Global aerodynamic design optimization based on data dimensionality reduction

• DOI: 10.1016/j.cja.2018.02.005
• Publication Date: 2018-02-15T18:24:40Z
• Authors: Yasong QIU; Junqiang BAI; Nan LIU; Chen WANG
• Keywords: Global Optimization; Derivative-Free Optimization

In aerodynamic optimization, global optimization methods such as genetic algorithms are preferred in many cases because of their advantage on reaching optimum. However, for complex problems which large number design variables needed, the computational cost becomes prohibitive, and thus original strategies required. To address this need, data dimensionality reduction method is combined with methods, forming a new system, aiming to improve efficiency conventional optimization. The system involves applying Proper Orthogonal Decomposition (POD) space while maintaining generality space. Besides, an acceleration approach samples calculation surrogate modeling applied reduce time providing sufficient accuracy. optimizations transonic airfoil RAE2822 wing ONERA M6 performed demonstrate effectiveness proposed system. both cases, we manage from 20 10 42 respectively. converges faster it takes 1/3 total traditional converge better design, significantly reducing overall improving method.