摘要
以电池后盖板为CAE模拟分析模型,利用正交试验设计方法,将减小制品翘曲变形量作为优化目标,得到各工艺参数对制品翘曲变形量的影响程度及最优化工艺参数组合。利用径向基函数RBF神经网络对制品翘曲量进行预测,建立了各工艺参数与制品翘曲变形之间非线性映射关系模型,并与BP神经网络进行了对比。结果表明:RBF神经网络模型,可以较准备地预测制品的翘曲变形,并且在精度、训练速度等方面优于BP网络。
A battery back plate was taken as CAE simulation analysis model with the aim of reducing its warpage.By using orthogonal design methods,the effects of process parameters factors on degree of warpage and the optimum process parameter combination can be obtained.A radial basis function（RBF） network model on warpage of injection molding is established,the prediction model based on RBF network is trained through injection molding CAE data and verified by additional data successfully.Another network model based on back propagation（BP） network is also trained for comparison.The results show that for the problem in this paper,the RBF network is much better than BP network in accuracy an speed of training.
出处
《塑料制造》
2011年第10期60-63,共4页
Plastics Manufacture
关键词
翘曲变形
正交试验
神经网络
CAE
Warpage
Orthogonal design methods
Neural network
CAE
