LAUSR

Abstract In this study an online adaption of reduced order models (ROM) for fluid flows is presented. Often ROMs are stabilized and tuned by use of an eddy viscosity which is identified in the offline phase. As novelty the authors estimate the eddy viscosity in the online phase by treating the viscosity as an uncertain parameter in the framework of an extended joint Kalman filter. The mode dependency of the eddy viscosity is estimated at the same time as the eddy viscosity itself using a generalization of typical models for the eddy viscosity of fluid flows. Effectiveness of the proposed method is shown using the Burgers Equation and the incompressible two dimensional Navier-Stokes-Equations.