Method for determining reliability of chassis bearing components

By using big data-driven methods, rainflow counting and Weibull distribution fitting, combined with Wiener stochastic processes and Monte Carlo methods, the error problem in predicting the fatigue reliability of chassis load-bearing components was solved, and more accurate damage patterns and failure time predictions were achieved.

CN114792054BActive Publication Date: 2024-11-12DONGFENG MOTOR GRP
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210473842.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-04-29
Publication Date
2024-11-12
Estimated Expiration
2042-04-29

AI Technical Summary

Technical Problem

Existing technologies have significant errors in predicting the fatigue reliability of chassis load-bearing components and fail to accurately account for the differences in damage between products manufactured in the same batch. In particular, they are not accurate enough in predicting failure phenomena under complex working conditions.

Method used

By acquiring long-period measured dynamic stress data, using rainflow counting and Weibull distribution fitting, combined with swarm intelligence optimization algorithm and Wiener stochastic process, a damage equivalent transformation model based on big data is established to calculate the cumulative damage and reliability of chassis load-bearing components. The Markov chain-Monte Carlo method is used for parameter estimation.

Benefits of technology

It improves the accuracy of reliability calculations for chassis load-bearing components, enabling more precise prediction of damage patterns and failure times throughout their entire lifecycle, and reduces calculation errors.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114792054B_ABST
    Figure CN114792054B_ABST
Patent Text Reader

Abstract

The invention discloses a method for determining the reliability of chassis bearing parts based on big data, comprising the following steps: (1) obtaining dynamic long-cycle dynamic stress data of vehicle chassis bearing parts, and obtaining the amplitude of the i-level dynamic stress spectrum by rain flow counting; (2) using a swarm intelligence algorithm to perform parameter fitting on the amplitude spectrum of the measured data of each day, obtaining the estimated value of the amplitude spectrum distribution parameter, and further calculating the expected cumulative damage and variance of the bearing parts; (3) dividing the measured data into a finite number of mileage intervals, regarding the multi-day data contained in the same mileage interval as a multi-level loading process, and using a nonlinear damage conversion model to obtain the equivalent damage statistics of each mileage interval; (4) regarding the cumulative damage accumulation of the bearing parts over the entire life cycle as a Wiener random process, taking the equivalent damage statistics of each mileage interval as the observed value, and using the Monte Carlo-Markov chain method to obtain the probability distribution and estimated value of the unknown parameters, and further obtaining the law of the reliability of the bearing parts over the entire life cycle based on the measured big data with the mileage. The invention improves the calculation accuracy of the reliability by equivalent load conversion and subjecting the cumulative damage increment to the Wiener random process.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of fatigue reliability of automobile parts, and in particular to a method for determining the reliability of chassis bearing parts based on big data. Background Art

[0002] At present, the commonly used method for determining the fatigue reliability of passenger car chassis bearing parts in engineering is to obtain the measured dynamic stress load spectrum under specific working conditions, calculate it based on the linear cumulative damage law, divide the load spectrum into several levels by rain flow counting, and perform maximum likelihood estimation on the amplitude spectrum parameters that obey the Weibull distribution by the least squares method. These are all based on the cumulative damage process as a deterministic process, and there are huge errors in the prediction of failure phenomena of components subjected to complex working conditions.

[0003] In the research direction of damage uncertainty, it is common to regard fatigue life under different stress levels as independent random variables and describe them with the characteristic quantity distribution function of the random variables. However, there is no very accurate solution for the calculation of cumulative damage, and the difference in damage amount between chassis bearing parts produced in the same batch is not taken into account when calculating the cumulative damage. Summary of the invention

[0004] The purpose of the present invention is to overcome the deficiencies of the above-mentioned prior art and provide a method for determining the reliability of chassis bearing components based on big data drive, thereby improving the calculation accuracy of the reliability of chassis bearing components.

[0005] To achieve the above object, the present invention provides a method for determining the reliability of a chassis bearing component, comprising the following steps:

[0006] (1) Obtain the long-term measured dynamic stress data of the vehicle chassis bearing parts and obtain the dynamic stress spectrum. According to the stress amplitude of the dynamic stress spectrum, the measured data of each day are counted for rain flow. The data are divided into i parts according to the maximum stress amplitude to obtain a multi-level dynamic stress spectrum.

[0007] (2) Assuming that the amplitude distribution of the i-level dynamic stress spectrum of each day obeys the Weibull distribution, the Weibull distribution parameters are fitted by the swarm intelligence optimization algorithm, and the expectation and variance of the cumulative damage of one day are obtained according to the Weibull distribution parameters;

[0008] (3) The measured data are divided into multiple mileage intervals, each of which contains measured data from multiple days. Within the same mileage interval, the load effects of multiple days are regarded as a multi-stage loading process. The equivalent cumulative damage statistics of each mileage interval are obtained using the damage equivalent conversion model.

[0009] (4) Assuming that the cumulative damage increment in each mileage interval obeys the Wiener random process, the probability density function of the failure time is obtained, and the reliability of the load-bearing component is obtained based on the probability density function of the failure time;

[0010] The unit damage is the damage caused by a specific load level acting once, the cumulative damage is the sum of the unit damage within a period of time or a mileage, the failure time is the time when the cumulative damage of the load-bearing component reaches the critical damage, and the reliability is the probability that the current service life of the load-bearing component is less than the failure time.

[0011] Furthermore, the long period must satisfy the requirement that the number of multi-level load spectrum cycles obtained after rain flow counting of the measured stress data of each day is ≥ 10 6 .

[0012] Furthermore, assuming that the dynamic stress amplitude of the data contained in each day obeys the Weibull distribution, the fitting method of the Weibull distribution parameters includes that when the root mean square error is minimized, the fitting result of the Weibull distribution parameters is optimal, and the root mean square error RMES is

[0013]

[0014] In the formula, Fw(x i ) is the cumulative distribution frequency of the Weibull distribution; F(x i ) is the actual cumulative frequency based on the stress spectrum;

[0015] The Weibull distribution function F(x) is

[0016]

[0017] Where a is the scale parameter of the Weibull distribution, b is the shape parameter of the Weibull distribution, and the Weibull distribution parameters include the size parameter and the shape parameter. i} is the stress amplitude of level i, and the corresponding number of cycles is n = {n1, n2, ..., n i}, xi is the i-th level stress amplitude, hereinafter referred to as the i-th level load, and ni is the number of cycles calculated by Weibull distribution.

[0018] Furthermore, the expected cumulative damage μ(D(n)) is

[0019] μ(D(n))=E(∑D i )=∑n i ·E(D i )

[0020] The variance of the cumulative damage σ 2 (D(n)) is

[0021] σ2 (D(n))=D(∑D i )=∑n i ·D(D i )

[0022] Where D(n) is the cumulative damage of load level i, D i is the unit damage of the i-th level load, E(Di) is the expectation of the cumulative damage of the i-th level load, D(Di) is the variance of the cumulative damage of the i-th level load, and ni is the number of stress cycles of the i-th level load.

[0023] Furthermore, the expected cumulative damage E(Di) of the i-th level load is

[0024]

[0025] The variance D(Di) of the cumulative damage of the i-th level load is

[0026]

[0027] In the formula, m and C are constants, which are related to the material properties, specimen form and loading method. That is, the SN curve of the material can be obtained through experiments or by searching existing literature. The expression is S m N = C, where S is the stress amplitude and N is the number of loadings when the material fails at this stress level, usually called fatigue life (generally characterized by the number of cycles). Γ is the Gamma function.

[0028] Furthermore, Di is calculated as follows:

[0029]

[0030] Where Di is the unit damage under stress level Si, and Ni is the fatigue life under stress level Si.

[0031] Furthermore, the long-period measured data are divided into multiple interval mileage sets, each interval mileage contains data from multiple days, and the load action process of multiple days in the same interval mileage is regarded as a multi-level loading process.

[0032] Furthermore, the damage conversion calculation formula for two consecutive days is as follows:

[0033]

[0034] Where σ j is the mean stress amplitude of the jth day within the interval mileage, that is, Nj is the stress amplitude σ j The corresponding fatigue life is obtained through the SN curve of the material. k is the kth level load on the jth day within the interval.

[0035] Furthermore, the expectation and variance of the cumulative damage on the jth day within the interval mileage are calculated, and the cumulative damage value of the interval mileage is obtained based on the damage conversion model.

[0036] Furthermore, the damage set of all interval mileages is obtained, which is recorded as follows: D W ={D w1 , D w2 , ..., D wm}, m is the number of equally divided interval miles, D wm is the cumulative damage value of the mth interval mileage.

[0037] Furthermore, each interval mileage damage value is an observation, and the total number of observations is , let the kth sample at t n-1 Time to t n The time increment is Δt kj , the cumulative damage increment is ΔDw kn =Dw kn -Dw kn-1 The damage increment follows the Wiener process, ΔDw~N(μΔt,σ 2 Δt).

[0038] Furthermore, let ν = σ -2 , μ and v are regarded as random variables, and a damage increment model with a random process is established. Let v~Gamma(β,α), and the probability density function of v is as follows:

[0039]

[0040] Furthermore, under the condition of given v, the conditional distribution probability density function of μ|υ~N(θ, λ|υ) under the condition of given v is:

[0041]

[0042] Furthermore, the joint probability density function of the two is:

[0043]

[0044] Furthermore, the probability density function of the damage increment is:

[0045]

[0046] Furthermore, according to the above calculation of cumulative damage, when the cumulative damage reaches or exceeds the specified failure critical value, the time T (i.e., the first arrival time) is defined as follows:

[0047] T=inf{t|x(t)≥ξ}

[0048] ξ is the critical failure value, which is determined by referring to engineering experience, and x(t) is the cumulative damage value.

[0049] Furthermore, the conditional distribution of v and μ specified when the Wiener process with random effects first arrives is:

[0050]

[0051] Furthermore, according to the probability density function of the damage increment and the conditional distribution of the Wiener process parameters, the calculation formula of the reliability can be obtained as follows:

[0052]

[0053] Further, the reliability R(t) calculation expression of the tested component is obtained:

[0054]

[0055] Where P(T>t) is the probability that the current usage time t is less than the expiration time T, f(t) is the probability density function of the current usage time t, f(t|μ, v) is the joint probability density function of t|μ and v, g(μ, v) is the joint probability density function of μ and v, T 2β is a T distribution with 2β degrees of freedom, and the cumulative damage increment ΔD(n)~N(μΔt,σ 2 Δt), μ is the drift parameter, σ is the diffusion parameter, v = σ -2 , v~Gamma(β,α), α is the shape parameter of Gamma distribution, β is the scale parameter of Gamma distribution, μ|v~N(θ,λ|v), ξ is the critical damage amount.

[0056] Furthermore, the Markov Chain Monte Carlo method (MCMC) is used to transform the prior distribution into the posterior distribution through the Bayesian formula to obtain the estimated parameters.

[0057] The beneficial effect of the present invention is to improve the calculation accuracy of the reliability of chassis bearing parts. The present invention obtains long-period measured dynamic stress data through unattended data acquisition equipment, counts rain flow for each day's measured data to obtain a multi-level spectrum, and the amplitude distribution of the measured data on the same day obeys the Weibull distribution. The distribution parameters are fitted using a swarm intelligence algorithm to calculate the cumulative damage statistics for each day: expectation and variance.

[0058] The long-period measured data are divided into multiple mileage intervals. The same mileage interval contains the measured data of multiple days. The load effects of multiple days in the same mileage interval are regarded as a multi-stage loading process. The nonlinear damage conversion model is used to obtain the equivalent damage statistics of each mileage interval.

[0059] The damage accumulation of the chassis load-bearing structure throughout its life cycle is regarded as a Wiener random process. The equivalent damage statistics of each mileage interval are taken as observations. MCMC is used to obtain the probability distribution and estimated values ​​of the unknown parameters, thereby obtaining the variation law of the reliability of the whole life cycle with mileage driven by measured big data. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 The figure is a flow chart of the reliability determination method of the present invention.

[0061] Figure 2 It is the strain-time variation curve of a certain measuring point of the load-bearing component.

[0062] Figure 3 .a)~3.d) are the cumulative damage change diagrams of measuring points 1~4 of the present invention.

[0063] Figure 4 .a)~4.d) are the Monte Carlo simulation convergence calculation results of α, β, θ, and λ.

[0064] Figure 5 This is the curve of reliability changing with mileage. DETAILED DESCRIPTION

[0065] The following specific implementations are used to further explain the technical solutions of the claims of the present invention in detail, so that those skilled in the art can more clearly understand the claims. The protection scope of the present invention is not limited to the following specific embodiments. The technical solutions of the claims of the present invention made by those skilled in the art but different from the following specific implementations are also within the protection scope of the present invention.

[0066] like Figures 1 to 5 As shown, a method for determining the reliability of a chassis bearing component comprises the following steps:

[0067] (1) In order to accurately evaluate the fatigue reliability of chassis bearing parts under user working conditions, strain sensors and data acquisition equipment are installed at key locations of chassis bearing parts. A large number of measured strain signals are obtained through long-term tracking tests with unattended equipment. Special data processing software is used to pre-process the signals to improve the signal-to-noise ratio and convert the electrical signals into stress values.

[0068] The dynamic stress data of each measuring point every day are counted by rain flow counting method to obtain the dynamic stress spectrum. The number of multi-level load spectrum cycles obtained after rain flow counting of the measured dynamic stress data of each day is ≥10 6 , which satisfies the sample size requirement in the law of large numbers.

[0069] The stress amplitude of each day's dynamic stress spectrum is divided into 16 equal parts to obtain a 16-level dynamic stress spectrum amplitude distribution. The rain flow counting results of the data of a certain day are selected and shown in Table 1. In this embodiment, i=16, and the stress amplitude is the difference between the maximum stress and the minimum stress in the dynamic stress spectrum. For the data within one day, the stress amplitude and the number of cycles are recorded as follows after rain flow counting: x={x1, x2....x 16}, the corresponding number of cycles is n = {n1, n2, ..., n 16}, where x is the stress spectrum amplitude and n is the corresponding number of cycles.

[0070] Table 1 Distribution of 16-level stress spectrum amplitudes of rain flow counts for a certain day

[0071]

[0072] (2) Assuming that the amplitude distribution of the 16-level dynamic stress spectrum of each day obeys the Weibull distribution, the Weibull distribution parameters are fitted by the swarm intelligence optimization algorithm and the 16-level dynamic stress spectrum amplitude distribution. The Weibull distribution function F(x) of the amplitude spectrum of the measured data of one day is:

[0073]

[0074] Where a is the scale parameter of the Weibull distribution, b is the shape parameter of the Weibull distribution, x = {x1, x2....x 16} is the 16-level stress amplitude spectrum.

[0075] The probability density function f(x) of the stress amplitude at level i is

[0076]

[0077] The fitting method of Weibull distribution parameters includes optimizing the fitting by using the swarm intelligence algorithm - Gray Wolf Algorithm, which can effectively improve the defects of data residual normality and variance homogeneity caused by the linearization of nonlinear distribution function by numerical method. The swarm intelligence algorithm directly optimizes the objective function in the solution space and will not change the nonlinear characteristics of the distribution function. The minimum root mean square error is used as the objective function. When the root mean square error is minimized, the fitting result of Weibull distribution parameters is optimal, as shown in Table 2. The root mean square error RMSE is

[0078]

[0079] In the formula, F W (x i ) is the cumulative distribution frequency of Weibull distribution; F(x i ) is the actual cumulative frequency based on the stress spectrum;

[0080] In this embodiment, the threshold value is set to 0 during the counting process. When the Weibull distribution is used to fit the stress spectrum amplitude, the unknown parameter c=0. The least square method and particle swarm algorithm are used to solve equation (1) respectively. The chi-square test is used as the criterion for judging the goodness of fit. In general, χ 2 The smaller the value, the smaller the difference between the theoretical distribution and the actual data distribution. In this embodiment, the test level α = 0.01 is used to obtain the critical value χ 2 The value is 7.042, and the calculated stress spectrum distribution parameters are consistent with χ 2 The values ​​are shown in Table 2, full load acceptance criteria.

[0081] Table 2 Weibull distribution parameter optimization fitting results

[0082]

[0083] The SN curve is commonly used in engineering to describe the direct relationship between fatigue life and external load (weak point stress). The formula is:

[0084] In the formula, m and C are constants, which are related to the material properties, specimen form and loading method. S represents the stress level, that is, X, and N is the number of cycles at this stress level.

[0085] The unit damage Di at a specific load level is:

[0086]

[0087] Where Di is the unit damage under stress level Si, Ni is the fatigue life under stress level Si, and C is a constant, where Ni and C can be obtained according to the SN curve library corresponding to the material.

[0088] Due to the randomness of the external load, the dynamic stress at the measuring point is also a random process. The stress amplitude obeys the Weibull distribution, and its high-order origin moment is:

[0089]

[0090] Where a and b are the scale parameter and shape parameter of the Weibull distribution, respectively, k is a constant greater than 1, and Γ(·) is the Gamma function, which is expressed as follows:

[0091]

[0092] Since Si (that is, xi) is a random variable, we can know that Di is also a random variable.

[0093] The expected cumulative damage E(Di) of the i-th level load is

[0094]

[0095] The variance D(Di) of the cumulative damage of the i-th level load is

[0096]

[0097] The damage of each unit follows an independent and identical distribution. Assuming that different rainflow cycles are independent of each other, under the action of random stress, with the number of stress cycles n as a parameter, {D(n); n = 1, 2, 3, ..., n i} is a stable independent incremental process. When n is large enough (≥10 6 ), according to the central limit theorem, the cumulative damage D(n) asymptotically obeys the normal distribution.

[0098] The expected cumulative damage μ(D(n)) obtained based on one day's measured data is:

[0099] μ(D(n))=E(∑D i )=∑n i ·E(D i )

[0100] The variance of the cumulative damage per day is σ 2 (D(n)) is

[0101] σ 2 (D(n))=D(∑D i )=∑n i ·D(D i )

[0102] Where D(n) is the cumulative damage of the i-th load, Di is the unit damage of the i-th load, E(Di) is the expectation of the cumulative damage of the i-th load, D(Di) is the variance of the cumulative damage of the i-th load, and n is the number of stress cycles corresponding to this load level.

[0103] (3) The long-term tracking measured data is divided into multiple mileage intervals. Each mileage interval contains data from multiple days. In the same mileage interval, the loading process of multiple days is regarded as a multi-level loading process. The unit damage conversion calculation formula for two adjacent days is as follows:

[0104]

[0105] Where σ j is the mean value of the stress amplitude on the jth day in the mileage interval, that is, Nj is the stress amplitude σ j The corresponding fatigue life is obtained through the SN curve of the material. According to step 3, the expectation and variance of the cumulative damage on the jth day are calculated. Then the cumulative damage value of the mileage interval is obtained. Thus, the damage set of a finite number of equal mileage intervals is obtained. It is recorded as follows: D w ={D w1, D w2 , ..., D wm}, m is the number of equally divided mileage intervals, D wm is the cumulative damage value of the mth mileage interval.

[0106] Since the data scale of this embodiment is as high as more than 50,000 kilometers, m in this embodiment is 10, that is, each mileage interval is 5000km of measured data. The accumulated damage of the four measuring points obtained after equivalent damage conversion is as shown in the attached figure. Figure 3 shown.

[0107] (4) Assuming that the cumulative damage increment of the load-bearing component throughout its life cycle obeys the Wiener random process, the probability density function of the failure time is obtained. Based on the probability density function of the failure time, the reliability of the load-bearing component is obtained. The failure time is the time when the cumulative damage of the load-bearing component reaches the critical damage, and the reliability is the probability that the current service life of the load-bearing component is less than the failure time.

[0108] For the long-term tracking data of user working conditions, it is often several months, and the actual mileage is tens of thousands of kilometers. The damage accumulation increment conforms to the Wiener random process. According to (3), a total of m observation samples are obtained, and the Kth sample is set at t n-1 Time to t n The time increment is Δt kj , the cumulative damage increment is ΔDw kn =Dw kn -Dw kn-1 , then the cumulative damage increment ΔD(n)~N(μΔt,σ 2 Δt), μ is the drift parameter, and σ is the diffusion parameter.

[0109] The increment distribution of the Wiener process only depends on the time difference, so the Wiener is a homogeneous independent increment process, the increments obey the normal distribution, and μ and σ can be constants or random variables.

[0110] Further considering the difference in degradation between products in the same batch, assuming that μ and σ are both random variables, let v = σ -2 , v~Gamma(β,α), α is the shape parameter of Gamma distribution, β is the scale parameter of Gamma distribution, then the probability density function of v is:

[0111]

[0112] Under the given v condition, μ|v~N(θ,λ|v), then the conditional distribution probability density function under the given v condition is:

[0113]

[0114] The joint probability density function of μ and v g(μ, v) is

[0115]

[0116] Assume that there are N observation samples in total, and let the i-th sample be at t j-1 Time to t j The time increment at a moment is Δt ij , the degradation increment is ΔDw kn According to the definition of Wiener process and the total probability formula of random variables, we can get ΔDw kn The probability density function is:

[0117]

[0118] Assuming the above model to be a Wiener process with random effects, the likelihood function of its unknown parameters is:

[0119]

[0120] When the performance degradation trajectory reaches or exceeds the specified critical failure value for the first time, this time is the first arrival time. Let ξ be the critical damage amount, then the mathematical definition of the failure time T is expressed as:

[0121] T=inf{t|X(t)≥ξ}

[0122] In the formula, the critical damage amount is taken as 0.7 according to engineering experience.

[0123] The joint probability density function f(t|μ, v) of the first arrival of the Wiener process with random effects t|μ and v is:

[0124]

[0125] Through the above derivation, the probability density function of the failure time T can be obtained as follows:

[0126]

[0127] The reliability R(t) of the bearing is

[0128]

[0129] Where P(T>t) is the probability that the current usage time t is less than the expiration time T, f(t) is the probability density function of the current usage time t, and T 2β is a T distribution with 2β degrees of freedom, and ξ is the critical damage amount.

[0130] The above formula contains multiple unknown parameters α, β, θ, λ, and its form is relatively complex, making it difficult to solve using the maximum likelihood method. Therefore, the Monte Carlo-Markov chain can be used to convert the prior distribution into the posterior distribution through the Bayesian formula to obtain parameter estimates, and subsequent calculations can be performed on this basis. At the same time, the autocorrelation is used to judge the convergence of the Monte Carlo simulation. Rapid convergence to 0 represents sampling convergence.

[0131] In this embodiment, the prior distribution is obtained by using the conjugate distribution method, assuming that the prior distribution of the unknown parameters is as follows:

[0132] θ~N(0.01, 0.001)

[0133] β~Gamma(0.003,0.001)

[0134] α~Gamma(0.0002,0.01)

[0135] λ~Gamma(0.0001, 0.01)

[0136] The MCMC algorithm was used to perform 10,000 samplings, and the Bayesian posterior distribution and estimated values ​​of the unknown parameters α, β, θ, and λ were obtained, as shown in Table 3.

[0137] Table 3 Estimated values ​​of model parameters α, β, θ, λ

[0138]

[0139] At the same time, the autocorrelation function of the above parameters is calculated to determine the convergence of MCMC sampling. If the autocorrelation function approaches 0 quickly, it can be considered that the sampling converges and the parameter estimation value is acceptable. The autocorrelation functions of the above four unknown parameters are shown in the appendix. Figure 4 shown.

[0140] Substituting the estimated values ​​of the four unknown parameters obtained in Table 3 into the reliability calculation formula, we can obtain the variation law of the reliability of the load-bearing structure over its entire life cycle with mileage.

Claims

1. A method for determining the reliability of chassis bearing components, characterized in that: The steps include: (1) Obtaining long-term measured data of dynamic stress of vehicle chassis bearing parts and converting it into a stress spectrum, processing the stress amplitude of each day's stress-time history data by rain flow counting method to obtain a multi-level spectrum; (2) Assuming that the amplitude of the dynamic stress data measured every day follows the Weibull distribution, the swarm intelligence algorithm is used to estimate the distribution parameters, and then the expectation and variance of the cumulative damage of one day are obtained; (3) The measured data are equally divided into multiple interval mileages, where each interval mileage contains measured data from multiple days. The strength degradation model is introduced, and the data within an interval mileage is regarded as multi-level loading. The statistics of equivalent damage at each interval mileage are calculated. (4) The damage accumulation of the load-bearing component during its entire life cycle is regarded as a Wiener process with random effects. The damage value at each interval mileage is used as the observation quantity. The probability distribution and estimated value of the unknown parameters of the Wiener process are obtained by the MCMC method, thereby obtaining the variation law of the reliability of the entire life cycle with time / mileage based on the measured dynamic stress data; In step 2), it is assumed that the dynamic stress amplitude within each day obeys the Weibull distribution. The fitting method of the Weibull distribution parameters includes that when the root mean square error is minimized, the fitting result of the Weibull distribution parameters is optimal. The root mean square error RMES is Among them, F W (x i ) is the cumulative distribution frequency of Weibull distribution; F(x i ) is the actual cumulative frequency based on the stress spectrum; The Weibull distribution function F(x) is Where a is the scale parameter of the Weibull distribution, b is the shape parameter of the Weibull distribution, and the Weibull distribution parameters include the size parameter and the shape parameter. i } is the stress amplitude of level i; The expected cumulative damage μ(D(n)) for one day is: μ(D(n))=E(∑D i )=∑n i ·E(D i ) The variance of the cumulative damage σ 2 (D(n)) is: σ 2 (D(n))=D(∑D i )=∑n i ·D(D i ) Where D(n) is the cumulative damage of the i-level load, Di is the unit damage of the i-level load, E(Di) is the expectation of the cumulative damage of the i-level load, D(Di) is the variance of the cumulative damage of the i-level load, n is the number of stress cycles after rainflow counting of one day's data, ni is the number of cycles corresponding to the i-level load inversely inferred using the Weibull distribution, that is, the function of converting the cumulative damage of one day into the number of rainflow cycles; In step 3), the measured data is divided into multiple mileage intervals, each mileage interval contains data from multiple days, and the multiple-day load action process contained in the same mileage interval is regarded as a multi-level loading process; The calculation formula for unit damage conversion between two consecutive days is as follows: Where σ j is the mean value of the stress amplitude on the jth day in the mileage interval, that is, Nj is the stress amplitude σ j The corresponding fatigue life, that is, the fatigue life of the material, is obtained through the SN curve of the material; according to step 3), the expectation and variance of the cumulative damage on the jth day are calculated, and then the cumulative damage value of the interval mileage is obtained, thereby obtaining a damage set of a finite number of equally spaced mileages, recorded as D W ={D w1 , D w2 , ..., D wm }, m is the number of equally divided interval miles, D wm is the cumulative damage value of the mth interval mileage; a total of m observation samples are obtained, and the kth sample is set at t n-1 Time to t n The time increment is Δt kj , the cumulative damage increment is ΔDw kn =Dw kn -Dw kn-1 , the damage increment follows the Wiener process, that is, the damage increment follows the normal distribution, ΔDw~N(μΔt,σ 2 Δt).

2. The method for determining the reliability of chassis bearing components according to claim 1, characterized in that: The long period must satisfy the requirement that the number of multi-level load spectrum cycles obtained after rain flow counting of the measured stress data of each day is ≥ 10 6 ; For the data within one day, the stress amplitude and the number of cycles are as follows after rain flow statistics: x={x1,x2…x i }, the corresponding number of cycles is n={n1,n2,…,n i } Where X is the stress spectrum amplitude and N is the corresponding number of cycles.

3. The method for determining the reliability of chassis bearing components according to claim 2, characterized in that: The expected cumulative damage E(Di) of the i-th level load is The variance D(Di) of the cumulative damage of the i-th level load is Where m and C are constants, which are related to the material properties, specimen form and loading method, that is, the SN curve S of the material m N = C, where S is the stress amplitude, N is the number of load cycles when the material fails, and Γ is the Gamma function; Where Di is calculated as follows: Where Di is the unit damage under stress level Si (corresponding to xi), Ni is the fatigue life under stress level xi, and C is a constant, where Ni and C can be obtained according to the SN curve library corresponding to the material.

4. The method for determining the reliability of chassis bearing components according to claim 3, characterized in that: The damage increment follows a normal distribution, and the distribution parameters μ and σ are random variables. Let ν = σ -2 , and the probability density function of ν~Gamma(β,α),ν is as follows: Under the condition of given ν, μ|v~N(θ,λ|v), the conditional distribution probability density function under the premise of given ν is: The joint probability density function of the two is: The probability density function of damage increment is:

5. The method for determining the reliability of chassis bearing components according to claim 4, characterized in that: When the cumulative damage reaches or exceeds the specified failure critical value, the time T, that is, the first arrival time, is defined as follows: T=inf{t|x(t)≥ξ} ξ is the critical failure value, which is determined by referring to engineering experience. The conditional distribution of ν and μ specified when the Wiener process with random effects first arrives is:

6. The method for determining the reliability of chassis bearing components according to claim 5, characterized in that: The joint density probability distribution function is obtained as The reliability R(t) of the bearing component is further obtained as Where P(T>t) is the probability that the current usage time t is less than the expiration time T, f(t) is the probability density function of the current usage time t, f(t|μ,ν) is the joint probability density function of t|μ and ν, g(μ,ν) is the joint probability density function of μ and ν, T 2β is a T distribution with 2β degrees of freedom, and the cumulative damage increment ΔD(m)~N(μΔt,σ 2 Δt), ΔD(m)=D wi -D wi-1 , μ is the drift parameter, σ is the diffusion parameter, ν = σ -2 , ν~Gamma(β,α), α is the shape parameter of Gamma distribution, β is the scale parameter of Gamma distribution, μ|ν~N(θ,λ|ν), ξ is the critical damage amount.

7. The method for determining the reliability of chassis bearing components according to any one of claims 3 to 5, characterized in that: The distribution parameters are converted into posterior distribution by Markov Chain Monte Carlo (MCMC) method through Bayesian formula to obtain parameter estimates, and then the variation law of life cycle reliability with mileage is obtained.