Research On Prediction Of Bearing Temperature For Electric Locomotive

Bearing failure caused by hot axle is the most common type of train failure,and axle temperature is an effective indicator to reflect train operation status.During the operation of the train,if the bearing heats up abnormally,the service life of the train bearing will be reduced in a short time,and the maintenance and replacement cost will be increased.In severe cases,the axle will break and cause an accident.The basis of vehicle fault detection caused by increased axle temperature is the measurement of axle temperature.In order to estimate axle temperature and diagnose faults,the following researches are carried out in this paper.1.This article takes the bearings of the HXD1 C electric locomotive running in real time in our section as the research object.First,based on the JK00430 on-board monitoring device used in our section,the bearing temperature data is collected by the temperature sensor mounted on it;2.Considering that the collected bearing temperature monitoring data is interfered by different factors such as environmental conditions and operating conditions,there are problems such as large noise,data distortion,data missing,and different dimensions,and it is inconvenient to analyze the collected data.Noise reduction is performed on the collected data.Processing,smoothing and normalization processing methods to obtain higher-quality data information after preprocessing;3.In view of the fact that the change of bearing temperature rise under different operating conditions of the train is affected by various factors,the Pearson correlation coefficient method,the Kendall correlation coefficient method and the Spearman correlation coefficient method are used to analyze the overall operation and performance of the rolling stock.Based on the correlation analysis of the bearing temperature under various operating conditions,We found that the reason that has the biggest influence on the bearing temperature is the ambient temperature,and the dynamic threshold estimation model of the shaft temperature is established according to the difference between the bearing temperature and the ambient temperature;4.Based on the multiple linear regression method,random forest method and gradient regression tree method,the axle temperature estimation model was established,and then different weight values were assigned according to the evaluation effect of each model,and the pairwise fusion and all fusion axle temperature estimation models were obtained.The single axle temperature estimation model and the fusion model are compared,and the best effect is the bearing temperature estimation based on the fusion of random forest and gradient regression tree;5.Finally,the established bearing temperature prediction model was tested,and the axial temperature alarm information detected by the HXD1 C fuel power locomotive detected by a certain segment of JK00430 is referred to,and then the temperature of these bearings is predicted by the model to compare the temperature threshold of the bearing temperature.The bearing that produces an abnormally temperature rise corresponds to the real fault information after the top wheel detection,one by one,as a method of verifying the accuracy of the model.