Fault state recognition of wind turbine gearbox based on generalized multi-scale dynamic time warping. (November 2021)

• Record Type: Journal Article
• Authors: Pang, Bin; Tian, Tian; Tang, Gui-Ji
• Is Part Of: Structural health monitoring. Volume 20:Number 6(2021)
• Page Start: 3007
• Page End: 3023
• Publication Date: 2021-11
• Subjects: Generalized multi-scale dynamic time warping -- Laplace Score -- random forest -- wind turbine gearbox -- fault diagnosis; 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: 10.1177/1475921720978622
• Languages: English
• ISSNs: 1475-9217

Abstract

Fault diagnosis of wind turbine gearbox is significant to ensure the operating efficiency and reduce the maintenance cost of wind farms. The key to achieve an accurate fault diagnosis is to extract the evidence of fault state identification effectively. Dynamic time warping has been widely used as a classifier for automatic pattern recognition as the dynamic time warping distance can indicate the similarity between two data sequences. The similarity between the analyzed signal and the template signal can be an evidence for characterizing the fault types of the analyzed signal; a generalized multi-scale dynamic time warping algorithm was accordingly developed in this article to quantitatively evaluate the condition information of wind turbine gearbox by calculating the generalized multi-scale dynamic time warping distances between the template signal (i.e. the vibration signal of wind turbine gearbox in normal state) and the testing signal (i.e. the vibration signals of wind turbine gearbox to be analyzed). Then, the sensitive features of the condition information evaluation results obtained via the generalized multi-scale dynamic time warping algorithm were selected by the Laplace Score approach to construct the eigenvector. Finally, random forest was introduced to realize the intelligent fault recognition of wind turbine gearbox. The analysis results of both experimental and engineering signals indicate that the presented method can accurately identify different fault states of wind turbine gearbox. In addition, the proposed method performs a higher accuracy of fault state classification compared with some existing methods.