Comparison of artificial neural networks and dynamic principal component analysis for fault diagnosis

Juan C. Tudón-Martínez, Ruben Morales-Menendez, Luis Garza-Castañón, Ricardo Ramirez-Mendoza

Research output: Chapter in Book/Report/Conference proceedingConference contribution

4 Citations (Scopus)

Abstract

Dynamic Principal Component Analysis (DPCA) and Artificial Neural Networks (ANN) are compared in the fault diagnosis task. Both approaches are process history based methods, which do not assume any form of model structure, and rely only on process historical data. Faults in sensors and actuators are implemented to compare the online performance of both approaches in terms of quick detection, isolability capacity and multiple faults identifiability. An industrial heat exchanger was the experimental test-bed system. Multiple faults in sensors can be isolated using an individual control chart generated by the principal components; the error of classification was 15.28% while ANN presented 4.34%. For faults in actuators, ANN showed instantaneous detection and 14.7% lower error classification. However, DPCA required a minor computational effort in the training step. © 2011 Springer-Verlag.
Original languageEnglish
Title of host publicationComparison of artificial neural networks and dynamic principal component analysis for fault diagnosis
Pages10-18
Number of pages9
ISBN (Electronic)9783642218217
DOIs
Publication statusPublished - 25 Jul 2011
Externally publishedYes
EventLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) -
Duration: 25 Jul 2011 → …

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume6703 LNAI
ISSN (Print)0302-9743

Conference

ConferenceLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Period25/7/11 → …

Fingerprint

Fault Diagnosis
Dynamic Analysis
Principal component analysis
Principal Component Analysis
Failure analysis
Artificial Neural Network
Fault
Neural networks
Actuators
Actuator
Sensors
Model structures
Sensor
Heat exchangers
Heat Exchanger
Identifiability
Control Charts
Historical Data
Principal Components
Testbed

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Tudón-Martínez, J. C., Morales-Menendez, R., Garza-Castañón, L., & Ramirez-Mendoza, R. (2011). Comparison of artificial neural networks and dynamic principal component analysis for fault diagnosis. In Comparison of artificial neural networks and dynamic principal component analysis for fault diagnosis (pp. 10-18). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6703 LNAI). https://doi.org/10.1007/978-3-642-21822-4_2
Tudón-Martínez, Juan C. ; Morales-Menendez, Ruben ; Garza-Castañón, Luis ; Ramirez-Mendoza, Ricardo. / Comparison of artificial neural networks and dynamic principal component analysis for fault diagnosis. Comparison of artificial neural networks and dynamic principal component analysis for fault diagnosis. 2011. pp. 10-18 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{d8b5c73df1a642c186553810ed86618d,
title = "Comparison of artificial neural networks and dynamic principal component analysis for fault diagnosis",
abstract = "Dynamic Principal Component Analysis (DPCA) and Artificial Neural Networks (ANN) are compared in the fault diagnosis task. Both approaches are process history based methods, which do not assume any form of model structure, and rely only on process historical data. Faults in sensors and actuators are implemented to compare the online performance of both approaches in terms of quick detection, isolability capacity and multiple faults identifiability. An industrial heat exchanger was the experimental test-bed system. Multiple faults in sensors can be isolated using an individual control chart generated by the principal components; the error of classification was 15.28{\%} while ANN presented 4.34{\%}. For faults in actuators, ANN showed instantaneous detection and 14.7{\%} lower error classification. However, DPCA required a minor computational effort in the training step. {\circledC} 2011 Springer-Verlag.",
author = "Tud{\'o}n-Mart{\'i}nez, {Juan C.} and Ruben Morales-Menendez and Luis Garza-Casta{\~n}{\'o}n and Ricardo Ramirez-Mendoza",
year = "2011",
month = "7",
day = "25",
doi = "10.1007/978-3-642-21822-4_2",
language = "English",
isbn = "9783642218217",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "10--18",
booktitle = "Comparison of artificial neural networks and dynamic principal component analysis for fault diagnosis",

}

Tudón-Martínez, JC, Morales-Menendez, R, Garza-Castañón, L & Ramirez-Mendoza, R 2011, Comparison of artificial neural networks and dynamic principal component analysis for fault diagnosis. in Comparison of artificial neural networks and dynamic principal component analysis for fault diagnosis. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 6703 LNAI, pp. 10-18, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 25/7/11. https://doi.org/10.1007/978-3-642-21822-4_2

Comparison of artificial neural networks and dynamic principal component analysis for fault diagnosis. / Tudón-Martínez, Juan C.; Morales-Menendez, Ruben; Garza-Castañón, Luis; Ramirez-Mendoza, Ricardo.

Comparison of artificial neural networks and dynamic principal component analysis for fault diagnosis. 2011. p. 10-18 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6703 LNAI).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Comparison of artificial neural networks and dynamic principal component analysis for fault diagnosis

AU - Tudón-Martínez, Juan C.

AU - Morales-Menendez, Ruben

AU - Garza-Castañón, Luis

AU - Ramirez-Mendoza, Ricardo

PY - 2011/7/25

Y1 - 2011/7/25

N2 - Dynamic Principal Component Analysis (DPCA) and Artificial Neural Networks (ANN) are compared in the fault diagnosis task. Both approaches are process history based methods, which do not assume any form of model structure, and rely only on process historical data. Faults in sensors and actuators are implemented to compare the online performance of both approaches in terms of quick detection, isolability capacity and multiple faults identifiability. An industrial heat exchanger was the experimental test-bed system. Multiple faults in sensors can be isolated using an individual control chart generated by the principal components; the error of classification was 15.28% while ANN presented 4.34%. For faults in actuators, ANN showed instantaneous detection and 14.7% lower error classification. However, DPCA required a minor computational effort in the training step. © 2011 Springer-Verlag.

AB - Dynamic Principal Component Analysis (DPCA) and Artificial Neural Networks (ANN) are compared in the fault diagnosis task. Both approaches are process history based methods, which do not assume any form of model structure, and rely only on process historical data. Faults in sensors and actuators are implemented to compare the online performance of both approaches in terms of quick detection, isolability capacity and multiple faults identifiability. An industrial heat exchanger was the experimental test-bed system. Multiple faults in sensors can be isolated using an individual control chart generated by the principal components; the error of classification was 15.28% while ANN presented 4.34%. For faults in actuators, ANN showed instantaneous detection and 14.7% lower error classification. However, DPCA required a minor computational effort in the training step. © 2011 Springer-Verlag.

U2 - 10.1007/978-3-642-21822-4_2

DO - 10.1007/978-3-642-21822-4_2

M3 - Conference contribution

SN - 9783642218217

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 10

EP - 18

BT - Comparison of artificial neural networks and dynamic principal component analysis for fault diagnosis

ER -

Tudón-Martínez JC, Morales-Menendez R, Garza-Castañón L, Ramirez-Mendoza R. Comparison of artificial neural networks and dynamic principal component analysis for fault diagnosis. In Comparison of artificial neural networks and dynamic principal component analysis for fault diagnosis. 2011. p. 10-18. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-642-21822-4_2