Mariusz JAGIEA1, Ernest A.

MENDRELA2
1

Opole University of Technology, Institute of Electromechanical Systems and Industrial Electronics

2
Louisiana State University, Electrical and Computer Engineering Department

Surrogacy-assisted back-emf optimization in PM-BLDC

in-wheel motor for operation with delta connection
Abstract. Delta connection of three-phase windings of brushless DC motors with the surface-mounted magnets contributes to rise of power loss due
to the zero-sequence voltage induced by triplen (3, 9, 15, 21, ) harmonics of main flux. Partial control over this undesired effect can be
accomplished by means of modification of winding distribution or an application of larger than normally angle of stack skew. This work attempts to
reduce these harmonics by means of application of small skew angle along with the magnetic circuit design using a finite element model. The
surrogacy-assisted two-objective genetic optimization of motor's magnetic circuit ensures small losses due to zero-sequence current and locates the
motor efficiency at 91 per cent of that of basic motor configuration with phases connected in wye.
Streszczenie. Skojarzenie pasm uzwoje w trjfazowych silnikach bezszczotkowych wzbudzanych magnesami trwaymi w trjkt skutkuje wzrostem
strat wywoanych skadow zerow siy elektromotorycznej rotacji indukowanej przez tzw. potrojone (3, 9, 15, 21, ) harmoniczne strumienia
gwnego. Redukcja tego zjawiska moe by osignita za pomoc odpowiedniego rozoenia cewek lub zastosowania wikszego ni normalnie
kta skosu rdzenia. W niniejszej pracy analizowana jest take moliwo redukcji skadowej zerowej prdu oraz zachowania wartoci parametrw
eksploatacyjnych maszyny metod optymalizacji obwodu magnetycznego. Optymalizacj przeprowadzono przy zastosowaniu wielokryterialnego
algorytmu genetycznego oraz metamodelu utworzonego metod Krigingu na podstawie oblicze polowych. W wyniku optymalizacji uzyskano silnik o
uzwojeniu skojarzonym w trjkt, ktrej sprawno jest mniejsza o 9 % od sprawnoci maszyny pracujcej z uzwojeniem skojarzonym w gwiazd.
(Minimalizacja harmonicznych napicia rotacji zerowej kolejnoci faz w silniku BLDC z wykorzystaniem metamodelu)

Keywords: permanent magnet motors, triplen harmonics, multiobjective optimization, metamodel.

Sowa kluczowe: silniki bezszczotkowe, skadowe symetryczne, optymalizacja wielokryterialna, metamodel.

Introduction
The surface-mounted permanent magnet (SMPM)
brushless DC (BLDC) motor is the most commonly used
type of PM brushless machine. It generates quasitrapezoidal back-emf waveform and performs best driven
from static converters providing quasi-rectangular phase
currents. The latter is realized traditionally via connection of
phases in wye and providing the 120o current conduction
period through phase windings.
Some applications, like the battery-powered appliances,
often do not provide an in-place DC voltage supply of
a preferred level. In such a case the most natural way of
driving motor from a lower voltage is to reconfigure phases
connection into delta [1], [2], [3]. Though, it is well known
that this relates with the occurrence of the zero-sequence
current component that encloses in the delta circuit [2]. It is
caused by the triplen harmonics of the back-emf. Typically,
the ratio of third-to-fundamental harmonic of back-emf is
between 20 and 30 per cent [1], [2], [3], [4]. This harmonic
adds significant ohmic power loss and produces only an
unfavorable braking torque [2]. Unlike the concentrated or
modular windings, which, due to a rich set of the back-emf
waveform harmonics [3], should not be considered for
operation with delta connection, the distributed windings are
often connected in delta as a restraint of the detrimental
impact of the triplen harmonics is easier. This is, however
restricted practically to the fractional-power machines.
The ability to control this effect by the latter winding type
relates with the use of larger number of slots that allows
modifications of coils' distribution as well as provides more
control over the magnitudes of higher air-gap flux density
harmonics. In motors with interior-permanent-magnet rotors
the distribution of air-gap flux density can be shaped by
adjusting angular variation of an equivalent air-gap
permeance. In this way the magnitudes of harmonics with
ordinals less than first due to slotting, can be reduced to
some extent. There are examples of such the designs
carried out to this date [1], [2], [3], [4], [5]. However, the
surface-mounted magnets limit that possibility due to nearly
unity relative magnetic permeability.

The content of the triplen harmonics in the phase backemf waveform can also be controlled by distribution of coils
so as to block propagation of detrimental flux density
harmonics, and also by means of an appropriate stator
a)

b)

Fig. 1. Physical motor: a) in-wheel assembly, b) winding and

skewed stator.

stack skew angle.

Because the effect of the above modifications may not be
satisfactory, this work also investigates the possibility of
controlling the content of triplen back-emf harmonics via
optimization of the magnetic circuit.
First part of this work presents a discussion on possible
winding distributions and stack skew angles in the considered motor. Further sections aim at redesigning the magnetic circuit using a finite element model and a multiobjective genetic algorithm connected with a metamodel.
1. Physical motor
The analysis is addressed to a three-phase outer-rotor
motor, designed for an in-wheel drive of a wheelchair (see
Fig. 1 and Table 1). In a basic application it operates with
the winding connected in wye and is supplied from a 24 V
battery. It has a one-layer distributed winding. This work
analyses the design variation that must be supplied from a
12 V battery having phases connected in delta, whilst the
requirements regarding the ratings remain the same.

PRZEGLD ELEKTROTECHNICZNY, ISSN 0033-2097, R. 89 NR 9/2013

257

Table 1. Specifications of basic motor with wye-connected winding

supplied from 24 V battery.
Parameter

Unit

Value

p
Number of slots
Turns per slot
Outer/Inner
diameter
Air-gap length

14
42
18

mm/mm

194/110

mm
rad
mm
rpm
A
%
rpm
Nm
Nm

0.5
2/42
40
196.0
7.1
83.0
164.5
1.2
41.5

sk
Length
n0
I10

10

n10
Tcmax

max

2. Motor model
a) Finite element modeling
To compute the motor quantities a comprehensive
quasi-three-dimensional, multi mesh-slice [6] (see Fig. 2),
circuit-driven finite element model of motor, that includes an
inverter drive and core losses, was elaborated. Equations of
the model can be written in form:
(1)

b) Zero-sequence components of voltage and current

Recalling the principles of dq0 theory for AC machines,
the zero-sequence voltage and the zero-sequence current,
are to be determined from formulas
(2ab)

Where
1 1 1 . The quantities of dq axes are out of
the considerations. Additional power loss due to zerosequence components, calculated as

Fig. 3. Calculated and measured waveforms of phase back-emf

and its zero-sequence component at rotational speed equal to 170
rpm.

3. Reduction of detrimental harmonics

a) Basic motor with delta connection
Table 2 summarizes quantities of motor with delta
connection supplied from a 12 V battery, obtained from
model outlined in section 3.
Table 2. Computed quantities for motor with basic single-layer
delta-connected winding supplied from 12 V battery.
Quantity
n10
n0
P0

max

will be used further to characterize various design versions

of motor.
A characteristic state of operation related with delta
connection is a continuous current conduction through
winding, even though the conduction periodS of the
o
converter switches are equal to 120 (apart from the
existence of the zero-sequence current). Consequently, it is
worth noticing that the efficiency of motor with delta

258

Fig. 2. Magnetic field distribution over mesh slices of a skewed

motor at a given instant of time using periodic symmetry with
respect to two poles.

where: S is the finite element matrix due to material

reluctivity, G is the finite element matrix due to material
conductivity, Q is a matrix that couples the phase linkage
fluxes with their circuit equations, R and L are matrices of
circuit resistances and inductances, respectively, denotes
stator length. Variables and are, respectively the vector
of unknown magnetic potentials attributed to a single meshslice and the vector of loop currents. Vector
represents
equivalent currents due to permanent magnets, and vector
e represents the voltage sources supplying the system [6].
Bracketed matrices are block-diagonal components of
global system of equations that result from skew
approximation by the mesh-slices. Equations (1) are
integrated using backward Euler time-stepping formula.
Torque is calculated using the method of local forces [7].
Validation of model against measurements of the phase
back-emf and the zero-sequence component of phase
back-emf is shown in Fig. 3.

(3)

connection cannot be the same as that of motor with wye

is
connection, even if the additional power loss
completely suppressed.

Unit

Value

%
rpm
A
rpm
W
Nm

69.0
135.6
9.6
171.5
19.6
34.3

As it can be deduced from results (see Tables 1 and 2),

nearly ten percent of input power is consumed by the power
loss due to triplen harmonics, which results in efficiency
being only eighty three percent of that of motor in Table 1 .
b) Modification of winding distribution
With the basic one-layer, fully-pitched winding depicted
in Fig. 4a the motor back-emf "mirrors" all of the air-gap flux
density harmonics. However, the unfavorable triplen
harmonics can effectively be suppressed using a special
two-layer winding having one coil split into two identical
coils (see Fig. 4b). The coils are shifted exactly by one-sixth
of the fundamental harmonic wavelength, therefore the
winding provides natural cancellation of all triplen back-emf
harmonics.

PRZEGLD ELEKTROTECHNICZNY, ISSN 0033-2097, R. 89 NR 9/2013

is equal to 0.955 only due to skew because both the onelayer and the above considered two-layer winding have
unity fundamental distribution and coil-span factors.
To reduce the triplen harmonics the short-pitch coils
could also be considered. In such the case the zero value of
3. In such the case
can be obtained for
is
also completely cancelled, but the corresponding
fundamental coil-span factor drops to 0.866. In order to get
rid of the cogging torque, the stator stack must be used,
which further reduces
to 0.827.
The conclusion is that an application of the considered
two-layer winding is one possible solution of the problem,
but shortening the coil-pitch of the basic winding, which
must be done by at least one-sixth of the fundamental
wavelength, results in too severe reduction of the
fundamental winding factor.

a)

b)

Fig. 4. Winding layout: a) one-layer winding arrangement in basic

motor, b) equivalent two-layer arrangement.

Table 3 summarizes motor quantities computed for the

motor operated with such the winding. As it can be noticed,
with the triplen harmonics completely cancelled the
efficiency rises to 91 percent of that for motor operating with
wye connection (see Table 1).
Table 3. Computed quantities for motor with modified two-layer
delta-connected winding supplied from 12 V battery.
Quantity
n10
n0
P0

max

Unit

Value

%
rpm
A
rpm
W
Nm

75.8
156.5
8.1
205.0
0.0
38.0

c) Application of skew
Reduction of the triplen harmonics can also be realized
via application stack skew. The functioning of skew is
twofold because alongside filtering the back-emf harmonics
it must reduce the cogging torque. Here, for zero values of
,
must be equal to double of a slot-pitch. Although
physically possible, such the value of
is too large as the
corresponding fundamental winding factor would be only
0.827. With the basic stack skew angle, equal to one slotpitch, the fundamental winding factor rises to 0.955, but the
winding factors of triplen orders are affected only slightly.
With respect to reduction of the cogging torque in the
considered motor, much better results can be obtained
using skew angle equal to one slot-pitch. A solution
preferred in larger machines and those with short stack
length, is to skew the magnets using segmentation of poles,
as shown in Fig. 6 [8], [9].
We analyze magnets split into four segments having the
displacement angles for the last two segments mirrored with
respect to those of the first two segments, as denoted in
Fig. 6. The segments have the same pole fill-factor as the
basic motor.

Fig. 6. Pole segmentation to approximate skew

Fig. 5. Results of computations: a) zero-sequence back-emf at noload operation, b) line current at no-load operation.

Figure 5 depicts variations of zero-sequence voltage

and no-load line current in motor with basic (one-layer) and
the modified (two-layer) winding. It can be seen that
although the zero-sequence current is cancelled the no-load
line current is still significant. This is the result of odd backemf harmonics of orders 5, 7, 11, ... whose magnitudes are
not affected by the modified winding.
In this motor the total fundamental winding factor

The computed maps of variations of peak values of the

zero-sequence voltage and cogging torque with 1 and 2
are shown depicted in Fig. 7.
It can be noticed that locations of the extrema of these
variations are entirely opposite, i.e. the zero-sequence
voltage has its minimum where the cogging torque has
minimum, and vice versa. Although in this figure we marked
five candidate solutions having significantly reduced
magnitude of
, the cogging torque is too high for any
practical application.
This consideration can be concluded such that practical
stack skew angle contributes to rise of motor efficiency, but
the triplen harmonics are not completely cancelled.
Skewing, using segmented poles, cannot be applied if high
ratio of reduction of the zero-sequence voltage and cogging
torque is expected.

(4)

PRZEGLD ELEKTROTECHNICZNY, ISSN 0033-2097, R. 89 NR 9/2013

259

estimates the percentage of the magnitude of third

harmonic with respect to the fundamental harmonic of radial
component of an air-gap flux density, at the centre of a
skewed machine measured in axial direction (see Fig. 8b).
It can be seen that this reduces the third harmonic of airgap flux density more effectively with higher pole-pitch fill
factors. This promises better utilization factor of motor.
Reduction of harmonics excited by an air-gap permeance
variation as the rotor rotates requires re-designing the slot
tooth-tips as well (see Fig. 8a). In this case search for the
optimal dimensions of the magnetic circuit must be done via
numerical optimization. Besides of reduction of voltage , it
is necessary to guarantee that the performance of motor will
not be worse than before optimization. Hence, the
optimization problem becomes two-objective.

Fig. 7. Results of computations: a) variation of u0 at no-load

operation in Volts b), maximum value of cogging torque vs. angles
of segmented poles in Nm.

c) Re-designing magnetic circuit with basic winding

Generally speaking, in motor with the basic one-layer
winding the triplen back-emf harmonics can be reduced by
reducing the magnitudes of spatial harmonics of an air-gap
flux density of the same ordinals. In a simplest manner this
can be realized by adjustment of pole fill-factor.
a)

4. Optimization
The use of metamodels, also known as surrogates, in
optimization of computationally demanding problems
described by the Maxwell equations, is much more efficient
than evaluation of the complete computer models [10]-[15].
To realize the goal of this work, the two responses of the
computer model f, presented in section 3 (formula (1)), with

(where
an input variable
is the feasible region), must be taken into account. These
responses are:

Fig. 8. Re-designing the magnetic circuit relying upon slanting

edges of poles and adjustment of dimensions of the stator tooth
tips: a) motor layout and design variables, b) percentages of third to
fundamental harmonic of radial component of an air-gap flux
density vs. relative pole width for two different pole shapes.

An attempt of suppression of the triplen back-emf

harmonics in this way was made in [1], where it was also
shown that the results may not be satisfactory in some
applications. Fig. 8a-b illustrates how this goal can be
realized via application of the magnets with slanted edges.
To justify the above we carried out an analysis which

260

where

is the maximum value of phase back-emf, and

is the maximum value of zero-sequence component
of the back-emf. Their corresponding surrogate models are:

(6a,b)

b)

(5)

where
denotes the Kriging metamodel [14]. In this case
the universal Kriging was used with a 0th order polynomial
regression model and a linear function to fit the variogram.
In order to generate the grid points for evaluation of
responses, the computer experiment was designed using
the Latin hypercube sampling [11], [13]. Due to costly
evaluation of responses, requiring calculation of at least one
half-period of the phase back-emf, the number of sample
points was kept as small as possible, whilst the quality of
local approximation was observed during a few separate
numerical experiments. For the considered model the least
possible number of sampling points over the feasible region
is equal to 25, requiring only 25 executions of model f. An
entire process of creation of the metamodel on a laptop PC
computer equipped with a 1,83 GHz dual core processor
and 3 GB RAM took approximately 170 minutes.
The further computational problem regards minimization
of the zero-sequence component of the back-emf and also
an absolute value of difference between maximum value of
the phase back-emf and its given value
. The latter
provides that the performance of the optimized motor will
not be worse than prior to optimization. Hence, the
objectives are:
(7a,b)

The problem for optimization is to find

(8) min

such that

subject to

7/90, /9, 0.9, 1.2 ,

/9, 43/300,
1.0, 3.0 being, respectively the vector of lower and upper

with

was
bounds of the design variables. In calculations
set equal to 10.5 V, which corresponds with the maximum

PRZEGLD ELEKTROTECHNICZNY, ISSN 0033-2097, R. 89 NR 9/2013

value of the back-emf in the motor operating at speed equal

to 170 rpm, prior to optimization (see Fig. 3).
Problem (8) was solved using an elitist multiobjective
genetic algorithm based on non-dominated sorting
approach [15]. The initial population size was set equal to
60. A maximum relative change in fitness function, equal to
0.5 %, was used to terminate the optimization. Three
executions of the optimization procedure were carried out
from among which the best Pareto set was selected (see
Fig. 9). Naturally, the solutions on the Pareto front are
equally optimal in sense of the fitness function (8). This
form of objective function was assumed in order to avoid
additional costly computations related with considering
loaded machine, although it does not represent the loaded
motor ideally (nonlinearity, reaction of armature). Moreover,
it is necessary to guarantee, that the motor operating at noload and close to or recovering energy during downhill ride,
will generate ohmic losses as small as possible. Therefore,
we feel justified to favor a set for which is closer to zero.

Through comparison of results summarized in Tables 3

and 4 it can be seen that the optimization has led to almost
the same efficiency as in case of use of the special twolayer winding (shown in Fig. 4b) in the motor having the
initial (unaffected) magnetic circuit.

Fig. 10. Results of computations: a) zero-sequence back-emf at noload operation, b) line current at no-load operation.

Fig. 9. Distribution of equally optimal solutions according to the

Pareto front.

As a consequence, for the candidate designs, we

selected the first left five solutions. Table 4 summarizes the
quantities calculated for these designs.
From our point of view the first candidate should be
most favored due to smallest
and the highest rated
speed. In this design the zero-sequence voltage is reduced
by the factor of ten as compared to the initial design (see
Fig. 10). Table 5 compares dimensions of the magnetic
circuits of motors, before and after optimization. The
corresponding shapes of the magnetic circuits are shown
depicted in Fig. 11. A positive side-effect of the magnetic
circuit modification for operation with delta connection is the
reduction of normal forces (see Fig. 11).

Motor with the optimized magnetic circuit and the basic

winding has much smaller no-load line currents (see Fig.
10b) resulting in smaller no-load losses than motor with the
modified two-layer winding. This result corresponds with the
fact that the re-designing the magnetic circuit affects all of
the back-emf harmonics, including those having orders 5, 7,
11, etc. Apart from that, it can be said that both motors are
nearly equally efficient operating with nominal load and
consequently, there are two the most effective ways to
reduce the detrimental zero-sequence component of phase
back-emf in the considered motor.
a)

Table 4. Quantities for five equally optimal candidate designs of

motor with delta-connected winding.
Quantity
n10
n0

max

Unit
%
rpm
A
rpm
W
Nm

b)
Force [N]

Number of candidate according to Fig. 10

1
2
3
4
5
75.4
75.5
75.7
75.8
75.8
153.5
151.8
150.3
149.6
148.2
8.4
8.3
8.3
8.3
8.2
203.0
199.7
196.7
195.2
193.2
0.2
0.4
0.5
0.6
1.0
32.6
32.6
32.6
32.4
32.3

Table 5. Dimensions of magnetic circuit before and after

optimization.
After optimization
Variable
Unit
Before optimization
(candidate 1)
/9
5/64
rad
/9
/9
rad
mm
0.5
1.1
mm
1.7
3.0

Fig. 11. Shapes of motor magnetic circuit considering distributions

of local forces over motor cross-section (central mesh-slice in axial
direction): a) before optimization (motor for operation with wye
connection), b) after optimization (motor for operation with delta
connection).

PRZEGLD ELEKTROTECHNICZNY, ISSN 0033-2097, R. 89 NR 9/2013

261

Conclusion
It has been shown that there are two the same efficient
methods of reducing the detrimental impact of triplen backemf harmonics PM-BLDC motors with the surface-mounted
magnets.
These methods rely upon usage of a special two-layer
winding or re-designing the magnetic circuit. It is apparent
that re-winding the stator is a standard technical process,
which requires neither detailed investigation nor a specific
knowledge. However, the direction towards optimization of
the magnetic circuit might be more advantageous in mass
production of motors. The motor with the optimized
magnetic circuit has some additional advantages like very
small no-load ohmic losses, reduced normal forces and no
permanent-magnet edges subjected to demagnetization by
the armature reaction field. This might be especially
advantageous in larger machines.
APPENDIX
The winding factors for motors with three-phase distributed
windings are given by:
(9)
(10)

cos

(11)
Nomenclature
winding distribution factor of order
coil-span factor of order
skew factor of order
n0
rotational speed at no load operation, rpm
rotational speed under 10 Nm load torque, rpm
n10
vector of phase currents, A
rms phase current under 10 Nm load torque, A
I10
number of pole-pairs, p
max electromagnetic torque at zero speed, Nm
cmax peak value of cogging torque, Nm

period, s
vector of phase voltages, V
skew anglerad
shift angle of q-th pole segment (q=1,2)rad
coil-cut anglerad
10 efficiency under 10 Nm load torque, including
inverter and core loss, %

pole-pitch, rad
REFERECES
[1] Y. B. Kim, H. S. Choi, C.-S. Koh, "A back EMF Optimization of
Double Layered Large-Scale Motor", IEEE Trans. Magn., Vol.
47, No. 5, pp. 908-1001, 2010.

262

[2] E. Schmidt, M. Susic, A. Eilenberger, "Design Studies on a

Permanent Magnet Synchronous Machine with Y and Dconnected Stator Winding", IEEE Trans. Magn., vol. 47, No. 5,
pp. 1042-1045, 2010.
[3] K. Atallah, J. Wang, and D. Howe., "Torque ripple minimization
in modular permanent magnet brushless machines, IEEE
Trans. Industry Appl., Vol. 39, No. 6, pp 1689-1695, 2003.
[4] G. Ombach , J. Junak:, Torque ripple optimization of skewed
IPM motor for field weakening operation, in Proc. of Int. Conf.
on Electrical Machines and Systems (ICEMS'11), pp. 1-7,
2011.
[5] R. Wrobel, P. H. Mellor, "Design considerations of a direct drive
brushless machine with concentrated windings", IEEE Trans.
Energy Conv., Vol. 23, no. 1, pp. 1-8, 2008.
[6] A. M. Oliveira, R. Antunes, P. Kuo-Peng, N. Sadowski, P.
Dular, "Electrical machine analysis considering field-circuitmovement and skewing effect", COMPEL, Vol. 23, No. 4, pp.
1080-1091, 2004.
[7] I. Nishguchi, A. Kameari, K. Haseyama, "On the local force
computation of deformable bodies in magnetic field", IEEE
Trans. Magn., Vol. 35, no. 3, pp. 1650-1653, 1999.
[8] G. Dajaku and D. Gerling, Skewing effect on the PM fluxlinkage high harmonics of the PM machines with delta winding,
in Proc. 13th Eur. Conf. Power Electronics and Applications
(EPE09), pp. 17, 2009.
[9] Lukaniszyn M, Jagiela M, Wrobel R., "Optimization of
permanent magnet shape for minimum cogging torque using a
genetic algorithm", IEEE Trans. Magn., Vol. 40, No. 2, pp.
1228-1231, 2004.
[10] G.
Pellegrino, F. Cupertino,
FEA-based
multi-objective
optimization of IPM motor design including rotor losses in
Proc. of Energy Conversion Congress (ECE'10), pp. 36593666, 2010.
[11] J.B. Kim, K.Y. Hwang, B. I. Kwon, "Optimization of Two-phase
In-wheel IPMSM for Wide Speed Range by Using the Kriging
Model Based on Latin Hypercube Sampling", IEEE Trans.
Magn., Vol. 47, no. 5, pp. 1078-1081, 2010.
[12] A. Kobetski, J. L. Coulomb, M. C. Costa, Y. Merchal,
U.
Jnsson, "Comparison of radial basis function approximation
techniques", COMPEL, Vol. 22, No. 3, pp. 616-629, 2003.
[13] J. Sacks, W. J. Welch, T. J. Mitchell, H. P. Wynn, "Design and
analysis of computer experiments", Statist. Science, Vol. 4, No.
4, pp. 409-435, 1989.
[14] G. Crevecoeur, P. Sergeant, L. Dupr, R.
Van de
Walle,
"A Two-level Genetic Algorithm for Electromagnetic
Optimization", IEEE Trans. Magn., Vol. 46, No. 7, pp. 25852595, 2010.
[15] K. Deb, A. Pratap, S. Agarwal, and T. Meyarivan, A Fast
Elitist Multiobjective Genetic Algorithm: NSGA-II, IEEE Trans.
Evol. Comp., Vol. 6, No.2, pp. 182-197, 2002.
Authors: dr. Mariusz Jagiela, D. Sc. (Eng.), Opole University of
Technology, Institute of Electromechanical Systems and Industrial
Electronics, ul. Proszkowska 76, 45-758 Opole, Poland, E-mail:
m.jagiela@po.opole.pl; dr. Ernest A. Mendrela, D. Sc. (Eng.),
Louisiana State University, Electrical & Computer Engineering
Department, Baton Rouge, LA 70808, USA, E-mail:
ermen1@lsu.edu.

PRZEGLD ELEKTROTECHNICZNY, ISSN 0033-2097, R. 89 NR 9/2013