Interpretable machine learning-based stretch formability prediction of magnesium alloys

In this study, we developed an interpretable prediction framework to access the stretch formability of AZ31 magnesium alloys through combining the extreme gradient boosting (XGBoost) model with the sparrow search algorithm (SSA). Eleven features are extracted from the microstructures (e.g., grain size (GS), maximum pole intensity (Imax), degree of texture dispersion (μ), radius of maximum pole position (r), and angle of maximum pole position (A)), mechanical properties (e.g., tensile yield strength (TYS), ultimate tensile strength (UTS), elongation-to-failure (EL), and strength difference (∆S)), and testing conditions (e.g., sheet thickness (t) and punch speed (v)) in the data collected literature and experiments. By Pearson correlation coefficient and exhaustive screening methods, ten key features (not including UTS) can be identified as the final inputs, thus enhancing the accuracy of prediction accuracy of Index Erichsen (IE) value, which serves as the model's output. The newly developed SSA-XGBoost model exhibits a better prediction performance with the goodness of fitting (R²) value of 0.91 compared to traditional machine learning models. A new dataset (4 samples) is prepared to validate the reliability and generalization capacity of this model, and the errors between the predicted and experimental IE values are below 5%. Based on this result, the quantitative relationship between the key features and IE values are established by Shapley additive explanations method and XGBoost feature importance analysis. Imax, TYS, EL, r, GS, and ΔS significantly influence the IE value among 10 input features. This work offers a reliable and accurate tool for predicting the stretch formability of AZ31 magnesium alloys and provides insights to developing the high-formable magnesium alloys.