Study On The Application Of ANN On The Prediction Of TWBs' Springback

The tailor welded blanks (TWBs) become more and more important to the automobile industry, for their advantages such as lowering manufacturing cost, vehicle weight reduction, environmental protection and safety improvement, etc. Despite of these advantages of TWBs, there are also some technical problems in welding and punching, and there are great difference of stress and strain in the each side of blank for the difference in the thickness between the two side of TWBs which makes the formability more complex compared with the blank, such as cracking, . wrinkling, weld-line movement and springback. Researchers have done a great deal of works on cracking, wrinkling and weld-line movement, but little work has done on springback, and only a few references are available. Along with the more and more strict demand on dimensional precision, especially the 2mm engineering, the research on springback TWBs is getting more and more important.In this thesis, the springback of longitudinal-welded TWBs was researched. And the principal contents studies are presented as follows:1. Taking U-shape TWBs into account, we study the influence of blank holder force and its distribution on the springback of TWBs through numerical simulation and experiment. The results show that, blank holder force and its distribution have a notable influence on the springback of TWBs, and there are some relation between TWBs' springback and weld-line movement. The springback quality can be reduced via adjusting blank holder force and its distribution.2. Several process parameters influencing the springback of TWBs are comprehensively studied. Combination of parameters is arranged with orthogonal design method, and the numerical simulation of TWBs bending is carried out, then the results are gained. Sequentially, how much the different parameters influencing TWBs' springback is studied. According to the results, several other parameters beside of blank holder force, such as the position of weld line, die profile radii and tools clearance etc. also influence the springback of TWBs strongly. 3. Nonlinear relationship between process parameters and TWBs' springback is established based on the artificial neural network (ANN) , whose accuracy is testified by the test samples.4. Analysis of TWBs' springback and process parameters optimization based on artificial neural network and orthogonal experiment design are carried out. Within the arrange of process parameters, artificial neural network model, as the substitute of numerical simulation, is used to gain the trends of TWBs' springback effected by different process parameters, and combined with the orthogonal experiment, is introduced to further optimize the process parameters.5. Application of artificial neural network on TWBs' springback prediction is studied. Aided with the powerful nonlinear mapping capability of artificial neural network model, within the value range of training specimens, a certain springback quality can be attained from the given suit of process parameters without numerical simulation software, so the time of process program is saved.Integration of orthogonal experiment and artificial neural network is applied on the TWBs' springback analysis and parametric optimization, which obviously save the time of optimizing parameters and improve the process design efficiency on thecondition of accuracy permission.