New strategies for the aerodynamic design optimization of aeronautical configurations through soft-computing techniques
AuthorsAndrés Pérez, Esther
IdentifiersPermanent link (URI): http://hdl.handle.net/10017/15761
DirectorSalcedo Sanz, Sancho
Aerodinámica-Proceso de datos
Aviones-Diseño y construcción
Description / Notes
Premio Extraordinario de Doctorado de la UAH en 2013
Lozano Rodríguez, Carlos, codir.
This thesis deals with the improvement of the optimization process in the aerodynamic design of aeronautical configurations. Nowadays, this topic is of great importance in order to allow the European aeronautical industry to reduce their development and operational costs, decrease the time-to-market for new aircraft, improve the quality of their products and therefore maintain their competitiveness. Within this thesis, a study of the state-of-the-art of the aerodynamic optimization tools has been performed, and several contributions have been proposed at different levels: -One of the main drawbacks for an industrial application of aerodynamic optimization tools is the huge requirement of computational resources, in particular, for complex optimization problems, current methodological approaches would need more than a year to obtain an optimized aircraft. For this reason, one proposed contribution of this work is focused on reducing the computational cost by the use of different techniques as surrogate modelling, control theory, as well as other more software-related techniques as code optimization and proper domain parallelization, all with the goal of decreasing the cost of the aerodynamic design process. -Other contribution is related to the consideration of the design process as a global optimization problem, and, more specifically, the use of evolutionary algorithms (EAs) to perform a preliminary broad exploration of the design space, due to their ability to obtain global optima. Regarding this, EAs have been hybridized with metamodels (or surrogate models), in order to substitute expensive CFD simulations. In this thesis, an innovative approach for the global aerodynamic optimization of aeronautical configurations is proposed, consisting of an Evolutionary Programming algorithm hybridized with a Support Vector regression algorithm (SVMr) as a metamodel. Specific issues as precision, dataset training size, geometry parameterization sensitivity and techniques for design of experiments are discussed and the potential of the proposed approach to achieve innovative shapes that would not be achieved with traditional methods is assessed. -Then, after a broad exploration of the design space, the optimization process is continued with local gradient-based optimization techniques for a finer improvement of the geometry. Here, an automated optimization framework is presented to address aerodynamic shape design problems. Key aspects of this framework include the use of the adjoint methodology to make the computational requirements independent of the number of design variables, and Computer Aided Design (CAD)-based shape parameterization, which uses the flexibility of Non-Uniform Rational B-Splines (NURBS) to handle complex configurations. The mentioned approach is applied to the optimization of several test cases and the improvements of the proposed strategy and its ability to achieve efficient shapes will complete this study.