A novel bearing current signal diagnosis method combining variational modal decomposition and improved random forests

Area of Science:

• Mechanical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Bearing failures are critical in rotating machinery, leading to costly downtime.
• Traditional fault diagnosis methods struggle with noisy bearing current signals.
• Accurate and robust fault diagnosis is essential for predictive maintenance.

Purpose of the Study:

• To develop a novel fault diagnosis approach for bearings using current signals.
• To enhance the accuracy and reliability of bearing fault detection.
• To address challenges posed by strong background noise in current signals.

Main Methods:

• Bearing current signals were decomposed using variational modal decomposition (VMD).
• Intrinsic mode functions (IMFs) were constructed as feature vectors using kurtosis.
• A random forest (RF) classifier was optimized with the whale optimization algorithm (WOA) for fault diagnosis.

Main Results:

• The proposed method achieved a classification accuracy of 95.11% for real damaged bearing fault types.
• The fault diagnosis accuracy for manually damaged bearings reached 93.83%.
• The WOA-optimized RF model demonstrated superior performance compared to traditional models.

Conclusions:

• The developed fault diagnosis approach shows high accuracy and strong generalization ability.
• VMD and WOA-optimized RF effectively handle noisy signals for reliable bearing fault detection.
• This method offers a promising solution for industrial bearing health monitoring.