A novel approach for studying crack propagation in polycrystalline graphene using machine learning algorithms

A machine learning model is proposed to predict the brittle fracture of polycrystalline graphene under tensile loading. The model employs a convolutional neural network, bidirectional recurrent neural network, and fully connected layer to process the spatial and sequential features. The spatial features are grain orientations and location of grain boundaries whereas sequential features are associated with the crack growth. Molecular dynamics modeling is used to obtain the fracture process in pre-cracked polycrystalline graphene sheet subjected to tensile loading. The data from molecular dynamic simulations along with novel image-processing techniques are used to prepare the data set required to train and test the proposed model. Crack growth obtained from the machine learning model shows a close agreement with the molecular dynamic simulations. The proposed machine learning model predicts crack growth instantaneously avoiding the computational costs associated with molecular dynamics simulations.