1998
Abstract
Abstract
1997
Abstract
This paper describes the computation of flow and heat transfer in a rotating cavity with a stationary outer casing, using both a steady-state, axisymmetric finite volume solver and a time-dependent, axisymmetric direct numerical simulation (DNS) procedure. The geometric configuration represents the turbine disc cooling-air system used in some gas turbine-engines. Steady flow computations, using a low Reynolds number k-e turbulence model, give reasonably good predictions of Nusselt numbers, however convergence histories suggest that some of the flows computed may in reality be unstable. DNS reveals a highly unstable secondary flow structure, and gives rise to time-averaged velocity distributions in better agreement with experimental data than were obtained with the k-e turbulence model. There is some evidence that the effects of unsteadiness on heat transfer may be comparatively small.
Abstract
This paper describes a combined theoretical, computational and experimental study of the flow in an adiabatic pre-swirl rotor-stator system. Pre-swirl cooling air, supplied through nozzles in the stator, flows radially outward, in the rotating cavity between the rotating disc and a cover-plate attached to it, leaving the system through blade-cooling holes in the disc. An axisymmetric elliptic solver, incorporating the Launder-Sharma low-Reynolds-number k-e turbulence model, is used to compute the flow. An LDA system is used to measure the tangential component of velocity, Vf, in the rotating cavity of a purpose-built rotating-disc rig. For rotational Reynolds numbers up to 1.2 x 106 and pre-swirl ratios up to 2.5, agreement between the computed and measured values of Vf is mainly very good, and the results confirm that free-vortex flow occurs throughout most of the rotating cavity. Computed values of the pre-swirl effectiveness (or the nondimensional temperature difference between the pre-swirl and blade-cooling air) agree closely with theoretical values obtained from a thermodynamic analysis of an adiabatic system.
Abstract
This paper describes a combined computational and experimental study of the heat transfer in a rotating cavity with a peripheral inflow and outflow of cooling air for a range of rotational speeds and flow rates. Measurements are made in a purpose-built rig, with one of the two rotating discs heated, and computations are conducted using an axisymmetric elliptic solver incorporating the Launder-Sharma low-Reynolds-number k-e turbulence model. Measured values of the tangential component of velocity, Vf, exhibit Rankine-vortex behaviour which is not accurately modelled by the computations. Both computed and measured values of the radial component of velocity, Vr, confirm the recirculating nature of the flow. In the outflow region, agreement between computed and measured values of Vr is mainly good, but in the inflow region the computations exhibit a “peaky” distribution which is not shown by the measurements. The measured and computed Nusselt numbers show that Nu increases as the magnitudes of the flow rate and the rotational speed increase. The computed Nusselt numbers (allowing for the effects of conduction through and radiation to the unheated disc) reproduce the measured trends but tend to underestimate the experimental values at the larger radii.
Abstract
Conditions in the internal-air system of a high-pressure turbine stage are modelled using a rig comprising an outer preswirl chamber separated by a seal from an inner rotor-stator system. Preswirl nozzles in the stator supply the "blade-cooling" air, which leaves the system via holes in the rotor, and disc- cooling air enters at the centre of the system and leaves through clearances in the peripheral seals. The experimental rig is instrumented with thermocouples, fluxmeters, pitot-tubes, and pressure taps, enabling temperatures, heat fluxes, velocities and pressures to be measured at a number of radial locations. For rotational Reynolds numbers of Re,phi about 1.2 x 1e6, the swirl ratio and the ratios of disc-cooling and blade-cooling flow rates are chosen to be representative of those found inside gas turbines. Measured radial distributions of velocity, temperature, and Nusselt number are compared with computations obtained from an axisymmetric elliptic solver, featuring a low-Reynolds-number k-epsilon turbulence model. For the inner rotor-stator system, the computed core temperatures and velocities are in good agreement with measured values, but the Nusselt numbers are underpredicted. For the outer preswirl chamber, it was possible to make comparisons between the masured and computed values for cooling- air temperatures but not for Nusselt numbers. As expected, the temperature of the blade-cooling air decreases as the inlet swirl ratio increases, but the computed air temperatures are significantly lower than the masured ones. Overall, the results give valuable insight into some of the heat transfer characteristics of this complex system.
Abstract
Recent computational research has shown that finite-volume, elliptic solution procedures are capable of predicting accurately the turbulent flow and heat transfer in rotating-disc systems with a radial outflow of cooling air, representing conditions which occur in the internal cooling systems of gas-turbine engines. These include idealised systems, such as the rotating cavity and rotor-stator system, and more realistic internal cooling-air systems in which both the disc-cooling flow and the pre-swirl blade-cooling flow are modelled. This paper describes previous and new computational results for direct-transfer (rotor-stator) and cover-plate (rotating cavity) pre-swirl systems. In the cover-plate system, pre-swirl air flows radially outward as a free vortex between a rotating-disc and a cover-plate attached to it, while in the direct-transfer system the pre-swirl air is confined between inner and outer seals near the top of a rotor-stator system. Adiabatic calculations for both systems show the relative effectiveness of the two configurations, and validation data for flow and heat transfer is available from a purpose-built rotating-disc rig. Rotating-disc systems involving a radial inflow of air give rise to a combined free and forced vortex flow structure which is difficult to predict with basic two-equation turbulence models. Such flows occur in a rotating cavity in which cooling air enters and leaves through a stationary casing at the periphery of the system. It is shown that a Richardson number correction to the dissipation- rate equation in the turbulence model can be used to give some improvement of the flow predictions.
1996
Abstract
Encapsulated thermochromic liquid crystal (TLC) can be used to determine the surface temperature of stationary or rotating bodies. For narrow-band TLC, with a colour change from red to blue over a bandwidth of 1 degree C, temperatures on stationary surfaces can be measured with an uncertainty of around +/- 0.1 degree C. For gas turbine applications, where centripetal accelerations in excess of 1e4 g are common, there is a belief held by many reserach workers that TLC is significantly affected by rotation: "rotational shifts" have been observed where temperature measurements made by thermocouples and TLC diverge with increasing rotational speed; shifts up to 7 degree C at around 1e4 g have been observed. In this paper, the surface temperature of a rotating disc is measured simultaneously by an infrared (IR) thermal imager and wide-band TLC with an effective temperature range from 48 to 58 degree C. Within the uncertainty of the measurements, which was around 0.5 degree C, it was concluded that there was no significant rotational shift for accelerations up to 16,000 g
Abstract
In some engines, corotating gas-turbine discs are cooled by air introduced at the periphery of the system. The air enters through holes in a stationary peripheral casing and leaves through the rim seals between the casing and the discs. This paper describes a combined computational and experimental study of such a system for a range of flowrates and for rotational Reynolds numbers of up to Ref = 1.5 x 106. Computations are made using an axisymmetric elliptic solver, incorporating the Launder-Sharma low-Reynolds-number k-e turbulence model, and velocity measurements are obtained using laser-Doppler anemometry. The stationary peripheral casing creates a recirculation region: there is radial outflow in boundary layers on the discs and inflow in the core between the boundary layers. The radial extent of the recirculation region increases as the flow rate increases and as the rotational speed decreases. In the core, the radial and tangential components of velocity, Vr and Vf, are invariant in the axial direction, and the measured values of Vf conform to a Rankine-vortex flow. The agreement between the computed and measured velocities is not as good as that found for other rotating-disc systems, and deficiencies in the turbulence model are believed to be responsible.
Synopsis
The flow in the internal cooling-air systems of gas turbines can be modelled using simple rotating-disc systems. In this paper, finite-volume methods and low Reynolds-number k-e turbulence models are used to compute the flow and heat transfer between two confined discs with a radial outflow of cooling air. The two discs can rotate at different speeds and the systems can be characterised by the rotation ratio, G, of the slower to the faster disc. Computational results for axisymmetric, incompressible, steady flow are compared with data from purpose-built experimental rigs. The co-rotating (G=+1), rotor-stator (G=0) and contra-rotating (G=-1) disc systems discussed represent problems encountered in internal-air systems, and detailed understanding of the behaviour of these systems should lead to improvements in gas-turbine design with consequent improvements in efficiency.
Abstract
The paper describes a combined experimental and computational study of the heat transfer from an electrically-heated disc rotating close to an unheated stator. A radial outflow of cooling air was used to remove heat from the disc, and local Nusselt numbers were measured, using fluxmeters at seven radial locations, for nondimensional flow rates up to Cw = 9680 and rotational Reynolds numbers up to Re,phi = 1.2 x 1e6. Computations were carried out using an elliptic solver with a low-Reynolds-number k-epsilon turbulence model, and the agreement between the measured and computed velocities and Nusselt numbers was mainly good.
1995
Abstract
A superposed radial outflow of air is used to cool two discs that are rotating at equal and opposite speeds at rotational Reynolds numbers up to 1.2 x 1e6. One disc, which is heated up to 100 degree C, is instrumented with thermocouples and fluxmeters; the other disc, which is unheated, is made from transparent polycarbonate to allow the measurement of velocity using an LDA system. Measured Nusselt numbers and velocities are compared with computations made using an axisymmetric elliptic solver with a low Reynolds-number k-epsilon turbulence model. Over the range of flow rates and rotational speeds tested, agreement between the computations and measurements is mainly good. As suggested by the Reynolds analogy, the Nusselt numbers for vontra-rotating discs increase strongly with rotational speed and weakly with flow rate; they are lower than the values obtained under equivalent conditions in a rotor-stator system.
Abstract
The paper describes a combined experimental and computational study of laminar and turbulent flow between contra-rotating discs. Laminar computations produce Batchelor-type flow; radial outflow occurs in boundary layers on the discs and inflow is confined to a thin shear layer in the mid-plane: between the boundary layers and the shear layer, two contra-rotating cores of fluid are formed. Turbulent computations (using a low Reynolds-number k-epsilon turbulence model) and LDA measurements provide no evidence for Batchelor-type flow, even for rotational Reynolds numbers as low as 2.2 x 1e4. Whilst separate boundary layers are formed on the discs, radial inflow occurs in a single interior core that extends between the two boundary layers; in the core, rotational effects are weak. Although the flow in the core was always found to be turbulent, the flow in the boundary layers could remain laminar for rotational Reynolds numbers up to 1.2 x 1e5. For the case of a superposed outflow, there is a source region in which the radial component of velocity is everywhere positive; radially outward of this region, the flow is similar to that described above. Although the turbulence model exhibited premature transition from laminar to turbulent flow in the boundary layers, agreement between the computed and measured radial and tangential components of velocity was mainly good over a wide range of nondimensional flow rates and rotational Reynolds numbers.
1994
Wilson, M. and Owen, J. M. (1994)
Axisymmetric computations of flow and heat transfer in a pre-swirl rotor-stator system
1st International Conference on Flow Interaction, Hong Kong, pp 447-450
1993
Wilson, M., Syson, B. J. and Owen, J. M. (1993)
Image processing techniques applied to wide band thermochromic liquid crystals
Eurotherm 32 (Heat Transfer in Single Phase Flow), Oxford, pp 41-49
1992
1989