Reliability Evaluation Based On Small Sample Of Gearbox Structure For High Speed Train

In order to improve the accuracy of structural reliability evaluation,a large amount of test data or sample information is often needed.However,for the rail vehicle and its key parts,the structure is complex and the manufacturing cost is high,so only a small amount of test data can be used to verify the rationality of its design index,so it is impossible to accurately evaluate it from the perspective of reliability.To on the basis of the existing data to achieve an accurate assessment of structural reliability,in this paper,the structure reliability assessment method based on small sample are studied,respectively from the small sample,extremely small sample and no failure data of small sample and so on three aspects to carry out the related work,and in view of the main problems of the traditional method,proposed the corresponding solution,the main research work is as follows:An improved Bayes method is proposed for the problem of small sample size and difficulty in making full use of a variety of prior information to obtain a more accurate prior distribution when the Bayes method is used for reliability evaluation of small sample.First,ML-II weighted average multi-source information fusion method was used to fuse various prior information to obtain a more accurate prior distribution.Secondly,the S-W method was used to test the consistency between the distribution types of the small sample and the prior distribution,and the K-S test method was used to test the compatibility between the distribution of the small sample and the prior distribution,so as to determine its origin from the same distribution.Thirdly,Monte Carlo method is used to expand the sample data to increase the sample capacity,and then the estimated parameter values of the distribution of the small sample are fitted,and the posterior distribution of the small sample is obtained by taking it into the Bayes formula.Finally,the method is used to evaluate the reliability of the gearbox with small sample.An improved Bootstrap reliability evaluation method was proposed to evaluate the reliability of very small samples by means of Bootstrap,which resulted in inaccurate evaluation results in the process of heavy sampling.Firstly,the virtual augmented sample method is used to augment the very small sample,even the small sample.Secondly,the small samples after augmentation were sorted and grouped according to their sizes,and the reliability of the data after augmentation was evaluated using the quartile difference method and the traditional Bootstrap method.Finally,the validity of the improved Bootstrap reliability evaluation method was verified based on the very small sample data of the gearbox.Aiming at the problem of small sample size and lack of prior information for reliability evaluation of small sample without failure data,a reliability evaluation method based on Bayes is proposed.Firstly,the Bootstrap method was used to small sample all master books.Secondly,the estimated failure probability of the mechanical structure under each heavy sampling truncation time is calculated by using the e-bayes method and the distribution curve method of multiple Bayes.Thirdly,according to the theory of reliability,we get the parameter estimation of the distribution of the sub-jack,and then get the estimation of the failure rate and reliability of the mechanical structure.Finally,this method is applied to the reliability assessment of the small sample of non-failure data of gearbox,and the advantages and disadvantages of the e-bayes method and the multiple Bayes method for the reliability assessment of the small sample of non-failure data are obtained through comparative analysis.