Investigation of ensemble Kalman filter-based data assimilation for turbulent flow in a 3×3 rod bundle

Abstract

Computational Fluid Dynamics (CFD) has advanced rapidly in reactor core thermal-hydraulics, serving as a vital tool for nuclear reactor safety and performance analysis. This study addresses uncertainties in complex turbulence simulations within reactor cores, proposing a novel approach to optimize Reynolds-Averaged Navier-Stokes (RANS) models using the Ensemble Kalman Filter (EnKF) data assimilation method. By simulating turbulent flow in a 3 × 3 rod bundle channel and using time-averaged velocity fields from Particle Image Velocimetry (PIV) as observational data, an observation operator maps local velocity measurements to full-field turbulence parameters, optimizing key RANS closure parameters. Results show that the EnKF-optimized RANS model significantly improves the accuracy of velocity and turbulent kinetic energy predictions, with relative errors below 1 % in most regions, outperforming the original model. This approach provides a data-driven optimization paradigm for high-fidelity reactor thermal-hydraulic simulations, supporting complex confined turbulence modeling and multi-physics coupling analysis. However, further validation is needed to explore additional factors affecting assimilation performance in nuclear reactor systems.