The invention discloses a multi-fidelity design optimization method for a shell structure of an autonomous
underwater vehicle. A multi-fidelity method of an addition bridging function based on a
Kriging model is adopted, a multi-start optimization method and a unique self-adaptive model updating strategy are combined, and accordingly design optimization of the shell structure is conducted. The multi-fidelity method is adopted to construct an agent model of which the maximum stress of the shell structure avoids the
instability of the
critical load, when the agent model is constructed, a large amount of low-fidelity calculation and analysis are used, the precision of the model is improved, and meanwhile the total calculation cost of the optimization process of the structure design is reduced. During
sequence optimization, a new agent model is adopted to update the strategy; meanwhile, an optimum point of the current model and an MSE maximum point of the current model are added into high and
low fidelity sampling points; through the MSE maximum point of the added model, the global space can be sufficiently explored, and accordingly the total precision of the agent model is improved; accordingly, an optimum design scheme is obtained, and the precision and efficiency of design optimization are improved.