Bayesian-Optimized Hybrid Kernel SVM for Rolling Bearing Fault Diagnosis.

Area of Science:

• Mechanical Engineering
• Signal Processing
• Machine Learning

Background:

• Rolling bearings are critical components in machinery, and their failure can lead to significant economic losses.
• Accurate fault diagnosis is essential for predictive maintenance and preventing catastrophic failures.
• Traditional methods struggle with the nonlinearity and nonstationarity of vibration signals from bearing faults.

Purpose of the Study:

• To propose a novel fault diagnosis model for rolling bearings.
• To address the challenges of ambiguous fault identification in bearing vibration signals.
• To enhance the accuracy and reliability of bearing fault diagnosis.

Main Methods:

• Feature extraction from vibration signals using Discrete Fourier Transform (DFT).
• Development of a hybrid kernel Support Vector Machine (SVM) combining polynomial and radial basis functions.
• Optimization of SVM parameters using Bayesian Optimization (BO) for improved performance.
• Validation using the Case Western Reserve University bearing dataset and laboratory experiments.

Main Results:

• The proposed model achieved a fault diagnosis accuracy of 100% on the Case Western Reserve University dataset.
• Compared to direct SVM input (85% accuracy), the optimized model showed significant improvement.
• Laboratory verification demonstrated 100% accuracy, with an average of 96.7% across five replicates.
• The Bayesian-optimized hybrid kernel SVM outperformed other diagnostic models in accuracy.

Conclusions:

• The proposed fault diagnosis model is highly effective and superior for rolling bearings.
• The combination of hybrid kernel SVM and Bayesian optimization offers a robust solution for complex fault identification.
• The method demonstrates significant potential for industrial applications in predictive maintenance.