In gas turbines some parts are exposed to combustion gases with temperatures well above the melting temperature of the material. Therefore, various cooling techniques are utilized in order to protect the parts exposed to these hot gases. One such technique, film cooling, is a common and well established way to protect the exposed parts. Film cooling involves the ejection of cold air on the surface of the parts that are to be protected, thus creating a film of colder air between the surface and the hot gases.
Computational Fluid Dynamics (CFD) is a way of calculating fluid flow, and can be used to calculate the effectiveness of a cooling film in film cooling applications. CFD is demanding in terms of computer power, especially when advanced methods are to be used. Even the simpler methods, such as Reynolds Average Navier-Stokes (RANS), can be quite demanding, time and computer power-wise, and require resources not always available. Finding ways of limiting the needed computer power is therefore of large interest.
The aim of this thesis is to reduce the computational time of film cooling CFD-simulations, by using reduced models. To achieve this, simulations has been conducted and compared to experiments. The investigated setup is of an enginelike equipment, where a guide vane is investigated for heat transfer coefficient and film effectiveness. The geometry in the experimental setup is constructed in such a way as to give the same pressure distribution around the guide vane as can be seen in a real gas turbine, although at lower temperatures than those in the real turbine. The CFD-simulations conducted on the test rig includes RANS-simulations using the realizable k- and the SST k-! turbulence models.
The reduced model contains only the central part of the vane. The walls of the test rig is replaced with periodic boundary conditions. This narrow model gives good agreement with the full model for heat transfer coefficient. Due to the large computational cost required to conduct simulations with cooling on the full model no comparison were made between the cooled narrow and cooled full model.
To further reduce the size of the computational domain, two additional models were investigated. The first one involves a reduction of the full domain to only include the section being studied, in this case the suction side of the guide vane.
This infers a reduction of the mesh size to less than ten percent of the size of what a mesh of the cooled full domain would be. The next step to reduce the size of the model and mesh is to make a narrow version of the already shortened model. The results for these two models show that they perform adequately to each other and (in the cases where a comparison is possible), to the full domain.
Linköping: Linköping University Electronic Press, 2013. , 44 p.
2013-05-31, Sal C3, C-huset, Campus Valla, Linköpings universitet, Linköping, 10:15 (English)
Storck, Karl, Dr.