Dynamic data driven applications systems (DDDAS) for multidisciplinary optimisation (MDO)

Abstract

[ES] Nowadays, the majority of optimisation processes that are followed to obtain new optimum designs involve expensive simulations that are costly and time comsuming. Besides, designs involving aerodynamics are usually highly constrained in terms of infeasible geometries to be avoided so that it is really important to provide the optimisers effective datum or starting points that enable them to reach feasible solutions. This MSc Thesis aims to continue the development of an alternative design methodology applied to a 2D airfoil at a cruise flight condition by combining concepts of Dynamic Data Driven Application Systems (DDDAS) paradigm with Multiobjec- tive Optimisation. For this purpose, a surrogate model based on experimental data has been used to run a multiobjective optimisation and the given optimum designs have been considered as starting points for a direct optimisation, saving number of evaluations in the process. Throughout this work, a technique for retrieving experi- mental airfoil lift and drag coefficients was conducted. Later, a new parametrisation technique using Class-Shape Transformation (CST) was implemented in order to map the considered airfoils into the design space. Then, a response surface model considering Radial Basis Functions (RBF) and Kriging approaches was constructed and the multiobjective optimisation to maximise lift and minimise drag was under- taken using stochastic algorithms, MOTSII and NSGA. Alternatively, a full direct optimisation from datum airfoil and a direct optimisation from optimum surrogate- based optimisation designs were performed with Xfoil and the results were compared. As an outcome, the developed design methodology based on the combination of surrogate-based and direct optimisation was proved to be more effective than a single full direct optimisation to make the whole process faster by saving number of evaluations. In addition, further work guidelines are presented to show potential directions in which to expand and improve this methodology.