关键词: 离散伴随优化, 非支配解优化, 多目标优化, 虚拟Pareto可行解集, 相关性分析

Abstract: To study multi-objective optimization in the high-dimensional target space, a multi-objective integrated design for the aerodynamic shape of large civilian aircraft is carried out by using the AMDEsign, a self-developed software for large-scale parallelization and distributed optimization of aircraft aerodynamic shape. The multi-objective optimization of the digitized model of a wide-body aircraft is performed by two typical modules of AMDEsign:Principal Componet Analysi (PCA) and the discrete adjoint method. The discrete adjoint method is combined with the virtual pareto solution set method to provide an effective directional choice for the weight coefficient. The design results show that the principal component analysis can effectively identify the correlation of the objective function, and the virtual feasible solution set method has high efficiency. The design also fully utilizes the advantages of high efficiency of discrete adjoint and the prediction capabilities of the guided weight function. The multi-point optimized configuration demonstrate significant improvements in cruise lift-drag ratio, drag convergence characteristics, and resistance divergence. The integrated design method proposed is shown to be simple, efficient and applicable in practice.

Key words: discrete adjoint optimization, non-dominated solution optimization, multi-objective optimization, virtual Pareto solution set, correlation analysis