An Effective Deep SVM Approach for Fault Diagnosis of 25 Hz Track Circuit

Track circuits, being vital components in railway signaling systems, are crucial for ensuring the smooth and safe operation of trains. Faults in these circuits can lead to train accidents, operational delays, and a significant threat to railway safety. Traditional fault diagnosis methods often struggle with complex or unidentified faults, but recent advancements in deep learning have revolutionized this domain. Specifically, the 25 Hz phase-sensitive track circuit, commonly used in Chinese high-speed railways, requires reliable fault diagnosis to guarantee uninterrupted and safe train operations. We focuses on collecting multi-dimensional feature time-series data from the 25 Hz phase-sensitive track circuit. By integrating deep learning for feature extraction and Support Vector Machine (SVM) for nonlinear classification, we have developed a model that accurately diagnoses faults in this critical circuit. Extensive experimental results demonstrate the model’s superior performance across various evaluation metrics, achieving a diagnostic accuracy of 97.525%, representing a significant improvement of approximately 1.5% points over Multi-LSTM, CNN-LSTM-Attention, and other existing methods. This approach combines the feature extraction capabilities of deep learning models with the robust nonlinear classification strengths of SVM, providing an effective solution for track circuit fault diagnosis and advancing the field of railway safety.