Prediction the Springback in Air-Bending Process Using Neural Network Metamodel

Abstract

To eliminate geometric deviations of sheet metal subjected to an air-bending process from the geometry required by the designer, it is necessary to predict the accurate value of the springback. Springback is the elastic deformation observed upon removal of the load during a bending process. In order to predict the springback amount, a multidimensional function should be approximated. In this paper, a neural network metamodel (NNM) based on the back propagation algorithm is introduced to predict the springback value. A verified nonlinear finite element model is developed to generate NNM training data. To select the training data for the NNM, computer generated D-optimal designs are utilized. The NNM developed model in this research can be used in determination of the springback value in sheet metal bending.