Wing shape optimization using FFD and twist parameterization

Numerical shape optimization for aerodynamic problems can lead to improved designs than what is possible by conventional methods. Gradient-based optimization methods require the development of adjoint codes which can be a lengthy exercise and is also not very mature for RANS-based approaches. Furthermore, they may lead to locally optimal solution only, which is not desirable, especially in a preliminary design stage. Gradient-free methods are in this sense attractive and they also have the potential to give globally optimal solutions. The use of Euler/RANS codes for modeling the flow in a design context can be computationally expensive due to the need to evaluate many designs. However, the use of parallel computers and sophisticated surrogate models makes this feasible today. A shape optimization exercise requires the development and coupling of several elements in an automatic chain. (1) A shape modeling system which converts the design variables into a shape. (2) A grid generation program that generates a surface grid and a volume grid. (3) A CFD solver, and (4) an optimizer. A shape parameterization system typically involves a CAD tool which must be coupled to a grid generator. Every time the shape design variables are changed, a new grid has to be generated without human intervention. For complex problems and especially involving RANS models, fully automatic grid generation can be difficult or impossible. An alternative approach is