Zeszyty Problemowe Maszyny Elektryczne Nr 4/2014 (104)

99

Rafa Piotuch, Ryszard Paka

West Pomeranian University of Technology Szczecin
Department of Power Systems and Electrical Drives

FEM BASED IPMSM OPTIMIZATION

OPTYMALIZACJA MES MASZYNY SYNCHRONICZNEJ Z MAGNESAMI
ZAGNIEDONYMI
Abstract: The work presented in this paper relates to the Interior Permanent Magnet Synchronous Motor
(IPMSM) optimization procedure programed in Matlab and Maxwell environments. The stator of the machine
is an mass-produced one with concentrated winding. In the first optimization stage geometry of IPMSM
machine was considered, concerning average torque value maximization and cogging torque minimization
with physical and technological constraints. By combining Matlab software and Maxwell application authors
used genetic algorithm for Finite Element Model optimization. Moreover Ld and Lq inductances were estimated
for evaluation of CPSR machine capabilities and selection for the best geometry among Pareto Front solutions.
Streszczenie: Artyku podejmuje temat optymalizacji maszyn synchronicznych z magnesami zagniedonymi
z wykorzystaniem narzdzi Matlab i Maxwell. Dokonano optymalizacji geometrii wirnika maszyn o zadanych
parametrach stojana, z uwzgldnieniem maksymalizacji wartoci redniej momentu elektromagnetycznego
i minimalizacji momentu zaczepowego oraz ogranicze geometrycznych i technologicznych. W programie
Matlab zaimplementowano algorytm genetyczny uyty do optymalizacji modelu MES stworzonego
w programie Maxwell. Ponadto wyznaczono indukcyjnoci w osiach d oraz q dla wybranych struktur wirnika
wskazanych frontem Pareto w celu wyboru optymalnego rozwizania zapewniajcego szeroki zakres pracy
maszyny przy staej mocy.
Keywords: PM electrical machines, IPMSM optimization, FEM, Pareto front
Sowa kluczowe: Maszyny z MT, optymalizacja, MES, Pareto front

1. Introduction
Thanks to the rapid advancement in the field of
power electronics, digital signal processors and
control algorithms PM excited and Switched
Reluctance Motors are finding more and more
applications and have replaced traction systems
of most present hybrid electrical vehicles
(HEVs) because they offer high performance
over other DC and AC machines [1, 2, 3].
In particular, owing to development of rareearth magnets with high energy product, it is
possible to develop high power density
machines with high overall efficiency.
Furthermore, extended high speed capabilities,
demanded in a highway cycle, are achieved
thanks to proper rotor geometry design, and
field weakening control strategies [3, 4, 5, 6].
Surface and radially-laminated Interior PM
synchronous machines with conventional
structure have limited or zero flux-weakening
capability [5, 7]. Properly designed IPMSMs
are capable of operating in CPSR (Constant
Power Speed Region) such machines perform
also inverse saliency that is q-axis inductance

is larger than d-axis inductance. Consequently,

it has an additive torque value, so called
reluctance torque, that may be exploited to
extend CPSR. Other positive feature of IPM
rotor is that centrifugal forces cannot damage
the magnet, thus whole construction is
mechanically robust, especially for high speeds.
To protect environment, there is a strong
demand to develop highly efficient motors with
high torque/mass ratio. Small cogging torque
reduces noise and mechanical vibrations which
cause increased reliability [8, 9, 10]. All these
requirements can be fulfilled by appropriately
designed PM motors [7, 11, 12]. Proposed
geometry has segmented magnet poles oriented
in the radial direction and iron bridges between
magnets that provides additional flux canals to
give the rotor inherent capability of flux
weakening. Such structures are referred as
Segmented IPM machines [5, 7]. Optimization
procedure with genetic algorithm has been
implemented in Maxwell and Matlab tools. The
final geometry has been analyzed regarding
inductances and air-gap magnetic flux density.

Zeszyty Problemowe Maszyny Elektryczne Nr 4/2014 (104)

100

2. PMSM mathematical model

3. Design problem

Torque developed in IPM motor can be divided

into PM-caused component and reluctance
torque as shown in Fig. 2. Mathematical model
of IPM machines can be described as follows:

The case of study is represented by a 4-pole

segmented IPMSM with fixed stator geometry
and winding parameters. Rotor is equipped with
NdFeB magnets (Br = 1.23 T, 3x7x40 mm).
Most important motor dimensions are presented
in Table 1. Before the optimization process
implementation, classical and segmented
IPMSM
geometries
were
considered.
According to the literature [4, 5] fractional
magnets arrangement may provide very wide
flux-weakening range with high overall
performance parameters. In such structures
d-axis current is still used for PM
demagnetization but it is also used to alter PM
flux-paths. The demagnetizing current causes
some part of PM flux to be canalized into the
rotor iron section between magnet poles.
In such a way PM flux passing through the airgap is efficiently reduced, while the PM-flux is
mostly preserved.
Initial design has been pre-optimized with
simple iterative process. Selected design
variables have been chosen according to the
geometrical constraints, and are presented in
Table 2. Analyzed geometry design variables
are also depicted in Fig. 1.

Tem =

3
pb PM I q + ( Ld Lq ) I d I q
2

(1)

The first term in the equation (1) is the PM

generated torque, and the second term is the
reluctance torque which is proportional to the
difference in stator inductances, Ld and Lq. In
the analyzed IPMSM, Lq is higher than Ld (due
to the lower reluctance in q-axis direction),
because the magnetic flux flowing along the
d-axis has to cross through the magnet cavities
in addition to the rotor air gap, while the
magnetic flux of the q-axis crosses only the air
gap. The d-axis is also basically magnetized
with PM [13, 14]. Such difference increases
torque and extends CPSR. Mathematical model
for IPM machines is presented by following
equation set:
dd
pb m q
dt
dq
+ pb m d
U q = RI q +
dt

U d = RI d +

d = Ld I d + PM
q = Lq I q

(2a)
(2b)
(2c)
(2d)

The control strategy applied for such motors

meets several limitations:
(3a)
U d 2 + U q2 < U N 2
Id 2 + Iq2 < I N 2

(3b)

For the high speed regions flux from magnets

gives high electro-motive force, which exceeds
supply voltage. Using field weakening
method, main flux is decreased by d-axis
negative current (2c) and thus it is possible to
stay in the voltage limit (3a). Optimum
flux-weakening condition can be written as:

PM

(4)
Ld
Such designs are called optimal fieldweakening IPM motor designs and theoretically
exhibit unlimited CPSR. Poles segmentation
provides physical reduction of the air-gap
magnet flux - PM - during flux-weakening,
thus very high ratio of Lq to Ld is not crucial to
extend CPSR. Authors try to consider Lq/Ld
ratio, and PM decrease in the high current
regions.
I rated =

Fig. 1. Initial geometry cross-section

Table 1. Machine parameters
Rotor Outer
Diameter
[mm]

Air gap
length
[mm]

70

Stator
Outer
Diameter
[mm]
120

Magnet
Thickness
[mm]
3.25

Zeszyty Problemowe Maszyny Elektryczne Nr 4/2014 (104)

101

Table 2. Design variables constraints

x1 [mm]
Flux Barrier
X position
4-8

x2 [mm]
Flux
Barrier
Y position
20-25

x3 [mm]
Flux
Barrier
Radius
2-3

x4 [mm]
Magnet
Inset
Radius
20-27

4. Optimization process
In the optimization procedure authors used
heuristic technique based on struggle between
individuals commonly known as Genetic
Algorithm (GA). GA mimics natural selection
process were the strongest individuals survive
and non-important features fades away during
struggle between genes according to present
demanding [12, 15]. Details of used genetic
algorithm have been described in detail in [16].

f1 ( x ) = max( Tcogg ( t ) )
f 2 ( x ) = mean( Tem )

(4a)
(4b)

After solving the MOP a set of optimal nondominated solutions were generated, creating
Pareto front shown in Fig. 4. It is a set of
models that acts as area for selection of best
individual. The final selection should be made
taking into account other features that the motor
should have [1].
4.2. Second optimization stage
For obtained set of models Ld and Lq
inductances values has been evaluated using
method presented in [13]. It is based on
Lagrange formalism, which, according to
Sobczyk [17], may be applied to PM excited
machines. A model with the highest ratio
of Lq/Ld has been selected for the final analysis.
Whole optimization algorithm has been
presented in Fig. 2 and has been described in
detail in [12, 15].

Fig. 2. Optimizatiom algorithm workflow

5. Results
The direct problem has been solved using a 2D
FE model of the motor; torque has been
calculated with the virtual work principle.
GA used 10 generations with 70 models
in population. First optimization stage lasted for
about 6 hours on typical performance PC
(I5, 8GB RAM, Windows 7). The mesh of the
model consisted of a few thousand of elements.
Pareto Front
0.08
Pareto optimal
Final model
Initial model

0.075
max(Tcogg) (Nm)

4.1. First optimization stage

The problem presented in the paper is a
Multiobjective Optimization Problem (MOP)
subject to a set of constraints. For the proposed
geometry the following objective functions
should be minimized, and are defined as:

0.07
0.065
0.06
0.055
0.05
-3.6

-3.4

-3.2

-3
-2.8
-2.6
-mean(Tem) (Nm)

-2.4

-2.2

Fig. 3. Pareto front for optimal geometrie

In the second optimization stage all

Pareto-Optimal geometries were compared
considering Lq/Ld ratio. The final geometry is
presented in Fig. 7 (Lq/Ld = 1.8).

Zeszyty Problemowe Maszyny Elektryczne Nr 4/2014 (104)

102

Torque waveforms
5
Final Model
Initial Model

Torque (Nm)

12
18
electrical angle (deg)

24

30

Fig. 4. Torque waveforms comparison

6. Summary

Torque waveforms
0.08
Final Model
Initial Model

0.06

Torque (Nm)

0.04
0.02
0
-0.02
-0.04
-0.06
-0.08

6
9
mechanical angle (deg)

12

15

Fig. 5. Cogging torque waveforms comparison

Air gap magnetic flux density
0.6
Bnormal

0.5

Btangent

0.4

B (T)

0.3
0.2
0.1
0
-0.1
-0.2

15

30
45
60
mechanical angle (deg)

Fig. 7. Final geometry cross-section; magnetic flux

denisty distribution

75

90

Fig. 6. Magnetic flux density waveforms (no-load

condition) final geometry

Maxwell and Matlab tools were combined for

the rotor geometry optimization process of
Segmented
Interior
Permanent
Magnet
Synchronous Motor. The connection between
these two packages allowed flexible FE model
geometry definition and analysis as well as
effective results evaluation. The whole
optimization process has improved average
torque value significantly with the same
cogging torque value prior to initial geometry.
The simulation points out the benefits of the
optimization using a genetic algorithm.
Non ideal torque waveforms and higher
harmonics (Fig. 6) in the air-gap magnetic flux
density make sinusoidal phase currents
improper for such IPM motors. Simple power
supply system applied during simulations
causes high torque ripples. It should be
emphasized that Maxwell can be connected
with Simplorer, therefor whole analysis of
a drive system, may be properly evaluated [18]
with cascade control structure commonly used
in practical applications. During simulations
there were encountered several problems.
The first one was proper mesh settings in order
to shorten whole optimization process, and
assure correct FEM calculation results.
Particularly, mesh should be more dense in the
air gap region, and in the stator core could be
thinner.

Zeszyty Problemowe Maszyny Elektryczne Nr 4/2014 (104)

The second problem was to select proper

control angle for each model to produce the
highest torque from the available current. The
control angle in each model was constant, thus
torque evaluation exhibit minor error (current in
each
phase
was
sinusoidal
with
1.6 A rms value). Lq/Ld ratio for Pareto-optimal
models varied slightly, but the ratio was not
evaluated for other analyzed models.
Accurate procedure needs proper power angle
evaluation
for
each
model
during
GA optimization which strongly extends
calculation time with proposed inductance
estimation procedure. Future work will be
focused on Ld and Lq saturated parameters
calculation, because nonlinear IPM motor
characteristics may affect CPSR performance.

7. Literature
[1]. Di Barba P., Mognaschi M. E.: Industrial design
with multiple criteria: shape optimization of
a permanent-magnet generator, IEEE Transaction
on Magnetics, vol. 45, no. 3, pp. 1482-1485, (2009)
[2]. Paplicki P., The new generation of electrical
machines applied in hybrid drive car, Electrical
Review, R. 86, no. 6, pp. 101-103, (2010)
[3]. May H., Paka R., Paplicki P., Szkolny S.,
Wardach M.: Comparative research of different
structures of a permanent-magnet excited
synchronous machine for electric vehicles, Electrical
Review, no. 12a/2012, pp. 53-55, (2012)
[4]. Di Barba P., Mognaschi M. E., Paka R.,
Paplicki P., Szkolny S.: Design optimization of
a permanent-magnet excited synchronous machine
for electrical automobiles, International Journal of
Applied Electromagnetics and Mechanics, IOS
Press, vol. 39, no. 1-4, pp.889-895, (2012)
[5]. Stumberger B, Hamler M., Trlep M., Jesenik
M.: Analysis of Interior Permanent Magnet
Synchronous Motor Designed for Flux Weakening
Operation, IEEE Transaction on Magnetics, Vol. 37,
No. 5, pp. 3644-3647, (2001)
[6]. Paka R., Paplicki P., Piotuch R.,Wardach M.:
Maszyna z magnesami o regulowanym wzbudzeniu
wybrane wyniki prac projektowych, Prace Naukowe
Instytutu
Maszyn,
Napdw
i
Pomiarw
Elektrycznych Politechniki Wrocawskiej, Vol 66,
no. 32, pp. 128-133 (2013)
[7]. Dutta R.: A Segmented Interior Permanent
Magnet Synchronous Machine with Wide FieldWeakening Range, The University of New South
Wales, PhD thesis, (2007)

103

[8]. Keyahani A., Murthy S. K., Studer C. B.,

Sebastian T.: Study of cogging torque in permanent
magnet machines, Electric Machines and Power
Systems, no. 27, pp. 665-678, (1999)
[9]. Wardach M.: Cogging torque reducing in
electric machine by poling modification of magnetic
circuit, Electrical Review, No. 2, pp. 131-133,
(2009)
[10]. Putek P., Slodika M., Paplicki P., Paka R.:
Minimization of cogging torque in permanent
magnet machines using the topological gradient and
adjoint sensitivity in multi-objective design,
International Journal of Applied Electromagnetics
and Mechanics, vol. 39, no. 1-4, pp. 933-940, (2012)
[11]. Putek P., Paplicki P., Slodika M., Paka R.,
Van Keer R., Application of topological gradient
and continuum sensitivity analysis to the multiobjective design optimization of a permanentmagnet excited synchronous machine, Electrical
Review, vol. 88, no 7a, pp. 256-260, (2012)
[12]. Caramia R. Piotuch R. Paka R.: Multiobjective
FEM based optimization of BLDC motor using
Matlab and Maxwell scripting capabilities, Archives
of Electrical Engineering, vol. 63(1), pp. 115-124,
(2014)
[13]. Piotuch R.: Inductance calculation considering
magnetic saturation of interior permanent magnet
synchronous machines, Informatyka, Automatyka
Pomiary w Gospodarce i Ochronie rodowiska
vol. 2. pp. 29-33 (2013)
[14]. Rahman M., Zhou P.: Determination of
saturated parameters of pm motors using loading
magnetic fields, IEEE, Transactions on Magnetics,
Vol. 27, No. 5, (1991)
[15]. Caramia R. Piotuch R. Paka R.: Multiobjective
geometry optimization of BLDC motor using an
evolutionary algorithm, Zeszyty problemowe
Maszyny Elektryczne vol. 100, part I. pp. 89-94.
(2013)
[16]. Deb K., Pratap A., Agarwal S., Meyarivan T.:
A fast and elitist multiobjective genetic algorithm:
NSGA-II, IEEE Transactions On Evolutionary
Computation, vol. 6, no. 2, April (2002)
[17]. Sobczyk T.: Problems of mathematical
modelling of permanent magnet synchronous
generator, Prace Instytutu Elektrotechniki, vol. 231,
pp. 99-123, (2007)
[18]. Paka R., Paplicki P., Piotuch R., Wardach M.:
Simulation results of a permanent magnet
synchronous motor with interior rotor which is
controlled by the hysteresis current controller,
Electrical Review, no. 02b/2013, pp. 147-149,
(2013)

104

Zeszyty Problemowe Maszyny Elektryczne Nr 4/2014 (104)

Authors

Acknowledgment

M.A. Rafa Piotuch, rafal.piotuch@zut.edu.pl

Prof. Ryszard Paka, rpalka@zut.edu.pl
West Pomeranian University of Technology
Szczecin, Department of Power Systems and
Electrical Drives,
ul. Sikorskiego 37, 70-313 Szczecin,
tel.: +48 91 449 48 73

This work has been created with the support of

the Ministry of Education and Science, Poland,
under grant N N510 508040.