Fault Diagnosis of High-Speed Railway Bogies Based on LSTM

The bogie abnormal state is usually reflected in the abnormal vibration of railway bogie. According to the different functions and performance degradation degree of the mechanical components in entire railway system, the abnormal vibration signal is reflected in different ways. This paper presents a Long Short-Term Memory (LSTM)model to analyze the vibration signal that has a deeply excavated internal correlation, since the bogie vibration data are time series data, the use of recurrent neural network (RNN)can extract internal information embedded in the bogie vibration signal well. LSTM network that is able to control the amount of data in input, memory, and output is an upgrade of RNN, the previous data and the future data are connected to analyze the characteristics of the current data by LSTM. And the accuracy rate of this model for multi-fault classification reaches 94 %.