中文摘要：

在基于支持向量回归的响应面可靠度计算方法基础上，采用试验设计中设计试验点的方法生成样本点，构造训练集，并对可靠度计算的迭代序列响应面方法进行改进。计算结果表明，与采用拉丁超立方抽样和常规梯度优化法求解的基于支持向量回归响应面方法相比，文中方法所需样本更少、收敛更快、结果更理想。

英文摘要：

On the basis of the response surface method for reliability computation based on support vector regression, the iterative sequence response surface method is improved for reliability computation. The proposed method generates samples by the technique for designing experimental points in experimental design, and then constructs the training sets. Two numerical examples are presented. Re- suits show that the present method possesses three main advantages over the response surface method based on support vector regression using the Latin Hypercube sampling and the ordinary gradient optimization method： requiting less samples, converging more rapidly, and producing better issues.