Vibration sensor-based bearing fault diagnosis using ellipsoid-ARTMAP and differential evolution algorithms.

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Rolling element bearings are critical components in rotating machinery.
• Effective fault classification is essential for preventing catastrophic failures and ensuring operational safety.
• Existing fault diagnosis methods may struggle with complex vibration signal patterns and overfitting.

Purpose of the Study:

• To propose a novel vibration sensor-based fault diagnosis method for rolling element bearings.
• To develop a robust and accurate fault classification system using an Ellipsoid-ARTMAP (EAM) network.
• To optimize the EAM network parameters using a differential evolution (DE) algorithm for improved performance.

Main Methods:

• Feature extraction from vibration signals using wavelet packet decomposition.
• Feature selection via a minimum-redundancy maximum-relevancy algorithm to reduce dimensionality.
• Fault classification using a DE-optimized EAM (DE-EAM) network with hyper-ellipsoid nodes.
• Parameter optimization using DE with dual objectives: classification accuracy and node count.

Main Results:

• The proposed DE-EAM classifier demonstrated reliable and accurate recognition of four types of rolling element bearing faults under various loads.
• The EAM network effectively depicted decision boundaries for dispersed samples, mitigating overfitting.
• A two-fold cross-validation scheme with randomized sample ordering confirmed the classifier's robustness.

Conclusions:

• The DE-EAM method offers a highly effective approach for fault diagnosis in rolling element bearings.
• The combination of EAM and DE provides a powerful tool for analyzing complex vibration data.
• This method enhances the safety and reliability of rotating machinery through accurate fault classification.