Synergistic enhancement of strength and corrosion resistance in Al-Mg-Si alloys through chemical composition design via machine learning approaches

In this work, a novel Al-Mg-Si alloy was designed using a non-dominated sorting genetic machine-learning algorithm (NSGA-II) and Shapley additive explanations (SHAP). The designed alloy in the T5 temper simultaneously exhibited enhanced mechanical properties and intergranular corrosion resistance, with an ultimate tensile strength of 408.4 ± 1.6 MPa, an elongation of 9.9 ± 0.4 %, and a maximum intergranular corrosion depth of 108.7 ± 9.5 μm. Compared to the high-strength 6110A in the T5 temper (comparative alloy), the designed alloy showed a 9 % improvement in strength and a 20 % improvement in corrosion resistance. Furthermore, for the designed alloy, the T6 temper did not significantly improve properties compared to the T5 temper, which indicates the alloy exhibits good adaptability to online quenching processes. The principles of NSGA-II for optimizing alloying elements were elucidated through SHAP. In addition, strengthening model and microstructures quantitative characterizations were conducted to reveal the synergistic enhancement mechanisms of strength and corrosion resistance. These findings provide valuable guidance for developing high-strength Al-Mg-Si alloys suited for online quenching.