Fault detection for chemical process based on robust PLS

Author

Hu Yi ; Ma Hehe ; Shi Hongbo

Author_Institution

Key Lab. of Adv. Control & Optimization for Chem. Processes of Minist. of Educ., East China Univ. of Sci. & Technol., Shanghai, China

fYear

2013

fDate

25-27 May 2013

Firstpage

4947

Lastpage

4952

Abstract

The training dataset collected from industrial processes usually contain some outliers, and partial least squares (PLS) regression will have poor properties due to the sensitiveness of PLS to outliers. Under this circumstance, a multivariate statistical process monitoring method based on robust PLS (RPLS) is developed. By means of weight strategy, RPLS can eliminate the effects of the outliers in the original data and construct precise model. Then, robust monitoring statistics and control limits are derived for process monitoring purposes. A case study of the Tennessee Eastman (TE) process illustrated that the proposed approach showed superior process monitoring performance compared to conventional PLS when the modeling data set contains outliers.

Keywords

chemical industry; fault diagnosis; least squares approximations; manufacturing processes; process monitoring; regression analysis; PLS regression; TE process; Tennessee Eastman process; chemical process; control limit; fault detection; multivariate statistical process monitoring method; partial least squares regression; weight strategy; Chemical processes; Data models; Electronic mail; Fault detection; Monitoring; Process control; Robustness; Fault detection; Outliers; Partial least squares; Robust partial least square;

fLanguage

English

Publisher

ieee

Conference_Titel

Control and Decision Conference (CCDC), 2013 25th Chinese

Conference_Location

Guiyang

Print_ISBN

978-1-4673-5533-9

Type

conf

DOI

10.1109/CCDC.2013.6561830

Filename

6561830