A rolling bearing fault diagnosis method based on an improved parallel one-dimensional convolutional neural network

Area of Science:

• Engineering
• Artificial Intelligence
• Machine Learning

Background:

• Rolling bearing fault diagnosis is crucial for industrial equipment reliability and preventing downtime.
• Existing methods struggle with accuracy in low signal-to-noise ratio environments, often below 92%.

Purpose of the Study:

• To develop an improved deep learning model for accurate rolling bearing fault diagnosis.
• To overcome the limitations of existing methods in noisy conditions.

Main Methods:

• Proposed an improved parallel one-dimensional convolutional neural network (CNN).
• Integrated a parallel dual-channel convolutional kernel, gated recurrent unit (GRU), and attention mechanism.
• Utilized global max-pooling and Softmax for classification.

Main Results:

• Achieved a superior fault diagnosis accuracy of 99.62%.
• Demonstrated significant performance improvement over traditional CNNs and benchmark methods.
• The model effectively captures global and local features while mitigating overfitting and parameter redundancy.

Conclusions:

• The proposed model offers a highly effective solution for rolling bearing fault diagnosis.
• The integration of dual-channel CNN, GRU, and attention mechanisms enhances diagnostic accuracy and reliability.
• This approach significantly advances the state-of-the-art in condition monitoring for industrial equipment.