采用Box-Behnken矩阵设计方法进行试验设计,并根据设计点的响应,利用最小二乘回归法建立响应面函数.将响应面函数中参数的梯度信息与其分散程度结合,得到各参数的灵敏度系数,再归一化灵敏度系数得到概率灵敏度;将响应面模型与一次二阶矩法相结合计算结构的可靠度.针对一个具体算例,分别采用基于响应面法与基于ANSYS的Monte Carlo法计算了结构的灵敏度及可靠度值,结果的一致性验证了该方法的有效性.
It is often necessary to study parameter sensitivity and structural reliability.The test design points of input-variables are obtained by the Box-Behnken matrix method in this paper.According to the responses of specimens,the coeffcients of the response surface function are identified by a least-squares regression.The steps of calculating probability sensitivity are as follows,firstly,computing the gradients of input-variables respectively,secondly,multiplying the gradient and the related range,and finally normalizing the above quantities.The structural reliability is analyzed by combining the response surface model and the first-order second-moment method.The sensitivity and the reliability of an example are analyzed by the response surface method and the Monte Carlo method in ANSYS.The results of two methods are consistent,which shows that the method in the paper is valid and feasible.