Regular Papers

International Journal of Control, Automation, and Systems 2024; 22(11): 3499-3508

https://doi.org/10.1007/s12555-024-0081-3

© The International Journal of Control, Automation, and Systems

A Novel Approach for Fault Detection and Diagnosis in Multi-mode Processes Based on PCA, Random Forest, and K-means Clustering

Julieth Mendoza-Díaz, Camilo Cueto-Barboza, Ivan Portnoy*, and Ana C. Torregroza-Espinosa

Universidad de la Costa

Abstract

The multi-mode nature of many processes poses a challenge to fault detection and diagnosis tasks, as their non-monotonous behavior renders the traditional monitoring approaches incapable of capturing such varying patterns. This study introduces the F-DINAMITE algorithm, a novel fault detection and diagnosis method incorporating principal component analysis, random forest, and k-means clustering, aiming to identify normal and faulty operational conditions while reducing false alarms and detection errors and providing fault diagnosis capabilities. The novel features of F-DINAMITE include the tailored selection of the number of clusters to minimize the false alarm rate and the automatic update of the fault database when new faults are identified. We validated the proposed method with a case study of a complex building HVAC system, where the algorithm analyzed faults in air dampers and cooling valve systems, testing its detection and diagnosis performance. Moreover, we assessed its false alarm rate while monitoring normal operation data. F-DINAMITE achieves a reduction in false alarm and misdetection rates, compared to those of conventional PCA, while successfully diagnosing faults. This methodology could significantly contribute to maintenance in complex processes by expediting fault detection and reducing associated costs. Nevertheless, future research must focus on conducting a more comprehensive assessment of F-DINAMITE’s performance with more real-life case studies or high amounts of synthetic data.

Keywords Diagnosis, fault detection, multi-mode processes, random forest.

Article

Regular Papers

International Journal of Control, Automation, and Systems 2024; 22(11): 3499-3508

Published online November 1, 2024 https://doi.org/10.1007/s12555-024-0081-3

Copyright © The International Journal of Control, Automation, and Systems.

A Novel Approach for Fault Detection and Diagnosis in Multi-mode Processes Based on PCA, Random Forest, and K-means Clustering

Julieth Mendoza-Díaz, Camilo Cueto-Barboza, Ivan Portnoy*, and Ana C. Torregroza-Espinosa

Universidad de la Costa

Abstract

The multi-mode nature of many processes poses a challenge to fault detection and diagnosis tasks, as their non-monotonous behavior renders the traditional monitoring approaches incapable of capturing such varying patterns. This study introduces the F-DINAMITE algorithm, a novel fault detection and diagnosis method incorporating principal component analysis, random forest, and k-means clustering, aiming to identify normal and faulty operational conditions while reducing false alarms and detection errors and providing fault diagnosis capabilities. The novel features of F-DINAMITE include the tailored selection of the number of clusters to minimize the false alarm rate and the automatic update of the fault database when new faults are identified. We validated the proposed method with a case study of a complex building HVAC system, where the algorithm analyzed faults in air dampers and cooling valve systems, testing its detection and diagnosis performance. Moreover, we assessed its false alarm rate while monitoring normal operation data. F-DINAMITE achieves a reduction in false alarm and misdetection rates, compared to those of conventional PCA, while successfully diagnosing faults. This methodology could significantly contribute to maintenance in complex processes by expediting fault detection and reducing associated costs. Nevertheless, future research must focus on conducting a more comprehensive assessment of F-DINAMITE’s performance with more real-life case studies or high amounts of synthetic data.

Keywords: Diagnosis, fault detection, multi-mode processes, random forest.

IJCAS
November 2024

Vol. 22, No. 11, pp. 3253~3544

Stats or Metrics

Share this article on

  • line

Related articles in IJCAS

IJCAS

eISSN 2005-4092
pISSN 1598-6446