Predict the phase formation of high-entropy alloys by compositions

Abstract

The existing high-entropy alloys' phase formation prediction models are established based on empirical thermophysical parameters. The process is complicated, and the accuracy of the descriptors seriously affects the final prediction results. In this article, we achieve the prediction of phase selection in high-entropy alloys by compositions. The high-entropy alloys compositions are mapped to the pseudo-two-dimensional periodic table, automatically extracting features through the convolutional neural network for classification. The results show that this method simplifies the prediction process while improving the prediction accuracy. The prediction accuracy of intermetallic compounds exceeds 89%, solid solutions and amorphous phases exceed 98%. The case study demonstrates the validity of our model. The phase composition of AlxFeCrNi (x = 0, 0.5, 1.0) high-entropy alloys are also accurately predicted and results in agreement with experiments are obtained.

Keywords

• Machine learning
• High-entropy alloys
• Compositions
• Phase formation
• Convolutional neural network