Abstract: Objectives The series multi ball composite shell structure can be used as the main structure of deep-sea space stations. To ensure the full performance of its precision testing equipment, this paper conducts optimization design research on the acoustic and vibration performance of underwater series multi ball composite shells. Methods Firstly, establish its finite element model, calculate and analyze its underwater acoustic radiation characteristics; Then, with the model quality as the constraint condition and the radiation noise under vertical excitation as the optimization objective, a uniform experimental design is established with the geometric parameters of the connecting structure as the design variables. The RBF(Radial Basis Function) neural network is used to establish a multidimensional mapping model between the design variables and the optimization objective. The PSO(Particle Swarm Optimization) algorithm is used to optimize the underwater radiation noise of the model, and verified through finite element method; Finally, underwater acoustic radiation experiments were conducted, and the experimental test values were compared with the simulation values, proving the accuracy of the simulation results. Results After optimization, the total underwater radiation sound power level of the series multi sphere composite shell under vertical excitation is reduced by 2.92 dB, and the mass is reduced by 0.061 t.Conclusions The research provided new ideas for optimizing structural noise design methods.