Fault detection of gearbox by multivariate extended variational mode decomposition-based time-frequency images and...

Area of Science:

• Mechanical Engineering
• Signal Processing
• Machine Learning

Background:

• Gearbox fault detection is crucial for industrial machinery maintenance.
• Traditional methods struggle with non-stationary, low signal-to-noise ratio data.
• Existing signal processing techniques lack robustness for multi-channel gearbox signals.

Purpose of the Study:

• To introduce a novel fault detection method for gearboxes.
• To enhance the accuracy and robustness of gearbox fault diagnosis.
• To compare the proposed method against established algorithms.

Main Methods:

• Utilizing multivariate extended variational mode decomposition (MEVMDTFI) for time-frequency image construction.
• Employing an incremental relevance vector machine (IRVM) algorithm for fault classification.
• Developing a combined MEVMDTFI-IRVM approach for gearbox fault detection.

Main Results:

• The MEVMDTFI-IRVM method demonstrated stable fault detection results for gearboxes.
• The proposed method significantly outperformed VMDTFI-IRVM, VMD-RVM, and traditional RVM algorithms.
• MEVMDTFI showed superior robustness for non-stationary, multi-channel signals with low SNR.

Conclusions:

• The MEVMDTFI-IRVM method offers a robust and effective solution for gearbox fault detection.
• This novel approach enhances diagnostic accuracy, particularly in challenging signal conditions.
• The study validates the superiority of MEVMDTFI-IRVM over existing fault detection techniques.