Determination of the initial billet geometry for a forged product using neural networks

In the paper, the authors propose a new technique to determine the initial billet geometry for the forged products using a function approximation in neural networks. A three-layer neural network is used and the back-propagation algorithm is employed to train the network. The billet geometry is determined by applying the ability of the function approximation of neural networks to determine an optimal billet which, satisfies the forming limitations; minimises incomplete filling in the die cavity, load, energy; and also secures a more uniform distribution of effective strain. The neural networks are applied to find the optimal initial billet size for a plane-strain rib-web product in cold forging and for an axisymmetric rib-web product in hot forging. The neural networks may reduce the number of finite-element simulation for the determination of the optimal die geometry of forging products and, further, they can be usefully applied in the physical modeling of the forging design. (C) 1997 Elsevier Science S.A.