Computational and experimental analysis of

the churning power losses in an industrial
planetary speed reducer
Franco Concli, Carlo Gorla
Politecnico di Milano, Dipartimento di Meccanica, Milan, Italy

Abstract
Advantages of planetary speed reducers are well known and due to their compact
design and power density they are suitable for a wide range of applications.
Efficiency is becoming more and more a main concern in the design of power
transmissions and the demand for high efficiency gearboxes is continuously
increasing.
For this reason it’s important to have some models in order to quantify the power
losses already during the design stage. Some theoretical or semi-empirical
models that allow to estimate losses like those of bearings, of seals, of gear
meshing (due to sliding) and, for ordinary gears, those of churning are available
in literature. In the case of planetary speed reducers, where the motion of the
planet carrier causes a rotatory motion of the planets around the axis of the
gearbox, the oil splash lubrication is an important source of losses.
This report introduces a multiphase CFD model for the prediction of the
churning losses characteristic of planetary gears. The analysis has been carried
out by means of an unsteady Volume of Fluid (VOF) model and implemented on
a commercial software (Fluent). The whole geometry of the speed reducer and
many operating conditions like the rotational speed, the oil level and the
operating temperature have been taken into account. Moreover the results of an
experimental testing campaign on an especially designed gearbox are presented
and compared with the computation ones in order to validate the model. The two
approaches gives results in good agreement.
Keywords: Planetary speed reducers, Churning losses, Multiphase flow
simulation, VOF

1 Intrroduction
In the last years, effficiency is be coming more and more a main
m concernn also in
the dessign of power transmissionss and the dem
mand for high efficiency geaarboxes
is conttinuously incrreasing. Just hhaving appropriate modelss to predict thhe final
efficienncy of the speeed reducer [11] it is possible to take the correct
c
choicees since
the deesign step, av
voiding wastee of time an
nd money. So
ome models for the
estimattion of the diifferent sourcees of losses liike those of bearings
b
[2], oof seals
[3], off gear meshin
ng (due to ssliding) [3] and,
a
for ordin
nary gears, thhose of
churninng [4] already
y exist. What is still missin
ng, is an appro
opriate modell for the
predicttion of the ch
hurning lossees in planetary
y speed reduccers. In this kkind of
gearingg, in fact, the motion of the planetary
y gears due to
t the planet carrier
rotationn is an additio
onal source off losses.
Aim off this study is to provide a model for thee correct estim
mation of this kind of
losses and, consequeently, for the ccorrect estimaation of the eff
fficiency of thee whole
transm
mission. The model
m
has beeen performed by mean of a computationnal fluid
dynam
mic (CFD) anaalysis that sim
mulates the behaviour
b
of the air-oil luubricant
mixturre.
To validate the num
merical model,, an industriall planetary speeed reducer hhas been
convenniently modifiied in order too be able to measure
m
the churning
c
lossees alone
causedd by the motio
on of the planeetary gears. This
T gearbox has
h been testeed on an
especiaally designed test rig.

2 Prooblem defin
nition
Figure 1 shows an example
e
of a planetary geaaring: it can be
b seen the suun gear,
the plaanetary gears, the planet carrrier and the ex
xternal crown.

Figure 1: planetary gearing

In the common con
nfiguration, thhe power flow
ws from the sun
s gear shaft
ft to the
planet carrier shaft. The planetaary gears hav
ve therefore two
t
rotatory motion
compoonents: the firsst one is a rottation around their axis and
d the second oone is a
motionn with a circu
ular path arouund the gearb
box axis due to the planett carrier
rotationn on which th
hey are mounteed.
The planetary speed
d reducers are generally oil splash lubricated and it is jjust the
interacction between the rotating elements and
d the lubrican
nt air-oil mixtture the
source of losses inveestigated by thhe presented model.
m
The inffluence param
meters are the oil level and its
i properties, such as the vi
viscosity
and thee density (fun
nctions of the temperature), the geometry and, of couurse, the
rotationnal speed. Th
he transmitted torque influences this kind
d of losses inddirectly:
increassing the transm
mitted torque means more load dependent power lossses (like
the meeshing ones and
a part of th
the bearing ones) and, con
nsequently, a higher
regimee temperature of the lubricannt.

3 Geoometry mod
difications
The iniitial geometry
y of the analyssed speed redu
ucer is shown in figure 2.

Fig
gure 2: initial geometry of the
t speed redu
ucer
i composed bby a sun gearr, 3 planetary gears and a ““double
The sppeed reducer is
disk” planet
p
carrier. Both, the inpuut and the outtput shaft, are mounted by m
mean of
2 beariings. Two contact seal aree also presentt to avoid leak
kage of lubriccant. In
order to evaluate thee churning pow
wer losses related to the mo
otion of the pllanetary
gears around
a
the gearbox axis duue to the rotattion of the plaaner carrier onnly, the
originaary geometry of the speedd reducer has been modifieed. The teethh of the
externaal crown have been removved by mach
hining and thee sun gear haas been
substituuted with a sm
maller and cyllindrical (with
hout teeth) com
mponent. Thee aim of
this moodifications iss to prevent thhe engaging off the gears, av
voiding slidingg losses
and chhurning lossess due to the rotation of the
t gears around their axiis. This

modified geometry has been uused both for the numerical model aand the
experim
mental testing
g campaign.

4 Mod
del set up
4.1 Geeometry and Mesh
The doomain for the CFD
C
analysis is the internaal free volume of the speed reducer
after thhe above desccribed geomeetry modifications. Figure 3 shows this volume
markedd in yellow.

Figure 3: modified geometry
g
of tthe speed redu
ucer: the com
mputational doomain is
marked in yellow
The coomputational domain
d
for thee CFD analysiis has been modelled by meean of a
3D cadd software an
nd discretizedd with a swep
pt mesh. Thiss meshing tecchnique
consistts in creating a mesh on onne side of the region, know
wn as the sourc
rce side,
and theen copying th
he nodes of thhat mesh, onee element layeer at a time, uuntil the
final siide, known as the target sidde, is reached.
The whhole model haas been discreetized with ab
bout 500000 hexahedral
h
cellls. This
kind off elements allows a larger aaspect ratio compared with
h the tetrahedrral cells
in whicch it will invaariably affectss the skewnesss of the cell, which is undeesirable
as it may
m impede acccuracy and coonvergence.
4.2 Paarameter settiing
A VO
OF multiphasee approach hhas been cho
osen for the analysis. Thhe VOF
formullation relies on
o the fact tthat two or more fluids (or phases) are not
interpeenetrating. Th
he fields for aall variables and propertiees are shared by the
phases and represen
nt volume-aveeraged values, as long as th
he volume fracction of
each off the phases iss known at eacch location. Thus
T
the variab
bles and propeerties in
any giiven cell aree either pureely representaative of one of the phaases, or
represeentative of a mixture
m
of thhe phases, dep
pending upon the volume ffraction
values.. The tracking
g of the interfface between the
t phases is accomplishedd by the
solutioon of a continu
uity equation ffor the volumee fraction of one
o of the phas
ases

1

where
is the density of the
phase,
the volume fraction of that phase,
and
the mass transfer from phase
to phase
and vice versa
respectively. This volume fraction equation has been solved be mean of an
explicit scheme discretisation in order to avoid the numerical diffusion. That
means solving
2

,

where
1 and are the indexes for the new time step and the preious one,
volume fraction computed with the Geo, is the face value of the
is the volume flux on
Reconstruction scheme, is the volume of the cell and
the face. The Geo-Reconstruction approach is an accurate scheme that assumes
that the interface between two fluids has a linear slope within each cell, and uses
this linear shape for the calculation of the advection of fluid through the cell
faces. This scheme avoids the numerical diffusion but needs an accurate grid.
The properties appearing in the above transport equations are determined by the
presence of the component phases in each control volume. In a two-phase
system, if the phases are represented, for example, by the subscripts air and oil,
and if the volume fraction of the second of these is being tracked, the density in
each cell is given by
1

3

All other properties are calculated in the same manner.
The two phases properties are summarized in table 1.
Table 1: properties of the two phases at 40°C
[Kg/m3]
1.225
1041

40°C

Air
Lubricant Oil

40°C [Kg/ms]
1.7894E-05
0.2082

As operating density has been chosen the lowest one.
A single momentum equation is solved throughout the domain, and the resulting
velocity field is shared among the phases

4

One lim
mitation of th
he shared-fiellds approximaation is that in
i cases wherre large
velocitty differencess exist betwe en the phasees, the accuraacy of the veelocities
compuuted near the interface can be adversely
y affected. The energy equuation si
also shhared among th
he phases

5

The VOF
V
model treats
t
energyy, , and tem
mperature, , as mass-avveraged
variablles. As with the
t velocity ffield, the accu
uracy of the temperature
t
nnear the
interfacce is limited in cases wheere large temp
perature differrences exist bbetween
the phaases.
For thhe pressure-velocity-coupliig a SIMPLE
E scheme haas been adoppted as
suggessted for flow
ws in closed domains. Th
his algorithm uses a relattionship
betweeen velocity an
nd pressure coorrections to enforce mass conservationn and to
obtain the pressure field.
f
In order to reprodu
uce the operaating conditio
ons of the speed reducer, a rigid
motionn of the mesh has been appplied. The mo
otion has been
n defined by m
mean of
an UD
DF (User defined Functionn). All the bo
oundaries has been set to no slip
walls: the internal boundaries,
b
coorresponding to
t the planet carrier and pllanetary
gears surfaces,
s
rotatte together w
with the mesh (marked in blue
b
in the fiigure 4)
while the
t other boun
ndaries, correesponding to the
t external crrown and the case of
the geearbox (markeed in red in the figure 4),
4 are station
nary in the aabsolute
referennce frame.

a)
b)
Figure 4: a) 2D scheematization off the computattional domain (section. A-A
A fig. 3)
b) 3D representation of tthe computatiional domain
The tim
me step for thee transient anaalysis has been
n evaluated ass

,

6

where
is the volume of the smallest cell in the computational domain
,
and is the velocity scale of the problem.
4.3 Operating conditions
The purpose of the simulations is to analyze the resistant torque on the planet
carrier shaft. This moment is calculated with a surface integral on the moving
walls with respect to the gearbox axis and it is composed by two parts: the first
given by pressure and the second by the viscous effects. Some simulations have
been computed with a different combinations of oil level, operating temperature
and rotational speed. Table 2 shows the combinations of parameters for each
simulation. is the operating temperature in °C,
the rotational speed of the
planet carrier in rpm and is the oil level in mm measured from the gearbox axis
(positive if higher than the axis, negative if lower).
Table 2: parameters for each simulation:
T [°C] L[mm] [rpm]
40

20

500

40

0

500

40

0

1500

40

20

1000

40

20

1500

40

0

1000

40

-20

500

40

-20

1000

40

-20

1500

20

0

1000

65

0

1000

The annalysis have been
b
stopped after the resisstant torque had
h no more ssensible
fluctuaations and hass stabilized. B
By multiplying
g the mean vaalue of the torrque by
the im
mposed rotatio
onal speed, it is possible to
t determine the churningg power
losses.

b)
a)
Figure 5: a) contou
urs of velocity
ty magnitude for the oil phase
p
b) conttours of
volume fraaction for the oil phase in th
he mid-section
n

5 Exp
perimental tests
In ordder to provide a validatioon of the nu
umerical model, a real inndustrial
planetaary speed redu
ucer has been modified as already
a
described and testedd. After
the moodifications, th
he input shaft and the outpu
ut shaft are co
ompletely unccoupled.
That alllows to movee the planet caarrier and rep
produce the co
ondition in whhich the
meshinng losses and the churning losses due to
o the rotationss of the gears around
their axxis are avoided.

g
Figure 6: schhematic layoutt of the test rig
A scheematic layout of
o the test rig is shown in figure
fi
6. The speed reducer is fixed
to the ground
g
by meean of a speciially designed
d structure. Th
he planet carriier shaft

is connnected to a HBM
H
T12 torqque meter with a telemetricc transmissionn of the
signal by mean of a double carddan shaft in orrder to avoid bending. Thee torque
meter, in turn, is con
nnected to a 355KW DC mottor by means of
o a coupling..

Fig
gure 7: test rigg with the testted speed redu
ucer
A transparent pipe allows
a
the moonitoring of the
t actual oil level. The opperating
temperrature can be imposed
i
by a special heatin
ng system (inssulated chambber with
a heatinng source) and measured byy mean of a th
hermocouple. The rotationaal speed
can be controlled by
y the motor PL
LC (programm
mable logic co
ontroller). For each of
the 3 oil
o levels, 4 different
d
tempperatures havee been tested. For all this ccoupled
operatiing condition
ns (oil level + temperaturre) the poweer losses havve been
measurred in a rotatio
onal speed rannge between 100
1 rpm and 1500 rpm.

6 Ressults
Beforee comparing the experimeental results with
w
the com
mputation onees, it is
necessaary to subtracct the seal lossses and the bearing
b
lossess from the m
measured
values.. The torque meter, in factt, measures th
he input pow
wer of the redu
ducer. A
little frraction of this power is disssipated by the two internal bearings
b
of thee planet
carrier shaft and by its seal. This part of lossess can be easily
y calculated bby mean
of som
me proved mod
dels [2], [3] ass function of th
he testing con
nditions.
Figure 8 to 13 show
w both the exxperimental an
nd the compu
utation results for the
different oil levels an
nd as a functioon of temperaature and rotattional speed.
Due too the high co
omputational ttime needed for the solutiion of the nuumerical
modelss, not all the experimentallly tested con
nditions have been also coomputed
with thhe CFD analyssis.
It can be seen thaat the numeri
rical results are
a in good agreement w
with the
numeriical ones.

Power losses L=+20mm
300

ω=500rpm (exp)

250

ω=1000rpm (exp)

P [W]

200

ω=1500rpm (exp)

150

ω=500rpm (CFD)

100

ω=1000rpm (CFD)

50

ω=1500rpm (CFD)

0
0,0

50,0
100,0
T [°C]
Figure 8: experimental and computation results for L=20mm

Power losses L=0mm

P [W]

250

ω=500rpm (exp)

200

ω=1000rpm (exp)

150

ω=1500rpm (exp)

100

ω=500rpm (CFD)
ω=1000rpm (CFD)

50

ω=1500rpm (CFD)

0
0,0

50,0
100,0
T [°C]
Figure 9: experimental and computation results for L=0mm

Power losses L=‐20mm
150

ω=500rpm (exp)

P [W]

125

ω=1000rpm (exp)

100

ω=1500rpm (exp)

75

ω=500rpm (CFD)

50

ω=1000rpm (CFD)

25

ω=1500rpm (CFD)

0
0,0

50,0
100,0
T [°C]
Figure 10: experimental and computation results for L=-20mm

P [W]

Power losses (L=20mm)
300
250
200
150
100
50
0

T=20°C (exp)
T=45°C (exp)
T=65°C (exp)
T=85°C (exp)
T=40°C (CFD)

0

500

1000

1500

2000

V [rpm]
Figure 11: experimental and computation results for L=20mm

P [W]

Power losses (L=0mm)
300
250
200
150
100
50
0

T=20°C (exp)
T=45°C (exp)
T=65°C (exp)
T=85°C (exp)
T=40°C (CFD)

0

500

1000
1500
2000
V [rpm]
Figure 12: experimental and computation results for L=0mm

P [W]

Power losses (L=‐20mm)
175
150
125
100
75
50
25
0

T=20°C (exp)
T=45°C (exp)
T=65°C (exp)
T=85°C (exp)
T=40°C (CFD)

0

500

1000
1500
2000
V [rpm]
Figure 13: experimental and computation results for L=-20mm

7 Conclusions
As reliable models to predict the churning losses of the planetary gears are still
not available, a CFD model has been applied in order to predict this important
component of losses of epicycloid gear reducers, starting from the geometry and
the operating conditions. The results of the model are well supported by the
experiments (except for extremely low temperatures). The losses increase, as
expected, with static lubricant level and rotational speed and decrease with the
temperature. The decrease rate of the power losses with the temperature is very
high for low temperature and decreases with the temperature growth.
The increase rate of the losses with the rotational speed is more than linear.
Future development are the investigation of other influence parameters like the
oil type and the geometry of the reducer.

References
[1] Concli, F., Gorla, C., Arigoni, R., Cognigni, E., Musolesi, M.,
Planetary Speed Reducers: Efficiency, Backlash, Stiffness, International
conference on gears, Munich 2010
[2] General Catalogue SKF –– SKF Group, December 2006
[3] Niemann, G., Winter, H., Maschinenelemente – Band 2: Getriebe Allgemein,
Zahnradgetriebe – Grundlagen, Stirnradgetriebe – 2. Auflage , Springer,
Berlin 2003
[4] ISO/TR 14179-1 and -2
[5] Patankar, S.V., Numerical heat transfer and fluid flow, Taylor&Francis, USA
1980
[6] Versteeg, H.K., Malalasekera, W., An introduction to computational fluid
dynamics – The finite volume method, Longman Group, London 1995
[7] Comini, G., Fondamenti di termofluidodinamica computazionale,
SGEditoriali, Padova 2004
[8] Concli, F., Gorla, C., Arigoni, R., Musolesi, M., Riduttori di precisione a
gioco ridotto ed alta efficienza, Organi di trasmissione – febbraio 2011,
Tecniche Nuove, Milano 2011
[9] Csobàn, A., Kozma, M., Influence of the Oil Churning, the Bearing and the
Tooth Friction Losses on the Efficiency of Planetary Gears, Journal of
Mechanichal Engineering 56(2010)4, pp.231-238