Investigating aerodynamic shape design space reconstruction using proper orthogonal decomposition (POD)

In order to increase optimization precision in the design process of aircraft, the design parameters have augmented; this makes the design process more complicated and extends the design cycle. Aiming at this issue, reconstruction method of aerodynamic design space using Proper Orthogonal Decomposition is investigated. Main objective is divided into two aspects: one is reduction of the number of aerodynamic design parameters in design process, the other is increase of the ratio of samples which satisfy design constraints. In the Hicks-Henne parameterization space and reconstructed space by POD, 20000 samples are chosen by LHS method. It is illustrated that the ratio of Hicks-Henne parameterization space satisfying thickness constraints is less than 25%, while the ratio of reconstructed space by POD is about 70%. Therefore, after space reconstruction by POD, sample quality is enhanced significantly. Moreover, the number of optimization parameters is decreased. The influence of traditional constraint handling methods and reconstructed design space is investigated with the optimization of Cl/Cd of RAE2822 airfoil. Traditional constraint handling methods include penalty function method and death penalty. It is shown that death penalty method is slightly better than penalty function method from optimization results. Additionally, there are no penalty factors in death penalty process, which is convenience to use. Comparing the optimization results before and after reconstructed design space, the improvements of Cl/Cd are 27.61% and 28.20% respectively for two traditional penalty handing methods in Hicks-Henne parameterization space, while the improvements are 28.51% and 30.63% respectively in reconstructed design space by POD. Therefore, the optimization precision is almost the same before and after space reconstruction, and the number of design parameters is decreased by half; this makes design efficiency improve significantly.