Research On Temperature Prediction Method Of Gearbox Bearing For High-speed Train

With the rapid development of high-speed railway industry,the running speed of highspeed trains is constantly increasing.Meanwhile,the safety problem of high-speed trains cannot be ignored.Bearing,as an important part of the bogie of high-speed train,is a vulnerable part of high-speed train because of its complex and varied operating conditions and large load.The working state of bearing will directly affect the safety of high-speed train.The existing onboard detection system of high-speed trains uses temperature sensors to collect shaft temperature data,and determines the occurrence of bearing faults and gives alarm by setting temperature threshold.Although this method can effectively avoid hot shaft accidents caused by rising bearing temperature,the train needs to slow down or stop immediately after an alarm occurs,which will destroy the original train operation order.Cause unnecessary losses and have negative social impact.In this paper,the temperature prediction method of gearbox bearing of high-speed train is studied,so as to realize the early warning of the temperature rise of bearing and ensure the train running safety.The main research contents are as follows:(1)Through the analysis of the mechanism of bearing heat generation and heat dissipation,the sensitive factors of gearbox bearing temperature change during the running of high-speed train are obtained.On this basis,the temperature prediction model of high-speed train gearbox bearings based on support vector machine was established to predict the temperature of four types of gearbox bearings,and the parameter combination of the optimal kernel parameter g and penalty coefficient C was obtained by particle swarm optimization algorithm.The results show that the temperature prediction model of gearbox bearing with support vector machine optimized by particle swarm optimization algorithm is more accurate,but the prediction accuracy will decrease when it is used for a long time.(2)A gearbox bearing temperature prediction model is established based on its variant-long and short term memory network by analyzing the shortcomings of the cyclic neural network.The gearbox bearing temperature prediction model based on the long and short term memory network and the gearbox bearing temperature prediction model based on the two-way long and short term memory network were established respectively to predict the four kinds of gearbox bearing temperature,and the prediction results of the two kinds of networks were compared and analyzed.Because the bidirectional long short-term memory network not only learns the characteristics of the past data but also makes reference to the future data information,the results show that when the bidirectional long short-term memory network is used to predict the temperature of gearbox bearing,the prediction effect is better than that of the standard long short-term memory network,and the prediction result is more accurate.(3)By combining the support vector machine prediction model of particle swarm optimization algorithm and the axial temperature prediction model of bidirectional long and short term memory network,the gearbox bearing temperature can be predicted more accurately.The Kalman filter algorithm can effectively avoid the error accumulation caused by a single model in the prediction because it updates the process of time and observation amount when making the combined prediction of the two models.The results show that the Kalman filter fusion prediction method has higher prediction accuracy for gearbox bearing temperature.The statistical process control method is introduced to verify the accuracy of the Kalman filter fusion prediction method.The results show that the Kalman filter fusion prediction method has certain reliability in the temperature prediction of high-speed train gear box bearing.