The volatile coefficient was adjusted according to the defect of slow convergence rate and trapping in local optimum of Ant Colony Algorithm(ACA).Volatile coefficient was initially endowed with a greater value to make the ants search for the better path,later,was decreased and self-adjusted by degrees to avoid the local convergence and obtain the global optimal path.The improved algorithm was used for the optimization design of concrete-filled steel tubular structure and the design model was established,in which the design parameter is section's character and the aiming function is the minimal cost.The pure flexural and axial-compressed members of concrete-filled steel tubular structure were used as the examples which were optimized for the design model.The results were compared with the ones of improved Genetic Algorithm in literate [4].The results showed that the column and the beam obtained the global optimal solution after 58 and 52 iterations separately.The algorithm jumped out of the analysis of the complicated-interactive mechanism between steel tube and confined-concrete,was simple and efficient.