Accelerated design of age-hardened Mg-Ca-Zn alloys with enhanced mechanical properties via machine learning

Precipitation-hardenable magnesium alloys have significant applications due to their lightweight and high specific strength properties. However, the wide compositions and aging treatment conditions pose challenges in efficiently identifying optimal combinations for rapid peak aging. In this study, 294 sets of data were collected for the age-hardening Mg-Ca-Zn alloys from the literature. By studying the suitability of various Machine Learning (ML) models, including linear models, support vector regression (SVR), random forest (RF), XGBoost, and AdaBoost, the alloys hardness was optimized using active learning based on the most suitable model. The results illustrate that the random forest model was the most effective model in predicting both the hardness and hardness variation of experimental data. The prediction of alloy hardness presents better performance compared to hardness variation. The alloy composition and aging process with the fast-aging response resulted in peak hardness of 71.10 Hv after aging at 175 degrees C for 8 h. This research demonstrates the potential of data-driven approaches in alloy design and optimization of age-hardening alloys.