Abstract
It is a difficult task to select the process variables and determine the optimum design using trial-and-error procedure in the metals industry. In this paper, an attempt has been made to combine finite element analyses (FEA) and artificial neural network (ANN) to study the effect of process variables on material flow behavior and required forming load in cold forging. The process variables are used as the inputs and the finite element results as the target outputs, a neural network model was established. In order to achieve a proper neural network model which can generalize the training data obtained from FE data well, some crucial training parameters, such as learning algorithms, hidden neurons and hidden layers, number of training data and error goal, have been investigated. Finally, the optimum neural network model was used to predict random unseen data and thus find the optimum process variables according to a certain decision rule.
