An optimized variational mode decomposition method and its application in vibration signal analysis of bearings. (September 2022)

• Record Type: Journal Article
• Authors: Gu, Jun; Peng, Yuxing; Lu, Hao; Chang, Xiangdong; Cao, Shuang; Chen, Guoan; Cao, Bobo
• Is Part Of: Structural health monitoring. Volume 21:Number 5(2022)
• Page Start: 2386
• Page End: 2407
• Publication Date: 2022-09
• Subjects: Rolling bearings -- variational mode decomposition -- genetic algorithm -- permutation entropy -- support vector machine; Structural health monitoring -- Periodicals; Structural stability -- Periodicals; Strength of materials -- Periodicals; Nondestructive testing -- Periodicals; Constructions -- Stabilité -- Périodiques; Résistance des matériaux -- Périodiques; Contrôle non destructif -- Périodiques; Electronic journals; 624.17
• DOI: 10.1177/14759217211057444
• Languages: English
• ISSNs: 1475-9217

Abstract

The performance of the rolling bearing of a spindle device is directly related to the safety and reliability of the operation of a mine hoist. To extract bearing vibration signal features effectively for fault diagnosis, a feature extraction method based on the parameter optimization of a variational mode decomposition (VMD) method and permutation entropy (PE) is proposed. In addition, a support vector machine (SVM) classifier is used to identify bearing fault types. An analogue signal is used to test the effect of noise and sampling frequency on VMD performance. Focused on the problem of the VMD method needing to determine the number of mode components K and a penalty factor α during the signal decomposition process, a genetic algorithm is used to optimize the parameter combination [K, α] with the minimum sample entropy as the indicator. By using mean squared error (MSE) and correlation coefficient, an evaluation indicator is constructed to determine the decomposition effects of the optimized VMD, centre frequency, empirical mode decomposition (EMD) and ensemble EMD (EEMD) methods. The normalized PE of the five mode components is used as an eigenvalue, which is used as the input parameter of the SVM. Two different experimental datasets are used to verify the effectiveness of the proposed method. The results show that the proposed method has better diagnostic accuracy than EMD, EEMD and a BP neural network in the case of limited samples and unknown sample inputs. It can provide a good reference for the diagnosis of a rolling bearing and has practical application value.