In order to improve the driving safety of the braking system of urban rail vehicles and reduce the probability of traffic accidents,a communication fault prediction method for the braking system of urban rail vehicles based on ST-CNN is proposed.The paper analyzes the framework structure of the spatiotemporal convolutional neural prediction model,establishes a spatio-temporal sequence matrix,and obtains the normal patterns and changing trends of communication data.It establishes a convo-lutional neural network for communication feature extraction,extracts communication features for alarm types,and implements dimensionality reduction processing.Communication features of different alarm types are quantified through pooling operations,and a fully connected layer is used to convert different alarm feature vector values to complete communication fault prediction for the braking system.The experimental results show that the overall stability of the F1 value of the proposed method is above 0.97,and the prediction time is below 0.5 s.The predicted value is highly similar to the actual value,which can meet the actu-al needs of urban rail transit.
关键词
城轨车辆/时空矩阵/卷积神经网络/经验模态分解/映射函数
Key words
urban rail vehicle/spatiotemporal matrix/convolutional neural network/empirical mode decomposition/mapping function
维普
万方数据