Transfer learning-motivated intelligent fault diagnosis framework for cross-domain knowledge distillation

Area of Science:

• Artificial Intelligence
• Machine Learning
• Control Systems

Background:

• Transfer learning has significantly advanced artificial intelligence (AI) applications across various domains.
• Intelligent fault diagnosis (IFD) using transfer learning enhances the reliability of industrial automation.
• Nonlinear systems present unique challenges for accurate fault diagnosis.

Purpose of the Study:

• To propose a novel transfer learning-based methodology for nonlinear system fault diagnosis.
• To enhance temporal sequence analysis capabilities within an IFD framework.
• To improve the stability and reliability of fault diagnosis in complex systems.

Main Methods:

• Integration of cross-domain knowledge distillation into an IFD scheme.
• Embedding twin-spiking neural networks (SNNs) for advanced temporal analysis.
• Transforming and transferring prior knowledge from feature extraction backbones to twin SNNs.

Main Results:

• The proposed methodology effectively reconstructs residual generators for IFD.
• Experimental validation on nonlinear systems in high-energy vehicle lithium batteries confirms effectiveness.
• Demonstrated practical applicability and improved diagnostic accuracy.

Conclusions:

• The developed transfer learning framework offers a robust solution for nonlinear system fault diagnosis.
• The integration of knowledge distillation and SNNs enhances diagnostic performance.
• This approach holds significant potential for improving the safety and efficiency of industrial automation, especially in critical applications like battery management.