GMRVGG: A Bearing Fault Diagnosis Method Based on Tri-Modal Image Feature Fusion.

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Bearing fault diagnosis is crucial for rotating machinery.
• Existing methods often suffer from limited data dimensionality and feature representation.
• This leads to suboptimal diagnostic accuracy in mechanical health monitoring.

Purpose of the Study:

• To propose an advanced bearing fault diagnosis method.
• To overcome limitations of single-modal data and single signal-to-image conversion.
• To enhance diagnostic accuracy through multi-modal feature fusion.

Main Methods:

• Utilized Gramian Angular Difference Field (GADF), Markov Transition Field (MTF), and Recurrence Plot (RP) for 1D vibration signal to 2D image conversion.
• Employed VGG16 backbone network for extracting and fusing shallow-to-deep features.
• Integrated a fully connected classifier for the final fault diagnosis.

Main Results:

• Achieved superior overall accuracies of 96.9% on CWRU and 95.8% on Ottawa datasets.
• Demonstrated robustness under severe Gaussian white noise and pink noise conditions.
• Significantly outperformed conventional single-modal approaches in diagnostic accuracy.

Conclusions:

• The proposed tri-modal image feature fusion method effectively addresses feature dimensionality and representation limitations.
• This establishes a highly reliable and robust solution for intelligent bearing fault diagnosis.
• The GADF-MTF-RP-VGG16 (GMRVGG) method offers a significant advancement in mechanical health monitoring.