Вы находитесь на странице: 1из 8

Proceedings of

2000 International Joint Power Generation Conference


Miami Beach, Florida, July 23-26, 2000

IJPGC2000-15004

INVESTIGATIONS OF FLOW CHARACTERISTICS


OF AN HP TURBINE STAGE INCLUDING
THE EFFECT OF TIP LEAKAGE AND WINDAGE FLOWS
USING A 3D NAVIER-STOKES SOLVER
WITH SOURCE/SINK-TYPE BOUNDARY CONDITIONS
Piotr Lampart and Andrzej Gardzilewicz
Institute of Fluid Flow Machinery
Polish Academy of Sciences, Gdask, Poland
e-mail: lampart@imppan.imp.pg.gda.pl

Sergey Yershov and Andrey Rusanov


Institute of Mechanical Engineering Problems
Ukrainian National Academy of Sciences, Kharkov, Ukraine
e-mail: yershov@online.kharkiv.net

plished thanks to the application of non-reflective boundary


conditions there, giving four values of invariants for injection
and one for extraction. The corresponding mass flow rates of
injected/extracted medium - equivalent to the intensity of
sources/sinks - are calculated from more simple onedimensional studies of leakage and windage flows.
Three computational examples of a high pressure turbine
stage with short height blading are enclosed in the paper. First,
computations are made without tip leakage and windage flows
with source/sink slots closed. Second, tip leakage slots are
open. Third, both tip leakage and windage flow slots are open,
and the effect of mixing of tip leakage and windage flows with
the main flow, on the performance of the stage is studied.

ABSTRACT
Energy conversion in turbomachinery flows takes place in
extremely complex three-dimensional geometries. Besides
blade-to-blade channels and axial gaps - the domain of the main
flow - there are places of completely different aspect ratios like
labyrinth seals and passages between the fixed and rotating
parts of the machinery where leakage and windage flows take
place and then mix in the blade-to-blade passage with the main
stream. 3D modelling of the entire impulse turbine stage or a
group of stages, that is accounting for all these areas simultaneously, only inlet and exit conditions assumed, is an extremely
difficult task.
The approach presented in the paper enables injection of
the medium at arbitrary velocities and angles, determination of
the effect of mixing of injected medium with the main flow which is the main source of entropy creation due to tip leakage
and windage flows, as well as interaction of injected streams
with other vortex flows - secondary flows or separations. 3D
viscous, compressible flow in the blade-to-blade passage and
axial gaps is modelled with the help of thin-layer Reynoldsaveraged Navier-Stokes equations with a modified BaldwinLomax algebraic model of turbulence. In addition to boundary
conditions typical for turbomachinery codes - total pressure and
temperature at the inlet to the stage, static pressure assumed at
the exit, no slip and no heat flux at the walls - the source/sinktype boundary conditions are assumed at some places at the
endwalls referring to design locations of injection of leakage
and windage flows into, or their extraction from the blade-toblade passage. Injection and extraction of the streams is accom-

INTRODUCTION
3D modelling of the entire turbine stage or a group of
stages, only inlet and exit conditions assumed, still remains an
extremely difficult task. The main difficulty lies in the complexity of turbomachinery geometries, and different aspect ratios
and flow scales between the main flow in the blade-to-blade
passages, tip leakage over shrouded rotor blades, leakage
through stator sealing glands, and windage flows in passages
between the fixed and rotating parts of the machinery. A picture
of a group of stages for an impulse turbine with indicated directions of the above mentioned streams is presented in Fig. 1.
Before full 3D computations of the entire flow geometry of
the impulse turbine shown in Fig. 1 are within easy reach, a
recent paper of Tajc, Polansky 1999, describes their CFD com-

Copyright 2000 by ASME

Ts
m
tan 1
= 2 L (1
sin 2 2 )
m
tan 2
0.5V22

where 1 and 2 are the swirl angles upstream and downstream


of the blade row, mL is the tip leakage mass flow rate compared
to the mass flow rate in the blade-to-blade passage m. The formula shows the determinant effect of the difference in swirl
velocity across the blade row, or in other words, the difference
in swirl velocity between the main stream and leakage jet at its
re-entry on the tip leakage losses.
However, the swirl velocity of the tip leakage can change in
the labyrinth seals. The tip leakage jet can also have some axial
component of the velocity on its re-entry, and it is expected that
the direction of the tip leakage jet re-entry as well as the angle
at which the windage jet is injected can have a significant influence on the entropy creation in the downstream mixing process.
Leakage and windage flows as mass injections or extractions
interact also with secondary flows and separations. It is highly
recommended that 3D codes for turbomachinery flow prediction should enable modelling of these phenomena.
Fig. 1. Impulse turbine geometry.
putations of tip leakage labyrinth seals using the commercial
software FLUENT with the RNG k- turbulence closure. The
authors consider a number of labyrinth seal configurations
which are then ranked according to the resultant computed
leakage flow rate. A notice on labyrinth seal computations can
also be found in a recent news release from NREC & NUMECA Int. Partnership, NREC News, 1999. The notice underlines an importance of, first, fine resolution of computational
grids to conform accurately to very small clearance gaps and
large cavities between the seal teeth and, second, choice of turbulence models since the majority of the cavity regions is
dominated by large secondary flows. CFD-based analysis can
minimise the leakage flow rates and improve performance of
labyrinth seals. NS computations also enable the evaluation of
the entropy creation processes in the labyrinth seals. However,
unlike for the leakage flow rate that is inherent to the labyrinth
seal geometry, most of entropy creation due to tip leakage takes
place not in the labyrinth seals but in the blade-to-blade passage
downstream of the tip leakage jet re-entry where the injected
leakage jet mixes with the main stream.
At the highest level of simplification, the relative enthalpy
losses due to tip and root leakages (referred to isentropic enthalpy drop across the turbine stage) can be assumed equal to
the relative mass flow rate of the tip and root leakages (referred
to the total mass flow rate of the main flow plus leakages),
Gardzilewicz 1984. Tip leakage losses over both shrouded and
unshrouded blades are discussed in detail in a landmark paper
of Denton 1993, on loss mechanisms in turbomachinery. The
author derived a formula for tip leakage losses making a number of simplifications, among others assuming a constant swirl
velocity of the leakage jet through the rotor, and neglecting the
axial velocity of the leakage jet at its re-entry, obtaining

COMPUTATIONAL METHOD
The computations whose results will be presented in the
paper are carried out with the help of a code FlowER - solver of
viscous compressible multi-stage turbomachinery flows developed by Yershov & Rusanov 1996ab, see also Rusanov & Yershov 1996, Yershov et al. 1998. The code has been tested on a
number of turbomachinery geometries and flow regimes for real
full-scale and model turbines, including tests held at the ERCOFTAC Workshops on Turbomachinery Flow Prediction, see
Yershov et al. 1997. The main features of the code will shortly
be described below.
Governing equations
Flow in the blade-to-blade passages and axial gaps is described by a set of thin-layer Reynolds-averaged unsteady
Navier-Stokes equations written in a curvilinear body-fitted
coordinate system ( - radial, - pitch-wise, - stream-wise)
rotating with an angular speed
R R
QJ E F G
+
+
+
= JH +
+
,
t

(1)

where Q is a conservative variable vector, E, F and G are flux


vectors; H is a source term vector; R and R are cross-stream
viscous terms (with stream-wise diffusion discarded)

2 v + 2 r
u
x

Q = v ; H = 2 u + 2 ry ;

w
0

Copyright 2000 by ASME

E=

uU + p
x

J vU + p y

wU + p
z

( h + p )U

R =

R =

;F =

1
U
3

1
V

uV + p
x

J vV + p y

wV + p
z

( h + p )V

;G=

uW + p
x

J vW + p y

wW + p
z

( h + p )W

0
1
u
x + 20
3

1
2 v
y + 0
3

1
w
z + 20
3

+ 20 u

u
v
w 1
+v
+w
+

Pr

0
1
u
x + 20
3

1
v
y + 20
3

1
w
z + 20
3

+ 20 u

u
v
w 1
+v
+w
+

Pr

Boundary conditions
As the boundary conditions at the inlet, span-wise distributions of total pressure, total temperature, pitch and yaw angles
are assumed, and also a span-wise distribution of static pressure
- as a voluntary condition to help evaluate the initial flow field.
At the exit from the computational domain we use optionally
either a span-wise distribution of static pressure, or static pressure at the mid-span section, with its span-wise distribution
calculated from the radial equilibrium equation. There is no slip
and no heat flux at the walls, and the static pressure is found
from Eq. (1) written along cross-flow gridlines at blade walls

p 1
=
J

+ const ; h =

p
1
= 2 (rx x + ry y ) +

+ x x + y y + z z

2
0

2 ( JU )
0
+

( JW ) 1
( JV )
2

,
+ +0

3

{( x x

2
0

+ y y + z z

2 ( JV )
0
+

( JW ) 1 2 ( JU )
+ +0

.

3

The assumption of complete spatial periodicity at free boundaries upstream of the leading edges and downstream of the trailing edges, combined with the concept of a mixing plane between the fixed and moving blade rows in the steady-state approach, or time-space periodicity with a sliding plane in the
unsteady approach are assumed.
Source/sink-type boundary conditions
In order to account for the effect of leakage over shrouded
blade tips and windage flows the computational domain is
modified at the endwalls where some places are permeable
boundaries of the domain - sources or sinks, see Fig. 2. In addition to the boundary conditions typical for turbomachinery
codes as presented in the preceding subsection, source/sinktype boundary conditions are used at places at the endwalls
referring to design locations of injection of leakage and windage flows into, or their extraction from, the blade-to-blade passage. The injection and extraction are accomplished with the

1 p u 2 + v 2 + w2 2 r 2
+
+ const ;
1
2

0 = 2x + 2y + 2z ; 0 = 2x + 2y + 2z ;
= x

+ y y + z z

and at endwalls

U = u x + v y + w z ; V = u x + v y + w z ; W = u x + v y + w z ;

+ x x + y y + z z

where p is the pressure, u,v,w - Cartesian components, U,V,W contravariant components of velocity, i - enthalpy, h - rothalpy

i=

{( x x

w
v
u
u
v
w
+y
+ z
; = x
+ y
+ z
.

J denotes the Jacobian of transformation from the Cartesian to


curvilinear coordinate system and is the specific heat ratio, Pr
- Prandtl number. The effective (molecular and turbulent) viscosity and conductivity are as in the Bussinesq hypothesis
where the laminar part is found from the formula of Sutherland
whereas in order to determine the turbulent part a modified
Baldwin-Lomax model is used. The original model of BaldwinLomax, 1978, was modified to better calculate turbulent viscosity in the regions of separation and wake.

Fig. 2. Computational domain with source/sink-type permeable


boundaries to simulate the effect of leakage over shrouded rotor
blade tips and windage flows; S - stator, R - rotor.

Copyright 2000 by ASME

application of non-reflective boundary conditions there, giving


four values of invariants for injection ( u n > 0 )

Qin, j , k = Qin, +j ,1k Qin, j , k = Hin, +j ,1k/ 2

[
]
[( F R ) ]
+[ G ]

( E R ) n

I + = u n + 2a / ( 1) = const ,
ut1 = const ,

ut 2 = const , S = p / = const

i , j + 1/ 2 , k

and one value for extraction ( un < 0 )

[
]
[ ( F R ) ]

[ G ]

( E R ) n

i 1 / 2 , j , k

(2)

i , j 1/ 2 , k

i , j , k + 1/ 2

I = u n 2a / ( 1) = const

i , j , k 1/ 2

where subscripts i,j,k refer to cell centres, i1/2, j1/2, k1/2 to


cell sides, n is a time instant.

where un , ut 1 , ut 2 are velocity components - one normal and


two tangential to the source/sink boundaries, p - pressure, density, S - entropy function, a - speed of sound, - isentropic
exponent, I , I + - left and right Riemann invariant, see also
Yershov et al. 2000. These invariants can be found, first, from
preliminary computations in the basic computational domain
without sources and sinks, giving p and at places referring to
the source/sink locations, with the density at the sink throat of
the tip leakage jet determined from isenthalpic conditions. The
corresponding mass flow rates of the injected/extracted fluid equivalent to the intensity of sources/sinks - can be calculated
from more simple 1D studies of leakage and windage flows.
Then, the needed velocity components can be obtained based
on the density and size of source/sink slots.

ENO reconstruction
In Godunov-type schemes inviscid fluxes are found from the
solution of the Riemann problem, Godunov et al. 1976. In our
method, initial values for the Riemann problem are found from
q
q
q ( , , , t ) = qm + ( m ) + ( m ) +
m
m

(3)

q
q
( m ) + (t tn )
t n
m

where q = (, u, v, w, p) is the primitive variable vector, qm - its


value at the cell centre, m, m, m - cell centre coordinates. The
spatial derivatives that appear in Eq. (3) are found from the
following ENO approximation written for the characteristic
variable vector , see Harten & Osher 1987,

Stage losses
The presented approach enables injection of the medium at
arbitrary velocities and angles, determination of the effect of
mixing of injected medium with the main flow, as well as interaction of injected streams with other vortex flows - secondary
flows or separations. The approach embodies the message that
most of the entropy creation due to tip leakage is inherent to the
blade-to-blade passages not to the tip clearance or labyrinth seal
itself. Mass averaged kinetic energy losses of the stage can be
found from a formula that takes into account leakage streams

def
1
min mod

= =

m + min mod ( m m 1 , m +1 m ),

(4)

m +1 min mod ( m +1 m , m + 2 m +1 )

where is the cell size, m=m-m-1 and


minmod(a, b) = sign(a) max{0, min[a,bsign(a)]}.

= i Gi / Gi
ex .+ s.

i + 1/ 2 , j , k

( J ) i , j , k

This formulation requires transformation from the primitive to


characteristic variables and the other way round. The choice of
constants , determines the order of the numerical scheme.
For =1/2, =1/2, we have an ENO scheme that is secondorder accurate everywhere in time and space. For =2/3,
=1/3, the scheme is locally third-order accurate, remaining at
least second-order everywhere, see Yershov 1994, and this
scheme is implemented in the present paper. The time derivative in Eq. (3) is found from the non-divergent form of Eq. (1)

ex .+ s.

where the summation extends on all streams that carry away the
fluid from the blading system (exit and sinks) and i is the kinetic energy loss, Gi - mass flow rate in a stream i. In the case
of nominal directions of leakage and windage flows and equal
intensities of respective sources and sinks, the formula reduces
to the summation over exit streams only. Exit velocity losses
are found as in non-source/sink computations.
NUMERICAL SCHEME

Explicit step
The applied scheme draws on cell-centred finite-volume
discretisation

R R

q
q
q
q
+ D
+ D
+ D
= T JH +
+
,
t

where matrices D can be found from Eq. (1), and T is the


transformation from primitive to conservative variables.

Copyright 2000 by ASME

Implicit step
One drawback of the explicit scheme - Eq. (2) - is its insufficient effectiveness in terms of computational costs. The process of convergence to a steady-state can be accelerated with the
aid of an implicit scheme of Beam & Warming, 1978

I +
(
1
+
x) J

Viscous fluxes
The viscous fluxes are calculated based on ENO approximation and a weighted linear interpolation, for example (subscripts i,k are left out as constant)
u
u
u
= ( j +1/ 2 j )
+ ( j + 1 j + 1/ 2 ) . (7)

j + 1/ 2
j + 1
j

A B C n
1
x
Q n 1 , (5)
+
+
RHS n +

Q =

1
1
+
x
+
x

where I = diag {1} is a unit matrix; , x are constants; Q - the


conservative variable vector. In order to obtain second-order
accuracy it is necessary to put down = 2, x = 1, and assure the
approximation of the right-hand side (RHS) term with secondorder accuracy in space. Solving Eq. (5) requires its factorisation (with regards to space coordinates) and diagonalisation of
matrices A, B, C. The employed factorised implicit scheme
works on the characteristic variable vector

COMPUTATIONAL RESULTS
The computed turbine stage is a typical impulse HP stage of
a 200MW steam turbine operating at the pressure drop of about
0.9, inlet temperature - 780K, flow rate - 170 kg/s, average reaction - 0.15; the aspect ratios are: span/chord - 0.8 (stator) and
2.0 (rotor), pitch/chord - 0.8, span/diameter - 0.08. Prior to
CFD computations, the stage was scrutinised with the aid of a
1D code to evaluate the mass flow rate of the main flow in the
blade-to-blade passage of the stator and rotor G1, G2 as well as
flow rates of leakages at the tip and root GT, GR and windage
flows GW, GW based on the given pressure drops and geometry
of labyrinth seals and passages. The results obtained from the
1D approach necessary for further 3D computations are as follows: GT=2.7%G1, GW=GW=1.2%G1. Then 3D computations
were made in three variants. First, without tip leakage and
windage flows with source/sink slots closed, second, with only
tip leakage slots open, third, with both tip leakage and windage
flow slots open. In source/sink computations, it was assumed by
way of example that the fluid is extracted and injected through
the sinks and sources in the radial direction (no axial and swirl
velocity), which is far from the real turbine situation. However,

+
+ n +1/ 3 = n ;
I +

(
1
)
+
x
J

+
+
I +
(1 + x ) J

+
+
I +
(1 + x ) J

n+2 /3

= n +1/ 3 ;

(6)

= n + 2 / 3 ,

where = ( ) / 2 , the diagonal matrix consisting


of eigenvalues of D. This formulation requires transformation
between the characteristic, primitive and conservative variables.

Fig. 3. Axial distribution of mass flow rate in the computational domain of the rotor computed without sources and sinks (left), computed with tip leakage (centre), computed with tip leakage and windage flows (right).

Fig. 4. Entropy function contours and velocity vectors in the rotor at 9% blade span from the root computed without sources and sinks (left), computed with tip leakage (centre), computed with tip leakage and windage flows (right).

Copyright 2000 by ASME

Fig. 5. Entropy function contours in meridional view in the computational domain of the rotor
at 3% (left), 12% (centre left), 48% (centre right) and 96% (right) blade-to-blade distance from the suction surface computed without sources and sinks (top), computed with tip leakage (centre), computed with tip leakage and windage flows (bottom).
the computational results presented comparatively below in
subsequent Figs. from 3 to 7 can be viewed as an illustration of
the described idea of flow solving, showing interesting effects
of leakage and windage flows on turbomachinery performance.
The axial distribution of mass flow rate in the rotor computed with source/sink slots closed can be considered nothing
more than a measure of convergence of the numerical algorithm. In source/sink computations, sink and source throats belong entirely to the rotor computational domain. In this case,
therefore, the axial distribution of mass flow rate in the rotor
illustrates the tip leakage mass flow rate equal to 2.7% of the
total mass flow (as assumed in the boundary conditions for
variant 2), or the summary mass flow rate for tip leakage and
windage flows equal to 3.9% of the total mass flow (variant 3)
by-passing the rotor blade-to-blade passage and not contribut-

ing to the rotor work, see Fig. 3.


Fig. 4 with entropy function contours and velocity vectors in
the rotor at 9% of the blade span from the root computed without sources and sinks, and with tip leakage, or tip leakage plus
windage flow sources and sinks shows an interesting flow feature - separation from the front part of the suction surface of the
rotor blade at the root. It seems that pitch/chord/stagger angle
optimisation and stator/rotor matching for that stage may not
have been executed with due care. The shape of the separation
zone undergoes changes with the presented computational variants. The separation zone is the smallest for non-source/sink
computations. It slightly changes with a tendency to increase in
size in computations with tip leakage. However, the changes are
not spectacular as the flow modification takes place at the opposite endwall, that is at the tip. The separation zone significan-

Copyright 2000 by ASME

dary flows, separation, and wake. The separation and secondary


flow vorticity at the root merge towards the mid-span section,
giving more loss than that due to secondary flows at the opposite endwall. Computations with the tip leakage (pictures in the
centre) show also the effect of mixing of the leakage stream
with the main flow adding to flow losses near the tip endwall.
The effect of separation at the root slightly increases. However,
as the high entropy boundary layer fluid is sucked out into the
tip leakage slot prior to the rotor, an interesting feature is observed that the intensity of secondary flows is reduced. The
span-wise extension of the secondary flow zone considerably
shrinks, compared to non-source/sink computations. Results of
source/sink computations with tip leakage and windage flows
(pictures at the bottom) confirm all the previous findings. The
separation zone conspicuously extends, intensity of secondary
flows is reduced. The zone of mixing due to the tip leakage is
seen to extend more significantly in the radial direction, compared to that of the windage flow. The effect of windage flow is
of a lesser consequence for the flow downstream of the rotor
blades than that of the tip leakage flow due to the fact that its
flow rate at the source throat was assumed only 1.2%G1, compared to 2.7%G1 for the tip leakage mass flow rate.
Fig. 7 is a quantitative reflection of the phenomena observed
on previous pictures. The figure shows a comparison of spanwise distribution of kinetic energy losses in the rotor and stage,
computed for three considered variants. In all cases the losses
are calculated at a section located 42% of the axial chord downstream of the rotor trailing edge. There will certainly be more
loss further downstream as the mixing processes is not yet accomplished at the assumed test section. The shape of graphs
undergoes considerable redistribution over the considered computational variants. For non-source/sink computations, similar
to pictures of entropy function contours, the maximum at the
mid-span is due to the merged root separation and secondary
flow vorticity. The second, lower maximum should be attributed to secondary flows at the opposite endwall. Source/sink
computations add losses near the endwalls as a result of interaction (mixing) of the injected fluid from the sources with the
main stream in the exit diffuser downstream of the rotor trailing
edge. Although tip leakage and windage flow losses can not be
easily separated from other losses, especially the tip leakage
loss is seen to have a great share of the total stage loss. The loss
maximum due to separation at the root increases with the increasing mass flow rate by-passing the blade-to-blade passage.
The maximum due to secondary flows at the tip is hardly discernible from other sources of loss, giving testimony to the decreased rate of secondary flows in source/sink computations.
The presented results are obtained assuming that the medium is extracted and injected through the sinks and sources in
the radial direction (no axial and swirl velocity). The investigations will be continued extending on extraction and, especially,
injection of tip leakage and windage jets also with axial and
swirl velocities according to the geometry of the tip leakage

Fig. 6. Entropy function contours in the rotor at 10%, 52%,


96% axial chord from the leading edge, and downstream of the
rotor at 10% and 45% axial chord from the trailing edge computed without sources and sinks (top), computed with tip
leakage (centre), with tip leakage and windage flows (bottom).
tly extends when tip leakage and windage flows are taken into
account. No wonder. Extraction of the fluid prior to the rotor
reduces the mass flow rate in the rotor blade-to-blade passage
and changes the angle of attack locally. Especially, extraction
of the fluid into the windage slot at the root acts to extend the
zone of separation at the root section, compared to nonsource/sink computations.
Figs. 5 and 6 illustrate contour of the entropy function
S=p/ in the rotor at pitch-wise subsequent sections between
the suction and pressure surface, and also at axially subsequent
sections beginning from the leading edge to the trailing edge
and downstream into the wake of the rotor for non-source/sink
computations, as well as for source/sink computations with tip
leakage and tip leakage plus windage flow. The pictures exhibit
characteristic features of subsonic flows in axial turbines. Nonsource/sink computations illustrate the development of secon-

Copyright 2000 by ASME

Fig. 7. Span-wise distribution of kinetic energy losses in the rotor (1), stage without the exit velocity (2) and with the exit velocity (3) computed without sources and sinks (left), computed with tip leakage (centre), computed with tip leakage and windage flows (right).
labyrinth seal and windage flow passages. It is expected that the
direction of the tip leakage jet re-entry as well as the angle at
which the windage jet is injected have a significant influence on
the entropy creation in the downstream mixing process.

Fluid Flow Machinery, Gdask, Poland, No. 161 (in Polish).


Godunov S.K., Zabrodin A.W., Ivanov M.A., 1976, Solving
multi-dimensional problems in gas dynamics, Nauka, Moscow
(in Russian).
Harten A., Osher S., 1987, Uniformly high-order accurate
non-oscillatory schemes, SIAM Journal of Numerical Analysis,
Vol. 24, No. 2.
NREC News, 1999, CFD analysis to improve performance
of labyrinth seals, NREC News, Vol 13., Issue 1, pp. 3-4.
Rusanov A.V., Yershov S.V., 1996, The new implicit ENO
method for 3D viscous multi stage flow calculations, Proc. 3rd
ECCOMAS Computational Fluid Dynamics Conf., Paris,
France, September 9-13.
Tajc L., Polansky J., 1999, Labyrinth seal flow computations, Rep. SKODA Energo, Prague, Czech Rep. (in Czech).
Yershov S.V., 1994, The quasi-monotonous ENO scheme
of increased accuracy for integrating Euler and Navier-Stokes
equations, Math. Modelling, Vol. 6, No. 11 (in Russian).
Yershov S.V., Rusanov A.V., 1996a, The high resolution
method of Godunov's type for 3D viscous flow calculations,
Proc. 3rd Colloq. Process. Simulation, Espoo, Finland, June
13-16.
Yershov S.V., Rusanov A.V., 1996b, The application package FlowER for the calculation of 3D viscous flows through
multistage turbomachinery, Certificate of Ukrainian state agen
cy of copyright and related rights, Kiev, Ukraine, February 19.
Yershov S.V., Rusanov A.V., Gardzilewicz A., Badur J.,
Lampart P., 1997, Calculations of Test Case 3 - Durham low
speed turbine cascade, Calculations of Test Case 9 - Highly
loaded transonic linear turbine guide vane cascade, Proc. V
ERCOFTAC Seminar and Workshop on 3D Turbomachinery
Flow Prediction, Courchevel, France, January 6-9.
Yershov S.V., Rusanov A.V., Gardzilewicz A., Lampart P.,
wirydczuk J., 1998, Numerical simulation of viscous compressible flows in axial turbomachinery, TASK Q., Vol. 2, No. 2.
Yershov S.V., Rusanov A.V , Lampart P., Gardzilewicz A.,
2000, A numerical method for Navier-Stokes simulation of flow
in axial multi-row turbine blade-to-blade passages with source/
sink-type boundary conditions for leakage flows, Proc. Seminar
Topical Problems in Fluid Mechanics'2000, February 16, Prague, Czech Rep.

CONCLUSIONS
Investigations of the effect of tip leakage over shrouded
rotor blades and windage jet on the flow through an HP stage of
an impulse turbine have been carried out using a 3D NavierStokes code with source/sink-type permeable boundary conditions implemented at places at the endwalls referring to design
locations of injection of leakage and windage flows into, or
their extraction from, the blade-to-blade passage. These approach enables tracing and quantitative evaluation of the process of mixing of tip leakage and windage flows with the main
stream, and their interaction with secondary flows and separations. The investigations have been conducted for the medium
extracted and injected through the sinks and sources in the radial direction with no axial and swirl velocity. More research is
required to find the effect of direction of the tip leakage jet reentry, or the effect of angle at which the windage jet is injected
on the entropy creation in the downstream mixing process.
ACKNOWLEDGEMENTS
Part of numerical calculations for this paper were carried
out on work stations of the TASK Centre in Gdask.
REFERENCES
Baldwin B.S., Lomax H., 1978, Thin layer approximation
and algebraic model for separated turbulent flows, AIAA Paper,
No. 257.
Beam R.M., Warming R.F., 1978, An implicit factored
scheme for the compressible Navier-Stokes equations, AIAA J.,
Vol. 16, No. 4.
Denton J.D., 1993, Loss mechanisms in turbomachines,
ASME J. Turbomachinery, Vol.115.
Gardzilewicz A., 1984, Selected problems of computeraided design of steam turbines and heat cycles, Rep. Institute of

Copyright 2000 by ASME