AbstractIn gas turbines, the ingress of hot annulus gas into cavities between stator and rotor discs is an important topic to engine designers. Rim seals are fitted at the periphery of these cavities to reduce the pressurised purge required to protect vulnerable, highly-stressed engine components. This thesis describes a computational study of ingress and egress through turbine rim seals. The computations were undertaken using time-domain and frequency-domain flow solvers available in DLR’s TRACE code. Several turbine configurations, incorporating different vane, blade and rim seal geometries were modelled and validation was performed against existing experimental results available from the University of Bath’s 1.5-stage axial turbine test rig.
The circumferential pressure distribution in the annulus was generally well-predicted across all configurations. The circumferential pressure difference, ΔCp, downstream of the vane was dependent upon the relative positions of the vanes and rim seal, however it was largely unaffected by the presence of the blades.
Egress from the upstream seal was shown to be entrained into the passage vortex. This feature moved radially outward of the hub, promoting little re-ingestion of upstream purge into the downstream wheel-space. Egress from the downstream wheel-space remained hub-bound due to weaker secondary flow features along the downstream vane.
The qualitative features of the flow structure in the seal clearances and wheel-spaces were captured, and a chute seal geometry demonstrated good quantitative prediction of ingress. However, the sealing effectiveness of radial overlap seals was over-predicted and it is speculated that the RANS turbulence model failed to accurately capture flow separation at the seal’s corners.
High levels of shear across the rim seals promoted the formation of large-scale structures at the periphery of the wheel-spaces, associated with increased levels of ingress. These transient structures are referred to as instabilities in this thesis. Unsteady time-domain computations showed that the presence and intensity of these structures are dependent upon both geometry and sealing flow rate. Structures were stronger for the upstream wheel-space than the downstream wheel-space due to a higher tangential shear gradient. The computed structures qualitatively reflected that measured in experiments, however their strength was generally over-predicted; this is again believed to be due to deficiencies in the RANS turbulence model.
Comparison of computational domains with circumferential sectors ranging from 30º to 360º indicate that the time-averaged features of the flow are largely unaffected by sector size. However, differences in large-scale flow structures present in the rim seals were pronounced with a 60º sector, suggesting that modelling an even number of blades in small sector simulations should be avoided.
A bladeless configuration was used to capture large-scale structures from a frequency domain computation for the first time. The harmonic balance solver captured similar behaviour to that of the time-domain solver, but with reduced computational cost. However, a reduction in the computed ingress was attributed to neglecting the transport equations for the higher harmonics of turbulence. Furthermore, in its current form the solver is not able to capture the non-linear interaction between the differing fundamental frequencies associated with blades and the large-scale structures.
The injection of high radius purge from the stator wall was shown to have a strong influence on the flow structure in the wheel-spaces. Increasing the proportion of purge supplied from these discrete positions disrupted the conventional Batchelor flow regime and led to the formation of recirculations. For the rim seal configuration studied, the presence of a buffer cavity restricted the influence of the high radius purge at the rim seal and resulted in no clear impact upon ingress.
|Date of Award||19 Nov 2019|
|Supervisor||Michael Wilson (Supervisor) & Carl Sangan (Supervisor)|
- gas turbine
- computational fluid dynamics
- rim seal