| The wheel is a basic component of the EMU travel system and its performance determines whether the EMU can operate efficiently and safely.The flourishing development of artificial intelligence and big data technology has provided a new direction for the use of EMU wheels.This paper collects and analyses the track data and inspection data of the wheels of EMU,uses data mining technology to study the degradation pattern of the wheels of EMU and the strategy of re-profiling,and visualizes the functions of wheel information management,trend prediction and re-profiling decision.The main elements are as follows:(1)A time series-random forest regression model is developed to analyse the trend of wheel dimensions of the EMU with the operation of the EMU.The model of ARIMA(Differential Sequence Adaptive Model)is trained using the historical mileage series of the EMU and enables prediction of the average daily mileage of the EMU for a future period.The model prediction accuracy is 0.9867,MSE is 0.0669 and MAE is 0.2148.The results were compared with those of support vector machine and decision tree algorithms to verify the effectiveness of the time series-random forest regression algorithm for predicting wheel size parameters.(2)A logistic regression-random forest regression model is established to realize the wheel state judgement and the prediction of the wheel turning volume within a single re-profiling cycle of the wheels of the EMU.Firstly,the historical wheel inspection data samples are tagged with the state,and the data imbalance of the inspection data samples is dealt with by the method of oversampling;after the samples are processed,a logistic regression model is established to realize the wheel state judgement.The random forest algorithm was trained using the data before and after re-profiling as the features and labels of the data samples,so that the trained model could simulate the wheel re-profiling rules,predict the wheel re-profiling volume for the wheel judged to be in urgent need of re-profiling,and predict the wheel diameter value after re-profiling.The accuracy of the model prediction is0.9971 with a mean relative error of 0.19.(3)The RFID tag is used to identify and manage the wheels of the EMU and to trace the full life cycle of the wheels.The back-end database based on My SQL is established to centralize the management of the operation data,history data and inspection data of the wheels of the EMU.The system is based on the Python language for visualization and programming of the functional modules;it enables human-machine interaction for wheel identification,wheel information management,wheel size prediction and tread re-profiling recommendations.The system enables the unified management of wheel data,the prediction of key wheel dimensions,the determination of wheel re-profiling status and the prediction of re-profiling volume.The system helps to manage wheel data in a standardised way and provides reference advice on wheel re-profiling management. |
