相关情况下可靠性灵敏度分析的矩方法

Reliability sensitivity for correlative variables by moment method

针对正态变量具有相关性的可靠性灵敏度分析问题,给出了线性极限状态情况下的解析方法及可靠性灵敏度的通用数字模拟算法,并提出了一种高效的矩方法。在所提的矩方法中,相关变量首先被等效转换为不相关的变量,利用转换的不相关变量空间中均值与方差、相关性特征参数的非耦合性,得到了失效概率对相关变量均值的灵敏度,并且通过利用基本变量方差相同时,相关性描述参数与均值、方差的自然解耦,得到了失效概率对相关系数的灵敏度。在所提方法的适用范围内,其精度和效率令人满意。

For the Reliability Sensitivity （RS） problem with normal correlative variables, an analytical method for linear performance function and a numerical simulation method are presented, and an efficient moment based RS method is presented as well to analyze the reliability sensitivity with the normal correlative variables. In the presented moment based RS method, the correlative variables are firstly transformed into independent variables equivalently. By use of the decoupling between the mean value and the variance, the characteristic parameter describing the correlativity, the RS of the failure probability with respect to the mean value of the correlative variables can be obtained. Furthermore, employing the natural deeoupiing between the correlative coefficient and the mean value, the RS of the failure probability with respect to the correlative coefficient can be obtained by constructed moment method in case that the basic variables have the same varianees. In the applicable range, the precision and the efficiency of the presented moment method are satisfied.