An Investigation Into a Novel Turbine Rotor Winglet - Proceedings

ASME

Proceedings of GT2006 of GT2006 ASME Turbo Expo 2006:Proceedings Power for Land, Sea and Air Turbo Expo 2006: Power for Land, Sea and Air May 8-11, 2006, Barcelona, Spain

May 8-11, 2006, Barcelona, Spain GT2006-90456

GT-2006-90456 An Investigation Into a Novel Turbine Rotor Winglet. Part 1: Design and Model Rig Test Results N W Harvey Compression Systems, Rolls-Royce plc

D A Newman, F Haselbach Turbine Systems, Rolls-Royce plc

L Willer Combustor & Turbines, Rolls-Royce Deutschland Ltd & Co KG ABSTRACT In modern gas turbines hot section components, the over tip leakage (OTL) flow that occurs between the stationary casing and rotating tip of a shroudless HP turbine is still a considerable source of loss of performance. The principal means of reducing this loss have been to minimise the tip gap and/or to apply a rotating shroud to the rotor. Tip clearance control systems continue to improve, but a practical limit on tip gap remains. Winglets have been identified by a number of researchers as having potential, but none have yet to enter commercial service. Harvey & Ramsden [1] analysed a novel design of one, which indicated that it could significantly reduce OTL loss. This paper presents the design of such a winglet as applied to the rotor blade of a research high pressure turbine carried out as part of the ANTLE (Advanced Near Term Low Emissions) technology demonstrator programme. The use of Computational Fluid Dynamics (CFD) calculations in the design process is discussed. In particular, the use of coarse meshes and idealised geometries, for computational speed, did involve some compromise with accuracy. Results from high speed model rig testing of this research turbine are presented. The turbine efficiency was measured for three different tip gaps over a range of conditions. In addition detailed measurements of the flow field were taken, principally exit area traverses and rotor surface static pressures. These experimental results are very encouraging and show a high potential for further development. Part II of this paper presents a post-test re-analysis of the rig results using the state of the art Rolls-Royce in-house CFD code HYDRA, good agreement being found between the two. NOMENCLATURE Ac/At OTL Clearance Area Parameter ANTLE Advanced Near Term Low Emissions c Chord cax Axial Chord Cd Discharge Coefficient, Ratio of Actual to Ideal Mass Flow CFD Computational Fluid Dynamics cp Specific Heat Capacity (constant pressure) cp∆T/T Turbine Stage Work (J/kg K)

∆H/U2 g h HP Mn NGV OTL p Re s V Vτ y Va/U y+ z

β η µ ρ ζ

Stage Loading Tip Clearance Gap Blade Span High Pressure Mach Number Nozzle Guide Vane Over Tip Leakage Pressure (static unless subscript denotes otherwise) Reynolds Number (=ρVc/µ) Pitch Velocity Skin Friction Velocity Distance Normal to the Surface in a Boundary Layer Turbine Flow Coefficient Non-dimensional Distance from the Surface = yVτ ρ µ Chord-wise distance Relative Angle Stage Efficiency Viscosity Density Row Kinetic Energy Loss Coefficient

Subscripts p Pressure Side rel Relative s Suction Side 0 Total/ Stagnation 2 Exit INTRODUCTION Most Rolls-Royce large Civil Aero Engines, in particular the Trent series, have single-stage HP Turbines with shrouded rotor blades. The shroud top geometry features two or three radial fins which act as a labyrinth seal to provide very low OTL loss, described in more detail later – see Figure 3 for a typical geometry. (The

1

Copyright © 2006 by ASME

physical size of these engines precludes the use of four or more radial fins, which can be achieved in some larger steam or gas turbines). However, future developments are likely to require hotter engine core temperatures, higher blade speeds and thus stresses, and reduced aerofoil numbers. In this environment it will be increasingly difficult to retain the full shroud. Partial shrouds are one solution already in use particularly in gas turbines for power generation. See for example the recent investigation by Nirmalan & Bailey [52] into “cut-back” shroud options using a high-speed linear cascade. These partial shrouds still have some (small) portion which is continuous around the circumference of the blade tip. Ultimately, though, it will not be possible to retain even this. Thus the work presented in this paper was undertaken to determine what could be achieved using a blade tip geometry, discontinuous at the tip but still overhanging it by as much as current fully shrouded geometries. An additional “subtext” for this paper has been to show how CFD can be used to design a novel concept even though the accuracy of its prediction of some parts of the flow field is open to question. In conjunction with this, part II of this paper demonstrates just how well current CFD can now calculate such flows. OVER TIP LEAKAGE Over tip leakage in axial flow turbomachinery has been the subject of extensive research since the advent of the gas turbine. An analysis of OTL loss is given in Denton [2], while VKI [3, 4] provide a thorough review of OTL flow in axial flow turbines.

Tip Gap Flow Field Figure 2 (taken from Denton [2]) illustrates the two typical regimes for flow through the tip gap, depending on the thickness of the aerofoil locally. These are sections through the blade, roughly normal to the camberline. The flow entering the gap from the blade pressure side separates from the tip and contracts to a jet (the “vena contracta”), which is largely lossless up to the minimum flow area. Figure 2 shows the flow for a sharp corner between the pressure surface and rotor tip. If the blade thickness is large enough (Figure 2a) the jet mixes out above the blade tip, increasing the loss and static pressure. Denton [2] states this occurs for a gap height/ local blade thickness of 4, while Heyes & Hodson [8] give a ratio of 6. The static pressure after this mixing is depressed by the blockage of the tip leakage vortex - see Morphis & Bindon [15] and Yaras & Sjolander [9, 10]. If the blade tip is thin enough, the jet will not reattach within the gap (Figure 2b) - this means that there is no pressure recovery in the gap and so the discharge coefficient (Cd) will be lower, Heyes et al. [12]. Maximum gap/ thickness ratios for this are 1.5 – 2.5, see Heyes & Hodson [8] and Denton [2]. For a typical turbine rotor with a plain tip this will only occur near the trailing edge. By using a cavity tip this effect can be achieved over the two side “squealers”, over most of the blade chord. This approach is now widely used, see VKI [4].

Shroudless Over Tip Leakage The basic form of over tip leakage in a shroudless axial flow turbine is illustrated in Figure 1. The pressure difference between the two surfaces of the aerofoil drives a leakage flow through the rotor tip/ casing clearance gap. Typically this is ejected as a strong jet which mixes with the main stream on the suction side, usually rolling up to form a vortex. This interacts in some way with the “classical” outer passage secondary flow vortex, see Sieverding [5] and Langston [16]. Detailed measurements of OTL flow have been made by Bindon [6], Moore & Tilton [7], Heyes & Hodson [8], Yaras et al. [9, 10], Morphis & Bindon [15] and recently Palafox et al. [11]. All of them studied rotor profiles in large scale, low speed, linear cascade. The last three sets modelled the relative motion between the casing and rotor with a moving end wall. The next sections give a summary of this research.

Figure 2. Tip gap flow for an unshrouded blade (Denton, [2]).

Figure 1. Illustration of OTL and outer passage secondary flows for a shroudless turbine rotor.

Effects of Secondary Flows and Relative Casing Motion The interaction between the OTL and outer passage secondary flows, and the effects of the relative motion between the rotor tip and the casing, have been observed to vary considerably: 1. The two vortices reinforce each other. Yamamoto [13] found in a low speed linear cascade that in the outer half of the passage the OTL and passage vortices are in close proximity and counter rotate. By the trailing edge, the outer passage vortex has moved to below the OTL vortex and near the aerofoil suction surface. Govardan et al. [14] obtained a similar result also in linear cascade, describing the interaction as intense. 2. With relative motion between the casing and the rotor the outer passage vortex is enhanced. Morphis & Bindon [15] and Yaras & 2


Sjolander [9, 10] found in their cascades that this reduced the driving pressure difference and “throttled” the OTL flow. They also concluded that pressure forces dominated the flow, rather than viscous ones. 3. Graham [18] obtained the opposite result in a water flow rig relative motion between the casing and rotor reduced the flow in the gap directly. This may have been related to the low values of Re in his experiment. Palafox et al. [11] obtained a similar result, and also found the passage vortex was weakened by the relative wall motion. 4. Chan et al. [19], using the same facilities as Yaras & Sjolander [9, 10], found that with a tip clearance of 5.5% of span the OTL vortex occupied almost the whole of the passage width by the trailing edge plane with only a small passage vortex. 5. Bindon & Morphis [20] found in their linear cascade that, uniquely, even by the trailing edge plane most of the OTL leakage flow remained in a flat high energy wall jet rather than rolling into a vortex. The cross passage secondary flow on the casing was weak and there was no indication of a passage vortex. The presence of the OTL jet, as opposed to the vortex, was confirmed in their testing of a full 1½ stage low speed annular rig, Morphis & Bindon [21, 22]. 6. Yamamoto et al. [23, 24] found, in a low speed 1½ stage turbine, that the rotor hub passage vortex confined the outer secondary flow and OTL vortices to near the casing. These two vortices were never present at the same time. The outer passage one was weak at a tip gap of 0.5% span and disappeared at a clearance of 1.9% span. Clearly there is a wide range of interactions possible in the flow field. A number have been identified here, and will be referred to later in the discussion of the analysis of the new winglet design. Loss Generation The tip clearance in a typical aero engine varies around its operating cycle and increases during the engine life due to wear and tear, see Stakolich & Stromberg [25]. The sensitivity of the turbine efficiency to changes in tip clearance is very important to the designer. It is usually expressed as an exchange rate: change of efficiency with change in clearance-to-span ratio, ∆η/∆g/h. Hourmouziadis & Albrecht [26] investigated a number of shroudless turbine rig and engine tests at MTU as well as other published studies, and found the OTL exchange rate is in the range 1.5 to 3.0, with a mean of about 2.0.

Figure 3. Perspective view of a typical shroud top geometry for a Rolls-Royce Civil HP turbine rotor. For shrouded turbines, Hartley [27] shows that for a geometry with two fins and two fences (see Figure 3) the exchange rate is reduced by a factor of 4 relative to a shroudless turbine, and by a factor of about 2 with just two fins present, see also Figure 14.

An explanation of the generality of the 2.0 ∆η/∆g/h exchange rate for shroudless rotors may be found by considering the control volume analysis of Denton [2] for the mixing of the gap flow with the passage flow. Denton presents an (incompressible flow) equation for the row kinetic energy loss coefficient ζ due to OTL: 1

2Cd  g   c  ς=    cos β2  h   s 

∫

 Vs     V2 

3

 Vp    Vp    1 −  1 −    Vs    Vs   

1

2

dz c

(1)

0

Most of the turbines considered by Hourmouziadis & Albrecht [26] have tip exit angles of 60° or above. For these Denton [2] calculated values of ζ of 4% or more for 1% g/h. The % change in stage efficiency is typically half that in the row loss (depending on reaction), giving a ∆η/∆g/h value of 2.0 or above. The dominance of the mixing process in the generation of OTL loss is (indirectly) demonstrated by Matsunuma [50] in a study of a micro-turbine which showed it to be independent of Re and inlet turbulence. Conclusions from OTL Research a) Apart from Morphis & Bindon, researchers agree that the OTL and passage secondary flows (when present) roll up into vortices soon after the blade trailing edge. There is then little scope for recovering the energy subsequently and it will be dissipated as loss. b) The effects of relative motion between the casing and the rotor are significant and must be included in any analysis. c) The loss is largely due to the mixing of the OTL flow with the suction side free stream; much less from mixing in the gap. OVER TIP LEAKAGE LOSS REDUCTION Based on the analysis of Denton [2], Harvey & Ramsden [1] discuss a number of options for reducing OTL loss. Three principal means of controlling the OTL loss of a shroudless HP turbine rotor at a given tip gap, other than by applying a full shroud, were identified. 1. Reduce the discharge coefficient Cd. 2. Aerodynamically off load the tip, in particular to reduce the suction side velocity Vs, shown by equation (1) to be very powerful. 3. Apply a winglet. These options are considered in the following sections. Reduced Discharge Coefficient This is the option most often pursued by researchers. Booth et al. [28] carried out an extensive investigation into rotor tip geometries aimed at reducing Cd. They found that a knife edge tip squealer had a Cd 25% lower than that for a plain tip, although the results were strongly dependent on tip gap Re and the details of the configurations. Heyes et al. [12] showed that a single suction side squealer gave the best reduction in Cd. Morphis & Bindon [15] gained a similar result with a contoured tip, which approximated to a suction side squealer. Researchers are still very active in this area. Recently Prakash et al. [17] demonstrated a 10% improvement in OTL loss exchange rate for a cavity tip blade with the pressure side squealer angled pointing into the OTL flow. While these results offer the possibility of reducing loss by directly reducing the OTL flow, there are a number of problems: a) Heyes et al. [12] showed that for any geometry to achieve a low Cd, the entry corner radius had to be kept below 0.5% chord. For a typical chord of 30 mm. the radius must be less than 0.15 mm. This dimension could not be maintained for an in-service engine. b) The vena contracta in the tip gap is a loss source. The lower the Cd, the larger the contraction and thus the mixing loss after it. In

3


addition, if the exit Mn is high enough, shocks will form over the contraction possibly adding to the loss, see Moore & Elward [29]. c) The presence of the vena contracta significantly increases the local heat transfer rates, especially at the reattachment point, as shown by a number of researchers such as Moore et al. [30], Metzger et al. [31] and Nasir et al. [32]. It is a major factor in the in-service problem of burnout at the blade tip pressure side, see Bindon [33]. Rather than try to reduce the Cd, it is suggested the tip pressure surface geometry should be sufficiently radiused to remove the vena contracta thereby reducing the potential for tip over heating. Yoon & Martinez-Botas [51] have also suggested rounding of this corner will improve the local effectiveness of film cooling rows placed at the tip. Rotor Tip Lean DeCecco et al. [34] and Yamamoto et al. [23, 24] originally suggested off-loading the tip to reduce OTL flow. Staubach et al. [35] achieved this by bowing their rotor thus applying a radial body force towards the casing, moving passage mass flow away from the tip. Their best result was with a tangentially bowed aerofoil only. The tip loss exchange rate was reduced by 40%. Figure 4, taken from [35], illustrates the (calculated) off-loading achieved. This does remain a mechanically challenging option for cooled turbines, however.

acceptable design solution the possibility of having two “aerofoils” at the rotor tip (mounted on a reduced size shroud), however, is. Despite the positive results that have been obtained for winglets, none is known to have entered commercial service. NEW WINGLET DESIGN From Harvey & Ramsden [1] and Harvey [37] it was concluded that the best opportunity for reducing OTL loss was to pursue a design of winglet that effectively doubled the number of rotor tips. It was also decided to aerodynamically off-load the rotor tip further by applying a limited, local tangential lean – which would not compromise the mechanical design in any engine application. This design would: a) Reduce the exit velocity of the OTL flow and of the free stream (on the suction side of the tip) that it mixes with. b) Reduce the OTL mass flow by off loading the aerofoil tip, but not by significantly changing the tip gap Cd - the latter rejected because of the risk of exacerbating in-service over heating problems.

p/po,rel

Figure 5. Illustration of winglet geometry, from [37].

Figure 4. Non-dimensional tip static pressure distribution for HP turbine rotor of Staubach et al. [35]. Winglets Winglets offer the possibility of modifying the local surface velocities at the rotor tip, in particular increasing Vp . The best result for one has been that of Patel [36] who obtained a stage efficiency improvement of 1.2% (at 3% g/h tip clearance). The tip loss exchange rate, however, was surprisingly unchanged from 2.0. Booth et al. [28] investigated a number of winglet designs in a water rig. Applying one of these (it is not clear which) to a low aspect ratio transonic turbine they achieved a 0.6% improvement in rotor efficiency at 3% g/h. Yaras & Sjolander [9,10] investigated winglets on the suction and pressure sides of a low turning aerofoil in linear cascade at 2.4% g/h. They obtained a reduction of 10% of OTL loss for each design. Staubach et al. [35] obtained a negative result with a winglet. They were primarily studying the effect of rotor lean on OTL, but found the winglet reduced stage efficiency by 0.35% at 1.7% g/h. Harvey & Ramsden [1] used equation (1) to estimate the effect of changing the pitch/ chord ratio s/c on the OTL loss. For the turbine they studied they predicted a 44% reduction in the loss for doubling the number of rotor tips. While doubling rotor numbers is not an

Application to ANTLE Research HP Turbine This winglet concept has now been applied to the rotor blade of the model HP Turbine of the ANTLE (Advanced Near Term Low Emissions) civil aero engine technology demonstrator. Its operating point is similar to that of the Rolls-Royce Trent 500 engine, see later, although the ANTLE HP rotor has about 23% fewer aerofoils. A numerical investigation of the rotor blade was undertaken, using a (then) standard turbomachinery CFD code used within RollsRoyce, over a range of tip gaps for the following geometries: 1. 2. 3.

Straight (radial) blade. Leant tip rotor, based on Staubach et al. [35]. Winglet tip.

The height of the winglet, illustrated in Figure 5, is about 8% of span. The basic elements of the geometry are as follows: 1. There are two “aerofoil” shapes at the rotor tip, which form a channel or “gutter”, along which there is an additional, chord-wise leakage flow from the leading edge. They are not equally spaced (pitch-wise), and thus are not equally loaded aerodynamically. 2. The pressure and suction side overhangs of the winglet almost halve the passage throat width at the tip. The pressure surface shape is intended to increase blockage, and thus lower the local static pressure driving the OTL flow. The

4


aerodynamic design did not include any cavities in the two “aerofoils” either side of the gutter. Cavities would likely be included in an engine design (to minimise weight and make use of the beneficial effects of squealers). However, for the model rig test these sides had to be solid – to aid the operation of the tip clearance measurement system. CALCULATION METHOD The CFD code used in this study is a Reynolds Averaged NavierStokes solver with a pressure correction method based on the algorithm of Moore [39]. The iterative method used is based on the SIMPLER pressure correction scheme. Stability in transonic regions is achieved using an upwinded pressure in the calculation of density. The calculations are based on a structured “letter-box” type of body fitted H-grid, which enables accurate representation of the full blade shape, which is then refined using mesh embedding. The capability of this method for the prediction of turbomachinery aerodynamics has been shown by Moore & Gregory-Smith [40], Moore & Moore [48] and Robinson & Northall [41].

Figure 6. Typical unstructured calculation grid. Blade-to-blade view of winglet, from [1].

Figure 7. Typical unstructured calculation grid. Axial view + pressure side tip detail, 50% cax. 7 nodes in tip gap, from [1].

Grid Details The calculation grids were first created as a coarse definition on the blade-to-blade plane, then stacked in the span-wise direction to produce three-dimensional grids. The unstructured grids were refined using mesh embedding, see Lapworth [42]. The maximum grid size available was about 160,000 nodes in the free stream. Typical unstructured grids are shown in Figures 6 and 7, the former being the blade-to-blade grid for a winglet, taken from Harvey & Ramsden [1]. Typical grid dimensions are 103 axially, 42 circumferentially and 39 to 53 radially with only 5 or 7 in the tip gap. The average near wall spacing y+ was 30 to 40 - acceptable with the use of wall functions. Generally the grid definition in the free stream is coarse, with refinement only in the boundary layer, the wake regions and at the tip. All the geometries were modelled with sharp corners at the tip. Boundary Conditions and Convergence Criteria All the calculations were “single row”, for the rotor only and were run to the same, fixed inlet relative stagnation conditions and relative whirl angle (derived themselves from a CFD calculation for the NGV) together with a specified exit static pressure profile. For simplicity the boundary layers were set to zero thickness at inlet to the grid. Thus the skewing of the hub end wall boundary layer, as it moves from the NGV exit onto the moving rotor end wall, was not modelled. Bindon [43] showed this increases rotor hub secondary flow, which as a result is underestimated here. All the boundary layers were set to be adiabatic and turbulent. Acceptable convergence was achieved when the residuals in all 3 velocity components and the static pressure had fallen by at least two orders of magnitude from their initial values. Usually this should be achieved after 100 iterations. However, the complexity of the grids used here meant that up to 400 were required for some calculations. The maximum error in mass flow conservation through the passage in any solution was 0.1%, and generally was within ± 0.05%. Turbulence Model and Wall Functions An algebraic mixing length model was used based on Prandtl's formulation for the length scale within a shear layer. Wall functions, described in more detail in Harvey et al. [44], were adopted to represent the near wall variation of the boundary layer based on a generalised expression for the law of the wall, see also White [45] or Spalding and Patankar [46]. They are valid for values of y+ up to the edge of the logarithmic region, say 100 to 200 depending on the magnitude of the local pressure gradient. Of particular importance is the ability to apply the wall function model in the buffer layer region, around y+ = 20, and to give good results irrespective of the y+ value which inevitably varies significantly over the aerofoil surface. CFD Validation Despite the extensive use of CFD, users must be realistic in their expectations of it. The CFD code used here is a general turbomachinery flow solver - validated for the calculation of the bulk flow field and secondary flow deviations, rather than loss. It does not calculate with equal accuracy all features of turbomachinery flow fields. The following ranking for these features is suggested (in order of decreasing calculation accuracy): Static pressure; Mass flow and exit angle distributions; Secondary flows; Overall entropy rise (in subsonic, attached flow); Shocks and separations; Local surface skin friction and/ or heat transfer rates. More details of the validation of the code are given in Harvey & Ramsden [1].

5


DESIGN RESULTS The CFD solutions generated for these tip geometries were not fully consistent. Results are presented, in the following sections, for: 1. 2.

Structured grid calculations for the radial and leant blades. Unstructured grid calculations for the leant and winglet tips.

Data is presented from these calculations in the form of: flow visualisation of the OTL; tip static pressure distributions; relative total pressure and exit whirl angle profiles at the trailing edge plane; OTL loss exchange rates.

1.1

0.9

p/p0,rel

OTL Flow Visualisation Figure 8 shows flow visualisation of the OTL flow for the straight blade (Figure 8a) and winglet (Figure 8b). Even though these are from structured and unstructured calculations respectively, the difference between them is much less than was seen for the rotor of the study of Harvey & Ramsden [1]. However, it does seem that the OTL vortex for the winglet tip rolls up less than for the plain tip.

Tip Pressure Distributions Figure 9 shows the calculated static pressure distributions around the blade tip, at 97% span. The data are presented as static pressure normalised by the mass averaged relative total pressure at inlet to the rotor, plotted against axial distance. Comparing the two structured grid calculations for the straight blade and leant tip, the effect of the lean in off-loading the tip, in particular on the suction surface, can be clearly seen. It is not as marked as that achieved by Staubach et al. [35], see Figure 4, though they did see the small increase in the loading on the pressure side. The static pressures from the unstructured grid calculation for the leant tip are very close to that of the structured grid. The small differences that can be seen on the suction surface are where the flow is transonic, which is not as well resolved by the unstructured grid.

0.7

Straight Blade Leant, Structured Leant, Unstructured Winglet Tip

0.5

0.3 10

14

18

22 26 Axial Distance (mm)

30

34

Figure 9. Calculated, non-dimensionalised tip static pressure distributions for ANTLE HP model blade at 97% span. Interestingly, the winglet does not off-load the suction surface very much more than the tip lean, comparing the two unstructured calculations – its biggest effect is on the pressure side. The net result is that the tip loading is halved, compared to the straight blade (and allowing for the differences between the calculation methods). a) Radial blade, plain tip

b) Blade with winglet Figure 8. Visualised OTL flow for ANTLE HP turbine model rig blade, from CFD calculation.

Flow Conditions in the Trailing Edge Plane Figures 10 and 11 show contours of relative total pressure and relative whirl angle respectively at the trailing edge grid plane for all four calculations. In both Figures care must be taken when reading across from the structured calculation for the straight blade to the unstructured one for the winglet. The differences between the two types of calculation for the leant blade must be taken into account. The variation in relative total temperature in the calculations is small, so in Figure 10 the deficit in relative total pressure can be taken as a measure of the loss. Generally the losses in the structured calculations are greater (both OTL and secondary flow loss cores larger) than those in the unstructured ones. Correspondingly, there is also a small bulk difference in the exit relative whirl angle. From Figures 10 and 11 the following qualitative conclusions can be drawn about the predicted effects of tip lean and the winglet: a. There is a strong interaction between the OTL and outer passage vortices. The latter, found directly below the OTL vortex at the trailing edge plane, is quite weak by comparison. However, since no inlet boundary layer was modelled it is very likely this is under estimated in the calculation. b. The depth and the span-wise extent of the OTL loss core (A) is progressively reduced by these tip treatments.

6


38

A

A

B

B

10c) Leant tip, unstructured grid.

10a) Radial blade, structured grid. A

A

B

B

10d) Winglet, unstructured grid.

10b). Leant tip, structured grid.

Figure 10. Calculated contours of p0,rel (kPa) in the outer half of the trailing edge plane for ANTLE model HP turbine blade. C

C

D

D

E

E

11a) Radial blade, structured grid.

11c) Leant blade, unstructured grid. C

C

D

D

E

E

11b) Leant blade, structured grid.

11d) Winglet, unstructured grid.

Figure 11. Calculated contours of βrel (º) in the outer half of the trailing edge plane for the ANTLE model HP turbine blade. 7


c.

However, with the winglet present the pitch-wise width of the OTL loss core does increase when the internal channel of the winglet is taken into consideration. d. Referring back to the flow visualisation of Figure 8 the OTL remains more of a jet at the tip, rather less a vortex. This is now closer to the OTL jet which Morphis & Bindon [21, 22] considered of benefit in the multi-stage environment. e. The change in whirl angle across the vortex is much reduced. In Figure 11a) from C to D it is at least 90º, while in Figure 11d) it has reduced to about 70º. It would be expected that the vortex being weaker would translate into reduced mixed out loss. f. The outer passage loss core (B), although smaller than that seen in the study of Harvey & Ramsden [1], seems to have been suppressed in the same way as for their rotor. This progressively improves with tip lean and then with the winglet. It is consistent with the reduction in local under turning (E) seen in the contours of relative whirl angle in Figure 11. Although not shown here, in VKI [4] velocity vector plots in the tip gap showed that the benefits of the winglet were not due to an increase in the calculated size of the tip separation bubble. % Tip Clearance / Span 0 % Loss of Stage Efficiency

0.0

0.5

1.4

1.8

Design Tip Gap

-1

-2

0.9

Straight Blade Leant, structured Leant, unstructured Winglet Tip

-3

Figure 12. Calculated OTL losses for ANTLE HP model turbine. Tip Loss Exchange Rates Figure 12 compares the calculated tip loss exchange rates. They are based on loss developed by the calculation grid exit, about 50% of axial chord downstream of the trailing edge of the blade. Two points must be made immediately. First, the OTL losses from the structured calculations are confirmed as being much greater than those from the unstructured ones. Secondly, although the qualitative comparisons of the flow field gave some hope of the winglet reducing the OTL loss, this is not reflected in the calculated tip loss slopes. There is a small reduction in the tip loss slope calculated for tip lean, about 7%, while the winglet appears to offer no further benefit This is disappointing, given the greater degree of off-loading of the tip, and it is a concern that the biggest changes in the loss are between the two calculation methods rather than between different geometries. Final Discussion of Design Study Despite the disappointingly small reductions in OTL loss predicted for the winglet, the HP model turbine rig test for the ANTLE programme went ahead with this geometry. As was noted before, loss is well down the ranking of CFD accuracy. The winglet was chosen largely on the basis of the very encouraging qualitative improvements in the

calculated flow field. In particular the off-loading of the rotor tip was comparable with that obtained by Staubach et al. [35]. Some further comments may be made about the design process, which was completed under significant time constraints. The exercise was carried out with computer-based tools that were stretched to the limit of their capability (and have now been superseded). In particular the geometry and subsequent mesh generation processes were highly labour intensive – and involved some idealisation of the geometries. For instance, fillet radii between the aerofoil and the winglet (present in the actual model blade) could not be modelled for the calculation. The grid refinement process could not easily be constrained to one locality and each such iteration usually resulted in a very large increase in the number of grid nodes. As a consequence, to keep to grid sizes that were practicable given the computational resources available at the time, the definition of the tip gap was very coarse (only 5 or 7 points in the gap, as previously noted). Ideally, much larger grids should have been used – not just to represent the geometry but also to resolve the mixing downstream of the aerofoil better. This was not possible in the time, and thus the final decision on the acceptability of the design was based mainly on the prediction of the static pressure distributions – the flow parameter that CFD generally predicts with the highest accuracy. In part II of this paper, results of analysis of the rotor blade including the full (manufactured) geometry using state-of-the-art CFD tools and high quality grids is presented. As will be seen, a much improved calculation of the flow field is then possible. COLD FLOW RIG TEST VALIDATION Figure 13 shows a general arrangement of the HP turbine cold flow model rig configured for the (shrouded) HP Turbine of the Trent 500 engine. Although the rig does not model the cooling air flows present in the engine, the aerofoils reproduce the engine nondimensional velocity triangles. The general arrangement of the rig (on the same test facility) for the ANTLE HP Turbine is very similar. The principal difference is that, of course, the rotor blade is unshrouded. Non-dimensional operating conditions for the Trent 500 HP model turbine, taken from Harvey et al. [38] are as follows: Stage Loading, ∆H/U2 Flow Coefficient, Va/U Pressure Ratio Inlet Turbulence Intensity NGV Re Rotor Re

1.54 0.39 2.4 ≈ 4% 6.8x105 4.4x105

The inlet mass flow is metered using a calibrated ISA nozzle upstream of the working section. The total temperature is measured in the inlet plenum chamber upstream of the rig working section. The inlet total pressure is derived from a continuity calculation using measured annulus wall static pressures in the inlet duct and the known rig mass flow and (zero) inlet flow angle. Back-up measurements of the total pressure and temperature are provided by rakes in the inlet. Downstream of the HP rotor there were two traverse planes: the rotor “close up” one and the far downstream plane at which the overall performance is evaluated. Area traverses using three-hole pneumatic pressure probes and hot wires, over 1½ NGV pitches, were made at the close-up plane at the design point only. The exit total pressure is determined, for performance evaluation, from a continuity calculation using the measured flow angles and annulus wall static pressures at the downstream plane and the known

8


Rotor Close-up Traverse Plane HP NGV Exit Annulus Static Tapings Inlet Temperature and Pressure Rakes

Exit Parallel Annulus Traverse Plane

Tip Clearance Measurement Rotor

Inlet Annulus Static Tapings

Exit Rakes NGV

HP Rotor Exit Annulus Static Tapings

Exit Annulus Static Tapings

Disc Cavity Pressure and Temperature Instrumentation Figure 13. General Arrangement of the Trent 500 HP turbine model rig.

RESULTS This paper presents only the basic aerodynamic results from the rig test, in the form of OTL loss exchange rates, rotor surface static pressures and radial profiles of flow conditions at the rotor exit. (More detailed data, from hot wire measurements, are presented in paper II). For reference, these results are compared with those from the Trent 500 HP model turbine rig tests which had a shrouded rotor with effectively two fins, see Rose et al. [47].

ANTLE Rig: small ANTLE Rig: intermediate ANTLE Rig: large Trent 500 Rig (fin tip gap) 6

ANTLE HP Rig Stage Efficiency Loss ∆η%

rig mass flow. The temperature drop across the turbine is determined from the power output, measured using both a dynamometer and shaft torquemeter. Tip clearances are measured using Caplong capacitance probes at 8 circumferential locations around the HP rotor casing To determine the turbine efficiency from the overall results a statistical analysis of all the available data is undertaken, interpreted via a best fit through flow model. This “χ2” method is described in detail in Rose et al. [47]. The repeatability of the measured turbine efficiencies was ± 0.25%.

Ac 2.π .g.r0 = At Throat Area and r0 is the mean radius of the casing over the rotor tip.

4 3 2

0

Shrouded: 2 Fins 3 Fins 2 Fins + 2 Fences Trent 500 Rig Clearance 0 Ratio of Clearance 0.02 Area0.04 to Throat Area0.06 (Ac/At)

(2)

In the case of a shrouded blade with radial fins, if the tip gap is that over the last fin, then this parameter is the ratio of the controlling flow area of the OTL to that of the main stream flow. Following Taylor et al. [49], this is used to compare shrouded and shroudless turbine blades. Figure 14 presents tip clearance exchange rates for a number of different shrouded blade geometries (two fins, three fins and two fins plus fences as shown in Figure 3) together with the corresponding curve for a shroudless blade. These are averaged results from Rolls-Royce in-house model and engine parts rig tests. Taking the intermediate of the three ANTLE tip gaps as the nominal value, the others are related as follows:

Shroudless

5

1

OTL Loss Exchange Rate The turbine performance was measured at three different tip gaps. Figure 14 plots the change in stage efficiency, derived from χ2 analysis against tip clearance parameter Ac/At where:

Ac/At 0.87 x nominal nominal 1.28 x nominal 0.68 x nominal

0.08

Figure 14. Summary of turbine rig OTL loss exchange rates. A number of points should be made about Figure 14: 1. The measured tip loss slope is about half that of a shroudless blade. However, the ANTLE rig was tested at relatively large tip gaps, and the line fit through the data requires a significant extrapolation to derive the zero gap efficiency. Given this the winglet blade can only be said to be as good as a shroud with two fins – a 45% reduction in loss slope, almost exactly as expected from equation [1]. 2. The dashed line indicates the clearance the Trent 500 rig was tested at. At this value, the winglet blade still falls short of the best shrouded geometry by 1% of stage efficiency. 3. However, the data for two fins and two fences is from engine parts rig testing with tip ejected cooling air present. This is understood to beneficially interact with the over tip leakage, see 9


Taylor et al [49]. This air was not present in the ANTLE model rig, and one of the questions remaining around the winglet is whether its performance can be further improved once this flow is included. 1

former effectively delivers an outer wall boundary layer into the next blade row that is about 20% of span in height. This might be expected to significantly increase the secondary flow in that row. By contrast the profile from ANTLE should have a much less deleterious effect. 100 5º

90

% Annulus Height

p/p0,rel

0.8 CFD: small tip gap CFD: large tip gap Data: nominal tip gap

0.6

70 ANTLE: nominal Ac/At ANTLE: 1.28 x Ac/At Trent 500: 0.68 x Ac/At

60

0.4 0

20

40 60 % Axial Chord

80

100 50

Figure 15. ANTLE HP blade mid-height static pressures. Rotor Surface Static Pressures Time averaged static pressures, obtained from Kulite probes mounted along the mid-height of the rotor aerofoil are plotted in Figure 15, non-dimensionalised by the mass averaged inlet relative total pressure. The data was taken at the nominal tip gap. Also plotted are the CFD predictions (unstructured grid) made at the small and large tip gaps. The match between calculation and measurement, especially on the suction surface, is remarkably good given the former are true predictions. There may be a systematic discrepancy on the pressure surface - which would indicate the rotor ran at slight positive incidence at this condition. Generally, however, this data does give confidence that the turbine was operating largely as expected. Measured Radial Profiles Radial profiles of total pressure (non-dimensionalised by the mass averaged value) and absolute whirl angle have been averaged from area traverses made in the rotor “close-up” plane, see Figure 13, at the nominal and large tip gaps. These are plotted in Figures 16 and 17 for the outer half of the blade span – to focus on the OTL region.

-50 -45 -40 -35 -30 -25 -20 (Over Turning) Exit Whirl Angle

-15

-10 -5 0 (Under Turning)

Figure 17. Turbine absolute exit whirl angle profiles. In Figure 17 the winglet at nominal clearance has the same maximum under turning as the Trent 500, but the radial extent of this region from the casing is significantly greater. The local maximum in under turning inboard of this, at 80% span in the Trent 500 rig, is associated with the outer passage secondary vortex. This seems to have been largely eliminated in the ANTLE turbine, as expected from the design calculations. Derived Radial Profiles Figures 18 and 19 show, respectively, radial profiles of the variation of row efficiency (about the mean value for each) and relative whirl angle. These have been derived from the χ2 through flow analysis of the measured data. [Note that care must be taken in interpreting the data at the walls, due to the limitations of the calculation in modelling the (highly 3-D) flow in this region]. 100

90

% Annulus Height

100 90 % Annulus Height

80

80 70 ANTLE: nominal Ac/At ANTLE: 1.28 x Ac/At Trent 500: 0.68 x Ac/At

60 50 0.94

80

ANTLE: nominal Ac/At ANTLE: 1.28 x Ac/At Trent 500: 0.68 x Ac/At

70

60

50

0.96

0.98

1

1.02

-20

1.04

Figure 16 indicates a very significant effect of the tip gap on the total pressure profile for the ANTLE turbine – even with the winglet present. However, the profile for the shrouded turbine (at the smaller tip gap) is much steeper than for ANTLE at the nominal tip gap. The

-10

-5

0

5

10

Variation in Row Efficiency (%)

Absolute Total Pressure (non-dimensionalised)

Figure 16. Turbine exit radial total pressure profiles.

-15

Figure 18. Turbine rotor row efficiency profiles. Examining Figure 18 provides the surprising result that the ANTLE blade profile at the nominal tip gap is really no worse than the shrouded blade (which is at a smaller tip gap). This does, though, provide some corroboration to the stage efficiency measurements.

10


The relative whirl angle profiles, shown in Figure 19, provide a very different picture again. As with the absolute whirl angles the effect of increasing the tip gap by 28% is large – the maximum under turning increases by about 4º. The shrouded blade exhibits large over turning at the casing while the ANTLE blade shows under turned flow – which would be expected for a shroudless blade. This behaviour of the shrouded blade may be as much due to the outer passage secondary flow as to the OTL itself. The relative effects of these profiles (on any downstream blade row) cannot be simply estimated. 100

% Annulus Height

90

80

70 2º 60

ANTLE: nominal Ac/At ANTLE: 1.28 x Ac/At Trent 500: 0.68 x Ac/At

50 -80 -78 -76 Over Turning

-74

-72 -70 Whirl Angle

-68

-66 -64 -62 Under Turning

Figure 19. Turbine relative exit whirl angle profiles. CONCLUSIONS A novel turbine blade winglet, first described by Harvey & Ramsden [1], has been applied to the shroudless rotor blade of the ANTLE HP turbine model rig. This has been investigated in a high speed cold flow rig and the results compared with those of the previous (shrouded) Trent 500 HP turbine model rig test. The winglet has been shown to be as good as a shroud with two fins in reducing OTL loss (45% below shroudless). Examination of the exit flow field corroborates this. In terms of row efficiency and total pressure profiles the tip flow field for the winglet is much the same as for the shrouded geometry. The exit whirl angle profiles, however, indicate a more uneven flow field than for the shroud. From a purely aerodynamic point of view, the winglet is now available for use in civil HP turbines in the case where the physical limits on circumferentially contiguous shrouds have been reached. Currently, however, Rolls-Royce continues to use fully shrouded HP turbine rotor blades even in its latest Trent engines (such as for the Boeing 787 Dreamliner) - not least for their excellent performance retention. The winglet cannot replace these until the OTL loss has been reduced to that of a full shroud with two fins and two fences. Possible areas for future improvement to achieve this are optimisation of the winglet shape, using cavities in the winglet sides and the inclusion of tip ejected cooling air - already beneficial for the shroud top geometries currently employed by Rolls-Royce. Issues around the mechanical design and cooling of the winglets are discussed further in part II of this paper. In addition, it has been shown how CFD tools can be used to successfully design such a novel concept even when some key features of the flow field, such as loss, are not accurately predicted. ACKNOWLEDGEMENT This work has been carried out with the support of Rolls Royce plc. The authors would like to thank them for their permission to publish this paper.

REFERENCES [1] Harvey N. W., Ramsden K., (2001), “A Computational Study of a Novel Turbine Rotor Partial Shroud”, ASME J. of Turbomachinery, Vol. 123, pp. 534-543. [2] Denton J. D., (1993), “Loss Mechanisms in Turbomachines”, ASME 93-GT-435. [3] VKI Lecture Series, (1997), “Secondary and tip-clearance flows in axial turbines”, VKI LS 1997-01. [4] VKI Lecture series, (2004), “Turbine Blade Tip Design and Tip Clearance Treatment”, VKI LS 2004-02. [5] Sieverding C. H.,(1985), “Secondary Flows in straight and annular Turbine Cascades”, Ed. Ucer, Stow and Hirsch. Thermodynamics & fluids of Turbomachinery, Nato Series, Vol. II, PP. 621-624. [6] Bindon J. P., (1988), “The Measurement and Formation of Tip Clearance Loss”, ASME 88-GT-203. [7] Moore J., Tilton J. S., (1988), “Tip Leakage Flow in a Linear Turbine Cascade”, ASME J. of Turbomachinery, Vol. 100, pp. 18-26. [8] Heyes F. J., Hodson H., (1992), “ Measurement and Prediction of Tip Clearance Flow in Linear Turbine Cascades”, ASME 92-GT-214. [9] Yaras M. I., Sjolander S. A., (1991), “Effects of Simulated Rotation on Tip Leakage in a Planar Cascade of Turbine Blades. Part I: Tip Gap Flow”, ASME 91-GT-127. [10] Yaras M. I., Sjolander S. A., (1991), “Effects of Simulated Rotation on Tip Leakage in a Planar Cascade of Turbine Blades. Part II: Downstream Flow Field and Blade Loading” ASME 91-GT-128. [11] Palafox P., LaGraff J. E., Oldfield M. L. G., Jones T. V., (2005), “PIV Maps of Tip Leakage and Secondary Flow Fields on a Low Speed Turbine Blade with Moving Endwall”, ASME GT2005-68189. [12] Heyes F. J. G., Hodson H. P., Dailey G. M., (1991), “The Effect of Blade Tip Geometry on the Tip Leakage Flow in Axial Turbine Cascades”, ASME 91-GT-135. [13] Yamamoto A., (1989), “Endwall Flow/Loss Mechanisms in a Linear Turbine Cascade with Tip Clearance”, ASME J. of Turbomachinery, Vol. 111, pp. 264-275. [14] Govardan M., Venktrayulu N., Vishnubhotla V. S., (1993), “Tip Clearance Effects on the Flow Field of an Axial Turbine Rotor Blade Cascade”, ISABE 93-7057. [15] Morphis, G., Bindon J., (1988), “The Effects of Relative Motion, Blade Edge Radius and Gap Size on Blade Tip Pressure Distribution in an Annular Cascade with Clearance”, ASME 88-GT-256. [16] Langston L. S., (2001), “Secondary Flows in Axial Turbines - A Review”, Heat Transfer in Gas Turbine Systems, Annals of the New York Academy of Sciences, Vol. 934, May 2001, pp. 11-26. [17] Prakash C., Lee C-P., Cherry D., Wadia A., (2005), “Analysis of Some Improved Blade Tip Concepts”, ASME GT2005-68333. [18] Graham J. A. H., (1985), “Investigation of a Tip Clearance Cascade in a Water Analogy Rig”, ASME 85-IGT-65. [19] Chan J. K. K., Yaras M. I., Sjolander S .A., (1994), “Interaction Between Inlet Boundary Layer, Tip-Leakage and Secondary Flows in a Low-Speed Turbine Cascade”, ASME 94-GT-250. [20] Bindon J. P., Morphis, G., (1990), “The Development of Axial Turbine Leakage Loss for Two Profiled Tip Geometries Using Linear Cascade Data”, ASME 90-GT-152. [21] Morphis, G., Bindon, J. P., (1994), “The Flow in a Second Stage Nozzle of a Low Speed Axial Turbine and its Effect on Tip Clearance Loss Development”, ASME 94-GT-145. [22] Morphis G., Bindon J. P., (1994), “The Performance of a Low Speed One and a Half Stage Axial Turbine with Varying Rotor Tip Clearance and Tip Gap Geometry”, ASME 94-GT-481.

11


[23] Yamamoto A., Tominaga J., Matsunuma T., (1994), “Detailed Measurements of Three-Dimensional Flows and Losses Inside an Axial Flow Turbine Rotor”, ASME 94-GT-348. [24] Yamamoto A., Matsunuma T., Ikeuchi K., (1994), “Unsteady Endwall/Tip-Clearance Flows and Losses due to Turbine Rotor-Stator Interaction”, ASME 94-GT-461. [25] Stakolich E. G., Stromberg W. J., (1983), “JT9D Performance Deterioration Results from a Simulated Aerodynamic Load Test”, AIAA Journal of Aircraft, Vol. 20, No. 8, pp. 650-658. [26] Hourmouziadis J., Albrecht G., (1987), “An Integrated Aero/Mechanical Performance Approach to High Technology Turbine Design”, AGARD-CP-421, “Advanced Technology for Aero Gas Turbine Components”. [27] Hartley R., (1996), “High Pressure Turbine Tip Clearance Performance Investigation”, MSc Thesis, Cranfield University. [28] Booth T. C., Dodge P. R., Hepworth H. K., (1981), “Rotor-Tip Leakage Part I - Basic Methodology”, ASME 81-GT-71. [29] Moore J., Elward K. M., (1992), “Shock Formation in Overexpanded Tip Leakage Flow”, ASME 92-GT-1. [30] Moore J., Moore J. G., Henry G., (1989), “Flow and Heat Transfer in Turbine Tip Gaps” ASME J. Turbomachinery, Vol. 111, pp. 73-79. [31] Metzger D., Rued K., Chyu M., (1989), “Influence of Clearance leakage on Turbine Heat Transfer at and near Blade Tips. Summary of Recent Results”, AIAA 89-0327. [32] Nasir H., Ekkad S. V., Bunker R. S., (2005), “Effect of Tip and Pressure Side Coolant Injection on Heat Transfer Distributions for a Plane and Recessed Tip”, ASME GT2005-68595. [33] Bindon J. P., (1987), “Pressure Distributions in the Tip Clearance Region of an Unshrouded Axial Turbine as Affecting the problem of Tip Burnout”, ASME 87-GT-230. [34] De Cecco, S., Yaras, M. I., Sjolander, S. A., (1995), “Measurements of the Tip-Leakage Flow in a Turbine Cascade with Large Clearances”, ASME 95-GT-77. [35] Staubach J. B., Sharma O. P., Stetson G., (1996), “Reduction of Tip Clearance Losses Through 3-D Airfoil Designs”, ASME 96-TA-13. [36] Patel K. V., (1980), “Research on a High Work Axial Gas Generator Turbine”, SAE 800618. [37] Harvey N. W., (1997), “Over Tip Leakage Control in Axial Flow Turbines”, MSc Thesis, Cranfield University. [38] Harvey N. W., Brennan, G., Rose, M. G., Newman, D. A., (2002), “Improving Turbine Efficiency using Non-Axisymmetric End Walls:

Validation in the Multi-Row Environment and with Low Aspect Ratio Blading.”, ASME GT-2002-30337. [39] Moore J. G., (1985), “Calculation of 3D Flow without Numerical Mixing”, AGARD-LS-140 on 3D Computation Techniques applied to Internal Flows in Propulsion Systems, pp 8.1-8.15. [40] Moore H., Gregory-Smith D. G., (1996), “Transition Effects on Secondary Flows in a Turbine Cascade”, ASME 96-GT-100. [41] Robinson C. J., Northall J. D., McFarlane C., (1989), “Measurement and Calculation of Three-Dimensional Flow in Axial Compressor Stators, with and without End Bends”, ASME 89-GT-6. [42] Lapworth B. L., (1993), “Three-Dimensional Mesh Embedding for the Navier-Stokes Equations Using Upwind Control Volumes”, Int. J. Numerical Methods in Fluids, 17, pp. 195-220. [43] Bindon J. P., (1980), “Exit Plane and Suction Surface Flows in an Annular Turbine Cascade with a Skewed Inlet Boundary Layer”, Int. Journal Heat & Fluid Flow, Vol. 2, No. 2. [44] Harvey N. W., Rose M. G., Coupland J., Jones T. V., (1998), “Measurement and Calculation of Nozzle Guide Vane End Wall Heat Transfer”, ASME paper 98-GT-66. [45] White F. M., (1991), “Viscous Fluid Flow”, McGraw-Hill. [46] Spalding D. B., Patankar S. V., (1967), “Heat and Mass Transfer in Boundary Layers”, Morgan-Grampian. [47] Rose M. G., Harvey N. W., Seaman P., Newman D. A., McManus, D., (2001), “Improving the Efficiency of the Trent 500 HP Turbine using Non-axisymmetric End Walls: Part II Experimental validation”, ASME 2001-GT-505. [48] Moore J., Moore J. G., (1991), “A Computational Study of Tip Leakage Flow and Losses in a Linear Turbine Cascade”, AGARD Conference Proceedings No. 510 on CFD Techniques for Propulsion Applications, San Antonio, Texas. [49] Taylor M. D., De Pablos T., Rose M. G., (1998), “Time Resolved HWA Measurements of the OTL Flow Field from a Shrouded Turbine HP Rotor Blade”, ASME 98-GT-564. [50] Matsunuma T., (2005), “Effects of Reynolds Number and FreeStream Turbulence on Tip Clearance Flow, ASME GT2005-68009. [51] Yoon J., Martinez-Botas R. , (2005), “Film Cooling Performance in a Simulated Blade Tip Geometry”, ASME GT2005-68863. [52] Nirmalan N. V., Bailey J. C., (2005), “Experimental Investigation of Aerodynamic Losses of Different Shapes of a Shrouded Blade Tip Section”, ASME GT2005-68903.

12
