Machine Learning Breakthrough: New Model Predicts Hardness of Solid-Solution Alloys

Investigators at Yanshan University have made a significant breakthrough in machine learning research, developing a new physical model that can predict the hardness of solid-solution alloys. The research, funded by organizations such as the National Natural Science Foundation of China and the Ministry of Education Yangtze River Scholar Professor Program, aimed to address the long-standing challenge of accurately predicting hardness based on microscopic electron structure, particularly for metal or alloys. By leveraging ridge regression, data interpolation, and machine learning techniques, the researchers created a model that outperforms traditional macroscopic solution strengthening models.

Key Takeaways:

• The research proposes a new energy-based physical model using ridge regression (RR) to explain the hardness of solid solution state alloys.
• The model's accuracy is validated based on hardness data from different normal solid solubility (NSS) or non-equilibrium solid solubility (NESS) Mg-X binary alloys.
• The model establishes a regression expression ΔHV = ΣAN(p)E(p), where A is a constant, N-p is the number of Mg-X ion pairs passed by free electrons, and E-p is the energy barrier for each Mg-X ion pair.
• The new model, developed using data interpolation and machine learning techniques, demonstrates better generalization performance compared to traditional macroscopic models in both NSS and NESS cases.
• The research concludes that the new model provides a novel microscopic approach to predicting the hardness of solid-solution alloys.
• The study has been peer-reviewed and published in the Journal of Alloys and Compounds (Elsevier).

Statistics:

• The research was funded by a total of four organizations, including the National Natural Science Foundation of China and the Ministry of Education Yangtze River Scholar Professor Program.
• The model is validated based on hardness data from 7 different Mg-X binary alloys.
• The number of Mg-X ion pairs passed by free electrons (N-p) ranges between 1.2 and 1.9.
• The energy barrier for each Mg-X ion pair (E-p) is a key parameter in the regression expression ΔHV = ΣAN(p)E(p).

Sources:

• Solid-solution Hardness Prediction Model of Mg-x Binary Alloys By Machine Learning. Journal of Alloys and Compounds, 2025;1040.
• Yanshan University, State Key Laboratory of Metastable Materials Science and Technology.
• National Natural Science Foundation of China (NSFC).
• Natural Science Foundation of Hebei Province.
• Hebei Provincial Key Research Projects.
• Innovation Groups Program.
• Ministry of Education Yangtze River Scholar Professor Program.