03.1 - Aerodynamics - CFD Methods and Validation
COMPARISON AND APPLICATION OF DIFFERENTS TECHNIQUES OF MULTI-OBJECTIVE OPTIMIZATION FOR AERODYNAMIC WING DESIGN
N. Gautierš, N. Manzanares-Filhoš, E.R. Da Silvaš; šUNIFEI, Brazil
In real-world engineering design problems the designer often seeks to optimize multiple and conflicting merit functions or objectives relating to the performance of the given design. Traditionally, multiobjective optimization has been carried out by forming a single, composite objective function comprising a weighted summation usually at the discretion of the designer of each of the objectives involved. Several techniques are available today for design through numerical optimization. In particular, in the field of aerodynamic design, the techniques more properly related to direct optimization include gradient based methods and more recent approaches such as automatic differentiation, control theory based methods, genetic algorithms (GAs) or based on population set. This paper discusses, the comparison of the different multi-objective optimization techniques, applied to aerodynamic wing design, although commercial code, using different population set algorithms, like: (i) the Non-dominated Sorting Genetic Algorithm II (NSGA-II); (ii) A single and Multi Objective Simulated Annealing (MOSA); (iii) Multi-Objective based on Game Theory coupled with Simplex algorithm (MOGT); (iv) Multi-Objective Particle Swarm (MOPSO) and (v) Multi Objective Genetic Algorithm (MOGA-II). All algorithms are capable of performing global optimization tasks efficiently. After the evaluation of each algorithm, will be chosen the one that obtained the best solutions found through the Pareto border or other criteria of the choice. One case study is presented for testing the efficiency and robustness of the proposed methodology with each technique an aerodynamic aircraft wing design application. Since the main objective of this paper is a prospective nature, was employed two types of solver fidelity flow computation, first a low-fidelity flow computation code solver for aerodynamic wing design, which was based on first order 3D panel method, for lift and induced drag coefficients calculation, an