Quantitative Prediction of Ti6Al4V Tribological Behavior Using Advanced Machine Learning Regression with Feature Engineering and Ensemble Models

The growth of industrial development has increased attention to sustainability and efficiency, resulting in greater research advances toward improved performance of materials. The tribological behavior of materials, specifically friction and wear, is one of the most primary topics of interest in material performance advancement. This work provides an advanced machine learning regression-based model for the quantitative prediction of the Coefficient of Friction (CoF) and wear rate of the Ti6Al4V alloy. The novel approach employs an extensive pipeline of advanced feature engineering to inform an ensemble model based on a dataset compiled from the literature. The optimized Gradient Boosting Regressor achieved F1 results in excess of 95% accuracy on an unseen data set (R2 = 0.944; RMSE = 0.020) for CoF predictions, and a stacking regressor/model markedly improved wear rate predictions (R2 = 0.730) compared to baseline models and the CoF predictions for clarification of real-time engineering applications. The ensemble regression model is designed to provide high-fidelity, quantitative benchmarks for Ti6Al4V, which can be used as critical tools for materials design and optimization. The methodology confirmed the models' important physical relevance through feature-importance analysis: Hardness × Load for the CoF models, and Sliding Distance for wear rate.