Fault diagnosis method for gearbox systems using OVMD and KLIRVM

Area of Science:

• Mechanical Engineering
• Signal Processing
• Machine Learning

Background:

• Gearbox systems are critical in industrial machinery.
• Accurate fault diagnosis is essential for preventing failures and ensuring operational efficiency.
• Existing diagnostic methods may struggle with non-stationary signals from damaged components.

Purpose of the Study:

• To develop an integrated methodology for precise gearbox fault diagnosis.
• To enhance the accuracy of identifying faults in rolling elements.
• To address the challenge of analyzing streaming data with locally varying signal features.

Main Methods:

• Orthogonalized Variational Mode Decomposition (OVMD) using Bessel basis functions to isolate non-stationary fault characteristics.
• Kernel Learning Incremental Relevance Vector Machine (KLIRVM) with adaptive kernel parameter learning and incremental updates.
• Ablation studies to validate the proposed OVMD-KLIRVM framework.

Main Results:

• OVMD effectively isolates transient fluctuations from damaged rolling elements.
• KLIRVM accurately characterizes locally varying signal features in streaming data.
• The proposed OVMD-KLIRVM framework demonstrates superior diagnostic accuracy compared to competing methods.

Conclusions:

• The integrated OVMD-KLIRVM methodology offers a robust solution for gearbox fault diagnosis.
• The adaptive nature of KLIRVM is crucial for handling dynamic signal variations.
• This approach significantly improves diagnostic precision in gearbox systems.