The shape parameter in Radius Basis Function (RBF) determines RBF's ability of approximation. Aimed to improve the ability, the error function of RBF is build, and then the shape parameter is chosen by optimization of the error function. The results of test functions show that the algorithm proposed here is much more adaptable to higher dimensional problems than Kriging model. Compare with Kriging, when the present algorithm is used in building Response Surface of wing structure, better approximation results are obtained in the computation of the mass of wing, the maximum stress of skin and the front edge displacement of wing. The effectiveness of the algorithm is further verified at the same time.