Development and Verification of Unstructured Adaptive Mesh Technique With Edge Compatibility

In a design study of the large-sized sodium-cooled fast reactors in Japan (JSFR), one key issue to establish an economically superior design is suppression of a gas entrainment (GE) phenomenon at a free surface in the reactor vessel. However, the GE phenomenon is highly non-linear and too difficult to be evaluated theoretically. Therefore, we are developing a high-precision CFD method to evaluate the GE phenomenon accurately. The CFD method is formulated on an unstructured mesh to establish an accurate modeling for a complicated shape of the JSFR system. As a two-phase flow simulation method, a high-precision volume-of-fluid algorithm is employed in the CFD method. In addition, physically appropriate formulations at gas-liquid interfaces are introduced into the CFD method. The developed CFD method is already applied to the simulation of a GE phenomenon in a basic GE experiment and the simulation results show good agreement with experimental results. Therefore, it is confirmed that the proposed CFD method can reproduce a GE phenomenon. However, for the simulation of the GE phenomenon in the JSFR, we still have one problem on a mesh subdivision. Though a fine mesh subdivision has to be applied to the regions where the GE occurs, it is difficult to preliminarily know the regions because the GE occurrence is strongly affected by a local instant flow pattern, i.e. a vortex generation. Therefore, an adaptive mesh technique is necessary to apply a fine mesh subdivision automatically to only the local GE occurrence regions in the large-sized JSFR. In this study, as one part of an adaptive mesh development, a two-dimensional unstructured adaptive mesh technique is developed and verified. In the proposed two-dimensional adaptive mesh technique, each cell is isotropically subdivided to reduce distortions of the mesh. In addition, a connection cell is formed to eliminate the edge incompatibility between a refined and a non-refined cells. A connection cell has several subdivision patterns and one of them is selected to be compatible with adjacent cells on every cell edge. Finally, the present unstructured adaptive mesh technique is verified by solving well-known driven cavity problem. As the result, the present unstructured adaptive mesh technique succeeds in providing a high-precision solution, although we employ a poor-quality distorted mesh at the initial state. In addition, the simulation error on the unstructured adaptive mesh at the steady state is much less than the error on the structured mesh consisting of a larger number of cells.