958

IEEE TRANSACTIONS ON MAGNETICS, VOL. 47, NO. 5, MAY 2011

Optimization of Interior PM Motors With Machaon Rotor Flux Barriers

Piergiorgio Alotto, Massimo Barcaro, Nicola Bianchi, and Massimo Guarnieri
Department of Electrical Engineering, University of Padova, I-35131 Padova, Italy
Interior permanent magnet (IPM) motors are normally designed with two or more flux barriers per pole. The form of such flux barriers has a direct impact on the torque developed by the IPM motor, with regards to both its average value and ripple. The Machaon
structure includes flux barriers of different shape, aimed at reducing the torque ripple. Their shape depends on the number of poles,
number of slots, winding arrangements, and PM volume used in the rotor. An optimization technique is adopted in order to determine
the best shape of the flux barriers with the objective of achieving a smooth torque with a high average value.
Index TermsMachaon, optimization, proper interior permanent magnet (IPM) motor, torque ripple.

I. IPM MOTOR WITH A MACHAON ROTOR STRUCTURE

HE STRUCTURE of a four-pole permanent magnet

(PM) motor with anisotropic rotor is shown in Fig. 1. The
rotor has two flux barriers per pole, which limit the -axis flux
without obstructing the -axis flux. PMs are inset within the
flux barriers in order to saturate the iron bridges and increase the
motor torque. This motor design is known as interior permanent
magnet (IPM) motor or PM assisted synchronous reluctance
motor. A common drawback of such motors is the high torque
ripple [1], [2], which is due to the interaction between the stator
MMF spatial harmonics and the rotor anisotropy. Such a high
torque ripple is intolerable for most applications. A reduction
of the torque ripple can be achieved by means of a suitable
choice of the number of flux barriers with respect to the number
of stator slots, according to a uniform distribution of the flux
barrier tips along the air gap [3]. Such a solution is patented [4]
and requires a number of flux barrier tips with equal spacing
along the air gap and with a number equal to the stator slots 4.
An alternative to such a solution is the adoption of the Machaon
rotor structure, formed by flux barriers of different shape in
adjacent poles [5], [6]. As an example, Fig. 2 shows a four-pole
rotor with two different flux barrier shapes. As far as the torque
behavior is concerned, the most important rotor parameters are
the angular positions of the flux barrier tips [1]. The best choice
depends on the number of poles, number of stator slots, winding
arrangement, and PM volume, and thus it is not a trivial task to
carry out manually through a trial-and-error method.
This paper demonstrates how a stochastic optimization technique can be successfully applied in order to determine the best
shape of the flux barriers with the twofold objective of achieving
a smooth and high average value torque.
A. Analytical Model

Fig. 1. Motor structure and main geometry symbols.

Fig. 2. Machaon rotor structure.

is
where is the harmonic order, is the number of poles,
the mechanical angle, and
is the current phase. The stator
magnetic potential
at the stator bore is

The numerical optimization relies on an analytical model of

the motor [7]. The stator current distribution expressed in the
rotor reference frame is
(1)

Manuscript received May 28, 2010; accepted August 23, 2010. Date of
current version April 22, 2011. Corresponding author: P. Alotto (e-mail:
alotto@die.unipd.it).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TMAG.2010.2073450

(2)
where is the stator inner diameter. In each rotor island (iron
region bounded by flux barriers), the magnetic potential
increases due to the magnetic flux crossing the flux barriers,
according to
(3)
where is a dimensionless coefficient related to the barrier geometry [5], [6],
and
is the

0018-9464/$26.00 2011 IEEE

ALOTTO et al.: OPTIMIZATION OF INTERIOR PM MOTORS WITH MACHAON ROTOR FLUX BARRIERS

959

TABLE I
MAIN GEOMETRICAL DATA OF THE MOTOR

angular position of the barrier. Therefore, the radial component

of the air gap flux density is
(4)
where is the air gap thickness.
Finally, the torque is obtained by integrating Lorentzs force
density

Fig. 3. Torque versus rotor angle (24-slot four-pole motor, non-chorded).

(5)
where
is the stack length. This analytical model considers
the effect of a single pole at a time. The computation of the
Machaon rotor structure is carried out by analyzing the different
poles separately and then superposing their effects since the flux
lines in the poles are supposed to not influence each other.
II. OPTIMIZATION ALGORITHM
As already stated, the main degrees of freedom characterizing
the behavior of the Machaon flux barrier arrangement are the
angles , , , and
shown in Fig. 2, i.e., the angular positions of the tips of the flux barriers. Thus, the design problem
can be cast as a scalar, four-parameter optimization problem.
The chosen optimizer is TRIBES [8], a variant of the
well-known particle swarm optimization (PSO) meta-heuristic.
While in classical PSO the topology and the type and quantity of social relationships of the swarm are user-specified
parameters, in TRIBES they evolve over time in response to
performance feedback. With this modification, PSO becomes
a parameter-free algorithm greatly improving its robustness
in solving arbitrary problems (at the cost of some efficiency
to tackle a specific problem). Although TRIBES has already
been applied to static electromagnetic problems, this is the first
time that it is used to solve real motor design problems, which
constitute a challenging benchmark for its performance and
robustness.
III. 24-SLOT FOUR-POLE MOTOR
All results refer to a motor whose data are reported in Table I.
The operating condition of the motor refers to a rms conductor
(i.e.,
current density of 6 A/mm and a current phase
-axis current equal to -axis current).
A. Non-Chorded Winding
The first result refers to a non-chorded winding [9]. The best
solution found by the optimizer is

Fig. 4. Torque harmonics (24-slot four-pole motor, non-chorded).

with an average torque

Nm and a torque ripple
.
Fig. 3 shows the torque behaviors versus mechanical angle
due to the first couple of barrier angles ( , : Machaon 1),
due to the second couple of barrier angles ( , : Machaon 2),
and the resulting total torque. The reduction of the torque ripple
achieved by combining the effects of the two barriers is evident.
The torque harmonic contents are shown in Fig. 4.
The final choice of the optimization process has the following
interpretation: Two sets of angles are chosen such that the torque
harmonics of order 6 and 12 are quite low. In addition, they are
chosen so that they are out of phase of 180 . Therefore, when
the two torque components are considered together, the final
solution exhibits almost zero torque harmonics of such orders.
An interesting comparison can be made with two flux barriers
per pole having uniformly distributed tips. According to [4], the
angles are
and
.
The average torque in this case is 3.74 Nm, which is similar
to the one found by the optimizer. Conversely, the torque ripple
is much higher at 41%. Another solution derived from [4] yields
and
.
The average torque in this case is 3.99 Nm, once more similar
to the one found by the optimizer, but the torque ripple is much
higher at 31%.

960

IEEE TRANSACTIONS ON MAGNETICS, VOL. 47, NO. 5, MAY 2011

TABLE II
IMPACT OF THE NUMBER OF SLOTS

Fig. 5. Torque versus rotor angle (24-slot fourpole motor, chorded).

Fig. 6. Torque harmonics (24-slot four-pole motor, chorded).

optimization algorithm can be interpreted. In fact, the fundamental harmonic and all the slot harmonics are only slightly reduced, while the other harmonics are much more. This is quantified by means of the pitch factor.
The consequence of the reduction of the MMF harmonics
yields a reduction of the corresponding torque harmonics. In the
case of a four-pole 24-slot motor, all torque harmonics that are
multiples of 6 but not multiples of 12 (corresponding to the slot
harmonics) are reduced by the chording. Taking advantage of
the reduction of the former harmonics, the effort of the optimization process is mainly on selecting the flux barrier angles
so as to reduce the harmonics multiple of 12. This is evident
in Fig. 6. The torque harmonics of sixth order are low for both
and ,
(compare the
the combinations with angles ,
amplitude of these harmonics to those of Fig. 4 referring to a
non-chorded winding). On the contrary, the torque harmonics
of twelfth order are high for both angle combinations. However,
these harmonics are out of phase of 180 degrees, so their sum
yields a drastic ripple reduction.
As a conclusion, the adoption of a chorded winding allows a
reduction of all torque harmonics: In the example under study,
chording allows to further reduce the torque ripple from 11% to
8.8%.
D. Thickness of the Flux Barriers

B. Current Phase
Further optimizations have been carried out with increased
to 55 and 65 . In all cases, even if the average torque decreases,
as expected in the flux-weakening operating region, the torque
ripple remains almost the same. In fact, the harmonic contents
of the stator MMF distribution is only slight, and therefore the
torque ripple is expected to be almost the same.
C. Chorded Winding
The second set of results refers to a chorded winding with a
pitch angle of 75 electrical degrees. The optimized solution is
found with the following angles:

Fig. 5 shows the torque behaviors versus rotor angle due to

the first couple of barrier angles ( , : Machaon 1), due to
the second couple of barrier angles ( , : Machaon 2), and
the resulting total torque. The reduction of the torque ripple
achieved by combining the effects of the two barriers is evident.
The torque harmonic contents are shown in Fig. 6.
With chorded winding, the torque harmonics are lower than
in the previous case, and thus the torque ripple of this Machaon
IPM motor is extremely low. Also in this case, the choice of the

The thickness of the flux barriers has a very limited impact

on both average torque and torque ripple. Variations of up to
50% of the flux barrier thickness have been explored, finding
that optimal angles differ slightly, but in all cases the average
torque and the torque ripple remain equal to
Nm and
close to
, respectively. Once again, the adoption
of a chorded winding yields a reduction of the torque ripple to
about 8.5%.
E. Number of Slots
The number of stator slots has a major impact on the torque
ripple. This is because the amplitudes of the MMF harmonics
decrease as the number of slots per pole and per phase increase.
In addition, the slot harmonics have a higher order. The impact
, is reported in Table II for to the
of the number of slots,
four-pole motor. Increasing the number of slots noticeably decreases the torque ripple, which is more than halved when
is
doubled. This has been found for both non-chorded and chorded
windings. Table II reports the optimized torque values for dif(in electrical degrees).
ferent chording angles
F. Number of Pole Pairs
The effect of the number of pole pairs has been investigated,
exploring also
. The main geometrical dimensions remain

ALOTTO et al.: OPTIMIZATION OF INTERIOR PM MOTORS WITH MACHAON ROTOR FLUX BARRIERS

Fig. 7. Pareto front in a torque-ripple graph. x-axis: T

(a) Non-chorded winding. (b) Chorded winding.

, y -axis: T

961

as in Table I, while the number of slots has been increased to

. With a non-chorded winding, the optimized solution
is found with the following angles:

resulting in
timized solution is

. With a chorded winding, the op-

resulting in

Fig. 8. Comparisons between analytical model and FEM for different motor
configurations.

IV. MULTIOBJECTIVE SOLUTION

The assistance of PM insets within the flux barriers causes
the saturation of the ribs at the ends of the flux barriers, with
a consequent torque increase. Fig. 7 shows the Pareto front in
a plane whose coordinates are the average torque
and the
torque ripple
. Fig. 7(a) refers to a non-chorded winding,
and Fig. 7(b) refers to a one-slot chorded winding. It is worth
noticing that the front can be approximated by two straight lines,
one almost orthogonal to the other, so that a convenient solution
can be found at the intersection between the two lines where it is
possible to find small torque ripple with a high average torque.
V. COMPARISON TO FINITE ELEMENT ANALYSIS
Some optimal solutions (found based on the analytical model)
have been validated by means of a finite element (FE) analysis.
A linear rotor and a smooth stator with currents imposed in the
points corresponding to the center of the slots [10] are assumed.
Fig. 8 shows some results for different motor configurations.
The comparison confirms a satisfactory agreement between analytical and FE analysis. The main difference is a slight reduction of the average torque with respect to the FE analysis, but
the torque ripple remains the same, confirming the validity of
the analytical formulation that can therefore be used within the
optimization procedure.
VI. CONCLUSION
The low computational cost of an analytical model of the IPM
motor allows the optimizer to quickly find the optimal solution
of the single-objective problem or the Pareto front of the multiobjective one. Moreover, several scenarios with different design parameters (number of poles and slots per pole, chording)
have been investigated (see Table II) and further variants can be

easily explored. Results, in terms of minimal torque ripple, are

confirmed by FEM validation. The optimization procedure provides considerable improvements with respect to patented solutions under the same modeling hypotheses.
REFERENCES
[1] A. Fratta, G. Troglia, A. Vagati, and F. Villata, Evaluation of torque
ripple in high performance synchronous reluctance machines, in Rec.
IEEE Ind. Appl. Soc. Annu. Meeting, Toronto, ON, Canada, October
1993, vol. 1, pp. 163170.
[2] T. Jahns and W. L. Soong, Pulsating torque minimization techniques
for permanent magnet ac motor drivesA review, IEEE Trans. Ind.
Electron., vol. 43, no. 2, pp. 321330, Apr. 1996.
[3] A. Vagati, M. Pastorelli, G. Franceschini, and S. Petrache, Design of
low-torque-ripple synchronous reluctance motors, IEEE Trans. Ind.
Appl., vol. 34, no. 4, pp. 758765, Jul.Aug. 1998.
[4] A. Vagati, Synchronous reluctance electrical motor having a low
torque-ripple design, U.S. Patent 5,818,140, Oct. 6, 1998.
[5] N. Bianchi, S. Bolognani, D. Bon, and M. Dai Pr, Rotor flux-barrier
design for torque ripple reduction in synchronous reluctance motors,
in Conf. Rec. IEEE 41th Ind. Appl. Soc. Annu. Meet., Tampa, FL, Oct.
812, 2006, vol. 3, pp. 11931200.
[6] N. Bianchi, S. Bolognani, D. Bon, and M. Dai Pr, Torque harmonic
compensation in a synchronous reluctance motor, IEEE Trans. Energy
Convers., vol. 23, no. 2, pp. 466473, Jun. 2008.
[7] N. Bianchi, Analysis of the IPM motorPart I, analytical approach,
in design, analysis, and control of interior PM synchronous machines,
in IEEE IAS Tutorial Course Notes, IAS Annu. Meeting, CLEUP, N.
Bianchi and T. M. Jahns, Eds., Seattle, WA, Oct. 3, 2004, pp. 3.13.33.
[8] L. Coelho and P. Alotto, Tribes optimization algorithm applied to the
Loneys solenoid, IEEE Trans. Magn., vol. 45, no. 3, pp. 15261529,
Mar. 2009.
[9] M. Liwschitz-Garik and C. C. Whipple, Electric Machinery. New
York: Van Nostrand, 1960, vol. II, A-C Machines.
[10] N. Bianchi, Electrical Machine Analysis Using Finite Elements, ser.
Power Electronics and Applications Series, CRC. Boca Raton, FL:
CRC, 2005.