Machine Learning-Driven Prediction of Microstructure-Mechanical Property Relationships in Mg-Al Alloys

Abstract

This study employs machine learning algorithms to predict the correlation between microstructural characteristics and mechanical properties in Mg-Al binary alloys, while optimizing the parameter ranges for synergistic strength-ductility regulation. Results demonstrate that eight microstructural descriptors, including grain size and Mg17Al12 phase distribution, predominantly govern the alloy's strength, plasticity, and their coordinated enhancement. The XGBoost model achieved prediction errors of merely 1.54% and 4.66% for strength and plasticity, respectively. Experimental validation revealed that Mg-14.4Al alloy solidified under 5.2 GPa pressure within optimized parameter ranges exhibited gradient-structure-induced strengthening-toughening effects. The Shapley Additive Explanations model further elucidated an inverse regulatory mechanism between Mg17Al12 phase volume fraction and strength-ductility metrics.

Keywords: Mg-Al alloys, Machine learning algorithms, Strength-ductility synergy, Microstructural parameter design