改进的重要抽样技巧

An improved importance sampling technique

提出了改进的重要抽样技巧．引入指数罚因子来自动搜索设计点，并用加权残值法逼近最佳的重要抽样密度函数；然后以核函数作为拟合密度函数实现拟合－适应重要抽样，再用极大似然法选取核函数的窗口宽度和敏感系数．此方法可以不需要显式的失效方程，因此适用于结构可靠性随机模拟，尤其对每次抽样都很费机时的复杂结构问题，本法具有很大的优越性．

An improved importance sampling technique is proposed. A penalty factor is introduced to search for the design point and the method of weighted residuals is utilized to obtain better importance sampling density function. The kernel function is employed as fitting density function in the fittingadaptive process. The window width and the sensitivity index are determined by means of maximum likelihood method. Without recourse to direct Monte Carlo simulation and explicit failure function, the proposed technique is suitable to timeconsuming structural reliability problems.