The Variable Blank-holder Force Curve Prediction Of Deep Drawing In Square Box Of Tailor-welded Blank

With the increase number of cars, oil energy consumption is becoming more and more big, now to solve the problem of world energy, reduce automobile fuel consumption has become one of the main goal of today’s car manufacturers. Tailor-welded sheet forming technology in the car on the successful application of effectively reduce the body weight, greatly improving the energy efficiency, accord with the requirement of today’s society for energy conservation and environmental protection. But tailor-welded sheet forming is more complex than ordinary veneer of forming, so in recent years, many scholars for welding plate forming technology carried out extensive research.Based on the welding plate box shape parts as the research object, using the finite element analysis software DYNAFORM material with different thickness of welding plate box form a study on numerical simulation analysis of deep drawing process. Poor first by using the orthogonal experiment analysis to determine the box shape parts forming of welding plate wrinkling, rupture, the weld-line movement in common defects of the three main influencing factors, and improved the welding plate box shaped pieces of various influencing factors of deep drawing properties of the optimal combination; And then through the finite element analysis to find a better loading ways of the variable blank-holder force curve to improve the performance of tailor-welded sheet forming and combining with the results of the study puts forward there is a optimal loading ways of the blank-holder force curve, can guarantee the welding plate box shape stamping do not produce wrinkles broke, when the weld-line movement and the smallest idea, and through the finite element simulation analysis was carried out preliminary validation.Established welding plate box shaped pieces of deep drawing process variable blank-holder force prediction of BP neural network model, realized with mold structure, different material thickness ratio, different material has a better case blank-holder force load curve forecasting. The comparison of the neural network prediction results and the simulation results are in good agreement, and the relative error is withen 20%.