구형 부품의 운영정보를 고려한 무고장 신형 전력반도체의 베이지안 신뢰도 특성 연구

초록

An important difference between classical and Bayesian statistics is that Bayesian statistics use a prior distribution. Even when there is insufficient data in the reliability analysis, successful results can be derived if an appropriate prior distribution is used. In this paper, by analyzing the failure information of old parts that have been discontinued, the gamma distribution and the simple uniform distribution were proposed as prior distributions for new parts by considering average values and deviations. Here, by mathematically transforming the likelihood function, the Bayesian reliability characteristics of new parts with no failure characteristics were studied. It was found that it is possible to calculate the quantitative reliability value of zero failures information, which cannot be done in classical statistics. Also, in the case of the prior distribution developed with the uniform distribution and the gamma distribution, both prior distributions can produce quantitative values, but the model of the prior distribution made from the gamma distribution is expected to give a more explicit effect to field technicians in interpreting the posterior distribution.

키워드