Fault Detection for Train-Controlled On-Board Equipment Using a Hybrid CNN-LSTM Model

With the continuous development of railway transport intelligence, the fault detection technology of train-controlled on-board equipment, as a key system to ensure train safety and efficient dispatching, has attracted much attention. In this paper, a fault detection method based on CNN-LSTM model is proposed for the local anomalies and long time-series dependence in the sensor data of train-controlled vehicle-mounted equipment. The method uses convolutional neural network to automatically extract the local tiny fluctuations in the equipment signals, and captures the global dynamic trend through the long and short-term memory network, which achieves the accurate identification of the abnormal state of the equipment. The experimental results show that the model can still maintain a high detection accuracy and low misjudgement rate in a high noise environment, which provides effective data support for the preventive maintenance and risk warning of train-controlled on-board equipment, and also provides a solid theoretical and technological basis for the intelligent upgrade of the railway system.