Turbulence model optimization of ship wake field based on data assimilation

The Reynolds-averaged Navier-Stokes (RANS) equations are primarily used to describe turbulent flows, and hence are essential for fluid simulation in Naval Architecture and Ocean Engineering. Parametric uncertainty, however, arises from coefficient closure models. Such coefficients are usually calibrated against experimental data for a set of simple flows, which are generally far from practical applications. This study applies a data assimilation technique called the ensemble Kalman filter (EnKF) to modify inaccurate coefficients for RANS k - ! turbulence model for the wake field of a ship, and its effectiveness is investigated as well. The present analysis shows that values of k - ! model coefficients can be better improved accurately and statistically by the EnKF. After data assimilation, the updated model outperforms its counterpart with default model coefficients in predicting the wake field. Then, several cases are carried out to demonstrate sensitivity for different assimilation parameters, such as ensemble size, sampling strategy, and observation. It not only presents the deep potential of data assimilation for marine hydrodynamics but also sheds light on future and more general applications.