Surrogate-Based Optimum Design for Stiffened Shells with Adaptive Sampling

For cylindrical stiffened shells under internal pressure and nonuniform axial compression, a load-controlled loading pattern based on the nonlinear explicit dynamic analysis is employed. Because of the nonuniform load, numerous variables are involved in structural design, whereas it is practically impossible to perform the optimization for the overall variables simultaneously, even if a surrogate model is adopted. The framework of a bistep surrogate-based optimization with adaptive sampling is presented in this study, which aims to build relatively high-fidelity surrogate models with less computational cost for complex engineering designs. A three-dimensional rigid frame is first established to validate the proposed method. Results indicate that the optimum design obtained by the proposed method is potentially close to the global optimum design. Then, a typical fuel tank configured for use in a launch vehicle is used to illustrate the proposed method. Comparisons of direct surrogate-based optimization and the proposed method for two examples show that the bistep surrogate-based optimization with adaptive sampling can achieve a larger weight saving with less computational cost. Consequently, it can be concluded that the proposed method provides an efficient and practical approach to finding the optimum design for structure with numerous variables and various variable types.