A Sequential Response Surface Method Using I nherited Uniform Design

This study presents a Sequential quadratic Response Surface Method(SRSM) using inherited Uniform Design(UD), to reduce the computational burden in optimization process and improve the computational efficiency of engineering optimization. The program comprises a low accuracy global quadratic response surface modeling and a precision increasing sequential quadratic response surface modeling, which were consturcted by the 1.5 and 1 times saturated sampling points, respectively. The initial sub-region of the sequential response surface is determined by the global response surface, according to the optimum point in the sample set. A new move limit criterion based on the relative position of normalized optimum is proposed to achieve sub-region planning and zooming. Three test mathematic functions and a buckling optimization design problem are solved to show the numerical performance of the proposed method. The calculated data show that the inherited UD and the move limit criterion introduced into approximate modeling can effectively improve the efficiency and the conergence accuracy of the method presented. The results exhibit the progressive in the computation-inte nsive engineering application.