Research On The Application Of Historical Data Of Track Irregularity

The track is the basis for ensuring the safe and smooth operation of the train at all times.Since the "13th Five-Year Plan",China’s railway construction has developed rapidly,which has brought a huge amount of work to railway engineering personnel in the management of track smoothness.The traditional track management method(afterthe-fact troubleshooting or periodic repair)is difficult to meet the established line state management standards under the existing "sunroof" operating time and manpowermaterial conditions.Preventive maintenance,also known as state repair,is the best way to maintain the track,that is,to guide the work department when and where the irregularity changes occur,and to solve the above problems reasonably.The fundamental way to implement preventive maintenance is to establish an irregular prediction model of the railway,that is,the trend of deterioration of the track smoothness.The purpose of this paper is to analyze the historical data of the track detector and solve the problem of mileage deviation between the historical data of the irregularity,that is,to ensure the contrast between the historical data,establish the corresponding prediction model,and determine the smoothness of the track in advance.The track irregularity prediction model can provide technical support for railway engineering personnel to implement the track state repair strategy,improve work efficiency and make full use of existing manpower and material resources and “sunroof” operation time.First,observe and analyze the original chord data of right vertical profile detected by the track detector.On the waveform of the double abscissa,it can be found that there are peaks and troughs of the original string mutation every 800 detection points or every 100 m.Combined with the working principle of the rail detector with a sampling interval of 0.125 m and the structural characteristics of the track,it is concluded that the weld between the adjacent rails of the 100 m length of the high-speed railway after the longtrack locking will cause the sudden change of the original chords of right vertical profile in conclusion.According to domestic and foreign research,the characteristic deviation of the track can be used as the absolute mileage of the key equipment to correct the mileage deviation between the track irregularity historical data.Therefore,extracting the weld as a key device is a prerequisite for correcting the mileage deviation between historical data.In this paper,the local(dynamic)standard deviation and the moving average function with a step size of 5 are used to preprocess the original chord data of right vertical profile,and the local features are strengthened to avoid the extraction of non-existent welds.The cubic spline interpolation model is used to extract the peak mutation.The maximum position is used as the weld seam;the loss function is established to match the characteristic welds extracted from the unsmooth historical data on the same track interval to avoid the uneven data misalignment between the adjacent two welds.Secondly,the previous detection data is a reference template,and the subsequent detection data is a test template.Taking the extracted weld mileage as the mileage reference,this paper divides the reference template and the test template into several segments with the weld as the demarcation point,and uses the Dynamic Time Warping(DTW)algorithm to perform point-to-point segmentation to align the entire interval deviation.Compared with the autocorrelation technology,the correction effect of the DTW algorithm’s mileage deviation is significantly improved,and the correlation coefficient reaches 0.98.Finally,on the basis of correcting the mileage deviation between the historical data of the irregularity,according to the demand of the railway engineering department for the model prediction data and the results of the existing research on the characteristics of the track degradation trend,the Holt-Winters model is established to realize the sampling points of the track.Forecasts are made for various irregularities.\The test results show that the prediction results of the model are very close to the measured values and meet the actual engineering application.Compared with the prediction results of historical data corrected without mileage deviation,the prediction accuracy of each irregularity index is increased by about 50%.