Research On Fault Warning And Fault Diagnosis Method For High Speed Train Axle

The high-speed train axle is the key component to support the train operation.It is of great significance to study the fault warning and diagnosis of the axle to ensure the safe and efficient operation of the train.Considering that the temperature difference between axle temperature and ambient temperature is an important parameter for the reaction axle state,this paper studies the axle fault warning method based on temperature difference variation analysis and the axle fault diagnosis method based on temperature difference estimation by using the monitoring data collected during train operation.The main research work of this paper is as follows:(1)A fault warning method for axle fault of high-speed trains based on temperature difference variation analysis is proposed.The traditional hot-axis fault discrimination is based on the alarm when the temperature difference reaches the alarm threshold.It is a remedy after the fault.As the high-speed train gradually enters the four-level repair,it is foreseeable that the development of the train's visual maintenance technology will become an important research direction.Therefore,this paper studies the fault warning method of faults before the occurrence of hot-axis faults.When the temperature difference reaches 30? and the train running condition is kept unchanged,the method determines whether it will reach the lower alarm threshold within the time interval according to the historical data of the temperature difference change,thereby achieving hot-axis fault warning.Based on the above content,this paper proposes a high-speed train axle fault warning method based on long short-term memory(LSTM)and a high-speed train axle fault warning method based on regression coefficient.The former method based on historical data of temperature difference changes,using LSTM for temperature difference estimation and fault warning based on estimated values.The latter method,firstly,adopts principal component analysis to carry out characteristic dimension reduction on the factors affecting the temperature difference change.Then,the new characteristic factors obtained after dimensionality reduction are fitted to the temperature difference by multiple linear regression,and the regression coeff-icient is extracted as the fault warning basis.Finally,the support vector machine with parameter optimization by genetic algorithm is used to classify the regression coeff-icients and realize fault warning.The experimental results show that the fault warning effect of the axle fault warning method based on regression coefficient is better.(2)A fault diagnosis method for axles of high-speed trains based on temperature difference estimation is proposed.Firstly,according to the historical data of the same running route of the same train,the working condition division based on the speed variation feature is carried out.Secondly,combined with the correlation between the factors affecting the temperature difference change and the temperature difference under different working conditions,the factors with higher correlation under different working conditions are retained.Then,the Back propagation neural network(BPNN)is used to establish the temperature difference estimation model based on characteristic factors under different working conditions.The model can reflect the mapping relationship between temperature difference and characteristic factors under different working conditions.Finally,according to the statistical process control theory,the temperature difference anomaly discrimination is performed on the estimated residual.So as to realize the axle fault diagnosis.The experimental results verify the effectiveness of the axle fault diagnosis method based on temperature difference estimation.