Research On Prediction Method Of Temperature Rise Of High Speed Train Bearing Based On ARIMA And BP Neural Network

Bearing as an important component in the safe operation of high-speed train,is in a complex service environment,which is prone to wear,crack,spalling and other faults,thus bringing a great threat to the operation and safety of high-speed train.Therefore,it is very important to carry out real-time monitoring of the condition of high-speed train bearings.At present,the real-time monitoring system of the on-board shaft temperature is mainly used to monitor the temperature change of the bearing.By setting the warning alarm threshold for early warning,remedial measures can be taken to effectively avoid dangerous accidents caused by shaft heat and fuel.Due to the complex operating conditions of high-speed trains,there are many factors affecting bearing temperature rise,and the phenomenon of false alarm and missing alarm frequently occurs in the existing shaft temperature monitoring system.Therefore,based on the temperature data of high-speed train motor stator,pinion gearbox bearing and large gearbox bearing,this paper studies the temperature rise prediction method of high-speed train bearing,so as to effectively guarantee the operation safety of high-speed train and reduce the maintenance cost.The main research contents are as follows:(1)Firstly,the prediction model of high-speed train shaft temperature based on ARIMA time series analysis is established.ARIMA model modeling only considers the historical temperature data of the high-speed train bearing,and the future temperature data can be predicted through the historical temperature data.Moreover,the temperature data of high-speed train bearing is complex,so the ARIMA time series analysis model has some limitations in the temperature prediction of high-speed train bearing.It can be seen from the prediction results that the prediction effect of the time series analysis model for the static prediction of bearing temperature is the best.In the dynamic prediction,with the increase of the prediction step,the hysteresis of the predicted data becomes more and more obvious,and the delay increases continuously,and the prediction performance decreases accordingly.(2)Secondly,the prediction model of high speed train shaft temperature based on BP neural network is established.Based on the principle of bearing temperature rise,the sensitive factors affecting the change of shaft temperature are determined.Establish data samples according to bearing history data.According to the BP neural network algorithm,the number of network layers of BP neural network,the number of nodes of hidden layer and the selection of transfer function between network layers are determined.After normalization processing according to the data sample set,the model is trained to establish the high-speed train bearing temperature prediction model based on the BP neural network.Finally,this model is used to predict the measured temperature data of three kinds of bearings of high-speed train,and the prediction error results are analyzed.The results show that the prediction effect of the high-speed train shaft temperature prediction model based on BP neural network is better and its generalization ability is better.(3)Finally,combining the advantages of the above two methods,a high-speed train shaft temperature prediction model optimized by ARIMA-BP neural network is established.In order to guarantee the high accuracy of the high-speed train bearing temperature,this article will ARIMA model combined with BP neural network forecasting model,using ARIMA model for prediction of BP neural network forecasting error correction,to further improve the prediction accuracy,the temperature prediction model was optimized and the ARIMA-BP neural network optimization of high-speed train axle temperature prediction model for high-speed train 3 kinds of bearing temperature sequence forecast analysis,results show that the optimization model can effectively improve the single model to predict the error accumulation,thus to further improve the high speed train overall prediction effect of three kinds of bearing temperature sequences.