THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS 89-GT-184

345 E. 47 St., New York, N.Y. 10017

rr. The Society shall not be responsible for statements or opinions advanced in papers or in dis-
C cussion at meetings of the Society or of its Divisions or Sections, or printed in its publications.
m Discussion is printed only if the paper is published in an ASME Journal. Papers are available
],[ ® from ASME for fifteen months after the meeting.
Printed in USA.
Copyright © 1989 by ASME

The Use of De-Swirl Nozzles to Reduce the Pressure

Drop in a Rotating Cavity with a Radial Inflow
P. R. FARTHING', J. W. CHEW 2 and J. M. OWEN'
'Thermo-Fluid Mechanics Research Centre 2Rolls Royce plc
School of Engineering and Applied Sciences Moor Lane
University of Sussex P. 0. Box 31
Brighton, BN1 90T Derby, DE2 8BJ
UK UK

ABSTRACT m inlet mass flow rate (negative for

radial inflow)
A combined theoretical and experimental study ms mass flow rate for the shroud boundary
is described in which de-swirl nozzles were used to layer
reduce the radial pressure drop in a rotating cavity N number of de-swirl nozzles
with a radial inflow of air. The nozzles, which p static pressure
were attached to the outer part of the cavity, were q speed, relative to disc, of fluid
angled such that the angular speed of the air at leaving de-swirl nozzles
inlet could be in the opposite direction to that of QL leakage flow rate
the cavity. Solutions of the momentum-integral r radial coordinate
equations were used to predict the resulting radial re radius of inner boundary of source
distributions of pressure throughout the cavity. region
Flow visualization was used to confirm the flow rs radius of stagnation point on disc
structure, and transducers attached to one of the Rer radial Reynolds number = C w /2nx
rotating discs in the cavity were used to measure Re cD rotational Reynolds number = S^b 2 ;v
the radial pressure distributions. Results are s axial spacing between discs
presented for 'swirl fractions' (that is, the ratio u radial component of velocity
of the angular speed of the air leaving the nozzles uo function of r in eqn (1)
to that of the cavity) in the range -0.4 to + 0.9, v tangential component of velocity
and for 0.01 < IC w I Rep ° • e < 0.5, where Cw and Re m relative to disc
are the nondinrensional flow rate and rotational v tangential component of velocity,
Reynolds number, respectively. The measured relative to disc, outside the boundary
pressures are in good agreement with the predicted layer
values, and the pressure drop across the cavity can vm mean tangential velocity, relative to
be significantly less than that associated with disc, of fluid in the shroud boundary
solid-body rotation. The flow rate produced by the layer
pressure drop across the cavity is not unique: x nondimensional radius = r/b
there are up to three possible values of flow rate z axial coordinate
for any given value of pressure drop. S boundary-layer thickness
oP s pressure drop across experimental rig
e angle of de-swirl nozzle to tangential
direction
NOMENCLATURE
aL laminar flow parameter = C w /Rep ° - 5
)-r turbulent flow parameter = C w /Re ii ° - 8
a inner radius of cavity
P absolute viscosity
b outer radius of cavity (inner radius v kinematic viscosity = u/P
of nozzle ring) p density
c inlet swirl
fraction T
effectivee inlet swirl fraction r,o^ T m,o radial and tangential shear-stress
tiff components on disc
C nondimensional pressure difference
p 2 4) angular coordinate
( ssure coef 0 angular velocity of cavity
C p c pressu re coefficient
_ ( pb - pref ) / P 02 b`
C w mass flow coefficient = m/ub
Subscript s
d outlet diameter of de-swirl nozzles
a value at inner radius of cavity
F friction factor
b value at outer radius of cavity
I constant in momentum-integral ref reference value
equations, eqn (7)

Presented at the Gas Turbine and Aeroengine Congress and Exposition—June 4-8, 1989—Toronto, Ontario, Canada
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, 1989

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/83350/ on 03/04/2017 Terms of Use: http://www.asme.org/abo

 0

1. INTRODUCTION In both cases the models included solutions of the

momentum-integral equations for the boundary layers
In a gas-turbine engine, cooling air for the and agreement with experiment was claimed. It should
turbine has to be bled from the compressor, and it be noted, however, that the model of Alberga et al
is sometimes required that the air should flow includes an empirical constant, and it is not
radially inwards through the cavity between two entirely clear how much tuning of this constant was
corotating compressor discs. Various devices may be required.
used to reduce the pressure drop across the cavity, The application of integral methods to this
and an investigation of the effect of fins attached problem was also discussed by Chew and Snell (1988).
to one of the discs was reported recently (Chew et In their paper a model is described for radial
al 1988). In this paper, the use of de-swirl inflow through a rotating cavity with the inlet
nozzles to control the pressure drop is considered. angular velocity of the air equal of that of the
Although de-swirler devices have been considered disc. The model is similar to that of Owen et al,
previously for application in gas-turbine engines but incorporates a more satisfactory treatment of
(see, for example, Scheper 1962, Brown and Manente the entraining boundary layer in the source region.
1977, Kraftwerk Union Aktiengesellschaft 1977), a Predictions from the model were shown to be in good
literature search has failed to reveal any detailed agreement with the measurements of Firouzian et al
study of their operation. and in fair agreement with the data of Alberga et
The de-swirl nozzles may be fitted to the al. Comparison of Chew and Snell's predictions with
peripheral shroud which rotates with the discs, and Graber et al's measurements of the pressure drop
the nozzles are angled so as to produce a tangential across the cavity also shows good agreement.
component of velocity of the air in the opposite In this paper, the integral method is extended
direction to the rotation of the discs. Thus the to include the effects of de-swirl as described in
air will enter the cavity with a lower angular the ner•t section. The experimental apparatus and
velocity than the discs themselves, and may even be comparison between measurements and theory are
rotating in the opposite direction to the discs. In presented in Sections 3, 4 and 5.
the present study the performance of such a system
has been investigated both theoretically and 2. THEORY
experimentally.
The flow structure in a rotating cavity with 2.1 Outline of the model
radial inflow has been established from The flow is treated as steady and axisymmetric
experimental and theoretical studies by a number of with the flow structure as shown in Fig.1 and
research workers (Firouzian et al 1985, Owen et al described above. It is also assumed that the flow
1985, Chew 1987 and Morse 1987). Referring to is rotationally dominated so that the radial and
Fig.1, the flow comprises a source region, axial components of velocity are generally small
Ekman--type boundary layers on the discs, an interior compared with a representative tangential velocity.
core of rotating fluid, and a sink layer near the Outside the boundary layers and the mixing region,
outlet (which in this case is assumed to be a viscous efFects are neglected and approximate
uniform cylindrical sink). On entering the cavity, analytical solutions are used to describe the flow.
the air mixes with the recirculating flow in the These are the free vortex relationship in the source
outer part of the source region and then flows region and the equations of the rotating core, which
radially inwards. Inside the source region, the flow imply zero axial and radial velocities and an
is similar to that of a free vortex, and the axially-uniform tangential velocity. These solutions
associated pressure drop can he very large. are coupled directly to solutions of the
In Fig. 1 the mixing is assumed to occur momentum-integral equations for the disc boundary
rapidly, close to the inlet; this assumption will layers. A simple treatment of the shroud boundary
be discussed further below. As the fluid travels layer is also incorporated, and this is coupled to
radially inwards, its angular momentum is conserved the core solution through mixing of the
and its tangential velocity increases. At some recirculating and inlet flows. Outlet effects are
radius r s , say, the fluid velocity reaches that of assumed to be confined to a small region and are
the disc, and downstream of this point the air will neglected in this analysis.
rotate faster than the discs and the flow will Some justification for the assumption of rapid
gradually be entrained into the disc boundary mixing of the nozzle flow and the recirculating flow
layers. When all the supplied flow has been is given by finite-difference solutions of the
entrained, these boundary layers become Reynolds-averaged Navier-Stokes equations (Chew 1987
nonentraining, and a rotating core develops in which and Morse 1937). In these studies the inlet was
there is little radial or axial motion. modelled by a narrow axisymmetric slot and the
Pressure and velocity measurements for radial solutions showed a rapid change of tangential
inflow through a rotating cavity have previously velocity near the inlet followed by a region in
been presented by Owen et al (1985), Firouzian et al which the flow is similar to that of a free vortex.
(1986), Alberga et al (1987) and Graber et al This aspect of the model is also consistent with the
(1987). In these experiments either flow entered the assumpt on of rotationally-dominated flow in which
cavity at the same tangential velocity as the disc the pressure gradient in inviscid regions must
or the inlet conditions were such that the balance the centrifugal force. Thus any axial
tangential velocity of the flow at inlet was variation in angular velocity will produce an axial
uncertain. The measurements are generally variation in pressure which, in turn, will produce
consistent with the flow structure described above, an axial motion tending to eliminate the axial
but Alberga et al and Graber et al give a different variation of velocity.
interpretation of the flow outside the boundary The details of the mathematical model are
layers. These workers put forward a mathematical described below. Incompressible flow is assumed as
model in which turbulent exchange is significant in this is adequate to model the experimental
the core region, whereas Owen et al and Firouzian et conditions considered later. Extensions to
al proposed a model which neglects any such effect. compressible flow requi1 es inclusion of an energy
-

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/83350/ on 03/04/2017 Terms of Use: http://www.asme.org/abo

 equation, but this is reasonably straightforward the disc, and p and v denote fluid density and
(see Chew and Rogers 1988, Chew and Snell 1988). kinematic viscosity.
With the usual boundary-layer assumptions,
2.2 The _momentum-int gra_l equations using eqns. (1) and (2), and using the radial
A cylindrical coordinate system (r,^,z) momentum equation for the inviscid flow region to
rotating with the discs is employed. The discs are eliminate the pressure, the following
located at z = 0 and z = s but, assuming symmetry, momentum-integral equations may be derived
only the half cavity 0 z < S/2 need be considered.
The momentum-integral equations for the disc
boundary layer may be derived by integrating the
full equations of motion, or by considering a r dr(rpuos) + P s_ 2 I 5 + 2pSVSI 4 = - Tr,o (5)
control volume of vanishing radial thickness across
the width of the boundary layer. Assumptions must be
made regarding the boundary-layer velocity profiles
` o v^) - r dr(rPuoE) + 2I,Pu 0 52s = - T m o
and the wall shear stress. The assumptions used here rz dr (r Pu l v
are generalizations of those of von Karman (1921)
for free-disc flow and may be written as follows (6`
u = u o (1 - z/t)(z/S)i/ (1) where C2 is the angular velocity of the disc. The co-
efficients I are given by
V = v(z/S)/ z (2)
49 34 3 4 9_1 2
13 Ia = Is =
T , 0 = 0.0225 p(s) 1 / 4 v(uo) + v 2 3/e (3) I1 120 ' 12 1656 144

u (7)
Tr ,o = v T m o (4)
Eqns. (5) and (6) express the conservation laws for
radial and angular momentum; for further details of
Here u and v are the radial and tangential velocity the derivation the reader is referred to Chew and
components in the rotating co-ordinate system, S is Rogers (1988). Substituting eqns. (3), (4) and (7)
the boundary-layer thickness, u o is a function of r into eqns.(5) and (6) gives two first-order
only, v is the tangential velocity at the differential equations in the unknowns u o , v and E,
all of which are functions of r. To close the
boundary-layer edge, T (P o and Tr o are the
tangential and radial components of shear stress at problem a further equation is required and
appropriate boundary conditions must be specified.
An additional equation giving v in the source
Mixing region region is provided by the solution for the flow
outside the boundary layer. This is the free-vortex
r b relation
r(v + Cr) = ceff ll b 2 (8)
Source region
where ceff is an effective inlet swirl fraction (to
be defined below) and b is the outer radius of the
r=r cavity. In the rotating core, there is no radial
s flow and the flow in the two non- entraining
Ekman-type layers must equal the net flow through
the cavity, m, so that
r = re
4rr I 1 P£u o = m (9)
Ekman layer
The boundary layer on the disc can be divided
into three distinct regions: the outflow region b >
Core r > r s , the entraining inflow region rs > r > re ,
and the nonentraining inflow region re > r > a. The
stagnation point on the disc occurs when the fluid
velocity given by eqn.(8) equals the disc speed
which, in the rotating frame, is zero.
Thus
Sink layer rs° = ceff b` (10)

r=a At this point there is no flow in the boundary

layer, arid starting conditions for both the inflow
arid outflow regions must be chosen to give zero or
r very small S and su o at this point. The position r
= re is defined as the point at which all the flow
z has been entrained into the inflow layer and is
determined from the solution. Starting conditions
for the nonentraining layer are provided by the
solution for the entraining layer at r = r e .
The pressure drop across the cavity may be
obtained from the radial-momentum equation in the
Fig.1 Schematic of radial inflow in a inviscid region which reduces to
rotating cavity

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/83350/ on 03/04/2017 Terms of Use: http://www.asme.org/abo

 0

shown that C p ; calculated from the above

d p = P (V + Or) (11) equations, is a function of only r,T mnd tte radius
dr r
ratio a/b. When c x 1, a similar analysis (details
of which are not given here) shows that C p c is a
where p denotes static pressure. function of T, a/b, c and s/h. Thus, for any
particular geomery, C c depends only on xT and c.
2.3 Treatment of the shroud and mixing regions
The mass flow in the boundary layer along the 2.5 Numerical solution of the equations
shroud is assumed to be constant and equal to that As in earlier studies, solutions of the
in the outflow boundary layer on the disc at r = b. integral equations (5) and (6) with the auxiliary
It is also assumed Ihat the boundary-layer thickness relations (3), (4) and (7) to ('11) were obtained
is small compared with b. A friction factor F is using a standard mathematical library routine for
defined such that the tangential shear at the ordinary differential equations. The computer
surface is - Fpv m 2 where v m is the mean angular program used was developed from that of Chew and
velocity of the fluid relative to the shroud. The Rogers (1988) and works with a nondiriensional
angular-momentum balance for the flow in the shroud representation of the equations. The solutions for
boundary layer then gives the following equation both the outflow and inflow regions of the layer
wPPre started at the stagnation point r = r s and
dv
proceed in the direction of the boundary-laver flow.
(l)
m ` d- n - nb Fpvm

Some iteration is required to find the effective
inlet swirl fraction ceff. Initially, a guessed
If F is assumed constant this equation may be value is used to obtain a solution for the outflow
integrated between z = 0 and z z s/2 to give region of the -disc boundary layer, and an improved
estimate of c e ff is then obtained using egns.(1?)
and (14). This process is repeated until
_ ?m s (v m ) z =o 3. sat isfactory convergence is obtained. For any
(v m)z=s/z rrbsPF(vm)z-o - ems t ) particular example, only a few seconds of CPU time
on e.ri IBM 3081 were required to obtain a solution.
Eqn. (13) was used to estimate the increase in
3. EXPERIMENTAL APPARATUS
angular momentum due to friction at the shroud, and
F and (v m ),_ o were then estimated from the
3.1 The rotating-cavityrig
disc-boundary-layer solution at r = h.
Simplified diagrams of the rotating cavity and
The inlet swirl fraction c of the fluid leaving
the de-swirl nozzles are shown in Figs. 2 and 3. The
the nozzles is defined as the tangential velocity of
cavity comprised two steel discs, of outer radius
the fluid in a stationary coordinate system divided
381 mm and thickness 12.7 mm, and a 1.5-mm - thick
by the disc speed. The effective inlet swirl
Paxolir peripheral shroud. The latter contained
fraction ceff is defined as the tangential velocity
thirty equispaced hole=, of 28.6 mm diameter, in its
of the mixed recirculation and inlet flows divided
mid--a*ial plane. To obtain an initial swirl fraction
by the disc speed. If c is less than unity, radial
of unity, and to generate a settling chamber for the
outflow occurs in the boundary layer on the disc,
air upstream of the de-swirl nozzles, the shroud was
and this results in the recirculation shown in Fig.
covered by a layer of porous rubber foam.
1. For a given flow rate, c may be estimated from
Each disc was mounted on a hollow tube with an
the nozzle geometry, as discussed in Section
inner radius of 38.1 mm. One tube was blanked off
3.1. A simple mixing sum then gives the following
and the other was connected, via stationary piping,
relationship for ceff
to the inlet of a centrifugal compressor, which
could produce flow rates up to 0.1 kg/s. The discs
v ^ could be rotated up to 2000 rev/min by means of a
(2m s + m) ceff = 2ms M)z=s/2 + 1 ] + me (14) variable-speed electric niotor.
The de-swirl nozzles were machined from a ring
of Rohacell, a lightweight, high-strength, plastic
Note that the treatment of the boundary-layers foam. Sixty equispaced holes were drilled at an
described above would not be expected to be valid angle of e = 30 0 to the tangential direction at each
for ceff < 0 • of three axial locations, as shown in Fig. 3.
Convergent nozzles were formed by machining a
2.4 Dimensional_analysis conical inlet section, with a 10° half-angle, and a
A pressure coefficient C p c , rotational parallel outlet section of diameter d = 5 mm. The
Reynolds number Ret, mass flow coefficient C w and outer radius of the cavity was taken to be the inner
throughflow parameter aT are defined as follows radius of the nozzle ring where r = b = 320 mm. The
complete nozzle assembly was bonded to the discs, as
s shown in Fig.4.
C p c = P,zb1 (Pb - Pref ) ' Rem = -b The inlet-swirl fraction is given by

(15) fib - gcose

c = fib (16)
m Cw
Cw ub' >'T = Re oe =
where q is the magnitude of the velocity, relative
to the disc, of the air leaving the nozzles at r
where subscript b refers to the outer radius and b. If N nozzles are used then
ref" refers to a suitable reference location.
For the special case where the inlet swirl
fraction, c, is unity, Chew and Snell (1988) have 4I 2pNN
q = nd (17)

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/83350/ on 03/04/2017 Terms of Use: http://www.asme.org/abo

 L

Hence, using the definition of C w , Re5 and sT given u,ei ar a datum from whl,h the other four pressures
in cqr (1F, it follows that were mea;c^red. Tire reference tube from each of the
for tierisdusers were connected, via a common tube,
to this datum tai:. To minimize imbalance, the
c = 1 -- 4 ose
( b^ u wj _ - 4cose (b3 ) r.i^i5au ,er_ , w r e located on a diametral line, with
1Ji8)
nN d Rex nN d Re 'cc on )ne radius and two on the other. Thus, two
datuir tips and two common tubes were used. The
electrical signals from, and the power supply to,
For given values of C w and Res, c can be varied by the transducers were brought out through a silver
altering N, and this was done in practice by placing sip rindassernbly, and the voltages were measured
rubber plugs in some of the 180 nozzles. by a Snlartron TMS 3510 data-logger controlled by e
Cobs (bulbous hubs i also made from Ro'iacell, PDP 11/41 minicomputer.
and identical to thor,e used by Fsrthing and Oven crlid-body rotat on occurred inside the common
18c,1, were attachec' to the centre of the discs. tote, whereas inside the cavity the pressure
The shahs wa chosen to be rehresentetlye of that difference could be less or greater than that
j,,ect in gas turbine design, end the inner and outer associated with - o1id-body rotation. As a
radii Hof the cob,, were 36.1 and 138 mm. The minimum csequence, the transducers were individually
al spacing between the cobs was 33 mm and the calibrated, off the rig, over the entire operating
.'i l _;e. tion is tended to a radius of 107 mm. To Barge of ± o kN;rn". The largest source of error was
r;'d_sce overall pressure drop, ten Rohacell -aused by 'zero drift': the sensitivity (volts/unit
c,er^tr,l ,-any.>, :is described by Chew et ai (1988),
w p,s,itiored between the cobs.

• J o ,t urentatio
on
Hole f 5 mm diameter were drilled axially
t . uyh one of the discs (the left-hand disc shown
n fl. 2) at radial locations of r r 172, 223, 274,
end 361 mni. These holes were used as
s`.atic -pressurc taps, and a Kulite XT--190 series
^iIii^ture differential-pressure transducer was
flush-mounted to each of the four outermost taps
lr,umbur:, 1 to 4). The innermo_-t tap (number 0) was

Air inlet
{(from atmosphere)
E""

)ud
-ling chamber

swirl nozzles in

Rohacell ring

ition of
in tap

ral vane

ral bung

e shaft

tion
Fig.2 Rotating cavity and location
of pressure taps (numbered Section X-X
0 to 4)

Dimensions in mm Fig.3 De-swirl nozzles used in

experiments

Dimensions in mm

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/83350/ on 03/04/2017 Terms of Use: http://www.asme.org/abo

 0

pressure) of the transducers remained constant but, of each photograph correspond to the exit of the
depending on the rotational speed, the voltage level de-swirl nozzles and the edge of the central vanes,
varied with time. The results were corrected for respectively. (The vanes prevented optical access
this zero drift, which was determined immediately below x = 0.5.) The opaque de-swirl nozzles
before each test. The maximum correction was 15% of prevented the illumination of the r - z plane
the reading. through the shroud, and so visualization was via the
The flow rate was measured, by Annubar transparent (left-hand) disc shown in Fig. 4 (which
differential-pressure devices, to an accuracy to appears as the right-hand disc of Fig. 5). The
f 3%. However, a relatively small amount cf air field-of-view was at an angle of approximately 50°
leaked into the system (which was at sub-atmospheric to the plane of illumination, and so the edges of
pressure) through the seals between rotating and the nozzle ring and the central vanes appear as
stationary components. A series of calibrations oblique lines at x = 1 and 0.5. For these
revealed that the leakage flow rate, QL, was photographs, all 180 de-swirl nozzles were used.
independent of rotational speed but was proportional Fig. 5 shows smoke patterns photographed at
to np s where oP s was the pressure drop across the different times for C w _ -480, Rem = 3.8 x 10 4 and c
system measured by a static pressure tap downstream Z 0.66; for this value of C w , the flow is expected
of the cavity. The measured flow rate was corrected to be laminar, and interpretation of the flow
to account for the leakage, which was less than 15%
of the measured value.
For flow visualization, the left-hand steel
disc in the cavity was replaced by one made from
transparent polycarbonate. A 4W argon-ion laser was
used to provide slit illumination through the
polyca.rbonate disc (see Fig. 4), and the air
entering the cavity was "seeded" with micron-sized
oil particle€ produced by a Concept smoke generator.
Vii+.es recordings and photographs were used to study
tho flow structure. in the cavity.

4. FLOW VISUALIZATION

4.1 Flow structure

Using the flow visualization apparatus
deii ibed in the preceeding section, photographs, as
shown in Fig. 5, were taken of the smoke patterns in
the r - plane of the cavity. The top and bottom

(a)
7 1
-

eye

From
laser
-I►

^o.5
(b)
Fig. 4 Photograph of the rotating-cavity rig
showing the de-swirl nozzles and the plane of Fig.5 Smoke patterns in the r - z plane
illumination. for IC w j = 480, Re m = 3.8 x 10 4
and c = 0.66.

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/83350/ on 03/04/2017 Terms of Use: http://www.asme.org/abo

 structure is aided by reference to Fig. 1. Fig. 5(a) structure. These measurements were compared with
shows that the smoke, after entering the cavity in the theoretical values of Firouzian et al (1985)
three jets, is beginning to fill the source region. where x e = r e /b and
The smoke pattern on the right-hand disc provides
evidence that there is radial outflow as well as
inflow inside the source region. Fig. 5(b) was xe = (c ILI) (19)
taken a few seconds later when smoke had penetrated
through the source region, Ekman layers and sink
layer, and the interior core appears as a black for laminar flow, and
region for x 0.7. The other black region, for x
0.9, is caused by recirculation in the source
x e = (c - 2.221XT1 5 /ax e 3 / 8 ) (20)
region; as discussed in Section 2.3, this
recirculation occurs when ceff < 1 .
The photographs shown in Fig. 5 provide for turbulent flow.
evidence that the usual source-sink flow-patterns These equations were derived from the sc-called
still occur even when the fluid enters the cavity linear momentum-integral equations: when v/0r<<1,
through de-swirl nozzles. Other photographs showed the Coriolis forces dominate over the nonlinear
that, as expected, the size of the source region inertial terms, and the latter can be neglected.
increased with increasing flow rate and with The radius r e , is taken to be the point where all
decreasing rotational speed. the fluid flowing radially inward has been entrained
into the boundary layers on the discs. For the
4.2 Size rof hesource re3ion experimental measurements shown in Figs. 6 and 7,
A series of tests were conducted for 200 < ICwl c r: 1 and consequently equations (19) and (20) are
2300 and 0.17 < Re P /10 5 < 4.6 with all the 180 expected to give similar' results to the numerical
nozzles. The radius of the edge of source region, snnlution of the full nonlinear equations described
r e , was measured, to an estimated accuracy of in Section 2.
t 10 mm, from video recordings of the flow
i i A

xe
Ie

UI ❑m

0.5L 0-51
0 0.5 0 1
(0-2-221 ATl 5/8 x e 3/1 )^z
C Th I
- I

Fig.6 Size of source region for laminar Fig.7 Size of source region for turbulent
flow flow

❑ e>perimental nieesu;sements ❑ experimental measurements

difference between maximum and — equation (20)

minimum values

— equation (19)

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/83350/ on 03/04/2017 Terms of Use: http://www.asme.org/abo

 0

Transition from laminar to turbulent flow is where, for convenience, pb was taken to be the
assumed to occur when Re r = 180, and the results for pressure recorded by transducer number 3 (this war
laminar and turbulent flow are shown in Figs. 6 and actually located at r = 325 mm where x = 0.280).
7, respectively. For some cases of laminar flow, the Tests were conducted for 550 CWI y' 2.1 10 and
source region appeared to oscillate, and this is 0.15 < Re 4,/10 6 / 1.4; by using N = 40, 80 or 160
indicated by the bar lines in Fig. 6 which join the (see equation (18)) it was possible to vary the
maximum and minimum values of x e . Similar swirl fraction in the range -0.4 < c + 0.9 for a
instabilities in the flow structure were observed by relatively large range of the flow parameter
Firouzian et al who noted that the size of the Fig. 8 shows a comparison between the variation
source region could increase or decrease of C with x obtained from the momentum-integral
periodically in a non-axisymmetric fashion. From equat ions, as discussed in Section 2 and the
Fig. 6 it is evident that, where no instabilities msiasured values at x = 0.52, 0.68 and 0.83. Tha
occur, there is reasonable agreement between the measurements were made at Pe (r = 0.61 x 10 6 with N =
theoretical predictions and the measured values. 40, and it can be see; that the agreement between
For the cases where the size of the source region the theoretical and experimental results is good.
fluctuated, reasonable agreement is obtained with The agreement was also good for most of the other
the maximum value of x e (i.e. the smaller size of results obtained, and this gave confidence in both
source region). For turbulent flow no oscillations the theoretical model and the experimental
were observed, and Fig. 7 shows reasonable agreement procedures.
between the experimental and theoretical results. Fig. 9 shows the variation of the pressure
coefficient for the cavity, C p c (based on the
5. PRESSURE DROP IN THE ROTATING CAVITY pressure difference between transducer 3 and the

Pressure transducers 1, 2 and 3, at the 25

locations shown in Fig. 2, were used to determine
the stat c pressure relative to the datum pressure 9
tap. The nondimensional pressure difference, C p , Cp 'C
was defined as

2
p
C = nzb ^
b p
(21)

2
15 t+++

C P pd(^ d
d0

0 7
xx
x
xxxxxx

^ p
tom'
O 05

3 4 5 ^ ❑ ^13
1 °
D ° o
0
0 0 01 02 03 0.4 05

0.4 0.6 0.8 1

x

Fig.9 Effect of c on variation of C p c

with IaTI
Fig.8 Var lation of C p with x for
curve measurements C
Rem = 0.61 x 10 6

c = -0.2, Ia T I = 0.144 1 • -0.4 — Solutions of

2 ® -0.2 integral
3 0 0 equations
e c = 0, IaTI = 0.120
4 ♦ +0.1
5 ❑ +0.2 --- Solid-body
0 c = +0.2, IaTI = 0.096
6 x +0.4 rotation
7 0 +0.6
Solutions of integral
8 + +0.8
equations
9 0 +0.9

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/83350/ on 03/04/2017 Terms of Use: http://www.asme.org/abo

 u

reference tap), with I'TI for a range of different experiments (where ICwl rather than C p c was
inlet swirl fractions. For solid-body rotation, the controlled), no such jump phenomena were observed.
value of C p c for the cavity is approximately 0.75.
For each value of c, up to three clusters of 16
experimental data can be seen: these correspond to
N = 40, 80 and 160. The experiments were carried I Cwl
out by fixing N then varying C w and Rey to keep c 10 3
constant for as large a range of xT as possible. 12
It can be seen from Fig. 9 that, for most
values of c, there is a maximum value of C pc : the
magnitude of this maximum and the value of Ti at
which it occurs increase with increasing c. These 8 ❑

effects can be explained by considering the mixing

process discused in Section 2.3: ceff tends to unity 0

as I5TI approaches zero, and ceff tends to c for

large values of !aTl. Also for large values of IXTI,
as the source region begins to fill the entire

e
cavity, free-vortex flow occurs and C p c tends to a
limiting value (C p c = c 2 (x r f - 1)). For the larger
0 05 1SC 2
values of c where mixing is weak, this limiting pc
value is approached monotonically as IaTI is
increased (see curve 9 in Fig. 9); for the smaller
values of c where mixing is significant, a turning (a) Re,, = 0.23 x 10 6
point is apparent (for example, see curve 6). 16
The minimum value of C p occurs when the I Cw l ❑
positive swirl in the outflowing boundary layer on
10 3 ❑
the discs in the source region neutralizes the
negative swirl created by the nozzles such that ceff 12 ❑
= 0: the fluid then moves radially inwards with
little or no residual swirl, thereby minimizing the
pressure drop across the cavity. For large ranges
of 1-H-1 and c, it is therefore possible to reduce
the pressure drop across the cavity to less than
that associated with solid-body rotation. Without
the dry -swirl nozzles, the pressure drop could be an
order of magnitude greater than that associated with
solid-body rotation (see Chew et al 1988).
In the experiments, the pressure drop was the
dependent variable: aT and c were varied in a
controlled fashion and the resulting value of C p c 0 05 15 2
Op ' s
was measured. In a gas-turbine engine, the pressure (b) Re = 0.38 x 10 6
drop between the compressor bleed-off and, for
example, the turbine-blade-cooling air is fixed:
the flow rate of cooling air is determined by the 16
available pressure difference. To illustrate the I 0w l
variation of flow rate with pressure difference, the 10 3
variation of IC w l with C p c is shown in Fig. 10 for
12
N = 80) and for various values of Re m . For the
experimental data, as IC w l increases c decreases in
the manner given by equation (18).
Fig. 10(a) shows the variation of IC w l with
C p c for Re q, z 0.23 x 10 6 the resulting 's-curve"
;

can be explained by referring to Fig. 9. The

turning point at ICI = 10 3 corresponds to the
approximate maximum value of C c in Fig. 9; the
turning point at IC w l = 5 x 10° corresponds to the
minimum value of O p c (where ceff = 0). For ICl >
5 x 10 3 (where ceff < 0), the fluid rotates in the
opposite direction to that of the discs, and the OL
pressure drop increases as IC w l increases. The 0 05 1 15 2
C pc
agreement between theory and experiment is good up
to IC w l = 5 x 10 3 the theoretical model is invalid
; (c) Re = 0.61 x 10 6
for ceff < 0. Similar results are shown in Figs.
10(b) and (c) for Re D/10 6 = 0.38 and 0.61, and again
the agreement between theory and experiment is good.
Fig. 10 Variation of IC with C PIC
The s-curve behaviour, illustrated in Fig. 10,
could result in unstable conditions inside an ❑ Experimental measurements
engine. As Cp C is increased in the engine, IC w l
might jump from a low to a high branch of the -- Solutions of integral equations
curve; conversely, a jump from a high to a low
branch could occur as C p c is reduced. In the

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/83350/ on 03/04/2017 Terms of Use: http://www.asme.org/abo

 0

6. CONCLUSIONS Chew, J.W., 1987. Computation of Flow and Heat

Transfer in Rotating Disc Systems. Proc. of the 2nd
Solutions of the momentum-integral equations ASME_JSME Thermal Engng Joint Conf_ vol.3, pp
have been used to calculate the effect of inlet 361-367.
swirl on the pressure drop in a rotating cavity with
a radial inflow of fluid. The theoretical values Chew, J.W. and Rogers, R.H., 1988. An Integral
were compared with measurements made inside a rig in Method for the Calculation of Turbulent Forced
which de-swirl nozzles were used to reduce the swirl Convection in a Rotating Cavity with Radial Outflow.
fraction of the incoming air. Flow visualization Int_ J. Heat and Fluid Flow, vol.9, pp 37-48.
confirmed that "conventional" source-sink flow
structure still occurred when the de-swirl nozzles Chew, J.W. and Snell, R.J., 1988. Predictions of the
were fitted, and pressure transducers attached to Pressure Distribution for Radial Inflow Between
one of the discs were used to measure the static Co-Rotating Discs. 33rd International ASME Gas
pressure inside the rotating cavity. Turbine Conference, Amsterdam. Paper No. 88-GT-61.
In the main, there is good agreement between
the theoretical and experimental values of pressure Chew, J.W., Farthing, P.R., Owen, J.M., and
distribution. The mixing that occurs between the Stratford, B., 1988. The Use of Fins to Reduce the
radially outward flow in the boundary layers on the Pressure Drop in a Rotating Cavity with a Radial
discs and the radially inward flow through the Inflow. 33rd International ASME Gas Turbine
de-swirl nozzles modifies the inlet swirl fraction. Conference, Amsterdam. Paper No. 88-GT-58.
However, if these two flows are suitably matched,
the effective swirl fraction can be reduced to zero, Farthing, P.R. and Owen, J.M., 1987. The Effect of
and the resulting pressure drop across the cavity is Disc Geometry on Heat Transfer in a Rotating Cavity
then negligibly small. For a constant value of with a Radial Outflow of Fluid. J _Eng. Gas Turbines
rotational speed, the variation of flow rate with and Power, vol. 110, pp 70-77.
pressure drop exhibits an "s-curve" behaviour in
which there are up to three different values of flow Firouzian, M., Owen, J.M., Pincombe, J.R. and
rate for each value of pressure drop. In a Rogers, R.H., 1985. Flow and Heat Transfer in a
gas-turbine engine, this behaviour may well result Rotating Cavity with a Radial Inflow of Fluid. Part
in flow rates which jump between different branches 1: The Flow Structure. Int. J. _Heat and Fluid Flow,
of the s-curve. vol.6, pp 228-234.
it should be pointed out that the pressure
measurements described above were all made inside Firouzian, M., Owen, J.M., Pincombe, J.R. and
the rotating cavity itself: the pressure drop Rogers, R.H., 1986. Ibid. Part 2: Velocity,
across the de-swirl nozzles is also important. This Pressure and Heat Transfer Measurements. Int. J.
pressure drop, and the matching of the nozzles and Heat and Fluid Flow, vol.7, pp 21-27.
the cavity, will be the subject of a future paper.
Graber, D.J., Daniels, W.A. and Johnson, B.V., 1987.
Disc Pumping Test. Report No. AFWAL-TR-87-2050,
ACKNOWLEDGEMENTS Wright-Patterson Air Force Base, Ohio, U.S.A.

We wish to thank the Science and Engineering Karman, T. von, 1921. Uber Laminare and Turbulente
Research Council, Rolls Royce plc and Ruston Gas Reibung. Z. angew. Math. Mech., vol.1, pp 233-252.
Turbines plc for funding the research described in
this paper. Kraftwerk Union Aktiengesellschaft, 1977.
Improvements in or Relating to Gas Turbine
Assemblies. UK Patent 1,541,533.

REFERENCES Morse, A.P., 1987. Numerical Prediction of

Turbulent Flow in Rotating Cavities, 32nd
International ASME Gas Turbine Conference, Anaheim.
Alberga, D.G., Stephens, G.E. and Johnson, B.V., Paper No. 87-GT-74.
1987. Comparison of Predicted and Measured
Velocities in a Compressor Disc Drum Model. 32nd Owen, J.M., Pincombe, J.R. and Rogers, R.H., 1985.
International ASME Gas Turbine Conference, Anaheim. Source-Sink Flow Inside a Rotating Cylindrical
Paper No. 87-GT-208. Cavity. J. Fluid Mech., vol.155, pp 233-265.

Brown, W.M. and Manente, J.C., 1977. Compressor Scheper, G.W., 1962. Turbine Rotor Ventilation
Bleed System. US Patent 4,008,977. System. US Patent 3,043,561.

10

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/83350/ on 03/04/2017 Terms of Use: http://www.asme.org/abo