Research on singular spectrum decomposition and its application to rotor failure detection

• Main Authors: Bin Pang, Guiji Tang, Yuling He
• Format: Article
• Language: English
• Published: JVE International 2018-09-01
• Series: Journal of Vibroengineering
• Subjects: singular spectrum decomposition; energy separation algorithm; variable speed condition; rotor; failure detection
• Online Access: https://www.jvejournals.com/article/19200

As an important part of rotating machinery, a healthy rotor is critical to ensuring optimal working conditions of the entire system. Considering that the vibration signal of rotor consists of different frequency components when the failure arises, a novel rotor failure detection method based on singular spectrum decomposition (SSD) is presented. The original vibration signal is adaptively decomposed into a number of singular spectrum components (SSCs) by the SSD method. Then, energy separation algorithm (ESA) is adopted to demodulate each singular spectrum component. Finally, the SSD-ESA time-frequency spectrum can be obtained and the fault features contained in the SSD-ESA time-frequency spectrum can be identified to determine the fault types. The effectiveness of SSD for harmonic separation was assessed through tones separation analyses, the results show that SSD is able to separate more harmonic pairs of different amplitude ratios than empirical mode decomposition (EMD). Furthermore, three simulations of multi-component signals were designed to investigate the use of SSD for signal decomposition. The SSD method was then applied to detect signatures caused by rotor oil film whirl in experimental signals and compared to both EMD and ensemble EMD (EEMD). The simulated analysis results reflect that SSD shows superiority to EMD and EEMD in inhibiting mode mixing and extracting the time-varying frequency components. The experimental analysis results demonstrate that the SSD based rotor failure detection method is an alternative method under both constant and variable speed conditions.