Blind Deconvolution Based on Correlation Spectral Negentropy for Bearing Fault

Area of Science:

• Mechanical Engineering
• Signal Processing
• Condition Monitoring

Background:

• Blind deconvolution methods like MED and OMEDA are sensitive to extreme values.
• MCKD and MOMEDA require prior fault knowledge, limiting their practical application.
• Effective fault diagnosis in rolling bearings is crucial for machinery health.

Purpose of the Study:

• To develop a novel blind deconvolution algorithm for improved rolling bearing fault diagnosis.
• To address the limitations of existing deconvolution methods, particularly the need for prior fault information.
• To enhance the detection of fault characteristics in rolling bearing vibration signals.

Main Methods:

• Proposed a new deconvolution algorithm based on Maximum Correlation Spectral Negentropy (CSNE).
• Utilized the Particle Swarm Optimization (PSO) algorithm to determine filter coefficients.
• Leveraged the periodicity and impact characteristics of bearing fault signals.

Main Results:

• The proposed PSO-CSNE algorithm effectively overcomes harmonic and random pulse signal interference.
• Demonstrated superior performance compared to existing blind deconvolution algorithms in simulations and experiments.
• Successfully identified fault characteristics without requiring prior fault knowledge.

Conclusions:

• The PSO-CSNE algorithm offers a robust and effective solution for blind deconvolution in rolling bearing fault diagnosis.
• This method provides a significant advancement over traditional techniques by eliminating the need for prior fault information.
• The algorithm shows strong potential for practical applications in condition monitoring and fault detection.