Intelligent development of high strength and ductile heat treatment-free Al-Si-Mg alloys for integrated die casting through the machine learning of experimental big data

Abstract

A group of twelve Al-xSi-yMg (x = 7–10, y = 0.3–0.6, in wt%) alloys were prepared by high pressure die casting, and the microstructure and tensile properties of the alloys were evaluated in the as-cast state and after natural ageing for 14 and 30 days, respectively. Based on the experimental big data, several machine learning (ML) models were applied for learning the relationship between the composition, natural ageing time and tensile properties of the die-cast alloys, and the performance of the ML models was evaluated by four parameters including mean absolute error (MAE), mean squared error (MSE), root mean squared error (RMSE) and R-square (R2). Among the applied ML models, the adaptive boosting (AdaBoost) was found as the most accurate and intelligent prediction of the yield strength (YS), ultimate tensile strength (UTS) and elongation (EL) of the die-cast alloys because of the lowest MAE of 1.06, 0.72 and 0.38 and the highest R2 of 0.96, 0.98 and 0.84. A high strength and ductile heat treatment-free Al8Si0.45Mg die-cast alloy was intelligently predicted by the AdaBoost model. The experimental validation showed that the alloy delivered the YS, UTS and EL of 150.6 ± 2.8 MPa, 288.2 ± 3.1 MPa and 11.74 ± 0.94 % in the as-cast state, and 160 ± 2.4 MPa, 294.5 ± 2.8 MPa and 10.87 ± 0.91 % after natural ageing for 30 days. The errors between the prediction and the experimental results were < 0.5 % for the strength and < 3.9 % for the ductility. This work provides a pathway for the intelligent development of high strength and ductile Al-Si-Mg heat treatment-free die-cast alloys for integrated die casting.