Early Fault Detection via Multiple Feature Fusion and Ensemble Learning

Early fault detection (EFD) of rotating machines is important to decrease the maintenance cost and improve the mechanical system stability. One of the key points of EFD is developing a generic and robust model for different equipment. Most existing EFD methods focus on learning fault representation...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE sensors journal Jg. 24; H. 5; S. 7196 - 7204
Hauptverfasser: Song, Wenbin, Wu, Di, Shen, Weiming, Boulet, Benoit
Format: Journal Article
Sprache:Englisch
Veröffentlicht: New York IEEE 01.03.2024
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Schlagworte:
Datasets
Early fault detection (EFD)
Ensemble learning
Fault detection
Feature extraction
Frequency domain analysis
Maintenance costs
Mechanical systems
Noise reduction
one-class support vector machine (SVM)
Representations
Rotating machinery
Rotating machines
stacked denoising autoencoder (SDAE)
Stacking
Support vector machines
Systems stability
ISSN:1530-437X, 1558-1748
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Early fault detection (EFD) of rotating machines is important to decrease the maintenance cost and improve the mechanical system stability. One of the key points of EFD is developing a generic and robust model for different equipment. Most existing EFD methods focus on learning fault representation by one type of feature. However, a combination of multiple features can capture a more comprehensive representation. In this article, we propose an EFD method based on multiple feature fusion with stacking architecture (M2FSA). The proposed method can extract generic and discriminative features to detect early faults by combining time domain (TD), frequency domain (FD), and time-frequency domain (TFD) features. To unify the dimensions of the different domain features, stacked denoising autoencoder (SDAE) is used to learn deep features in three domains. The proposed method is tested on three bearing datasets and a motor dataset. The results demonstrate that the proposed method is better than existing methods in both sensibility and reliability.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1530-437X
1558-1748
DOI:10.1109/JSEN.2024.3353732