Fault detection using multiscale PCA-based moving window GLRT

M. Ziyan Sheriff, Majdi Mansouri, M. Nazmul Karim, Hazem Nounou, Mohamed Nounou

Research output: Contribution to journalArticle

33 Citations (Scopus)

Abstract

The presence of measurement errors (noise) in the data and mode l uncertainties degrade the performance quality of fault detection (FD) techniques. Therefore, an objective of this paper is to enhance the quality of FD by suppressing the effect of these errors using wavelet-based multiscale representation of data, which is a powerful feature extraction tool. Multiscale representation of data has been used to improve the FD abilities of principal component analysis. Thus, combining the advantages of multiscale representation with those of hypothesis testing should provide further improvements in FD. To do that, a moving window generalized likelihood ratio test (MW-GLRT) method based on multiscale principal component analysis (MSPCA) is proposed for FD. The dynamical multiscale representation is proposed to extract the deterministic features and decorrelate autocorrelated measurements. An extension of the popular hypothesis testing GLRT method is applied on the residuals from the MSPCA model, in order to further enhance the fault detection performance. In the proposed MW-GLRT method, the detection statistic equals the norm of the residuals in that window, which is equivalent to applying a mean filter on the squares of the residuals. This means that a proper moving window length needs to be selected, which is similar to estimating the mean filter length in data filtering. The fault detection performance of the MSPCA-based MW-GLRT chart is illustrated through two examples, one using synthetic data, and the other using simulated Tennessee Eastman Process (TEP) data. The results demonstrate the effectiveness of the MSPCA-based MW-GLRT method over the conventional PCA-based and MSPCA-based GLRT methods, and both of them provide better performance results when compared with the conventional PCA and MSPCA methods, through their respective charts T2 and Q charts.

Original languageEnglish
Pages (from-to)47-64
Number of pages18
JournalJournal of Process Control
Volume54
DOIs
Publication statusPublished - 1 Jun 2017

Fingerprint

Fault Detection
Multiscale Analysis
Fault detection
Principal component analysis
Principal Component Analysis
Generalized Likelihood Ratio Test
Chart
Representation of data
Hypothesis Testing
Filter
Testing
Synthetic Data
Measurement errors
Measurement Error
Feature Extraction
Statistic
Feature extraction
Wavelets
Filtering
Statistics

Keywords

  • Fault detection
  • Generalized likelihood ratio test
  • Moving window
  • Multiscale principal component analysis
  • Tennessee Eastman Process

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Modelling and Simulation
  • Computer Science Applications
  • Industrial and Manufacturing Engineering

Cite this

Fault detection using multiscale PCA-based moving window GLRT. / Sheriff, M. Ziyan; Mansouri, Majdi; Karim, M. Nazmul; Nounou, Hazem; Nounou, Mohamed.

In: Journal of Process Control, Vol. 54, 01.06.2017, p. 47-64.

Research output: Contribution to journalArticle

@article{20975c29a2c346158c30c0138fb8fd8a,
title = "Fault detection using multiscale PCA-based moving window GLRT",
abstract = "The presence of measurement errors (noise) in the data and mode l uncertainties degrade the performance quality of fault detection (FD) techniques. Therefore, an objective of this paper is to enhance the quality of FD by suppressing the effect of these errors using wavelet-based multiscale representation of data, which is a powerful feature extraction tool. Multiscale representation of data has been used to improve the FD abilities of principal component analysis. Thus, combining the advantages of multiscale representation with those of hypothesis testing should provide further improvements in FD. To do that, a moving window generalized likelihood ratio test (MW-GLRT) method based on multiscale principal component analysis (MSPCA) is proposed for FD. The dynamical multiscale representation is proposed to extract the deterministic features and decorrelate autocorrelated measurements. An extension of the popular hypothesis testing GLRT method is applied on the residuals from the MSPCA model, in order to further enhance the fault detection performance. In the proposed MW-GLRT method, the detection statistic equals the norm of the residuals in that window, which is equivalent to applying a mean filter on the squares of the residuals. This means that a proper moving window length needs to be selected, which is similar to estimating the mean filter length in data filtering. The fault detection performance of the MSPCA-based MW-GLRT chart is illustrated through two examples, one using synthetic data, and the other using simulated Tennessee Eastman Process (TEP) data. The results demonstrate the effectiveness of the MSPCA-based MW-GLRT method over the conventional PCA-based and MSPCA-based GLRT methods, and both of them provide better performance results when compared with the conventional PCA and MSPCA methods, through their respective charts T2 and Q charts.",
keywords = "Fault detection, Generalized likelihood ratio test, Moving window, Multiscale principal component analysis, Tennessee Eastman Process",
author = "Sheriff, {M. Ziyan} and Majdi Mansouri and Karim, {M. Nazmul} and Hazem Nounou and Mohamed Nounou",
year = "2017",
month = "6",
day = "1",
doi = "10.1016/j.jprocont.2017.03.004",
language = "English",
volume = "54",
pages = "47--64",
journal = "Journal of Process Control",
issn = "0959-1524",
publisher = "Elsevier Limited",

}

TY - JOUR

T1 - Fault detection using multiscale PCA-based moving window GLRT

AU - Sheriff, M. Ziyan

AU - Mansouri, Majdi

AU - Karim, M. Nazmul

AU - Nounou, Hazem

AU - Nounou, Mohamed

PY - 2017/6/1

Y1 - 2017/6/1

N2 - The presence of measurement errors (noise) in the data and mode l uncertainties degrade the performance quality of fault detection (FD) techniques. Therefore, an objective of this paper is to enhance the quality of FD by suppressing the effect of these errors using wavelet-based multiscale representation of data, which is a powerful feature extraction tool. Multiscale representation of data has been used to improve the FD abilities of principal component analysis. Thus, combining the advantages of multiscale representation with those of hypothesis testing should provide further improvements in FD. To do that, a moving window generalized likelihood ratio test (MW-GLRT) method based on multiscale principal component analysis (MSPCA) is proposed for FD. The dynamical multiscale representation is proposed to extract the deterministic features and decorrelate autocorrelated measurements. An extension of the popular hypothesis testing GLRT method is applied on the residuals from the MSPCA model, in order to further enhance the fault detection performance. In the proposed MW-GLRT method, the detection statistic equals the norm of the residuals in that window, which is equivalent to applying a mean filter on the squares of the residuals. This means that a proper moving window length needs to be selected, which is similar to estimating the mean filter length in data filtering. The fault detection performance of the MSPCA-based MW-GLRT chart is illustrated through two examples, one using synthetic data, and the other using simulated Tennessee Eastman Process (TEP) data. The results demonstrate the effectiveness of the MSPCA-based MW-GLRT method over the conventional PCA-based and MSPCA-based GLRT methods, and both of them provide better performance results when compared with the conventional PCA and MSPCA methods, through their respective charts T2 and Q charts.

AB - The presence of measurement errors (noise) in the data and mode l uncertainties degrade the performance quality of fault detection (FD) techniques. Therefore, an objective of this paper is to enhance the quality of FD by suppressing the effect of these errors using wavelet-based multiscale representation of data, which is a powerful feature extraction tool. Multiscale representation of data has been used to improve the FD abilities of principal component analysis. Thus, combining the advantages of multiscale representation with those of hypothesis testing should provide further improvements in FD. To do that, a moving window generalized likelihood ratio test (MW-GLRT) method based on multiscale principal component analysis (MSPCA) is proposed for FD. The dynamical multiscale representation is proposed to extract the deterministic features and decorrelate autocorrelated measurements. An extension of the popular hypothesis testing GLRT method is applied on the residuals from the MSPCA model, in order to further enhance the fault detection performance. In the proposed MW-GLRT method, the detection statistic equals the norm of the residuals in that window, which is equivalent to applying a mean filter on the squares of the residuals. This means that a proper moving window length needs to be selected, which is similar to estimating the mean filter length in data filtering. The fault detection performance of the MSPCA-based MW-GLRT chart is illustrated through two examples, one using synthetic data, and the other using simulated Tennessee Eastman Process (TEP) data. The results demonstrate the effectiveness of the MSPCA-based MW-GLRT method over the conventional PCA-based and MSPCA-based GLRT methods, and both of them provide better performance results when compared with the conventional PCA and MSPCA methods, through their respective charts T2 and Q charts.

KW - Fault detection

KW - Generalized likelihood ratio test

KW - Moving window

KW - Multiscale principal component analysis

KW - Tennessee Eastman Process

UR - http://www.scopus.com/inward/record.url?scp=85015863708&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85015863708&partnerID=8YFLogxK

U2 - 10.1016/j.jprocont.2017.03.004

DO - 10.1016/j.jprocont.2017.03.004

M3 - Article

VL - 54

SP - 47

EP - 64

JO - Journal of Process Control

JF - Journal of Process Control

SN - 0959-1524

ER -