Machine learning assisted microstructure optimization and tensile properties improvement of Al-Si-Mg alloy

• Employing an excellent prediction accuracy and fitting effect machine learning model for tensile properties of cast Al-Si-Mg alloy. And developing the high tensile performance (ultimate tensile strength, yield strength and Elongation are 356 MPa, 307 MPa and 9.3 %) via machine learning. • Comparing and ranking the importance of microstructure features by SHapley Additive exPlanations analysis and microstructure characterization. • Discussing and explaining the reason for the strengthen by calculation of precipitate strengthen and microstructure characterization. This work quantitatively investigated the relationship between microstructure and tensile properties using a data-driven machine learning model, then strengthen the alloy by optimization. First, a database of cast Al-Si-Mg alloy, including microstructure parameters and tensile properties, was established and applied to six different machine learning models. Among them, the Back-Propagation Artificial Neural Network (BP-ANN) model demonstrated superior accuracy and generalization capability and was selected for subsequent analysis and optimization. The feature importances were ranked using SHapley Additive exPlanations (SHAP) analysis, and the four most important features for ultimate tensile strength (UTS), yield strength (YS) and elongation were listed. Furthermore, a multi-objective optimization is adopted to optimize the microstructure parameters to the best ranges to strengthen the alloys, and the UTS, YS and elongation of validation results are 356 MPa, 307 MPa and 9.3 %. Finally, strengthen mechanism was further explained based on microstructure analysis. The uniform distribution of Mg 2 Si precipitates controlled by Mg content and aging temperature in grains is the main reason for improvement of tensile properties. The refining of β-Fe phase incoherent to α-Al grain reduces detrimental effect; Refining π-Fe phase and Si particle can transform the brittle fracture into ductile fracture, contribute to the improvement the tensile properties.