Kevin G. Wang, Philip S. Beran, Shunxiang Cao
DOI Number: N/A
Conference number: IFASD-2017-212
Adjoint-based mesh adaptation is applied to the prediction of flutter with a direct, Hopf bifurcation method. Adjoint variables are computed with respect to the objective of flutter speed, and regarded as an a posteriori indicator of the relative importance of local conditions to this objective. The magnitude of adjoint variables is then used to drive mesh adaptation for improved accuracy. The methodology is demonstrated on the model problem of a onedimensional thin panel in supersonic flow. The structure is modeled using the von Karman plate theory, the fluid pressure distribution on the panel is modeled using piston theory, and the combined system of equations is discretized using a finite difference method. Mesh adaptation is used to take an initially uniform mesh and locally refine it to better resolve parts of the panel that are judged to be relatively more important to flutter prediction. Three mesh-adaptation criteria are compared and contrasted based on: (1) strict adjoint-based mesh adaptation using the adjoint variables, (2) feature-based mesh adaptation using the solution curvature, and (3) a relaxed adjoint-based mesh adaptation using a combination of the adjoint variable and the solution curvature. The results show that for coarse meshes, criteria leveraging adjoints clearly outperform feature-based adaptation for predicting the critical dynamic pressure at flutter onset.