Study On The Method Of Predicting The Quantity Of Forming Formwork For Pre - Longitudinal Beam Of Passenger Car

Under the lightweight automobile demand, high and super-high strength steels are being used in producing automobile covering parts more and more often,which makes the springback of sheet metal more prominent and affect the size precision of parts severely. In order to guide the mold making and stamp springback control panel. Generally,it is considered in two aspects:One is optimizing the process parameters to control the rebound.The other is using CAE software for sheet stamping phenomenon by numerical simulation. But the above two methods can only guide preliminary design process of mold. When die finishing, it is only the mold designer’s experience that can be relied on, which is both time-consuming and laborious. In the meanwhile, the effective components of the initial model can’t be fully used twice, just take it as a kind of direct result, learn the repairmodulus perceptually without the establishment of an effective model to use in the further development of the mold. In this paper, the modeling method of mathematical regression model will be used. And molding cavity by regression model using BP neural network and support vector machine will be used respectively to make regression prediction.The traditional methods of controlling springback is too complicated.Crafty, material, geometry and other factors will be taken into consideration to measure them comprehensively, the more complex the model is, more difficult to set up a model. This article tries to find a regression law between the repair modulus and parts cavity based on real repair modulus.Since the regarding factors are not much, data processing method is simple, and the method can be used generally,it provides a new idea for the prediction of repair modulus predictionfinishing die.The study object of this paper is a passenger car left/right front longitudinal section made of B340material.The paper first uses the DYNAFORM software to carry on numerical simulation to predict springback parts and its general rebound value, which provides a reference value for the initial modulus of repairing mold designer. Then a mathematical model is set up according to the model designed by mold engineers, to explore a new research method for the repair quantity. Firstly,looking for parts and forming regression of die through in-depth analysis of part features and process. In order to study the regularity of the regression, I used two intelligent algorithms-SVM and BP neural network to fit training. Then using the trained network to make mold prediction of mold position modified repeatedly by mold designers. The article also compares the forecasting results of SVM and BP neural network and both their advantages and disadvantages in predicting the repair module. Finally, improve the prediction accuracy, the eventual prediction error was controlled in3mm through the improvement of data acquisition and optimization modeling methods. Because of the performance of stamping process parameters and materials are not considered, this method is simple and practical, strong applicability, has the reference value to other kinds of U shaped parts as repair mode.