This repository stores code for my masters thesis to create machine learning models for optimizing SUPG-parameters for singularly perturbed convection diffusion problems.
can be found in the readme of the repository and here. Here is a short description.
FEM-routines are implemented using Fenicsx - in particular Dolfinx and UFL. Additional computations are done using NumPy. Visualizations use Pyvista and Matplotlib.
Pytorch is used to implement the Neutral Networks.
There is a minimal documentation for the code implemented here. Documentation FEniCsx and Pytorch can be found on the respective websites.
In order to possibly improve on the models provided in the literaure, it is expected that some experiamentation will be necessary. Here visualizations and observations will be documented ordered by release.