To use this software, I highly recommend using Docker. I have prebuilt a Docker image which contains the necessary packages to run all the code here.
After installing Docker, execute the following commands in a terminal to launch JupyterLab:
docker pull cms6712/finite_element_skin
docker run --rm --workdir /skin -it -p 8888:8888 cms6712/finite_element_skin jupyter lab --allow-root --ip 0.0.0.0
If you already have an application on port 8888, you can change "8888:8888" to "{YOUR DESIRED PORT}:8888" in the command above. If a browser does not open immediately after running this second command, copy and paste the last URL into a browser:
http://127.0.0.1:{YOUR DESIRED PORT}/lab?token=.....
If you did not change the port forwarding the second command, leave {YOUR DESIRED PORT} at 8888. Make sure to copy the entirety of the token, which serves as a password!
Also beware that when you shut down JupyterLab, all files will be deleted unless you remove the --rm option from the second command.
Install your favorite distribution of conda. I recommend miniconda3. Install the following packages from conda-forge using a terminal:
conda create -n fenics fenics-dolfinx==0.7.1 jupyterlab scikit-learn pyvista trame pandas seaborn gmsh tqdm python-gmsh -c conda-forge -y
Activate the environment:
conda activate fenics
Install two more packages:
pip install sklearn_evaluation xgboost
And run
jupyter lab
Main.ipynb contains demonstrations of all the methods developed here. If you wish to run Main.ipynb from scratch, be sure to delete all files in the cache folder. This may take a while, since it will be performing nonlinear solving from scratch.
Simulation Analysis.ipynb contains the code used to process code and generate figures. data.pkl is the processed data, which is available in the Docker container.
wound_geometries.py is a helper file where we generate all the geometries and meshes used.
Finally, to_run.py is the Python script which can be used to perform simulations en masse. It takes a single argument, an integer in the range 0 to 11663 inclusive, and is invoked by
python to_run.py {SIMULATION_ID}
It generates the mesh in .xdmf format in the mesh/ folder, the displacement field stored as a NumPy array in the results/ folder, the von Mises stress in the vm_results/ folder, and the interactive PyVista HTML figures in the html/ folder.