You are on page 1of 7

Abstract -- This paper compares the torque densities, i.e.

torque per volume (T/V) and torque per weight (T/G), between
permanent magnet (PM) and electrically excited (EE)
machines for low speed applications when they have the same
overall size and copper loss. The machines are optimized
individually and analytically. The optimal split ratio and flux
density ratio are derived. The influence of pole number and
machine size is investigated as well. The analytical models are
verified by finite element (FE) analyses. It shows that PM
machines exhibit more than 2 times torque densities of EE
machines. For EE machines, there is an optimal pole number
to maximize torque densities and they prefer large volume
applications. The comparison also shows that the T/G designs
have significantly higher split ratio than the T/V designs to
balance torque and weight.

Index TermsAnalytical, optimization, comparison, split
ratio, torque density ratio, permanent magnet machines,
electrically excited machines.

I. NOMENCLATURE
A
a
, A
f
Stator slot and field winding areas
B
m
Maximum flux density in the lamination
B

Average airgap flux density over pole arc


D
o
, D

, D
r
Outer, airgap, and rotor diameters
D
ry
Diameter at outer surface of rotor yoke
G Total active material weight
G
a
, G
sc
, G
r
Active material weights of armature
winding, stator core and rotor
h
o
Tooth tip height
h
ps
Pole shoe height of EE machines
k
a
, k
f
Armature and field winding packing factors
L
ef
Effective axial length
L
a
, L
f
Armature and field end winding lengths
N
a
I
a
, N
f
I
f
Armature and field magnetomotive forces
p Number of pole pairs
p
cu
, p
cua
, p
cuf
Total, armature and excitation copper losses

p
Pole arc to pole pitch ratio
Flux density ratio
, ' Mechanical and equivalent airgap lengths

,
r
Airgap and rotor split ratios
Conductor resistivity

s
,
r
Stator and rotor flux leakage coefficients

0
Permeability in the air
Average relative permeability of cores
II. INTRODUCTION
LECTRICALLY excited (EE) and permanent magnet
(PM) machines are two most widely employed
electrical machines in various applications, such as traction
vehicles, renewable energies, and industrial automations.

W. Q. Chu, Z. Q. Zhu, and J.T. Chen are with Department of
Electronic and Electrical Engineering, University of Sheffield, Sheffield
S1 3JD, U.K. (W.Q.Chu@sheffield.ac.uk, Z.Q.Zhu@sheffield.ac.uk,
xjtaha@hotmail.com).
Some comparative studies have been made between EE
and PM machines. In [1] and [2], the comparison is on the
system performance for wind power applications. In [3],
they are compared in terms of system mass and efficiency
for the man-portable power system using the genetic
algorithm (GA). In this paper, their torque densities will be
quantitatively compared when they have the same overall
size and copper loss. For fair comparison, it is necessary to
optimize the machines individually.
The optimizations of EE machines have been discussed
in [4]-[6]. In [4], the new concept of stator-output
coefficient between torque and stator volume is proposed to
analyze synchronous machines. Optimal airgap flux density
and slot width are derived. The influence of voltage, slot
depth, number of poles and airgap diameter is investigated
as well. In [5], several sizing equations for electrical
machines are developed. One is
o
3
I
c]
sizing equation,
which relates the output torque to outer diameter and axial
length. The other two sizing equations,
o
2.5
I
c]
and

2.5
I
c]
, are developed by combining
o
3
I
c]
sizing
equation with the conventional

2
I
c]
sizing equation. It
shows that the output torque is a function of airgap flux
density, armature current density, split ratio and flux
density ratio. The optimal split ratio has been derived
analytically while the influence of flux density ratio has
been investigated as well. However, the output torque
coefficients in [4] and [5] have some limits. Firstly, it is less
meaningful when they are not related with the machine
volume,
o
2
I
c]
. Secondly, the torque coefficients are
developed based on the given airgap flux density, which
will be shown later in this paper that it is inappropriate for
EE machines. Thirdly, the optimal quantities are derived by
optimizing the stator only. However, the optimal quantities
may be different when both stator and rotor are considered,
especially for EE machines. Furthermore, the pole number
is fixed in [4] and [5] but it is an important design
parameter to be investigated. In [6], the EE machine is
optimized in terms of weight based on GA. However, GA is
less insightful than the analytical methods.
The PM machine optimizations in terms of torque per
volume are intensively addressed in [7]-[12] based on
parallel-side stator teeth and full-pitch magnets. The
optimal split ratio is analytically derived in [7], [8] and [10].
In [10], the optimal airgap flux density is also developed.
This paper is devoted to the comparisons between EE
and PM machines in terms of two torque densities, T/V and
T/G, for low speed applications when they have the same
machine size, copper loss and stator configuration. Both
machines are optimized analytically and individually. The
optimal split ratio and flux density ratio are derived
analytically as well. The influence of pole number and
Analytical Optimization and Comparison of
Torque Densities between Permanent Magnet
and Electrically Excited Machines
W. Q. Chu, Z. Q. Zhu, J. T. Chen
E
978-1-4673-0141-1/12/$26.00 2012 IEEE
1190

machine size is also investigated. The analytical models are
verified by finite element (FE) analyses. The quantitative
and comprehensive comparisons are made between T/V and
T/G designs as well to aid machine designs.
III. MACHINE FEATURES AND BASIC EQUATIONS
For fair comparison, EE and PM machines have the
same machine size, copper loss, and stator configuration,
which has full-pitch distributed overlapping armature
windings and rectangular semi-closed stator slots. For EE
machines, the field windings have rectangular cross-
sections, as illustrated in Fig. 1. The other parameters are
detailed in Table I.
Two sets of dimensions, D1 and D2, will be investigated
to study the influence of machine size. Since the armature
winding in a large machine is usually pre-formed, its
packing factor is set as high as the field windings, which
are pre-formed as well. The flux leakage coefficients are
higher in T/V designs and large machines due to the deeper
slots. B
m
is relatively low since it is determined by the open
circuit field.

Fig. 1. Cross-section of EE machines.
TABLE I
SUMMARY OF MAIN PARAMETERS
Profile dimension set D1 D2
Pole arc to pole pitch ratio p 0.75
Maximum B in the laminations Bm 1.4T
Conductor resistivity 1.724*10
-8
m
Pole shoe height of EE machines hps 0.1
Outer diameter Do 420mm 2500mm
Active axial length Lef 110mm 1100mm
Airgap length 2mm 4mm
Total copper loss pcu 300W 25kW
Armature winding packing factor ka 0.45 0.7
Field winding packing factor kf 0.7 0.7
Tooth tip height ho 4mm 8mm
For T/V designs
Stator flux leakage coefficient s 1.08 1.35
Rotor flux leakage coefficient r 1.08 1.3
For T/G designs
Stator flux leakage coefficient s 1.08 1.15
Rotor flux leakage coefficient r 1.05 1.1
Assuming a rectangular open-circuit airgap flux density
distribution, as illustrated in Fig. 2, the average torque can
given as:
I = Sno
p
I
c]

6
B
6
N
u
I
u
(22) (1)
The armature copper loss is given by:
p
cuu
= S6(N
u
I
u
)
2
p(I
c]
+ I
u
)(k
u
A
u
) (2)
According to the geometric and magnetic relationships,
the total stator slot area is given by:
A
u
=
n
o
2
z
6
2
(1 - o
s
o
p
y)(1 - [
s
-z
6
-
no
s
o
p
yz
6
2p
) (3)
y = B
6
B
m
(4)
z
6
=
6

o
(5)
[
s
= 2b
o

o
(6)

Fig. 2. Iillustration of ideal open-circuit airgap flux density distribution.
For PM machines, their rotor is copper loss free and able
to achieve wider range of B

by using surface-mounted PM
(SPM) or interior PM (IPM) configurations. Hence, its
torque can simply depend on the stator and is given as:
I =
Sn
22
o
p
I
c]

o
z
6
B
m
y_
p
cuu
A
u
k
u
S6p(I
c]
+I
u
)
(7)
Different from PM machines, the open-circuit airgap
flux density in EE machines is produced by the field
winding, which has copper loss and requires more space.
The relationship between excitation copper loss and
excitation MMF is given by:
p
cu]
= 4(N
]
I
]
)
2
p(I
c]
+ I
]
)(k
]
A
]
)
(8)
Another difference between PM and EE machines is the
influence of MMF drops in the stator and rotor. For PM
machines, firstly, the equivalent airgap is much bigger than
the mechanical one, since the magnet is thick and its
permeability is close to the air. Secondly, the magnets are
small and very easy to be accommodated. Hence, the
influence of the MMF drops in the stator and rotor is low
and can be easily compensated by using slightly more
magnet, although it is also important to consider the cost.
However, for EE machines, the influence of the MMF
drops in the stator and rotor is much higher due to the
smaller equivalent airgap. Meanwhile, the field winding
area is proportional to the square of excitation MMF.
Hence, the MMF drops in the stator and rotor will result in
a significant increase of field winding area and hence affect
the machine design. Therefore, for EE machines, it is
important and necessary to consider the MMF drops in the
stator and rotor.
In the actual field, the MMF drops in stator and rotor are
complicated. However, a simplified method is suggested
here to approximate the influence by using an average
relative permeability and equivalent airgap length:
o
i
= o + (1 - z
6
z

)
o
p (9)
z

6

(10)
Therefore, B

and torque of EE machines are given as:


B
6
=
p
0
N
]
I
]
2po'
=
p
0
2po'
_
p
cu]
A
]
k
]
4p(I
c]
+ I
]
)
(11)
I =
n
162
o
p

o
z
6
ppo'
_
k
u
k
]
I
c]
2
p
cuu
A
u
p
cu]
A
]
(I
c]
+I
u
)(I
c]
+ I
]
)

(12)
1191

The field winding area can be obtained based on
geometric and magnetic relationships as:
A
]
= n(
o
z
6
)
2
(z

-o

o
p
y)(1 - [

- z

)2
(13)
[

= 2(o + b
ps
)(
o
z
6
) (14)
IV. COPPER LOSS OPTIMIZATION OF EE MACHINES
As can be seen in (12), the torque of EE machines is the
function of both armature and excitation copper losses,
which should be optimized first to maximize the torque.
The total copper loss p
cu
is fixed due to the efficient and
cooling requirements. The output torque is rewritten as:
I =
no
p
162

o
z
6
ppo'
_
k
u
k
]
I
c]
2
p
cuu
A
u
(p
cu
- p
cuu
)A
]
(I
c]
+ I
u
)(I
c]
+I
]
)

(15)
It can be concluded that the torque of EE machines
peaks when p
cua
=p
cuf
=p
cu
/2.
According to (7), since PM rotors are copper loss free,
they can achieve much higher torque when having the same
total copper loss as EE machines. However, in order to
emphasize the differences due to the rotor, all optimal
designs in this paper are obtained when they have the same
armature copper loss as p
cua
=p
cu
/2.
V. MAXIMUM TORQUE PER VOLUME (T/V)
For most applications, higher torque is desired when the
machine size is fixed. Hence, EE and PM machines are
optimized and compared first in term of torque per volume.
According to (3), (7), (12), and (13), the torque is the
function of

, , and 2p. However, for EE machines, is


limited by

and 2p. Hence, two analytical models are


developed to aid the optimizations:
(a) Given B

model: the model is developed by


assuming that B

is given and can be arbitrary. The optimal


split ratio
op
, optimal flux density ratio
op
and hence
maximum torque are derived.
(b) Max. B

model: The EE rotor is optimized for


maximum flux density ratio
m
. Based on this,

is
optimized to maximize torque.
For PM machines, their performance can be obtained
based on Given B

model. However, for EE machines,


there are two cases: (a) when
op

m
, the peak torque is
predicted by Given B

model. (b) when


op
>
m
, the peak
torque is given by Max. B

model.
A. Given B

model
According to (3) and (7), the torque can be derived as:
I
2
=
n
3
S2
(
o
2
B
m
o
p
I
c]
)
2
p
cuu
k
u
p(I
c]
+I
u
)
z
6
3
y
2

(1 -o
s
o
p
y)(
1 - [
s
- z
6
2
-
no
s
o
p
yz
6
4p
)
(16)
The torque variation with split ratio and flux density
ratio is shown in Fig. 3. It can be seen that there is a
combination of optimal split ratio and flux density ratio to
maximize the torque. Assuming is given,
op
is derived
by letting T/

=0 as:
z
6op
=
S
4
1 - [
s
|1 + no
s
o
p
y(2p)]
(17)
Since
s
,
s
and
p
are constant,
op
simply depends on
and 2p. As shown in Fig. 4,
op
is lower due to the thicker
stator yoke when either is higher or 2p is lower. However,
when 2p is higher, the influence of on
op
is lower.

Fig. 3. Torque variation based on Given B when 2p=20 and in D1.

Fig. 4. Variations of op based on Given B model when size is D1.
In the similar way, by letting T/=0,
op
can be
obtained as:
y
op
=
1
8o
s
o
p
nz
6
_
S|2p(1 -[
s
- z
6
) +nz
6
] -
_
9|2p(1 - [
s
- z
6
) + nz
6
]
2
-64pnz
6
(1 - [
s
-z
6
)
_ (18)
The influence of

and 2p on
op
is illustrated in Fig. 5.
Either lower 2p or higher

will result in larger


op
. It also
shows that for optimal designs, whose

<0.75, their
op
is
high and varies modestly.

Fig. 5. Variations of op based on Given B model when size is D1.
The global optimal designs can be obtained by letting
both T/

=0 and T/=0:
z
6op
=
(1 - [
s
)(6p +n)
4(2p + n)
(19)
0.5
0.6
0.7
0.
0.5
0.6
0.7
0.8
0.9
1
160
180
200
220
240
S
p
lit ratio
F
l
u
x
d
e
n
s
it
y
r
a
t
io
T
o
r
q
u
e

(
N
m
)
0.50
0.55
0.60
0.65
0.70
0.75
0.5 0.6 0.7 0.8 0.9 1
O
p
t
i
m
a
l

s
p
l
i
t

r
a
t
i
o
Flux density ratio
2p=8 2p=10
2p=12 2p=14
2p=16
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.4 0.5 0.6 0.7 0.8 0.9
O
p
t
i
m
a
l

f
l
u
x

d
e
n
s
i
t
y

r
a
t
i
o
Split ratio
2p=8 2p=10
2p=12 2p=14
2p=16
1192

y
op
=
4p
o
p
(6p +n)
(20)
The performance of global optimal designs is
represented by the PM designs in Figs. 7-12. It can be seen
that the optimal split ratio, flux density ratio and hence
torque increase when the number of poles is higher, since
the stator yoke is thinner. Due to the assumption of constant
flux leakage coefficients, which is for the sake of
understanding and simplicity, the predicted torque increases
monotonously with the pole number. Actually, the flux
leakages and hence flux leakage constants will increase due
to higher inter-pole leakage and deeper stator slot when the
pole number is higher. Hence, the torque will be lower.
However, updating the leakage constants as suggested in
Table I, analytical models are still applicable.
However, for EE machines, their performance depends
on
m
, which will be derived later.
B. Max. B

model
To achieve higher , the EE rotor needs to be optimized.
According to (13), it can be seen that A
f
depends on
r
when
and

are given. The optimal rotor split ratio


rop
to
maximize A
f
can be obtained as:
z
op
= (1 - [

+ o

o
p
y)2
(21)
A
]m
= n
6
2
(1 - [

-o

o
p
y)8 (22)
Combined with (11),
m
can be given as:
y
m
=
(1 - [

)
o + o

o
p
(23)
o =
2po'B
m
p
0

o
z
6
_
S2p(I
c]
+I
]
)
np
cu]
k
]
(24)
For the T/V designs, the machine space is fully utilized
to achieve higher torque. The equivalent airgap length
can be approximated as:
o

= _
o + (1 - u.4)
o
p (small machines)
o + (1 - u.S)
o
p (large machines)
(25)
In Fig. 6, it shows that
m
increases when

is larger,
since more rotor space is available, while when 2p
increases,
m
decreases, since B

is inverse proportional to
2p as shown in (11). Therefore, lower 2p is preferred.
However, lower 2p results in longer end windings, thicker
yokes, and hence lower MMFs. Therefore, there will be an
optimal 2p to maximize the torque.

Fig. 6. Variation of m based on Max. B model when size is D1
Based on (23), the torque can be given as:
I
2
=
n
3
64
(
o
2
B
m
I
c]
)
2
p
cuu
k
u
z
6
3
p(I
c]
+ I
u
)
_
(1 - [

)z
6
b + k
c
z
6
_
2

_1 -
(1 - [

)z
6
b + k
c
z
6
_ _1 - [
s
- z
6
-
n
2p
(1 - [

)z
6
2
b + k
c
z
6
_
(26)
b =
oz
6
o
s
o
p
=
2po'B
m
o
s
p
0

o
o
p
_
S2p(I
c]
+ I
]
)
np
cu]
k
]
(27)
Since k

1,
r
<<1 and

<1, hence, (k

-1+
r
)

0 and
(26) can be simplified as:
I
2

z
6
5
(b + k
c
z
6
)
3
_1 -[
s
- z
6
-
n
2p
(1 - [

)z
6
2
b + k
c
z
6
_ (28)
Mathematically, three solutions can be derived from (28)
However, only one is the actual
op
, which is a real number
and between 0-1.
C. Comparisons
The performance of optimal T/V EE and PM designs is
summarized and compared in Figs. 7-12. It shows that
op

increases when 2p increases for both machines, since the
stator yoke is thinner. However, EE machines have higher

op
due to lower . Their decreases when 2p increases
while PM machines can maintain
op
. Consequently, EE
machines have lower torque even when they have the same
armature copper loss as PM machines. When having the
same copper losses, the armature copper loss of PM
machine is doubled and hence their torque can be increased
by 2 times.
It also shows that the pole number influences PM and
EE machines in significantly different ways. For PM
machines, their torque increases monotonically with 2p.
However, for EE machines, the torque increases first and
then decreases. It is the compromise between the influence
of yokes, end windings and airgap flux density. Hence,
there is an optimal 2p for EE machines to maximize torque.
The comparison between the small size machines D1
and large size machines D2 shows that the differences
between EE and PM machines based on the large machine
D2 are less, which suggests that EE machines prefer large
volume applications.

Fig. 7. Variation of with 2p in D1 for max. T/V.
0
0.1
0.2
0.3
0.4
0.5
0.5 0.6 0.7 0.8
M
a
x
i
m
u
m

f
l
u
x

d
e
n
s
i
t
y

r
a
t
i
o
Split ratio
2p=8 2p=10
2p=12 2p=14
2p=16
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
6 8 10 12 14 16 18 20 22 24
F
l
u
x

d
e
n
s
i
t
y

r
a
t
i
o
Number of poles
PM
EE
1193


Fig. 8. Variation of op with 2p in D1 for max. T/V.

Fig. 9. Variation of torque with 2p in D1 for max. T/V.

Fig. 10. Variation of with 2p in D2 for max. T/V.

Fig. 11. Variation of op with 2p in D2 for max. T/V.

Fig. 12. Variation of torque with 2p in D2 for max. T/V.
VI. MAXIMUM TORQUE PER WEIGHT (T/G)
The foregoing optimizations are aimed for maximization
of T/V. Generally, higher torque results in heavier
machines, although the machine volume is fixed. However,
for some applications, such as wind power generation, since
the mass influences system performance significantly, they
prefer higher T/G. Therefore, it is also important to discuss
the optimization in terms of T/G.
A. Analytical model
The developed Given B

model is adequate for the


investigation, since T/G is meaningless when the rotor is
unable to provide required B

. Only the weight of active


materials, i.e. copper, PM, stator and rotor cores, are
considered.
B. Optimization
According to (7), when B

is given, the relationship


between torque and weight can be given as:
I
0

z
6
A
u
0
u
+ 0
sc
+ 0

=
z
6
k
uu
A
u
+ (0
sc
+ 0

)A
u
(29)
As will be shown later, the variation of

is modest and
assumed as constant. Hence, the simplified design principle
can be obtained by letting (T/G)/A
a
=0 as:
0
uop
= k
uu
A
uop
= (0
sc
+0

) (30)
In other words, the T/G peaks when the armature
winding weight equals with the total weight of other parts.
C. Results and comparisons
The optimal T/G designs of PM machines are shown in
Fig. 13. T/G and
op
increase when the pole number is
higher. When is higher, the rotor and stator laminations
are heavier. It needs more armature winding area to balance
the weight. Hence,
op
is lower. T/G peaks when =0.5,
since further increase of results in much thicker magnet
and hence lower T/G.
The optimal T/G designs of EE machines are shown in
Fig. 14 and significantly different from PM machines. For
EE machines, when the number of poles increases, their
yokes and end windings decrease while the field winding
area increases. G
sc
+G
r
decreases firstly and then increases.
Hence, T/G increases firstly and then decreases. There will
be an optimal pole number to maximize T/G. When the pole
number is low,
op
increases modestly. The higher is, the
lower
op
is. However, when the pole number increases
further,
op
is forced to increase sharply, since the field
0.5
0.6
0.7
0.8
0.9
6 8 10 12 14 16 18 20 22 24
O
p
t
i
m
a
l

s
p
l
i
t

r
a
t
i
o
Number of poles
EE
PM
50
100
150
200
250
300
6 8 10 12 14 16 18 20 22 24
T
o
r
q
u
e

(
N
m
)
Number of poles
PM_analytical
PM_FE
EE_analytical
EE_FE
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
12 14 16 18 20 22 24 26 28 30 32 34 36 38 40
F
l
u
x

d
e
n
s
i
t
y

r
a
t
i
o
Number of poles
PM EE
0.5
0.6
0.7
0.8
0.9
12 14 16 18 20 22 24 26 28 30 32 34 36 38 40
O
p
t
i
m
a
l

s
p
l
i
t

r
a
t
i
o
Number of poles
PM EE
200
225
250
275
300
325
350
12 14 16 18 20 22 24 26 28 30 32 34 36 38 40
T
o
r
q
u
e

(
k
N
m
)
Number of poles
PM_analytical
PM_FE
EE_analytical
EE_FE
1194

winding requires more space. Consequently, T/G decreases
rapidly. When is higher, the turning happens at lower pole
number. The higher also results in lower maximum T/G.
The optimal T/G designs of PM and EE machines are
compared in Figs. 15-16. It shows that PM machines have
higher T/G than EE machines even when their armature
copper losses are the same. It can also be seen that the
differences between PM and EE machines in D2 are less
than the ones in D1.

(a) Torque per weight T/G

(b) Optimal split ratio
Fig. 13. Optimal T/G designs of SPM machines when size is D1.

(a) Torque per weight T/G

(b) Optimal split ratio
Fig. 14. Optimal T/G designs of EE machines when size is D1.

Fig. 15. Comparison of optimal T/G designs when =0.3 and size is D1.

Fig. 16. Comparison of optimal T/G designs when =0.6 and size is D2.
VII. MAX. T/V VERSUS T/G
The optimal designs in terms of two torque densities,
T/V and T/G, are compared in Fig. 17 and Table II. It can be
seen that the different criterions have led to the entirely
different design principles and outcomes.
For T/V designs, to achieve higher torque, they make use
of the most space, while for T/G designs, they are the
optimal balances between torque and weight. The most
remarkable difference is that
op
for T/G is more than 20%
higher than the one for T/V, as illustrated in Fig. 17.
For example, four designs are detailed and compared in
Table II, the torques of T/V designs are 20% and 16%
higher than the T/G designs for PM and EE machines,
respectively, while the T/V designs are 77% and 74%
heavier for PM and EE machines, respectively. In total, the
T/G designs of PM and EE machines exhibit 47% and 45%
higher torque per weight, respectively.
It suggests that the T/V designs are heavier and more
expensive than T/G designs as well. Although T/V designs
exhibit higher torque, their torque per weight and torque per
cost are significantly lower than T/G designs. Therefore, for
the weight- and cost-sensitive applications, such as wind
power generation, the T/G designs are more suitable.
Furthermore, the power factors of T/V designs are also
lower than T/G designs, since the slot leakage inductance
increases due to deeper slots. This makes T/G designs even
more promising than T/V designs.
It also shows that higher flux density ratio is demanded
for T/V designs than T/G designs. As shown in Fig. 13, PM
machines have maximum T/G when =0.5, which is lower
than
op
for T/V designs given in (20), Fig. 7 and Fig. 10.
For EE machines, it can be seen from Fig. 14 that their
optimal flux density ratio for T/G designs is even lower.
2.0
2.5
3.0
3.5
4.0
4.5
12 14 16 18 20 22 24
T
o
r
q
u
e

p
e
r

w
e
i
g
h
t

(
N
m
/
k
g
)
Number of poles
=0.4
=0.5
=0.6_analytical
=0.6_FE
=0.7
0.6
0.7
0.8
0.9
1.0
12 14 16 18 20 22 24
O
p
t
i
m
a
l

s
p
i
l
i
t

r
a
t
i
o
Number of poles
=0.4
=0.5
=0.6
=0.7
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
6 8 10 12 14 16 18 20 22 24
T
o
r
q
u
e

p
e
r

w
e
i
g
h
t

(
N
m
/
k
g
)
Number of poles
=0.2
=0.25
=0.3_analytical
=0.3_FE
=0.35
0.6
0.7
0.8
0.9
1.0
6 8 10 12 14 16 18 20 22 24
O
p
t
i
m
a
l

s
p
l
i
t

r
a
t
i
o
Number of poles
=0.2
=0.25
=0.3
=0.35
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
6 8 10 12 14 16 18 20 22 24
T
o
r
q
u
e

p
e
r

w
e
i
g
h
t

(
N
m
/
k
g
)
Number of poles
EE
PM
0
5
10
15
20
25
16 18 20 22 24 26 28 30 32 34 36
T
o
r
q
u
e

p
e
r

w
e
i
g
h
t

(
N
m
/
k
g
)
Number of poles
EE_analytical
EE_FE
PM_analytical
PM_FE
1195


Fig. 17. Comparison of optimal split ratios for T/V and T/G designs when
=0.5 and size is D2.
TABLE II
COMPARISON OF FOUR DESIGNS (SAME ARMATURE COPPER LOSS)
Design PM_T/V PM_T/G EE_T/V EE_T/G
Machine size D2
Flux density ratio 0.5
Number of poles 24
Torque (kNm) 288.30 239.05 288.30 247.15
Split ratio 0.698 0.867 0.698 0.860
Total weight (Ton) 25.22 14.19 29.33 16.82
T/G (Nm/kg) 11.42 16.84 9.82 14.68
Total cost (k) 106.77 57.10 111.27 57.54
T/cost (Nm/) 2.70 4.18 2.59 4.29
Power factor 0.528 0.704 0.511 0.611
Material price (/kg) Copper: 8.5, PM: 40, Lamination: 1.25

(a) PM_T/V (b) PM_T/G

(c) EE_T/V (d) EE_T/G
Fig. 18. Cross-sections of four designs in Table II.
VIII. FE VERIFICATION
2D non-linear FE analyses have been carried out by
using the OPERA FE software to verify the analytical
models. As can be seen from Figs. 9, 12, 13 and 14, the
analytical and FE predicted torques agree well.
IX. CONCLUSIONS
In this paper, the optimizations and comparisons
between EE and PM machines in terms of T/V and T/G
have been discussed. Optimal split ratio and flux density
ratio have been derived analytically. The influence of pole
number and machine size has been discussed as well to aid
the machine designs.
The study shows that PM machines can have more than
2 times torque densities of EE machines when they have
the same copper loss. The T/G designs have significantly
higher split ratio than the T/V designs. For EE machines,
they have an optimal number of poles to maximize the
performance and prefer large volume applications.
X. REFERENCES
[1] N. Bianchi and A. Lorenzoni, Permanent magnet generators for
wind power industry: an overall comparison with traditional
generators, Int. Conf. Opportunities Advances Int. Elect. Power
Generation, 1996, pp. 49-54.
[2] H. Polinder, F. F. A. van der Pijl, G. J. de Vilder, and P. J. Tavner,
Comparison of direct-drive and geared generator concepts for wind
turbines, IEEE Trans. Energy Convers., vol. 21, no. 3, pp. 725-733,
2006.
[3] M. Bash, S. Pekarek, S. Sudhoff, J.Whitmore, and M. Frantzen, A
comparison of permanent magnet and wound rotor synchronous
machines for portable power generation, Power Energy Conf.
Illinois, 2010, pp. 1-6.
[4] J. H. Walker, Output coefficient of synchronous machines: A new
concept, Proc. Inst. Elect. Eng., vol. 115, no. 12, pp. 1-5, 1968.
[5] V. B. Honsinger, Sizing Equations for Electrical Machinery, IEEE
Trans. Energy Convers., vol. EC-2, no. 1, pp. 116-121, 1987.
[6] M. Celebi, Weight optimization of a salient pole synchronous
generator by new genetic algorithm validated by finite element
analysis, IET Elect. Power Appl., vol. 3, no. 4, pp.324-333, 2009.
[7] J. D. Ede, Z. Q. Zhu, and D. Howe, Optimal split ratio for high-
speed permanent magnet brushless DC motors, Proc. Int. Conf.
Elect. Mach. Sys., 2001, vol. 2, pp. 909-912.
[8] Y. Pang, Z. Q. Zhu, and D. Howe, Analytical determination of
optimal split ratio for permanent magnet brushless motors, IEE
Proc. Elect. Power Appl., vol. 153, no. 1, pp. 7-13, 2006.
[9] Y. Tang, Y. Xu, J. Hu, J. Zou, and S. Li, Optimization of split ratio
to design the PM brushless DC motor, Proc. Int. Conf. Elect. Mach.
Sys., 2009, pp. 1-5.
[10] L. J. Wu, Z. Q. Zhu, J. T. Chen, Z. P. Xia, and G. W. Jewell,
Optimal split ratio in fractional-slot interior permanent-magnet
machines with non-overlapping windings, IEEE Trans. Magn., vol.
46, no. 5, pp. 1235-1242, 2009.
[11] D. Evans, Z. Azar, L. J. Wu, and Z. Q. Zhu, Comparison of optimal
design and performance of PM machines having non-overlapping
windings and different rotor topologies, Proc. IET Int. Conf. Power
Electron., Mach., Drives, 2010, pp. 1-7.
[12] Y. Shen, Z. Q. Zhu, and L. J. Wu, Analytical determination of
optimal split ratio for overlapping and none-overlapping winding
external rotor PM brushless machines, IEEE Int. Elect. Mach. Drive
Conf., 2011, pp. 41-46.
XI. BIOGRAPHIES
W. Q. Chu received B. Eng. and M. Sc. degrees in electrical engineering
from Zhejiang University, Hangzhou, China in 2004 and Huazhong
University of Science and Technology, Wuhan, China in 2007,
respectively, and is currently working toward the Ph.D. degree in the
Department of Electronic and Electrical Engineering, The University of
Sheffield, UK. From 2007 to 2009, he was with Delta Electronics
(Shanghai) Co., Ltd. His major research interests include the design and
analysis of novel machines for wind power applications.
Z. Q. Zhu started his research career in 1982 and has been with the
University of Sheffield since 1988, where he is currently Professor of
electrical machines and control systems, Head of the Electrical Machines
and Drives Research Group and Academic Director of Sheffield Siemens
Wind Power Research Centre. His research interests include design and
control of permanent magnet brushless machines and drives for
applications ranging from automotive to renewable energy.
J. T. Chen received B.Eng. and M.Sc. degrees in electrical engineering
from Huazhong University of Science and Technology, Wuhan, China, in
2001 and 2004, respectively, and Ph.D. degree in electrical engineering
from The University of Sheffield, Sheffield, UK, in 2009. Currently, he is
with Midea Welling Motor Technology (Shanghai) Co., Ltd. China. From
2004 to 2006, he was Engineer at Delta Electronics (Shanghai) Co., Ltd.
His major research interests include the design and analysis of novel
permanent magnet brushless machines for automotive and renewable
energy applications.


0.5
0.6
0.7
0.8
0.9
1.0
16 18 20 22 24 26 28 30 32 34 36 38 40
O
p
t
i
m
a
l

s
p
l
i
t

r
a
t
i
o
Number of poles
T/G_EE
T/G_PM
T/V_PM
1196
Powered by TCPDF (www.tcpdf.org)