Key points are not available for this paper at this time.
Abstract An efficient optimization method based on approximate gradient analysis was proposed to reduce the substantial computational burden in expensive optimization problems, in which the objective functions are evaluated with high-fidelity analyses in one-dimensional linear search, while the gradients are evaluated with an approximation model during the optimization process. Taking advantage of the similarity of CFD models with different fidelities, we employ a less time-consuming low-fidelity model as its main section, and use the gradient information of the high-fidelity model at the initial point as an amendment. The proposed procedure was applied to the aerodynamic optimization of a well-designed supercritical wing, resulting in obvious drag reduction, but the CPU time is only 28.5% as its SQP counterpart. The investigation indicates the proposed method is suitable for aerodynamic improvement of an existing baseline configuration to meet specified engineering requirements.
Liu et al. (Thu,) studied this question.