Fault Diagnosis of Planetary Gearboxes Based on LSTM Improved via Feature Extraction Using VMD, Fusion Entropy, and...

Area of Science:

• Mechanical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Planetary gearboxes are critical in many mechanical systems, but their complex vibration signals pose challenges for fault diagnosis.
• Effective fault feature extraction is essential for accurate and efficient diagnosis of gearbox failures.

Purpose of the Study:

• To propose a novel feature extraction method for planetary gearbox fault diagnosis.
• To improve diagnostic accuracy and efficiency by addressing signal complexity, feature redundancy, and interference.

Main Methods:

• Variational Mode Decomposition (VMD) for processing nonlinear and non-stationary vibration signals.
• A novel fusion entropy incorporating refined composite multi-scale entropies for comprehensive feature extraction.
• Random Forest (RF) for feature importance calculation and selection.
• Long Short-Term Memory (LSTM) networks for fault classification.

Main Results:

• The proposed fusion entropy method demonstrated higher accuracy than single entropy values.
• RF-based feature selection effectively reduced interference and improved diagnostic efficiency.
• The overall fault diagnosis method achieved high accuracy across varying rotational speeds and noise levels.

Conclusions:

• The developed method offers a robust solution for planetary gearbox fault diagnosis.
• The combination of VMD, fusion entropy, RF, and LSTM provides a powerful tool for analyzing complex machinery vibrations.