Multigrid sequential data assimilation for the Large Eddy Simulation of a massively separated bluff-body flow

Abstract

The potential of sequential Data Assimilation (DA) techniques to improve the numerical accuracy of Large Eddy Simulation (LES) performed on coarse grid is assessed. Specifically, this paper evaluates the performance of the Multigrid Ensemble Kalman Filter (MGEnKF) method, recently introduced by Moldovan, Lehnasch, Cordier and Meldi (Journal of Computational Physics, 2021). The international benchmark referred to as BARC (Benchmark of the Aerodynamics of a Rectangular 5:1 Cylinder) is chosen as test configuration, as it includes several complex flow dynamics encountered in turbulence studies. The results for the statistical moments of the velocity and pressure flow field show that the data-driven techniques employed are able to significantly improve the predictive features of the solver for reduced grid resolution. In addition, it was observed that, despite the sparse and asymmetric distribution of observation in the data-driven process, the DA augmented LES exhibits symmetric statistics and a significantly improved accuracy also far from the observation zone.