Application of artificial neural network and constitutive equations to describe the hot compressive behavior of 28CrMnMoV steel

Isothermal hot compression of 28CrMnMoV steel was conducted on a Gleeble-3500 thermo-mechanical simulator in the temperature range of 1173–1473 K with the strain rate of 0.01–10 s−1 and the height reduction of 60%. Based on the experimental results, constitutive equations and an artificial neural network (ANN) model with a back-propagation learning algorithm were developed for the description and prediction of the hot compressive behavior of 28CrMnMoV steel. Then a comparative evaluation of the constitutive equations and the trained ANN model was carried out. It was obtained that the relative errors based on the ANN model varied from −3.66% to 3.46% and those were in the range from −13.60% to 10.89% by the constitutive equations, and the average absolute relative errors were 0.99% and 4.09% corresponding to the ANN model and the constitutive equations, respectively. Furthermore, the average root mean square errors of the ANN model and the constitutive equations were obtained as 1.43 MPa and 5.60 MPa respectively. These results indicated that the trained ANN model was more efficient and accurate in predicting the hot compressive behavior of 28CrMnMoV steel than the constitutive equations.

► The hot compressive behavior of 28CrMnMoV steel was investigated over a wide range of temperatures and strain rates. ► The material shows strain hardening, strain rate hardening and thermal softening. ► A BP ANN model and constitutive equations were employed to predict the flow stress. ► It was found that the ANN model has a better prediction precision.