A multi-model fusion machine learning approach for Fe-containing heat resistant cast Al-Si alloy design

A voting model based on multi-model fusion machine learning approach was proposed to predict the composition of the Fe-containing heat resistant cast Al-Si alloy. The results show that the voting model has a higher prediction accuracy (R 2 = 0.9289, MAE = 23.73) and robustness (Test R 2 / Train R 2 = 0.95) than the single algorithm models. By using this model combined with genetic algorithm, a novel heat resistant cast Al-7.59Si-3.48Cu-2.48Mg-1.29Fe-0.54Mn-0.19Ti (wt.%) alloy was successfully designed and experimentally validated. The as-cast alloy demonstrated excellent tensile properties, with a tensile strength of 138.6 MPa at 300 °C, which improved to 159.3 MPa after T6 heat treatment. The good heat resistance of T6 alloy is mainly attributed to solid solution strengthening resulting from the dissolution of the θ-Al 2 Cu and β-Mg 2 Si phases, as well as the reduced stress concentration at the particle/matrix interface and the dispersion strengthening caused by the spheroidization of the eutectic Si and Q-Al 5 Cu 2 Mg 8 Si 6 phases, together with the rounding and coarsening of the α-Al(Fe,Mn)Si phase. The established framework also serves as a valuable reference for the efficient and interpretable design of other complex alloy systems.