A Deep Learning-Based Method for Bearing Fault Diagnosis with Few-Shot Learning

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Deep learning for fault diagnosis often struggles with limited sample data.
• Rolling bearing faults are critical in industrial machinery maintenance.

Purpose of the Study:

• To develop an accurate fault diagnosis method for rolling bearings under small sample conditions.
• To leverage deep learning and data augmentation for enhanced diagnostic performance.

Main Methods:

• Vibration signals converted to time-frequency images using continuous wavelet transform.
• Feature extraction using a Convolutional Neural Network (CNN) integrated with Kolmogorov-Arnold Networks (KANs).
• Data augmentation via diffusion networks to address limited sample sizes.
• Feature aggregation using a Feature Attention Network (FAN) module.

Main Results:

• The proposed KANs-CNN with diffusion models achieves superior accuracy in small sample fault diagnosis.
• The integration of KANs effectively extracts complex, nonlinear features.
• The FAN module enhances feature representation by aggregating multi-level information.

Conclusions:

• The KANs-CNN and diffusion network approach offers a robust solution for rolling bearing fault diagnosis with limited data.
• This method demonstrates significant potential for improving industrial equipment reliability and maintenance strategies.