Self-Supervised Joint Learning Fault Diagnosis Method Based on Three-Channel Vibration Images.

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Accurate bearing fault diagnosis is crucial for rotating machinery reliability.
• Deep learning methods for intelligent fault diagnosis often require extensive labeled data, posing industrial challenges.
• Reducing reliance on labeled data is essential for practical fault diagnosis applications.

Purpose of the Study:

• To propose a novel self-supervised joint learning (SSJL) fault diagnosis method.
• To leverage unlabeled data for learning robust fault features.
• To improve diagnostic accuracy, especially with limited labeled data.

Main Methods:

• Developed a self-supervised joint learning (SSJL) approach for fault diagnosis.
• Utilized three-channel vibration images to enhance feature representation.
• Combined self-supervised learning with supervised learning to maximize data utilization.

Main Results:

• The proposed SSJL method demonstrated higher diagnostic accuracy with small amounts of labeled data.
• The method effectively learned fault features by transforming data into three-channel vibration images.
• Experimental validation on motor bearing datasets confirmed the superiority over existing methods.

Conclusions:

• The SSJL method effectively reduces the need for large labeled datasets in bearing fault diagnosis.
• Transforming data into three-channel vibration images improves feature recognition and diagnostic performance.
• This approach offers a promising solution for intelligent fault diagnosis in industrial settings.