Rod bundle flows are commonplace in nuclear engineering, and are present in light water reactors (LWRs) as well as other more advanced concepts. Inhomogeneities in the bundle cross section can lead to complex flow phenomena, including varying local conditions of turbulence. Despite the decades of numerical and experimental investigations regarding this topic, and the importance of elucidating the physics of the flow field, to date there are few publicly available direct numerical simulations (DNS) of the flow in multiple-pin rod bundles. Thus a multiple-pin DNS study can provide significant value toward reaching a deeper understanding of the flow physics, as well as a reference simulation for development of various reduced-resolution analysis techniques. To this end, DNS of the flow in a square 5 × 5 rod bundle at Reynolds number of 19,000 has been performed using the highly-parallel spectral element code Nek5000. The geometrical dimensions were representative of typical LWR fuel designs. The DNS was designed using microscales estimated from an advanced Reynolds-Averaged Navier-Stokes (RANS) model. Characteristics of the velocity field, Reynolds stresses, and anisotropy are presented in detail for various regions of the bundle. The turbulent kinetic energy budget is also presented and discussed.

The Pressure Water Reactor (PWR) primary flow rate is determined using an enthalpy balance between the primary and the secondary circuits. Hot leg bulk temperature is necessary to estimate the enthalpy of the primary circuit. This bulk temperature is difficult to evaluate as the temperature is measured in the hot leg using a small number of local sensors in a region where its heterogeneousness is large. Measuring the bulk temperature in a real-life configuration is, to the authors’ knowledge, still a challenging task. Thus, Computational Fluid Dynamics (CFD) is an interesting approach to quantify the difference which may exist between the measured temperature and its bulk value. Deciding which turbulence model one has to use in order to have a satisfactory prediction of the fluid motion and the temperature field is not obvious. Indeed, the flow in the hot legs has strong secondary motions which may need a high order turbulence closure. The main objective of the present work is to compare a standard model such as the k-ε with a more sophisticated one (the Reynolds Stress Model from Speziale, Sarkar and Gatski (SSG)). EDF in-house CFD tool Code_Saturne is used for all the computations. The data coming from the so called “Banquise” experiment (1/5 th scale, Re = 100.000, PIV, thermal tracing) representing the upper plenum of a 1300 MWe PWR and its hot legs are used for comparisons. It is, at the present stage, difficult to clearly distinguish the quality of the two models. Both give globally satisfactory results with the criteria used in the present article. One finds however that the SSG model is superior at the beginning of one of the two hot legs which has a direct impact on the prediction of the temperature field.

Swirling and secondary flows play an important role in several locations of the primary circuit of a PWR. After a brief reminder of the experimental period during the initial designs, few examples such as the lower plenum, the upper Plenum and the hot legs, the fuel assemblies and the U-bend downstream the steam generator are given to illustrate this statement. Although experiments will remain mandatory to understand these complex flows, CFD could bring some light either to perform parametric studies or to analyze the flow structures if not possible in the experiment or in a reactor configuration. In order to have a “clean” approach, one has to first validate the CFD tools on few academic cases dominated by swirling/rotation/secondary effects. A full program for a better understanding of these kinds of flows with CFD computations has been launched at EDF. Examples of the very first calculations are given in the present paper. Although no mature conclusions can be drawn at this stage with the available simulations, one can argue that first moment closures such as k-epsilon or k-omega models can perform well in some configurations while Reynolds Stress Models performs almost always rather well. LES is clearly superior but more expensive. Even the growth of the computing power will not allow us to perform LES computations whatever the configuration, thus focusing on RSM model remains important.

In Pressurized Water Reactors (PWR), Steam Generator (SG) tubes constitute one of the three barriers which preserve the environment from radioactivity. Excessive tube vibrations under fluid forces, due to the steam-water mixture flow across the tube bundle, can lead to the failure of some tube. Several methods have been proposed to estimate some upper bounds for these forces. These bounds are applicable at the design stage and are helpful to avoid tube failures. Most of the available methods are based on experimental results that have been obtained on tube bundles installed in scaled test-facilities. Unlike this popular test-based approach, one combines here Computational Fluid Dynamics (CFD) to High Performance Computing (HPC), in order to estimate fluid forces in a simple case by applying the Direct Numerical Simulation (DNS) method to solve the Navier-Stokes equations. In the first paragraph, one summarizes the general standard method which allows one to derive the auto-power spectral density of the displacement response at any point of an SG tube, departing from the cross-power spectral densities of fluid forces between any two points along the tube. In the second paragraph, one recalls the equivalent dimensionless spectrum, which was proposed by Axisa et al. in the early nineties, and which still remains a useful reference in the domain. One then applies DNS to the test case of a single infinite cylinder, which is submitted to a single phase cross-flow in a rectangular channel. The Reynolds number is equal to 3900. One presents the time dependent tensors of fluid pressure and viscous stresses, and uses this tensor to estimate the field of non stationary forces that are applied by the fluid, per unit length, at a set of equidistant locations along the tube. Even if they do still require experimental validations, our computation results are more abundant and detailed than standard experimental results, as well as more flexible to use. They therefore provide an interesting additional source of information. They already allow us to try to get new insights into quantities that would be, in any case, very difficult to obtain experimentally. Lift, drag, and even the forces acting in the direction of the tube axis, are computed, and can be distinguished one from the other. Fluid forces due to viscous stresses can also be compared to the ones caused by pressure. The degree of correlation of the forces along the tube can also be examined.

This paper focuses on the geometry/solver interface for the CFD Software Code_Saturne ® which has been developed by EDF R&D since 1997 to replace its FE and BFC solvers. The solver includes various RANS models, as well as LES features and is parallelized through a domain splitting technique. It is based on a cell centered unstructured finite volume scheme, and accepts cells of any shape. This opened the possibility of using non conforming meshes, making it easier to build meshes with well-controlled quality and refinement even for complex geometries. Adjacent boundary faces of non-conforming input meshes may be automatically split into their intersecting subsets so as to build a conforming mesh of polyhedra with an arbitrary number of faces per cell. This also extends to the handling of periodic boundary conditions as a geometrical condition. We will explain how this is handled and illustrate the algorithm’s behavior on different complex grid examples.

This paper presents a numerical study to tackle thermal striping phenomena occuring in piping systems. It is here applied to the Residual Heat Removal (RHR) bypass system. A large Eddy Simulation (L.E.S.) approach is used to model the turbulent flow in a T-junction. The thermal coupling between the Finite Volume CFD Code_Saturne and the Finite Element thermal code Syrthes , gives access to the instantaneous field inside the fluid and the solid. By using the instantaneous solid thermal fields, mechanical computations (as presented in (Stephan et al 2002)) are performed to yield the instantaneous mechanical stresses seen by the pipework T-junction and elbow.