Development of Constitutive Models for Extrapolative Prediction of Nb–Ti Micro Alloyed Steel

Abstract

The experimental stress–strain data from hot compression tests at temperatures from 900 to 1050 °C and strain rates of 0.1, 1, and 10 s−1 are used to develop the constitutive models to predict flow stress of Nb–Ti micro alloyed steel. The Arrhenius‐type model considering compensation of strain, Bayesian regularization neural network model, and integrated model are investigated. The results show that although the Arrhenius‐type model considering compensation of strain can predict the correct variation trends under different deformation conditions, the accuracy is far from being satisfactory. On the other hand, the Bayesian regularization neural network model shows high accuracy in the training data range, but rather uncertainty for the data outside the training data range. By adding stress predicted by the Arrhenius‐type model considering compensation of strain and characteristic points (critical stress, peak stress, and steady–state stress) to neural network model's inputs, the integrated model can result in accurate prediction for hot deformation behavior of Nb–Ti micro alloyed steel, showing more promising potential for industrial application.