A GPU based 3D RANS multi-block slover with the capability of predicting heat transfer in fully resolved boundary layer flows based on the turbomachinery CFD code TBLOCK has been developed. A diabatic wall treatment and the scalar dissipation scheme have been integrated. The standard one-equation Spalart-Allmaras turbulence model with trip term is used for turbulence modeling. For validation, steady CFD results from the developed GPU version of TBLOCK are compared to a well-documented experiment of a highly loaded turbine cascade. The experiment covers a range of different Mach numbers and Reynolds numbers with static pressure and heat flux distributions on suction and pressure side of the blade. The predicted pressure and heat flux distribution show good agreement to the experimental data. A significant reduction of calculation time could be achieved compared to the authors’ CPU based MPI version of TBLOCK.

This content is only available via PDF.
You do not currently have access to this content.