THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS

345 E. 47th St, New York, NY. 10017 97-GT-243

The Society shall not be responsible for statements or opinions advanced impapers or dictation at meetings of the Society or of its Divisions or
Sections, or printed In its publications. Discussion is primed only It the paper is published in an ASME Journal. Authorization to photocopy
material lot Internal, or personal use under circumstance not falling within the fair use provisions °Me Copyright Act is granted by ASME to
libraries and other users registered with the Copyright Clearance Center (CCC) Transactional Reporting Service provided that the base tee of $0.30
per page is paid directly to the CCC, 27 Congress Street Salem MA 01970. Requests for special permission or bulk reproduction should be addressed
to the ASME Technical Put:ening Department.

Downloaded from http://asmedigitalcollection.asme.org/GT/proceedings-pdf/GT1997/78705/V003T09A047/2409962/v003t09a047-97-gt-243.pdf by Shanghai Jiaotong University user on 25 December 2024

CopyrigM 0 1997 by ASME All Rights Reserved Printed in U.S.A

FLOW IN A "COVER-PLATE" PRE-SWIRL ROTOR-STATOR SYSTEM

Hasan Karabay, Jlan-Xln Chen, Robert Pilbrow

Michael Wilson and J Michael Owen
l ifl III
School of Mechanical Engineering
University of Bath
Bath BA2 7AY, UK

ABSTRACT
axial clearance between rotor and stator
This paper describes a combined theoretical, computational and static temperature
experimental study of the flow in an adiabatic pre-swirl rotor-stator To total temperature in stationary frame
system. Pre-swirl cooling air, supplied through nozzles in the stator, total temperature in rotating frame
flows radially outward, in the rotating cavity between the rotating V„ V0 , radial, tangential and axial components of velocity in
disc and a cover-plate attached to it, leaving the system through stationary frame
blade-cooling holes in the disc. An aicisymmetric elliptic solver, nondimensional radius (rib)
incorporating the Launder-Sharma low-Reynolds-number k-e axial distance from stator
turbulence model, is used to compute the flow. An WA system is swirl ratio (Vo / Or)
used to measure the tangential component of velocity. V., in the ratio of specific heats
rotating cavity of a purpose-built rotating-disc rig. For rotational turbulent energy dissipation rate
Reynolds numbers up to 1.2 x 10 6 and pre-swirl ratios up to 2.5, adiabatic effectiveness (equation 7)
agreement between the computed and measured values of V. is S. ideal adiabatic effectiveness (equation 8)
mainly very good, and the results confirm that free-vortex flow occurs turbulent flow parameter (C., Re., 46)
throughout most of the rotating cavity. Computed values of the pre- Ii dynamic viscosity
swirl effectiveness (or the nondimensional temperature difference 13 static density
between the pm-swirl and blade-cooling air) agree closely with fraction of disc-cooling air that enters blade-cooling
theoretical values obtained from a thermodynamic analysis of an passages
adiabatic system. a angular speed of rotating disc

NOMENCLATURE Subscripts

outer radius of disc blade-cooling flow

co specific heat at constant pressure disc-cooling flow
nondimensional flow rate (m/mb) edge of source region
ee difference between computed and theoretical eff effective value
adiabatic effectiveness 0 value when p p= 0
turbulent kinetic energy 13,1711 pm-swirl flow, ideal pm-swirl flow
mass flow rate of cooling air sealing flow or location of stagnation point
Ma Mach number 1 upstream of pre-swirl nozzles
radius downstream of pm-swirl nozzles
ro inner, Outer radius of annular pre-swirl chamber 2' upstream of blade-cooling passages
Re, rotational Reynolds number (pf2b 240 2 inside blade-cooling passages

Presented at the International Gas Turbine 8s Aeroengine Congress 8z Exhibition

Orlando, Florida — June 2–June 5, 1997
This paper has been accepted for publication in the Transactions of the ASME
Discussion of it will be accepted at ASME Headquarters until September 30,1997
 outer seal th 5 outer shroud cover—plate outer shroud

lilp
t sealing air sealing air
A
outer shroud (rub)

thb
pre—swirl ltlb
air blade—cooling .8 blade—cooling

Downloaded from http://asmedigitalcollection.asme.org/GT/proceedings-pdf/GT1997/78705/V003T09A047/2409962/v003t09a047-97-gt-243.pdf by Shanghai Jiaotong University user on 25 December 2024

air stator air

stationary
rotating
rotor—stator
system
04 rotating
cavity
inner shroud
inner shroud

inner seal rotating

inner seal disc

stator annular
din pre—swirl
rotating pre—swirl chamber
disc air

rotating
inner cylinder
to rp ri

disc—cooling
air
ihd

a) direct—transfer system b) cover—plate system

Fig. 1 Schematic diagram of pre—swirl systems
— primary flow - - - secondary flow • rotating .."/ stationary

1 INTRODUCTION inviscid rotating fluid (which is not a free vortex) between two
nonentraining Elcman-type layers.
Fig 1 shows a schematic diagram of two pre-swirl systems that are
commonly used to supply cooling air to gas-turbine blades. In the so- The radial extent of the source region depends principally on the pre-
called direct-transfer (or large-radius) system, the pre-swirl nozzles swirl ratio, Pp, of the incoming air and on the turbulent flow
are located in a stationary casing, or stator, at a radial position, r o, parameter, kr, where Bp and ?vr are defined in the Nomenclature. For
approximately equal to the inlet radius of the blade-cooling passages, sufficiently large values of AT, the source region fills the entire
rb. In the cover-plate (or small-radius) system, the pre-swirl nozzles cavity; this is usually the case for the pre-swirl cooling systems
are located in the stator radially inward of the blade-cooling considered here. Another feature of rotating cavities is that the flow
passages, and the air flows outward in the clearance between the in the boundary layers is radially outward when the inviscid core
turbine disc and a cover-plate attached to it. rotates slower than the discs, and it is radially inward when the core
rotates faster. There is, therefore, a stagnation point in the boundary
By swirling the cooling air in the direction of rotation of the disc, the layer at the radial location where V o = ar in the inviscid core.
temperature of the air entering the blade-cooling passages is reduced.
In both systems, air is used to remove the windage heating from the Computational and experimental studies in pre-swirl systems, with
turbine disc and to provide a sealing flow to reduce the ingress of hot and without heat transfer, have been conducted by a number of
mainstream gas at the periphery of the system. For the direct- research workers, and the reader is referred to Meierhofer and
transfer system, most, or all, of the disc-cooling air and some of the Franklin (1981), El-Oun and Owen (1989), Chen, Owen and Wilson
ingested mainstream gas ends up in the blade-cooling passages. In (1993a,b), Wilson, Pilbrow and Owen (1995) and Popp,
the cover-plate system, an inner seal is used to prevent this Zimmermann and Kutz (1996). Details of the flow and heat transfer
undesirable contamination of blade-coolant with either disc-cooling in rotor-stator systems in general, and the subject of ingress in
air or ingested gas. particular, are given by Owen and Rogers (1989).

In the cover-plate system, a free vortex is created as the air flows The work described here is concerned principally with an adiabatic
radially outward in the rotating cavity between the disc and the pre-swirl cover-plate system. In Section 2, the performance of an
cover-plate. For radial outflow in rotating cavities (see Owen and adiabatic system is analysed, and expressions are derived for the pre-
Rogers 1995), there is usually a source region near the inlet, where swirl effectiveness in both direct-transfer and cover-plate systems.
inviscid free-vortex flow occurs between two entraining boundary The computational method and the experimental apparatus are
layers; radially outward of the source region, there is a core of described in sections 3 and 4, respectively, and the computational

2
 2' 2 there is computational and experimental evidence (see El-Oun and
mb Owen 1989 and .et al 1993a, b) that, even when there are discrete
blade-cooling holes rather than the annular slot considered here, most
or all of the disc-coolin g air ends up in the blade-coolin g passages,
and some of the pm-swirl air leaves the system throu gh the peripheral
rotating seals.
blade—cooling
passages
It is convenient to define O as the ratio of the disc-coolin g air that

Downloaded from http://asmedigitalcollection.asme.org/GT/proceedings-pdf/GT1997/78705/V003T09A047/2409962/v003t09a047-97-gt-243.pdf by Shanghai Jiaotong University user on 25 December 2024

enters the blade-coolin g passages to the total flow that enters these
passages. For the cover-plate s ystem, = 0 ; for the direct-transfer
system it is assumed that = ma/nib. If the disc-cooling air enters the
r2 system with zero swirl, equation (2) becomes
stationary
Fe-Swill nozzles

% (T02 - To° = C22r22 [1- a -rr2

i I (3)

(It should be noted that the effect of the disc-coolin g air could be
reduced if it were swirled at inlet to the s ystem.)

V
Also,
-e-tri-c .cOr;;;;;;;; f. t7;;;;;;;I-;;;;;;
TT;.--Aci-s y-mm y -1 2
C p Taz = C p T2 (I +— Ma2 ) (4)
2

results and the experimental velocity measurements are discussed in

= Cp T2 + 1/2 (V12 + + Vz2)2 (S)
Section 5. Conclusions are presented in Section 6.
It is convenient to define T a as the total temperature in the rotatin g
2 ANALYSIS OF AN ADIABATIC PRE-SWIRL
frame, which is the value that would be measured b y a total-
SYSTEM
temperature probe inside a blade-cooling passa ge; it is this
temperature that controls the heat transfer from the blade to the air.
2.1 Pre-swirl effectiveness
By definition
Fig 2 shows the control volume used in the anal ysis. Stations I and
I' are respectivel y immediatel y upstream and downstream of the pm-
Cp Tt2 = Cp T2 1/2V 2 +1/2 Vi22 ,2 (6)
swirl nozzles, and 2' and 2 are respectively immediately upstream of
and inside the blade-cooling passages; rr = r i and r2- = r2. From I to where, for normal blade-coolin g passages, V a = 0.
2' the flow is taken to be isentropic, and from 2' to 2 work is done to •
adjust the tangential speed of the air, V or, to that of the disc, V. It is also convenient to define the nondimensional pm-swirl
effectiveness, e, as
From the first law of thermod ynamics for an adiabatic open s ystem,
the rate of work done on the air is equal to the rate of increase of its
e , 01 2
total enthalp y. Also, as the air moves from stations to 2, the (7)
1 2
moment exerted by the rotatin g surfaces equals the rate of change of T ie r2
angular momentum of the air ; the product of this moment and the
angular speed of the disc is equal to the rate of work done on the air. which is the nondimensional difference in the total temperature
Hence, between the stationar y pm-swirl nozzles and the rotatin g blade-
cooling passages. A positive value of e indicates that the pm-swirl
C0 (T02 Toy) = Cl (r2V02- (1) system is effective in reducin g the total temperature of the blade-
cooling air.
or, as Tor = Tol and Vo = flr2,
Using equations (3), (5) and (6), it follows that the ideal
effectiveness, et, is given by
.- 2 ]
cri gat - To0 = 0 2r22 [i - p.(1rj (2)
2
Er= 2(1- CB p.(2-1-) -1 (8)
ri
where gip'= V ifr i. Orl is the inlet pre-swirl ratio.

It should be noted that equation (2) is valid for viscid and inviscid For the direct-transfer s ystem, where tla = me/al b and r i = r2,
flow providing all the blade-coolin g air ori ginates from the pre-swirl
nozzles. This condition should alwa ys be true for the cover-plate
es= 2(1--
Ind )13 0. -1 (9)
system, but it will not be true for a direct-transfer s ystem where the me
disc-coolin g and blade-coolin g flows mix, as shown in Fig I. In fact,

3
For the cover-plate system, where 0= 0, 13 1? > 1, there will be a radius, r = r„ say, where V./Or = 1. It
follows from equation (12) that

Er= 2/3 p.( 1 ) —1 (1 0) a si rj

r, = pv
7
For r > r1, air is entrained into the outflowing boundary layers until,
If op, (rj/r2)2 is kept the same for both systems (that is, the inlet at r = re , say, the entire mass flow rate, mb, has been entrained. For
angular momentum is the same), and if there is no disc-cooling flow, r > r, nonentraining Elunan-type layers are formed and angular
then the effectiveness of both systems will be the same. Equation momentum in the core is no longer conserved; the radius r e therefore

Downloaded from http://asmedigitalcollection.asme.org/GT/proceedings-pdf/GT1997/78705/V003T09A047/2409962/v003t09a047-97-gt-243.pdf by Shanghai Jiaotong University user on 25 December 2024

(8) holds for all adiabatic systems whether the flow is laminar, corresponds to the edge of the source region.
turbulent or inviscid and whether the flow is compressible or
incompressible. It is useful to estimate the value of r e which, according to Owen and
Rogers (1995), will depend on the turbulent flow parameter, Xtb and
El-Oun and Owen (1989) used an "unmixed theory" (where 4) = 0) for on 13v. They proposed the following theoretical correlation for the
the adiabatic effectiveness of a direct-transfer pre-swirl system nondimensional radial extent of the source region ic e :
(where ri = r2). Their theory was based on the Reynolds analogy in
which the work was transferred by shear stresses in the boundary 2.35
( ) 1"
layer on the disc. If a fluid is brought to rest, or to the speed of a — =1 (14)
moving surface, by viscous shear then the adiabatic surface IX x )
temperature is related to the total temperature of the fluid by a
recovery factor, R, say. For the case where R = I, when viscous where xe.,, is the value of Ice for the case where the flow enters with
recovery and isentropic recovery are identical, equation (8) becomes zero swirl (13v = 0). The value of x e.„, is based on the free-disc
identical to the equation derived by El-Oun and Owen. entrainment rate, where

Although pressure losses do not affect the pre-swirl effectiveness XC,o = I.37SATb 5113 (15)
directly, in practice they will reduce the amount of swirl that can be
imparted to the air when the overall pressure drop in the system is for the case where the flow is entrained equally into the boundary
fixed by engine conditions. The pressure loss in the pre-swirl nozzles layers on both discs. The value of x, in equation (14) is given from
is expected to increase with the square of the tangential velocity at equation (13) where
inlet, and this loss will therefore be much larger in the cover-plate
system, thereby reducing the inlet swirl ratio. However, as shown s 2 = f3p, xr,2 (16)
above, when disc-cooling air is introduced, the effectiveness of the
direct-transfer system will be reduced. It is therefore unclear which Equation (14) is only valid for x. < 1 and for x e < I. It is easy to
of these two systems has the better adiabatic performance in practice. show that xe = 1 when

Whatever the relative adiabatic performance, the effects of heat = 0.437(1 - (flv x p2) 1.18 ) 1 .66 (17)
transfer and hot gas ingress are likely to increase significantly the
temperature of the blade-cooling air in gas turbines. However, the providing Ap• xp2 < I. For Xtb greater than this value, the source
adiabatic effectiveness provides a datum from which to measure the region will fill the entire cavity and free-vortex flow will occur
performance of all practical pre-swirl systems. between the pre-swirl nozzles and blade-cooling passages. If
fivxp2 > I, there will be radial inflow in the boundary layers, with
2.2 Source region in a rotating cavity with radial outflow associated recirculating flow throughout the cavity. For most engine-
In a rotating cavity with a radial outflow of air, there is a source operating conditions, it is expected that Pp• Xp 2 < 1 and that Atb will
region where angular momentum is conserved and the incoming flow be greater than the value given by equation (17).
is entrained into boundary layers on the two discs. If the incoming
flow is rotating faster than the disc (0 > 1) then there will be radial
inflow in the boundary layers; radial outflow in the boundary layers 3 COMPUTIONAL MODEL
will only occur at the larger radii where pc 1. For free-vortex flow,
rVp is constant, and if The steady-state, axisymmetric finite-volume solver used in this work
is the same as that described by Wilson, Pilbrow and Owen (1995)
A V for a related study of a direct-transfer pre-swirl rotor-stator system.
.
P fir] The Launder-Sharma low-Reynolds-number k-e turbulence model
was used to close the coupled system of Reynolds-averaged Navier-
it follows that Stokes and energy equations, and the flow was assumed to be
incompressible. Turbulent heat transfer was represented using a
turbulent Prandtl number Pit equal to 0.9. A staggered grid was
V • (12) used with the axial and radial velocity components stored mid-way
=13 { 11)2
r =15 between the grid points, where the other solution variables were
located (pressure, tangential velocity, turbulence kinetic energy and
Equation (12) will be referred to as the equation for an "ideal free dissipation rate, and total enthalpy). The cover-plate and inner
vortex". shroud (Fig. lb) were represented by block obstructions within the

4
computational grid and the equations were solved using the r = 142 mm; at this radius, the spacers should cause the minimum of
SIMPLEC pressure-correction algorithm. disturbance to the designed flow. When Pp < 2.5, the drag exerted
by the spacers will tend to increase the angular momentum of the air
The inlet nozzles and blade-cooling holes of the experimental rig for r > 142 mm)
were represented in the axisymmetric model by equivalent-area
annular slots on the stator and rotor, with centrelines at r = r p and The mass flow rates of the pre-swirl, blade-cooling and sealing flow,
r = rb respectively. The axial velocity V, of the pre-swirl air was mb and in„ respectively, could be independently controlled, and
assumed uniform at inlet and deduced from the prescribed mass flow the flow rates were measured, with an uncertainty of t 3%, by orifice

Downloaded from http://asmedigitalcollection.asme.org/GT/proceedings-pdf/GT1997/78705/V003T09A047/2409962/v003t09a047-97-gt-243.pdf by Shanghai Jiaotong University user on 25 December 2024

rate mp. Similarly, the axial velocity at the blade-cooling slot and the plates made to British Standards (8S1042). The rotor assembly
radial velocity at the outer seal were calculated from known mass could be rotated up to 7000 rev/min by a variable-speed electric
flow rates mb and in, respectively; global mass balance was achieved motor, and the speed could be measured with an uncertainty of t 1
by ensuring that mp = ins + m,. The inlet tangential velocity V ci, was rev/min. (For the tests conducted on this rig, a speed of 7000 rev/min
fixed to give the required swirl ratio B. and Neumann (zero normal corresponds to Re. 2 x 10 6.)
derivative) boundary conditions for Vo were used at the two outlets.
The remaining velocity components at flow boundaries were taken to It should be pointed out that, for the results presented below, the
be zero, and no-slip conditions were applied at all solid surfaces. For ideal swirl ratio, Pp', was used (see Section 2). This was based on
the computations of the adiabatic effectiveness, the static temperature the measured mass flow rate, m r„ and the continuity equation:
was prescribed at inlet, and Neumann conditions were applied knowing the angle and diameter of the pre-swirl nozzles, giv could be
elsewhere. calculated.

The Launder-Sharma turbulence model required a very fine grid near The WA measurements were made using a TSI back-scatter optical
the boundaries, with y+ < 0.5 for the near-wall grid nodes, and the system with a Bragg-cell frequency shift and a single-channel IFA-
grid-spacing increased geometrically away from walls (including the 750 burst correlator. A Spectra-Physics 4W argon-ion laser was
cover-plate and shroud) with expansion factors of about 11. In total, used, and for the tests reported here up to 350 mW on the green line
a 223 by 223 axial by radial grid was used, with eight points covering (514.5 nm) was used to measure the tangential component of
each of the inlet and blade-cooling slots. About 55 axial grid nodes velocity, Vs, in, or near, the mid-plane of the rotating cavity for
were within the cover-plate, with the remaining points divided 0.57 < x < I.
equally between the rotor-stator system and the rotating cavity: about
70 radial grid points covered the annular pre-swirl chamber (Fig. lb). The processor could handle Doppler frequencies up to 90 MHz with
Convergence of the iterative method was improved using the Gosman signal-to-noise ratios as low as -5dB. The beam spacing was 50 mm,
distributive damping term and a fixed V-cycle multigrid algorithm and a converging lens with a focal length of 250 mm produced an
(Vaughan et al, 1989). Computation times were typically around 12 optical probe volume 340tun long and 34pin diameter, The inlet air
hours on a Silicon Graphics R10000 processor. was seeded with micron-sized oil particles produced by a Tsl 9306
six-jet atomiser. From these measurements, and those made using
this LDA equipment in other rotating-disc rigs (see, for example,
4 EXPERIMENTAL APPARATUS Gan er a! 1996), the uncertainty in the measured values of Vo / fir is
expected to be approximately t 1%.
A schematic diagram of the rig is shown in Fig I. The direct-
transfer rig (from which the present version was adapted) is Velocity measurements were made over the following ranges of
described by Wilson et al (1995), and so only the salient details of nondimensional parameters :
the cover-plate rig are described here.
5.3 x 105 < Re* <1.25 x 106
The outer radius of the system, b, was approximately 207mm, and 6500 < <26000
the radial location of the pre-swirl nozzles and blade-cooling holes 0.173 < Xtb <0.363
were rp = 90 mm and rb = 200 mm. There were 19 pre-swirl nozzles 1.18 <14, < 2.5
of 7.92 mm diameter, angled at 20 6 to the tangential direction, and
60 blade-cooling holes of 7.7 mm diameter, with their axes normal to
5 COMPUTATION AND MEASUREMENTS
the disc. The axial spacing between the cover-plate, which was 5mm
thick, and the stator was 10 mm, and between the cover-plate and the
rotor disc the spacing was also 10 mm. The inner and outer radii of
the annular pre-swirl chamber were r i = 80 mm and r0 = 100 mm, 5.1 Flow structure
and air entered the rotating cavity through the clearance at the centre Figs 3 and 4 show the computed streamlines and contours of V, / fir
of the cover-plate. for Xtb = 0.178 and By = 1.27 and for Xtb = 0.229 and Illp• = 2.51,
respectively. The geometry corresponds to the experimental rig
The cover-plate was made from transparent polycarbonate, and there described in Section 4.
was a window in the stator to provide optical access for WA
measurements in the rotating cavity between the cover-plate and the Values of Rep, XT,b and Pp are given in the legends of Fig. 3
disc. Axial restraint of the cover-plate was provided by six and 4; the nondimensional flow rate of the disc-cooling air, Cw.d, can
cylindrical spacers of 12 mm diameter bonded to the cover-plate and be found from the difference between C,,, p and C.,b. The flow
to the disc at a mean radius of 142 mm. (For an inlet swirl ratio of structure in the rotating cavity is determined by the magnitude of
= 2.5, the design condition, free vortex flow would make D = I at Res, XT,b and P p; the magnitude of C w4 has an insignificant effect on

5
 -0-

Downloaded from http://asmedigitalcollection.asme.org/GT/proceedings-pdf/GT1997/78705/V003T09A047/2409962/v003t09a047-97-gt-243.pdf by Shanghai Jiaotong University user on 25 December 2024

0.4 0.4
00 025 0.5 0.75 1.0 00 025 0.5 0.75 10 00 0.25 0.5 0.75 1.0 00 0.25 0.5 0.75 10
z/S LS z/S
a) streamlines b) contours of Vstar a) streamlines b) contours of Vs/fir
Fig. 3 Computed flow structure for cover—plate system: Fig. 4 Computed flow structure for cover—plate system:
Res = 1.09x106, Cw,b = 1.21x104, Res = 0.553x106, Cw,b = 0.902x1($,
Cw,p = 1.55x104, Arr,b = 0.178, Pp' = 1.27 Cw,p = 1.55x104 , A.tb = 0.229, f3p' = 2.51

the flow in the rotating cavity although it obviously has a strong Fig 4a shows the computed streamlines for km, = 0.229 and
effect on that in the rotor-stator system. Pp' = 2.51. In the rotating cavity, the separation zone is larger than
that shown in Fig 3a and the source region extends throughout most
Fig 3a shows the computed streamlines in three regions : the rotor- of the cavity. (For (3y = 2.51, equation (17) gives XT, b = 0.173; for
stator system (on the left-hand side of the figure); the rotating-cavity kr.t, > 0.173 the source region should fill the cavity, as shown in
(on the right-hand side); and the pre-swirl annulus (near the bottom). Fig 4a.)
In the rotor-stator system, there is a source region for x < 0.6, radially
outward of which there is the expected radial outflow on the rotating Fig 4b shows that Vo/LIr = 1 at x r4 0.65, which coincides with the
cover-plate and inflow on the stator, with a core of rotating fluid edge of the separation zone on the cover-plate. For x > 0.65, there is
between the two boundary layers. The flow in rotor-stator systems is little axial variation of Vo outside the boundary layers on the discs.
well understood and will not be discussed further here.
Fig 5 shows the variation of Vo/L2r with 12 for three sets of A.r.b and
The pre-swirl flow enters the system at z/s = 0 and x = 0.44 with a 130,. It is instructive to plot results against 12 as free-vortex flow can
swirl ratio of I3 0. =1.27. There are recirculation zones on either side be identified by straight lines passing through the origin at 1 2 =O.
of the swirling jet with a stagnation point on the outer wall of the The "ideal free vortex", which corresponds to equation (12), is
pre-swirl annulus (x = 0.49, z/s -s 0.42). For z/s < 0.42, there is an plotted from 12 = 0 to 12 = 5.24, the latter value being the radial
axial flow in the outer boundary layer towards the pre-swirl nozzles. location of the pre-swirl nozzles. The experimental measurements
It is this flow that enters the rotor-stator system through the clearance were obtained from nine tests carried out over many days. The
between the outer annular wall and the stator. nondimensional parameters varied slightly from test to test, but the
nominal values indicated on the figure are accurate to around ± 2%.
Most of the pre-swirl flow enters the rotating cavity, impinging on
the rotating disc and separating from the cover-plate: the separation The "computed free vortex" in Fig 5 is a free-vortex curve drawn
zone extends to x s 0.57. For x < 0.75, there is a source region, through the computed values of Valr at a single radius; a value of
radially outward of which the flow divides into two nonentraining 1 2 = 2.2 was arbitrarily chosen. The "computed free vortex" shows
Elcman-type layers. (For Xtb = 0.178 and Po, = 1.27, equation (14) the radial extent to which the results conform to a free vortex. It is
gives xe = 0.87, which overestimates the size of the source region also possible to compute the "effective pre-swirl ratio-, Pp.er,
shown in Fig 3a.) The flow leaves the rotating cavity through the corresponding to the value of the computed free vortex at 1 2 = 5.24.
annular slot in the rotating disc, at x = 0.97, half the flow coming Also shown on Fig 5 are the locations of the centres of the blade-
from the boundary layer on the disc and the other half from the cooling slot, at 1 2 = 1.06, and the cover-plate supports, at 1 2 = 2.13.
boundary layer on the cover-plate via the peripheral shroud.
Referring to Fig 5a, where 13 0, = 1.25 and krj, = 0.18, the
Fig 3b shows the contours of V0/fir for this case. For x > 0.57, computations show free-vortex flow from 1.8 < x < 4
radially outward of the separation zone, there is little axial variation (0.5 < x < 0.75) with 130.dr Pr. The shear near x = 1 is associated
of V0 outside the boundary layers on the two discs. with the boundary layer on the shroud, and the departure from the

6
 The experimental data shown in Fig 5a are consistent with a free
vortex at the larger values of 1 2 and illustrate the shear in the
z/840.13 boundary layer near x = I. Downstream of the cover-plate supports,
the measurements show an increase in Vo above the ideal free-vortex
curve. In this region, where Vo/flr r 0.5, the supports have evidently
imparted angular momentum to the fluid, as suggested in Section 4.

In Fig 5b, where By = 2.5 and 43, = 0.23, the computations and

Downloaded from http://asmedigitalcollection.asme.org/GT/proceedings-pdf/GT1997/78705/V003T09A047/2409962/v003t09a047-97-gt-243.pdf by Shanghai Jiaotong University user on 25 December 2024

measurements lie on a free-vortex curve that is significantly lower
than the ideal one. For this case, where Vo/fir a 0.9 near the cover-
plate supports, the supports appear to have no significant effect on
the flow. For these values of 13y and Xtb, the source region fills the
0 1 2 3 4 5 6 a .* 7 entire cavity, and the computed and measured values of Vo/fir show
a) Bpi = 1.25, Atb = 0.18, Res = 1.23 x 106 that free-vortex behaviour occurs up to the edge of the boundary layer
on the shroud at x = 1.
3.0 v,/fit
Fig Sc, where By = 2.5 and Xtb = 0.36, shows computed and
2.5
7./S40.8 measured results similar to those in Fig 5b. These results illustrate
the fact that 13y is the dominant parameter as far as V,/Or is
2.0
concerned, although km does have an effect on the size of the source
region. The measurements and computations for results that are not
1.5
shown here suggest that, for a fixed value of Am, the ratio of
1.0 Bo.cu / By decreases significantly as By increases; for a fixed value
of By, the ratio increases as Am increases.
0.5

0.0 5.2 Pre swirl effectiveness

-

0 1 2 3 4 Fig 6 shows a comparison between the computed values of e, using

b) 2.5, Xtb = 0.23, Res = 0.55 x 106 the definition in equation (7), and the theoretical curve for Er, using
equation (10). The geometry was based on the rig described in
3.0 ti./Or section 4 where rdr2= 0.45.

2.5
4/5-0.8 It should be pointed out that the authors were unable to make any
reliable experimental measurement of e for the adiabatic case.
2.0
Although total temperature probes were fitted in the rig to measure
To, and Ta, it was not possible to conduct experiments under
1.5
adiabatic conditions. Unlike the rig used by El-Oun and Owen
(1989), the one used here was designed for heat-transfer
I.0
experiments. Windage heating from the heater assembly and heat
input from the bearings created heat transfer from the disc to the
0.5
cooling air even when no external heating was used. Heat transfer
results, and the measurement of the temperature of the blade-cooling
0.0
1 4 air, will be the subject of a subsequent paper.
0 2 3 5 6
ON, 2.5, Xtb = 0.36, Re. b = 0.55 x 106
Fig 6 shows good agreement between the computed and theoretical
variation of e with By in the range By = 0 to 4.60. It is convenient
Fig 5 Comparison between computed and measured variation of
to define a nondimensional error, ee, based on the difference
V4,/Or with 12
between 0' and S. From equation (7) it follows that
o experimental data;
computation;
e r go -To
ideal free vortex; (18)
1 2 2
computed free vortex; -2 0 1.2
centre of blade-cooling slot;
centre of cover-plate support where tat is the theoretical value of the blade-cooling temperature,
based on equation (10), and T,,2 is the computed value. For 15
free-vortex curve at 1 2 s1.8 corresponds to the edge of the source computed values, ee range from 10 4 (for By = 2.5) to 0.019 (for
region, which was discussed above. (It should be noted that angular Bp' = 4.6). As the theoretical value given in equation (10) is exact,
momentum is not conserved outside the source region in the core ee represents the magnitude of the numerical errors in the computed
flow between Orman-type layers - see Owen and Rogers, 1995.) values of e.

7
 thermodynamic analysis of an adiabatic system. The analysis shows
1.0 that ei depends only on the parameter pp, (rs/rb)2, and for rs/rb = 0.45
and 0 513p, 5 4.6 the error between et and 0 was less than 2% of the
"dynamic temperature", fl 2rb2t2cp.
0.5
ACKNOWLEDGEMENTS

The authors thank the Engineering and Physical Sciences Research

Downloaded from http://asmedigitalcollection.asme.org/GT/proceedings-pdf/GT1997/78705/V003T09A047/2409962/v003t09a047-97-gt-243.pdf by Shanghai Jiaotong University user on 25 December 2024

0. 0 Council and European Gas Turbines Ltd for funding the research
described in this paper, and the Turkish Government and Kocaeli
University for providing the financial support for Hasan Karabay. We
also wish to thank the reviewers for their constructive comments.
-0.5
— equation (10) (rir, .4 0.45)
o computed results REFERENCES

Chen, J., Owen, J.M. and Wilson, M. 1993a. Parallel-computing

techniques applied to rotor-stator systems: fluid dynamics
Fig. 6 Comparison between computed and theoretical variation of computations, in Numerical Methods in Laminar and Turbulent
e with pp
Flow, g, 899 - 911 (Pineridge Press, Swansea).

Chen, J., Owen, J.M. and Wilson, M, 19936. Parallel-computing

For the experimental rig, the above conditions correspond to techniques applied to rotor-stator systems: thermal computations, in
f2r2 < 84 m/s, and the "dynamic temperature" (f2 2r22/2cp) is less Numerical Methods in Thermal Problems, g, 1212 - 1226,
than 3.5 °C; a value of es = 0.019 therefore corresponds to 0.07 °C in (Pineridge Press, Swansea).
the rig. In an engine, where dynamic temperatures can be in excess
of 50° C, a value of es = 0.019 would correspond to around 1 °C. For El-Gun, Z and Owen, J.M., 1989. Pre-swirl blade-cooling
practical purposes, the agreement between the computed and effectiveness in an adiabatic rotor-stator system. J. Turbomachinery,
theoretical values of 0 can be regarded as satisfactory. LI 522 - 529.

In engines, the effectiveness of cover-plate systems will be reduced Gan, X., Mirraree, L, Owen, J.M., Rees, D.A.S. and Wilson, M., 1996.
by heat transfer from the turbine disc to the cooling air. For direct- Flow in a rotating cavity with a peripheral inlet and outlet of cooling air.
transfer systems, the ingestion of disc-cooling air and mainstream gas ASME Int. Gas Turbine and Aero Engine Congress, Birmingham, UK,
into the blade-cooling passages is likely to have an even greater 10-13 June 1996. ASME Paper No 96-GT-309.
effect on the temperature of the blade-cooling air. However, even
when heat transfer is significant, the adiabatic effectiveness provides Meierhofer, B. and Franklin, C.J. , 1981. An investigation of a
a useful datum from which to measure pit-swirl performance. preswirled cooling airflow to a gas turbine disk by measuring the air
temperature in the rotating channels. ASME Int. Gas Turbine Coat,
6 CONCLUSIONS Houston, Paper No 81-GT-132.

Computations, made using an elliptic solver incorporating the Owen, J.M. and Rogers, R.H., 1989. Flow and heat transfer in
Launder-Shanna low-Reynolds-number turbulence model, and rotating disc systems: Vol. 1: Rotor-stator systems. Research Studies
velocity measurements, made using an WA system in a purpose- Press, Taunton, UK and John Wiley, New York, USA.
built experimental rig, have been used to study the flow in an
adiabatic cover-plate system. Owen, J.M. and Rogers, R.H., 1995. Flow and heat transfer in
rotating disc systems: Vol. 2: Rotating cavities. Research Studies
The computations confirm that the flow between the cover-plate and Press, Taunton, UK and John Wiley, New York, USA.
the rotating disc is similar to that in a rotating cavity with a radial
outflow of air. For sufficiently large values of km, the source region Popp, 0., Zimmermann, H. and Kutz, J., 1996. CFD analysis of
fills most of the cavity and, outside the boundary layers, the flow coverplate receiver flow. ASME Int. Gas Turbine and Aeroengine
behaves as a free vortex. The agreement between the measured and Cong., Birmingham, UK. Paper No. 96-GT-357.
computed values of V,/(2r in the rotating cavity is mainly very good.
Supports, attached to the cover-plate on the experimental rig, could Vaughan, CM., Gilham, S. and Chew, J.W., 1989. Numerical
disturb the flow when Vs < Or in the vicinity of the supports; at the solutions of rotating disc flows using a non-linear multigrid
design pre-swirl ratio (13v = 2.5), the effect of the supports was algorithm. Proc. 6th Int. Conf. Num. Meth. Laminar Turbulent
insignificant. The computed values of Vs/fir approximated to a free Flow, pp 66-73, (Pineridge Press, Swansea).
vortex with an effective pre-swirl ratio of Il ya. For > 1.25, the
ratio of (31,41113s. is less than unity and it decreases as 13 increases. Wilson, M., Pilbrow, R.G. and Owen, J.M., 1995. Flow and heat
transfer in a pre-swirl rotor-stator system. ASME hit. Gas Turbine
Values of the computed nondimensional pre-swirl effectiveness, 0, and Aeroengine Congress, Houston, June 1995. Paper No 95-GT-
were compared with theoretical values, et, obtained from a 239.