We introduce a learning algorithm to discover neural network parameterized cohesive laws using displacement fields.
To run the examples, execute python main.py --case ex1 for the first example and python main.py --case ex2 for the second.
The following libraries are required:
| Package | Version (>=) |
|---|---|
| numpy | 1.25.2 |
| torch | 2.0.1 |
| nvidia-warp | 1.6.0 |
@article{gavris2026discovering,
title={Discovering neural cohesive zone laws from displacement fields},
author={Gavris, Georgios Barkoulis and Sun, WaiChing},
journal={Computer Methods in Applied Mechanics and Engineering},
volume={452},
pages={118733},
year={2026},
publisher={Elsevier}
}