A New Technique of Springback Prediction by Combining FEM Calculation and Artificial Neural Network

The study of deformability of metal sheets subjected to complex deformation states has become a topic of present interest; this fact is also a consequence of the emergence of new materials meant for auto bodies fabrication, whose forming processes imply knowledge about the influence of their modified mechanic characteristics on their behaviour during the deformation, and also on the quality of products. By predicting the quality of a formed product it is possible to evaluate the measure of the influence of its errors on the functioning of the final assembly the product belongs to. The paper presents a method to predict springback using the knowledge discovered and recorded after one experimental test is set up. The neural network method is used for prediction of the springback that offer a perfectible and dynamic model which can be enriched with new experimental or FEM simulation data.