Skip to content

Codes for paper "Automatic Parameterization for Aerodynamic Shape Optimization via Deep Geometric Learning", published on AIAA Aviation Forum 2023

License

Notifications You must be signed in to change notification settings

kfxw/CFD_Mesh_Parameterization

Repository files navigation

Automatic Parameterization for Aerodynamic Shape Optimization via Deep Geometric Learning

Codes for the published paper on AIAA Aviation Forum 2023

[paper] [preprint] [video]

Dependencies

Please refer to requirements.txt for the packages needed for this project. Please note that the codes use NVIDIA GPU and Pytorch for acceleration by default.

Get Started

We provide commands so that you can try out DMM and LSM models in a few minutes.

# to obtain a DMM model that parameterizes NACA-3414
mkdir exp_dmm_2d
python DMM_airfoil_2d.py -workspaceDir exp_dmm_2d -type naca -profile 3414 -reconIter 600 2>&1 | tee exp_dmm_2d/log.log
# or to obtain a parameterization of CLARK-Y
python DMM_airfoil_2d.py -workspaceDir exp_dmm_2d -type uiuc -profile clarky-il -reconIter 600 2>&1 | tee exp_dmm_2d/log.log

# to train and obtain a LSM model
mkdir exp_lsm_2d
python LSM_airfoil_2d_train.py -workspaceDir exp_lsm_2d -trainingEpoch 21 -visualize 2>&1 | tee exp_lsm_2d/log.log

Citation

If you find this project is useful, please cite:

@inbook{doi:10.2514/6.2023-3471,
  author = {Zhen Wei and Pascal Fua and Michaël Bauerheim},
  title = {Automatic Parameterization for Aerodynamic Shape Optimization via Deep Geometric Learning},
  booktitle = {AIAA AVIATION 2023 Forum},
  publisher = {AIAA},
  doi = {10.2514/6.2023-3471},
  URL = {https://arc.aiaa.org/doi/abs/10.2514/6.2023-3471}
}

Additionally, if you find the Latent Space Model (LSM) is useful, please cite at the same time:

@article{doi:10.2514/1.J062533,
  author = {Wei, Zhen and Guillard, Benoit and Fua, Pascal and Chapin, Vincent and Bauerheim, Michaël},
  title = {Latent Representation of Computational Fluid Dynamics Meshes and Application to Airfoil Aerodynamics},
  journal = {AIAA Journal},
  year = {2023},
  doi = {10.2514/1.J062533},
  URL = { https://doi.org/10.2514/1.J062533}
}

Licence

This project has a BSD-style licence, as found in the LICENCE file. Redistributions and use of this project is permittet for academic purposes only. No commercial use is allowed.

About

Codes for paper "Automatic Parameterization for Aerodynamic Shape Optimization via Deep Geometric Learning", published on AIAA Aviation Forum 2023

Resources

License

Stars

Watchers

Forks

Languages