Reduced Order Model for Steady Aerodynamics Applications Based on Navier-Stokes Residual Vector Minimization

Abstract

Reduced order models are popular techniques in the aerospace engineering field to obtain accurate aerodynamic field solutions at a computational cost that is orders of magnitude lower than the high fidelity simulations. Of particular interest are the techniques based on proper orthogonal decomposition and direct minimization of the residual vector resulting from the discretization of the system of governing differential equations, e.g. Navier-Stokes equations for aerodynamics problems. This type of reduced order model methods have been already successfully applied within the German Aerospace Center using TAU-DLR as a computational fluid dynamics solver. % However, recent research has highlighted how the TAU-DLR implementation does not enable a complete exploitation of some advanced reduced order model techniques like hyperreduction. In particular, a reduced order model framework based on TAU-DLR does not allow the evaluation of the residual vector on the submesh determined by the hyperreduction indices. The goal of this research is to test and assess the prediction accuracy and computational time reduction of a consistent hyperreduction implementation that leverages the functionalities of the new in-house solver CODA which is capable to perform the residual assessment on a subset of computational grid cells. The hyperreduction is included in a reduced order model based on proper orthogonal decomposition and direct residual vector minimization. In particular, such a reduced order model is tested for the prediction of transonic flowfields around both an airfoil (RAE2822) and an aircraft (NASA-CRM) geometry.