Abstract

Recent developed data-driven techniques have demonstrated their great ability to complement or even replace conventional simulation methods owing to the massive computational advantages of machine learning models, without compromising on the accuracy of the results. The ultimate goal of the study is to develop a data-driven crystal plasticity framework for RVE based studies. To begin with, a machine learning model is used to effectively predict the stress-strain response and texture evolution of single crystals under uniaxial tension. Stress-strain and texture data for single crystals obtained from crystal plasticity fast Fourier transformation simulations are divided into training and test datasets, which are used to train, validate and test the machine learning model. The model will accept the strain and three Euler angles representing the input crystallographic texture as input and will produce the stress and three Euler angles representing the evolved crystallographic texture as outputs. The model will also be used to predict the stress-strain response and texture evolution of polycrystals under uniaxial tension and also other varied loading conditions, such as simple shear and biaxial loading. The model will be verified by the typical rolling and annealing texture components that have not been used for training, which allows the transferability of the framework to practical engineering problems.