Você está na página 1de 9

Turbine Blade Tip Heat Transfer in

Low Speed and High Speed Flows


Andrew P. S. Wheeler1
e-mail: a.wheeler@qmul.ac.uk
School of Engineering and Materials Science,
Queen Mary, University of London,
Mile End Road,
London, E1 4NS, UK

Nicholas R. Atkins
Whittle Laboratory,
University of Cambridge,
Cambridge, CB3 0DY, UK

Li He
Department of Engineering Science,
University of Oxford,
Parks Road, Oxford OX1 3PJ, UK

In this paper, high and low speed tip flows are investigated for a high-pressure turbine
blade. Previous experimental data are used to validate a computational fluid dynamics
(CFD) code, which is then used to study the tip heat transfer in high and low speed
cascades. The results show that at engine representative Mach numbers, the tip flow is
predominantly transonic. Thus, compared with the low speed tip flow, the heat transfer is
affected by reductions in both the heat-transfer coefficient and the recovery temperature.
The high Mach numbers in the tip region M 1.5 lead to large local variations in
recovery temperature. Significant changes in the heat-transfer coefficient are also observed. These are due to changes in the structure of the tip flow at high speed. At high
speeds, the pressure side corner separation bubble reattachment occurs through supersonic acceleration, which halves the length of the bubble when the tip-gap exit Mach
number is increased from 0.1 to 1.0. In addition, shock/boundary-layer interactions
within the tip gap lead to large changes in the tip boundary-layer thickness. These effects
give rise to significant differences in the heat-transfer coefficient within the tip region
compared with the low speed tip flow. Compared with the low speed tip flow, the high
speed tip flow is much less dominated by turbulent dissipation and is thus less sensitive to
the choice of turbulence model. These results clearly demonstrate that blade tip heat
transfer is a strong function of Mach number, an important implication when considering
the use of low speed experimental testing and associated CFD validation in engine blade
tip design. DOI: 10.1115/1.4002424

Introduction

The design of high-pressure HP turbine blade tips has a significant impact on both the aerodynamic loss and the heat load of
unshrouded turbine blades. For these reasons, there has been a
significant amount of research into the thermofluid dynamics of
turbine blade tip flows. Of particular interest has been the accurate
experimental measurement and computational fluid dynamics
CFD prediction of blade tip Nusselt numbers since these values
can be used directly in the turbine design process provided that
the flow physics are simulated correctly. In subsonic conditions,
this requires matching the engine-scale Reynolds numbers. For
transonic conditions such as when blade-exit Mach numbers exceed 0.9 as what commonly occurs in HP turbines, the effects of
compressibility become significant, particularly within the tip gap
itself.
A significant amount of the studies into the tip heat transfer
have involved experimental and computational investigations in
low speed or subsonic turbine cascades 1. Recently, low speed
work has shown that significant differences in the predicted heat
transfer can be obtained depending on the choice of turbulence
model 2,3.
Far fewer experimental studies have been performed at engine
Mach numbers due to the complexities of testing in high speed
flows. The water table experiments of Moore et al. 4 and Moore
and Elward 5 showed that the formation of the vena-contracta at
the entrance to the tip gap was able to accelerate the flow to
supersonic conditions when the gap exit Mach number exceeded
0.8. Furthermore, the experimental and numerical tests of Chen et
al. 6 on a two-dimensional tip gap in transonic flow showed that
for an exit Mach number of 1.0, the peak Mach number in the gap
was 1.4. They also showed that when the tip flow speed was
1
Corresponding author.
Contributed by the International Gas Turbine Institute IGTI of ASME for publication in the JOURNAL OF TURBOMACHINERY. Manuscript received February 11, 2010;
final manuscript revised February 22, 2010; published online April 26, 2011. Editor:
David Wisler.

Journal of Turbomachinery

increased from subsonic to supersonic flow, the separation bubble


at the inlet to the gap reduced significantly in length from 1.5 g to
less than 1.0 g.
Early tip heat-transfer measurements in high speed rotating facilities were performed by Metzger et al. 7 and Dunn and Kim
8. These tests were performed at stage pressure ratios of 2.3 and
1.7, respectively, and the blade-exit Mach numbers are estimated
to have been in the region of 0.70.8 9. Later heat-transfer measurements on a flat tip were presented by Molter et al. 10 at a
higher stage pressure ratio of around 4. The heat transfer was
measured at four locations on the tip and was compared with
CFD. In this case, the blade flow field was transonic and showed
significant differences from previous subsonic HP turbine results,
such as those of Ameri and Bunker 11.
Thorpe et al. 12 took high resolution measurements of the tip
heat transfer along the camber line of a HP turbine blade in a
transonic turbine facility using an array of 17 thin film heattransfer gauges. In this case, the stage pressure ratio was 3.12, and
the blade-exit Mach number was around 1.0. These results provide
a detailed data set, which will be used in this paper for validation
purposes. Other tip heat-transfer measurements in transonic turbine stages are presented by Didier et al. 13 and Chana and
Jones 14.
Typically, single stage HP turbine engine blade-exit Mach numbers are in the region of M = 0.9 1.1, with corresponding peak
suction surface values of M = 1.1 1.3, and thus the peak tip Mach
numbers will exceed this. It will be shown in this paper that even
high subsonic exit blading M 0.8, which does not reach sonic
conditions on the blade surface, may have regions of transonic
flow within the tip gap itself. Despite this, little has been published on the effect of compressibility on the tip heat transfer.
Therefore, this paper aims to determine the key differences between high and low speed tip-gap flows and what effects these
differences have on the tip heat transfer for a flat blade tip.
In this paper, high speed and low speed tip flows are compared.
Fully turbulent CFD predictions are initially validated against experimental data, which show good agreement with the measurements in regions of turbulent flow over the aft portion of the tip

Copyright 2011 by ASME

OCTOBER 2011, Vol. 133 / 041025-1

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 12/06/2014 Terms of Use: http://asme.org/terms

Fig. 2

Quasi-three-dimensional tip model mesh

results to the k- model. The HYDRA code with the Spalart


Allmaras model has been used previously to study tip leakage
flows from high speed rig tests by Willer et al. 15, who studied
the tip loss, and by Atkins et al. 16, who studied unsteady tip
heat transfer. Both investigations found good agreement with experimental results.
In order to test the ability of the CFD code to predict blade heat
transfer, a series of calculations was compared with previous experimental data, as discussed next.
Fig. 1 1 and 1/2 stage mesh top and typical cascade blade
passage mesh bottom

and blade surface. The fully turbulent CFD overpredicts the heat
transfer in regions of laminar and transitional flows, which occur
near the leading edge of the tip and blade surface. Three turbine
blade profiles are then tested numerically. The three blades are
designed to have the same loading distribution and Reynolds
number but different exit Mach numbers 0.98, 0.67, and 0.1.
Both the standard k- and SpalartAllmaras turbulence models are
used. Comparisons of the predicted tip heat transfer at the three
Mach numbers show a significant drop in tip heat load as the
Mach number is increased. This drop is driven by the transonic
nature of the tip flow. The results indicate that the high speed tip
flow is much less dominated by turbulent dissipation compared
with the low speed flow. Thus, an improved picture for turbine tip
flows is presented.

Simulating a Transonic Blade Tip Flow

The 3D computational meshes used in this investigation were


created using the PADRAM Rolls-Royce in-house meshing program. In order to compare the numerical scheme with experimental data, a 1 and 1/2 stage mesh was used, as shown in the upper
part of Fig. 1. The actual blade tip corner radius was modeled, and
the calculation was run using measured boundary conditions. The
high-pressure blade HPB mesh featured 2.5 106 cells, with 52
grid lines in the tip gap and y+ values at the first cell from the
surface of approximately 30 on the blade tip and casing with wall
functions. Typical grid sizes for the single row study shown in
the lower part of Fig. 1 were over 3 106 cells per blade passage,
again with y+ values of approximately 2030 when wall functions
were implemented. A quasi-three-dimensional tip model was also
tested numerically, and the mesh for this was created using a code
written in MATLAB; this mesh is shown in Fig. 2.
The Rolls-Royce in-house CFD code, HYDRA, was used to solve
the flow. This program is able to solve the steady and unsteady
Reynolds-averaged NavierStokes equations in three dimensions.
For the current investigation, steady calculations were performed.
The viscous surfaces were isothermal in all cases.
The CFD calculations used fully turbulent implementations of
both the SpalartAllmaras and the k- turbulence models. As
such, no attempt was made to model transitional flow. Recent
work has shown that the standard k- turbulence model gives the
best match with low speed experimental data 3; however, this
work also showed that the k- model gave qualitatively similar

2.1 Transonic Linear Cascade Blade Surface Heat Flux


Comparison. First, a comparison with the experiments of Doorly
17 and Nicholson 18 was performed their data are also available in Ref. 19. Both Doorly 17 and Nicholson 18 tested the
midspan flow field of a transonic turbine cascade see Table 1.
The experimentally measured isentropic Mach number distribution of Nicholson 18 is shown in Fig. 3. Several CFD calculations are also shown; these have different turbulence modeling
and mesh densities. Two SpalartAllmaras calculations are shown
in the figure, one with wall functions, where y + 20 denoted as
SAWF, and the other where the laminar sublayer is fully resolved
y + 1 denoted as SA. A k- calculation with wall functions,
again y + 20 is also shown denoted as KEWF. The predictions
agree well with the experiment. The SA calculation with fully
Table 1 Details of RT27a cascade
Mexit
Reexit = VexitCx / L
in
out
Tw / Tg
Pitch-to-chord ratio

0.96
1.55 106
42.75 deg
68 deg
1.5
0.86

Fig. 3 Isentropic Mach number distributions at midspan, comparison of CFD and experiment

041025-2 / Vol. 133, OCTOBER 2011

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 12/06/2014 Terms of Use: http://asme.org/terms

Transactions of the ASME

Fig. 4 Nusselt number distributions at midspan, comparison


of CFD and experiment

resolved boundary layers gives the closest agreement with experiment in the diffusing part of the blade suction surface. Over the
remainder of the blade surface, there is very little difference between the CFD calculations.
The experimentally measured Nusselt number distributions of
Doorly 17 and Nicholson 18 are shown in Fig. 4. The experiments were performed at high turbulence intensity Tu= 5% and
low turbulence intensity Tu 0.2%. The turbulence intensity has
a significant effect on the Nusselt number on the suction surface
because at low turbulence levels the boundary layer is laminar for
approximately 40% of the suction surface length. The higher turbulence level significantly increases the suction surface Nusselt
number by promoting transition nearer the leading edge. For this
reason, the CFD calculations with fully turbulent boundary layers
SAWF, SA, and KEWF differ significantly with the experimental data over the early suction surface. There is a much better
agreement over the late suction surface after 50% surface length,
where transition is most likely to be complete. A laminar CFD
calculation is also shown. It is important to note that it was not
possible to fully converge this calculation; however, it is interesting to note that this calculation agrees well with the low turbulence level results of Doorly 17 on the early suction surface. On
the pressure surface, the effect of turbulence intensity is less pronounced. The fully turbulent CFD calculations tend to lie in the
range of the experimental data.
The results show that in regions of fully turbulent flow, both the
SpalartAllmaras and the k- turbulence model give a reasonable
agreement with experiment. In regions of laminar or transitional
flow, the fully turbulent CFD clearly overpredicts the heat transfer
significantly as might be expected.
2.2 Transonic Turbine Stage Blade Tip Heat Flux
Comparison. In order to test the veracity of the computational
code for tip heat transfer on a transonic blade, stage calculations
of the Oxford Rotor Facility ORF geometry were performed and
compared with the experimental data presented by Thorpe et al.
12. The working section of the ORF contains a 0.55 m diameter
shroudless high-pressure turbine stage and allows the simulation
of engine representative Mach and Reynolds numbers as well as
the appropriate gas-to-wall temperature ratio. There are 36 inlet
nozzle guide vanes upstream of the rotor disk, which has 60
blades and rotates at 8910 rpm. The rotor exit Mach number is
0.98, and the exit Reynolds number based on axial chord is 1.55
106. The static tip clearance is 2.5% of the blade height. The
Journal of Turbomachinery

Fig. 5 Comparison of experimental and predicted blade tip


heat flux with the mapping between experimental gauge locations and the CFD mesh

stage inlet total temperature to blade surface temperature is 1.3.


The midspan blade profile is the same as that tested by Doorly
17 and Nicholson 18 discussed previously.
Steady CFD calculations were performed with a mixing-plane
treatment between blade rows. Thus, the circumferential averaged
potential field effects of the upstream and downstream vanes and
the effect of relative casing motion were simulated in these calculations. The tip edge radius was modeled in the CFD geometry
in order to match the experimental tip geometry, as described by
Atkins et al. 16.
Figure 5 shows a comparison of the CFD SAWF with the
measured tip heat flux at the gauge locations of Thorpe et al. 12,
which are shown in the lower part of the figure. The experimental
data show a falling trend of heat load from gauge locations 13
620% axial chord, then rising until a maximum at gauge 6
50% axial chord, and then falling from gauges 713 5090%
axial chord. At the location of gauge 1, the turbulent calculation
predicts a heat flux to within a few percent of the experimental
data. However, moving rearward along the camber line, the CFD
overpredicts the heat flux. From gauge location 6 onward, the
CFD prediction falls and follows both the trend and absolute level
of the experiments very well, to within a few percent at locations
7 and 8, and to approximately 10% of the local measured level
through to gauge 12 6% of the maximum measured value.
In short, the CFD gives a good match with experiment over the
mid and aft portions of the tip, while it overpredicts the heat
transfer near the leading edge. In order to understand why, it is
OCTOBER 2011, Vol. 133 / 041025-3

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 12/06/2014 Terms of Use: http://asme.org/terms

Fig. 7 Isentropic Mach number distributions for blades A, B,


and C

Fig. 6 Predicted heat flux and Mach number contours and


streamlines SpalartAllmaras+ WF

useful to investigate further the differences between the tip flow


near the leading edge and the flow over the aft portion of the tip,
as described next.
2.2.1 Near the Leading Edge. The flow that passes over the
first four gauge locations comes from near the leading edge see
Fig. 6. Unlike the majority of the tip flow, this boundary-layer
flow does not separate as it enters the tip gap but rather experiences a long region of acceleration. Furthermore, the streamwise
Reynolds numbers at gauges 14, based on the local flow conditions, are below the critical value for transitional flow on a flat
plate with a zero pressure gradient ReL 3 105. Considering
that the leading edge of the tip region is subject to the ingestion of
a complex sheared casing boundary layer, as well as unsteady
wake passing events, a transitional boundary layer is to be expected in this region these unsteady effects are not simulated in
the CFD with a mixing-plane treatment. To illustrate this, a laminar solution has been obtained using the fully converged turbulent prediction as an initial condition to estimate the heat flux,
which would result from a fully laminar boundary layer in this
region. It should be noted that the solution is not fully converged
and provides an estimate only. The laminar values are plotted in
Fig. 5 and show excellent qualitative agreement with the trend of
the experimental data at the first four gauge locations, albeit with
an absolute value of approximately 50% of the measured values.
This is a strong indication that the tip boundary layer close to the
leading edge is transitional rather than fully turbulent, as was also
observed on the blade suction surface.
2.2.2 Mid Chord and Aft. Figure 6 shows the computationally
predicted blade tip surface heat flux and isentropic Mach number
contours. Mach number contours are also shown on planes parallel to the tip flow over the mid and aft portions of the tip. In this
region, the tip surface boundary layer separates from the pressure
side edge of the tip, causing a region of low heat flux. This is
followed by a reattachment indicated by a white dashed line,
which causes a rise in the heat transfer.
The flow can be divided into a region where this reattachment
occurs in subsonic flow region Y, over the mid portion of the tip,
and the region where the reattachment occurs in supersonic flow
separation, over the aft portion of the tip region X.

In region Y, where there is a subsonic reattachment, there is a


high level of heat flux downstream of the reattachment. The peak
heat flux is roughly 50% higher than the tip area-average heat
flux. In addition, the separation length is large.
In region X, where there is a supersonic reattachment, the separation length reduces significantly compared with the region of
subsonic reattachment. The Mach number contours in this region
show that the flow accelerates over the separation bubble to a
Mach number of over 1.5. The flow remains supersonic downstream of the bubble and then decelerates across a normal shock at
the exit of the tip gap. The level of heat flux downstream of the
bubble is significantly lower than over the mid portion of the tip.
The reasons for this will be shown later.

High Speed Versus Low Speed Tip Flow

In order to compare the transonic tip flow with an analogous


low speed tip flow, it was necessary to design low speed blades
that matched the chord-to-gap ratio, the blade loading, and the
Reynolds number of the transonic blade. The tip section 95%
span of the Oxford rotor blade was used as the baseline high
speed geometry blade A, which has an exit Mach number of
Mexit = 0.98. The predicted Mach number distribution for this can
be seen in Fig. 7. The peak Mach number is 1.15, and the flow is
supersonic between approximately 25% and 75% of the suction
surface length.
Figure 7 also shows the isentropic Mach number distribution of
two other blade profiles: blade B with Mexit = 0.67 and blade C
with Mexit = 0.1 typical low speed blade, which were designed to
match the blade loading and Reynolds number of the transonic
blade A. The blade profiles can be seen in Fig. 8. In general,
reducing the Mach number requires an increase in blade thickness.
Therefore, a further constraint on the profile design was made so

Fig. 8 Predicted tip isentropic Mach number for blades A, B,


and C normalized by exit Mach number

041025-4 / Vol. 133, OCTOBER 2011

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 12/06/2014 Terms of Use: http://asme.org/terms

Transactions of the ASME

Fig. 10 Predicted variation in area-averaged tip heat load


Fig. 9 Predicted tip heat flux for blades A, B, and C Tg / Tw
= 1.5

that the trailing-edge thickness-to-chord ratio was held constant.


This ensured that the trailing-edge loss would not increase significantly. Figure 7 shows that the suction surface distributions are
matched well. There is a poorer match on the pressure surface,
which was a consequence of maintaining a constant trailing-edge
thickness-to-chord ratio while matching the suction surface velocity distribution.
3.1 Effect of Mach Number on Tip Heat Flux. In order to
study the effect of Mach number on the tip heat transfer, threedimensional CFD calculations of the blade profiles with a tip gapto-axial-chord ratio of 5% were performed. This matched the Oxford rotor blade described previously. The wall temperature was
specified in these calculations so that Tg / Tw = 1.5. Calculations
were performed with both SpalartAllmaras and k- turbulence
models with wall functions. This helps to eliminate effects due to
the choice of turbulence model since this has been shown to be
important at low speed conditions in previous studies 2,3. For
these calculations, both the blade and the casing wall were
stationary.
The predicted tip surface isentropic Mach number contours for
blades A, B, and C are shown in Fig. 8. The black lines indicate
contours of M = 1.0. The figure shows that much of the blade A tip
is supersonic with the peak Mach number exceeding 1.5. This
occurs over the separation bubble near the pressure surface edge
of the tip. For blade B, there is a small supersonic patch, but the
majority of the tip is subsonic. For blade C, the flow is predominantly incompressible with the Mach number never exceeding 0.2.
Figure 8 shows that the tip surface pressure distributions are
similar for the three blade profiles. The main differences in the
pressure distributions occur near the pressure side edge and are
due to changes in the separation bubble geometry. There is very
little difference between the pressure distributions for the two turbulence models on the high speed blade Mexit = 0.98. However,
for the low speed blade Mexit = 0.1, the pressure distributions
differ more significantly, and again this is mainly due to differences in the prediction of the pressure side edge separation
bubble.
Figure 9 shows the predicted tip surface heat flux contours for
the three blade profiles. For the front portion of the tip, between
0% and 10% axial chords, the heat flux remains relatively unchanged for the three blades. However, over the remainder of the
tip, there is a marked increase in the heat flux as the Mach number
is reduced. The overall changes in tip heat load can be seen in Fig.
10. Reducing the Mach number from 0.98 to 0.1 increases the heat
load by 60%.
It is interesting to note that the difference in predicted heat load
between the KEWF calculation and the SAWF, as a proportion of
their mean, does not rise at higher Mach numbers. In fact, there is
a drop in this difference from 24% at M = 0.1 to 20% at M
= 0.98 see Fig. 10. This indicates that the high speed tip flow is
Journal of Turbomachinery

less sensitive to the choice of turbulence model compared with the


low speed tip flow, as was also observed in the tip Mach number
distributions shown in Fig. 8.
In order to understand further the change in heat load with
Mach number, it is useful to consider that the surface heat flux is
given by
q = hTaw Tw

At low Mach numbers, the driving temperature difference is


constant over the tip surface. This is because the adiabatic wall
temperature is given by the gas total temperature, which does not
vary significantly over the tip surface. However, for a transonic
tip, the adiabatic wall temperature is given by the local recovery
temperature, which is a function of local Mach number and recovery factor. Therefore, both the driving temperature difference and
the heat-transfer coefficient must be considered.
The driver temperature and heat-transfer coefficient are both
shown in Fig. 11 for blade A. These have been calculated from
CFD calculations at two different wall temperatures: at the baseline wall temperature, Tw2 = 2 / 3Tg and the increased wall temperature of Tw1 = Tg. Thus, the heat-transfer coefficient and adiabatic wall temperature were given by h = q1 q2 / Tw2 Tw1 and
Taw = q1 / q2Tw2 Tw1 / q1 / q2 1.
Figure 11 shows significant spatial variations in the adiabatic
wall temperature, which give rise to a maximum drop in the driver
temperature of over 25% in the region of the pressure side edge of
the tip when Tg / Tw = 1.5. This occurs in the region of peak tip
Mach number over the separation bubble see Fig. 8. This par-

Fig. 11 Predicted adiabatic wall temperature and Nusselt number for blade A Mexit = 0.98, Tg / Tw = 1.5

OCTOBER 2011, Vol. 133 / 041025-5

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 12/06/2014 Terms of Use: http://asme.org/terms

Fig. 12 Predicted turbulent to laminar viscosity ratio T / L


for blades A and C, cut at 4% tip gap above tip surface

tially explains why the lower speed blades blades B and C exhibit high area-average heat loads, although this contributes a relatively small amount.
The most significant cause of the increased heat flux at low
speeds is due to an increase in heat-transfer coefficient. As shown
in Fig. 10, the change in average recovery temperature h
Tw Trec accounts for only 7% of the increased areaaveraged heat load as Mexit is reduced from 0.98 to 0.1, while the
change in heat-transfer coefficient h Tw Trec contributes
93% to this change when Tg / Tw = 1.5. It is important to note that
the effect of changes in recovery temperature on heat load is dependent on Tg / Tw. For smaller Tg / Tw, changes in recovery temperature will affect the heat load more significantly.
The main reason for the increased heat-transfer coefficient at
low speeds can be seen in Fig. 12, which shows the turbulent
viscosity ratio on a cut plane at 4% of the tip gap above the tip
surface. The figure shows that for blade C, the turbulent viscosity
in the tip gap is roughly twice that of blade A between 30% and
100% axial chords. This increased turbulence will tend to increase
both the momentum and heat transfer within the tip, thus increasing the tip heat-transfer coefficients. Figure 12 also shows that the
KEWF calculations give a higher turbulent viscosity compared
with the SAWF calculations for both blades A and C note that the
color scale for KEWF is 2.7 times larger than SAWF. This causes
the increased heat flux in the KEWF calculations observed earlier
in Fig. 9.
The generation of turbulence in the tip gap is highly dependent
on the nature of the bubble reattachment. The structure of the
separation bubble for blades A and C can be observed in Fig. 13.
This figure shows the Mach number contours on a cut plane parallel to the flow indicated in the figure. It can be seen in Fig. 13
that the reattachment on blade A Mexit = 0.98 occurs predominantly in a supersonic flow, which must also locally accelerate,
while in blade C Mexit = 0.1 the reattachment occurs in a region
of pressure recovery brought about by turbulent mixing.
Figure 13 shows the predicted heat transfer in this region from
the SAWF calculations. For the high speed blade A Mexit = 0.98,
the rapid thinning of the accelerating boundary-layer downstream
of the bubble leads to an increased heat-transfer coefficient. This
is terminated by a rapid thickening of the boundary layer due to
the interaction of the normal shock, which causes a drop in the
heat-transfer coefficient. Thus, the strong pressure gradients in the
high speed tip flow give rise to significant variations in the local

Fig. 13 Mach number contours on a cut plane through the tip


gap and surface heat-transfer contours for blades A and C
SAWF

heat transfer, which are not observed in the low speed tip flow.
Figure 13 also highlights that the tip width-to-gap ratio w/g of
the low speed Blade C is greater than Blade A. The variations of
w/g for both blades A and C are shown in Fig. 14. The variations
of w/g for both blades A and C have been computed based on the
tip streamlines, rather than the actual blade width, since it is the
streamwise tip width which is of most importance to the tip flow.
Blade C has a 2030% higher width-to-gap ratio between 030%
axial chord, and a 50100% higher width-to-gap ratio between
50% and 70% axial chords. Therefore, the effects of the difference
in width-to-gap ratio are likely to be most significant over the
latter portion of the tip.
The following section uses a quasi-three-dimensional model of
the tip flow to investigate the effects of both width-to-gap ratio
and Mach number on the tip flow.

Quasi-Three-Dimensional Tip Flow

In order to investigate the effects of Mach number and widthto-gap ratio on the bubble reattachment, quasi-three-dimensional
CFD calculations of a gap flow were performed. The computational domain and mesh are shown in Fig. 2. A baseline width-togap ratio of w / g = 5.0 was chosen, and the tip Reynolds number
based on the tip width and the exit velocity was Re= 4 105.
These values are typical of the Oxford rotor blade described
above. A SpalartAllmaras turbulence model was used without
wall functions y + 1.
The effect of the width-to-gap ratio on the quasi-threedimensional gap flow can be seen in Fig. 15. The figure shows
Mach number contours of the predicted gap flow from calculations at five different width-to-gap ratios, with a constant isentro-

Fig. 14 Width-to-gap ratio against axial chord blades A and C

041025-6 / Vol. 133, OCTOBER 2011

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 12/06/2014 Terms of Use: http://asme.org/terms

Transactions of the ASME

Fig. 16 Effect of pressure ratio on quasi-three-dimensional tip


flow w / g = 5

Fig. 15 Effect of width-to-gap


dimensional tip flow Mexit = 0.9

ratio

on

quasi-three-

pic exit Mach number of 0.9. For ease of viewing, the figures are
plotted so that they appear to have the same aspect ratio; however,
the true aspect ratios are indicated below the contour plots. At
w / g = 1.75, a normal shock exists downstream of the bubble reattachment. Doubling the tip width w / g = 3.5, causes the shock to
move upstream as a proportion of the tip width. In addition, the
flow reaches a significantly higher peak Mach number of over 1.5.
Increasing the tip width further continues to move the normal
shock upstream.
Between w / g = 3.5 and 14, the qualitative tip flow structure
does not vary significantly, and thus the difference in width-to-gap
ratio between the high speed blade A and low speed blade C is
unlikely to have a significant effect on the tip heat transfer.
At w / g = 28, the tip separation is confined to roughly 10% of
the tip surface, and the remainder of the gap flow is akin to a fully
developed pipe flow, where the tip and casing boundary layers
extend to fill the gap. When scaled to the widest part of the Oxford rotor RT27a blade, the w / g = 28 case corresponds to a gap
to span ratio of approximately 0.7%.
The effect of compressibility on the quasi-three-dimensional
gap flow can be seen in Fig. 16. This figure shows contours of the
predicted Mach number distribution within the gap at a series of
isentropic exit Mach numbers from 0.1 to 1.0 w / g = 5.0. Between Mexit = 0.1 and 0.6, there is a gradual reduction in bubble
height but very little change in the qualitative nature of the flow.
This reduction in bubble height is coupled with an increase in the
acceleration over the initial part of the separated shear layer. At
Mexit = 0.8, the peak Mach number reaches 1.2, and a normal
shock forms downstream of the bubble. At Mexit = 1.0, the peak
Mach number exceeds 1.5, and there is a dramatic reduction in the
bubble length.
Figure 17 plots the bubble height and length with Mach number. The bubble height reduces by a factor of 2 when the tip-gap
exit Mach number is increased from 0.1 to 1.0. In contrast, the
bubble length remains roughly constant as the Mach number is
increased until there is a supersonic reattachment of the bubble,
which causes the bubble length to reduce suddenly by a factor of
2. A similar shortening of the bubble was observed by Chen et al.
Journal of Turbomachinery

6, who saw a reduction in bubble length from 1.5 g to less than


1 g when the flow speed was increased from subsonic to supersonic.
Figure 16 also shows that for Mexit = 1.0, oblique shocks form
downstream of the bubble, which reflect from the tip and casing,
causing a rapid thickening of both boundary layers. Between the
shock reflections, expansion waves cause a significant thinning of
the boundary layers. The supersonic region is terminated by a
normal shock, which causes a further thickening of the boundary
layer. This shock structure is very similar to that observed by
Moore et al. 4 and Moore and Elward 5.
The variation in turbulent viscosity in the tip gap as the exit
Mach number is increased is plotted in Fig. 18. From Mexit = 0.1 to
0.8, there is a gradual reduction in the production of turbulence.
This corresponds to the increasing acceleration of the flow over
the separated shear layer near the inlet to the gap, simply due to
the effect of compressibility. However, at Mexit = 1.0, there is a
sudden drop in the turbulence production when the bubble reattachment occurs in supersonic accelerating flow. Thus, the bubble
reattachment is brought about by this accelerating pressure gradient, which causes a significant foreshortening of the bubble and in
turn produces much less turbulent mixing.

Discussion

The high speed tip flow structure is shown in Fig. 19. This
figure has been constructed from an inviscid duct calculation using the predicted bubble displacement thickness from a gap exit
Mach number of Mexit = 1.0 and w / g = 5 see Fig. 16. As the flow

Fig. 17 Separation bubble height and length from the quasithree-dimensional tip flow w / g = 5

OCTOBER 2011, Vol. 133 / 041025-7

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 12/06/2014 Terms of Use: http://asme.org/terms

1
dV dA
=
V
A M 2 1

Fig. 18 Effect of pressure ratio on turbulent viscosity w / g


= 5

enters the tip gap, the flow separates from the pressure side corner.
The separation accelerates the flow to supersonic like a
converging-diverging nozzle; thus, as the separation begins to reattach, the change in area forces the mainstream flow to accelerate. This, in turn, aids the reattachment process, and the separation
is closed much earlier than would occur through turbulent mixing
alone. This extra mechanism for supersonic reattachment cannot
occur in subsonic flow.
Compression waves form due to the curvature of the reattaching boundary layer see Fig. 19. These coalesce into an oblique
shock, which reflects from the tip and casing. Between these
shock reflections, the flow rapidly accelerates. These rapid accelerations and decelerations cause large local thinning and thickening of the tip boundary layer. This supersonic flow is terminated
by a normal shock, which causes a further thickening of the tip
boundary layer.
This basic structure occurs over a significant portion of the tip
on both the high speed cascade blade see Fig. 13 and on the
rotating turbine blade see Fig. 6. In essence, this high speed tip
flow is dominated by strong local pressure gradients, while the
low speed tip flow is governed by the overall pressure gradient
across the tip width. This can be understood by considering the
combined energy and continuity equations as

Fig. 19 Transonic tip flow Mach contours

Thus, for the same change in boundary-layer displacement thickness, the mainstream flow will experience a greater acceleration at
higher Mach numbers. This acceleration will then reduce the rate
of boundary-layer growth and thus moderate the increase in acceleration. This effect was observed in Fig. 16. The figure showed
that as the exit Mach number was increased, from Mexit = 0.1 to
0.8, the acceleration over the separation bubble increased, which
in turn gradually reduced the height of the bubble.
In the high speed tip flow, Eq. 2 also shows that downstream
of the bubble throat, the coupling between the inviscid and viscous flow fields is contrary to that in the low speed tip flow. This
is because while in subsonic flow, the reduction in blockage due to
the boundary-layer reattachment causes a streamwise pressure
rise, and in supersonic flow the reduction in blockage causes a
pressure drop, which further aids the boundary-layer reattachment. At the same time, the increased acceleration of the separated
shear layer and the reduction of bubble length lead to a reduction
in turbulence production for the high speed tip flow. Therefore,
the high speed tip flow is much more dominated by strong local
pressure gradients than turbulent dissipation or the overall pressure difference across the tip.
Considering the 3D blade tip, within the supersonic regions,
this acceleration and reduction of turbulence aid the accurate prediction of the heat flux. As the flow field is predominantly inviscid, the physics are well captured by the relatively simple one
equation model with a wall function type approach; hence, the
heat-transfer coefficient is well predicted by the associated Reynolds analogy. Within the subsonic regions on the blade tip, the
heat-transfer coefficient is governed by a complex combination of
subsonic turbulent reattachment and highly skewed boundary layers, for which the basic flow structure is dependent on the turbulence modeling.
It is also worth noting that as blade-exit Mach numbers increase
above 0.8, the results indicate that the blade tip will start to choke
while the blade mass flow will continue to increase. Consequently,
we would expect the tip leakage mass flow rate as a proportion of
the mainstream flow to reduce at exit Mach numbers greater than
0.8, which will tend to reduce losses.

Conclusion

In this paper, low speed and high speed tip flows were compared for a high-pressure turbine blade at the same Reynolds number. At high speed Mexit = 0.98 the majority of the tip flow was
transonic, while at low speed Mexit = 0.1 the flow was essentially
incompressible. The results showed a 60% drop in tip heat load
for the high speed blade. This was due to a combination of reduced driver temperature and reduced heat-transfer coefficient.
High Mach numbers over the pressure side corner separation
bubble led to a 25% reduction in the local driver temperature
difference when Tg / Tw = 1.5. This effect will be amplified at
lower gas-to-wall temperature ratios. However, the most significant cause of the drop in heat transfer was due to a reduction in
heat-transfer coefficient. This occurred in the high speed flow due
to a reduction in turbulent mixing. The turbulence was found to
reduce as a result of an increase in the acceleration of the mainstream flow over the separation bubble combined with a halving
of the bubble length brought about by the supersonic acceleration
of the reattaching boundary layer.
The results indicate that for turbine tip flows at engine Mach
numbers, the tip flow structure is significantly different from that
suggested by low speed cascade testing. At high speeds, the tip
heat flux is both quantitatively and qualitatively different compared with the incompressible tip flow. Fundamentally, the high
speed tip flow is much more dominated by large local pressure
gradients and less dependent on turbulent dissipation than the low
speed tip flow. The results also indicate that the accuracy of CFD

041025-8 / Vol. 133, OCTOBER 2011

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 12/06/2014 Terms of Use: http://asme.org/terms

Transactions of the ASME

predictions for high speed tip heat transfer is likely to be less


dependent on the choice of turbulence model than that for low
speed tip flows.

Acknowledgment
The authors would like to thank Rolls-Royce plc for their support and provision of the CFD and meshing code.

T o, T g
Tu
Tw
V
w
y+

L
T

5
6

Nomenclature
A
Cx
g
h
k
kaw
L
M, Misen
Nu
q
Re
s
S0
Taw

area
axial chord
tip-gap height
heat-transfer coefficient
conductivity at wall temperature
conductivity at adiabatic wall temperature
separation bubble length
Mach number, isentropic Mach number
Nusselt number= hCx / k
surface heat flux
Reynolds number
surface distance from leading edge
total surface length
adiabatic wall temperature, recovery
temperature
gas stagnation temperature
turbulence intensity
wall temperature
velocity
tip width parallel to tip flow
nondimensional wall distance
swirl angle
separation bubble displacement thickness
density
laminar viscosity
turbulent viscosity

7
8
9
10
11
12
13
14
15
16
17
18

References
1 Bunker, R. S., 2004, Blade Tip Heat Transfer and Cooling Techniques, VKI
Lecture Series, Paper No. 2004-02.
2 Krishnababu, S. K., Newton, P. J., Dawes, W. N., Lock, G. D., Hodson, H. P.,

Journal of Turbomachinery

19

Hannis, J., and Whitney, C., 2007, Aero-Thermal Investigations of Tip Leakage Flow in Axial Flow Turbines, Part 1Effect of Tip Geometry and Tip
Clearance Gap, ASME Paper No. GT2007-27954.
Tang, B., Palafox, P., Oldfield, M., Gillespie, D., and Cheong, B., 2008, Computational Modeling of Tip Heat Transfer to a Super-Scale Model of an Unshrouded Gas Turbine Blade, ASME Paper No. GT2008-51212.
Moore, J., Moore, J. G., Henry, G. S., and Chaudhry, U., 1989, Flow and
Heat Transfer in Turbine Tip Gaps, ASME J. Turbomach., 111, pp. 301309.
Moore, J., and Elward, K. M., 1993, Shock Formation in Overexpanded Tip
Leakage Flow, ASME J. Turbomach., 115, pp. 392399.
Chen, G., Dawes, W. N., and Hodson, H. P., 1993, A Numerical and Experimental Investigation of Turbine Tip Gap Flow, 29th Joint Propulsion Conference and Exhibit, AIAA Paper No. 93-2253.
Metzger, D. E., Dunn, M. G., and Hah, C., 1990, Turbine Tip and Shroud
Heat Transfer, ASME Paper No. 90-GT-333.
Dunn, M. G., and Kim, J., 1995, Turbine Blade Platform, Blade Tip and
Shroud Heat Transfer, Presented at the 12th International Symposium for
Airbreathing Engines.
Dunn, M. G., and Haldeman, C. W., 2000, Time-Averaged Heat Flux for a
Recessed Tip, and Platform of a Transonic Turbine Blade, ASME Paper No.
2000-GT-0197.
Molter, S. M., Dunn, M. G., Haldeman, C. W., Bergholz, R. F., and Vitt, P.,
2006, Heat-Flux Measurements and Predictions for the Blade Tip Region of a
High Pressure Turbine, ASME Paper No. GT2006-90048.
Ameri, A. A., and Bunker, R. S., 2000, Heat Transfer and Flow on the
First-Stage Blade Tip of a Power Generation Gas Turbine: Part 2Simulation
Results, ASME J. Turbomach., 122, pp. 272277.
Thorpe, S. J., Yoshino, S., Thomas, G. A., Ainsworth, R. W., and Harvey, N.
W., 2005, Blade-Tip Heat Transfer in a Transonic Turbine, Proc. Inst. Mech.
Eng., Part A, 219, pp. 421430.
Didier, F., Dnos, R., and Arts, T., 2002, Unsteady Rotor Heat Transfer in a
Transonic Stage, ASME J. Turbomach., 124, pp. 614622.
Chana, K. S., and Jones, T. V., 2003, An Investigation of Turbine Tip and
Shroud Heat Transfer, ASME J. Turbomach., 125, pp. 513520.
Willer, L., Haselbach, F., Newman, D. A., and Harvey, N. W., 2006, An
Investigation Into a Novel Turbine Rotor Winglet, Part 2: Numerical Simulation and Experimental Results, ASME Paper No. GT2006-90459.
Atkins, N. R., Thorpe, S. J., and Ainsworth, R. W., 2008, Unsteady Effects on
Transonic Turbine Blade-Tip Heat Transfer, ASME Paper No. GT200851177.
Doorly, J. E., 1985, The Development of Heat Transfer Measurement Techniques for Application to Rotating Turbine Blades, Ph.D. thesis, Department
of Engineering Science, University of Oxford, Oxford, UK.
Nicholson, J. H., 1981, Experimental and Theoretical Studies of the Aerodynamic and Thermal Performance of Modern Gas Turbine Blades, Ph.D. thesis, Department of Engineering Science, University of Oxford, Oxford, UK.
Moss, R. W., Ainsworth, R. W., and Garside, T., 1998, Effects of Rotation on
Blade Surface Heat Transfer: An Experimental Investigation, ASME J. Turbomach., 120, pp. 530540.

OCTOBER 2011, Vol. 133 / 041025-9

Downloaded From: http://turbomachinery.asmedigitalcollection.asme.org/ on 12/06/2014 Terms of Use: http://asme.org/terms

Você também pode gostar