Some Observations on the Computational Sensitivity of Rotating Cavity Flows

Heat transfer inside rotating cavities plays an important role in gas turbine engineering. Flows in both compressors and turbine internal flow cavities exhibit self-generated large-scale inertial wave structures, and buoyancy effects are often important. Across the open literature on the topic, there seems to be no clear consensus on what the most suitable modelling fidelity is — although it is a widely held opinion that URANS approaches are less suitable than LES, many authors have succeeded in getting reasonable heat transfer results with URANS. There is also little knowledge of the validity of hybrid URANS/LES type approaches (such as DES) when it comes to predicting the heat transfer in these flows, and furthermore, on the sensitivity of the flow model validity to local driving aerothermal mechanisms in different parts of the cavity. This paper presents the results of a systematic investigation of a rotating cavity with axial throughflow at a Grashof number of 3 × 109. It is found that, for the case investigated, the disk Ekman layers remain laminar. This causes the disk heat transfer to be relatively insensitive to the modelling fidelity used with URANS, DES, and LES giving similar results. The effect of the disk thermal boundary condition is also investigated — it is found to have a significant effect on the direction of the near-wall flow at high radii, despite the large-scale flow structure within the cavity remaining essentially unchanged. This feedback of the disk heat transfer to the near-disk aerodynamics implies that conjugate heat transfer computations of rotating cavities may be worth investigating. On the shroud, URANS fails to resolve the heat transfer enhancement from small-scale buoyancy driven streaks, whilst these are captured by LES. DES also captures these streaks, as the URANS layer within which they are located returns a very small eddy viscosity, and behaves in a similar manner to LES.