Está en la página 1de 7

1

The 12th International Symposium on Transport Phenomena and


Dynamics of Rotating Machinery
Honolulu, Hawaii, February 17-22, 2008

ISROMAC12-2008-20111

NUMERICAL SIMULATION OF THE FLOW IN A TWO STAGE
TURBINE DRIVING A LIQUID OXYGEN PUMP


1.S. Rao, M.Saravanakumar

ChieI Science OIIicer, CFD Engineer.
Tel 91 80 66294500; Fax 91 80 66294700
js.raoaltair.com saravana.kumaraltair.com




ABSTRACT
This paper describes computational investigations oI
two stage liquid oxygen impulse turbine oI a high speed
cryogenic turbo-pump used in the Iourth stage oI a geo
stationary launch vehicle. The computational domain
consists oI nozzle block, rotor, stator and the secondary
rotor. CFX 5.6, the commercial code used in the
calculation Ior simulating three-dimensional viscous Ilow-
Iields, with a SST k-e two-equation turbulence model was
adopted to enable calculations with turbulence.
In the calculation, the Iinite-volume method and a
structured grid system were used Ior the solution
procedure oI the discretized governing equations Ior this
problem. Stage or Mixing Plane Model is used to simulate
interaction between the stator and rotor in the turbine. The
results obtained Ior pressures, Mach numbers and
temperatures across the stationary and moving blade rows
are presented. Static pressure distributions on the rotor
blades are presented. Taking into account the pressure
distribution at pressure and suction surIaces, the torque
developed Ior a single blade in each stage is determined.
The total power oI the LOX Turbine was also calculated.
Turbine perIormance is studied at operating speed

INTRODUCTION
High speed cryogenic turbo-pumps are commonly
used in the last stage propulsion systems oI space launch
vehicles. These engines use liquid hydrogen and oxygen;
sometimes both the oxygen and hydrogen pumps are
combined into the same engine along with the turbine and
in other applications, two separate turbo-pumps may be
used to handle these cryogenic liquids.

Previous studies on turbo pump have been carried out
by Majumdar et.al |1|.This paper presents a numerical
model oI internal Ilows in a rocket engine Turbopump
developed with the Generalized Fluid System Simulation
Program, GFSSP Ior analyzing steady state and time-
dependant Ilow rates, pressures, temperatures oI a turbo
pump. The axial thrust and internal Ilow distribution oI the
Fastrac Turbopump were calculated during the start
transient oI a component test. GFSSP computes the time-
dependant Ilow in the internal or secondary Ilow circuits
oI the Turbopump.
The aerodynamic design oI the turbines driving the
pumps oI the Vinci expander cycle engine is discussed by
Helmers et.al |2|. Focus is put on the turbine driving the
hydrogen pump oI the engine. A turbine stage perIormance
is simulated in 3D viscous, turbulent analysis. In this paper
a standard setting oI all numerical schemes, viz., higher
order spatial discretization, explicit three-stage Runge-
Kutta time stepping, standard k-c turbulence model, steady
state calculation with tangential averaging at the rotor
stator interIace was used.
Osamu et.al |3|, presented three-dimensional viscous
Ilows in a turbine maniIold Ior a Iuel turbo pump oI a
rocket engine, which breathes the gas Irom the tangential
direction and then distributes it through a volute and radial
slits to the inlet oI turbine nozzle, has been numerically
calculated and visualized. The computational domain
consists oI sub domains Ior the volute, radial slits and the
inner passage. UPACS, the computer code used in the
calculation Ior simulating three-dimensional viscous Ilow-
Iields, has an advantage that it can be applied to
complicated conIigurations with the technique oI large-
scale parallel computation. The computed viscous Ilow
2
Iield is presented to understand the Ilow structure in the
maniIold
CFD analysis oI an axial Ilow turbine remains a
challenging problem due to the model Iidelity required to
capture the complicated Ilow physics, such as,
shock/boundary layer separation, shock wave reIlection,
interaction between the shock wave and tip clearance Ilow,
etc. 3D unsteady calculations with the sliding mesh
approach make it possible to model the aerodynamic
interaction between the rotor and stator. Furthermore
modeling oI the turbine is commonly complicated by the
Iact that stator and rotor posses an unequal number oI
blades in order to reduce vibrations. Hence geometrical
periodicity cannot be applied in the simulation process and
the entire stator and rotor have to be considered. As a
result this leads to an enormous computational eIIort in
terms oI memory and CPU demand and requires the use oI
modern high perIormance computers see |4, 5|. ThereIore
in practical design a transient computation oI an entire
turbine is still too expensive and time consuming.
Consequently commercial designers are Iorced to revert to
steady state simulations.
A simple method is a stage averaging method, where
at the interIace between the stationary and rotating Irame
oI reIerence the Ilow variables are averaged in
circumIerential direction prior to their exchange. The
advantage oI this approach is that regardless oI the actual
pitch only one blade passage oI the stator and rotor needs
to be modeled which leads to a huge reduction oI memory
demand see |6|. As a drawback the approach neglects all
transient interaction eIIects. In this work, the Ilow is
assumed as steady and compressible. In order to save
computer power these simulations are perIormed steady
state mixing plane technique by applying an averaging
procedure in the circumIerential direction. Here, this
approach is adopted.

NOMENCLATURE
P Pressure (Pa)
m Mass Ilow rate (kg/s)
T Temperature (K)
k Turbulent kinetic energy (m
2
/s
2
)
e SpeciIic Dissipation rate
c Dissipation rate
R Gas constant (J/kg K)
N Speed(rpm)
q Non-dimensional radius
(r-r
h
)/(r
c
-r
h
)
q
T
EIIiciency
C Absolute velocity (m/s)
p Density (kg/m
3
)
n No oI blades
CoeIIicient oI expansion
C
P

Static pressure coeIIicient
1
1
C
2
1
P P

=

0
Dimensionless Mass Ilow
0
2
reI
0
P r
RT n m


=

Subscripts
1 Inlet oI the rotor
0 Total
C Casing
H Hub
R Radial component
U CircumIerential component
X Axial component
ReI ReIerence radius

GEOMETRY AND COMPUTATIONAL DOMAINS
LOX Turbine geometric model dealt with in the
present study is shown in Fig.1. The geometry oI the liquid
oxygen turbine was created with UniGraphics CAD
package. It consists oI inlet maniIold, nozzle block, Iirst
rotor, stator, second rotor, and exhaust maniIold.


Fig. 1: LOX Turbine Geometric Model

!nlet manifold
Nozzle block
First Rotor
Stator
Second Rotor
Exhaust manifold
3
Tip clearance
Blade to blade view
The LOX turbine inlet maniIold is oI toroidal shape
with single gas entry. The gas Irom LH2 turbine exhaust is
carried to the LOX turbine inlet maniIold by a pipe. The
inlet maniIold distributes the gas throughout the
circumIerence oI the nozzle block. The nozzle row
converts pressure energy into velocity. Most oI pressure
drop occurs in the nozzle. The shape oI the nozzle
passages is convergent. The rotor deIlects the Ilow and
hence a torque is imparted to the rotor. The turbine is a
two stage velocity compounded impulse turbine and has
two rotors. The stator between the two rotors changes the
direction oI the gas Ilow Irom Iirst rotor outlet to the
second rotor inlet. The exhaust maniIold is used to collect
the Ilow Irom the rotor. The collected Iluid in the maniIold
goes to atmosphere through a dump nozzle.

Fig. 2: LOX Turbine Computational domain

Due to the cyclic symmetric geometry oI the turbine
model, the computational grids were created Ior one blade
domain in the nozzle block nozzle block, Iirst rotor, stator
and second rotor. This allowed a signiIicant reduction oI
modeling and computation time. Fig. 2 shows the
Computational domain created Irom the geometry oI two
stage LOX turbine.

DESCRIPTION OF THE NUMERICAL DOMAIN
The structured computational mesh was created using
the pre-processor ICEMCFD 5.1. This mesh generator is
well suited Ior creating complex structured grids as the
domain can be split into several sub domains in a top-
down approach allowing Ior good control oI each sub
domain oI the mesh. When the mesh is complete the
subdomains can be clustered in lager blocks Ior output.
The computational mesh consists oI 8 blocks. Each vane
or blade is gridded with one O-type grid around the airIoil
and an H-type grid Ior the major part oI the channel. Finer
grids are implemented near the leading and trailing edges
to capture the Ilow details around the vanes, Ior better
resolution oI boundary layer. To consider tip clearance,
the grid in the tip clearance region above the rotor tip is
constructed by simply extending the blade span to the
casing, creating another block between the tip and the
casing. The tip clearance block is then discretized as
shown in the Fig.3. The Iull size oI the 3D grid is about 1
million grid cells.
Grid points are clustering towards the blade surIace
and end walls to ensure y values at the Iirst cell are in the
20~100 range over the whole Ilow Iield.








Fig. 3: Computational Grid

BOUNDARY CONDITIONS AND MODEL SETUP
In CFX-Pre, boundary conditions, solver model,
convergence criteria are deIined and a deIinition Iile is
created. The total pressure, Ilow direction, and total
temperature were imposed as the inlet boundary
conditions. Static pressure was applied as the outIlow
boundary condition. The periodic boundary conditions |5|
are used to take care oI the periodicity oI Ilows in the
circumIerential direction. No slip boundary conditions Ior
hub, tip and blade surIaces are applied.

Boundary conditions Ior LOX turbine analysis are as
shown in Fig. 4. Fluid properties used in the turbine are
Gas constant R 2171 J/kg K
SpeciIic heat at constant pressure 7918 J/kg K
The operational requirements Ior LOX turbine at
reIerence conditions are summarized in Table 1.

Table 1: Operational requirements

Inlet stagnation pressure 5.6 bar
Inlet stagnation temperature 650 K
Turbine speed N 15000 rpm
Power P 481 kW
Neridional view
Nozzle block
First Rotor
Stator
Second Rotor
4
Mass Ilow m 1.17 kg/s

Real Ilow in a turbo-pump is quite complex, being
viscous, unsteady and three-dimensional. CFD programs
oIIer diIIerent approaches to investigate such a Ilow Iield
numerically, all oI them using multiple reIerence Irames:
Frozen rotor model
Mixing plane model
Sliding mesh model
The three approaches above diIIer in the way that the
interIace between moving and nonmoving cell zones is
treated. In the Irozen rotor model, the coupling between
the cell zones is obtained by maintenance oI absolute
velocity in the global coordinate, i.e. the velocities are just
switched between relative and absolute Irames. Thus, one
obtains a snapshot` oI the Ilow Iield at one Iixed rotor
position.

!nflow
Total Pressure: 5.6 bar
Total Temperature:650 K
Outflow
Shroud
Hub
Periodic Surface
Blade
!nterface:
Stage or Nixing plane model

Fig.4: Boundary Conditions

For the mixing plane approach |7|, Ior each Iluid zone
a steady state solution is calculated. The two adjacent
Irames are coupled by exchange oI Ilow Iield variables at
the interIace. Flow Iield data are averaged
circumIerentially Ior both Irames at the interIace and then
passed to the adjacent zone as boundary conditions. II
Ilow variations in the pitch wise direction are neglected in
the rotor/stator interIace region, one can assume that the
relative Ilow Iield in successive rotors and stators is
steady. This technique is reIerred to as the mixing plane
approach, and has been widely used since the Iirst
publications oI Denton et. al |8|. This approach not only
neglects the unsteady eIIects oI rotor/stator interaction, but
also introduces an error generated by the artiIicial mixing
process applied at the interIace. It is however the current
approach in industry Ior multistage simulation.
Both the Irozen rotor and the mixing plane models
assume that the Ilow Iield is steady, with the rotor/stator
eIIects being accounted Ior by approximate means. On the
other hand, the sliding mesh model assumes that the Ilow
Iield is unsteady, and hence models the interaction with
complete Iidelity. Most oIten, the unsteady solution that is
sought in a sliding mesh simulation is time-periodic. Note,
since the sliding mesh model requires an unsteady
numerical solution, it is computationally more demanding
than the Irozen rotor and mixing plane approaches.

NUMERICAL METHOD
The 3D viscous, compressible Ilow analyses, and the
associated result acquisition, were perIormed with
commercially available CFX-5.6 Ilow simulation soItware.
CFX-5 uses a coupled solver, which solves the
hydrodynamic equations (Ior u, v, w, and p) as a single
system. This solution approach uses a Iully implicit
discretization oI the equations at any given time step. For
steady state problems the time-step behaves like an
acceleration parameter, to guide the approximate solutions
in a physically based manner to a steady-state solution.
This reduces the number oI iterations required Ior
convergence to a steady state, or to calculate the solution
Ior each time step in a time dependent analysis. A
convergence criterion oI 0.0001 is deIined. The deIinition
Iile is solved in CFX-Solver, and the governing continuity,
momentum, energy and turbulence equations are solved
until the required convergence criterion is reached.

TURBULENCE MODELLING
One oI the main problems in turbulence modeling is
the accurate prediction oI Ilow separation Irom a smooth
surIace. Standard two equation turbulence models oIten
Iail to predict the onset and the amount oI Ilow separation
under adverse pressure gradient conditions. In this
analysis, the most prominent two-equation model Shear
Stress Transport (SST) k- based model was chosen with
an automatic near-wall treatment, which utilizes a wall
Iunction where y~11 to ensure suIIicient boundary layer
resolution where the mesh would not be adequately
reIined. This model was designed to give a highly accurate
prediction oI the onset and the amount oI Ilow separation
under adverse pressure gradients by the inclusion oI
transport eIIects into the Iormulation oI the eddy-viscosity.
This results in a major improvement in terms oI Ilow
separation predictions. The SST model is recommended
Ior high accuracy boundary layer simulations and adopted
in this analysis.
RESULTS
The steady-state distribution shows the average
pressure distribution across the surIace oI the Turbine
stage. Fig.5 shows Static Pressure distribution over the
nozzle block, rotor, stator and the secondary rotor 50
span Irom casing. In the turbine stages, Most oI pressure
drop occurs in the nozzle block and the pressure almost
remains constant in the Iirst rotor, stator and second stator.
5
Suction side vortex
Tip leakage vortex

Fig.5: Contour plots oI Static Pressure distribution


Fig.6: Distribution oI static pressure coeIIicient
in First rotor



Fig.7: Distribution oI static pressure coeIIicient
in Second rotor
Fig.6 shows distribution Ior static pressure coeIIicient
on Iirst rotor blade surIace at q 0.7 and 0.9 spanwise
locations. In the rotor blade, Ilow deceleration occurred
near the leading edge Ior both pressure and suction
surIace. On the pressure surIace, aIter the initial
deceleration to about 3 surIace distance, the Ilow
continuously accelerated toward the trailing edge. On the
suction surIace, the Ilow accelerated Irom 5 surIace
distance to about 30 surIace and then decelerated toward
the trailing edge. For the second rotor, the pressure
distribution was similar to that Ior the Iirst rotor as shown
in Fig.7. At q 0.9, the static pressure on the suction
surIace is decreased more compared to 0.7 due to the
eIIect oI large tip clearance, but it has a little eIIect on the
pressure surIace.















Fig.8: Tip leakage Ilow

The existence oI tip clearance between the rotor blade
and casing wall is a major source oI perIormance
distortion Ior axial turbomachines. Fig.8 shows the stream
lines oI velocity magnitude that are released Irom the
upstream oI the Iirst rotor blade. In the tip clearance the
stream lines near the pressure surIace are being entrained
in to the suction side through region oI the gap.
This entrainment is driven by spanwise pressure
gradient due to the unloading oI the blade at the tip. These
streams then exit Irom the gap and wrap around the
leakage vortex core that results Irom the leakage Ilow
Iurther upstream. A close look on the suction surIace
shows that the leakage vortex acts as a blockage to the
passage Ilow. These complex Ilows clearly aIIect the
pressure distributions on both sides oI rotor blade as
clearly seen in Figs.6 and 7.

6


Fig.9: Contour plots oI Mach number at q 0.5

Fig. 9 shows the steady-state Mach number contours
Ior 50 span Irom the casing. The Ilow is accelerated in
the Iirst stator with the Mach number increasing Irom 0.2
to 0.8 (Mass Ilow average value) and then decelerated
across rotor and stator It can also be seen Irom the
contours oI Mach number that there may be a region oI
Ilow that is trying to separate near the trailing edge oI the
rotor blade. At that point, the Mach number is very low,
and thereIore some separation would be expected to occur
near this point somewhere in the Ilow Iield. Upon
investigation oI a contour plot oI axial velocity this region
oI separation was Iound to be near the hub at the trailing
edge oI the rotor.

Zone of heat damage

Fig.10: Contour plots oI Static Temperature

Most oI temperature drop occurs in the nozzle block
and the temperature almost remains constant in the Iirst
rotor, stator and second stator. Fig.10 shows distribution
Ior static temperature on turbine Iirst blade surIace at q
0.5. The tip clearance is responsible Ior a signiIicant
portion oI perIormance losses in turbines and the blade tip
is known to be susceptible to damage Irom excessive
thermal loading as shown in Fig.10. The power calculated
on the turbine blade is 470 kW.

LOX TURBINE PERFORMANCE MAP
The numerically calculated perIormance map oI the
LOX Turbine is shown in Fig.11 and 12 Ior operating
conditions.
0.5
0.6
0.7
0.8
0.9
1
1.1
1 1.5 2 2.5 3 3.5
Pressure Ratio
D
i
m
e
n
s
i
o
n
l
e
s
s

M
a
s
s

F
l
o
w
Turbine Operating Speed
15000 rpm

Fig.11: Turbine characteristics map Pressure ratio Vs Flow


The adiabatic eIIiciency and total pressure ratio are
calculated using the mass averaged values oI the total
pressure and total temperature at the inlet and exit stations
oI turbine. In Fig 11 it can be seen that the pressure ratio
increases as the Ilow rate increases up to the pressure ratio
2.1 and remains constant Ior Iurther increase in Ilow rate.
Adiabatic eIIiciency varies with the pressure ratio at a
given speed as shown in Fig12.
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
0.55
1 1.5 2 2.5 3 3.5
Pressure Ratio
E
f
f
i
c
i
e
n
c
y
Turbine Operating Speed
15000 rpm

Fig.12: Turbine characteristics map


Pressure ratio Vs EIIiciency
7

Three-dimensional viscous Ilows in a LOX impulse


turbine oI a high speed cryogenic turbo-pump has been
numerically calculated and visualized. The interaction
between the stator and rotor in the turbine was simulated
by Stage or Mixing Plane Model. The results obtained Ior
pressures, Mach numbers and temperatures across the
stationary and moving blade rows are presented. Static
pressure distributions on the rotor blades are presented
Visualization with streamlines is quite useIul to understand
the general Ilow structure in the tip clearance.


ACKNOWLEDGMENTS
The authors are thankIul to Altair Engineering India
Ior their support.

klllkl\lS klllkl\lS klllkl\lS klllkl\lS


1. Majumdar, A. K., and Van Hooser, K. P.,
Numerical Prediction oI Transient Axial Thrust
and Internal Flows in a Rocket Engine
Turbopump, Paper No. AIAA 99-2189, 35th
AIAA/ASME/ SAE/ASEE Joint Propulsion
ConIerence June 20-24, 1999, Los Angeles,
CaliIornia
2. Helmers, L., Steen. J., Ljungkrona, I., Brodin. S.,
and Johnsson, R., Turbine Design and
PerIormance at Large Tip Clearance oI
Unshrouded Rotor Cascades, AIAA-2003-4766,
39th Joint Propulsion ConIerence, Huntsville,
Alabama, 20-23 July 2003
























3. Osamu, N., Toshio, N, Kazuhisa, S., Numerical
Simulation oI the Flow in a ManiIold oI Axial
Turbine, Proceedings oI PSFVIP-4., June 3-5,
2003, Chamonix, France.
4. Ruprecht, A., Bauer, C., Gentner, C., and Lein, G.,
Parallel Computation oI Stator-Rotor Interaction
in an Axial Turbine, ASME PVP ConIerence,
CFD Symposium, Boston, 1999
5. Gier, J., Ardey, S., Eymann, S., Reinmller, U.,
and Niehuis, R., 2002, Improving 3D Flow
Characteristics in a Multistage LP Turbine by
means oI Endwall Contouring and AirIoil Design
ModiIication, Part 2: Numerical Simulation and
Analysis, ASME Paper GT-2002-30353
6. Brost, V., Ruprecht, A., MaihIer, M., Rotor-
Stator Interactions in an Axial Turbine, a
Comparison oI Transient and Steady State Frozen
Rotor Simulations., Institute Ior Fluid Mechanics
and Hydraulic Machinery, University oI Stuttgart,
Germany
7. CFX 5 User`s Guide, ANSYS Inc
8. Denton, J. & Singh, U. (1979), Time Marching
Methods Ior Turbomachinery Flow Calculations,
VKI-LEC-SER-1979-7, von Karman Institute
9. Miner, S. M., CFD Analysis oI the First Stage
Rotor and Stator in a Two-Stage Mixed Flow
Pump, International Journal oI Rotating
Machinery, March 30, 2004
10. Ferziger, J. H., and Peric, M., Computational
Methods Ior Fluid Dynamics, Springer, 1996

También podría gustarte