Online estimation method of bogie fatigue life based on axlebox vibration
By constructing the transfer function between axle box vibration and bogie dynamic stress, combined with the fatigue damage calculation method of the frequency domain method, the online estimation of bogie fatigue life based on a small number of sensors is achieved, solving the problems of high cost and low reliability in the prior art, and improving the reliability and accuracy of the evaluation.
Patent Information
- Application Number
- CN202111587981.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-12-23
- Publication Date
- 2025-03-21
- Estimated Expiration
- 2041-12-23
AI Technical Summary
The prior art requires a large number of sensors when evaluating the fatigue life of the bogie, resulting in high cost and low reliability, and the inability to effectively deal with misjudgment and misjudgment.
The online bogie fatigue life estimation method based on axle box vibration is used to construct the transfer function between axle box vibration and bogie dynamic stress, and combine the fatigue damage calculation method of the frequency domain method to achieve fatigue damage evaluation.
It reduces the cost of bogie structure health monitoring, improves the reliability of fatigue damage assessment, and can effectively predict the overall error of dynamic stress less than 25%, which is acceptable in engineering applications.
Smart Images
Figure CN114357614B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of railway vehicle structure health monitoring, and in particular relates to an online estimation method for bogie fatigue life based on axle box vibration. Background Art
[0002] The new generation of intelligent high-speed railways have installed temperature-vibration composite sensors in parts such as axle boxes and gear boxes. How to use a small number of sensors to grasp more vehicle operating conditions has become a top priority in vehicle health monitoring research.
[0003] The current method of evaluating the fatigue life of bogies is mainly to arrange sensors in the areas of concern. Since the bogie system has many areas of concern, a large number of sensors need to be arranged on the bogie. The number of railway vehicles is huge. On the one hand, too many sensors are not economical. On the other hand, over-reliance on sensors is not conducive to improving the reliability of bogie fatigue damage monitoring and cannot deal with "misjudgment" and "missed judgment" phenomena. Based on this, the present invention proposes a method for implementing bogie fatigue damage assessment based on a small number of sensors, reducing the cost of bogie structural health monitoring and improving the reliability of fatigue damage assessment. Summary of the invention
[0004] The purpose of the present invention is to provide an online estimation method for the fatigue life of a bogie based on axle box vibration. The method obtains the stress PSD of the focus position according to the axle box vibration acceleration PSD combined with the transfer function, and realizes fatigue damage assessment by using the fatigue damage calculation method of the frequency domain method.
[0005] To achieve the above object, the present invention adopts the following technical solution:
[0006] The online estimation method of bogie fatigue life based on axle box vibration includes the following steps:
[0007] (1) Constructing the transfer function between axle box vibration and bogie dynamic stress
[0008] First, the transfer function is calibrated in the laboratory stage, and the input is obtained: the axle box vibration acceleration signal, and the output is the dynamic stress at the key position of the bogie; the mutual power density of the input and output and the autopower spectrum density of the input are calculated, and the transfer function H(f) is calibrated in combination with formula (1). In the line measurement stage, the input power spectrum density S is calculated according to the axle box vibration signal. ij (f), when i≠j, S ij (f) is the cross power spectral density between the i-th vibration signal and the j-th vibration signal. When i=j, S ij (f) represents the auto-power spectrum density of the i-th vibration signal, and the input power spectrum density S ij (f) Combined transfer function H j(f) Substitute into formula (2) to calculate the output power spectrum density S y (f);
[0009]
[0010]
[0011] (2) Fatigue damage estimation based on frequency domain method
[0012] First, the input power spectrum density S is calculated according to the vibration signal. ij (f); then calculate the output power spectral density based on the input power spectral density and the transfer function constructed in step (1) Secondly, according to the output power spectrum density S y (f) Combined with the frequency domain fatigue damage calculation method, the fatigue damage of the area of interest can be estimated, and the damage per unit time Among them, v p is the peak crossover frequency, C and k are material constants, and p(s) is the probability density function of the rainflow amplitude.
[0013] Furthermore, the mean value of the transfer function constructed in step (1) may be corrected by multiplying the transfer function by the mean error of the transfer function.
[0014] Furthermore, the input power spectrum density S ij (f) and output power spectral density S y (f) are calculated using fast Fourier transform.
[0015] Furthermore, the rainflow amplitude probability density function p(s) is calculated using a suitable distribution model, such as the Dirlik method and the Zhao Baker method.
[0016] The Dirlik method combines an exponential distribution and two Rayleigh distributions, and its p(s) is expressed as:
[0017]
[0018] Among them, G1, G2, G3, R and Q are parameters. G3 = 1-G1-G2; Z is the normalized amplitude, x m represents the mean frequency,
[0019] The Zhao Baker method linearly combines a Weibull distribution and a Rayleigh distribution, and its p(s) expression is:
[0020]
[0021] W is the weight coefficient, α and β are Weibull parameters, α = 8-7r, Γ(·) is the gamma function,
[0022] Furthermore, the window size of the Dirlik method is 4s, and the window size of the Zhao Baker method is 0.5s.
[0023] The present invention has the following beneficial effects:
[0024] (1) The transfer function obtained by the cross power spectrum density and the auto power spectrum density can be used to establish the transfer relationship between the dynamic stress PSD and the axle box acceleration PSD in the frequency domain. Considering the nonlinearity, the transfer function can fit the measured values well in terms of the prediction of dynamic stress PSD, RMS value and irregularity factor. The overall prediction error of dynamic stress is less than 25%, which is acceptable in engineering applications.
[0025] (2) The Dirlik method and the Zhao Baker method can estimate the dynamic stress amplitude distribution well, especially for narrow-band processes. The two distribution models have high prediction accuracy for the signal distribution of the combination of large stress amplitude distribution and small stress amplitude distribution. The smaller window size can better characterize the non-stationary characteristics of dynamic stress. The Dirlik method and the Zhao Baker method have the highest prediction accuracy when the window length is 4 s and 0.5 s, respectively.
[0026] In summary, the present invention provides an alternative method for structural health monitoring, which can estimate structural fatigue damage online based on axle box acceleration. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] Figure 1 Schematic diagram of measured data, (a) is the measured signal of the three-axis vibration acceleration of the left axle box, (b) is the measured signal of the three-axis vibration acceleration of the right axle box, and (c) is the measured dynamic stress of the antenna beam.
[0028] Figure 2 Schematic diagram of the cross-power spectrum density between the axis vibration signal and the antenna beam strain signal.
[0029] Figure 3 is the autopower spectral density of the axle box vibration signal.
[0030] Figure 4 Schematic diagram of the transfer function between the antenna beam and the axle box signal, where the thick black line represents the average transfer function.
[0031] Figure 5Comparison diagram of the dynamic stress PSD predicted by the average transfer function and the measured dynamic stress PSD.
[0032] Figure 6 It is a comparison chart of the irregularity factor and RMS of the predicted value and the measured value, as well as the corresponding error diagram.
[0033] Figure 7 is the average error of the transfer function.
[0034] Figure 8 Comparison diagram of the dynamic stress PSD of the antenna beam predicted by the modified transfer function and the measured dynamic stress PSD.
[0035] Fig. 9 Comparison diagram of the simulated and measured values of RMS and irregularity factor achieved using the corrected transfer function, as well as a schematic diagram of the error.
[0036] Fig.10 It is a comparison chart of the probability density function of the measured stress amplitude obtained by the rain flow counting method and the probability density function of the simulated stress amplitude calculated by the frequency domain method.
[0037] Fig.11 Schematic diagram of the stress amplitude probability density function calculated based on time-domain rainflow counting and frequency-domain power spectral density with different window lengths. (a) is the time-domain rainflow counting and Dirlik method, and (b) is the time-domain rainflow counting and Zhao Baker method.
[0038] Fig.12 Comparison diagram of fatigue damage calculated by time-domain rain flow counting, Dirlik method and Zhao Baker method. DETAILED DESCRIPTION
[0039] The method for online estimation of fatigue life of a bogie based on axle box vibration provided in this embodiment comprises the following steps:
[0040] (1) Constructing the transfer function between axle box vibration and bogie dynamic stress
[0041] First, the transfer function is calibrated in the laboratory stage, and the input is obtained: the axle box vibration acceleration signal, and the output is the dynamic stress at the key position of the bogie; the mutual power density of the input and output and the autopower spectrum density of the input are calculated, and the transfer function H(f) is calibrated in combination with formula (1). In the line measurement stage, the input power spectrum density S is calculated according to the axle box vibration signal. ij (f), when i≠j, S ij (f) is the cross power spectral density between the i-th vibration signal and the j-th vibration signal. When i=j, S ij (f) represents the auto-power spectrum density of the i-th vibration signal, and the input power spectrum density S ij(f) Combined transfer function H j (f) Substitute into formula (2) to calculate the output power spectrum density S y (f);
[0042]
[0043]
[0044] (2) Fatigue damage estimation based on frequency domain method
[0045] First, the input power spectrum density S is calculated according to the vibration signal. ij (f); then calculate the output power spectral density based on the input power spectral density and the transfer function constructed in step (1) Secondly, according to the output power spectrum density S y (f) Combined with the frequency domain fatigue damage calculation method, the fatigue damage of the area of interest can be estimated, and the damage per unit time Among them, v p is the peak crossover frequency, C and k are material constants, and p(s) is the probability density function of the rainflow amplitude.
[0046] The statistical characteristics of a signal are usually expressed by the spectral moment and spectral width coefficient of the signal. The i-th order spectral moment m i for:
[0047] The irregularity factor r is: Its value range is 0 to 1. The larger the value of r is, the more likely the signal is a narrowband random process. Peak crossover frequency v p And the zero line forward crossing frequency v0 can be expressed as:
[0048] The reliability of this embodiment is tested. Since the antenna beam is far away from the rear wheel pair, the vibration of the antenna beam is highly correlated with the vibration of the front wheel pair. Three-dimensional vibration acceleration sensors are installed on the left and right axle boxes. The three-dimensional vibration acceleration sensors collect three-dimensional vibration signals (including lateral, longitudinal and vertical) as input excitation of the antenna beam. The uniaxial strain gauge is installed in the middle position of the antenna beam, and the dynamic stress at this position is measured as the output. A 6-input single-output system is established.
[0049] There are 16 test sites in total. The data measured from sites 2 to 3 are as follows: Figure 1 shown.
[0050] Figure 2 The cross power spectrum density of the axle box vibration signal and the antenna beam stress signal is shown. The cross power spectrum density describes the correlation between the two signals in the frequency domain.
[0051] Figure 3 Shows the autopower spectral density of the input
[0052] Based on the cross power spectrum density of the input and output signals and the auto power spectrum density of the input signal, the transfer function of each time period can be calculated, and the average transfer function can be further calculated. Figure 4 shown.
[0053] Figure 5 The measured stress PSD and simulated stress PSD are shown in the four selected time periods without considering the correction of the transfer function. It can be found that there are large differences between the predicted values and the simulated values in some frequency bands.
[0054] By comparing the predicted and measured RMS values and irregularity factors of dynamic stress, the accuracy of the method proposed in this embodiment is verified. Figure 6 As shown, the overall error of the predicted results of RMS and irregularity factor is less than 35% relative to the measured values.
[0055] Figure 7 The average error distribution of the transfer function in the frequency domain determined by dividing the measured PSD by the predicted PSD for each segment is further illustrated. The predicted results tend to underestimate the frequency amplitude of 60Hz and overestimate the amplitude of the dominant frequency of 75Hz. Therefore, the transfer function is further corrected by multiplying the transfer function by the average error of the transfer function. Figure 6 (c) and Fig. 9 (c) It can be found that the prediction error of the irregular factor of the predicted output power spectrum is significantly reduced, and the overall prediction error of the irregular factor is less than 20%, thereby better improving the accuracy of this embodiment.
[0056] Figure 8 The predicted stress PSD of the test section using the modified transfer function is given. Compared with Figure (5), the prediction accuracy is higher.
[0057] like Fig. 9 As shown in Figure 2, the predicted RMS value and irregularity factor are comparable to the measured values, and the overall error is still at an acceptable level. This indicates that the modified transfer function can better characterize the transfer relationship between the antenna beam stress and the axle box acceleration.
[0058] By using the dynamic stress power spectrum density obtained above and combining the Dirlik and Zhao Baker method, the probability density function of the rain flow stress amplitude can be calculated: Fig.10 shown.
[0059] Fig.11The stress amplitude density function calculated based on the time-domain rainflow counting and the frequency-domain power spectrum density is shown. In the figure, the bar graph is the probability density function calculated based on the time-domain rainflow counting results, and the solid line represents the S y (f) The calculated probability density function of the stress amplitude is the average probability density function obtained by adding the probability density functions of each time period and dividing by the total number of time periods.
[0060] Fig.12 The fatigue damage calculated by time-domain rainflow counting, Dirlik method and Zhao Baker method is shown. The total damage calculated by time-domain rainflow counting is 2.84×10 -4 As the baseline, it can be found that as the window size continues to increase, the fatigue damage calculated by the two methods continues to decrease. When the window size of Dirlik is 4s and the window size of Zhao Baker method is 0.5s, the calculation results are closest to the time domain calculation results.
[0061] In summary, the transfer function method can be used to establish the transfer relationship between the dynamic stress of the bogie and the axle box acceleration, the stress PSD of the bogie frame focus area can be calculated by the axle box vibration acceleration PSD, and the fatigue damage can be calculated by the stress PSD. At present, the axle box acceleration sensor has been used as a common configuration for the new generation of rail vehicles, which can estimate the online fatigue damage of the bogie. It provides a basis for the fatigue life management of key components of railway vehicles.
[0062] The above description is only a preferred implementation manner of the present invention, but the protection scope of the present invention is not limited thereto, and any modification and replacement based on the technical solution and inventive concept provided by the present invention should be included in the protection scope of the present invention.
Claims
1. An online estimation method for bogie fatigue life based on axle box vibration is characterized by The steps include: (1) Constructing the transfer function between axle box vibration and bogie dynamic stress First, the transfer function is calibrated in the laboratory stage, and the input is obtained: the axle box vibration acceleration signal, and the output is the dynamic stress at the key position of the bogie; the mutual power density of the input and output and the autopower spectrum density of the input are calculated, and the transfer function H(f) is calibrated in combination with formula (1); During the line measurement phase, the input power spectrum density S is calculated based on the axle box vibration signal. ij (f), when i≠j, S ij (f) is the cross power spectral density between the i-th vibration signal and the j-th vibration signal. When i=j, S ij (f) represents the auto-power spectrum density of the i-th vibration signal, and the input power spectrum density S ij (f) Combined transfer function H j (f) Substitute into formula (2) to calculate the output power spectrum density S y (f); (2) Fatigue damage estimation based on frequency domain method First, the input power spectrum density S is calculated according to the vibration signal. ij (f); then calculate the output power spectral density based on the input power spectral density and the transfer function constructed in step (1) Secondly, according to the output power spectrum density S y (f) Combined with the frequency domain fatigue damage calculation method, the fatigue damage of the area of interest can be estimated, and the damage per unit time Among them, v p is the peak crossover frequency, C and k are material constants, and p(s) is the probability density function of the rainflow amplitude.
2. The method for online estimation of bogie fatigue life based on axle box vibration according to claim 1 is characterized in that: The mean value of the transfer function constructed in step (1) can be corrected by multiplying the transfer function with the average error of the transfer function to improve the prediction accuracy of the output power spectral density.
3. The method for online estimation of bogie fatigue life based on axle box vibration according to claim 1, characterized in that: The input position is selected on the axle box, the output position is the weak position of the bogie, and the transfer function of input and output is constructed.
4. The method for online estimation of bogie fatigue life based on axle box vibration according to claim 1, characterized in that: The rain flow amplitude probability density function p(s) is calculated by selecting the Dirlik method and the Zhao Baker method distribution model.
5. The method for online estimation of bogie fatigue life based on axle box vibration according to claim 4 is characterized in that: The Dirlik method has the highest prediction accuracy when the window size is 4s, and the Zhao Baker method has the highest prediction accuracy when the window size is 0.5s.