Related Experiment Video
Updated: Jul 5, 2025
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
42.9K
Graph Multi-Scale Permutation Entropy for Bearing Fault Diagnosis
Qingwen Fan1, Yuqi Liu1, Jingyuan Yang2
1School of Mechanical Engineering, Sichuan University, Chengdu 610017, China.
Sensors (Basel, Switzerland)
|January 11, 2024
Summary
This study introduces graph multi-scale permutation entropy (MPEG) for accurate bearing fault diagnosis. The novel method enhances vibration analysis for improved machine health monitoring.
Area of Science:
- Mechanical Engineering
- Signal Processing
- Machine Learning
Background:
- Bearing faults are a major cause of failure in rotating machinery, necessitating accurate diagnostic techniques.
- Vibration-based monitoring is a common approach for detecting these faults.
- Graph signal processing offers potential for extracting critical fault features.
Purpose of the Study:
- To propose a novel method, graph multi-scale permutation entropy (MPEG), for enhanced bearing fault diagnosis.
- To evaluate the effectiveness of MPEG against conventional and other graph entropy methods.
Main Methods:
- Vibration signals are converted into visibility graphs.
- Graph coarsening generates multi-scale graph representations.
- Permutation entropy is calculated on these graphs to extract fault features.
- Support Vector Machine (SVM) is employed for classification of bearing fault types.
Main Results:
- The proposed MPEG method demonstrates higher accuracy in bearing fault diagnosis compared to existing techniques.
- The method exhibits superior robustness and de-noising capabilities.
- Validation was performed using both open-source and laboratory datasets.
Conclusions:
- Graph multi-scale permutation entropy (MPEG) is an effective technique for bearing fault diagnosis.
- The method offers significant improvements in accuracy, robustness, and noise reduction for machine health monitoring.
