Bearing fault diagnosis method based on enhanced VMD and adaptive-optimized SDAE

Area of Science:

• Mechanical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Motor rolling bearings are critical in industrial production.
• Vibration signal extraction and fault diagnosis are challenging due to operating conditions and noise.
• Existing methods often struggle with complex noise environments.

Purpose of the Study:

• To propose an adaptive variational mode decomposition (VMD) approach optimized by dung beetle algorithm for signal decomposition and feature extraction.
• To develop a robust fault diagnosis model for motor rolling bearings.
• To enhance the accuracy and noise resistance of bearing fault diagnosis.

Main Methods:

• Adaptive VMD based on dung beetle optimization algorithm for signal decomposition.
• Composite optimization indicator function (Tanimoto coefficient, permutation entropy, kurtosis) as fitness function.
• Composite multiscale permutation entropy for feature extraction.
• Enhanced Sine Cosine Algorithm (SCA) with Cauchy chaotic mutation for optimizing stacked denoising auto-encoders (SDAE) hyperparameters.
• Utilized CWRU open bearing dataset for validation.

Main Results:

• The proposed method effectively decomposes and extracts vibration signals even in strong noise.
• Feature vectors were successfully created using composite multiscale permutation entropy.
• The optimized SDAE model achieved high prediction accuracy for fault diagnosis.
• Demonstrated superior adaptability and noise resistance compared to other existing methods.

Conclusions:

• The developed fault diagnosis method shows significant promise for industrial applications.
• The integration of dung beetle optimization with VMD offers a flexible and robust signal processing technique.
• The approach provides a reliable solution for bearing fault diagnosis under challenging noisy conditions.