Is there a need for fully converged CFD solutions? Global extremum seeking applied to aerodynamic shape optimisation

Abstract

Optimisation of aerodynamic shapes using computational fluid dynamics approaches has been successfully applied over a number of years, however the typical optimisation approaches employed utilise gradient algorithms that guarantee only local optimality of the solution. While numerous global optimisation techniques exist, they are usually too time consuming in practice. In this paper we show that interpreting the convergence of computational fluid dynamics solvers as plant dynamics allows recent results in global extremum seeking to be deployed. This alleviates the computational burden of requiring full convergence of the computation fluid dynamics solver. The approach is demonstrated on a simple example involving drag minimisation on a NACA aerofoil.