Evaluating mechanical and tribological performance of B4C reinforced aluminum composites with predictive machine learning models

Abstract

This study presents a comprehensive investigation into the mechanical and tribological performance of aluminum alloy AA8011 reinforced with boron carbide particles, developed through ultrasonic-assisted stir casting. The influence of varying boron carbide content (0–10 wt.%), applied load (10–50 N), and sliding speed (1–6 m/s) on wear rate and friction coefficient was examined using an extensive factorial experimental design totaling 180 test runs. Microstructural analysis revealed that 6 wt.% boron carbide yielded the most uniform particle dispersion, resulting in a 38.59% improvement in hardness and significantly enhanced wear resistance. Worn surface characterization confirmed the formation of protective tribofilms and reduced material degradation under increased load and speed. To model and predict tribological behavior, six supervised machine learning algorithms were applied. Gaussian Process Regression achieved the highest accuracy in predicting wear rate (R² = 0.955), while Extreme Gradient Boosting performed best for friction coefficient prediction (R² = 0.994). Feature importance analysis identified load and speed as the most influential parameters, surpassing the effect of reinforcement content. The combined experimental and data-driven approach not only deepens understanding of wear mechanisms such as abrasive, oxidative, and delamination wear but also reduces the need for repetitive testing by 40%, offering a scalable framework for accelerated material development in automotive and aerospace applications.

Topics & Concepts