Applied Machine Learning Method to Predict Crack Propagation Path in Polycrystalline Graphene Sheet

A machine learning model is developed to predict the crack propagation path in polycrystalline graphene sheets. The dataset used for training the machine learning (ML) model is obtained from the molecular dynamics (MD) simulations. A training set of 700 samples has been used to train the ML model. Each training sample consists of an input image which contains the information of the initial configuration of precracked polycrystalline graphene sheet and an output image which contains the information of crack path. After training, the ML model can predict the crack propagation path in polycrystalline graphene sheet instantaneously, thus avoiding the computational costs involved with MD simulations.