CFD - Technische Universität Berlin (Germany)
Local Project ID:
HPC Platform used:
Hermit/Hornet/Hazel Hen of HLRS
As part of the collaborative research centre CRC 1029, impingement cooling is studied at the chair of computational fluid dynamics at the Technical University of Berlin. The project aims at a more efficient cooling of turbine blades. This is necessary since future combustion concepts within gas turbines bring much higher thermal loads. Large scale direct numerical simulations (DNS) are carried out using the Cray supercomputers Hermit, Hornet and Hazel Hen of the HLRS.
In 2012, a visionary collaborative research centre (CRC 1029, http://www.sfb1029.tu-berlin.de) started at the Technical University of Berlin (TU Berlin). Due to a radical change of the combustion concept within gas turbines, an increase of the overall efficiency from 40% to 50% is aimed. The success of this technology can save huge amounts of natural resources and therefore has an important economical and environmental impact. The new combustion concept leads to higher turbine entry temperatures and stronger temperature fluctuations. For this reason turbine blade cooling needs to be improved.
Modern turbine blades are cooled from outside by film cooling and from inside by impingement cooling. The latter is the target of this research project. Experimental preliminary studies at the TU Berlin showed that heat transfer efficiency can be improved when the cooling air supply is unsteady. In order to explore the underlying physics of this phenomenon, direct numerical simulations (DNS) were carried out.
DNS is the most precise method within computational fluid dynamics and enables the visualization and quantification of vortices and heat transfer. The precision can be afforded only by the world’s most powerful supercomputers. This project started on the Cray XE6 system Hermit and was continued on the Cray XC40 machine Hornet/Hazel Hen of the HLRS. The shift from Hermit to Hornet reduced the required computing time by a factor of 4 due to the modern architecture. In addition, the increased total resources allowed the parallel usage of up to 16384 cores. The biggest simulations have a Reynolds number of 8000 and were carried out on more than 1 billion grid points (1024 x 1024 x 1024). Around a quarter million time steps were needed to provide statistical information.
In the first step, a continuous impinging jet was analysed. Figure 1 shows the vortical structure of the jet as well as heat transfer at the impinging plate. It can be seen that vortex rings develop naturally within the shear layer. The formation of these rings is periodically and fits with the experimentally found frequencies, where an increase of heat transfer efficiency could be observed. The rings are transported streamwise and travel along the impinging wall. The large vortex rings cause a high local heat transfer as indicated by the white arrows. Consequently, an increase of heat transfer can be obtained by increasing these vortical structures. This is done by the application of a pulsation at the nozzle inlet, without changing the average mass flow of cooling fluid.
In the second step, such a simulation was carried out. Figure 2 shows a comparison of the strength of the vortices. The upper image corresponds to the continuous air supply and the lower one to the pulsed case. The results prove that the frequencies occurring naturally enable the generation of much bigger vortex rings, which are the key to higher heat transfer at the impinging plate and the basis for the new combustion concept for future gas turbines.
F. Haucke, W. Nitsche, R. Wilke, and J. L. Sesterhenn. Experimental and numerical investigations regarding pulsed impingement cooling. In Deutscher Luft- und Raumfahrt Kongress, Rostock, Germany, 2015.
R. Wilke and J. Sesterhenn. Direct numerical simulation of heat transfer of a round subsonic impinging jet. In Active Flow and Combustion Control 2014, pages 147–159. Springer, 2015.
R. Wilke and J. Sesterhenn. Numerical simulation of impinging jets. In High Performance Computing in Science and Engineering ’14, pages 275–287. Springer, 2015.
R. Wilke and J. Sesterhenn. Numerical simulation of subsonic and supersonic impinging jets. In High Performance Computing in Science and Engineering ’15, pages 349–369. Springer, 2016.
R. Wilke and J. Sesterhenn. Statistics of fully turbulent impinging jets. arXiv preprint arXiv:1606.09167, 2016.
R. Wilke. The impinging jet. Unpublished doctoral thesis, Technische Universität Berlin, 2017.
Research Team & Contact Information:
Robert Wilke, Jörn Sesterhenn
Prof. Dr. Sc. techn. habil. Jörn Sesterhenn
Technische Universität Berlin
Department of Computational Fluid Dynamics
Müller-Breslau-Str. 15, D-10623 Berlin (Germany)
e-mail: Joern.Sesterhenn [at] TU-Berlin.DE