Exploring multi-stage shape optimization strategy of multi-body geometries using Kriging-based model and adjoint method

Abstract This paper deals with an efficient and high-fidelity design strategy for wing-body configuration. According to the nature of the design space and the number of design variables, aerodynamic shape optimization is carried out at each design stage by using a selective optimization strategy. In the first stage, global optimization techniques are applied to wing planform design with a few geometric design variables. In the second stage, local optimization techniques are employed to wing surface design with many design variables, which can maintain a sufficient design space with a high Degree of Freedom (DOF) geometric change. For global optimization, the Kriging method in conjunction with a Genetic Algorithm (GA) is used. A searching algorithm exploiting Expected Improvement (EI) design points is introduced to efficiently enhance the quality of the initial Kriging model for the wing planform design. For local optimization, a discrete adjoint method is adopted to obtain sensitivity information by fully hand-differentiating the three-dimensional Euler and N–S equations on an overset mesh topology. By the successive use of the global and local optimization methods, the drag of a multi-body aircraft configuration can be minimized for inviscid and viscous flow conditions while the baseline lift and wing weight are maintained. Throughout the design process, the performances of the test models are improved, compared to those with the single stage design approach. The performance of the proposed multi-stage design framework is evaluated by the drag decomposition method.