Combining Machine Learning and Molecular Dynamics to Predict Mechanical Properties and Microstructural Evolution of FeNiCrCoCu High-Entropy Alloys

Abstract

Compared with traditional alloys, high-entropy alloys have better mechanical properties and corrosion resistance. However, their mechanical properties and microstructural evolution behavior are unclear due to their complex composition. Machine learning has powerful data processing and analysis capabilities, that provides technical advantages for in-depth study of the mechanical properties of high-entropy alloys. Thus, we combined machine learning and molecular dynamics to predict the mechanical properties of FeNiCrCoCu high-entropy alloys. The optimal multiple linear regression machine learning algorithm predicts that the optimal composition is Fe33Ni32Cr11Co11Cu13 high-entropy alloy, with a tensile strength of 28.25 GPa. Furthermore, molecular dynamics is used to verify the predicted mechanical properties of high-entropy alloys, and it is found that the error between the tensile strength predicted by machine learning and the tensile strength obtained by molecular dynamics simulation is within 0.5%. Moreover, the tensile-compression asymmetry of Fe33Ni32Cr11Co11Cu13 high-entropy alloy increased with the increase of temperature and Cu content and the decrease of Fe content. This is due to the increase in stress caused by twinning during compression and the decrease in stress due to dislocation slip during stretching. Interestingly, high-entropy alloy coatings reduce the tensile-compression asymmetry of nickel; this is attributed to the reduced influence of dislocations and twinning at the interface between the high-entropy alloy and the nickel matrix.

Keywords

• high entropy alloy
• machine learning
• molecular dynamics
• mechanical properties