摘要
鉴于夹芯注射成型充模流动过程属多相分层流动,其影响因素错综复杂,很难用线性关系将工艺参数与层间界面形状关联起来;人工神经网络具有很强的信息综合能力,具有良好的非线性逼近功能两方面的情况。针对具体层间界面形状,基于贝叶斯正则化神经网络预测工艺参数,并借助MPI软件的co-injection模块检验。结果表明其误差完全达到了工程实用的精度,证明提出的贝叶斯正则化神经网络可应用于研究夹芯注塑中的非线性函数映射问题。
Melt flow of sequential co-injection molding belong to multifluid flow, whose influencing factors were so complex that it was difficult to linearly correlate processing parameter with layer interface of sequential co-injection. Artificial neural network （ANN） was capable of synthesizing the information and treating the nonlinear relation all right. For the specific layer interface, the processing parameter was forecasted based on bayesian-regularization neural networks （BRNN）, which was certificated by means of the co-injection module in the software of Mgldflow Plastics Insight. The error reached the precision of engineering practice, which indicated that BRNN established could be applied to the research on nonlinear problem of sequential co-injection molding.
出处
《工程塑料应用》
CAS
CSCD
北大核心
2008年第8期41-43,共3页
Engineering Plastics Application
关键词
夹芯注射成型
层间界面
人工神经网络
贝叶斯正则化神经网络
sandwich injection molding, layer interface, artificial neural network, bayesian-regularization neural networks
