An improved multiple maximum scatter difference for process fault detection. (8th July 2019)

• Record Type: Journal Article
• Authors: Guo, Xiaoping; Liu, Shiyang; Li, Yuan
• Is Part Of: Journal of chemometrics. Volume 33:Number 7(2019)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2019-07-08
• Subjects: fault detection -- feature space -- local neighborhood standardization (LNS) -- multi‐mode processes -- multiple maximum scatter difference (MMSD); Chemistry -- Mathematics -- Periodicals; Chemistry -- Statistical methods -- Periodicals; 542.85
• DOI: 10.1002/cem.3142
• Languages: English
• ISSNs: 0886-9383

Abstract

Abstract: Multiple maximum scatter difference (MMSD) method used in face recognition is first applied to process fault detection and an improved method is also proposed in this paper. MMSD method is used to extract feature space of normal data of processes and construct a statistic in the feature space for fault detection. However, MMSD method can be much effective for single‐mode processes. So for multi‐mode processes, an improved method—improved multiple maximum scatter difference (IMMSD)—is proposed, in which a data processing method local neighborhood standardization (LNS) is applied. The three‐phase flow facility process, a numerical example, and the Tennessee Eastman process are used to illustrate the efficiency of the proposed methods. Abstract : Multiple maximum scatter difference (MMSD) method used in face recognition is first applied to process fault detection. In order to more effectively detect faults in multi‐mode processes, an improved method IMMSD is proposed, in which a data processing method local neighborhood standardization (LNS) is applied. The three‐phase flow facility process, a numerical example, and the Tennessee Eastman process are used to illustrate the efficiency of the proposed methods.