Learning to predict crystal plasticity at the nanoscale: Deep residual networks and size effects in uniaxial compression discrete dislocation simulations
Zijiang Yang, Stefanos Papanikolaou, Andrew C. Reid, Wei-keng Lao, Alok Choudhary, Carelyn E. Campbell, Ankit Agrawal
The increase of dislocation density in a metallic crystal undergoing plastic deformation influences the mechanical properties of the material. This effect can be used to examine the related inverse problem of deducing the prior deformation of a material sample from its subsequent behavior. We quantify the deformation of sample materials using strain profiles. In this work, we propose a deep residual network to predict the initial strain deformation level of the sample of interest using its strain profile. The results show that the proposed deep learning model significantly outperforms a traditional machine learning model as well as accurately produces statistical predictions of the stress-strain curves for samples of interest. In addition, by visualizing the filters in convolutional layers and saliency maps, we observe that the proposed model is able to learn the significant features and capture the salient region from strain profiles of samples.
, Papanikolaou, S.
, Reid, A.
, Lao, W.
, Choudhary, A.
, Campbell, C.
and Agrawal, A.
Learning to predict crystal plasticity at the nanoscale: Deep residual networks and size effects in uniaxial compression discrete dislocation simulations, Nature Communications, [online], https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=928638
(Accessed October 21, 2021)