摘要
针对传统BP神经网络具有易陷入局部极小等缺陷,采用遗传算法(GA)对BP神经网络(初始权值、阈值)进行了优化,将人工智能技术和激光扫描测量技术有机结合,建立了金属板材数字化渐进成形回弹预测的遗传神经网络模型,对计算结果与BP神经网络预测结果进行比较,表明遗传神经网络预测值与实测值之间具有很高的相关性和精确度,该模型可用于预测渐进成形工艺参数与回弹量之间的映射关系,为金属板材数字化渐进成形回弹量的预测开辟了一条新的途径.
Artificial neural networks were introduced to the process of incremental sheet NC forming （ISF）. There were some disadvantages in BP （backpro pagation） neural networks, such as easily falling into local minimum point, BP networks were optimized by genetic algorithm （GA）. By combination of artificial intelligence technology with laser-scanning measuring, built was genetic neural network model for incremental sheet metal NC （numerical control） forming springback prediction. The calculated results were compared with those of traditional BP neural network. The results showed that the prediction precision was precise and the pertinence between the predicted GA-BP and measured values were considerably high. Thus, this model can be used to predicate the relation between the process parameters of ISF and springback and provides a new way to predicate the springback of ISF.
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2008年第1期121-124,共4页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(50175034)
关键词
渐进成形
回弹预测
遗传算法
BP神经网络
incremental sheet NC forming （ISF）
springback prediction
genetic algorithm
BP neural network
作者简介
韩飞（1977-），男，博士研究生；武汉，华中科技大学材料成形及模具技术国家重点实验室（430074）．E-mail：hanfei_hust@126．com
