Fault Diagnosis for Rotating Machinery Using Vibration Measurement Deep Statistical Feature Learning

Area of Science:

• Engineering
• Computer Science
• Data Science

Background:

• Rotating machinery requires effective fault diagnosis for maintenance.
• Detecting faults and fault patterns in machinery is a significant challenge.

Purpose of the Study:

• To present a deep statistical feature learning model for rotating machinery fault diagnosis.
• To improve the accuracy of fault classification using vibration measurements.

Main Methods:

• Vibration signals from rotating machinery were analyzed in time, frequency, and time-frequency domains.
• A Gaussian-Bernoulli deep Boltzmann machine (GDBM) was developed by stacking real-value Gaussian-Bernoulli restricted Boltzmann machines (GRBMs).
• The GDBM was applied to gearbox and bearing fault diagnosis.

Main Results:

• The proposed deep learning approach achieved 95.17% fault classification accuracy for gearboxes and 91.75% for bearings.
• The model outperformed standard methods like support vector machines and individual GRBMs.
• The best fault classification rate was achieved using the proposed deep statistical feature learning model.

Conclusions:

• Deep learning combined with statistical feature extraction offers significant potential for diagnosing rotating machinery faults.
• The GDBM model provides an effective solution for complex fault diagnosis tasks.
• This approach enhances the reliability and efficiency of machinery maintenance.