Sequential fuzzy diagnosis method for motor roller bearing in variable operating conditions based on vibration...

Area of Science:

• Mechanical Engineering
• Signal Processing
• Machine Learning

Background:

• Motor roller bearings are critical components in machinery.
• Operating under unsteady speed and load conditions complicates fault diagnosis.
• Traditional methods struggle with non-stationary vibration signals.

Purpose of the Study:

• To propose a novel intelligent fault diagnosis method for motor roller bearings operating under unsteady conditions.
• To develop a feature extraction technique that is independent of rotation speed and load.
• To enhance the accuracy and sensitivity of bearing condition diagnosis.

Main Methods:

• Utilized pseudo Wigner-Ville distribution (PWVD) for time-frequency analysis.
• Employed relative crossing information (RCI) for automatic feature spectrum extraction.
• Applied ant colony optimization (ACO) clustering to obtain synthesizing symptom parameters (SSP).
• Developed a fuzzy diagnosis method based on sequential inference and possibility theory.

Main Results:

• The RCI method successfully extracted instantaneous feature spectra, independent of speed and load variations.
• Synthesizing symptom parameters (SSP) demonstrated higher diagnostic sensitivity compared to original symptom parameters (SP).
• The proposed method accurately identified bearing conditions, reflecting feature spectrum characteristics effectively.

Conclusions:

• The novel intelligent fault diagnosis method provides a robust solution for motor roller bearings under non-stationary conditions.
• The SSP and fuzzy diagnosis approach offer improved precision and sensitivity for machine condition monitoring.
• This method enhances the reliability and predictive maintenance capabilities for industrial machinery.