Combining Machine Learning and Molecular Dynamics to Predict Mechanical Properties and Microstructural Evolution of...

Area of Science:

• Materials Science
• Computational Materials Science

Background:

• High-entropy alloys (HEAs) offer superior mechanical and corrosion properties compared to traditional alloys.
• Complex compositions of HEAs hinder a full understanding of their mechanical behavior and microstructural evolution.

Purpose of the Study:

• To predict the mechanical properties of FeNiCrCoCu high-entropy alloys using machine learning and molecular dynamics.
• To determine the optimal composition for enhanced mechanical performance.

Main Methods:

• Utilized machine learning, specifically multiple linear regression, for predictive modeling.
• Employed molecular dynamics simulations for verification of predicted properties.
• Investigated the influence of temperature, composition (Fe, Cu content), and alloying on tensile-strength and tensile-compression asymmetry.

Main Results:

• The optimal composition predicted by machine learning is Fe33Ni32Cr11Co11Cu13, with a tensile strength of 28.25 GPa.
• Molecular dynamics simulations confirmed ML predictions with less than 0.5% error.
• Tensile-compression asymmetry in Fe33Ni32Cr11Co11Cu13 HEAs increases with temperature and Cu content, and decreases with Fe content.

Conclusions:

• Machine learning and molecular dynamics are effective tools for studying HEA properties.
• The Fe33Ni32Cr11Co11Cu13 alloy exhibits significant tensile strength and specific tensile-compression asymmetry characteristics.
• High-entropy alloy coatings can mitigate tensile-compression asymmetry in nickel by altering dislocation and twinning effects at interfaces.