Adaptation of reduced order models for applications in aeroelasticity.

摘要

Aeroelasticity is the study of the interaction and coupling of aerodynamic, elastic, and inertial forces. There is a class of problems in aeroelasticity that involves the repeated analyses of the fluid-structure system for many different parameter combinations. Due to the nonlinear fluid behavior in the transonic flight regime, these analyses generally require the use of very large high fidelity models to obtain accurate results. Such models incur a large computational cost primarily due to their size. Although current computational fluid dynamics codes, for example, are routinely used to compute point solutions for complex, viscous flows, their use in related disciplines such as aeroelasticity is limited due to the large computational expense required by the repeated analyses. Of particular interest is the prediction of flutter, which is a dynamic aeroelastic instability that produces increasing oscillations that often lead to structural failure. For aircraft, the accurate prediction of flutter is essential for high performance and safe designs. In order for a flutter analysis to be useful in a design process, in flight testing, or in developing flutter suppression systems, it must be relatively quick and efficient so that many design iterations and contingencies can be examined. These flutter analyses would be more feasible if low order models of the full aeroelastic system were available.; The major computational cost of an aeroelastic analysis is mainly attributable to the fluid system. Thus, this thesis presents a methodology for constructing and adapting reduced order models of the fluid that preserve the accuracy of the high-fidelity simulations while substantially reducing the computational expense of the analysis. The proposed approach is to use an empirical ROM construction technique known as proper orthogonal decomposition. Since the constructed ROMs are valid only within a limited state-space, a ROM adaptation strategy is presented to widen the range under which the ROM is applicable. The POD method and its adaptation to varying parameters are introduced and demonstrated on a simple three dimensional wing. Then, these methods are tested on a detailed three dimensional model of a full aircraft to demonstrate the relevance of these methods to industrial applications.