Mixture design optimization and machine learning-based prediction of Al-Mg alloy composite reinforced by Zn nanoparticles: A molecular dynamics study

Abstract

Al-Mg alloy is widely used in various industries due to its practicality. It is crucial to achieve the highest possible tensile properties for this alloy. In this study, first, Zn nanoparticles were combined with the aluminum-magnesium alloy to modify their tensile properties. Then, a mixture design of experiment was utilized to find the best combination of Al-Mg-Zn alloy, which was validated through molecular dynamics simulation. Mechanical tests were also conducted on Al-1%Mg-1%Zn alloy at different temperatures and strain rates. The results suggested that when the temperature increases, the tensile properties decrease, while with higher strain rates they increase. Furthermore, four machine learning-based algorithms including Generalized Linear Model (GLM), Artificial Neural Network (ANN), Decision Tree (DT), and Random Forest (RF) were compared to determine the simplest and the most accurate one for the prediction of mechanical properties of Al-Mg-Zn alloy and due to the highest values of correlation coefficient, the ANN algorithm was found as the most accurate one for the prediction of mechanical properties. The GLM algorithm was also known as an appropriate algorithm to predict the mechanical properties. However, the DT and RF algorithms had low prediction precisions. Therefore, the parameters that were influential in the prediction precision of DT and RF algorithms were changed, and the best value of each parameter was obtained. It was found that the DT algorithm with a maximum depth of 6 has the highest precision prediction of mechanical properties. Furthermore, the RF algorithm had the highest accuracy for the prediction of mechanical properties with No. of trees = 4 and maximum depth = 4.