Artificial intelligence approach to the internal variable-based rheological model for steels

The paper is a continuation of the authors' earlier work dealing with application of artificial neural networks to the prediction of yield stress in hot forming of metals. At present, the task of the network is to predict a time-derivative of the dislocation density during hot deformation. The inputs are the state of the material defined by the current dislocation density and by the time-integral of strain, the current strain rate and temperature. The flow stress curve is determined from the dislocation density vs. strain function, which is calculated using a finite difference technique in which the time-derivative of the dislocation density is supplied by the artificial neural network. Examples of calculations are presented for the axi-symmetrical compression of low carbon steel.