Mileage Calibration And TQI Prediction Of Track Irregularity Data

The track irregularity data is one of the important data for railroad management departments to grasp the quality of rails.However,the complexity of the circuit detection environment which will lead to a certain distortion in the detected irregularity data.Moreover,there is a deviation in the track irregularity data mileage value because of the slipping,locking,and equipment failure of the wheels of the train,which will seriously affect the judgment of the railway managers on the actual state of the track.Therefore,it is necessary to design a suitable algorithm to preprocess the data.At the same time,the prediction of the orbital quality state can prevent the potential hazards of the orbit in advance.The research content of this thesis consists of the following three aspects: preprocessing of the track irregularity data,prediction of TQI(Track Quality Index),and a platform for displaying and analyzing track irregularity data has been developed.In this thesis,different forms of track irregularity data anomalies are dealt with in different ways.In view of the existence of redundant points and mileage intermittent errors in track irregularities.First,contrasting the before and after mileage value in sample point,and removing the error data to extend the mileage.Then filtering out the burrs in the data according to the threshold based on cubical smoothing algorithm with five-point approximation.Aiming at the calibration problem of mileage drift,this thesis proposes a method of waveform section matching based on wavelet transform.Firstly,we compare the curvature information in the ledger data with the curvature of the original data on the canvas to convert the pixels and data points in proportion,and obtain the standard data by artificial interaction.Then,the standard data and the uncalibrated data are segmented,and the Cubic Spline Interpolation is used to expand the data,and the feature vectors are extracted by the wavelet transform,then the feature vectors are normalized.Finally,the Euclidean distance matching waveform segment is used to generate irregularity data with relatively correct mileage values.After the track irregularity data is preprocessed,the quality state of the track can be predicted.In the thesis,the grey prediction and SVM combined model are used to predict.According to the characteristics of TQI data,the background value of the grey model is modified,and the data to be predicted is weighted.The prediction model is obtained by using the least square method to estimate the constant parameters of the grey model.The residual time series value is obtained by comparing the data predicted by the preliminary prediction model with the real data,and the SVM model is trained by the residual sequence value,and the final prediction value is corrected.Finally,a platform for displaying and analyzing track irregularity data is built.The system mainly includes the display and analysis of the original data of track irregularity and ledger data.Using this platform,the data abnormity can be processed,and the standard data can be obtained under the artificial interactive mode,then the data of other batches can be calibrated intelligently according to the standard data,and multiple batches of data can be compared.