中文摘要：

准确预测和有效控制拼焊板成形回弹，研究各因素的影响及其交互作用具有重要意义。以U形件拼焊板为研究对象，通过数值模拟、正交试验和神经网络相结合的方法，考察多个工艺参数对拼焊板回弹的交互作用，建立拼焊板回弹的BP神经网络预测模型，对U形件拼焊板的回弹进行预测和控制。结果表明，板料参数、焊缝位置、压边力、模具间隙、凹模圆角等均对拼焊板回弹有重要的影响，并存在交互作用；建立的BP神经网络模型能很好地预测U形件拼焊板在各参数影响下的回弹变化趋势，和给定一组工艺参数下的拼焊板回弹量，为拼焊板的回弹控制提供了可靠的依据。神经网络技术在拼焊板回弹预测中的应用，为拼焊板成形优化研究提出了新思路。

英文摘要：

It＇s great significance to exactly predict and efficiently control the springback in Tailor welded blanks （TWBs） forming, and to study the interaction of various process parameters influencing springback. Taken U-shaped TWBs into account, interaction of several process parameters influencing springback of TWBs is comprehensively studied through numerical simulation and orthogonal experiment. According to the characteristic of TWBs forming, BP neural network model adapted to predict springback is built up, which is used to predict the springback of U-shaped TWBs. The results show that springback of TWBs is mutually influenced by parameters of blank, position of welded-line, blank holder force, clearance of tools, die profile radii etc. , tendency of springback affected by every parameter respectively and the springback under the case of giving a group of process parameters can be preferably predicted by BP neural network prediction model, which supplies a foundation to control springhack of TWBs. Neural network technology applied to springback prediction of TWBs gives a new way to research optimization of TWBs forming.