Facial reconstruction

Search LJMU Research Online

Browse Repository | Browse E-Theses

Fault diagnosis of rolling bearing of wind turbines based on the Variational Mode Decomposition and Deep Convolutional Neural Networks

Xu, Z, Li, C and Yang, Y (2020) Fault diagnosis of rolling bearing of wind turbines based on the Variational Mode Decomposition and Deep Convolutional Neural Networks. Applied Soft Computing Journal, 95. ISSN 1568-4946

Fault Diagnosis of Rolling Bearing of Wind Turbines Based on the Variational Mode Decomposition.pdf - Accepted Version
Available under License Creative Commons Attribution Non-commercial No Derivatives.

Download (4MB) | Preview


Machine learning techniques have been successfully applied in intelligent fault diagnosis of rolling bearings in recent years. However, in the real world industrial application, the dissimilarity of data due to changes in the working conditions and data acquisition environment often cause a poor performance of the existing fault diagnosis methods. Consequently, to address these inadequacies, this paper developed a novel method by integrating the Convolutional Neural Networks (CNNs) with the Variational Mode Decomposition (VMD) algorithms. Named as “Variational Mode Decomposition with Deep Convolutional Neural Networks (VMD-DCNNs)”, the method, in an end-to-end way, directly processes raw vibration signals without artificial experiences and manual intervention to realize the fault diagnosis of rolling bearings. In addition, the CNN technique is used to extract features from each Intrinsic Mode Function (IMF) in order to address the deficiency in extracting features from a single source and to achieve an effective and efficient fault diagnosis of rolling bearings under different environments and states. The value of parameter K of the VMD-DCNNs model is optimized by considering time complexity and generalization ability of the model. Lastly, bearing experiments are conducted to verify the superiority of the VMD-DCNNs in diagnosing fault under different conditions. The visualizations of the signals in the convolutional layer explain the reasonability in selecting the value of parameter K and they also indicate that the translational invariances in a raw IMF component have been learned by the VMD-DCNNs model.

Item Type: Article
Uncontrolled Keywords: Science & Technology; Technology; Computer Science, Artificial Intelligence; Computer Science, Interdisciplinary Applications; Computer Science; Deep learning; Convolutional Neural Networks; VariationaL Mode Decomposition; Fault diagnosis; Rolling bearing; Artificial Intelligence & Image Processing; 0102 Applied Mathematics; 0801 Artificial Intelligence and Image Processing; 0806 Information Systems
Subjects: T Technology > TA Engineering (General). Civil engineering (General)
T Technology > TC Hydraulic engineering. Ocean engineering
Divisions: Engineering
Publisher: Elsevier
SWORD Depositor: A Symplectic
Date Deposited: 27 Jun 2022 11:00
Last Modified: 27 Jun 2022 11:00
DOI or ID number: 10.1016/j.asoc.2020.106515
URI: https://researchonline.ljmu.ac.uk/id/eprint/17160
View Item View Item