Innovative multilevel techniques for structural optimization.

We address the structural optimization problem of sizing an aircraft fuselage. The problem consists in computing the dimensions of the dierent elements constituting a fuselage mini- mizing the total mass subject to some mechanical constraints. Mathematically, the problem may be formulated as a very large nonlinear optimization problem subjected to several non- linear inequality constraints. We show that this problem possesses a natural hierarchical structure that can be exploited by a multilevel approach. This approach is innovative in the industrial sector and represents a promising alternative to the commonly employed decom- position strategies. Hence, we propose a multilevel procedure which embeds the Recursive Multilevel Trust Region method [1] into an Augmented Lagrangian framework. Some results on both academic and industrial test cases are presented.