基于BP神经网络的W形微弯曲回弹预测  

Prediction of the micro W-bending’s springback based on the BP neural network

作　　者：刘晓宇[1] 陆小龙[1] 黄茜 杜怡君 LIU Xiao-yu;LU Xiao-long;HUANG Xi;DU Yi-jun(School of Mechanical Engineering,Sichuan University,Chengdu 610065)

机构地区：[1]四川大学机械工程学院

出　　处：《机械设计》2019年第10期14-17,共4页Journal of Machine Design

基　　金：四川省科技计划面上资助项目(2018JY0573);四川大学专职博士后研发基金资助项目(2018SCU12066)

摘　　要：能准确预测并有效控制微弯曲回弹对于提高其尺寸精度和成形质量具有重要意义。首次以W形微弯曲成形为研究对象,开展基于BP神经网络的微弯曲回弹预测研究。首先采用基于I优化准则的试验设计方法对影响成形精度的4个因素进行了试验,得到56组数据。再分别利用40组、8组、8组数据对神经网络进行了训练、验证和预测。结果表明,BP神经网络能快速预测回弹量,且预测精度满足实际工程要求,可为W形微弯曲成形尺寸精度的控制提供参考。Accurately predicting and effectively controlling the springback is very important for improving the micro Wbending’s dimensional accuracy and forming quality.In this article,the micro W-bending is taken as the research object for the first time,and the research is conducted on the prediction of its springback.Firstly,the experimental design based on the I-optimal criterion is carried out to explore how the four factors affect the forming accuracy,and 56 sets of data are identified.Then,40,8 and 8 sets of data are employed for the training,validation and test of the neural network respectively.The results show that the BP neural network predicts the springback amount rapidly,and the prediction accuracy meets the requirements of practical engineering,which provides reference for controlling the dimensional accuracy of the micro W-bending.

关 键 词：回弹 BP神经网络 W形微弯曲 成形精度 尺寸效应 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]