gnn-fracture

Prediction and control of fracture paths in disordered architected materials using graph neural networks

Pipeline

To generate the Voronoi networks and their corresponding finite element meshes, run generation/generateVoronoiLattices.py. Adding the flag --plot=1 plots each Voronoi network as it is generated. Their mechanical response under mode-I loading is then evaluated using the open-source finite element code ae108. The training dataset can be found here.
To train the model, run learning/train_gru.py, and to evaluate the model on the dataset run learning/evaluate_gru.py.
To optimize the fracture length, run optimization/optimize.py. This computes and stores the parameters generating optimal Voronoi networks.

Name		Name	Last commit message	Last commit date
Latest commit History 16 Commits
data		data
generation		generation
learning		learning
optimization		optimization
LICENSE		LICENSE
README.md		README.md