This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2657493, IEEE
Transactions on Industry Applications

Analysis of a Dual-Rotor, Toroidal-Winding,

Axial-Flux Vernier Permanent Magnet Machine
Tianjie Zou, Student Member, IEEE, Dawei Li, Member, IEEE, Ronghai Qu, Senior Member, IEEE,
Jian Li, Member, IEEE, and Dong Jiang, Senior Member, IEEE
State Key Laboratory of Advanced Electromagnetic Engineering and Technology
School of Electrical & Electronic Engineering
Huazhong University of Science and Technology, Wuhan, 430074, China
E-mail:daweili@ hust.edu.cn

Abstract—In this paper, a vernier permanent magnet (VPM)
machine with dual-rotor, toroidal-winding, axial-flux topology is
analyzed. Through the combination of toroidal windings with the
rotor-stator-rotor topology, the end winding length of the proposed
machine is significantly reduced compared with that of regular
VPM machine. Based on the airgap permeance function, the back-
EMF and torque expressions are derived, and the nature of the
machine is revealed. The influence of pole ratio (ratio of rotor pole
pair number to stator pole pair number) and main geometric
parameters such as slot opening, magnet thickness etc., on
electromagnetic performance is then analytically investigated in
detail. Both the quasi-3-dimensional (quasi-3D) finite element
analysis (FEA) and 3D FEA are applied to verify the derived
equations. Finally, a prototype of the proposed machine has been
manufactured and experimentally evaluated. The test results agree
well with both theoretical and FEA analyses. With the current
density of 5.8 A/mm2 and slot fill factor of 0.35, the torque density
of the proposed machine prototype can reach 31.9 kNm/m3.
Index Terms—Dual-rotor, toroidal-winding, axial flux, vernier
machine, torque equation, quasi-3D FEA.
I. INTRODUCTION
W ith the merits of high torque density and compact
mechanical structure, vernier permanent magnet (VPM)
machines are attracting more and more research interests [1-4].
Due to the so-called magnetic gearing effect [1], the rotor of a
VPM machine rotates at a much lower speed than that of the
stator magnetic field, which makes this kind of machine a
promising alternative for low-speed, high-torque, direct-drive
applications such as wind power and ship propulsion.
Since firstly proposed in 1995 [5], VPM machines have been
extensively studied in the past few years. Many novel VPM
machine topologies [6-9] were proposed, such as dual-
excitation [6], outer rotor [7], dual stator [8], consequent pole
[9], etc. Besides, some researchers paid their attention on
theoretical analysis of VPM machine performance, including
back-EMF [10], torque capability and quality [11] as well as
power factor [12]. Most recently, practical application-based
optimal design of VPM machines have also become a research
hotspot [13-14].
Despite the strength in torque capability, regular VPM
machines with integral slot distributed windings (ISDW) tend to
suffer from long end turns. The end windings may occupy
considerable percentage of total machine volume and therefore
offset the advantage in torque density. In order to shorten the
end turns, none-overlapping winding and toroidal winding have
been adopted for VPM machines [15-17]. Non-overlapping
windings are suitable for VPM machines with special designed
stator topology, i.e., the stator with auxiliary tooth-slot structure
[15-16]. [17] proposed a dual-rotor VPM machine, in which
toroidal windings were applied to take advantage of the dual
airgap structure.
So far, most researches of VPM machines are conducted on
radial flux topologies. In recent years, axial flux permanent
magnet (AFPM) machines have become an important
alternative to radial flux PM machines [18-21]. With inherent
advantage in torque density, AFPM machine with high pole
number are becoming an attractive solution for low speed
applications, especially when the machine aspect ratio is a
dominant consideration [20-21]. Up to now, A few topologies
of axial flux VPM (AFVPM) machines have been found in
literature [22-26]. [22] proposed a dual stator, axial flux, spoke
type VPM machine with improved power factor. In [23],
comparative analyses were performed on a novel five-disk
axial-flux-modulated machine and its radial flux counterpart,
which shares the same operation principle with VPM machine.
This paper will provide a deep analysis on the nature of AFVPM
machines, including detailed back-EMF and torque derivation
based on geometric parameters, as well as these parameters’
effect upon machine performance.
First, a dual-rotor, toroidal-winding AFVPM machine is
presented in this paper, as illustrated in Fig. 1, which combines
the advantages of AFPM and VPM machines. Compared with
regular VPM machines, this machine topology has advantage in
end winding length. Then, the machine structure and operation
principle will be introduced In Section II. Section III will be
devoted to the back-EMF and torque derivation of the proposed
machine. Some design parameters, such as slot opening,
diameter ratio (ratio of inner to outer diameter), magnet
thickness etc., will be investigated based on the analytical
equations. Quasi-3-dimensional (quasi-3D) finite element
analysis (FEA) [27-28] and 3D FEA will be used to verify the
theoretical analysis in Section IV. The electromagnetic
performance of the AFVPM machine will be compared with a
designed regular dual rotor AFPM machine in Section V.
Finally, a prototype of the proposed machine with optimized

0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2657493, IEEE
Transactions on Industry Applications

geometric parameters has been designed and manufactured. to slot pitch remains constant. However, due to manufacturing
Experimental validation will be given in Section VI. difficulties, a radially proportional structure for an axial flux
machine may not be a cost-effective option when silicon steel
Rotor core Winding
sheet are used to fabricate iron cores. Therefore, rectangular
Magnets
instead of trapezoidal slots are chosen for the proposed machine.
On the rotor side, trapezoidal shaped magnets are preferred to
keep the pole arc unchanged and thus remain a balanced
magnetic loading along radial direction. Fig. 3 gives the sketch
of the proposed machine, with some main geometric parameters
tabulated in Table I.
Rotor yoke
N S N S N
Stator core Rotor core hss
Fig. 1. Exploded view of the proposed AFVPM machine. Stator yoke hys Lt

II. MACHINE STRUCTURE AND OPERATION PRINCIPLE g

ws
wp(r)
hm N S N S N
The exploded sketch of the proposed dual rotor AFVPM Rotor yoke hyr
machine is shown in Fig. 1. Briefly, the proposed machine is (a)
constructed in a similar way with conventional dual rotor AFPM
machine, including two rotors and one stator sandwiched in N
Slot
between. The surface mounted PMs on the two rotor disks are wp(r)
S
p
circumferentially aligned with each other and of opposite ws N
θs
polarities, forming a North-North (NN) type TORUS topology Slot θs S
[18]. Benefited from the rotor-stator-rotor structure combined N
Ri
with the NN type magnet polarity arrangement, toroidal Ro
Ri
Ro
windings can be used to significantly shorten the end winding (b) (c)
length. Fig. 3. Sketch of the proposed AFVPM machine with main geometric
Fig. 2 illustrates the flux distribution along cylindrical surface parameters. (a)cylindrical view. (b)top view of the stator. (c)top view of the
rotor.
of the proposed machine. Two strands of flux lines, driven by
PMs on different rotors, respectively, pass through two axial TABLE I
airgaps into the stator core and then travels circumferentially in LIST OF MAIN GEOMETRIC PARAMETERS
the stator back iron. It can be clearly seen that the flux lines Symbol Meaning Symbol Meaning
exhibit the property of mirror symmetry. Therefore, the hss stator slot depth ws stator slot width
analyzing process can be simplified by focusing on only one hys stator yoke thickness wp(r) stator slot pitch
rotor and half the stator. In comparison, the rotor-stator-rotor hyr rotor yoke thickness θs stator mechanical angle
hm PM thickness αp PM pole arc coefficient
topology applied in radial flux VPM machines will lead to a g airgap length Ri machine inner radius
more complex design process, since the inner and outer rotors Lt machine axial length Ro machine outer radius
should be designed individually, as well as the stator teeth
facing the inner and outer air gaps. In order to illustrate the strength of toroidal windings applied
in the dual-rotor AFVPM machine topology, a designed 22 rotor
pole pair, 2 stator pole pair case is taken as a specific example.
+B +B -C -C +A +A -B -B +C +C -A -A The geometric parameters of the stator are summarized in Table
II, in which the stator slot pitch wp(r) is calculated at average
Symmetrical Axis
radius. Fig. 4 shows three optional winding configurations per
-B -B +C +C -A -A +B +B -C -C +A +A
pole per phase for the same 24-slot stator, schematically. Then,
the copper loss and material consumption of the three cases for
Fig. 2. Flux distribution along half the cylindrical surface of the proposed
machine. producing identical stator magnetomotive force (MMF) are
calculated. The slot fill factor is assumed to be 0.5.
Different with regular PM machines, the slots and teeth of the
TABLE II
stator for VPM machine are not only for embedding windings
THE STATOR GEOMETRIC PARAMETERS FOR WINDING
or serving as part of the magnetic path, but also working as flux TYPE COMPARISON
modulators, which have significant influence on torque Item Value Item Value
performance of VPM machines. In order to maintain the same Ro 110mm Ri 66mm
modulation effect along radial direction of an AFVPM machine, hys 26mm hss 16mm
ws 11.5mm wp(r) 23.0mm
trapezoidal slot is desirable since the ratio of slot opening width

0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2657493, IEEE
Transactions on Industry Applications

remanence and thickness of the PM, respectively. With only the

major term of Fc concerned, the MMF harmonics with the order
higher than i=1 will be neglected in the following analysis.
Fc1 
Fc ( s , t )  
i 1,3,5... i
sin(i p ) cos(iPr θs  iωt )
2 (1)
 Fc1 cos( Pr θs  ωt )
(a) (b) (c) 4 Br 
Fc1  hm sin(  p ) (2)
Fig. 4. Sketch of three optional winding configurations. (a) toroidal winding. (b) dual  0  r 2
side concentrated winding. (c) dual side, single layer lap winding.
The flux modulation effect of the proposed AFVPM machine
The comparison results of the three winding configurations
is presented by introducing the airgap permeance function (3),
are summarized in Table III. The dual side concentrated
where Z is the number of stator slots, Λ0 and Λ1 stand for
winding is not a cost-effective option, since the needed phase
constant and fundamental term of the airgap permeance per unit
current is much larger than that of the toroidal winding due to
area, respectively. The coefficients of the airgap permeance
its poor winding factor, which will cause significant increase in
function can be further expressed by using the conformal
copper loss. If lap winding is adopted, there will be long end
turns, as well as difficult winding embedment and reshaping. mapping method in (4)~(7) [29], where μ0 is the permeability of
vacuum and μr the relative permeability of PMs. ws, wp(r) and g
Through rough calculation, the copper weight and loss by using
are the geometric parameters that have been described and
toroidal winding is reduced by ~49% when compared with that
summarized in Fig. 3 and Table I. Since ws is constant for
of single layer lap winding.
rectangular slots, the term ws /wp(r) is dependent on the radial
TABLE III position r. Hence, the airgap permeance is a function of r, which
COMPARISON OF THE THREE WINDING CONFIGURATIONS
results in varying flux modulation effect from the inner part to
Toroidal Concentrated Lap
Item the outer part of the proposed machine.
winding winding winding
Turns in series per phase 256 256 256
Winding factor 0.97 0.25 0.97 (s , r )  0 (r )  1 (r ) cos( Zs ) (3)
Peak phase current 10A 38.8A 10A 0 ws
Total copper weight 3.77kg 2.96kg 7.45kg 0 (r )  [1  1.6 ] (4)
Relative copper weight 1 0.78 1.98 g' wp (r )
Total copper loss 51.7W 610.9W 102.4W
0 2 0.78125 ws
Relative copper loss 1 11.8 1.98 1 (r )  { }sin[1.6 ]
g  0.78125  2[ ws ]2
' wp ( r ) (5)
wp ( r )
III. ANALYTICAL DERIVATION OF AXIAL FLUX VPM hm
g'  g  (6)
MACHINES r
Since the proposed machine exhibits the property of mirror 1 ws 2  12
symmetry from the viewpoint of flux distribution, the derivation   {1  [1  ( ) ] } (7)
2 2g '
will be focused on half the whole model. That is to say, the
following analyzing process is also applicable for regular In order to investigate the characteristics of Λ0 and Λ1, a
AFVPM machine with only one surface mounted rotor and one specific model of AFVPM machine has been built, with its
single-side open-slotted stator. It should be noted that the iron major fixed parameters listed in Table IV. Fig. 5 and Fig. 6 give
cores are assumed to have infinite permeability with no the variation of Λ0 and Λ1 with slot opening ratio (defined as ws
saturation considered. The flux leakage between PMs is also /wp(r) at average radius) and radial position r/Ro, respectively.
neglected. Besides, the radial component of flux is assumed to It should be noted that ws cannot exceed the minimum slot pitch
be negligible. wp_min, i.e., wp at innermost radius. It can be seen in Fig. 5 that
Λ0 drops with the increase of slot opening while remains almost
A. Airgap Flux Density Excited by PMs
unchanged along radial direction.
According to the magnetic circuit theory in PM machines, Meanwhile, the situation for Λ1 is more complicated as shown
one piece of PM can be regarded as one constant MMF in series in Fig. 6. At the inner part of the machine where r/Ro is small,
with a magnet reluctance. With the origin for θs illustrated in Λ1 decreases after an initial increase and reaches its peak value
Fig. 3 at the initial time, the Fourier series of the MMF Fc when ws /wp-avg is ~0.55. Meanwhile, this tendency gradually
developed by PMs can be expressed by (1), where Fc1 is the disappears with the increase of radius r, which can be attributed
amplitude of the fundamental term, Pr the number of PM pole to the limitation of ws since the range of ws /wp keeps decreasing
pairs, ω the electrical angular velocity and αp the ratio of pole along radial direction and cannot reach an optimal value at the
arc to pole pitch. Fc1 is given by (2), where Br and hm are the outer part of the machine.

0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2657493, IEEE
Transactions on Industry Applications

The variation of airgap permeance along radial direction is where k is the diameter ratio, i.e., ratio of inner to outer diameter.
further considered in the calculation of airgap flux density. As N(θs) is the winding function of one phase expressed by (11),
illustrated in [10], the airgap flux density Bg is the production of where Ps is the number of stator pole pairs. Generally, the stator
PMs’ MMF and the airgap permeance per unit area, which slot number Z, stator pole pair number Ps and rotor pole pair
yields (8). By substituting (1)-(3) into (8), Bg can be represented number Pr in a VPM machine should satisfy the relationship
by (9), which is similar with that in [10]. However, it should be expressed by (12).
noticed that Bg is related to radial position of the evaluated point, Do / 2 2
which is different from that for radial flux VPM machines.  ph   [  B ( , r, t ) N ( )d ]rdr
kDo / 2 0
g s s s (10)
TABLE IV
LIST OF MAIN MAJOR FIXED PARAMETERS 2 Ns
Symbol Meaning Quantity
N ( s )  
j 1,3,5... j Ps
kwj cos( jPs s ) (11)
Do stator outer diameter 220mm
Di stator inner diameter 132mm Z =Pr  Ps (12)
hm PM thickness 3mm
g airgap length 1mm By substituting (9), (11) and (12) into (10), λph can be
Br PM remanence 1.20T represented by (13)-(15), where kwj is the winding factor of jth
αp PM pole arc coefficient 0.9
Z stator slot number 24 harmonic, Ns the number of turns in series per phase. In the
Pr Rotor pole pair number 22 following analysis, the proposed AFVPM machine satisfies
Z=Pr+Ps, since larger torque capability can be achieved [1].
-4
x 10
Ns
3.5
 ph  2Fc1kw cos ωt ( f  0  f 1 ) (13)
3 Pr
Λ0(H/mm2)

Do / 2
0 2 1.6 ws

2.5
f0  0 (r ) rdr  Do (1  k )(1 
2
) (14)
2
kDo / 2
8g ' wp  avg
1.5 Do / 2
Pr 1
f 1   1 (r ) rdr (15)
1

Ps kDo / 2
2
0.8

0.7 0.8

r/R 0.6 0.2 0.3 0.4 0.5 0.6

The back-EMF expression can then be obtained by (16),
o 0.1
w /w
s p-avg where Ω stands for mechanical angular speed of the rotor. For
Fig. 5. Λ0 vs slot opening ratio ws /wp-avg & relative radial position r/Ro. the proposed three-phase, surface-mounted AFVPM machine
x 10
-4 with negligible cogging torque and reluctance torque, the
1
electromagnetic torque Te can be expressed by (17). Since the
0.8
attention of this paper is focused on steady-state performance of
Λ1(H/mm2)

0.6
the proposed machine from the viewpoint of electromagnetic
0.4

0.2
field, the input current waveform is assumed to be sinusoidal.
0
Therefore, Te is further expressed by (18), where Iph is the
1
amplitude of phase current. As average electrical loading Aavg is
0.9

0.8
given by (19), Te can be rewritten by (20).
0.7 0.7 0.8
0.4 0.5 0.6
d
r/R
o
0.6 0.1 0.2 0.3
w /w
e ph (t )    ph  2kw N s ΩFc1 ( f  0  f 1 )sinωt =E ph sin ωt (16)
s p-avg
dt
Fig. 6. Λ1 vs slot opening ratio ws /wp-avg & relative radial position r/Ro.
ea (t )ia (t )  eb (t )ib (t )  ec (t )ic (t )
Te  (17)
Bg (s , r, t )  Fc (s , t )(s , r ) (8) 
Bg ( s , r , t )  Fc1{ 0 (r ) cos( Pr θs  ωt ) 3 E ph I ph
Te   3kw Fc1 N s I ph ( f  0  f 1 ) (18)
 1 (r ) 2 Ω
 cos[(Z Pr )θs + ωt ] (9) 3N s I ph 3 2 N s I ph
2 Aavg   (19)
 1 (r ) 2 Davg  (1  k ) Do
 cos[(Z Pr )θs  ωt ]}
2 2
Te  kw Fc1 Aavg (1  k ) Do ( f  0  f 1 ) (20)
B. Back-EMF and Torque Expression 2
It can be seen from (16) and (18) that both the back-EMF and
The winding function theory is applied to expediently derive
torque are proportional to the term fΛ + fΛ , which is decided by
the back-EMF and torque expression. For the proposed AFVPM 0 1

the main geometric parameters of the proposed machine. fΛ is

machine, the flux linkage per phase λph can be calculated by (10), 0

directly expressed with these parameters in (12), while fΛ in (13) 1

0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2657493, IEEE
Transactions on Industry Applications

can be evaluated by numerical integration. Essentially, fΛ and 0 copper loss of the proposed machine is expressed by (21), viz,
fΛ are proportional to constant and fundamental term of the real
1

permeance over the entire airgap region. Analogous to the Pr:Ps=5:1

3.0
analytical equations for radial flux VPM machines [10], fΛ can 0
Pr:Ps=11:1
Pr:Ps=17:1
be regarded as the coefficient for a conventional AFPM machine,

f0f1 (10-6H)
while fΛ is the additional term brought in by the vernier structure.
1
2.5

It is easily seen that the parameters affecting fΛ and fΛ are D0, k, 0 1

ws, g’, Pr and Ps. By now, the influence of these parameters on 2.0
output torque of the proposed AFVPM machine can be
investigated analytically. 1.5
Fig. 7 gives the variation of fΛ and fΛ with slot opening ratio 0 1

for different values of diameter ratio k. It can be seen that fΛ 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
exceeds corresponding fΛ as ws /wp-avg gets larger than 0.45. That
0 slot opening ratio ws / wp-avg
is to say, the torque capability of an AFVPM machine can be Fig. 9. fΛ0 + fΛ1 vs slot opening ratio ws /wp-avg with different pole ratio. Ps =2,
more than twice that of its conventional counterpart. The g’=4mm.
variation of fΛ + fΛ with slot opening ratio at different diameter

0 1

ratios is illustrated in Fig. 8. It can be seen that fΛ + fΛ drops 0 1

PCu  Cu JAavg (1  k 2 ) Do 2 (21)
with the increase of k, while the optimal values of ws /wp-avg 4
remain around 0.6 for the pole ratio of 22/2. The relationship of where ρCu is the resistivity of copper, J the current density and
fΛ + fΛ with slot opening ratio for different pole ratios is shown
0 1
Aavg the average electrical loading described in (19). In the
in Fig. 9, from which it can be seen that the optimal value of ws following analysis, PCu is kept 80W.
/wp-avg increases when the pole ratio gets larger. The diameter ratio k is a key parameter for AFPM machines.
The tendency shown in Fig. 8 does not mean that k should be as
2.0
f,k=0.4 small as possible in a practical design, since it does not take the
1.6
f,k=0.6 copper loss into account. The pole ratio, defined as the ratio of
f,k=0.8
f,k=0.4 rotor to stator pole pair number, is an important design
f f (10-6H)

1.2
f,k=0.6 parameter for VPM machines, which must be carefully chosen.
f,k=0.8
Therefore, the influence of k and pole ratio on torque
performance is investigated through the analytical torque
1

0.8
equation. Fig. 10 and Fig. 11 give the torque variation with k for
0

0.4 different combinations of Ps and Pr. When Do and Ps are fixed

at 220mm and 2, respectively, the proposed machines with
0.0 higher pole ratio seems to achieve larger average torque. Fig.
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
11 shows the torque variation with k when the pole ratio is fixed
slot opening ratio ws / wp-avg
at 11/1. It can be seen that the lower stator pole pair number
Fig. 7. fΛ0 & fΛ1 vs slot opening ratio ws /wp-avg with different diameter ratios. Pr =22, leads to larger torque and the optimal k for the pole ratio of 11/1
Ps =2, g’=4mm.
is around 0.6.
3.5 The airgap structures, such as slot opening ratio, magnet
k=0.3
3.0 k=0.4 thickness, airgap length etc., also have significant influence on
k=0.5
k=0.6 torque performance of the proposed machine. The effects of
f f (10-6H)

2.5 k=0.7
k=0.8
these parameters will be investigated through the derived
2.0 k=0.9 equations in association with FEA in section IV.
1

1.5
120 Pr/Ps=5:1
0

Pr/Ps=11:1
1.0 Pr/Ps=17:1
110
0.5
Torque(Nm)

100
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
slot opening ratio ws / wp-avg
90
Fig. 8. fΛ0 + fΛ1 vs slot opening ratio ws /wp-avg with different diameter ratio. Pr =22, Ps
=2, g’=4mm. 80

In order to investigate the electromagnetic torque

70
performance with the derived equations, the thermal condition 0.2 0.3 0.4 0.5 0.6 0.7 0.8
should be fixed in the variation of design parameters. The DC Diameter Ratio k
Fig. 10. Torque vs k with different pole ratio. Ps =2, ws /wp-avg=0.5

0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2657493, IEEE
Transactions on Industry Applications

models have been built. Fig. 13 gives the comparison of torque

140 variation with slot opening between analytical method and FEA.
Ps=1
Ps=2 It can be seen that the results match well. Moreover, it should
Ps=3
120 Ps=4
be noted that for AFPM machines, the range slot opening is
Torque(Nm)

dependent on the diameter ratio k. The larger the diameter ratio

100
is, the wider range of slot opening can be obtained. For k=0.6,
the optimal slot opening ratio is ~0.63. In practical case, the
selection of slot opening and diameter ratio should also take into
80
account the room for end windings and tooth saturation at the
inner part of the machine.
60
0.2 0.3 0.4 0.5 0.6 0.7 0.8 Fig. 14 shows the torque variation with PM thickness for
Diameter Ratio k different values of k. In each curve, the torque increases rapidly,
Fig. 11. Torque vs k with different stator pole pair number. Pole ratio=11/1. reaching its peak value at the PM thickness of around 2.6mm
ws /wp-avg=0.5. and then drops slowly. The gap between analytical results and
FEA gets larger since PM thickness exceeds 10mm, which can
IV. COMPARISON OF ANALYTICAL METHOD WITH QUASI-3D be attributed to considerable flux leakage between adjacent
FEA AND 3D FEA
magnets. The variations of torque with airgap length and pole
A. Quasi-3D FEA arc coefficient are illustrated and compared in Fig. 15 and Fig.
As axial flux machines have inherent 3D flux path, the most 16, respectively, in which the results coincide well. Overall,
accurate solution to verify the proposed equations is to apply 3D sufficient accuracy has been verified within a wide variation
FEA. However, the simulation of 3D FEA models will be too range of corresponding design parameters.
time-consuming when the design parameters are changing. 140
Therefore, the so-called quasi-3D FEA is employed to make
120
more efficient comparison with the analytical results. As
illustrated in Fig. 12, the real 3D model is “cut” into several 100

layers along concentric cylindrical surfaces at certain radius,

Torque(Nm)

80
with each segment being “straightened” to a rectangular model. k=0.45, analytical
60
These rectangular models can be regarded as linear machines k=0.60, analytical
k=0.75, analytical
and analyzed with 2D FEA. The overall performance of the 40 k=0.45, Quasi-3D FEA
whole machine can then be obtained by summing the 20
k=0.60, Quasi-3D FEA
k=0.75, Quasi-3D FEA
performance of all the layers. Essentially, this method is a multi-
0
layer 2D FEA. In this part, 5-layer quasi-3D FEA is conducted 4 8 12 16 20
to balance the computation time and accuracy. By adopting the Slot opening(mm)
quasi-3D FEA instead of conventional 2D FEA, the variation of Fig. 13. Torque vs slot opening with different diameter ratio k. g=1mm.
geometric parameters along radial direction is simplified but
still taken into account, which makes the comparison more 100
convincing.
80
Torque(Nm)

60
k=0.45, analytical
40 k=0.60, analytical
k=0.75, analytical
k=0.45, Quasi-3D FEA
20 k=0.60, Quasi-3D FEA
k=0.75, Quasi-3D FEA
0
0 5 10 15 20
PM thickness(mm)
Fig. 14. Torque vs PM thickness with different diameter ratio k. g=1mm.

Fig. 12. Schematic of transforming the 3D model of the proposed machine to

corresponding quasi-3D model.

To verify the derived equations, several quasi-3D FEA

0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2657493, IEEE
Transactions on Industry Applications

140 150

120 Analytical
100
Quasi-3D FEA
100 3D FEA
50

Back-EMF(V)
Torque(Nm)

80
0
60 analytical
Quasi-3D FEA -50
40
-100
20
-150
0
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 0 50 100 150 200 250 300 350

Airgap length(mm) Rotor Position(elec.deg)

Fig. 15. Torque vs airgap length. Diameter ratio k=0.6. Fig. 17. Comparison of back-EMF waveforms.

120 analytical
115 Quasi-3D FEA
3D FEA
110
110

Torque(Nm)
Torque(Nm)

100
105
analytical
FEA 90
100

80
95
0 2 4 6 8 10 12 14
0.65 0.70 0.75 0.80 0.85 0.90 0.95
Time(ms)
Pole arc coefficient Fig. 18. Comparison of torque waveforms.
Fig. 16. Torque vs pole arc coefficient αp. Diameter ratio k=0.6.

B. 3D FEA
In order to validate both the theoretical and quasi-3D FEA
results, a 3D FEA model has been built, whose geometric
parameters have been optimized based on the foregoing analysis.
The major machine parameters are listed in Table VI, in which
the electrical loading is calculated at the average radius. Fig. 17
illustrates the comparison of back-EMF waveforms between the
analytical, quasi-3D FEA and 3D FEA results. It can been seen
that the waveforms match well with each other. Fig. 18 gives
the comparison of torque waveforms. It should be noted that Fig. 19. Magnetic field plot of the 3D FEA model and the outermost layer of quasi-
3D FEA model.
only the average torque is considered in theoretical analysis.
With the electrical loading of 164 A/cm, the analytically
V. ELECTROMAGNETIC PERFORMANCE COMPARISON WITH
calculated torque value is 110.3Nm, ~4% larger than that from
REGULAR AFPM MACHINE
3D FEA. The torque density of the 3D model can reach 32.6
kNm/m3. The torque ripples of the quasi-3D model and 3D In order to further investigate the electromagnetic
model are 0.5% and 2.3%, respectively. The performance performance, i.e., torque capability and power factor, of the dual
comparison of the three methods, along with experimental test rotor AFVPM machine, a regular 24 slot, 20 pole dual rotor
results, is summarized in Table VII. AFPM machine with the same outer diameter is designed for
comparison. The sketch of the two machines is illustrated in Fig.
The magnetic field plot of the 3D model and the outermost 20. It can be seen that the regular machine comprises a half-
layer of quasi-3D model is shown in Fig. 19. It can been seen closed slot stator embedded with dual side fractional slot
that the flux density distribution of the two models coincide well. concentrated winding (FSCW). The main design parameters of
Under the rated phase current of 16.7A, the peak value of flux the two machines are listed in Table V. To make a relatively fair
density is around 1.65T in both the stator and rotor yoke. comparison, the stator and rotor iron cores of the counterpart are
designed to keep a similar maximal flux density with that of the
AFVPM machine. Fig. 21 gives the magnetic field plots of the
two machines on load condition.

0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2657493, IEEE
Transactions on Industry Applications

TABLE V AFVPM machine is not an exception.

Main Design Parameters of the Two Machines
AFVPM Regular AFPM 150
Item AFVPM machine
machine machine Regular AFPM machine
100
Number of slots 24 24

Back-EMF(V)
Number of rotor pole pair 22 10 50
Number of stator pole pair 2 10
Stator outer diameter 220mm 220mm 0
Diameter ratio 0.6 0.6 -50
Average slot opening ratio 0.5 0.9
Total axial length 91mm 94mm -100
Stator yoke thickness 26mm 16mm
-150
Rotor yoke thickness 12.5mm 9mm
PM thickness 3.0mm 4.0mm 0 60 120 180 240 300 360
PM pole arc coefficient 0.9 0.9 Rotor Position(elec.deg)
Airgap length 1.0mm 1.0mm
2 Fig. 22. Back-EMF waveforms of the two machines at 320 rpm.
Single slot area 184mm 180mm2
Turns in series per phase 256 256
Peak phase current 16.7A 16.7A 120 1.0
Core material 50WW350 50WW350
100
Magnet material N40UH N40UH 0.8

Power Factor
Torque(Nm)
80
0.6
60

40 0.4
AFVPM machine (Torque)
20 Regular AFPM machine (Torque)
AFVPM machine (PF) 0.2
Regular AFPM machine (PF)

2 4 6 8 10 12 14 16 18
(a) (b) Phase Current(A)
Fig. 23. Comparison of output torque and power factor along with phase current.

VI. PROTOTYPE AND EXPERIMENTAL TEST OF THE PROPOSED

(c)
Fig. 20. Sketch of the AFVPM machine and the regular counterpart.
(a) 1/4 model of the dual-rotor AFVPM machine.
AFVPM MACHINE
In order to validate the foregoing analysis, a dual-rotor,
(d) toroidal-winding AFVPM prototype machine has been designed
and manufactured, as illustrated in Fig. 24. Both the stator and
rotor cores are made of tape wound laminated silicon steel. The
core material with 0.5mm thickness and 3.5 watt rated loss per
kilogram is named 50WW350 by its manufacturer. The main
design parameters are listed in Table VI. As an experimental
prototype, this machine is fabricated with a relatively low slot
fill factor of 0.35 for easy winding embedment.

(b) 1/4 model of the regular AFPM machine.

Fig. 21. On load Magnetic field plots along middle layer of the two quasi-3D FEA
models.

Fig. 22 shows the phase back-EMF waveforms of the two

machines at 320 rpm through Quasi-3D FEA. Due to the flux
modulation effect, the fundamental amplitude of back-EMF
induced by the AFVPM machine is ~86% larger than that of the
regular one. The variations of output torque and power factor (a) stator and rotor cores (b) stator with toroidal windings
along with phase current are compared in Fig. 23. It can be seen
that with the same phase current and similar copper loss, the
output torque of the AFVPM machine can be increased by more
than 80% compared with the counterpart. The power factor of
the AFVPM machine, as expected, drops quickly with the
increase of phase current, and is much lower than that of the
regular AFPM machine. It should be noted that the low power
factor has been a major challenging for application extension of (c) rotor with surface mounted PMs (d) prototype
most vernier machine topologies [8-12], in which the presented Fig. 24. The proposed AFVPM prototype machine.

0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2657493, IEEE
Transactions on Industry Applications

TABLE VI When the prototype is driven at rated condition, i.e., 320 rpm
Main Parameters of the Prototype and test bench
with 16.7A phase current, the electromagnetic performances are
Item Value summarized in Table VII. The measured power factor is close
Number of slots 24 to the predicted value of 3D FEA. As the end winding
Number of rotor pole pair 22
Stator outer diameter 220mm inductance is not considered in the analytical calculation and
Diameter ratio 0.6 Quasi-3D FEA, the power factor calculated by these two
Total axial length 91mm methods differ considerably from the measured one.
Magnet thickness 3.0mm
Airgap length 1.0mm
Slot depth 16.0mm
Slot width 11.5mm
Slot opening ratio 0.5
Pole arc coefficient 0.9
Electrical loading 164A/cm
Current density 5.8A/mm2
Slot fill factor 0.35
Turns in series per phase 256
Rotation speed 320rpm
Peak phase back-EMF 150V
DC voltage of the inverter 550V
Core material 50WW350
Magnet material N40UH
Fig. 26. Measured three phase back-EMF waveforms. Rotation speed=320rpm.

The on-load test bench is shown in Fig. 25. As the neutral

point of the three phase winding is drawn out of the casing, the 120
Analytical
phase back-EMF waveform can be measured, which is given in 100
Quasi-3D FEA
Measured
Fig. 26. It can be seen that the back-EMF waveform is Torque(Nm)
sinusoidal and its amplitude agrees well with both the analytical 80

and FEA results shown in Fig. 17. The torque vs. current curves 60
are compared in Fig. 27. It can be seen that the Quasi-FEA
40
predicted and experimentally measured curves match well,
which are both slightly lower than the analytical one. 20

0
Prototype 2 4 6 8 10 12 14 16 18
Phase Current(A)
Fig. 27. Average torque vs. phase current.

100
Torque transducer Current probe
90
Dyno 80
machine Oscilloscope
Efficiency(%)

70 FEA Calculated
Measured
Fig. 25. On-load test bench of the AFVPM prototype machine. 60

Fig. 28 gives the comparison of FEA predicted and measured 50

efficiency at 320 rpm. The DC copper loss with end windings 40

counted, stator and rotor iron loss and PM loss are included in
30
the FEA calculated efficiency. When the phase current is small, 2 4 6 8 10 12 14 16
the measured efficiency is significantly lower than the measured Phase Current(A)
one, due to relatively high no-load mechanical loss. As the Fig. 28. FEA predicted efficiency and measured efficiency vs. phase current. Rotation
phase current increases, the gap between the two efficiency speed = 320rpm.
curves narrows. At rated load, the measured efficiency is 86.0%,
TABLE VII
~4.4% lower than the FEA predicted one. For this prototype, PERFORMANCES COMPARISON OF THE PROPOSED MACHINE
copper loss is the main loss component, which is independent Item Analytical Quasi-3D 3D Measured
from the rotation speed. Since the output power is proportional Avg. Torque /Nm 110.3 105.6 106.1 103.8
to both torque and speed, low speed high torque machine Torque density /kNm/m3 33.9 32.7 32.6 31.9
Power factor 0.62 0.64 0.59 0.57
topologies such as vernier machines always tend to have Calculation time <1s 40min 8hour -
relatively poor efficiency.

0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2657493, IEEE
Transactions on Industry Applications
VII. CONCLUSION
In this paper, a dual-rotor, toroidal-winding, axial flux vernier
permanent magnet (AFVPM) machine has been and analyzed.
Benefited from the rotor-stator-rotor topology, the end winding
length is significantly reduced by adopting toroidal winding
configuration. Detailed back-EMF and torque expressions for
AFVPM machines have been analytically derived. Performance
evaluation has been made through both analytical calculation
and quasi-3D FEA within a wide range of machine parameters
and the results turn out to match well. With adequate accuracy,
the proposed analytical method shows strength in calculation
time. It has been found that some design parameters, such as
pole ratio, diameter ratio, slot opening, PM thickness and airgap
length have significant influence on torque performance of the
proposed machine. For the proposed machine with pole ratio of
11/1, the optimal diameter ratio and average slot opening ratio
are around 0.60 and 0.63, respectively.
3D FEA based on an optimized machine model has also been
conducted to validate the forgoing analyzing results. Through
comparison with a designed regular AFPM machine, the
analyzed AFVPM machine shows obvious advantage in torque
density. Finally, a dual-rotor AFVPM prototype machine has
been manufactured. Test results show good agreement with the
predicted ones, and the advantage in torque density of the
proposed machine has been experimentally demonstrated.
VIII. ACKNOWLEDGEMENT
This work was supported by two grants from National
Natural Science Foundation of China (NSFC) under Project No.
51337004 and Project No. 51407082.
REFERENCES
[1] A. Toba and T. Lipo, “Generic torque-maximizing design methodology
of surface permanent-magnet vernier machine,” IEEE Trans. Ind. Appl.,
vol. 36, no. 6, pp. 2032-2035, Jun.2005.
[2] R. Qu, D. Li, and J. Wang, “Relationship between magnetic gears and
vernier PM machines,” in Proc. Int. Conf. Electrical Machines & Systems,
Beijing, China, Aug. 18–20, 2011, pp:1-6.
[3] D. Li, R. Qu, and Z. Zhu, “Comparison of Halbach and dual-side vernier
permanent magnet machines,” IEEE Trans. Magn., vol. 50, no. 2, pp.
7019804, Feb. 2014.
[4] S. L. Ho, S. Niu, and W. N. Fu, “Design and comparison of vernier
permanent magnet machines”, IEEE Trans. Magn, vol. 47, no. 10, pp.
3280-3283, Oct. 2011.
[5] A. Ishizaki, et al., “Theory and optimum design of PM vernier motor,” in
Proc. IEE Int. Conf. Electrical Machines and Drives ’95, 1995, pp.208–
212.
[6] A. Toba and T. Lipo, “Novel dual-excitation permanent magnet vernier
machine,” in Conf. Rec. 34th IEEE IAS Annu. Meeting, Oct. 1999, vol. 4,
pp. 2539-2544.
[7] J. Li, K. T. Chau, J. Jiang, C. Liu and W. Li, “A new efficient permanent-
magnet vernier machine for wind power generation,” IEEE Trans. Magn.,
vol. 46, no. 6, pp.1475-1478, Jun. 2010.
[8] D. Li, R. Qu, and T. Lipo, “High power factor vernier permanent magnet
machines,” IEEE Trans. Ind. Appl., vol. 50, no. 6, pp. 3664-3674,
Nov./Dec. 2014.
[9] D. Li, R. Qu, J. Li, and W. Xu, “Consequent pole, toroidal winding, outer
rotor vernier permanent magnet machines, ” IEEE Trans. Ind. Appl., vol.
51, no. 6, pp.4470-4481, Nov./Dec. 2015.
0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
[10] B. Kim, and T. Lipo, “Operation and design principles of a PM vernier
motor,” IEEE Trans. Ind. Appl., vol. 50, no. 6, pp. 3656-3663, Nov./Dec.
2014.
[11] D. Li, R. Qu, J. Li, L. Wu and W. Xu, “Analysis of torque capability and
quality in vernier permanent magnet machines,” IEEE Trans. Ind. Appl.,
vol. 52, no. 1, pp. 125-135. Jan. /Feb. 2016.
[12] L. Wu, R. Qu, D. Li, and Y. Gao, “Influence of pole ratio and winding
pole numbers on performances and optimal design parameters of surface
permanent-magnet vernier machines,” IEEE Trans. Ind. Appl., vol. 51, no.
5, pp. 3707-3715. Sept. /Oct. 2015.
[13] S. Gerber, R. Wang, “Design and evaluation of a PM vernier machine,” in
Proc. 2015 IEEE Ener. Con. Cong. & Exp., Sep. 2015, pp.5188- 5194.
[14] B. Kim, T. Lipo, “Design of a surface PM vernier motor for a practical
variable application,” in Proc. 2015 IEEE Ener. Con. Cong. & Exp., Sep.
2015, pp.776- 783.
[15] K. Okada, N. Niguchi, and K. Hirata, “Analysis of a vernier motor with
concentrated windings,” IEEE Trans. Magn., vol. 49, no. 5, pp. 2241–
2244, May. 2013.
[16] J. Yang, G. Liu, W. Zhao, Q. Chen, Y. Jiang, L. Sun, and X. Zhu,
“Quantitative comparison for fractional-slot concentrated-winding
configurations of permanent-magnet vernier machines,” IEEE Trans.
Magn., vol. 49, no. 7, pp. 3826–3829, Jul. 2013.
[17] S. Niu, S. Ho, W. Fu, and L. Wang, “Quantitative comparison of novel
vernier permanent magnet machines,” IEEE Trans. Magn., vol. 46, no. 6,
pp. 2032-2035, June. 2005.
[18] S. Huang, J. Luo, F. Leonardi, and T. Lipo, “A comparison of power
density for axial flux machines based on general purpose sizing
equations,” IEEE Trans. Energ. Convers., vol. 14, no. 2, pp. 185-192, Jun.
1999.
[19] A. Cavagnino, M. Lazzari and F. Profumo, “A comparison between the
axial flux and the radial flux structures for PM synchronous motors”, IEEE
Trans. Ind. Appl., vol. 38, no. 6, pp. 1517-1524, Nov./Dec. 2002.
[20] D. J. Patterson, J. L. Colton, B. Mularcik, B. J. Kennedy, S. Camilleri, and
R. Rohoza, “A comparison of radial and axial flux structures in electrical
machines,” in Proc. IEEE IEMDC, May 2009, pp. 1029–1035.
[21] A. Di Gerlando, G. Foglia, M. Iacchetti, and R. Perini, “Axial flux PM
machines with concentrated armature windings: Design analysis and test
validation of wind energy generators,” IEEE Trans. Ind. Electron., vol. 58,
no. 9, pp. 3795–3805, Sep. 2011.
[22] F. Zhao, T. Lipo, B. Kwon, “A novel dual-stator axial-flux spoke-type
permanent magnet Vernier machine for direct-drive applications,” IEEE
Trans. Magn., vol. 50, no. 11, pp. 8104304, Nov. 2014.
[23] L. Li, W. N. Fu, S. Niu and Y.Li, “A quantitative comparison study of
power electronic driven flux-modulated machines using magnetic field
and thermal field co-simulation” , IEEE Trans. Ind. Electron., vol. 62, no.
10, pp. 6076-6084, Oct. 2015.
[24] T. Zou, R. Qu, J. Li, and D. Li, “A consequent pole, dual rotor, axial flux
vernier permanent magnet machine”, Int. Conf. Ecological Vehicles and
Renewable Energies, Monte Carlo, Monaco, Mar. 31–Apr. 2, 2015, pp:1-
9.
[25] T. Zou, R. Qu, J. Li, and D. Li, “Analysis and design of a dual-rotor axial-
flux vernier permanent magnet machine”, in Proc. 2015 IEEE Ener. Con.
Cong. & Exp., Sep. 2015, pp. 3906- 3913.
[26] F. Zhao, M. Kim, B. Kwon, and J. Baek, “A small axial-flux vernier
machine with ring-type magnets for the auto-focusing lens drive system,”
IEEE Trans. Magn., vol. 52, no. 7, pp. 8204604, July. 2016.
[27] A. Egea, G. Almandoz, J. Poza, G. Ugalde, and A. J. Escalada, “Axial-
flux-machine modeling with the combination of FEM-2-D and analytical
tools,” IEEE Trans. Ind. Appl., vol. 48, no. 4, pp. 1318-1326, July./Aug.
2012.
[28] A. Parviainen, M. Niemelä, and J. Pyrhönen, “Modeling of axial flux
permanent-magnet machines”, IEEE Trans. Ind. Appl., vol. 3640, no. 5,
pp. 1333-1340, Sep./Oct. 2004.
[29] Z. Q. Zhu and D. Howe, “Instantaneous magnetic-field distribution in
brushless permanent-magnet dc motors, part III effect of stator slotting,”
IEEE Trans. Magn., vol. 29, no. 1, pp. 143-151, Jan. 1993.