The Effect of Purge-Mainstream Density Ratio on Turbine Cavity Flows

Abstract

Gas turbines are prevalent across many sectors within engineering, particularly in aerospace and power generation. A gas turbine consists of three sub-assemblies: a compressor, combustor and a turbine. It functions based on an open-loop Brayton cycle with the cycle efficiency increasing with the turbine entry temperature. In modern gas turbines, the turbine entry temperature exceeds the metallurgical limits of the materials used to manufacture the turbine discs. Consequently, air is bled from the compressor and used to cool the cavities between the turbine stators and rotors to avoid damaging the discs. The system through which this air is delivered is known as the Secondary Air System (SAS). The SAS is designed to purge the rotor-stator wheel-space with relatively cool air and prevent hot gases from the annulus from penetrating into the cavity, known as ingress. Ingress is believed to be caused by many potential mechanisms that arise from fluid interactions near the periphery of the cavity, known as the rim seal. Research into ingress has primarily been conducted using cold-flow, low Technology Readiness Level (TRL) rigs and seeded air-to-air interactions to track the flow of simulated coolant gases in the engine. However, the hot annulus gases have a lower density that the superposed gases in the cavity that gives rise to a purge-mainstream density ratio. This ratio is around 1.7 for state-of-the-art technology, but could rise to 2.7 with the transition to alternative fuel sources such as hydrogen. Gases with different densities interact via different mechanisms to those that are closer to homogenous mixtures, such as through naturally-occurring instabilities. Therefore, the fundamental interactions exhibited in industrial turbines may not be accurately reflected by research facilities conducting experiments near a unity density ratio.

To understand the impact that increasing the purge-mainstream density ratio has on the sealing performance of a turbine cavity, a series of experimental campaigns were conducted using a low TRL test rig. The campaigns were chosen to understand and identify the impact of the purge-mainstream density ratio under three conditions: standard operation, the effect of flow coefficient, and the interactions with the blade. Using the Large Annulus Rig, these effects were studied from both a time-averaged and unsteady perspective. To simulate the purge-mainstream density ratio observed in engines, CO₂ gas was introduced in increasing proportions to the purge flow. This enabled purge-mainstream density ratios in the range 1.00 ≤ DR ≤ 1.54 to be studied while maintaining isothermal conditions. Additional methodologies were developed using fundamental principles in chemistry to calculate localised mixture properties, and to convert measurements to non-dimensional parameters that can be used in low-order models. The cavity behaviour was studied using time-averaged pressure, swirl and sealing effectiveness measurements, in addition to unsteady pressure measurements taken in the rim seal to determine the effect of increasing the purge-mainstream density ratio on the fluid interactions in the rim seal. Measurements of sealing effectiveness were extended into the annulus for the standard operation case to track the flow emerging from the cavity into the main gas path. Furthermore, modifications to the design of the Large Annulus Rig were introduced to enhance the measurement capabilities without impacting concurrent programmes, in addition to other modifications that would have further expanded the capabilities and reach of this study. The Ingress Wave Model was applied to corroborate findings from the unsteady measurements and enable a more thorough understanding of the unsteady behaviour in the rim seal.

In standard operation ("on-design"), there was notably no change to the distribution of pressure coefficient in the rotor-stator cavity as a result of increasing the purge-mainstream density ratio, with some discrepancy at low flow rate and high density ratio being attributed to the definition of the pressure coefficient. The swirl ratio in the inner cavity was similarly unaffected by increasing the purge density, though the outer cavity experienced a suppression of the swirl ratio at lower flow rates when using denser coolant gas. Increasing the purge-mainstream density ratio was shown to improve sealing performance when considering the original definition of the non-dimensional sealing flow parameter, Φ₀. The non-dimensional unsteadiness in the rim seal was shown to weaken with increasing density ratio from both a modelled and experimental approach. The energy attributed to the large-scale structures was also shown to reduce with increasing density ratio, which was consistent across the range of sealing flow rates tested. Comparison between the Ingress Wave Model and synchronisation potential of the rim seal instabilities to the blade pressure field showed strong agreement at the "on-design" condition, strengthening the idea that the amplification of ingress from the traditional behaviour is related to synchronisation between two pressure fields.

Extending the study to consider flow coefficients at "off-design" conditions highlighted a persistent independence of the pressure coefficient in both cavities to the density ratio, beyond the definition-induced error. Further investigations were conducted into the suppression rate of the pressure coefficient across the range of flow coefficients which highlighted a more significant suppression rate around the region identified in the literature as being impacted by the blade. The swirl in the inner cavity continued to be unaffected by the increase in purge density and showed a suppression consistent with Batchelor-type flow. The suppression first observed in the outer cavity at the "on-design" condition occurred at higher flow rates for higher flow coefficients. This aligned with the sealing effectiveness which showed that the suppression occurred at the maximum amplification of ingress. Increasing the density ratio continued to increase the sealing effectiveness when considering the original Φ₀ definition. At C_F ≈ 0.43, the effectiveness in the outer cavity was initially observed to tend to ε_m → 0.5, such was the intensity of the amplified ingress. The sealing behaviour across the range of flow coefficients was shown to be consistent with behaviour observed in literature and increasing the density ratio had a similar effect to increasing the non-dimensional sealing flow rate. The non-dimensional unsteadiness continued to show a decrease in intensity with increasing density ratio. The energy of the large-scale structures continued to show a monotonic decrease at two of the flow rates tested, though at the highest flow rate, energy was seen to concentrate at the large-scales with increasing flow rate up to the flow rate of maximum amplified ingress. The Ingress Wave Model and synchronisation potential of the unsteadiness continued to show strong agreement, indicating that synchronisation was a significant contributor to the behaviour observed.

At conditions approaching RI (i.e. C_F → 0), the pressure coefficient in both cavities was persistently independent when considering an increased purge density. The swirl ratio in the inner cavity was also independent of the density ratio as per higher flow rates. The outer cavity showed an independence at low density ratio, but at high density ratio, the swirl ratio was seen to increase in the C_F ≈ 0.13 tests. With no annulus flow, the outer cavity showed both an independence to the density ratio and a monotonic decrease as purge flow rate was increased, consistent with suppression of Batchelor-type flow. In both the case of a bladed and bladeless rotor, there was a significant amplification of ingress at low flow rates for C_F ≈ 0.13. A similar behaviour was observed across a flow coefficient range in the "blade effect" region identified in literature, however, as this is also affecting the bladeless rotor, the mechanism must be different. With no annulus flow, the sealing effectiveness became independent of the density ratio when using a bladed rotor, and exhibited a classical sealing behaviour, indicating a fundamental change to the interaction. When the blades were removed, the sealing effectiveness became more buoyancy-driven and increasing the density ratio reduced sealing effectiveness for the first time. The intensity of the unsteadiness in the rim seal was at least an order of magnitude lower than at the higher flow coefficients tested, underscoring that the unsteadiness had a lower impact on the sealing behaviours in this flow coefficient range. The Ingress Wave Model and synchronisation potential disagreed at C_F ≈ 0.13, highlighting that the amplification of ingress in this region was not due to synchronisation to salient pressure fields. This was supported by the amplification increasing when the blades were removed.

Date of Award	20 May 2026
Original language	English
Awarding Institution	University of Bath
Sponsors	Siemens Energy Global GmbH & Co & EPSRC
Supervisor	Carl Sangan (Supervisor), James Scobie (Supervisor) & Gary Lock (Supervisor)