Process Monitoring Based on Independent Component Contribution  

作　　者：吕小条 宋冰 侍洪波 谭帅 

机构地区：[1]Key Laboratory of Advanced Control and Optimization for Chemical Processes of Ministry of Education, East China University ofScience and Technology, Shanghai 200237, China

出　　处：《Journal of Donghua University(English Edition)》2017年第3期349-354,共6页东华大学学报（英文版）

基　　金：National Natural Science Foundations of China(Nos.61374140,61403072,61673173);Fundamental Research Funds for the Central Universities,China(Nos.222201717006,222201714031)

摘　　要：Independent component analysis( ICA) has been widely applied to the monitoring of non-Gaussian processes. Despite lots of applications,there is no universally accepted criterion to select the dominant independent components( ICs). Moreover, how to determine the number of dominant ICs is still an open question. To further address this issue,a novel process monitoring based on IC contribution( ICC) is proposed from the perspective of information storage. Based on the ICC with each variable,the dominant ICs can be obtained and the number of dominant ICs is determined objectively. To further preserve the process information, the remaining ICs are not useless. As a result,all the ICs are regarded to be divided into dominant and residual subspaces. The monitoring models are established respectively in each subspace, and then Bayesian inference is applied to integrating monitoring results of the two subspaces. Finally, the feasibility and effectiveness of the proposed method are illustrated through a numerical example and the Tennessee Eastman process.Independent component analysis（ ICA） has been widely applied to the monitoring of non-Gaussian processes. Despite lots of applications,there is no universally accepted criterion to select the dominant independent components（ ICs）. Moreover, how to determine the number of dominant ICs is still an open question. To further address this issue,a novel process monitoring based on IC contribution（ ICC） is proposed from the perspective of information storage. Based on the ICC with each variable,the dominant ICs can be obtained and the number of dominant ICs is determined objectively. To further preserve the process information, the remaining ICs are not useless. As a result,all the ICs are regarded to be divided into dominant and residual subspaces. The monitoring models are established respectively in each subspace, and then Bayesian inference is applied to integrating monitoring results of the two subspaces. Finally, the feasibility and effectiveness of the proposed method are illustrated through a numerical example and the Tennessee Eastman process.

关 键 词：Independent Eastman Tennessee Bayesian preserve illustrated criterion universally remaining integrating 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]