Time-Frequency Multi-Domain 1D Convolutional Neural Network with Channel-Spatial Attention for Noise-Robust Bearing...

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Bearing faults are critical in industrial machinery.
• Existing 1D CNN models struggle with noise in vibration signals.
• Accurate fault diagnosis is essential for predictive maintenance.

Purpose of the Study:

• To develop a noise-robust bearing fault diagnosis model.
• To improve accuracy in high-noise industrial environments.
• To enhance feature extraction from vibration signals.

Main Methods:

• Proposed a time-frequency multi-domain 1D CNN (TF-MDA) model with attention.
• Utilized parallel CNN modules for simultaneous time and frequency domain feature extraction.
• Incorporated physics-informed preprocessing and channel/spatial attention modules.

Main Results:

• The TF-MDA model demonstrated superior performance compared to existing models.
• Achieved high accuracy across a range of signal-to-noise ratios (-6 to 6 dB).
• The attention mechanism effectively enhanced noise-robustness.

Conclusions:

• The TF-MDA model offers a robust and accurate solution for bearing fault diagnosis.
• The multi-domain approach and attention mechanism are key to its effectiveness.
• The model is suitable for industrial applications with high noise levels.