CN-121980955-A - Small sample mechanism motion precision reliability assessment method based on data fusion

CN121980955ACN 121980955 ACN121980955 ACN 121980955ACN-121980955-A

Abstract

The invention relates to a small sample mechanism motion precision reliability assessment method based on data fusion, which comprises the steps of collecting distribution information of random variables, utilizing the distribution information to determine motion precision parameters as initial priori samples, expanding the initial priori samples by combining small sample test data, implementing prior distributed super-parameter estimation, acquiring posterior distribution parameter values fused with simulation and test information according to the super-parameter estimation values and statistical values of test samples, determining point estimation values of the motion precision distribution parameters, inputting the point estimation values into a mechanism motion precision reliability model, and acquiring motion precision reliability estimation values of a mechanism. According to the invention, the test information and the priori information are fused by a Bayesian method, so that the statistical inference of the motion precision parameter is more stable, and the stability and the accuracy of the mechanism reliability evaluation result under the small sample test condition are improved.

Inventors

YIN YIN
YAN YUMENG
ZHOU JIYUAN
LIANG TAOTAO
WEI XIAOHUI
NIE HONG
CHENG YUQIAN

Assignees

南京航空航天大学

Dates

Publication Date: 20260505
Application Date: 20260211

Claims (9)

1. The small sample mechanism motion precision reliability assessment method based on data fusion is characterized by comprising the following steps of: collecting distribution information of random variables, and determining a motion precision parameter as an initial priori sample by utilizing the distribution information; The initial prior sample is combined with small sample test data to expand the prior sample, super-parameter estimation of prior distribution is implemented, posterior distribution parameter values fusing simulation and test information are obtained according to the super-parameter estimation value and the statistical value of the test sample, and the posterior distribution parameter values are used for determining point estimation values of motion precision distribution parameters; And inputting the point estimated value into a mechanism motion precision reliability model to obtain a mechanism motion precision reliability estimated value.
2. The method for evaluating the motion precision reliability of a small sample mechanism based on data fusion according to claim 1, wherein determining the motion precision parameter as an initial prior sample comprises: Generating a plurality of input sample points by using the distribution information of the random variables, and performing batch simulation on the plurality of input sample points by using a simulation model to obtain corresponding motion precision parameters, wherein the initial prior sample is formed based on the motion precision parameters; Wherein the random variables include joint clearances and assembly deviations of the mechanism.
3. The method for evaluating motion accuracy and reliability of a small sample mechanism based on data fusion according to claim 1, wherein obtaining the small sample test data comprises: acquiring various typical working conditions, wherein the typical working conditions are joint clearances and assembly deviation working conditions with different values; And carrying out a mechanism dynamics test on various typical working conditions to obtain the small sample test data, namely a mechanism movement precision parameter set measured under various typical working conditions.
4. The method for evaluating motion accuracy and reliability of a small sample mechanism based on data fusion according to claim 1, wherein performing a priori sample expansion on the initial a priori sample in combination with small sample test data comprises: Step 1, training a BP neural network model by taking random variables in the initial priori sample and the small sample test data as input data and taking motion precision parameters as training labels, judging regression coefficients of the BP neural network model in the training process, and finishing the training if the regression coefficients are larger than a preset threshold value; Step 2, inputting a plurality of input sample points corresponding to the distribution information into the trained BP neural network model to obtain a prediction sample; step 3, calculating the mean value and variance of the prediction samples; and 4, repeating the step 2-3 until reaching a repeated threshold value, and obtaining a plurality of groups of mean values and variances for forming a priori mean value sample and a priori variance sample.
5. The method for estimating motion accuracy and reliability of a small sample mechanism based on data fusion according to claim 4, wherein performing the super-parametric estimation of the prior distribution comprises: And calculating the prior distributed super-parameter estimation by using the prior mean value sample and the prior variance sample, and obtaining the super-parameter estimation value, the prior normal distribution parameter value and the prior inverse gamma parameter value.
6. The method for evaluating the reliability of the motion precision of a small sample mechanism based on data fusion according to claim 1, wherein determining the point estimation value of the motion precision distribution parameter comprises: determining a statistical value of the test sample based on the small sample test data; Acquiring the posterior distribution parameter value according to the super parameter estimation value and the statistical value of the test sample; And determining the point estimation value of the motion precision distribution parameter by using the posterior distribution parameter value.
7. The method for evaluating motion accuracy and reliability of a small sample mechanism based on data fusion according to claim 1, wherein obtaining the posterior distribution parameter value comprises: ; Wherein, the 、、 And As a posterior distribution parameter, the distribution of the parameters, 、、 And The method is used for obtaining the super-parameters, namely unknown parameters contained in the prior distribution, In order to test the amount of sample, For the mean value of the test sample, Is the test sample variance.
8. The method for evaluating the reliability of the motion precision of a small sample mechanism based on data fusion according to claim 1, wherein determining the point estimation value of the motion precision distribution parameter comprises: ; Wherein, the As a point estimate of the motion accuracy distribution parameter, 、、 And Is a posterior distribution parameter.
9. The method for evaluating the motion precision and reliability of a small sample mechanism based on data fusion according to claim 1, wherein the mechanism motion precision and reliability model comprises: ; ; Wherein, the For the motion accuracy reliability estimation of the mechanism, As the threshold value of the motion accuracy, As a parameter of the accuracy of the movement, 、 Is that Is a function of the mean and variance of (a), 、 Is that Is a function of the mean and variance of (a), As a function of the probability, As a function of the mechanism, Is that Is used for the average value of (a), Is that Is a variance of (c).

Description

Small sample mechanism motion precision reliability assessment method based on data fusion Technical Field The invention relates to the technical field of reliability evaluation of complex mechanisms of aircrafts, in particular to a small sample mechanism motion precision reliability evaluation method based on data fusion. Background In order to verify the mechanism movement principle, expose design defects and faults and eliminate potential factors which cause the mechanism to fail, a reliability test is required to be carried out on the landing gear retraction mechanism, so that movement reliability analysis is completed. The landing gear retraction system is a complex system of a collector, electricity and liquid, is limited by test cost and test period, and can only perform small sample tests. Key influencing factors (joint clearance, assembly deviation and the like) of the motion precision of the retraction mechanism are inevitably influenced by uncertainty factors in the design, manufacture and use processes, so that the characterizability of small sample test data on the actual motion precision parameter distribution of the mechanism is further weakened, accurate and stable reliability is difficult to calculate, and the reliability of results is limited. Therefore, there is a need to develop a reliability estimation method specifically for the movement characteristics of a jack under small sample conditions. Disclosure of Invention In order to solve the problems in the prior art, the invention aims to provide a small sample mechanism motion precision reliability assessment method based on data fusion, which improves the mechanism reliability assessment result robustness under the condition of small samples. In order to achieve the above object, the present invention provides the following solutions: the small sample mechanism motion precision reliability assessment method based on data fusion comprises the following steps: collecting distribution information of random variables, and determining a motion precision parameter as an initial priori sample by utilizing the distribution information; The initial prior sample is combined with small sample test data to expand the prior sample, super-parameter estimation of prior distribution is implemented, posterior distribution parameter values fusing simulation and test information are obtained according to the super-parameter estimation value and the statistical value of the test sample, and the posterior distribution parameter values are used for determining point estimation values of motion precision distribution parameters; And inputting the point estimated value into a mechanism motion precision reliability model to obtain a mechanism motion precision reliability estimated value. Optionally, determining the motion precision parameter as an initial prior sample comprises: Generating a plurality of input sample points by using the distribution information of the random variables, and performing batch simulation on the plurality of input sample points by using a simulation model to obtain corresponding motion precision parameters, wherein the initial prior sample is formed based on the motion precision parameters; Wherein the random variables include joint clearances and assembly deviations of the mechanism. Optionally, obtaining the small sample test data comprises: acquiring various typical working conditions, wherein the typical working conditions are joint clearances and assembly deviation working conditions with different values; And carrying out a mechanism dynamics test on various typical working conditions to obtain the small sample test data, namely a mechanism movement precision parameter set measured under various typical working conditions. Optionally, performing a priori sample expansion on the initial a priori sample in combination with small sample test data includes: Step 1, training a BP neural network model by taking random variables in the initial priori sample and the small sample test data as input data and taking motion precision parameters as training labels, judging regression coefficients of the BP neural network model in the training process, and finishing the training if the regression coefficients are larger than a preset threshold value; Step 2, inputting a plurality of input sample points corresponding to the distribution information into the trained BP neural network model to obtain a prediction sample; step 3, calculating the mean value and variance of the prediction samples; and 4, repeating the step 2-3 until reaching a repeated threshold value, and obtaining a plurality of groups of mean values and variances for forming a priori mean value sample and a priori variance sample. Optionally, performing the super-parametric estimation of the prior distribution comprises: And calculating the prior distributed super-parameter estimation by using the prior mean value sample and the prior variance sample, and obtaining the super-