Research

Design Optimization of Vortex Generators for Aircraft

Wing oscillation, which has harmful effects on safe and comfortable flight, can be caused when an aircraft encounters disturbances such as air turbulence and wind gusts. An array of small vanes is frequently employed as vortex generators (VGs) and equipped on the upper surface of the wing to alleviate the wing oscillation.

Design optimization of VGs was performed by using the supercomputing system to propose the effective design suppressing the oscillation independent of the engineer’s experience. The optimized VG not only suppressed the oscillation by mixing air but also minimized the aerodynamic drag acting on VG itself because VG generally increases the drag and fuel consumption. Moreover, beneficial design rules were extracted from the optimization results and should contribute the next generation aircraft design.

Surrogate Models for Efficient Global Optimization

Evolutionary algorithms (EAs) such as genetic algorithm (GA) and particle swarm optimization (PSO) are well-developed and frequently employed for design optimization in various engineering fields because they can robustly find out the global optimal solution with population-based exploration. However, optimization in real-world problems is sometimes time consuming and computationally expensive in the evaluation of objective functions which represent performances of products. Numerical simulation of flow fields by means of computational fluid dynamics (CFD) described above is a typical example of expensive optimization problems.

Surrogate models are applicable to optimize these expensive problems. In surrogate-base optimization, an expensive objective function is replaced by a surrogate model which estimates objective function values at any design point from a few sample points where real objective function values are evaluated through expensive computations. EAs explore the optimal solution on the surrogate model, thus, it is important to construct more accurate surrogate models with less sample points. In order to solve expensive optimization problems efficiently, the expected improvement (EI) of penalty-based boundary intersection (PBI) named EPBII/EIPBII, Kriging model with coordinate transformation (KCT), and RBF/Kriging hybrid model (RK) have been proposed by the author, and the EI coupled with mutual information (EI-MI) is currently under development.

Google Sites

Report abuse