Study On Pre-Warning Method Of Abnormal Temperature Rise Of Bearing On High Speed Train

Bearings are the key components which ensure the safe operation of high-speed trains.The harsh service condition can easily result in wear,peeling,cracks,and other failures,which poses a great threat to the safety of high-speed train operations.It is necessary to implement condition monitoring on bearings.Temperature monitoring is the main measure of condition monitoring for high speed trains,which alarm based on pre-set threshold.This method can effectively avoid serious accidents caused by hot-box.However,once an alarm occurs,it will lead to vehicle operation accidents.This has caused economic losses and negative social impact.Therefore,It is both important in engineering and scientific research to research on pre-warning method of abnormal temperature Rise(ATR)of bearing on high speed train to propose a set of methods for early detection of abnormalities and for continuous tracking and pre-warning.Because of the complex operating conditions of high speed trains,it is difficult to build an accurate thermal model of bearing,and the abnormal temperature rise samples are scarce,which leads to the difficult to establish a supervised ATR diagnosis model.In addition,there are many factors that affect bearing temperature,and the diagnosis accuracy and prediction accuracy for ATR of bearings are low.In order to solve these problems,this thesis carried out the research on the ATR diagnosis model,the temperature prediction model and the ATR pre-warning strategy for bearings on high speed trains,the abnormal condition of bearing was detected as soon as possible,by the ATR diagnosis model,and then the temperature prediction model was employed to determine bearings' future condition,combining with the pre-warning strategy,the operational safety for high-speed train is ensured,and more time is also reserved for hot-box troubleshooting.The main research results are as follows:(1)A diagnosis model based on feature clustering of correlational measuring points(CMP)for ATR of high-speed train bearing is proposed.For the CMP have similar location and operation conditions,the temperature changes of CMP are also similar under normal conditions,based on the K-means and DB-SCAN fusion,an ATR diagnosis model of the bearing at CMP is constructed.The verification results show that the model solves the problem that subjective setting of the discriminant threshold parameter of the classical cluster diagnosis model,and the misjudgment rate is reduced.(2)A diagnosis model is presented based on the historical feature under operating condition for ATR of high-speed train bearing.the sensitive factors of temperature rise of bearings are determined,through the analysis of the temperature rise mechanism of the bearings.Based on the historical service data,the relationship between operating conditions and maximum temperature rise is further analyzed.Using support vector machine regression,a model for estimating bearing temperature rise based on historical operating data is constructed,and the difference between the actual temperature rise of the bearing and the estimated temperature rise is diagnosed by the statistical process control(SPC)method.The results of the verification show that the model has an accurate application effect.(3)A prediction model of high speed train bearing temperature is presented based on multifunction hybird of least square regression(LSR).According to the statistical analysis of the temperature signal of bearings on high speed train,the characteristics of the bearing temperature change are obtained.According to this,different basis functions are selected and the trend of the bearing temperature is predicted based on its historical neighborhood.Then the prediction error distribution of different basis functions is analyzed,and the bearing temperature prediction model is built based on the multi-function hybird using the LSR.The model has accurate prediction accuracy under various operating conditions,and effectively improves the deficiency of the poor adaptability of prediction model based on the single basis function.(4)A two-level pre-warning strategy for ATR of high-speed trains is developed based on a combination of diagnosis and prediction models.The diagnosis model for ATR based on feature clustering of CMP and the diagnosis model for ATR based on the historical feature under operating condition are used to diagnose the ATR of bearing from the two dimensions of time and space,then make a primary level warning for diagnosed ATR bearing.At the same time,the ATR is predicted based on multi-function hybrid of LSR.If the forecast results exceed the set alarm threshold the secondary level pre-warning is carried out to realize the effective prewarning of the ART of the high-speed train.