Stationary Wavelet-Fourier Entropy and Kernel Extreme Learning for Bearing Multi-Fault Diagnosis.

Area of Science:

• Mechanical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Rotating machine health monitoring relies heavily on bearing fault diagnosis.
• Intelligent methods often combine feature extraction with machine learning for improved accuracy.
• Shannon entropy features have shown promise in enhancing bearing fault diagnosis.

Purpose of the Study:

• To introduce a novel Shannon entropy feature for bearing fault diagnosis.
• To evaluate the effectiveness of the proposed feature combined with a shallow learning classifier.
• To compare the new method against existing techniques in terms of accuracy and efficiency.

Main Methods:

• Developed a new feature: stationary wavelet packet Fourier entropy (SWPFE).
• Combined multi-scale stationary wavelet packet analysis with Fourier amplitude spectrum.
• Utilized a shallow kernel extreme learning machine (KELM) classifier for fault diagnosis.
• Validated the method on experimental vibration signal databases of rolling element bearings.

Main Results:

• The SWPFE method achieved higher accuracy than SWPPE and SWPDE.
• The proposed method required fewer features for effective diagnosis.
• The SWPFE method demonstrated reduced dependence on user expertise due to no hyperparameter calibration.

Conclusions:

• The SWPFE feature offers a more accurate and efficient approach to bearing fault diagnosis.
• This method enhances the reliability of rotating machinery health monitoring.
• The technique presents a valuable advancement in intelligent fault diagnosis systems.