Towards the accurate prediction of axial flow and heat transfer in a tightly spaced bare rod bundle configuration

The understanding of flow behaviour and temperature distribution in rod bundles during operation is of great importance for the safety and economical design of existing nuclear reactors and as well as new nuclear technologies. In order to study the thermal-hydraulics phenomena in the fuel assembly geometries, researchers all over the world are increasingly using Computational Fluid Dynamics (CFD) as a reliable research tool. Nevertheless, the prediction capabilities of the most commonly used Reynolds Average Navier-Stokes (RANS) models must be assessed. In this regard, in the present research work, an extensive effort has been put forward and is described in three main steps. As a first step, a wide range of RANS and unsteady RANS computations have been performed to design a numerical experiment for a closely-spaced bare rod bundle in order to perform a direct numerical simulation (DNS), which will serve as a reference for the validation purpose. As a second step, the DNS of this bare rod bundle has been performed for three different Prandtl number (Pr) fluids, i.e. Pr = 2, 1 and 0.025. This DNS has been performed using a higher-order spectral element code and consists of 660 million grid points. Accordingly, an extensive database has been generated for validation purposes. Finally, this database is used to assess the prediction capabilities of different linear and non-linear RANS models. This article will provide a good overview of all three steps and will also highlight the main lessons learned from this research.