Triboinformatic analysis and prediction of B₄C and granite powder filled Al 6082 composites using machine learning...

The traditional methods for fabricating and evaluating wear properties are inherently time-consuming and financially demanding. To address these challenges, machine learning (ML) has emerged as a potent approach in predicting the mechanical and tribological behavior of advanced materials, including Al-based composites. The primary aim of this study is to combine experimental methodologies with ML algorithms to accurately predict the wear and coefficient of friction for B₄C-granite composites, thereby aiding in the design and manufacturing of materials with enhanced wear performance. The composites were synthesized using stir casting, and wear behaviour was experimentally evaluated under dry sliding conditions using a pin-on-disc tribometer, resulting in a dataset comprising 81 samples. The experiments revealed that wear loss increased with higher load and lower reinforcement percentage, reaching to 0.315 g at 2.5%, 30 N, 1.67 m/s and 1200 m, compared to minimum wear loss of 0.029 g at 7.5%, 10 N, 0.83 m/s and 600 m. Seven different supervised regression-based ML models were applied to accurately predict wear characteristics, with hyperparameter tuning conducted to ensure a robust comparative analysis. The developed model's results were evaluated utilizing a number of statistical metrics to identify the most reliable algorithm for wear and COF prediction. These models training and validation has been performed using experimental data, demonstrated strong potential for predicting tribological behavior with high accuracy, thereby reducing the need for extensive physical testing. Among all the approaches, the Fuzzy logic model achieved the highest predictive performance with highest R² of 0.9638 and lowest MAE of 0.0023 for wear loss and R² value of 0.9833 and lowest MAE of 0.0059 for COF, respectively. In addition, the Pearson coefficient correlation map establishes that reinforcement percentage have strong negative correlation of (- 0.57) and (- 0.50) with wear loss and COF.