Machine learning-assisted prediction of the hardness of additively manufactured and heat-treated Ti-6Al-4V alloy

• A novel multi-source dataset integrating experimental, literature, and Stratified Bootstrap combined with Gaussian Copula Noise synthetic data was developed to overcome data scarcity in SLM Ti-6Al-4V hardness modeling. • Unlike prior studies focusing on tensile properties or as-built conditions, this work systematically models post-SLM heat-treatment–hardness relationships. • The proposed ensemble framework achieved high predictive accuracy (R² ≈ 0.92; MAE ≈ 7.8 HV), representing a significant improvement in generalization across diverse heat-treatment conditions. • Explainable ML (SHAP) quantitatively confirmed heat-treatment temperature and time as dominant drivers, aligning model predictions with metallurgical phase-transformation mechanisms. • The framework enables data-driven optimization of post-SLM heat-treatment design, reducing reliance on costly trial-and-error experimentation in industrial settings. The optimization of post-processing heat treatment for selective laser melted Ti-6Al-4V remains challenging due to the strong nonlinear coupling between thermal history, microstructure, and hardness. Existing predictive models are typically limited by small datasets and narrow process coverage, particularly for post-heat-treatment hardness. In this study, a machine learning framework was developed to predict the Vickers hardness of heat-treated SLM Ti-6Al-4V using a curated multi-source dataset integrating experimental measurements (19 samples), literature-derived data (42), and 200 synthetically generated samples via Stratified Bootstrap combined with Gaussian Copula Noise. Fifteen regression models were systematically benchmarked using cross-validation. Among them, the Voting Regressor achieved the highest predictive accuracy (R² ≈ 0.92, MAE ≈ 7.8 HV), demonstrating robust generalization across diverse heat-treatment conditions. Explainable artificial intelligence analysis revealed that microstructural characteristics and heat-treatment parameters are the dominant drivers of hardness, in agreement with phase-transformation mechanisms governing α′ decomposition and α+β stabilization. The proposed framework provides a quantitative and interpretable tool for rational heat-treatment design of SLM Ti-6Al-4V, reducing reliance on empirical trial-and-error approaches and enabling data-driven process optimization.