Consistency Identification Of Accelerated Test Failure Mechanism And Bayesian Statistical Analysis

In order to realize the rapid and accurate evaluation of the life and reliability of long life products at normal stress level.In this paper,the Bayes theory was used to conduct a comprehensive and deep research for the statistical analysis method of test data for the constant stress accelerated test.The main research conclusions are as follows:Firstly,based on the acceleration coefficient constant principle,the problem of failure mechanism consistency for Log-normal distribution,Weibull distribution and mixed exponential distribution is proposed in accelerated lifetime test(ALT),and the specific judgment method is given.Taking a Wiener degradation model and inverse gaussian model as research object,an identification method of failure mechanism consistency based on acceleration coefficient constant principle was proposed in accelerated degradation test(ADT),and the specific judgment method is given.Secondly,taking exponential distribution and Weibull distribution as research object in ALT,based on the prior distribution of the parameters are assumed to Jeffreys non-information distribution and gamma distribution,based on the shape parameter of Weibull distribution is subject to discrete distribution and continues distribution,the posterior distribution of the parameters are obtained by Bayes formula.The Bayes estimation of the parameters are obtained under the squared loss function,and the confidence interval of the reliability is constructed by the Bootstrap method.Engineering case shows that the reliability evaluation method under Bayes is more reasonable,operational and superiority.Thirdly,the Wiener process with random parameters is used to describe the individual differences of the products to overcome the heterogeneity of the products and improve the prediction accuracy.Based on the prior distribution of the parameters are assumed to normal-gamma distribution and Jeffreys non-information prior distribution,the posterior distribution of the parameters are obtained by Bayes formula.The EM algorithm is used to obtain the estimated value of the hyperparameter.Based on degraded sample under accelerated stress is mapped to the normal stress by acceleration coefficient,the method of reliability assessment at normal stress level is established.The feasibility of the proposed method is quantitatively analyzed by Monte Carlo simulation test.Fourthly,the inverse gaussian process with random parameters is used to describe the individual differences of the products to overcome the heterogeneity of the products and improve the prediction accuracy.Based on the prior distribution of the parameters are assumed to subject normal-gamma distribution,the posterior distribution of the parameters are obtained by Bayes formula.The EM algorithm is used to obtain the estimated value of the hyperparameter.Based on degraded sample under accelerated stress is mapped to the normal stress by acceleration coefficient,the method of reliability assessment at normal stress level is established.The feasibility of the proposed method is quantitatively analyzed by Monte Carlo simulation test.