and Iinite
element model oI the SRM motor in Flux2D
. Experimental
characterization techniques as well as Iinite element techniques
are used to obtain the single-phase Ilux-linkage data. Transient
electromagnetic analysis using the Iinite element method,
coupled with a drive model and optimization strategy in
Matlab/Simulink
.
,
.
,
.
c
c
+
c
c
+ = (1)
The electromagnetic system equation (1) can be rewritten
to:
( )
( )
(
c
c
c
c
=
m
i
i R u
i
i dt
di
.
,
. .
,
1
(2)
For a given SRM characterization (Ilux-linkage (i,)),
the phase current i can be calculated Irom (2).
In order to derive an expression Ior the electromagnetic
torque oI a SRM, an energy-balance must be written. The
electrical energy is used Ior electromagnetic torque
production or is stored into the magnetic Iield. Magnetic
co-energy W
co
is used to derive a general Iormula Ior the
electromagnetic torque production:
( )
( ) ( )
}
c
c
=
c
c
=
=
i
cst i
co
m
di
i i W
i T
0
.
, ,
,
(3)
Equation (3) gives an expression Ior instantaneous torque
in terms oI the diIIerential change in co-energy with
respect to position, evaluated at a constant phase current.
Graphically, the mechanical energy W is the surIace,
enclosed in the energy-conversion graph, by exciting a
phase. Fig. 3 shows an example (
m1
100 rad/s,
m2
250 rad/s, i
reI
10 A, u 310 V). Due to the high back-
emI at rotor speed
m2
, the reIerence current cannot be
sustained. The average torque is equal to the area
enclosed by the change in co-energy divided by the
change in position. The instantaneous torque is the
diIIerential co-energy as the change in position tends
toward zero. Each phase produces a torque, according to
(3). The global electromagnetic torque generation T
SR
oI a
SRM can be approximated by (superposition method):
( ) ( ) , ,
1
i T i T
S
n
k
mk SR
=
= (4)
with: n
S
number oI excited stator phases
20 30 40 50 60 70 80
0
2
4
6
8
10
12
rotor position ()
p
h
a
s
e
c
u
r
r
e
n
t
(
A
)
0 2 4 6 8 10 12
0
0.2
0.4
0.6
0.8
phase current (A)
f
l
u
x
-
l
i
n
k
a
g
e
(
W
b
)
wm1 = 100 rad/s
wm2 = 250 rad/s
W1
W2
Figure 3. Phase current and energy conversion during single-phase
excitation (m1 100 rad/s, m2 250 rad/s, ireI 10 A, u 290 V,
optimal aON and aOFF Ior maximum torque)
Using (4) as an expression Ior the electromagnetic torque
production, a three-dimensional graph, representing the
electromagnetic torque Ior one phase, can be built (Iig. 4).
Two calculation methods can be used to construct this
torque-graph:
- partial derivative (to ) oI the co-energy
- integral over current-range oI the partial
derivative (to ) oI the Ilux-linkage (Iurther called
p
(i,))
( )
dt
i d ,
Figure 2. Electrical equivalent oI a
stator phase
-
3
0 -
2
6 -
2
2 -
1
8 -
1
4 -
1
0
-
6
-
2
2
6 1
0 1
4 1
8 2
2 2
6 3
0
0
4
8
1
2
-25
-20
-15
-10
-5
0
5
10
15
20
25
E
M
t
o
r
q
u
e
(
N
.
m
)
Rotor position ()
Phase
current (A)
Figure 4. Single phase torque behaviour oI a SRM (Flux2D
)
This graph Iully describes the torque behaviour oI a SRM
as Iunction oI phase current and rotor position. Fig. 4
clearly shows that most oI the torque generation is
concentrated in a narrow interval between unaligned and
aligned position. (2) and (3) are the Iundamental
equations to represent the behaviour oI phase current and
electromagnetic torque during the rotation oI a SRM.
These equations depend on both current and position
dependant partial derivatives p
i
and p
, being:
( )
( )
( )
( )
c
c
=
c
c
=
,
, ;
,
,
i
i p
i
i
i p
i
(5)
Instead oI characterizing a SRM by means oI its Ilux-
linkage graph (i,), another approach consists oI using
the partial derivatives p
i
(i,) and p
(i,). Parameter p
i
(i,) represents the phase inductance (Fig. 5) and p
(i,)
the back-emI coeIIicient (Fig. 6). The inductance p
i
(i,)
remains nearly constant when phase current is high, due
to the high saturation in the overlapping poles. For low
phase current levels, no saturation occurs and a linear
relation between inductance and rotor position exists
during pole overlapping.
0
4
8
1
2
3
0 3
3 3
6 3
9 4
2 4
5 4
8 5
1 5
4 5
7 6
0 6
3 6
6 6
9 7
2 7
5 7
8 8
1 8
4 8
7 9
0
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
p
i
(H)
phase
current (A)
rotor position ()
Figure 5. Flux-linkage partial derivative pi (i,) (inductance)
3
0 3
3 3
6 3
9 4
2 4
5 4
8 5
1 5
4 5
7 6
0 6
3 6
6 6
9 7
2 7
5 7
8 8
1 8
4 8
7 9
0
0
4
8
1
2
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
p
[Wb/rad]
rotor position ()
phase
current (A)
Figure 6. Flux-linkage partial derivative p (i,) (back-emI coeIIicient)
Using these partial derivatives, diIIerent control strategies
could be analyzed and Iinally optimized. E.g. the start oI
the inductance raising can be used to calculate an eIIicient
turn-on angle a
ON
to activate the phase. BeIore this pole
overlap, the torque generation is minimal compared to the
Joule-losses in the phase winding. Another optimization
could be perIormed Ior maximum mean torque production
using the area in the energy-conversion diagram.
4. SRM single-phase characterization
The Ilux-linkage partial derivatives (5) Iully describe the
SRM electromagnetic behaviour. This data can be
achieved using Iinite element modelling with measured
material magnetization (BH)-data or by Ilux-linkage
measurement on an experimental test setup.
4.1 Finite element analysis
A multiparametric (i,) model is built and solved through
magnetostatic computation with two-dimensional
triangular elements. The uniIorm mesh pattern is created
using mesh
density on
points. Due to
symmetry, only
halI a 8/6 motor
has to be
modelled. Fig. 7
shows a part oI
the meshed
SRM. A rotating
airgap (0.38 mm) connects the stator with the moving
rotor. Fig. 1 shows the equiIlux-lines at diIIerent rotor
positions. Because oI the length oI the rotor, which is 3
times the rotor diameter, a two-dimensional model is
used. For shorter SRM`s, the end-eIIects can be modelled
by a current and position dependant correction Iactor K
ee
(|14|).
Figure 7. Two-dimensional FE-model (Flux2D
)
( ) ( ) ( ) , . , ,
2
i K i i
ee D
= (6)
Using a stacking Iactor oI 0.97, the eIIective stack iron
length is taken into account. The magnetization (BH)-data
oI the stack material is measured on a standard Epstein
Irame and is stored in the materials database oI the FE-
analysis. The eIIect oI punching on the magnetic material
characteristics is quite complex (|16|) and can be
modelled by modiIying the permeability in the aIIected
zone. This modiIied permeability is Iunction oI the
induction and the distance to the punched border.
4.2 Experimental test setup
DiIIerent papers are published in which automated SRM-
characterization systems are presented (|9-10|). A test
setup is built using a PMSM load machine and torque
transducer (Iig. 8). The load machine operates in speed-
control mode and the SRM in current-control mode. A
braking resistor at the load machine side (generating)
dissipates the energy Irom the SRM (motoring). The test
setup is equipped with a DS1104 dSpace
-controller
board. This setup enables the measurement oI static
torque, Ilux-linkage and torque-speed characteristics.
PMSM
load machine
torque
transducer
SRM
test machine
reduction
Figure 8. Experimental test setup
Miller |11| describes a direct method to measure the
magnetization curves (-i-) using the voltage equation:
( )
dt
i d
i R u
,
. + = (7)
Flux-linkage is calculated by integrating the source
voltage minus the resistive voltage drop. To obtain proper
results, the resistance R must be chosen careIully to obtain
zero Ilux at zero current (Iig. 9).
0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05 0.055
-60
-40
-20
0
20
40
60
time [s]
M
e
a
s
u
r
e
d
q
u
a
n
t
i
t
i
e
s
[
V
]
;
[
A
]
;
[
W
b
]
phase voltage
phase current (x2)
flux-linkage (x40)
Figure 9. Single phase Ilux-linkage characterization (aligned position)
(measurement)
Automation is achieved through a position controlled load
machine, which gradually moves the rotor oI the SRM.
Low voltage excitation is used to minimize the eIIect oI
induced currents in coupled conducting paths.
Fig. 10 compares the calculated single-phase static torque
with the measurements on the test setup.
0
2
4
6
8
10
12
14
-30 -25 -20 -15 -10 -5 0
Rotor position ()
T
o
r
q
u
e
(
N
.
m
)
5A meas. 5A FEM 10A meas. 10A FEM
Figure 10. Single-phase torque generation (simulation vs. measurement)
5. SRM drive modelling techniques
A model oI the complete SRM with drive consists oI the
motor-model (electromagnetic system equation (2) and
(4)), a phase-activation control strategy, a hysteresis
current controller, a speed controller and a load-model.
The hysteresis current controller uses soIt-chopping (ON,
FREEWHEEL and OFF-state) to maintain the reIerence
current. In all simulation results, these angles are
optimized to obtain maximum mean torque Ior every
speed and reIerence current. A Iirst technique uses the
single-phase Ilux-linkage characterization data, obtained
Irom measurements or FE-analysis. For multiple phase
excitations, a superposition oI the torque generation oI
each single phase is used. A second and more accurate
modelling technique calculates the torque-generation at
every time-step using a direct coupling between a model
oI the SRM-drive and a FE-analysis oI the SRM-
geometry.
5.1 Superposition of single-phase
characterization data
The SRM with drive is modelled in Matlab/Simulink
p
h
(
A
)
0 1 2 3 4 5 6 7 8 9 10
0
0.2
0.4
0.6
0.8
ph
(A)
F
l
u
x
-
l
i
n
k
a
g
e
(
W
b
)
0 10 20 30 40 50 60 70 80 90
6
8
10
12
rotor position (deg)
T
(
N
m
)
phase current
phase inductance (*100)
Figure 11. SRM control using superposition oI single-phase
characterization data (Matlab/Simulink
)
5.2 Direct coupling with FE-analysis
Previous modelling technique uses single-phase Ilux-
linkage data to calculate the torque, generated by
simultaneously excited phases, by superposing the shiIted
torque characteristics. In this way, the computationally
intensive electromagnetic Iinite element analysis is de-
coupled Irom the time-stepping simulation. In reality, due
to adjacent excitations, diIIerent polarization oI the stator
occurs, resulting in diIIerent Ilux-paths. This results in a
higher reluctance Ior excitation combination AD,
compared with AB, BC and CD. Fig. 12 illustrates some
Ilux-patterns during simultaneous phase excitations.
C
+
C
-
B
-
B
+ D
+
D
-
A
+
A
-
C
+
C
-
B
-
B
+ D
+
D
-
A
+
A
-
C
+
C
-
B
-
B
+ D
+
D
-
A
+
A
-
excitation CD (11) excitation AB (41) excitation AD (55)
Figure 12. Simultaneous phase excitations
The coupling between drive model and FE-analysis is
represented in Iig. 13. The input parameters Ior the FE-
analysis are the applied phase-voltages and the rotor
speed. The output parameters are the phase-currents and
the instantaneous electromagnetic torque. At every time-
step (t
s
50 s), data is exchanged between the drive
model and the FE-analysis. The coupled electric circuit
model consists oI phase coils with a programmable
voltage source and a switch to obtain a unidirectional
current. A transient magnetic analysis is used, in which a
time-stepping process is employed. This technique allows
to split the drive and load model in Matlab/Simulink
p
h
(
A
)
w
ref
= 150 rad/s - sim.
w
ref
= 150 rad/s - meas.
w
ref
= 250 rad/s - sim.
w
ref
= 250 rad/s - meas.
Figure 14. Validation oI hysteresis currentcontroller ( iref 6A)
2 current transducers (simulation with superposition method vs.
measurements)
With the topology oI a Iour phase SRM, a choice between
two (A-C and B-D) or Iour (one Ior each phase) current
transducers can be made. The number oI current
transducers has a signiIicant inIluence on the torque
perIormance oI the SRM drive. Fig. 14 clearly shows that
chopping already starts beIore the reIerence phase current
is achieved, due to a remaining current in the alternating
phase. This loss oI surIace in the energy-conversion plane
results in a loss oI mean torque production.
Fig. 15 compares the instantaneous torque calculation oI
both modelling techniques with the measurements. The
torque-dip Ior AD excitation is clearly visible in the FE-
coupling and the measurements but does not appear in the
single-phase superposition method. This torque dip is
caused by the increased reluctance due to equal
polarization oI phases A and D (Iig. 12). The diIIerence
between measurement and simulation is due to the
material uncertainties, such as eIIect oI punching,
temperature, .
18
20
22
24
26
28
30
0 10 20 30 40 50 60
Rotor position ()
T
o
r
q
u
e
(
N
m
)
measurement FE-coupling Superposition
AD AB BC CD
Figure 15. Torque during simultaneous excited phases
(iref 14A ; m 50 rad/s ; u 290 V ; aON 32.9 ; aOFF 58.8)
5.4 Estimation of iron losses in SRM`s
The losses in a switched reluctance motor mainly consist
oI iron and copper losses. The prediction oI copper losses
is relatively simple when the current waveIorm is known.
Estimation oI iron losses is much more complicated. The
iron losses are Iunction oI the peak value and time-
derivative oI the induction (B) waveIorm in the stack
material. Figure 16 illustrates the iron losses oI stack
material during sinusoidal excitation with varying
Irequency and induction peak value, measured on a
standard Epstein Irame.
1
10
100
1000
10000
100000
10 100 1000
frequency (Hz)
P
[
m
W
/
k
g
]
0.2T 0.4T 0.6T 0.8T 1T 1.2T 1.4T 1.6T 1.8T 2T 0.1T 0.05T
Figure 16. Iron losses oI stack material during sinusodal excitation
(measurement)
The iron losses can be expressed using the Steinmetz
equation:
B f C P
m Fe
`
. . = (8)
with: C
m
, , : Steinmetz parameters
However, inside the stator and rotor oI a SRM, the Ilux
and induction distribution is not sinusoidal at all. The
induction waveIorms in diIIerent motor parts (rotor core,
rotor pole, stator core, stator pole) can be modelled using
the direct coupling with the FE-model or by an analytical
approach.
When applying the direct coupling between SRM-drive
and FE-model oI the motor geometry, the induction
waveIorms can be calculated Ior every operating point.
This is realized by integrating a single-turn coil in the
stator pole, stator core, rotor pole and rotor core (Iig. 17).
The induced coil voltage e
coil
is returned to
Matlab/Simulink
= (10)
with:
( )
dt
dt
dB
B B
f
T
eq
2
0
2
min max
2
.
1
.
2
}
|
.
|
\
|
T
f
r
6
= (T: induction waveIorm period)
To validate the estimation oI iron losses, measurements oI
electrical, mechanical and Joule energy must be
perIormed on the test setup. Iron losses can be calculated
using (11):
Joule mech el Fe
P P P P = (11)
When analysing (11), the iron losses are very small
compared to the electrical and mechanical energy. This
error propagation can result in a global error on the iron
losses measurement oI up to 10. For this reason,
accurate measurement oI SRM iron losses is very
diIIicult. Fig. 20 compares the simulation results, using
the superposition oI harmonic components and the
modiIied Steinmetz equation, with the measurements on
the test setup. The simulation results, using the modiIied
Steinmetz equation, give the best results.
Iron Iosses
0
20
40
60
80
100
120
140
25,00 75,00 125,00 175,00 225,00
rotor speed (rad/s)
P
f
e
(
W
)
Modified Steinmetz - 5A Harm. superposition - 5A measurement - 5A
Modified Steinmetz - 7.5A Harm. superposition 7.5A measurement 7.5A
Figure 20. Estimation oI iron losses (simulation vs. measurement)
6. Conclusion
DiIIerent nonlinear modelling techniques Ior SRM`s are
discussed in this paper. A Iirst method uses the
characterization data (measured or obtained by FEM) oI a
single phase and superposes the torque characteristics to
model simultaneous phase-excitations. A second method
uses a coupling between the drive model and a FE-
analysis to calculate the torque behaviour oI
simultaneously excited adjacent phases. Those results are
more accurate because the mutual inductance is taken into
account. Similarly, the induction waveIorm in diIIerent
motor parts can be calculated with higher accuracy using
the direct coupling. For optimization oI the SRM control,
this time-consuming method must be replaced by (less
accurate) analytical modelling techniques, programmed in
Matlab/Simulink
.
Control oI SRM`s can be deIined through diIIerent
strategies, which optimize one single parameter oI the
drive concept. A Ilexible optimization platIorm generates
optimized drive parameters through an input Iunction
(application demands) and a set oI rules (relations
between drive parameters and input Iunction). Those rules
are generated using the nonlinear model. The accuracy oI
the optimization procedure strongly depends on the model
accuracy.
8. Acknowledgement
The authors wish to thank the Flemish Government (IWT)
Ior granting the research project 'Bepaling van de
optimale stuur- en regelparameters voor systemen met
SR-motor aandrijving. Ontwerp van een
ontwikkelingsplatIorm.
References
|1| N. J. Nagel, R. D. Lorenz, Modelling oI a switched
reluctance motor using an operating point analysis and the
unsaturated torque equation, IEEE Trans. on Industrv
Applications, vol. 36, May/June 2000, pp. 714-722.
|2| W. M. Chan, W. F. Weldon, Development oI a simple
nonlinear switched reluctance motor model using measured
Ilux linkage data and curve Iit, in Conf. Rec. IEEE-IAS
Annual Meeting, New Orleans, LA, Oct. 5-9, 1997, pp 318-
325.
|3| V. Hrabovcova, P. RaIajdus, M. Licko, L. Janousek,
Modelling oI the dynamic operation oI the switched
reluctance motor drive by Simulink, in Conf. Proc.
SPEEDAM, Sorrento (It.), June 3-5, 1998, pp. 61- 66.
|4| V. Vuji, S. N. Vukosavic, A simple nonlinear model oI the
switched reluctance motor, IEEE Trans. on Energv
Conversion, vol. 15, Dec. 2000, pp. 395-400.
|5| C. Roux, M. M. Morcos, A simple model Ior switched
reluctance motors, IEEE Power Engineering Review, Oct.
2000, pp. 49- 52.
|6| S. Mir, I. Husain, M. E. Elbuluk, Switched reluctance
motor modelling with on-line parameter identiIication,
IEEE Trans. on Industrv Applications , vol. 34, July/August
1998, pp. 776-783.
|7| J. Mahdavi, G. Suresh, B. Fahimi, M. Ehsani, Dynamic
modelling oI non-linear SRM drive with Pspice, in Conf.
Rec. IEEE-IAS Annual Meeting, New Orleans, LA, Oct. 5-
9, 1997, pp 661-667.
|8| I.A. Viorel, G. Henneberger, I. Tomescu, I. Vintiloiu, A
new approach to the computation oI the switched
reluctance motor dynamic behaviour, in Conf. Proc.
ICEM2000, Espoo (Fi.), August 28-30, 2000, pp. 1605-
1608.
|9|A. D. Cheok, N. Ertugrul, Computer-based automated test-
measurement system Ior determining magnetization
characteristics oI switched reluctance motors, IEEE Trans.
on Instrumentation and Measurement, vol. 50, June 2001,
pp. 690-696.
|10| V. TriIa, A. Graur, O. Rabulea, Labview virtual
instrumentation Ior a switched reluctance motor, in Proc.
Ee2001, Novi Sad (Yu.), Oct. 31 Nov. 2, 2001, pp. 283-
285.
|11| T. J. E. Miller, Electronic control of switched reluctance
machines (Newnes Power Engineering Series, OxIord,
2001).
|12| T. J. E. Miller, Switched reluctance motors and their
control (OxIord University Press and Magna Physics
Publications (U.K), 1993).
|13| P. N. Materu, R. Krishnan, Estimation oI switched
reluctance motor losses, IEEE Trans. on Industrv
Applications, vol. 28, N 3, May/June 1992, pp. 668-679.
|14| A. Matveev, V. Kuzmichev, E. Lomonova, A new
comprehensive approach to estimation oI end-eIIects in
switched reluctance motors, Proc. ICEM2002, Brugge
(Belgium), August 26-30, 2002.
|15| J. Reinert, R. Inderka, R. W. De Doncker, A novel method
Ior the prediction oI losses in switched reluctance
machines, Proc.EPE97 , Trondheim, 1997, pp 3608-3612.
|16| F. Ossart, E. Hug, O. Hubert, C. Buvat, R. Billardon, EIIect
oI punching on electrical steels: experimental and
numerical coupled analysis, IEEE Trans. on Magnetics, vol.
36, N 5, Sept. 2000, pp. 3137-3140.
Biographies
F. Dhulster was born in Tielt,
Belgium, on April 25, 1974. He
received the degree oI industrial
engineer in mechanical
engineering Irom the Katholieke
Hogeschool in Ostend, Belgium,
in 1996, and the M. Sc. Degree
in mechanical engineering Irom
Katholieke Universiteit Leuven,
Heverlee, Belgium, in 1998.
Currently he is perIorming his
Ph.D.-study at the Hogeschool
West-Vlaanderen dept. PIH in cooperation with the
Katholieke Universiteit Leuven dept. ESAT. His
interests are optimization and modelling oI Switched
reluctance machines and voltage sags.
K. Stockman was born in
Kortrijk, Belgium, on September
24, 1972. He received the
degree oI industrial engineer in
electrical engineering Irom
'Provinciale Industriele
Hogeschool, Kortrijk, Belgium,
in 1994. Since 1995 he has been
with the department oI electrical
engineering oI the Hogeschool
West-Vlaanderen, dept. PIH,
Kortrijk, Belgium. Currently he
is perIorming his Ph.D.-study at the Katholieke
Universiteit Leuven. His interests are adjustable speed
drives, voltage sags and control engineering.
R. J. M. Belmans (S`77-M`84-
SM`89) received the M.S.
degree in electrical engineering
and the Ph.D. degree Irom the
Katholieke Universiteit Leuven,
Leuven, Belgium, in 1979 and
1984, respectively, and the
special Doctorate and the
Habilitierung Irom RWTH
Aachen, Germany, in 1989 and
1993, respectively.
He is currently a Full ProIessor
at the Katholieke Universiteit
Leuven, teaching in the areas oI electrical machines,
power electronics, and variable-speed drives. His
research interests include power quality, electrical energy
systems, electrical machine design, and vibrations and
audible noises in electrical machines. He was with the
Laboratory Ior Electrical Machines oI RWTH Aachen as
a Von Humboldt Fellow Irom October 1988 to September
1989. From October 1989 to September 1990, he was a
Visiting ProIessor at McMaster University, Hamilton,
ON, Canada. He obtained the Chair oI the Anglo-Belgian
Society at London University Ior 1995-1996. He is
currently a Visiting ProIessor at Imperial College,
London, U.K.
Dr. Belmans is a Fellow oI the Institution oI Electrical
Engineers, U.K., the International Compumag Society,
and the Koninklijke Vlaamse Ingenieursvereniging. He is
chairman oI the board oI Elia, the Belgian transmission
system operator.