Proceedings of ASME Turbo Expo 2013: Turbine Technical Conference and Exposition

GT2013
June 3-7, 2013, San Antonio, Texas, USA

GT2013-95768

INVESTIGATION OF A RADIAL-INFLOW BLEED AS A POTENTIAL FOR

COMPRESSOR CLEARANCE CONTROL

N. R. Atkins∗
Whittle Laboratory, Dept. of Engineering, University of Cambridge

ABSTRACT δ Radial displacement [m]

The mismatch in thermal response between a High Pressure G Clearance (gap) [m]
Compressor (HPC) drum and casing is a limiting factor in the h̄ Lumped heat transfer coeff. [W m−2 K−1 ]
reduction of compressor clearance. An experimental test rig has p Pressure [bar]
been used to demonstrate the concept of radial inflow to reduce r Local radius [m]
the thermal time constant of HPC discs. The testing uses a simu- s Axial rotor-rotor disc spacing [m]
lated idle - Maximum Take Off (MTO) - idle transient in order to ṁ Mass flow rate [kg s−1 ]
measure the thermal response directly. The testing is fully scaled v Velocity [m s−1 ]
in the dimensionless sense to engine conditions. A simple clo- S Blade span [m]
sure model based on lumped capacitance is used to illustrate the t Time [seconds or minutes]
scope of potential benefits. The proof-of-concept testing shows T Temperature [K]
that HPC disc time constant reductions of the order 2 are feasible v Velocity [m−1 ]
with a radial-inflow bleed of only 4% of bore flow at scaled MTO V Volume [m3 ]
conditions. Using the experimental results, the simple closure α Coeff. of linear thermal expansion [K−1 ]
modelling suggests that for a stage with a significant mismatch ρ Density (metal or fluid) [kg m−3 ]
in thermal response, reductions in 2D axis-symmetric clearance ω Rotational speed [rad s−1 ]
of as much as 50% at MTO conditions may be possible along τ Time constant [s]
with significant scope for improvements at cruise conditions.
Dimensionless
Grb Grashof number based on shroud radius b
NOMENCLATURE Reφ Rotational Reynolds number
a Disc inner radius [m] Re x Axial Reynolds number
A Surface area [m2 ] Ro Rossby number
b Disc outer radius [m] y+ Normalised wall distance
c Specific heat capacity [J kg−1 K−1 ] β Swirl ratio
C Closure [m] Θ Normalised temperature

Sub & superscripts

∗ The
bore Bore flow conditions
author completed the experimental work as a Visiting Research Fellow
casing Casing value
at the University of Sussex Thermo Fluid Mechanics Research Centre. Corre-
spondence to [email protected] cf Centrifugal component
drum Drum value
idle, decel Idle conditions

1 Copyright © 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 in f low Radial-inflow conditions active control, the dominant term is usually axisymmetric and
inlet Bore-flow inlet conditions is set by the need to accommodate the relative thermal and cen-
exit Bore-flow exit conditions tripetal growth of the compressor drum and casing. In general,
metal Metal surface value the casing has a lower thermal mass and is subjected to higher
MT O, accel Maximum take-off conditions heat transfer coefficients than the drum. The drum typically has
r Radial component a longer thermal time constant as it has a comparatively higher
radial Radial inflow temperature thermal mass and lower heat transfer coefficients in the unventi-
s Shaft lated inter-disc cavities.
shroud Shroud (r = b) value This mismatch in thermal response leads to a transient maxi-
x Axial component mum clearance value after an acceleration and to a transient min-
0 Total or stagnation conditions imum clearance value after a deceleration. Occurring just after
φ Tangential component take-off, the transient maximum clearance can be thought of as a
∞ Ambient (sea level) conditions point of minimum surge margin. At the post deceleration mini-
∗ Effective value (temperature or radius) mum clearance point, the rotor drum is hot, but the engine is in
a low power (and rpm) state. Considering a situation such as an
Abbreviations aborted landing, the engine must be capable of a rapid accelera-
CBC Cold Build Clearance tion back to MTO conditions, the worst case is known as a Full
FHRS Full Hot Re-Slam Hot Re-Slam (FHRS). The Cold Build Clearance (CBC) is thus
HPC High Pressure Compressor governed by the need to accommodate this near instantaneous
LHS Left Hand Side centripetal growth. The actual CBC policy depends on the level
MTO Maximum Take-Off conditions of rub which is acceptable for a given design. Each rub opens up
PRT Platinum Resistance Thermometer the CBC further (the strip clearance) reducing the efficiency and
RHS Right Hand Side stability of the compressor.
A reduction in the disc radial growth time constant relative
to that of the casing reduces the magnitude of the transient ex-
INTRODUCTION cursions, which allows a tighter CBC. This gives a benefit at the
Reductions in compressor blade clearance improve machine maximum clearance point and the potential for a reduction in
efficiency and stability. The next generations of airframes and en- clearance at cruise. This paper explores the use of a small bleed
gines have many technology drivers towards smaller core sizes. of radial inflow to decrease compressor disc time constants.
These include the push towards ultra high-bypass ratio engines,
geared turbofans and fully optimised short haul aircraft. In the
medium term, embedded and distributed engines and open rotor Radial bleeds
architectures are also likely to require reduced core sizes com- Radial off-takes are often used as part of a secondary air sys-
pared to current engines. Higher bypass ratios in particular mean tem design. In this case it is often beneficial to reduce pressure
smaller blade heights in the late stages of HP compressors for an losses with de-swirl vanes or vortex reducers (as used in the IAE
approximately constant radius. In general terms, reduced core V2500 for example [1]).
sizes can lead to increased clearance losses. Young and Snowsill used Computational Fluid Dynamics
This project is part of the NEWAC (NeW Advanced core (CFD) modelling to optimise a radial-inflow off-take in an in-
Concepts) project and is motivated by the potential use of inter- termediate pressure compressor [2]. In this study the bleed flow
cooling between the intermediate pressure and high pressure is taken between the compressor spools and it is essential to min-
compressor. Such a configuration would drive a significant re- imise the pressure drop.
duction the HPC blade height relative to current practice even A comparatively small bleed of radial inflow (relative to a
for a nominally constant thrust and constant casing radius. dedicated off-take) offers the potential to increase the heat trans-
With current clearance control technologies any reduction in fer coefficients on the disc faces and to speed up the response
HPC blade height means the clearance gap, G, becomes a larger of the drum. In comparison to a dedicated off-take the pressure
proportion blade span, S . This gives greater aerodynamic loss, drop is not so important for the small bleeds being considered
and reduced stability. As such, improved clearance control is an here. The bleed is likely to be taken towards the later stages of a
important enabling technology for many future engine develop- compressor with a total pressure margin relative to the bore flow.
ments. The penalty will be related to the parasitic heating, extra windage
Compressor clearance is a stack up of several sources. The and thermodynamic cycle cost of using higher pressure air. It is
source terms include blade untwist, radial growth and both expected that the potential increase in surge margin and potential
steady-state and transient non-axisymmetry. However, without reduction in cruise clearance would offset these additional loses.

2 Copyright © 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 closure characteristic. Finally, it details a simple lumped capaci-
δdrum
δ casing tance closure model. This model is used to explore the scope of
potential improvements in HPC clearance.
Closure is the relative movement between the drum (blade
tips) and the casing, as illustrated in Figure 1. The running clear-
ance, G, is the summation of the CBC and the closure1 . In this
paper, the closure and clearance are defined such that an increase
in the drum radius relative to the casing gives a reduction in clear-
ance:
FIGURE 1. Definition of radial movement sign convention

C = δcasing − δdrum , and G = CBC + C

Unfortunately, such detailed modelling is beyond this proof-of-
concept study and is part of ongoing research. A typical closure characteristic is shown in Figure 2. In order to
The nature of the rotationally dominated flow field between illustrate how the CBC is set, each of the various transients are
the compressor discs means that a small bleed flow will be con- described in detail below.
fined to a thin layer next to the discs, effectively coating them
with a film of hot air with enhanced heat transfer coefficients. Acceleration When accelerating from idle to MTO con-
The rotationally dominated flow field amplifies the effect so that ditions the growth of the rotor due to centripetal acceleration
a bleed of only a few percent can increase the heat transfer coef- is almost instantaneous, increasing the closure (decreasing the
ficients by an order of magnitude. clearance). The casing heats quicker than the drum, which
reduces the closure (increasing the clearance). As mentioned
Proof-of-concept testing above, during acceleration the engine is at high power where
This paper demonstrates the potential benefits with proof- surge margin is an important issue. Approximately 20 seconds
of-concept experimental testing in a facility which is fully scaled after the acceleration, the casing reaches steady state at its max-
to engine conditions in the non-dimensional sense. imum radius. However, the response of the drum lags behind,
The considerable effort of testing at fully scaled conditions and this gives rise to the maximum clearance in the cycle. This
is thought a necessity for rotational flow fields. With external maximum clearance is not just a test-bed anomaly - in flight, this
flows, it is often only the regime of the flow field that affects the maximum clearance is exactly what happens just after take-off
scaling of an aerothermal problem. As such, experimental testing when the aircraft is climbing at almost maximum power. It is par-
with low speed analogs is often highly effective. However, for ticularly undesirable as it causes a significant reduction in surge
rotationally-dominated internal cavity flows, the dimensionless margin at one of the most critical points in the flight envelope.
parameters have a first order effect on the bulk flow field (such
as the radial pressure gradient and the Coriolis forces). It is still Deceleration When decelerating, the reverse situation
an open question as to how rotationally dominated aerothermal occurs. The drum shrinks instantaneously with the reduction in
problems scale at engine levels of rotational Reynolds numbers. spool speed, decreasing closure (increasing clearance). The cas-
An existing facility has been extensively modified in order ing then cools and contracts rapidly, increasing closure (decreas-
to simulate a transient cycle which is fully scaled in the dimen- ing clearance) once again. However, the drum cools and shrinks
sionless sense to engine conditions. With this mode of operation, at a much slower rate. This gives rise to the characteristic over-
the thermal time constants can be measured directly at the correct shoot.
conditions with low uncertainty.
Importantly, the direct measurement of time constant avoids
the need for Nusselt number measurements. Cavity flow testing Hot re-slam acceleration This transient might not be
suffers from considerable inherent uncertainties in Nusselt num- a problem on its own, as the engine is in a low power condi-
ber measurement which could render the results almost worth- tion without risk of surge. However, the engine may need to
less. The direct measurement of time constant used in this study be brought back to full power before the drum has had time to
avoids the need for Nusselt number measurements. cool. For example, this is exactly the situation which would
occur during an aborted landing. Bringing the engine back to
full power when the drum is still hot means that the centrifgual
CLOSURE CHARACTERISTICS growth increases the closure almost instantaneously (decreases
This section introduces the definition of closure used in this
work. It then gives further details of a typical engine test cycle
1 After a rub the CBC term should be replaced by the new strip clearance

3 Copyright © 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 the clearance). Often known as a hot re-slam, this rapid increase temperature and the shroud metal temperature, Θ. It is given by
in closure has the potential to grind the blade tips into the casing. the following expression:
As well as the retention of engine performance and operability,
certification requires that the CBC clearance be set such that the ∗
T idle = T bore,idle + Θidle T shroud,idle − T bore,idle


engine can survive such an event.
If the CBC is set sufficiently large, the blade tips would not
rub the casing at all even with a full hot re-slam (FHRS), where Using Heaviside step function notation, the model for the effec-
the engine is brought from idle back to MTO conditions at the tive temperature of the drum throughout the transient cycle is
worst point in the overshoot. then
However, this policy would give an unacceptably high clear-  " − (t − t0 )
!#
ance at cruise. As such, a compromise is often set so that the
∗
T drum (t) = T ∞ + H(t − t0 ) T idle
∗
− T ∞ 1 − exp ...
τidle
blade tips would rub during an extreme event, but the engine " !#
  − (t − t1 )
would survive. This reduces the clearance penalty carried at +H(t − t1 ) T MT
∗
− T ∗
idle 1 − exp ...
cruise. A simplistic rub policy is illustrated in Figure 2, where
O
τaccel
" !#
the CBC is set such that the clearance between the blade tips and   − (t − t2 )
∗
−H(t − t2 ) T MT ∗
O − T idle 1 − exp ...
casing at the worst case point is half the centrifugal growth of a τdecel
FHRS.
" !#
  − (t − t3 )
A decrease in the time constant of the drum reduces the over- +H(t − t3 ) T cruise
∗ ∗
− T idle 1 − exp
τcruise
shoot. This allows for a tighter CBC clearance with potential
efficiency and surge margin benefits across the complete cycle. Radial displacement This temperature time-history is
used to predict the radial displacements. These are modelled us-
ing an effective radius, r∗ , the coefficient of thermal expansion, α,
Lumped modelling
and the radial growth due to centrifugal acceleration, δc f . The ef-
A simple lumped capacitance model is used in this study
fects of temperature dependent material properties are neglected.
as a convenient method to illustrate the scope of potential im-
The centrifugal growth term is calculated from the instantaneous
provements. The model has been iteratively matched to a full
response after an acceleration from idle to MTO conditions:
thermo-mechanical prediction.
h i
δdrum = αrdrum
∗ ∗
T drum (t) − T ∞ + δc f
Time constant The mechanical response of each stage
is dominated by a single first order time constant and a temper-
ature step. Each disc or casing section is modelled as a lumped Model parameter matching/estimation The model
thermal mass, and the overall time constant for thermal growth parameters which must be determined for each stage of the drum
in the radial direction is given by the following expression: or casing segments are then Θ MT O , Θcruise , Θidle and the time
constants, τaccel , τdecel and τcruise . These parameters have been
ρVc found iteratively by matching the simple lumped model to a full
τ= 2-D thermo-mechanical prediction of the entire HPC module,
h̄A
which was provided by Rolls-Royce. This full 2-D model is
where h̄ represents the effective average heat transfer coefficient built by matching to actual tip clearance measurements during
and ρ, V and A are the density, volume and surface area of a an engine development test. The clearance gap, G, is measured
disc or casing segment. This definition of the overall average directly using capacitance probes above the blade tips of each
heat transfer coefficient includes both convection and the effect stage.
of conduction through the drive arms and down the rotor discs.
Results The matched closure model is shown for a sin-
Effective temperature The temperature of the drum gle stage in Figure 2, and the iteratively matched parameters for
and casing is modelled as a summation of step functions. The the drum are given in Table 1. The clearance and the closure
inner side of the casing and the shroud of the drum experience are normalised by the blade span, S . It can be seen that a sin-
main annulus temperatures. The outer side of the casing is ex- gle time constant is sufficient to model the behaviour to within a
posed to the temperature of the void or various upstream bleeds few percent of the absolute closure. It is possible to match the
and off-takes. The disc bore is cooled by the through flow. As steady state behaviour with any combination of drum and casing
such, the radial movement of each stage is driven by an inter- parameters. However, matching both the transient and steady
mediate temperature. This effective temperature, T ∗ , is modelled behaviour gives a fully determined problem with a unique solu-
here using a fraction of the difference between the bore flow air tion. The value of Θ required to match the characteristic at idle

4 Copyright © 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Closure high degree of coupling between the metal temperatures and the
−2 flow field makes it a challenge for reliable CFD modelling.
Recently, Vinod et al. [3] compared an integral momentum
C/S [%]

−3 boundary layer method with CFD predictions based on vari-

Lumped model
ous turbulence models. They used both axi-symmetric and 3-
−4 Full thermo−mech model dimensional ∼20◦ sector models. They compared their mod-
Max clearance at high power
Hot re−slam point
elling results against several test cases. In general they found
−5 CBC set at 1/2 FHRS that eddy-viscosity models had mixed results but performed al-
0 500 1000 1500 2000 2500 3000 3500 4000 most as well as Reynolds Stress Model (RSM) predictions in nar-
idle MTO idle cruise row aspect cavities such as those in the present study. In partic-
Clearance ular the HYDRA based predictions using the Spalart-Allmarras
(S-A) turbulence model performed well, even when using a wall
3
function treatment. Good agreement was also found using the in-
tegral method and they concluded that this suggests that the core
G/S [%]

2
flow is relatively insensitive to turbulence modelling. As such,
HYDRA predictions using the S-A model with wall functions
1
have been used in this work to help visualise the flow field.
Owen and Wilson [4] give a comprehensive review of the
0 field of rotating disc systems. Therefore, a only a brief overview
0 500 1000 1500 2000 2500 3000 3500 4000
Time [s] is given here. Important early reference works were published
by Hide [5], Wormley [6], and Owen et al. [7]. These ground-
FIGURE 2. Closure and clearance characteristic during a transient test
breaking works established the characteristics of the flow field.
bed cycle
Firouzian et al. [8, 9] presented flow and heat transfer measure-
ments in a large scale test rig with a Perspex outer shroud. Far-
thing [10] studied the use of de-swirl nozzles to reduce losses and
Idle MTO Test bed eq. cruise
pressure drops in cavities with radial inflow. Günther et al. [11]
Θidle 0.41 Θ MT O 0.42 Θcruise 0.2 studied steady state heat transfer with and without radial inflow
in a rig similar to that used for the study presented in this paper.
τdecel 350 s τaccel 74 s τaccel 149 s

TABLE 1. Fitted drum model parameters Visualisation - CFD setup The Rolls-Royce HYDRA
(Ver. 6) solver was used in steady mode. A grid dependence
study was used to set the cell count of the multi-block struc-
and MTO conditions is ∼ 0.4. This corresponds to within a few tured mesh topology. The Spalart-Allmaras turbulence model
percent of the measured data at a normalised radius of r/b = 0.8 was used with wall functions and y+ values where kept to ap-
(presented below). Both the idle and MTO test bed conditions proximately 30 on the surfaces. The prediction shown in Figure 3
are completely equivalent to those in flight as the aircraft/engine was run using the measured boundary conditions from the MTO
is in low level flight with sea level inlet temperature and pres- condition steady state test cases with a radial inflow of 4.4% of
sure. In contrast, the cruise condition shown here is a test-bed- bore flow (see Table 4 - Case 12). The CFD solution is used here
equivalent operating point where the the engine inlet pressure is to illustrate the flow field and is discussed in detail below.
about 3 times higher and the inlet temperature is approximately
100 K higher than those in-flight. Dimensionless scaling
The next sections discuss the cavity flow field with radial The pertinent geometry of the cavity is shown in Figure 4,
inflow. and the governing dimensionless parameters are given in Table 2.
The key dimensions are: r s = 0.070 m; a = 0.080 m; b = 0.230 m;
s = 0.048 m. The rotational Reynolds number (or the inverse of
CAVITY FLOW FIELD the Ekman number) represents the relative important of Coriolis
Instrumentation access is inherently limited in the study of forces over viscous forces within the cavity [12]. The Rossby
radial inflow within a rotating cavity. Testing has traditionally number at the cob radius a represents the degree to which the
required a trade off between the Reynolds number and the res- flow is driven by the cavity flow or the axial through flow, which
olution of the measurements. Similarly, the full annular nature, is characterised by the axial Reynolds number.
the rotational dominance on turbulence production and the often The baseline cavity flow, that of a rotating cavity with axial

5 Copyright © 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Radial inflow bleed

s
b

a
rs
Bore flow

FIGURE 4. Geometry definitions

average flow field in the outer part of the cavity approaches

solid body rotation. However, it has been shown by Long and
Childs [15] that the hot shroud drives natural convection currents
which increase with increasing Grashof number and lead to large
scale structures and instabilities. In the present study the Grashof
number has been matched by controlling the shroud temperature
with a radiant heater array.
The toroidial vortex is also prone to instability and the vortex
has been shown to break down and enter the cavity as a precess-
ing radial arm [13].
The introduction of radial inflow generally washes out these
instabilities. Even if the radial inflow has a low velocity relative
to the cavity, it can have considerable angular momentum in the
absolute frame of reference. In the absence of viscous losses,
conservation of angular momentum means that the velocity of
FIGURE 3. Predicted flow structure in a rotating cavity with radial the inflow increases with the inverse of its reducing radius. This
inflow. The bore flow is from left to right. The LHS image shows a causes the flow within the cavity to spin up to a high tangential
contour and two iso-surfaces of swirl ratio β = vφ /(ωr) which illustrate velocity approaching the behaviour of a free vortex. With this
the quasi 2-dimensional flow structure. The RHS image shows a contour increase in rotational speed the Coriolis forces act to stiffen the
and iso-surface of vr set at -0.1 ms−1 which indicates the extent (or limit) flow to perturbations in any direction other than parallel to the
of the radial inflow within the cavity. axis of rotation similar to geostrophic conditions in atmospheric
flows. This free-vortex like behaviour means that when com-
pared to the baseline case the flow in the upper part of the cavity
ρbore v x 2(a−r s )
Axial Reynolds No Re x = µbore is much more stable and is almost completely 2D in nature.
ρbore ωb2 Without rotation, the radial bleed might be expected to mix
Rotational Reynolds No Reφ = µbore out like a jet across the axial extent of the cavity. However, any
Ro = b2 flow into the rotationally dominated core is suppressed by the
ωa = 2a(a−r s ) Reφ
vx Re x
Rossby No
Coriolis force. An increase in boundary layer thickness as the in-
ρ2bore ω2 (b(b−a))3 β(T shroud −T bore )
Grashof No Grb = µ2bore
flow moves down the disc would require the core fluid to occupy
a volume with a reduced axial extent. Continuity would require
TABLE 2. Governing dimensionless groups for cavity flows an element of fluid which reduces in axial extent to change ra-
dius. Conservation of angular momentum means that this change
in radius is resisted in just the same way as one experiences resis-
throughflow, has been studied extensively (for example Owen et tance when trying to move their legs whilst spinning on a chair.
al. [13] and Johnson et al. [14]) and features a range of regimes As such, the radial bleed flow can only move radially inwards at
depending on the Rossby number and cavity geometry. Shear the walls where the viscous forces dominate.
between the throughflow and the cavity flow drives a torroidal This flow field is illustrated in Figure 3. The left hand image
vortex. The rotating cavity flow suppresses this vortex, and the shows contours of swirl ratio, β = vφ /(rω). Two iso-surfaces have

6 Copyright © 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 been overlaid to illustrate the 2D nature of the cavity flow field. MTO Idle
Although the radial inflow is confined to a thin layer on the disc
surface, the swirling flow still imparts momentum to the 2D core, Pinlet 4.5 bar 2.5 bar
which spins up in a free vortex type manner. At the lower extent T bore 340 K 300 K
of the cavity, this circumferential free vortex type structure mixes T radial ≈ T shroud 390 K 370 K
with the axial bore flow. The right hand image shows the same ṁbore 0.7 kg s−1 0.2 kg s−1
solution, but this time contours of the radial velocity component ω 730 rad s−1 370 rad s−1
are shown. The scale is centred about zero so the extent of the
green region indicates that the flow is almost entirely quiescent TABLE 3. Nominal test conditions
in the radial direction. Again, an iso-surface has been overlaid to
illustrate the flow structure; in this case the iso-surface is set at Radial inflow
a radial velocity, vr , of -0.1 ms−1 . This delineates the region of Case Type Ro Reφ Re x Grb per cavity
radial inflow within the cavity. It can be seen that the inflow is ×10−6 ×10−5 ×10−12 ṁin f low /ṁbore
confined to an annular region near the shroud (the source region)
1 Accel. 0.58 8.3 1.5 2.4 n/a
before it diffuses axially towards the disc surfaces, where it is 2 Decel. 0.44 2.7 0.35 0.41 n/a
confined to the disc surfaces until it reaches the bore flow. 3 Accel. 0.57 8.5 1.5 2.6 4.4%
4 Decel. 0.47 2.6 0.37 0.36 7.8%

5 0.45 2.6 0.36 0.37 n/a

PROOF-OF-CONCEPT TESTS
6 Idle 0.47 2.6 0.37 0.36 7.8%
This section gives details of the test facility. The working 7 (steady) 0.45 2.9 0.42 0.41 10%
section (Figure 5h) features a simplified version of an HPC bore 8 0.48 2.8 0.41 0.36 15%
geometry which is approximately 1:1 scale to a current engine.
9 0.58 8.2 1.5 2.5 n/a
It was developed as part of the ICAS-GT projects. Thorough de-
10 MTO 0.57 8.6 1.5 2.5 1.6%
tails and test data from the previous investigations can be found 11 (steady) 0.60 8.2 1.5 2.7 3.0%
in Alexiou [16, 17] and Miche [18]. 12 0.59 8.2 1.5 2.6 4.4%
The facility in this study was designed to match the dimen-
sionless conditions of a large civil turbofan HPC bore. The ratio
of the axial to rotational Reynolds number is fixed by the Rossby TABLE 4. Dimensionless test conditions
number and the physical dimensions. The dimensionless groups
are not independent and cannot be matched to the engine values ble 5. For these proof-of-concept tests, the rig has been exten-
simultaneously unless the geometry is directly scaled. As the sively modified, completely re-instrumented, and new air sup-
test rig is almost 1:1 scale with the reference engine a very close plies and heating systems have been installed. As well as the
match is indeed possible. The Rossby number was fully matched modifications to deliver the radial inflow, the pressure, temper-
to reference engine values and the rotational Reynolds number ature and rig speed have all been increased in order to match
was set as close to the engine values as possible (based on the the dimensionless engine conditions. A steady state thermo-
power limits of the laboratory compressor plant) whilst reducing mechanical model was built using the Rolls-Royce SC03 FE
the axial Reynolds number appropriately. The nominal test con- code to check the integrity of the rig.
ditions for idle and MTO conditions are given in Table 3, and The new installation uses two oil-free screw compressors to
accurate dimensionless conditions for each of the 12 test cases deliver the cold idle and warm MTO condition bore flows. A
used in this paper are detailed in Table 4. The inlet bore flow was real time control system (Figure 5n) allows control of the non-
kept nominally constant for the idle and MTO conditions and the dimensional conditions and transient operation.
radial inflow is additional. In a potential engine design case it
is perhaps more likely that the total, or exit, bore flow would be
kept constant. Working Section
In summary, the Rossby number was engine matched, the A sectional view of the test section is shown in Figure 6. The
rotational Reynolds numbers are within 10% and the axial darkest grey region is the simulated shaft, the mid grey section
Reynolds number are within 20% of the engine values. is the outer drum and the light grey section indicates the rotating
assembly.
The outer drum supports the rotating section via a grease-
Facility packed bearing. The other end of the rotating assembly is sup-
The various components and subsystems are indicated in ported by an oil-fed axial-roller bearing to accommodate differ-
Figure 5 and the key to the various components is given in Ta- ential thermal growth. The outer drum also locates a 24 kW radi-

7 Copyright © 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 ant heater array. The heaters are driven from a three-phase power o) n) m) l) k)
controller, which is connected to the central control system.
The warm MTO and cold idle supplies were fed to the facil-
ity via ISO standard orifice plates. At idle conditions, only the
cold supply was fed to the test section, whilst the hot supply was
vented via the bypass line (Figure 5b). At MTO conditions the
two supplies were combined adjacent to the rig to minimise the
thermal lag of the pipe work. The pressure within the working
a)
section was modulated by a butterfly valve at exit (Figure 5d);
again, this was controlled by the central control system. b)
j)
The bore flow was introduced into the rig via a small
plenum, Figure 6p. The flow then passes through eight 20 mm di-
c)
ameter transfer holes which help to produce uniform conditions d)
in the downstream plenum. The flow then mixes in the rotating e)
plenum, before it is transferred to the bore through a further six i)
22 mm diameter angled transfer holes, Figure 6l. The additional
pressure drop ensures a uniform flow field and gives engine rep-
f)
resentative turbulence (N.B. this is the opposite direction of flow
to previous work).
Two sided, pressure-balanced labyrinth seals have been
added to the rig. Three are used to seal the bore-flow inlet-path,
at the outer radius (Figure 6r), on the inner side of the transfer g) h)
holes (Figure 6q) and close to the roller bearing (Figure 6m). A
further seal is used to seal the radial-inflow delivery. The seals A) z) y) x)
indicated by r, q and g feature 3 fins on either side of the balance
cavity. The inner seal, q, has 4 fins on the flow side and 2 on the
vent. The balance cavities are fed from an auxiliary air supply
via proportional valves (Figure 5A, z), which are also connected p)
to the central control system. The cold-build clearances are of
w)
the order of 0.1 mm, and the system achieves flow side pressure v)
differentials in the region of +/-10 mbar, which ensures that the u)
leakage path is negligible. t)
s)
r)
Radial-Inflow Supply
The radial inflow is metered via a pair of ISO standard ori-
fice plates, one for idle conditions (low flow) and one for MTO
q)
conditions (high flow) (Figure 5x and y). The radial-inflow air is
supplied to the test section via a 10 kW inline heater bank (Fig-
FIGURE 5. Sussex Multiple Cavity Rig
ure 5q) after an actuated control valve (Figure 5p).
The inflow air is routed through the central stationary shaft
(Figure 6k) into 4 feed pipes (Figure 6j). It then passes into the General Instrumentation The static pressure measure-
rotating frame through a spigot (Figure 6j), which is sealed via a ments use Scanivalve DSA transducer units and the mass flow
pressure-balanced seal (Figure 6g). The radial-inflow air is then rate measurements use Rosemount differential transducers.
fed into an annular cavity (Figure 6v) sealed on all faces by con- The rotating frame thermocouple locations are indicated in
tinuous silicone O-rings. It then flows via cross-drilled transfer Figures 8 and 9. In addition to the metal surface temperatures, the
holes (Figure 6u) to a series of axial galleries. These galleries radial-inflow air temperature is measured at two locations. The
feed threaded ports. The can be either sealed or fitted with an thermocouple beads are mounted via a plug of epoxy resin to re-
insert. The inserts are cross-drilled in order to deliver the ra- duce conduction errors. Glass fibre insulated K-type thermocou-
dial inflow to the rotating cavities (Figure 6t). The testing in this ples are used. The thermocouple wires are routed circumferen-
paper uses sixteen 5 mm diameter inflow ports for each inflow tially along an isotherm for at least 10 wire diameters to remove
cavity. conduction errors. Each bead is peened into the surface, which

8 Copyright © 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 x) w) v) u) t) s)
a Cold (Idle) supply line (orifice plate out of view)
b Hot (MTO) bypass line
c Mixed inlet to the test section
d Back-pressure control valve
e Oil breather (return) r)
f Oil separator tank
q)
g Radial-inflow pre-heater
p)
h Test section (See Figure 4)
i Test-section purge-air blower
a)
j Temperature controlled instro. cupboard
k Motor control and DC brake system b)
l Motor and TASC speed control unit
m Hot (MTO) supply line and orifice plate
n Controller, instro and electrical systems c) d) e) f ) g) h) i) j) k) l) m) n) o)

o Drum and radial-inflow power controllers

p Radial-inflow mass-flow control valve
q Radial-inflow inline heaters
s, t & v Balance seal supplies
u Oil return Bore-flow
path
w Drum heater hook-ups
x High flow radial-inflow orifice plate mass flow tube
y Low flow radial-inflow orifice plate mass flow tube
z Balance seal proportional control valves
A Balance seal supply
Radial-inflow
path

TABLE 5. Key to Figure 5

FIGURE 6. 2D cross-section and the flow paths for the bore flow and
radial inflow
gives a cold welded metal-to-metal contact. Extension wires are
not used and the thermocouple wires are routed through the discs
to the telemetry unit (Figure 6b). The peening itself leaves a de- from the Sussex TSW facility test programme which used the
pression of a few tenths of a millimetre. This is back filled with a same measurement chain, where repeated testing indicated real-
skin of epoxy resin to give a smooth aerodynamic surface. This istic bias uncertainty in the region of ±0.5 K [19]. A 60 channel
small disturbance in conductivity is minor when compared to the Datatel radio telemetry system has been used to measure the ro-
potentially significant local heat transfer enhancement of a sur- tating frame thermocouples. The system comprises 4 individual
face protrusion or sharp edged depression. A simple 1D model units connected in a parallel bus (Figure 6b). Each unit has an
of the thermal disturbance of the thermocouple itself (the embed- in house developed cold junction reference with a PRT mounted
ding error) has been carried out for previous installations and the in close thermal contact to the gold pins used to terminate the
disturbance gives errors of the order of 0.1 K. This can effectively thermocouple junctions at the telemetry amplifier modules. The
be ignored for this proof-of-concept test. resistance of the PRT is measured on a dedicated 4-wire resis-
The junctions are welded in house from a single batch of tance channel.
wire. The entire data acquisition chain including the thermo-
couples, cold junctions, amplifiers and acquisition cards are end-
to-end calibrated by comparison against a national-laboratory- EXPERIMENTAL RESULTS AND DISCUSSION
standard reference system. This reference system gives a com- Data from a baseline (LHS) and radial-inflow transient idle-
parison uncertainty of 20 mK which is safely an order of magni- MTO-idle run (RHS) are shown in Figure 7. The upper graphs
tude lower than the uncertainty of the thermocouples. Analysis show a subset of metal temperatures (for clarity) and the lower
of the calibration data shows that both the thermocouples and figures show the variation of non-dimensional conditions. The
cold junctions give typically single point uncertainties in the re- transients between the idle and MTO conditions are achieved in
gion of 0.1 K. under 30 seconds.
Errors such as system drift are not assumed to be normal,
so there is no reduction in uncertainty through averaging (al-
though it does remove the sample to sample noise). All of the
error sources are simply added to give a worst case uncertainty
estimate of approximately ±0.5 K. This is consistent with data

9 Copyright © 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Transient data
a Telemetry unit stationary antenna
b Telemetry unit
The time constants are measured by fitting the transient data
c Belt pulley to a simple first order model of the form
e Grease-pack rotor bearing
f Inner grease seal  −t 
g Radial-inflow balance seal T (t) = T (0) + ∆T 1 − e τ
h Radial-inflow transfer spigot
i Shaft location bearing (no longer in use)
where τ is the time constant and ∆T is the overall temperature
j Radial-inflow delivery tubes - 4 OFF
k Shaft cavity (stationary) step at each model point (or radius). The fitting uses a Nelder-
l Bore-flow transfer holes Mead least squares minimisation over the entire transient to re-
m Shaft inner balance-seal 1 ject subtle changes due to small adjustments in the rig control
n Oil fed roller bearing system. The temperature data in the following figures is plotted
o Shaft
in a normalised form, Θ, which is directly equivalent to that used
p Bore-flow inlet
q Shaft inner balance-seal 2
in the simple model. It is calculated using the bore flow inlet
r Shaft outer balance-seal air temperature, T inlet , and T shroud , which is the measured metal
s Outer casing and end plates temperature at r = b:
t Radial-inflow galleries and delivery holes
u Radial-inflow cross-drilled entrance to galleries T (r) − T inlet
v Radial-inflow supply cavity T (r) = T inlet + Θ(r) (T shroud − T inlet ) i.e. Θ(r) =
T shroud − T inlet
w Exit labyrinth seal
x Exhaust
A subset of the raw and fitted metal surface temperature traces
for the baseline and radial-inflow transients is shown on the LHS
TABLE 6. Key to Figure 6 of Figure 8 (cases 1 & 2) and Figure 9 (cases 3 & 4). The mea-
sured time constants for all measurement points are shown on the
r/b = 1 RHS of the corresponding figures. The data is plotted so that the
Baseline r/b = 0.9 Radial-inflow
140 140
difference between the upstream and downstream disc faces can
r/b = 0.7
r/b = 0.3 be seen.
120 120
Inlet The fitted model is shown as a dashed line in all cases. The
Temperature [°C]

100 100 fitted response is close to the raw data, particularly for the radial-
80 80 inflow deceleration (case 4). However, several of the traces show
60 60
evidence of both longer time constants, and time constants which
change with time. This represents the thermal signatures of the
40 40
various sections of the test facility, including the inlet pipework
20 20 and routing, as well as the thermal mass of the outer drum itself.
0 20 40 60 80 0 20 40 The effect of the thermal mass of the inlet ducting can also be
seen in the raw run plots of Figure 7 for example. The tran-
Dimensionless (see legend for scaling) [−]

Ro sient bore flow and heater systems were designed specifically

3 Re x10−7 3
φ with low thermal mass to minimise these effects, and in the few
−5
2.5 Re x x10 2.5 cases where these longer time constants are evident, they are only
2 Gr x10−12 2
a few percent of the dominant first order response at most. The
b
technique clearly resolves the subtle changes in time constant,
1.5 1.5
both radially and axially across the discs, driven by the differing
1 1 local aerothermal boundary conditions.
0.5 0.5

0 0
Discussion The time constant data for the four cases
0 20 40 60 80 0 20 40 shown in Figures 8 & 9 are compared in Figure 10. At the radial-
Time [min] Time [min]
inflow levels shown here, there is a clear and significant reduc-
tion in the disc time constant. The concept of disc time con-
stant reduction through the introduction of radial inflow works
FIGURE 7. Transient run data for baseline (LHS - cases 1 & 2) and as expected.
radial inflow (RHS - cases 3 & 4). Top to bottom: disc temperatures; As seen in the steady state data presented in the following
dimensionless conditions sections, the heat transfer in the cob region of the disc is less

10 Copyright © 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Baseline: Accel. temperatures Time constant − τ Radial inflow: Accel. minflow /m bore = 4% Time constant − τ
1 1 1 1

0.9 0.9
0.9 0.9

0.8 0.8
0.8 0.8
0.7 0.7

0.7 0.7
0.6 0.6
r/b [−]

r/b [−]
Θ [−]

Θ [−]
0.5 0.6 0.5 0.6

0.4 0.4
0.5 0.5

0.3 0.3
r/b = 1
r/b = 1 0.4 0.4
r/b = 0.9
r/b = 0.9
0.2 0.2 r/b = 0.7
r/b = 0.7
r/b = 0.5
r/b = 0.5
0.3 r/b = 0.3 0.3
0.1 r/b = 0.3 0.1 Upstream face
Upstream face
Downstream face Downstream face
0 0.2 0 0.2
0 5 10 15 0 5 10 0 5 10 15 0 5 10
Time [minutes] Time [minutes] Time [minutes] Time [minutes]

Baseline: Decel. Temperatures Time constant − τ Radial inflow: Decel. minflow/m bore = 8% Time constant − τ
1 1 1 1

0.9 0.9
0.9 0.9

0.8 0.8
0.8 0.8
0.7 0.7

0.7 0.7
0.6 0.6
r/b [−]

r/b [−]
Θ [−]

Θ [−]

0.5 0.6 0.5 0.6

0.4 0.4
0.5 0.5

0.3 0.3
r/b = 1
0.4 r/b = 0.9 0.4
r/b = 1
0.2 r/b = 0.9 0.2 r/b = 0.7
r/b = 0.7 r/b = 0.5
r/b = 0.5 0.3 r/b = 0.3 0.3
0.1 0.1
r/b = 0.3 Upstream face
Upstream face
Downstream face
Downstream face
0 0.2 0 0.2
0 5 10 15 0 5 10 0 5 10 15 0 5 10
Time [minutes] Time [minutes] Time [minutes] Time [minutes]

FIGURE 8. Baseline transients (Upper) Accel. (Lower) Decel. LHS: FIGURE 9. Radial-inflow transients (Upper) Accel. (Lower) Decel.
Raw and fitted metal temperature time histories. RHS: Local time con- LHS: Raw and fitted metal temperature time histories. RHS: Local time
stant at each model point location (cases 1 & 2). constant eat each model point location (cases 3 & 4).

affected by the radial inflow than the diaphram region. There terms of time constant reduction. The absence of this effect at
is also less of an effect at MTO conditions than idle, where the MTO conditions suggests that it is simply due to the dominance
change in time constant is only ∼10 seconds. This is perhaps of the time constant of the thermal mass of the thick outer rim at
to be expected, as the wetted area of the cob region sees more lower power (i.e. lower heat transfer coefficients and lower heat
bore flow in comparison to the rest of the disc. This suggests that flux down the disc overall). The outer rim of the test section is
with increasing axial flow Reynolds number, Re x , the through- considerably thicker than an engine drive arm, and it is thought
flow dominates the cob region heat transfer, which is three times that this effect is undoubtedly test rig specific and unlikely to be
higher at MTO than idle conditions. a factor in the engine architecture (the disc diaphragm and cob
At idle conditions, the near-shroud region is less affected in are fully representative). This effect was seen and anticipated

11 Copyright © 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Time constant − τ Idle conditions − steady state
1 1
Baseline − case 5
7.8% inflow − case 6
0.9 0.9 10% inflow − case 7
15% inflow − case 8
0.8
0.8
0.7
Accel.

r/b [−]
0.7
0.6
r/b [−]

0.6
0.5

0.5 0.4

0.4 Decel. 0.3

0.2
0.3 0 0.2 0.4 0.6 0.8 1
Θ [-]
0.2
0 1 2 3 4 5 6 7 8 9 10
Time [minutes]
MTO conditions − steady state
1
FIGURE 10. Comparison of the local time constant values between
0.9
the baseline conditions (cases 1 & 2) and the radial-inflow testing (cases
3 & 4). Radial inflow - blue symbols, baseline - red symbols.
0.8

0.7
r/b [−]

during the rig design stages from the results of the 2D thermo-
mechanical modelling. However, a reduction in thermal mass of 0.6
the rim was rejected due to the increased complexity and risk to
the project. 0.5

0.4
Steady-state data and discussion Baseline − case 9
1.6% inflow − case 10
The steady state temperature data is shown in normalised 0.3
3% inflow − case 11
form in Figure 11 at idle (upper) and MTO (lower) conditions. 4.4% inflow − case 12
It can be clearly seen that radial inflow raises the average tem- 0.2
0 0.2 0.4 0.6 0.8 1
perature of the complete disc. It has a significant effect on the Θ [-]
radial temperature profile even at levels as low as 1.6% of bore
flow per cavity. As introduced earlier, this effect can be explained FIGURE 11. Comparison of radial temperature profiles for three lev-
through the dominance of Coriolis forces away from the viscous els of radial inflow as a percentage of bore flow at idle (cases 5, 6, 7 &
boundary layers at these levels of rotational Reynolds number. 8) and MTO scaled conditions (cases 9, 10, 11 & 12).
The rotational dominance keeps the radial inflow confined to thin
layers on the disc faces. With conservation of angular momen-
tum, the circumferential velocity of this radial inflow increases Increasing the level of radial inflow both increases the nor-
as it progresses down the disc. It is this relative velocity in the malised temperature of the disc and further reduces the temper-
circumferential direction that increases the heat transfer coeffi- ature gradient on the outer parts of the disc. This behaviour is
cient. most pronounced in the near shroud and diaphragm regions, but
Relative to the baseline case, the radial inflow reduces the reduces towards the cob region as the axial flow becomes domi-
normalised temperature gradient near the shroud and to a lesser nant.
extent on the diaphragm regions of the disc. Conversely, it in-
creases the radial temperature gradients in the cob region. It is
Bore flow pressure drop The bore flow pressure drop
noted that this may introduce additional thermal stress in a criti-
data from the steady state tests has been included in Table 7. The
cal area of the disc.

12 Copyright © 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

  
p0,inlet − pexit Radial inflow Idle - 7.8% radial inflow MTO - 4.4% radial inflow
Case Type Ro Reφ Re x p0,inlet per cavity
×10−6 ×10−5 ×100 [%] ṁin f low /ṁbore
Θidle 0.9 Θ MT O 0.75
τdecel 350/1.4 s τaccel 74/2 s
5 0.45 2.6 0.36 0.36 n/a
6 Idle 0.47 2.6 0.37 0.71 7.8%
7 (steady) 0.45 2.9 0.42 0.77 10%
8 0.48 2.8 0.41 0.89 15% Test bed eq. cruise

9 0.58 8.2 1.5 2.21 n/a Θcruise 0.5 0.6 0.7

10 MTO 0.57 8.6 1.5 2.20 1.6% τaccel 149/1.8 s
11 (steady) 0.60 8.2 1.5 2.35 3.0%
12 0.59 8.2 1.5 2.37 4.4% TABLE 8. Estimated radial-inflow model parameters

TABLE 7. Pressure loss results

Scoping study
Clearance At both idle and MTO conditions the normalised tempera-
ture value of about Θ = 0.4, which matched the baseline closure
3.5 Baseline
characteristic in the lumped modelling, corresponds to a radial
Radial−inflow estimate
height of r/b = 0.8. As an estimate, the normalised temperature
3
at a radial height r/b = 0.8 is taken as that which drives the ra-
dial growth. It is clear that any amount of radial inflow brings a
2.5
significant increase in the temperature of the discs, which brings
an increase in the steady state radial growth. For the cases tested
2
G/S [%]

here of an inflow of 7.8% bore flow at idle and 4.4% of bore flow
at MTO conditions, the normalised temperature values increase
1.5 from the baseline level of about 0.4 up to 0.95 and 0.7 respec-
Θcruise = 0.5
tively. The time constant improvement is taken as a factor of 2
Θcruise = 0.6
1 for the MTO conditions and a factor of 1.4 for the idle conditions.
Θcruise = 0.7
These estimated model parameters are summarised in Table 8.
0.5 The prediction of the lumped model with the estimated
radial-inflow parameters is shown in Figure 12. The reduction
0 in the high power clearance is significant with an almost 50%
0 500 1000 1500 2000 2500 3000 3500 4000
Time [s] reduction from 2.15% to 1.25% of span. The hotter steady state
drum temperature also drives a beneficial reduction in the steady
FIGURE 12. Comparison of baseline and estimated radial-inflow state MTO clearance. The reduction in post deceleration over-
model shoot allows for a tighter CBC, but the actual cruise clearance
depends entirely on the steady state drum temperatures. The ex-
perimental testing shows that the radial inflow always increases
definition of pressure losses is complicated by the rotating flow. the steady state disc temperatures. As such, the predicted cruise
As a comparative reference, the inlet total pressure, p0,inlet , is cal- clearance is shown for three levels of normalised (test-bed equiv-
culated from the measured average static pressure in the cavity alent) cruise disc temperatures of 0.5, 0.6 and 0.7. The reduction
upstream of the angled transfer holes (Figure 6l) and the mea- in cruise clearance increases almost linearly with increasing disc
sured mass flow rate. The exit static pressure, pexit , is measured temperature suggesting significant potential for improvement.
on the shaft just upstream of the exit transfer holes which can be However, this would need confirmation with full 2-D thermo-
seen in Figure 6. The data is expressed as the dynamic pressure mechanical testing. Also, this simple analysis does not take in
lost as a percentage of the inlet total pressure. At idle conditions account the effects of increased windage, the impact of the in-
the dynamic pressure lost from inlet to outlet increases by 0.5% creased pressure drop and parasitic heating on the downstream
of the inlet total pressure with the addition of 15% of bore flow components.
via the radial ports. The increase in pressure loss at MTO condi-
tions is 0.2% of the inlet total pressure for an addition of 4.4% of
bore flow through the radial ports. CONCLUSIONS
The proof-of-concept tests have demonstrated that the use of
radial inflow does indeed reduce the time constant of HPC discs.

13 Copyright © 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 At MTO conditions, a radial-inflow level of 4.4% bore flow (ap- “Source–sink flow inside a rotating cylindrical cavity”.
proximately 0.04% of main annulus flow) reduced the disc time Journal of Fluid Mechanics, 155, pp. 233–265.
constant by a factor of 2 at the outer diaphragm. At idle, a radial- [8] Firouzian, M., Owen, J. M., Pincombe, J. R., and Rogers,
inflow level of 7.8% bore flow reduced the disc time constant by R. H., 1985. “Flow and heat transfer in a rotating cavity
a factor of up to 1.4. Together with previous literature, a CFD with a radial inflow of fluid Part 1: The flow structure”. In-
solution of the flow field has been used to explain how the domi- ternational Journal of Heat and Fluid Flow, 6(4), pp. 228
nance of the Coriolis forces in the rotating flow field amplify the – 234.
effect of the small radial bleed. [9] Firouzian, M., Owen, J. M., Pincombe, J., and Rogers, R.,
This amplification has two physical sources: i) the small 1986. “Flow and heat transfer in a rotating cylindrical cav-
bleed enters the cavity with significant angular momentum rela- ity with a radial inflow of fluid Part 2: Velocity, pressure
tive to the bore flow and accelerates with decreasing radius driv- and heat transfer measurements”. International Journal of
ing high swirl velocity and heat transfer coefficients; ii) the flow Heat and Fluid Flow, 7(1), pp. 21 – 27.
can only travel inwards within the viscous boundary layers, coat- [10] Farthing, P. R., Chew, J. W., and Owen, J. M., 1991. “The
ing the disc surfaces without significant mixing. use of deswirl nozzles to reduce the pressure drop in a rotat-
A simple lumped model of an HPC stage has been built and ing cavity with a radial inflow”. Journal of Turbomachin-
calibrated against engine test data. Using the results of the ex- ery, 113(1), pp. 106–114.
perimental testing, a set of radial-inflow model parameters have [11] Günther, A., Uffrecht, W., and Odenbach, S., 2012. “Local
been estimated. The model shows that a small radial-inflow bleed measurements of disk heat transfer in heated rotating cavi-
has significant potential for reductions in compressor clearance ties for several flow regimes”. Journal of Turbomachinery,
at high power conditions, and is likely to improve steady state 134(5), p. 051016.
cruise clearance through increased steady state disc tempera- [12] Greenspan, H. P., 1980. The theory of rotating fluids. Cam-
tures. bridge University Press.
[13] Owen, J. M., and Pincombe, J. R., 1979. “Vortex break-
down in a rotating cylindrical cavity”. Journal of Fluid
ACKNOWLEDGEMENTS Mechanics, 90, pp. 109–127.
The present work was performed as part of the NEWAC re- [14] Johnson, B. V., Lin, J. D., Daniels, W., and Paolillo, R.,
search program (European Commission Contract No. AIP5- CT- 2004. “Flow characteristics and stability analysis of vari-
2006-030876). The collaborative support of Work Package 3.4 able density rotating flows in compressor disk cavities”. In
partners and the financial support from the EU is gratefully ac- Proceedings of the ASME Turbo Expo 2004, Paper Num-
knowledged. Thanks is also given to Simon Davies and Dr Chris ber GT2004-54279.
Long of the University of Sussex TFMRC and to Dr Tim Scan- [15] Long, C., and Childs, P., 2007. “Shroud heat transfer mea-
lon, Leo Lewis and Jeff Dixon of Rolls-Royce plc. surements inside a heated multiple rotating cavity with ax-
ial throughflow”. International Journal of Heat and Fluid
Flow, 28(6), pp. 1405 – 1417.
REFERENCES [16] Alexiou, A., 2000. “Flow and heat transfer in gas turbine
[1] Various, 2005. The Jet Engine. Rolls-Royce PLC. H.P. compressor internal air systems”. PhD thesis, Univer-
[2] Young, C., and Snowsill, G. D., 2003. “CFD optimization sity of Sussex.
of cooling air offtake passages within rotor cavities”. Jour- [17] Alexiou, A., Hills, N. J., Long, C. A., Turner, A. B., and
nal of Turbomachinery, 125(2), pp. 380–386. Millward, J. A., 2000. “Heat transfer in HP compressor
[3] Vinod Kumar, B. G., Chew, J. C., and Hills, N. J., 2012. gas turbine internal air systems: a rotating disc-cone cavity
“Rotating flow and heat transfer in cylindrical cavities with with axial throughflow”. International Journal for Experi-
radial inflow”. In Proceedings of ASME Turbo Expo 2012, mental Heat Transfer, 13(4), pp. 299–328.
Paper Number GT2012-86770. [18] Miché, N. D. D., 2008. “Flow and heat transfer measure-
[4] Owen, J. M., and Wilson, M., 2001. “Some current research ments inside a heated multiple rotating cavity with axial
in rotating-disc systems”. Annals of the New York Academy throughflow”. PhD thesis, University of Sussex.
of Sciences, 934(1), pp. 206–221. [19] Coren, D. D., Atkins, N. R., Turner, J. R., Eastwood, D. E.,
[5] Hide, R., 1968. “On source-sink flows in a rotating fluid”. Davies, S., Childs, P. R. N., Dixon, J., and Scanlon, T. S.
Journal of Fluid Mechanics, 32, pp. 737–764. “An advanced multi-configuration stator well cooling test
[6] Wormley, D. N., 1969. “An analytical model for the incom- facility”. In Proceedings of ASME Turbo Expo 2010, Paper
pressible flow in short vortex chambers”. Journal of Basic Number GT2010-23450.
Engineering, 91(2), pp. 264–272.
[7] Owen, J. M., Pincombe, J. R., and Rogers, R. H., 1985.

14 Copyright © 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use