摘要
针对某汽车塑件注塑成型时成型末端翘曲量较大导致尺寸变差的问题,结合注塑成型CAE工艺分析后发现,引起产品充填末端翘曲变形大的主要原因为注塑后冷却收缩不均,针对此问题,将CAE仿真分析和RBF神经网络的预测分析相结合,对注塑工艺参数中的保压工艺和冷却工艺进行了优化设计,CAE分析方案采用(冷却+填充+保压+翘曲),RBF神经网络采用聚类法和梯度算法,应用改善翘曲的L_(27)(38)设计试验方案进行神经网络训练和检验,应用混合正交法(L_(36)(2*6 3*2))进行二次水平密化优选参数,通过优化,找到了改善翘曲的注塑工艺方案,优化的注塑工艺方案能较好的指导产品的批量生产,对其它同类注塑产品的生产有较好的实践参考意义。
Aimed at the problem of overproof difference caused by large warpage of forming end in injection molding of automobile parts, combined with injection molding CAE process analysis, the main cause of warpage at the end of product filling was uneven shrinkage after injection. To solve this problem, CAE simulation analysis and RBF neural network prediction are combined. The optimization of the pressure holding process and cooling process in the injection molding process parameters were carried out. CAE analysis program（ cooling ＋ filling ＋ pressure ＋ warpage）, RBF neural network adopts clustering method and gradient algorithm, the L27 （38） design test scheme for improving warpage was trained and tested by neural network, and the orthogonal optimization （ L36 （2 ＊ 6 3 ＊ 2） ） was used to optimize the parameters. Through optimization,the improvement of warpage of injection molding process was found,the optimized injection process scheme could better guide the batch production of the products,and had a good reference for the production of other similar injection products.
出处
《塑料》
CSCD
北大核心
2017年第3期121-125,共5页
Plastics
基金
广西中青年教师能力提升项目(KY2017YB638)
关键词
塑件
CAE分析
RBF神经网络
梯度法
聚类法
plastic parts
CAE analysis
RBF neural network
gradient method
clustering method
