DATA DRIVEN MULTI-FIDELITY APPROXIMATE MODELING OPTIMIZATION OF STIFFENED CONICAL SHELL FOR LARGE LAUNCH VEHICLE

In order to improve the load-bearing efficiency of the stiffened conical shell in large launch vehicle, the lightweight design of the stiffened conical shell was carried out via a data-driven multi-fidelity approximate modeling optimization method.Aiming at the problems such as low efficiency and insufficient accuracy of the single fidelity approximate modeling optimization method, a data-driven multi-fidelity approximate modeling optimization framework was built based on variable-fidelity expected improvement (VF-EI) point criterion，and accordingly the optimization design of stiffened conical shell structure was carried out.Based on the finite element models of stiffened conical shells with different mesh sizes, a Co-Kriging multi-fidelity approximate model for the collapse load of stiffened conical shells was established.In the optimization iteration, multi-fidelity sampling points were generated by using VF-EI point criterion, and the global and local approximation accuracy of Co-Kriging multi-fidelity approximation model was improved sequently.Moreover, the optimization efficiency and accuracy of the proposed method were demonstrated by comparing with radial basis function approximation model and Kriging model. Besides, 11.5% weight reduction of the optimized stiffened conical shell structure is obtained compared with the initial design, which has certain engineering application value.