Self-Adaptive Spectrum Analysis Based Bearing Fault Diagnosis

Area of Science:

• Mechanical Engineering
• Signal Processing
• Machine Learning

Background:

• Bearing faults are critical in rotating machinery, posing challenges due to non-linear and unstable signals.
• Conventional fixed-dimension feature extraction methods often lead to information loss or redundancy, impacting diagnostic performance.

Purpose of the Study:

• To propose a novel Self-adaptive Spectrum Analysis (SSA) method and a diagnostic framework for enhanced bearing fault diagnosis.
• To address limitations of traditional signal analysis and feature extraction in non-linear and unstable bearing data.

Main Methods:

• Signals are decomposed into more analyzable components.
• Self-adaptive Spectrum Analysis (SSA) is developed for adaptive fault feature extraction, creating non-fixed dimension feature vectors.
• Support Vector Machine (SVM) is employed for fault feature classification.

Main Results:

• The SSA-based diagnosis framework demonstrates superior performance compared to conventional methods.
• Experimental results confirm higher reliability and robustness of SSA across different working conditions.
• The proposed method achieves enhanced accuracy and stability, even with varying sample division schemes.

Conclusions:

• The SSA method, combined with signal decomposition, offers a more reliable and robust approach to bearing fault diagnosis.
• The SSA-based diagnostic framework provides higher accuracy and stability under diverse operating conditions and data sampling strategies.