Time-Frequency Feature Extraction And Dynamic Diagnosis Method For Track Short-Wave Defects

The state of the high-speed railway track directly determines whether the rail-wheel system can operate safely.Short-wave defects,such as high/low joints,wheel burn and rail corrugation,are likely to cause high-frequency excitation of the rail-wheel system,decreasing the travelling comfort and the safety running of the train.In order to alleviate the abnormal vibration,noise,bogie and line component failure problems caused by track short-wave defects,the track maintenance division have to increase the investment in track maintenance,but the large-scale maintenance has brought a lot of work pressure to the division,as well as the economic costs.In order to prevent the influence to the rail-wheel system of high-frequency excitation caused by the short-wave defects,reduce the work pressure of manual inspection,and rationally arrange the planning of the maintenance work,the study of the dynamic diagnosis method of the track short-wave problem is particularly important.The research content of this thesis is the time-frequency feature extraction and dynamic diagnosis method of track short-wave defects based on vehicle dynamic response data,which including: time-frequency analysis method and instantaneous frequency extraction method of non-stationary signal,diagnosis method of impact track short-wave defects,diagnostic method of periodically track short-wave defects,and detection method of out-of-roundness of wheel.In order to analyze and extract the time-frequency characteristics of the vehicle dynamic response data caused by the impact short-wave defects of railway track,a Wigner Ville Decomposition(WVD)analysis method based on Ensemble Empirical Mode Decomposition(EEMD)is proposed.Firstly,we decompose the dynamic response signal by EEMD,and then use the WVD to calculate the time-frequency distribution of each Intrinsic Mode Function(IMF)got by EEMD.The distribution of the final signal is obtained by the addition of time-frequency distribution of each component.This method,combined both the advantages of EEMD decomposition and WVD method,can well describe the time-frequency characteristics of impact short-wave defects of railway track,and provide a foundation for the evaluation of track impact short-wave irregularity.The data of rail impact defects are analyzed,and the frequencies of the vehicle dynamic response vibration data caused by high/low joints,rail scratches,falling blocks and other impact track short-wave defects are generally below 500 Hz.At the same time,the difference of vehicle dynamic response frequency distribution of welded joints with different heights is studied.When there are no short-wave defects in the welded joint area,the vehicle runs more smoothly.In the frequency domain,the vehicle dynamic response signals are mainly distributed in a wide frequency range.When the joint begins to appear short-wave defects,the vibration amplitude of the vehicle dynamic response signal increases and begins to concentrate on a band with limited bandwidth;when the joint defects develops further,the composition of the vibration signal becomes more concentrated and fixed near a frequency.Finally,after a long time of cumulative wear of millions of times,a wave is formed behind the joint.This is the very process of contact filtering.In order to diagnose the impact short-wave defects such as high/low joints,wheel burn and spalling of rail head,according to the frequency response characteristics of vehicle dynamic response data,the range of bandpass filtering is given.And the track impact index,a method for diagnosing short-wave defects on rail surfaces,using vehicle dynamic response data is proposed.This method demodulating the high-frequency vibration data into a stable low-frequency signal can avoid the randomness of using the amplitude of the vehicle dynamic response data to evaluate the track short-wave impact defects,and make the threshold easy to determine.After research and calculation,it is found that under the condition that the roughness of the surface of the rail joint is relatively smooth,the height of the joint and the magnitude of the track impact index satisfy a certain linear positive correlation.In order to study the dynamic response characteristics of the vehicle,obtain the time-frequency distribution with high time-frequency resolution and extract the instantaneous frequency curve,when the train passed by the rail corrugation section,the synchro-squeezed wavelet transform and the synchro-squeezed STFT transform are studied and improved.A window adaptive method has been introduced in this method and relatively continuous instantaneous frequency curve can be obtained based on the optimized and improved method,which is also made it convenient to observe the characteristics of the vehicle dynamic response signal.It is found that in the rail corrugation section,the dynamic response signal of the vehicle is relatively simple and the signal is periodically in the time domain.In the section where the grinding trace exists,the vehicle dynamic response signal exhibits obvious nonlinear harmonic resonance characteristics,and its response frequency is mainly expressed as the multiplier or submultiple of the wave frequency of the section in which it is located.In order to diagnose periodic short-wave defects such as rail corrugation or grinding trace,a diagnostic method for characterizing short-wave defects on rail surfaces using rail corrugation index and energy factor is proposed.The pass band,at the calculation of corrugation index,is determined according to the relationship of frequency,speed and wavelength.The window length is determined according to the influence distance of the corrugation,and the corrugation index is obtained according to the RMS of the filtered component at the calculated segment.Since the rail corrugation exhibits pseudo-periodicity,it is proposed to use the energy factor to characterize the vehicle dynamic response data,which concentrated near the frequency corresponding to the corrugation wavelength.The energy factor is calculated from the ratio of energy in the 3 Hz range near the main frequency to the total energy in the section which the train has a constant velocity,and 3 Hz range from the main frequency ridge to the total energy in the acceleration or deceleration section.The larger the energy factor,the stronger the periodic short-wave irregularity,that is,the more obvious the rail corrugation feature.The wavelength rail corrugation is determined if the speed and frequency determined,that means,the corrugation wavelength is the ratio of the speed to the main frequency in the section which the train has a constant velocity and 1-norm of a result,which is a dot division of the speed curve and the first frequency ridge line,divided by n,in the acceleration or deceleration section.Combined with the vehicle dynamic response data such as axle box acceleration,wheel-rail force and wheel-rail noise data,a method of rail corrugation diagnosis method based on multisource data is proposed.A comparison matrix is constructed by analytic hierarchy process,and the weight of axle box acceleration,wheel-rail force and wheel-rail noise data are obtained,which is used for the calculation of corrugation index of vehicle dynamic response data.And finally,a comprehensive evaluation method of the rail corrugation using multisource data is obtained.Compared with a single source data,this method could catch the short-wave defects at early period.When the train runs at high speed,the out-of-roundness of wheel will cause high-frequency wheel-rail impact,which will influence the fatigue life of the wheel-rail system components such as axles,wheels and rails,and will also interfere with the inspection data.In order to timely discover the influence of out-of-roundness of wheel and reduce the ineffectiveness of the detection data,the generalized resonance demodulation method is proposed to realize the out-of-roundness diagnosis of the wheel based on the dynamic response of the vehicle.Firstly,the high-resolution time-frequency spectrum of the signal is obtained by synchro-squeezed short-time Fourier transform,and then the time-frequency information,which could use for the separation of the detail and approximate part of the vibration signal caused by the impact,is automatically extracted from the time-frequency spectrum;After the pass band is determined,by the norm criterion,the generalized Hilbert envelope demodulation method is used to demodulate the dynamic response signal,and then the information of out-of-roundness of wheel is extracted.The results show that this method can effectively diagnose the out-of-roundness of wheel.