A method for predicting the service life of primary steel springs of railway vehicles under time-varying loads

Abstract

The present invention relates to a method for predicting the service life of primary steel springs on rail vehicles under time-varying loads. The method comprises: collecting time-varying vibration accelerations from the axlebox region; combining the test data to form a new feature channel; segmenting the feature channel into feature samples, estimating the distribution characteristics of the square root values ​​based on the PSD of the feature samples, summarizing the power spectrum density of the time-varying vibration load using the normal tolerance upper limit method and the Johnson method; and predicting the service life of primary steel springs on rail vehicles based on the summarized spectrum of the time-varying vibration load. Compared with the prior art, the present invention quantitatively expresses the energy and frequency domain distribution characteristics of non-normal time-varying vibration loads during rail vehicle service, thereby improving the accuracy of service life prediction for primary steel springs on rail vehicles.

Description

Technical Field

[0001] The present invention relates to the field of vibration fatigue of rail transit vehicle structural parts, and in particular to a method for predicting the service life of a primary steel spring of a rail vehicle under time-varying loads. Background Art

[0002] The bogie is one of the most critical components of an EMU train. It serves as the running gear, supporting the vehicle body and carrying its weight while guiding it along the track. Whether the performance of each bogie component meets operational requirements directly impacts vehicle safety and dynamic performance. Furthermore, various bogie parameters directly determine vehicle stability and ride comfort. As a key component of the bogie's primary suspension, the performance of the primary spring directly determines vehicle safety and ride comfort.

[0003] During the service life of a railway vehicle, diverse track structures and varying speeds on the same line result in short-term temporal variations in the vibration loads on the primary springs. Furthermore, the constantly changing wheel-rail contact surface conditions and the aging of key rubber components in the suspension system contribute to long-term temporal variations in the vibration loads on the primary springs. Existing methods typically predict the service life of primary springs based on the vibration loads of a specific track type or a specific period. This can lead to significant errors, reducing the accuracy of primary spring service life predictions. Summary of the Invention

[0004] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a method for predicting the service life of primary steel springs of railway vehicles under time-varying loads. The method can obtain an inductive spectrum that expresses the energy and frequency domain distribution characteristics of the time-varying vibration load during the service of the railway vehicle, ensure that the inductive spectrum and the time-varying vibration load have equal damage capacity, and use the inductive spectrum of the time-varying vibration load to predict fatigue life and solve the service life of the primary steel spring. The method is simple and reliable, thereby improving the accuracy of the service life prediction of the primary springs of railway vehicles.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] A method for predicting the service life of a primary steel spring of a railway vehicle under time-varying loads comprises the following steps:

[0007] S1. Test the rail vehicle and collect the time-varying vibration acceleration of the axle box area to obtain an original data set. The original data set includes M×N test data, where M represents the number of vibration acceleration measurement points in the axle box area, N represents the number of tests on the rail vehicle, and X(m,n) represents the test data corresponding to measurement point m during the nth test. The length of the test data is length, where M>1, N>1, m=1, 2, ..., M, and n=1, 2, ..., N.

[0008] S2. Traverse the test data in the original data set. If two test data satisfy the preset merging conditions, merge them into a new test data. Repeat this step until there is no test data that meets the merging conditions, and obtain a channel data set. The channel data set includes P feature channels. represents the qth test data under feature channel p, p = 1, 2, ..., P, q = 1, 2, ..., N P , N P Indicates the number of test data for feature channel p, N P >0, the capacity Q of feature channel p P =N P ×length;

[0009] S3, traverse the feature channels in the channel data set, divide the feature channels into multiple feature samples according to the preset step size Δw, perform PSD estimation on each feature sample, and calculate the upper limit of the power spectrum density of each feature channel at each spectral line, where, Represents the upper limit of the power spectral density of characteristic channel p at spectral line k;

[0010] S4, select the maximum value of each spectral line in all characteristic channels as the inductive spectrum value, That is, the time-varying vibration acceleration summary spectrum of the axle box area is obtained, where represents the inductive spectrum value at spectral line k;

[0011] S5. Using the time-varying vibration acceleration induction spectrum as the load spectrum for fatigue analysis of the primary spring of a railway vehicle, the stress power spectrum density G is solved through the finite element model of the primary spring structure of the railway vehicle. σ (f), based on the stress power spectrum density G σ (f) Determine the service life of the primary steel spring of a railway vehicle.

[0012] Furthermore, step S2 is specifically as follows:

[0013] S21. Calculate the mean and variance of each test data respectively;

[0014] S22. Obtain two test data sets from the original data set, denoted as test data a and test data b. If both meet the pre-set merging condition, proceed to step S23. If not, repeat step S22 until no two test data sets in the original data set meet the merging condition.

[0015] S23: Delete the test data a and the test data b from the original data set, put the new test data obtained by merging the two back into the original data set, and execute step S22.

[0016] Furthermore, in step S21, the calculation process of the mean and variance of the test data X(m,n) is as follows:

[0017] Split the test data X(m,n) of length length into L parts, where L is the preset parameter value, and obtain the data set {X(m,n)1, X(m,n)2, ..., X(m,n) L}, calculate X(m,n)1, X(m,n)2, ..., X(m,n) respectively. L The root mean square of the data set {Rms(m,n)1, Rms(m,n)2, …, Rms(m,n) L}, the data set {Rms(m,n)1, Rms(m,n)2, ..., Rms(m,n) L The mean and variance of} are used as the mean of the test data X(m,n) and variance S 2 (m,n).

[0018] Furthermore, step S22 is specifically as follows:

[0019] Get two test data from the original data set, record them as test data a and test data b, and construct statistics F(a,b) and t(a,b):

[0020]

[0021] Among them, S 2 (a) and S 2 (b) represents the variance of test data a and test data b, respectively. and Represent the means of test data a and test data b respectively. If F(a, b) and t(a, b) meet the preset merging conditions, step S23 is executed. If not, step S22 is repeated until any two test data in the original data set do not meet the merging conditions. The merging conditions are:

[0022]

[0023] Among them, α' is the preset confidence level.

[0024] Furthermore, in step S23, the length of the new test data is the sum of the lengths of test data a and test data b, length(a) and length(b). and The average value of is the mean of the new test data, S 2 (a) and S 2 The mean of (b) is the variance of the new test data.

[0025] Furthermore, in step S3, the process of determining the upper limit of the power spectrum density of the characteristic channel p at the spectrum line k is as follows:

[0026] S31, according to the preset step size Δw, the capacity is Q P The feature channel p is divided into Num p feature samples, Num p =(Q P / Δw);

[0027] S32, performing power spectrum density estimation on each feature sample respectively, and calculating the square root of the PSD estimation of each feature sample;

[0028] S33, at the spectral line k, if the Num of the characteristic channel p p If the PSD estimated square root of the characteristic samples conforms to the normal distribution, then the first calculation process is used to calculate If it conforms to the non-normal distribution, it is calculated according to the preset second calculation process

[0029] Furthermore, the first calculation process is as follows:

[0030] First calculate as follows:

[0031]

[0032] in, Num is the number of feature channels p p The mean and root mean square of the PSD estimated by the characteristic samples at the spectral line k, Z represents the number of characteristic channels p p The normal distribution of the square root of the PSD estimate of the characteristic samples, Z β is the β penalty point of the normal distribution Z, The non-centrality is The degrees of freedom are Q p -1 non-central t distribution, α is the preset confidence level of the interval estimate, and β is the preset probability;

[0033] based on Calculate the upper limit of the power spectrum density of the characteristic channel p at the spectrum line k

[0034]

[0035] Get the upper limit of the power spectrum density of the characteristic channel p at the spectrum line k

[0036] Furthermore, the second calculation process is as follows:

[0037] First, based on the Johnson system distribution function, the Num of the feature channel p is p The square root of the PSD estimate of the characteristic samples is converted from a non-normal distribution at the spectral line k to a normal distribution:

[0038] z=γ+ηK i (x;λ,ε)

[0039]

[0040] Among them, x represents the Num of feature channel p p The square root of the PSD estimate of the characteristic sample is non-normally distributed at spectral line k, z represents the standard normal distribution obtained by transformation, γ, η, λ, ε are Johnson system parameters, and the value of i ranges from 1 to 3, which is determined according to the characteristics of the non-normal distribution x;

[0041] Under the preset probability β, solve the quantile F' of the corresponding standard normal distribution z, and based on the inverse of the Johnson system distribution function, the upper limit of the probability β non-normal data can be obtained as follows:

[0042]

[0043] Where μ = 0, σ = 1, z β represents the β penalty point of the standard normal distribution z;

[0044] based on Calculate the upper limit of the power spectral density of characteristic channel p at spectral line k as follows:

[0045]

[0046] Get the upper limit of the power spectrum density of the characteristic channel p at the spectrum line k

[0047] Furthermore, in step S5, the stress power spectrum density G is solved σ (f) as follows:

[0048]

[0049] Where H is the frequency response function of the spring stress response due to the axlebox acceleration excitation, f = k * Δf, where Δf depends on the number of analysis points during PSD estimation, and ∑ is the summation formula, which refers to the sum of the stresses generated by the axlebox's lateral, longitudinal, and vertical acceleration spectra on the primary spring.

[0050] Furthermore, in step S5, the service life of the primary steel spring of the railway vehicle is calculated as follows:

[0051] Based on the stress power spectrum density Gσ (f) Based on the Dirlik rainflow amplitude probability density function, the fatigue cumulative damage D is solved:

[0052]

[0053] Where S is the stress amplitude, P(S) is the probability density estimation function of Dirlik rain flow amplitude, according to G σ (f) is obtained, T is the load duration, v p is the peak crossing times, C and m are the material SN curves;

[0054] When the fatigue cumulative damage D=1, the structure fails due to fatigue. The service life of the primary spring of the railway vehicle is calculated based on the fatigue cumulative damage D:

[0055] T life =T / D

[0056] Among them, T life It is the service life of the primary steel spring of rail vehicles.

[0057] Compared with the prior art, the present invention has the following beneficial effects:

[0058] (1) By designing statistics and merging conditions, similar test data are merged, which realizes the merging of time-varying vibration load data in the same area of ​​the axle box, and plays a role in quickly and effectively processing the long-term time-varying vibration data in the axle box area.

[0059] (2) The characteristic channel is divided into characteristic samples. According to the normal or non-normal distribution of the square root of the PSD estimation value of the characteristic sample, the upper limit of the power spectrum density of the characteristic channel is calculated by the normal tolerance upper limit method and the Johnson method, and then the time-varying vibration load induction spectrum is obtained. Therefore, the induction spectrum has the advantage of better meeting the excitation requirements of the reliability test.

[0060] (3) A finite element model of the primary steel spring structure of a rail vehicle is established. The dynamic stress response spectrum of the steel spring is solved by applying the time-varying vibration load induction spectrum, and the fatigue damage is calculated to determine the service life of the primary steel spring. The service life of the primary steel spring under the time-varying vibration load can be quickly predicted with higher prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 is a flow chart of the present invention;

[0062] Figure 2 This is a schematic diagram of the primary steel spring structure of a rail vehicle. DETAILED DESCRIPTION

[0063] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0064] In the drawings, components with identical structures are denoted by the same reference numerals, and components with similar structures or functions are denoted by similar reference numerals. The size and thickness of each component shown in the drawings are arbitrary and are not limited by the present invention. For clarity, some components in the drawings are exaggerated.

[0065] Example 1:

[0066] A method for predicting the service life of primary steel springs of railway vehicles under time-varying loads, such as Figure 1 As shown, the following steps are included:

[0067] S1. Test the rail vehicle and collect the time-varying vibration acceleration of the axle box area to obtain an original data set. The original data set includes M × N test data, where M represents the number of vibration acceleration measurement points in the axle box area, N represents the number of tests on the rail vehicle, and X(m,n) represents the test data corresponding to measurement point m during the nth test. The length of the test data is length, where M>1, N>1, m=1, 2, ..., M, and n=1, 2, ..., N.

[0068] S2. Traverse the test data in the original data set. If two test data meet the preset merging conditions, merge them into a new test data. Repeat this step until there is no test data that meets the merging conditions. The channel data set is obtained. The channel data set includes P feature channels. represents the qth test data under feature channel p, p = 1, 2, ..., P, q = 1, 2, ..., N P , N P Indicates the number of test data for feature channel p, N P >0, the capacity Q of feature channel p P =N P ×length;

[0069] S3, traverse the feature channels in the channel data set, divide the feature channels into multiple feature samples according to the preset step size Δw, perform PSD estimation on each feature sample, and calculate the upper limit of the power spectrum density of each feature channel at each spectral line, where Represents the upper limit of the power spectral density of characteristic channel p at spectral line k;

[0070] S4, select the maximum value of each spectral line in all characteristic channels as the inductive spectrum value, That is, the time-varying vibration acceleration summary spectrum of the axle box area is obtained, where represents the inductive spectrum value at spectral line k;

[0071] S5. Using the time-varying vibration acceleration induction spectrum as the load spectrum for fatigue analysis of the primary spring of a railway vehicle, the stress power spectrum density G is solved through the finite element model of the primary spring structure of the railway vehicle. σ (f), based on the stress power spectrum density G σ (f) Determine the service life of the primary steel spring of a railway vehicle.

[0072] (1) Obtain the time-varying vibration acceleration of the axle box area of ​​the rail vehicle during service. In this embodiment, within a maintenance cycle of the rail vehicle, the axle box area vibration acceleration test is performed every 10,000 kilometers of running mileage, M measuring points are set in the axle box area, and a total of N tests are performed to obtain M×N sets of test data. X(m,n) represents the test data corresponding to the measuring point m during the nth test. The values ​​of M and N can be set as needed. In order to facilitate the explanation of the technical solution of this application, the following example is taken as M=4 and N=6. X(m,n) takes test values ​​of 0s to 1000s. Assuming that 1s takes a value, the length of the test data is 1000. However, it can be understood that 4, 6, and 1000 do not represent actual implementation data, but are just schematic values ​​for illustration. The original data set is as follows:

[0073] Table 1: Example of original dataset

[0074] 1 2 3 4 5 6 1 X(1,1) X(1,2) X(1,3) X(1,4) X(1,5) X(1,6) 2 X(2,1) X(2,2) X(2,3) X(2,4) X(2,5) X(2,6) 3 X(3,1) X(3,2) X(3,3) X(3,4) X(3,5) X(3,6) 4 X(4,1) X(4,2) X(4,3) X(4,4) X(4,5) X(4,6)

[0075] (2) Step S2 is specifically as follows:

[0076] S21. Calculate the mean and variance of each test data respectively;

[0077] The calculation process of the mean and variance of the test data X(m,n) is as follows:

[0078] Split the test data X(m,n) of length length into L parts, where L is the preset parameter value, and obtain the data set {X(m,n)1, X(m,n)2, ..., X(m,n) L}, calculate X(m,n)1, X(m,n)2, ..., X(m,n) respectively. L The root mean square of the data set {Rms(m,n)1, Rms(m,n)2, …, Rms(m,n) L}, the data set {Rms(m,n)1, Rms(m,n)2, ..., Rms(m,n) L The mean and variance of} are used as the mean of the test data X(m,n) and variance S 2 (m,n).

[0079] L can be set as needed, l = 1, 2, ..., L, so X(m,n) l It is actually a data segment of length / L, including multiple test values. For example, if L is 10 and X(m,n)1 is the test value from 0 to 100s, the root mean square Rms(m,n)1 of X(m,n)1 can be obtained.

[0080] The root mean square of L parts is used to obtain the data set {Rms(m,n)1, Rms(m,n)2, ..., Rms(m,n) L}, and then calculate the mean and variance to get the mean of the test data X(m,n) and variance S 2 (m,n).

[0081] S22. Obtain two test data sets from the original data set, denoted as test data a and test data b. If both meet the pre-set merging condition, proceed to step S23. If not, repeat step S22 until no two test data sets in the original data set meet the merging condition.

[0082] Step S22 is specifically as follows:

[0083] Get two test data from the original data set, record them as test data a and test data b, and construct statistics F(a,b) and t(a,b):

[0084]

[0085] Among them, S 2 (a) and S 2 (b) represents the variance of test data a and test data b, respectively. and Represent the means of test data a and test data b respectively. If F(a, b) and t(a, b) meet the preset merging conditions, step S23 is executed. If not, step S22 is repeated until any two test data in the original data set do not meet the merging conditions. The merging conditions are:

[0086]

[0087] Among them, α' is the preset confidence level, such as 0.95, F α' / 2 (L-1,L-1),F 1-α' / 2 (L-1,L-1),t 1-α' / 2 The value of (2L-2) can be determined by substituting the values ​​of α' and L into the F and t distribution functions.

[0088] S23, delete test data a and test data b from the original data set, put the new test data obtained by merging the two back into the original data set, and execute step S22; the length of the new test data is the sum of the lengths of test data a and test data b, length(a) and length(b). and The average value of is the mean of the new test data, S 2 (a) and S 2 The mean of (b) is the variance of the new test data.

[0089] In step S2, if two pieces of test data meet the merging conditions, their variances and means are basically the same, and it is considered that the two pieces of test data belong to the same subject. The data can be merged to form new data, namely feature channel data. In fact, after the merger is completed, the test data belonging to the same subject will be placed in the same feature channel. In extreme cases, a feature channel includes 1 or M×N pieces of test data.

[0090] (3) In step S3, the process of determining the upper limit of the power spectrum density of the characteristic channel p at the spectrum line k is as follows:

[0091] S31, according to the preset step size Δw, the capacity is Q P The feature channel p is divided into Num p feature samples, Num p =(Q P / Δw); For example, the feature channel p includes 4 test data, the capacity Q P is 4000 and the step size Δw is 20, then 200 feature samples are obtained.

[0092] S32, performing power spectrum density estimation on each feature sample respectively, and calculating the square root of the PSD estimation of each feature sample;

[0093] Power Spectral Density (PSD) is used to describe the frequency domain distribution characteristics of the vibration energy of the test data. Therefore, PSD is estimated for each sample, and the horizontal axis is the frequency (Hz) and the vertical axis is the spectrum value (g 2 / Hz), and then taking the square root of the vertical axis value to obtain the square root of the PSD estimate for each feature sample. For feature channel p, after completing this step, taking spectral line k = 10 (essentially corresponding to frequency value) as an example, the square root of the PSD estimate for 200 feature samples at spectral line 10 can be determined.

[0094] S33, at the spectral line k, if the Num of the characteristic channel p p If the PSD estimated square root of the characteristic samples conforms to the normal distribution, then the first calculation process is used to calculate If it conforms to the non-normal distribution, it is calculated according to the preset second calculation process

[0095] Analyze each feature channel separately, taking feature channel p as an example:

[0096] At spectral line k = 1, the square root of the PSD estimate of the 200 characteristic samples of characteristic channel p is normally distributed or non-normally distributed, so the first calculation process or the second calculation process can be selected to solve the upper limit of the power spectral density of characteristic channel p at spectral line 1. The situation of characteristic channel p at spectral line k = 2 is further analyzed, and the first calculation process or the second calculation process is selected to solve the upper limit of the power spectral density of characteristic channel p at spectral line 2, until the upper limit of the power spectral density of characteristic channel p at each spectral line is determined. According to actual needs, the commonly used spectral line interval k = 0 to 1000 can be selected for analysis.

[0097] Similarly, it is necessary to perform the above analysis on all characteristic channels in the channel data set to calculate the upper limit of the power spectrum density of each characteristic channel at each spectral line.

[0098] The above process determines the upper limit of the power spectral density of one characteristic channel at each spectral line, and then determines the upper limit of the power spectral density of the next characteristic channel at each spectral line. Similarly, after completing the upper limit calculation of the power spectral density of all characteristic channels at one spectral line, the upper limit calculation of the power spectral density of all characteristic channels at the next spectral line can be continued.

[0099] The first calculation process is the normal tolerance upper limit method, as follows:

[0100] First calculate as follows:

[0101]

[0102] in, Num is the number of feature channels p p The mean and root mean square of the PSD estimated by the characteristic samples at the spectral line k, Z represents the number of characteristic channels p p The normal distribution of the square root of the PSD estimate of the characteristic samples, Z β is the β penalty point of the normal distribution Z, The non-centrality is The degrees of freedom are Q p -1 non-central t distribution, α is the preset confidence level of the interval estimate, and β is the preset probability;

[0103] based on Calculate the upper limit of the power spectrum density of the characteristic channel p at the spectrum line k

[0104]

[0105] Get the upper limit of the power spectrum density of the characteristic channel p at the spectrum line k

[0106] The second calculation process is the Johnson method, as follows:

[0107] First, based on the Johnson system distribution function, the Num of the feature channel p is p The square root of the PSD estimate of the characteristic samples is converted from a non-normal distribution at the spectral line k to a normal distribution:

[0108] z=γ+ηK i (x;λ,ε)

[0109]

[0110] Among them, x represents the Num of feature channel p p The square root of the PSD estimate of the characteristic sample is non-normally distributed at spectral line k, z represents the standard normal distribution obtained by transformation, γ, η, λ, ε are Johnson system parameters, and the value of i ranges from 1 to 3, which is determined according to the characteristics of the non-normal distribution x;

[0111] Under the preset probability β, solve the quantile F' of the corresponding standard normal distribution z, and based on the inverse of the Johnson system distribution function, the upper limit of the probability β non-normal data can be obtained as follows:

[0112]

[0113] Where μ = 0, σ = 1, z β represents the β penalty point of the standard normal distribution z;

[0114] based on Calculate the upper limit of the power spectral density of characteristic channel p at spectral line k as follows:

[0115]

[0116] Get the upper limit of the power spectrum density of the characteristic channel p at the spectrum line k

[0117] (4) Taking spectrum line k as an example, obtain the upper limits of the power spectrum density of P characteristic channels at spectrum line k, and select the maximum value as the time-varying load induction spectrum value at spectrum line k. Similarly, determine the induction spectrum value of each spectrum line to obtain the time-varying vibration acceleration induction spectrum of the axle box area.

[0118] (5) In step S5, the time-varying vibration acceleration summary spectrum is used as the fatigue analysis load spectrum of the primary spring of the railway vehicle, and the stress power spectrum density G is solved through the finite element model of the primary spring structure of the railway vehicle. σ (f), based on the stress power spectrum density G σ (f) Determine the service life of the primary steel spring of a railway vehicle. The structure of the primary steel spring of a railway vehicle is as follows: Figure 2 shown.

[0119] Solve the stress power spectrum density G σ (f) as follows:

[0120]

[0121] Where H is the frequency response function of the spring stress response due to the axlebox acceleration excitation, f = k * Δf, where Δf depends on the number of analysis points during PSD estimation, and ∑ is the summation formula, which refers to the sum of the stresses generated by the axlebox lateral, longitudinal, and vertical acceleration spectrum on the primary spring.

[0122] The service life of the primary steel spring of a railway vehicle is calculated as follows:

[0123] Based on the stress power spectrum density G σ (f) Based on the Dirlik rainflow amplitude probability density function, the fatigue cumulative damage D is solved:

[0124]

[0125] Where S is the stress amplitude, P(S) is the probability density estimation function of Dirlik rain flow amplitude, according to G σ (f) is obtained, T is the load duration, v p is the peak crossing times, C and m are the material SN curves;

[0126] When the fatigue cumulative damage D=1, the structure fails due to fatigue. The service life of the primary spring of the railway vehicle is calculated based on the fatigue cumulative damage D:

[0127] T life =T / D

[0128] Among them, T life It is the service life of the primary steel spring of rail vehicles.

[0129] The present invention collects multiple time-varying vibration load data from the axle box area as test data; combines the multiple time-varying vibration load data from the axle box area to form new characteristic channel data; divides the characteristic channels into characteristic samples, estimates the distribution characteristics of the square root values ​​based on the PSD of the characteristic samples, calculates the power spectral density tolerance upper limit of each characteristic channel at each spectral line using the normal tolerance upper limit method and the Johnson method, and summarizes the time-varying vibration load summary spectrum; and predicts the service life of the primary steel spring of a rail vehicle based on the time-varying vibration load summary spectrum. Compared with the existing technology, the present invention quantitatively expresses the energy and frequency domain distribution characteristics of the non-normal time-varying vibration load during the service life of the rail vehicle, thereby improving the prediction accuracy of the service life of the primary steel spring of the rail vehicle.

[0130] The above describes in detail the preferred embodiments of the present invention. It should be understood that those skilled in the art can make numerous modifications and variations based on the concepts of the present invention without inventive effort. Therefore, any technical solutions that can be derived by those skilled in the art through logical analysis, reasoning, or limited experimentation based on the concepts of the present invention and the prior art should be within the scope of protection defined by the claims.

Claims

1. A method for predicting the service life of a primary spring of a railway vehicle under time-varying loads, characterized in that: The following steps are involved: S1. Test the rail vehicle and collect the time-varying vibration acceleration of the axle box area to obtain an original data set. The original data set includes M×N test data, where M represents the number of vibration acceleration measurement points in the axle box area, N represents the number of tests on the rail vehicle, and X(m,n) represents the test data corresponding to measurement point m during the nth test. The length of the test data is length, where M>1, N>1, m=1, 2, ..., M, and n=1, 2, ..., N. S2. Traverse the test data in the original data set. If two test data satisfy the preset merging conditions, merge them into a new test data. Repeat this step until there is no test data that meets the merging conditions, and obtain a channel data set. The channel data set includes P feature channels. represents the qth test data under feature channel p, p = 1, 2, ..., P, q = 1, 2, ..., N P , N P Indicates the number of test data for feature channel p, N P >0, the capacity Q of feature channel p P =N P ×length; S3, traverse the feature channels in the channel data set, divide the feature channels into multiple feature samples according to the preset step size Δw, perform PSD estimation on each feature sample, and calculate the upper limit of the power spectrum density of each feature channel at each spectral line, where, Represents the upper limit of the power spectral density of characteristic channel p at spectral line k; S4, select the maximum value of each spectral line in all characteristic channels as the inductive spectrum value, That is, the time-varying vibration acceleration summary spectrum of the axle box area is obtained, where represents the inductive spectrum value at spectral line k; S5. Using the time-varying vibration acceleration induction spectrum as the load spectrum for fatigue analysis of the primary spring of a railway vehicle, the stress power spectrum density G is solved through the finite element model of the primary spring structure of the railway vehicle. σ (f), based on the stress power spectrum density G σ (f) Determine the service life of the primary steel spring of a railway vehicle.

2. The method for predicting the service life of a primary spring of a railway vehicle under time-varying loads according to claim 1, characterized in that: Step S2 is specifically as follows: S21. Calculate the mean and variance of each test data respectively; S22. Obtain two test data sets from the original data set, denoted as test data a and test data b. If both meet the pre-set merging condition, proceed to step S23. If not, repeat step S22 until no two test data sets in the original data set meet the merging condition. S23: Delete the test data a and the test data b from the original data set, put the new test data obtained by merging the two back into the original data set, and execute step S22.

3. The method for predicting the service life of a primary spring of a railway vehicle under time-varying loads according to claim 2, characterized in that: In step S21, the calculation process of the mean and variance of the test data X(m,n) is as follows: Split the test data X(m,n) of length length into L parts, where L is the preset parameter value, and obtain the data set {X(m,n)1, X(m,n)2, ..., X(m,n) L }, calculate X(m,n)1, X(m,n)2, ..., X(m,n) respectively. L The root mean square of the data set {Rms(m,n)1, Rms(m,n)2, …, Rms(m,n) L }, the data set {Rms(m,n)1, Rms(m,n)2, ..., Rms(m,n) L The mean and variance of} are used as the mean of the test data X(m,n) and variance S 2 (m,n).

4. The method for predicting the service life of a primary spring of a railway vehicle under time-varying loads according to claim 3, characterized in that: Step S22 is specifically as follows: Get two test data from the original data set, record them as test data a and test data b, and construct statistics F(a,b) and t(a,b): Among them, S 2 (a) and S 2 (b) represents the variance of test data a and test data b, respectively. and Represent the means of test data a and test data b respectively. If F(a, b) and t(a, b) meet the preset merging conditions, step S23 is executed. If not, step S22 is repeated until any two test data in the original data set do not meet the merging conditions. The merging conditions are: Among them, α' is the preset confidence level.

5. The method for predicting the service life of a primary spring of a railway vehicle under time-varying loads according to claim 4, characterized in that: In step S23, the length of the new test data is the sum of the lengths of test data a and test data b, length(a) and length(b). and The average value of is the mean of the new test data, S 2 (a) and S 2 The mean of (b) is the variance of the new test data.

6. The method for predicting the service life of a primary spring of a railway vehicle under time-varying loads according to claim 1, characterized in that: In step S3, the process of determining the upper limit of the power spectrum density of the characteristic channel p at the spectrum line k is as follows: S31, according to the preset step size Δw, the capacity is Q P The feature channel p is divided into Num p feature samples, Num p =(Q P / Δw); S32, performing power spectrum density estimation on each feature sample respectively, and calculating the square root of the PSD estimation of each feature sample; S33, at the spectral line k, if the Num of the characteristic channel p p If the PSD estimated square root of the characteristic samples conforms to the normal distribution, then the first calculation process is used to calculate If it conforms to the non-normal distribution, it is calculated according to the preset second calculation process 7. The method for predicting the service life of a primary spring of a railway vehicle under time-varying loads according to claim 6, characterized in that: The first calculation process is as follows: First calculate as follows: in, Num is the number of feature channels p p The mean and root mean square of the PSD estimated by the characteristic samples at the spectral line k, Z represents the number of characteristic channels p p The normal distribution of the square root of the PSD estimate of the characteristic samples, Z β is the β penalty point of the normal distribution Z, The non-centrality is The degrees of freedom are Q p -1 non-central t distribution, α is the preset confidence level of the interval estimate, and β is the preset probability; based on Calculate the upper limit of the power spectrum density of the characteristic channel p at the spectrum line k Get the upper limit of the power spectrum density of the characteristic channel p at the spectrum line k 8. The method for predicting the service life of a primary spring of a railway vehicle under time-varying loads according to claim 6, characterized in that: The second calculation process is as follows: First, based on the Johnson system distribution function, the Num of the feature channel p is p The square root of the PSD estimate of the characteristic samples is converted from a non-normal distribution at the spectral line k to a normal distribution: z=γ+ηK i (x;λ,ε) Among them, x represents the Num of feature channel p p The square root of the PSD estimate of the characteristic sample is non-normally distributed at spectral line k, z represents the standard normal distribution obtained by transformation, γ, η, λ, ε are Johnson system parameters, and the value of i ranges from 1 to 3, which is determined according to the characteristics of the non-normal distribution x; Under the preset probability β, solve the quantile F' of the corresponding standard normal distribution z, and based on the inverse of the Johnson system distribution function, the upper limit of the probability β non-normal data can be obtained as follows: Where μ = 0, σ = 1, z β represents the β penalty point of the standard normal distribution z; based on Calculate the upper limit of the power spectral density of characteristic channel p at spectral line k as follows: Get the upper limit of the power spectrum density of the characteristic channel p at the spectrum line k 9. The method for predicting the service life of a primary spring of a railway vehicle under time-varying loads according to claim 1, characterized in that: In step S5, the stress power spectrum density G is calculated. σ (f) as follows: Where H is the frequency response function of the spring stress response due to the axlebox acceleration excitation, f = k * Δf, where Δf depends on the number of analysis points during PSD estimation, and ∑ is the summation formula, which refers to the sum of the stresses generated by the axlebox's lateral, longitudinal, and vertical acceleration spectra on the primary spring.

10. The method for predicting the service life of a primary spring of a railway vehicle under time-varying loads according to claim 1, characterized in that: In step S5, the service life of the primary steel spring of the railway vehicle is calculated as follows: Based on the stress power spectrum density G σ (f) Based on the Dirlik rainflow amplitude probability density function, the fatigue cumulative damage D is solved: Where S is the stress amplitude, P(S) is the probability density estimation function of Dirlik rain flow amplitude, according to G σ (f) is obtained, T is the load duration, v p is the peak crossing times, C and m are the material SN curves; When the fatigue cumulative damage D=1, the structure fails due to fatigue. The service life of the primary spring of the railway vehicle is calculated based on the fatigue cumulative damage D: T life =T / D Among them, T life It is the service life of the primary steel spring of rail vehicles.

Images