Application of efficient surrogate modeling to aeroelastic analyses of an aircraft wing

A nonlinear, transient surrogate model is constructed to replace the computational fluid dynamics solver within the aeroelastic analysis of a detailed aircraft wing with variable fuel tank levels during cruise flight. This surrogate model allows high-fidelity analyses at a fraction of the original computational cost with satisfactory accuracy. During the model construction a parameter reduction via proper orthogonal decomposition is combined with a radial basis function artificial network for system identification and nonlinear effects. Transient forced motion computational fluid dynamics analyses based solely on the structural dynamic behavior of the detailed wing model are used as training inputs. To assess the accuracy of the approach the reduced-order model results are compared with high-fidelity full-scale aeroelastic computations. A small local reduction in accuracy can be seen for highly nonlinear effects like dynamic shock movements. However, variations in tank fill level or angle of attack show sufficient agreement of the model with the full-scale analyses. An interpolation of an intermediary angle of attack shows higher errors due to large changes in aerodynamic phenomena between the sample points. (C) 2016 Elsevier Masson SAS. All rights reserved.