Research On Springback In Stamping Based On Wavelet Neural Network

With the request of lighter vehicle body, both the high-strength steel sheet and the aluminum alloy sheet had been widely used in recent years, which made the springback of stamping parts more significant. Compared with the low carbon steel sheet stamping parts, it was difficult to meet precision requirement. In addition, there were many complex beam parts in sheet stamping forming, which were easily twisted when the external force was unloaded. It was difficult to guarantee the numerical simulation accuracy of stamping and springback. Therefore, the control of the springback of complex high-strength steel and aluminum alloy stamping parts had become a key in current research.In order to control the springback of stamping parts, and improve the optimization efficiency, the wavelet neural network metamodel was used to the research. The S-rail of NUMISHEET 96 was taken into account, and the finite element model was set up, then the numerical simulation was taken based on the finite element software Dynaform. Taking four process parameters as influencing factors, and springback as forming target, the Latin hypercube was used to obtain samples, then the wavelet neural network metamodel of influencing factors and forming target was built. The optimal process parameters were got by using the improved particle swarm optimization algorithm. The results show that the wavelet neural network metamodel can describe the relationship between process parameters and springback of sheet forming, and the springback of S-rail can be greatly reduced after the optimization.The front side member model of NUMISHEET 2011 was taken into account to study the twist springback influenced by material parameters of high-strength steel sheet, the stamping, trimming and springback process were numerically simulated. A new method to evaluate the torsion springback was proposed. Then the orthogonal experiment was carried out to determine the influence of major factors on twist springback. At last, the principle for the choice of high-strength materials was proposed in order to reduce the twist springback on the basis of the results of the orthogonal experiment.Aiming at the big twist springback appearing after the stamping process of high-strength steel beam parts, the flex-rail was taken as the research object. With the evaluation index of twist springback which was proposed as forming target, the process parameters that had effect on the twist springback of flex-rail were analyzed by using the orthogonal experiment. Then four key parameters were screened out. Then the optimal process parameters was got by the application of the flex-rail’s wavelet neural network metamodel of key parameters and forming target, which was built combined with Latin hypercube design. The results show that the optimized parameters can effectively reduce the twist springback of flex-rail, and the research provides a reference for the control of twist springback.