Proceedings of ASME Turbo Expo 2013: Turbine Technical Conference and Exposition

GT2013
June 3-7, 2013, San Antonio, Texas, USA

GT2013-94483

STUDY OF AN AIR-OIL PUMP AND SEPARATOR SOLUTION FOR AERO ENGINE

LUBRICATION SYSTEMS

Johan Steimes1 François Gruselle1 Patrick Hendrick1

1 Aero-Thermo-Mechanics Department
Ecole polytechnique de Bruxelles
Université Libre de Bruxelles
Brussels, Belgium
Email: [email protected]

ABSTRACT R Radius.
Lubrication system of aero engines have to manage a two- e Fraction of entrained liquid.
phase flow of oil and air created in air-sealed bearing chambers. u Velocity.
A part of this flow goes through scavenge pumps and a de-aerator d Droplet diameter.
before returning to the tank, the other part passes through a de- σ Surface tension.
oiler before leaving the engine. The Aero-Thermo-Mechanics F Droplet distribution.
Department of Université Libre de Bruxelles is developing an δ Rosin-Rammler reference diameter.
original and integrated system to pump and separate a liquid- n Spread parameter.
gas flow. This system helps to reduce the complexity and the oil Γ Gamma function.
consumption of aero-engine lubrication systems by replacing the ṁ Mass flow rate.
de-oiler, the de-aerator and scavenge pumps. The prototype is q̇ Volume flow rate.
composed of an axial part (blades and metallic foam) that per- PLDR PASS Low Diameter Radial attack.
forms the separation and a radial part that pressurize the liq- PHDR PASS High Diameter Radial attack.
uid. De-oiling efficiency is measured with a radio-tracer mea- PHDD PASS High Diameter frontwards attack.
surement system. This paper presents the test results of four con- PHDI PASS High Diameter backwards attack.
figurations (different radii, flow paths, size of the metallic foam).
Droplet size measurements at the inlet and the outlet of the pro-
totype are also presented. A physical interpretation has been
INTRODUCTION
developed and is presented.
Lubrication systems in aero engines have two primary
functions: cooling and reduction of friction in bearings and
gears. To achieve them, different solutions exist depending
NOMENCLATURE on a set of constraints (integration, weight, cost, performance,
ω Angular velocity. reliability, severe and emergency operating conditions, etc.).
Cp Pressure coefficient. Therefore, a unique solution or architecture of aero engine
CR Radius coefficient. lubrication system does not exist. Nevertheless, they all have to
P Pressure. pump and separate an air/oil mixture [1]. Furthermore, a lot of
ρ Density. other applications have to deal with the pumping of a two-phase

1 Copyright
c 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 04/07/2014 Terms of Use: http://asme.org/terms

 flow of gas and liquid: oil extraction [2], flow in nuclear power
plants, pulp and paper processing [3, 4], etc. For this reason,
the Aero-Thermo-Mechanics department (ATM) of Université
Libre de Bruxelles (ULB) is studying a device to simultaneously
separate a gas/liquid flow and pressurize the liquid part of the
mixture [5–7].

Context
In aircraft gas turbine engines, the lubrication and cooling of
the bearings is performed by oil injection in air sealed chambers,
generating a two-phase flow of oil and air. This air is provided
by low-pressure compressor bleed air. Figure 1 shows a typical
engine lubrication system. A vent line collects the sealing air,
and brings it to a de-oiler where the oil droplet laden air is
cleaned before leaving through the overboard. In this example,
scavenge volumetric pumps collect an oil/air mixture from the FIGURE 1. AERO ENGINE LUBRICATION SYSTEM [1].
bearings compartments and the gearboxes. This mixture is
de-aerated in a cyclone separator located at the entrance of the
oil tank. The two-phase flow in the vent and scavenge lines is
respectively a slug / churn flow and an annular flow. This can perature. All those future evolutions impact the behaviour of the
be determined with two-phase flow maps, based on the working de-oiler, pumps and de-aerators.
conditions and the size of the pipes [8].

The de-oiling, de-aeration and pressure performances are

Content
important parameters for the scavenge part of lubrication sys-
tems. Insufficient pressure leads to oil leaks through the seals. This paper presents an integrated pump and separator
A poor de-aeration efficiency impacts the operation of the feed solution that could be installed in the scavenge lubrication
pump, and thus, the lubrication and cooling of the bearings. The system to replace the scavenge pumps, the cyclone separator,
de-oiling efficiency is directly linked to the oil consumption of and the de-oiler. It will insure the de-aeration, de-oiling and
the engine as the overboard vent represents the only outlet of the pressure function of the scavenge system. This would lead
scavenge system. This consumption has to be lowered regarding to a more reliable, simple and efficient system than the ones
environmental and economical factors, but also regarding the currently used. This prototype is also used to perform an in
limited in-flight oil quantity. depth study of the metallic foams used in de-oilers (as not much
relevant literature has been published [8]).
Two typical de-oilers designs exist. One uses a rotating
porous media to perform the separation and the other uses One of the difficulties related to oil/air separation is the mea-
centrifugal separation in an hollow shaft. The porous media surement and prediction of the separation efficiency. The ATM
is usually a metallic foam, realized by compressing together department of ULB has actively worked on this problem by ex-
metallic powders. This foam is available in different porosities perimental and theoretical ways and the results will be presented
(grades). The higher grade performs a better droplet separation, in this paper. The Pump And Separation System (PASS) is pre-
but presents an higher pressure drop [9]. sented and the influence of key parameters of lubrication system
is studied. A comparison of the size and the flow path of the
The need to reduce the fuel and oil consumption of aircraft metallic foam will be presented. It will be shown that the oil con-
engines will deeply change the architecture and the working con- sumption is linked to the size and distribution of the oil droplets
ditions. The labyrinth seals currently used will be replaced by in the inlet two-phase flow. The de-aeration efficiency is affected
brush seals which are more efficient [10]. They reduce the air by the working conditions and the results are presented herein.
flow rate and thus deeply change the nature of the two-phase flow A physical interpretation of the working principle of the pump
in the bearing chambers. The more-electric aircraft, the reduced and separator system will be presented. This study is a first step
fuel consumption, and the increase of rotating speed will change towards a model predicting the separation efficiency of current
the heat fluxes in the lubrication system and increase the oil tem- aero engine breathers.

2 Copyright
c 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 04/07/2014 Terms of Use: http://asme.org/terms

 TEST FACILITIES q̇oil q̇air T N Pout oil
Test bench The ATM Department has been working for
a long time on lubrication systems and owns a certified test bench (l/h) (m3 /h) (◦C) (RPM) (kPa)
[11]. This bench was modified to allow the study of aero engine
Min 160 0 20 1500 100
scavenge lubrication systems. It provides a gas/liquid mixture,
in varying proportions, to the inlet of the pump and separator Max 3500 300 180 4500 350
prototype. This is performed through mixing chambers that are
TABLE 1. TEST BENCH WORKING CONDITIONS
fed with air from a compressor and hot oil from a tank (through
a volumetric lubrication pump). The purpose of those cham-
bers is to simulate the air/oil mixture coming from the bearing
chambers. The air flow rate is measured by a vortex flow me- performs a continuous measurement of the fluid density of
ter (35 − 318 m3 /h, ±1, 5% of reading), the oil flow is measured multiphase flows. The liquid fraction is computed from the
by a turbine flow meter ( 450 − 4500 l/h, ±0.6% of reading). density of pure fluid (these measurements have been confirmed
The inlet pressure is taken as the sump pressure, with calibrated by visual observation of the flow downstream of the Coriolis
pressure transducers (0 -5 bars, 0,2% full scale precision). Pt100 meter, see [6]). This method is widely used in the industry and
thermal sensors are used (75, +200◦C, ±1◦C). Pt100 and pres- is able to measure a change in density of 0.2 kg/m3 [12].
sure transducers are also used at the air (0 - 2 bars, 0,2% full scale
precision) and oil (0 - 5 bars, 0,2% full scale precision) outlets. Measuring a very low oil consumption is challenging as it is
Data’s are collected with a 4-20 mA National Instruments acqui- due to very small oil droplets in an important air flow. The values
sition card and are post-processed with LabView. included in this paper were measured by a weighting method
and a radio-tracer method. For the first one, a filter (produced
by Eaton) is connected to the air outlet of the prototype and
captures the oil droplets. The filter is then weighted (on a ±0.1g
precision balance) to compute the de-oiling efficiency. This
kind of filter has a separation efficiency of over 60% for 1.5 µm
solid particles and a good oil adsorption capabilities for a low
pressure drop [13]. The second method uses a radio-active tracer
and has been developed in collaboration with Delta Service
Industriels (DSi) who has a great experience with this method
on car turbochargers [14]. The oil is loaded to a radioactivity
of 1 MBq with a Germanium-69 tracer. This isotope has a
half-life of 39 hours and emits gamma rays between 511 and
1336 kEv. The tracer is placed on a molecule of oil with a
boiling point of 385◦C. The outlet of the prototype is connected
to a filter surrounding a large probe that measures the emitted
radioactivity. This filter is separated from the bench by the
means of a lead wall. The increase in radioactive concentration
in the filter is linked to the (radio traced) oil quantities. This
method allows to reach higher working temperatures and to have
an online measurement of the oil consumption. Depending on
FIGURE 2. SCHEMATIC OF THE LUBRICATION TEST BENCH. the oil consumption, the tests last between 10 and 20 minutes
to have accurate measurements. The two oil consumption
measuring methods were in good accordance.
The maximum and minimum working conditions of the
scavenge system on the ATM bench are showed in Tab. 1.
The bench is also equipped with an Helos laser diffrac-
tion (LD) system, from Sympatec [15]. This system includes
Measure of the de-aeration, de-oiling and granu- two lenses to measure droplet sizes between 0.9 − 175 µm and
lometry The de-aeration (at the liquid outlet) and de-oiling 9−1750 µm. As it is inherent to LD systems, the lenses have dif-
(at the gas outlet) efficiencies are, with the back-pressure, the ferent working distances between the detector and the measuring
most interesting parameters to measure for our application. zone (respectively 96 and 566 mm). The detector is connected
The liquid volume fraction (pipe liquid content) at the liquid to a PC for post-processing. This LD system was chosen due to
outlet is measured with an Emerson Coriolis flow meter, which its large particle size spectrum and for its simplicity that fits the

3 Copyright
c 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 04/07/2014 Terms of Use: http://asme.org/terms

 the liquid film (Patm ). No difference was observed in the de-
aeration efficiency of the three PASS configurations with a high 100 28°C
40°C
radius. Figure 4 presents the comparison of the liquid fraction 70°C
110°C
at the liquid outlet of the PLDR and PHDR (influence of the di-

Liquid Fraction (%)

95
ameter and of the closed impeller for rotation speeds between
1500 and 4500 RPM). It also shows the influence of the flow
90
rate on the liquid fraction for the PHDR. It is clear that the PASS
with the closed rotor performs a much better de-aeration. One
of the reasons can be that the open rotor allows the liquid film 85

to be torn apart at the interface between the stator and the rotor.
It is difficult to determine the impact of the impeller radius as it 80
is taken into account in the non-dimensional pressure coefficient.
0 5 10 15 20 25 30 35
One can also notice that increasing the oil flow rate for a constant Cp(%)
C p (i.e. same rotational speed and pressure) decreases the liquid
fraction.
FIGURE 5. DE-AERATION EFFICIENCY FOR DIFFERENT
TEMPERATURES.

100
performances will be affected by the changes in oil flow rate (but
this will also change the back-pressure and the rotational speed)
and the temperature of the oil.
Liquid Fraction (%)

95

90
Physical interpretation It has been shown experimen-
tally that the de-aeration efficiency depends on the rotational
85 speed and on the back-pressure: they determine the position of
the gas-liquid interface. With the assumptions that the gas-liquid
PLDR
PHDR 1275 l/h separation is perfect, that the axial and radial velocities are neg-
80 PHDR 2550l/h
PHDR 3250l/h ligible relative to the tangential velocity and by neglecting the
0 5 10 15 20
Cp(%)
25 30 35 friction losses, the momentum equation inside the impeller leads
to Eqn. 2. It links the position of the gas-liquid interface (R, see
Fig. 3) to the pressure coefficient C p .
FIGURE 4. DE-AERATION EFFICIENCY OF THE PHDR AND
PLDR.
1 R
C p = 1 − ( )2 (2)
2 R2
Figure 5 shows the influence of the temperature on the
liquid fraction. The Coriolis flow meter is not able to measure According to this definition, a constant value of C p means
de-aeration rates lower than 80% and this explains the data a constant film thickness (R) that should lead to the same de-
located at a constant value of 80%. Below this value, and aeration efficiency. Figures 4 and 5 show that this is not the case:
according to the visualization cell placed after the sensor, increasing the oil flow rate lowers the liquid fraction. This may
reducing the C p also reduces the liquid fraction further (this be due to the higher number of separated droplets impacting the
value of 80% should be lowered). Increasing the temperature gas-liquid interface. In this case air re-entrainment is more likely
for a constant C p has an important impact until 70◦C. Up to to occur. High flow rates also mean high drag forces on the air
110◦C, the influence is limited, but the effect could be more bubbles and the buoyancy effect has thus a lower impact on the
important if the values lower than 80% were taken into account. separation. These hypothesis need to be further more studied.
Furthermore, increasing the temperature leads to small changes Increasing the temperature lowers the viscous losses. Figure 6
in the pressure and a constant need to adjust the oil flow rate. shows the decrease in dynamic viscosity (of Mobil Jet Oil II)
This explains the spread of the results at higher temperatures. and in liquid fraction with the temperature. This trend is able
to explain the behaviour of Fig. 5 as there is a big impact of
As a conclusion, it is preferable to design a PASS system the viscosity between low and mid temperatures (70◦C), but not
with a closed rotor. Nevertheless, on a full engine cycle, the any more at higher temperatures (110◦C). The viscous losses in

5 Copyright
c 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 04/07/2014 Terms of Use: http://asme.org/terms

 the liquid film (Patm ). No difference was observed in the de-
aeration efficiency of the three PASS configurations with a high 100 28°C
40°C
radius. Figure 4 presents the comparison of the liquid fraction 70°C
110°C
at the liquid outlet of the PLDR and PHDR (influence of the di-

Liquid Fraction (%)

95
ameter and of the closed impeller for rotation speeds between
1500 and 4500 RPM). It also shows the influence of the flow
90
rate on the liquid fraction for the PHDR. It is clear that the PASS
with the closed rotor performs a much better de-aeration. One
of the reasons can be that the open rotor allows the liquid film 85

to be torn apart at the interface between the stator and the rotor.
It is difficult to determine the impact of the impeller radius as it 80
is taken into account in the non-dimensional pressure coefficient.
0 5 10 15 20 25 30 35
One can also notice that increasing the oil flow rate for a constant Cp(%)
C p (i.e. same rotational speed and pressure) decreases the liquid
fraction.
FIGURE 5. DE-AERATION EFFICIENCY FOR DIFFERENT
TEMPERATURES.

100
performances will be affected by the changes in oil flow rate (but
this will also change the back-pressure and the rotational speed)
and the temperature of the oil.
Liquid Fraction (%)

95

90
Physical interpretation It has been shown experimen-
tally that the de-aeration efficiency depends on the rotational
85 speed and on the back-pressure: they determine the position of
the gas-liquid interface. With the assumptions that the gas-liquid
PLDR
PHDR 1275 l/h separation is perfect, that the axial and radial velocities are neg-
80 PHDR 2550l/h
PHDR 3250l/h ligible relative to the tangential velocity and by neglecting the
0 5 10 15 20
Cp(%)
25 30 35 friction losses, the momentum equation inside the impeller leads
to Eqn. 2. It links the position of the gas-liquid interface (R, see
Fig. 3) to the pressure coefficient C p .
FIGURE 4. DE-AERATION EFFICIENCY OF THE PHDR AND
PLDR.
1 R
C p = 1 − ( )2 (2)
2 R2
Figure 5 shows the influence of the temperature on the
liquid fraction. The Coriolis flow meter is not able to measure According to this definition, a constant value of C p means
de-aeration rates lower than 80% and this explains the data a constant film thickness (R) that should lead to the same de-
located at a constant value of 80%. Below this value, and aeration efficiency. Figures 4 and 5 show that this is not the case:
according to the visualization cell placed after the sensor, increasing the oil flow rate lowers the liquid fraction. This may
reducing the C p also reduces the liquid fraction further (this be due to the higher number of separated droplets impacting the
value of 80% should be lowered). Increasing the temperature gas-liquid interface. In this case air re-entrainment is more likely
for a constant C p has an important impact until 70◦C. Up to to occur. High flow rates also mean high drag forces on the air
110◦C, the influence is limited, but the effect could be more bubbles and the buoyancy effect has thus a lower impact on the
important if the values lower than 80% were taken into account. separation. These hypothesis need to be further more studied.
Furthermore, increasing the temperature leads to small changes Increasing the temperature lowers the viscous losses. Figure 6
in the pressure and a constant need to adjust the oil flow rate. shows the decrease in dynamic viscosity (of Mobil Jet Oil II)
This explains the spread of the results at higher temperatures. and in liquid fraction with the temperature. This trend is able
to explain the behaviour of Fig. 5 as there is a big impact of
As a conclusion, it is preferable to design a PASS system the viscosity between low and mid temperatures (70◦C), but not
with a closed rotor. Nevertheless, on a full engine cycle, the any more at higher temperatures (110◦C). The viscous losses in

5 Copyright
c 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 04/07/2014 Terms of Use: http://asme.org/terms

 0.05 100 250
Dynamic viscosity PHDR
Liquid fraction PLDR

200

Oil consumption (g/h)

Dynamic viscosity (St)

Liquid fraction (%)

150

100

50

0 80 0
0 20 40 60 80 100 120 50 100 150 200
Temperature (°C) 3
Air flow rate (Nm /h))

FIGURE 6. LIQUID FRACTION AND VISCOSITY IN FUNC- FIGURE 7. OIL CONSUMPTION FOR THE PHDR AND PLDR
TION OF TEMPERATURE. WITHOUT METALLIC FOAMS.

the impeller and the volute change the position of the interface 40
as more viscous flow needs a thicker liquid film to produce the PHDI
35 PHDD
same pressure. The hypothesis of neglecting the viscous losses PHDR
should be reconsidered to lead to a more accurate modelling of 30 PLDR
the position of the interface. Oil consumption (g/h) 25

20

De-oiling efficiency 15
Without the metallic foam Tests without the metallic
10
foam, performed on the PHDR and PLDR, allow to understand
how the increase of the radius of the inlet and the blades affects 5
the de-oiling efficiency. Figure 7 shows the oil consumption in
0
function of the air flow rate for the two prototypes. The influ- 50 100 150 200
Air flow rate (Nm3/h))
ence of the diameter is important and increases with the air flow
rate. At the nominal value, the oil consumption is three times
higher with the PLDR. As expected, increasing the radius has an FIGURE 8. DE-OILING EFFICIENCY OF THE FOUR PROTO-
important effect on the cleanliness of the flow reaching the foam. TYPES.

With the metallic foam Tests with the metallic foam the end of the blades is not more centrifuged by the vanes and is
represent the oil consumption of the prototypes after three sep- likely to easily tear out droplets from the liquid film. These big
aration stages. Figure 8 shows the results for all the prototypes droplets are then taken to the bottom of the rotor, and can more
in function of the air flow rate. The error bars represent the easily pass through the foam. Nevertheless building a rotating
standard deviation for all the performed tests. Some changes u-turn is difficult so this solution will not be studied any more
of the oil consumption were observed with small changes of (especially regarding the results of the other configurations).
temperature of the test bench (±5◦C), but they are not always
affecting the consumption the same way. The small differences The other configurations perform quite well. Regarding the
can be interpreted as measurement uncertainties, but big changes inlet oil flow rate, this corresponds to de-oiling efficiencies of
are representative of an abnormal behaviour of the prototype. 99,9998 to 99,9994%. There is no relevant difference between
the oil consumption of the PHDD, PHDR and PLDR. The
The oil consumptions of the PHDI are always higher than results obtained without the metallic foam seem to indicate
the consumption of the 3 other prototypes. Furthermore, the that whatever the flow path of the mixture, and unregarding the
standard deviation shows that these results are not repeatable. diameter of the prototype (until a certain limit), it is the metallic
This is due to a design error. The air that is making a u-turn at foam that regulates the de-oiling efficiency.

6 Copyright
c 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 04/07/2014 Terms of Use: http://asme.org/terms

 40
The PHDR can be changed to a configuration in which only PHDD 3 foams
PHDD 1 foam
35
the first or last slice of foam performs the separation (the flow
goes through a small/all the length of the vanes). Figure 9 shows 30

Oil consumption (g/h)

that there is no clear difference in the oil consumption of the two
25
layouts. The difference observed at higher flow rates is still in
the range of the measurement uncertainties. 20

15

10
40
PHDR 3 foams 5
PHDR 1 frontal foam
35
PHDR backwards foam 0
50 100 150 200
30 3
Air flow rate (Nm /h))
Oil consumption (g/h)

25

20 FIGURE 10. IMPACT OF THE NUMBER OF METALLIC FOAMS

ON THE DE-OILING EFFICIENCY OF THE PHDD.
15

10

5
6
0 PHDR 1 foam 2550l/h
50 100 150 200
5 PHDD 1 foam 2550l/h
Air flow rate (Nm3/h))
PHDR 3 foams 2550l/h
PCore ! Patm (kPa) 4 PHDR 3 foams 1275l/h
PHDR 3 foams 3250l/h
FIGURE 9. IMPACT OF THE NUMBER OF METALLIC FOAMS 3
ON THE DE-OILING EFFICIENCY OF THE PHDR.
2

1
The same study was performed on the PHDD, but in this
0
case the prototype is studied only using the front foam (see Fig.
3 ). Figure 10 shows that no significant difference is observed !1
between the configuration with full foam and front foam. 0 50 100 150 200 250
Air flow rate (Nm3/h)
As a primary conclusion, one can consider that the flow
path in the metallic foam has no impact on the de-oiling effi- FIGURE 11. PRESSURE DROP BETWEEN THE INLET OF THE
ciency of the prototype. Furthermore, reducing the size of the METALLIC FOAM AND THE AIR OUTLET.
metallic foam is possible as it does not affect the performances.
Nevertheless, reducing the number or the size of the metallic
foams increases the pressure drop through the system. This
pressure drop can be crucial as it will affect the pressure (and the
air flow rate) in the bearing chambers. This can be seen on Fig. Studying the influence of these parameters on the de-oiling
11. The most important pressure drop comes with the PHDD efficiency helps to have a deeper understanding of the behaviour
configuration with one metallic foam (lowest cross-section). of engine breathers. The influence of the air flow rate and of the
Pressure drop in porous media increases with the speed of the temperature is showed in Fig. 12. These results are presented for
flow and the length of the media. Reducing the number of foams the PHDR. This figure resumes that increasing the air flow rate
reduces the cross-section: for the same flow rate, the pressure and the temperature increases the oil consumption of the PASS.
drop due to the flow speed in the foam increases more than the The more the air flow rate increases, the more the temperature
reduction due to the length. impacts the oil consumption. Figure 13 shows that increasing
the oil flow increases the oil consumption. The trend of the
Aero engine lubrication systems can work from -50 to curves is the same for different temperatures, but shifted to
180◦C. The temperature and the oil flow rate is expected to higher oil consumptions.
increase in the future while the air flow rate will be lowered.

7 Copyright
c 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 04/07/2014 Terms of Use: http://asme.org/terms

 40
flow rate is equal to 160 l/h and at a temperature of 33◦C. The in-
PHDR 34°C crease of the air flow rate has an important impact on the droplet
PHDR 55°C
35
PHDR 70°C distribution. Increasing the air flow rate from 60 to 120 m3 /h
30 changes the cumulative distribution for 100 µm droplets from 3
Oil consumption (g/h)

to 30% of the total droplet volume. This means that there is 30%
25
of the volume contained in droplets with a diameter lower than
20 100 µm. As bigger droplets take a lot of volume, this volume
15
distribution does not show that there is a lot of small droplets
(that are harder to separate). The Sauter-Mean Diameter is a pa-
10
rameter used to characterize this effect1 . In this case, the SMD is
5 50 µm. Increasing the oil flow rate has not always a clear impact
0
on the SMD, but it increases the total number of droplets. This
50 100 150 200 250 explains why increasing the oil flow rate increases the oil con-
3
Air flow rate (Nm /h))
sumption. Figure 15 shows the influence of the air and oil flow
rates on the SMD. No measures could be obtained at higher tem-
FIGURE 12. DE-OILING EFFICIENCY FOR TWO WORKING peratures or flow rates as they are too many droplets to perform
TEMPERATURES. a correct measurement with the LD system.

40
PHDR q 80m3/h 100
air

Cumulative distribution ! volume (%)

120m3/h
35 PHDR q 160m3/h
air 90 100m3/h
80m3/h
30 80
60m3/h
Oil consumption (g/h)

70
25
60
20
50

15 40

10 30

20
5
10
0
0 500 1000 1500 2000 2500 3000 3500 4000 0
Oil flow rate (l/h)) 0 100 200 300 400 500 600 700 800 900 1000
d (µm)

FIGURE 13. DE-OILING EFFICIENCY IN FUNCTION OF THE

FIGURE 14. DROPLET DIAMETER DISTRIBUTION AT THE IN-
OIL FLOW RATE FOR TWO AIR FLOW RATES.
LET OF THE PASS.

Granulometry study A granulometry study was per-

Measurements performed at the outlet of the prototype are
formed at the inlet and the outlet of the separator to get a deeper
presented on Fig. 16. The outlet flow is taken as a whole and
understanding of the phenomenon behind the behaviour of
guided to the measurement system with a 2 m pipe. There is a
the de-oiling efficiency. The changes in size and number of
single droplet distribution at the outlet of the PASS for all the
the inlet droplets is suspected to have an important impact on
working conditions and oil consumptions. No particles with a
the de-oiling performances. The inlet flow is an annular flow.
diameter bigger than 3 µm leave the separator. The SMD is al-
Compared to other annular flows studied in the literature [18],
ways around 1 µm but does not take into account the droplets
it shows a higher liquid mass flux and a lower air mass flux. A
with a diameter lower than 0, 5µm. A study performed by Wil-
new system had to be developed to extract the droplets from the
lenborg et al. [8], with a more accurate system, computed a SMD
flow and deviate the liquid film [19]. The laser is located just
of 0, 5 µm for the same metallic foam. If the really small droplets
before the inlet of the separator.
were taken into account by the Sympatec diffraction system, the
same value could be found.
Figure 14 shows the volume droplet distribution at the inlet
for different air flow rates. The prototype has two tangential in-
lets and the values given here are for one of the inlets. The oil 1 It represents a diameter of a sphere having the same volume/surface area

ratio as the whole droplet distribution.

8 Copyright
c 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 04/07/2014 Terms of Use: http://asme.org/terms

 300
complicated as most correlation found in the literature are semi-
qoil = 160l/h empirical. A recent study [21] proposed an improved version of
q = 320l/h
250
oil
qoil = 640l/h
the correlation developed by Azzopardi [18] and is computed as
Eqn. 4. More information on the different parameters can be
200 found in [21] and [18]. The dependence of the mean droplet size
SMD (µm)

with the liquid Reynolds number is not unanimous (as Fig. 15

150 showed it). Nevertheless, of all the proposed parameters, a value
of A = 0.44 and B = 18.6 estimates the SMD with the least error.
100
The spread parameter should then be 1.75 [19].
50

S
0
50 60 70 80 90 100 110 120 130
ρGas (uSGas )2 λA −0,58 euLiquid
3
d32 = λA [A( ) + B( S )] (4)
Air flow rate (m /h) σ uGas

FIGURE 15. SMD AT ONE OF THE INLETS OF THE PASS FOR The modelling of the separation process inside the proto-
DIFFERENT WORKING CONDITIONS. type, without the metallic foam, is presented in [5]. Nevertheless,
the test result of the present paper showed that the de-oiling
100
efficiency is imposed by the metallic foam and having a model
of its separation process would be helpful to design future proto-
Cumulative Volume distribution (%)

90
80m3/h types. As shown by the test results in Fig. 16, the granulometry
80 160m3/h
200m/h at the outlet of the prototype is made of droplets with a diameter
70 Rosin Rammler distribution
lower than 3 µm with a SMD is about 0.5 to 1 µm. The droplet
60 distribution at the outlet can be modelled by a Rosin-Rammler
50 distribution that fits the experimental data with a SMD of 0.9 µm
40 and a spread parameter of 2.3 (see Fig. 16).
30

20
Modelling the de-oiling efficiency could consist in applying
10
a cut-off size to the droplet distribution in the inlet if some as-
sumptions can be verified: no droplet coalescence, break up or
0
0 0.5 1 1.5 2 2.5
d (µm)
3 3.5 4 4.5 5
creation in the prototype and no droplet is removed from the liq-
uid film. Nevertheless, one has to be certain that no droplets are
created by the shear flow in the metallic foam. The exact number
FIGURE 16. DROPLET DIAMETER DISTRIBUTION AT THE of each droplet is needed to compute the oil consumption deriv-
OUTLET OF THE PASS. ing from this distribution. This data is not given by the diffrac-
tion system. Assuming that there is no droplet fragmentation,
Inlet and outlet flow prediction For future modelling computing the area under the inlet curve between 0 and 3 µm
of the flow in the PASS, it is interesting to predict the flow in the overestimates the oil consumption has not every droplet between
inlet pipes. It can be modelled as a Rossin-Rammler distribution, 1 and 3 µm participates to the oil consumption (see Fig. 16).
defined in Eqn. 3 [20], in function of the particle diameter d. This first step towards a metallic foam model has to be compared
to experimental results to be confirmed and will be the subject of
future papers.
d
Fm (d) = 1 − exp[−( )n ] (3)
δ
SUMMARY AND PERSPECTIVES
where δ and n are two empirical constants called, respec- Different prototypes of a Pump And Separation System
tively, the Rosin-Rammler reference diameter and the spread pa- have been tested on an aero engine lubrication test bench by the
rameter. The reference diameter [20] is computed based on the ATM department of ULB. The system works successfully for
Sauter Mean Diameter (d32 )2 . Modelling the d32 can be quite different working conditions. Four configurations were tested
allowing to study changes in diameter, open or closed rotor and
the size and attack of the metallic foam. Test results highlighted
2δ = d32 Γ(1 − n1 ) that the diameter is playing a role only when the separator

9 Copyright
c 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 04/07/2014 Terms of Use: http://asme.org/terms

 is used without the metallic foam. Otherwise, it is possible [6] Gruselle, F., Steimes, J., and Hendrick, P., 2011. “An in-
to reduce the size of the PASS as it is possible to reduce the novative two-phase flow pump and separator solution”. In
thickness and the diameter of the metallic foam without affecting Proceedings of ASME Turbo Expo 2011, Vancouver.
the de-oiling efficiency of the prototype. It was also showed [7] Steimes, J., Ruelle, A., Gruselle, F., and Hendrick, P., 2011.
that the future changes in the lubrication system will impact the “Study of air-oil separation and pumping devices for gas
separation efficiency as the air/oil flow rates and temperature turbine engines”. In Proceedings of 20th conference of the
have an important impact on the de-oiling efficiency. With the International Society for Air Breathing Engines (ISABE),
help of laser diffraction measurements it was also shown that the Goteborg.
increase of oil consumption with the air flow rate is due to the [8] Willenborg, K., Klingsporn, M., Tebby, S., Ratcliffe, T.,
increase of the number of small droplets in the inlet pipes. Gorse, P., Dullenkopf, K., and Wittig, S., 2008. “Experi-
mental analysis of air/oil separator performance”. Journal
The de-aeration efficiency is dependent on the rotational of Engineering for Gas Turbines and Power, 130(6).
speed and on the back-pressure. An open rotor badly affects the [9] ERG, 2011. “Company website:”.
separation performances. At the same working-points, the liquid http://www.ergaerospace.com/index.html.
fraction is decreased by 15% by comparison with a closed rotor. [10] Chew, J., Lapworth, B., and Millener, P., 1995. “Mathe-
The last one is showing a good de-aeration efficiency at a wide matical modeling of brush seals”. International Journal of
spectrum of working points. This efficiency is also affected by Heat and Fluid Flow, 16(493-500).
the flow rate and the temperature of the oil. [11] Hendrick, P., Berten, O., Gruselle, F., Raimarckers, N., and
Saive, G., 2008. “Test benches of innovative systems and
A physical interpretation of the different phenomenon components for aircraft gas turbine engines”. In Proceed-
occurring inside the prototype was presented. A first step ings of Lubmat conference, San Sebastian.
towards a model of the oil consumption is presented. It will be [12] Emerson, 2011. “Company website:”. www.emerson.com.
useful to predict the performances for working conditions that [13] Eaton, 2011. “Company website:”. www.eaton.com.
are difficult to test on the bench (low atmospheric pressure, low [14] Delvigne, T., Benot, D., Jerome, O., Philippe, C., and Car-
oil temperatures, high back pressure, ...). lier, P., 2005. “A new methodology for on line lubricant
consumption measurement”. In 2005 SAE Brasil Fuels &
Future works include a new PASS design, with smaller di- Lubricants Meeting, Rio de Janeiro.
mensions and a closed rotor to optimize the size and the weight. [15] Sympatec, 2011. “Company website”.
Improvements of the theoretical study will also be done through http://www.sympatec.com/.
defining new accurate correlations for the inlet and outlet flows. [16] Chan, A., Barreca, S., and Hartlen, R., 1991. “An ex-
The modelization of the oil consumption from the droplet distri- perimental study of centrifugal pump performance under
bution at the outlet of the metallic foam needs to be assessed. steam-water two-phase flow conditions at elevated pres-
sures”. Cavitation and Mutiphase Flow Forum, 109,
pp. 111–117.
REFERENCES [17] Retimet, D., 2011. “Company website:”.
[1] Linke-Diesinger, A., 2008. Systems of commercial turbo- http://www.dunlop-equipment.com/.
fan engines: an introduction to systems functions. Springer [18] Azzopardi, B. J., 1985. “Drop sizes in annular two-phase
Berlin Heideleberg. flow”. Experiments in Fluids, 3, pp. 53–59.
[2] Maier, W., Chochua, G., and Biba, Y., 2010. “An innovative [19] Steimes, J., Diakodimitris, C., di Matteo, M., and Hendrick,
two-phase flow pump and separator solution”. In Proceed- P., 2012. “A laser diffraction study on droplet diameters in
ings of ASME Turbo Expo 2010, Glasgow. two lubrication systems”. In Proceedings of 16th Int Symp
[3] Chan, A., Kawaji, M., Nakamura, H., and Kukita, Y., 1999. on Applications of Laser Techniques to Fluid Mechanics,
“Experimental study of two-phase pump performance using Lisboa.
a full size nuclear reactor pump”. Nuclear Engineering and [20] Lefebvre, A., 1988. Atomization and Sprays. CRC Press
Design, 193, pp. 159–172. Taylor & Francis Group.
[4] Furuya, O., 1985. “An analytical model for prediction [21] Simmons, M. J. H., and Hanratty, T. J., 2001. “Droplet
of two-phase (non-condensable) flow pump performance”. size measurements in horizontal annular gas-liquid flow”.
ASME Journal of Fluids Engineering, 107(1), pp. 139– International Journal of Multiphase Flow, 27(5), pp. 861 –
147. 883.
[5] Gruselle, F., Steimes, J., and Hendrick, P., 2011. “Study of
a two-phase flow pump and separator system”. Journal of
Engineering for Gas Turbines and Power, 133(6).

10 Copyright
c 2013 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 04/07/2014 Terms of Use: http://asme.org/terms