surrounding text describing how the action 'performs the change in the parameters'. If the implementation follows Eq. (2), the learned policy operates on a different dynamical system than described; if the implementation uses STATE + α ⊙ ACTION, the equation must be corrected. Please fix this sign error and state explicitly which update rule was actually used in the experiments.
[Section III.C, hybrid optimisation case, Tables IV and VI] The statement that 'the improvement in aerodynamic efficiency is only slightly lower for every value of σ (around a 10%)' is contradicted by the pointwise comparison of Table VI with Table IV. For σ=30, improvement drops from 94±41 to 79±36 (16%) and best-median from 186 to 171 (8%); for σ=100, improvement drops from 84±42 to 62±35 (26%) and best-median from 176 to 151 (14%); for σ=1000, improvement drops from 54±35 to 26±22 (52%) and best-median from 137 to 110 (20%). These are also the cases with the largest time reductions (92.6%, 94.8%, and 89.7%). The claim of 'comparable performance' should therefore be restricted to the purely aerodynamic case and to small σ (roughly σ≤20), or the results should be reframed as a trade-off in which high values of σ sacrifice more performance in exchange for larger time savings.
[Section III.A and Table V] The number of evaluated airfoils for the same DRL/XFoil purely aerodynamic setup is 1456 in Table II but 1982 in Table V. The text attributes differences in evaluated airfoils to XFoil non-convergence and says the number 'fluctuates a bit', but a 526-airfoil gap (about 36%) is far larger than that wording suggests and changes the evaluation set across comparisons. Please report the evaluation-set construction, give per-agent convergence counts, and either use a fixed common evaluation set or show explicitly that the qualitative conclusions are insensitive to the differing sets.
[Section III.A, PSO baseline] The comparison with PSO uses Xoptfoil2 default parameters without tuning, while the DRL agents have had hyperparameters tuned, and the authors explicitly state that they cannot guarantee that PSO results would not change after thorough hypertuning. Because the headline claim that DRL outperforms PSO depends on this baseline, the statement should be softened or supplemented with a sensitivity check (for example, a reasonable PSO budget or tuned swarm parameters). As written, the comparison is not a fully fair apples-to-apples comparison, and the reader cannot assess how much of the DRL advantage is due to the DRL method itself versus the untuned PSO configuration.

minor comments (6)

[Section I] There is a typo in the introduction: 'As advanced' should be 'As noted' or 'As mentioned'.
[Section III.A] The word 'certifying' is used twice to describe single evaluation results; 'indicating' or 'consistent with' would be more accurate given the stochastic and baseline-dependent nature of the comparison.
[Section III.A, timing comparison] The timing paragraph says DRL takes 0.0147 s/airfoil for 'all 1566 airfoils', while Table II reports 1456 evaluated airfoils for the same DRL setup; clarify which evaluation set is used for the timing comparison.
[Section III.C, solver times] The statement that a call to XFoil takes 73 ms and a call to the smallest NeuralFoil model takes 4 ms is attributed to 'NeuralFoil's GitHub'; please provide the exact version or URL and state whether these times include any environment overhead or are pure solver-call times.
[Appendix B] The hyperparameter section says an Optuna search was conducted for the NeuralFoil agent but does not report the resulting configuration or comparison; add those results to support the claim that the Optuna configuration 'closely aligns' with the empirically tuned parameters.
[Conclusions] The conclusion restates the 86% time-reduction figure but does not mention the σ-dependent performance loss in the hybrid case; it should be aligned with the corrected, qualified claim.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the central claims are empirical evaluations against external XFoil/UIUC baselines, with self-citations appearing only in contextual roles.

full rationale

The paper's derivation chain is self-contained in the relevant sense: the DRL agent is trained by maximizing the incremental reward in Eq. (5), where CL/CD comes from XFoil or NeuralFoil and the structural term is the Gaussian kernel of Eq. (4), and performance is then measured on the UIUC/Aerosandbox evaluation set with XFoil (Tables II–VI). No fitted parameter is renamed as a prediction; the transfer-learning comparison is a measured head-to-head of improvement/best/ΔMT and of training-step counts, with the time reduction of Eq. (6) being a bookkeeping ratio of solver calls times published per-call costs (73 ms XFoil, 4 ms NeuralFoil). The σ sweep is a hyperparameter study of the imposed reward, not a fitted input used to generate the reported metrics. The PSO comparison is acknowledged in Sec. III A to use untuned Xoptfoil2 defaults ('Since the main point of this paper is DRL, we do not conduct a thorogh hyperparameter optimisation for PSO...'), which is a baseline-fairness limitation rather than circularity. Self-citations exist (Refs. 25, 35, 39, 40), but they are contextual: the Gaussian regularizer is said to be 'inspired by' Refs. [39,40] (co-authored by Rubio), and pyLOM and the certification-pipeline reference are not load-bearing to any derivation. The skeptical point about Table VI is a real internal-consistency issue—the text's 'around a 10%' lower improvement is contradicted at σ=30 (−16%), σ=100 (−26%), and σ=1000 (−52%)—but numerical inconsistency is a correctness concern, not circularity. No self-definitional, fitted-input-renamed-as-prediction, or uniqueness-imported step could be exhibited.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The central claims rest on the fidelity of XFoil and NeuralFoil, the expressiveness of CST parameterization, the validity of maximum thickness as a structural proxy, the scalarized reward formulation, and the fairness of the PSO baseline. No new physical entities are introduced.

free parameters (6)

sigma (Gaussian penalty strength) = scanned values: 0,2,5,10,15,20,30,100,1000; headline comparison uses 15
Controls the trade-off between aerodynamic efficiency and maximum-thickness preservation; the paper states the choice of sigma=15 is arbitrary and selected for illustration.
gamma (PPO discount factor) = 0.3
Tuned empirically and via Optuna; the paper reports gamma=0.3 yields faster learning than higher values.
PPO clip range = 0.2, 0.3, or 0.6 depending on training stage
Increased from the default 0.2 to improve convergence in some agents.
PPO entropy coefficient = 0, 0.001, or 0.005 depending on agent
Adjusted to encourage exploration after pretraining.
Episode max length = 100
Chosen to balance exploration and number of training episodes; affects the alpha step size in Eq. (2).
CST parameter bounds = Table VII
Selected based on empirical testing to avoid extreme or unrealistic geometries.

assumptions (6)

domain assumption XFoil accurately computes CL/CD for 2D subsonic airfoils at AoA=2, Ma=0.5, Re=1e6.
Used as the ground-truth solver for both training and evaluation; all performance claims depend on XFoil fidelity.
domain assumption NeuralFoil is a sufficiently accurate surrogate of XFoil to provide a useful pretraining signal.
Cited as trained on millions of XFoil simulations; the transfer-learning strategy relies on this.
domain assumption CST parameterization with the given bounds can represent practical airfoil shapes.
Reduces design-space dimensionality; the agent can only find shapes within this parameterization.
domain assumption Maximum thickness is a sufficient proxy for structural integrity.
The paper explicitly states this is a simple proxy; if false, the hybrid optimization does not actually preserve structural integrity.
ad hoc to paper The scalarized reward lambda * kappa * CL/CD adequately represents the multi-objective aerodynamic/structural optimization problem.
The Gaussian kernel and product form are chosen by the authors, not derived; the paper notes Pareto optimality is not guaranteed.
ad hoc to paper PSO with Xoptfoil2 default parameters is a fair baseline for comparison.
No tuning was performed; authors acknowledge results may differ after hypertuning, so the DRL-vs-PSO comparison is conditional on this assumption.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Transfer learning-enhanced deep reinforcement learning for aerodynamic airfoil optimisation subject to structural constraints." pith.science (2026). https://pith.science/paper/6R3PRWST

@misc{pith2026250502634,
  author       = {Pith},
  title        = {Pith review of: Transfer learning-enhanced deep reinforcement learning for aerodynamic airfoil optimisation subject to structural constraints},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6R3PRWST}},
  note         = {Machine review of arXiv:2505.02634}
}

abstract

The main objective of this paper is to introduce a transfer learning-enhanced deep reinforcement learning (DRL) methodology that is able to optimise the geometry of any airfoil based on concomitant aerodynamic and structural integrity criteria. To showcase the method, we aim to maximise the lift-to-drag ratio $C_L/C_D$ while preserving the structural integrity of the airfoil -- as modelled by its maximum thickness -- and train the DRL agent using a list of different transfer learning (TL) strategies. The performance of the DRL agent is compared with Particle Swarm Optimisation (PSO), a traditional gradient-free optimisation method. Results indicate that DRL agents are able to perform purely aerodynamic and hybrid aerodynamic/structural shape optimisation, that the DRL approach outperforms PSO in terms of computational efficiency and aerodynamic improvement, and that the TL-enhanced DRL agent achieves performance comparable to the DRL one, while further saving substantial computational resources.

Figures

Figures reproduced from arXiv: 2505.02634 by the authors.

**Figure 1.** Basic DRL training diagram. 3. RL Specs III: Reward function Once the episode has been defined, we need to specify the reward function, which estimates the quality of any given action. For pedagogical reasons, we now outline the construction of such reward function in a step-by-step way. As a starting point, we consider the maximisation of the lift-to-drag ratio CL/CD –where both aerodynamic coefficients are com… view at source ↗

**Figure 2.** The Gaussian regularisation term λ = exp(−σ[x−1] 2 ) as a function of the hyperparameter σ. This term penalises geometries where the airfoil’s maximum thickness deviates from the initial one, as captured by a maximum thickness quotient x that deviates from one. 4. RL specs IV: the agent To train an agent with reinforcement learning, one needs to choose the specific DRL algorithm. This choice may depend, among other… view at source ↗

**Figure 4.** Illustration of a DRL policy (20 steps) on [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗

Figures from the paper (3 more)

**Figure 5.** Figure 5: Boxplot (showing medians, interquantile range [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]

**Figure 7.** Figure 7: Scatter plot of the DRL-based hybrid optimisation, [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]

**Figure 8.** Figure 8: Episode mean reward during training with different [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 42 canonical work pages

[1]

author author F. K. \ Lu ,\ title Airfoil geometries in various flight vehicles , \ in\ https://eaglepubs.erau.edu/introductiontoaerospaceflightvehicles/chapter/airfoil-geometries/ booktitle Introduction to Aerospace Flight Vehicles \ ( publisher Embry-Riddle Aeronautical University Press ,\ year 2022 )\ Chap.\ chapter Airfoil Geometries ,\ note accessed:...

work page 2022
[2]

Bak , author N

author author C. Bak , author N. Gaudern , author F. Zahle , \ and\ author T. Vronsky ,\ title title Airfoil design: Finding the balance between design lift and structural stiffness , \ 10.1088/1742-6596/524/1/012017 journal journal Journal of Physics: Conference Series \ volume 524 ,\ pages 012017 ( year 2014 ) NoStop

work page doi:10.1088/1742-6596/524/1/012017
[3]

Skinner \ and\ author H

author author S. Skinner \ and\ author H. Zare-Behtash ,\ title title State-of-the-art in aerodynamic shape optimisation methods , \ @noop journal journal Applied Soft Computing \ volume 62 ,\ pages 933--962 ( year 2018 ) NoStop

work page 2018
[4]

author author P. G. \ et al. ,\ title title A review on deep reinforcement learning for fluid mechanics , \ @noop journal journal arXiv preprint arXiv:1908.04127 \ ( year 2019 ) NoStop

work page Pith review arXiv 1908
[5]

Secanell \ and\ author A

author author M. Secanell \ and\ author A. Suleman ,\ title title Numerical evaluation of optimization algorithms for low-reynolds-number aerodynamic shape optimization , \ @noop journal journal AIAA Journal \ volume 43 ,\ pages 2262--2267 ( year 2005 ) NoStop

work page 2005
[6]

Lyu , Z.; Xu \ and\ author J

author author Z. Lyu , Z.; Xu \ and\ author J. Martins ,\ title title Benchmarking optimization algorithms for wing aerodynamic design optimization , \ in\ @noop booktitle 8th International Conference on Computational Fluid Dynamics \ ( organization Citeseer ,\ year 2014 ) NoStop

work page 2014
[7]

Pironneau ,\ title title On optimum design in fluid mechanics , \ @noop journal journal Journal of Fluid Mechanics \ volume 64 ,\ pages 97--110 ( year 1974 ) NoStop

author author O. Pironneau ,\ title title On optimum design in fluid mechanics , \ @noop journal journal Journal of Fluid Mechanics \ volume 64 ,\ pages 97--110 ( year 1974 ) NoStop

work page 1974
[8]

author author P. D. \ Frank \ and\ author G. R. \ Shubin ,\ title title A comparison of optimization-based approaches for a model computational aerodynamics design problem , \ @noop journal journal Journal of Computational Physics \ volume 98 ,\ pages 74--89 ( year 1992 ) NoStop

work page 1992

Show all 51 references

[9]

Jameson , author L

author author A. Jameson , author L. Martinelli , \ and\ author N. Pierce ,\ title title Optimum aerodynamic design using the navier-stokes equations , \ @noop journal journal Theoretical and Computational Fluid Dynamics \ volume 10 ,\ pages 213--237 ( year 1998 ) NoStop

1998
[10]

Jichao , author D

author author L. Jichao , author D. Xiaosong , \ and\ author J. R.R.A.Martins ,\ title title Low-reynolds-number airfoil design optimization using deep-learning-based tailored airfoil modes , \ @noop journal journal Aerospace Science and Technology \ volume 121 ,\ pages 567 ( ...

2022
[11]

Viquerat , author J

author author J. Viquerat , author J. Rabault , author A. Kuhnle , author H. Ghraieb , author A. Larcher , \ and\ author E. Hachem ,\ title title Direct shape optimization through deep reinforcement learning , \ @noop journal journal Journal of Computational Physics \ volume 4...

2021
[12]

Bizzarrini , author F

author author N. Bizzarrini , author F. Grasso , author D. P. \ Coiro , et al. ,\ title title Genetic algorithms in wind turbine airfoil design , \ @noop journal journal EWEA, EWEC2011, Bruxelles, Belgium \ volume 14 ( year 2011 ) NoStop

2011
[13]

\ Wang , author B.-Q

author author Y.-Y. \ Wang , author B.-Q. \ Zhang , \ and\ author Y.-C. \ Chen ,\ title title Robust airfoil optimization based on improved particle swarm optimization method , \ @noop journal journal Applied Mathematics and Mechanics \ volume 32 ,\ pages 1245--1254 ( year 201...

2011
[14]

Fourie \ and\ author A

author author P. Fourie \ and\ author A. Groenwold ,\ title title The particle swarm optimization algorithm in size and shape optimization , \ @noop journal journal Structural and Multidisciplinary Optimization \ volume 23 ,\ pages 259--267 ( year 2002 ) NoStop

2002
[15]

Nejat , author P

author author A. Nejat , author P. Mirzabeygi , \ and\ author M. Panahi ,\ title title Airfoil shape optimization using improved multiobjective territorial particle swarm algorithm with the objective of improving stall characteristics , \ @noop journal journal Structural and M...

2014
[16]

Wang , author J

author author J. Wang , author J. Periaux , \ and\ author M. Sefrioui ,\ title title Parallel evolutionary algorithms for optimization problems in aerospace engineering , \ @noop journal journal Journal of Computational and Applied Mathematics \ volume 149 ,\ pages 155--169 ( ...

2002
[17]

Arulkumaran , author M

author author K. Arulkumaran , author M. P. \ Deisenroth , author M. Brundage , \ and\ author A. A. \ Bharath ,\ title title Deep reinforcement learning: A brief survey , \ @noop journal journal IEEE Signal Processing Magazine \ volume 34 ,\ pages 26--38 ( year 2017 ) NoStop

2017
[18]

author author T. P. \ Dussauge , author W. J. \ Sung , author O. J. \ Pinon Fischer , \ and\ author D. N. \ Mavris ,\ title title A reinforcement learning approach to airfoil shape optimization , \ @noop journal journal Scientific Reports \ volume 13 ,\ pages 9753 ( year 2023 ) NoStop

2023
[19]

Lou , author R

author author J. Lou , author R. Chen , author J. Liu , author Y. Bao , author Y. You , \ and\ author Z. Chen ,\ title english title Aerodynamic optimization of airfoil based on deep reinforcement learning , \ 10.1063/5.0137002 journal journal Physics of Fluids \ volume 35 ,\ ...

doi
[20]

author author S. L. \ Brunton , author B. R. \ Noack , \ and\ author P. Koumoutsakos ,\ title title Machine learning for fluid mechanics , \ @noop journal journal Annual Review of Fluid Mechanics \ volume 52 ,\ pages 477--508 ( year 2020 ) NoStop

2020
[21]

Vinuesa , author S

author author R. Vinuesa , author S. L. \ Brunton , \ and\ author B. J. \ McKeon ,\ title title The transformative potential of machine learning for experiments in fluid mechanics , \ @noop journal journal Nature Reviews Physics \ volume 5 ,\ pages 536--545 ( year 2023 ) NoStop

2023
[22]

Vinuesa \ and\ author S

author author R. Vinuesa \ and\ author S. L. \ Brunton ,\ title title Enhancing computational fluid dynamics with machine learning , \ @noop journal journal Nature Computational Science \ volume 2 ,\ pages 358--366 ( year 2022 ) NoStop

2022
[23]

Le Clainche , author E

author author S. Le Clainche , author E. Ferrer , author S. Gibson , author E. Cross , author A. Parente , \ and\ author R. Vinuesa ,\ title title Improving aircraft performance using machine learning: a review , \ @noop journal journal Aerospace Science and Technology \ volum...

2023
[24]

author author S. L. \ Brunton , author J. Nathan Kutz , author K. Manohar , author A. Y. \ Aravkin , author K. Morgansen , author J. Klemisch , author N. Goebel , author J. Buttrick , author J. Poskin , author A. W. \ Blom-Schieber , et al. ,\ title title Data-driven aerospace...

2021
[25]

Lacasa , author A

author author L. Lacasa , author A. Pardo , author P. Arbelo , author M. S \'a nchez , author N. Bascones , author P. Yeste , author A. Mart \' nez-Cava , author G. Rubio , author I. G \'o mez , author E. Valero , et al. ,\ title title Towards certification: A complete statist...

2025
[26]

Liu , author M

author author Z. Liu , author M. Zhang , author D. Sun , author L. Li , \ and\ author G. Chen ,\ title english title A deep reinforcement learning optimization framework for supercritical airfoil aerodynamic shape design , \ 10.1007/s00158-024-03755-5 journal journal Structura...

doi
[27]

Liu , author R

author author J. Liu , author R. Chen , author J. Lou , author H. Wu , author Y. You , \ and\ author Z. Chen ,\ title english title Airfoils optimization based on deep reinforcement learning to improve the aerodynamic performance of rotors , \ 10.1016/j.ast.2023.108737 journal...
[28]

Yan , author J

author author X. Yan , author J. Zhu , author M. Kuang , \ and\ author X. Wang ,\ title title Aerodynamic shape optimization using a novel optimizer based on machine learning techniques , \ https://doi.org/10.1016/j.ast.2019.02.003 journal journal Aerospace Science and Technol...

2019 doi
[29]

Noda , author K

author author T. Noda , author K. Okabayashi , author S. Kimura , author S. Takeuchi , \ and\ author T. Kajishima ,\ title title Optimization of configuration of corrugated airfoil using deep reinforcement learning and transfer learning , \ 10.1063/5.0134198 journal journal AI...

doi
[30]

Bhola , author S

author author S. Bhola , author S. Pawar , author P. Balaprakash , \ and\ author R. Maulik ,\ title title Multi-fidelity reinforcement learning framework for shape optimization , \ https://doi.org/10.1016/j.jcp.2023.112018 journal journal Journal of Computational Physics \ vol...

2023
[31]

author author Y. Y. \ Liu , author J. X. \ Shen , author P. P. \ Yang , \ and\ author X. W. \ Yang ,\ title title A cnn-pinn-drl driven method for shape optimization of airfoils , \ @noop journal journal Engineering Applications of Computational Fluid Mechanics \ volume 19 ,\ ...

2025
[32]

author author H. P. \ Buckley , author B. Y. \ Zhou , \ and\ author D. W. \ Zingg ,\ title title Airfoil optimization using practical aerodynamic design requirements , \ @noop journal journal Journal of Aircraft \ volume 47 ,\ pages 1707--1719 ( year 2010 ) NoStop

2010
[33]

Sharpe \ and\ author R

author author P. Sharpe \ and\ author R. J. \ Hansman ,\ title title Neuralfoil: An airfoil aerodynamics analysis tool using physics-informed machine learning , \ @noop journal journal arXiv preprint arXiv:2503.16323 \ ( year 2025 ) NoStop

2025 arXiv
[34]

author author M. Drela ,\ title title Xfoil: An analysis and design system for low reynolds number airfoils , \ in\ @noop booktitle Low Reynolds Number Aerodynamics: Proceedings of the Conference Notre Dame, Indiana, USA, 5--7 June 1989 \ ( organization Springer ,\ year 1989 )...

1989
[35]

Eiximeno , author A

author author B. Eiximeno , author A. Miró , author B. Begiashvili , author E. Valero , author I. Rodriguez , \ and\ author O. Lehmkuhl ,\ title title Pylom: A hpc open source reduced order model suite for fluid dynamics applications , \ https://doi.org/10.1016/j.cpc.2024.1094...
[36]

author author B. M. \ Kulfan ,\ title title Universal parametric geometry representation method , \ @noop journal journal Journal of aircraft \ volume 45 ,\ pages 142--158 ( year 2008 ) NoStop

2008
[37]

Towers , author A

author author M. Towers , author A. Kwiatkowski , author J. Terry , author J. U. \ Balis , author G. De Cola , author T. Deleu , author M. Goul \ a o , author A. Kallinteris , author M. Krimmel , author A. KG , et al. ,\ title title Gymnasium: A standard interface for reinforc...

2024 arXiv
[38]

note Https://stable-baselines3.readthedocs.io/en/master/guide/rl \_tips.html\#tips-and-tricks-when-creating-a-custom-environment NoStop
[39]

Huergo , author G

author author D. Huergo , author G. Rubio , \ and\ author E. Ferrer ,\ title title A reinforcement learning strategy for p-adaptation in high order solvers , \ @noop journal journal Results in Engineering \ volume 21 ,\ pages 101693 ( year 2024 ) NoStop

2024
[40]

Huergo , author M

author author D. Huergo , author M. de Frutos , author E. Jan \'e , author O. A. \ Marino , author G. Rubio , \ and\ author E. Ferrer ,\ title title Reinforcement learning for anisotropic p-adaptation and error estimation in high-order solvers , \ @noop journal journal arXiv p...

2024 arXiv
[41]

Mnih , author A

author author V. Mnih , author A. P. \ Badia , author M. Mirza , author A. Graves , author T. Lillicrap , author T. Harley , author D. Silver , \ and\ author K. Kavukcuoglu ,\ title title Asynchronous methods for deep reinforcement learning , \ in\ @noop booktitle Internationa...

2016
[42]

Schulman , author F

author author J. Schulman , author F. Wolski , author P. Dhariwal , author A. Radford , \ and\ author O. Klimov ,\ https://arxiv.org/abs/1707.06347 title Proximal policy optimization algorithms , \ ( year 2017 ),\ http://arxiv.org/abs/1707.06347 arXiv:1707.06347 [cs.LG] NoStop

2017 arXiv
[43]

Haarnoja , author A

author author T. Haarnoja , author A. Zhou , author P. Abbeel , \ and\ author S. Levine ,\ title title Soft actor-critic: Off-policy maximum entropy deep reinforcement learning with a stochastic actor , \ in\ @noop booktitle International conference on machine learning \ ( org...

2018
[44]

Fujimoto , author H

author author S. Fujimoto , author H. Hoof , \ and\ author D. Meger ,\ title title Addressing function approximation error in actor-critic methods , \ in\ @noop booktitle International conference on machine learning \ ( organization PMLR ,\ year 2018 )\ pp.\ pages 1587--1596 NoStop

2018
[45]

Raffin , author A

author author A. Raffin , author A. Hill , author A. Gleave , author A. Kanervisto , author M. Ernestus , \ and\ author N. Dormann ,\ title title Stable-baselines3: Reliable reinforcement learning implementations , \ http://jmlr.org/papers/v22/20-1364.html journal journal Jour...

2021
[46]

author author P. D. \ Sharpe ,\ title AeroSandbox: A Differentiable Framework for Aircraft Design Optimization ,\ @noop Master's thesis ,\ school Massachusetts Institute of Technology ( year 2021 ),\ note available at https://dspace.mit.edu/handle/1721.1/140023 NoStop

2021
[47]

Kennedy \ and\ author R

author author J. Kennedy \ and\ author R. Eberhart ,\ title title Particle swarm optimization , \ in\ @noop booktitle Proceedings of ICNN'95-international conference on neural networks ,\ Vol. volume 4 \ ( organization ieee ,\ year 1995 )\ pp.\ pages 1942--1948 NoStop

1995
[48]

Mukesh , author K

author author R. Mukesh , author K. Lingadurai , \ and\ author K. Elamvaluthi ,\ title title Influence of optimization algorithm on airfoil shape optimization of aircraft wings , \ @noop journal journal Applied Mechanics and Materials \ volume 232 ,\ pages 614--619 ( year 2012...

2012
[49]

Hoyos , author J

author author J. Hoyos , author J. H. \ J \' menez , author C. Echavarr \' a , \ and\ author J. P. \ Alvarado ,\ title title Airfoil shape optimization: Comparative study of meta-heuristic algorithms, airfoil parameterization methods and reynolds number impact , \ in\ @noop bo...

2021
[50]

author author D. P. \ Jochen Guenzel ,\ @noop title Xoptfoil2: An optimization tool for airfoil design , \ howpublished https://github.com/jxjo/Xoptfoil2 ( year 2025 ),\ note accessed: 2025-01-15 NoStop

2025
[51]

Akiba , author S

author author T. Akiba , author S. Sano , author T. Yanase , author T. Ohta , \ and\ author M. Koyama ,\ title title Optuna: A next-generation hyperparameter optimization framework , \ in\ @noop booktitle Proceedings of the 25th ACM SIGKDD International Conference on Knowledge...

2019

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.

[1] [1]

author author F. K. \ Lu ,\ title Airfoil geometries in various flight vehicles , \ in\ https://eaglepubs.erau.edu/introductiontoaerospaceflightvehicles/chapter/airfoil-geometries/ booktitle Introduction to Aerospace Flight Vehicles \ ( publisher Embry-Riddle Aeronautical University Press ,\ year 2022 )\ Chap.\ chapter Airfoil Geometries ,\ note accessed:...

work page 2022

[2] [2]

Bak , author N

author author C. Bak , author N. Gaudern , author F. Zahle , \ and\ author T. Vronsky ,\ title title Airfoil design: Finding the balance between design lift and structural stiffness , \ 10.1088/1742-6596/524/1/012017 journal journal Journal of Physics: Conference Series \ volume 524 ,\ pages 012017 ( year 2014 ) NoStop

work page doi:10.1088/1742-6596/524/1/012017

[3] [3]

Skinner \ and\ author H

author author S. Skinner \ and\ author H. Zare-Behtash ,\ title title State-of-the-art in aerodynamic shape optimisation methods , \ @noop journal journal Applied Soft Computing \ volume 62 ,\ pages 933--962 ( year 2018 ) NoStop

work page 2018

[4] [4]

author author P. G. \ et al. ,\ title title A review on deep reinforcement learning for fluid mechanics , \ @noop journal journal arXiv preprint arXiv:1908.04127 \ ( year 2019 ) NoStop

work page Pith review arXiv 1908

[5] [5]

Secanell \ and\ author A

author author M. Secanell \ and\ author A. Suleman ,\ title title Numerical evaluation of optimization algorithms for low-reynolds-number aerodynamic shape optimization , \ @noop journal journal AIAA Journal \ volume 43 ,\ pages 2262--2267 ( year 2005 ) NoStop

work page 2005

[6] [6]

Lyu , Z.; Xu \ and\ author J

author author Z. Lyu , Z.; Xu \ and\ author J. Martins ,\ title title Benchmarking optimization algorithms for wing aerodynamic design optimization , \ in\ @noop booktitle 8th International Conference on Computational Fluid Dynamics \ ( organization Citeseer ,\ year 2014 ) NoStop

work page 2014

[7] [7]

Pironneau ,\ title title On optimum design in fluid mechanics , \ @noop journal journal Journal of Fluid Mechanics \ volume 64 ,\ pages 97--110 ( year 1974 ) NoStop

author author O. Pironneau ,\ title title On optimum design in fluid mechanics , \ @noop journal journal Journal of Fluid Mechanics \ volume 64 ,\ pages 97--110 ( year 1974 ) NoStop

work page 1974

[8] [8]

author author P. D. \ Frank \ and\ author G. R. \ Shubin ,\ title title A comparison of optimization-based approaches for a model computational aerodynamics design problem , \ @noop journal journal Journal of Computational Physics \ volume 98 ,\ pages 74--89 ( year 1992 ) NoStop

work page 1992