CFD simulations of large-scale reorganizations in PWR rod bundle flows

International audienceA coolant flow is used in Pressurized-Water Reactors (PWR) cores to extract the heat generated by nuclear reactions. Its temperature must be as homogeneous as possible in order to avoid a localized boiling, which would deteriorate the behaviour of the reactor. This flow is injected in the interstices of rod bundles, which namely are arrays of cylinders held together by support grids. While the initial velocity flow field is mostly aligned with the cylinder axes, mixing vanes are placed on the support grids in order to vastly increase inter-channel velocities and thus improve the overall thermal mixing. As depicted inthe sketch in figure 1, large-scale structures are generated in cross-section planes orthogonal to the rod axes in the wake of the mixing grid, and their layout as well as their evolution downstream of the grids have a large impact on the flow boiling margin. A thorough investigation of these coolant flow large-scale structures thus constitutes a key element of PWR safetyanalyses.Both experiments and Computational Fluid Dynamics (CFD) simulations have been used regularly in the last decades to investigate these large-scale flow structures. Notable experiments include the AGATE facility operated at the CEA Grenoble [3] as well as the OECD/NEA-KAERI benchmark test [2]. An example of a large-scale cross-flow pattern observed in the 5x5 rod bundle flow of the AGATE facility is shown in figure 2. Interestingly, this pattern spontaneously reorganized itself in thefar wake of the mixing grid so as to rotate from a mostly 45 and#9702; angle to a 135 and#9702; one. This phenomenon can be related to earlier observations of a pattern change in the wake of a mixing grid by Shen et al. [4], therein dubbed as a velocity inversion. In addition to the results obtained by Bieder et al. [1], CFD simulations of rod bundle flows based on a Large-Eddy Simulation (LES) sub-grid scale model have been performed, both in a reduced 3x3 rod bundle and in a 5x5 one, with the aim of reproducing such experimental large-scale reorganizations of the cross-section flow.Furthermore, a method of coupling between the 3D axial coordinate and the time variable through the Taylor frozen turbulence hypothesis is being used to advance a physical explanation to the large-scale reorganization phenomena. 2DDirect Numerical Simulations (DNS) have thus been performed in rod bundle cross-section geometries from an initial condition based on a 2D slice of the 3D flow in the immediate wake of the grid in LES simulations. A comparison between the evolution over time of the 2D simulated flow in a cross-section geometry and the axial evolution of a steady 3D flow cross-section is then carried out, revealing interesting parallels