| Autor: |
Zhijian Wang, Huihui He, Junyuan Wang, Wenhua Du |
| Jazyk: |
angličtina |
| Rok vydání: |
2019 |
| Předmět: |
|
| Zdroj: |
IEEE Access, Vol 7, Pp 154986-155001 (2019) |
| Druh dokumentu: |
article |
| ISSN: |
2169-3536 |
| DOI: |
10.1109/ACCESS.2019.2945409 |
| Popis: |
The Singular spectrum decomposition (SSD) method has been widely used in gearbox fault diagnosis. However, there are two defects in the SSD method. SSD method uses the normalized mean square error as the stopping criterion for decomposition, and there is an over-decomposition phenomenon; the noise has a great influence on SSD and extracting fault features is difficult in a strong background noise environment. In view of the deficiencies of the SSD algorithm, the paper proposes a novel stopping criterion to make the SSD method adaptively stop. Firstly, the SSD method is improved by using the cumulative percent variance contribution rate of the principal component analysis method (PCA) as the decomposition stop criterion. Secondly, calculate the correlation coefficient between the decomposed singular spectral components (SSC) and the raw signal. Eliminating weakly correlation signal. Thirdly, due to the component signal contains noise, and employ a single filter has limitations. So, the paper uses the Auto Regressive (AR) filter filters the decomposed high-frequency component signal and the Savitzky Golay (SG) filter filters the decomposed low-frequency component signal. Finally, applies mutual information entropy (MIE) to distinguish the SSCs components distinguish two parts: high-frequency part and low-frequency part. FFT transforms and extracts fault features. The simulation signal and the composite fault signal extraction of the wind turbine gearbox test bench shows the effectiveness and superiority of the method. |
| Databáze: |
Directory of Open Access Journals |
| Externí odkaz: |
|
