A rolling bearing fault detection method based on compressed sensing and a neural network

Area of Science:

• Mechanical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Traditional Nyquist sampling poses challenges due to high sampling rates, demanding equipment, and large data volumes.
• Increased data complexity complicates information transmission and storage in signal processing applications.
• Efficient fault detection in rolling bearings is crucial for industrial machinery maintenance.

Purpose of the Study:

• To develop an efficient rolling bearing fault signal detection method.
• To reduce data storage and transmission burdens associated with high-frequency sampling.
• To leverage compressed sensing and neural networks for enhanced signal processing.

Main Methods:

• Utilized compressed sensing theory to acquire undersampled data.
• Employed a neural network to predict signal components from sampled data.
• Reconstructed the original signal using predicted observations for secondary compression.

Main Results:

• Successfully demonstrated a method for detecting rolling bearing faults.
• Achieved significant data reduction through compressed sensing and neural network prediction.
• Enabled secondary compression of the signal, minimizing storage and transmission needs.

Conclusions:

• The proposed method effectively detects rolling bearing faults while reducing data requirements.
• Combining compressed sensing with neural networks offers a promising approach for efficient signal processing.
• This technique alleviates the limitations of traditional high-frequency sampling methods.