Particle swarm optimization algorithm to solve the deconvolution problem for rolling element bearing fault diagnosis.

Area of Science:

• Mechanical Engineering
• Signal Processing
• Machine Condition Monitoring

Background:

• Rolling element bearing fault diagnosis relies on extracting fault impulses from vibration signals.
• Traditional deconvolution techniques like MED, MEDA, MCKD, OMEDA, and MOMEDA are used but have limitations.

Purpose of the Study:

• To introduce particle swarm optimization (PSO) for solving deconvolution filter coefficients in bearing fault diagnosis.
• To enhance the extraction of fault-related impulses from vibration signals.

Main Methods:

• Utilizing particle swarm optimization (PSO) with a generalized spherical coordinate transformation to determine deconvolution filter coefficients.
• Developing PSO-based versions of Minimum Entropy Deconvolution (PSO-MED) and Optimal MED Adjusted (PSO-OMEDA).
• Developing PSO-based versions of Maximum Correlated Kurtosis Deconvolution (PSO-MCKD) and Multipoint Optimal MED Adjusted (PSO-MOMEDA).

Main Results:

• PSO-MED and PSO-OMEDA effectively handle large random impulses and deconvolve periodic impulses.
• PSO-MCKD and PSO-MOMEDA perform well even with inaccurate fault periods.
• Validated effectiveness on simulated and experimental bearing fault signals, showing superior performance over traditional methods and EEMD/fast kurtogram.

Conclusions:

• Particle swarm optimization offers a robust approach to deconvolution for rolling element bearing fault diagnosis.
• The proposed PSO-based deconvolution methods significantly improve fault detection accuracy and reliability.
• PSO-based deconvolution presents a promising advancement in condition monitoring of rotating machinery.