Predictive Modeling of Vickers Hardness Using Machine Learning Techniques on D2 Steel with Various Treatments.

Area of Science:

• Materials Science
• Mechanical Engineering
• Computer Science

Background:

• Hardness is a critical mechanical property for material selection and performance assessment.
• Machine learning (ML) offers efficient methods for predicting material characteristics.
• Traditional hardness testing requires precise diagonal measurements of indentations.

Purpose of the Study:

• To predict Vickers hardness values using only scanned monochromatic images of indentation imprints.
• To evaluate the effectiveness of non-deep ML regression methods for this task.
• To eliminate the need for manual diagonal measurements in hardness assessment.

Main Methods:

• Regression techniques including decision trees, adaptive boosting, extreme gradient boosting, and random forest were applied.
• The models were trained on 54 images of D2 steel in various conditions (commercial, quenched, tempered, TiNbN coated).
• Vickers hardness was predicted directly from image data, bypassing geometrical measurements.

Main Results:

• The Random Forest model achieved the best performance.
• Key performance metrics included a Root Mean Square Error (RMSE) of 0.95, Mean Absolute Error (MAE) of 0.12, and a Coefficient of Determination (R2) of approximately 1.
• The Random Forest model significantly outperformed other regression methods evaluated.

Conclusions:

• ML algorithms can accurately predict Vickers hardness from indentation images.
• This image-based ML approach offers a rapid and precise alternative to traditional measurement methods.
• The findings suggest potential for enhanced quality control in industrial material assessment.