Efficiency of the Double Roller Full Toroidal Variator

Leonardo De Novellis1 , Giuseppe Carbone1, Luigi Mangialardi1

1
Department of Mechanical Engineering, Politecnico di Bari, Italy
E-mail: l.denovellis@poliba.it, carbone@poliba.it, lmm@poliba.it

Keywords: CVT, toroidal variator, traction performance, spin losses.

SUMMARY. The paper deals with the theoretical evaluation of the performance of a new toroidal
traction drive for CVT transmissions, the so-called Double roller Full Toroidal Variator, which con-
sists of a couple of counter-rotating rollers arranged in a typical full-toroidal cavity.
The innovative features of this new geometry may improve the efficiency of the toroidal variator
over a wide range of speed ratios in comparison to the existing toroidal traction drives.

1 INTRODUCTION
Recent developments in the automotive field are related to the design of drive-trains with the aim
of improving the exploitation of the thermal engine, according to the requirements of reduction of
fuel consumption and polluting emissions [1, 2]. To achieve these targets, chain/belt Continuously
Variable Transmissions (CVTs) have been widely studied [3, 4, 5, 6, 7] either theoretically either ex-
perimentally; however, limitations of the maximum transmittable torque and of control possibilities
made the toroidal traction variators a valid alternative for the development of CVT drivetrains.
A toroidal traction drive is constituted of input and output disks, which are coupled, respectively,
with drive and driven shafts and shaped in such a way to realize a toroidal cavity. A power roller,
rotating inside the toroidal cavity, is employed to transfer torque from the drive disk to the driven
one, by means of shearing action of the oil film under elasto-hydrodynamic conditions. The tilting
of the power roller allows shifting manouvers.

Figure 1: The toroid geometries: Single roller Half Toroidal Variator (a), Single roller Full Toroidal
Variator (b), Double roller Full Toroidal Variator (c).

Referring to the toroidal variators which are actually on the market for automotive applications,
the basic geometric distinction is on the position of the roller tilting center (point O in Fig. 1), which
can coincide or not with the center of the toroidal cavity [8]. In the first case we have the so-called
full-toroidal variator [see Fig. 1(b)] whereas in the second case we have the half-toroidal variator
[see Fig. 1(a)].

1
 In Fig. 1(c) a new patented toroidal geometry [9], the so-called Double roller Full Toroidal Vari-
ator (or DFTV), is shown: two counter-rotating rollers are arranged inside the full-toroidal cavity,
with the aim of reducing the spin losses at the roller-disk contact; also, the roller conical shape
allows to balance the normal forces, thus making unnecessary the employment of a thrust roller-
bearing. In this way, the main advantages of the single-roller toroidal geometries can be combined
with significant improvements of the overall transmission efficiency.

2 GEOMETRY AND KINEMATICS

In this section we define the main geometric and kinematic features of the Double roller Full
Toroidal Variator.
Looking at Fig. 2, r0 is the first principal radius of the input and output discs whereas r11
and r33 are the second principal radii of the input and output discs respectively. Each power roller
has a conical part and a toroidal part: the rollers are in contact along the conical part, whilst the
toroidal part is shaped as a typical half-toroidal roller and it is in contact with the input or output
disk. Therefore, we call r2 the curvature radius of a section perpendicular to the roller axis and
r22 is the second principal radius. For practical aspects, it is useful to define the conformity ratio
CR = r22 /r0 and the aspect ratio k = e/r0 , where e is the eccentricity.

Figure 2: The Double roller Full Toroidal Variator (DFTV) geometric quantities.

An important control parameter is the tilting angle γ, therefore we can express the dimensionless
discs curvature radii formulation as a function of the cone angle θ = π/2 and of the tilting angle γ
(see Fig. 2) as
r1
r11 = (1)
cos(θ + γ)
r3
r33 = (2)
cos(θ − γ)

2
where r1 = r0 (1 + k − cos(θ + γ)) and r3 = r0 (1 + k − cos(θ − γ)). From geometric relations
we also obtain r2 = (r0 + e) sin(α/2).
We define the geometric speed ratio srID as the ratio of the radial coordinates r1 and r3 , namely
r1
srID = (3)
r3

Figure 3: The Double roller Full Toroidal Variator (DFTV) kinematic quantities.

Said ω1 the input angular velocity and ω3 the rotational speed of the output disk, we define the
speed ratio sr as
|ω 3 |
sr = (4)
|ω 1 |
Percentual differences of the velocities of the sliding contact pairs are taken into account by defining
input and output creep coefficients, respectively Crin and Crout , as

|ω1 | r1 − |ω2 | r2
Crin = (5)
|ω1 | r1

|ω ′2 | r2 − |ω 3 | r3
Crout = (6)
|ω ′2 | r2

3
and the slip coefficient sc = |ω ′2 | / |ω 2 |where ω 2 and ω′2 are the angular velocities of the two
counter-rotating rollers. The transmission ratio can be expressed as a function of the overall creep
coefficient Cr :
sr = (1 − Cr )sc srID (7)
with (1 − Cr ) = (1 − Crin )(1 − Crout ).
Fig. 3 shows the velocities vector triangles: for each roller i of the roller pair we can define points
Θi , which represents the point of intersection of the tangents to the toroidal cavity at the roller-disc
contact points, and Ωi , defined as the intersection point of the roller and disk rotation absolute
axes. The angular velocities of the roller relative to the input and output disks, ω 21 = ω 2 − ω1 and
ω23 = ω ′2 −ω3 have spin velocity components, ω 21spin and ω 23spin , which can be directed inwards
or outwards. By choosing a proper value of the cone incidence angle α, it is possible to have zero
spin velocities at srID = 1 and the spin coefficients can be expressed as:

ω21spin 1 + k − cos(θ + γ)
σ21 = = sin(θ + γ) − (1 − Crin ) (8)
ω1 1+k
ω23spin 1 1 + k − cos(θ − γ)
σ23 = = sin(θ − γ) − (9)
ω3 (1 − Crout ) 1+k

3 THE THEORETICAL MODEL

3.1 Traction coefficients and efficiency
Following the same approach proposed in [10], let us define the traction coefficient µ = FT /FN
as the ratio between the tangential force at the roller-disk interface FT and the normal load FN (see
Fig. 4) and the spin momentum coefficient χ = MS / (FN r), where MS is the spin momentum
and r the disk radial coordinate of the contact point. From the equilibrium equations applied to the
counter-rotating rollers we get:

(µin − µc )(1 + k) sin(α/2) + r̃1 χin sin θDT − tBa = 0 (10)

(µc − µout )(1 + k) sin(α/2) + r̃3 χout sin θDT − tBb = 0 (11)

where we have considered the angle θDT = (π − α)/2 (see Fig. 2) and the radial coordinates
in dimensionless form, namely r̃1 = r1 /r0 and r̃3 = r3 /r0 . The terms tBa and tBb represent
the dimensionless torque losses due to roller bearings on the variator roller axes, which have been
evaluated according to the SKF technical documentation [11]. From the equilibrium equations of
the disks we can evaluate the effective torque coefficients at the input and ouptut variator side:

tin = µin + χin sin(θ + γ) (12)

tout = µout − χout sin(θ − γ) (13)

with tin = Tin / (mnFN r1 ), tout = Tout / (mnFN r3 ) can be interpreted as effective torque co-
efficients at the input and ouptut variator side. In the previous relations we have considered the
possibility to arrange n set of rollers inside m toroidal cavities. Moreover, we can express the thrust
force acting along the disk axial direction FD as a function of the normal load FN and of the tilting
angle γ:

FDin = nFN sin(θ + γ) (14)

FDout = nFN sin(θ − γ) (15)

4
 Figure 4: The forces acting on the DFTV elements.

With these positions, a straightforward expression of the overall mechanical efficiency ν =

Pout /Pin can be obtained:
µout − χout sin(θ − γ)
ν = sc (1 − Cr ) (16)
µin + χin sin(θ + γ)

3.2 Contact model

A fully flooded isothermal contact model [10] has been utilized to calculate the shear stresses at
the interface of the elements in contact, and thus, the traction and spin coefficients. We distinguish
the contact conditions at the roller-disk interface, where the contact area is elliptical and at the roller-
roller interface along the conical surfaces, where instead the contact area is rectangular.
To study the contact between two curved profiles a and b, we use to define an equivalent Young’s
modulus E ′ , expressed as a function of the elasticity modulus E and Poisson’s ratio ν of each
element in contact [12]:
−1
1 − νa2 1 − νb2

′
E =2 + (17)
Ea Eb
and an equivalent curvature radius ρeq = (1/ρeqx + 1/ρeqy )−1 , with
1 1 1
= + (18)
ρeqx rax rbx
1 1 1
= − (19)
ρeqy ray rby
in the case of the disk-roller contact and rax = r2 and rbx = r11 , ray = r22 and rby = r0 , whilst, in
the case of the rectangular patch, ray = rby = r2 / cos(α/2) and ray = rby = ∞.
The results of elasto-hydrodynamic theory are employed to describe the lubrication regime on
the contact area. In particular, we have considered the viscosity-pressure effect with the Roeland’s

5
formula:
 1−(1+p/cp )Z1
η η∞
= (20)
η0 η0
Because of the very high shear rates, the lubricant shows a non linear relation between the shear
stresses τij and the shear strain rates ε̇ij , according to the common rule used in the plasticity theory
to split the shear strain along the different directions, we write
∂vi ∂vj τij
+ = Γ (τe ) (21)
∂xj ∂xi τe

where the function Γ (τe ) is represententative of the non-linear behavior of the traction oil and
√
2
τe = (τij τij )1/2 (22)
2
is the equivalent stress. By following the rheological model proposed by Bair and Winer, the function
Γ (τe ) takes the form  
τL 1
Γ (τe ) = ln (23a)
η 1 − τe /τL
where τL = τL0 + ap is the limiting shear stress. The estimation of the shear strain is carried
out by considering, as already done before, that under EHL conditions most part of the contact is
characterized by an almost constant thickness of the lubricant film. Therefore, we can write

∂vx ∂vz vx
+ ≈ (24)
∂z ∂x h
∂vy ∂vz vy
+ ≈
∂z ∂y h
which allows us to determine, once known the pressure distribution at the interface, the shear stresses
in the lubricant, and in turn, the traction, spin, slip and efficiency performance of the variator.

Absolute viscosity at the atmospheric pressure η0 = 3.25 × 10−3 Pa s

Viscosity-pressure index Z1 = 0.85
Pressure-viscosity coefficient ζ = 1.71 × 10−8 Pa−1
Limiting shear stress at atmospheric pressure τL0 = 0.02 × 109 Pa
Limiting shear stress constant a = 0.085
Pole pressure constant of Roelands viscosity model cp = 1.96 × 108 Pa
Pole viscosity of Roelands viscosity model η∞ = 6.31 × 10−5 Pa s

Table 1: Fluid properties at T = 99◦ C.

The pressure distribution is evaluated recalling the Hertz’ contact theory

p
p̃ = p̃max 1 − X 2 − Y 2 (25)
p
with p̃ = pΛ2 /FN and the characteristic length [12] Λ = 3 6FN r0 / (πE ′ ).

6
 The shear stresses at the input and output contact area are defined in dimensionless form as:
 
6 v21i −
η|v21 |
τ̃21i = ℜτ̃L 1 − e hτL (26)
π |v21 |
 
6 v23i −
η|v23 |
τ̃23i = ℜτ̃L 1 − e hτL (27)
π |v23 |

with ℜ = r0 /Λ whereas i is the x or y component, depending on the considered direction and

τ21i = τ21i Λ2 /FN and τ23i = τ23i Λ2 /FN . Also, the relative velocities are defined as a function of
the creep and spin coefficients, namely
v21x σ21 ãy v21y σ21 ãx
= Crin − Y, = X (28)
|ω1 | r1 ℜr̃1 |ω1 | r1 ℜr̃1
v23x Crout σ23 ãy v23y σ23 ãx
=− − Y, = X (29)
|ω3 | r3 1 − Crout ℜr̃3 |ω3 | r3 ℜr̃3

2 2

1/2 2 2

1/2
and simply |v21 | = |ω1 | r1 v21x + v21y and |v23 | = |ω3 | r3 v23x + v23y ; furthermore, the
terms
η |v21 | η̃ |v21 |
= ω̃1 (1 + k) (30)
hτL Hin τ̃L ω1 r1
η |v23 | η̃ |v23 |
= ω̃1 (1 + k)(1 − Cr )sc srID (31)
hτL Hout τ̃L ω3 r3
with Hin and Hout being the oil film thicknesses (refer to [10] for their analytical expression).
By integrating the shear stresses over the contact area we can evaluate the traction coefficients at
the input and output sides
Z 1 Z 2π
µin = ãXin ãYin dR τ̃21X Rdψ (32)
0 0
Z 1 Z 2π
µout = −ãXout ãYout dR τ̃23X Rdψ (33)
0 0

and the spin momentum coefficients as

1 2π
ãXin ãYin
Z Z
χin = dR (ãXin τ̃21Y cos ψ − ãYin τ̃21X sin ψ)R2 dψ (34)
ℜr̃1 0 0
1 2π
ãXout ãYout
Z Z
χout = dR (ãXout τ̃23Y cos ψ − ãYout τ̃23X sin ψ)R2 dψ (35)
ℜr̃3 0 0

where 
X = R cos ψ
, 0 ≤ R ≤ 1, 0 ≤ ψ ≤ 2π (36)
Y = R sin ψ
According to the Hertz’ theory, the contact pressure distribution at the roller-roller interface can
be expressed as:
2w′
q
pc = a2h − x2 (37)
πa2h

7
where w′ is the load ′
p per unit length acting on the cylinders, namely w = Fn /hc while the semicon-
tact length ah = 8w′ Rx /(πE ′ ). By integrating the shear stress along the contact length we obtain
the expression of the traction coefficient:

2hc E ′ ah
Z
µc = τ̃c dx (38)
FN 0

− η̃ ω̃c
 
where τ̃c = τ̃L 1 − e Hc τ̃L and ω̃c = ωc η0 /E ′ .

4 RESULTS
The performance of the DFTV variator have been evaluated considering the following geometric
quantities: r0 = 40mm, r2 = 18.7mm, CR = 0.8, α = 43deg, k = 0.286, m = 2, n = 3. Fluid
properties are listed in Tab. 1.
The effective output torque coefficient tout is depicted in Fig. 5 as a function of the overall
creep coefficient Cr : calculations have been carried out by keeping constant the input disk rotational
velocity at the value ω1 = 2000RPM and the normal load FN = 20kN at different gear ratios srID .

0.1

0.08
srID = 1 srID = 2

0.06
tout

srID = 0.5

0.04

0.02

0
0 0.02 0.04 0.06 0.08 0.1
Cr

Figure 5: The effective output torque tout as a function of the overall creep coefficient Cr for the
three geometric ratios (srID = 0.5 - black curve, srID = 1 - blue curve, srID = 2 - red curve).
Calculations are done at normal load FN = 20kN.

At the underdrive (srID = 0.5) and overdrive (srID = 2) ratios, the torque coefficient linearly
increases with the creep until a certain creep threshold value is reached (see the black and red curves
in Fig. 5), which is about Crs = 0.03. Then the traction curve suddenly reaches a saturation value,
which represents the maximum effective output torque deliverable by the variator. We point out
that the maximum transmittable torque toutmax = 0.09 does not change with the speed ratios, since
all the traction curves overlap at large creep values. Also, at the extreme ratios (srID = 0.5, 2) the
DFTV toroidal geometry shows a symmetric behaviour, since the traction curves are almost perfectly
superposed either in the linear either in the saturation regions. Different trend is shown at the unit

8
speed ratio (blue curve in Fig. 5) where the traction curve presents a very large slope at small creep
values, since in this working condition the spin and creep velocities are comparable in magnitude
and the elastic deformation of the variator elements have been neglected [10], the contribution to the
overall creep coefficient is notably small.

srID = 1
0.9 srID = 2
srID = 0.5
Ν

0.8

0.7
0 0.02 0.04 0.06 0.08 0.1
tin

Figure 6: The overall efficiency ν as a function of the effective input torque tin for the three geo-
metric ratios (srID = 0.5 - black curve, srID = 1 - blue curve, srID = 2 - red curve). Calculations
are done at normal load FN = 20kN.

We have evaluated the variator efficiency ν as a function of the effective input torque tin at
different geometric speed ratios (see Fig. 6).
The highest efficiency is obtained at the unit ratio srID = 1 (see the blue curve in Fig. 6) since,
as already explained, the spin losses are almost completely negligible in this working condition. The
maximum efficiency is achieved at a value of the input torque coefficient tin = 0.09, where the
threshold between linear and saturation regions of the traction curve is located.
The influence of the spin losses at the roller-disk interface is clear if one looks at the efficiency
values at srID = 0.5 (black curve in Fig. 6) and srID = 2 (red curve in Fig. 6) where the maximum
value of the efficiency is about 4% lower than the maximum efficiency at the unit speed ratio.

5 CONCLUSIONS
We have presented a theoretical analysis of a recently patented geometry of a toroidal traction
drive, the Double roller Full Toroidal Variator, which has been designed in such a way to combine
the advantages of existing toroidal geometries in order to enhance the efficiency of the variator. In
particular, the DFTV variator consists of a set of counter-rotating conical rollers which are placed
into a toroidal cavity: the geometric characteristics of the power rollers allow to reduce the spin
losses in a wide range of speed ratios and to get rid of the thrust ball-bearing, which contributes to
the torque losses of the half-toroidal type geometry.
In order to evaluate the performance of the toroidal traction drive, we have developed a fully-
flooded isothermal contact model, based on teh results of EHL lubrication theory. Our calculations

9
have shown the effectiveness of the Double roller Full Toroidal Variator in terms of reduction of spin
losses and improvement of the overall efficiency.

References
[1] Mantriota G., “Fuel consumption of a vehicle with power split CVT system,” Int. J. Veh. Des.,
37, 327-342 (2005).
[2] Brace C., Deacon M., Vaughan N. D., Horrocks R. W. and Burrows C. R., “The Compromise
in Reducing Exhaust Emissions and Fuel Consumption from a Diesel CVT Powertrain over
Typical Usage Cycles,” in Proc. Int. Congress on Continuously Variable Power Transmission
CVT’99, Eindhoven, The Netherlands, September 16-17, 27-33 (1999).

[3] Srnik J. and Pfeiffer F., “Dynamics of CVT Chain Drives,” Int. J. Veh. Des., 22, 54-72 (1999).
[4] Carbone G. Mangialardi L. and Mantriota G., “The influence of pulley deformations on the
shifting mechanisms of MVB-CVT,” J. Mech. Des., 127, 103-113 (2005).
[5] Sorge F., “Influence of pulley bending on metal V-belt mechanics,” in Proc. Int. Conference on
Continuously Variable Power Transmission, Japanese Society of Automotive Engineers, Paper
No. 102 (9636268), Yokohama, Japan, September 11-12, 9-15 (1996).

[6] De Novellis L. and Carbone G., “Experimental investigation of chain link forces in continu-
ously variable transmissions,” J. Mech. Des., doi: 10.1115/1.4002764, (2010).
[7] Kanehara S., Fujii T. and Kitagawa T., “A study of a Metal Pushing V-Belt type CVT - Part 3:
What Forces Act on Metal Blocks?,” SAE paper 940735, (1994).
[8] Imanishi T., Machida H., “Development of Powertoros Unit Half-Toroidal CVT (2) - Compar-
ison between Half-Toroidal and Full-Toroidal CVTs,” Motion & Control, No. 10, (2001).
[9] Durack M. J., “Full Toroidal Traction Drive,” World Intellectual Property Organization, Inter-
national Publication Number: WO 2011/041851 A1.
[10] Carbone G., Mangialardi L. and Mantriota G., “A comparison of the performances of full and
half toroidal traction drives,” Mech. Mach. Theory, 39, 921-942, (2004).
[11] SKF, SKF Technical Documentation.
[12] Hamrock B.J., Fundamentals of fluid film lubrication, Series in Mechanical Engineering, Book,
McGraw-Hill, 1994 (ISBN 0-07-025956-9).

10