The present paper describes the efforts on the construction of a computational framework for 2-D and 3-D aerodynamic optimizations. The creation of the framework is an attempt to generate a design environment capable of coupling various tools from different levels of complexity and with diverse functionalities. The conceptual framework is developed to be inserted into daily activities of an aerodynamic CFD group. The framework is implemented for both Windows and Linux-running platforms, and it

more »

... s augmented by a user-friendly graphical interface. Usage of the framework is illustrated in the paper by 2-D and 3-D aerodynamic optimization of a cruise configuration for different flight conditions. The aspects investigated include the influence of the number of individuals on the aerodynamic coefficients, the effect of boundary layer transition location on the final optimized shape, and the benefits of the solver fidelity level as compared to the computational cost. Moreover, the use of a neural network is evaluated in order to analyze the benefits that this methodolgy can bring to the implemented framework in terms of computational cost. * Graduate Student, ap-antunes@uol.com.br. applications, a set of subroutines, written in Fortran-90, and templates are developed. Furthermore, in order to confer a certain level of flexibility in the use of the framework, it is implement for both Windows and Linux-running platforms. Moreover, a GUI is also developed to serve as a user-friendly interface between the user and the coupled tools. At the present stage of the framework development, only the aerodynamic activities are integrated. However, in a second moment, other disciplines will be coupled to it in order to create an MDO environment. The framework provides an improvement in the aerodynamic design efficiency, through the integration of various design tools with an optimization algorithm. Such integration intends to eliminate most of the time spent during the preparation of the simulation. The saved time can be used to increase the number of simulations or to perform more thorough and critical analyses of the results. The optimization process might be simply defined as finding the most suitable solution for a specific problem being analyzed. Many times the problem at hand might appear simple at a first glance. However, the complexity increases with the number of problem variables and the number of objectives to be achieved. A large amount of numerical and mathematical methods can be used to perform the optimization of a problem. One of the reasons for the existence of a vast number of techniques is the fact that many of them have been developed to deal with specific types of optimization problems. The available optimization methodologies can be classified as: • classical methods; • evolutionary methods. Examples of classical methods are geometric programming, 2 nonlinear programming, 3 linear programming, 4 and conjugate gradient methods, 5 among others. On the other hand, genetic algorithms, 6 artificial life, 7 and ant colony 8 techniques can be cited as examples of evolutionary methods. There are advantages and disadvantages on both approaches. Usually, classical methods are blamed for getting stuck on suboptimal solutions and for having optimal solution which depend on the chosen initial condition. On the other hand, the evolutionary methods are more expensive due to the fact that they work with a large number of solutions. In the present case, the decision was to use genetic algorithms (GA's) 910 as the main optimization methodology. GA's are search optimization methods that use principles of natural genetics and natural selection. In such methods, the possible solutions for a certain problem are represented by some form of biological population, which evolve over generations to adapt to an environment by selection, crossover and mutation. Instead of working with a single solution at each iteration of the process, a GA works with a number of solutions, known as a population. The optimization processes can be classified, with regard to the objective function, as single-objective optimizations and multi-objective optimizations. For the first optimization type, a single objective function drives the search for the best, or most feasible, solution, and the optimization is expected to obtain a single solution. On the other hand, for multi-objective optimizations, the process is driven by more than one objective function. In the real world, most of the optimization problems are multi-objective and, usually, the objectives are conflicting with each other. Such conflicting characteristics do not allow for a single solution, but one obtains a set of solutions, which are commonly known as a Pareto front. The two important goals in multi-objective optimization are: • to find solutions as close as possible to the Pareto-optimal solutions; • to find solutions as diverse as possible in the resulting non-dominated front. Results for single-objective and multi-objective optimizations are presented in the paper.