On the understanding and prediction of tribological properties of Al-TiO 2 nanocomposites using artificial neural network

Due to the influence of the manufacturing process on the composite’s wear properties, mathematical models cannot accurately predict the wear rates and coefficients of friction of composite materials. As a result, this work provides a deeper comprehension of the tribological properties of Al-TiO₂ nanocomposites with varying TiO₂ content tested at varying sliding loads as well as improved predictability. Accumulative roll bonding (ARB) was used to create Al-TiO₂ nanocomposites that had good TiO₂ nanoparticle dispersion in the matrix. The pin-on-disc test was used to measure their wear rates and coefficient of friction. A neural network model was used to predict the wear rates and the coefficient of friction because there was a correlation between the composite morphology, hardness, and microstructure, as well as the evolution of the tribological properties. Due to the uniform distribution of TiO₂ nanoparticles within the composite and the saturation of grain refinement in the Al matrix, it was experimentally demonstrated that wear rates decrease as the number of ARB passes increases until a plateau is reached. After five ARB passes, the composite containing 3% TiO₂ nanoparticles achieved the maximum hardness improvement of 153.7%. While the same composite’s wear rates decrease from 3.7 × 10⁻³ g/m for pure Al to 1.1 × 10⁻³ g/m at 5 N load. For each of the produced composites that were subjected to four distinct wear loads, the artificial neural network model was able to accurately predict the wear rates and coefficient of friction, achieving determination coefficient R2 values of 0.9766 and 0.9866, respectively, for the wear rates and coefficient of friction.