geometric_parameterization_ML

A test case for geometrical parameterization using machine learning.

Steps for reproducing the simulation

You can follow the steps to run the case:

1. generate_mesh.ipynb
2. simulation_data/data_generation
3. data_preprocessing.ipynb
4. prediction
   1. EZyRB_PODI_RBF.ipynb, EZyRB_PODI_GPR.ipynb and EZyRB_PODI_ANN.ipynb. The three techniques are utilized to interpolate the POD coefficients.
   2. EZyRB_AE_RBF.ipynb, EZyRB_AE_GPR.ipynb and EZyRB_AE_ANN.ipynb. The three techniques are utilized to interpolate the latent space coordinates.

Python packages

1. The simulations are carried out by OpenFOAM v2306.
2. It is recommended to use conda to manage the Python packages used in the case. The required packages are listed in the file package-list.txt. You can install it by running: conda create --name --file package-list.txt
3. The EZyRB packages for PODI can be installed by running: pip install git+https://github.com/mathLab/EZyRB. But there might be conflicts between the packages. Please create a new env if any problems with importing packages occur.