This thesis describes the computational study of the flow and heat transfer in a direct-transfer pre-swirl rotor-stator system. Pre-swirl cooling air enters the system at low radius through angled pre-swirl nozzles, located on the stator, impinges on the rotor and flows outward in the wheel-space between stator and rotor, and leaves the system through receiver holes, located on the rotor.
Computations were carried out using a 3D incompressible model, with one discrete pre-swirl nozzle on the stator and cyclic symmetry boundary conditions applied at the tangential faces of the domain. To permit steady-state computations, an annular outlet was used on the rotor that matched the centerline radius and total area of the receiver holes. The Reynolds-averaged Navier-Stokes equations in cylindrical polar coordinates were solved in primitive-variables using the finite-volume method, hybrid differencing and the SIMPLE pressure-correction scheme. The low-Reynolds-number Launder-Sharma turbulence model was used primarily and the Morse κ – ε model was also tested.
An axisymmetric numerical investigation was conducted to study the effect of the swirl ratio and other flow parameters on the flow and heat transfer in system, with computation times reduced by a factor of around 7 compared with the corresponding 3D computations.
The computations were also compared with data obtained from a complementary experimental study. The range of flow parameters tested in the experiments and used in the computations were: for rotational Reynolds numbers, 0.77×106 < Reφ < 1.2×106; for non-dimensional pre-swirl flow rates, 0.6×104 < cw,p < 2.8×104 (giving 0.12 λT,p = cw,pReφ-0.8 < 0.4); for pre-swirl ratio, 0.5 < βp < 3.
The computed and measured values of (tangentially-averaged) non-dimensional tangential velocity, Vφ/Ωr, and static and total pressure coefficient are mainly in good agreement. The computed results suggest that free-vortex flow occurs between the pre-swirl inlets and the receiver outlet. The results show a significant loss in total pressure near the pre-swirl inlets. An expression has been derived for calculating the discharge coefficient for the receiver outlet, and there is good agreement between measured and computed values.
The computed local Nusselt number, Nu, is compared with measured values. There is reasonably good agreement between computation and measurement for the level of Nu apart from the impingement region and radially outward of the receiver outlet. There is a large peak in Nu near the inlet radius, due the behaviour of the low-Reynolds number turbulence model in the impingement region. The measured effects of Reφ, λT,p and βp on the level of Nu are reproduced well by the computations.
|Date of Award||5 Sep 2003|
|Supervisor||Michael Wilson (Supervisor) & Mike Owen (Supervisor)|