1 Introduction

Modern low emission combustors produce turbine gas path radial temperature profiles which are relatively flat. This has resulted in endwall region over-temperature problems, specifically in the blade tip and platform regions. Because of this, and the desire to improve system efficiency by cooling flow reductions, the issue of endwall heat transfer has taken on greater significance in turbine design. Predicting heat transfer on the blade platform is difficult because of the highly complex nature of the flow field, which features seal purge air introduced into regions where secondary flows are already significant (Sharma and Butler, 1987). In addition, high performance turbines feature closely spaced blade rows and thus are characterized by inherently unsteady intra-blade regions. Here, the seal air emerges into a low-momentum endwall flow dominated by convected wakes from upstream vanes and potential disturbances from highly loaded downstream rotors.

Thermal management of internal disc cavities and structures is a critical durability issue. The literature contains many interesting computational and experimental works addressing this issue. These studies have focused on a variety of phenomena including windage effects resulting from facing surfaces which are alternately stationary and rotating, conduction from wetted surfaces to internal engine structural components, and leakage of hot mainstream gas through rim seals. Due to the extremely high temperature of the turbine mainstream flow, the most significant issue in this thermal management problem is the prevention of hot gas ingress.

The primary function of the purge air flowing from the disk cavities through the vane-rotor seal is to prevent hot gas ingress into these cavities. Researchers have focused considerable effort on the seal purge flow from this perspective (Green and Turner, 1994; and Hills et al., 1997). The recent review of the literature by Johnson et al. (1994) gives a complete discussion of seal ingestion including the extent to which studies have focused on the influence of mainstream flows. Von Karman's (1921) work on viscous pumping formed the basis for early approaches addressing the cavity seal flow problem. These studies expanded to include the effect of cavity geometry, surface irregularities (e.g. fasteners), gap sizes and forms, and rotational rates. These studies reveal some information about the expected nature of the flow leaving the seal, but are idealized to the extent that external flow effects were not considered. Thus, further studies of seal flow including the presence of external flows have been conducted with variations in the swirl of the external flow and pressure asymmetry of the external flow.

The intent of these studies has been to understand how to optimize the rim seal design. As this seal flow degrades the thermal efficiency of the gas turbine cycle, the goal is to minimize the seal flow rate required to eliminate ingress. However, less effort has been expended studying the effect of the seal flow on aerodynamics and heat transfer in the mainstream flow outside of the disk cavity.

Although the primary function of the seal flow is to prevent ingress, a beneficial side effect can be the enhancement of platform film cooling. However, the effectiveness of this cooling is dependent on the trajectory of the emergent seal flow, with this path not well understood. An understanding of the parasitic purge flow, which must be tolerated, may well allow optimization of ejection characteristics to maximize the platform cooling effect of this flow.

It is expected that the horseshoe vortex formed at the blade leading edge will be a significant mechanism in the transport of the seal cooling flow. The horseshoe vortex itself can play a crucial role in blade heat transfer, producing a scraping effect that enhances blade suction surface heat transfer and adversely affects surface temperature

(Blair 1994). Because of this, suppression of the horseshoe vortex has been a topic of interest in turbine design, and the purge flow may also have a role to play in this process.

This research effort is focused on developing a better understanding of the seal flow-mainstream flow interaction. With such knowledge, the motivation for rim seal design improvement may well grow to include not only the prevention of ingress, but also the optimization of blade heat transfer and aerodynamic performance.

1.1 Literature Survey

A single study by Granser and Schulenberg (1990) is concerned with the cooling effect of a purge flow introduced through a slot upstream of a blade row. This film cooling study stands alone in the literature as the only available work on the usefulness of purge flow in platform cooling. The issue remains of interest, however, as the work of Granser and Schulenberg was limited to a linear cascade of two stationary blades. Aside from this work, understanding of the complex flow issues and forces that contribute to the seal flow interaction with the mainstream flow must be garnered from other sources.

While most cavity flow work is dedicated to understanding heat transfer within the seal cavity or the cavity mechanisms that contribute to cavity ingress, there are some that consider the nature of external flows. Again these works concentrate on the cavity effects of the external flow and use highly simplified simulated mainstream flows. The review by Johnson et al. (1994) cites the literature as providing only one work including the effects of both vanes and blades. Nevertheless, for completeness, some of the work focused on cavity flows will be discussed here.

As the interaction of seal purge flow with secondary flow structures may influence the effectiveness of any hoped for external blade platform cooling benefits, information on secondary flow structures, film cooling, blade row flow mixing, and interaction of film cooling with secondary flows is of interest here.

The basic geometry to be considered is presented, Figure 1.1, in an engine cross section taken from Wilson et al. (1995). The typical inner engine cavities and cooling air flow paths are shown. Within the figure, the cavity space is called a 'wheel-space' and the cool air delivered to these cavities is called disc cooling air. Wilson et al. measured and predicted heat transfer within the cavity space, which is not of primary interest here, but contributed to an understanding of the motivation for these studies. They choose to study a simple cavity geometry which is common in the literature, and is sketched in Figure 1.2.

1.1.1 Seal Exit Profiles

Seal cavity flow work provides insight to possible velocity and temperature profiles of the flow exiting the seal gap. The study of Ko and Rhode (1992) is among those that present details of flow within the seal gap. They predict purge air exit velocity profiles into an external cross flow with near zero radial components at the upstream seal edge and very strong radial components at the downstream edge. All of their work generated time average results and did not predict ingress at any axial point along the seal exit plane. The computational studies by Hills et al. (1997) also provide insight to flow within the seal gap, adding the complication of external pressure asymmetry with upstream bluff bodies to simulate vanes and thus predicting a circumferential variation in the radial and tangential velocity components. Maximum ingress is predicted in the negative circumferential direction from the low pressure zone and maximum egress is predicted just to the positive side of the low pressure zone. The circumferential direction was taken as positive in the direction of rotor rotation. As with this study, all available information on the nature of the seal exit flow field has been predicted computationally.

Some experimental investigations that may add to understanding the nature of the seal exit flow have observed lobe patterns of recirculation within the wheel space. Work by Phadke and Owen (1988) and Nordquist et al. (1990) show non-uniformity existing within the wheel space flow field. These flow patterns certainly affect the way in which the seal flow interacts with the main stream flow. Interaction with the external flow may be one of the most important factors in understanding seal flows according to Chew et al. (1991).

Chew (1991) reviews selected works by Phadke (1982), Phadke and Owen (1983a, 1983b, 1988), Bayley and Owen (1970) and others to develop a mathematical model for prediction of cavity and seal flows. The work focused on the minimum seal flow rate to prevent ingress at the seal. The developed theory was applied to the experimental measurements of the investigators listed. Pressure ratio and seal gap dimension both influence the ingress problem, but it was found that they are linked together and not independent. Also Chew claims a non-linear relationship between ingress and rotational Reynolds number, which had not been proposed in previous theoretical studies. But the most significant conclusion, in light of the present research, is that a need exists for understanding the impact that pressure asymmetries in the external flow have on the level of ingress. He postulates that these pressure effects may be even more important than the seal flow Reynolds number. Four different geometries were considered in this discussion, but the main geometry of concern is of the same style as that used in the Purdue Research Turbine, and sketched in Figure 1.2.

To address these questions Chew et al. (1994) developed a mathematical model to describe cavity flow which included flow within the seal gap and the influence of an external flow described by an experimentally determined pressure variation due to upstream nozzle guide vanes. The result is a distinct circumferential variation in the seal gap through flow, with a location of inflow which is unexpected, and attributed to a strong circumferential component of velocity within the seal gap. Their primary conclusion from the study of an asymmetric external flow is that low seal flows are dominated by the vane induced pressure field, with the higher seal flow rate interaction with the mainstream flow is postulated to be important in influencing seal effectiveness.

Hartland et al. (1998) are concerned with the effect of circumferential pressure distributions on the seal flow. When the mainstream is of relatively high pressure then the seal flow pressure must be high to prevent ingress. Conversely, when the mainstream flow is of relatively low pressure than the high pressure seal flow will leak heavily into the mainstream. Thus some work has been done to profile the upstream endwall surface to cause a local reduction in pressure in those regions where the mainstream is normally high pressure. Hartland et al. make the point within theory development that circumferential pressure variations result from pressure waves that travel nearly normal to the flow direction or the exit blade angle, with the high pressure wave initiation point at the blade trailing edge. The low pressure waves are clocked one half of a vane pitch from the high pressure wave, and emanating from the vane suction surface.

Also influencing seal effectiveness, and hence the strength with which the purge will emerge into the main flow, is the swirl of the mainstream flow. Izenson et al. (1994) conclude that the magnitude of the external tangential velocity relative to the wheel speed can be increased to the point where significant increase in seal flow rates is required to maintain seal effectiveness.

Unfortunately, the seal flow literature is focused on predicting ingress, and the limited information available that presents egress details is far from enough to give a picture of the flow profile that should be expected from a seal. Additionally, a study of the circumferential temperature profile of the exiting seal flow may be useful but has not been found in the literature.

With a similar cavity geometry to that shown in Figure 1.2, a numerical study on heat transfer within the cavity was conducted by Ko and Rhode (1992). They predict a pattern of strong flow recirculation just within the cavity under the seal gap. This recirculation is credited for an area of increased temperatures, since the recirculation introduces thermal mixing, taking warm fluid near the exit and mixing it with cool fluid further inside the cavity. This indicates that care must be taken in using the rim seal flow for cooling as this mixing effect will minimize surface film cooling efficiency, with the fluid exiting the seal gap having been warmed within the cavity. The heat transfer studies within literature concentrate on effects applied to surfaces within the cavity.

1.1.2 Vane Exit Flows

The vane exit flow is convected across the seal gap. Vane exit flow structures and the vane developed pressure field are of particular interest. It is important to know the pressure field and the velocity field into which the seal flow is introduced, since the interaction of the two may influence downstream flow development.

Measurements of pressure, velocity and flow angle from a low speed single stage turbine nozzle guide vane by Zaccaria et al. (1996) were made at stations near the exit of the nozzle passage. The measurements give a clear picture of the flow structures near the endwall. The suction side of the passage shows strong vortical structures at both the tip and the endwall. These structures result from the passage vortex, as explained by Sharma and Butler (1987), Langston (1980) and Sieverding (1985). The structure results from a pressure gradient from the pressure side of the vane to the suction side of the vane which initiates a cross-passage flow within the boundary layers of the endwalls. By continuity, this wall flow sets-up two regions of circulation of opposite rotational direction within the passage. Once established, these circulations are then pushed towards the suction side of the passage, so that a survey of the passage exit shows that these structures have a strong vortex core on the suction side of the passage.

Figure 1.3 taken from Zaccaria et al. (1996) shows a secondary flow velocity vector plot at the suction surface and endwall corner. Point P is at the center of a counter clockwise rotation described as the passage vortex flanked by a weaker rotation labeled Hss, which is conjectured to be the remnant of the suction side leg of the horseshoe vortex.

An earlier study by Zaccaria and Lakshminarayana (1995) surveyed the development of these nozzle exit structures at downstream locations where the pressure side flow and the suction side flow interact. The interesting result was to see that the strong pressure differential within the wake served to annihilate the passage vortex core found in the suction side - endwall corner. Their results, Figure 1.4, show a series of structures of opposite rotation of the passage vortex across the entire blade span. The represented measurement location is 0.025 chords downstream of the trailing edge. These structures shown in the secondary (perturbation) velocity plot of Figure 1.4 are very tight vortices with the strongest appearing at the hub wall. It is this flow structure that is convected downstream toward the seal gap and into the rotor blade passage.

These experimental investigations of the vane flow field can be used to understand the type of structures expected to exist within the measurement region of the intra-stage space. It is important to note that the passage vortex which might be expected to impact the seal flow is not present at the downstream seal location, but rather the wake flow field which has structures counter to the rotation of the passage vortex exist within the measurement region.

1.1.3 Rotor Blade Passage Flow

These vane studies also serve as a model for the type of flow structures that exist within a rotor blade passage. The flow structure tracking within a vane row show that boundary layer fluid is swept into the passage vortex. The complementary rotor blade passage structure is then likely to capture all of the cooling flow emanating from the seal gap and entering the rotor passage.

An ethylene injection and detection technique is used by Denton and Usui (1981) to track flow through a low speed turbine. They claim that tracking through the nozzle row shows that "all fluid that enters the nozzle row close to the casing wall ends up in the core of the passage vortex." In fact the presented concentration distributions show a distinct migration from pressure to suction side for both the casing side injection and the hub side injection. The most interesting finding here is that only the lower half of the boundary layer is completely swept to the suction side, and the upper portion of that same boundary layer remains on the pressure side, reattaches and forms a new boundary layer beyond the passage vortex separation line.

When the tracking fluid passed through the downstream rotor row, it was expected that the rotor surface boundary layers with their low momentum fluid would result in sweeping the tracking fluid through a great circumferential distance. The fact that this was not witnessed in large concentrations resulted in the conclusion that very thin boundary layers exist on these surfaces, thus maximizing the effect that mainstream flow can have on surface heat transfer.

In fact the heat transfer studies of Hamabe et al. (1993) show high values of Nusselt number on the blade suction side near the trailing edge and the hub wall. This is the same location where the passage vortex is observed. The alignment of this flow structure with the region of high Nusselt numbers suggests that the scraping effect of the passage vortex is enhancing heat transfer at this point. In addition to high heat transfer at the trailing edge, their studies show relatively large Nusselt numbers at the leading edge, just to the suction side of the stagnation line, on the hub wall. These two areas are important to consider when projecting the influence that seal flow could have on heat transfer.

1.1.4 Film Cooling and Secondary Flows

Angled injection holes positioned at various vane endwall axial locations were used by Goldman and McLallin (1977) for a study of the secondary flow effects of platform film cooling. Vane row exit pressure profiles were used to determine the size of the vortex cores convecting off of the vane trailing edge along the end wall. These regions were measured for both endwall cooled and endwall un-cooled cases. The results were shown as pressure contours which indicate the size of the vortex core generated at the vane end wall interface along the suction side of the vane. These clearly show that the size of the vortex core is affected by the cooling flow. It appears that high pressure ratio coolant injection results in weaker vortex cores, given by a smaller area of reduced pressure. The investigators attribute the improvements to higher momentum fluid being introduced in the case of high injection pressure ratios. The injection pressure ratio is measured as the ratio of the inlet pressure of the coolant to the inlet pressure of the primary flow. It should also be noted that this coolant flow had an impact on the secondary flows. The lower momentum injection was carried along by the primary flow induced secondary structures, but higher momentum injection tended to disrupt the secondary structures generated by the primary flow stream.

Goldman and McLallin (1977) also note that a strong vortical structure is generated at the vane suction surface-end wall interface. The coolant delivered upstream is carried into these structures by the end wall flow migration from pressure surface to suction surface. Also note that the coolant injection studied was concerned with the angle of coolant injection, not the inclination of the hole which was 15° to the endwall, but with the angle of injection relative to the streamline direction. Directing the coolant at an angle relative to the flow direction resulted in redirecting the through-flow streamlines.

The direct influence of endwall cooling on secondary flow structures was summarized by saying, "[that] increased coolant momentum decreased under-turning in the passage vortex regions." Two studies on endwall flow injection trajectory in a turbine, Moore and Smith (1984) and Gaugler and Russell (1984), revealed little more than additional qualitative understanding of where path-lines typically exist within a single vane passages.

Granser and Schulenberg (1990) present results from a two vane cascade experiment. Their results identify the rim seal flow as useful for film cooling of the vane platform. The inlet flow in this case is uniform. Thus the measured film cooling effectiveness is uniform across the passage, and at the passage inlet. However, due to the passage vortex, the downstream cooling effectiveness is skewed advantageously towards the suction side of the passage. Flow in the near seal area is not presented in detail, but the effect of blowing rate and momentum addition is discussed with a focus on the limit at which the blowing ratio has a beneficial cooling effect. The resultant of cooling flow penetration through the mainstream boundary layer is an identifiable limit on the cooling effectiveness.

1.2 Research Objective

The overall objective of this research program is to gain an understanding of how purge flow introduced through a first stage turbine rim seal interacts with the external flow field so that downstream flow development can be understood for the purpose of generating meaningful heat transfer predictions. If improved platform cooling and/or secondary flow suppression is to be garnered from what is a necessary evil from a performance perspective, then a capability must be demonstrated to model these complex seal flows.

Unsteady viscous Computational Fluid Dynamics (CFD) simulations thus become the obvious tool of choice, and investigations of similar flows have already shown promise in compressor geometries (Heidegger et al., 1994). However, without data with which to confirm the predictions and calibrate the codes, the usefulness of such simulations is limited. To address this, the current research program is focused on providing unique experimental data to characterize the vane-blade seal purge flow behavior and its effect on secondary flow structures. To recreate the nature of this complex unsteady flow field, experiments are conducted in the Purdue Research Turbine, with the flow field in the seal region downstream of the first vane and upstream of the first rotor being characterized with Particle Image Velocimetry (PIV).

The specific objectives of the experimental work are to quantify the velocity field in the intra-stage space, including the velocity field that develops downstream of this seal injection point, since this initial interaction of the main flow and the injection flow reveals mixing information from the seal flow interaction with the horseshoe vortex. Measurements within this region are made such that variations of the flow field with circumferential location are captured. The measurements are intended to reveal the influence of upstream flow structure generated from the vane row, on flow development downstream of the seal flow.

Figure 1.1 A typical turbine cross-section with cooling flow paths

(from Wilson, 1995).

Figure 1.1 A typical turbine cross-section with cooling flow paths

(from Wilson, 1995).

Figure 1.2
Figure 1.3 Flow field within a turbine nozzle ( from Zaccaria et al., 1996).
Figure 1.4 Flow field at the exit of a turbine nozzle (Zaccaria and Lakshminarayana, 1995).
0 0

Post a comment