Hot deformation behavior of 30MnB5V steel: Phenomenological constitutive model, ensemble learning algorithm, hot processing map and microstructure evolution

The hot deformation behavior of 30MnB5V steel was experimentally investigated and theoretically analyzed using phenomenological constitutive models and an ensemble learning algorithm. Hot compression experiments were conducted with an MMS-200 thermo-mechanical simulation machine on the 30MnB5V steel at strain rates of 0.01–10 s−1 and temperatures of 950–1150 °C. The characteristics of the true stress-strain curves under different conditions and the corresponding microstructural evolutions were examined in detail. The results indicate that raising the temperature or lowering the strain rate promotes dynamic recrystallization. As temperature decreases or strain rate increases, the degree of recrystallization softening declines, and stress-strain curves transition from recrystallization to sustained work hardening or dynamic recovery types. Subsequently, the strain-compensated Arrhenius model, modified Johnson-Cook (J-C) model, and Bayesian-optimized LightGBM (BO-LightGBM) model were constructed sequentially to predict the flow stress of the tested steel. The accuracy of the three models was evaluated using the coefficient of determination (R2), mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE). The results reveal that the accuracy of the BO-LightGBM model is significantly higher than that of the Arrhenius model and the modified J-C model. Finally, three-dimensional power dissipation maps and hot processing maps were established under various strain conditions. The results suggest that the tested steel demonstrates good workability at deformation temperatures of 1075–1125 °C and strain rates of 0.03–0.22 s−1.