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/TMAG.2015.2481862, IEEE Transactions on Magnetics

> FOR CONFERENCE-RELATED PAPERS, REPLACE THIS LINE WITH YOUR SESSION NUMBER, E.G., AB-02 (DOUBLE-CLICK HERE) < 1

Analytical Solution of the Magnetic Field and EMF calculation in

Ironless BLDC Motor Used for Flywheel
Xiangdong Liu1, Hengzai Hu1, Jing Zhao1, Anouar Belahcen2, Liang Tang3, Lei Yang3
1
School of Automation, Beijing Institute of Technology, Beijing, 100081, China
2
Department of Electrical Engineering and Automation, Aalto University, Aalto, Espoo, FI-00076, Finland
3
Beijing Institute of Control Engineering, Beijing, 100190, China

The ironless brushless DC (BLDC) motors are widely used for flywheel due to their light mass, high efficiency and high inertia to
mass ratio. In this paper, a 2D analytical solution for predicting the magnetic field distribution in ironless BLDC motor is presented
and compared with 2D finite element method. The results calculated by the proposed 2D analytical method and 2D FEM are in good
agreement. It is also found that the intensity and direction of the magnetic field change largely in the coil region, which makes the
electromotive force (EMF) of strands in one coil package different from each other. Hence, three different models to calculate the EMF
of ironless BLDC motor are proposed and compared. Meanwhile, the axial end-effect and influence of end winding on EMF are both
considered, and then a correction factor is proposed for ironless BLDC motor. The effectiveness of the proposed methods is validated
by experimental results. Due to its accuracy and computational speed, the method can be used in the design and optimization of
ironless BLDC motors.

Index Terms—Analytical solution, ironless BLDC motor, EMF calculation, axial end-effect, correction factor.

accurate and thus, preferred for initial design and optimization.

I. INTRODUCTION Recent comprehensive reviews on the analytical methods can

D UE to the advantages of high efficiency, light mass and

small volume, the flywheel based on brushless DC
be found in the literature [17], [18]. The magnetic field
distribution is usually obtained by solving the Maxwell's
(BLDC) motor with ironless stator is widely used in the equations in all the studied subdomains through the separation
attitude control system of the micro-satellite [1]-[3]. Different of variables technique, and the use of adequate boundary and
from the ordinary BLDC motors, the stator of ironless BLDC interface conditions [19]-[24]. In [8], the magnetic field in an
motor is made not of iron but of a nonmagnetic plastic ironless BLDC motor is analyzed by an approximate
material which makes the motors have the advantages: (1) no analytical model through specifying a constant potential
iron loss in the stator resulting in higher efficiency; (2) no distribution on the pole surface. Only the magnetic field in air
cogging torque, which would arise from the interaction region is solved and this model doesn't have adequate
between PMs and stator iron; (3) lower motor mass [4]-[7]. accuracy in some special applications. In [19], the magnetic
But on the other hand, the nonmagnetic material makes the field of a double-side Halbach permanent magnet synchronous
ironless stator lack the flux focusing capability for guiding the motor/generator (PMSM/G) with ironless stator used for
magnetic flux paths [8]. This makes the intensity and direction flywheel energy storage system is analyzed by 2D analytical
of the magnetic field change largely and the magnetic circuit method and the electromotive force (EMF) is calculated to
becomes more complex. Meanwhile the axial end-effect will estimate the machine performances. In [20], [21], the
be larger compared with the iron BLDC motor. analytical calculation of the magnetic field in an ironless axial-
Different approaches have been developed to predict the flux PMSM is presented and verified by FEM solution. In [18],
magnetic field distribution in permanent magnet motors. At [22], [23], the magnetic field is also calculated by subdomain
the present time, the numerical methods for field computation, method in a slotless PM motor and compared with FEM
such as the finite element method (FEM) or the finite results.
difference method, can provide accurate results accounting for The EMF is a very important parameter to evaluate the
the iron saturation and accurate geometrical shapes [8-15], but electromagnetic performance of electrical machines and is
these numerical methods are generally difficult to provide heavily investigated in many publications [17]-[21], [24]-[28].
straightforward physical relationships between the In [19]-[21], due to the double-side rotor structure of radial-
electromagnetic performances and the design parameters and flux or axial-flux ironless PM motor, the magnetic flux density
are often very time-consuming, especially for 3D transient in the coil region is uniform. In these cases the EMF could be
analysis. Furthermore, due to the larger rate of change of the calculated easily based on Faraday's law. In [18], [27], the
magnetic field in ironless PM motors, the regions should be full-pitched windings are assumed to be a cylindrical area and
subdivided into finer elements to be able to correctly model concentrated on their average radius. In such an approximation
the skin effect [16]. The analytical models are fast and enough the EMF is calculated without accounting for the difference
between the separate strands. Meanwhile, the effect of
Manuscript received December 11, 2014 (date on which paper was different magnetization patterns on the EMF is investigated.
submitted for review). Corresponding author: Jing Zhao (e-mail: However, for ironless BLDC motor, the EMF is sensitive to
[email protected]).
Digital Object Identifier inserted by IEEE radial location of the coil regions, due to the large difference

0018-9464 (c) 2015 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/TMAG.2015.2481862, IEEE Transactions on Magnetics

> FOR CONFERENCE-RELATED PAPERS, REPLACE THIS LINE WITH YOUR SESSION NUMBER, E.G., AB-02 (DOUBLE-CLICK HERE) < 2

in the rate of change of the magnetic field. The EMF where M r is the radial component of the magnetization vector
calculation methods of iron motors are not suitable for ironless and M  is the tangential component. er and e are the radial
BLDC motor. In addition, as for the ironless BLDC motor
and tangential unit vectors respectively. The radial and
used for flywheel, the length to diameter ratio is always made
tangential components can be expressed as Fourier series [22],
small to maximize the inertia. The axial end-effect has a huge
[29] as follows
impact on the EMF and cannot be ignored. Meanwhile, the 
influence of end winding on EMF calculation should also be
considered.
M r    
n 1,3,5,
M rn cos  np  (2a)

In this paper, an ironless BLDC motor with external rotor is 

investigated. Firstly, the analytical magnetic field solution of M     

n 1,3,5,
M  n sin  np  (2b)
the ironless BLDC is obtained in PM and air regions. The
results are verified by 2D FEM computations. Secondly, the where M rn and M  n are the nth spatial harmonics of the
field expression is used to calculate EMF without accounting radial and tangential magnetization components respectively.
for axial end-effect and end winding. Three EMF analytical n is the spatial harmonic order, p is the number of pole-pairs
calculation models are proposed and compared. Thirdly, the and θ is the angular position.
axial end-effect of ironless BLDC motor is considered by For the radially magnetized PM
introducing a correction factor. The factor is calculated by 2D 4 Br p n p
analytical solution and 3D static analysis. Then the EMF M rn  sin (3a)
0 n 2
accounting for axial end-effect and end winding is obtained.
M  n ( )  0 (3b)
Finally, a prototype is manufactured. The effectiveness of the
proposed methods is validated by experimental tests. where Br is the remanent flux density of the PM, αp is the
magnet pole-arc to pole-pitch ratio and 0 is the free space
II. ANALYTICAL SOLUTION OF IRONLESS BLDC MOTOR permeability. For parallel magnetized PM [18]
A. Analytical Model of Ironless BLDC Motor Br p
M rn   A1n  A2 n  (4a)
The analytical model of ironless BLDC motor shown in 0
Fig.1 is based on the following assumptions: 1) the Br p
permeability of rotor iron is infinite and the relative M n   A1n  A2 n  (4b)
0
permeability of nonmagnetic stator material is unity, which
where
can be regarded as vacuum; 2) axial end-effect is ignored and
  p 
the problem is simplified to a 2-D case (this assumption will sin   np  1 
be later corrected by an additional coefficient); 3) the  2p 
A1n  (5a)
demagnetization characteristic of PM is linear and its relative 
permeability is constant. 4) the relative permeability of the air-  np  1 p
2p
gaps between the PMs is assumed to be equal to that of the
PMs.    p 
 sin   np  1 
μr→∞ αpπ/p I : PM region   2p 
II : air region (air-gap and for np  1
A2 n    p (5b)
  np  1 2 p
ironless stator)
Rm Rr Rr : inner radius of rotor
Rm : inner radius of PM 
2π/p μr : relative permeability 1 for np =1
αp : magnet pole-arc to pole-
pitch ratio The problem domain is divided into two regions as shown
p : number of pole-pairs
in Fig. 1: Region I is the PMs and Region II is the air-gap. In
II I different regions, the vector fields B and H are coupled by
the following constitutive relations:
In region I B  0 r H   0 M (6)
In region II B  0 H  (7)
where  r is the relative permeability of the PMs.
The magnetic flux density B is resolved by introducing the
Fig. 1 The analytical model of ironless BLDC motor
magnetic vector potential A . For the 2-D case, the magnetic
B. PM model and governing equations vector potential reduces to its axial component Az and the
In this paper, the radial magnetization and parallel radial and tangential components of the flux density are
magnetization of the PMs are both considered. In polar defined by
coordinates, the magnetization vector M of a PM can be 1 Az A
Br  and B   z (8)
given by r  r
M  M r er  M  e (1) The governing equation in Region I is a Poisson equation and

0018-9464 (c) 2015 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/TMAG.2015.2481862, IEEE Transactions on Magnetics

> FOR CONFERENCE-RELATED PAPERS, REPLACE THIS LINE WITH YOUR SESSION NUMBER, E.G., AB-02 (DOUBLE-CLICK HERE) < 3

in Region II is a Laplace equation, i.e.,

1  R  
np 1 np 1

 r 
 2 Az 1 Az 1  2 Az   M r  B   r ,       F1n    G1n  m   sin  np 
In region I   2   0  M  (9) n 1,3,5... Yn   Rr   r  
r 2
r r r  2
r    


In region II
 2 Az 1 Az 1  2 Az
  2 0 (10)
 
n 1,3,5...
tn  r  sin  np 
r 2 r r r  2
(17b)
C. The boundary conditions
1   r  
np 1 np 1

 r 
In the interfaces between the two regions, the field vector Br   r ,      F2 n  
n 1,3,5... Yn  R
 G2 n    cos  np 
B and H should follow   r  Rm  
n   B1  B2   0 (11a) (17c)
1   r  
np 1 np 1
n   H1  H2   K (11b) 
 r 
B   r ,       F2 n    G2 n    sin  np 
where n is the unit vector normal to the interface between two n 1,3,5... Yn   Rr   Rm  

regions. For the ironless BLDC motor analysis model in this
(17d)
paper, the relationships can be written as
where
 H   r ,   r  R  0 Yn    r  1 Rm2 np Rr2 np    r  1
 r (18a)
 Br  r ,  r  R  Br  r ,   r  R  0 np  M rn  npM  n 
 m m
(12) np  1
 
 np   1
2
 H   r ,  r  Rm  H   r ,   r  Rm 
ln   (18b)
   0  M  1  M r1 
 Br  r ,   r  0    np  1
 2
   r  1 ln   r npkn  ln  Rm Rr
D. Analytical solution  np 1  np 1
np  1
By using the method of separation of variables, the F1n   (18c)
 r  1  k1  ln Rm  ln Rr   l1   l1 Rm Rr
2 2
np  1
analytical solutions of Laplace equation and Poisson equation
are obtained as  r  1 ln Rmnp 1 Rr np 1    r npkn  ln  np  1
In Region I, for np  1 , G1n   (18d)
k1   r  1 ln Rm  ln Rr   r l1 np  1

Az  r ,     A n r np  Bn r  np  kn r  sin  np  (13)
 2ln   npkn  ln  Rm Rr
 np 1  np 1
np  1
n 1,3,5,... F2 n   (18e)
l1  Rm Rr  1 np  1
2 2
and for np =1
Az  r ,     An r  Bn r 1  kn r ln r  sin  (14)  npkn  ln np  1
G2 n   (18f)
where  2k1  ln Rm  ln Rr  np  1
 0  M  n  npM rn 
 np  1 npkn np  1
  np   1
2 sn  r    (18g)
kn   (15) k1  ln r  ln Rm  np  1
 0  M  1  M r1 
 np  1  kn np  1
 2 tn  r    (18h)
In Region II  k1  ln r  1  ln Rm  np  1
 The developed analytical solution has been applied to an
Az  r ,    A
n 1,3,5,...
n r np  Bn r  np  sin  np  (16) external-rotor ironless BLDC motor with radial magnetization
or parallel magnetization. The main parameters of the motor
where An , Bn , A1 , B1 , An , Bn are integral constants and are listed in Table I. Note that for this first part, only the
can be determined by applying the boundary conditions. parameters related to the PM and the rotor are needed. The
Hence, the complete solution of the magnetic field in coil parameters are needed in later Sections. The radial
Region I and Region II can be deduced as component and tangential component of the magnetic flux
1   r  density on different lines shown in Fig.1 are calculated by the
R  
np 1 np 1

Br   r ,      F1n    G1n  m   cos  np  analytical method and validated by FEM, as shown in Figs.
n 1,3,5... Yn  R  r  
  r 2~9. In FEM, the relative permeability of rotor core μr is 2000
 instead of infinite. From the comparison between the
 
n 1,3,5...
sn  r  cos  np  analytical results and the 2D FEM results, it can be seen that
there is a very good agreement for both radial and parallel
(17a)
magnetization cases. And the radial component decreases
significantly with the decrease of the radius in Region II. In
other words, the radial component in the winding region

0018-9464 (c) 2015 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/TMAG.2015.2481862, IEEE Transactions on Magnetics

> FOR CONFERENCE-RELATED PAPERS, REPLACE THIS LINE WITH YOUR SESSION NUMBER, E.G., AB-02 (DOUBLE-CLICK HERE) < 4

which locates between r=21.6mm and r=24.5mm changes 1.2

significantly. So the winding should not be considered as a Analytical FEM

0.8 r=25.5mm

Radial component of flux density (T)

lumped parameter when the electromagnetic performances, r=26mm
such as EMF and electromagnetic force, are calculated. The 0.4
r=26.5mm
r=27mm
asymmetry of phase EMF and the circulating current may also r=27.5mm
occurs due to the placement error of strands. Fig. 10 shows the 0.0
magnetic flux distribution for radial magnetization and parallel
magnetization. -0.4
TABLE I
PARAMETERS OF THE IRONLESS BLDC MOTOR -0.8
Valu
Symbol Parameter Unit
e -1.2
p Number of pole pairs 7 - -90 -45 0 45 90 135 180 225 270
Q Total number of slots 12 - Angle (elec. deg)
αp Magnet pole-arc to pole-pitch ratio 1.0 - Fig. 4 Tangential component of the magneic flux density in Region I for radial
Rr Inner radius of the rotor 25.0 mm magnetization.
hm Height of PM 3 mm 1.2
L1 Axial length of the winding 12 mm
θs Coil pitch 30 mech. deg 0.8 Analytical FEM

Radial component of flux density (T)

Rso Outer radius of the coil region 24.5 mm r=25.5mm
Rsi Inner radius of the coil region 21.8 mm r=26mm
0.4 r=26.5mm
θci Maximum span angle of coil region 29 mech. deg
r=27mm
θco Minimum span angle of coil region 20 mech. deg r=27.5mm
μr Relative recoil permeability of the 1.043 0.0
PMs
Br Remanent flux density of the PMs 1.28 T -0.4

1.6
-0.8
1.2
Radial component of flux density (T)

Analytical FEM
r=25.5mm -1.2
0.8 -90 -45 0 45 90 135 180 225 270
r=26mm
r=26.5mm Angle (elec. deg)
0.4
r=27mm Fig. 5 Tangential component of the magneic flux density in Region I for
0.0 r=27.5mm parallel magnetization.
0.6
-0.4 Analytical FEM
R=22mm
Radial component of flux density (T)

0.4 R=22.5mm
-0.8
R=23mm
-1.2 0.2 R=23.5mm
R=24mm
-1.6 R=24.5mm
-90 -45 0 45 90 135 180 225 270 0.0
Angle (elec. deg)
Fig. 2 Radial component of the magneic flux density in Region I for radial
-0.2
magnetization.
1.6
-0.4
1.2
Analytical FEM
Radial component of flux density (T)

r=25.5mm -0.6
0.8
r=26mm -90 -45 0 45 90 135 180 225 270
r=26.5mm Angle (elec. deg)
0.4
r=27mm Fig. 6 Radial component of the magneic flux density in Region II for radial
0.0 r=27.5mm magnetization.
0.6
-0.4 Analytical FEM
R=22mm
Radial component of flux density (T)

-0.8 0.4 R=22.5mm

R=23mm
-1.2 0.2 R=23.5mm
R=24mm
-1.6 R=24.5mm
-90 -45 0 45 90 135 180 225 270 0.0
Angle (elec. deg)
Fig. 3 Radial component of the magneic flux density in Region I for parallel
-0.2
magnetization.

-0.4

-0.6
-90 -45 0 45 90 135 180 225 270
Angle (elec. deg)
Fig. 7 Radial component of the magneic flux density in Region II for parallel
magnetization.

0018-9464 (c) 2015 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/TMAG.2015.2481862, IEEE Transactions on Magnetics

> FOR CONFERENCE-RELATED PAPERS, REPLACE THIS LINE WITH YOUR SESSION NUMBER, E.G., AB-02 (DOUBLE-CLICK HERE) < 5

1.2 density in Region II and dS is the normal vector of the coil

Analytical FEM
Tangential component of flux density (T)

R=22mm surface area. According to Faraday’s law, the back EMF

0.8 R=22.5mm induced in each coil is equal to the negative derivative of the
R=23mm
0.4 R=23.5mm flux linked by the coil, i.e.,
R=24mm d  r ,  c 
0.0
R=24.5mm e  r , c    (22)
dt
-0.4
The total back EMF is calculated by adding the EMF
induced in all coils in series. The EMF induced in adjacent
-0.8 slots is phase shifted so they should be summed as vectors.
Considering the characteristic of magnetic flux distribution
-1.2 of ironless BLDC motor, different EMF calculation methods
-90 -45 0 45 90 135 180 225 270
Angle (elec. deg) are presented in this section. Three calculation models, shown
Fig. 8 Tangential component of the magneic flux density in Region II for in Fig. 11, are proposed and their performances are compared
radial magnetization. in details in the following section.
1.2
Analytical FEM
Tangential component of flux density (T)

R=22mm
0.8 R=22.5mm
R=23mm
0.4 R=23.5mm
R=24mm
R=24.5mm
0.0

-0.4

-0.8 (a) single coil model

-1.2
-90 -45 0 45 90 135 180 225 270
Angle (elec. deg)
Fig. 9 Tangential component of the magneic flux density in Region II for
parallel magnetization.

A A A A A A slot center line

C C

(b) single-layer concentric coils model

(a) (b)
Fig. 10 FEM-computed magnetic flux density distribution: (a) radial
magnetization; (b) parallel magnetization.

III. BACK EMF CALCULATION OF IRONLESS BLDC MOTOR l ... 1

1 ...
...
... 2 ...
Considering the characteristic of magnetic field in the
...

...
...

... s ...
winding region, the EMF calculation of ironless BLDC motor
slot center line
is further investigated. If the rotor makes a rotation α, the flux
density radial component in Region II can be written as (c) multi-layer concentric coils model

Fig.11 Calculation models of the back EMF: (a) Model 1: single coil model;
Br   r ,      
n 1,3,5...
f n  r  cos np     (19) (b) Model 2: single-layer concentric coils model; (c) Model 3: multi-layer
concentric coils model.
where
A. Model 1: single coil model
1   r   r  
np 1 np 1

f n  r    F2 n   Similar to the calculation method of slotless BLDC motor

 G2 n    (20)
Yn   Rr   Rm   in [1], the coils in Model 1 are assumed to be concentrated on
 
the slot center line at the average radius of the slot. The
with   r t   0 the rotor positional angle, r the rotor magnetic linkage linked with the coils in the slot can be
angular velocity, and  0 the initial rotor position. In order to expressed as
s / 2
calculate the induced EMF, it is required to obtain the   N s Lef  Rav f n  Rav  cos np     d (23)
 s / 2
magnetic flux linked by a single coil using
  r ,  c    B  dS (21) where N s is the number of coil turns, Rav   Rso  Rsi  2 is

where  c is the coil span angle, B is the magnetic flux the average radius of the coil area, Lef is the effective length

0018-9464 (c) 2015 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/TMAG.2015.2481862, IEEE Transactions on Magnetics

> FOR CONFERENCE-RELATED PAPERS, REPLACE THIS LINE WITH YOUR SESSION NUMBER, E.G., AB-02 (DOUBLE-CLICK HERE) < 6

of the coil side, and  s is the span angle of the slot. 6

B. Model 2: single-layer concentric coils model 4

In model 2, the coils are assumed to be evenly distributed
on the coil region. The coil turn density can be obtained by
2
N
c  s (24)

EMF (V)
Sc Analytical FEM
0 Model 1
where Sc is the cross area of the coil region. The magnetic Model 2
linkage linked with the coils in a slot can be expressed as Model 3-(1,1)
-2
Model 3-(2,2)
    c Lef  
Rso  co c / 2
rf r cos np     d  rdrd c (25)
 c / 2 n  
Model 3-(3,3)
Rsi ci 
-4
C. Model 3: multi-layer concentric coils model
In model 3, the coil region is divided into ( l  s ) regions and -6
0 60 120 180 240 300 360
the coils of each region are assumed to be concentrated on the Angle (elec. deg)
center of the subregion. Different numbers of subregions are Fig. 13 The computed EMF waveforms with different calculation methods
shown and illustrated in Fig. 12. compared with that obtained by FEM.

IV. AXIAL END-EFFECT

In the previous analysis, the 2D analytical solution and
l  s  1,1 l  s  1, 2  EMF are obtained without accounting for the axial end-effect.
In actual motor, the axial end-effect may have different impact
... on the magnetic field distribution and EMF when the motors
...

...
...
...

...

have different axial length, outer diameter and pole pairs.

l  s   2, 2  l  s  l, s  Meanwhile considering the fringe effect of PM, the end
Fig. 12 Some samples of the subregion. winding is also in the region of rotor magnetic field and
The magnetic linkage linked with the coils in a slot can be influences the EMF calculation. In order to evaluate the axial
expressed as end-effect and account for the influence of end winding on
s l EMF, a correction factor Ke is introduced by carrying out a 3D
f n  Rij  cos np     d
cij / 2
   N sij Rij Lef  static FEM. For a specific motor, the Ke varies with the radial
 cij / 2
1 1 (26)
radius and axial length. The calculation of Ke is defined as
1  i  l ,1  j  s Lef
Br  3 D  r , 0, z  dz
where N sij is coil number of the coil subregion, Rij is the K e  r , Lef   0

Br   r , 0   L1
 100% (27)
radius of the coil subregion centre and  cij is the span angle of
the coil subregion defined by the indices i and j. The number where Br  r ,   is the radial component obtained by 2D
of subregions depends on the shape and size of the coil region. analytical solution and Br  3 D  r ,  , z  is obtained by 3D static
Without accounting for axial end-effect and influence of
FEM.
end winding, that is Lef=L1, the EMF waveforms of ironless
  0 is selected as the position for the estimation of the
BLDC motor with parallel magnetization calculated by the
presented three 2D analytical models and 2D FEM are shown parameter Ke. The 2D and 3D analysis models are shown in
in Fig. 13. From Fig. 13, it can be seen that the EMF Fig. 14. For the same motor parameters shown in Table I
calculated by Model 1 has the maximum error compared with except axial length, Ke at different radial radii of the motors
FEM result, while the EMF calculated by Model 2 is in good with different axial lengths is calculated and shown in Table II
agreement with FEM result. Meanwhile it is shown that by and Fig. 15. Due to the effective length Lef is different for
increasing the number of subregions, a more accurate result different motors, the Ke shown in Table II is under the
can be obtained through Model 3. It should be noted that assumption that Lef=L1. When the position becomes far from
Model 2 is only suitable for regular shape of coil region in the surface of PM, Ke decreases with the radial radius
polar coordinate, while Model 3 is more convenient in actual decreasing. Moreover, with the increasing axial length, the
motor with any irregular shape of coil region. axial end-effect becomes smaller and the Ke is getting closer to
100 percent. The radial component of the flux density along
the axial direction with the fringe effect of PM calculated by
2D analytical solution and 3D FEM with three different axial
lengths are shown in Fig. 16. It can be seen that the axial end-
effect is about 10mm inside the PM and about 5mm outside
the PM. The flux density in the end part calculated by 3D
method is either larger or smaller than that computed by 2D

0018-9464 (c) 2015 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/TMAG.2015.2481862, IEEE Transactions on Magnetics

> FOR CONFERENCE-RELATED PAPERS, REPLACE THIS LINE WITH YOUR SESSION NUMBER, E.G., AB-02 (DOUBLE-CLICK HERE) < 7

method depending on the radial radius. The magnetic field in 0.5

the end part decreases sharply. The larger the radius, the faster
the magnetic field decreases. When the axial length is larger

Radial component of flux density (T)

0.4
than 20mm, the magnetic flux densities in the middle part
calculated by 2D method and 3D method are coincident.
Hence, for the ironless BLDC motor used for flywheel with 0.3

relative smaller axial length, the axial end-effect should be

considered. 0.2
3D FEM 2D FEM
r=22mm r=22mm
TABLE II
r=22.5mm r=22.5mm
THE KE WITH DIFFERENT RADII AND AXIAL LENGTHS r=23mm r=23mm
Radial radius (mm) 0.1 r=23.5mm r=23.5mm
Axial length (mm)
r=24.5 r=24 r=23.5 r=23 r=22.5 r=22 r=24mm r=24mm
12 100.50 95.03 91.15 88.25 85.7 83.58 r=24.5mm r=24.5mm
20 102.26 98.72 95.32 93.95 92.27 91.69 0.0
-5 0 5 10 15 20 25 30 35 40 45
40 101.23 99.50 98.39 97.18 96.28 96.10
Axial distance (mm)
60 101.20 99.90 98.70 97.85 97.58 97.14
80 100.96 100.38 99.25 98.61 98.22 97.82 (b) 40mm
0.5
100 100.56 100.39 99.51 98.84 98.78 98.54

Radial component of flux density (T)

0.4

0.3
 0
3D FEM 2D FEM
0.2
 0 r=22mm r=22mm
r=22.5mm r=22.5mm
r=23mm r=23mm
0.1 r=23.5mm r=23.5mm
r=24mm r=24mm
Fig. 14 The 2D and 3D analysis models. r=24.5mm r=24.5mm
102
0.0
0 10 20 30 40 50 60 70 80 90 100
100
Axial distance (mm)
105
98 (c) 100mm
Fig. 16 The radial component of flux density in axial direction with different
Correction factor Ke (%)

100
96 axial length of PM: (a) L1=12mm, (b) L1=40mm, (c) L1=100mm.
95 It should be noted that the correction factor Ke also will
94

change when the pole number p and radius of PM Rm are

90
92
changed. Table III and Fig. 17 show the Ke when the Rm
85
90
becomes 40mm and other parameters are the same as the data
80
in Table I. Table IV and Fig. 18 show the Ke when only 2p
100
becomes 8 and other parameters are the same as the data in
88

80 22
60 22.5 86 Table I.
23
40
23.5 TABLE III
Axial length (mm) 20 24
84
THE KE WITH DIFFERENT RADII AND AXIAL LENGTHS WHEN RM =40mm
0 24.5 Radial radius (mm)
Radial radius (mm)
Fig. 15 The correction factor Ke with different radial radii and axial lengths. Axial length (mm)
r=39.5 r=39 r=38.5 r=38 r=37.5 r=37
0.5
L1=12mm 12 114.7 107.5 100.4 96.1 92.2 89.2
End part End part 20 108.96 105.2 100.96 98.45 95.21 93.24
Radial component of flux density (T)

40 104.34 102.15 100.76 99.66 97.93 96.97

0.4
60 102.95 102.18 100.52 99.2 98.65 97.73
80 101.63 101.5 100.53 99.91 99.28 98.8
0.3 100 101.35 101.43 100.4 99.63 99.17 98.68
TABLE IV
THE KE WITH DIFFERENT RADII AND AXIAL LENGTHS WHEN 2p=8
3D FEM 2D FEM
0.2
r=22mm
Radial radius (mm)
r=22mm Axial length (mm)
r=22.5mm r=22.5mm r=24.5 r=24 r=23.5 r=23 r=22.5 r=22
r=23mm r=23mm 12 117.92 106.46 100.92 96.48 91.48 88.66
0.1 r=23.5mm r=23.5mm 20 110.65 104.77 101.27 98.17 95.57 93.46
r=24mm r=24mm 40 105.28 103.1 100.52 98.62 97.82 97.08
r=24.5mm r=24.5mm
60 103.22 102.2 100.89 99.65 98.56 97.9
0.0
-4 -2 0 2 4 6 8 10 12 14 16 80 102.31 101.85 100.44 100.02 99.32 98.66
Axial distance (mm) 100 101.79 101.39 100.52 99.79 99.2 98.9
(a) 12mm

0018-9464 (c) 2015 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/TMAG.2015.2481862, IEEE Transactions on Magnetics

> FOR CONFERENCE-RELATED PAPERS, REPLACE THIS LINE WITH YOUR SESSION NUMBER, E.G., AB-02 (DOUBLE-CLICK HERE) < 8

115 110
Section AA
θco
Correction factor Ke (%)

110
θci
105 105
L2 L1 Rso
100
A A
95 Rsi
100
90

85
100 95
80 22
60 22.5
40 23 Fig. 19 Top view of a coil package and cross section of the coil package.
23.5 90
Axial length (mm) 20
0 24.5
24
Radial radius (mm)
Fig. 20 shows the coil region shape, and the actual slots are
Fig. 17 The Ke with different radial radii and axial lengths when Rm=40mm.
rectangular. The coils are assumed to be evenly distributed
and the winding size is also measured, as shown in Fig. 20.
115
Fig. 21 shows the EMF waveforms calculated by 2D FEM and
120
Model 3 with or without accounting for the axial end-effect
and end winding. The EMF calculated by Model 3 is in good
Correction factor Ke (mm)

115 110

110
agreement with the 2D FEM result. But they have a large
105
difference from the result which accounts for the axial end-
105
100
effect and end winding. The correctness of this result will be
95
shown in the next section.
100
90

85
100 95
80 22
60 22.5
40 23
23.5 90
Axial length (mm) 20 24
0 24.5 Radial radius (mm)
Fig. 18 The Ke with different radial radii and axial lengths when 2p=8.
To calculate the EMF more accurately, the axial end-effect
and the influence of end winding should be both considered.
For the influence of end winding, the effective length Lef is
always adopted. Fig 19 shows the shape and cross section of
coil package of ironless motor with non-overlapping winding.
Lef can be calculated by Fig. 20 The coil region shape and winding size in actual motor.
L  L2 6
Lef  1 (28) 2D analytical method
2 2D FEM
4 2D analytical method accouting
where L2 is the maximum length of the coil package. for end effect and end winding
Then the magnetic flux calculation can be corrected by the 2
coefficient Ke with accounting for axial end effect and end
EMF (V)

winding. The equation (26) can be rewritten as 0

s l
   K e  Rij , Lef  N sij Rij Lef  f n  Rij  cos np     d
cij / 2

 cij / 2 -2
1 1

1  i  l ,1  j  s
-4
(29)
and the EMF can be derived by equation (22). -6
0 60 120 180 240 300 360
Angle (elec. deg)
Fig. 21 The EMF waveforms of actual motor.

V. EXPERIMENT
In order to validate the effectiveness of the proposed
method, a prototype of ironless BLDC motor with parallel
magnetized PMs is manufactured and experimental
investigations are carried out. This choice of parallel
magnetization is motivated by the ease of implementation of
the magnets. The experimental setup is shown in Fig. 22. The

0018-9464 (c) 2015 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/TMAG.2015.2481862, IEEE Transactions on Magnetics

> FOR CONFERENCE-RELATED PAPERS, REPLACE THIS LINE WITH YOUR SESSION NUMBER, E.G., AB-02 (DOUBLE-CLICK HERE) < 9

no-load EMF waveform is obtained at speed of 3018rpm and China under Project 20121101120024 ， in part by Basic
is compared with the analytical result, as shown in Fig.23. The Research Foundation of Beijing Institute of Technology under
difference between the measured and calculated is about 3.4%, Grant 20110642015 and 20120642013 and 20130642105, and
which may be caused by the position error of the coils. in part by Excellent Young Scholars Research Fund of Beijing
Institute of Technology.

REFERENCES
[1] M. C. Chou, C. M. Liao, S. B. Chien, F. H. Shien, J. R. Tsai, and H. C.
Chang, "Robust current and torque controls for PMSM driven satellite
reaction wheel," IEEE Trans. Aero. Elec. Sys., vol. 47, no. 1, pp. 58-74,
Jan. 2011.
[2] X. Zhou and J. C. Fang, "Precise braking torque control for attitude
control flywheel with small inductance brushless DC Motor," IEEE
Trans. Power Syst., vol. 28, no. 11, pp. 5380-5390, Nov. 2013.
[3] R. P. Praveen, M. H. Ravichandran, V. T. S. Achari, V. P. J. Raj, G.
Madhu, and G. R. Bindu, "A novel slotless Halbach-array permanent-
magnet brushless DC motor for spacecraft applications," IEEE Trans.
Ind. Electron., vol. 59, no. 9, pp. 3553-3560, Sep. 2012.
[4] F. Caricchi, F. Crescimbini, F. Mezzetti, and E. Santini, "Multistage
Fig. 22 Experimental test system. axial-flux PM machine for wheel direct drive," IEEE Trans. Ind. Appl.,
6
vol. 32, no. 4, pp. 882-888, Jul./Aug. 1996.
Experimental result
[5] R. J. Wang, M. J. Kamper, K. V. Westhuizen, and J. F. Gieras, "Optimal
2D analytical method accouting
4
for end effect and end winding design of a coreless stator axial flux permanent-magnet generator," IEEE
Trans. Magn., vol. 41, no. 1, pp. 55–64, Jan. 2005.
2
[6] E. Spooner, P. Gordon, J.R. Bumby and C.D. French, "Lightweight
ironless-stator PM generators for direct-drive wind turbines," IEE Proc.
Electr. Power. Appl., vol. 152, no. 1, pp. 17-26, Jan. 2005.
EMF (V)

0 [7] I. Stamenkovic, N. Milivojevic, N. Schofield, M. Krishnamurthy, and A.

Emadi, "Design, analysis, and optimization of ironless stator permanent
-2 magnet machines," IEEE Trans. Power Electron., vol. 28, no. 5, pp.
2527-2538, May 2013.
[8] U. K. Madawala and J. T. Boys, "Magnetic field analysis of an ironless
-4
brushless DC machine," IEEE Trans. Magn., vol. 41, no. 8, pp. 2384-
2390, Aug. 2005.
-6 [9] M. Ooshima, S. Kitazawa, A. Chiba, T. Fukao, and D. G. Dorrell,
0 60 120 180 240 300 360
"Design and analyses of a coreless-stator-type bearingless
Angle (elec. deg)
motor/generator for clean energy generation and storage systems," IEEE
Fig. 23 The comparison of EMF between experiment result and analytical Trans. Magn., vol. 42, no. 10, pp. 3461-3463, Oct. 2006.
result. [10] P. Zheng, J. Zhao, R. R. Liu, C. D. Tong, and Q. Wu, "Magnetic
characteristics investigation of an axial-axial flux compound-structure
VI. CONCLUSION PMSM used for HEVs," IEEE Trans. Magn., vol. 46, no. 6, pp. 2191-
2194, Jun. 2010.
In this paper, an analytical method for the computation of [11] P. Rasilo, A. Belahcen, and A. Arkkio, "Importance of iron-Loss
the magnetic field in an ironless BLDC motor with external modeling in simulation of wound-Field synchronous machines," IEEE
Trans. Magn., vol. 48, no. 9, pp. 2495-2504, Sep. 2012.
rotor is presented. In the air region, the magnitude and [12] L. Y. Li, D. G. Pan, and X. Z. Huang, "Analysis and optimization of
direction of the magnetic field changes largely, which makes ironless permanent-magnet linear motor for improving thrust," IEEE
the magnetic flux linked with each coil to be different and the Trans. Plasma Sci., vol. 41, no. 5, pp. 1188-1192, May. 2013.
[13] H. Kim, Y. You, and B. Kwin, "Rotor shape optimization of interior
calculation of EMF induced in the winding more complex. permanent magnet BLDC motor according to magnetization direction,"
Three different models are proposed to calculate the EMF. IEEE Trans. Magn., vol. 49, no. 5, pp. 2193-2196, May. 2013.
The results obtained by the three models are compared with [14] A. Belahcen, P. Rasilo, and A. Arkkio, "Segregation of iron losses from
rotational field measurements and application to electrical machine,"
those obtained by 2D FEM. Then the axial end-effect is also IEEE Trans. Magn., vol. 50, no. 2, pp. 7022104, Feb. 2014.
considered and the EMF accounting for this effect is [15] H. Z. Hu, J. Zhao, X. D. Liu, Z. Chen, Z. X. Gu, and Y. Sui, " Research
recalculated. Finally, a BLDC motor prototype was on the torque ripple and scanning range of an arc-structure PMSM used
for scanning system," IEEE Trans. Magn., vol. 50, no. 11, pp. 8105504,
manufactured, and experiment is performed to validate the Nov. 2014.
effectiveness of the proposed method. The results indicate that [16] B. L. J. Gysen, K. J. Meessen, J. J. H. Paulides, and E. A. Lomonova,
the end-effect and the influence of end winding should not be "General formulation of the electromagnetic field distribution in
machines and devices using fourier analysis," IEEE Trans. Magn., vol.
ignored for the ironless BLDC motor with short axial length. 46, no. 1, pp. 39-52, Jan. 2010.
[17] Z. Q. Zhu, L. J. Wu, and Z. P. Xia, "An accurate subdomain model for
ACKNOWLEDGMENT magnetic field computation in slotted surface-mounted permanent-
magnet machines," IEEE Trans. Magn., vol. 46, no. 4, pp. 1828–1837,
This work was supported in part by the National Natural Apr. 2010.
Science Foundation of China under Project 51307008，in part [18] A. Rahideh and T. Korakianitis, "Analytical calculation of open-circuit
magnetic field distribution of slotless brushless PM machines," Electr.
by the Key laboratory for Intelligent Control & Decision of Power Energy Syst., vol. 44, no. 1, pp. 99-114, 2013.
Complex Systems of Beijing Institute of Technology, in part [19] J. H. Choi, S. M. Jang, H. I. Park, K.H. Kim "Characteristic analysis on
by Ph.D. Programs Foundation of Ministry of Education of synchronous machine with double-side permanent magnet rotor for

0018-9464 (c) 2015 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/TMAG.2015.2481862, IEEE Transactions on Magnetics

> FOR CONFERENCE-RELATED PAPERS, REPLACE THIS LINE WITH YOUR SESSION NUMBER, E.G., AB-02 (DOUBLE-CLICK HERE) < 10

flywheel energy storage system in EV," in IEEE Vehicle Power and

Propulsion Conf., 2012, pp. 1223-1227.
[20] Y. K. Huang , B. Y. Ge , J. N. Dong , H. Y. Lin, J. G. Zhu, and Y. G.
Guo, "3-D analytical modeling of no-load magnetic field of ironless
axial flux permanent magnet machine," IEEE Trans. Magn., vol. 48, no.
11, pp. 2929-2932, Nov. 2012.
[21] P. Virtic, P. Pisek, T. Marcic, M. Hadziselimovic, and B. Stumberger,
"Analytical analysis of magnetic field and back electromotive force
calculation of an axial-flux permanent magnet synchronous generator
with coreless stator," IEEE Trans. Magn., vol. 44, no. 11, pp. 4333-4336,
Nov. 2008.
[22] Z. Q. Zhu, D. Howe, and C. C. Chan. "Improved analytical model for
predicting the magnetic field distribution in brushless permanent-magnet
machines," IEEE Trans. Magn., vol. 38, no. 1, pp. 229-238, Jan. 2002.
[23] S. R. Holm, H. Polinder, and J. A. Ferreira, "Analytical modeling of a
permanent-magnet synchronous machine in a flywheel." IEEE Trans.
Magn., vol. 43, no. 5, pp. 1955–1967, May 2007.
[24] T. J. E. Miller and R. Rabinovici, “Back-EMF waveforms and core
losses in brushless DC motors,” Proc. Inst. Elect. Eng. Elect., vol. 141,
no. 3, pp. 144–154, May 1994.
[25] M. Markovic and Y, Perriard, "An analytical formula for the back emf
of a slotted BLDC motor," IEEE IEMDC Conf., Vol. 2, pp. 1534-1539,
May 2007.
[26] C. Bi, Z. J. Liu, and S. X. Chen, "Estimation of back-EMF of PM BLDC
motors using derivative of FE solutions," IEEE Trans. Magn., vol. 36,
no. 4, pp. 697–700, July 2000.
[27] A. Rahideh, M. Mardaneh, and T. Korakianitis, "Analytical 2-D
calculations of torque, inductance, and back-EMF for brushless slotless
machines with surface inset magnets," IEEE Trans. Magn., vol. 49, no. 8,
pp. 4873–4884, Aug. 2013.
[28] V. Simon-Sempere, M. Burgos-Payan, and J. Cerquides-Bueno,
"Influence of manufacturing tolerances on the electromagnetic force in
permanent-magnet motors," IEEE Trans. Magn., vol. 49, no. 11, pp.
5522–5532, Nov. 2013.
[29] K. F. Rasmussen, "Analytical prediction of the magnetic field from the
surface mounted permanent magnet motor," in Proc. IEEE IEMDC,
Seattle, WA, May, 1999.

0018-9464 (c) 2015 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.