The present thesis was written by Karim Osman at the University of Manchester in pursuance of the degree of Master of Philosophy in 2011. It presents the benchmarking and development of methodologies to reduce the sensitivity of turbulence models to the near wall mesh size in order to ease simulations of turbulent convective heat transfer.One of the main difficulties in studying convective heat transfer is the modelling of turbulent boundary layers. A review of the most up-to-date techniques revealed two main approaches: the wall functions method and the down to the wall resolution. The last one is more accurate than the first one and it is also more expensive because it needs a very fine mesh near the wall. Nevertheless, both techniques suffer from strong deficiencies when near wall meshes do not obey precise criteria.We designed verification test cases in order to show clearly the limits of both methods and to test some widely used solution like the "scalable wall functions". It has appeared that perturbation of the mesh near the wall has a dramatic effect on the results and that the scalable wall functions are able to correct this problem. However this technique does not better the result when the mesh is fine.For this reason we decided to propose an hybrid treatment of the near wall turbulence called "adaptative wall function" based on an early work of Kalitzin et al. and a recent version of a down to the wall model. This new model was implemented in Code_Saturne and tested on several cases featuring heat transfer. Our approach was proved to be robust and accurate on highly non-uniform meshes and those with nearwall cell spacing not generally recommended for wall-function approaches. We also investigated some ideas that can improve our result like the zonal correction method.
|Date of Award||1 Aug 2012|
- The University of Manchester
|Supervisor||Dominique Laurence (Supervisor)|