Research On Springback Forecasting Of Sheet Metal Stamping Based On A Combination Forecasting Method Of Artificial Neural Network

Sheet metal stamping is a very important process for the manufacture of sheet metal parts and is widely applied in many industrial fields. In sheet metal forming fields, the final shape of part is commonly controlled by the springback after stamping, so the springback problem is the hotspot of research all the time. In the springback research, there are two aspects: the prediction of springback and the control of springback. If we can predict the springback exactly, we can control the springback better. Therefore it is significant to do further research on the prediction of springback.Traditional method to eliminate the geometric error is trial-and-error to modify the surfaces of the die and establish the stamping technics. It not only wastes time, but also has large expenses. If we can choose the better die shape and establish the better stamping technics in advance, the cost can be greatly reduced and the production cycle can be shortened. The appearance and development of the computer-aided engineering based on finite element method provide modern means for solving the springback problem. The former researches separately invest the material parameters; die shape and the stamping process, so it is lack the synthesis consideration. At the same time, the prediction of springback is localized in single parameter single method or multiple parameters single method. Therefore, research on the mathematical model of prediction of springback based on the numerical simulation to improve the stamping precision becomes the main task of this thesis.In this thesis, the ponderance of stamping-springback issue and the significance of research on this subject to the whole sheet metal stamping field are summarized. Internal and overseas articles related to the theory of sheet metal stamping, springback control and prediction is discussed. Then, a new method based on the orthogonal design and the numerical simulation to invest the parameters of springback and the springback prediction model based on uniform design are presented. The method considers the synthesis influence of geometry, material and technics. In this thesis, we introduce the combination forecasting into the field and use a new method based on BP Neural Networks to predict the springback.