基于新型多可信度多项式混沌-Kriging模型的高维全局气动优化设计方法

基于代理模型的高效全局优化设计方法已经成为满足先进飞行器气动设计要求最重要的途径之一。然而随着飞行器设计指标不断提升，基于代理模型的精细化气动优化方法需要利用更高可信度的CFD模拟和更多的独立设计变量，面临严峻的维数灾难挑战，严重限制了其在飞行器复杂气动设计中的应用。基于此，申请人前期提出了自适应多可信度多项式混沌-Kriging模型，较分层Kriging等显著提升了建模效率和泛化能力。项目拟结合偏最小二乘和基于数据驱动的多层多可信度多项式混沌-Kriging发展一种新型的高维多可信度代理模型。新方法将不仅显著提高对复杂高非线性模型响应的泛化能力，还能有效缓解代理模型面临的维数灾难难题。同时，项目拟发展一种基于该新型模型的全局并行加点优化方法，进一步提升高维优化效率。基于这些研究，项目将突破代理优化方法遭遇的瓶颈，建立高效的高维全局精细化设计理论和方法，为发展先进飞行器设计方法奠定基础。

Surrogate-based efficient global design optimization method has become one of the most important techniques to meet the advanced design requirements of aircraft. However, with enhanced design indices, surrogate-based global aerodynamic optimization method is required using the higher-fidelity Computational Fluid Dynamics (CFD) simulation and more independent design variables, which is faced with the serious challenges of the curse of dimensionality, and severely limits its application in the complex aerodynamic design of the aircraft. Based on these findings, a novel adaptive multi-fidelity polynomial chaos-Kriging (AMF-PCK) surrogate model was proposed in the preceding research of the applicant, which significantly improves the modeling efficiency and the generation ability compared with the popular hierarchical Kriging and so on. This project, by combining the partial least squares (PLS) analysis and a data-driven multi-level multi-fidelity polynomial chaos-Kriging (MLMF-PCK) model, aims to develop a novel high-dimensional multi-fidelity surrogate modeling approach. This method will not only significantly improve the generation ability of surrogating the complex highly-nonlinear and/or high-order model’s response and landscape, but also efficiently alleviate the problem of the curse of dimensionality encountered by the surrogate model. Meantime, this project plans to develop a parallel MLMF-PCK-based global infill-optimization method, further improving the efficiency of the high-dimensional aircraft design optimization. Based on these proposed studies, the project will break through the bottleneck of the surrogate-based optimization method, establish the high-dimensional global optimization design theory and method, and lay the foundation for the development of the advanced aircraft design method.