Professional Documents
Culture Documents
Design Procedure of A Turbopump Test Bench
Design Procedure of A Turbopump Test Bench
AT
Abstract
ROT ING M
O
N The high complexity of turbopumps for liquid rocket engines and their demanding requirements
AC
SYMPOSIA
necessitate that their design process is accompanied by extensive experimental investigations and
HINERY
validation tests. This paper presents the design procedure for a rocket turbopump test bench, where
water is used as a surrogate for the cryogenic fluids usually used in rocket engines. Scaling methods,
that allow for a comparison of tests under varying conditions, are reviewed from literature and
applied to derive the necessary dimensions of the test bench. The resulting test bench design is
ISROMAC 2017 shown in detail and its capabilities to support the turbopump design process are assessed. Further,
the operational envelope of the derived test bench design is evaluated with respect to later tests of
International
Symposium on
different pumps.
Transport Phenomena Keywords
and turbopump — liquid oxygen scaling — test bench design
Dynamics of Rotating
Machinery 1 Technical University of Munich, Department of Mechnical Engineering, Chair of Turbomachinery and Flight Propulsion,
Divison Space Propulsion, Munich, Germany
Maui, Hawaii 2 Institute of Space Propulsion Lampoldshausen, German Aerospace Center (DLR), Hardthausen, Germany
3 Cryogenic Engineering Centre, Indian Institute of Technology Kharagpur, India
December 16-21, 2017 *Corresponding author: julian.pauw@ltf.mw.tum.de
available software tool EcosimPro® . In order to be able to use to model flows with free surfaces where gravitational forces
the test bench for other configurations as well, an outlook on have a large impact.
the generalized operational envelope of the test bench will
complete this paper, so that the possible use for future test
inertia force v2
campaigns with different configurations can be estimated. Fr = = (4)
gravitational force gLc
2. SCALING METHODS The Reynolds number relates between the acting inertia
In order to test a pump in a test facility at conditions that forces and the viscous forces. Equivalent Reynolds num-
differ from the normal operating conditions, the similarity of bers predict a similar development of boundary layers inside
the flow passing through the pumps needs to be established two compared pumps. The generation of boundary layers
both at the test bench and at the operational conditions. Ac- is closely linked to the surface roughness of the parts that
cording to [12, 13, 14], this similarity can be reached if four conduct the flow. Especially for geometrically scaled models
different parameters are comparable: (1) the geometry un- it is difficult to generate a comparable surface roughness of
der investigation, (2) a comparable establishment of velocity all pump components. For the purpose of comparing differ-
triangles at pump inlet and outlet, (3) a similar dynamic be- ent radial machines, the Reynolds number is often defined as
haviour of the pump and (4) comparable thermodynamic the product of peripheral velocity U3 at the impeller outlet
properties of the fluids. and the corresponding tip radius R3 divided by the kinematic
viscosity υ. When comparing the scaled measurements in
Following Sigloch [13, 14], geometric comparability of two detail, it is strongly recommended to calculate the Reynolds
turbo machines is fulfilled if the dimensions of the parts number locally to guarantee comparable boundary layers all
conducting the flow follow a certain ratio in all spatial di- over the pump [14, 13, 15, 16].
mensions. This offers not only comparable interaction of
the fluid with the static and dynamic parts of the machine, 2
U3 Rt,3 ωRt,3
inertia force
but also yields similar velocity triangles at every correspond- Re = = = (5)
viscous force υ υ
ing point for friction-free conditions because the velocity
vectors are then governed by passage dimensions and the The Euler number describes the ratio between pressure forces
rotational speed only. It is important to make sure that not and inertia forces. For the scaling of pumps, it is inevitable
only the rotor and blade geometries are comparable, but also to keep the Euler number the same for model and prototype.
the gap geometries and clearance distances between rotor
and housing.
pressure force ∆p
Eu = = 2 (6)
inertia force ρv
dimension of prototype Xm
λgeometr y = = (1) Further, the thermal properties of the fluids, i.e. the spe-
dimension of model Xa
cific heat, the enthalpy and the thermal diffusivity, largely
influence the flow characteristics. This is especially true for
velocities of prototype vm machines that exhibit a large pressure difference. It is advis-
λvelocity = = (2) able to make use of fluids with comparable properties and to
velocities of model va
closely track property changes.
In the case of single-phase flows, i.e. in the case of non-
cavitating flows, the forces acting on the pump are the fol- The above scaling methods need to be incorporated in order
lowing: inertia forces, pressure forces, viscous forces and the to design a test bench that can be used to test a LOX turbop-
gravitational force. Additionally, forces resulting from the ump with water. For ground tests of liquid rocket engine
elasticity of the mechanical system, especially the shaft, can turbopumps, the gravitational effects are negligible. The dif-
act on the pump components. The flow through two different ferences in height and the resulting effects are comparably
pumps can only be considered similar if the ratio of the forces small. The Froude scaling is thus irrelevant. The Reynolds
acting on these pumps is constant in all spatial directions. scaling is of high importance. At least for this application, it
is sufficient to achieve fully turbulent conditions in the model
forces at prototype Fm and the prototype because the boundary layers that develop
λF or ce = = (3) at turbulent conditions are assumed to be comparable. Fully
forces at model Fa
turbulent conditions can be considered for Ret,3 > 106 . Due
The forces acting in pumps with single-phase flows can be to the high rotational speeds of liquid rocket engine turbop-
related by the Froude number, the Reynolds number and the umps, this requirement can be fulfilled without any problems.
Euler number. These characteristic numbers should be the Figure 1 sums up the properties that need to be scaled for
same for the model and the prototype. liquid rocket engine turbopumps.
The Froude number is defined as the ratio of inertia forces to Obeying the above given similarity conditions, the operating
gravitational forces. It is commonly used in hydro-mechanics properties can be transferred between model and prototype
Design Procedure of a Turbopump Test Bench — 3/11
Geometric
Similar velocity
In order to show comparable behaviour with water as a sur-
scaling of
rotating parts Fluid- triangles rogate compared to cryogenic fluids in terms of cavitation, it
dynamic is necessary to match the same flow coefficient ϕ as well as
scaling of
flow
the cavitation number σ [14, 15].
2Q
ϕ= (12)
ADt ω
Figure 1. Comparable properties for fluid-dynamic
similarity
(pst − pv )
σ= (13)
2ρl Dt2 ω2
with the help of well-known scaling laws for pumps[14].
However, in experiments with several fluids, especially with
For the volume flow rate: cryogenic fluids, that operate close to their critical point like
liquid hydrogen and liquid oxygen, it was possible to observe
Qm Qa
= (7) that the inducers of pumps are operating without cavitation
Nm Dt,m 3 Na Dt,a 3 at NPSH values smaller than the NPSHR which would be
expected for an ideal fluid. This effect is commonly called
For the head rise:
thermal suppression head (TSH). By definition, the TSH cal-
Hm Ha culates to the difference between the available NPSH A and
= (8)
Nm 2 Dt,m 2 Na 2 Dt,a 2 the NPSHideal f luid that would be expected for an ideal fluid.
NPSHideal f luid equals the NPSHt ank minus friction losses
For the input power: in the inlet tubing [18].
Pm Pa
= (9)
Nm 3 Dt,m 5 ρm Na 3 Dt,a 5 ρa TSH = NPSHavailable − NPSHideal f luid (14)
growth over a time-varying pressure field. The resulting DB Table 2. LOX turbopump detailed design parameters
parameter is similar to the formulation suggested by Ruggeri
and Moore [20] and is considered very convenient for esti- Property Value Unit
mating the thermal operational boundaries of the test bench
at an early stage in the test bench design process, as it is Pump Characteristics
√
based solely on thermodynamic properties of the bulk fluid Specific speed NS 24.34 r pm · m2/ s
rather than empirical correlations. For the detailed investiga- Design head coefficient ψst, pump 0.56 -
tion of a specific pump, it is of high importance to take into Hydraulic efficiency ηhyd 0.88 -
account all available prediction models, especially those for Pump overall efficiency ηtot 0.75 -
which better validation data is available in literature. Shaft power Psha f t −197 kW
√ Inducer Characteristics
Rt ω3/2 Cp,l Tl ρ2l αl √
DB = (16) Design suction specific speed NSS 421.42 r pm · m2/ s
h2f g ρ2v Design head coefficient ψtot,inducer 0.11 -
Design flow coefficient ϕinducer 0.089 -
Number of blades Zinducer 2 -
3. TEST OBJECT Hub-to-tip ratio, inlet Dh,1 /Dt,1 0.40 -
Hub-to-tip ratio, outlet Dh,2 /Dt,2 0.64 -
One main goal of the test bench under construction is to
Tip diameter Dt,1 70.83 mm
establish a result validation loop within the numerical design
process for turbopumps at TUM. The pump to be developed Impeller Characteristics
is a liquid oxygen turbopump designed for a liquid hydrogen
and liquid oxygen expander cycle engine. The desired opera- Design head coefficient ψtot,impeller 0.50 -
tional parameters are given in Table 1. The resulting thrust Design flow coefficient ϕimpeller 0.10 -
level is in the order of magnitude of the VINCI upper stage Number of blades Zimpeller 6 -
engine. Tip-to-outlet ratio Dt2 /Dt3 0.72 -
Hub-to-outlet ratio Dh2 /Dt3 0.46 -
Tip diameter Dt3 100.284 mm
Table 1. LOX turbopump nominal operating conditions
F Water Pipe
Main Throttle Massflowmeter Pump Shaft
Valve
Temperature
Pressure
Optional Water Heating System Pressure
Temperature
Electrical
Water Heater
Main Tank
Inducer &
Bearing
Pressure Control Impeller under
Auxiliary Pump Unit
System Investigation Motor
Bearing
Torque
Speed
F
Data
Massflowmeter
Temperature
Pressure
Figure 3. Drawing of the test bench layout
rates and rotational speeds in order to generate a pump per- The scaled design head rise at Nmax equals
formance chart. Therefore, it is necessary to acquire the
following data: the pressure difference over the investigated
pump, the mass flow rate, the shaft torque and the rotational Ha Nm 2
Hm, Nma x ≈ = 33.30 m (18)
speed of the shaft. The torque measurements give informa- Na 2
tion about the efficiency of the pump under investigation.
At the same operating point, the shaft power calculates to
A second goal is to experimentally observe the cavitation
behaviour of the inducer and the pump with water tests in
order to predict the cavitation behaviour under liquid oxygen Pa Nm 3 ρm
conditions. Therefore, as evident from the cavitation scaling Pm, Nma x ≈ = 3.29 kW (19)
Na 3 ρ a
theory, the water needs to be heated up in order to operate
the facility with a liquid which is close to its boiling point. The dimensions of the new test facility have to be chosen
in such a way that those parameters can be satisfied. A
wider range of operation above those limits is desirable for
potentially subsequent expansions of the test facility.
4.2 Test Facility Dimensioning
In order to test the cavitation performance of the pump in-
For an operation of the pump in non-cavitating conditions, ducer, it is necessary that the mass flow coefficient ϕ, the cav-
the scaling methods of equation 7, equation 8 and equation itation number σ and the thermodynamic properties match
9 have to be applied. The maximal rotational speed of the the ones of the LOX application. Further, the turbopump
motor that drives the pump has been chosen to Nmax = has to be operated in regions where Re > 106 . As for non-
5500 r pm. The following considerations will approve this cavitating conditions, the geometry of the LOX hardware is
choice. Further, it is in good agreement with the rotational used without dimensional scaling.
speeds of comparable test facilities. Consequently, sub-scale The minimal necessary rotational speed for which Re = 106
tests will be performed. The geometry stays the same as is satisfied can be described, according to Equation 5, by
in the LOX hardware. The hardware designed for the LOX
turbopump will be used on the test bench without any scaling:
λgeom = 1. The efficiency of the pump on the test bench and υRe
ω≥ 2
= 400.55 r ad/s or N ≥ 3825.00 r pm (20)
the efficiency of the original LOX pump are assumed to be Rt,3
equal: ηm = ηa .
The thermodynamic properties are assumed to be compara-
With the properties of the pump under development at TUM, ble if the Dimensionless Bubble (DB) parameter, as denoted
given in Table 1 and Table 2, the maximal mass flow rate at in Equation 16, is equal or close to equal. For the LOX tur-
Nmax based on the design point of the original LOX hardware bopump, with the operational parameters given in Table 1
calculates to and the dimensions of the impeller given in Table 2, this
parameter calculates to DBLOX,re f = 0.306 for the impeller
of the TUM design. The plot in Figure 4 shows the LOX
Q a Nmax DB reference value DBLOX,re f as a constant line. Together
Q m, Nma x ≈ = 24.75 m3/h (17) with this constant value, the DB value calculated for different
Na
Design Procedure of a Turbopump Test Bench — 6/11
rotational speeds is shown in dependence of the water tem- inductive mass flow meters. The decay of the water quality
perature. It is clearly visible that for each rotational speed is slowed down as well due to the fact that no permanent
of the turbopump on the test bench, a specific water tem- access of light is present in the system. A major drawback of
perature has to be set in order to satisfy the equality of the a closed-loop system is that, in case of occurring cavitation,
Dimensionless Bubble (DB) parameter. With increasing ro- vapour bubbles can persist in the system and be sucked into
tational speed, the required water temperature rises. For the inlet again.
the maximal chosen rotational speed of the motor Nmax , the The drawing in Figure 3 shows an overview of the circuit.
water temperature iteratively calculates to TNma x = 368 K. It Necessary sensor positions, actuators and their positions
has to be pointed out that this value for the bulk water tem- within the loop are shown as well.
perature at the pump inlet only serves as design constraint
The water reservoir is realized by a stainless steel tank with
to the water heating system at this early point in the test
a volume of Vt ank = 2000 l. It is designed to withstand the
bench design procedure.
mechanical and thermal loads of water at pt ank,max = 4 bar
and Tt ank,max = 100◦ C. The tank is equipped with a EPDM
104 membrane filled with pressurized air. This membrane has a
1500rpm
3100rpm volume of Vmembr ane = 500 l. The air pressure within this
5500rpm membrane and therefore also within the tank is variable and
LOX-Reference can be controlled and regulated electronically. In operational
102 modes where cavitation occurs, it is very likely that bubbles
are transported into the tank and might disturb the pump
DB [-]
valves are equipped with an electronically controllable ac- tests with heated water. Therefore, a water heating system
tuation unit. This makes it possible to set the pressure drop is included in the test bench setup. This heating system is
across each valve independently. Further, the mass flow rate designed as a second auxiliary circuit that can be decoupled
can be controlled. This feature is needed in order to do a from the pump circuit. Thus, it is possible to heat the water
performance mapping of the pump under investigation and contained in the main tank to a desired temperature. The
it is a crucial requirement for cavitation scaling. water is heated up by an electrical heater with a power of
The mass flow rate is sensed at two locations. One inductive Pht,el = 60 kW. The water in the second auxiliary water
mass flow meter is placed in the inlet tubing directly in front circuit is driven by a separate pump. The heating system is
of the pump inlet. A second inductive mass flow meter is designed to control the temperature within ∆T1 = ±1 K and
placed after the pump. At the same position, a measurement can attain a maximal water temperature of Tmax = 100◦ C.
orifice is placed. This is where the pressure drop across a For means of flow control in the auxiliary heating circuit, the
defined through-flow area is measured. Thus, for a given fluid water temperature at the heater outlet, the static pressure at
density, the mass flow rate can be calculated in a second, inde- heater inlet and heater outlet as well as the mass flow rate
pendent way. This allows for a comparison of measurement are closely monitored. The mass flow rate is detected by a
results and the improvement of the measurement accuracy measuring orifice. The bladder inside the main tank serves
in all measurement ranges. as a compensation reservoir for volumetric changes due to
the heating of the water. A dedicated cooling system is not
For the controlled operation of the test bench, pressures and
implemented, but the insulation of the tank and the tubing
temperatures are captured at different locations of the water
is designed in a way that continuous heating is necessary to
circuit. The static pressure is monitored, as depicted in Figure
attain a constant high temperature.
3, in the pump inlet section, in the pump outlet section and
Figure 5 shows a CAD plot of the main circuit of the test
directly after the main throttle valve configuration. At all
facility that is currently being constructed at the Division
three locations, the pressure measurements are averaged over
Space Propulsion, TUM.
the circumference of the horizontally placed pipes in order
to compare for gravity effects. The measured static pressure
difference between pump inlet and pump outlet can be used
to calculate the head rise of the pump. The static pressure
difference over the throttling valve configuration allows for
a safe operation of the same. In addition, the tank pressure
is also monitored. Additionally, temperatures are detected at
all mentioned pressure sensor locations. Especially for tests
with heated water, these temperature readings yield valuable
information for the thermal control system.
The pump is driven by a three-phase alternating current (AC)
electric motor with a maximal power of Psha f t,max = 12 kW.
The motor reaches its maximal torque Tsha f t,max = 27 Nm
at its design speed Nr e f = 3100 r pm. The maximal speed of
the motor, without the use of any additional transmission, is
Nmax = 5500 r pm. The motor is connected to the shaft by a
flexible coupling that dampens the temporary high torque
during start-up. The shaft is held in position by a bearing
unit in overhung configuration - the bearings are positioned
between the pump and the drive unit. The bearing unit is de-
signed as an arrangement of a fixed bearing close to the pump Figure 5. CAD view of the turbopump test bench at TUM
and a floating bearing close to the drive unit. All loads on
the drive unit are closely monitored. This includes the torque
of the shaft, the rotational speed and the axial force acting 4.4 Numerical Design Methods
on the fixed bearing. Further, a bearing monitoring system In order to enhance the test facility development described
is established by permanent observation of the bearing race in the previous sections, the test bench was modelled nu-
temperatures. All bearings are run with grease lubrication merically in parallel with the physical setup. Therefore, the
and the grease quality is observed in fixed intervals. Further, water circuit including all pipes, valves and the tank have
the pump housing is equipped with three acceleration sen- been modelled in the software tool EcosimPro® . Especially
sors for the investigation of potential instabilities. The data all components that are foreseen to be electronically con-
obtained from those sensors can also be used to monitor the trolled were investigated numerically in detail. Therefore,
bearing operation. the heating system and the tank, including the membrane
As shown in section 2, in order to reproduce the cavitation bladder, were represented as detailed numerical models. For
behaviour of the LOX turbopump, it is necessary to run all components, the pressure drop across those components
Design Procedure of a Turbopump Test Bench — 8/11
50
80
60%
50% 40 1500 rpm
3100 rpm
60
5500 rpm 70%
H [m]
H [m]
60% 30
80%
40
70% 90%
80% 20
100%
90%
20 100% 10
0 0
0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35
Q [m3/h] Q [m3/h]
Figure 6. Numerically obtained system performance chart. Figure 7. Numerically obtained pump performance chart
The percentile values denote the level of opening of the for N = 1500 r pm, N = 3100 r pm & N = 5500 r pm. The
throttle valve configuration. system performance chart for different opening levels of the
throttle valve configuration are plotted in gray.
Operational envelope of the test facility at TUM Hmax NSS,min Pmax Qmax
10-1
Specific Diameter DS
100
101
102
10-3 10-2 10-1 100 101 102 103 104
Specific Speed NS
bench have been plotted in a Specific Speed - Specific Diam- For the second approach, a detailed model of the inlet, the
eter (NS − DS ) chart. The visualization of the boundaries is inducer, the impeller, the diffusor and the volute of the pump
shown in Figure 8. For this figure, the maximal mass flow developed at TUM are under development.
rate has been chosen to m max = 25 kgs . In general, the mass
flow rate of the test facility is not limited, but, according
NOMENCLATURE
to the system performance chart in Figure 6, the necessary
minimal head rise of the investigated pump increases with Symbols
increasing mass flow rate. A area
α thermal diffusivity
The numerical implementation of all components with Cp specific heat
EcosimPro® in parallel with the physical construction of the D diameter
test facility has shown to be beneficial. It was possible to DB Dimensionless Bubble parameter
investigate component dimensions at question numerically η efficiency
and the a priori understanding of the test facility was greatly Eu Euler number
improved. Especially the dimensioning of the water heating F force
system and the development of the inlet pressure control Fr Froude number
system in the tank were supported by numerical studies. Im- g gravitational acceleration. g = 9, 81 m/s2
plementations of simple circuit components like straight and H head rise
bent pipes, valves and tanks are available and the empirical hf g heat of vaporization
correlations are of good quality for the full operating region λ scaling factor
of the test bench. The preliminarily calculated performance m mass flow rate
chart derived from the generic pump model shows the same N rotational speed min−1
order of magnitude as the analytically scaled values. Any- √
Q
how, especially the implementation of the pump, based on NS specific speed. NS = N · H 0.75 √
Q
the generic pump component from the ESPSS library, does NSS suction specific speed. NSS = N · N PSH 0.75
not yield the desired amount of details. Two possibilities P power
have been identified in order to perform improved numer- p pressure
ical analyses: (1) Experimental measurement of the pump ψ head coefficient. ψ = 2gH/U32
performance map and implementing this information in a ϕ flow coefficient
module or (2) detailed modeling of the single pump compo- U peripheral velocity
nents. Both approaches are currently being followed at TUM. Q volume flow rate
Design Procedure of a Turbopump Test Bench — 10/11