Numerical techniques for solving the navier-stokes equations on complex geometries
- Autores: Concepcio Lifante Navarrete
- Directores de la Tesis: Carlos David Pérez Segarra (dir. tes.), Asensio Oliva Llena (dir. tes.)
- Lectura: En la Universitat Politècnica de Catalunya (UPC) ( España ) en 2006
- Idioma: inglés
- Tribunal Calificador de la Tesis: Esteban Codina Macia (presid.), Joaquim Rigola Serrano (secret.), José Manuel Cejudo López (voc.), Alberto Coronas Salcedo (voc.), Manel Soria Guerrero (voc.)
- Materias:
- Texto completo no disponible (Saber más ...)
- Resumen
Fluid Dynamics and Heat Transfer problems are involved in multiple processes in the nature and industry, Their study and resolution are basic for the improvement and optimization of different industrial equipments. Computational Fluid Dynamics deals with this kind of problems discretizing the domain to study into small cells or control volumes. Usually this approximation to the domain is performed employing orthogonal control volumes. Therefore, the domain which is really considered is not the original one but an approximation. There are some techniques as blocking-off ones which try to solve this drawback, however, not always are appropiate: the number of control volumes is significantly increased if an accurate approximation is required, or for instance among other problems, turbulent cases can not be solved properly because a false roughness appears.
The physical phenomena that want to be studied show complex geometries and require more accurate approximations. The first part of the thesis consists of the development of a numerical methodology in order to solve the Navier-Stokes equations on body-fitted, or boundary-fitted, grids. Once this section is done, a multiblock conservative scheme has been developed, consisting of splitting the original domain into different domains o blocks, in order to be solved independently. Therefore, new boundary (interpolation) conditions, conservatives, which are responsible for the information exchange must be introduced. This kind of techniques allows the optimization of the number and location of the grid nodes, the possibility of meshing each zone independently considering only the complexity of the flow in this domain zone, the possiblity of employing finer meshes or the possibility of performing faster execucions. These last two properties are consequency of the use of the memory and processsor of different computers, specially important due to the increasing computing power of the beowulf clusters (compounded by per
