Rotor Fault Diagnosis Based on Characteristic Frequency Band Energy Entropy and Support Vector Machine

Area of Science:

• Mechanical Engineering
• Signal Processing

Background:

• Rotors are critical mechanical components prone to defects.
• Effective rotor fault diagnosis is essential for system reliability.
• Current methods face challenges in accurate fault signature extraction and classification.

Purpose of the Study:

• To develop an accurate rotor fault diagnosis technique.
• To propose a novel evaluation index, characteristic frequency band energy entropy (CFBEE), for feature extraction.
• To employ support vector machine (SVM) for automated fault type identification.

Main Methods:

• Utilized improved singular spectrum decomposition (ISSD) and Hilbert transform (HT) for time-frequency spectrum (TFS) analysis.
• Calculated CFBEE from the ISSD-HT TFS to create fault feature vectors.
• Employed SVM for automatic classification of rotor fault types.

Main Results:

• ISSD demonstrated superior performance over empirical mode decomposition (EMD) in extracting signal sub-components.
• The ISSD-HT TFS provided more accurate time-frequency information than EMD-HT TFS.
• Experimental results confirmed the method's ability to accurately identify rotor defect types.

Conclusions:

• The proposed ISSD-HT TFS method combined with CFBEE and SVM offers an effective approach for rotor fault diagnosis.
• This technique outperforms existing methods in accuracy and feature extraction capabilities.
• The developed method contributes to enhanced mechanical system monitoring and maintenance.