Numerical Modelling Techniques for Turbine Blade Internal Cooling Passages

Abstract

Numerical modelling of internal cooling passages in gas turbine blades is a challenging task due to their physical characteristics, such as rounded duct corners, the presence of rib turbulators and their staggered locations between surfaces. This results in complex fluid dynamic phenomenon such as counter-rotating vortices and other secondary flow structures that can drive the heat transfer. Heat transfer mechanisms in such passages are inherently coupled with momentum transport and diffusion. Current industry practices for numerical modelling of such passages use unstructured mesh generation tools, steady Reynolds-averaged Navier-Stokes (RANS) equations and two-equation turbulence models such as k-ε and k-ω SST. This paper investigates two generic, engine-representative rib geometries using current numerical practices to determine their limitations. Three mesh generation tools and two turbulence models are compared across two rib geometries. The results are qualitatively and quantitatively compared to detailed experimental Nusselt numbers on the passage walls. It was found that as long as the rib geometry results in a secondary flow that directly impinges onto the wall, the meshing tools and turbulence models agree reasonably well with experiments. When the passage includes wall-wrapped ribs resulting in more complex secondary flows, this decreases the validity of the numerical tools, suggesting that more sophisticated modelling techniques are required as rib geometries continue to evolve.

Heat Transfer and Pressure Drop of an Angled Discrete Turbulator at Elevated Reynolds Numbers Up to 900,000

“…The use of coarser grid away from the ribs in the entry section (in comparison with the mesh around the ribs) may have also introduced numerical inaccuracies in capturing the flow development at the inlet. However, this inlet effect is not expected to have a significant The numerical approach in the current investigation is unable to capture the flow unsteadiness and the effect of the anisotropic characteristics of turbulence, which may play an important role in the turbulent characteristics of swirling flow and heat transfer near the ribbed walls, as also discussed by Dhopade et al [36]. It is the opinion of the authors that further evaluation of the various numerical approaches and a sensitivity study of thermal boundary conditions, similar to the works of Lörstad [3] and Iaccarino et al [42], would be beneficial for a comprehensive numerical capability assessment.…”

“…A mass flow inlet with a uniform velocity profile was applied at the inlet of the computational domain, which approximately represents the uniform flow velocity and temperature exiting the heater mesh shortly downstream of the wind tunnel inlet. wall temperature, similar to the approach of Dhopade et al [36], while adiabatic conditions were applied for the top and side walls. A no-slip boundary condition was imposed on all walls and the air is considered an incompressible ideal fluid, since the Mach number does not exceed 0.2.…”

Grid Generation and Fluid-Solid Coupling Methods for the Investigation of Gas Turbine Rotor Blade Internal Cooling

“…However, the meshing process for a complex 3D model with such a large mesh size is usually very timeconsuming. Even for experienced engineers, it may take days to weeks to create a mesh for a very complex internal cooling geometry [9,10]. When considering extra uncertainties coming from badly parametrised entities in practical applications, non watertight models, etc, a great waste of engineering hours and long meshing turn around time are frequent.…”

“…Although, this problem can be overcome by means of multi-block structured mesh [22], in which the computational domain is sub-divided into multiple blocks (see Figure 2.1 (a)), allowing much more geometrical flexibility. This process is usually very time-consuming and even for experienced engineers, it may take days to weeks to create mesh for a very complex internal cooling geometry [23]. Therefore, an unstructured based, automatic and less user-demanding method is preferable.…”