High-temperature tensile characteristics and constitutive models of ultrahigh strength steel

In this investigation, the isothermal tensile experiments over wide ranges of deformation parameters (strain rate and tensile temperature) are conducted for studying the high-temperature tensile behaviors of an ultrahigh strength steel. The influences of deformation parameter on high-temperature tensile behaviors, fracture characteristics and deformation mechanisms are analyzed. Moreover, Arrhenius-type phenomenological (AP) model developed by the regression method or the Nelder-Mead (NM) simplex method, and the artificial-neural-network (ANN) model developed by combining genetic algorithm (GA) and back propagation learning algorithm (BP) are proposed, respectively. The results show that the high-temperature tensile behavior of the studied steel exhibits the typical work hardening and dynamic recovery characteristics. The necking capability increases with the strain rate decreasing and tensile temperature increasing. However, the large deep dimples dramatically deteriorate the loading capability during the localized necking, leading to the poor elongation to fracture at low strain rate. Both for the modeling and verifying data, the AP model developed by the NM simplex method shows the relatively high relative coefficient (higher than 0.9963), low average absolute relative error (lower than 1.6692%) and narrow error band (controlled in +/- 6.8MPa), compared with the AP model developed by the regression method and the GA-BP ANN model.

Automated summary

The study reveals that the high-temperature tensile behavior of the steel shows typical work hardening and dynamic recovery characteristics. The necking capability increases with decreasing strain rate and increasing tensile temperature. However, the presence of large deep dimples significantly reduces the loading capability and leads to poor elongation to fracture at low strain rates. The Arrhenius-type phenomenological (AP) model developed by the Nelder-Mead (NM) simplex method shows high accuracy and narrow error bands compared to other models.