Analysis of Phase Formations and Mechanical Properties in Complex Concentrated Alloys by Machine Learning Approach

TL;DR

This study employs machine learning models to classify phase formations and predict mechanical properties of complex concentrated alloys, achieving high accuracy and demonstrating the potential of ML in alloy design.

Contribution

It introduces new classification and regression models for CCAs, utilizing multi-task SISSO and random forest techniques with high predictive accuracy.

Findings

Classification accuracy over 85%

Correlation coefficient > 0.9 for hardness and yield stress

SISSO models with correlation > 0.85

Abstract

The mechanical properties of complex concentrated alloys (CCAs) depend on their forming phases and corresponding structures, the prediction of the phase formation for a given CCA is essential to its discovery and applications. 541 sample were collected from previous studies, comprising 61 amorphous, 164 single-phase crystalline, and 361 multi-phases crystalline CCAs. We proposed three classification models to category and understand the phase selection of CCAS. Also, a two-objective regression model was constructed to predict the hardness and compressive yield stress of CCAs. All three classification models have accuracies higher than 85%, and correlation coefficient of random forest regression model is greater than 0.9 for both of two objectives. In addition, we proposed four descriptors via multi-task SISSO method to predict the mechanical properties of CCAs, the average correlation coefficient of SISSO models is higher than 0.85. The present work demonstrates the great potential of machine learning approach in the prediction of target properties in CCAs.