This talk will address a hybrid data-driven reduced-order model for segregated fluid-structure interaction solvers in an arbitrary Lagrangian-Eulerian (ALE) approach at a high Reynolds number in the Finite Volume Method. The reduced-order model (ROM) is based on the classical proper orthogonal decomposition (POD), where a reduced basis is formed using energy-dominant POD modes. The quantities of interests comprise the velocity, pressure, eddy viscosity, and mesh motion considering the airfoil's two degrees of freedom structural motion. Galerkin projection is used to obtain the reduced space of both velocity and pressure. Data-driven approaches such as radial basis functions and neural networks/recurrent networks are used for predicting the temporal coefficients of the grid motion and eddy viscosity respectively.The effectiveness of this hybrid methodology is demonstrated in the case study of time-dependent problems governed by the 2D incompressible Navier-Stokes equations of flow-induced vibrations (FIV) of an airfoil at Reynolds number Re =10 000 000. The results show the stability and accuracy of the proposed method with respect to the high-dimensional model by capturing transient flow fields and, more importantly, the forces acting on the foil. At the end of the talk, initial results on optimal sensor placement for state reconstruction in the avenues of hyper-reduction will be discussed.
A hybrid data-driven reduced-order model for segregated fluid-structure interaction solvers in an ALE approach at high Reynolds number
Research Group:
Speaker:
Valentin Nkana
Institution:
SISSA
Schedule:
Friday, February 23, 2024 - 14:00
Abstract:
