Aplicação de regressão de vetores de suporte na otimização em flambagem e pós-flambagem de estruturas compósitas laminadas

Abstract

Laminated composite materials are applied in many industrial sectors, particularly in competition automotive and aerospace fields, since they have strength-to-weight and stiffness-to-weight ratios much higher than the metals in general. Structures made by these materials are usually thin and hence they are subject to buckling. Traditional design requirements usually take into account the buckling, but in some cases the design is conservative since the structure can still be functional in the postbuckling regime. However, the behavior in this regime is nonlinear, in addition of being difficult to evaluate when the failure of the structure takes place, which makes the analysis more complex and computational expensive if compared to a linear buckling analysis. Within this context this work is inserted, which aims to find the orientations of the fibers that maximize the buckling and postbuckling load of composite structures using metamodels in the optimization process to alleviate the computational cost. Two metamodeling techniques are used and tested: artificial neural networks and support vector regression, with emphasis on the latter. In combination with the metamodels, two recently developed metaheuristics, the harmony search algorithm and the firefly algorithm, are employed. Several problems, with different levels of difficulty, are presented and discussed. The best optimization results were obtained with the firefly algorithm associated with the support vector regression metamodel, showing that these techniques are promising to solve this class of problems. One of the main contributions of this thesis is the adaptation/implementation of support vector regression for layup orientation sequence problems of composite structures, in particular for buckling and postbuckling optimizations. Moreover, advances were made in the modeling of the behavior and optimization in postbuckling regime using failure and damage criteria for composites.