Periodical sparse low-rank matrix estimation algorithm for fault detection of rolling bearings.

Area of Science:

• Mechanical Engineering
• Signal Processing
• Condition Monitoring

Background:

• Early bearing fault detection is critical for preventing catastrophic failures.
• Repetitive defect impulses indicative of bearing faults are often obscured by significant background noise in vibration signals.

Purpose of the Study:

• To propose a novel algorithm for extracting weak, repetitive transient signals from noisy environments.
• To enhance the early detection of bearing faults through improved signal processing techniques.

Main Methods:

• Developed a periodical sparse low-rank (PSLR) matrix estimation algorithm.
• Revealed periodical group sparsity and low-rank properties of fault transients in the time-frequency domain.
• Incorporated non-convex penalty functions and an iterative ADMM-MM algorithm with a Gini-guided fault information thresholding (FIT) scheme.

Main Results:

• The PSLR algorithm effectively extracts defect impulses from noisy vibration signals.
• Simulated and real-world signal analyses validated the algorithm's performance.
• The proposed FIT scheme enhanced the extraction of fault transients compared to traditional methods.

Conclusions:

• The novel PSLR algorithm demonstrates superior performance in extracting bearing fault transients from noisy signals.
• This method offers a promising approach for reliable early bearing fault detection in condition monitoring.