In response to the curse of dimensionality in the lightweight and high-dimensional optimization design of complex structures of Blended wing body fusion underwater gliders (BWBUG), a multi-source data-driven optimization design of the skeleton structure of BWBUG was carried out. Firstly, a multi fidelity global optimization based on HK model (MFGO-HK) algorithm is adopted to solve the computationally expensive black box optimization problem. Secondly, establish a parameterized model of the skeleton structure and determine the high and low-fidelity finite element simulation model of the skeleton structure through finite element simulation analysis. Finally, with the optimization goal of lightweight skeleton structure and the constraint conditions of strength, stiffness, stability, and external interference, an engineering optimization mathematical model and optimization framework for skeleton structure were established. The MFGO-HK was used to optimize the durable skeleton, resulting in a 70.3% reduction in weight compared to the initial plan. Compared with other optimization algorithms, MFGO-HK has stronger optimization ability and can effectively balance optimization accuracy and computational cost.