SURROGATE MODELLING FOR EFFICIENT DESIGN OPTIMIZATION OF COMPOSITE AIRCRAFT FUSELAGE PANELS

摘要

This paper considers a new methodology for large scale optimization that involves structural design analyses by finite element method (FEM) models of a composite aircraft fuselage barrel. A natural decomposition of the overall design optimization problem into two levels, i.e. the fuselage barrel level and the level of the individual fuselage panels, exists in these analyses. Therefore we implemented a variant of the multi-level optimization methodology known as BLISS (Bi-Level Integrated System Synthesis), which we first apply to a relatively simple test case based on the 10-bar truss optimization problem. In the fuselage barrel optimization, the decomposition into two levels allows for fast analysis with relatively coarse models of the whole fuselage barrel, while much more detailed models are used for the panel level analyses. These detailed panel models may include specific composite material properties like lay-ups and fiber orientations and detailed geometric aspects of frames and stringers. For further speed-up of the optimization process we apply surrogate modeling methods for the representation of the sub-system behavior. In this way the design variables that originate from the different model levels can be incorporated in a computationally efficient manner.