Extrapolative prediction of the hot strength of austenitic steels with a combined constitutive and ANN model

Abstract

An integrated phenomenological and artificial neural network (IPANN) model developed previously by Hodgson et al. [P.D. Hodgson, L.X. Kong, C.H.J. Davies, J. Mater. Process. Technol. 87 (1999) 132–139] significantly improves the accuracy of the prediction of the hot strength of a commercial 304 stainless steel in comparison with either the phenomenological or the ANN model because of the integration of information developed from a phenomenological constitutive model. In the present work, the Estrin–Mecking constitutive model [Y. Estrin, H. Mecking, Acta Metall. 32 (1984) 57–70] was combined with the IPANN model to predict extrapolatively the hot strength of a plain-carbon austenitic steel with a carbon content of 0.79 wt.%, deformed at temperatures from 900 to 1100°C and at strain rates between of 1 and 30 s −1. The ANN model was able to predict the hot strength over a wider range of deformation conditions using the experimental data and the data from the physical model as ANN training data set. Although, the prediction is not as accurate as if a complete experimental data set had been available, the technique does provide an accurate approach to predict extrapolatively the hot strength of steels with the ANN model.