The phase selection via machine learning in high entropy alloys

Abstract

The phase formation of high entropy alloys (HEAs) is diversity due to the breadth of the composition space. Most of the studies about phase formation are based on features of solid solution. However, the establishment of the link between compositions and phase formation plays an important role in accelerating novel HEAs development. To achieve this aim, we employed the support vector machine (SVM) to build phase selection models with both the composition dataset and the thermodynamic parameters dataset. The accuracies of both models are above 85%. A dataset with more than a thousand data has been established, which covers 18 elements and most of the HEAs families. The feature dataset was transferred by calculation of thermodynamic parameters. The similar accuracies of the models with compositions dataset and with thermodynamic parameters proved that the prediction of phase formation in HEAs can be carried out directly from compositions. Our models provide a universal approach of phase formation prediction in HEAs.