Received: 14 November 2022

DOI: 10.1049/elp2.12298

ORIGINAL RESEARCH
- Accepted: 6 January 2023

- IET Electric Power Applications

Magnetic equivalent circuit and finite element modelling of

anisotropic rotor axial flux permanent magnet synchronous
motors with fractional slot distributed winding

Attila Nyitrai | Miklós Kuczmann

Department of Power Electronics and Electric Abstract

Drives, Széchenyi István University, Győr, Hungary
Axial flux motors have been in the focus of many research work due to their promising
properties when compared to conventional radial flux machines. Axial flux permanent
Correspondence
Attila Nyitrai, Department of Power Electronics and
magnet synchronous motors (AFPMSM) studied in the literature were mostly limited to
Electric Drives, Széchenyi István University, surface‐mounted permanent magnet machines (SPMSM), excluding only a few studies
Egyetem tér 1, Győr 9026, Hungary. about axial flux internal permanent magnet (AFIPM) or permanent magnet‐assisted
Email: [email protected]
synchronous reluctance (AFPMaSynRM) machines. Since numerous rotor topologies
have been proposed and investigated in case of radial flux motors, analogous design
methodologies can be developed in case of axial flux motors with rotor saliency. The
parametrisation and modelling of a magnetically anisotropic rotor axial flux motor are
presented. This approach is described generally and the method is applied to an example
motor geometry. The results were validated by FEA (Finite Element Analysis), using the
2D multi‐slice and 3D modelling approaches. The modelling error was calculated and the
analytical and numerical methods were compared. It was found that the proposed model
was suitable for calculating important parameters of the anisotropic rotor axial flux
permanent magnet synchronous motor (AnR‐AFPMSM) during the design process.

KEYWORDS
AC motors, electric motors, permanent magnet motors

1 | INTRODUCTION 1.1 | Traction drive requirements

This article presents the research work about the analytical Application in traction drive systems sets several special re-
and numerical modelling of magnetically anisotropic rotor quirements for an electric motor design compared to the
permanent magnet axial flux motors (AnR‐AFPMSM) pre- standard industrial ones [2–5]. In addition to strict geometrical
sented in Figure 1. Generally, axial flux (AF) motors have constraints (packaging), specific power, mechanical requir-
received much attention in recent years, mainly due to the ements, noise and vibration characteristics, IP classes and
possible application of the high torque‐density and compact transient winding temperature rise during standard drive cycles,
design they offer in traction drives [1]. However, despite AF a set of electromagnetic design aspects need to be taken into
motors demonstrated several advantages when compared to account when designing traction motors. Most importantly, the
radial flux (RF) motors, due to severe manufacturing diffi- motor must be able to operate in a wide speed range demanded
culties, the proportion of AF motors used in industrial and by the system requirements. The efficiency requirements must
especially in traction applications is still very low. Little be met during the whole operation range. In addition to those
attention has been paid especially on the axial flux motors above, electric motors have strict cost requirements in the
with anisotropic rotor structures. highly competitive automotive industry. Supply chains need to

-
This is an open access article under the terms of the Creative Commons Attribution‐NonCommercial‐NoDerivs License, which permits use and distribution in any medium, provided the
original work is properly cited, the use is non‐commercial and no modifications or adaptations are made.
© 2023 The Authors. IET Electric Power Applications published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology.

IET Electr. Power Appl. 2023;17:709–720. wileyonlinelibrary.com/journal/elp2 709

710
- NYITRAI AND KUCZMANN

ratio need to be defined. The characteristic current (Ich) in case

of an IPMSM is as follows:

λPM
Ich;IPMSM ¼ ; ð2Þ
Ld

where λPM is the permanent magnet flux linkage and Ld is the

d‐axis inductance. The characteristic current in case of a
PMaSynRM is

λPM
Ich;PMaSynRM ¼ ; ð3Þ
Lq

where λPM is the permanent magnet flux‐linkage and (Lq) is the

q‐axis inductance. The saliency ratio:

Lq
ξ¼ ; ð4Þ
Ld

where Lq and Ld are the d‐ and q‐axis inductances, respectively.

F I G U R E 1 3D geometry of the axial flux motor with an anisotropic Some authors already presented a unified terminology [7]
rotor. The 3D coil geometry of the phase winding is indicated. in order to avoid misunderstandings when comparing different
types of motors with rotor saliency, described in Section 1.2.
be planned for years; therefore, reliable and foreseeable prices Since the scope of the current article is AFPM rotor saliency in
of raw materials are of great importance. general, IPM and PMaSynRM terminologies are not used
Since rare Earth elements, as major compounds of high separately. Instead, a unified terminology is used, where the
energy product NdFeB magnets, are supplied by only a few rotor d‐axis is always the axis of the permanent magnets,
participants on the market, the price of permanent magnets, regardless of the proportion of the torque contribution of the
which accounts for approximately up to 70% of the raw ma- magnets.
terial cost of a simple PM traction motor [6], is highly volatile Therefore, the characteristic current:
and unpredictable. Therefore, serious attempts have been made
to avoid or at least reduce the use of rare Earth elements λPM
(REE) in traction motors in order to avoid cost issues during a Ich;AnR−AF PM ¼ ; ð5Þ
Ld
longer manufacturing period. In addition to that, REE mining
especially, due to the weak regulation of the processes in the
supplier countries, has a significant environmental load. where λPM is the permanent magnet flux‐linkage and (Ld) is the
The present research addresses the REE issue by studying the d‐axis (PM axis) inductance.
modelling process of novel axial flux motors with interior Adequate operation with high CPSR requires the following
permanent magnets. This kind of motors is suitable for the two conditions.
reduction of the PM material necessary compared to SPM
motors in case of radial flux machines by utilising the reluc- � The saliency ratio ξ is high enough to generate reluctance
tance torque component originated in rotor saliency. Another torque.
key feature is the field‐weakening capability of such motors, � The characteristic current Ich is close to the rated current of
which will be presented in the next sections. the machine: In = Ich, which is the condition of optimum
flux‐weakening.

1.1.1 | Constant power speed ratio Efficient field‐weakening is not achievable if the charac-
teristic current is significantly larger than the rated current. In
Traction applications require a significant constant power speed this case, a large d‐axis current would be necessary and it
ratio (CPSR). CPSR is the ratio of the maximum speed at which would decrease the motor efficiency and result in a higher
the motor can provide the rated power and the base speed [7]: temperature rise of the motor.

ωmax;Pn
CPSR ¼ : ð1Þ
ωbase 1.1.2 | High‐speed efficiency

In order to determine if the machine is capable of field High‐speed efficiency is important in automotive (e‐mobility)
weakening operation, the characteristic current and the saliency applications [8]. Since conventional surface permanent magnet
NYITRAI AND KUCZMANN
- 711

axial flux motors have less effective flux‐weakening capability counterparts, the already established design and analysis met-
as a consequence of the previously described high character- hods of radial flux machines can be adapted. Only a few au-
istic current, the high‐speed efficiency of these motors is thors have studied anisotropic rotor AF motors; AFIPMSM
rather low compared to radial flux IPMSMs or PMaSynRMs. and AFPMaSynRM machines were compared in [7] consid-
Therefore, the system efficiency of traction drives consisting ering the requirements of traction applications. A 12s10p axial
conventional AF motors is lower than systems with radial flux flux interior permanent magnet motor design was presented in
machines. [11]. The authors designed a motor with a double rotor single
stator (DSSR) structure. The motor has double‐layer concen-
trated windings and a rotor using V‐shaped magnet arrange-
1.2 | AFPMSM terminology ment. In order to better utilise the rotor volume, Tekgun et al.
applied a shift angle between the two opposite magnet layers
As mentioned in Section 1.1.1, the definition of the characteristic and the stators accordingly. It was found that the torque
current is not unified for different motor types due to the density of the AFIPM was lower and the efficiency and CPSR
different definition of direct and quadrature axes. In this section, were higher than the comparable AFSPM – similarly to radial
a unified terminology is presented and applied further in the flux motors. It was also found that the less magnet material was
article. The most simple type of permanent magnet snychronous necessary compared to the AFSPM motor to achieve the same
motors has surface‐mounted PMs. Since the magnetic perme- torque. In [12], the authors presented the design procedure of
ability of the PMs is close to that of air, such rotor structures have an AFPMaSynRM where the resulting design was compared to
different magnetic reluctances in the axis of the magnets (polar the AFIPM described in [11]. It was found that the AFPMa-
or direct axis) than in the axis between the magnets (interpolar or SynRM had higher CPSR and maximum operating speed. The
quadrature axis). Therefore, any rotor with different d‐ and magnet weight was reduced by approximately 50%; however,
q‐axis inductances is capable of producing reluctance torque. the motor current was increased by 33%. The design was
This property of a rotor is called magnetic saliency or anisotropy. optimised using the Taguchi‐method.
Traditionally, a type of wound rotor synchronous motors,
where the poles are salient structures (in contrary of the whole
rotor being cylindrical), is called salient pole motors. However, 1.3 | Modelling of axial flux motors
the rotor surface of the synchronous reluctance motors, inte-
rior permanent synchronous motors (IPMSM) or permanent Generally, there are different aspects of electric motor design
magnet‐assisted synchronous reluctance motors (PMaSynRM) and modelling where the following approaches can be identi-
are cylindrical. However, from the inductance point of view, fied [13–17].
these structures are magnetically salient. Permanent magnet‐
assisted synchronous reluctance motors were developed by � Analytical sizing procedures based on electromagnetic and
inserting PMs (mostly ferrites in industrial applications first) in thermal utilization numbers;
the flux barriers of SynRMs. However, these motors differ � Analytical design methods;
from IPMs only in the saliency ratio, the amount of PMs and � Analytical performance evaluation equations‐based optimi-
therefore the values of PM torque to reluctance torque ratio, zation methods;
CPSR and power factor. The torque production mechanisms � Finite element model‐based optimization methods;
are the same and they are geometrically very similar; therefore, � Reluctance network or magnetic equivalent circuit (MEC)
the modelling considerations are basically the same. On the model‐based parameter identification;
other hand, some authors [9, 10] distinguish them by the di- � Finite element method‐based parameter identification.
rection of the direct and quadrature axes according to the
different terminologies used in case of PMSMs or SynRMs. The characteristics of an anisotropic rotor machine can be
Other authors mentioned that sometimes it is difficult to evaluated from the flux‐linkage maps in a two‐axis (d/q)
distinguish IPMs and PMaSynRMs for the first sight, since model. The d‐ and q‐axis inductances (or flux‐linkage maps)
basically it depends on the proportion of PM and reluctance are fundamental parameters for tuning the motor controller
torques [7]. Therefore, in this article, axial flux motors with parameters too. Therefore, these parameters are not only
different d‐ and q‐axis reluctances are called anisotropic rotor important to determine torque curves or to analyse field‐
axial flux permanent magnet motors (AnR‐AFPMSM), weakening operation, but they are necessary in system simu-
regardless of the PM and reluctance torque ratios. In every lations and control algorithm evaluation as well. The present
case, the d‐axis is defined as the axis of the PMs and therefore article describes methods of parameter identification from the
the PM flux linkage (λPM). list above.
Few authors have already presented studies about param-
eter identification of anisotropic rotor AF motors: parameter
1.2.1 | AnR‐AFPMSM topologies measurements and identification methods of an axial flux
motor (described as AFIPM in the article) were presented in
Since the physical behaviour of axial flux motors with rotor [18]. Based on the prototype measurements, the author found
anisotropy is expected to be analogous to that of radial flux appropriate d‐ and q‐axis equivalent circuits. Due to their
712
- NYITRAI AND KUCZMANN

geometry and the resulting magnetic flux paths, electromag-

netic modelling of axial flux motors is generally a three‐
dimensional problem. However, amongst the 3D modelling
of these machines [19], several approaches were developed in
order to reduce the task to a set of 2D models. An analytical
quasi‐3D model of an AFPMSM was described in [20]. The
model was validated by 2D and 3D FEA. The model was built
using geometrical linearisation: the 3D geometry was built‐up
using linear motor geometry equivalents (slices). It is called
the multi‐slice method in later articles [21]. In such models, a
parametric reluctance network model or MEC is updated at
every slice of the 3D geometry and the results are summed up
to get resultant motor parameters. A quasi‐3D analytical model
using the multi‐slice method was provided by the authors of
[22], where the validity domain of the proposed method was
presented. The disadvantage of the analytical method is the
difficult application to new, more complex geometries. A non‐
linear analytical model of axial flux motors was presented in
[5]. The authors proposed an equivalent circuit model with
analytical calculation of the air gap and the core permeances.
The method was validated by FEA. The authors in [23]
F I G U R E 2 Sectional view of an AnR‐AFPMSM with the permanent
described a combined 2D FEA and analytical method, where magnets and winding. The coils of one phase winding (A) are indicated by
they introduced correction factors to take into account air‐gap blue colour.
flux‐density deviations from the ideal curve. The method was
validated by tests; however, the analysed motor had a very
simple geometry. A reluctance network model (magnetic
equivalent circuit) was provided in [24]. The authors described
a model that could be extended to any kind of AF‐SPMSM.
The novelty of the method was that it enabled calculating the
machines at different rotor angles using a static system.

2 | STRUCTURE AND
PARAMETRIZATION

The AnR‐AFPMSM analysed in this article has a double rotor–

single stator configuration (DRSS). This configuration has an
advantage of zero residual magnetic pull on the rotor. The
yokeless stator is constructed from segments in case of the 3D
analysis. However, the structure can be manufactured either
from segments or one part. In case of double‐rotor yokeless
interior stator axial‐flux motors, the stator core is usually
manufactured from segments. The mechanical (dynamical, F I G U R E 3 Geometrical parameters of the anisotropic rotor axial flux
motor calculated at the mean diameter in case of a two‐magnet‐layer
NVH) behaviour of the two designs is significantly different.
configuration.
The rotor has flux‐barriers in the structure and permanent
magnets inserted into the magnet pockets. In the simplest ri
form, the magnets have a ring sector shape presented in P n ðri Þ ¼ Pðrmean Þ ⋅ ; ð6Þ
rmean
Figures 1 and 2.
where ri is the radius of the ith slice and rmean is the mean
radius.
2.1 | Geometrical parametrisation The analysed geometry in this paper consists of two layers
of flux barriers (and permanent magnets). However, it is not
The geometrical parametrisation is presented in Figure 3. The the optimum number regarding the reduction of torque ripple;
main geometrical parameters of the rotor flux barriers are al- it is used in this study due to its simplicity. More layers can be
ways indicated at the mean diameter of the motor. Using the added in the same way without modification of the described
multi‐slice method, the actual parameter values at the different methods. The values of the main geometry parameters are
diameters need to be recalculated by included in Table 1.
NYITRAI AND KUCZMANN
- 713

2.2 | Winding configuration higher number of poles usually in an outer rotor RF

configuration.
In the following section, a winding configuration suitable for Basically, similarly to radial flux electric motors, axial‐flux
the AnR‐AFPMSM is described. motors can be designed with concentrated or distributed wind-
In traction applications, the number of motor poles are ings. Apart from the differences in the electro‐magnetic prop-
mostly limited by the maximal switching frequency of the drive erties of the two winding methods, due to the geometrical
that is still feasible and economical. The pole‐slot combina- configuration of the axial‐flux motor, the distributed winding
tions were studied by many authors before and the aspects of needs considerable space at the inner diameter that needs to be
the selection of the pole and slot numbers in case of taken into account during the design. From the manufacturing
concentrated windings were summarised in [25]. Usually, and packaging point of view, concentrated winding has clear
higher torque‐density designs can be achieved by increasing the advantages [26]. Since the space necessary for the winding head
number of poles. Currently, the usual number of poles of increases with the slot pitch, a two‐slot‐pitch distributed winding
permanent magnet automotive traction motors with two‐stage can be a compromise that is still feasible when the motor has
gearboxes varies from 6 to 10. Special vehicles of electric appropriate geometrical proportions.
aviation, ship propulsion or certain direct‐drives can apply a Since the present motor has an anisotropic rotor structure,
in order to generate suitable reluctance torque, the winding of
TABLE 1 Geometrical parameters of the AnR‐AFPMSM. synchronous reluctance motors should be studied. Generally,
Geometrical parameters
SynRMs are designed with distributed windings. It was pre-
sented in [27] that the mutual coupling of the windings had an
ID Value Description
important role in reluctance torque generation. It was also
DSo 280 mm Stator outer diameter proved in [28] that the integer slot distributed winding (ISDW)
DSi 185 mm Stator inner diameter can generate higher reluctance torque compared to the frac-
tional slot concentrated winding (FSCW) in SynRMs because
DRo 280 mm Rotor outer diameter
of the poor mutual coupling in case of FSCWs. The authors in
DRi 185 mm Rotor inner diameter [26] presented a comparative study of concentrated and
L 172 mm Motor total length
distributed windings in IPMs, but mostly regarding the loss
components and found that distributed windings have gener-
P1 55 mm Rotor core length ally an advantage in less PM eddy‐current losses.
P2 60 mm Stator core length Therefore, the authors in [27, 29] proposed a PM‐assisted
P3 1 mm Airgap length
radial flux synchronous reluctance motor with 24 slots and 10
poles. The motor had a fractional slot distributed winding
P4 26.4 mm Magnet 1 width (FSDW). The 24S10P winding layout was derived from the
P5 55 mm Magnet 2 width well‐known 12S10P fractional slot concentrated winding
P6 5 mm Magnet 1 thickness
(FSCW) layout. The winding layout was compared to different
topologies. Compared to the integer slot distributed winding
P7 8 mm Magnet 2 thickness (ISDW), the FSDW was found to reduce the torque ripple (due
P8 34.8 mm Magnet 1 position to its MMF harmonics characteristics) while keeping a
reasonable amount of reluctance torque.
P9 60.5 mm Magnet 2 position
In the present study, the winding layout of the radial‐flux
P10 42° Barrier 1 angle motor presented in [27, 29] was adopted to an axial‐flux
P11 24° Barrier 2 angle AnR‐AFPMSM. The winding arrangement is presented in
Figure 4. The winding factors are determined using the star of
P12 22 mm Tooth width
slots concept (Figure 5) described in [30]. During the

FIGURE 4 Winding configuration of the AnR‐AFPMSM.

714
- NYITRAI AND KUCZMANN

T A B L E 2 Fractional slot distributed winding parameters of the

AnR‐AFPMSM.

Winding parameters
Type FSDW

Number of slots 24

Winding layers 2

Configuration AA∣bb∣aC∣Ba∣cc∣AA∣Cb∣aC∣BB∣cc∣bA∣Cb …

calculation of inductances, one turn per phase was consid-

ered. The winding parameters are summarized in Table 2.

3 | MAGNETIC EQUIVALENT CIRCUIT

MODELING

The electromagnetic properties, such as inductances and tor-

F I G U R E 5 Winding configuration of the AnR‐AFPMSM used in the
que, can be traditionally calculated by analytical models. These
star of slots analysis.
models can be magnetic equivalent circuits (MEC) or reluc-
tance networks (RN). The MEC is generally a simple circuit of
magnetic reluctances and MMF sources. Reluctance networks applied the modified winding function theory. If the motor has
on the other hand apply a finer decomposition of the geom- significant magnetic saturation, a non‐linear model is neces-
etry, taking into account geometric alteration such as stator sary: a saturating lumped‐parameter reluctance network model
slots and teeth during the rotation [31]. The quasi‐3D MEC of radial flux interior permanent magnet motors was presented
models can take into account the slotting, magnetic saturation, in [34]. In the study, the d‐ and q‐axis inductances and the
leakage fluxes and cross‐saturation of the d‐ and q‐axes [15]. generated torque were compared with the results of the finite
The reluctances of the magnetic equivalent circuit in the element analysis. It was found that the proposed model was
current model are determined using the flux tube concept. suitable to predict the motor performance under high satura-
The reluctance of a flux tube can be calculated by the tion. In the present article, a saturating MEC was implemented;
following formula: however, the cross‐saturation effects were neglected. The
Z l cross‐saturation effects on radial flux IPM motor performance
1
R¼ dx; ð7Þ were analysed in [9].
0 μðBÞS x
Applying the multi‐slice method, the reluctances of the
where μ(B) is the non‐linear flux‐density‐dependent perme- magnetic equivalent circuit can be calculated by the following
ability of the material and S(x) is the cross‐sectional area along equations described in [15]:
the flux tube.
The permanent magnet MMF can be calculated by the 1 Θ
RΘ ¼ ; ð10Þ
following formula [32]: μðBÞ Lz ⋅ log RRo
i

Br lPM where RΘ is the reluctance in the tangential direction, Lz is the

FPM ¼ ; ð8Þ
μPM width in axial direction and Ro and Ri are the outer and inner
diameters of the slice, respectively.
where Br is the remanent flux‐density, lPM is the magnet length
The axial reluctances can be calculated by [15]
and μPM is the PM magnetic permeability.
The fluxes can be calculated by the following matrix
equation: 1 l
Rz ¼ ; ð11Þ
μðBÞ Rav ⋅ Θ ⋅ Lslice
Φ ¼ RðΦÞ−1 Θ; ð9Þ
where Rz is the reluctance in the axial direction.
where Φ is the flux vector and R is the reluctance matrix and The two‐axis MEC model is presented in Figures 7 and 8
Θ is the MMF vector. based on the simplified geometries in Figure 6. The reluctances
In order to prepare a magnetic equivalent circuit of the and MMF sources are applied according to the figures in order
AnR‐AFPMSM, the similar models developed by authors for to determine d‐ and q‐axis flux linkages and the permanent
radial flux machines can be adapted. An inductance‐based magnet flux linkage. Saturation generally occurs in the main
analysis of IPM machines was presented in [33]. The author components of the q‐axis: the q‐axis model is necessarily non‐
NYITRAI AND KUCZMANN
- 715

F I G U R E 6 Simplified geometrical representation of the d‐ and q‐axis models. The position of the d axis is indicated by blue colour, the position of the q‐
axis is indicated by red colour.

linear. In the d‐axis, the model can be built with linear reluc- rffiffi 2 cosðΘÞ −sinðΘÞ 1
3
2 4
tance elements assuming saturation of the saturation bars; C¼ ⋅ cosðΘ − 2π=3Þ −sinðΘ − 2π=3Þ 1 5: ð13Þ
however, the more accurate modelling approach is to use the 3
cosðΘ − 4π=3Þ −sinðΘ − 4π=3Þ 1
same non‐linear material models as in the q axis. In Figures 7
and 8, the theoretical contour lines of the flux barriers are The transformation can be applied to analytical and nu-
indicated by blue dashed lines. merical model results. Generally, the solution of a magneto-
static analysis is an inductance matrix in the abc frame of
reference. In the current study, the 2D and 3D finite element
4 | APPLICATION OF THE dq‐MODEL models resulted in non‐diagonal Labc inductance matrices.
ON THE AnR‐AFPMSM However, the analytical magnetic equivalent circuit model does
not take into account cross‐saturation; therefore, the resulting
The following calculations were carried out applying the dq‐axis matrices were diagonal.
modelling theory; therefore, sinusoidal flux‐linkages were � �
assumed. Applying the finite element method, the direct‐ and Ldd ðiÞ Ldq ðiÞ
LðiÞdq ¼ : ð14Þ
quadrature‐axis inductances can be derived from the flux link- Lqd ðiÞ Lqq ðiÞ
ages or can be transformed from the inductance matrix resulted
from the magnetostatic or transient magnetic analysis. The apparent inductances can be also calculated by the flux
linkages and currents as follows.

4.1 | Calculation of d‐ and q‐axis

inductances
Ψd − λPM
Ldd ¼ ; ð15Þ
The inductance matrix can be transformed from the abc frame id
of reference to the dq frame of reference by the following
equation [35]: Ψd − λPM
Ldq ¼ ; ð16Þ
iq
Ldq ¼ C T Labc C; ð12Þ
Ψq
Lqd ¼ ; ð17Þ
where the transformation matrix: id
716
- NYITRAI AND KUCZMANN

FIGURE 8 The d‐axis reluctance network model.

FIGURE 7 The q‐axis reluctance network model.

Ψq
Lqq ¼ : ð18Þ
iq where m is the number of phases. By introducing the saliency
ratio ξ = Lq/Ld:
If the cross‐saturation effect is neglected, the inductance
matrix can be simplified as follows: m �
� � Tdq ¼ p λPM Iq þ ð1 − ξÞLd id iq : ð22Þ
Ldd ðiÞ 0 2
LðiÞdq ¼ ð19Þ
0 Lqq ðiÞ:

4.2 | Torque calculation 5 | FINITE ELEMENT MODELING

The electromagnetic torque can be calculated from the classical In the present paper, the motor parameters were calculated
equations (derived from the cross product of the flux‐linkage from 2D and 3D magnetostatic and transient magnetic ana-
and current vectors): lyses. The general method to calculate the flux maps or dq
inductances is to set up a linear space of d‐ and q‐axis currents
m � and apply the transformed phase currents to the finite element
Tdq ¼ p Ψd Iq − Ψq Id ; ð20Þ
2 model. The result is a flux‐linkage map in the d‐ and q‐axes or
inductance matrices in the abc stationary reference frame.
where Tdq is the electromagnetic torque from the dq model The currents were transformed between the two refer-
and Ψd and Ψq are the d‐ and q‐axis flux linkages, respectively. ence frames by applying the following equations. The d‐ and
Substituting the flux linkages in the above equation, q‐axis currents can be calculated from the RMS stator cur-
the maximum electromagnetic torque can be calculated from the rent Is.
inductances and currents directly using the following equation:
pffiffi
id ¼ − 3Is sinðγÞ; ð23Þ
m � � pffiffi
Tdq ¼ p λPM Iq þ Ld − Lq id iq ; ð21Þ iq ¼ 3Is cosðγÞ; ð24Þ
2
NYITRAI AND KUCZMANN
- 717

where γ is the angle between the stator current phasor and the geometry was remeshed adaptively in order to minimise energy
rotor direct axis. The phase currents can be calculated by the errors (adaptive refinement). By applying the adaptive refine-
following equations. ment, the size of the initial mesh approximately doubled during
pffiffi the solution process (Table 4). The torque‐current angle curves
i1 ¼ 2Is sinðθ þ γÞ; ð25Þ are presented at different current values (Figure 9) and sepa-
pffiffi rated to torque components (Figure 10).
i2 ¼ 2Is sinðθ þ γ þ 2π=3Þ; ð26Þ The torque‐current angle curves are presented in Figure 10.
pffiffi The shift of the current angle at maximum torque can be
i3 ¼ 2Is sinðθ þ γ þ 4π=3Þ; ð27Þ observed as the saturation has greater effects at higher current
values.
where θ is the electrical angle between the phase A axis and the
rotor direct axis. The phase currents were calculated from the TABLE 4 2D finite element model parameters.
d‐ and q‐axis current values and applied to the model directly 2D FE model parameters
as excitations.
Analysis type Magnetostatic

Non‐linear residual 10−5

5.1 | 3D models Mesh type Triangle

Number of mesh elements 5735/10709 (refined)

In order to validate 2D multi‐slice finite element models and
analytical (MEC) models, a high‐fidelity 3D finite element
model has been built in Ansys AEDT (Maxwell) software. The
main parameters of the model are presented in Table 3. The
parameterised geometry used is presented in Figure 1. The 3D
model is a full geometry transient electromagnetic model. The
rotor has been defined as a rotating component and time‐
dependent current excitation has been applied to the phase
windings. The flux linkages and phase inductances were
computed at every time step. The torque curve has been
computed by applying a parameter sweep of current angle (γ)
and stator current (Is). In case of the 3D finite element analysis,
the solution errors can be generally reduced by applying suit-
able mesh operations to increase the mesh density locally or by
using an adaptively refined mesh from a magnetostatic solu-
tion. Mesh refinement in every time step of a transient analysis
would be too time‐consuming for a simple computer.

FIGURE 9 Torque‐current curves at different current values.

5.2 | 2D equivalent models
The 2D multi‐slice method is very effective to evaluate the main
characteristics of the motor without running a full 3D transient
electromagnetic analysis. In the present study, a 2D slice of the
3D geometry based on a parameterised 2D geometry (Figure 3)
was used. The main parameters of the magnetostatic analysis is
presented in Table 4. In order to calculate the motor parameters
in case of different stator currents (Is) and current angles (γ), a
parametric study was prepared. At each single analysis, the

TABLE 3 3D finite element model parameters.

3D FE model parameters
Analysis type Transient

Non‐linear residual 10−5

Element type Tetrahedra

Number of mesh elements 515,696 F I G U R E 1 0 Comparison of the torque components calculated using
the different models.
718
- NYITRAI AND KUCZMANN

FIGURE 13 The d‐axis flux linkages in function of d‐ and q‐axis

F I G U R E 1 1 D‐axis inductances in function of the peak stator current currents.
at different current angle values.

FIGURE 14 The q‐axis flux linkages in function of d‐ and q‐axis

currents.
F I G U R E 1 2 Q‐axis inductances in function of the peak stator current
at different current angle values.

should be selected to create flux linkage and torque maps in

The flux linkages and the inductances were calculated at order to use them in motor control applications. A parametric
every parameter setup: total currents from 0 to 5000A and finite element model was prepared that is suitable to take into
current angles from 0° to 90°. The d‐ and q‐axis inductance account material non‐linearity and cross‐saturation effects. In
curves are shown in Figures 11 and 12, respectively. The Table 5, the calculated parameter values of the MEC and the
flux linkages are presented in the flux‐linkage plots Figures 13 2D FEM are given at 50% of the peak stator current used in
and 14. this article in parameter sweeps. The flux linkages were
calculated directly from the MEC model and the inductances
were derived by Equations (15) and (18), ignoring cross‐
6 | COMPARISON OF THE MODELS coupling terms. However, the equations cannot be used,
when Id or Iq are zero or close to zero. In that case, other
The motor parameters by applying different analytical and calculations with modified current angles need to be carried
numerical models were compared. It was found that the out. In case of the 2D or 3D finite element models, the pa-
analytical magnetic equivalent circuit method is suitable for rameters can be either calculated from the flux linkages or by
sizing of AnR‐AFPMSMs. However, the simple MEC method the transformation of the phase inductances applying Equa-
described in this article does not take into account cross‐ tion (12). The torque values at a given operating point can be
saturation behaviour. Therefore, the finite element method calculated from Equation (20) or (21). In case of the finite
NYITRAI AND KUCZMANN
- 719

T A B L E 5 Comparison of the 2D FEA and MEC models of the

AnR‐AFPMSM at 50% current. C O N F L I C T O F I N TE R E S T S TA TE M E N T
The authors do not have a conflict of interest to disclose.
Model comparison
Parameter 2D FEM MEC DA TA AVA I L A B I L I T Y S TA TE M E N T
Phase current, I[A] 2500
Data sharing is not applicable to this article as no datasets were
generated or analysed during the current study.
D‐axis current, Id[A] −1767.67

Q‐axis current, Iq[A] 1767.67 O R CI D

PM flux linkage, λPM[Wb] 0.027 0.0214
Attila Nyitrai https://orcid.org/0000-0003-0418-847X

D‐axis flux linkage, ξd[Wb] 0.0163 0.0158

Q‐axis flux linkage, ξq[Wb] 0.0273 0.0268 R EF E R E N C E S

1. Kahourzade, S., et al.: A comprehensive review of axial‐flux permanent‐
D‐axis inductance, Ld[μH] 5.09 3.17 magnet machines. Can. J. Electr. Comput. Eng. 37(1), 19–33 (2014).
Q‐axis inductance, Lq[μH] 13.33 15.17 https://doi.org/10.1109/cjece.2014.2309322
2. Chis, M., et al.: Efficiency optimization of EV drive using fuzzy logic.
Torque, T[Nm] 552.0 564.6 Conf. Rec. – IAS Annu. Meet. (IEEE Ind. Appl. Soc.) 2, 934–941 (1997)
3. Lei, G., et al.: A review of design optimization methods for electrical
machines. Energies 10(12), 1962 (2017). https://doi.org/10.3390/
en10121962
element method, the torque values were computed directly by
4. Sieklucki, G.: Optimization of powertrain in EV. Energies 14(3), 1–12
the Maxwell stress or the virtual work methods. It was found (2021). https://doi.org/10.3390/en14030725
that taking the 3D finite element model as the reference, the 5. Kano, Y., Kosaka, T., Matsui, N.: A simple nonlinear magnetic analysis
2D methods have an approximately 10% error in terms of for axial‐flux permanent‐magnet machines. IEEE Trans. Ind. Electron.
electromagnetic torque, where the additional error of the 57(6), 2124–2133 (2010). https://doi.org/10.1109/tie.2009.2034685
6. Widmer, J.D., Martin, R., Kimiabeigi, M.: Electric vehicle traction motors
nonlinear MEC model is within 5% compared to the 2D FEM.
without rare earth magnets. Sustain. Mater. Technol. 3(7–13), 7–13
(2015). https://doi.org/10.1016/j.susmat.2015.02.001
7. Huynh, T.A., Hsieh, M.F.: Comparative study of PM‐assisted SynRM and
7 | CONCLUSION IPMSM on constant power speed range for EV applications. IEEE
Trans. Magn. 53(11), 1–6 (2017). https://doi.org/10.1109/tmag.2017.
2707125
In the present study, an anisotropic dual‐rotor axial flux per-
8. Oki, S., Ishikawa, S., Ikemi, T.: Development of high‐power and high‐
manent magnet synchronous motor (AnR‐AFPMSM) was efficiency motor for a newly developed electric vehicle. SAE Technical
presented. The motor topology was synthesised from the ad- Papers 1(1), 104–111 (2012). https://doi.org/10.4271/2012‐01‐0342
vantageous properties of radial flux interior PM motors and 9. Armando, E., et al.: Accurate modeling and performance analysis of
yokeless axial flux motors. Based on the results of research work IPM‐PMASR motors. IEEE Trans. Ind. Appl. 45(1), 123–130 (2009).
https://doi.org/10.1109/tia.2008.2009493
with FSDW configurations and the numerous studies on IPM
10. Pellegrino, G., et al.: Performance comparison between surface‐mounted
and PMaSynRM machines, an axial flux motor topology was and interior PM motor drives for electric vehicle application. IEEE
proposed. A simple non‐linear analytical magnetic equivalent Trans. Ind. Electron. 59(2), 803–811 (2012). https://doi.org/10.1109/tie.
circuit model was developed and in addition 2D and 3D finite 2011.2151825
element models were developed and the results compared. It 11. Tekgun, B., et al.: Design of a novel interior permanent magnet axial flux
machine. In: 2017 IEEE Energy Conversion Congress and Exposition,
was found that the proposed motor topology can be considered
ECCE 2017, pp. 314–320 (2017)
in traction drives and the analysis methodologies can be 12. Tarek, M.T.B., Yilmaz, S.: Design of a novel axial flux permanent magnet
adopted. The topology has the advantages of axial flux and assisted synchronous reluctance motor. In: 2019 IEEE Energy Conver-
interior PM rotor radial flux motor designs in terms of possible sion Congress and Exposition, ECCE 2019 (2019)
high torque density and high saliency ratio and therefore 13. Paar, C., Muetze, A.: Quantification of external heat load on HEV in-
tegrated IPMs using the air‐gap shear stress. IEEE Trans. Ind. Appl.
increased reluctance torque and a possibility to achieve high
53(3), 1909–1919 (2017). https://doi.org/10.1109/tia.2017.2656069
CPSR while using a reduced amount or grade of permanent 14. Krasopoulos, C.T., Beniakar, M.E., Kladas, A.G.: Multicriteria PM motor
magnet material. On the other hand, precise analysis of other design based on ANFIS evaluation of EV driving cycle efficiency. IEEE
aspects, such as efficiency, thermal and mechanical behaviours, Trans. Transport. Electrif. 4(2), 525–535 (2018). https://doi.org/10.
needs to be carried out in the scope of future research. 1109/tte.2018.2810707
15. Benlamine, R., et al.: Modeling of an axial‐flux interior PMs machine for
an automotive application using magnetic equivalent circuit. In: 2015
AUTH O R CO NT R I B U TI O N S 18th International Conference on Electrical Machines and Systems,
Attila Nyitrai: Conceptualization; Data curation; Investiga- ICEMS 2015, pp. 1266–1271 (2018)
tion; Methodology; Validation; Visualization; Writing – original 16. Chen, Q., et al.: Design and multi‐object optimisation of axial flux
draft. Miklós Kuczmann: Supervision; Validation; Writing – interior PMSM for EV and HEV applications. J. Eng. 2017(13),
2215–2220 (2017). https://doi.org/10.1049/joe.2017.0724
review and editing.
17. Nyitrai, A., Gergely, S., Sándor, R.H.: Parameter determination and drive
control analysis of axial flux permanent magnet synchronous motors.
ACKNO W L E DG EMEN T S Period Polytech. Electr. Eng. Comput. Sci. 66(2), 205–214 (2022).
There is no funding to report for this submission. https://doi.org/10.3311/ppee.19714
720
- NYITRAI AND KUCZMANN

18. Cavagnino, A., et al.: Axial flux interior PM synchronous motor: parameters 29. Gamba, M., Pellegrino, G., Vagati, A.: A new PM‐assisted synchronous
identification and steady‐state performance measurements. IEEE Trans. reluctance machine with a nonconventional fractional slot per pole
Ind. Appl. 36(6), 1581–1588 (2000). https://doi.org/10.1109/28.887209 combination. J. Electr. Eng. 15(1), 97–104 (2015)
19. Rhyu, S.H., et al.: Design and analysis of axial flux permanent magnet motor 30. Bianchi, N., Dai Pré, M.: Use of the star of slots in designing fractional‐
for electric bicycles with hybrid stator core. In: 2017 20th International slot single‐layer synchronous motors. IEE Proc. Elec. Power Appl.
Conference on Electrical Machines and Systems, ICEMS 2017 (2017) 153(3), 459 (2006). https://doi.org/10.1049/ip‐epa:20050284
20. Parviainen, A., Niemelä, M., Pyrhönen, J.: Modeling of axial flux 31. Vuljaj, D., et al.: Modelling of cross saturation effect in interior perma-
permanent‐magnet machines. IEEE Trans. Ind. Appl. 40(3), 1333–1340 nent magnet synchronous machines using magnetic equivalent circuits.
(2004). https://doi.org/10.1109/tia.2004.834086 In: Proceedings – 2018 23rd International Conference on Electrical
21. Tiegna, H., et al.: Analytical modeling of the open‐circuit magnetic field Machines, ICEM 2018, pp. 840–846 (2018)
in axial flux permanent‐magnet machines with semi‐closed slots. IEEE 32. Ullah, N., et al.: Analytical modelling of open‐circuit flux linkage, cogging
Trans. Magn. 48(3), 1212–1226 (2012). https://doi.org/10.1109/tmag. torque and electromagnetic torque for design of switched flux permanent
2011.2171979 magnet machine. J. Magn. 23(2), 253–266 (2018). https://doi.org/10.
22. Tiegna, H., Amara, Y., Barakat, G.: Validity domain of a quasi‐3D mul- 4283/jmag.2018.23.2.253
tislice analytical model for synchronous axial flux machines with trape- 33. Ge, H.: Fractional slot concentrated winding interior permanent magnet
zoidal magnets. EPJ Appl. Phys. 70(1), 1–11 (2015). https://doi.org/10. machines with reluctance torque: Inductance‐based methodology for
1051/epjap/2015140495 comprehensive analysis, design, and control
23. Egea, A., et al.: Axial‐flux‐machine modeling with the combination of 34. Lovelace, E.C., Jahns, T.M., Lang, J.H.: A saturating lumped‐
FEM‐2‐D and analytical tools. IEEE Trans. Ind. Appl. 48(4), 1318–1326 parameter model for an interior PM synchronous machine. IEEE
(2012). https://doi.org/10.1109/tia.2012.2199450 Trans. Ind. Appl. 38(3), 645–650 (2002). https://doi.org/10.1109/tia.
24. Ahmed, H., et al.: A simple and efficient quasi‐3D magnetic equivalent 2002.1003413
circuit for surface axial flux permanent magnet synchronous machines. 35. Drobnič, K., Gašparin, L., Fišer, R.: Fast and accurate model of interior
IEEE Trans. Ind. Electron. 66(11), 8318–8333 (2019). https://doi.org/ permanent‐magnet machine for dynamic characterization. Energies
10.1109/tie.2018.2884212 12(5), 783 (2019). https://doi.org/10.3390/en12050783
25. Libert, F., Soulard, J.: Investigation on pole‐slot combinations for
permanent‐magnet machines with concentrated windings. In: Interna-
tional Conference on Electrical Machines, pp. 5–8 (2004)
26. Choe, Y.Y., et al.: Comparison of concentrated and distributed winding in
an IPMSM for vehicle traction. Energy Proc. 14, 1368–1373 (2012). How to cite this article: Nyitrai, A., Kuczmann, M.:
https://doi.org/10.1016/j.egypro.2011.12.1103 Magnetic equivalent circuit and finite element modelling
27. Zhu, S., et al.: A PM‐assisted synchronous reluctance motor with two of anisotropic rotor axial flux permanent magnet
slot‐pitch winding. In: 23rd International Conference on Electrical Ma-
synchronous motors with fractional slot distributed
chines and Systems, ICEMS 2020, pp. 206–211 (2020)
28. Gamba, M., et al.: Synchronous Reluctance Motor with concentrated winding. IET Electr. Power Appl. 17(5), 709–720
windings for IE4 efficiency. In: 2017 IEEE Energy Conversion Congress (2023). https://doi.org/10.1049/elp2.12298
and Exposition, ECCE 2017, pp. 3905–3912 (2017)