Axle box acceleration adaptive feature extraction-based corrugation identification method and device

Abstract

The invention discloses a corrugation identification method and device based on axle box acceleration adaptive feature extraction. The method comprises the following steps: acquiring axle box vibration acceleration detection data; for axle box vibration acceleration detection data, adopting a kernel function to extract time-frequency characteristics in a fuzzy domain; the kernel function is used for removing a mutual component of the axle box vibration acceleration detection data and retaining a self component of the axle box vibration acceleration detection data; and identifying the rail corrugation section according to the time-frequency characteristics. The accuracy of rail corrugation detection can be improved, meanwhile, the automation level of rail corrugation recognition is improved, and the flow of rail corrugation recognition is simplified.

Description

Technical Field

[0001] The invention belongs to the technical field of railway engineering infrastructure detection, and specifically relates to a corrugation identification method and device based on axle box acceleration adaptive feature extraction. Background Art

[0002] The comprehensive detection train axle box vibration acceleration detection system usually constructs a corrugation index to diagnose rail corrugation by calculating the time domain statistical indicators of the axle box acceleration, such as the mean, maximum value, and moving effective value; in addition, traditional time-frequency analysis methods such as short-time Fourier transform and wavelet transform are used to diagnose rail corrugation. Although rail corrugation can be diagnosed based on axle box vibration acceleration, the following problems still need to be solved: Due to the cross-term interference problem in the traditional time-frequency feature extraction method, the accuracy of the extracted time-frequency features is not high and the time-frequency resolution is low. In addition, in the process of rail corrugation identification based on the traditional time-frequency feature extraction method, there is still the problem of cumbersome procedures. Summary of the invention

[0003] The embodiment of the present invention provides a corrugation identification method based on axle box acceleration adaptive feature extraction, which is used to improve the accuracy of rail corrugation detection, improve the automation level of rail corrugation identification, and simplify the process of rail corrugation identification. The method includes:

[0004] Obtain axle box vibration acceleration detection data;

[0005] For the axle box vibration acceleration detection data, a kernel function is used in a fuzzy domain to extract time-frequency features; the kernel function is used to remove the mutual components of the axle box vibration acceleration detection data and retain the self-components of the axle box vibration acceleration detection data;

[0006] According to the time-frequency characteristics, the rail corrugation section is identified.

[0007] The embodiment of the present invention further provides a corrugation identification device based on axle box acceleration adaptive feature extraction, which is used to improve the accuracy of rail corrugation detection, while improving the automation level of rail corrugation identification and simplifying the process of rail corrugation identification. The device includes:

[0008] A data acquisition module is used to obtain the axle box vibration acceleration detection data;

[0009] A time-frequency feature extraction module is used to extract time-frequency features of the axle box vibration acceleration detection data using a kernel function in a fuzzy domain; the kernel function is used to remove the mutual components of the axle box vibration acceleration detection data and retain the self-components of the axle box vibration acceleration detection data;

[0010] The rail corrugation section identification module is used to identify the rail corrugation section according to the time-frequency characteristics.

[0011] An embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned corrugation identification method based on adaptive feature extraction of axle box acceleration when executing the computer program.

[0012] An embodiment of the present invention further provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the corrugation identification method based on adaptive feature extraction of axle box acceleration is implemented.

[0013] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the corrugation identification method based on adaptive feature extraction of axle box acceleration is implemented.

[0014] In the embodiment of the present invention, the axle box vibration acceleration detection data is obtained; the time-frequency characteristics of the axle box vibration acceleration detection data are extracted by using a kernel function in a fuzzy domain, and the kernel function is used to remove the mutual components of the axle box vibration acceleration detection data and retain the self-components of the axle box vibration acceleration detection data; and the rail corrugation section is identified according to the time-frequency characteristics. Compared with the technical solution in the prior art that uses traditional time-frequency feature extraction technologies such as short-time Fourier transform and wavelet transform to extract time-frequency characteristics of axle box vibration acceleration, the kernel function is used in the fuzzy domain to extract time-frequency characteristics. Since the kernel function removes the mutual components of the axle box vibration acceleration detection data and retains the self-components of the axle box vibration acceleration detection data, it has a better effect in depicting the details of non-stationary signals. The extracted time-frequency characteristics can accurately reflect the vibration acceleration response of the axle box of the comprehensive detection train caused by the rail corrugation disease as an input excitation, which can improve the accuracy of rail corrugation detection. At the same time, the method in the embodiment of the present invention improves the automation level of rail corrugation identification and simplifies the process of rail corrugation identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0016] Figure 1 It is a processing flow chart of a corrugation identification method based on axle box acceleration adaptive feature extraction in an embodiment of the present invention;

[0017] Figure 2It is a time domain waveform diagram of the axle box vibration acceleration detection data before mileage correction in an embodiment of the present invention;

[0018] Figure 3 It is a time domain waveform diagram of the axle box vibration acceleration detection data after mileage correction in an embodiment of the present invention;

[0019] Figure 4 It is a frequency domain waveform diagram of the axle box vibration acceleration detection data in an embodiment of the present invention;

[0020] Figure 5 This is a waveform diagram of the axle box vibration acceleration after filtering using a 3rd order Butterworth filter in an embodiment of the present invention;

[0021] Figure 6 It is a time-frequency feature diagram of the axle box vibration acceleration detection data extracted by using a kernel function in the fuzzy domain in an embodiment of the present invention;

[0022] Figure 7 This is an example diagram of the results of identifying rail corrugation with similar section lengths but different wavelengths in an embodiment of the present invention;

[0023] Figure 8 This is an example diagram of the results of identifying rail corrugation with similar wavelengths but different segment lengths in an embodiment of the present invention;

[0024] Fig. 9 It is a flow chart of the overall technical solution for the specific implementation of the corrugation identification method based on the adaptive feature extraction of the axle box acceleration in the embodiment of the present invention;

[0025] Fig.10 Schematic diagram of a corrugation identification device based on adaptive feature extraction of axle box acceleration in an embodiment of the present invention. DETAILED DESCRIPTION

[0026] To make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. Here, the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0027] In order to improve the accuracy of rail corrugation detection, an embodiment of the present invention provides a corrugation identification method based on adaptive feature extraction of axle box acceleration. Figure 1 FIG. 4 is a processing flow chart of a corrugation identification method based on axle box acceleration adaptive feature extraction in an embodiment of the present invention. Figure 1 As shown, the method may include:

[0028] Step 101, obtaining axle box vibration acceleration detection data;

[0029] Step 102: extracting time-frequency features of the axle box vibration acceleration detection data using a kernel function in a fuzzy domain; the kernel function is used to remove the mutual components of the axle box vibration acceleration detection data and retain the self-components of the axle box vibration acceleration detection data;

[0030] Step 103: Identify the rail corrugation section according to the time-frequency characteristics.

[0031] Depend on Figure 1 It can be seen from the process shown that the corrugation identification method based on the adaptive feature extraction of axle box acceleration in the embodiment of the present invention first obtains the axle box vibration acceleration detection data, and then extracts the time-frequency features of the axle box vibration acceleration detection data using a kernel function in a fuzzy domain. The kernel function is used to remove the mutual components of the axle box vibration acceleration detection data and retain the self-components of the axle box vibration acceleration detection data. Finally, the rail corrugation section is identified according to the time-frequency features. Compared with the technical solution of extracting the time-frequency features of the axle box vibration acceleration using traditional time-frequency feature extraction technologies such as short-time Fourier transform and wavelet transform in the prior art, the kernel function is used to extract the time-frequency features in the fuzzy domain. Since the kernel function removes the mutual components of the axle box vibration acceleration detection data and retains the self-components of the axle box vibration acceleration detection data, it has a better effect in depicting the details of non-stationary signals. The extracted time-frequency features can accurately reflect the vibration acceleration response of the axle box of the comprehensive detection train caused by the rail corrugation disease as an input excitation, which can improve the accuracy of rail corrugation detection. At the same time, the method in the embodiment of the present invention improves the automation level of rail corrugation identification and simplifies the process of rail corrugation identification.

[0032] It should be noted that, since the kernel function used in the fuzzy domain in the present invention has the characteristic of being able to automatically adjust according to the local characteristics of the signal as time changes, for the convenience of description, in the following embodiments, a kernel function that effectively removes mutual components and retains self-components is used in the fuzzy domain to extract time-frequency features, which is called adaptive optimal kernel technology.

[0033] During specific implementation, the axle box vibration acceleration detection data is first obtained.

[0034] In the embodiment of the present invention, it is considered that the axle box vibration acceleration detection data collected from the train at different times may have a certain degree of error in spatial mileage. Therefore, the axle box vibration acceleration detection data can be preprocessed to improve the accuracy of the recognition result. The data preprocessing can include, for example, mileage correction processing.

[0035] Generally, a comprehensive inspection train is used to collect axle box vibration acceleration inspection data, and track geometry inspection data can be collected at this time. The track geometry inspection data in multiple inspection processes are basically consistent, so the mileage of the axle box vibration acceleration inspection data and the track geometry inspection data of each inspection can be aligned according to the speed curve, thereby realizing the mileage alignment of the axle box vibration acceleration inspection data collected multiple times.

[0036] In one embodiment, the axle box vibration acceleration detection data is corrected for mileage before the kernel function is used to extract the time-frequency features in the fuzzy domain, and the steps are as follows:

[0037] Obtain track geometry detection data;

[0038] From the axle box vibration acceleration detection data, a subsequence with the same data volume as the track geometry detection data is selected each time, and the correlation coefficient between the subsequence and the track geometry detection data is calculated until all the axle box vibration acceleration detection data are selected, and the correlation coefficients between multiple subsequences and the track geometry detection data are obtained;

[0039] The track mileage corresponding to the subsequence with the largest correlation coefficient is determined as the track mileage of the axle box vibration acceleration detection data, and the axle box vibration acceleration detection data after mileage correction is obtained.

[0040] For example, there is axle box vibration acceleration detection data in the target file. The speed curve is matched by the principle of maximizing the correlation coefficient. Assume that the speed channel y of the track geometry detection data in the reference file has m data points, and the speed channel x of the axle box vibration acceleration detection data in the target file has n data points. Usually n>m, n-m+1 subsequences x(·) with a data volume of m are taken from the speed channel x of the axle box vibration acceleration detection data. The correlation coefficients of each subsequence x(·) and the speed channel y of the track geometry detection data are calculated respectively. The method for calculating the correlation coefficient r(k) of the kth subsequence and the track geometry detection data is shown in formula (1):

[0041]

[0042] Where i is the number of the data point in the velocity channel sequence, is the mean value of the speed channel of the axle box vibration acceleration detection data, is the mean of the velocity channel of the track geometry detection data.

[0043] According to the above steps, n-m+1 correlation coefficients are obtained, and the track mileage corresponding to the subsequence with the largest correlation coefficient is selected as the track mileage of the axle box vibration acceleration detection data. Figure 2 is a time domain waveform diagram of the axle box vibration acceleration detection data before mileage correction in an embodiment of the present invention, Figure 3 This is a time domain waveform diagram of the axle box vibration acceleration detection data after mileage correction in an embodiment of the present invention. Figure 2 and Figure 3 By comparison, it can be seen that according to the above mileage correction steps, the mileage automatic alignment of multiple axle box vibration acceleration detection data can be achieved, thereby improving the efficiency of mileage correction of axle box vibration acceleration detection data.

[0044] In order to further improve the accuracy of the recognition results, data preprocessing may also include: obtaining the upper and lower cutoff frequencies of the filtering process through spectrum analysis, and using a filter to filter the axle box vibration acceleration detection data to reduce noise interference.

[0045] In one embodiment, for the axle box vibration acceleration detection data, before using the kernel function to extract the time-frequency features in the fuzzy domain, the following steps are performed:

[0046] Perform spectrum analysis on the axle box vibration acceleration detection data to obtain a frequency domain waveform diagram of the axle box vibration acceleration detection data;

[0047] Determine the upper cutoff frequency and the lower cutoff frequency of the bandpass filter according to the frequency domain waveform diagram;

[0048] Based on the upper cut-off frequency and the lower cut-off frequency, a band-pass filter is used to perform filtering processing on the axle box vibration acceleration detection data.

[0049] In one embodiment, performing spectrum analysis on the axle box vibration acceleration detection data may include: performing spectrum analysis on the axle box vibration acceleration detection data using a fast Fourier transform method.

[0050] In one embodiment, using a bandpass filter to filter the axle box vibration acceleration detection data may include: using a 3rd order Butterworth filter to filter the axle box vibration acceleration detection data.

[0051] For example, the fast Fourier transform method is used to perform spectrum analysis on the axle box vibration acceleration detection data x. The calculation method is shown in formula (2), and the frequency domain waveform diagram of the axle box vibration acceleration detection data is obtained. Figure 4 It is a frequency domain waveform diagram of the axle box vibration acceleration detection data in an embodiment of the present invention.

[0052]

[0053] Where t is the mileage position, f is the frequency, x(t) is the axle box vibration acceleration at the mileage position t, X(f) is the axle box vibration acceleration at the frequency f, and j is the imaginary unit.

[0054] like Figure 4As shown in the figure, the spectrum composition and bandwidth of the axle box vibration acceleration are observed from the frequency domain waveform diagram. The main frequency components of the signal and the frequency range of the noise are determined by analyzing the spectrum diagram, and the upper and lower cutoff frequencies f of the bandpass filter can be obtained. low and f cut .

[0055] After using fast Fourier transform to perform spectrum analysis, based on the upper and lower limits f low and f cut As the cutoff frequency of the bandpass filter, a third-order Butterworth filter is used to filter the axle box vibration acceleration detection data x to obtain the filtered axle box vibration acceleration detection data x fiter . Figure 5 : is a waveform diagram of the axle box vibration acceleration after filtering using a 3rd order Butterworth filter in the embodiment of the present invention. low =1Hz, f cut =200Hz. Because the third-order Butterworth filter has a flat frequency response characteristic, the frequency response remains flat in the passband and decays rapidly at the cutoff frequency. The transfer function H(s) of the standard third-order Butterworth filter is as follows:

[0056]

[0057] Wherein: s is a complex frequency domain variable, a0, a1, a2, a3 and b0, b1, b2, b3 are the coefficients of the third-order Butterworth filter.

[0058] In the embodiment of the present invention, considering that the traditional time-frequency feature extraction technology, such as Cohen-type time-frequency distribution, usually has cross-term interference and cannot automatically adjust the processing parameters or basis functions according to the characteristics of different data, it will affect the accuracy of identifying faults from the extracted features. To this end, the embodiment of the present invention uses the adaptive optimal kernel technology to extract the time-frequency features of the axle box vibration acceleration detection data, which can avoid the cross-term interference that occurs in the traditional time-frequency feature extraction process.

[0059] The essence of suppressing cross terms is to retain the autocomponent of the signal to the greatest extent, while suppressing the mutual component to the greatest extent. Since the positions of the autocomponent and the mutual component are different in the fuzzy domain of the signal, the autocomponent is distributed near the origin, and the mutual component is far away from the origin. Therefore, a kernel function that can effectively remove the mutual component and retain the autocomponent in the fuzzy domain is designed in the embodiment of the present invention, which is the adaptive optimal kernel technology. The use of the adaptive optimal kernel technology to extract the time-frequency features of the axle box vibration acceleration detection data not only makes the kernel function more compatible with the axle box vibration acceleration detection data, but also avoids the cross-term interference problem existing in the traditional time-frequency feature extraction technology, thereby improving the accuracy and time-frequency resolution of the extracted time-frequency features.

[0060] In an embodiment of the present invention, a kernel function is used to extract time-frequency features of the axle box vibration acceleration detection data in a fuzzy domain, and the kernel function is used to remove the mutual components of the axle box vibration acceleration detection data and retain the self-components of the axle box vibration acceleration detection data.

[0061] In one embodiment, a kernel function is used to extract time-frequency features of the axle box vibration acceleration detection data in a fuzzy domain, wherein the kernel function may be a radial Gaussian kernel function, which is a two-dimensional function of Gaussian type along any radial section.

[0062] In another embodiment, for the axle box vibration acceleration detection data, a short-time fuzzy function can be used to determine the fuzzy domain, and a kernel function can be used in the fuzzy domain to extract time-frequency features.

[0063] For example, in order to design a kernel function that can better match the signal, the kernel function to be sought is defined as a two-dimensional function that is Gaussian along any radial section, called the radial Gaussian kernel function. Its polar coordinate representation is shown in formula (4):

[0064]

[0065] Where: μ is the angle between the radial and horizontal directions; σ(μ) is the expansion function that controls the radial Gaussian kernel function in the radial angle direction; θ is the frequency resolution expansion parameter that controls the radial Gaussian kernel function in the frequency domain; τ is the time resolution expansion parameter that controls the radial Gaussian kernel function in the time domain.

[0066] It should be noted that a larger θ makes the kernel function's range in the frequency domain wider, making it more sensitive to changes in different frequencies, able to capture subtle frequency features, and having a higher frequency resolution; a smaller θ makes the kernel function's range in the frequency domain narrower, making it more focused near a specific frequency, and having a lower frequency resolution. A larger τ makes the kernel function's range in the time domain wider, thus having a better resolution in the time domain, and being able to capture changes in the signal at different time points; a smaller τ makes the kernel function more focused in time, sacrificing some time resolution and paying more attention to the local details of the signal.

[0067] When using the adaptive optimal kernel technology to extract the time-frequency characteristics of the axle box vibration acceleration, for a set of axle box vibration acceleration detection data x(t), the calculation of the short-time fuzzy function A(t; θ; τ) of x(t) in polar coordinate expression is defined, as shown in formula (5):

[0068]

[0069] Where: x(·) is the axle box vibration acceleration detection data; t is the mileage position of the axle box vibration acceleration detection data; u is the integral variable; ω(·) is a symmetric window function, where t is the center position of ω(·); x*(·) is the conjugate function of x(·); ω * (·) is the conjugate function of ω(·); j is the imaginary unit; A(t; θ; τ) is the short-time ambiguity function.

[0070] It should be noted that when |u|>T (window length), ω(u)=0, then at any mileage position t at a certain moment, the fuzzy function A(t; θ; τ) of the signal can only be calculated using formula (5) within the range of [tT, t+T] intercepted by the window function ω(u).

[0071] Using the radial Gaussian kernel function defined by formula (4) as the basis function and the fuzzy function A(t; θ; τ) of the axle box vibration acceleration detection data defined by formula (5), the adaptive optimal kernel time-frequency feature extraction method is obtained. The specific calculation method is shown in formula (6):

[0072]

[0073] Among them, the time-frequency characteristics of the axle box vibration acceleration detection data obtained by the adaptive optimal kernel time-frequency feature extraction method are TF AOK (t, f). Figure 6 It is a time-frequency characteristic diagram of the axle box vibration acceleration detection data extracted based on the adaptive optimal kernel technology in an embodiment of the present invention.

[0074] After the above data preprocessing and feature extraction process, the time-frequency characteristics of the axle box vibration acceleration without noise can be obtained. The results are as follows: Figure 6 shown.

[0075] Since the kernel function of the adaptive optimal kernel technology can automatically adjust the local characteristics of the signal as time changes, it is better in depicting the details of non-stationary signals. The extracted time-frequency characteristics can accurately reflect the vibration acceleration response of the train axle box caused by the rail corrugation disease as input excitation. After obtaining the time-frequency energy distribution of the axle box vibration acceleration detection data, the change characteristics of the energy peak at the same frequency and mileage position are observed to identify the energy concentrated area as the abnormal section of rail corrugation.

[0076] In the embodiment of the present invention, the rail corrugation section is identified according to the time-frequency characteristics.

[0077] In one embodiment, identifying the rail corrugation section according to the time-frequency characteristics may specifically include the following steps:

[0078] Extract the characteristic frequency f corresponding to each energy peak position of the axle box vibration acceleration time-frequency characteristicsi , Mileage position t i , Comprehensive test train running speed v i ; The number of energy peaks of the time-frequency characteristics of the axle box vibration acceleration is N, i = 1, 2, ... N;

[0079] According to the characteristic frequency f corresponding to each energy peak position of the axle box vibration acceleration time-frequency characteristics i And comprehensive detection of train running speed v i , calculate the wavelength λ of the rail corrugation disease corresponding to each energy peak position of the axle box vibration acceleration time-frequency characteristics i ;

[0080] The characteristic frequency f corresponding to the energy peak i According to mileage location i Arrange in ascending order to form a four-dimensional array of length N (t i , f i ,λ i , v i );

[0081] If two adjacent mileage positions t i and t i+1 The distance between them is less than the set first threshold, and the corresponding characteristic frequency f i and f i+1 The absolute value of the difference between the two adjacent mileage positions is less than the set second threshold, then the two adjacent mileage positions t i and t i+1 Merge and identify them as the same corrugation section.

[0082] For example, the number of energy peaks of the time-frequency characteristics of the axle box vibration acceleration is N, i = 1, 2, ... N. First, the characteristic frequency f corresponding to each energy peak position is extracted. i , Mileage position t i And the corresponding comprehensive detection train speed v i , usually the unit of comprehensive detection train running speed is km / h; then combined with the comprehensive detection train running speed v i , the time-frequency characteristics of the axle box vibration acceleration extracted using the adaptive optimal kernel technology f i Calculate the wavelength λ of the rail corrugation disease corresponding to each energy peak position according to formula (7): i :

[0083]

[0084] Where: The calculated wavelength is in meters, which can also be converted into millimeters according to the application scenario.

[0085] The characteristic frequency f corresponding to the energy peak obtained by the above steps i According to mileage location i Arrange in ascending order to form a four-dimensional array of length N (t i , f i ,λ i , v i ), i = 1, 2, ... N, the corrugation section is identified according to the following method: if two adjacent mileage positions t i and t i+1 The distance between them is less than the first threshold, which is usually 5 meters in applications, and the two adjacent mileage positions t i and t i+1 The corresponding characteristic frequency f i and f i+1 If the absolute value of the difference between the two is smaller than the set second threshold, which is usually 2 Hz in applications, the two are combined and identified as the same corrugation section.

[0086] Finally, the rail corrugation section identification result chart is exported. Figure 7 1 is an example diagram of the results of identifying rail corrugation of rails with similar section lengths but different wavelengths in an embodiment of the present invention. Figure 8 This is an example diagram of the results of identifying rail corrugation of similar wavelengths but different segment lengths in an embodiment of the present invention.

[0087] Since the railway site pays more attention to the longer sections of rail corrugation during maintenance, and in order to avoid false alarms caused by interference from other short-wave rail diseases, the threshold is set to 10 meters when exporting the rail corrugation section identification results, that is, corrugation sections below 10 meters are not exported. The corrugation identification method based on adaptive feature extraction of axle box acceleration in the embodiment of the present invention has been verified for its effectiveness using field test data. During the verification process, the axle box vibration acceleration responses caused by on-site rail corrugation diseases of different wavelengths and different section lengths were compared. The rail corrugation section identification results are as follows: Figure 7 and Figure 8 As shown, the method in the embodiment of the present invention can provide a reference for rail corrugation identification and maintenance.

[0088] Fig. 9 The flowchart of the overall technical scheme for the specific implementation of the corrugation identification method based on the adaptive feature extraction of axle box acceleration in the embodiment of the present invention is as follows. First, the axle box vibration acceleration detection data can be preprocessed, and the preprocessing includes: mileage correction, frequency domain analysis, and filtering processing. Then, the adaptive optimal kernel technology is used to extract the time-frequency features. Finally, the rail corrugation section is identified according to the time-frequency features. The steps include: calculating the wavelength, identifying the section, and outputting the results.

[0089] In summary: the wave wear identification method based on axle box acceleration adaptive feature extraction proposed in the embodiment of the present invention obtains axle box vibration acceleration detection data; a kernel function is used to extract the time-frequency characteristics of the axle box vibration acceleration detection data in the fuzzy domain, and the kernel function is used to remove the mutual components of the axle box vibration acceleration detection data and retain the self-components of the axle box vibration acceleration detection data; according to the time-frequency characteristics, the rail wave wear section is identified, which can improve the accuracy of rail wave wear detection. At the same time, the method in the embodiment of the present invention improves the automation level of rail wave wear identification and simplifies the process of rail wave wear identification.

[0090] The present invention also provides a corrugation identification device based on axle box acceleration adaptive feature extraction, as described in the following embodiments. Since the principle of the device to solve the problem is similar to that of the corrugation identification method based on axle box acceleration adaptive feature extraction, the implementation of the device can refer to the implementation of the corrugation identification method based on axle box acceleration adaptive feature extraction, and the repeated parts will not be repeated.

[0091] Fig.10 FIG. 1 is a schematic diagram of a corrugation identification device based on axle box acceleration adaptive feature extraction in an embodiment of the present invention. Fig.10 As shown, the device comprises:

[0092] The data acquisition module 1001 is used to obtain the axle box vibration acceleration detection data and the track geometry detection data;

[0093] The time-frequency feature extraction module 1002 is used to extract the time-frequency features of the axle box vibration acceleration detection data by using a kernel function in a fuzzy domain; the kernel function is used to remove the mutual components of the axle box vibration acceleration detection data and retain the self-components of the axle box vibration acceleration detection data;

[0094] The rail corrugation section identification module 1003 is used to identify the rail corrugation section according to the time-frequency characteristics.

[0095] In one embodiment, the time-frequency feature extraction module 1002 is specifically used to:

[0096] For the axle box vibration acceleration detection data, a radial Gaussian kernel function is used in the fuzzy domain to extract time-frequency features. The radial Gaussian kernel function is a two-dimensional function of Gaussian type along any radial section.

[0097] In one embodiment, the time-frequency feature extraction module 1002 is specifically used to:

[0098] For the axle box vibration acceleration detection data, the short-time fuzzy function is used to determine the fuzzy domain, and the kernel function is used to extract the time-frequency features in the fuzzy domain.

[0099] In one embodiment, the data acquisition module 1001 is further used for:

[0100] While acquiring the axle box vibration acceleration detection data, the track geometry detection data is acquired;

[0101] From the axle box vibration acceleration detection data, a subsequence having the same data volume as the track geometry detection data is selected each time, and the correlation coefficient between the subsequence and the track geometry detection data is calculated until all the axle box vibration acceleration detection data are selected, thereby obtaining the correlation coefficients between the multiple subsequences and the track geometry detection data;

[0102] The track mileage corresponding to the subsequence with the largest correlation coefficient is determined as the track mileage of the axle box vibration acceleration detection data, and the axle box vibration acceleration detection data after mileage correction is obtained.

[0103] In one embodiment, the data acquisition module 1001 is further used for:

[0104] Perform spectrum analysis on the axle box vibration acceleration detection data to obtain a frequency domain waveform diagram of the axle box vibration acceleration detection data;

[0105] Determine the upper cutoff frequency and the lower cutoff frequency of the bandpass filter according to the frequency domain waveform diagram;

[0106] Based on the upper cut-off frequency and the lower cut-off frequency, a band-pass filter is used to perform filtering processing on the axle box vibration acceleration detection data.

[0107] In one embodiment, the rail corrugation section identification module 1003 is specifically used to:

[0108] Extract the characteristic frequency f corresponding to each energy peak position of the time-frequency feature i , Mileage position t i , Comprehensive test train running speed v i ; The number of energy peaks of the time-frequency characteristics of the axle box vibration acceleration is N, i = 1, 2, ... N;

[0109] According to the characteristic frequency f corresponding to each energy peak position of the time-frequency characteristics i And comprehensive detection of train running speed v i , calculate the wavelength λ of the rail corrugation disease corresponding to each energy peak position of the time-frequency characteristics i ;

[0110] The characteristic frequency f corresponding to each energy peak position of the time-frequency feature i According to mileage location i Arrange in ascending order to form a four-dimensional array of length N (t i , f i ,λ i , v i );

[0111] If two adjacent mileage positions t i and t i+1 The distance between them is less than the set first threshold, and the corresponding characteristic frequency f i and f i+1 The absolute value of the difference between the two adjacent mileage positions ti and t is less than the set second threshold, then the two adjacent mileage positions ti and t i+1 Merge and identify them as the same corrugation section.

[0112] An embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned corrugation identification method based on adaptive feature extraction of axle box acceleration when executing the computer program.

[0113] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the corrugation identification method based on adaptive feature extraction of axle box acceleration is implemented.

[0114] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the corrugation identification method based on the adaptive feature extraction of axle box acceleration is implemented.

[0115] In summary, compared with the technical solution of extracting time-frequency features of axle box vibration acceleration using traditional time-frequency feature extraction technologies such as short-time Fourier transform and wavelet transform in the prior art, a kernel function is used in the fuzzy domain to extract time-frequency features. Since the kernel function removes the mutual components of the axle box vibration acceleration detection data and retains the self-components of the axle box vibration acceleration detection data, it is better in depicting the details of non-stationary signals. The extracted time-frequency features can accurately reflect the vibration acceleration response of the axle box of the comprehensive detection train caused by the rail corrugation disease as input excitation. The rail corrugation section is identified according to the time-frequency features, which can improve the accuracy of rail corrugation detection. At the same time, the method in the embodiment of the present invention improves the automation level of rail corrugation identification, simplifies the process of rail corrugation identification, and can be used in axle box vibration acceleration detection devices for various types of tracks. It has the advantages of simple installation, easy use, and high accuracy.

[0116] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0117] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0118] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0119] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0120] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A corrugation identification method based on axle box acceleration adaptive feature extraction, characterized in that: include: Obtain axle box vibration acceleration detection data; For the axle box vibration acceleration detection data, the kernel function is used in the fuzzy domain to extract the time-frequency features; The kernel function is used to remove the mutual component of the axle box vibration acceleration detection data and retain the self-component of the axle box vibration acceleration detection data; According to the time-frequency characteristics, the rail corrugation section is identified.

2. The method according to claim 1, characterized in that For the axle box vibration acceleration detection data, the kernel function is used in the fuzzy domain to extract the time-frequency features, including: For the axle box vibration acceleration detection data, a radial Gaussian kernel function is used in the fuzzy domain to extract time-frequency features. The radial Gaussian kernel function is a two-dimensional function of Gaussian type along any radial section.

3. The method according to claim 1, characterized in that For the axle box vibration acceleration detection data, the kernel function is used in the fuzzy domain to extract the time-frequency features, including: For the axle box vibration acceleration detection data, the short-time fuzzy function is used to determine the fuzzy domain, and the kernel function is used to extract the time-frequency features in the fuzzy domain.

4. The method according to claim 1, characterized in that For the axle box vibration acceleration detection data, before using the kernel function to extract the time-frequency features in the fuzzy domain, it also includes: Obtain track geometry detection data; From the axle box vibration acceleration detection data, a subsequence with the same data volume as the track geometry detection data is selected each time, and the correlation coefficient between the subsequence and the track geometry detection data is calculated until all the axle box vibration acceleration detection data are selected, and the correlation coefficients between multiple subsequences and the track geometry detection data are obtained; The track mileage corresponding to the subsequence with the largest correlation coefficient is determined as the track mileage of the axle box vibration acceleration detection data, and the axle box vibration acceleration detection data after mileage correction is obtained.

5. The method according to claim 1, characterized in that For the axle box vibration acceleration detection data, before using the kernel function to extract the time-frequency features in the fuzzy domain, it also includes: Perform spectrum analysis on the axle box vibration acceleration detection data to obtain a frequency domain waveform diagram of the axle box vibration acceleration detection data; Determine the upper cutoff frequency and the lower cutoff frequency of the bandpass filter according to the frequency domain waveform diagram; Based on the upper cut-off frequency and the lower cut-off frequency, a band-pass filter is used to perform filtering processing on the axle box vibration acceleration detection data.

6. The method according to claim 1, characterized in that According to the time-frequency characteristics, the rail corrugation section is identified, including: Extract the characteristic frequency f corresponding to each energy peak position of the time-frequency feature i , Mileage position t i , Comprehensive test train running speed v i ; The number of energy peaks of the time-frequency characteristics of the axle box vibration acceleration is N, i = 1, 2, ... N; According to the characteristic frequency f corresponding to each energy peak position of the time-frequency characteristics i And comprehensive detection of train running speed v i , calculate the wavelength λ of the rail corrugation disease corresponding to each energy peak position of the time-frequency characteristics i ; The characteristic frequency f corresponding to each energy peak position of the time-frequency feature i According to mileage location i Arrange in ascending order to form a four-dimensional array of length N (t i , f i ,λ i , v i ); If two adjacent mileage positions t i and t i+1 The distance between them is less than the set first threshold, and the corresponding characteristic frequency f i and f i+1 The absolute value of the difference between the two adjacent mileage positions is less than the set second threshold, then the two adjacent mileage positions t i and t i+1 Merge and identify them as the same corrugation section.

7. A corrugation identification device based on axle box acceleration adaptive feature extraction, characterized in that: include: A data acquisition module is used to obtain the axle box vibration acceleration detection data; A time-frequency feature extraction module is used to extract time-frequency features of the axle box vibration acceleration detection data using a kernel function in a fuzzy domain; the kernel function is used to remove the mutual components of the axle box vibration acceleration detection data and retain the self-components of the axle box vibration acceleration detection data; The rail corrugation section identification module is used to identify the rail corrugation section according to the time-frequency characteristics.

8. The device according to claim 7, characterized in that The time-frequency feature extraction module is specifically used for: For the axle box vibration acceleration detection data, a radial Gaussian kernel function is used in the fuzzy domain to extract time-frequency features. The radial Gaussian kernel function is a two-dimensional function of Gaussian type along any radial section.

9. The device according to claim 7, characterized in that The time-frequency feature extraction module is specifically used for: For the axle box vibration acceleration detection data, the short-time fuzzy function is used to determine the fuzzy domain, and the kernel function is used to extract the time-frequency features in the fuzzy domain.

10. The device according to claim 7, characterized in that The data acquisition module is also used to: Obtain track geometry detection data; From the axle box vibration acceleration detection data, a subsequence with the same data volume as the track geometry detection data is selected each time, and the correlation coefficient between the subsequence and the track geometry detection data is calculated until all the axle box vibration acceleration detection data are selected, and the correlation coefficients between multiple subsequences and the track geometry detection data are obtained; The track mileage corresponding to the subsequence with the largest correlation coefficient is determined as the track mileage of the axle box vibration acceleration detection data, and the axle box vibration acceleration detection data after mileage correction is obtained.

11. The device according to claim 7, characterized in that The data acquisition module is also used to: Perform spectrum analysis on the axle box vibration acceleration detection data to obtain a frequency domain waveform diagram of the axle box vibration acceleration detection data; Determine the upper cutoff frequency and the lower cutoff frequency of the bandpass filter according to the frequency domain waveform diagram; Based on the upper cut-off frequency and the lower cut-off frequency, a band-pass filter is used to perform filtering processing on the axle box vibration acceleration detection data.

12. The device according to claim 7, characterized in that Rail corrugation section identification module is specifically used for: Extract the characteristic frequency f corresponding to each energy peak position of the time-frequency feature i , Mileage position t i , Comprehensive test train running speed v i ; The number of energy peaks of the time-frequency characteristics of the axle box vibration acceleration is N, i = 1, 2, ... N; According to the characteristic frequency f corresponding to each energy peak position of the time-frequency characteristics i And comprehensive detection of train running speed v i , calculate the wavelength λ of the rail corrugation disease corresponding to each energy peak position of the axle box vibration acceleration time-frequency characteristics i ; The characteristic frequency f corresponding to each energy peak position of the time-frequency feature i According to mileage location i Arrange in ascending order to form a four-dimensional array of length N (t i , f i ,λ i , v i ); If two adjacent mileage positions t i and t i+1 The distance between them is less than the set first threshold, and the corresponding characteristic frequency f i and f i+1 The absolute value of the difference between the two adjacent mileage positions is less than the set second threshold, then the two adjacent mileage positions t i and t i+1 Merge and identify them as the same corrugation section.

13. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 6 is implemented.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

15. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images