Rolling Bearing Remaining Useful Life Prediction Based on CNN-VAE-MBiLSTM.

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Predictive Maintenance

Background:

• Predicting the remaining useful life (RUL) of rolling bearings is critical for preventing equipment failure and optimizing maintenance schedules.
• Real-world industrial environments present challenges like signal interference and process complexity, hindering accurate RUL prediction.
• Existing methods often struggle with feature extraction and capturing complex temporal dependencies in multi-axis sensor data.

Purpose of the Study:

• To propose a novel and robust Remaining Useful Life (RUL) prediction model for rolling bearings.
• To leverage deep learning techniques for automated feature extraction and sequential data analysis.
• To enhance the accuracy and reliability of RUL predictions in complex industrial machinery.

Main Methods:

• A hybrid deep learning model, CNN-VAE-MBiLSTM, integrating Convolutional Neural Network (CNN), Variational Autoencoder (VAE), and Multiple Bi-directional Long Short-Term Memory (MBiLSTM) networks.
• Utilizing CNN-VAE for automated extraction of low-dimensional features from the time-frequency spectrum of multi-axis signals.
• Employing MBiLSTM to capture sequential characteristics and inter-axis feature differences for precise RUL prediction.

Main Results:

• The CNN-VAE component effectively extracts salient features, reducing dimensionality and designer bias.
• The MBiLSTM component accurately predicts RUL by analyzing extracted features and their temporal dependencies across multiple axes.
• Validation on an industrial case demonstrated superior accuracy and anti-noise capabilities compared to existing methods.

Conclusions:

• The proposed CNN-VAE-MBiLSTM model offers a significant advancement in rolling bearing RUL prediction.
• The integration of CNN-VAE and MBiLSTM provides robust feature extraction and accurate RUL forecasting.
• This approach holds promise for improving the reliability and efficiency of predictive maintenance strategies in rolling machinery.