In view of the problem of mathematical nonlinear regression, it is necessary to continuously improve the design model and design parameters, so that the design effect is the best. The response surface method is often used to optimize the minimum volume problem under the constraint conditions, which means that the simulation model is fitted by the function relation and the optimization direction is determined by the screening test. In order to optimize the shape of the classic 10-bar truss structure,based on the response surface, the APDL (ANSYS parametric design language) program was written and run, and the initial stress and volume of the truss were obtained. The change of node coordinates was used to reflect the change of shape. Each factor was set at three levels and encoded as -1, 0 and 1, and the center and level were adjusted by response surface method (RSM). Based on the adjusted level, the best neighborhood was found, and then the optimal solution of the volume under the allowable stress in the best neighborhood was solved by the optimization software. Finally, the determined node vertical coordinate value was returned to the APDL program to verify whether the constraint condition was satisfied. If it was satisfied, the volume was the best solution at this time. The results indicated that the iterations number was only 3 times when using the response surface method, which could quickly locate the optimal solution and reduce the previous design cost. By comparing the experimental design method in literature[9], it was found that the RSM was more efficient in optimizing the shape of the structure. In engineering applications, it is efficient to locate the design test program and find the best matching design under constraint conditions.