Abstract: The coupling of quality states between different processes in metal 3D printing and the inherent conflicts among multiple quality indicators make it difficult to perform continuous quality prediction and construct a multi-objective function encompassing multiple objectives, thereby restricting the intelligent optimization of process parameters. Therefore, a research on an intelligent optimization system for metal 3D printing process parameters based on a multi-objective genetic algorithm is proposed. A random forest is employed to build a 3D printing quality prediction model, which quickly predicts the quality indicators corresponding to any parameter combination in the digital space through bootstrap sampling and decision tree ensemble, while considering the quality transfer effects between layers and processes, thus achieving continuous quality prediction from single process to multiprocess. A sensitivity function is introduced to quantitatively characterize the influence degree of various process parameters on the target quality function, and a multiobjective function is designed with the minimization of the total deformation energy of the component and the best quality of the metal component as optimization objectives. A multiobjective genetic algorithm is used for global search, obtaining the Pareto optimal solution set through fast nondominated sorting, crowding distance calculation, and elite retention strategy, and the analytic hierarchy process is employed to determine the optimal process parameter combination. Test results show that the system can predict the printing quality under different printing process parameters; the hypervolume indicators are all above 0.8, with the maximum value close to 1.0, indicating excellent solving performance and the ability to determine the optimal process parameters.