文献类型：期刊文献

英文题名：Fault detection in time-varying chemical process through incremental principal component analysis

作者：Gao, Yong[1];Wang, Xin[2];Wang, Zhenlei[1];Zhao, Liang[1]

机构：[1]East China Univ Sci & Technol, Ministiy Educ, Key Lab Adv Control & Optimizat Chem Proc, Shanghai 200237, Peoples R China;[2]Shanghai Jiao Tong Univ, Ctr Elect & Elect Technol, Shanghai, Peoples R China

年份：2016

卷号：158

起止页码：102

外文期刊名：CHEMOMETRICS AND INTELLIGENT LABORATORY SYSTEMS

收录：;EI(收录号：20242616330937);WOS:【SCI-EXPANDED(收录号:WOS:000390504900012)】；

基金：This work is supported by the Major State Basic Research Development Program of China (2012CB720500), National Natural Science Foundation of China (61174118), Shanghai Natural Science Foundation (14ZR1421800, 16ZR1407300) and the State Key Laboratory of Synthetical Automation for Process Industries (PAL-N201404).

语种：英文

外文关键词：Time-varying; Slow ramp fault; Incremental PCA; Fault detection

摘要：Incremental principal component analysis (IPCA) is proposed to improve the detection performance of a slow ramp fault in the time-varying chemical process. Conventional monitoring methods of the time varying process such as recursive method and moving window strategy, which update the monitoring model and control limit when the newly monitored sample is detected as a normal one, track the slow ramp fault and lose the ability to detect this kind of fault. In this study, the incremental principal components (IPCs) describing time-varying information are proposed to extract the normal time-varying information. This study proposes IPCA method based on IPCs for process monitoring of the time-varying processes. The monitoring model remained unchanged because the normal time-varying information has already been identified by IPCs. The method can distinguish between the slow ramp fault from the normal time-varying process. Two numeric case studies demonstrate the efficiency of the method. Application of the method to an acetylene hydrogenation reactor is also provided. (C) 2016 Published by Elsevier B.V.