Modified multiscale weighted permutation entropy and optimized support vector machine method for rolling bearing...

Area of Science:

• Mechanical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Rolling bearing vibration signals are complex, non-linear, and non-stationary, making fault diagnosis challenging with conventional methods.
• Extracting sensitive features and recognizing patterns in these signals is crucial for effective fault diagnosis.

Purpose of the Study:

• To propose a novel intelligent fault-diagnosis method for rolling bearings.
• To enhance the accuracy and reliability of bearing fault diagnosis through advanced signal processing and machine learning techniques.

Main Methods:

• A novel non-linear technique, generalized composite multiscale weighted permutation entropy (GCMWPE), was developed for sensitive feature extraction from multiple scales.
• Supervised Isomap (S-Iso) was employed for dimensionality reduction of the extracted features.
• A support vector machine (SVM) optimized by the marine predators algorithm (MPA-SVM) was utilized for pattern recognition and fault diagnosis.

Main Results:

• The proposed GCMWPE method provides more stable entropy values by using a generalized composite coarse-grained structure.
• The MPA-SVM effectively diagnoses and identifies bearing states using the reduced feature set.
• Experimental validation on two bearing fault datasets confirmed the method's high diagnostic accuracy.

Conclusions:

• The integrated approach of GCMWPE, S-Iso, and MPA-SVM offers a robust and accurate solution for rolling bearing fault diagnosis.
• This intelligent method overcomes the limitations of conventional signal processing techniques for complex vibration signals.
• The study demonstrates significant potential for improving machinery health monitoring and predictive maintenance.