基于改进支持向量回归的旋片泵自相关过程监控  

The Process Monitoring of Vane Pump Autocorrelation Based on Improved Support Vector Regression

作　　者：郑辉[1] 姜美玲 王东菲 ZHENG Hui;JIANG Meiling;WANG Dongfei(Department of Industrial Engineering,Tianjin University of Science and Technology,Tianjin,300222)

机构地区：[1]天津科技大学工业工程系,天津300222

出　　处：《工业工程与管理》2022年第6期9-13,共5页Industrial Engineering and Management

基　　金：科技部创新方法工作专项资助(2019IM020300);2018年国家社科基金年度项目(18BTQ032);天津市教委科研计划项目成果“面向社会网络的个性化知识服务研究”(2017SK007)。

摘　　要：为了更好地实现对旋片泵自相关过程有效监控,基于混沌遗传算法(chaotic genetic algorithm,CGA),提出优化传统支持向量回归机(support vector regression,SVR)的方法,简称为CGASVR。并运用CGA-SVR对单级旋片泵的噪音和温度两个质量特性自相关统计过程进行监控,实践证明该方法不仅可以有效滤除自相关,而且能克服SVR模型的超参数过早陷入局部最优这一缺陷。进而基于CGA-SVR模型构建的多变量残差MEWMA控制图,在旋片泵自相关监控过程中有更好体现。In order to effectively monitor the autocorrelation process of the rotary vine pump,a new method of optimizing traditional support vector regression(SVR)was proposed based on chaotic genetic algorithm(CGA). CGA-SVR was used to monitor the autocorrelation statistical process of the noise and temperature of the single-stage rotary vane pump. The practice proves that this method can not only filter out the autocorrelation effectively,but also overcome the defect that the super parameters of the SVR model fall into the local optimum precociously. Furthermore,the multi-variable residual MEWMA control chart based on CGA-SVR model is better reflected in the autocorrelation monitoring process of rotary vane pump.

关 键 词：遗传算法 支持向量回归机 旋片泵 自相关 MEWMA控制图 

分 类 号：TH166[机械工程—机械制造及自动化]