Author:
Yin Hongxu,Wu Yuguang,Mu Zhijun,Geng Hongbin,Li Xiaobo
Abstract
Abstract
To settle the issue of conventional principal component analysis similarity factor cannot take advantage of higher-order statistics information of process variables, we presented an improved statistics principal analysis similarity factor (SPASF) to identify fault patterns in our work. In the improved SPASF approach, process data is first converted from original space into a new statistics space by the means of statistics pattern analysis (SPA) technology, and principal component analysis (PCA) is then employed to extract principal components in statistics space. At last, the pattern of snapshot dataset is recognized by measuring the similarity of principal components derived from statistics snapshot dataset and statistics historical fault dataset. The effectiveness of suggested SPASF based approach is verified through a case study on continuous stirring tank reactor (CSTR) by the means of recognizing fault patterns of snapshot datasets.
Subject
General Physics and Astronomy
Reference18 articles.
1. Multivariate statistical process monitoring and control: recent developments and applications to chemical industry;Liang;Chin. J. Chem. Eng.,2003
2. A multi-level approach for complex fault isolation based on structured residuals;Ye;Chin. J. Chem. Eng.,2011
3. Survey on data driven fault diagnosis methods;Li;Control and Decision,2011
4. A comparison study of basic data-driven fault diagnosis and process monitoring methods on the benchmark tennessee eastman process;Shen;Journal of Process Control,2012
5. Survey on data-driven industrial process monitoring and diagnosis;Qin;Annual Reviews in Control,2012