Assessment of hybrid turbulence models for the simulation of ribbed channel with heat transfer

Authors

Wojciech Sadowski - Chair of Thermal Turbomachines and Aeroengines, Ruhr University Bochum, Germany
Federico Lo Presti - Chair of Thermal Turbomachines and Aeroengines, Ruhr University Bochum, Germany
Francesca di Mare - Chair of Thermal Turbomachines and Aeroengines, Ruhr University Bochum, Germany

Abstract

Manufacturing periodic ribs in cooling channels is a widely used technique to enhance turbulent heat transfer. Examples of applications include cooling channels of gas turbine blades, heat exchangers and gas-cooled reactor fuel elements. Currently, the improvement of such systems is a relevant topic of industrial research and numerical simulations including heat transfer are of great importance. The heat transfer enhancing mechanism introduced by the ribs, is characterized by the periodic interruption of the growth of the boundary layer, inducing the formation of separation and reattachment zones. The prediction of separation zones is one of the main challenges in turbulence modelling, with some approaches failing to predict even the existence of such areas. Moreover, there is a wide consensus that RANS simulations are not capable of reproducing experimental results when it comes to the prediction of heat transfer. While in such cases the use of scale-resolving turbulence modelling approaches would be desirable, the computational cost is a major limitation in an industrial context. Moreover, cooling channels are typically part of larger configurations (e.g., in a gas turbine) so that the computational cost of the whole simulation has to be taken into account. It would be valuable to have a single modelling approach that is able to capture relevant physics in areas of interest while being less expensive than classical LES. Ideal candidates for this purpose is a wide range of hybrid RANS/LES approaches that has been developed over the years. These approaches try to bridge the robustness of time averaged equations in regions like boundary layers with accuracy of LES in separation regions. The most important factor for a successful hybrid RANS/LES computation is generating a numerical mesh with appropriate resolution for both RANS and LES regions. For RANS, an appropriate resolution is the one that guarantees mesh independence of results. Such criterion is not appropriate for LES, where a standard quality indicator is resolving at least of 80% of turbulent energy. The problem is further complicated by the presence of a moving blurry interface region between the two formulations and different methods of blending two types of models together. Determining a proper mesh resolution for such simulations is still an open research topic. In the current work an LES simulation of the flow through a ribbed duct with heat transfer will be performed. The quality of the LES will be carefully verified using quality criteria based on e.g., total resolved turbulence energy, turbulence viscosity ratio. The mesh resolution required for an accurate scale-resolving simulation and its results will be used as a baseline reference for the study of hybrid models performance. Optimal meshing strategies for hybrid models, balancing numerical cost and modelling accuracy, will be discussed. Mesh sizing will be gradually increased globally and the performance of RANS/LES models will be assessed, including prediction of velocity, velocity fluctuations and heat transfer. In this context, the influence of local mesh refinement on main flow features (e.g. reattachment length) will also be tested.