摘要
针对注塑生产过程中人工质量检测存在的效率低、成本高等问题,提出了一种基于注塑加工过程数据来对产品尺寸是否合格进行预测判定的方法。先对清洗后的数据集采用5折交叉验证筛选出LR(logistic regression)模型、SVM(support vector machine)模型等5个分类模型,再以ROC(receiver operating characteristic)曲线和AUC(area under curve)值作为性能评估指标,综合比较和分析了5个分类模型在不同特征选取方法下的分类性能。结果表明:基于树模型特征选取与LR分类模型组合对本文的数据集表现出了优良的分类性能,准确率可达96.42%,具有一定的工程应用价值。
Aiming at the problems of low efficiency and high cost of manual quality inspection in injection molding production process, a method for predicting and judging whether the product size was qualified based on injection molding process data was proposed. Firstly, five classification models such as LR(logistic regression model) and SVM(support vector machine) model were selec-ted by 50% cross validation on the cleaned data set, and then the classification performance of the five classification models under different feature selection methods was comprehensively compared and analyzed with ROC(receiver operating characteristic) curve and AUC(area under curve) value as performance evaluation indicators. The results show that the combination of tree model feature selection and LR classification model shows excellent classification performance for the data set, and the accuracy can reach 96.42%, which has a certain engineering application value.
作者
宋建
王宇峰
梁家睿
李东
SONG Jian;WANG Yu-feng;LIANG Jia-rui;LI Dong(Guangdong Advanced Polymer Manufacturing Technology and Equipment Key Laboratory,South China University of Technology,Guangzhou 510640,China;Key Laboratory of Polymer Processing Engineering,the Ministry of Education,South China University of Technology,Guangzhou 510640,China;Enterprise Technology Center of Kingfa Sci.&Tech.Co.,Ltd,Guangzhou 510663,China)
出处
《科学技术与工程》
北大核心
2022年第27期12000-12005,共6页
Science Technology and Engineering
基金
国家重点研发计划(2019YFB1704900)。
关键词
合格性预测
注塑成型
分类模型
特征提取
qualification prediction
injection molding
classification model
feature extraction