Bearing Fault Diagnosis via Stepwise Sparse Regularization with an Adaptive Sparse Dictionary.

Area of Science:

• Mechanical Engineering
• Signal Processing
• Condition Monitoring

Background:

• Vibration monitoring is crucial for bearing fault diagnosis.
• Sparsity constraint-based regularization effectively extracts transients from noisy signals.
• Conventional methods face a tradeoff between sparsity and data fidelity.

Purpose of the Study:

• To address the limitations of conventional sparse regularization methods.
• To propose a novel stepwise sparse regularization (SSR) method.
• To improve the accuracy of bearing fault diagnosis.

Main Methods:

• Modeled bearing fault diagnosis as a multi-parameter optimization problem.
• Introduced a stepwise sparse regularization (SSR) method with an adaptive sparse dictionary.
• Implemented sparsity-enhanced and fidelity-enhanced optimization steps.

Main Results:

• The SSR method adaptively determines time indexes and the number of atoms.
• Achieved high-precision reconstruction amplitudes by removing the regularization term.
• Demonstrated superior reconstruction accuracy compared to other sparse regularization methods under various noise conditions.

Conclusions:

• The proposed SSR method overcomes the tradeoff between sparsity and data fidelity.
• SSR provides more accurate results for bearing fault diagnosis.
• This method offers enhanced capability for extracting repetitive transients from noisy vibration signals.