Rolling Bearing Fault Detection System and Experiment Based on Deep Learning.

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Frequent small-scale accidents highlight the inadequacy of existing methods for preventing bearing failures.
• Timely detection of faults in rolling bearings is crucial for preventing catastrophic failures and ensuring operational safety.

Purpose of the Study:

• To develop and validate a novel deep learning model for accurate rolling bearing fault diagnosis.
• To enhance the identification and analysis of multi-state vibration signals under diverse working conditions.

Main Methods:

• Utilized a combination of deep transfer learning and metric learning techniques.
• Employed SSAE-based (Stacked Sparse Autoencoder) similarity measurement criteria integrated with deep transfer learning.
• Analyzed bearing multi-state vibration signals under different working conditions.

Main Results:

• The proposed LCM-SSAE (Layer-wise Convolutional Metric Learning-Stacked Sparse Autoencoder) method achieved a 0.6 percentage point higher detection accuracy compared to the SSAE-based method.
• Demonstrated the model's effectiveness in reducing inter-domain differences and improving the distinguishability of boundary data samples.
• Validated the proposed deep learning model through bearing fault diagnosis analysis.

Conclusions:

• The developed LCM-SSAE deep learning model is suitable and effective for rolling bearing fault detection.
• The proposed method exhibits robustness and superior performance in identifying bearing faults compared to existing techniques.
• This research contributes a validated deep learning solution for enhancing the reliability and safety of rolling bearing systems.