A Sparsity-Promoted Method Based on Majorization-Minimization for Weak Fault Feature Enhancement.

Area of Science:

• Mechanical Engineering
• Signal Processing
• Condition Monitoring

Background:

• Rotating machinery faults generate weak transient impulses often obscured by noise.
• Early-stage fault detection is challenging due to weak fault signatures and significant interference.
• Traditional sparse representation methods like the Majorization-Minimization (MM) algorithm face issues with sparse basis selection and computational complexity.

Purpose of the Study:

• To develop a robust method for enhancing weak fault features in rotating machinery.
• To improve the efficiency and accuracy of fault diagnosis in the presence of strong background noise.
• To overcome the limitations of traditional MM algorithms in fault diagnosis.

Main Methods:

• A modified Majorization-Minimization (MM) algorithm was proposed, integrating an impulsive feature-preserving factor and a penalty function factor.
• A modified Majorization iterative method was applied to solve the convex optimization problem, yielding sparse coefficients.
• Envelope analysis was performed on the sparse coefficients for weak fault feature extraction, omitting the reconstruction step.

Main Results:

• The proposed method effectively enhances weak fault features and extracts them from strong background noise.
• The modified MM algorithm demonstrated superior detection results and efficiency compared to the traditional MM algorithm.
• Simulated and experimental signals from bearings and gearboxes validated the method's effectiveness.

Conclusions:

• The developed modified MM algorithm offers an efficient and accurate approach for weak fault feature extraction in rotating machinery.
• The method's ability to preserve impulsive features and reduce noise makes it suitable for early fault diagnosis.
• Eliminating the need for sparse basis selection and the reconstruction step significantly increases detection efficiency.