Answers are grounded in this paper and cite their sources.
Evaluates ensemble machine-learning models for predicting mechanical properties in titanium alloys, suggesting new analytics techniques.
Titanium alloys are widely used in aerospace, biomedical, and energy applications because of their high strength-to-weight ratio, corrosion resistance, and biocompatibility, but experimental characterisation remains resource-intensive. This study evaluates XGBoost, LightGBM, and Random Forest using an expert-curated dataset of 283 Ti-alloy entries from 105 peer-reviewed studies. Instead of complete-case analysis, a target-specific strategy used all available rows for each property: 237 (YM, 83.7%), 271 (YS, 95.8%), 182 (UTS, 64.3%), 257 (DAR, 90.8%), and 136 (HV, 48.1%) entries, avoiding the loss of more than 75% of records. Predictors comprised elemental atomic fractions, thermophysical descriptors (mixing entropy, Mo and Al equivalencies, valence electron concentration, and atomic size mismatch), harmonised processing indicators, comment-derived flags, and auxiliary measured properties with median imputation. Phase-stratified 80/20 train-test splits produced held-out sets of 28–55 samples. All targets achieved positive held-out test R2 values: UTS 0.924 (LightGBM), HV 0.914 (XGBoost), YS 0.893 (XGBoost), DAR 0.832 (XGBoost), and YM 0.810 (Random Forest). The results provide a quantitative baseline for Ti-alloy materials informatics.
No takes yet. Share an insight, caveat, or question.
Mihail Kolev (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: