Dynamic Threshold Estimation And Fault Diagnosis Of High Speed Train Axle Temperature Based On Data

The axle temperature is an important index to reflect the performance of the axle.In view of the existing axle fault detection system based on fixed axle temperature threshold has high false negative and false alarm rate.And in the monitoring data,the normal axle temperature data and the fault axle temperature data sample are seriously unbalanced,the normal amount of data is far greater than the amount of fault data.Based on the train monitoring data,this paper establishes a dynamic threshold estimation model for the axle temperature adapted to the environment and running state,and the fault diagnosis method of the hot axis,which can effectively improve the reliability and stability of the monitoring equipment,and provide a new method and means for the axle fault diagnosis of the high-speed train.The main contents of this paper include the following four aspects:1、Through the statistical analysis of the monitoring data of train operation,the importance analysis model of axle temperature and other monitoring factors is established,and the influence factors and correlation degree of the axle temperature change in the whole running process of the train and under the three operating conditions of traction,inerting and braking respectively are explored.2、According to the importance analysis model of train under the condition of traction,braking and laziness,the axial temperature estimation model based on multiple regression,the axle temperature estimation model based on the random forest and the axial temperature estimation model based on the gradient regression tree are established respectively.On this basis,according to the model estimation error,a model fusion based axle temperature estimation is established,which enhances the generalization ability and robustness of the model,and the experimental results prove the effectiveness of the proposed method.3、Aiming at the limitation of the traditional off-line axle temperature estimation,a model of train running condition based on fuzzy membership function is established.On this basis,an estimation algorithm of on-line axle temperature threshold is proposed by combining the model of working condition axle temperature estimation and fuzzy logic,and the automatic analysis of the axle temperature threshold is realized.In addition,using the historical monitoring data of axle temperature,a model of axial temperature prediction based on LSTM is proposed,which lays the foundation for the fault early warning of axle temperature,and the experimental results prove the effectiveness of the method.4、According to the hot axle fault diagnosis of train,the factors that affect the train thermal shaft fault are analyzed.Based on the measured data in the running process of a large number of high speed trains,the data based fault diagnosis model of the train hot axle is established.The experimental results show that the accuracy and time efficiency of the model are high.