Physics-Informed Neural Networks for Shell Structures

We introduce a framework to simulate the mechanical small-strain response of shell structures via PINNs as described in 'Physics-Informed Neural Networks for shell structures'.

To run the studies, simply clone this repository via

git clone https://github.com/jhbastek/PhysicsInformedShellStructures.git

and run main.py with the indicated study. For the strong form, simply run main_strong.py. You may choose your own setting by providing the corresponding values in params.py. To consider different surfaces, simply add the corresponding charts in similar style as the given charts to src.geometry.py.

For further information, please first refer to the publication, or reach out to Jan-Hendrik Bastek.

Requirements

The requirements are fairly basic. Though we do not foresee any compatibility issues, we provide each version that was verified to work correctly.

• Python (tested on version 3.7.1)
• Python packages:
  • Pytorch (1.11.0 with CUDA 11.7)
  • NumPy (1.19.2)
  • SciPy (1.5.2)
  • Matplotlib (3.3.2, only required for plotting)

Citation

If this code is useful for your research, please cite our publication.

@article{Bastek2023,
author = {Bastek, Jan-Hendrik and Kochmann, Dennis M.},
doi = {10.1016/j.euromechsol.2022.104849},
issn = {09977538},
journal = {European Journal of Mechanics - A/Solids},
pages = {104849},
title = {{Physics-Informed Neural Networks for shell structures}},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0997753822002790},
volume = {97},
year = {2023}
}