A data assimilation method for recovering turbulent flows using heterogeneous experimental data

Data assimilation in turbulent fields serves as a crucial means of acquiring high-fidelity training data for turbulence modeling. Prevailing data assimilation methodologies predominantly concentrate on individual experimental observations, while little attention has been paid to the influence of heterogeneous experimental data types on assimilation outcomes, and identifications are rarely made to find out which experimental observations can best facilitate the high-fidelity inversion of flow fields. This paper presents a data assimilation approach based on Ensemble Kalman Filter, which assimilates diverse experimental observations into Reynolds averaged turbulent fields. The impact of different types of experimental observations on data assimilation outcomes is thoroughly investigated, and we further explore which types of experimental observations can maximally facilitate the inversion of the physical flow field. Experimental measurements are applied to S809 airfoil and two-dimensional backward-facing step cases to update the eddy viscosity field, and assimilation outcomes are assessed accordingly. Results indicate that combining wall pressure and specific velocity profiles yields superior results compared to using a single type of measurement. Multi-Dimensional Scaling technique is employed to evaluate the consistency between assimilated fields and experimental measurements providing further evidence for superiority of results obtained by our heterogeneous assimilation.