Abstract: The operating conditions of high-speed vehicles were harsh and changeable, and the reliability of suspension system was related to the safety and riding comfort of trains. When the suspension system of the high-speed vehicles fails, the vibration signal was non-linear and non-stationary. Therefore, this paper proposeded a fault identification method for vehicle suspension system based on Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) model. The paper established a vehicle track coupling dynamic model with suspension system on the SIMPACK platform, and obtained vibration signals of various components of the vehicle system under healthy and various fault states, and took the vibration acceleration signal of the components associated with the faulty component as the input of the model, through the construction of a CNN-LSTM model, extracted features and predicted classification of time-series signals, thereby achieved fault recognition of the vehicle suspension system, evaluated the method by constructing fault datasets for different operating conditions. The experimental results showed that the fault recognition accuracy of the method can reach up to 98 % under the same speed level, and the fault recognition accuracy can reach up to 99 % under different speed levels, which verifies the effectiveness of the method.