High precision modeling with stacked neural network to predict the mechanical property of aluminum alloy

The compositions and microstructures of aluminum alloys determine their mechanical properties. This study used fourteen machine learning models, including random forest and gradient boosting regression, to predict the ultimate tensile strength (UTS) of aluminum alloys. Due to dataset and model limitations, individual model predictions lacked accuracy. Thus, a novel stacked neural network (SNN) was proposed, utilizing multiple single models as base models. This SNN improves prediction accuracy and generalization by adjusting the network structure. The 31 combinations of base models provide the most effective integration strategy. The SNN's predicted tensile strength closely matches experimental data, confirming its effectiveness and accuracy.