sensors

Article
Torque Ripple Minimization of the Permanent
Magnet Synchronous Machine by Modulation of the
Phase Currents
Cezary J˛edryczka * , Dawid Danielczyk and Wojciech Szelag
˛
Institute of Electrical Engineering and Electronics, Poznan University of Technology, 60-965 Poznan, Poland;
dawid.p.danielczyk@doctorate.put.poznan.pl (D.D.); wojciech.szelag@put.poznan.pl (W.S.)
* Correspondence: cezary.jedryczka@put.poznan.pl




Received: 24 March 2020; Accepted: 20 April 2020; Published: 23 April 2020


Abstract: This paper deals with the torque ripple minimization method based on the modulation of
the phase currents of the permanent-magnet synchronous motor (PMSM) drive. The shape of the
supply current waveforms reducing the torque ripple of the machine considered was determined on
the basis of finite element analysis (FEA). In the proposed approach, the machine is supplied by a
six-leg inverter in order to allow for the injection of zero sequence current harmonics. Two test PMSMs
with fractional-slot concentrated windings (FSCW) and surface-mounted permanent magnets (SPMs)
have been examined as a case study problem. Wide-range fractional analyses were performed using
developed numerical models of the electromagnetic field distribution in the considered machines.
The results obtained show that the level of torque ripple in FSCW PMSMs can be effectively reduced
by the modulation of the phase currents under the six-leg inverter supply.

Keywords: permanent-magnet synchronous motor; torque ripple minimization; finite element analysis

1. Introduction
It can be stated that the progress in the development of electrical machines (EMs) in recent
decades has mostly been related to the improvements and the progress made in the development of
the materials used in their magnetic circuits—like permanent magnet materials with high magnetic
energy density and low-loss electrical steel sheets, or soft magnetic composites—allowing for the
operation of electrical machines at higher frequencies. On the other hand, the dynamic progress in
computational methods and the power of computers has allowed for precise analysis of electrical
machines’ performance. In particular, the Finite Element Method (FEM) is one of the widely used
methods employed for determining electromagnetic field distribution, which has led to progress in the
design methods of EMs. The availability of the professional tools for the combined analysis of coupled
electromechanical and thermal phenomena allows for the “optimal” design of electrical machines.
It can be stated that “state of art” design methods and tools for the analysis of electrical machines’
performance allow us to synthesize construction characterized by high efficiency and/or power density
and/or low cost of production, properties which should be treated as goals or constraints during
the design process. Currently, besides the above discussed figures of merit, the quality of energy
conversion in electrical machines is becoming a point of interest for many research and design teams.
The reduction in unwonted phenomena related to energy conversion processes—like the influence on
the grid (higher distortion of currents or voltages in the case of generators), noise and vibration, as well
as a reduction in torque ripples—is within the scope of research activities and efforts taking place in the
design stage of modern EMs. This paper is focused on one of the discussed aspects of energy conversion
quality, i.e., torque ripple minimization in permanent-magnet synchronous motors (PMSMs). The need

Sensors 2020, 20, 2406; doi:10.3390/s20082406 www.mdpi.com/journal/sensors

Sensors 2020, 20, 2406 2 of 11

to provide high quality energy conversion in electrical drive systems is obvious. Lowering the value of
the electromagnetic torque ripple decreases the levels of vibration and noise, as well as extending the
life expectancy of the drive system. In many direct drive applications, such as elevator propulsion
units, providing high quality torque (without the torque ripples) is essential. Therefore, reducing the
torque ripples in electrical drive systems has been a point of interest for many research and design
teams in recent decades. In this type of research, two major trends can be observed: (1) the reduction
in torque ripples by the control system side [1–5] and (2) the improvement of the torque quality at the
design stage by the optimization of the electromagnetic circuit of the electrical machine. In the first
trend, the current harmonic injection is one of the commonly applied techniques [1–3]. Due to the
technical difficultly of directly measuring the torque ripple in such approaches, torque observers are
usually employed [2]. To reduce the torque ripples, harmonics are injected into the control system
at the level of the rotary reference currents; this is mainly done by modifying the torque producing
current iq . To find the iq harmonics that will reduce the torque ripples, many different optimization
techniques and algorithms are used. For example, the Cuckoo behavior-based algorithm has been
successfully employed using the method described in [1], while the neural network (NN)-based
parameter estimators have been applied to reduce the torque ripple in motor drives using the approach
described in [2]. Nevertheless, the model-based predictive current control methods suffer from a need
for either complex algorithms for model parameter identification or a high amount of data for training
the NN [6]. Consequently, problems of the reliability of the measured input data and the desired high
control loop frequency arise.
On the other hand, the origins of the fluctuations in electromagnetic torque values lie inside the
machine. In general, the torque level, as well as it ripples, are effects of the Maxwell stress tensor
distribution inside the air gap of the machine, which depend on the interaction between the magnetic
field’s temporal and spatial distributions excited by rotor magnets and the magnetomotive force (mmf )
of the temporal and spatial distributions, which is generated by the winding. Therefore, the second
group of studies tend to minimize the torque ripple and focus on the design of the machine’s magnetic
circuit. Among others, these methods were based on the stator and rotor skewing, the profiling the
magnetization of the permanent magnets, as well as introducing non-uniform air gap lengths or
increasing the number of phases employed in order to suppress the cogging torque, as well as to reduce
the torque ripple level of the machine [7,8].
The study presented in this paper is in some ways rooted in both of the above discussed trends in
the research. As discussed in [9], the reduction in torque ripples by supply current shaping has already
been investigated by, among others, Razek in 1990 [10]. In terms of the torque ripple minimization
methods given in [9], the difficulties in practical realization have been pointed out as being the main
disadvantages of such methods. It should be noted that the injection of 3rd integer multiplier harmonics
(triplen harmonics) is not possible in PMSMs driven by conventional three-leg (see Figure 1a) inverters
when the phase windings of the machine are arranged in a star connection. On the other hand,
due to the decreasing cost of electronic power switches, the six-leg inverters are now an interesting
alternative to three-leg systems in terms of increased reliability [11–13]. The application of knowledge
from the literature regarding the dual-fed open-ended winding PMSM (see Figure 1b) enables the
development of new, more effective methods for improving the energy conversion quality in electrical
machines (by modulating the phase currents by higher harmonics and, in particular, by injecting
zero-sequence harmonics).
 Sensors 2020, 20, x FOR PEER REVIEW 3 of 11
Sensors 2020, 20, 2406 3 of 11

(a) (b)
Figure
Figure 1. 1. Conventional3-leg
Conventional 3-leginverter
inverterdriving
driving a permanent-magnet
permanent-magnet synchronous
synchronousmotor
motor(PMSM) (a)(a)
(PMSM)
and
and dual-fed
dual-fed open-endedwinding
open-ended windingPMSM
PMSM(b).
(b).

2. Torque Ripple
2. Torque Minimization
Ripple MinimizationofofPMSM
PMSMby
by Current
Current Modulation Method
Modulation Method
In In
contrast
contrast totoother
otherapproaches
approachesexploiting
exploiting phase currentmodulation,
phase current modulation,ininthe theproposed
proposed method,
method,
to find the amplitudes and phases of the current harmonics that will allow us to
to find the amplitudes and phases of the current harmonics that will allow us to improve the energy improve the energy
conversion
conversion quality
quality in the
in the PMSM,
PMSM, the precise
the precise assessment
assessment oftorque
of the the torque
rippleripple is carried
is carried out byoutan by an
analysis
analysis
of the of the
magnetic magnetic
field field distribution,
distribution, determined usingdetermined using FEM.
FEM. Numerical Numerical
models basedmodels
on FEM,based
due toonthe
FEM,
high due toresulted
reliability the highfrom reliability
solving resulted
partial from solving
difference partial describing
equations difference equations describing
the magnetic the
field in space
magnetic field in space and time domains, are now commonly used at the
and time domains, are now commonly used at the design stage of EMs [14–17]. In general, exploiting design stage of EMs
[14–17]. In
FEM-based general,
models exploiting
allows FEM-based
researchers to avoidmodels
buildingallows
a highresearchers to avoid building
number of expensive prototypes,a high
which
number of expensive prototypes, which would otherwise be needed to verify the design process. In
would otherwise be needed to verify the design process. In the studied problem of torque ripple
the studied problem of torque ripple minimization, the “optimal” shape of the phase current
minimization, the “optimal” shape of the phase current waveforms could also have been determined
waveforms could also have been determined by using simpler and less computationally expensive
by using simpler and less computationally expensive analytical methods, allowing the torque ripple
analytical methods, allowing the torque ripple level assessment to take place. Nevertheless, to
level assessment to take place. Nevertheless, to superimpose the rotor and stator fields (formed by
superimpose the rotor and stator fields (formed by each phase current) as proposed in the field
each phase current) as proposed in the field reconstruction method [18], for example, the magnetic
reconstruction method [18], for example, the magnetic saturation effect must be neglected.
saturation effect must
Consequently, the be neglected.
resulting Consequently,
current waveforms themay
resulting
not current
minimizewaveforms
the torquemayripple
not minimize
level
thesignificantly
torque ripple level significantly [19]. To mitigate the above discussed problem
[19]. To mitigate the above discussed problem in the proposed method, the calculation in the proposed
method,
of the the calculation
torque of the(treated
ripple factor torque asripple factor
a goal (treated
function for as
theacurrent
goal function
waveform for the current
shape waveform
optimization
shape optimization
process) of the two process) of the two
components of components of the electromagnetic
the instantaneous instantaneous electromagnetic
torque of a PMSM torque of a
are
PMSM are determined [20] on the basis of the finite element analysis
determined [20] on the basis of the finite element analysis (FEA) of the machine: (FEA) of the machine:
( ) + TTr (t()t)
T (Tt) t==TT00 + (1) (1)
r
where T0 is the constant (average) component and Tr is a periodic component dependent on time or
where T0 is theangle.
the electrical constant (average) component and Tr is a periodic component dependent on time or
the electrical angle.
There are several, known from the literature, definitions of the torque ripple factor (TRF) which
areThere
used are
[20]several, known from
in the analysis of thethe literature,
torque qualitydefinitions of the
of electrical torque ripple
machines. In the factor (TRF ) which
most commonly
areapplied
used [20] in the the
approach, analysis
torqueof the torque
ripple factor isquality
calculatedof electrical machines.
as the difference In the
between the most
maximumcommonly
and
applied approach,
the minimum the in
value torque
torqueripple factor
T, when is calculated
related as thevalue.
to its average difference between
However, the maximum
in practice, when theand
thetorque
minimum value in
is measured ortorque T, when
calculated relatedmanner
in a discrete to its average
(e.g., in value. However,
the proposed methodin practice,
exploiting when
FEM)the
the proper
torque determination
is measured of the
or calculated in amaximum and minimum
discrete manner (e.g., in values of the method
the proposed torque isexploiting
difficult. FEM)
To
themitigate the above discussed
proper determination of theproblem,
maximum in the
andpresented
minimum work, theof
values torque rippleislevel
the torque is expressed
difficult. by
To mitigate
the root mean square (RMS) value of T , related to its average value, T :
the above discussed problem, in the presented work, the torque ripple level is expressed by the root
r 0

mean square (RMS) value of Tr , related to its average T value, T0 :

TRF = r RMS ⋅ 100% (2)
T0
TrRMS
TRF = ·100% (2)
where T0 and the RMS value of Tr are calculated T0for an electrical period of supply voltage at the
steady-state operation of the studied machine.
where T0 and the RMS value of Tr are calculated for an electrical period of supply voltage at the
To determine the value of the instantaneous electromagnetic torque T(t) of a PMSM, the integral
steady-state operation of the studied machine.
of the tangential ft component of the Maxwell stress tensor, over the cylindrical surface of the radius
r and length Li, placed in the air gap of the machine is calculated:
Sensors 2020, 20, 2406 4 of 11

To determine the value of the instantaneous electromagnetic torque T(t) of a PMSM, the integral
of the tangential ft component of the Maxwell stress tensor, over the cylindrical surface of the radius r
and length Li , placed in the air gap of the machine is calculated:

Z2π
T (t) = rLi ft (t, ϑ)dϑ (3)
0

where the angle ϑ describes the location of the dϑ on the circumference of the surface.
The tangential component of the Maxwell stress tensor ft is expressed by the normal (Bn ) and
tangential (Bt ) components of the magnetic flux density vector B in the air gap of the machine, calculated
by the FEM:
1
ft (t, ϑ) = Bn (t, ϑ)Bt (t, ϑ) (4)
µ0
In general, the distributions Bt (ϑ) and Bn (ϑ) can be represented by complex periodic functions that
depend upon the magnetic circuit structure of the machine, i.e., the number of poles, the number of
stator slots, the magnetization pattern and geometry of the magnets, and the geometric and magnetic
properties of the stator. Additionally, they are also time-dependent functions defined by winding
structures and their phase currents. The interaction between the temporal and spatial distributions
of these functions contributes to the generation of the average T0 and pulsating Tr electromagnetic
torque components. Despite the high computational complexity, only the application of the numerical
FEM model on the machine allows all of the abovementioned dependencies for torque ripple level
assessment to be taken into account.
In the proposed approach, it is assumed that the machine is supplied by the phase current
waveforms, represented by this sum of the harmonics:

l
2π 2π
  X    
ik (t) = Im1 sin ωt + (k − 1) + Imn sin n (ωt) + (k − 1) + ϕn (5)
3 3
n=3

where k = 1, 2, 3 for the phase a, b, c, respectively, n is an odd integer greater than 1, l is the maximum
harmonic order taken into consideration, ϕn is the phase of the n harmonic and Im1 is the amplitude of
the fundamental harmonic of the phase current.
The amplitudes Imn are defined by their percentage IPn content in relation to the Im1 :

IPn Im1
Imn = (6)
100%
In order not to introduce any asymmetry into the magnetic circuit, the even harmonics of the
current waveform were assumed to be equal to zero in the presented studies.
The values IPn and ϕn for considered harmonics can be represented as a components of vectors IP
and ϕ, which sought next to minimize the Tr component of (1). The goal function used in the process
of finding the optimal “shape” of the current waveform has the following form:

TRFh (Ip, ϕ)
F g (Ip, ϕ) = (7)
TRFr

where TRFh (IP ,ϕ) is the value of the torque ripple factor calculated for injected harmonics IP , ϕ,
and TRFr is the torque ripple factor calculated for the reference case, i.e., a machine supplied by the
fundamental harmonics of the phase currents.
The components of vectors IP , ϕ can be treated as the design variables for the optimization
process. Since the number of design variables is high (for example, equal to 18 considering l = 19)
and the goal function call is quite expensive in terms of computation time (about 20 min for each
Sensors 2020, 20, 2406 5 of 11

goal function evaluation), in the proposed method, the process of finding the optimal current shape is
performed in two main stages. Firstly, the values of particular harmonic phases are each determined
for the assumed values of their amplitudes. Next, the values of the amplitudes are searched for using
sequential parametrical calculations to narrow the design space.
The proposed method of torque ripple minimization using current modulation has been explained
Sensors 2020, 20, x FOR PEER REVIEW 5 of 11
in more detail and demonstrated in the case study problem described in Section 3. The PMSM of
fractional-slot concentrated windings (FSCW) was intentionally selected for the case study. Despite
lower winding cost and simpler manufacturing technology), the FSCW PMSMs suffer from parasitic
the many advantages compared to machines with distributed windings (including lower winding cost
effects caused by the high distortion of the magnetomotive force distribution [21]—in particular,
and simpler manufacturing technology), the FSCW PMSMs suffer from parasitic effects caused by the
increased eddy current losses in the rotor conductive elements, as well as an increased level of
high distortion of the magnetomotive force distribution [21]—in particular, increased eddy current
torque ripples.
losses in the rotor conductive elements, as well as an increased level of torque ripples.
3. Case Study Problem
3. Case Study Problem
Next, the proposed method has been applied to demonstrate the effect of harmonic injection on
Next, the proposed method has been applied to demonstrate the effect of harmonic injection on an
an FSCW PMSM with a surface-mounted permanent magnet (SPM) rotor. The structure of the
FSCW PMSM with a surface-mounted permanent magnet (SPM) rotor. The structure of the magnetic
magnetic circuit and the main parameters of consideration in the case study problem machine are
circuit and the main parameters of consideration in the case study problem machine are shown in
shown in Figure 2. A tested PMSM consists of nine stator slots (ns) and 10 magnetic poles (np). Due to
Figure 2. A tested PMSM consists of nine stator slots (ns ) and 10 magnetic poles (np ). Due to the high
the high value of the least common multiple of np and ns, the value of the cogging torque is low.
value of the least common multiple of np and ns , the value of the cogging torque is low. Moreover,
Moreover, the shape and dimensions of the magnetic circuit in the machine are initially optimized
the shape and dimensions of the magnetic circuit in the machine are initially optimized by FEA in
by FEA in order to reduce the torque ripple level.
order to reduce the torque ripple level.

(a) (b)
Figure2.2.Structure
Figure Structureof
ofthe
themagnetic
magneticcircuit
circuit(a)
(a)and
andmajor
majorparameters
parameters(b)
(b)ofofstudied
studiedPMSM.
PMSM.

4.4.Numerical
NumericalFEM
FEMModel
Modelof
ofPMSM
PMSMand
andResults
Resultsof
ofSimulations
Simulations
To
Toassess
assess the
the impact of current
impact of current harmonics
harmonicson onthe
thelevel
levelof of electromagnetic
electromagnetic torque
torque ripples,
ripples, the
the numerical model of studied PMSM was developed in the ANSYS
numerical model of studied PMSM was developed in the ANSYS Maxwell environment. The Maxwell environment.
The considered
considered domainwas
domain was subdivided
subdividedintointoabout 25,000
about triangular
25,000 elements.
triangular The applied
elements. finite element
The applied finite
(FE) mesh and exemplary magnetic flux density distribution are shown in Figure
element (FE) mesh and exemplary magnetic flux density distribution are shown in Figure 3. A 3. A time-stepping
algorithm was employed
time-stepping algorithmand was period Tf ofand
theemployed the the
fundamental
period Tharmonics was divided into 1500 steps.
f of the fundamental harmonics was
Equations (2) and
divided into 1500(5)steps.
wereEquations
implemented in the(5)
(2) and developed model in order
were implemented to developed
in the perform a current
model waveform
in order to
shape optimization.
perform a current waveform shape optimization.
Sensors 2020,20,
Sensors2020, 20,2406
x FOR PEER REVIEW 66of
of11
11
Sensors 2020, 20, x FOR PEER REVIEW 6 of 11

Figure3.3. Applied
Figure Applied FE
FE mesh
mesh and
and exemplary
exemplaryplot
plotof
ofmagnetic
magneticflux
fluxdensity.
density.
Figure 3. Applied FE mesh and exemplary plot of magnetic flux density.
4.1.
4.1. Optimization
Optimization ofofthe
theCurrent
CurrentWaveform
WaveformShape
Shape
4.1. Optimization of the Current Waveform Shape
The
Theinfluence
influenceofofparticular
particular phase
phase current
currentharmonics
harmonics waswasexamined
examined first;first;
the harmonics of 3rd,
the harmonics of 5th,
3rd,
7th, The 11th
influence of particular phase current harmonics was examined first; the harmonics of 3rd,
5th,9th,
7 th, 9th,and 13th
11th orders
and 13thwere then
orders injected
were theninto the phase
injected intocurrent
the phasewaveform.
currentThe calculations
waveform. The
5th, 7performed
were th, 9th, 11th
for theand 13th
rated orders
value of were thencurrent
fundamental injected into thearbitrary
harmonic, phase current
set waveform.
values of amplitudesThe
calculations were performed for the rated value of fundamental current harmonic, arbitrary set
calculations
and different were
phasesperformed
ofand
particular forcurrent
the rated value of
harmonics. Thefundamental
maximum current
torque per harmonic,
ampere arbitrary
control schemeset
values of amplitudes different phases of particular current harmonics. The maximum torque per
values
was of amplitudes
assumed. To study and different
the impact phases
of the of particular
injection of the current harmonics.
harmonics, the Fast The maximum
Fourier torque
Transform per
(FFT)
ampere control scheme was assumed. To study the impact of the injection of the harmonics, the Fast
ampere was
method control scheme on
performed wastheassumed.
calculated ToTstudy the impact of the injection of the harmonics,
of thethe Fast
Fourier Transform (FFT) method r waveforms.
was performed on theThe results
calculatedofTthe performed
r waveforms. FFT results
The torque
of the
Fourier
ripple Transform
waveforms (FFT) method was performed on the calculated Tr waveforms. The results of the
performed FFT ofare
theillustrated
torque ripple in Figure 4.
waveforms are illustrated in Figure 4.
performed FFT of the torque ripple waveforms are illustrated in Figure 4.

Figure4.4.Harmonic
Figure Harmoniccontent
contentof
oftorque
torqueripple
ripplefor
forodd
oddharmonic
harmonicinjection.
injection.
Figure 4. Harmonic content of torque ripple for odd harmonic injection.
When
Whencomparing
comparing the the obtained
obtained results
results to
tothe
thereferenced
referencedcase case(the
(themachine
machinewas wassupplied
suppliedby bythe
the
When
fundamental comparing
harmonic the obtained
denoted as results
Ref in to
Figure the referenced
4), it can be case
seen (the
that machine
all the was
studied supplied
harmonics by the
have
fundamental harmonic denoted as Ref in Figure 4), it can be seen that all the studied harmonics have
fundamental
an harmonic denoted as Ref in Figure 4),onlyit can betheseen that the
all the studied harmonics have
animpact
impactonon the torque
the ripple
torque level.
ripple Nevertheless,
level. Nevertheless, inonly cases
in theof cases 5th,
of9th
theand5th,13th
9thharmonics
and 13th
an
was impact on
a reduction the torque ripple level. Nevertheless, only in the cases of the 5th, 9th and 13th
harmonics was in the amplitudes
a reduction in theofamplitudes
particular torque ripple harmonics
of particular torque ripple observable.
harmonics At this first stage,
observable. At
harmonics
the proper was
phases a reduction
of these in the
particular amplitudes
harmonics of particular
were searched torque
for. ripple harmonics observable. At
this first stage, the proper phases of these particular harmonics were searched for.
this first stage, the proper
Next, phases of these particular findharmonics were searched for.torque ripple level)
Next,parametrical
parametricalanalyses
analyseswerewereconducted
conductedto to findthe theoptimal
optimal(in (in terms
terms ofof torque ripple level)
Next,
amplitudes parametrical analyses were conducted to find the optimal (in terms of were
torque ripple level)
amplitudes of the particular harmonics. The odd harmonics up to the 19th order were all
of the particular harmonics. The odd harmonics up to the 19th order all considered.
considered.
amplitudes
The of the particular harmonics. The odd harmonics up tothousand
the 19th order were all considered.
Thetorque
torquewaveforms
waveformswere weredetermined
determined by byFEA
FEAfor forover
overfour
four thousanddifferentdifferentcombinations
combinationsof of
The torque
values waveforms were determined by FEA for over four thousand different combinations of
values of odd harmonic amplitudes. Finally, after narrowing the search space, the followingvalues
of odd harmonic amplitudes. Finally, after narrowing the search space, the following values
determined:IP3IP3==16.3, IP5==1.125, IP7==1.125, IP9== 3.31, IP11 == 0.075,
values of odd harmonic amplitudes. Finally, after narrowing the search space, the following values
of
ofharmonic
harmonicamplitudes
amplitudeswere weredetermined: 16.3,IP5 1.125,IP7 1.125,IP9 3.31,IP11 0.075,
of harmonic amplitudes were determined: IP3 = 16.3, IP5 = 1.125, IP7 = 1.125, IP9 = 3.31, IP11 = 0.075,
IP13 = 0.15, IP15 = 0.24, IP17 = 0.08, IP19 = 0.24 in % of Im1. The determined torque ripple waveforms Tr and
IP13 = 0.15, IP15 = 0.24, IP17 = 0.08, IP19 = 0.24 in % of Im1. The determined torque ripple waveforms Tr and
Sensors 2020, 20, 2406 7 of 11

Sensors 2020, 20, x FOR PEER REVIEW 7 of 11

Sensors 2020, 20, x FOR PEER REVIEW 7 of 11
IP13 = 0.15, IP15 = 0.24, IP17 = 0.08, IP19 = 0.24 in % of Im1 . The determined torque ripple waveforms Tr
their
and
their harmonic
their harmonic
harmonic content for
content
content fundamental
forfor fundamental
fundamental (reference) and
(reference)
(reference) andoptimized
and optimizedcurrent
optimized currentharmonics
current harmonics(optimized)
harmonics (optimized)
(optimized)
are
are shown
areshown in Figure
shownininFigure 5.
Figure5.5.

Figure
Figure 5.5.Torque
Figure5. Torque ripple
Torqueripple waveforms
ripplewaveforms and
waveformsand their
andtheir harmonic
theirharmonic content
harmoniccontent (base
content(base structure
(basestructure ofofnnsns/n
structureof /np equal to
equaltoto9/10).
p pequal
s/n
9/10).
9/10).

When
When comparing
Whencomparing
comparingthe the torque
thetorque values
torquevalues
valuesof of the
ofthe ripple
theripple factor
factorTTT
ripplefactor RF, calculated for the machine supplied
RF RF, calculated for the machine supplied
by reference
reference sinusoidal
sinusoidal current
current (0.65%)
(0.65%)
by reference sinusoidal current (0.65%) to the value of RF to
to the
the value
value of
of TTTRF (whichwas
RF (which obtained
wasobtained
obtainedfor for the
forthe optimized
theoptimized
optimized
current
current shape (0.04%)), it can be concluded that the proposed current harmonic injectionmethod
current shape
shape (0.04%)),
(0.04%)), it it
cancan bebe concluded
concluded thatthat
the the proposed
proposed current
current harmonic
harmonic injection
injection method will
method
will
allow allow
the the
torque torque
ripple ripple
level level
to be to be
reducedreducedby by
more
will allow the torque ripple level to be reduced by more than 16 times. more
than than
16 16
times. times.
To
To verify
Toverify the
verifythe proposed
theproposed approach
proposedapproachapproachfor for reducing
forreducing
reducingthe the torque
thetorque ripples
torqueripples
ripplesby by using
byusing
usingthethe current
thecurrent harmonic
currentharmonic
harmonic
injection, the
the above
above optimization
optimization calculations
calculations were
were repeated
repeated
injection, the above optimization calculations were repeated for the second structure of PMSM.for the
for second
the secondstructure
structure of PMSM.
of The
PMSM.The
machine
The with
machine the
with same
the same stator geometry
stator geometry and and anann pnvalue
value equal
equal
machine with the same stator geometry and an np pvalue equal to eight was examined. The winding toto eight
eight was
was examined.
examined. The winding
winding
layout
layout
layoutwaswas
was also
also changed.
changed.
also changed. TheThe determined
determined
The determined values values
of theof
values the
the harmonic
harmonic
of amplitudes
harmonic amplitudes for
for the second
amplitudes for the second
structure
the second
structure
are as are
follows: as I P3 =
follows:
10, IIP5 = 10, I
0.75, P5I=
P7 =
0.75, I
1.05, P7 =IP9 =
1.05, I P9 =IP11
2.4, 2.4, =I =
0.05,
P11 0.05,
IP13 I=
structure are as follows: IP3 = 10, IP5 = 0.75, IP7 = 1.05, IP9 = 2.4, IP11 = 0.05, IP13 = 0.05, IP15 = 0.15, IP17 =0.05,
P3 = = 0.05,
P130.05, IP15I =
P15 = 0.15,
0.15, I I P17 =
P17 = 0.05,
P19 ==0.15.
IP19 0.15.
IP19 = 0.15.
The
The structure
Thestructure
structure ofof
of the
the PMSM
the PMSM
PMSMthat that
was was
was examined
thatexamined secondsecond
examined is shown
second isisinshown
Figure in
shown 6, Figure
inwhile
Figure the6,6, while
while the
determined the
determined
torque rippletorque
determined waveforms
torque ripple
ripple Trwaveforms TTr r and
and their harmonic
waveforms and their harmonic
contents
their contents
contents for
for fundamental
harmonic fundamental
(reference)
for fundamental and (reference)
optimized
(reference)
and
andoptimized
current harmonics
optimized current harmonics
are shown
current harmonics are
in Figureareshown
7.
shown in
inFigure
Figure7.7.

(a)
(a) (b)
(b)
Figure
Figure
Figure 6.6.6.
Structure
Structure and
Structure(a)
and major
and parameters
major
major ofofstudied
parameters
parameters (b) PMSM
of studied
studied (second
PMSMPMSM structure
(second
(second ofofnns/n
structure
structure p of ns /nto
equal 9/8).
p equal
s/np equal to 9/8).
to 9/8).
Sensors 2020, 20, x FOR PEER REVIEW 8 of 11

Sensors 2020, 20,

Sensors 2020, 20, 2406
x FOR PEER REVIEW 88 of
of 11
11

Figure 7. Torque ripple waveforms and their harmonic content (second structure of ns/np equal to
9/8). 7. Torque ripple waveforms and their harmonic content (second structure of ns /np equal to 9/8).
Figure
Figure 7. Torque ripple waveforms and their harmonic content (second structure of ns/np equal to
4.2.
4.2. Influence
9/8).
Influence of
of the
the Current
Current and
and the
the Torque
TorqueAngle
AngleValues
Values
The
The optimization
optimization of
of the
the supply
supply current
current harmonics was was carried
carried out
out forfor thethe rated
rated operational
4.2. Influence of the Current and the Torque Angle harmonics
Values operational
conditions of the studied machines (i.e., rated speed, torque and current
conditions of the studied machines (i.e., rated speed, torque and current values). In practice, values). In practice, the PMSM the
The
machines optimization
usually work
PMSM machines of the
withwork
usually supply
variable
with current
torque harmonics
loadstorque
variable was
(corresponding carried out for
to the supply to
loads (corresponding the rated
current operational
value), current
the supply as well
conditions
as the torque
value), ofangle
as well the
as studied
γ, which
the torque machines
can
anglebe γ, (i.e.,
slightly
which rated
can speed,
different torque
to 90
be slightly and current
◦ . To assess
different the90°.
to values).
robustness
To assessof Inthepractice,
the proposedthe
robustness
PMSM
approach,machines
the torque
of the proposed usually
approach, work
ripple thefactor with
torque variable
was ripple
determined torque
factorforwas loads
the (corresponding
current value
determined in the
for the to the
rangevalue
current supply
from in0.8thecurrent
to 1.2 of
range
value),
the as
rated well
current as the
and torque
γ in the angle
range γ, which
from 86 ◦can
to be
90 ◦ .slightly
The different
characteristics to 90°.
of
from 0.8 to 1.2 of the rated current and γ in the range from 86° to 90°. The characteristics of TRF, T To
RF , assess
obtained theasrobustness
a function
of
of the relative
the proposed
obtained approach,
as acurrent,
function ofthe
have thetorque
been ripple
compared
relative factor
to
current, was determined
the have
characteristics
been compared for the
of the to current
reference value in the
sinusoidal
the characteristics range
supplied
of the
from
machine0.8 to 1.2
reference(i.e., of the
the machine
sinusoidal rated current
withoutmachine
supplied and γ
harmonic(i.e.,in the range
injection); from 86° to
see Figurewithout
the machine 90°. The
8a,b, respectively, characteristics
harmonicforinjection); of
the machines Tsee
RF,

obtained
with 9/10
Figure as a function of
9/8 ns /np ratio.
andrespectively,
8a,b, the relative current, have been
for the machines with 9/10 and 9/8 ns/np ratio. compared to the characteristics of the
reference sinusoidal supplied machine (i.e., the machine without harmonic injection); see
Figure 8a,b, respectively, for the machines with 9/10 and 9/8 ns/np ratio.

(a) (b)
Figure8.8.Torque
Figure Torqueripple
ripple
(a) vs.vs.relative
relativecurrent
currentcharacteristics
characteristics
forfor different
different values
values of the
of the
(b) torque
torque angleangle
γ (a)γ
(a) machine
machine s /nn
1 of1nof
no. no. ps/n p equal
equal to 9/10
to 9/10 andand
(b) (b) machine
machine 2 of2 nofs /nnps/n
no.no. p equal
equal to to 9/8.
9/8.
Figure 8. Torque ripple vs. relative current characteristics for different values of the torque angle γ
(a)
Asmachine
As can
canbe no. 1 of
benoted,
noted, thens/n
the p equalobtained
resultstoobtained
results 9/10 and (b)
show
show machine
that no.
forfor
that 2 ofthe
all the
all ns/nconsidered
p equal to
considered 9/8.conditions
conditions (i.e., (i.e.,
currents and
currents
values),
γand when the
γ values), whendetermined current harmonics
the determined are injected,
current harmonics areboth machines
injected, bothexhibit
machinesa significantly
exhibit a
lower As can
level be
of noted,
torque the results
ripples. obtained
Simulations show
also that
show for
that all the
further considered
optimization
significantly lower level of torque ripples. Simulations also show that further optimization conditions
of the (i.e.,
injection currents
of
of the
the
and
currentγ values),
harmonics
injection when
of the current the
is possible. determined
Extending
harmonics current
the design
is possible. harmonics
space using
Extending are
thethe injected,
values
design both machines
of theusing
space current exhibit
theamplitude
values ofand a
the
significantly
the lower
torqueamplitude
current angle wouldlevel of
andallow torque
the for ripples.
the definition
torque Simulations
angle wouldof theallow also
amplitudes show
for the that further
of particular optimization
definitionharmonics of the
as nonlinear
of the amplitudes of
injection
functions of
of the
the current harmonics
fundamental harmonicis possible.
amplitude Extending
and torquethe design
angle. space using
Consequently,
particular harmonics as nonlinear functions of the fundamental harmonic amplitude and torque the
the values
suppressionof theof
current
the
angle. amplitude
torque ripples will
Consequently, and
be the
themore torque
effective
suppression angle
of thewould
across the allow
whole
torque forwill
range
ripples ofthe bedefinition
the torque of theacross
demanded
more effective amplitudes
of thethemachine.
wholeof
particular
range of theharmonics as nonlinear
torque demanded functions
of the machine.of the fundamental harmonic amplitude and torque
angle. Consequently, the suppression of the torque ripples will be more effective across the whole
range of the torque demanded of the machine.
Sensors 2020, 20, 2406 9 of 11

Sensors 2020, 20, x FOR PEER REVIEW 9 of 11

4.3. Infuence of Triplen Harmonics

4.3. Infuence of Triplen Harmonics
In the last stage, the impact of triplen harmonics was examined. The calculations were performed
In the last stage, the impact of triplen harmonics was examined. The calculations were
for the optimized current shape (i.e., determined harmonics phases and amplitudes, including triplen
performed for the optimized current shape (i.e., determined harmonics phases and amplitudes,
harmonics), in case (a) and, in case (b), for the phase current waveforms without triplen harmonics,
including triplen harmonics), in case (a) and, in case (b), for the phase current waveforms without
IP3 = harmonics,
i.e.,triplen IP9 = IP15 =i.e.,
0. The
IP3 = torque
IP9 = IP15ripple waveforms
= 0. The andwaveforms
torque ripple their harmonic contents
and their for both
harmonic machines
contents for
studied were compared in Figure 9a,b, respectively.
both machines studied were compared in Figure 9a,b, respectively.

(a) (b)
Figure
Figure 9. 9. Comparisonsofoftorque
Comparisons torqueripple
ripplewaveforms
waveforms andand their
their harmonic
harmoniccontents,
contents,illustrating
illustratingtriplen
triplen
current
current harmonics’impact
harmonics’ impactfor
for(a)
(a)machine
machineno.
no. 11 with /nppequal
with nss/n equaltoto9/10,
9/10,and
and(b)
(b)machine
machine no.
no.2 with
2 with
ns /n
nsp/nequal
p equal toto9/8.
9/8.

It can
It can bebe notedthat
noted thattriplen
triplenharmonics
harmonics have a significant
significantinfluence
influenceon onthe
theeffectiveness
effectiveness of of
torque
torque
ripple
ripple minimization
minimization by by current
current harmonic
harmonic injection.
injection. When When comparing
comparing the
the TRF T RF values,
values, the the
lack lack
of of
triplen
triplen harmonics increases the torque ripple level from 0.04% to 1.02% for the
harmonics increases the torque ripple level from 0.04% to 1.02% for the first machine, and from 0.11%first machine, and
to from
0.88%0.11% to machine
for the 0.88% forwith
the machine
an ns /np with
ratio an ns/nto
equal p ratio equal to 9/8. It should be noted that the
9/8. It should be noted that the elimination of
elimination of the triplen harmonics in machine
the triplen harmonics in machine no. 1 causes an increase no. 1 causes in
antorque
increaseripples
in torque ripples
above the above
level ofthethe
level of the reference supply conditions (i.e., fundamental harmonic current). This indicates
reference supply conditions (i.e., fundamental harmonic current). This indicates that, for the three-leg that, for
the three-leg inverter-powered PMSMs, the current shape optimization should be carried out as an
inverter-powered PMSMs, the current shape optimization should be carried out as an independent
independent research task, performed prior to assuming a lack of triplen harmonics.
research task, performed prior to assuming a lack of triplen harmonics.

5. 5. Conclusion
Conclusions
In this paper, a method for torque ripple minimization by current modulation in a PMSM drive
In this paper, a method for torque ripple minimization by current modulation in a PMSM drive
has been presented and discussed. In the proposed approach, to find the optimal shape of the phase
has been presented and discussed. In the proposed approach, to find the optimal shape of the phase
current waveforms, the torque ripple level is evaluated by means of a detailed numerical model of
current waveforms, the torque ripple level is evaluated by means of a detailed numerical model of the
the magnetic field inside the machine, ensuring the results obtained are highly reliable. Despite
magnetic field inside the machine, ensuring the results obtained are highly reliable. Despite higher
higher computational complexity, in relation to approaches based on analytical models, the
computational
utilization ofcomplexity,
FEM based in relation
models to approaches
allows the saturationbased on analytical
phenomena, models,
as well the utilization
as other of FEM
parasitic effects
based models allows the saturation phenomena,
occurring in PMSMs, to be taken into account. as well as other parasitic effects occurring in PMSMs,
to be taken into account.
To demonstrate the universality and effectiveness of the method, an analysis of the impact of
To demonstrate the
the phase current harmonic universality
injection and
oneffectiveness of the method,
the electromagnetic an analysis
torque ripple of the
level of the impact
permanentof the
phase current harmonic injection on the electromagnetic torque ripple level
magnet synchronous machine driven by the six-leg inverter was carried out. Two FSCW PMSMs of of the permanent magnet
synchronous
SPM rotor machine
structure driven by the six-leg
and different numbers inverter
of poleswaswere
carried out. Two
examined as FSCW
a case PMSMs of SPM To
study problem. rotor
structure
reduce and different
the level numbers
of torque of poles
ripples, were examined
the waveforms of theas a case
phase study problem.
currents To reduce
were modulated the level
by higher
of harmonics.
torque ripples,A wide range fractional
the waveforms analysis
of the phase was performed.
currents were modulatedThe results
by higherobtained proveAthe
harmonics. wide
effectiveness
range fractional of the method,
analysis showing The
was performed. thatresults
the smoothness of the
obtained prove thetorque waveform
effectiveness of thecan be
method,
significantly
showing improved
that the by theofcurrent
smoothness harmonic
the torque injection.
waveform canComparing the value
be significantly of the torque
improved by theripple
current
factor ofinjection.
harmonic the machine supplied the
Comparing by the
valuereference sinusoidal
of the torque current
ripple factor(0.65%)
of theto TRF for the
machine optimized
supplied by the
current shape (0.04%), it can be concluded that the torque ripple level in the studied
reference sinusoidal current (0.65%) to TRF for the optimized current shape (0.04%), it can be concluded PMSM with an
ns/n
that p ratio equal to 9/10 was reduced by more than 16. In the case of the second tested PMSM, the
the torque ripple level in the studied PMSM with an ns /np ratio equal to 9/10 was reduced by
torque ripples were reduced by around 13. Simulations also show that the torque ripple level
more than 16. In the case of the second tested PMSM, the torque ripples were reduced by around
Sensors 2020, 20, 2406 10 of 11

13. Simulations also show that the torque ripple level depends on the current value (consequently,
the effective torque value) as well as the torque angle. Nevertheless, the results obtained show that
for all considered conditions (i.e., variation of the currents and torque angle γ values), when the
determined current harmonics are injected, both machines studied exhibited significantly lower levels
of torque ripples than the reference sinusoidal supplied machines.
Our ongoing task is to practically validate the effectiveness of the current harmonic injection
method on the PMSM test drive supplied by a six-leg inverter. The results obtained will provide the
scope for further publications by the authors.

Author Contributions: Conceptualization, C.J. and D.D.; Methodology, C.J. and W.S.; Software, D.D. and C.J.;
Validation, D.D. and C.J.; Formal analysis, D.D. and C.J.; Data curation, D.D. and C.J.; Writing—original draft
preparation, D.D., Writing—review and editing, W.S. and C.J.; Visualization, D.D.; Supervision, C.J. All authors
have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Brock, S.; Lukichev, D.V.; Demidova, G.L. Minimizing Torque Ripple in PMSM Drive by Cuckoo Search
Algorithm. In Proceedings of the 2017 19th European Conference on Power Electronics and Applications
(EPE’17 ECCE Europe), Warsaw, Poland, 11–14 September 2017.
2. Flieller, D.; Nguyen, N.K.; Wira, P.; Sturtzer, G.; Abdeslam, D.O.; Mercklé, J. A Self-Learning Solution for
Torque Ripple Reduction for Nonsinusoidal Permanent-Magnet Motor Drives Based on Artificial Neural
Networks. IEEE Trans. Ind. Electron. 2014, 61, 655–666. [CrossRef]
3. Gómez-Espinosa, A.; Hernández-Guzmán, M.V.; Bandala-Sánchez, M.; Jiménez-Hernández, H.;
Rivas-Araiza, A.E.; Rodríguez-Reséndiz, J.; Herrera-Ruíz, G. A New Adaptive Self-Tuning Fourier Coefficients
Algorithm for Periodic Torque Ripple Minimization in Permanent Magnet Synchronous Motors (PMSM).
Sensors 2013, 13, 3831–3847. [CrossRef]
4. Xu, Y.; Shi, T.; Yan, Y.; Gu, X. Dual-Vector Predictive Torque Control of Permanent Magnet Synchronous
Motors Based on a Candidate Vector Table. Energies 2019, 12, 163. [CrossRef]
5. Kwon, J.-W.; Lee, J.; Zhao, W.; Kwon, B.-I. Flux-Switching Permanent Magnet Machine with Phase-Group
Concentrated-Coil Windings and Cogging Torque Reduction Technique. Energies 2018, 11, 2758. [CrossRef]
6. Pietrowski, W. Approximation of Electromagnetic Torque-Time Characteristic by Means of Artificial Neural
Network. Przeglad ˛ Elektrotechniczny 2009, 85, 9–11.
7. Islam, R.; Husain, I.; Fardoun, A.; McLaughlin, K. Permanent Magnet Synchronous Motor Magnet Designs
with Skewing for Torque Ripple and Cogging Torque Reduction. In Proceedings of the 2007 IEEE Industry
Applications Annual Meeting, New Orleans, LA, USA, 23–27 September 2007; pp. 1552–1559.
8. Jedryczka, C.; Szelag, W.; Piech, Z.J. Multiphase Permanent Magnet Synchronous Motors with Fractional
Slot Windings: The Future of Low Speed Drives? COMPEL–Int. J. Comput. Math. Electr. Electron. Eng. 2016,
35, 1937–1948. [CrossRef]
9. Jahns, T.M.; Soong, W.L. Pulsating Torque Minimization Techniques for Permanent Magnet AC Motor
Drives—A Review. IEEE Trans. Ind. Electron. 1996, 43, 321–330. [CrossRef]
10. Marchand, C.; Razek, A. Optimal Torque Operation of Digitally Controlled Permanent Magnet Synchronous
Motor Drives. IEE Proc. B Electr. Power Appl. 1993, 140, 232–240. [CrossRef]
11. An, Q.; Liu, J.; Peng, Z.; Sun, L.; Sun, L. Dual-Space Vector Control of Open-End Winding Permanent Magnet
Synchronous Motor Drive Fed by Dual Inverter. IEEE Trans. Power Electron. 2016, 31, 8329–8342. [CrossRef]
12. Rovere, L.; Formentini, A.; Calzo, G.L.; Zanchetta, P.; Cox, T. IGBT-SiC Dual Fed Open End Winding PMSM
Drive. In Proceedings of the 2017 IEEE International Electric Machines and Drives Conference (IEMDC),
Miami, FL, USA, 21–24 September 2017; pp. 1–7.
13. Chu, L.; Jia, Y.; Chen, D.; Xu, N.; Wang, Y.; Tang, X.; Xu, Z. Research on Control Strategies of an Open-End
Winding Permanent Magnet Synchronous Driving Motor (OW-PMSM)-Equipped Dual Inverter with a
Switchable Winding Mode for Electric Vehicles. Energies 2017, 10, 616. [CrossRef]
Sensors 2020, 20, 2406 11 of 11

14. Hussein, I.; Al-Hamouz, Z.; Abido, M.A.; Milhem, A. On the Mathematical Modeling of Line-Start Permanent
Magnet Synchronous Motors under Static Eccentricity. Energies 2018, 11, 197. [CrossRef]
15. You, Y.-M. Optimal Design of PMSM Based on Automated Finite Element Analysis and Metamodeling.
Energies 2019, 12, 4673. [CrossRef]
16. Cupertino, F.; Leuzzi, R.; Monopoli, V.G.; Cascella, G.L. Design Procedure for High-Speed PM Motors Aided
by Optimization Algorithms. Machines 2018, 6, 5. [CrossRef]
17. Baranski, M.; Szelag, W.; Lyskawinski, W. An Analysis of a Start-up Process in LSPMSMs with Aluminum and
Copper Rotor Bars Considering the Coupling of Electromagnetic and Thermal Phenomena. Arch. Electr. Eng.
2019, 68, 933–946.
18. Zhu, W.; Fahimi, B.; Pekarek, S. A Field Reconstruction Method for Optimal Excitation of Permanent Magnet
Synchronous Machines. IEEE Trans. Energy Convers. 2006, 21, 305–313. [CrossRef]
19. Sutthiphornsombat, B.; Khoobroo, A.; Fahimi, B. Mitigation of Acoustic Noise and Vibration in Permanent
Magnet Synchronous Machines Drive Using Field Reconstruction Method. In Proceedings of the 2010 IEEE
Vehicle Power and Propulsion Conference, Lille, France, 1–3 September 2010; pp. 1–5.
20. Gieras, J.F.; Wang, C.; Lai, J.C. Noise of Polyphase Electric Motors; CRC Press: Boca Raton, FL, USA, 2018.
21. Magnussen, F.; Lendenmann, H. Parasitic Effects in PM Machines with Concentrated Windings. IEEE Trans.
Ind. Appl. 2007, 43, 1223–1232. [CrossRef]

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).