An Intelligent Machinery Fault Diagnosis Method Based on GAN and Transfer Learning under Variable Working Conditions.

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Machine Learning

Background:

• Intelligent fault diagnosis is crucial for mechanical equipment safety.
• Current models require extensive, uniform data, which is often unavailable in real-world scenarios.
• Data scarcity and distribution shifts across working conditions hinder diagnostic model accuracy and generalizability.

Purpose of the Study:

• To develop a robust fault diagnosis method overcoming data limitations and working condition variations.
• To enhance the accuracy and applicability of diagnostic models in diverse operational environments.

Main Methods:

• A novel approach combining Generative Adversarial Networks (GANs) for data augmentation and transfer learning for knowledge migration.
• GANs generate synthetic fault data to expand limited datasets.
• A deep residual network with domain-adapted regularization extracts and transfers fault characteristics across different working conditions.

Main Results:

• The proposed method effectively expands datasets using GANs, addressing data insufficiency.
• The deep transfer fault diagnosis model successfully extracts and migrates fault characteristics.
• Validation using bearing fault data demonstrates significant improvement in diagnostic accuracy under varying conditions.

Conclusions:

• The developed method mitigates the impact of insufficient monitoring data.
• It enables effective migration of fault diagnosis knowledge across different working conditions.
• This approach enhances the practical application of intelligent fault diagnosis for mechanical equipment.