Cast steel tubular joints have the advantages of excellent integrity and low stress concentration levels， making them widely used in spatial structures. However， current design codes do not provide verification formulas for such joints， which forces engineers to rely on finite element analysis for safety validation. This often leads them to search for reasonable design schemes through a trial-and-error approach. To improve the design efficiency and quality， this paper first proposes a novel shape optimization method， which develops the parametric modeling of joints based on subdivision surface， realizes the automatic finite element analysis of joints through secondary development in ABAQUS， and adjusts the shape of the joints with the genetic algorithm. Secondly， the objective function of the shape optimization problem under multiple load cases is constructed using four objective merging methods， namely linear weighted method， compromise programming method， ε-constraint method， and minimax method， respectively， and the shape optimization problem under multiple load cases is transformed into a single-objective optimization problem. Finally， the above methods are applied to a cast steel tubular joint in a cylinder shell， and the results demonstrated that the proposed method can reduce the peak Mises stress of the joint by 44%~60%， effectively reducing the stress level of the joints without increasing the joint volume significantly. At the same time， among the objective merging methods， the maximum minimization method can balance different load cases most effectively and exhibit similar peak Mises stress under different load cases， making the most use of the material.