From Sub-Channel Analysis to Two-Phase Flow CFD: Improving Thermal-Hydraulics Analysis of Nuclear Reactor Cores

For safety analyses of nuclear reactor cores, a correct prediction of the thermal-hydraulic characteristics is crucial. These predictions are all the more difficult as the flows generally involve liquid and vapor phases simultaneously. Moreover, the geometry of nuclear cores is often quite complex. The typical situation is that of a rod bundle, the characteristic length of the gap between two rods being much smaller than the size of the bundle. The traditional approach to the simulation of such flows is called the sub-channel analysis. The flow is assumed to have a privileged direction, and the cross-flow inertia effects are neglected. Moreover, a lumped-geometry approach is generally adopted, whereby a single discretization cell is used to represent the volume between several rods. This leads to efficient solution methods but forbids a precise description of local or global three-dimensional effects. As the computational power of modem computers steadily increases, a finer description of the flows in nuclear reactor cores becomes possible. Indeed, there is a current trend in the nuclear industry toward a CFD-like description of these flows as shown by Paillère, et al., (1998) and Rautaheimo, et al., (1999). However, the numerical method used for the simulation must satisfy some specific requirements: • The use of unstructured meshes must be possible to allow an easy description of the geometry between the rods. • The numerical method must be suitable for variable density (thermally expandable) flows. Cavendish, Hall, and Porsching (1994) present a covolume method designed to meet these requirements. This method relies on the construction of a dual or Voronoi mesh. The pressure and the thermodynamic variables are computed at the vertices of the primal mesh. Additionally, the velocities are computed in the normal direction to each face of the primal mesh, or, equivalently, along each edge of the dual mesh. The continuity equation is integrated by parts on the dual polytopes, while the momentum equations are discretized by finite differences on the primal mesh. This Cavendish, Hall, and Porsching covolume numerical method has been used to solve typical problems of nuclear reactor thermal-hydraulics analysis. The physical models are those of the CORETRAN code (1999). The first numerical results demonstrate the efficiency of the method and validate the new approach.