Simple Analytical Expressions for the Force and Torque

of Axial Magnetic Couplings

Thierry Lubin, Mezani Smail, Abderrezak Rezzoug

To cite this version:

Thierry Lubin, Mezani Smail, Abderrezak Rezzoug. Simple Analytical Expressions for the Force
and Torque of Axial Magnetic Couplings. IEEE Transactions on Energy Conversion, 2012, 11 p.
�10.1109/TEC.2012.2183372�. �hal-00673920�

HAL Id: hal-00673920

https://hal.science/hal-00673920v1
Submitted on 24 Feb 2012

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
 1

Simple Analytical Expressions for the Force and

Torque of Axial Magnetic Couplings
Thierry Lubin, Smail Mezani, and Abderrezak Rezzoug

Abstract—In this paper, a theoretical analysis of an axial

magnetic coupling is presented, leading to new closed-form
expressions for the magnetic axial-force and torque. These
expressions are obtained by using a two-dimensional (2-D)
approximation of the magnetic coupling geometry (mean radius
model). The analytical method is based on the solution of
Laplace’s and Poisson’s equations by the separation of variables
method. The influence of geometrical parameters such as number
of pole pairs and air-gap length is studied. Magnetic field
distribution, axial force and torque computed with the proposed
2-D analytical model are compared with those obtained from 3-D
finite elements simulations and experimental results.

Index Terms— Torque transmission, axial magnetic coupling,

analytical model, axial force.
Fig. 1. Geometry of the studied axial-type magnetic coupling (p = 6)

I. INTRODUCTION
and poorly flexible for the first step of design stage.

M AGNETIC couplings are of great interest in many

industrial applications. They can transmit a torque from
a primary driver to a follower without mechanical contact. As
Analytical methods are, in general, less computational time
consuming than numerical ones and can provide closed-form
solutions giving physical insight for designers. So, they are
the torque could be transmitted across a separation wall, axial useful tools for first evaluations of magnetic couplings
field magnetic couplings are well suited for use in isolated performances and for the first step of design optimization.
systems such as vacuum or high pressure vessels. Moreover, Three-dimensional analytical models for ironless permanent
they present a maximum transmissible torque (pull-out torque) magnet couplings have been proposed in the literature [1-16].
giving an intrinsic overload protection. The proposed models are developed for axial magnetic
Axial magnetic couplings consist of two opposing discs couplings with parallelepiped magnets or cylindrical tile
equipped with rare earth permanent magnets as shown in Fig. magnets. As the magnets are in free space (with no other
1. The magnets are magnetized in the axial direction. They are magnetic materials present), analysis is based either on the
arranged to obtain alternately north and south poles. The flux amperian model with Biot-Savart law or on the coulombian
is closed by soft-iron yokes. The torque applied to one disc is method with equivalent surface charges. Although these
transferred through an air-gap to the other disc. The angular methods give very accurate results, they are not suitable for the
shift between the two discs depends on the transmitted torque study of magnetic couplings with iron-core structures.
value. The main drawback of axial-type magnetic couplings is An alternative analytical method to compute the torque for
the significant value of the axial attractive force between the magnetic couplings with iron yokes is based on boundary
two discs. value problems with Fourier analysis. This method consists in
An accurate knowledge of the magnetic field distribution is solving directly the Maxwell’s equations in the different
necessary for predicting the torque and the axial force. The regions (air-gap, magnets....) by the separation of variables
magnetic field can be evaluated by analytical methods [1-22] method [17], [18]. The magnetic field distribution is obtained
or by numerical techniques like finite elements [23-26]. in each region by using boundary and interface conditions. The
Finite elements simulations give accurate results considering torque and the force are then computed by using the Maxwell
three dimensional (3-D) effects and nonlinearity of magnetic stress tensor. In [19] and [20], two-dimensional (2-D)
materials. However, this method is computer time consuming analytical models for radial-type magnetic couplings were
developed and closed-form expressions for the torque was
Manuscript received October 31, 2011. given and used for design optimization. In [21] and [22], quasi
T. Lubin, S. Mezani and A. Rezzoug are with the Groupe de Recherche en 3-D analytical models are proposed to compute the
Electrotechnique et Electronique de Nancy (GREEN), Université Henri
Poincaré, 54506 Nancy, France (e-mail: [email protected]).
performances of axial-flux permanent magnets machines. A
 2

modulation function is defined to take into account the radial

dependence of the magnetic field.
In this paper, we propose new formulas for the torque and
the axial force of an axial-type magnetic coupling with iron
yokes (fig. 1). The analytical study is based on the solution of
2-D Laplace’s and Poisson’s equations in air-gap and
permanent magnets regions by using the separation of
variables method. The torque expression is used to study the
influence of geometrical parameters (number of pole pairs and
air-gap length). In order to study the accuracy of the proposed
formulas, the results are compared with those obtained from 3-
D finite elements simulations and experimental results. Fig. 2. 2-D model of the axial magnetic coupling at the mean radius of the
magnets Re =(R1+R2)/2.

II. PROBLEM DESCRIPTION AND ASSUMPTIONS with

As shown in Fig. 1, the geometrical parameters of the Br
M = M z ez = ± ez (2)
studied magnetic coupling are the inner and outer radii of the µ0
magnets R1 and R2, the air gap length e, and the magnets where M is the magnetization vector, Br the remanence of the
thickness h. The pole-arc to pole-pitch ratio of the permanent magnets, ez the unit vector along the axial direction and ±
magnets is α. The number of pole-pairs is p. indicates the magnetization direction.
Analytical study of axial magnetic couplings is complicated
because of the three-dimensional nature of the magnetic field III. 2-D ANALYTICAL MODEL
distribution. However, in order to simplify the analysis and to
By using the separation of variables method, we now
carry out closed-form expressions for the axial force and
consider the solution of Poisson’s equations for PMs regions
torque, the 3-D problem is reduced to a 2-D one by
and Laplace’s equation for the air-gap region.
introducing a cylindrical cutting surface at the mean radius of
the magnets Re =(R1+R2)/2 at which the magnetic field will be A. Solution of Poisson’s Equation in the PMs Regions
computed [21], [22]. (Regions I and III)
Fig. 2 shows the resulting 2-D model by considering the Poisson’s equation in the magnets region (region III) can be
unrolled cylindrical cutting surface. With this approach, we written in a cylindrical coordinates system as
neglect the radial component of the magnetic field and we
consider that the axial and tangential components do not
1 ∂ 2 AIII ∂ 2 AIII µ0 ∂M z  h + e ≤ z ≤ 2h + e
depend on the r-coordinate. Moreover, for simplicity, we + =− for  (3)
Re2 ∂θ ∂z Re ∂θ 0 ≤ θ ≤ 2π / p
2 2
adopt the following assumptions:

1) The iron yokes have infinite magnetic permeability, where µ0 is the permeability of the vacuum and Mz is the axial
2) The magnets are axially magnetized with relative recoil magnetization of the magnets.
permeability µ r = 1 . Knowing that the tangential component of magnetic field at
z = 2h + e is null (soft-iron yoke with infinite permeability)
As shown in Fig.2, the whole domain is divided into three and considering the continuity of the axial component of the
regions: the PMs regions (regions I and III) and the air-gap flux density at z = h + e , we obtain the following boundary
region (region II). The magnets of region III are shifted by an conditions
angle δ (torque angle) from the magnets of region I. Due to the ∂AIII
periodicity of the magnetic field distribution, the studied =0 (4)
∂z z = 2 h + e
domain is limited by 0 ≤ θ ≤ 2π/p.
AIII (θ , h + e) = AII (θ , h + e) (5)
A magnetic vector potential formulation is used in 2D
cylindrical coordinates to describe the problem. According to
the adopted assumptions, the magnetic vector potential in each where AII (θ , z ) is the magnetic vector potential in the air-gap
region has only one component along the r-direction and only region.
depends on the θ and z-coordinates. The electromagnetic The distribution of the axial magnetization Mz is plotted in
equations in each region expressed in term of the magnetic Fig.3, δ is the relative angular position between the magnets of
vector potential are region I and region III. The axial magnetization can be
∇ 2 A = − µ0 ∇ × M in Regions I and III (PMs) expressed in Fourier’s series and replaced in (3)
 2 (1)
∇ A = 0 in Region II (air-gap)
 3

One can apply the same procedure for region I by

considering a zero value for δ. This leads to the following
expression for the magnetic vector potential
 kp 
∞ ch  z
 Re 
AI (θ , z ) =
k =1
∑ I
(ak
 kp 
+ K k ) cos ( kpθ )
ch  h 
 Re 
(12)
 kp 
∞ ch  z 
R
+ ckI  e  sin ( kpθ )
∑
k =1
 kp 
ch  h 
 Re 
The integration constants akI and ckI in (12) are determined
Fig. 3. Magnetization distribution along θ-direction (region III).
using a Fourier series expansion of AII (θ , h) over the interval
∞ [0, 2π/p]
M z (θ ) = ∑M k sin ( kp (θ − δ ) ) (6) 2π / p
2p
4 Br  π
k =1


akI + K k =
2π ∫ AII (θ , h) cos(kpθ )dθ (13)
Mk = cos  k (1 − α )  with k = 1,3,5, 7.... (7) 0
kπµ0  2  2π / p
2p
ckI =
2π ∫ AII (θ , h)sin(kpθ )dθ (14)
Taking into account the boundary conditions (4) and (5), the 0
general solution of the magnetic vector potential in Region III The expressions of the coefficients akI and ckI are given in
can be written as the appendix.

AIII (θ , z ) = B. Solution of Laplace’s Equation in the Air-Gap Region

(Region II)
 kp 
∞ ch  ( z − 2h − e )  Laplace’s equation in the air-gap region can be written in a
 Re  + K cos(kpδ )) cos kpθ
∑ (a
k =1
III
k
 kp 
k ( ) (8) cylindrical coordinates system as
ch  h 
 Re  1 ∂ 2 AII h ≤ z ≤ h + e
∂ 2 AII
+
= 0 for  (15)
 kp  ∂θ
Re2 ∂z 0 ≤ θ ≤ 2π / p
2 2

∞ ch  ( z − 2h − e ) 
 Re  + K sin(kpδ )) sin kpθ
∑ (c
The continuity of the tangential component of the magnetic
+ III
k k ( ) field at z = h and at z = h + e leads to the following
k =1
 kp 
ch  h  boundary conditions
 Re 
with
∂AII ∂AI ∂AII ∂AIII
Re = and = (16)
K k = µ0 Mk (9) ∂z ∂z ∂z ∂z
2 z =h z =h z = h+e z =h+ e
kp
The integration constants akIII and ckIII are determined By taking into account the boundary conditions (16), the
using a Fourier series expansion of AII (θ , h + e) over the general solution of the magnetic vector potential in the air-gap
interval [0, 2π/p] can be written as
AII (θ , z ) =
2π / p  kp   kp 
2p ch  ( z − h − e )  ch  ( z − h ) 
akIII + K k cos(kpδ ) =
2π ∫
0
AII (θ , h + e) cos(kpθ )dθ (10)
∑
∞
R
( −akII e
kp
R
 e
 kp 
 + b II e  e
k
R
kp
R
 kp 
 ) cos kpθ
( )
k =1
sh  e  sh  e 
2π / p
2p  Re   Re 
ckIII + K k sin(kpδ ) =
2π ∫
0
AII (θ , h + e)sin(kpθ )dθ (11)
∞
 kp 
ch  ( z − h − e ) 
 kp 
ch  ( z − h ) 
R  + d II Re  Re
+ (−ckII e  e
∑  )sin kpθ
R
k ( )
k =1
kp  kp  kp  kp 
The expressions of the coefficients akIII and ckIII are given sh  e  sh  e 
 Re   Re 
in the appendix. (17)
 4

The integration constants akII , bkII , ckII and d kII are closed-form expression for the electromagnetic torque which
depends directly on the geometrical parameters.
determined using Fourier series expansions of ∂AI ∂z h and
∂AIII ∂z h + e over the air-gap interval [0, 2π/p]  3
16 Br2 3   R1 
 sin2  α π  sh2 a
Te = R2 1 −     sin ( pδ )
3π µ0   R2    2  sh ( 2 (1 + ν ) a )
2π / p  
2p ∂AI
akII =
2π ∫
0
∂z h
cos(kpθ ) dθ (18)
with
(25)

2π / p h e
2p ∂AIII a= p and ν= (26)
bkII =
2π ∫
0
∂z h+e
cos( kpθ )dθ (19) Re 2h

2π / p As expected, the torque presents a sinusoidal characteristic

2p ∂AI
ckII =
2π ∫ ∂z h
sin(kpθ )dθ (20) with the relative angular position δ. Its maximum value (pull-
0 out torque) is obtained at the angle δ=π/2p.
2π / p
2p ∂AIII
d kII =
2π ∫
0
∂z h+e
sin(kpθ ) dθ (21)
B. Axial-Force
Axial magnetic force is an important parameter for the
design of an axial magnetic coupling. This attractive force
The expressions of these coefficients are developed in the
must be known because it affects directly the rotor structure
appendix.
and bearings. Indeed, the bearing lifetime depends on the
The axial and tangential components of the magnetic flux
bearing load. By using the Maxwell stress tensor, the axial
density in the air-gap can be deduced from the magnetic vector
force expression is
potential by
2π
R 2 − R12
∫ (B )
1 ∂AII ∂A
F= 2 IIz (θ , ζ ) − BII2 θ (θ , ζ ) dθ
2
BIIz =− BIIθ = II (22) (27)
Re ∂θ ∂z 4 µ0
0

Substituting (22) into (27), the analytical expression for the

IV. AXIAL-FORCE AND TORQUE EXPRESSIONS axial force becomes

( )
A. Electromagnetic torque
π R22 − R12 ∞
The electromagnetic torque is obtained using the Maxwell
∑ (( Z + X k ) − (Wk + Yk ) )
2 2
F= (28)
stress tensor. A line at z = ζ ∈ [ h, h + e] in the air-gap region 4 µ0
k
k =1
is taken as the integration path so the electromagnetic torque is
expressed as follows Considering only the fundamental component of the
magnetic field in the air-gap (k = 1), we can derive a closed-
2π form expression for the axial force
R23 − R13
Te =
3µ0 ∫B 0
IIθ (θ , ζ ) B IIz (θ , ζ )dθ (23)

 R  2
sh2 ( a )
Incorporating (22) into (23), the analytical expression for F=
8 Br2
R22 1 −  1  sin 2  α π 

 R  
π µ0   2  sh 2 ( 2(1 + ν )a )
  2  (29)
the electromagnetic torque becomes 
× ( cos ( pδ ) ch ( 2(1 + ν )a ) + 1)
(
π R23 − R13 ) ∞
Te =
3µ0 ∑ (W X
k =1
k k + Yk Z k ) (24)
From (25) and (29), we can see that the torque and the axial
force dependence on the design parameters are explicit. For
engineering purpose, it is important to have simple relations to
where the coefficients Wk, Xk, Yk and Zk are given in the
study rapidly the effects of the geometrical parameters on the
appendix.
coupling performances. This is developed in the following
The torque can be computed with a good precision by
subsection.
considering only the fundamental components of the the flux
density distribution in the air-gap (k = 1). This is especially
true for large number of PM pole-pairs and/or large air-gap.
Considering the first harmonic approximation, we can derive a
 5

V. RESULTS OBTAINED WITH 2-D ANALYTICAL MODEL Fig. 5 corresponds to the full load condition (δ = 15°). We
In this section, we use the proposed 2-D analytical model to can observe clearly on Fig. 5a the distortion of the flux lines
compute the magnetic field distribution in the air-gap for due to the angular displacement of the upper magnets. Fig. 5b
different angular position between the two discs. For each and Fig. 5c show respectively the axial and the tangential
position, the torque and the axial force are calculated by components of the flux density in the middle of the air-gap.
respectively using (25) and (29). Then, the influence of some For this angular position, the torque reaches its maximum
geometrical parameters on the coupling performances is value Te = 94 Nm (pull-out torque) as indicated in fig. 7 and
investigated (particularly the air-gap length and the pole-pairs the axial force is still attractive (F = 423N) as shown in fig. 8.
number). The geometrical parameters of the studied device are As it can be observed in fig.8, the axial force reaches a null
given in Table I. These parameters correspond to the one value for an angle slightly higher to half the pole pitch (around
which give a pull-out torque of around 90 Nm (obtained using 17°).
(25)) when we consider an air-gap length of 3 mm and a 6
pole-pairs.
A. Flux density distribution and torque calculation for e =
3mm and p=6
Figs. 4a and 4b show respectively the flux lines (for two
pole pitches) and the axial component of the flux density in the
middle of the air-gap under no-load condition (δ = 0°). The
magnetic flux density distribution along the air-gap is
computed by using (17) and (22). We can observe that the flux
lines are almost axial along the air-gap (the tangential
component of the flux density is null in the middle of the air-
gap). For this position, the torque is then equal to zero and the
axial force is attractive and reaches its maximum value. By
using (29), we obtain F ≈ 2500N as shown in Fig. 8. (a)

(b)

(a)

(c)
(b)
Fig. 4. No load condition (δ = 0°): (a) magnetic flux lines, (b) axial Fig. 5. Full load condition (δ = 15°): (a) magnetic flux lines, (b) axial
component of the flux density in the middle of the air-gap. component of the flux density, (c) tangential component of the flux density.
 6

Fig. 6 corresponds to an angular displacement δ = 30°

(unstable position). In this position, the magnets of the two
discs are in opposed direction and the flux lines repel each
other. The flux density presents only a tangential component in
the middle of the air-gap. The torque is then equal to zero and
the axial force is now repulsive as shown in fig. 8. The axial
force value computed with (29) gives F = -1628 N.
Fig. 7 and Fig. 8 summarize the variation of torque and
axial force as a function of the angular displacement δ. As
shown previously, the maximum torque occurs at an angular
shifting of half pole pitch angle. We can observe that the first
harmonic approximation gives accurate results (the error is
less than 5%) compared to the ones obtained by taking into Fig. 7. Torque versus the angular displacement δ for e = 3mm and p = 6.
account 10 harmonic terms in (24) and (28).

Fig. 8. Axial force versus the angular displacement δ for e = 3 mm and p = 6.

(a) B. Influence of the air-gap length

The length of the air-gap has a significant influence on the
characteristics of the axial magnetic coupling. Fig. 9 and Fig.
10 show respectively the pull-out torque and the maximal axial
force as a function of the air-gap length. The results have been
computed by using (25) and (29). The geometrical parameters
are the ones given in table I and we have considered a number
of pole pairs p = 6. As shown in fig. 9, the pull-out torque of
the magnetic coupling decreases quickly as the distance
between the magnets increases. The maximum torque is almost
divided by two when the air-gap is increased from 2mm to
7mm. In the same way, the maximal axial force is reduced
when the air-gap length increases (fig. 10).

(b) C. Influence of the number of pole pairs

The variation of pull-out torque and maximal axial force
Fig. 6. Magnets in opposed direction (δ = 30°): (a) magnetic flux lines, (b)
tangential component of the flux density in the middle of the air-gap. versus the number of pole pairs are respectively shown in fig.
11 and fig. 12. The results have been computed by using (25)
TABLE I and (29). For the study, we have considered several air-gap
PARAMETERS OF THE STUDIED AXIAL COUPLING lengths. The other geometrical parameters are those given in
Symbol Quantity value Table I. Fig. 11 shows that all the curves present a maximum
which depends on the air-gap length. The optimum value of
R1 Inner radius of the magnets 30 mm
R2 Outer radius of the magnets 60 mm
the number of pole pairs is shifted to the right when the air-gap
h Magnets thickness 7 mm is reduced. This result is well known for magnetic couplings.
e Air-gap length variable For the studied coupling (Table I), the optimal number of pole
α PMs pole-arc to pole-pitch ratio 0.9 pairs is p = 6 if we consider an air-gap length e = 5mm. We
p Pole-pairs number variable
Br Remanence of the permanent magnets 1.25 T can observe in Fig. 12 that the maximum axial force decreases
when the number of pole pairs increases.
 7

We have shown here that the torque formula (25) can

predict the effects of the geometrical parameters on the
coupling performances and from fig. 11, we can choose
rapidly the optimum value of the number of pole pairs when
the other geometrical parameters are given.
In the next subsection, we investigate the precision of the 2-
D approximation (25), by comparing the previous analytical
results with 3-D FEM simulations and experimental results.

Fig. 12. Maximum axial force versus the number of pole pairs for several air-
gap values.

VI. 3-D FEM SIMULATIONS AND EXPERIMENTAL RESULTS

In order to show the limits of the formulas (25) and (29), the
analytical results have been compared to 3-D FEM simulations
in one hand and to experimental results in another hand. For
the 3-D finite element simulations, we have used COMSOL
multiphysics software.
Fig. 9. Pull-out torque versus air-gap length for p = 6.
For the experimental validation, we have manufactured an
axial magnetic coupling prototype using sector type NdFeB
magnets glued on iron yokes. The geometrical parameters of
the prototype are those of Table I. We choose a number of
pole pairs p = 6 that corresponds practically to the optimum
value for an air-gap value e = 5mm.
Fig. 13 shows the axial magnetic coupling placed on the test
bench. The axial coupling is inserted between two electrical
machines. In fig. 13, the air-gap value is e = 9.5mm.
A. 3-D FEM results
Fig. 14 and Fig. 15 show respectively the pull-out torque
and the maximal axial force as a function of the air-gap length
obtained with 3-D finite elements analysis and with the 2-D
analytical model. The number of harmonic terms used in (24)
and (28) is N = 10. The geometrical parameters are those
Fig. 10. Maximum axial-force versus air-gap length for p = 6.
given in Table I. For this study, the pole pair number is fixed
to p = 6. As expected for this type of device, the 2-D analytical
prediction gives higher values for both pull-out torque and
maximal axial force as compared to 3-D FE analysis.
This is mainly due to the 3-D effects which are not taken
into account in the proposed model (the radial dependence of
the magnetic field is not considered). The error on the pull-out
torque prediction ranges from 22% for e=2mm to 37% for
e=12mm. The error on the maximal axial force prediction is
less important and ranges from 8% for e=2mm to 34% for
e=12mm.
Figs. 16 show the pull-out torque versus the number of pole
pairs computed with 3 methods ((24), (25) and3-D FEM). The
results are given for three values of the air-gap lengths. As it
Fig. 11. Pull-out torque versus the number of pole pairs for several air-gap can be observed in Figs. 16, although the analytical formula
values. (25) predicts higher torque values, the number of pole pairs
which corresponds to the maximal value of the torque is
almost the same for the 3 methods in use.
 8

(b)
Fig. 13. Axial magnetic coupling prototype placed on the test bench (e =
9.5mm).

(c)
Fig. 16. Pull-out torque versus the number of pole pairs obtained with 3-D
Fig. 14. Pull-out torque versus the air-gap length for p = 6: 3-D FEM and 2-D FEM and 2-D analytical model: (a) e = 2mm; (b) e = 6mm; (c) e = 10mm.
analytical results.
Fig. 17 presents the synthesis of Fig. 16 and gives the
optimal value of the pole-pair number versus air-gap lengths.
This is an important result since we can observe in fig. 17 that
the analytical formula (25) is suitable in the determination of
the optimum value of the pole-pair number with the air-gap
value when the other geometrical parameters are fixed.
B. Experimental results
Fig. 18 compares the measured values of the axial flux
density and the ones obtained with the proposed 2D analytical
model for no load condition (δ=0). The measurements are
made along the θ coordinate at the mean radius of the magnets
Re = 45 mm. For this test, the air-gap is fixed at e = 9.5mm. A
Fig. 15. Maximum axial-force versus the air-gap length for p = 6, 3-D FEM Hall probe placed on a XY table is used to measure the
and 2-D analytical results. magnetic field distribution. As the magnetic flux density is
measured at the mean radius Re, we can observe very good
agreement between experimental results and the ones obtained
with the 2-D analytical model.
To show the limits of the 2-D analytical model, we have
measured the radial dependence of the axial flux density at a
center line of a pole for no-load condition (δ=0). The air-gap is
fixed at e = 9.5mm. The results are shown in fig. 19. As it can
be observed, the axial flux density shows large variations
along the radial expanse of the magnet. This is due to the large
value of the air-gap. This result can not be predicted by the 2-
D analytical model which neglects the radial dependence of
the magnetic field. We can note a good agreement between 3-
(a)
D FEM simulations and experimental results.
 9

(a)
Fig. 17. Optimal value of the pole-pair number versus air-gap dimension
computed with 3-D FEM and 2-D analytical model (25).

(b)
Fig. 20. Measured and computed static torque versus the angular
Fig. 18. Measured and computed (2D analytical model) axial flux density in displacement δ for p = 6: (a) e = 4mm; (b) e = 9.5mm
the middle of the air-gap at the mean radius Re =(R1+R2)/2 for e = 9.5 mm.
VII. CONCLUSION
In this paper, we have proposed new simple analytical
expressions for computing axial force and torque of an axial
magnetic coupling. These expressions are determined by the
solution of 2-D Laplace’s and Poisson’s equations (mean
radius model) in the different regions (air-gap and magnets).
Although the proposed 2D analytical model shows some
lack of accuracy compared to 3D finite-element simulations
and experimental results (error of around 30% on the pull-out
torque prediction), we have shown that it can be used to
determine rapidly the optimal value of the pole-pair number
when the other geometrical parameters are given.
Fig. 19. Measured and computed axial flux density along a radial lines in the
Moreover, the proposed analytical formulas can be useful
middle of the air-gap at a center line of a pole for e = 9.5 mm. tools for the first step of design optimization since continuous
derivatives issued from the analytical expressions are of great
Figs. 20 show the comparison between the measured values importance in most optimization methods.
of the static torque and the calculated ones by using the 2-D
analytical model (25) and 3-D FEM. The static torque was APPENDIX
measured by suspending weights from a wire attached to a rod • The development of (10) and (11) gives
(a rotor is locked and the other one can rotate). The relative  kp 
angular position δ was measured using an encoder with a ch  e 
R
+ bkII e  e 
R 1 R
resolution of 4096 steps per revolution (0.088 degree). Two akIII + K k cos(kpδ ) = − akII e (A.1)
kp  kp  kp  kp 
values of the air-gap dimension were considered (e = 4mm and sh  e  sh  e 
e = 9.5mm). It can be noticed that the experimental  Re   Re 
measurements are in good agreement with the results obtained  kp 
ch  e 
with 3-D FE simulations. As shown previously, analytical
ckIII + K k sin(kpδ ) = −ckII
Re 1
+ d kII
Re  Re  (A.2)
formula (25) gives higher values of around 30% for the pull- kp  kp  kp  kp 
sh  e  sh  e 
out torque.  Re   Re 
 10

The development of (13) and (14) gives [8] E. P. Furlani, “Analysis and optimization of synchronous couplings,” J.
Appl. Phys., vol. 79, pp. 4692–4694, 1996.
 kp  [9] E. P. Furlani and M. A. Knewtson, “A three-dimensional field solution
ch  e 
R
akI + K k = − akII e  e  + bkII e
R R 1 for permanent-magnet axial-field motors,” IEEE Trans. Magn., vol. 33,
(A.3) no. 3, pp. 2322–2325, May. 1997.
kp  kp  kp  kp 
sh  e  sh  e  [10] P. Elies and G. Lemarquand, “Analytical optimization of the torque of a
 Re   Re  permanent-magnet coaxial synchronous coupling,” IEEE Trans.Magn.,
vol. 34, no. 4, pp. 2267–2273, Jul. 1998.
 kp  [11] J. F. Charpentier and G. Lemarquand, “Optimal design of cylindrical
ch  e 
R
ckI = −ckII e  Re  + d II Re 1
(A.4)
air-gap synchronous permanent magnet couplings,” IEEE Trans.Magn.,
kp  kp 
k
kp  kp  vol. 35, no. 2, pp. 1037–1046, Mar. 1999.
sh  e  sh  e  [12] J. F. Charpentier, N. Fadli, and J. Jennane, “Study of ironless permanent
 Re   Re  magnet devices being both a coupling and an axial bearing for naval
propulsion,” IEEE Trans. Magn., vol. 39, no. 5, pp. 3235–3237, Sep.
• Expressions of the coefficients akII , bkII , ckII and d kII for the 2003.
air-gap region [13] H. L. Rakotoarison, J. P. Yonnet, and B. Delinchant., “Using
Coulombian approach for modelling scalar potential and magnetic field
The development of (18) to (21) gives of a permanent magnet with radial polarization,” IEEE Trans. Magn.,
kp  kp  kp  kp  vol. 43, no. 4, pp. 1261–1264, Apr. 2007.
akII = akI th  h  bkII = −akIII th  h  [14] R. Ravaud and G. Lemarquand, “Comparison of the Coulombian and
Re  Re  Re  Re 
Amperian current models for calculating the magnetic field produced by
kp  kp  kp  kp  arc-shaped permanent magnets radially magnetized,” Prog.
ckII = ckI th  h  d kII = −ckIII th  h  (A.5) Electromagn. Res., vol. 95, pp. 309–327, 2009.
Re  Re  Re  Re  [15] R. Ravaud, G. Lemarquand, V. Lemarquand, and C. Depollier,
• Expression of the coefficients Wk, Xk, Yk and Zk given in (24) “Permanent magnet couplings: Field and torque three-dimensional
expressions based on the Coulombian model,” IEEE Trans. Magn., vol.
and (28) 45, no. 4, pp. 1950–1958, Apr. 2009.
 kp   kp  [16] R. Ravaud, V. Lemarquand, and G. Lemarquand, “Analytical design of
sh  (ζ − h − e )  sh  (ζ − h )  permanent magnet radial couplings,” IEEE Trans. Magn., vol. 46, no.
R R
Wk = − akII  e  + b II  e
k
 11, pp. 3860–3865, Nov. 2010.
 kp   kp  [17] B. L. J. Gysen, K. J. Meessen, J. J. H. Paulides, and E. A. Lomonova,
sh  e  sh  e 
“General formulation of the electromagnetic field distribution in
 Re   Re 
machines and devices using Fourier analysis,” IEEE Trans. Magn., vol.
 kp   kp 
ch  (ζ − h − e )  ch  (ζ − h ) 
46, no. 1, pp. 39-52, Jan. 2010.
R [18] T. Lubin, S. Mezani, and A. Rezzoug, “Exact analytical method for
X k = ckII  e  − d II Re  Re  magnetic field computation in the air-gap of cylindrical electrical
k
 kp  kp  kp  machines considering slotting effects,” IEEE Trans. Magn., vol. 46, no.
sh  e  sh  e 
 Re   Re  4, pp. 1092-1099, Apr. 2010.
(A.6) [19] R. M. Hornreich and S. Shtrikman, “Optimal design of synchronous
 kp   kp 
sh  (ζ − h − e )  sh  (ζ − h )  torque couplers,” IEEE Trans. Magn., vol. 14, no. 5, pp. 800–802, Sept.
R R 1978.
Yk = −ckII  e  + d II e  e 
R
k [20] J. Fontchastagner, Y Lefèvre, and F. Messine, “Some co-axial magnetic
 kp  kp  kp  couplings designed using an analytical model and an exact global
sh  e  sh  e 
 Re   Re  optimization code,” IEEE Trans. Magn., vol. 45, no. 3, pp. 1458–1461,
Mar. 2009.
 kp   kp 
ch  (ζ − h − e )  ch  (ζ − h )  [21] A. Parviainen, M. Niemelä, and J. Pyrhönen, “Modeling of axial flux
R R permanent-magnet machines,” IEEE Trans. Ind. Appl., vol. 40, no. 5,
Z k = −akII  e  + b II 
k
e 
pp. 1333–1340, Sep./Oct. 2004.
 kp   kp  [22] J. Azzouzi, G. Barakat, and B. Dakyo, “Quasi-3-D analytical modeling
sh  e  sh  e 
 Re   Re  of the magnetic field of an axial flux permanent-magnet synchronous
machine,” IEEE Trans. Energy. Convers., vol. 20, no. 4, pp. 746-752,
Dec. 2005.
REFERENCES [23] C. Ferreira and J. Vaidya, “Torque analysis of permanent magnet
[1] J. P. Yonnet, “Permanent magnet bearings and couplings,” IEEE Trans. coupling using 2D and 3D finite elements methods,” IEEE Trans.
Magn., vol. 17, no. 1, pp. 1169–1173, 1981. Magn., vol. 25, pp. 3080–3082, Jul. 1989.
[2] E. P. Furlani, “Formulas for the force and torque of axial couplings,” [24] W. Wu, H. C. Lovatt, and J. C.Dunlop, “Analysis and design
IEEE Trans. Magn., vol. 29, no. 5, pp. 2295–2301, Sep. 1993. optimisation of magnetic couplings using 3D finite element modelling,”
[3] J. P. Yonnet, S. Hemmerlin, E. Rulliere, and G. Lemarquand, IEEE Trans. Magn., vol. 33, no. 5, pp. 4083–4085, Sept. 1997.
“Analytical calculation of permanent magnet couplings,” IEEE Trans. [25] R. Wang, E. P. Furlani, and Z. J. Cendes, “Design and analysis of a
Magn., vol. 29, no. 6, pp. 2932–2934, Nov. 1993. permanent magnet axial coupling using 3D finite element field
[4] E. Furlani, S. Reznik, and A. Kroll, “A three-dimensional field solution computations,” IEEE Trans. Magn., vol. 30, no. 4, pp. 2292–2295, Jul.
for radially polarized cylinders,” IEEE Trans. Magn., vol. 31, no. 1, pp. 1994.
844–851, Jan. 1995. [26] T. F. Chan, W. Wang, and L. L. Lai, “Performance of an Axial-Flux
[5] E. P. Furlani, R. Wang, and H. Kusnadi, “A three-dimensional model Permanent Magnet Synchronous Generator From3-D Finite-Element
for computing the torque of radial couplings,” IEEE Trans. Magn., vol. Analysis,” IEEE Trans. Energy. Convers., vol. 25, no. 3, pp. 669-676,
31, no. 5, pp. 2522–2526, Sep. 1995. Sep. 2010.
[6] Y. D. Yao, G. J. Chiou, D. R. Huang, and S. J. Wang, “Theoretical
computations for the torque of magnetic coupling,” IEEE Trans. Magn.,
vol. 31, no. 3, pp. 1881–1884, May. 1995.
[7] R. Waring, j. Hall, K. Pullen, and M. R. Etemad, “An investigation of
face type magnetic couplers,” Proc. Inst. Mech. Engrs, Part. A., vol.
210, no. 4, pp. 263–272, 1996.
 11

Thierry Lubin was born in Sedan, France, in 1970.

He received the M.S. degree from the University
Pierre et Marie Curie, Paris 6, France in 1994 and the
Ph.D. degree from the University Henri Poincaré,
Nancy, France, in 2003.
He is currently a lecturer of Electrical Engineering
at the University of Nancy at the Groupe de
Recherche en Electrotechnique et Electronique de
Nancy. His interests include modeling and control of
electrical machines, and applied superconductivity in
electrical devices.

Smail Mezani was born in Algiers, Algeria, in

1974. He received the engineer diploma and the
magister degree from the University of Sciences
and Technology Houari Boumediene, Algiers,
Algeria in 1996 and 1999 respectively. He obtained
the Ph.D. degree from the Institut National
Polytechnique de Lorraine, France, in 2004.
He is currently a lecturer at the University Henri
Poincaré of Nancy, France, at the Groupe de
Recherche en Electrotechnique et Electronique de Nancy where his research
interests include the applications of superconductors in electromechanical
devices.

Abderrezak Rezzoug received the electrical

engineer degree from ENSEM INPL, Nancy,
France in 1972, and the Dr. Ing. diploma and the
Ph.D. degree from INPL, in 1979 and 1987
respectively.
After working at the INPL as an assistant Professor
until 1991, he is currently a Professor of Electrical
Engineering at the University Henri Poincaré,
Nancy, France. As a member of the Groupe de
Recherche en Electrotechnique et Electronique de Nancy, his main subjects of
research concern superconducting applications to electrical devices, and the
control and diagnosis of electrical machines.