A noval RUL prediction method for rolling bearing: TcLstmNet-CBAM

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Predictive Maintenance

Background:

• Accurate remaining useful life (RUL) prediction for rolling bearings is crucial for mechanical systems.
• Existing deep learning methods struggle with feature extraction and prediction accuracy due to reliance on limited temporal dependencies.

Purpose of the Study:

• To develop an advanced deep learning method for enhanced RUL prediction in rolling bearings.
• To address limitations in feature extraction and accuracy of current predictive models.

Main Methods:

• A novel TcLstmNet-CBAM model is proposed, integrating Temporal Convolutional Network (TCN) for long-term dependencies and Long Short-Term Memory (LSTM) for short-term dependencies.
• A Convolutional Block Attention Module (CBAM) is employed for multi-dimensional feature weighting, prioritizing critical information.
• The method was validated on PHM2012 and XJTU-SY rolling bearing datasets.

Main Results:

• The TcLstmNet-CBAM method achieved a Mean Absolute Error (MAE) of 2.287 and Root Mean Square Error (RMSE) of 3.123.
• Experimental results demonstrated superior performance compared to other prevalent deep learning prediction methods.
• The approach enables more comprehensive feature extraction and emphasizes key features for improved RUL prediction.

Conclusions:

• The proposed TcLstmNet-CBAM method effectively predicts the remaining useful life of rolling bearings.
• The integration of TCN, LSTM, and CBAM significantly enhances RUL prediction accuracy.
• This study validates the effectiveness and superiority of the TcLstmNet-CBAM approach for bearing prognostics.