Studies on High Temperature Wear Behavior of Additively Manufactured Ti-6Al-4V Alloy Using Machine Learning Models

This work focuses on tribological performance of titanium alloy processed by laser bed fusion (LBF) at elevated temperature. The specimens were prepared using the laser bed fusion method, and further heat treatment (annealing) was carried out at temperature of 860 and 960 °C. Microstructure of as-printed and heat-treated alloy specimens was studied using optical microscope as well as scanning electron microscope (SEM). Tribological behavior has been determined by means of pin-on-disk tribometer in dry condition at elevated temperature, with EN31 as counter material. Results show that COF gets reduced by 4.5 and 15.2%, respectively, with increase in load and pin temperature. Also, friction coefficient gets reduced by 1.75% when annealing temperature increases from 860 to 960 °C. However wear rate gets increased by 75 and 47% with increase in load and pin temperature. Eight diverse machine learning algorithms, Neural Network, Extra Trees, Random forest regressor, K-Nearest Neighbor (KNN), Decision Tree, Histogram-based Gradient Boosting, XG Boost and Cat Boost were applied to the experimental data. Performance calculations verified that ML-based models can successfully predict the variation of Wear and COF. The Extra Trees regression model outperforms other machine learning models, achieving the highest accuracy with an R² of 99.79% and the lowest MSE of 0.0824 for COF, and R² of 94.67% with an MSE of 0.0003 for wear. This highlights its superior predictive capability and minimal error. Different wear mechanisms were identified by analyzing the worn surfaces and wear transition. Additionally, a wear mechanism map was developed to identify safe loading condition.