Development of Fully-Automatic Parallel Algorithms for Mesh Handling in the OpenFOAM
The current development to set up an automatic procedure for automatic mesh generation and automatic mesh motion for internal combustion engine simulation in OpenFOAM®-2.2.x is here described. In order to automatically generate high-quality meshes of cylinder geometries, some technical issues need to be addressed: 1) automatic mesh generation should be able to control anisotropy and directionality of the grid; 2) during piston and valve motion, cells and faces must be introduced and removed without varying the overall area and volume of the cells, to avoid conservation errors. In particular, interpolation between discrete fields is frequent in computational physics: the use of adaptive and non-conformal meshes necessitates the interpolation of fields between different mesh regions. Interpolation problems also arise in areas such as model coupling, model initialization and visualisation. This paper discusses the efficient implementation of the sliding interface, an algorithm to handle motion and topological changes in a moving-mesh Finite Volume Method (FVM) framework, that has been implemented to work with the newest mesh handling strategy of OpenFOAM®-2.2.x in a massively parallel environment. The algorithm performs a consistent second order interpolation between flow regions connected through non-conformal interfaces, opening the use of the code to a wide range of applications: the simulation of the piston motion trough scavenging ports in a two-stroke engine, the generation of cylinder grids characterized by a high quality mesh near the valve region, the reduction of the mesh size in external flow calculations. Consistent interpolation across non conformal interfaces represents a very important factor in LES, where discretization error and interpolation errors during mesh motion can have a significant impact on the quality of the results. Finally, in order to automate the mesh generation process of complex engine geometries, the algrithm for mesh motion has been applied to work with STL geometries, that are used for run-time generation of hexahedra and split-hexahedra grids on IC geometries, by a fully parallelised algorithm with automatic domain decomposition, without the loss of any geometric feature.