베이지안 접근법을 이용한 소표본 HALT data에 대한 강도분포의 모수 추론

Abstract

HALT(Highly Accelerated Life Test) technique is accelerated methods which use stresses higher than the field environments to expose and then improve design weakness which can be explained stress-strength model. Because HALT conducted at the design phase, there are some constraint to analyze the data. For analysis HALT data through step stress tests, It was very important to parameter estimation of strength distribution accurately. When estimates parameters of strength distribution, previously used to MLE(Maximum Likelihood Estimation). But If we use the MLE for small samples HALT data, parameter‘s uncertainty is more increased than large sample. In this paper we proposed the Bayesian method for modeling HALT data. In case of Complete data which detect operating limits for all specimens was compared between existing Bayesian and MLE by using simulation method. For Censored case, to find the parameters of proposed posterior distribution, Metropolis-Hastings was used and also compared with MLE.