Lightweight Ghost Enhanced Feature Attention Network: An Efficient Intelligent Fault Diagnosis Method under Various...

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Deep neural networks show promise for bearing fault diagnosis but are computationally intensive.
• Complexity of current models hinders practical application and automated diagnostic tools.

Purpose of the Study:

• To develop a computationally efficient and accurate fault diagnosis framework for rolling bearings.
• To address the limitations of existing deep learning approaches in terms of computational resources and diagnostic speed.

Main Methods:

• Introduced a tri-channel preprocessing module (FFT, FWEO, Signal Envelope Analysis) for feature extraction.
• Developed the Lightweight Ghost Enhanced Feature Attention Network (GEFA-Net) using Ghost Module and Efficient Pyramid Squared Attention (EPSA).
• Reduced model complexity and enhanced feature representation through linear operations and attention mechanisms.

Main Results:

• GEFA-Net achieved superior diagnostic accuracy (98.53% and 99.98%) on experimental datasets.
• Significantly reduced model parameter count by 63.74% compared to MobileVit.
• Demonstrated accelerated detection processes and reduced computational complexity.

Conclusions:

• The proposed framework offers a robust and efficient solution for bearing fault diagnosis under variable operating conditions.
• GEFA-Net shows significant potential for practical industrial applications requiring automated diagnostics.
• The methodology effectively balances diagnostic accuracy with computational efficiency.