PINN-PC Framework for Composite Materials

This repository contains the implementation of the Physics-Informed Neural Network with Polynomial Chaos (PINN-PC) framework, as presented in our paper:

"Physics-Informed Neural Networks to Propagate Random Field Properties of Composite Materials"

The main idea of the PINN-PC framework is to use PINNs to approximate the coefficients of the PCE model. The model architecture is illustrated in Figure 1.

Citation

If you use this framework in your research, please consider citing our paper:

@inproceedings{bonnet2024pinnpc,
  title={Physics-Informed Neural Networks to Propagate Random Field Properties of Composite Materials},
  author={D. Bonnet-Eymard, A. Persoons, P. Gavallas, M. GR Faes, G. Stefanou, D. Moens},
  booktitle={Proceedings of ISMA-USD 2024},
  year={2024},
  organization={KU Leuven, Aristotle University of Thessaloniki, TU Dortmund University},
  doi={10.5281/zenodo.13907104}
}