Title:Enhancement of Large Eddy Simulation for the prediction of an intake flow rig using sequential Data Assimilation

Abstract:A Data Assimilation (DA) strategy based on an ensemble Kalman filter (EnKF) is used to enhance the predictive capabilities of scale resolving numerical tools for the analysis of flows exhibiting cyclic behaviour. More precisely, an ensemble of numerical runs using Large Eddy Simulation (LES) for the compressible steady-state flow rig is augmented via the integration of high-fidelity data. This observation is in the form of instantaneous velocity measurements, which are sampled at localized sensors in the physical domain. Two objectives are targeted. The first one is the calibration of an unsteady inlet condition suitable to capture the cyclic flow investigated. The second one is the analysis of the synchronization of velocity field predicted by the LES with the available observation. In order to reduce the computational costs required for this analysis, a hyper-localization procedure (HLEnKF) is proposed and it is integrated in the library CONES, tailored to perform fast online DA. The proposed strategy performs a satisfactory calibration of the inlet conditions, and its robustness is assessed using two different prior distributions for the free parameters optimized in this task. DA state estimation is efficient in obtaining accurate local synchronization of the inferred velocity fields with the observed data. The modal analysis of the kinetic energy of the flow field provides additional information on the quality of the reconstruction of the velocity field, which shows improvements. Thus, the HLEnKF shows promising features for the calibration and the synchronization of scale-resolved turbulent flows, opening perspectives of applications for complex phenomena using advanced tools such as digital twins.