Study Of Track Dynamic-Detected Data De-noising Method Based On Wavelets And Application

At present, track inspection train are applied in detecting the geometry condition of the line tracks dynamically. After the processing and analyzing of the test data, the smoothness of the track can be evaluated, and then the plans of maintenance and repairmen are made to ensure the smoothness of tracks. However, as the electricity signals are used for detecting signals when track inspection train doing the detecting work, it may lead to the occurrence of noise in detecting signals due to these factors like the unstable voltage, the electromagnetic interference, inadequate shielding for instruments and so on. It is obviously that the justifiability and the accurateness of the evaluation about the smoothness of tracks will be impacted. Thus the data denoising should be done to improve the quality of test data obtained from track inspection vehicles before evaluating of the smoothness condition.Wavelet analysis is the newest effective tool to remove data noise nowadays. The original track inspection data can be multi-scale decomposed by wavelet transform, then valid information in low frequency can be extracted and the interference information in high frequency can be abandoned, and the reconstruction is made without distortion, so that the reliability of the analysis about the dynamic smoothness condition of tracks is improved. In this thesis, the main goal is to apply the wavelet analysis with its de-noise function to the de-noising process for track inspection data. And the research is focusing on how to use the wavelet analysis method to weaken and eliminate the noise of the raw track inspection measurement data, so it can provide guarantees for the accuracy of the subsequent data analysis and research of the detecting data, resulting in purifying the original track inspection data. The whole content of this research can be concluded as following three parts:(1) The essence and methods of data de-noising based on wavelet analysis theory;(2) A few wavelet de-noising methods applied into the track inspecting data has been put forward, including optimal selection of wavelet basis function, the choice and innovation of threshold and threshold function and so on;(3) The evaluation methods of smoothness condition of tracks in our country have been summarized, and examples are made to illustrate the influence of evaluating results with de-noising and without data de-noising.To the content above, the research based on MATLAB wavelet toolbox is illustrated at first. With the understanding about the essence of wavelet de-noising, the characteristics of track inspection data are analyzed and by combining the de-noising effect of real examples, the wavelet de-noising method which is suitable for track inspection data, the optimum resolve and reconstruction scales, the threshold and threshold function are confirmed. Secondly, on the basis of concluding the evaluation methods about dynamic track smoothness, the software about evaluating and analyzing the dynamic smoothness of tracks is developed. By contrasting the peak value management parameters and average value management parameters before and after de-nosing which is calculated by the real measured data, it reveals the effect of the nosise on the evaluating about dynamic smoothness of tracks, and indicate that it is essential to de-noise the original track inspection data, and proving that applying the wavelet analysis theory into the de-noising of track inspection data is reasonable and proper.