Improving the prediction accuracy of constitutive model with ANN models

The unified constitutive model developed by Estrin and Mecking (1984) has successfully been used in hot rolling to provide information for the control of strip thickness. It has presented a high accuracy in predicting the hot strength of austenitic steels. However, the materials can show quite different properties under different deformation conditions and the constitutive models are not able to be generalised to cover a wide range of compositions and deformation conditions, therefore, the potential of those model is limited. In this work, the robustness of the unified constitutive model is enhanced by incorporating an artificial neural network model to predict the flow strength of austenitic steels with carbon content ranging from 0.0037 to 0.79%.