Today, high temperatures are needed for the inlet gas of turbines to increase the efficiency of gas turbines. The vanes and blades of a turbine that are exposed to hot gas must be cooled, and film cooling is now being widely applied to gas turbines because of its high cooling effectiveness. Recently, numerous researchers have conducted numerical simulations of film cooling flows using Reynolds-averaged Navier–Stokes (RANS) simulation, detached eddy simulation (DES), and large eddy simulation (LES) turbulence models, along with structured and/or unstructured meshes. In fact, for cooling designers, both the prediction accuracy and turnaround time (TAT) of their mesh generations are important to speedily and repeatedly optimize the locations of film cooling holes and shapes. Several researchers have examined the prediction accuracy and effectiveness of film cooling flows in case of RANS in order to minimize the TAT of mesh generation such as for unstructured meshes. However, few papers have considered LES.

In this paper, the open-source computational fluid dynamics (CFD) toolkit OpenFOAM [17] is used as the solver, and LES is used to numerically simulate the film cooling flow into the zero-pressure gradient cross-flow on a flat plate from cylindrical holes, which is one of the most typical and basic test cases. LES is used because of its ability to capture the flow unsteadiness and non-uniformity observed in film cooling flows. In addition, surface-adjusted octree hexahedral meshes, which are relatively easy to generate automatically are used. The mesh generation tool used in this study is snappyHexMesh, which is included in OpenFOAM. The experimental data presented by Sinha et al. [15] are used. Because the film cooling effectiveness strongly depends on certain parameters and conditions, those of the simulated film cooling flow are set to the experimental parameters and conditions as closely as possible. In this study, the density ratio DR is set to 2.0, and the blowing ratio BR is set to 0.5 and 1.0 with the steady cross-flow inlet. In addition, a case is considered where BR = 0.5 (DR = 2.0), with an unsteady turbulent cross-flow inlet.

In the case of LES, the results obtained using the octree hexahedral meshes are compared with those of the experiment and my previous film cooling study (Fujimoto [32]), which was conducted using multiblock structured meshes. The results show that the prediction accuracy using the octree hexahedral meshes is equivalent to that using the multiblock structured meshes, even though the mesh generation TAT is significantly smaller.

This content is only available via PDF.
You do not currently have access to this content.