Accelerating phase prediction of refractory high entropy alloys via machine learning. (1st December 2022)

• Record Type: Journal Article
• Authors: Qu, Nan; Zhang, Yan; Liu, Yong; Liao, Mingqing; Han, Tianyi; Yang, Danni; Lai, Zhonghong; Zhu, Jingchuan; Yu, Liang
• Is Part Of: Physica scripta. Volume 97:Number 12(2022)
• Publication Date: 2022-12-01
• Subjects: refractory high entropy alloys -- phase classification -- machine learning -- decision tree; Physics -- Periodicals; 530.05
• DOI: 10.1088/1402-4896/aca2f2
• Languages: English
• ISSNs: 0031-8949

Abstract

Abstract: The unique high-temperature properties of refractory high entropy alloys (HEAs) are mainly depended on their phase formation. Therefore, a new approach to predict the phase formation has to be proposed, in order to accelerate the development of refractory HEAs. Here, we use machine learning to build classifiers to predict the phase formation in refractory HEAs. Our dataset containing 271 data only consists of as-cast refractory HEAs data. We simplify the input parameters to element content, and refine the phase formation outputs into five classes. Decision tree has been employed to build our phase classifier, due to its great advantages in solving classification problem. Both training and test accuracy of phase formation prediction achieve 90% using our classifier. The five single phase prediction accuracies are above 97%. Our phase classifier performs effectively in multi-phases classification and prediction of refractory HEAs, and establishes a direct relation between compositions and refractory phase formation.