高维高可靠性问题是工程中广泛存在而亟待解决的难题，目前尚缺乏有效的可靠性评估方法。广义Pareto分布方法(GPD)是基于样本信息对随机参数的尾部直接进行建模，对于问题的维度以及非线性程度依赖较低，然而传统GPD方法中确定阈值的高额计算成本限制该方法在工程领域的应用。本项目将基于GPD方法并结合Kriging代理模型，开展高维高可靠性问题评估方法研究，致力于解决以下四个难点问题：第一，对Kriging模型的适用性进行研究，并进一步降低建立Kriging模型的计算成本；第二，基于Copula函数实现重要抽样函数的高精度自适应确定；第三，针对高维可靠性问题，提出基于分位点的广义Pareto分布函数尾部近似，得到高维可靠性问题的高效计算方法；第四，提出基于分位点信息的可靠性优化解耦方法，实现针对高非线性、高可靠性问题的可靠性优化设计。

There exist lots of high-dimension and high-reliability problems in practical engineering, which is still a challenging research topic. The generalized Pareto distribution (GPD) method builds the model of the tail part of the random parameter based on samples, which is independent of the dimension and nonlinearity of the problems. However, the high computational cost of the threshold value in the GPD method hinders its application in engineering problems. The major target of this project is to develop the estimation method for high-dimension and high-reliability problems based on the GPD method, combining the surrogate model to improve the computational efficiency. In the purpose of solving these problems, the applicability of the Kriging model is investigated to obtain an efficient update strategy for Kriging model used in the importance sampling (IS) method. Meanwhile the adaptive determination for the optimal importance sampling density is also studied. The quantile based GPD method is proposed to solve the high dimensional problems, and the calculation for the quantiles is adopted from the sequential IS method. A new method for reliability based design optimization is also proposed based on the quantiles.