Machine Learning-Based Comprehensive Prediction Model for L12 Phase-Strengthened Fe-Co-Ni-Based High-Entropy Alloys

L1(2) phase-strengthened Fe-Co-Ni-based high-entropy alloys (HEAs) have attracted considerable attention due to their excellent mechanical properties. Improving the properties of HEAs through conventional experimental methods is costly. Therefore, a new method is needed to predict the properties of alloys quickly and accurately. In this study, a comprehensive prediction model for L1(2) phase-strengthened Fe-Co-Ni-based HEAs was developed. The existence of the L1(2) phase in the HEAs was first predicted. A link was then established between the microstructure (L1(2) phase volume fraction) and properties (hardness) of HEAs, and comprehensive prediction was performed. Finally, two mutually exclusive properties (strength and plasticity) of HEAs were coupled and co-optimized. The Shapley additive explained algorithm was also used to interpret the contribution of each model feature to the comprehensive properties of HEAs. The vast compositional and process search space of HEAs was progressively screened in three stages by applying different prediction models. Finally, four HEAs were screened from hundreds of thousands of possible candidate groups, and the prediction results were verified by experiments. In this work, L1(2) phase-strengthened Fe-Co-Ni-based HEAs with high strength and plasticity were successfully designed. The new method presented herein has a great cost advantage over traditional experimental methods. It is also expected to be applied in the design of HEAs with various excellent properties or to explore the potential factors affecting the microstructure/properties of alloys.