Comparative study on constitutive relationship of as-cast Ti60 titanium alloy during hot deformation based on Arrhenius-type and artificial neural network models

• The influence of strain on flow stress was incorporated in material constants.
• The predictability of two constitutive models was evaluated.
• The Arrhenius-type constitutive model is broken down at high strain rates.
• The ANN model can track the experimental data at all regimes.

Hot compression of as-cast Ti60 titanium alloy at the deformation temperatures ranging from 970 to 1120 °C with an interval of 30 °C, the strain rates ranging from 0.01 to 10.0 s−1 and the height reduction of 75% is conducted on a Gleeble-3500 thermo-mechanical simulator. The experimental stress–strain data from hot compression are employed to develop the Arrhenius-type constitutive model incorporating the strain effect and artificial neural network (ANN) model with a back-propagation learning algorithm. The strain compensated constitutive model can track the experimental data across the whole hot working domain other than that at high strain rates (⩾1 s−1). It is possibly associated with the deformation mechanisms at high strain rates, where microstructure exhibits bands of flow localization and longitudinal cracking, are far different from that at low strain rates (⩽0.1 s−1) where dynamic recrystallization occurs. Comparison of the predicted results of flow stress based on the ANN model and those acquired from the strain compensated constitutive model has been performed. It is found that the relative error of the ANN model varies from −3.42% to 4.33% while that of the strain compensated constitutive model ranges from −14.65% to 13.63%, and the average absolute relative error is 2.41% and 8.45% corresponding to the ANN model and strain compensated constitutive model, respectively. These results sufficiently indicate that the ANN model is more accurate and efficient in terms of predicting the flow stress of as-cast Ti60 titanium alloy.