A novel data-driven strategy for rapid design of synergistic heat treatment in titanium bimetal

Titanium bimetals composed of alloys with distinct compositions and microstructures offer a pathway to overcome the strength-ductility trade-off, yet their synergistic heat treatment design remains constrained by empirical trial-and-error approaches. To address this, a data-driven inverse design strategy integrating XGBoost regression (XGBR) with the non-dominated sorting genetic algorithm II (NSGA-II) was proposed. A dataset of 12,732 literature-derived samples trained four XGBR models to predict tensile strength (σ) and elongation (δ) for high-strength (HTS) and high-ductility (LTS) alloys, achieving R² values above 0.85. Coupling NSGA-II optimization with grey relational analysis (GRA) enabled comprehensive evaluation of Pareto front solutions and selection of optimal heat treatment parameters. Microstructural characterization revealed α/β phase evolution under the designed thermal conditions, providing physical insight into the data-driven predictions. Experimental validation on a representative HTS/LTS titanium bimetal composed of Ti-3.65Al-6.17Cr-4.08Mo-7.99V-4.01Zr (HTS) and Ti-4.22V-6.37Al-0.24C (LTS, wt.%) demonstrated that the designed heat treatment (810 °C/0.4 h-460 °C/11.5 h) yielded a tensile strength of 1625 MPa in the HTS alloy and an elongation of 18.55 % in the LTS alloy, with deviations from target values within 10 %. These results establish the XGBR-NSGA-II strategy as an effective and generalizable framework linking data-driven optimization with experimental characterization for the rapid and accurate design of titanium bimetals.