中文摘要：

为解决金属板材折弯成形的回弹及凸模设计问题,提出了一种基于遗传算法的反向传播神经网络预测方法.根据板料的折弯试验,建立了板料回弹预测的遗传神经网络正、逆模型.采用对象-正模型-逆模型学习法,用正模型研究板料折弯回弹规律,用逆模型预测折弯凸模的半径几何值.通过对某一起重机吊臂工件的折弯成形实验,证明提出的预测方法在模具设计和工件成形方面均有满意的效果.

英文摘要：

A prediction method based on genetic algorithm（GA） and back propagation neural network （BPNN） was proposed to solve springback problem of sheet metal after air-bending forming and punch design. Based on production experiment, the positive model and contrary model of springback prediction were developed by using GA and BPNN. Adopting the ＂object-positive model-contrary model＂ learning method, sheet springback law could be further studied with positive model and punch radius could be predicted by contrary model. Manifested by the experiment for air-bending forming of work-pieces, the prediction method proposed yields satisfactory performance in designing punch and forming workpiece.