摘要
为保证注塑制品的质量与精度要求,以检测仪外壳的翘曲变形量和体积收缩率为优化目标,采用有限元分析软件Moldflow对其进行模拟分析。选取熔体温度、模具温度、注射压力、保压压力、保压时间、冷却时间6个工艺参数及结构参数(浇口直径)作为输入量,翘曲变形量和体积收缩率作为输出量,建立深度神经网络(DNN),并且,对网络进行改进。将混合水平正交试验得到的数据作为样本,对神经网络进行训练和测试,得到输入量和输出量之间的非线性映射关系。结合非支配排序遗传算法(NSGA-Ⅱ)对浇口直径及工艺参数进行优化,优化后,塑件的翘曲变形量为0.3684 mm,体积收缩率为6.236%,与优化前相比,分别降低了67%、39%。
In order to ensure the quality and precision of the injection products,the warping deformation and volume shrinkage of the detector shell were taken as the optimization targets,and the finite element analysis software Moldflow was used to simulate and analyze them.The deep neural network(DNN)was established and improved by selecting six technological parameters and structural parameters(gate diameter),including melt temperature,mold temperature,injection pressure,holding pressure,holding time and cooling time as the input,and the warping deformation and volume shrinkage as the output.The neural network was trained and tested by using the data obtained from mixed horizontal orthogonal experiment,and the nonlinear mapping relationship between input and output was obtained.The gate diameter and process parameters were optimized with non-dominant sequencing genetic algorithm(NSGA-Ⅱ).The optimized warping deformation was 0.3684 mm,and the volume shrinkage was 6.236%,which were 67% and 39% lower than those before optimization,respectively.
作者
王常晶
范希营
刘欣
李春晓
王德炤
WANG Changjing;FAN Xiying;LIU Xin;LI Chunxiao;WANG Dezhao(School of Mechanical and Electrical Engineering,Jiangsu Normal University,Xuzhou,Jiangsu 221116,China)
出处
《塑料》
CAS
CSCD
北大核心
2023年第2期181-186,共6页
Plastics
基金
江苏师范大学研究生科研创新计划资助项目(2021XKT0358)。
关键词
注塑成型
CAE分析
深度神经网络
非支配排序遗传算法
多目标优化
injection molding
CAE analysis
deep neural network
non-dominant sequencing genetic algorithm
multiobjectiveoptimization
作者简介
王常晶(1997-),女,在读硕士研究生,主要从事模具设计及注塑成型等方面的研究;通信作者:范希营(1965-),男,硕士,教授,主要从事模具生产智能化等方面的研究。E-mail:fxy8441@163.com。
