基于Logistic回归的数据富裕环境下制造过程质量动态监控  

Manufacturing Process Quality Dynamic Monitoring Based on Logistic Regression Model in Data Rich Environment

作　　者：张帅[1,2] 吕晨[1] 杨剑锋 ZHANG Shuai;LYU Chen;YANG Jianfeng(Henan University of Engineering,Zhengzhou Henan 451191,China;Business School,Zhengzhou University,Zhengzhou Henan 450001,China)

机构地区：[1]河南工程学院,河南郑州451191 [2]郑州大学商学院,河南郑州450001

出　　处：《机床与液压》2023年第11期104-108,共5页Machine Tool & Hydraulics

基　　金：国家自然科学基金联合基金项目(U1904211);国家自然科学基金面上项目(71672182);国家社会科学基金(20BTJ059);河南省科技攻关项目(232102211040)。

摘　　要：针对数字化工厂“数据丰富,信息贫瘠”环境下制造车间生产过程存在异常信息利用不充分而造成监控效率低的问题,提出基于SPC技术和Logistic回归模型的制造过程质量监控方法。将关键工序中相关质量数据采集到MES系统,根据过程质量数据绘制T^(2)控制图,然后利用Logistic回归模型挖掘过程异常信息,并通过T 2和Logistic回归的联合优化实现数字化工厂制造过程质量监控的动态调整。以某薄膜晶体管液晶显示器等离子薄膜沉积生产工序为实际案例,验证了该制造过程质量监控方法的可行性和有效性。因此,在T^(2)控制图的基础上通过Logistic回归模型考虑过程异常信息能够提高制造过程质量监控的灵敏性和鲁棒性。Aiming at the problem of the abnormal information collected from manufacturing process is difficult to use and analysis based on“data rich,information poor”environment of manufacturing workshop in the digital factory,a novelty process quality monitoring technique was proposed based on SPC and Logistic regression.T 2 control chart was established by the related data from key stage in MES system.The Logistic regression model was applied to dig abnormal data,and the manufacturing process dynamic monitoring model was optimized by T^(2)and Logistic model jointly.Taking a plasma-enhanced chemical vapour deposition process of TFT-LCD manufacturing company as example,the feasibility and effectiveness of the process quality monitoring method were verified.The result shows that considering the abnormal information by using Logistic model based on T^(2)control chart can improve the sensitivity and robust of process quality monitoring.

关 键 词：数字化工厂 过程质量监控 T^(2)控制图 LOGISTIC回归 

分 类 号：TH166[机械工程—机械制造及自动化] TG506[金属学及工艺—金属切削加工及机床]