Uncovering stress fields and defects distributions in graphene using deep neural networks

In our latest article, “Uncovering stress fields and defects distributions in graphene using deep neural networks”: https://doi.org/10.1007/s10704-023-00704-z , we showed that conditional generative adversarial networks (cGANs) could transform complex deformation fields into stress fields by eliminating the need to evaluate elasticity distributions and develop complex nonlinear constitutive relations.

Moreover, cGANs demonstrated remarkable generalizability beyond the training samples when predicting defect distributions. The network accurately predicted the existence of a crack in a material sample even though cracked samples had not been used during the training stage.

The MATLAB scripts used to generate LAMMPS data/input files as well as postprocessing the numerical results of the simulations are available here: https://github.com/nuwan-d/deep_generative_neural_net

The trained neural networks and complete data set are available here: https://doi.org/10.5281/zenodo.7834444.