Multiobjective optimization of an aircraft wing design with respect to structural and aeroelastic characteristics using neural network metamodel

Abstract

The multidisciplinary design optimization (MDO) is an important concept that focuses on bringing together disciplines involved with machine design. It is possible to use MDO in any stage of an aircraft design, that is in the conceptual, preliminary, or detailed design, as long as the numerical models are fitted to each of these stages. This work describes the development of a multidisciplinary design optimization applied to flexible aircraft wings, with respect to structural and aeroelastic characteristics. As tools for the aircraft designer, the numerical models must be fairly accurate and fast. Therefore, metamodels for the critical flutter speed prediction of aircraft wings were considered, thereby reducing significantly the computational cost of the optimization. For this purpose, artificial neural networks (NN) metamodeling is evaluated, based on their inherent properties in dealing with complex mappings. The NN metamodel is prepared using an aeroelastic code based on finite-element model coupled with linear potential aerodynamics. Results of the metamodel performance are presented, from where one can note that the NN is well suited for flutter prediction. Multiobjective optimization (MOO) using the genetic algorithm (GA) based on non-dominance approach is considered. The objectives were the maximization of critical flutter speed and minimization of structural mass. One case study is presented to evaluate the performance of the MOO, revealing that overall optimization process actually achieves the search for the Pareto frontier.