This paper develops a graph neural network (GNN) model to predict the properties of polycrystalline materials by considering the physical interactions among neighboring grains. Using the magnetostriction of Tb0.3Dy0.7Fe2 alloys, the model achieves a low prediction error (~10%) across diverse microstructures and quantifies the importance of each grain's features. This microstructure-graph-based GNN model enables accurate and interpretable predictions.
Publisher
npj Computational Materials
Published On
Jul 09, 2021
Authors
Minyi Dai, Mehmet F. Demirel, Yingyu Liang, Jia-Mian Hu
Tags
graph neural network
polycrystalline materials
magnetostriction
Tb0.3Dy0.7Fe2 alloys
microstructure
prediction model
material properties
Related Publications
Explore these studies to deepen your understanding of the subject.
