신뢰성 해석을 위한 입력변수의 상관성을 고려한 아카이케 정보척도를 기반 코플라 추정 방법

Abstract

For reliability analysis, statistical modeling of input random variables such as joint cumulative distribution function is required. In some statistical models in engineering applications, input random variables are considered independent even though correlation exists between the variables. Copula is widely employed for modeling multivariate distributions since it captures the dependence between the marginal variables. Among many copula estimation methods, semiparametric method is preferred because assumption for marginal distributions is not required, thus estimation of copula is considered to be more accurate than other methods. However, semiparametric method needs assumption of marginal distributions while performing reliability analysis. In this paper, Akaike information criterion is employed for estimating marginal distributions and maximum likelihood estimation for estimating copula. The proposed method can use discrete information without any assumption for marginal distributions and is capable of performing reliability analysis when correlation exists between input random variables.