Bayesian Modeling And Optimization Based On Reliability Life Test Data

Product quality is not only the lifeline of the company,but also the key to winning customers in the global market competition.In recent years,the competition in product quality in various industries has become increasingly fierce,and reliability is one of the most important quality dimensions.It plays a vital role in improving product life,reducing warranty costs,and achieving expected product functions.Generally speaking,in the analysis process of reliability life data,data fitting and reliability improvement are two key issues.Many practical results show that: in reliability design,the limitations of randomized design and censoring problems inevitably lead to large deviations in the data fitting stage;In the reliability improvement stage,due to the uncertainty of the parameters,the output products often fluctuate,so it is necessary to consider the compliance of the optimization results.Considering the complexity of current product design and the diversity of customer needs,in the subsequent test phases,researchers also need to pay attention to finding an appropriate compromise for the optimum conditions on the premise that the product has both a quality and a reliability characteristic.In the data fitting stage,most studies have ignored the fitting problem of censored data.In addition,due to the limitation of calculation tools,the random effect of shape parameters has not been considered;In the reliability improvement stage,most researchers did not pay attention to the fluctuation of life expectancy.Combining Bayesian method,NLMM model,GLFP model,genetic algorithm and other technologies,this paper mainly studies the reliability improvement of accelerated life test.The detailed research content is as follows:(1)Bayesian modeling and analysis of Accelerated Life Data considering random effects.In the reliability design of accelerated life tests,the limitations of randomized design and censored data inevitably lead to significant deviations in low percentile estimation.Given the above problems,this paper combines Bayesian sampling technique and NLMM to propose a method for estimating reliability life.(2)Bayesian modeling and analysis based on multiple lifetime.For reliability lifetime data,when a single Weibull distribution cannot explain multiple failures,a sufficiently flexible model needs to be used to fit the data.In view of the above problems,under the framework of Bayesian statistical modeling,a new optimization model is constructed combining random effects and posterior probability method.A practical example reveals that the proposed method can obtain more robust and reliable estimation results when considering the effects of multiple failure modes and non-randomization.(3)Bayesian modeling and optimization of product quality and reliability characteristics considering random effects.In practical situations,products may have both a quality and reliability characteristics.Under the structure of sub-sampling,this paper combines Bayesian methods to construct analysis models of quality and reliability characteristics;and then a multivariate process capability index is constructed using the posterior sample values of the model to determine the compromise between the two characteristics.