Abstract: A successive quadratic response surface approximate optimization method(SRSM) is presented to reduce the iterations of model analysis and to improve the optimization efficiency. Based on uniform design, SRSM creates a quadratic approximate model including a global response surface with low accuracy and some local surfaces with the accuracy being gradually improved during iterations. Global response surface determines the initial sub-region, while local ones are used to obtain the optimal solution in the sub-region. In this program, successive sub-region is firstly updated by a new moving limit criteria obtained from the relative position of normalized optimum. Then the successive response surface constructed by inherited and added samples in the sub-region is iteratively resolved until the optimal design is achieved. Five well-known test functions and one engineering design problem are used to verify the validity of the method. The buckling optimization of cylinder shell with holes is efficiently carried out by SRSM. All results exhibit the potential application of this method for computation-intensive engineering problem.