IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 1

Optimal Design of a Dual Stator Winding

Induction Motor with Minimum Rate
Reduction level
M. Ayaz Khoshhava, H. Abootorabi Zarchi and G.R. Arab Markadeh
Abstract— Flux estimation of Single Winding NSi Number of winding turns per phase
Induction Machines (SWIMs) operating in zero speed I Current
region is challenging. Since, in this operation region δi Current density
the voltage drop on the stator resistance value which kF Fill factor
varies with operating point, is not negligible. Cross section area of the wire used for the
Furthermore, in this operation region because of the asi
interference between higher band frequencies of dc-
stator ith winding set
rejection filter with low-pass filter of the stator diameter of the wire used for the ith
dSi
currents, the dc-offset of measured currents may winding set
distort the true values of stator current. Dual Stator The amplitude of the flux density of the ith
Bgi
Winding Induction Machines (DSWIMs) which have winding set
two sets of three phase windings with different The flux angle of each winding set respect
number of pole pairs and adapt a standard squirrel Ɵ
to the A winding axis of ABC winding set
cage rotor, have overcame the associated problems the difference angle among ABC and XYZ
of SWIMs in zero speed region. Nevertheless, δ
winding flux densities
DWSIMs have lower power rating than SWIMs if they
use the same stator and rotor frames. In this paper an Bts Stator teeth flux density
optimal design procedure is proposed for DSWIMs BC Yoke flux density
which benefits the advantages of DSWIMs and Output power generated by the ith winding
surmounts the rate reduction problem as much as Pi
possible. In this regard, the optimal flux levels ratio of
sets
winding sets and the best pole pair ratio are aci ampere-conductor per meter
determined, first. Afterward, the winding electrical power generated by the ith
specifications are designated to have maximum PEi
winding set
output power, while the DSWIM advantages
ωSi Electrical speed of the ith winding flux
especially the zero speed region operation capability,
remain valid. A DSWIM is designed, simulated in Cosφgap Air-gap power factor
ANSYS/MAXWELL and fabricated based on the ηgap Air-gap efficiency
proposed method and utilizing commercially th
available standard stator and rotor frames. fi i winding nominal frequency
Experimental assessments verify that the DSWIM ns Synchronous speed
prototype has improved the rate reduction problem
Subscripts
up to 19% in comparison to similar proposed
DSWIMs. s, r Stator and rotor
g Air-gap
Index Terms—Dual Stator Induction Machine ABC, Two pole winding and six pole winding
(DSWIM), optimal flux ratio, Optimal pole number XYZ
selection, rate reduction improvement. i=1 and 2 ABC and XYZ
NOMENCLATURE
DS Internal diameter I. INTRODUCTION
Do External diameter
W
S
Tooth width
Number of slots
G NERALLY speaking, Induction Machines (IMs) are
the most frequent electrical machines in the world
industry and have found applications in various motoring
pi Number of poles of ith winding set and generating applications such as wind energy
dC Core depth harvesting systems [1], pumps and fans, elevators and
d Slot depth even electrical vehicles [2]. Since, they possess
L Core length numerous advantages over other types of electrical
A Slot area
machines as involving low fabrication and preservation
lg Air-gap length
expenses and showing high reliable performance in
kw Winding coefficient
different environmental conditions [3]. In addition, there
ZSi Number of conductors per phase
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
is no requirement for rare magnetic materials for IM
manufacturing which results in the ability to operate in
higher temperature and higher speeds in comparison with
Permanent Magnet (PM) machines [4]. However,
SWIMs operation in zero speed region is problematic.
Indeed, in this region, in order to avoid flux saturation,
the amplitude and frequency of stator voltage must be
decreased significantly. Consequently, the stator
resistance (a varying value with operating point) voltage
drop is not negligible and its effect must be taken into
consideration for flux estimation. Additionally, reduction
of stator current frequency will stimulate the offset
problem of integral operator in flux calculation
equations. Thus, the flux estimation in zero speed
involves some errors.
Recently, various kinds of IMs which have two sets of
three phase windings have been the center of attention of
many researches. Since, they have lots of superiorities,
such as higher reliability and broader range of operation
speeds over SWIMs [5]. These induction machines are
classified into two main groups. The first, which is
known as Brushless Doubly Fed IM (BDFIM), has two
sets of three phase windings wounded for different pole
pairs in their stator and a specially designed rotor called
nested loop rotor [6]. Utilizing different pole pairs
cancels the direct electromagnetic coupling amongst two
winding sets, however implementing a nested loop rotor
in BDFIMs results in indirect electromagnetic coupling
between stator winding sets. One of the winding sets of a
BDFIM, called power winding, is directly connected to
the grid and the other, which is named as the control
winding, is fed through a power converter. Consequently,
despite SWIMs, BDFIMs do not require a rated power
converter. The other kind of IMs with two sets of three
phase windings named Dual Stator Winding IMs
(DSWIMs), adapt standard squirrel cage rotors and they
are further categorized into two groups: equal pole pairs
and unequal pole pairs. In the first group there exists
direct electromagnetic coupling among stator winding
sets. The second type has been first introduced by Lipo et
al. in 1998 [7]. In this type of DSWIMs, since the pole
pairs are unequal and a squirrel cage rotor is utilized,
there exists no direct and indirect coupling between
winding sets [8]. Consequently, their control system is
simpler in comparison with BDFIMs and DSWIMs with
equal pole pairs. Furthermore, the independent operation
of stator winding sets brings the ability of convenient
operation in zero speed region. Indeed, in this operation
region, one of the winding sets works in generating mode
and the other works in motoring mode and supplies the
load torque and the other winding set torque. Principally,
DSWIMs with unequal pole pairs have some advantages
over other sort of IMs with two stator winding sets as low
2
construction costs, absence of circulating currents and
wide speed operation region including zero speed [9].
This paper is assigned to the optimal design of a
DSWIM with unequal pole pairs. Generally, there are
two main approaches for DSWIM design. In the first
approach the whole machine parameters including
windings and stator and rotor frame specifications are
determined, which may result in high fabrication costs
[10-13]. Other approaches utilize standard rotor and
stator frames and determine the winding parameters such
that a design objective such as loss minimization or
torque maximization, fulfilled [14, 15]. Obviously,
adapting standard and commercially available frames
leads to simple and low-cost implementation. In [10] a
DSWIM with unequal pole pairs is designed optimally
using Genetic Algorithm (GA) approach. The objective
function is considered to be the core loss minimization
and all of the rotor and stator parameters are determined
as the best chromosome in GA. In [12] GA is also applied
to design a DSWIM with equal pole pairs. However, the
objective is to minimize the reactive power of the control
winding. In [14] a DSWIM which has equal pole pairs, is
designed and fabricated using a standard SWIM frame.
In this machine one of the winding sets works in the
motoring and the other works in generating mode and
supplies a DC load. The 3.4kW DSWIM in [15] which
has a two pole and a six pole winding, is designed for
generation application and it is experimentally realized
by rewinding a 5.5 kW SWIM. Despite the valuable
advantages of DSWIMs over SWIMs, DSWIMs are
derated in comparison with SWIMs. Indeed, if a unique
stator frame is utilized to fabricate an SWIM and a
DSWIM, the DSWIM outputs lower power compared to
the SWIM. As in the above mentioned case in [15] which
has about 36% lower power than SWIM.
In this paper a DSWIM with unequal pole pairs is
designed for motoring application based on a 5.5hp
SWIM frame. The design objective is to benefit the
DSWIM advantages, in addition to minimize the rate
reduction level of the DSWIM.
The rest of the paper is organized as follows: Section
II describes the optimal design procedure. Section III
presents the simulation and experimental results and
section IV is assigned to conclusion remarks.
II. DSWIM OPTIMAL DESIGN
The flowchart of the proposed DSWIM design method
is depicted in Fig.1. As it is shown, in the first step the
optimal pole pair ratio and pole pair numbers are
selected. After that, the flux density ratio for two
windings which optimizes the total air-gap flux density is
determined. The next step is to assign the air-gap flux
density amplitude generated by each winding set. These
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 3

values are determined such that the flux saturation in
stator and rotor cores is avoided. Afterward, the slot area
allocation to each of the winding sets is done to maximize
the output power and to guaranty the zero speed operation
capability. Finally, based on the allocated area and
considering the number of turns for winding sets, the
cross-section area of conductors are calculated. The
design process is fulfilled based on an available three
phase IM frame, the parameters of which are tabulated in
table I and the structure of its rotor and stator slots is
illustrated in Fig. 2.
A. Pole Number Selection
One of the key conditions for independent operation of
In this pole combination the winding set operation is
independent of each other in normal operation condition.
However, in case of saturation, when the two pole
winding saturates it produces a third harmonic flux
density in the air gap even though a third harmonic
current does not flow. This third harmonic flux density
rotates synchronously with the fundamental component
produces a corresponding component in the rotor current
which rotates synchronously with the fundamental. This
third harmonic could couple with a second stator winding
that is wound with six poles.

dcS
winding sets in a DSWIM is to prevent flux saturation.
Therefore, the total flux level should be minimized.

Stator
Furthermore, total flux level minimization results in a

dS
AS
high total power density of the iron utilized for the stator
and rotor core fabrication. Since, total flux minimization WS

lg
permits higher flux levels for each of the winding sets. Wr

Consequently, each winding set generates higher power

Ar
level and the total power, which is the sum of the output

dr
powers of winding sets, increases [16]. The air-gap flux

Rotor
density in a DSWIM is sum of the independent flux
densities of each winding set [17] and can be written as:
Bgt  Bg cos(p1  )  Bg cos(p2    )
1 2
(1)
In which subscripts 1 and 2 stand for ABC and XYZ
winding sets, respectively. Indeed, ABC and XYZ are the Fig. 2. The structure of the rotor and stator slots of the utilized
frame
low pole number and high pole number windings, 1
Maximum Total Flux Density (p.u.)

respectively. Figure 3 illustrates the maximum value of 0.8

Bgt for various pole pair ratios, while δ changes from 0 to 0.6

2π radian. In this figure, 1 to 6 stand for the pole pairs 0.4

1
2
3
ratio of winding sets. As it is clear, the DSWIM drive 4
0.2
system should keep δ equal to 180o and the optimal pole
5
6
0
pair ratio for a DSWIM with different pole pairs is 1:3. 0 1 2 3
 (rad)
4 5 6 7

Fig. 3. The maximum value of the total flux densities for various pole
Optimal pole pair selection
pairs ratios in a DSWIM.
TABLE I
Machine Frame parameters
Optimal flux density ratio Stator Parameters Rotor Parameters
determination Parameters Value Parameters Value
DS 104.8982 Dr 104.2508 mm
mm
Air-gap flux density level selection Dos 167.0156 lg 0.3241 mm
considering core and teeth saturation
avoidance mm
LS 139.0014m Lr 141.7015 mm
AS 89.0112 mm2 Ar 59.1021mm2
Optimal slot area allocation for winding dS 15.5391 mm dr 16.7502 mm
sets for a specified frame
and Ss 36 Sr 28
WS 6.9012 mm Wr 10.1901 mm
In general, if the number of slots are constant, as the
Determination of Winding specification
including number of turns and cross number of pole pairs in an electrical machine increases,
section area
the number of slots per pole decreases and the winding
function involves more harmonics. Since for a purely
and
sinusoidal winding function infinite number of slots is
and
required and by decreasing the number of slots this
Fig. 1. The flowchart of the proposed design process for DSWIMs function is less similar to a purely sinusoidal waveform.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 4

In addition, as the number of poles are increased, the determined values and based on (2) is shown in Fig. 6.
power factor is decreased and the iron loss is increased
[19]. Accordingly, the number of poles should be

Maximum Total Flux density (p.u)

minimized. Therefore p1 and p2 are selected as 2 and 6, 0.1
10
respectively.
B. Optimal Flux Level Selection 10
0

As it is desired to design a DSWIM by rewinding an

0 1 2

available three phase SWIM, the maximum rated flux 10 10


10

density of the DSWIM ought to be the same as the SWIM Fig. 4. The maximum value of the total flux density for various
values of α.
before rewinding. 2
If the ratio of the amplitude of the flux density of two 1.5
pole winding to the six pole winding considered to be α

B(T)
and let δ to be 180o, equation (1) can be rewritten as: 1

1 0.5
Bgt  Bg1 (cos( )  cos(3 )) (2)
 0
0 1000 2000 3000 4000 5000 6000 7000 8000
If Bg1 supposed to be 1 per unit, the maximum total H(AT/m)

flux density of the DSWIM for various α values (1 < 𝛼 < Fig. 5. Cogent non-oriented processed electrical steel of the type M600
100) is depicted in Fig.4 in a logarithmic scale. As it is B-H curve
1

clear the optimal flux density ratio is: B2 B1 Bgt

Airgap Flux Density (T)

0.5

Bg1
6 (3) 0

Bg2
-0.5

In the employed frame, the stator and rotor cores are -1

0 1 2 3 4 5 6
made up of the cogent non-oriented processed electrical  (rad)

Fig. 6. Air-gap flux densities

steel of the type M600, the B-H curve of which is
illustrated in Fig. 5. Figure 5 indicates that selection of
the flux densities more than 1.4T results in the saturation
of the stator and rotor cores.
One of the most challenging fragments in an IM
machine core is the stator teeth. The ratio of the stator
teeth flux density to the air-gap flux density provides
[18]:
Bts  DS
 (4)
Bg WS S S
Pursuant to table I, this ratio will be 1.3228.
The other important step of design process is
determination of the flux density in stator yoke. The ratio
of the yoke flux density to the air gap flux density can be
written as [18]:
BC  DS
 (5)
Bg 2 pi dCS
According to the machine parameters stated in table I,
this ratio is 1.768 and 0.589 for two pole and six pole
windings, respectively.
In consonance with Fig. 5 and equations (4) and (5),
the maximum value for Bgt is specified as 0.7865.
Considering Fig. 4 and (3), the flux densities of two pole
and six pole winding are determined.
Bg1  0.9082T
(6)
Bg2  0.1514T
The total air gap flux density of a DSWIM with the
C. Winding Parameter Selection
This section is assigned to the determination of the
optimal winding parameters including the number of
turns and the cross section area assigned to each winding
set. Actually, these parameters are characterized to
maximize the DSWIM output power, while the
experimental and technical restrictions of DSWIM
operation especially in zero speed ranges are satisfied.
The output power of a DSWIM is sum of each winding
set output powers.
Ptot  P1  P2 (7)
In (7) Pi is calculated as [17]:
Pi  1.11 2 Bgi aci kw  DS2 LS nS cos
(8)
Where, aci is defined as:
Z Si I i
aci  (9)
 DS
If the conductor currents are defined in terms of
current density and their cross section area, (9) can be
rewritten as follows:
2 N Si aSi i
aci  (10)
 DS
According to the utilized frame structure and its
cooling system, δi is considered to be 6 A/mm2. In
consonance with table I and equations (7-10) and
considering the efficiency and the power factor to be 0.85
and 0.8, respectively, the total output power is
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
determined as:
Ptot  19.251(0.9082N S1 aS1  0.1514N S 2 aS 2 ) (11)
From (11) it is clear that, the greater the product of the
number of turns and conductor cross section areas, the
higher the output power.
However, the total slot area limitation should be
considered. The total number of slots of the utilized
frame is 36. Consequently, the conductors of each phase
of stator windings should be distributed in 12 slots. As a
result the total slat area constraint can be considered as:
N S aS  N S aS  6AS k F
1 1 2 2
(12)
According to table I and considering the slot fill factor
to be 0.45, the total product of the number of turns and
the conductor cross section areas will be 240.3 mm2.
Equations (11) and (12) indicate that the larger the two
pole winding is in comparison to the six pole winding,
the higher total output power is produced. Moreover, if
there exists just the two pole winding set and the six pole
one is omitted, the total output power is maximized.
Which results in a three phase SWIM and the advantages
of DSWIMs over the IMs, are canceled. Consequently,
the six pole winding should be designed as small as
possible, while the advantages of a DSWIM over an
SWIM remain valid.
One of the most significant advantages of DSWIMs
over SWIMs is their ability to be accurately controllable
in very low and zero speed ranges. Which is
accomplished by the capability of precise flux estimation
capability of DSWIMs in very low speed ranges. As
stated in pervious sections, in very low and zero speed
operation ranges one of the winding sets operates in
generating and the other in the motoring operation mode.
Therefore they could be supplied by a voltage with a
minimum frequency, in which the flux estimation is
possible [8].
Considering in the zero speed operation region, the
two poles winding set operates in the motoring operation
mode and the six pole winding set works in the
generating operation mode. Since the mechanical speed
is zero, the load power is zero and the produced electrical
power by winding sets are equal to each other.
PE1  PE 2 (13)
Where, PEi is:
PEi   2.22 3 Bgi aci kw  DS2 LSSigap cosgap (14)
Substituting (14) into (13) and omitting the equal
values from both sides of the equation results in:
Bg1 NS1 aS1S1  Bg2 NS2 aS2 S2 (15)
In order to estimate the correct flux values in zero
speed region, the flux densities are set to their nominal
values. Accordingly (15) is rewritten as (16).
6 NS1 aS1 f1  NS2 aS2 f 2 (16)
5
The nominal frequency of ABC and XYZ winding sets
are considered to be 50 and 150Hz respectively. The
minimum frequency of the ABC winding set, in which the
flux could be precisely estimated, is considered to be one
twentieth of the nominal frequency that is 2.5Hz as in [8].
If f1 is set to a constant value in (16), in order to
minimize the N S aS and hence maximize the total output
2 2
power, f2 should be as large as possible. Since it is desired
to maintain the air gap flux at its rated value v/f ratio must
remain approximately constant. Consequently, the
applied voltage (v) must be increased if the frequency is
increased. Nevertheless, increasing f2, significantly
increases the iron losses. Hence, f2 should be limited.
According to the above discussions the maximum
value of f2 in the zero speed region is selected as small as
on third of nominal frequency, which is 50 Hz.
Consequently from (16) the product of the number of
turns and the conductor cross section area will be:
N S1 aS1  184.846mm 2
(17)
N S2 aS2  55.454mm 2
If the number of turns of each winding sets are
considered to be equal to 96 for each phase (six series of
16 turn windings), the diameter of conductors, according
to the standard and commercially available conductors,
will be:
d S1  1.55mm
(18)
d S2  0.9mm
As a result, from (11) the total output power will be
3345.08 W. According to the determined diameters for
winding wires, the nominal rms currents for ABC and
XYZ windings are 9A and 5A, respectively.
According to the number of turns, utilized frame
dimensions and determined flux densities, the windings
nominal voltages can be determined as:
  DS 
E Si  4.44N Si f i k w B gi   LS (19)
 pi 
According to (19) the ABC and XYZ windings nominal
line voltages are 300V and 50V, respectively.
In order to evaluate the effect of α on the DSWIM
output power, this parameter is varied from 1 to 100 and
the output power is calculated, while the number of turns
and winding cross-section area are set to the designed
values. Figure 7 illustrates the output power for various
α values. As illustrated, the optimal value of α in term of
output power is 6.
The detail power rating and nominal parameters of the
designed DSWIM are addressed in table II.
III. EXPERIMENTAL AND SIMULATION RESULTS
The designed DSWIM simulated in
ANSYS/MAXWELL environment and fabricated using
 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 6

an available frame with the parameters given in tables I CW current DC Link Capacitors
sensors
and II. 2D simulations has been performed. Since, the
machine under study has symmetrical structure and all of
PW Inverter
the cross-sections are identical at the axial length of CW Inverter
Line Voltage
machine, the 2D analysis has an acceptable accuracy. sensors
Figure 7 illustrates the various parts of the fabricated
DSWIM prototype. The following will discuss the
Encoder TMS Board
PW current
simulation and experimental results in no-load, full-load Board
Analog Filter
sensors
and zero speed conditions.
The experimental test bench is depicted in Fig.8. The Resistive Load
DSWIM is supplied through a DSP-based completely
DC Machine DSWIM
modular prototype system which has two voltage source
inverters supplying the power (ABC) and control (XYZ)
windings, voltage and current sensor boards and a
TMS320F28335 signal processor board designed by
Texas Instrument Co. In order to measure the rotor speed
Encoder
a 1024 pulse incremental encoder provided by OPKON
Co. is utilized. The stator phase currents and line voltages
of two winding sets are measured through LEM LTS-6- Fig.8. Experimental test setup
NP current and LEM LV-25-P voltage sensors,
respectively. There is also provided an analog second- A. No-load condition
order low pass filter with 2.6 kHz cut-off frequency to The fabricated DSWIM is tested in the no-load
filter the measured stator current and voltage signals. The condition utilizing the test bench shown in Fig. 8. In
inverters switching frequency is chosen 10 kHz. The addition, some simulations performed for no-load
supply voltage for each of the boards is 24 V DC. As condition in ANSYS/MAXWELL.
depicted, a DC generator supplying a resistive load is In this operation condition the DC generator is
utilized as the DSWIM load. disconnected. The ABC voltage source inverter supplies
TABLE II its corresponding winding set a 300V 50Hz voltage
Designed DSWIM nominal parameters passed through the low pass filter. Similarly, the XYZ
Value
Parameter voltage source inverter provides a 50 V 150 Hz. This
ABC XYZ
Pole number 2 6 voltage is also filtered and supplies the XYZ winding set.
Output Power 2845W 500W The no-load ABC winding currents sampled through two
Voltage Frequency 50 Hz 150 Hz sets of current sensors and the currents obtained from
RMS Line Voltage 300 V 50 V
Wire diameter 1.55mm 0.9mm simulation results are depicted in Fig.9 (a) and (b),
Phase Current 9A 5A respectively. Figure 10 shows the similar waveforms for
RMS number of turns 96 96 the XYZ winding set. It should be noted that these current
Efficiency 85% waveforms are depicted in steady-state operation
Power factor 0.8
3500 condition. As it is illustrated, the current waveforms and
amplitude are approximately the same in experimental
Output power (W)

3000 and simulation results.

Figures 11 and 12 illustrate the speed and torque curves
2500 in the no-load condition for experimental tests and also
simulation results. These figures confirm that the
2000
0 20 40 60 80 100 simulated and fabricated DSWIMs show similar transient
 and steady-state behavior. As illustrated after 4.5s the
Fig.7. DSWIM output power for various values of α transient operation condition is terminated and the
Stator ABC
Frame Winding Coupling
steady-state no-load speed is about 2920 rpm. The
Set Bearing Rotor Frame
Bearing steady-state electromagnetic torque is about 0.74 N.m.
ABC
Terminals XYZ This electromagnetic torque is produced to overcome the
Winding Set
no-load losses. Figure 13 shows the core losses in no-
load condition. These losses are obtained from simulation
XYZ
Terminals
results in ANSYS/MAXWELL environment. As it is
1
(a) (b) clear the steady-state core loss is approximately 80W. In
Fig.7. the fabricated DSWIM a) Stator b) Rotor order to determine the mechanical losses, both winding
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 7

sets are disconnected and the rotor is coupled to the DC TABLE III
Fabricated DSWIMs rate reduction ratio
machine. The DSWIM is rotated by DC machine at 2920
DWIM output SWIM Rate reduction
rpm. In this condition, the DC machine input power is Ref
power power Ratio
measured. Now, subtracting the DC machine losses the This
3335W 4048W 17.6%
DSWIM mechanical losses are determined that is 145W. work
Consequently, the no-load losses equals 225W. [17] 1126W 2200W 48.8%
[15] 3473W 5500W 36.8%
Dividing the no-load losses to no-load angular speed, the

Two pole Winding Current (A)

no-load electromagnetic torque is 0.735 N.m. 2 IA IB IC

1
B. Full-load condition
0
The designed machine is also simulated for nominal
-1
operating condition in ANSYS/MAXWELL
-2
environment. In this condition the nominal line voltages, 0 0.01 0.02 0.03 0.04 0.05 0.06
time (s)
300V and 50V, are applied to the ABC and XYZ winding (a)
sets, respectively, and an 11 Nm load is applied to the

Two Pole Winding Current (A)

2
IA IB IC
rotor. The flux distribution and flux density distribution
1
are depicted in Figures 14 and 15, respectively. As it was
also expected from (5) the most challenging part is the 0

stator yoke. -1
Figure 16 illustrates the output power obtained from the 0 0.01 0.02 0.03 0.04 0.05 0.06
simulated DSWIM. The steady-state power is about time (s)
(b)
3.3kW. In order to determine the rating of the fabricated Fig.9. ABC winding currents in no-load condition a) simulation b)
DSWIM, a DC generator supplying a resistive load is experimental
coupled to the DSWIM shaft. At first, the nominal line Six pole Winding Current (A)
IX IY IZ
voltages are applied to the stator winding sets, the DC 4

machine field winding is supplied by its nominal DC 2

voltage (230V) and the variable resistive load is set 0

at its highest value. Then, the load resistance is decreased -2

gradually, until the nominal currents determined in -4
0 0.01 0.02 0.03 0.04 0.05 0.06
section II (9A and 5A) flow through the DSWIM stator time (s)
windings. Figures 17 and 18 illustrate the line currents (a)
for ABC and XYZ winding sets, respectively. These
Six Pole Winding Current (A)

4 IX IY IZ
figures confirm that in the full-load condition the
2
simulated and fabricated DSWIMs have similar current
amplitude and waveforms. In this condition the load 0

resistance is 18Ω and the DC generator armature voltage -2

is 230V. Consequently, the supplied power to the load is -4

0 0.01 0.02 0.03 0.04 0.05 0.06
2939W. The DC machine total losses in this condition is time (s)

396W. Thus, the DSWIM output power which is the sum (b)
of the load power and the DC machine total losses, is Fig.10. XYZ winding currents in no-load condition a) simulation b)
experimental
3335W. Consequently, the DSWIM power rating in the 3000
Rotor Speed (rpm)

design process, simulation and experimental tests are

2000
approximately the same.
As stated previously, the DSWIM prototype with the 1000 Experimental
designed specifications discussed in section II is realized Simulation
0
by rewinding a 5.5hp SWIM. Therefore, the rate 0 1 2 3 4 5 6
time (s)
reduction ratio equals: Fig.11. Rotor speed in no-load condition (experimental)
P  PDSWIM 2
Rate Re duction  SWIM  100  17.6% (20) Experimental
Simulation
PSWIM 1.5
Te(N.m)

In (20) PDSWIM is the DSWIM output power and PSWIM 1

stands for the SWIM output power before rewinding. The 0.5

rate reduction ratio of some similar works are compared 0

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
in table III. As shown in this table the fabricated time (s)
prototype based on the proposed design method has Fig.12. Electromagnetic torque in no-load condition (experimental)
improved the rate reduction problem about 19%.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 8

1000 motoring mode, while the other winding set produces a

800 negative torque and works in the generating mode. The

Core loss (W)

90
600 80 total torque, which is sum of the torques produced by
400 70
winding sets and the load torque are shown in Fig. 20 (c).
200 The rotor speed is shown in Fig. 20 (d). As it is clear the
0
0 2 4 6 8
proposed machine drive system can appropriately supply
time(s) the load torque in the zero speed region.
Fig.13. Core loss in no-load condition (simulation)
20

Two pole winding Current (A)

IA IB IC
15

10

-5

-10

-15
0 0.01 0.02 0.03 0.04 0.05 0.06
time (s)

(a)
20

Two pole Winding Current (A)

IA IB IC
15
10
5
0
Fig.14. Flux distribution in full-load condition -5
-10
-15
0 0.01 0.02 0.03 0.04 0.05 0.06
time (s)

(b)
Fig.17. ABC winding currents in full-load condition a) simulation b)
experimental
10
IX IY IZ
Six pole Winding Current (A)

-5

0 0.01 0.02 0.03 0.04 0.05 0.06

time (s)
Fig.15. Flux density distribution in full-load condition (a)
20
10
IX IY IZ
15
Six pole Winding Current (A)

3,300
10 5
(kW)

5
out

0
P

-5 -5

-10 0 0.01 0.02 0.03 0.04 0.05 0.06

0 2 4 6 8 10 time(s)
tim e (s )
Fig.16. the output power in full-load condition in simulation (b)
Fig.18. XYZ winding currents in full-load condition a) simulation
C. Zero-Speed operation b) experimental
In order to evaluate the operation of the fabricated V/Hz VSI
machine in zero speed region, a scalar (V/Hz) control ABC
DSWIM
scheme is implemented. It should be noted that this
control scheme is adapted just for the zero speed
operation region. The block-diagram of this control + Speed
V/Hz
Regulator
system which is similar to that of [8], is shown in Fig. 19. -
Saturation
VSI
XYZ
As illustrated, in this mode of operation the frequency of
the ABC winding set is fixed at the value considered in
Fig.19. V/Hz control system for zero speed operation region
the design process ( f 1*  2.5Hz ) and the output torque is
adjusted by the supply frequency (and hence the voltage) IV. CONCLUSION
of the XYZ winding set. Figures 20 (a-d) show the results
A procedure for optimal design of DSWIMs with
obtained in the zero speed region. Figures 20 (a) and (b)
unequal pole pairs was proposed. In the proposed
illustrate the torque produced by the ABC and XYZ
method, the optimal pole pair ratio and optimal air-gap
winding sets, respectively. As it is clear, the ABC
flux ratio were determined which were 1:3 and 1:6,
winding set produces a positive torque and acts in the
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
respectively. In addition, the windings characteristics
were specified to maximize the output power, as well as
guaranteeing DSWIM advantages over SWIM. Adapting
the proposed procedure, a DSWIM with unequal pole
pairs was designed. The performance of the designed
machine was validated by some 2D simulations in
ANSIS/MAXWELL. Additionally, the designed
machine laboratory prototype was fabricated by
rewinding a 5.5hp SWIM. The simulation results and the
experimental tests revealed that the designed DSWIM is
able to output more than 3335W power. Consequently,
the designed machine had improved the rate reduction
problem existing in DSWIMs by about 19%, in
comparison to likewise proposed DSWIMs.
0
-0.5
(N.m)
-1
XYZ
T
-1.5
-2
0 1
10
(N.m)
5 IECON Conf., 2013, pp. 2732-2737.
ABC
T
0
0 1
10
Total Torque(N.m)
5 IEEE EPE Conf., 2013, pp. 1-10.
0 Load Torque
0 1
10
Rotor Speed (rpm)
0 in Dual Stator Winding Induction Machines,” in IEEE Trans.
-10
-20
0 1
(d)
Fig.20. zero speed region results a) ABC winding generated torque b)
XYZ winding generated torque c) DSWIM generated torque and the
load torque d) Rotor speed (experimental)
REFERENCES
[1] G. Cimuca, S. Breban, M. M. Radulescu, C. Saudemont, and B.
Robyns, “Design and Control Strategies of an Induction-
Machine-Based Flywheel Energy Storage System Associated to
a Variable-Speed Wind Generator,” in IEEE Trans. Energy
Conversion, vol. 25, NO. 2, 2010, pp. 562–534.
[2] S. Jurkovic, K. M. Rahman, J. C. Morgante and P. J. Savagian
“Induction Machine Design and Analysis for General Motors e-
Assist Electrification Technology,” in IEEE Trans. Industry
Applications, vol. 51, NO. 1, 2015, pp. 631–639.
[3] G. von Pfingsten, S. Steentjes, “Operating Point Resolved Loss
Calculation Approach in Saturated Induction Machines,” in
IEEE Trans. Industrial electronics. vol. 64, NO. 3, 2017, pp.
2538–2546.
9
[4] M. Centner and U. Schäfer, “Optimized Design of High-Speed
Induction Motors in Respect of the Electrical Steel Grade,” in
IEEE Trans. Industrial electronics. vol. 57, NO. 1, 2010, pp. 288–
295.
[5] S. Basak and C. Chakraborty, “Dual Stator Winding Induction
Machine: Problems, Progress, and Future Scope,” in IEEE
Trans. Industrial electronics. vol. 62, NO. 7, 2015, pp. 4641–
4652.
[6] F. Bl´azquez, C. Veganzones, D. Ram´ırez, and C. Platero,
“Characterization of the Rotor Magnetic Field in a Brushless
Doubly-Fed Induction Machine,” in IEEE Trans. Energy
Conversion, vol. 24, NO. 3, 2009, pp. 599–607.
[7] A. Munoz-Garcia; T. A. Lipo,“ Dual stator winding induction
machine drive,” IEEE Industry Applications Conf. 1998, pp. 601-
608.
[8] A. R. Munoz and T. A. Lipo, “Dual stator winding induction
machine drive,” IEEE Trans. Ind. Appl., vol. 36, no. 5, 2000, pp.
1369–1379.
[9] Z. Wu and O. Ojo “Optimal Magnetic Design of the Stator
Windings of Dual Stator Winding Squirrel-Cage Induction
Machines,” in Proc. IEEE Energy Conversion Congress, 2009,
pp. 256-261.
[10] H. keshtkar and H. Abootorabi Zarchi “Improved Design of a
2 3 4 5 Dual Stator Winding Induction Generator for Wind Power
time(s)
(a) Applications,” in Proc. IEEE ICEE Conf., 2016, pp. 1278-1283.
[11] Lucian Nicolae Tutelea, Sorin Ioan Deaconu, Ion Boldea and
Nicolae Budişan “Design, Control and 2D-FEM Validation for a
Double Stator Winding Induction Generator,” in Proc. IEEE
[12] L. Lingshun, H. Yuwen and H. Wenxin “Optimal Design of Dual
Stator--Winding Induction Generator with Variable Speed Based
2 3 4 5
on Improved Genetic Algorithm,” in Proc. IEEE Electrical
time(s) Machines and Systems Conf., 2005, pp. 2345-2348.
(b) [13] L. Nicolae Tutelea, S. Ioan Deaconu, N. Budisan and I. Boldea
“Double Stator Winding Induction Generator for Wind and Hydro
Applications: 2D-FEM Analysis and Optimal Design,” in Proc.
DSWIM Torque [14] V. Chandrasekaran and T. Manigandan “An Innovative
Approach for Energy Conservation in Induction Motor,”
in Proc. ICoMMS Conf., 2015.
2 3 4 5
time(s) [15] J.A. Barrado Rodrigo*, X. Munté, H. Valderrama-Blavi and F.
(c) González-Molina “Design and Testing of a Dual Stator Winding
Induction Generator,” in Proc. IEEE SSD Conf., 2015, pp. 1-6.
[16] Juan M. Guerrero, and O. Ojo, “Total Air gap Flux Minimization
Power Electronics. vol. 24, NO. 3, 2009, pp. 787–795.
[17] Z. Wu, “AN INVESTIGATION OF DUAL STATOR WINDING
INDUCTION MACHINES,” PhD dissertation, Tennessee
Technological University, 2006.
2 3 4 5
time (s) [18] Essam, S. Hamdi, Design of Small Electrical Machines. New
York: John Wiley, 1995, ch. 5.
[19] F. Liang, D. W. Novotny, R. W. Fei, and X. Xu, “Selection of the
Pole Number of Induction Machines for Variable Speed
Applications,” IEEE Trans. Ind. Appl., vol. 31, no. 2, 1995, pp.
304–310.
Mojtaba Ayaz Khoshhava received the B.Sc. and
M.Sc. degrees in Electrical Engineering from Ferdowsi University of
Mashhad and Shiraz University, Iran, in 2012 and 2014,
respectively. He is currently a Ph.D. student at Faculty of
Engineering, Ferdowsi University of Mashhad. His fields of research
include electrical machine design, power electronics and variable-
speed drives.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
Hossein Abootorabi Zarchi received Ph.D
degree from Isfahan University of Technology, Isfahan/Iran. He
was a Visiting Ph.D. Student with the Control and Automation
Group, Denmark Technical University, Denmark, from May 2009 to
February 2010. He is currently an Assistant Professor in the Faculty
of Engineering, Ferdowsi University of Mashhad, Mashhad/Iran.
His research interests are renewable energies, electrical machines
and applied nonlinear control in electrical drives.
10
Gholamreza Arab Markadeh received the B.Sc.,
M.Sc., and Ph.D. degrees in Electrical Engineering from Isfahan
University of Technology, Iran, in 1996, 1998, and 2005,
respectively. He is currently an Associate Professor in the Faculty
of Engineering, Shahrekord University. His fields of research
include nonlinear control, power electronics, and variable-speed
drives. He is the Editor-in-chief of Journal of Dam and Hydroelectric
Powerplant. Dr. Arab Markadeh was the recipient of the IEEE
Industrial Electronics Society IECON’04 best paper presentation
award in 2004.