U-shaped Part Springback Forecasting Based On The Finite Element Method And The Neural Network

As an important processing method in the modern industrial production, stamping forming has been widely used in many fields. Residual stress will be produced in the process of sheet stamping because of elastic factors, which made the sheet moved to the opposite direction of stamping after unloading, the process was called springback. Springback is a kind of forming defects, which will influence not only the precision of the forming part, but also the assembly process and service life. Because of that, accurately forecasting and controlling the springback would be necessary. Springback is a hot issue in academic circles in the recent years because of its many factors, its complex reasons and its involving subject extensively. Thus precise forecast and analysis of springback is of great importance for improving product precision, prolong service life, and shorten the development cycle.On the basis of the researching metal forming and bending theories, the following works have been done in the paper:1. Computational results of sheet forming springback were gotten by finite element analysis software which was called ANSYS/LS-DYNA and DYNAFORM. Through the comparison of both tools, DYNAFORM was chosen to forecast springback in the paper.2. Blank-holder force, radius of moulds, the gap between moulds and friction coefficient were determined as the main factors that affect springback, and then each level of the five factors was determined. After that, orthogonal test was conducted. For the data was thinning enough, the orthogonal test was arranged twice. Through the analysis of the test results, the weights of parameters affect springback were gotten.3. On the basis of orthogonal test results, a BP neural network was build. After the learning and training, the forecasting results of network were very close with the finite element results, so the network could be used to forecast springback instead of the finite element method.Finite element analysis method, orthogonal test and BP neural network were comprehensive used in the paper. The results showed that combination of finite element and neural network could be used to forecast springback, it would be a new way to forecast and control springback. The method can not only save a lot of time, but also guarantee the accuracy of predicted results.