For the exhaust device of the soft recovery system in intelligent ammunition, the sparse grid response surface model and multi-objective genetic algorithm (MOGA) are combined to optimize its design. The initial parameter model is used as the boundary condition of the flow field, and the pressure distribution on the wall of the real flow field is analyzed in combination with the interior ballistic condition. Through the fluid-solid coupling analysis, the stress and deformation distribution of the exhaust pipe is obtained, and the final model is modified to obtain the optimal solution of the size parameters of the exhaust pipe. The results show that the design can lay a theoretical foundation for the fine analysis and test of the recovery system.