A neural network model for prediction of static recrystallization kinetics under non-isothermal conditions

This paper presents a new scheme to apply artificial neural network (ANN) in prediction of static recrystallization (SRX) kinetics. Firstly, based on empirical data of SRX kinetics, a mathematical model is suggested to construct a neural network. Then, an appropriate neural network is trained on the basis of the first and the second derivatives of recrystallized fraction with respect to time. Finally, a thermo-mechanical finite element analysis is coupled with the proposed ANN to predict static recrystallization kinetics in hot rolling process of AA5083. To verify the model, the predicted results and the experimental data are compared and it is observed that there is a good consistency between the two sets of results.