Assessment of source term and turbulence model combinations for film cooling in turbines

Authors

Michael Müller  - Deutsches Zentrum für Luft- und Raumfahrt, Germany
Christian Morsbach - Deutsches Zentrum für Luft- und Raumfahrt, Germany

Abstract

Modern gas turbines require an active cooling system with fluid taken from the compressor to withstand the high turbine inlet temperature, above the melting point of the alloys used in the first rows of a high-pressure turbine. This fluid is no longer part of the full thermodynamic cycle and the turbine will extract less energy from it. Therefore, using cooling fluid sparingly will increase the efficiency. However, insufficient cooling can drastically reduce the lifespan of the turbine. Computational Fluid Dynamics can help optimize the amount and distribution of cooling fluid. Calculating the effect of film cooling in a turbine usually requires a computational grid resolving those cooling holes. This significantly increases the number of cells and in turn the computational cost of the simulation. A reduced order model introducing a source term for the mass, momentum and energy equation was implemented in TRACE, DLR’s in-house flow solver for turbomachinery applications. The sources can be realized either in the volume around the injection point or directly on the surface of the cooling hole. This approach is intended for use in RANS simulations with resolved boundary layers. However, this class of models has known issues predicting e.g. the film cooling efficiency, especially on coarser grids. In the paper we want to adress the error of injecting the cooling fluid with a source term separately from the error when transporting and mixing the cooling film. Firstly, we will compare the accuracy and grid dependency of the injection using volume source terms and surface source terms for cylindrical and laid-back fan shaped cooling hole geometries. One strategy to reduce the grid dependency, the thickened-hole model proposed by Bizzari et al for LES simulations of effusion cooling in combustion chambers, will also be investigated in this context. Secondly, we will compare different turbulence and heat transport models. One of the reasons for deficits in accuracy of RANS simulations of film cooling holes, especially further downstream of the injection, is the choice of those models. Reynolds Stress Models combined with an anisotropic heat transport model can provide a significantly better prediction of the mixing and diffusion of the cooling fluid.