Abstract
Although vibration-based fault diagnosis is widely used for wind turbines, extracting reliable features from noisy and weak impulsive signals remains a major challenge. Time-domain entropy methods are useful for vibration analysis; however, they become less reliable when weak impulsive fault characteristics are masked by noise. This study addresses this limitation by proposing spectral symbolic fuzzy entropy (SSFE), a weak-fault feature-extraction method for vibration-based fault diagnosis. SSFE first converts the vibration signal from the time domain to the spectral domain, where spectral analysis reveals fault-related frequency components more clearly. It then applies symbolic dynamic filtering to the spectral representation to suppress low-intensity spectral fluctuations while important signal characteristics are retained. Finally, fuzzy entropy is calculated from these symbolic spectral patterns to quantify their irregularity and complexity. The proposed method is evaluated against three entropy-based benchmark approaches, including fuzzy entropy (FE), spectral fuzzy entropy (SpecFE) and symbolic fuzzy entropy (SFE). The experimental results show that SSFE provides the strongest performance among the compared methods for extracting weak fault features.
| Original language | English |
|---|---|
| Journal | Journal of Physics: Conference Series |
| Volume | 3325 |
| Issue number | 1 |
| DOIs | |
| State | Published - 2026 |
| Event | 7th International Conference on Mechanical Instrumentation and Automation, ICMIA 2026 - Chengdu, China Duration: 26 Jun 2026 → 28 Jun 2026 |
Fingerprint
Dive into the research topics of 'Weak fault feature extraction from vibration signals for wind turbine rotary components using spectral symbolic fuzzy entropy'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver