基于GRNN-NSGA的带嵌件注塑件多目标优化  

Multi-objective Optimization of Injection Molded Parts with Insert Based on GRNN-NSGA

作　　者：李春晓 范希营 郭永环 刘欣 曹艳丽 李璐璐 LI Chun-xiao;FAN Xi-ying;GUO Yong-huan;LIU Xin;CAO Yan-li;LI Lu-lu(School of Mechanical and Electrical Engineering,Jiangsu Normal University,Xuzhou 221116,China;CNPC Engineering Technology R&D Company Limited,Beijing 102206,China)

机构地区：[1]江苏师范大学机电工程学院,江苏徐州221116 [2]中国石油集团工程技术研究院有限公司,北京102206

出　　处：《塑料工业》2021年第2期84-88,共5页China Plastics Industry

基　　金：国家自然科学基金(51475220);江苏师范大学研究生科研创新计划(2020XKT188)。

摘　　要：带嵌件的注塑产品成型过程相较于传统注塑产品较为复杂,产品成型周期和产品质量难以预测。针对这一问题,以带嵌件的静电检测盒为例,运用广义神经网络(GRNN)和非支配排序遗传算法(NSGA-Ⅱ),对注塑成型过程进行控制与优化。以熔体温度、模具温度、注射时间、冷却时间、保压压力和保压时间为输入层,体积收缩率、X方向翘曲变形、Z方向翘曲变形作为输出层,建立GRNN模型。利用正交试验设计得到的样本对神经网络模型进行训练和测试,运用NSGA-Ⅱ对建立的模型进行优化,最终三个目标值分别降低了30.96%、22.76%、15.62%,表明该方法可以对注塑成型过程进行预测和控制。Comparing with traditional injection molding products,the molding process of injection molding products with embedded parts was more complex,and the molding cycle and quality of products were difficult to predict.In order to solve this problem,taking:the electrostatic detection box with embedded parts as an example,the injection molding process was controlled and optimized by using general regression neural network(GRNN) and non-dominant sequencing genetic algorithm(NSGA-Ⅱ).Taking melt temperature,mold temperature,injection time,cooling time,holding pressure and holding time as input layer and volume shrinkage,X direction warping deformation and Z direction warping deformation as output layer,GRNN model was established.The neural network was trained and tested by using the samples obtained from orthogonal experimental design.NSGA-Ⅱ was used to optimize the established model.And the final results show that the three target values reduce 30.96%,22.76% and 15.62%,respectively.It indicates that the method could predict and control the injection molding quality.

关 键 词：注塑成型 嵌件 广义回归神经网络 非支配排序遗传算法 多目标优化 

分 类 号：TQ320.66[化学工程—合成树脂塑料工业]