Experimental and machine learning-based analysis of wear behaviour in CNT and titanium reinforced copper matrix composites

This study investigates the dry sliding wear behaviour of copper-based metal matrix composites reinforced with carbon nanotubes (CNTs) and micro-sized titanium particles, combining experimental tribological analysis with data-driven modelling. Nine composite formulations were fabricated using stir casting and tested under applied loads of 1, 2, and 3 N using a pin-on-disc apparatus in accordance with ASTM G99. Experimental results demonstrate a clear load-dependent increase in wear depth, with the C9 composite (1.5 wt.% CNT + 5 wt.% Ti) exhibiting the lowest wear depth (120–188 µm), corresponding to an approximately 61% reduction compared to unreinforced copper. Machine learning models including Linear Regression, Polynomial Regression, Support Vector Regression (SVR), Random Forest, and Artificial Neural Network (ANN) were evaluated to analyse wear trends. Polynomial regression achieved perfect fitting (R² = 1.00), indicating overfitting under data-limited conditions, whereas SVR provided the best balance between accuracy and generalisation (R² = 0.94, MAE = 22 µm, RMSE = 34 µm). Feature importance analysis identified applied load as the dominant parameter governing wear behaviour. The study highlights the importance of cautious model selection and validation when applying machine learning to experimentally constrained tribological datasets, supported by microstructural observations.