A Self-Attention Legendre Graph Convolution Network for Rotating Machinery Fault Diagnosis

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Rotating machinery health is critical for industrial operations.
• Traditional deep learning methods miss relational information in fault diagnosis.
• Accurate fault diagnosis ensures safety and efficiency.

Purpose of the Study:

• To develop an advanced deep learning model for rotating machinery fault diagnosis.
• To overcome limitations of existing methods in feature extraction.
• To improve the stability and efficiency of fault detection.

Main Methods:

• Proposed a Legendre graph convolutional network (LGCN) integrated with self-attention graph pooling (SA-LGCN).
• Transformed vibration signals from Euclidean to non-Euclidean space (graph signals).
• Utilized a fast local spectral filter based on Legendre polynomials.

Main Results:

• The SA-LGCN model demonstrated significant advantages in fault diagnosis accuracy.
• The method showed improved load adaptability under various working conditions.
• Achieved superior performance in 10 different planetary gearbox fault tasks.

Conclusions:

• The SA-LGCN model offers a robust solution for rotating machinery fault diagnosis.
• The approach effectively captures relational information in vibration signals.
• This method enhances diagnostic capabilities for industrial equipment.