Cooling of the turbine section in gas turbine engines: combined experimental and numerical modeling using HPC infrastructure

Silvia Ravelli, Giovanna Barigozzi
University of Bergamo

Raffaele Ponzini

It is well known that the performance of a gas turbine engine is strongly influenced by the temperature at the inlet of the turbine section. Engine manufacturers have been continuously increasing turbine inlet temperature (TIT), since the last decades. Modern high performance gas turbines are designed to operate at TIT values (on the order of 1500-1800 K), well beyond the maximum allowable metal temperatures of about 1200 K. Consequently, high pressure turbine vanes and blades must be cooled by extracting air from the compressor stages. Elaborate cooling systems to protect the airfoil from severe thermal environment have been conceived, especially for the most critical regions, i.e the trailing edge. Here the focus is on a vane cooling scheme consisting of a trailing edge cutback with two rows of cooling holes, both located on the vane pressure side.
Computational Fluid Dynamics simulations were carried out to predict both thermal and aerodynamic performance of the cooled vane. Computational resources were provided by CINECA through a 2010-2011 LISA Initiative (Laboratory for Interdisciplinary Advanced Simulation a partnership between CINECA and Regione Lombardia) allowing to take advantage of High Performance Computing (HPC) platforms. Tailored experimental measurements for model validation purposes were collected in the wind tunnel installed at the Energy System and Turbomachinery Laboratory of Bergamo University, in the context of a national research project (PRIN 2007).
Results from different computational approaches will be presented and compared against experimental data. The investigated cases include the real-scale cooled vane with full details and also a simplified geometry with a smaller number of grid cells. The grid size ranged from ten to hundred million of cells. Numerical simulations were run using Fluent (Ansys.Inc; version 14.5) according to the adiabatic/conjugate approach in a steady/unsteady environment. RNG k-epsilon turbulence model provided closure. Finally, in order to evaluate the industrial feasibility of parametric studies and optimization of such a complex application, different HPC infrastructures were tested performing strong and weak scalability on up to 512 computational cores. Resulting synthetic indices (speed-up and efficiency) proved to be helpful in the evaluation of cost/effective benefits of CFD analysis.