Reliability Analysis And Statistical Inference Of Life Test Data With Competing Risk

With the rapid development of modern science and technology,people have higher requirements for product quality.Reliability is the core of product quality,so product reliability analysis and evaluation is a key issue in academic and engineering research.Due to the complexity of the internal structure of the product and the diversity of the external environment,the failure mechanism is complicated.Products often have multiple failure causes during the failure evolution process,and there are competitive failure characteristics.Therefore,it is necessary to carry out reliability statistical modeling on product life data with competitive failure characteristics,and establish data analysis methods for product reliability analysis and life prediction.In addition to competitive failure characteristics,modern industrial products often have the characteristics of high reliability and long life due to the improvement of manufacturing level,which makes it impossible to obtain sufficient failure samples in the process of product development and test verification.Therefore,the reliability test of life products is often carried out in the form of censored test.The so-called censored life test refers to the test method in which the failure time of all samples cannot be observed during the test,and the reliability test immediately stops when the specified conditions are met.Common censoring test methods include type-I(timing)censoring,type-II(fixed number)censoring,and progressive censoring life test.Different censoring strategies have different advantages and disadvantages.People can choose different censoring strategies for product life testing according to the purpose and conditions of the test.Based on two different life distribution models,firstly,this paper studies the stress-strength model of product reliability analysis in type-II censoring situation,and then studies the reliability statistical modeling and reliability analysis of different products under a kind of generalized censoring strategy for life distribution products with competitive failure risk.The specific research contents are as follows1.For multicomponent systems and products,it is assumed that the products stress and strength follow the Rayleigh life distribution.This paper studies the reliability of the system through two methods: classical estimation and constructed pivot quantity estimation in the case of the same and different stress and strength location parameters.Secondly,the thesis combined with mathematical theoretical methods to prove the rationality of the constructed pivotal quantities.Finally,the computer software is used to simulate the results of theoretical estimation,and the actual data set is analyzed.It is shown that the pivot estimation has more advantages than the classical estimation.2.In the case of competitive failure risk,assuming that the product failure time follows the Burr XII distribution,this paper studies a class of generalized progressive hybrid censored life parameters statistical inference problem.Classical estimation and Bayesian estimation are used to estimate unknown parameters.In the case of frequency statistical inference,this paper proves the existence and uniqueness of the parameter maximum likelihood estimation.At the same time,associated confidence interval estimates of parameters are constructed based on observed Fisher information matrix.Furthermore,the Bayesian estimation of parameters and the corresponding confidence interval estimation of the highest posterior density are established based on the introduction of a flexible prior distribution.Finally,the accuracy of the theoretical conclusions and the stability of the algorithm are tested by experimental simulation and real data sets analysis,and it is proved that the results of Bayesian estimation are better than classical estimation.