Planetary Gears Feature Extraction and Fault Diagnosis Method Based on VMD and CNN.

Area of Science:

• Mechanical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Planetary gears are critical components in many mechanical systems.
• Effective fault diagnosis is essential for ensuring operational reliability and preventing catastrophic failures.
• Extracting local weak fault features from vibration signals is challenging.

Purpose of the Study:

• To propose a novel feature extraction and fault diagnosis method for planetary gears.
• To enhance the accuracy and efficiency of fault identification and classification.
• To address the challenge of detecting local weak features in planetary gear systems.

Main Methods:

• Variational Mode Decomposition (VMD) was employed to decompose vibration signals into intrinsic mode components.
• Singular Value Decomposition (SVD) was utilized to extract local feature information from partitioned mode matrices.
• A Convolutional Neural Network (CNN) was trained using the extracted singular value vector matrices for fault classification.

Main Results:

• The proposed method successfully extracted local weak feature information from planetary gear vibration signals.
• Accurate identification and classification of different fault states were achieved, with a 100% recognition rate.
• The VMD-based method demonstrated superior performance compared to ensemble empirical mode decomposition (EEMD) in terms of recognition rate and training time.
• The method proved effective for degradation recognition in planetary gears.

Conclusions:

• The integrated VMD-SVD-CNN approach is an effective technique for planetary gear fault diagnosis.
• The method excels at extracting subtle fault signatures, leading to high diagnostic accuracy.
• This approach offers a promising solution for condition monitoring and predictive maintenance of planetary gearboxes.