Estimation of Flow Field in Natural Convection with Density Stratification by Local Ensemble Transform Kalman Filter

Abstract

For estimating thermal flow in a nuclear reactor during an accident accurately, it is important to improve the accuracy of computational fluid dynamics simulations. The temperature and flow velocity are not homogeneous and have large variations in a reactor containment vessel because of its very large volume. In addition, Kelm’s work pointed out that the influence of variations of initial and boundary conditions was important. Therefore, it is necessary to set the initial and boundary conditions taking into account the variations of these physical quantities. However, it is a difficult subject to set such complicated initial and boundary conditions. Then, we can obtain realistic initial and boundary conditions and an accurate flow field by data assimilation, and we can improve the accuracy of the simulation result. In this study, we applied data assimilation by a local ensemble transform Kalman filter to a simulation of natural convection behavior in density stratification, and we performed a twin model experiment. We succeeded in estimating the flow fields and improving the simulation accuracy by the data assimilation, even if we applied the boundary condition with error for the true condition.