Research On Extreme Small Sample Reliability Evaluation Method And Its Application On Center Sill And Body Bolster Of C70 Gondola Car

In order to guarantee safety operation of C70 gondola car under speed increase and heavy haul, reliability of the whole car structure must be evaluated. Center sill and body bolster are important component of gondola car, and it is a big threat to the safety operation when the component occurs fracture, so the fatigue limit with high confidence level and high reliability must be gained first. Because the dimension of center sill and body bolster was large, their processing was complex, and the test time was long, so there were only three stress pairs in the up and down test, the fatigue limit with high confidence level could not be gained by classical statistic method. Extreme small sample reliability evaluation method must be used in such circumstance.Firstly, several extreme small sample reliability evaluation methods were introduced. Through comparing their advantages and disadvantages, Bayes method and virtual expanding method were chosen as the processing method of extreme small sample test data was determined.According to the needs of the engineering application, Bayes extreme small sample reliability evaluation method at normal distribution form was given, and the acquisition method of prior distribution was introduced. Bootstrap method, which was widely used in small sample reliability evaluation, was described in detail, a virtual sample expanding method, which the reliability test sample size is expanded from 2~3 to 12, was proposed on the basis of predecessor's research results.Then, normal distribution data and weathering steel abutting joint test data were used to verify the two methods, and the results showed that, it was feasible to process the extreme small sample test data by using the two methods.At last, the fatigue limit distribution parameter of C70 gondola car center sill and body bolster was gained by using the above methods, Monte Carlo method and Matlab. Fatigue limit with 90% confidence level and 95% reliability was calculated, and this would provide basic analytical data to increase the structure reliability and operation reliability of C70 gondola car.