The AirfRANS dataset makes available numerical resolutions of the incompressible Reynolds-Averaged Navier–Stokes (RANS) equations over the NACA 4 and 5 digits series of airfoils and in a subsonic flight regime setup. Readthedocs documentation is available here.

• Access to 1000 simulations.
• Reynolds number between 2 and 6 million.
• Angle of attack between -5° and 15°.
• Airfoil drawn in the NACA 4 and 5 digits series.
• Four machine learning tasks representing different challenges.

Install with

pip install airfrans

From python:

import airfrans as af

af.dataset.download(root = PATH_TO_SAVING_DIRECTORY, unzip = True)

You can also directly download a ready-to-use version of the dataset in the PyTorch Geometric library Finally, you can directly download the dataset in the raw OpenFOAM version here, or in the more friendly pre-processed version here.

From python:

import airfrans as af

dataset, dataname = af.dataset.load(root = PATH_TO_DATASET, task = TASK, train = True)

The tasks are the one presented in the associated paper. You can choose between 'full', 'scarce', 'reynolds' and 'aoa'. The dataset loaded in this case is the same as the one you can directly access via the PyTorch Geometric library. If you want more flexibility about the sampling of each simulation for the inputs or targets, please feel free to build a custom loader with the help of the 'Simulation' class presented in the following. We highly recommend to handle those data with a Gemetric Deep Learning library such as PyTorch Geometric or Deep Graph Library.

The 'Simulation' class is an object to facilitate the manipulation of AirfRANS simulations. Given the root folder of where the directories of the simulations have been saved and the name of a simulation you can easily manipulate it.

import airfrans as af

name = 'airFoil2D_SST_57.872_7.314_5.454_3.799_13.179'
simulation = af.Simulation(root = PATH_TO_DATASET, name = name)

See the documentation for more details about this object.

This project is licensed under the ODbL-1.0 License.

The original paper accepted at the 36th Conference on Neural Information Processing Systems (NeurIPS 2022) Track on Datasets and Benchmarks can be found here and the preprint here. Disclaimer: An important update correcting an inconsistency in the Machine Learning experiments proposed in the main part of the NeurIPS version of the paper has been done. Please refer to the ArXiv version for the up to date version.

Please cite this paper if you use this dataset in your own work.

@inproceedings{
bonnet2022airfrans,
title={Airf{RANS}: High Fidelity Computational Fluid Dynamics Dataset for Approximating Reynolds-Averaged Navier{\textendash}Stokes Solutions},
author={Florent Bonnet and Jocelyn Ahmed Mazari and Paola Cinnella and Patrick Gallinari},
booktitle={Thirty-sixth Conference on Neural Information Processing Systems Datasets and Benchmarks Track},
year={2022},
url={https://arxiv.org/abs/2212.07564}
  }