Prediction of springback in the air V-bending of metallic sheets

Abstract

Springback is a critical phenomenon in design and analysis of sheet metal forming process of metallic sheets. An accurate prediction of elastic recovery of material allows to design forming tools which take into account springback compensation. Springback is influenced by many factors including mechanical properties of material, friction conditions, temperature and geometry of bending die. In this paper, the investigations are focused on the analysis of an intelligent air bending process using an artificial neural network (ANN). The air bending experiments were carried out in a designed semi closed 90° V-shaped die. The tests were conducted on three grades of sheet metals: aluminium 1070, brass CuZn37 and deep-drawing quality steel sheet DC04. The results of experimental tests were used as a training set for back-propagation learning of a multilayer artificial network built in Statistica Neural Network program. For all materials tested, an increase of the springback coefficient is observed when the bend angle increases. The results of neural prediction are in a good agreement with the experiments. The correlation coefficient of ANN prediction to the experimental results is equal to about 0.99.