Remaining Useful Life Prediction Model for Rolling Bearings Based on MFPE-MACNN

Area of Science:

• Mechanical Engineering
• Data Science
• Artificial Intelligence

Background:

• Rolling bearing degradation prediction is crucial for machinery maintenance.
• Existing deep learning models struggle with complex time-series feature learning.
• Redundant information in degradation data hinders accurate predictions.

Purpose of the Study:

• To develop a robust remaining useful life (RUL) prediction model for rolling bearings.
• To address challenges in feature extraction and information redundancy in degradation data.
• To improve the accuracy of RUL prediction using advanced deep learning techniques.

Main Methods:

• Resonance sparse decomposition to separate signal components.
• Multiscale fusion permutation entropy (MFPE) for feature extraction.
• Locally linear embedding for dimensionality reduction.
• Multiscale convolutional attention neural network (MACNN) for RUL prediction.

Main Results:

• The proposed MFPE and MACNN model effectively reduces redundant information.
• The MACNN model successfully learns features across multiple time scales.
• The method demonstrates superior performance in RUL prediction accuracy compared to existing models.

Conclusions:

• The developed model offers an effective solution for rolling bearing RUL prediction.
• The integration of MFPE and MACNN enhances feature representation and prediction accuracy.
• This approach provides a valuable tool for predictive maintenance in mechanical systems.