Domain adaptive fault diagnosis algorithm based on multi-graph convolution for rotating machinery.

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Industrial bearings face variable operating conditions, challenging existing fault diagnosis methods.
• Current deep learning approaches often fail to capture complex structural relationships in fault data, limiting generalization.
• Lack of domain adaptation hinders model performance across different working environments.

Purpose of the Study:

• To propose a novel domain adaptation multi-graph convolutional network (DAM-GCN) for robust bearing fault diagnosis.
• To enhance the generalization ability of deep learning models for bearing fault detection under varying conditions.
• To improve the accuracy and reliability of bearing fault diagnosis in industrial settings.

Main Methods:

• Feature extraction using Convolutional Neural Networks (CNN).
• Multi-graph construction (Top-k, k-NN, Radius graphs) to capture local, similarity, and density data characteristics.
• Contrastive learning to improve feature distinguishability and expressiveness.
• Joint optimization of classification and domain alignment losses for domain adaptation.

Main Results:

• The DAM-GCN effectively captures multi-perspective fault structure characteristics.
• Contrastive learning enhanced feature distinguishability for better comparison.
• Domain alignment minimized distribution and graph structure differences between source and target domains.
• Experimental results demonstrate superior performance compared to existing state-of-the-art methods.

Conclusions:

• The proposed DAM-GCN method significantly enhances bearing fault diagnosis under variable operating conditions.
• The multi-graph approach combined with contrastive learning and domain adaptation offers improved generalization.
• This method provides a more robust and effective solution for industrial bearing fault diagnosis.