Surrogate-Based Structural Optimization of Composite Shell Cylinder for Launch Vehicles

Lightweight and high-buckling-load capacity stiffened structures, such as lattice, skin-stringer, and sandwich panels, are commonly employed in rocket design. Although individual structures have been widely optimized, a comprehensive comparative study is essential for practical design selection. This study aims to explore the global optimal solution, representing the lightest achievable design, through a systematic comparison of different structural architectures. The optimization objective is to minimize structural mass under the primary constraint of critical buckling load. To reduce the computational cost of FEM-based buckling analysis, a Hierarchical Kriging (HK) surrogate model is used. HK models are constructed based on the relationships between samples of design variables generated by Latin Hypercube sampling（LHS） and FEM buckling analysis. The predictive accuracy of the HK model is maintained by incorporating an EF function-based update strategy. The optimization process is iterated, updating the HK model, until the error between the estimated buckling load from the HK model and the buckling load obtained from the FEM analysis is less than 1% at the optimal solution. The results demonstrate that the sandwich structure represents the global optimal solution among the architectures studied. Furthermore, the distinct characteristics identified for each structure emphasize the importance of selecting the appropriate stiffening architecture based on specific design objectives.