Application of neural networks to represent the rheological behaviour of nickel-based superalloys under varying hot deformation conditions

In this work, neural networks are employed to represent the rheological behaviour of nickelbased superalloys under varying hot deformation conditions, that approximate thermo-mechanical cycles of industrial hot forging operations. A feed-forward back-propagation neural network has been trained and then tested on rheological data, obtained through hot compression experiments, where the strain rate has been varied continuously during the deformation step. A good agreement between calculated and experimental data has been obtained, proving the feasibility of this new approach.