A swirl-ratio-conditioned neural operator for predicting and reconstructing tornado-like vortex
Physics of Fluids · 01.11.2026 · 38 (11)
DOI: 10.1063/5.0352865 ↗Kurz gefasst
Fachartikel in der Zeitschrift Physics of Fluids, online erschienen am 18.11.2026. Verfasst von Han Zhang und 3 weiteren Autor:innen. Die Arbeit nennt 34 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Basic Research Program of Jiangsu Province, Jiangsu Youth Talent Support Program.
Abstract (Originalsprache)
Rapid prediction of near-ground tornado vortices and the reconstruction of wind fields from sparse observations are essential for the wind-resistant design and disaster assessment. High-fidelity large-eddy simulation resolves vortex structures but requires tens of hours per case and repeated simulations for observation matching. We propose a physics-data integrated conditional neural operator that uses swirl ratio as the conditional input and spatial coordinates as queries for simulating tornado-like vortices (TLVs), with training constrained by flow-field data, governing-equation residuals, and the ground boundary condition. A branch-trunk architecture learns the operator mapping from swirl ratio to radial, tangential, and axial velocity fields. For unseen swirl ratios, zero-shot prediction preserves the vortex core, annular high-tangential-velocity region, near-ground inflow, and central updraft topology, with an overall tangential-velocity relative L2 error of 9.5% and near-real-time inference. Radial and axial errors occur mainly in the near-ground boundary layer, at the vortex-core edge, and around axial-flow-reversal regions, where steep gradients amplify small spatial displacements. For wind fields with unseen and unknown-swirl ratios, without retraining the main model, sparse-observation reconstruction reduces the tangential-velocity error to about 4% for numerical data and about 7.7% for external independent experimental data, with both reconstructions completed within seconds. These results show that conditioning on swirl ratio enables the network to learn the balance among rotational momentum, radial inflow, and the changes in the vortex-core size, rotational intensity, central updraft, and the axial-flow-reversal structure. The proposed operator therefore reconstructs the vortex features across swirl ratios and data sources, thereby providing an efficient computational path for simulating TLVs.
Förderung
- National Natural Science Foundation of China
- Basic Research Program of Jiangsu Province
- Jiangsu Youth Talent Support Program
Zitieren
APA
Zhang, H., Wang, H., Liu, Z., & Wang, X. (2026). A swirl-ratio-conditioned neural operator for predicting and reconstructing tornado-like vortex. Physics of Fluids, 38(11). https://doi.org/10.1063/5.0352865
Vancouver
Zhang H, Wang H, Liu Z, Wang X. A swirl-ratio-conditioned neural operator for predicting and reconstructing tornado-like vortex. Physics of Fluids. 2026;38(11). doi:10.1063/5.0352865
BibTeX
@article{Zhang2026,
author = {Zhang, Han and Wang, Hao and Liu, Zhenqing and Wang, Xu},
title = {A swirl-ratio-conditioned neural operator for predicting and reconstructing tornado-like vortex},
journal = {Physics of Fluids},
year = {2026},
volume = {38},
number = {11},
doi = {10.1063/5.0352865}
}
EUVIMED