Series Arc Fault Detection under Vibration Condition Based on N-M-M-B

Area of Science:

• Electrical Engineering
• Mechanical Engineering
• Signal Processing

Background:

• Mechanical faults in motors generate vibrations that can induce series arc faults in electrical contacts.
• Series arc faults compromise power supply reliability.
• Detecting these faults under vibration is crucial for system integrity.

Purpose of the Study:

• To develop and validate a method for detecting series arc faults under varying mechanical vibration conditions.
• To identify the specific motor fault type causing the vibration and subsequent arc fault.

Main Methods:

• Generated series arc faults under controlled vibration conditions, collecting fault phase current and arc voltage signals.
• Utilized noise-assisted multivariate empirical mode decomposition and correlation coefficients for fault feature signal selection.
• Employed a hybrid deep learning model (multi-scale convolutional neural network and bidirectional long short-term memory network) for fault detection and classification.

Main Results:

• The proposed method effectively detects series arc faults under vibration.
• The model can preliminarily identify the motor fault type responsible for the mechanical vibration.
• The developed model demonstrates robust noise immunity and good generalization capabilities.

Conclusions:

• The developed series arc fault detection method is effective and reliable, even in the presence of mechanical vibrations.
• The system offers a dual capability of fault detection and preliminary fault source identification.
• The approach provides a significant advancement in ensuring power supply reliability in dynamic motor environments.