Diagnosis of Compound Fault Using Sparsity Promoted-Based Sparse Component Analysis.

Area of Science:

• Mechanical Engineering
• Signal Processing
• Fault Diagnosis

Background:

• Compound faults in rotating machinery complicate fault diagnosis.
• Traditional blind source separation methods like Independent Component Analysis (ICA) and Sparse Component Analysis (SCA) struggle with underdetermined conditions and inadequately sparse vibration signals.

Purpose of the Study:

• To develop a sparsity-promoted approach for effective separation and diagnosis of compound faults in rotating machinery, especially under underdetermined conditions.
• To address the challenge of inadequately sparse vibration signals in fault diagnosis.

Main Methods:

• A sparsity-promoted approach using wavelet modulus maxima to obtain sparse observation signals.
• Utilizing a potential function to estimate the number of source signals and the mixing matrix.
• Employing the shortest path method for source signal separation.

Main Results:

• The proposed method successfully separated simulated and actual vibration signals from faulty roller bearings (outer-race, inner-race, rolling element flaws).
• Acquired fault features closely matched theoretical values, e.g., inner-race feature frequency of 101.3 Hz vs. theoretical 101 Hz.
• Demonstrated superior performance compared to traditional SCA for inadequately sparse vibration signals.

Conclusions:

• The proposed method is effective for separating compound faults in rotating machinery, even in underdetermined scenarios.
• This approach significantly improves fault diagnosis accuracy by enhancing signal sparsity.
• It offers a more robust alternative to traditional SCA when dealing with challenging vibration data.