The paper investigates whether lower prediction error in machine-learning surrogates for fluid dynamics necessarily implies better physical fidelity. Using paired CFD simulations and experimental PIV data for flow around a NACA4418 airfoil, the authors apply a mean‑preserving input intervention that removes velocity fluctuations from selected flow regions. They find that removing fluctuations from the most energetic 10% of cells alters forecasts more than random removal, indicating sensitivity to fluctuation energy. Additionally, a convolutional neural operator (CNO) shows lower overall velocity‑field error but higher two‑component fluctuation‑energy error compared to a reference, and an output attenuation stress test reveals discrepancies between benchmark errors and domain‑summed fluctuation energy. These results suggest that aggregate prediction scores alone are insufficient to assess physical correctness, motivating the inclusion of complementary physical diagnostics.
arXiv:2606. 16765v1 Announce Type: new Abstract: Evaluating neural operators for 3D turbulent flow requires validated datasets with physical benchmarks.
arXiv:2609.06660v1 Announce Type: cross Abstract: Accurate aerodynamic prediction is critical for designing fuel-efficient and safe transportation systems such as aircraft and automobiles, yet tradit...
The paper introduces a correction framework that grounds a CFD-trained deep learning surrogate model for aerospace aerodynamics using wind‑tunnel pressure‑sensor (PSP) data. By training a correction network on spatially registered PSP measurements at two Mach numbers, the authors adjust the surrogate’s predictions without retraining its core parameters, achieving improved agreement with experimental pressure distributions—especially at the wing suction peak and shock location. The grounded surrogate matches measurements within 2.3–2.7% of the Cp range on unseen angles of attack and outperforms simple interpolation between measured states.
The paper introduces VATO, a Vortex-Force-Aware Transformer Operator designed to predict unsteady separated flows around aerofoils more accurately. VATO couples a Vortex Force Map (VFM) method with a geometry-aware neural operator, offering two variants: VATO‑S, which adds training-only supervision of local VFM force contributions, and VATO‑A, which prioritizes force-relevant source locations for residual cross attention. Evaluated on CFD data for double‑edged‑plate aerofoils, VATO‑S and VATO‑A reduce velocity, pressure, and vorticity errors by up to 15.8%, 7.5%, and 31.2% respectively, and improve aerodynamic force predictions even beyond the training range.
arXiv:2608. 04222v1 Announce Type: cross Abstract: Turbulence is a central testbed for machine learning on physical dynamics because its governing laws are known exactly.
arXiv:2609.17160v1 Announce Type: new Abstract: Machine-learning surrogate models offer a promising alternative to high-fidelity Computational Fluid Dynamics (CFD) simulations for aerodynamic analysi...
arXiv:2606. 16602v1 Announce Type: new Abstract: Neural operator models trained on simulation data often lose accuracy when applied to experimental measurements due to the sim-to-real gap.
Generative AI emulators are increasingly used in scientific domains where we already have strong theory, benchmarks, and physical intuition. This raises a central evaluation and interpretability question: when a foundation-style model can reproduce known continuum dynamics, what internal mechanism supports that behavior, is the internal behaviour consistent with known physics, and how does it relate to where the emulator succeeds or fails?
arXiv:2608. 19632v1 Announce Type: new Abstract: Physics-informed neural networks (PINNs) embed governing partial differential equations directly into the training loss, offering a promising alternative to costly CFD solvers for unsteady flows.
The paper explores using residual learning with an LSTM neural network to enhance unsteady aerodynamic load predictions for aeroelastic applications. By training the network on the difference between high‑fidelity CFD lift data and a physics‑based Wagner model for the NLR 7301 airfoil in transonic flow, the residual approach consistently outperforms a direct neural‑network model in most tests, especially in generalization scenarios. The study demonstrates that residual learning can effectively augment classical low‑order aerodynamic theories by learning a lower‑variance correction to the baseline physics model.
The paper presents a dataset of roughly 10,000 RANS simulations of steady, incompressible, two‑dimensional subsonic flow around downforce‑generating NACA 4‑digit airfoils, aimed at automotive and motorsport regimes. It explores geometry‑conditioned neural surrogates—comparing neural fields, neural ODEs, MLPs, graph models, and spectral conditioning—and introduces Airfoil2Vec, a spectral geometry encoder that fuses joint contour, camber, and thickness spectra to predict continuous pressure and velocity fields. The surrogate models are evaluated on angle‑of‑attack interpolation, unseen NACA geometries, and non‑NACA airfoils, achieving accurate aerodynamic predictions with orders‑of‑magnitude speedups over traditional CFD.
arXiv:2609. 38977v1 Announce Type: cross Abstract: Neural surrogates have emerged as fast alternatives to the numerical simulation of three-dimensional turbulence.