Fault detection and isolation system for train power lines using machine learning

This work presents a hybrid failure detection and isolation system for train power lines, combining a convolutional neural network (CNN) classifier with root mean square (RMS) feature extraction. Under varying noise levels, the system recognizes and isolates defects, such as ground faults, open circuits, and short circuits. The technique successfully lowers noise sensitivity by employing RMS to preprocess time-domain electrical signals. These characteristics train the CNN model, producing accurate fault condition classifications. According to experimental data, the suggested method achieves better accuracy and robustness than conventional deep learning models like support vector machines and K-nearest neighbors. The framework's potential for real-time application in railway power systems is highlighted by its efficacy under various noise levels.