基于改进的BP神经网络的注塑成型翘曲优化设计  

Warpage optimization of injection molding based on improved BP neural network

作　　者：时慧焯[1] 王希诚[1] 

机构地区：[1]大连理工大学工业装备结构分析国家重点实验室,辽宁大连116024

出　　处：《化工学报》2011年第9期2562-2568,共7页CIESC Journal

基　　金：国家自然科学基金项目(11072048)~~

摘　　要：注塑成型是制造塑料产品应用最广泛的一种方法。整个注塑成型过程一般分为注射、保压和冷却3个阶段。成型过程中的翘曲变形是注塑制品一种严重的缺陷。由于注塑制品质量主要受工艺条件影响,所以如何确定最佳工艺条件来减少翘曲变形成为改进注塑制品质量的一个关键。以模具温度、熔体温度、注射时间、保压时间、保压压力和冷却时间为设计变量,运行Moldflow软件进行制品的翘曲变形分析,用BP神经网络模型来建立翘曲变形与设计变量的函数关系,加权形式的期望提高加点准则实现序列的迭代优化设计。这种加点准则能调整局部和全局搜索,在保证计算效率的同时提高对全局最优解的逼近程度。通过实例验证,所提出的优化方法能有效地减小注塑制品的翘曲变形。Injection molding is the most widely used process for producing plastic products.In general,injection processing can be divided into three stages： filling,packing and cooling,in which warpage defect is one of the most important quality problems.Since the quality of the injection-molded parts are mostly influenced by process conditions,how to determine the optimum process conditions to reduce warpage becomes the key to improving the quality of parts.In this study,the mold temperature,melt temperature,injection time,packing time,packing pressure and cooling time were regarded as process parameters（design variables）.Moldflow Plastic Insight software was used to analyze the warpage of the injection molding parts.BP neural network model was used to construct an approximate function relationship between warpage and the process parameters,replacing the expensive simulation analysis in the optimization iterations.The adaptive process was performed by improved Expected Improvement（EI）which was a weighted infill sample criterion.This criterion could balance local and global search and tend to find the global optimal design.Numerical results showed that the proposed adaptive optimization method could effectively reduce the warpage of the injection molding parts.

关 键 词：注塑成型 翘曲 优化 BP神经网络 期望提高 

分 类 号：TQ320.66[化学工程—合成树脂塑料工业]