A Compound fault diagnosis for rolling bearings method based on blind source separation and ensemble empirical mode...

Area of Science:

• Mechanical Engineering
• Signal Processing
• Condition Monitoring

Background:

• Compound faults in rolling bearings generate complex signals with significant noise, hindering traditional diagnostic methods like FFT, wavelet transform, or empirical mode decomposition.
• Accurate separation of individual fault signatures is crucial for reliable rolling bearing diagnostics.

Purpose of the Study:

• To develop and validate a new method for effectively separating weak fault signals within mixed signals from rolling bearings.
• To enhance the accuracy and clarity of compound fault diagnosis in rolling bearings.

Main Methods:

• The proposed method utilizes ensemble empirical mode decomposition (EEMD) to decompose mixed vibration signals into multiple intrinsic mode functions (IMFs).
• Independent component analysis (ICA) is applied to the selected IMFs, identified via a cross-correlation criterion, for signal separation.
• The combined EEMD-ICA approach facilitates the isolation of individual fault characteristics from complex signals.

Main Results:

• The EEMD-ICA method successfully separated compound fault signals, making fault features more extractable and identifiable.
• Experimental validation confirmed the method's effectiveness for outer race defects, roller defects, and unbalance faults.
• The technique demonstrated superior performance in separating weak fault signals compared to conventional methods.

Conclusions:

• The proposed EEMD-ICA method offers a robust solution for diagnosing compound faults in rolling bearings by effectively separating mixed signals.
• This approach significantly improves the ability to detect and identify various bearing faults, enhancing machinery health monitoring.