Inscrição na biblioteca: Guest
TSFP DL Home Arquivos Comitê executivo

4D-VARIATIONAL DATA ASSIMILATION FOR POD REDUCED-ORDER MODELS

Gilles Tissot
PPRIME Institute, CEAT, 43 route de l'aerodrome, 86000 Poitiers, France

Laurent Cordier
Department Fluides, Thermique, Combustion, CEAT Institut PPRIME, CNRS-Universite de Poitiers-ENSMA, UPR 3346 43 rue de I'Aerodrome, F-86036 Poitiers CEDEX, France

Nicolas Benard
Department of Fluids, Thermique et Combustion Institut PPRIME (CNRS UPR3346, Université de Poitiers, ISAE-ENSMA) Bd Marie & Pierre Curie, BP 30279, 86962 Futuroscope, France

Bernd R. Noack
Berlin Institute of Technology MB1 Strasse des 17. Juni 135, D-10623 Berlin, Germany; Departement Fluides, Thermique, Combustion Institut PPRIME, CNRS UPR 3346 CEAT, 43 rue de I'Aerodrome, F-86036 Poitiers, FRANCE

Resumo

In flow control, reduced-order models based on Proper Orthogonal Decomposition (POD ROM) are often used as surrogate model for deriving a control law. However, these models are in general too fragile to be used in closed-loop control where the dynamics is strongly modified by the control. Here, a 4D-Variational data assimilation approach (4D-Var) as classically used in meteorology is used to estimate at best the state of the system from inhomogeneous sources of information coming from a model, noisy observations and a background solution. Two complementary strategies (strong constraint 4D-Var and weak constraint 4D-Var) are assessed in the case of a cylinder wake flow with data coming from numerical simulation and PIV experiments.