Characteristics of work hardening and constitutive models comparison of powder metallurgy Al-5.6Zn-2Mg alloy during hot compression

The hot compression tests of Al-5.6Zn-2Mg aluminum alloy were conducted on a universal testing machine at temperature of 300–500 °C and strain rate of 0.1–0.0001 s⁻¹. The work hardening rate curves for the σ_c and ε_c for the onset of dynamic recrystallization were identified. The correlation among the key features σ_c (ε_c), σ_p (ε_p) and σ_SS, and the Z coefficient are determined. Four constitutive models include the Arrhenius-type model, modified Johnson Cook (MJC), modified Zerilli-Armstrong (MZA), and an artificial neural network (ANN) developed. The results showed that the ANN and Arrhenius-type models had the lowest AARE values of 0.486% and 3.36%, while the MZA and MJC models had higher AARE values of 8.84% and 3.93%, respectively. The Arrhenius-type model was found to be the most appropriate prediction model due to its ability to handle the nonlinear relationship among factors, but the MJC model could be a simpler alternative in cases where material properties are unknown or experimental data are limited. The MZA model was found to be unsuitable for estimating flow stress in hot compression. In addition, the highest predictive performance is seen in the best-trained ANN model, with an AARE of 0.486% and an R value of 0.99.