Marquette University

e-Publications@Marquette
Electrical and Computer Engineering Faculty Electrical and Computer Engineering, Department
Research and Publications of

7-1-2016

Optimal Design of IPM Motors With Different

Cooling Systems and Winding Configurations
Alireza Fatemi
Marquette University

Dan M. Ionel
University of Wisconsin - Milwaukee

Nabeel Demerdash
Marquette University, [email protected]

Thomas W. Nehl
General Motors

Accepted version. IEEE Transactions on Industry Applications, Vol. 52, No. 4 ( July-August 2016):
3041-3049. DOI. © 2016 Institute of Electrical and Electronics Engineers (IEEE). Used with
permission.
Marquette University

e-Publications@Marquette

Faculty Research and Publications/Department

This paper is NOT THE PUBLISHED VERSION; but the author’s final, peer-reviewed manuscript. The
published version may be accessed by following the link in th citation below.

IEEE Transactions on Industry Applications, Vol. 52, No. 4 (July/August 2016): 3041-3049. DOI. This
article is © Institute of Electrical and Electronics Engineers (IEEE) and permission has been granted for
this version to appear in e-Publications@Marquette. Institute of Electrical and Electronics Engineers
(IEEE) does not grant permission for this article to be further copied/distributed or hosted elsewhere
without the express permission from Institute of Electrical and Electronics Engineers (IEEE).

Optimal Design of IPM Motors With Different Cooling Systems and

Winding Configurations

Alireza Fatemi
Department of Electrical and Computer Engineering, Marquette University, Milwaukee, WI
Dan M. Ionel
Department of Electrical and Computer Engineering, University of Kentucky, Lexington, KY
Nabeel A. O. Demerdash
Department of Electrical and Computer Engineering, Marquette University, Milwaukee, WI
Thomas W. Nehl
General Motors Global Research and Development, Pontiac, MI

Performance improvement of permanent magnet (PM) motors through optimization techniques has been widely
investigated in the literature. Oftentimes the practice of design optimization leads to derivation/interpretation of
optimal scaling rules of PM motors for a particular loading condition. This paper demonstrates how these derivations
vary with respect to the machine ampere loading and ferrous core saturation level. A parallel sensitivity analysis using
a second-order response surface methodology followed by a large-scale design optimization based on evolutionary
algorithms are pursued in order to establish the variation of the relationships between the main design parameters
and the performance characteristics with respect to the ampere loading and magnetic core saturation levels
prevalent in the naturally cooled, fan-cooled, and liquid-cooled machines. For this purpose, a finite-element-based
platform with a full account of complex geometry, magnetic core nonlinearities, and stator and rotor losses is used.
Four main performance metrics including active material cost, power losses, torque ripple, and rotor PM
demagnetization are investigated for two generic industrial PM motors with distributed and concentrated windings
with subsequent conclusions drawn based on the results.

SECTION I
INTRODUCTION
The design and modeling of permanent magnet (PM) motors has consistently been a subject of special
interest in the literature due to the distinctive features of such machines, including but not limited to high
efficiency, and high power density, which makes them suitable for high-performance applications.1,2 Recent
design trends rely on large-scale design optimization techniques based on electromagnetic (EM) finite-
element (FE) analysis of design candidates for high-fidelity calculation of their performance metrics.3 In
some studies, coupled EM/thermal models have been utilized to accurately model the multiphysics nature
of the design problem.4,5,6,7 In many others, a fixed current density with reference to the machine's cooling
system, and accordingly an engineering guess of the operating temperatures of various components are
assumed.8,9,10,11,12,13 Although the coupled EM/thermal models offer relatively reasonable predictions of
temperature distributions inside such machines, which can be used either to update the material
properties for a subsequent iteration of the EM FE analysis until a convergence is reached,4 or to determine
the maximum current density limit based on the winding temperatures,5 the uncoupled approach is faster.
The uncoupled optimization approach can also be as effective, provided that the design problem is well
defined and that the thermal performances of the optimal designs are examined at the final stages of the
design process. The post-design optimization thermal modeling14 is usually done in order to identify the
parameters of the cooling system such as the type of surfaces or the coolant flow rate.15 Since more
efficient designs are achieved through the design optimization, the cooling system is anticipated to remove
lesser magnitudes of losses than those associated with motors constructed on the basis of traditional
design methods.

The design optimization practice is often accompanied by establishing the optimal scaling rules for
achieving a particular improvement in the machine performance. Accordingly, design recommendations for
mitigation of torque ripples, minimization of power losses, and reducing dependency on rare-earth PMs,
have been often reported in the literature for naturally cooled (NC) machines.9,10,11,16,17 Nevertheless,
increasing the stator winding current density or the machine's ampere loading elevates the flux level and,
hence, the saturation level throughout the machine's magnetic circuit. This not only affects the machine
performance, but also alters the correlation between them and the geometric design parameters
associated with the machine cross-section. In,18 the influence of ampere loading and magnetic saturation
on the cogging torque, back electromotive force, and the torque ripple of PM machines was investigated.
However, the impacts of the geometric design parameters on these performance metrics was not
considered. In,19 nonlinear scaling rules for low power density brushless PM synchronous machines were
developed without taking the effects of ampere loading and magnetic core saturation into account.

This paper contains further contribution through a parallel investigation of the relations between the
geometric design parameters and the performance metrics of interior PM (IPM) motors for a wide range of
ampere-loading levels determined by the machine's cooling system.20 This is accomplished through a
systematic sensitivity analysis and large scale design optimization. Accordingly, it will be demonstrated that
one set of design rules, such as those derived in,19 cannot be generalized to all three classes of cooling
systems.

In order to account for the distinctive performance characteristics of the integral and fractional winding
IPM machines, two mainstream case-study industrial IPM machine configurations with distributed and
fractional slot concentrated windings (FSCW) are considered. In essence, while FSCW machines offer
reduced copper losses owing to shorter end windings,21,22 they suffer from higher core losses as a result of
the space harmonics introduced by more “discrete” nature of the stator magnetomotive force (MMF)
waveform.23 Furthermore, FSCW machines are less susceptible to the torque ripple due to their intrinsically
lower cogging torque. This will lead to different correlations of the geometric design parameters with these
performance metrics as will be discussed in the paper.

SECTION II
BENCHMARK DESIGNS
The parametric cross-sections of the studied machines and their full-fledged 2-D time-stepping FE (TSFE)
models are discussed in this section.

A. Parametrized Cross Sections

Two generic industrial IPM motors with similar rotor configurations, though with different slot-pole
combinations are considered.

1. The first case-study is a 48-slot, 8-pole motor with a single-layer distributed winding configuration,
see Fig. 1(a). For brevity, this design will be referred to as the 48S8P machine.
2. The second case study is a 12-slot, 10-pole motor with a double-layer concentrated winding
configuration, see Fig. 1(b). Hereafter, this design will be denoted as the 12S10P machine.
Fig. 1. Cross-sections of the parameterized models of the studied IPM motors consisting of (a) single-layer distributed, and (b) all-tooth-
wound concentrated windings.

The design parameters in Fig. 1 are rationalized according to Table I in order to avoid geometric conflicts
between the structures of various components of the two motors. The parameterized FE model of the
48S8P motor is comprised of a total of ten independent design variables, five residing in the rotor, four
residing in the stator, in addition to the air-gap height. The 12S10P motor under investigation has one less
independent design variable due to its open slot configuration, thus eliminating the parameter associated
with the tooth tip.
Table I Independent Design Variables and Their Bounds

The design variables are confined by upper and lower bounds, also listed in Table I, either to prevent the
unintended intersection of various boundary surfaces of components or to address mechanical constraints,
e.g., minimum air-gap height or yield stress of the rotor bridges.19 Nevertheless, wide bounds are
designated to allow a full exploration of the design space during the optimization procedure. For both
machines, the stator outer and the rotor inner (shaft) diameters are fixed to 260 and 111 mm., respectively.

B. TSFE Model
The 2-D TSFE models of the two machines were developed in ANSYS Maxwell. The core laminations are
assumed to be nonoriented silicon steel of 0.36 mm thickness. The remenance and coercivity of the NdFeB
PMs, and the resistivity of the copper of the stator windings are evaluated corresponding to temperatures
of 100 and 150°C, respectively. This is in order to account for the lowest magnet field and highest winding
resistance.

The TSFE analysis with sinusoidal current excitation is utilized to compute the following performance
metrics of interest.

1. Active material cost (AMC): The normalized AMC is given below:

where the mass, m, is in kg and the steel cost is considered as the one-unit reference in this
normalized/per-unit formulation. It can be assumed that the AMC is an approximate indication of
the total cost of each motor configuration, given that the manufacturing expenses of the same
 motor topology with identical winding configuration is usually comparable for different design
solutions.

2. Power losses: The power losses consist of copper losses in the stator windings with approximations
of the end winding copper losses, core losses including hysteresis and eddy current losses in the
stator and rotor laminated cores, and the eddy current losses in the rotor PMs for the 12S10P
machine with the assumption that these eddy currents are resistance limited. For accurate
calculation of the core losses, the frequency domain core loss model introduced in24 is utilized. The
core losses have been obtained on an element by element basis for an excitation current frequency
corresponding to a rotor speed of 1800 r/min.
3. Torque ripple: Following the calculation of the average torque per unit length, Tave, using TSFE
analysis, the stack lengths of the designs are adjusted to produce the desired average torque of 300
N·m for a stator current density specifically chosen according to the cooling system of the machine.
The torque ripple, Tr, is subsequently determined from the torque profile over a full fundamental ac

cycle using (2)

4. PM demagnetization: PM demagnetization effects are characterized by the minimum flux density in
the rotor PMs over a complete ac cycle. Demagnetization is considered at the magnet piece
level. Using NdFeB PMs with reduced heavy earth dysprosium (Dy) content25 can offer local
demagnetization protection, assuming that reasonable measures, such as introducing air pockets
around vulnerable areas,26 are adopted at the final stages of the design.

The excitation current densities considered for the TSFE analysis accounts for the type of the cooling
system. Typical numbers are given in Table II.27 Here, the fixed current densities of 4, 8, and 16 A/mm2 are
assumed for NC, fan-cooled (FC), and liquid-cooled (LC) machines, respectively.

Table II Typical Current Density Ranges for Different Cooling Systems

The phase angle of the current vector is chosen so as to ensure maximum torque per ampere (MTPA)
operation. Since the design candidates can be saturated, particularly in FC and LC classes, the torque angle
at MTPA for each individual design is numerically calculated by sampling and interpolating the generated
torque at multiple phase angles of the stator current phasor.11,28

SECTION III
SENSITIVITY ANALYSIS
The variation of the relationships between the geometric design parameters and the performance metrics
can be understood by carrying out a sensitivity analysis at the three levels of stator winding current density.
In the following sections, the sensitivity analysis methodology and the results are presented.
A. Methodology
For each cooling system, five second-order response surfaces were defined corresponding to the cost,
copper losses, core losses, torque ripple, and degree of PM demagnetization. The core losses and copper
losses were treated separately since the design variables have distinctive effects on these two loss
components.

The regression coefficients pertaining to these response surfaces indicate how the variation of the design
parameters within their permissible ranges influences the machine performance metrics. The class of
central composite designs29 was used for design of experiments (DOE) in order to calculate the regression
coefficients associated with each second-order response surfaces, y, given in (3),

where βo, βi, βii, and βij are the regression coefficients for the n design variables, xi, expressed in the coded
form, ci, according to (4)

In Fig. 2, the regression coefficients of the sensitivity analysis are shown for the five examined performance
metrics at the three different levels of current densities accounting for the three classes of cooling systems.
Since the purpose of the sensitivity analysis is to provide a measure of importance of each design
parameter with respect to other design variables, the regression coefficients are normalized to the variable
with the maximum influence in each group of metrics. Furthermore, the DOE is conducted so as to ensure
no main effect or two-factor interaction is aliased with any other main effect or two-factor interaction of
the design variables, i.e., so called resolution V designs.29 Therefore, the regression coefficients can be
examined independently for each group of metrics.
Fig. 2. Normalized regression coefficients from the sensitivity study indicating the effect of the design variables on, AMC, copper loss,
core loss, minimum PM flux density, and torque ripple for (a) the 48S8P machine, and (b) the 12S10P machine.

Each coefficient is associated with a design variable, i.e., input, to indicate how its variation within the
permissible range denoted in Table I would influence the machine's performance, i.e., output. A positive
coefficient, (+), indicates that an increase in input will raise the output whereas a negative coefficient, (-),
constitutes the opposite trend between the input and output. Accordingly, a close examination of Fig.
2 leads to the observation that some trends vary drastically, or even change direction, with respect to the
ampere loading for the two 48S8P and 12S10P machines.

B. Discussion of the Results of the Sensitivity Analysis

The sensitivity analysis for each performance metric is summarized in this section.

1) AMC
According to Fig. 2(a.1) and (b.1), for both machine configurations.

1. The strongest correlation with AMC in the NC and FC classes is attributed to the yoke height, which is
opposite to the slot depth for a given stator bore.
2. The yoke height correlation with AMC is superseded by that of the air-gap height in the LC class.
3. The positive correlation of the tooth stem width with AMC monotonically decreases as the ampere
loading level increases.
2) Copper Losses
According to Fig. 2(a.2) and (b.2).

1. The considerable variations of the influences of the design parameters on the copper losses as the
ampere loading level increases should be noted for both machine configurations.
2. For the 48S8P machine, the air-gap height constitutes the strongest correlation with copper losses in the
NC class. In the FC and LC classes, the split ratio, ksi, the tooth-stem width, and the yoke height become
more influential.
3. Similar trends exist for the 12S10P machine, except for the diminished influence of the yoke height.

3) Core Losses
According to Fig. 2(a.3) and (b.3).

1. The influences of the design parameters on the core losses are sensitive to the ampere loading to such
an extent that some of these relationships, i.e., those associated with the stator inner diameter and
yoke height, are reversed as the ampere-loading increases.
2. For both machine configurations, in the NC class, the design parameters associated with the stator
constitute the strongest correlation with the core losses.
3. As the ampere loading increases, the design parameters associated with the rotor become more
influential.
4. The strong influence of the rotor q-axis bridge on the core losses, especially in the 12S10P motor, should
be noted.

4) Torque Ripple
According to Fig. 2(a.4) and (b.4).

1. Except for the air-gap height which maintains a strong negative correlation with torque ripple under any
ampere loading conditions, the two machine configurations have distinctive torque ripple
characteristics.
2. For the 48S8P machine, the width of the q-axis bridge becomes more influential than the air-gap height
as the level of ampere-loading increases. Meanwhile, the negative correlation of the widths of the tooth
stem and tooth tips with torque ripple diminishes as the saturation level of the magnetic core is
elevated.
3. As opposed to the 48S8P machine, the 12S10P machine is intrinsically less susceptible to torque ripple
under any loading conditions.
4. It is interesting to note the reversal of the correlation between the torque ripple and the tooth stem
width under heavy magnetic core saturation in the 12S10P motor.

5) PM Demagnetization
According to Fig. 2(a.5) and (b.5).

1. For both machine configurations and under all loading conditions, the PM demagnetization is strongly
correlated to the PM height and tooth-stem width.
 2. The correlation factors of the split ratio, ksi, and the yoke height monotonically increase as the level of
ampere loading increases.
3. Although the air-gap height constitutes a negative correlation with Bpm in NC machines, this relationship
reverses as the ampere-loading level increases.

SECTION IV
STATISTICAL ANALYSIS OF THE OPTIMIZEDDESIGNS
The sensitivity analysis merely reveals the independent effects of the design variables on the machine
performance. To understand how the interactions between these variables influence the final optimal
design which is subject to a unique set of objectives and constraints, a close examination of the design
optimization process and results is in order. Nevertheless, the sensitivity study suggests that the optimal
designs for each class should have distinctive features which will be examined in this section.

A large-scale design optimization algorithm has been developed and utilized to optimize the two motor
configurations for each level of current densities. The design optimization is followed by a statistical
analysis on the optimized designs to find the range of change, i.e., the statistical distribution of the design
variables in the optimized designs for each class of cooling systems.

A. Large-Scale Design Optimization

Six runs of a large-scale design optimization, each composed of 6600 candidate designs, were carried out to
optimize the machine model at the three aforementioned levels of current densities.

The optimization algorithm relies on the TSFE model developed in Section II. A combined multi-objective
optimization with differential evolution,30 was implemented as the optimization search algorithm.

The fitness function of the optimization problem has been defined based on the previously discussed
performance metrics. For a generic industrial use, it practically consists of the two following objectives: 1)
minimization of the AMC given in (1); and 2) minimization of the power losses.

Due to the opposite correlations of the design parameters, except the air-gap height, hg, with AMC and
power losses in Fig. 2, the two objectives are conflicting.

Furthermore, the two following constraints are introduced for reliable operation: 1) less than 15% torque
ripple; and 2) less than 70% PM demagnetization.

Each run of the optimization was started with an initial generation of 200 members and was carried out
over 800 generations of eight members each.30 Fig. 3(a)–(b)shows the optimization results in terms of the
conflicting objectives for the 48S8P and 12S10P machines, respectively. The design candidates indicated
in Fig. 3 are compared in terms of AMC and power losses at the same rated load operating point. It is
evident that increasing the stator winding current density and, thus, the ampere loading of the machine,
complicates the design of the cooling system by increasing the total dissipated losses. Nevertheless, in
many applications, higher torque density can translate into reduced system cost, size, and weight, thus
promoting such designs with sophisticated cooling systems.14,26 Furthermore, the extremes of the power
losses and AMC of the three classes of cooling systems are nonoverlapping. That is, the same efficiency
characteristics of NC machines cannot be obtained by LC machines. Similarly, the AMC in LC machines can
be reduced to values not achievable by NC machines.

Fig. 3. Optimization results for the three case-study cooling systems (a) 48S8P, and (b) 12S10P machines. The normalized AMC is defined
according to (1).

It is interesting to notice that due to higher core losses and larger masses of PM in their construction, the
Pareto-optimal designs of the 12S10P configuration are located further apart from the origin as compared
to the 48S8P configuration, see Tables IV and V.

B. Scaling Rules of the Optimum Candidate Designs

To differentiate the distinctive optimal design values between the three cooling classes, the statistical
distributions of the design variables in the Pareto-optimal designs should be investigated. For this purpose,
from each run of optimization, 500 superior designs were selected based on a three-step process. First, the
designs which do not violate the constraints on the torque ripple and on the PM demagnetization are
identified and separated in set P. Subsequently, a strength value s(di) is designated to each of the
designs, di in P, according to (5)

where “#” is the cardinality of the set,31 and k is the number of the designs in P. The designs in P are
thereafter ranked in a descending order based on their strength value, sdi, to determine the top 500
superior designs.

The distribution of the design parameters in the selected designs with respect to their bounds can be
described by “box plots” shown in Fig. 4. The rectangular boxes in Fig. 4represent the first, the second, and
the third quartiles of the distribution of design values with respect to the upper and lower bounds
designated for each design parameter in Table I. The distances between the different parts of these boxes
indicate the degree of dispersion and skewness in the value of the optimal designs. The circles in the
middle, and the whiskers represent the average, the maximum, and the minimum of the optimal design
value for each parameter.

Fig. 4. Distribution of the design variables in the optimized designs for (a) 48S8P, and (b) 12S10P machines. See Table III for the average
values of the optimal design parameters.

The representative cross-sections which were generated based on the means of the design variables
in Table III are provided in Fig. 5 which allows one to visualize the distinctive design features of each
machine. These features will be subsequently explained with reference to the statistical and sensitivity
analyses.

Fig. 5. Optimized cross sections derived based on the mean of the design variables in the 500 Pareto-optimal designs, see Table III, for the
three current density levels: (a) 48S8P, and (b) 12S10P machines.

Table III Average Design Parameters in the Selected Optimal Designs

According to Fig. 4, the optimal ranges of the design variables are broader in the 48S8P machine. This stems
from the imposed criterion on the torque ripple of the selected designs, which leads to the dispersion of
the chosen design candidates in the Pareto-front vicinity.

Three different trends can be recognized in the variation of the design parameters which are addressed in
the following discussion.
1) Parameters With Increasing Trends in Both Motor Configurations
As can be seen in Fig. 4, the split ratio, ksi, the ratio kwt, and the yoke height, hg, relatively grow with the
increase of the ampere loading. This is in line with the sensitivity analysis and the fact that decreasing
copper losses, as the major loss component in the case study PM machines according to Table IV, takes
precedence over decreasing core losses. The 𝑘𝑘𝑑𝑑𝑝𝑝𝑝𝑝 ratio varies widely in the 48S8P machine due to the
insignificant correlation of this design variable with the optimization objectives according to the sensitivity
analysis in Fig. 2(a). The stronger negative correlation of 𝑘𝑘𝑑𝑑𝑝𝑝𝑝𝑝 ratio with AMC and copper and core losses in
12S10P machine, has resulted in its maximization in the selected 12S10P designs, as shown in Fig. 4(b).
Similarly, the 𝑘𝑘𝑤𝑤𝑝𝑝𝑝𝑝 ratio and αpm are concentrated toward their higher bounds due to their negative
correlations with the two objectives as shown in Fig. 2.

Table IV Mean of the Ratio of Copper Losses to Core Losses in the Pareto-optimal Designs

2) Parameters With Decreasing Trends in Both Motor Configurations

The PM height, hpm, in Fig. 4 monotonically decreases as the ampere loading increases in both 48S8P and
12S10P machines. This can be explained by considering the results of the sensitivity analysis in Fig. 2, and
the change of the masses of various components in the selected designs for the three cooling systems, as
listed in Table V. Accordingly, the PM mass has been reduced in the selected designs proportionately to the
masses of other components by reducing the only parameter associated with the PM mass which constitute
a positive correlation with AMC, i.e., hpm. Meanwhile, the excessive PM demagnetization as a result of
reducing PM height has been prevented by introducing a constraint on the minimum flux density of the
PMs. The air-gap height, hg, in the optimum designs is minimized due to its positive correlation with AMC
and major component of the losses, i.e., copper losses.

Table V Mean of the Masses of the Optimized Designs Normalized With Respect to the Values Obtained for NC Class
3) Parameters With Conflicting Trends in the Two Motor Configurations
For the 48S8P machine, while the 𝑘𝑘𝑤𝑤𝑞𝑞 ratio monotonically increases in Fig. 4(a) as the torque ripple
becomes less of an issue, it decreases constantly in the 12S10P machine, Fig. 4(b), in order to reduce the
core losses. On the one hand, according to the sensitivity analysis, the core loss has a stronger correlation
with this design variable, and on the other hand, the core loss constitute a considerable portion of the
overall losses in the 12S10P machine, see Table IV.

The tooth-tip width of the 48S8P machine widely varies at higher ampere loading due to saturation. The
tooth tips are larger at low ampere loading to alleviate the torque ripple, see Fig. 2(a.4). Their variation is
ineffective when the machine saturates. Meanwhile, the mean of the tooth-tip width in Fig. 4(a) is reduced
to mitigate the core losses, although this reduction in width is not significant since the overall ratio of core
to copper losses is insignificant, see Table IV.

SECTION V
CONCLUSION
A parallel sensitivity analysis was carried out on two case-study IPM machines with concentrated and
distributed stator configurations and with different cooling systems. It was demonstrated that the
correlation between the main design variables and various performance metrics, particularly core losses,
copper losses, and torque ripple can be significantly affected by the machine's ampere loading and
magnetic core saturation. In some cases, these trends can even be reversed.

The distribution of the optimal design values were investigated for each case study in a practical
optimization problem where the interaction of the performance metrics and design variables occur.
Noticeable difference in the optimal design values were observed and the trends were classified for the NC,
FC, and LC machines.

Acknowledgment
The authors would like to thank ANSYS Inc. and Motor Design Limited for the software support.

FOOTNOTES
Paper 2015-EMC-0898.R1, presented at the 2015 IEEE International Electric Machines and Drives Conference,
Coeur d’Alene, ID, USA, May 10-13, and approved for publication in the IEEE TRANSACTIONS ON INDUSTRY
APPLICATIONS by the Energy Systems Committee of the IEEE Industry Applications Society. This work was
supported by General Motors Global Research and Development, and by the Midwest Energy Research
Consortium.

REFERENCES
1
D. Dorrell, M. Hsieh, M. Popescu, L. Evans, D. Staton and V. Grout. "A review of the design issues and
techniques for radial-flux brushless surface and internal rare-earth permanent-magnet motors" IEEE
Trans. Ind. Electron., vol. 58, no. 9, pp. 3741-3757, 2011
2
E. Armando, P. Guglielmi, G. Pellegrino, M. Pastorelli and A. Vagati. "Accurate modeling and performance
analysis of IPM-PMASR motors" IEEE Trans. Ind. Appl., vol. 45, no. 1, pp. 123-130, 2009
3
Y. Duan and D. Ionel. "A review of recent developments in electrical machine design optimization methods with
a permanent-magnet synchronous motor benchmark study" IEEE Trans. Ind. Appl., vol. 49, no. 3, pp.
1268-1275, 2013
4
Y. Wang, D. Ionel and D. Staton. "Ultrafast steady-state multi-physics model for pm and synchronous reluctance
machines" IEEE Trans. Ind. Appl., vol. 51, no. 5, pp. 3639-3646, 2015
5
W. Jiang and T. Jahns. "Coupled electromagnetic/thermal machine design optimization based on finite element
analysis with application of artificial neural network" Proc. IEEE Energy Convers. Congr. Expo., pp. 5160-
5167, 2014
6
S. Semidey, Y. Duan, J. Mayor, R. Harley and T. Habetler. "Optimal electromagnetic-thermo-mechanical
integrated design candidate search and selection for surface-mount permanent-magnet machines
considering load profiles" IEEE Trans. Ind. Appl., vol. 47, no. 6, pp. 2460-2468, 2011
7
J. Wang and D. Howe. "Design optimization of radially magnetized, iron-cored, tubular permanent-magnet
machines and drive systems" IEEE Trans. Magn., vol. 40, no. 5, pp. 3262-3277, 2004
8
G. Sizov, D. Ionel and N. Demerdash. "Multi-objective optimization of pm ac machines using computationally
efficient - FEA and differential evolution" Proc. IEEE Int. Elect. Mach. Drives Conf., pp. 1528-1533, 2011
9
G. Lei, C. Liu, J. Zhu and Y. Guo. "Techniques for multilevel design optimization of permanent magnet motors"
IEEE Trans. Energy Convers., vol. 30, no. 4, pp. 1574-1584, 2015
10
C. Xia, L. Guo, Z. Zhang, T. Shi and H. Wang. "Optimal designing of permanent magnet cavity to reduce iron loss
of interior permanent magnet machine" IEEE Trans. Magn., vol. 51, no. 12, pp. 1-9, 2015
11
N. Bianchi and A. Canova. "FEM analysis and optimisation design of an IPM synchronous motor" Proc. Int. Conf.
Power Electron., Mach. Drives, pp. 49-54, 2002
12
F. Parasiliti, M. Villani, S. Lucidi and F. Rinaldi. "A new optimization approach for the design of IPM synchronous
motor with wide constant-power region" Proc. XIX Int. Conf. Elect. Mach., pp. 1-7, 2010
13
F. Dubas, C. Espanet and A. Miraoui. "Design of a high-speed permanent magnet motor for the drive of a fuel
cell air-compressor" IEEE Conf. Veh. Power Prop., p. 8, 2005
14
M. Popescu, I. Foley, D. A. Staton and J. E. Goss. "Multi-physics analysis of a high torque density motor for
electric racing cars" Proc. IEEE Energy Convers. Congr. Expo., pp. 6537-6544, 2015
15
A. Boglietti, A. Cavagnino, D. Staton, M. Shanel, M. Mueller and C. Mejuto. "Evolution and modern approaches
for thermal analysis of electrical machines" IEEE Trans. Ind. Electron., vol. 56, no. 3, pp. 871-882, 2009
16
P. Zhang, G. Sizov, M. Li, D. Ionel, N. Demerdash, S. Stretz and A. Yeadon. "Multi-objective tradeoffs in the
design optimization of a brushless permanent-magnet machine with fractional-slot concentrated
windings" IEEE Trans. Ind. Appl., vol. 50, no. 5, pp. 3285-3294, 2014
17
P. Zhang, G. Sizov, D. Ionel and N. Demerdash. "Establishing the relative merits of interior and spoke-type
permanent-magnet machines with ferrite or ndfeb through systematic design optimization" IEEE Trans.
Ind. Appl., vol. 51, no. 4, pp. 2940-2948, 2015
18
Z. Azar, Z. Zhu and G. Ombach. "Influence of electric loading and magnetic saturation on cogging torque, back-
EMF and torque ripple of pm machines" IEEE Trans. Magn., vol. 48, no. 10, pp. 2650-2658, 2012
19
Y. Duan and D. Ionel. "Nonlinear scaling rules for brushless pm synchronous machines based on optimal design
studies for a wide range of power ratings" IEEE Trans. Ind. Appl., vol. 50, no. 2, pp. 1044-1052, 2014
20
A. Fatemi, D. M. Ionel and N. A. O. Demerdash. "Identification of design rules for interior pm motors with
different cooling systems" IEEE Int. Elect. Mach. Drives Conf. (IEMDC), Coeur d’Alene, ID, pp. 1228-
1234, 2015
21
J. Cros and P. Viarouge. "Synthesis of high performance pm motors with concentrated windings" IEEE Trans.
Energy Convers., vol. 17, no. 2, pp. 248-253, 2002
22
A. EL-Refaie. "Fractional-slot concentrated-windings synchronous permanent magnet machines: Opportunities
and challenges" IEEE Trans. Ind. Electron., vol. 57, no. 1, pp. 107-121, 2010
23
M. Nakano and H. Kometani. "A study on eddy-current losses in rotors of surface permanent magnet
synchronous machines" Proc. IEEE 39th IAS Annu. Meeting Conf. Record Ind. Appl. Conf., vol. 3, pp. 1696-
1702, 2004
24
D. Ionel, M. Popescu, M. McGilp, T. Miller, S. Dellinger and R. Heideman. "Computation of core losses in
electrical machines using improved models for laminated steel" IEEE Trans. Ind. Appl., vol. 43, no. 6, pp.
1554-1564, 2007
25
K. Hirota, H. Nakamura, T. Minowa and M. Honshima. "Coercivity enhancement by the grain boundary
diffusion process to nd amp;8211;fe amp;8211;b sintered magnets" IEEE Trans. Magn., vol. 42, no. 10,
pp. 2909-2911, 2006
26
S. Jurkovic, K. Rahman, N. Patel and P. Savagian. "Next generation voltec electric machines; design and
optimization for performance and rare-earth mitigation" SAE Int. J. Alt. Powertrains, vol. 4, no. 2, pp.
336-342, 2015
27
T. J. Miller and J. R. Hendershot. Design of Brushless Permanent-Magnet Machines. 2010, Motor Design Books
LLC
28
G. Pellegrino and F. Cupertino. "FEA-based multi-objective optimization of IPM motor design including rotor
losses" Proc. IEEE Energy Convers. Congr. Expo., pp. 3659-3666, 2010
29
R. H. Myers and D. C. Montgomery. Response Surface Methodology: Process and Product Optimization Using
Designed Experiments. 2009, Wiley
30
A. Fatemi, D. M. Ionel, N. A. O. Demerdash and T. W. Nehl. "Fast multi-objective cmode-type optimization of
pm machines using multicore desktop computers" IEEE Trans. Ind. Appl., 2016
31
Y. Wang and Z. Cai. "Combining multiobjective optimization with differential evolution to solve constrained
optimization problems" IEEE Trans. Evol. Comput., vol. 16, no. 1, pp. 117-134, 2012