A wheel non-circularization anti-interference detection method based on vehicle-mounted vibration

Abstract

The application discloses a wheel non-circularization anti-interference detection method based on vehicle-mounted vibration, which comprises the following steps: obtaining an axle box vibration acceleration signal and a wheel set rotating speed signal; performing equal-angle resampling on the axle box vibration acceleration signal according to the wheel set rotating speed signal to obtain sampling data; calculating a weighted moving average value according to a weight vector of the sampling data; performing frequency domain analysis processing on the weighted moving average value to obtain a spectrum or a time-frequency spectrum and judging whether a peak value exceeds a set value; if yes, there is a periodic wheel non-circularization defect; otherwise, the next step is entered; judging whether a peak value of an envelope spectrum or an envelope time-frequency spectrum exceeds a set value; if yes, there is a local wheel non-circularization defect; otherwise, the wheel state is normal. The application solves the low-order resolution problem caused by the period number loss of a traditional time synchronous average method, reduces the influence of abnormal impact and avoids the amplitude loss caused by the accumulated phase error.

Description

Technical Field

[0001] This invention relates to the field of out-of-roundness detection technology, specifically including a method for detecting wheel out-of-roundness based on vehicle vibration. Background Technology

[0002] Wheel non-circularity is a problem of non-uniform wear and damage along the longitudinal direction of the wheel circumference. It includes: periodic wear around the wheel circumference, i.e., wheel polygons; and localized non-circularity problems such as wheel flats and scratches. Severe wheel non-circularity can lead to severe vibration, noise, and structural safety issues. Furthermore, once the initial polygon forms, its development is relatively rapid. Therefore, wheel non-circularity is a type of failure that railway operators are highly concerned about.

[0003] To address this, various methods have been developed for inspection, primarily categorized into trackside and vehicle-mounted inspection technologies. Vehicle-mounted technology, with its advantages of direct path, high signal-to-noise ratio, and long-term tracking capability, has been widely adopted in bogie online monitoring systems. However, current vehicle-mounted inspection mainly focuses on threshold control for mid-to-late stage vibrations. Considering that mid-to-late stage wheel non-circularity vibrations and noise may have already caused damage, online monitoring of wheel non-circularity in the initial stage is of great significance.

[0004] In recent years, many scholars have studied various vibration-based diagnostic methods for non-circular wheel wear. Regarding signal processing methods, Empirical Mode Decomposition (EMD) and its improved versions have been widely used to account for the modulation nonlinearity of the wheel response, including EMD, ACMD, VMD, and HHT. For the quantitative identification of non-circular wheel wear, the inertial integration method has been used for quantitative diagnosis, but it has not considered structural resonance, which has a significant impact on quantitative diagnosis.

[0005] As the direct target of wheel-rail excitation, the method of diagnosing wheel wear based on axle box vibration acceleration is easily affected by factors such as short-wave irregularities in the rail, wheel speed fluctuations, and structural resonance. Rail corrugation, in particular—a type of harmonic excitation on the rail—has vibration characteristics very similar to polygonal wear on the wheel, making misdiagnosis based on wheel polygonal wear very easy. For the diagnosis of early-stage non-circular wheel wear, the influence of these interfering factors is even more challenging, making it difficult to accurately detect early-stage non-circular wheel wear using axle box vibration acceleration. Summary of the Invention

[0006] To address the aforementioned shortcomings in existing technologies, this invention provides a method for detecting wheel non-circularity based on vehicle vibration. This method solves the problems of weak early wheel non-circularity characteristics, which are easily submerged by random noise caused by rail roughness and structural resonance; interference frequencies generated by equally spaced impacts on the track or other rotating equipment are easily mistaken for wheel non-circularity; and fluctuations in rotational speed can cause nonlinear modulation of the wheel non-circularity response, leading to spectral leakage and ambiguity in the spectral characteristics.

[0007] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows: a method for detecting non-circular wheel interference based on vehicle vibration, comprising the following steps:

[0008] S1. Obtain the axle box vibration acceleration signal x(t) n ) and wheelset speed signal

[0009] S2, based on wheelset speed signal For the axle box vibration acceleration signal x(t) n Perform isotropic resampling to obtain the sampled data;

[0010] S3. Calculate the weight vector of the sampled data;

[0011] S4. Calculate the weighted moving average based on the weight vector;

[0012] S5. Perform frequency domain analysis on the weighted moving average to obtain the spectrum or time spectrum containing the order and amplitude characteristics of the wheel's periodic non-circularity, as well as the envelope spectrum or envelope time spectrum containing the wheel's local non-circularity wear information.

[0013] S6. Determine whether the peak value of the spectrum containing the order and amplitude characteristics of the wheel's periodic non-circularity exceeds the set value; if so, there is a periodic wheel non-circularity defect, i.e., a wheel polygonal defect; otherwise, proceed to step S7.

[0014] S7. Determine whether the peak value of the envelope spectrum or the envelope time spectrum containing information on local non-circular wear of the wheel exceeds the set value; if so, there is a local non-circular defect of the wheel, i.e., flat scars or scratches; otherwise, the wheel condition is normal.

[0015] Furthermore, the specific method of step S2 is as follows:

[0016] According to the formula:

[0017]

[0018] n=0,1,…,N-1,l=0,1,…,L-1

[0019] θ l =lφ w(t N-1 ) / (L-1)

[0020] Obtain the axle box vibration acceleration x(θ) after equiangular resampling l ); where θ l The uniformly discrete sample represents the angle variable; L is the total number of angle samples; x(t) n ) represents the axle box vibration acceleration signal; N represents the total number of equal-time sampling points; t n The sampling time interval is φ; l is the l-th angle sample; Interpolate(·) is the interpolation function; φ w (t N-1 ) is the phase function.

[0021] Furthermore, the specific implementation of step S3 is as follows:

[0022] S3-1, According to the formula:

[0023] b ij =|x(θ+2πi)-x(θ+2πk)|

[0024] Obtain the element b in the i-th row and j-th column of the deviation matrix B. ij Where x(θ+2πi) is the i-th data point within the moving window that is resampled at equal angles within a sampling interval, and x(θ+2πi) = x(θ l x(θ+2πk) is the k-th data point within the same window as x(θ+2πi), where i≠k; θ represents the coordinates of the isoangular resampled data point; π represents pi.

[0025] S3-2, According to the formula:

[0026]

[0027] Obtain the element w in the c-th row and d-th column of the weight matrix W. cd ;

[0028] S3-3. Obtain the corresponding weight vector p(m) based on the weight matrix W.

[0029] Furthermore, the specific implementation of step S4 is as follows:

[0030] According to the formula:

[0031]

[0032] Obtain the weighted moving average of the sampled points within a sampling interval. Where Q is the number of revolutions of the wheel, and M is... wx is the number of revolutions of the moving window during equal-angle resampling within a sampling interval; x(θ+2πm) is the m-th data point within the moving window during equal-angle resampling within a sampling interval.

[0033] Furthermore, the specific implementation of step S5 is as follows:

[0034] S5-1. According to the formula:

[0035]

[0036] Obtain the spectral FFT(f) containing the order and amplitude characteristics of the wheel's periodic non-circularity; or

[0037] According to the formula:

[0038]

[0039] The time-frequency spectrum STFT(θ,f) containing the order and amplitude characteristics of the wheel's periodic non-circularity is obtained; where e is the natural constant. This is the weighted moving average of the data within the corner domain; f represents the frequency; j represents the imaginary unit.

[0040] S5-2, According to the formula:

[0041]

[0042] Obtain the analytical signal Where H[·] denotes the Hilbert transform;

[0043] S5-3, According to the formula:

[0044]

[0045] Obtain the envelope

[0046] S5-4. According to the formula:

[0047]

[0048] The envelope spectrum FFT(f) containing information on local non-circular wear of the wheel is obtained;

[0049] S5-5, According to the formula:

[0050]

[0051] The envelope time spectrum STFT(θ,f) containing information on local non-circular wear of the wheel is obtained.

[0052] The beneficial effects of this invention are as follows: the introduction of an improved moving average method solves the problem of low order resolution caused by the loss of cycle number, and retains most of the cycle number; the introduction of weighted averaging improves the part of the averaging result related to wheel non-circularity and reduces the impact of abnormal impact. Attached Figure Description

[0053] Figure 1 This is a flowchart of the present invention;

[0054] Figure 2 This is a block diagram illustrating the principle of the present invention;

[0055] Figure 3 The schematic diagram is used to verify the data from vehicle dynamics simulations; among which, Figure 3 (a) is the data before it has been filtered by this invention; Figure 3 (b) is the STFT time-frequency expression of axle box acceleration; Figure 3 (c) is the time-frequency representation of the STFT method; Figure 3 (d) is the time-frequency representation of the traditional order tracking method (COT); Figure 3 (e) shows the time-frequency representation of the WTSMA method;

[0056] Figure 4 This is a schematic diagram illustrating the verification of the present invention under track corrugation; wherein, Figure 4 (a) is the data before WTSMA filtering; Figure 4 (b) is the measured roughness between the wheel and the rail; Figure 4 (c) Data obtained after using the STFT method; Figure 4 (d) is the data obtained after using the order tracking method; Figure 4 (e) is a schematic diagram of time-frequency expression after using WTSMA; Figure 4 (f) is the spectrum after using WTSMA. Detailed Implementation

[0057] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0058] like Figure 1 As shown, a method for detecting non-circular wheel interference based on vehicle vibration includes the following steps:

[0059] S1. Obtain the axle box vibration acceleration signal x(t) n ) and wheelset speed signal

[0060] S2, based on wheelset speed signal For the axle box vibration acceleration signal x(t) n Perform isotropic resampling to obtain the sampled data;

[0061] S3. Calculate the weight vector of the sampled data;

[0062] S4. Calculate the weighted moving average based on the weight vector;

[0063] S5. Perform frequency domain analysis on the weighted moving average to obtain the spectrum or time spectrum containing the order and amplitude characteristics of the wheel's periodic non-circularity, as well as the envelope spectrum or envelope time spectrum containing the wheel's local non-circularity wear information.

[0064] S6. Determine whether the peak value of the spectrum containing the order and amplitude characteristics of the wheel's periodic non-circularity exceeds the set value; if so, there is a periodic wheel non-circularity defect, i.e., a wheel polygonal defect; otherwise, proceed to step S7.

[0065] S7. Determine whether the peak value of the envelope spectrum or the envelope time spectrum containing information on local non-circular wear of the wheel exceeds the set value; if so, there is a local non-circular defect of the wheel, i.e., flat scars or scratches; otherwise, the wheel condition is normal.

[0066] The specific method for step S2 is as follows:

[0067] According to the formula:

[0068]

[0069] n=0,1,…,N-1,l=0,1,…,L-1

[0070] θ l =lφ w (t N-1 ) / (L-1)

[0071] Obtain the axle box vibration acceleration x(θ) after equiangular resampling l ); where θ l The uniformly discrete sample represents the angle variable; L is the total number of angle samples; x(t) n ) represents the axle box vibration acceleration signal; N represents the total number of equal-time sampling points; t n The sampling time interval is φ; l is the l-th angle sample; Interpolate(·) is the interpolation function; φ w (t N-1 ) is the phase function.

[0072] The specific implementation method of step S3 is as follows:

[0073] S3-1, According to the formula:

[0074] b ij =|x(θ+2πi)-x(θ+2πk)|

[0075] Obtain the element b in the i-th row and j-th column of the deviation matrix B. ij Where x(θ+2πi) is the i-th data point within the moving window that is resampled at equal angles within a sampling interval, and x(θ+2πi) = x(θ l x(θ+2πk) is the k-th data point within the same window as x(θ+2πi), where i≠k; θ represents the coordinates of the isoangular resampled data point; π represents pi.

[0076] S3-2, According to the formula:

[0077]

[0078] Obtain the element w in the c-th row and d-th column of the weight matrix W. cd ;

[0079] S3-3. Obtain the corresponding weight vector p(m) based on the weight matrix W.

[0080] The specific implementation method of step S4 is as follows:

[0081] According to the formula:

[0082]

[0083] Obtain the weighted moving average of the sampled points within a sampling interval. Where Q is the number of revolutions of the wheel, and M is... w x is the number of revolutions of the moving window during equal-angle resampling within a sampling interval; x(θ+2πm) is the m-th data point within the moving window during equal-angle resampling within a sampling interval.

[0084] The specific implementation method of step S5 is as follows:

[0085] S5-1. According to the formula:

[0086]

[0087] Obtain the spectral FFT(f) containing the order and amplitude characteristics of the wheel's periodic non-circularity; or

[0088] According to the formula:

[0089]

[0090] The time-frequency spectrum STFT(θ,f) containing the order and amplitude characteristics of the wheel's periodic non-circularity is obtained; where e is the natural constant. This is the weighted moving average of the data within the corner domain; f represents the frequency; j represents the imaginary unit.

[0091] S5-2, According to the formula:

[0092]

[0093] Obtain the analytical signal Where H[·] denotes the Hilbert transform;

[0094] S5-3, According to the formula:

[0095]

[0096] Obtain the envelope

[0097] S5-4. According to the formula:

[0098]

[0099] The envelope spectrum FFT(f) containing information on local non-circular wear of the wheel is obtained;

[0100] S5-5, According to the formula:

[0101]

[0102] The envelope time spectrum STFT(θ,f) containing information on local non-circular wear of the wheel is obtained.

[0103] like Figure 2 As shown, the average of the measured signal's periods is calculated, not by averaging the signal from start to finish. A weight is assigned to the outlier degree of each data point in each revolution compared to other synchronous data points in that revolution; the greater the outlier degree, the lower the weight.

[0104] In one embodiment of the present invention, such as Figure 3 As shown, Figure 3 (a) is the data before it has been filtered by this invention; Figure 3 (b) shows the STFT time-frequency expression of axle box acceleration; abnormal shocks are assigned lower weights due to their high outlier values, thus reducing their impact; the time-frequency expression of this invention, the WTSMA method, is shown in... Figure 3 In (e), it can be seen that it is similar to... Figure 3 (c) shows the time-frequency expression of the STFT method and as Figure 3(d) shows a significant advantage over the traditional order tracking (COT) method in time-frequency representation. COT is a classic technique that directly performs order analysis after isoangular resampling without the need for averaging filtering. Time-frequency representations show that in both STFT and COT, interference components such as impact at track joints, sleeper spacing, and structural resonance are significant, interfering with the diagnosis of wheel non-circularity. In this invention, these components are suppressed to varying degrees, with only those related to wheel non-circularity clearly visible. Wherein, Angle represents angle; Inherent vibration represents natural vibration; gear meshing represents gear meshing; Frequency represents frequency; Rail weldingjoint represents rail welding joint; Sleeper passing represents sleeper passing; Roughness level represents roughness level; WPW-related represents polygon-related; order represents order; and ABA represents axle box vibration acceleration.

[0105] As can be seen from the above, interferences such as periodic asynchronous components, structural resonance components, and random track irregularities can be effectively reduced by the WTSMA method, thereby improving the robustness and discrimination ability of the non-circular characteristics of the wheel under diverse interferences.

[0106] In the ABA-based wheel non-circularity recognition framework, rail corrugated wear is considered a special defect that is easily misdiagnosed as wheel polygonal wear. To verify the resistance of WTSMA to rail corrugation, we analyzed ABA data containing this rail defect, such as... Figure 4 As shown. The results indicate that in STFT and COT, track corrugation may be misdiagnosed as wheel non-circularity, such as... Figure 4 As shown; where, Figure 4 (a) is the data before WTSMA filtering; Figure 4 (b) is the measured roughness between the wheel and the rail; Figure 4 (c) Data obtained after using the STFT method; Figure 4 (d) is the data obtained after using the order tracking method; Figure 4 (e) is a schematic diagram of time-frequency expression after using WTSMA; Figure 4 (f) is the spectrum after applying WTSMA. The response of track corrugation is obvious, similar to the characteristics of wheel non-circularity; both are observed as horizontal bright lines in the time-frequency representation. Figure 4 (e) and Figure 4 As shown in (f), the WTSMA method effectively suppresses rail corrugation. Rail corrugation represents rail corrugation; wheelroughness represents wheel roughness; and wheel roughness represents rail roughness.

[0107] This invention introduces an improved moving average method, which is less susceptible to typical interferences and less prone to misdiagnosis, solving the problem of low order resolution caused by cycle number loss; it reduces the impact of abnormal impacts and has a certain ability to suppress rail corrugation interference; it uses an automatic algorithm to calculate accurate speed from gear meshing vibration to avoid amplitude loss caused by accumulated phase error; it is applicable to various types of non-circular wheel anti-interference detection, including wheel polygons, wheel scratches, flat marks, etc., and has strong adaptability.

Claims

1. A method for detecting non-circularity of wheels based on vehicle vibration, characterized in that, Includes the following steps: S1. Acquire the axle box vibration acceleration signal Wheelset speed signal ; S2, based on wheelset speed signal vibration acceleration signal of axle box Perform isotropic resampling to obtain the sampled data; S3. Calculate the weight vector of the sampled data; S4. Calculate the weighted moving average based on the weight vector; S5. Perform frequency domain analysis on the weighted moving average to obtain the spectrum or time spectrum containing the order and amplitude characteristics of the wheel's periodic non-circularity, as well as the envelope spectrum or envelope time spectrum containing the wheel's local non-circularity wear information. S6. Determine whether the peak value of the spectrum containing the order and amplitude characteristics of the wheel's periodic non-circularity exceeds the set value; if so, there is a periodic wheel non-circularity defect, i.e., a wheel polygonal defect; otherwise, proceed to step S7. S7. Determine whether the peak value of the envelope spectrum or the envelope time spectrum containing information on local non-circular wear of the wheel exceeds the set value; if so, there is a local non-circular defect of the wheel, i.e., flat marks or scratches; otherwise, the wheel condition is normal. The specific implementation method of step S3 is as follows: S3-1, According to the formula: The first deviation matrix B is obtained. i Line number j Column elements ;in, The moving window within which isoangular resampling occurs within a sampling interval. i Data points, ; To and The first in the same window k There are 10 data points, of which ; Represents the coordinates of the isoangular resampled data points; π Represents pi; The vibration acceleration of the axle box after equiangular resampling; S3-2, According to the formula: Obtain the first weight of the weight matrix W c Line number d Column elements ; S3-3. Obtain the corresponding weight vector based on the weight matrix W. ; The specific implementation method of step S4 is as follows: According to the formula: Obtain the weighted moving average of the sampled points within a sampling interval. ;in, Q The number of revolutions of the wheel. The number of revolutions of the moving window wheel during equal-angle resampling within a sampling interval; The moving window within which isoangular resampling occurs within a sampling interval. m Data points.

2. The method for detecting non-circular wheel interference based on vehicle vibration according to claim 1, characterized in that, The specific method of step S2 is as follows: According to the formula: Obtain the axle box vibration acceleration after isoangular resampling ;in, Uniformly discrete samples representing angular variables; L It is the total number of angle samples; This is the vibration acceleration signal of the axle box; For common use N Equal-time sampling points; The sampling time interval is equal. For the first l One angle sample; Interpolate(·) is the interpolation function; It is the phase function.

3. The method for detecting non-circularity of wheels based on vehicle vibration according to claim 2, characterized in that, The specific implementation method of step S5 is as follows: S5-1. According to the formula: The spectrum containing the order and amplitude characteristics of the wheel's periodic non-circularity was obtained. ;or According to the formula: The time spectrum containing the order and amplitude characteristics of the wheel's periodic non-circularity is obtained. ;in, e It is a natural constant; This is the weighted moving average of the data within the corner domain; = ; Indicates frequency; j Indicates the imaginary unit; S5-2, According to the formula: Obtain the analytical signal ;in, Represents the Hilbert transform; S5-3, According to the formula: Obtain the envelope ; S5-4. According to the formula: Obtain the envelope spectrum containing information on localized non-circular wear of the wheel. ; S5-5, According to the formula: Obtain the envelope time spectrum containing information on local non-circular wear of the wheel. .

Images