A Genetic Algorithm for Multiobjective Design Optimization in Aerodynamics and.pdfVIP

A Genetic Algorithm for Multiobjective Design Optimization in Aerodynamics and Electromagnetics ¨ 1 ¨ 1 2 1 R. A. E. Makinen , P. Neittaanmaki , J. Periaux and J. Toivanen Abstract. The solution of a multiobjective multidisci- suitable for multiobjective problems; see, for example, plinary design optimization (MDO) using a genetic algo- [3]. In this kind of problems, a design is optimal when rithm (GA) is considered. The objective functions in the it is nondominated or by using another term Pareto opti- optimization problem measure the aerodynamic feasibil- mal. A design is nondominated if no other feasible design ity based on the drag and lift coefficients and the elec- exists which is better with respect to any objective and at tromagnetic feasibility based on the backscatter of the least equally good with respect to all the other objectives. two-dimensional airfoil designs. The flow and backscat- Usually, in the practical solution of multiobjective ter are modeled by the thin-layer Navier–Stokes equa- problems, the task is to find several Pareto optimal solu- tions and the time-harmonic Maxwell equations, respec- tions. This can be computationally a very laborious task. tively. Numerical experiments illustrate the above evolu- Genetic algorithms (GA) can be adapted to this kind of tionary methodology on a parallel computer. problems and they are naturally parallel. Thus, GAs can be used efficiently in parallel computers which can offer