Corrosion Behaviour and Machine-Learning-Based Prediction of CNT and Micro- Titanium Reinforced Copper Composites

This Paper examines the corrosion behaviour of copper-based metal matrix composites reinforced with hybrid combinations of carbon nanotubes (CNTs) and micro-titanium particles, and evaluates the use of machine-learning models for corrosion prediction. Ten compositions (C0–C9) were fabricated by varying CNT content (0.5–1.5 wt.%) and titanium content (1–5 wt.%). Immersion corrosion tests were conducted in 0.5 N and 1 N acidic media for 24–96 hours. Results showed a clear reduction in corrosion rate with increasing reinforcement levels, with the unreinforced copper sample (C0) exhibiting the highest corrosion, while samples C8 and C9 showed the lowest values (≈ 0.001 mm/y), indicating a significant improvement in corrosion resistance. To enable predictive modelling, four regression approaches, Linear Regression, Polynomial Regression (Poly2), Support Vector Regression (SVR), and Random Forest (RF), were trained using the experimental dataset. Polynomial regression consistently provided the highest accuracy (R² >0.95 in most cases), while SVR showed poor predictive capability with negative R² values for several samples. Residual analysis and Q–Q plots confirmed that polynomial regression exhibited the most stable and normally distributed error behaviour. SEM surface morphology supported the corrosion results, revealing severe pitting and degradation in C0, whereas C8 and C9 showed smooth, minimally damaged surfaces, confirming strong corrosion protection. The study demonstrates that CNT–titanium hybrid reinforcement significantly enhances corrosion resistance in copper MMCs, and that polynomial regression offers a reliable machine-learning tool for forecasting corrosion trends. The combined experimental–computational approach provides a framework for accelerated design and screening of corrosion-resistant metal composites.