Adjacency-based, non-intrusive model reduction for Vortex-Induced Vibrations

Abstract

Vortex-induced vibrations (VIV) pose computationally expensive problems of high practical interest to several engineering fields. In this work we develop a non-intrusive, reduced-order modelling methodology for two-dimensional (2D) VIV simulations. We consider an elliptical, non-deformable solid mounted on springs, subject to a laminar, incompressible flow. Approximating the Arbitrary Lagrangian-Eulerian (ALE) incompressible Navier-Stokes (NS) formulation, a discrete-time, quadratic-bilinear data-driven model structure is assigned for the velocity flowfield prediction. The full-order, data-driven model operators have a predefined sparsity pattern, following the adjacency-based sparsity of the discretized ALE-NS operators. Thus, the data-driven operators inference requires solving many low-dimensional least squares problems, isolating the contribution of "nearest neighbours" for each degree of freedom. Numerical aspects such as data centering and regularization are extensively discussed. With this approach, Dirichlet boundary conditions at the inlet and the fluid/solid interface can be enforced on the full-order, non-intrusive level. Consequently, a non-intrusive reduced-order model (ROM) for the velocity flowfield is obtained after linear projection. The resulting ROM is coupled with the first-principle, 2D solid oscillation equations and is simulated using an implicit time integration scheme. The coupled solution is then mapped on the deformed grid through the inverse ALE map. This methodology is showcased for two testcases, at different Reynolds numbers (Re = 90, 180). Numerical results indicate a successful coupling between the data-driven velocity flowfield and the solid oscillation, with prediction errors of less than 3% for both the flowfield and the solid oscillation. A comparative study with respect to the ROM dimension indicates the robustness and potential of the approach.