Local and Global Search Methods for Design in Aeronautics.

Optimization is used at all stages of aircraft design. In the early phases (also called preliminary or conceptual design) multicriteria Pareto optimization [11] based on heuristic laws is carried out. Conversely, in the late design phases, accurate structural and aerodynamics predictions are required to steer small changes of a baseline shape through mono-objective optimization. This article is devoted to the local and global optimization methods actually used during this second part of the design process. Through discretization, parametrization, and numerical simulation, the design problem can be formulated mathematically as a finite-dimensional optimization. Hence a good knowledge of global and local optimization algorithms is important to aerodynamic design engineers. The classical algorithms which are currently used at Onera for shape design are presented. Besides, many local optimization algorithms require the gradient of the functions of interest with respect to the design parameters. The different ways to compute those derivatives - often called “sensitivities” - are also described. Numerous 2D and 3D applications were dealt with at Onera using the methods described; they are presented in a companion article [14] and also briefly described in a course [1].