Inverse measure-based tail modeling approaches for structural reliability estimation

摘要: Sampling-based reliability estimation with expensive computer models may be computationally prohibitive. One way to alleviate the computational expense in high designs is extrapolate estimates from observed levels unobserved levels. Classical tail modeling approaches provide a class of enable this extrapolation using asymptotic theory by approximating region cumulative distribution function (CDF). This paper proposes an alternate based on inverse measure, which can complement classical modeling. The proposed approach applies nonlinear transformation CDF measure and approximates transformed quadratic polynomial. Accuracy efficiency are competing factors selecting sample size. Yet, as our numerical studies reveal, accuracy lost reduction power very small method. method demonstrated two engineering examples true statistical distributions.