Insight into the hot deformation behavior and recrystallization mechanism of Mg-Y-RE alloys based on machine learning

The hot deformation behavior of magnesium (Mg) alloys is significantly governed by the multi-physics coupling effects of temperature (T), strain rate (ε) and strain (ε), resulting in flow behavior that exhibits pronounced nonlinearity and multi-scale complexity. This study systematically investigates the hot deformation behavior of Mg-Y-Nd-(Sm)-Zr alloys. Sm alloying promotes recrystallization. The flow stress of Sm-containing alloys declines sharply towards a steady state after reaching its peak value. To overcome the limitations of the Arrhenius-type constitutive (AC) model in predicting complex nonlinear flow behavior, the AC and data hybrid informed neural network (ACINN) model is developed. This approach enhances the predictive accuracy and extends the applicability of the traditional AC model. The evolution of microstructure and recrystallization behavior under hot deformation conditions are investigated based on results from electron backscatter diffraction (EBSD) and transmission electron microscopy (TEM). The relationship between the power dissipation factor (η) and recrystallization behavior is further examined using K-means clustering analysis. The results demonstrate that dynamic recrystallization (DRX) behavior varies with the η value, comprising four distinct regimes: dynamic recovery (DRV), discontinuous dynamic recrystallization (DDRX) dominance, continuous dynamic recrystallization (CDRX) dominance and complete dynamic recrystallization. This analysis presents a new perspective for studying the hot deformation processes of Mg alloys.