关键词: 拼焊板, 压边力预测, BP神经网络, 遗传算法, 盒形件

Abstract: Numerical simulation and neural network technology were used for the blank holder force(BHF) prediction of tailor welded blank(TWB) box part in deep drawing. The influence of BHF loading type on the formability of TWB box part was analyzed by numerical simulation， and a preferable BHF loading type was obtained. Then BP neural network model suitable for BHF prediction was developed. At last， the neural network model was optimized by the genetic algorithm. The elitism strategy was introduced into the course of gene selection， and finally an ideal BHF curve was obtained by the neural network model optimized by the genetic algorithm， which can predict BHF value varying in deep drawing process and can lay the technical foundation for intelligent stamping.

Key words: objective, interval particle, intervaltailor welded blank, blank holder force prediction, BP neural network, genetic algorithm, box part