提出采用响应面法进行有限元模型修正.该方法根据方差分析或先验信息进行设计参数筛选,以显式的响应面模型逼近目标函数与设计参数之间复杂的隐式函数关系,通过响应面自身特性进行迭代修正.在此基础上,根据样本点对目标函数贡献的差异性提出通过加权来构造改进的二次响应面,改进后的响应面可以更快地迭代收敛,计算量小,可推广至非线性等复杂模型,某算例验证该方法的有效性.
A new method for finite element model updating based on response surface method(RSM)is presented in this paper. Firstly variables are choosed based on variance analysis or priori information, then RSM is applied as a close-form multinomial which provides the relationship between concerned features and selected design variables, finally the RSM iterates for self-characteristics. Besides, the im- proved RSM is submitted considering the different contribution made for performance function by sample points. The improved RSM has faster convergence and fewer computation to be popularized to nonlinear problems, example verifies the effectiveness.