Research on predicting flow stress of 7075 aluminum alloy using machine learning models

Introduction Accurate prediction of flow stress during the hot deformation of 7075 aluminum alloy is essential yet challenging, as conventional constitutive models are often inaccurate and artificial neural network (ANN) approaches are computationally complex. Methods Hot compression experiments on as-rolled 7,075 aluminum alloy were carried out using a TA DIL805D thermal simulator over a temperature range of 573–733 K and strain rates between 0.001 and 1.0 s -1 . The resulting experimental data were subsequently used to train four machine learning models—decision tree, random forest, support vector machine, and XGBoost—for predicting the flow stress of annealed 7,075 aluminum alloy. Model performance was evaluated through residual analysis and several statistical indicators, including mean absolute error (MAE), mean squared error (MSE), average absolute relative error (AARE), correlation coefficient (R), and coefficient of determination ( R 2 ). Results The results demonstrate that, compared with previously reported artificial neural network (ANN) models, these four machine learning approaches achieve comparable predictive accuracy (up to 99.9%). Discussion While offering a simpler and more efficient model construction process.