Adaptive reduced order modelling for steady aerodynamics flows

The present work introduces an Adaptive Reduced Order technique, that aims at reconstructing the flow field around lifting bodies by adaptively selecting a Reduced Order Method that provides the highest accuracy among the existing model reduction approaches. The proposed adaptive approach automatically selects the most accurate order reduction methods between the classical snapshot Proper Orthogonal Decomposition and the Isomap Manifold Learning. The choice of the best method is performed by estimating the reconstruction error as different reduced order solutions are used. Problems of relevance to the aerodynamics field are considered to evaluate the performances of the proposed approach such as the parametric study of airfoils and wings aerodynamic performance.