Application Combining VMD and ResNet101 in Intelligent Diagnosis of Motor Faults

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Motor failures pose significant risks to critical industrial equipment like wind turbines and electric vehicles.
• Effective motor fault diagnosis is crucial for ensuring operational safety and reliability.
• Existing methods often require separate steps for signal analysis and feature extraction.

Purpose of the Study:

• To develop an end-to-end intelligent fault diagnosis system for motors.
• To integrate signal decomposition, feature extraction, and health status recognition into a unified framework.
• To improve the accuracy and efficiency of motor fault detection.

Main Methods:

• Employed variational mode decomposition (VMD) for adaptive signal processing of motor vibration signals.
• Utilized a residual neural network 101 (ResNet101) for deep learning-based feature extraction and classification.
• Developed a framework unifying signal acquisition, feature extraction, and fault identification.

Main Results:

• The proposed system demonstrated high performance in identifying six common motor faults.
• Achieved a prediction accuracy rate of 94% on a dataset from the Federal University of Rio de Janeiro (UFRJ).
• The integrated approach effectively handled complex vibration signals under variable working conditions.

Conclusions:

• The combined VMD and ResNet101 approach offers a robust and accurate method for intelligent motor fault diagnosis.
• This system provides a more effective solution for mechanical fault diagnosis engineering.
• The research highlights the potential for wide application in industrial settings requiring high motor reliability.