A Machine Learning Approach to Predict Tribological Properties of Heat-Treated Al7075/3 wt% Carbon Nanotube Composites

Abstract

Abstract. This research studies the wear characteristics of various heat-treated Al7075/3 wt% carbon nanotubes (CNTs) metal matrix composites processed by the casting process. The wear characteristics of both untreated Al7075 alloy and Al7075/3 wt% CNTs composites under as-cast, solution-treated, and T6 heat-treated conditions were evaluated using a pin-on-disc apparatus in dry sliding conditions. The influence of process parameters such as heat treatment, applied load, and sliding velocity on wear-rate and friction coefficient was studied. The wear mechanisms under high-stress conditions were examined using scanning electron microscopy (SEM) and energy-dispersive spectroscopy (EDS) of the worn surfaces and revealed that delamination, abrasion, oxidation, and adhesion were the dominant wear mechanisms. Different machine learning models, including random forest (RF), gradient boosting (GB), support vector machine (SVM), neural network (NN), XGBoost (XGB), and light gradient boosting machine (LGBM), have been trained for the tribological parameters prediction. For wear-rate prediction, the XGBoost showed the best overall performance with the lowest root-mean-square error (RMSE) of 0.082 and mean absolute error (MAE) of 0.065, along with the highest R2 score of 99.1%, indicating excellent predictive accuracy and minimal error. The gradient boosting achieved the lowest RMSE of 0.013 and MAE of 0.010, along with the highest R2 score of 97.6% for the average coefficient of friction.