Intelligent gearbox diagnosis methods based on SVM, wavelet lifting and RBR

Area of Science:

• Mechanical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Intelligent gearbox diagnosis faces challenges with data acquisition and sample size limitations.
• Variability in fault types and features in complex field environments complicates diagnosis.

Purpose of the Study:

• To develop an effective gearbox fault diagnosis method addressing data scarcity and fault diversity.
• To integrate wavelet lifting, support vector machine (SVM), and rule-based reasoning (RBR) for enhanced diagnostic accuracy.

Main Methods:

• Gearbox vibration signals were analyzed using wavelet packet decomposition to extract energy coefficients.
• Support vector machine (SVM) was employed for initial fault pattern recognition.
• Wavelet lifting was utilized for noise filtering and fault feature extraction.
• Rule-based reasoning (RBR), based on expert knowledge, was applied for detailed fault type identification.

Main Results:

• Support vector machine (SVM) demonstrated effectiveness in gearbox fault pattern recognition with small sample sizes.
• Wavelet lifting successfully filtered noise while preserving critical fault impulse characteristics.
• Rule-based reasoning (RBR) accurately identified detailed fault types based on expert-derived rules.

Conclusions:

• The combined approach of SVM, wavelet lifting, and RBR provides a robust and effective solution for gearbox fault diagnosis.
• This integrated method overcomes limitations of traditional techniques, offering improved reliability in complex operational settings.