Abstract：Aiming at the uncertainty modeling of FE model for structure, D-opt experiment design produced sample space. Supported Vector Regression (SVR) method was presented to construct the implicit mapping between structure response and uncertainty parameters. The computational process of forward transfer of uncertainty parameters has been clarified. Three level testing criteria was used to evaluate extensively precision and generalization ability of RSM based SVR.Compare to polynomial RSM, The efficiency and accuracy of the proposed method were proved by the case study of GARTEUR.
2010, Vol. 29 
