A Kriging model-based method for evaluating the kinematic reliability of a gun's coordinated electromechanical and hydraulic system

Abstract

The present invention belongs to the field of weapon technology and relates to a method for evaluating the kinematic reliability of a coordinated electromechanical and hydraulic system of an artillery gun based on a Kriging model. Points are selected through a Kriging method, a MCMC method, a K-means method, and a sampling distance restriction method. A kinematic displacement accuracy evaluation index and a feed line angle accuracy evaluation index are designed, and true response values ​​of the two evaluation indexes are obtained. The present invention designs a coordinator kinematic accuracy evaluation index, considers the importance of posture deviations in different directions to the reliability of the electromechanical and hydraulic system, and avoids information redundancy caused by the concentration of training samples in the point selection rules. The K-means method can realize parallel computing of multiple computers, and a kinematic reliability accuracy evaluation method for the electromechanical and hydraulic system of the artillery coordinator is established, so that high computing efficiency is achieved while ensuring the accuracy of reliability calculation.

Description

Technical Field

[0001] The present invention belongs to the field of weapon technology, and in particular relates to a motion reliability assessment technology for an electro-hydraulic system of a gun coordinator. Background Art

[0002] Artillery, as a conventional weapon, continues to play a vital role in modern warfare. The coordinator's electro-mechanical and hydraulic systems, as a crucial mechanical substructure of self-propelled artillery, play a crucial role. If the coordinator's positioning error exceeds the specified coordination accuracy, the artillery's firing mission can be impacted. Research on the kinematic reliability of artillery coordinators is crucial for improving the survivability of self-propelled artillery on the battlefield.

[0003] The kinematic reliability calculation of the coordinator's electro-hydraulic system primarily involves constructing a performance function and then solving for the structural reliability using primary reliability, secondary reliability, or Monte Carlo methods. To achieve high-precision reliability, the Monte Carlo method is generally used. However, the numerous parameters that influence the kinematic accuracy of the coordinator's electro-hydraulic system, and their influencing relationships, are complex. This results in a high computational complexity and significant labor and resource consumption when using the Monte Carlo method for reliability calculations.

[0004] Currently, to solve the above problems, a surrogate model is usually used to approximate the performance function, and then the reliability is solved through the Monte Carlo method. However, traditional sampling methods make it difficult to obtain good sample points, resulting in poor accuracy of the constructed model. The accuracy of the surrogate model can be improved by increasing the number of sample points, but this will increase the test cost. The calculation accuracy of the reliability depends largely on the accuracy of the surrogate model. Therefore, obtaining a high-precision surrogate model while ensuring a low test cost is very important for calculating the kinematic reliability of the coordinator electromechanical hydraulic system. At present, there is little research in China on the kinematic reliability evaluation of the coordinator electromechanical hydraulic system based on the Kriging model. Summary of the Invention

[0005] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a method for evaluating the kinematic reliability of a coordinated electromechanical and hydraulic system of a gun based on a Kriging model.

[0006] In order to achieve the purpose of the present invention, the present invention is implemented by adopting the following technical solutions.

[0007] A method for evaluating the kinematic reliability of a coordinated electromechanical and hydraulic system of a gun based on a Kriging model, characterized by comprising the following steps:

[0008] S1. Determine N random variables and their distribution types based on the relevant parameters that affect the coordinator's motion;

[0009] S2. Use Latin hypercube sampling to extract N0 training samples that meet the distribution type for each random variable, and form the initial training sample set in the standard normal space. And get the corresponding true response value Get the initial true response value matrix;

[0010] S3, use Monte Carlo random sampling method to extract N random variables for each MC A random sample that meets the respective distribution type is used as the corresponding candidate sample point;

[0011] S4, based on the initial training sample set S DoE The sample data and true response values ​​in are used to build the Kriging model;

[0012] S5, using the Markov Chain Monte Carlo method to extract candidate sample points with U(x)<2 from the candidate sample points generated in step S3, and using the k-means clustering algorithm to divide the extracted candidate sample points into K groups;

[0013] S6. Select K best training sample points from the K groups of candidate sample points formed in step S5 according to the method of selecting sample points. And obtain the true response value of each optimal training sample point

[0014] S7, determine whether the K best training sample points meet the current S DoE The distance criterion of the training sample points in :

[0015] If not, remove the non-compliant points from the corresponding K-means group. After removing the non-compliant points, go to step S6 and reselect based on the distance criterion.

[0016] If satisfied, proceed to step S8;

[0017] The distance criterion is expressed by a point selection formula, which is:

[0018] max{R(x new ,x DoE ;θ)}≥[r],

[0019] Where x DoE Indicates the training points that constitute the current Kriging model, [r] is usually between 0.95 and 1;

[0020] S8, K best training sample points and the true response value of each optimal training sample point Add them to the initial training sample set and the initial true response value matrix respectively to obtain the updated training sample set and true response value matrix, and update the Kriging model;

[0021] S9. Calculate the estimated failure probability P of the coordinator motion reliability based on the current Kriging model and the training sample set. f , error coefficient e ε and coefficient of variation δP f ;

[0022] S10, judgment error coefficient e ε Is it satisfied Requirements:

[0023] If not satisfied, return to step S5;

[0024] If satisfied, execute step S11;

[0025] Where, e ε is the error coefficient, [e ε ] is the threshold value, which is determined according to the actual accuracy requirement;

[0026] S11. Determine the coefficient of variation δP f Whether δP is satisfied f Requirements for ≤[δ]:

[0027] If it is not satisfied, the number of random sample points is increased by Monte Carlo method;

[0028] If satisfied, proceed to step S12;

[0029] Where [δ] is the threshold;

[0030] S12. Calculate and output the coordinator motion reliability results: failure probability Coefficient of variation and error coefficient e ε .

[0031] As a preferred solution of the present invention, the relevant parameters include the manufacturing error of the key dimensions of the coordinator, the clearance error of the key hole-axis fit, the projectile mass error, the oil pump speed, the leakage rate of the coordinating cylinder and the leakage rate of the flipping cylinder.

[0032] As a further solution of the present invention, the key dimensions of the coordinator include the dimensions of the coordinating arm and the dimensions of the flipping mechanism.

[0033] As a preferred embodiment of the present invention, the true response value The method for obtaining , comprising the following steps:

[0034] S41. Use Latin hypercube sampling to extract N0 training samples that meet the distribution type for each random variable, and form the initial training sample set in the standard normal space.

[0035] S42, the coordinator motion displacement accuracy γ D Expressed as:

[0036]

[0037] Among them, η X , η Y , η Z Represent the reliability error sensitivity coefficients corresponding to the X, Y and Z axes respectively;

[0038] S43, the coordinator motion feed line angle accuracy γ ε Expressed as:

[0039]

[0040] Among them, α, β, are the angular deviations of the three directions X, Y, and Z of the bullet feeding line, respectively, α ,λ β 、 They are the three angle deviations α, β, The corresponding sensitivity coefficient;

[0041] S44, the coordinator motion displacement accuracy γ D and the coordinator motion feed line angle accuracy γ ε As a standard value, a function function of the coordinator motion accuracy is established, and the expression of the function function is:

[0042] y(x)=ξ-|γ-γ′|,

[0043] Where ξ is the allowable error, which can be determined according to the accuracy requirements, and γ′ is the motion displacement γ D Or the angle of the moving bullet line γ ε The ideal value of γ is the displacement accuracy of the projectile. D Or the angle accuracy of the moving bullet feeding line γ ε .

[0044] As a preferred embodiment of the present invention, the method for establishing the Kriging model comprises the following steps:

[0045] In the S51 and Kriging models, the performance function G(x) consists of two parts: a deterministic part and a random part, namely:

[0046] G(x)=Γ(β,x)+z(x)=gT (x)β+z(x),

[0047] S52, given an initial training sample set S DoE , basis function g(x) and correlation coefficient function R(x i ,x j ;θ), Kriging predicted value and Kriging predicted values and Kriging equation The expression of is completely dependent on the parameter θ. The widely accepted maximum likelihood method is used to determine the optimal value of θ.

[0048] As a preferred embodiment of the present invention, the specific operation steps of step S5 are as follows:

[0049] In the MCMC method (Markov chain), if the sequence x0, x1, x2, ..., x K+1 Produced in {x k+1 ∣x k}, and the conditional probability satisfies: P(x k+1 =x|x k ,x k-1 ,…)=P(x k+1 =x|x k ), x k+1 The distribution of depends only on the current state x k , does not depend on the historical state, then x0, x1, x2, ..., x K+1 is a Markov chain. Where x0 is the given initial condition, according to the conditional probability distribution, the state x k+1 The desired probabilistic properties are conditioned by the previous state x k Given;

[0050] When k approaches infinity, x k It is independent of the initial value, that is, when k increases, the random vectors in the Markov chain will converge to the steady-state distribution π(x). At this time, the Markov chain is considered to be in a convergent state. When the Markov chain does not reach the convergent state, it is considered that the probability density distribution of each state cannot reach a stable distribution. When evaluating, the first M sampling values ​​that have not reached the steady-state distribution should be removed, and the last nM sampling values ​​should be used for evaluation: The estimated value of E[F] The variance is: The estimated value of V(F) is From this, a Markov chain can be constructed;

[0051] Using the K-means clustering algorithm, based on the principle of minimum Euclidean distance between the data sample in each group and the cluster center, under the condition of a given cluster number K value, n sample data are divided into K groups through multiple iterations. The calculation formula of Euclidean distance can be expressed as:

[0052]

[0053] Where: E is the sum of the mean square errors of all objects in the cluster sample; K is the number of clusters; xj is a point in the cluster; mi is the mean of cluster Ci;

[0054] According to the calculated Euclidean distance from the data sample to the cluster center, the samples are classified according to the minimum distance principle, and the n sample data are divided into K groups through multiple iterations.

[0055] As a preferred solution of the present invention, the k-means clustering algorithm includes the following steps:

[0056] S71, determine the number of clusters K, input the data vector group X, which contains data vectors X1, X2, X3, ..., X n ;

[0057] S72. Randomly select K vectors from the n data vectors as initial cluster centers;

[0058] S73, calculating the Euclidean distance of each vector to the K initial cluster centers, and assigning the vector to the cluster where the initial cluster center with the smallest distance value is located;

[0059] S74. If the data vector in the cluster changes, recalculate the centers of the K clusters;

[0060] S75. Repeat steps S73 to S74 until the centers of the clusters no longer change or the center distance is less than a specified threshold.

[0061] As a preferred embodiment of the present invention, the method for selecting sample points includes the following steps:

[0062] S81, calculating the learning function value of each candidate sample point corresponding to each random variable according to the initial Kriging model;

[0063] S82. Find the minimum learning function value corresponding to each group of random variables in the K groups;

[0064] S83. Determine whether the minimum learning function value corresponding to each group of random variables is less than 2:

[0065] If the minimum learning function value is less than 2, then the K candidate sample points corresponding to the minimum learning function value in each group of candidate sample points corresponding to each random variable are used as the K best training sample points of the random variable;

[0066] If the minimum learning function value is ≥ 2, continue to judge.

[0067] As a preferred embodiment of the present invention, the specific process of establishing the point selection formula includes the following steps:

[0068] S91, use the threshold [L] (the threshold [L] is determined by experience according to the parameters and dimensions) to directly new And the training sample set x that constitutes the Kriging model j (j=1,…,N) is limited in spacing, that is:

[0069]

[0070] Based on the above formula, we get a new A hypersphere with [L] as the radius and [L] as the center;

[0071] S92. Convert the constraint of the distance between two points into a correlation function problem, and restrict the correlation function between training samples, that is:

[0072] max{R(x new ,x DoE ;θ)}≥[r],

[0073] Where x DoE Indicates the training points that constitute the current Kriging model. [r] is usually between 0.95 and 1.

[0074] As a preferred solution of the present invention, the failure probability P f :

[0075] The coefficient of variation

[0076] The error coefficient e ε :

[0077] Where x MC,n (n=1,…,N MC ) is derived from the probability density function f X (x) is an independent and identically distributed random sample point, N MC is the number of sample points.

[0078] Beneficial effects

[0079] Compared with the prior art, the present invention has the following advantages:

[0080] The present invention designs a coordinator motion accuracy evaluation index, takes into account the importance of posture deviations in different directions to the reliability of the electromechanical and hydraulic system, and uses point selection rules to avoid information redundancy caused by the concentration of training samples. The K-means method can realize parallel calculation of multiple computers, and establishes a motion reliability accuracy evaluation method for the electromechanical and hydraulic system of the artillery coordinator, so that it has higher computing efficiency while ensuring the accuracy of reliability calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] Figure 1 is a basic flow chart of the method of the present invention;

[0082] Figure 2 This is a structural diagram of the coordinator of the present invention;

[0083] Figure 3 This is a posture comparison diagram of the coordinator described in the present invention;

[0084] Figure 4 This is a schematic diagram of the projectile center of mass displacement error of the present invention;

[0085] Figure 5 This is a schematic diagram of the angle error of the bullet feeding line described in the present invention. DETAILED DESCRIPTION

[0086] The present invention will be further described with reference to the accompanying drawings and embodiments.

[0087] As an embodiment of the present invention, Figures 1 to 5 As shown, a method for evaluating the motion reliability of a coordinated electromechanical and hydraulic system of a gun based on the Kriging model includes the following steps:

[0088] S1. According to the relevant parameters affecting the coordinator movement: the manufacturing error of the key dimensions of the coordinator, the clearance error of the key hole-shaft fit, the projectile mass error, the oil pump speed, the leakage rate of the coordination cylinder, and the leakage rate of the flip cylinder, N random variables and distribution types are determined according to these parameters; wherein the manufacturing error of the key dimensions of the coordinator includes the dimensions of the coordination arm and the flip mechanism, such as Figure 2 As shown;

[0089] S2. Use the LHS method to construct the initial training sample set in the standard normal space The coordinator motion displacement accuracy γ D Expressed as:

[0090]

[0091] Where η X , η Y , η ZThey represent the reliability error sensitivity coefficients corresponding to the X, Y, and Z axes respectively. The angular deviations of the three directions of the bullet feeding line, X, Y, and Z, are α, β,

[0092] The coordinator motion feed line angle accuracy γ ε Expressed as

[0093]

[0094] Where λ α ,λ β 、 They are the three angle deviations α, β, The corresponding sensitivity coefficient;

[0095] When obtaining the corresponding true function value, the motion displacement accuracy γ of the projectile at the end of the movement D and the angle accuracy of the moving bullet feeding line γ ε As a standard value, establish the function of the coordinator motion accuracy:

[0096] y(x)=ξ-|γ-γ′|,

[0097] Where ξ is the allowable error, which can be determined according to the accuracy requirements, and γ′ is the motion displacement γ D Or the angle of the moving bullet line γ ε The ideal value of γ is the displacement accuracy of the projectile. D Or the angle accuracy of the moving bullet feeding line γ ε . And get the real function value corresponding to SDoE like Figures 3 to 5 shown.

[0098] S3, use Monte Carlo random sampling method to extract N random variables for each MC A random sample that meets the respective distribution type is used as the corresponding candidate sample point;

[0099] S4, based on the initial training sample set S DoE The sample data and true response values ​​in are used to build the Kriging model;

[0100] In the process of constructing the approximate expression of the state function using the Kriging method, the performance function G(x) consists of two parts: a deterministic part and a random part, namely: G(x) = Γ(β,x) + z(x) = g T (x)β+z(x).

[0101] In a given sample set S DoE , basis function g(x) and correlation coefficient function R(x i ,xj ; θ), Kriging predicted value and Kriging predicted values and Kriging equation The expression of is completely dependent on the parameter θ; this method uses the widely recognized maximum likelihood method to determine the optimal θ value

[0102] S5. Use MCMC method to generate sample points with U(x)<2. If the sequence x0, x1, ..., x K+1 Produced in {x k+1 ∣x k}, and the conditional probability satisfies: P(x k+1 =x|x k ,x k-1 ,…)=P(x k+1 =x|x k ), x k+1 The distribution of depends only on the current state x k , does not depend on the historical state, then x0, x1, x2, ..., x K+1 is a Markov chain. Where x0 is the given initial condition, according to the conditional probability distribution, the state x k+1 The desired probabilistic properties are conditioned by the previous state x k Given;

[0103] When k approaches infinity, x k It is independent of the initial value, that is, when k increases, the random vectors in the Markov chain will converge to the steady-state distribution π(x). At this time, the Markov chain is considered to be in a convergent state. When the Markov chain does not reach the convergent state, it is considered that the probability density distribution of each state cannot reach a stable distribution. When evaluating, the first M sampling values ​​that have not reached the steady-state distribution should be removed, and the last nM sampling values ​​should be used for evaluation: The estimated value of E[F] The variance is: The estimated value of V(F) is From this, a Markov chain can be constructed;

[0104] Using the K-Means clustering algorithm, under the condition of a given cluster group number K value, n sample data are divided into K groups through multiple iterations. The algorithm uses Euclidean distance as the similarity measurement method. The objective function can be expressed as:

[0105] Where: E is the sum of the mean square errors of all objects in the cluster sample; K is the number of clusters; xj is a point in the cluster; mi is the mean of the cluster Ci;

[0106] The basic steps of the K-Means clustering algorithm are:

[0107] 1) Determine the number of clusters K and input the data vector group X, which contains data vectors X1, X2, X3, ..., X n ;

[0108] 2) Randomly select K vectors from n data vectors as initial cluster centers;

[0109] 3) Calculate the Euclidean distance of each vector to the K initial cluster centers and assign it to the cluster where the initial cluster center with the smallest distance value is located;

[0110] 4) If the data vector in the cluster changes, recalculate the centers of the K clusters;

[0111] 5) Repeat steps 3) to 4) until the centers of the clusters no longer change or the center distance is less than the specified threshold.

[0112] S6. Calculate the learning function value of each candidate sample point corresponding to each random variable according to the initial Kriging model. After the candidate samples are divided into K groups, find the minimum learning function value corresponding to each group of random variables in the K groups. If there is a minimum learning function value less than 2, then the K candidate sample points corresponding to the minimum learning function value in the candidate sample points corresponding to each random variable in each group are As the best sample point of the random variable; and obtain the true response value of each best sample point

[0113] Point selection principle: interval P i For example, at this time, the accuracy of the model in the interval is Δ i N , if in the interval P i Select a new sample point Updating the Kriging model will improve the model accuracy within this interval by Δ i N+1 Indicates that the model accuracy is improved Expressed as:

[0114]

[0115] Where, Δ i N can be considered as a constant, is with Related functions;

[0116] The expression (point selection formula) is:

[0117]

[0118] Where N G Indicates the number of integral points, v j and w j are the integral points and their corresponding weights respectively;

[0119] S7, use the point selection formula to select K candidate sample points With the current S DoE If the conditions are not met, go to step S6, remove the points that do not meet the conditions from the corresponding K-means group, and reselect them according to the point selection formula; if the conditions are met, go to step S8;

[0120] Among them, the construction process of the point selection formula is:

[0121] Use the threshold [L] to directly calculate the training sample x new and the training samples x that constitute the Kriging model j (j=1,…,N) is limited in spacing, i.e.

[0122]

[0123] Based on the above formula, we can get a new A hypersphere with [L] as the radius and [L] as the center;

[0124] The constraint on the distance between two points is converted into a correlation function problem, and the correlation function between training samples is restricted, that is:

[0125]

[0126] Where x DoE Indicates the training points that constitute the current Kriging model, [r] is usually between 0.95 and 1;

[0127] S8, K best training sample points and the true response value of each optimal training sample point Add them to the initial training sample set and the initial true response value matrix respectively to obtain the updated training sample set and true response value matrix, and update the Kriging model;

[0128] S9. Calculate the estimated failure probability P of the coordinator motion reliability based on the current Kriging model and the training sample set. f , error coefficient e ε and coefficient of variation δP f , the specific calculation process is:

[0129] The absolute error between the true value P of the failure probability and the estimated value P is:

[0130]

[0131] The I(x) value in the above formula cannot be obtained based on the Kriging model, but Satisfying the 0-1 distribution, it can be deduced according to the above formula expectations, i.e.

[0132]

[0133] From this, we can get an upper limit value A of the expected absolute error of failure probability,

[0134]

[0135] Error coefficient e ε for:

[0136]

[0137] Multiple I G≤0 (X) The mean of the random sample values ​​is used as P f valuation,

[0138]

[0139] Where x MC,n (n=1,…,N MC ) is derived from the probability density function f X (x) is an independent and identically distributed random sample point, N MC is the number of sample points;

[0140] According to the central limit theorem, when N MC →∞, According to the distribution converges to the normal distribution,

[0141]

[0142] so,

[0143]

[0144] The coefficient of variation is,

[0145]

[0146] δP f ≤[δ];

[0147] S10, judgment error coefficient e ε Is it satisfied Requirements:

[0148] If not satisfied, return to step S5;

[0149] If satisfied, execute step S11;

[0150] Where, e ε is the error coefficient, [e ε ] is the threshold value, which is determined according to the actual accuracy requirement;

[0151] S11. Determine the coefficient of variation δP f Whether δP is satisfied f Requirements for ≤[δ]:

[0152] If it is not satisfied, the number of random sample points is increased by Monte Carlo method;

[0153] If satisfied, proceed to step S12;

[0154] Where [δ] is the threshold;

[0155] S12. Calculate and output the coordinator motion reliability results: failure probability Coefficient of variation and error coefficient e ε .

[0156] As application examples of the present invention:

[0157] Specify the mechanism to be analyzed as the artillery coordinator electromechanical hydraulic system, determine the variables of the functional function and its probability distribution information, such as Figure 2 The figure shows the structure of the gun coordinator. The example takes into account the clearance between the two shaft holes of the coordination mechanism, the hydraulic oil temperature of the hydraulic system, the speed of the hydraulic pump, the leakage rate of the coordination hydraulic cylinder and the leakage rate of the flip hydraulic cylinder.

[0158] Table 1 Random variable distribution table

[0159]

[0160]

[0161] Table 2 compares the calculation results of the Monte Carlo method (MCS), the AK-MCS-U method, the AK-MCS-EFF method, and the proposed method. The proposed method performs calculations for k = 2, k = 3, and k = 4, respectively. Using the MCS results as a reference, the failure probability estimates obtained by the proposed method are closest to those obtained by the reference method and are lower than those of the other two methods. Furthermore, the number of function calls is 336, 340, and 338, respectively, which is also the lowest among all methods. These results demonstrate that the proposed method is more accurate and efficient than the AK-MCS-U and AK-MCS-EFF methods.

[0162] Table 2 Comparison of calculation results

[0163]

[0164] 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.

Claims

1. A method for evaluating the kinematic reliability of a coordinated electromechanical and hydraulic system of a gun based on a Kriging model, characterized in that: The steps include: S1. Determine N random variables and their distribution types based on the relevant parameters that affect the coordinator's motion; S2. Use Latin hypercube sampling to extract N0 training samples that meet the distribution type for each random variable, and form the initial training sample set in the standard normal space. And get the corresponding true response value Get the initial true response value matrix; S3, use Monte Carlo random sampling method to extract N random variables for each MC A random sample that meets the respective distribution type is used as the corresponding candidate sample point; S4, based on the initial training sample set S DoE The sample data and true response values ​​in are used to build the Kriging model; S5, using the Markov Chain Monte Carlo method to extract candidate sample points with U(x)<2 from the candidate sample points generated in step S3, and using the k-means clustering algorithm to divide the extracted candidate sample points into K groups; S6. Select K best training sample points from the K groups of candidate sample points formed in step S5 according to the method of selecting sample points. And obtain the true response value of each optimal training sample point S7, determine whether the K best training sample points meet the current S DoE The distance criterion of the training sample points in : If not, remove the non-compliant points from the corresponding K-means group. After removing the non-compliant points, go to step S6 and reselect based on the distance criterion. If satisfied, proceed to step S8; The distance criterion is expressed by a point selection formula, which is: max{R(x new ,x DoE ;θ)}≥[r], Where x DoE Indicates the training points that constitute the current Kriging model, [r] is usually between 0.95 and 1; S8, K best training sample points and the true response value of each optimal training sample point Add them to the initial training sample set and the initial true response value matrix respectively to obtain the updated training sample set and true response value matrix, and update the Kriging model; S9. Calculate the estimated failure probability P of the coordinator motion reliability based on the current Kriging model and the training sample set. f , error coefficient e ε and coefficient of variation δP f ; S10, judgment error coefficient e ε Is it satisfied Requirements: If not satisfied, return to step S5; If satisfied, execute step S11; Where, e ε is the error coefficient, [e ε ] is the threshold value, which is determined according to the actual accuracy requirement; S11. Determine the coefficient of variation δP f Whether δP is satisfied f Requirements for ≤[δ]: If it is not satisfied, the number of random sample points is increased by Monte Carlo method; If satisfied, proceed to step S12; Where [δ] is the threshold; S12. Calculate and output the coordinator motion reliability results: failure probability Coefficient of variation and error coefficient e ε .

2. The method for evaluating the kinematic reliability of a gun coordination electro-hydraulic system based on a Kriging model according to claim 1 is characterized in that: The relevant parameters include the manufacturing error of the key dimensions of the coordinator, the clearance error of the key hole-shaft fit, the projectile mass error, the oil pump speed, the leakage rate of the coordinating cylinder and the leakage rate of the flipping cylinder.

3. The method for evaluating the kinematic reliability of a gun coordination electro-hydraulic system based on a Kriging model according to claim 2 is characterized in that: The key dimensions of the coordinator include the dimensions of the coordinating arm and the dimensions of the flipping mechanism.

4. The method for evaluating the kinematic reliability of a gun coordination electro-hydraulic system based on a Kriging model according to claim 1, wherein: The true response value The method for obtaining , comprising the following steps: S41. Use Latin hypercube sampling to extract N0 training samples that meet the distribution type for each random variable, and form the initial training sample set in the standard normal space. S42, the coordinator motion displacement accuracy γ D Expressed as: Among them, η X , η Y , η Z Represent the reliability error sensitivity coefficients corresponding to the X, Y and Z axes respectively; S43, the coordinator motion feed line angle accuracy γ ε Expressed as: Among them, α, β, are the angular deviations of the three directions X, Y, and Z of the bullet feeding line, respectively, α ,λ β 、 They are the three angle deviations α, β, The corresponding sensitivity coefficient; S44, the coordinator motion displacement accuracy γ D and the coordinator motion feed line angle accuracy γ ε As a standard value, a function function of the coordinator motion accuracy is established, and the expression of the function function is: Where ξ is the allowable error, which can be determined according to the accuracy requirements, and γ′ is the motion displacement γ D Or the angle of the moving bullet line γ ε The ideal value of γ is the displacement accuracy of the projectile. D Or the angle accuracy of the moving bullet feeding line γ ε .

5. The method for evaluating the kinematic reliability of a gun coordination electro-hydraulic system based on a Kriging model according to claim 1 is characterized in that: The method for establishing the Kriging model comprises the following steps: In the S51 and Kriging models, the performance function G(x) consists of two parts: a deterministic part and a random part, namely: G(x)=Γ(β,x)+z(x)=g T (x)β+z(x), S52, given an initial training sample set S DoE , basis function g(x) and correlation coefficient function R(x i ,x j ;θ), Kriging predicted value and Kriging predicted values and Kriging equation The expression of is completely dependent on the parameter θ; the optimal value of θ is determined by the widely accepted maximum likelihood method.

6. The method for evaluating the kinematic reliability of a gun coordination electro-hydraulic system based on a Kriging model according to claim 1 is characterized in that: The specific operation steps of step S5 are as follows: In the MCMC method (Markov chain), if the sequence x0, x1, x2, ..., x K+1 Produced in {x k+1 ∣x k }, and the conditional probability satisfies: P(x k+1 =x|x k ,x k-1 ,…)=P(x k+1 =x|x k ), x k+1 The distribution of depends only on the current state x k , does not depend on the historical state, then x0, x1, x2, ..., x K+1 is a Markov chain; where x0 is the given initial condition, according to the conditional probability distribution, the state x k+1 The desired probabilistic properties are conditioned by the previous state x k Given; When k approaches infinity, x k It is independent of the initial value, that is, when k increases, the random vectors in the Markov chain will converge to the steady-state distribution π(x). At this time, the Markov chain is considered to be in a convergent state. When the Markov chain does not reach the convergent state, it is considered that the probability density distribution of each state cannot reach a stable distribution. When evaluating, the first M sampling values ​​that have not reached the steady-state distribution should be removed, and the last nM sampling values ​​should be used for evaluation: The estimated value of E[F] The variance is: The estimated value of V(F) is From this, a Markov chain can be constructed; Using the K-means clustering algorithm, based on the principle of minimum Euclidean distance between the data sample in each group and the cluster center, under the condition of a given cluster number K value, n sample data are divided into K groups through multiple iterations. The calculation formula of Euclidean distance can be expressed as: Where: E is the sum of the mean square errors of all objects in the cluster sample; K is the number of clusters; x j is a point in the cluster; m i is the mean of cluster Ci; According to the calculated Euclidean distance from the data sample to the cluster center, the samples are classified according to the minimum distance principle, and the n sample data are divided into K groups through multiple iterations.

7. The method for evaluating the kinematic reliability of a gun coordination electro-hydraulic system based on a Kriging model according to claim 6 is characterized in that: The k-means clustering algorithm includes the following steps: S71, determine the number of clusters K, input the data vector group X, which contains data vectors X1, X2, X3, ..., X n ; S72. Randomly select K vectors from the n data vectors as initial cluster centers; S73, calculating the Euclidean distance of each vector to the K initial cluster centers, and assigning the vector to the cluster where the initial cluster center with the smallest distance value is located; S74. If the data vector in the cluster changes, recalculate the centers of the K clusters; S75. Repeat steps S73 to S74 until the centers of the clusters no longer change or the center distance is less than a specified threshold.

8. The method for evaluating the kinematic reliability of a gun coordination electro-hydraulic system based on a Kriging model according to claim 1 is characterized in that: The method for selecting sample points comprises the following steps: S81, calculating the learning function value of each candidate sample point corresponding to each random variable according to the initial Kriging model; S82. Find the minimum learning function value corresponding to each group of random variables in the K groups; S83. Determine whether the minimum learning function value corresponding to each group of random variables is less than 2: If the minimum learning function value is less than 2, then the K candidate sample points corresponding to the minimum learning function value in each group of candidate sample points corresponding to each random variable are used as the K best training sample points of the random variable; If the minimum learning function value is ≥ 2, continue to judge.

9. The method for evaluating the kinematic reliability of a gun coordination electro-hydraulic system based on a Kriging model according to claim 1 is characterized in that: The specific process of establishing the point selection formula includes the following steps: S91, use the threshold [L] to train the sample set x new And the training sample set x that constitutes the Kriging model j (j=1,…,N) is limited in spacing, that is: Based on the above formula, we get a new A hypersphere with [L] as the radius and [L] as the center; S92. Convert the constraint of the distance between two points into a correlation function problem, and restrict the correlation function between training samples, that is: max{R(x new ,x DoE ;θ)}≥[r], Where x Do E represents the training points that constitute the current Kriging model, and [r] is usually between 0.95 and 1.

10. The method for evaluating the kinematic reliability of a gun coordination electro-hydraulic system based on a Kriging model according to claim 1 is characterized in that: The failure probability P f : The coefficient of variation The error coefficient e ε : Where x MC,n (n=1,…,N MC ) is derived from the probability density function f X (x) is an independent and identically distributed random sample point, N MC is the number of sample points.

Images