Bulletin of the JSME Vol.11, No.

6, 2017
Journal of Advanced Mechanical Design, Systems, and Manufacturing

Modelling of steady-state mechanical power losses in

planetary gear trains of automatic transmissions
Venkatakrishna JANAKIRAMAN*, Ahmet KAHRAMAN** and David TALBOT**
* The Department of Mechanical and Aerospace Engineering, The Ohio State University
210 W. 19th Avenue, Columbus, OH 43210, USA
E-mail: janakiraman.7@osu.edu
** The Department of Mechanical and Aerospace Engineering, The Ohio State University
210 W. 19th Avenue, Columbus, OH 43210, USA

Received: 18 September 2017; Revised: 23 October 2017; Accepted: 12 November 2017

Abstract
In this study, steady-state mechanical power loss model of an automatic transmission gear train consisting of
multiple stages of planetary gear sets is developed. Load dependent (mechanical) power losses at the internal
(ring-planet) and external (sun-planet) gear meshes and planet bearing interfaces are included through
elastohydrodynamic lubrication (EHL) based power loss formulations. At the end, efficiency of an 8-speed
automatic transmission is studied under representative operating conditions to quantify the contribution of
various loss mechanisms to the total loss under variable speed, load and temperature conditions. Impact of gear
surface roughness amplitudes on the resultant power losses is also described.

Keywords : Power loss, Efficiency, Planetary gear trains, Multi-stage

1. Introduction

In a planetary gear set, friction induced mechanical power losses occur at all its rolling-sliding contacts. Such
contacts consist of external and internal gear meshes (mostly helical) and planet bearings that are mostly cylindrical roller
(caged or full-complement) bearings. Computation of these power loss components require detailed load distributions
along the contact zones, and hence, the load carried by each contact interface, requiring a power flow formulation that
incorporates the power losses as well. Here, such a power flow formulation will be combined with EHL based power loss
models of gear meshes and planet bearings to predict mechanical losses of planetary gear trains of automatic
transmissions
Mechanical power loss models for gears and gear trains can be divided into two distinct categories, the first addresses
the power losses in simple gear systems like a pair of gears in mesh or a single planetary gear stage. These models may
utilize an empirical friction coefficient as reviewed in Martin (1978) or utilize physics-based models to calculate power
losses (Xu et al. (2007), Li and Kahraman (2010), Li and Kahraman (2011)). The second category of studies utilize a
kinematics argument to obtain the power loss or efficiency in complex gear trains by utilizing an assumed gear mesh
efficiency value for the individual meshes (Pennestri and Freudenstein (1993), del Castillo (2002), Chen and Angeles
(2011)). Also, these models do not include losses of planet bearings while they have been shown to be a significant source
of power loss (Talbot et al. (2012)), nor have a mechanism to consider the load independent losses in the system.
This study is aimed at developing a generalized power flow formulation to predict gear mesh and bearing loads with
power losses included. EHL models of external and internal gear meshes and planet bearings will be combined with this
power flow formulation to determine the mechanical power loss of a multi-stage planetary gear train.
While this paper is focused on multi-stage planetary gear trains consisting of multiple single-planet planetary stages,
double-planet planetary stages can be easily included as well. The principles and formulations introduced in this study
are easily extended to include double-planet planetary stages.

Paper No.17-00476
[DOI: 10.1299/jamdsm.2017jamdsm0080] © 2017 The Japan Society of Mechanical Engineers
1
 Janakiraman, Kahraman and Talbot,
Journal of Advanced Mechanical Design, Systems, and Manufacturing, Vol.11, No.6 (2017)

2. Generalized Power Flow Model

A power flow model that captures the mutual dependence of power flow and power losses is developed with the
following fundamental assumptions.
(i) Power loss does not cause any change in kinematics or direction of power flow in the system, and
(ii) components of power loss can be converted to equivalent loss torques to be applied to one or more elements.

This power flow model uses three distinct steps for the computation of power flow: (i) a complete solution of the
kinematics, (ii) calculation of power flow in the absence of power loss, and (iii) an iterative procedure to recalculate the
power flow with the addition of power losses by treating the gear train as a multi-body system and establishing torsional
equilibrium conditions. The immediate parameters required for this power flow model are tooth counts, number of planet
branches and the kinematic configuration of the system. Detailed component level design information is also needed for
the power loss models being employed here.

2.1 Global Planetary Kinematics Formulation

Let Zim and N m be the tooth count of gear i ( i = s, p, r ) and number of planet branches in stage m ∈ [1, z ] ,
respectively. The gear mesh speed equations for stage m can be written as
m
 ωs 
{}
m
Zs Zp −( Z s + Z p ) 0  ω p 
= 0 .
−( Z r − Z p ) Z r   ωc 
0 (1)
 −Z p 0
 
 ωr 

where ω is the absolute angular velocity, the subscripts s, p, c and r denote sun, planet, carrier and ring, and superscript
m denotes the planetary gear stage. Equation (1) can be written in compact form as

Λ mωm = 0 (2)

and extended to a z-stage planetary gear train as

 Λ1  0   ω1  0 
     
       =   . (3)
 z  z 0 
 0  Λ  2 z×4 z ω 4 z×1  2 z×1

The additional 2z equations required for the solution of kinematics come from the configuration information of the
connections between the components, inputs, outputs and fixed members. The configuration of a planetary gear train is
determined by sets of connections (C), inputs (I), outputs (O), fixed members (F), and unconnected members (UC). Each
consists of elements im, where i represents the component (s, p, c, r) and m denotes the planetary stage.
• For fixed member im in the set F, one sets ωim = 0 .
• For any pair of connected elements im and jn (member i of stage m and member j of stage n) in a set of C, one
writes ωim − ωnj = 0 indicating that these two connected members have the same angular velocity.
• In regards to sets of I and O, only input or output member speed is typically specified, not both. ωim = ω I / O
for a member im that is assigned an input or output duty.
Appending equations defined by sets I, O, F and C to Eq. (2), the global kinematics matrix is given as
Λ G ωG = ω
ˆG (4)

which is solved for the unknown vector of component speeds ωG , given vector of known (input) speeds ω
ˆG.

2.2 Global Planetary Torque Balance Formulation

Consider each stage of the planetary gear train as a multi-body system as shown in Fig. 1. Torque equilibrium
equations for stage m are written in matrix form as

[DOI: 10.1299/jamdsm.2017jamdsm0080] © 2017 The Japan Society of Mechanical Engineers

2
 Janakiraman, Kahraman and Talbot,
Journal of Advanced Mechanical Design, Systems, and Manufacturing, Vol.11, No.6 (2017)

m
 NZ s  m
1 0 0 0
Zp
0   Ts  m
   Tp   −Tˆs 
0 1 0 0 1 −1  T   ˆ 
 NZ r   c  −T p 
0 1 0 −
NZ s
−  T  = ˆ  (5)
0  T r   −Tc 
 Zp Zp 
 p/s   −Tr 
ˆ
 NZ r 
0 0 0 1 0 T p / r 
 Z p 

where Tim and Tˆim represents the external torque and the loss torque on component i of stage m, Timj is the torque
exerted on component i by component j of stage m, and N m is the number of planet branches in stage m.

Fig. 1: Multi-body equilibrium torques in a planet branch of a single-planet planetary gear set.

Writing Eq. (5) in matrix form as

Θ m Tm = T
ˆm, (6)

the global torque balance matrix of a z-stage gear train becomes

Θ1  0   T1  Tˆ1 
     
        =   . (7)
 z  z ˆz
 0  Θ  4 z×6 z T 6 z×1 T 4 z×1

An additional 2z equations required for the solution of torques are provided by sets of C, I, O and UC:
• For each unconnected element im of set UC, one writes:
Tim = 0 . (8a)
• For each connection set in C, an equation can be written. For example, a set with three connected elements im-
jn-kp yields:

[DOI: 10.1299/jamdsm.2017jamdsm0080] © 2017 The Japan Society of Mechanical Engineers

23
 Janakiraman, Kahraman and Talbot,
Journal of Advanced Mechanical Design, Systems, and Manufacturing, Vol.11, No.6 (2017)

Tim + T jn + Tkp =
0. (8b)
• More than one member may be connected to the input or output. That is, a connected pair or a chain of
connected pairs can be connected to the input or output, which must be captured by its torque balance equation.
For instance, for a connected pair im-jn that is connected to the input or output whose torque is known:
Tim + T jn =
TI O . (8c)
In this study, the output torque TO will be user-defined. Combining all the above equations for connections, inputs,
outputs and unconnected components with Eq. (7), one obtains the matrix equation

ΘG TG = Tˆ G (9)

where TG is the vector of external torques on the components and mesh torques and T ˆ defines the known (input or
G
output) torques and the loss torques on the components.
In this formulation, power loss components computed must be converted to equivalent loss torques to be applied on
one or more elements such that they can be included in Eq. (9) to solve for the power flow including the power losses.
Bearing mechanical power losses loss Pˆpm,b of a planet bearing is converted to a loss torque applied to the its planet as

Pˆpm,b
Tˆpm,b = − (10)
ωmp c

where ωmp c is the relative angular velocity of a planet of planetary stage m with respect to its carrier. Any gear mesh
power loss is converted to a loss torque to be applied on the driving gear (driving when viewed in the carrier frame of
that stage). If the driving gear is fixed in the ground/reference frame, the loss torque is then applied to the driven member
of the mesh. If the component on which the loss torque is to be applied is a central member ( i = s, r ), then

Pˆim
, gm
, gm = −
Tˆim (11a)
ωim

where Tˆim ˆm
, gm and Pi , gm are the mesh loss torque and the gear mesh power loss to be applied on central gear i of
planetary stage m and ωim is the absolute angular velocity of the component which the loss torque is applied to. If the
component on which the loss torque is to be applied is a planet, then

Pˆpm, gm
Tˆpm, gm = − . (11b)
ωmp c

2.3 Iterative Power Flow Computation Methodology

The power flow in the gear train impacts the power losses that occur in the gear train through the gear mesh and
bearing forces. Power losses in turn affect the power flow in the gear train. For example, at a given output power, when
power losses are included the input power has to increase thus changing power flow. In order to capture this mutual
dependence, an iterative computational procedure is utilized. The following steps are applied for this:
(i) Calculate gear train kinematics and torque balance, first assuming there is no power loss in the drive train, i.e.
Eq. (9) is solved with Tˆ containing only no loss torque such that ( i = s, p, c, r )
G
z
=
PI P=
O
1
2 ∑ ∑ Tim ωim . (12)
m =1 i

(ii) Calculate power loss components using analytical models with component speeds and no-loss torques from
step (i). Convert these power loss components to loss torques using the procedure of previous section.
(iii) Apply the calculated loss torques to Eq. (9) and recalculate torques TG applied to the components.
(iv) With these new component torques TG , recalculate the gear mesh and bearing power loss components and
the loss torques to solve for an updated TG . Compare this TG to that of previous iteration to check for
convergence. If converged (i.e. TG ( n ) − TG ( n − 1) ≤ εT where εT is a user-defined torque error

[DOI: 10.1299/jamdsm.2017jamdsm0080] © 2017 The Japan Society of Mechanical Engineers

24
 Janakiraman, Kahraman and Talbot,
Journal of Advanced Mechanical Design, Systems, and Manufacturing, Vol.11, No.6 (2017)

threshold), stop the algorithm; else go back to step (iii) for another iteration.
With the converged steady-state power flow solution in hand, the steady-state mechanical power loss PL of the drive
train at a speed and torque state are defined as
z
PL = ∑ ∑ Tim ωim . (13)
m =1 i

This iterative procedure converges naturally due to the monotonic nature of power loss for a given output power. At
a given output power, when power losses are included, the input power is increased. This increased input power in turn
causes a rise in power losses and so on. In other words, with every successive iteration in the algorithm the power loss is
increased albeit in smaller steps.

2.4 Mechanical Power Loss Models

2.4.1 Gear Mesh Mechanical Power Loss Model
Gear mesh power losses are predicted by using a mixed EHL formulation, which provides sliding and rolling
components along the tooth surface interfaces. Li and Kahraman (2010) used considered spur and helical gear pairs as
thin slices of spur gears and employed a line contact, rough surface, mixed-EHL model that follows the contact from the
start of active profile to the tip of the driving gear tooth. The changes in the normal load were captured via a gear load
distribution model while the changes in rolling and sliding velocities were calculated by using the involute geometry. As
this method is rather computationally demanding, an alternate method of developing closed-form friction formulae up
front by regression of a larger number of EHL analyses covering proper ranges of contact parameter for gears (Xu et al.
(2007), Li and Kahraman (2011)). The same approach was used by Talbot and Kahraman (2014) which was shown to
agree with their earlier experiments (Talbot et al. (2012)) well. The same approach will be adapted here.
For a gear mesh i-p ( i = s, r ) of stage m, the gear load distribution model discretizes all contact lines spanning
multiple tooth pairs in K number of contact segments and predicts load distribution at L number of mesh positions
covering a complete mesh cycle. For each contact segment κ ( κ ∈ [1, K ] ) at each rotational position λ , load intensity
Qκλ (force per unit length), contact segment length wκλ , and the sliding velocity ûκλ are used to determine the
average sliding mechanical power loss of this gear mesh as

1 L K
 sliding L ∑ ∑ κλ κλ κλ κλ
 Pˆim = µ Q w uˆ . (14)
 , gm
λ= 1 κ= 1

Here μ κλ is the sliding friction coefficient defined for a typical automatic transmission fluid in the manner of Li and
Kahraman (2011), separating the asperity and viscous friction as
µ = µ a + µv ;

a3ϑ+ a4 ln( Seq ) a5 + a6 ln( Req ) + a7 ln( Seq ) a9 ln( Req )

µ a = 3ϑa1P ( SR )a2 P U P Sco a8 ( SR ) Seq
exp  a10 + a11ϑU + ( SR )(a12 Sco + a13 Seq ) + a14 P  ; (15a)

µ p  a ln( ϑ) + a2G + a3 ln( Req ) a4 P a5 ln(U ) a P + a ln( Seq )

sinh  v h  = ( SR) 1 G P Req 6 7 Sco a8U
 τ 0 
exp  a9 + ( SR)(a10G + a11P ) + a12UP + a13 PSco 

Where a1 − a14 and a1 − a13 are regression constants, and

ur η0 u η uˆ ℜeq p
ϑ= , U = r 0 , SR = , G = αEeq , Req = , P= h ,
Eeq Sco Eeq  ref ur ℜref Eeq
(15b)
Sco Seq Rq1Rq 2
Sco = , Seq = , Sco = Rq12 + Rq 22 , Seq = .
 ref  ref Rq1 + Rq 2

Here, ur is the rolling velocity, ℜeq is the equivalent radius of curvature, Eeq is the equivalent Young’s modulus,

[DOI: 10.1299/jamdsm.2017jamdsm0080] © 2017 The Japan Society of Mechanical Engineers

25
 Janakiraman, Kahraman and Talbot,
Journal of Advanced Mechanical Design, Systems, and Manufacturing, Vol.11, No.6 (2017)

α is the pressure-viscosity coefficient, η0 is the ambient dynamic viscosity, τ0 is the Eyring stress, Rq1 and Rq 2
are the RMS surface roughnesses of the contacting surfaces ,  ref = 1 μm , and ℜref =5 mm .
Likewise, the rolling power loss of the same gear mesh is given by

1 L K
 rolling L ∑ ∑ κλ κλ
 Pˆim = Ψ w . (16)
 , gm
λ= 1 κ= 1

Here Ψ κλ is the rolling power loss density of contact segment κ at mesh position λ , defined for the same fluid as
Li and Kahraman (2011)
b11 ln(U ) +b12 Req b3U +b4 ln(G ) +b5 ln( Req ) +b6 ln( Sco )
=Ψ exp b1 + b2 ( SR)U  P Scob13 P ϑ
(17)
b8 ( SR ) +b9G +b10 ln( Req )
( SR)b7ϑU

Here, b1 − b10 are regression constants, and all other parameters used here are defined as in Eq. (15).
The total mechanical power loss of the same gear mesh is then

=Pˆim  ˆm 
, gm  Pi , gm  +  Pˆim  . (18)
sliding 
, gm 
rolling

Note that when the gear mesh power loss for say a sun-planet mesh is to be applied to the sun, then it includes the
power losses of all sun-planet meshes in that stage. If it is to be applied to the planet, then only the power loss in one sun-
planet mesh is used in Eq. (11b).

2.4.2 Bearing Power Loss Model

Bearing mechanical power losses are calculated using the model of Talbot et al. (2013). First planet bearing forces
and moments from the power flow calculation are used to predict the load distribution along the rolling elements of the
bearing. Similar to the gear mesh mechanical loss model, a line contact EHL model can be used to calculate the power
loss in each discretized segment κ ( κ ∈ [1, K ] ) covering all rolling element segments in contact as

∑ {[ Ψ κ wκ ]inner race + [ Ψ κ wκ ]outer race }

K
=
Pˆpm,b (19)
κ=1

where Ψ κ is the rolling power loss density at the inner and outer races obtained from a regression of EHL simulations
as,
b ξ+b ln ( ξ ) +b6 ln (U ) +b7 Sco
= (
Ψ exp b1 + G b2 ξ + b3 Sco  ξ 4 5 ) (20a)
G
b8 ln (U ) +b9 P
P
( )
b10 ln Req +b11 ln ( Sco )

where
ur η0
ξ= . (20b)
ph Sco

Here, b1 − b11 are regression constants, and all other parameters are defined as in Eq. (15).

3 An Example Analysis
An example planetary drive train from a rear-wheel-drive automatic transmission (Scherer et al. (2009)), shown in
Fig. 2, is analyzed in this section for its mechanical power losses. As detailed gear set design information was not
available, representative designs of gear meshes and planet bearings were carried out to define them as listed in Tables 1
and 2. Bearing contact surfaces were assumed to have roughnesses of Rq = 0.1 μm, while the baseline gear surface
roughnesses were taken to be Rq = 0.5 μm, representative of hard ground surfaces. The lubricant (automatic
transmission fluid) inlet temperature was 90°C. The 6th gear range (G6) is not included in this analysis since it is a direct
drive.

[DOI: 10.1299/jamdsm.2017jamdsm0080] © 2017 The Japan Society of Mechanical Engineers

26
 Janakiraman, Kahraman and Talbot,
Journal of Advanced Mechanical Design, Systems, and Manufacturing, Vol.11, No.6 (2017)

G GR A B C D E
1 4.70 X X X
2 3.13 X X X
3 2.10 X X X
4 1.67 X X X
5 1.28 X X X
6 1.00 X X X
7 0.84 X X X
8 0.67 X X X
REV -3.30 X X X
Fig 2: Configuration and clutching sequence of the 8-speed automatic transmission example (Scherer et al. (2009)).

Table 1: Gear parameters of example gear train. All units are in mm unless specified.
Stage 1 Stage 2 Stage 3 Stage 4
Sun Planet Ring Sun Planet Ring Sun Planet Ring Sun Planet Ring
Number of Teeth 48 25 96 48 25 96 69 22 111 23 31 85
Planet Branches 4 3 3 4
Normal Module 1.40 1.40 1.27 1.69
Normal Pressure Angle (o) 22.5 22.5 20 22.5
Normal Helix Angle (o) 20 20 22.5 22.5
Outside Diameter 75.18 40.64 144.27 75.18 40.64 144.27 94.23 32.00 149.35 45.72 61.47 153.67
Root Diameter 68.07 33.53 151.13 68.07 33.53 151.13 88.90 26.67 154.94 37.08 52.58 162.56
Operating CD 54.86 54.86 60.69 49.78
Face Width 31.75 25.40 20.32 38.10
Transverse Backlash 0.30 0.41 0.30 0.41 0.41 0.56 0.41 0.71

Table 2: Planet bearing parameters. All units are in mm

St. 1 St. 2 St. 3 St. 4
# of Rolling Elements 13 13 13 15
Inner Race Diameter 17.87 17.87 10.78 24.25
Outer Race Diameter 25.50 25.50 15.37 34.42
Roller Diameter 3.81 3.81 2.28 5.08
Roller Length 20.32 20.32 19.05 35.56

[DOI: 10.1299/jamdsm.2017jamdsm0080] © 2017 The Japan Society of Mechanical Engineers

27
 Janakiraman, Kahraman and Talbot,
Journal of Advanced Mechanical Design, Systems, and Manufacturing, Vol.11, No.6 (2017)

Fig 3: Power loss maps of all forward gears (except 6).

[DOI: 10.1299/jamdsm.2017jamdsm0080] © 2017 The Japan Society of Mechanical Engineers

28
 Janakiraman, Kahraman and Talbot,
Journal of Advanced Mechanical Design, Systems, and Manufacturing, Vol.11, No.6 (2017)

All gear meshes and bearings of the gear train are simulated within each gear range to obtain baseline power loss
maps within the input speed and input torque within the ranges of ω I ∈ [ 750, 4500] rpm and TI ∈ [ 75, 450] Nm,
respectively. Figure 3 shows baseline mechanical power loss maps for all forward gears (except G6 direct drive). In line
with measurements Talbot et al. (2012) and predictions of Talbot and Kahraman (2013) for a single stage planetary gear
set, the mechanical power loss maps at each gear range exhibit a nearly linear relationship with the torque and speed
where the deviations from linearity is originated from the effect of load (normal force) and speed (rolling and sliding
velocities) on the lubricated gear and bearing contacts as depicted in Eq. (15, 17, 20). Power loss in G2 is the largest in
comparison to the other gear ranges while G7 yielding the minimum power loss values. In each gear range, the maximum
power loss is observed at the maximum speed and toque (maximum power input) condition of (TI , ω I ) = ( 450, 4500 ) .
At this condition, PL = 2.71 kW for G2 and PL = 0.72 kW for G7 with the rest of the gear ranges having power losses
in between.
As a vehicle operating under normal driving conditions (city or highway) might not expose the transmission gear
train to all (TI , ω I ) values spanned by Fig. 3, a more reasonable input power condition of (TI , ω I ) = (150, 2250 )
which represents an input power of 35.3 kW, will be considered in the following figures.
500
(a) Tot.
PL [W]

St.1
St.2

250 St.3
St.4

0
500
(b) Gear Mesh
PL [W]

Bearing

250

0
250
(c)
External Mesh
PL [W]

Internal Mesh

0
1 2 3 4 5 7 8

Fig 4: Power loss (a) total with stage breakdown, (b) bearing vs gear mesh and (c) external vs internal mesh.

[DOI: 10.1299/jamdsm.2017jamdsm0080] © 2017 The Japan Society of Mechanical Engineers

29
 Janakiraman, Kahraman and Talbot,
Journal of Advanced Mechanical Design, Systems, and Manufacturing, Vol.11, No.6 (2017)

Figure 4(a) shows the comparison of total mechanical power loss in all gear ratios (G1 to G8) and the contribution
from each planetary gear stage of the planetary gear train at the operating condition of (TI , ω I ) = (150, 2250 ) . Total
mechanical power loss is largest in gear 2 and smallest in gear 7 with 465 W and 129 W, respectively. It is noted here
that only one stage is loaded in G1 and G8, there are two loaded stages in G2, G3, G4 and G7 while all four stages are
loaded in G5. Although, only one stage is loaded in G1 and G8, G8 has lower power losses as the mesh entraining velocity
is greater in the case of G8 compared to G1 due to the kinematic configuration. This leads to a thicker lubricant film
thereby causing reduced asperity interactions and hence reduced power loss. G2 has the highest power loss and G7 the
least even though only 2 stages are the sources of power loss in both. This is because the magnitude of power flow
through the meshes is higher in G2. Also note that mesh entraining velocities are the least in stage 4 (particularly external
mesh) in general, this leads to increased power loss due to increased asperity interactions which is the consequence of a
thin fluid film. In G5, all stages contribute to power loss but the mesh power is very small in stage 3 and stage 4 and the
entraining velocity is quite high in stage 1 and stage 2, this combination reduces the resulting total loss. In Fig. 4(b),
contributions from gear meshes and bearings are compared. Bearing power losses are between 1/5 and 1/3 of the total
power loss, indicating that the models neglecting bearing losses would be inaccurate. In Fig. 4(c), external gear meshes
are seen to contribute more to power loss than internal gear meshes in lower gear ranges while they are more balanced
in higher gear ratios.

500
(a) 90C
PL [W]

40C

250

0
500
(b) 90C
PL [W]

40C

250

0
250
(c) 90C
PL [W]

40C

0
1 2 3 4 5 7 8
Fig 5: Effect of temperature on (a) total loss, (b) gear mesh loss and (c) bearing loss.

[DOI: 10.1299/jamdsm.2017jamdsm0080] © 2017 The Japan Society of Mechanical Engineers

10
2
 Janakiraman, Kahraman and Talbot,
Journal of Advanced Mechanical Design, Systems, and Manufacturing, Vol.11, No.6 (2017)

500
(a) Rq=0.50 μm

PL [W]
Rq=0.25 μm
Rq=0.10 μm

250

0
500
(b) Baseline
PL [W]

SF Suns
SF Planets

250

0
1 2 3 4 5 7 8
Fig 6: (a) Gear surface roughness on mesh power loss and (b) Power loss in baseline vs superfinished suns vs
superfinished planets.

Figure 5 shows the effect of lubricant temperature on the mechanical power losses of the gear train. Two oil
temperatures considered here 40°C and 90°C correspond to 30.6 and 8.0 cSt oil viscosities. In Fig. 5(a), the total losses
at these two temperatures are almost the same. This has to do with the way oil viscosity impacts gear and bearing losses.
In Fig. 5(b), gear mesh loss is reduced by reducing the oil temperature. Increased viscosity increases the EHL film
thickness, in the process reducing the asperity contact of rough gear surfaces and hence the sliding friction coefficient
governed by Eq. (15a). While some of these improvements are offset by the increase in rolling losses according to Eq.
(17), there is still a net reduction in gear power losses at 40°C. Meanwhile, bearing losses increase consistently at lower
temperatures since the rolling losses defined by Eq. (20a) are increased.
Figure 6(a) presents a comparison of mesh power losses at three different gear surface roughness levels of
Rq = 0.50 , 0.25 and 0.10 μm. The roughest surface represents ground surfaces while Rq = 0.25 μm is intended for
honed or grind-polished surfaces. The smoothed gear surfaces are representative of chemically polished (superfinished)
surfaces. Power losses are reduced to about one-third by superfinishing all gear tooth surfaces. If superfinishing is applied
selectively to the sun gears, the combined roughness of the external meshes are reduced to lower the mesh power loss by
about 15% as shown in Fig. 6(b) while superfinishing all planets impacts the composite roughnesses of both external and
internal meshes, in the process reducing the power losses by about 30%.

4. Conclusions
This study proposed a physics-based methodology to predict the mechanical power losses of multi-stage planetary
gear trains of automatic transmissions. While the method is capable of capturing both load dependent and load
independent losses, this study was focused on the mechanical losses only. The methodology is capable of predicting the
power losses in bearing and gear meshes using an EHL-based formulation such that effects associated with operating
conditions, lubricant parameters and surface finish conditions can be included. The two-way relationship between
power losses and the resultant power flow conditions was captured through an iterative process. Parametric studies on an
example 8-speed transmission were presented to show the sensitivity of the model to parameters dictating lubricated
contacts of gears and bearings of a planetary gear train.

[DOI: 10.1299/jamdsm.2017jamdsm0080] © 2017 The Japan Society of Mechanical Engineers

11
2
 Janakiraman, Kahraman and Talbot,
Journal of Advanced Mechanical Design, Systems, and Manufacturing, Vol.11, No.6 (2017)

References
Chen, C., and Angeles, J., “Virtual Power Flow and Mechanical Gear-Mesh Power Losses of Epicyclic Gear Trains”, Journal of
Mechanical Design, Vol.129, No.1 (2011), pp.107-113.
del Castillo, J. M., “The analytical expression of the efficiency of planetary gear trains”, Mechanism and Machine Theory, Vol.37,
No.2 (2002), pp.197-214.
Li, S., and Kahraman, A., “A Method to Derive Friction and Rolling Power Loss Formulae for Mixed Elastohydrodynamic
Lubrication”, Journal of Advanced Mechanical Design, Systems, and Manufacturing, Vol.5, No.4 (2011), pp.252-263.
Li, S., and Kahraman, A., “Prediction of Spur Gear Mechanical Power Losses Using a Transient Elastohydrodynamic Lubrication
Model”, Tribology Transactions, Vol.53, No.4 (2010), pp.554-563.
Martin, K. F., “A Review of Friction Predictions in Gear Teeth”, Wear, Vol.49, No.2 (1978), 201-238.
Pennestri, E., and Freudenstein, F., “The Mechanical Efficiency of Epicyclic Gear Trains”, Journal of Mechanical Design, Vol.115,
No.3 (1993), pp.645-651.
Scherer, H., Bek, M. and Kilian, S., “ZF new 8-speed automatic transmission 8HP70-basic design and hybridization”, SAE
International Journal of Engines, Vol.2, No.1 (2009), pp.314-326
Talbot, D., Kahraman, A., “A Methodology to Predict Power Losses of Planetary Gear Sets,” Proceedings of International Gear
Conference (2014), pp.625-636.
Talbot, D., Kahraman, A., and Singh, A., “An Experimental Investigation of the Efficiency of Planetary Gear Sets”, Journal of
Mechanical Design, Vol.134, No.2 (2012), 021003.
Talbot, D., Li, S., and Kahraman, A., “Prediction of Mechanical Power Loss of Planet Gear Roller Bearings under Combined Radial
and Moment Loading”, Journal of Mechanical Design, Vol.135, No.12 (2013), 121007.
Xu, H., Kahraman, A., Anderson, N.E., and Maddock, D., “Prediction of Mechanical Efficiency of Parallel-axis Gear Pairs”, Journal
of Mechanical Design, Vol.129, No.1 (2007), pp.58-68.

[DOI: 10.1299/jamdsm.2017jamdsm0080] © 2017 The Japan Society of Mechanical Engineers

12
2