Novel image-free arc detection method for pantograph-catenary systems based on direct DWT-ANN signal analysis

Area of Science:

• Electrical Engineering
• Railway Systems Engineering
• Signal Processing

Background:

• Arc faults in pantograph catenary systems are a major concern for electric railway operations, impacting safety and reliability.
• Current detection methods using image processing and deep learning face computational delays, hindering real-time application.
• The need for efficient, real-time arc fault detection without additional sensors is critical for railway infrastructure.

Purpose of the Study:

• To develop and validate a novel, image-free method for real-time arc fault detection in electric railway systems.
• To evaluate the effectiveness of Discrete Wavelet Transform-Artificial Neural Networks (DWT-ANN) in analyzing traction current signals for arc fault transients.
• To establish a computationally efficient and robust solution for monitoring railway systems.

Main Methods:

• The proposed method analyzes measured traction current signals, bypassing image processing limitations.
• Discrete Wavelet Transform (DWT) is employed to extract transient arc features from current waveforms.
• Artificial Neural Networks (ANN) are utilized for classifying these extracted features, with specific mother wavelets (Daubechies db9, Symlet sym8) identified for optimal performance.

Main Results:

• The DWT-ANN framework successfully extracts transient arc features directly from traction current signals.
• The ANN achieved high classification accuracy, indicated by a strong regression coefficient.
• Algorithm robustness was confirmed through consistent detection performance across different wavelet bases during validation.

Conclusions:

• The DWT-ANN approach provides a computationally efficient and practical solution for real-time arc fault detection in electric railways.
• This image-free method enhances the reliability and safety of railway operations by enabling timely identification of arc faults.
• The study demonstrates the feasibility of using traction current signal analysis for robust arc fault monitoring.