Flow and heat transfer in the row-1 upstream rotor-stator disc cavity of a large 3600-rpm industrial gas turbine was investigated using an integrated approach. A 2D axisymmetric transient thermal analysis using aero engine-based correlations was performed to predict the steady state metal temperatures and hot running seal clearances at ISO rated power condition. The cooling mass flow and the flow pattern assumption for the thermal model were obtained from the steady state 2D axisymmetric CFD study. The CFD model with wall heat transfer was validated using cavity steady state air temperatures and static pressures measured at inlet to the labyrinth seal and four cavity radial positions in an engine test which included the mean annulus static pressure at hub radius. The predicted wall temperature distribution from the matched thermal model was used in the CFD model by incorporating wall temperature curve-fit polynomial functions. Results indicate that although the high rim seal effectiveness prevents ingestion from entering the cavity, the disc pumping flow draws air from within the cavity to satisfy entrainment leading to an inflow along the stator. The supplied cooling flow exceeds the minimum sealing flow predicted from both the rotational Reynolds number-based correlation and the annulus Reynolds number correlation. However, the minimum disc pumping flow was found to be based on a modified entrainment expression with a turbulent flow parameter of 0.08. The predicted coefficient of discharge (Cd) of the industrial labyrinth seal from CFD was confirmed by modifying the carry-over effect of a correlation reported recently in the literature. Moreover, the relative effects of seal windage and heat transfer were obtained and it was found that contrary to what was expected, the universal windage correlation was more applicable than the aero engine-based labyrinth seal windage correlation. The CFD predicted disc heat flux profile showed reasonably good agreement with the free disc calculated heat flux. The irregular cavity shape and high rotational Reynolds number (in the order of 7×107) leads to entrance effects that produce a thicker turbulent boundary layer profile compared to that predicted by the 1/7 power velocity profile assumption.
