A genetic algorithm for sizing optimization of geometrically non-linear dome structures

Afonso C.C. Lemonge¹; Patricia H. Hallak¹

¹ Federal University of Juiz de Fora

doi:10.20906/CPS/CILAMCE2015-0175

Resumo

In this paper a genetic algorithm is proposed to solve the weight minimization problem of a dome considering sizing design variables. The dome is assembled from standard modules composed by ten bars which are grouped into seven distinct design variables. Previous work done by the authors considered the geometrically nonlinear behavior of the domes and in this paper loads due the self-weight are included in the analyses during the optimization process. In this way, the load conditions is the sum of the static and the self-weight loads, observing that the self-weight loads vary along the optimization process since the cross-sectional areas of the bars change. In addition, cardinality constraints are introduced to define alternative member groupings of the bars by using a genetic special encoding. This type of optimization can be considered an interesting feature because it allows the designer to automate the search for the best solutions for a given maximum number of distinct cross-sectional areas. Several comparisons are performed using a 120-bar dome as test-bed including linear and non-linear analysis, with and without the consideration of the self-weight loads. Finally a trade-off curve is provided in order to give to the designer several options of optimized structures.

Palavras-chave: Geometrically non-linear analysis; Structural optimization; Genetic algorithms