Vickers hardness prediction from machine learning methods.

Area of Science:

• Materials Science
• Computational Materials Science
• Solid State Physics

Background:

• Predicting material hardness is crucial for industrial applications but challenging due to complex plastic behavior modeling.
• Existing hardness models are often complex, inaccurate for extrapolation, or require specialized coding skills.
• Mechanical properties like bulk modulus, shear modulus, Young's modulus, and Poisson's ratio are key indicators of material strength.

Purpose of the Study:

• To develop an accurate and accessible machine learning model for predicting material hardness.
• To utilize mechanical properties as input for hardness prediction.
• To facilitate the discovery of new superhard materials.

Main Methods:

• A Gradient Boosting Regressor (GBR) machine learning model was developed.
• The model was trained using an experimental Vickers hardness database of 143 diverse materials.
• Input mechanical properties were derived from theoretical elastic tensors.

Main Results:

• The GBR model successfully predicted material hardness with high accuracy.
• The model's predictions showed good agreement with available experimental data.
• Exploration of the Materials Project database identified potential new superhard materials.

Conclusions:

• The developed machine learning model offers a reliable and user-friendly method for hardness prediction.
• This approach simplifies the search for new superhard materials, benefiting industrial research.
• An accessible online application is available for broader use in materials science research.