Alexandria Engineering Journal (2020) 59, 589–600

H O S T E D BY
Alexandria University

Alexandria Engineering Journal

www.elsevier.com/locate/aej
www.sciencedirect.com

ORIGINAL ARTICLE

A fully coreless Multi-Stator Multi-Rotor

(MSMR) AFPM generator with combination of
conventional and Halbach magnet arrays
Asiful Habib a, Hang Seng Che a,*, Nasrudin Abd Rahim a, Mahdi Tousizadeh a,
Erwan Sulaiman b

a
Higher Institution Centre of Excellence (HICoE), UM Power Energy Dedicated Advanced Center (UMPEDAC), University of
Malaya, Kuala Lumpur, Malaysia
b
Research Centre for Applied Electromagnetics, Universiti Tun Hussein Onn Malaysia (UTHM), Johor, Malaysia

Received 25 October 2019; revised 14 January 2020; accepted 23 January 2020

Available online 21 February 2020

KEYWORDS Abstract An axial flux permanent magnet (AFPM) generator is known to be a good candidate for
Axial flux; both low and high-speed application. In this paper, a new design of fully coreless multi-stator multi-
Coreless; rotor (MSMR) AFPM generator has been presented with conventional and Halbach magnet
Permanent magnet genera- arrangement combined with an ironless (epoxy) rotor. For MSMR topology of AFPM machine,
tor; back iron is still present in the middle rotor and to maintain the same distribution of magnetic flux,
Multi stator-multi Rotor; magnets are using on both side of the middle rotor. This paper suggests replacing the middle iron
Conventional array; rotors with a single epoxy rotor which reduces the weight of the machine, hence increase power den-
Halbach array sity. On the other hand, for the elimination of iron and magnets in the middle rotor, conventional
magnet arrays are used to maintain a continuous flux path. In addition, Halbach array is adopted
on the external epoxy rotor to reduce flux leakage on the external sides of the machine. The perfor-
mance of the proposed generator is investigated in terms of voltage, current, power, power density,
and torque ripple. The analytical design approach is first presented and subsequently validated
using ANSYS MaxwellÒ electromagnetic finite element analysis (FEA) software. It is found that
a fully coreless MSMR AFPM generator with conventional and Halbach magnet arrays gives
higher power density and lower torque ripple with a reduced axial length which is favorable in wind
power and pico-hydro applications.
Ó 2020 The Authors. Published by Elsevier B.V. on behalf of Faculty of Engineering, Alexandria
University. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
licenses/by-nc-nd/4.0/).

1. Introduction
* Corresponding author.
E-mail address: [email protected] (H.S. Che).
AXIAL flux permanent magnet (AFPM) machines with core-
Peer review under responsibility of Faculty of Engineering, Alexandria
less topology have some noticeable advantages compared to
University. other conventional PM machines, such as higher power/torque
https://doi.org/10.1016/j.aej.2020.01.039
1110-0168 Ó 2020 The Authors. Published by Elsevier B.V. on behalf of Faculty of Engineering, Alexandria University.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
590 A. Habib et al.

densities, torque-to-weight ratios, and geometrically higher The paper is organized in the following sequence: the fun-
aspect ratios [1]. An interesting feature of the AFPM machine damental concepts of Halbach and conventional magnet
is the ease to cascade multiple stages of the machine to attain arrangements are first discussed in Section 2. The analytical
higher torque/power, such as multi-stator multi-rotor calculations are given in Section 3. The conventional and fun-
(MSMR) configuration. Among the different AFPM topolo- damental structure of a MSMR coreless AFPM, as well as the
gies [2], multi-stator multi-rotor (MSMR) configuration is con- proposed MSMR topology is presented in Section 4. All the
sidered to be not only highly efficient but also has high comparison, results and discussions for the proposed fully
mechanical strength and high power density without increasing coreless MSMR have been given in Section 5. Finally, in Sec-
diameter of the machine [2,3]. Ship propulsions, aircraft tion 6 conclude the whole design.
propulsions, pumps, wind energy generation, low and high-
speed PM generators are the application of multistage AFPM 2. Conventional and Halbach array
machines [4]. The construction of a multistage AFPM machine
can be made using either double stator single rotor (DSSR) or The arrangement of magnets is an important design criterion
single stator double rotor (SSDR) configurations [5]. The for AFPM machine, and two arrangements are considered
MSMR AFPM machines can have different topologies such here: the ’conventional array’ and the Halbach array. For con-
as slotted or slotless, iron core or ironless, NN or NS topolo- ventional array, the magnets are arranged with alternating
gies while their flux paths are the same as their single-stage north and south poles, either in radial (Fig. 1(a)) or tangential
structures [2]. In particular, coreless AFPM is gaining popular- direction (Fig. 1(b)). The ’Halbach Array’ can be considered as
ity in applications that require high power/torque density. a combination of two conventional arrays (radial and tangen-
Being coreless, the machine does not experience eddy cur- tial), as illustrated in Fig. 1(c). The arrangement of the magnet
rent (hence hysteresis losses) and has lower cogging torque. poles in the Halbach array helps to strengthen the field in one
This allows the machine to run at higher efficiency compared side of the array while canceling out the field on the other side
to other conventional machines [6]. Moreover, the coreless (Fig. 1(d)) [19].
structure reduces the weight of the machine significantly mak- According to the structure of the machine, the two-disc
ing it portable, which facilitates its deployment. In the litera- rotors associated with magnets have an axial gap between
ture, coreless AFPM machine usually refers to ‘‘stator them to fit the coil with an optimized physical gap between
coreless” where ferromagnetic structure (i.e. back-iron) is still the magnet discs. Normally iron core rotors enhance the mag-
used in the rotor [6–14]. A fully coreless design, where both netic field and as a result, higher air-gap magnetic flux density
rotor and stator uses non-magnetic materials, will reduce leads to a higher power. In the case of a conventional array,
the machine weight as well as cost. Furthermore, mechanical the magnetic field is active for both side of the array and the
stress also reduces as the machine does not produce attractive use of epoxy for the rotors leaves one side of the field unused.
forces between the two rotors and stator except between the
permanent magnets on the rotors. However, the use of a non-
magnetic rotor will result in significant flux leakage on the
back surface of the rotor and deteriorates the performance
of the machine. The use of a Halbach magnet array provides
an interesting solution to the issue: by amplifying the mag-
netic field on the useful side of the rotor and canceling the
magnetic field on the back of the rotor. Several works on Radial arrangement of north &
the use of Halbach array for AFPM has been presented in south pole
(a)
[15–19].
Though multistage AFPM machines are not widely found
in the literature, several research had been done on it, which N

shows its scope of possibility for research. In the literature

[5,20–25] the review, modeling, design, analysis have been Tangential arrangement of north &
south pole
found on multistage AFPM machine, where they have differ- (b)
S
ent stator (slot/slotless) and rotor (NN and NS type) structure.
To the best of the author’s knowledge a fully coreless/ironless
multistage AFPM machine has not been reported in the previ-
ous literature.
It is interesting to design a MSMR fully coreless AFPM Both Radial &Tangential
arrangement of north & south pole
generator using combinations of conventional and Halbach (c)
magnets with an ironless (Epoxy) rotor, to get a high power
density in AFPM generator. For this purpose, relatively lower
speed of 500 rpm has been chosen that will reduce the number
of poles which in turn alleviate the design and fabrication com-
Halbach arrangement of north &
plexity. The design has been carried out with respect to torque, south pole
voltage, current, power, machine weight and power density. (d)
The whole design process presented in the analytical equation
and validated with 3-D finite element analysis (FEA) software Fig. 1 Schematic of conventional and Halbach array flux
named ANSYS Maxwell. pattern.
Multi-Stator Multi-Rotor (MSMR) AFPM generator 591

That is why for conventional array iron core rotors are prefer- Wpm Wpm
ap
able for an AFPM machine. For ironless rotor structure, Hal-
bach array can be useful by canceling out the fluxes on the
back of the rotor and strengthen up the field to the active site
X X
of the array, thus help to obtain high magnetic flux density in
the air gap that leads to the high power density [26]. Fig. 2
shows the schematic of rotor poles with an opposite arrange-
ment (N-S type) and the associated flux paths for the Halbach la la
‘la’ for Halbach array ‘la’ for Conventional array
epoxy rotor generator. The epoxy materials used in stator and
rotor are temperature and pressure resistant. Apart from that, (a) (b)
the pole pair formation due to magnets is different from one
another. In conventional array, two magnets create one pole, Fig. 3 Schematic of pole pair wavelength (a) Halbach and (b)
whereas in Halbach array four magnets create one pole. conventional.
Fig. 3 shows the pole pair wavelength (la) of both arrays. Here
magnet width to pole pitch ratio (ɑP) defines the gap between
the magnets. For conventional array an optimum gap (ɑP) The rotating speed Ns is used to determine the pole number
between the magnets is necessary to get a high air-gap mag- using (2) and from pole number, the coil number can be easily
netic flux, on the other hand for Halbach array to strengthen determined by (3).
the field in one side of the array, the gap (ɑP) should be as less 120f
as possible. Ns ¼ ð2Þ
p

3. Analytical design of AFPM generator 3

Q¼ p ð3Þ
4
For analytical design, some inceptive design assumptions where f is the frequency, p is the pole number and Q is the
based on sizing equations are necessary. The parameters for number of the coil. The electrical loading (Am) is used for
initial design are the phase number (m), output power (Po), the sizing equation [27].
speed (N), magnetic loading (Bg), electrical loading (Am), ratio
of inner and outer diameter (k), power factor (cosu). For Am ¼ 4mNcph Iph =pDout ð1 þ kÞ ð4Þ
machine design, the main parameter to be decided first is diam- where Ncph is the number of conductor per phase Iph is the cur-
eter (Dout). Based on the sizing equation the Dout can be deter- rent per phase. From Eq. (5) Nc can be determined as
mined as [3,27]:
!13 paw Dout ð1 þ kÞAm
Nc ¼ ð5Þ
p0 4QIph
Dout ¼ p

 ð1Þ
K B A g f 1  k2 1þk
2 size g m p 2 where aw is the number of parallel current paths.
where Ksize = ke.ki.kp For sinusoidal waveform, the values of
ke, ki, kp ki are briefly explained in [3,27] where maximum 3.1. Magnet geometry
parameter’s value required in Eq. (1) comes from the very ini-
tial assumption. The air gap magnetic flux density Bg is nor- The air-gap magnetic flux density Bg and PM axial height hpm
mally dependent on the magnet geometry as well as magnet has a strong contribution as shown in
grade.
l Bg ðLsc þ 2gÞ
The ratio of inner and outer diameter (k) is considered as 
hpm ¼ m  ð6Þ
B
another important parameter for AFPM machine design in 2 0:9Br  Kpmg
maximizing the output power. From previous research, opti-
mized values of k were chosen for different AFPM design, where lm is the magnet permeability, Br is PM residual flux
p p density, g is the air-gap length and Kpm is PM leakage flux fac-
out of which, 1/ 3 and 1/ 2.5 are common for the most of
the configurations [28–30]. In the present study, the value of tor. As the machine is pure coreless the inner and outer diam-
p eter is actually the magnet outer and inner diameter. In this
k is chosen 1/ 3, for getting maximum power.
case, the magnet has a trapezoidal shape and is defined by
the inner and outer magnet width as shown in Fig. 4. The outer
magnet width (wpmo), inner magnet width (wpmi) and magnet
Epoxy Rotor
length (lpm) can be calculated as
N

w pmi

S
Epoxy Rotor
l pm

h pm
Fig. 2 Schematic of Poles with an opposite arrangement (N-S w pmo
type) and the associated flux paths for Halbach Array in SSDR
AFPM. Fig. 4 Schematic geometry of magnet.
592 A. Habib et al.


2pro  nM ap 1 2sinhðBzÞ
wpmo ¼ ð7Þ Bx ðx; zÞ ¼ Bg cosðbxÞ ð12Þ
nM b expðbt=2Þ


2pri  nM ap 2coshðBzÞ
wpmi ¼ ð8Þ Bz ðx; zÞ ¼ Bg sinðbxÞ ð13Þ
nM expðbt=2Þ
where Bx is the tangential component (along the x-axis), Bz is
lpm ¼ wpmo  wpmi ð9Þ
the normal component (along z-axis). For the proposed design
where nM is the total number of the magnet. the significance of normal component is higher as the design is
dual disc rotor topology of axial flux machine. At a glance, the
3.2. Air gap magnetic flux value of Bz Bg is minimum in the middle of the two rotor discs
and closer to the magnets the value becomes maximum. For
The main difference between the Halbach and conventional conventional array the equation [32] is given by
Z
array is on the magnetic flux density and distribution, as 1
explained in Section 2. Geometry and other parameters can Bg ðcÞ ¼ rBx Bz dS ð14Þ
l0
be kept the same for better comparison except for the magnet s

size. For the different characteristics of the two topologies, the where l0 is the permeability of free space, r is the radius of the
size of the magnet cannot be the same for a fixed rotor diam- rotor, S is the integration surface of the mid plane of the air-
eter. The number of magnets used for Halbach configuration is gap. Bx and Bz depend on the optimized geometry of the
double that of the conventional array. In order to maintain the machine parameter, like ɑP.
same total magnet volume, the axial height of the magnets are
adjusted while the magnet length was kept constant for both. 3.3. Stator coil geometry
Though the wavelengths are same for the both conventional
and Halbach, they have different effect for magnet width to
For higher efficiency and lower cost, non-overlapping concen-
pole pitch ratio (ɑP). For conventional rotor, optimal ɑP is
trated windings have been used since it needs less volume of
required for less torque ripple with maximum average torque
copper which reduces the copper losses and increases generator
[31]. On the other hand, Halbach rotor needs to minimize
efficiency. In this design single layer trapezoidal coil shape is
the gap between magnets (ɑP = 1) to ensure effective can-
chosen for shortening the end winding length (le) as compared
celling of magnetic field on one side and the strengthening of
to active coil length (lact) of the coil that helps to maximize the
flux on the other side. Thus there are two different equations
coil flux linkage. A shorter le reduce the resistive losses in the
for calculating Bg. The equations of the air gap magnetic flux
inactive part of the coil. The calculation of le and wc are done
density of Halbach array and conventional array can be writ-
based on the design for non-overlapping windings [33]. Fig. 6
ten as (10)–(14).
shows the schematic of the coil geometry. On the other hand,

 sinðp=nMl Þ only the stator coil axial height (Lsc) and coil cross-section area
BgðhÞ ¼ Br 1  exp bhpm ð10Þ
p=nMl (Sw) is considered because the coil height has an impact on the
geometry of the axial height of the machine via the air-gap
where b is the 2p/la and la is the spatial period (wavelength) of
length from magnet associated rotor disk to another side of
the array, nMl is the number of magnet per wavelength. For la
the rotor disk. An optimized Lsc is required for higher output
the value differs for the two different topologies of Halbach
power. A large Lsc increases the total air-gap length as well as
array and conventional array. For Halbach array, four mag-
the active area length of the magnetic flux density and finally
nets create one full wavelength or one pole pair as there is
will decrease the magnetic flux density in the air-gap. On the
no gap between the magnets, refer to (11). In Fig. 3, the sche-
matic is shown for la [26].
la ¼ 4wpm ð11Þ
From Eq. (10) the peak value of magnetic flux density at the
active surface of Halbach array can be calculated. As seen in m
hp
Fig. 5, the tangential and normal component of the magnetic Qm
c
flux in the space between two discs are given by [26] Ls
ri
le
ravg
Wpmi
Rotor Qravg lact
lpm ro

z Wpm
N
t/2 Wc
C Wpmo B
x mid plane
t A

t/2
S

Rotor Fig. 6 Schematic of coreless non-overlapping winding and 3D

geometry of magnet (green color) and coil (red color) together
Fig. 5 Plane region for air-gap magnetic flux density. with their position in the machine.
Multi-Stator Multi-Rotor (MSMR) AFPM generator 593

Table 1 Difference of figures for two topology.

Conventional MSMR Proposed MSMR

Complete model for Axial length

Complete model for Axial length

Whole model for different parts

Whole model for different parts

Flux Path Flux Path

12
Magnet
Magnet width to 10
pole pitch ratio
(ɑp) 8
RC =41.10Ω
Current

6
Epoxy Rotor

2 RCI =38.5Ω

0
0 5 10 15 20 25 30 35 40 45 50
Internal Embedded Conventional Magnet with epoxy
Resistance
Rotor
Conventional MSMR Proposed MSMR Fixed Current line @5.23A-p
Magnet
Fig. 8 Resistive load of different MSMR for a fixed stator
Unity ɑp current.
2Iph Nc
Sw ¼ ð15Þ
Kf aw Ja

2Sw Q
Lsc ¼ ð16Þ
Rotor ks pDin
External Surface mounted Halbach magnet with
epoxy rotor where Kf is the fill factor, Ja is the current density, and kS ks is
the space utilization factor. For coil wounding, Kf is an impor-
Fig. 7 Schematic of Internal and external rotor with conven- tant factor as it signifies the cross-section of the coil Sw. As a
tional and Halbach magnet arrays respectively. rule of thumb, the value of Kf for manually constructed wind-
ing with circular cross section is around 0.55–0.78 [34] The fac-
other hand, very short axial length will require a higher Sw in a
tor ks is related to the mechanical strength of the stator
result large width of coil wc. For a fixed diameter the high wc
structure. The coreless structure should be mechanically strong
increases the difficulty to fit all the coils in the limited circular
enough to hold the coils and against the attraction of magnets.
space. The Lsc and Sw can be obtained as follows:
Winding factor (Kw) is another important parameter for the
594 A. Habib et al.

Table 2 Common input parameters for all four topology.

Parameter Name Unit Direct Cascade MSMR Proposed MSMR
General parameters Number of Pole (P) – 12
Number of rotors 3
Number of coreless stators 2
Number of Coils (C) – 18
Rotational Speed (N) rpm 500
No of Turns per Coil (NC) – 250
Specific design parameters Ratio of Inner & Outer diameter (k) – 0.577
Outer Diameter (Do) mm 180
Inner Diameter (Di) mm 104
Air-gap between rotor & stator (tg) mm 0.5
Density of magnet(NdFeb) g/cm3 7.5
Iron Density (steel M19G) g/cm3 7.4
Epoxy Density g/cm3 1.25
Copper Density g/cm3 8.96
Magnet Length (ML) mm 30
Total Magnet Volume (MV) mm3 353.3e3
Magnet Weight kg 2.6
Stator coil weight kg 3.158
Stator epoxy weight kg 2.344e4
Coil Axial thickness (tCh) mm 8.5
Coil Bandwidth (CBW) mm 17
Conventional rotor thickness (tCR) mm 8
Halbach rotor thickness (tHR) mm 10
Modified Rotor thickness (tMod) 10
Magnet Width pole Pitch ratio (ɑP) – 0.66 (C) 0.66(C) & 1(H)
Number of Magnet (Mn) – 12  4 (C) 12  1(C) & 24  2(H)
Magnet Upper Width (MWU) mm 19.35 (C) 19.35(C) & 14.51(H)
Magnet Lower Width (MWL) mm 29.72 (C) 29.72(C) & 22.63(H)
Magnet Axial Thickness (tMh) mm 10 (C) 10(C) & 9.98(H)
Rotor Weight kg 1.074 0.1135
Total weight of the Generator kg 6.83 5.87
Total Axial Length of Machine (TL) mm 2  (2tCR + 2tMh (2tHR + 2tMh (H) + tMod
(C). + 2tg + tCh) = 91 + 4tg + 2tCh) = 68.96
Load Resistance X 41.10 38.5
*(C): Internal rotor with conventional magnet array and (H): External rotor with Halbach magnet array

Rotor
For ironless rotor, this magnetic pull is due to the magnetic
Halbach Magnets
attractions between the magnets, while for the case of iron
z
rotor, additional attraction exists between the magnets and
t/2
N mid plane
x
t the iron rotors.
t/2
The attraction force due to two magnets placing in a dis-
Conventional Magnets
tance of g apart, with a magnet surface area of SPM, a magnet
z t/2
x thickness of hpm and an air-gap magnetic flux density Bg can be
S mid plane t
t/2 expressed as follows:
2  Bg 2 hpm 2 SPM
Rotor F¼ ð17Þ
l0 g2

Fig. 9 Plane region for proposed MSMR generator.

4. MSMR coreless AFPM generator

coil design as well as the total torque and the power, which is Multistage AFPM generator can be based on double stator
typically A value between 0.9  Kw  1 [35]. single rotor (DSSR) or single stator double rotor (SSDR) con-
figurations [5]. Here, the SSDR configuration is selected, where
3.4. Magnetic pull on rotor discs N-stage MSMR machine will have N stators and (N + 1)
rotors. For the purpose of discussion in this paper, a two-
In a SSDR AFPM generator there exist magnetic pull between stage (N = 2) MSMR machine is considered.
the two rotors which exerts mechanical stress to the structure.
Multi-Stator Multi-Rotor (MSMR) AFPM generator 595

4.1. Conventional direct cascade

The simplest form of MSMR machine is to directly cascade

two or more AFPM machine together [2]. For example, two
sets of SSDR machines can be simply combined to yield a
two-stage MSMR to increase the torque/power while main-
taining the machine’s diameter. As illustrated in the left col-
umn of Table 1, the direct cascade MSMR has 2 stators and
effectively 3 rotors (2 external rotors and 1 internal rotor).
Due to direct cascading, the internal rotor is essentially
back-to-back rotors of a SSDR machine. The stator coils are
single layer concentrated windings of trapezoidal geometry
which are embedded in epoxy and covered with composite
material hardener to ensure sufficient mechanical strength.

4.2. Proposed topology

Here, a fully coreless MSMR AFPM machine is proposed

where the iron rotors are replaced with epoxy rotors to
improve the torque and power density. There are two signifi-
cant modifications to the convention direct cascade MSMR
machine:

1. Halbach arrays are used on the external rotors to avoid

leakage flux toward the exterior of the machine.
2. The back-to-back internal rotors are combined into a single
internal epoxy rotor with conventional magnet array to
simplify design and save materials

By analyzing the flux path, it can be observed that the pro-

posed topology retains similar path with the conventional
direct cascade MSMR machine. However, the proposed topol- Fig. 11 The stress test results using ‘‘Von Mises” method in
ogy is more compact and has lower weight due to the elimina- AutoCAD Mechanical for the 4 mm C-I Rotor.
tion of the back iron on the rotor. The magnet volume is kept
constant as previous direct cascaded MSMR, by adjusting the
magnet thickness of each rotor disc. Hence the thickness of the
magnets of two external discs rotor and the thickness of the are kept at 0.5 mm on both sides for both topologies. For sta-
magnets of internal rotor are same whereas the total number tor part, the proposed topology maintains the same stator coil
of magnets are increased here from previous direct cascade numbers, windings layout and geometry used in the direct cas-
MSMR. Fig. 7 shows the difference between the external rotor cade MSMR.
and internal rotor of the proposed MSMR AFPM machine. In addition, the current loading for the stator coil is fixed
The rotor structure is formed by trapezoidal shape magnets, here, in order to get the performance of the modified rotor
rotor core, and shaft. The gap between the rotor and stator design. To make a fair comparison with the direct cascade
MSMR, a fixed stator current of 5.23 Apeak is considered here
based on the stator wire gauge, which is AWG16. Fig. 8 repre-
2 sents the different load resistance for both MSMR with a fixed
current loading. Since the back emf of the proposed MSMR
1.5 and conventional MSMR generators are slightly different,
proper load resistance need to be selected to obtain the same
Deflection

1 current loading. As seen in Fig. 8, the current drawn from

the generator reduces when load resistance increases. For a
0.5 fixed 5.23 Apeak stator current (black color line), the value of
the load resistance is 38.50 X for proposed MSMR and
0 41.10 X for conventional MSMR. As the two different magnet
0 5 10 15 20 25 30 arrangements (conventional and Halbach) are used in the
-0.5 design, the proposed machine is named as ‘‘fully coreless
Rotor Thickness
MSMR Halventional AFPM’’ generator. The generator is
C-I Rotor H-I Rotor C-E Rotor H-E Rotor designed with 12 poles, and 18 coils (both stators) for a speed
of 500 rpm at 50 Hz. The input parameters for the proposed
Fig. 10 Mechanical deflection (mm) of rotor vs. rotor thickness and direct cascade design are shown in Table 2. The weights
(mm) for four different topologies. of machines are obtained by multiplying the volume of the
596 A. Habib et al.

Upper (external rotor) disc

with Halbach magnets

Internal rotor disc

with Conventional
magnets

Lower (external rotor) disc

with Halbach magnets

Fig. 12 Air-gap magnetic flux density in the mid of air-gap between the internal and external rotors (Proposed Design).

Proposed MSMR

Fig. 13 Simulated flux path for the conventional (left) and

proposed (right) MSMR AFPM generator (hiding the external
rotor discs).

Fig. 15 Magnetic flux density of the stator coils (flux linkage) for
Conventional MSMR Proposed MSMR.

components (using dimensions obtained from FEA software)

with the densities of the respective materials.
However, the plane region of the whole model for proposed
design has been given in Fig. 9 for better understanding of the
design.

4.3. Analysis of mechanical deflection of rotor

For the construction of the MSMR AFPM machine, the two

opposite rotors attached with magnets are attracted to each
other. Although attraction force works for both ferromagnetic
and non-ferromagnetic rotor, for non-ferromagnetic rotor it is
slightly less. The strong attraction force between the magnets
Fig. 14 Magnetic flux density of the stator coils (flux linkage) for imposes an opposite pulling force on the rotor that tends to
conventional MSMR. bend the rotor structure leading to a collision between the
Multi-Stator Multi-Rotor (MSMR) AFPM generator 597

250 Phase B Phase C Phase A

200

150

100

50
Voltage

-50

-100

-150

-200

-250
0 5 10 15 20 25
Time(ms)
Fig. 19 FFT analysis of magnetic flux density in the air-gap for
Proposed_MSMR Conventional_MSMR
both side of the Internal rotor (Proposed MSMR).

Fig. 16 Voltage waveform for both the proposed and direct

cascade MSMR from the FEA analysis.

6 Phase B Phase C Phase A

2
Current

-2

-4 Fig. 20 FFT analysis of magnetic flux density in the air-gap for

-6
both side of the Internal rotor (Direct cascade MSMR).
0 5 10 15 20 25
Time(ms)

Proposed_MSMR Conventional _MSMR ferromagnetic and non-ferromagnetic material. In order to

do that four rotors have been chosen for bending test with dif-
Fig. 17 Current waveform for both proposed and direct cascade ferent combination of magnet arrangement and material selec-
MSMR from the FEA analysis. tion. The two different magnet arrangements (Conventional
and Halbach), are combined with two different materials (iron
and epoxy) of the rotor, to give the following four types of
-40
rotor:
-35

-30 (i) Conventional magnet array with iron rotor (C-I Rotor)
(ii) Conventional magnet array with epoxy rotor (C-E
Torque(N-m)

-25
TR= 13.2% TR= 18.6% Rotor)
-20
(iii) Halbach magnet array with iron rotor (H-I Rotor)
-15 (iv) Halbach magnet array with epoxy rotor (H-E Rotor)
-10

-5
The major difference among the four topologies lies in the
arrangement of the magnets and disparity of materials for the
0
5 10 15 20 25 rotor discs. Fig. 10 shows the deflection of the rotor from von
Time(ms)
Mises stress analysis for the four topologies, while Fig. 11
Proposed Conventional_MSMR
shows the particular von Mises test result from the Auto-
Fig. 18 Torque waveform for proposed MSMR. CAD Mechanical.
It is evident that deflection is higher in non-ferromagnetic
rotor compared with ferromagnetic material. Thus, to ensure
that deflection is within the acceptable range thickness of
rotor and stator or can reduce the air-gap distance between non-ferromagnetic rotor should be greater than that of the fer-
them; as a result, a non-uniform and unbalanced magnetic flux romagnetic one. According to Fig. 10, the thickness of 8 mm
density will be formed in the air-gap. Hence, mechanical and 4 mm is chosen for C-I and H-I Rotor respectively that
deflection is another important point of consideration to comes with negligible deflection values which is 10% of allow-
design optimum rotor thickness for the four topologies. Eq. able bending. On the other hand, with the maximum allowed
(17) is showing the force calculation for the magnetic pull on deflection of 0.2 mm, the thickness of the C-E and H-E Rotor
rotor discs. To find an optimum thickness of the rotor, the are selected to be 18 and 10 mm respectively which is safe
force is applied to the various thickness of the rotor for both bending for the epoxy rotors.
598 A. Habib et al.

Table 3 Performance comparison of two MSMR AFPM generator.

Machine Name Generator Power Power density Average Torque Density Torque per unit Torque
Weight (kg) (Watt) (W/kg) torque (savg) (Nm/kg) Volume (Nm/mm3) Ripple (TR)
Direct cascaded 6.84 1659 242 31.69 4.63 1.10  105 18.6%
MSMR
Proposed MSMR 5.87 1510 257 28.85 4.91 1.32  105 13.2%
(Halventional)

generator. Here it is important to mention that the current

250 (Fig. 17) is same for the both design as it was designed with
same stator current loading. In terms of torque ripple, the pro-
200 posed MSMR machine shows lower torque ripple of 13.2%
compared to 18.6% in the direct cascade MSMR machine
150 (Fig. 18). Figs. 19 and 20 show the FFT of the magnetic flux
density of the two machines. It can be concluded that the flux
100
density is the direct cascade
50
MSMR is higher which translate to higher voltage and tor-
que generated. However, there is slightly higher 3rd and 7th
0 harmonic current in the machine which can contribute to the
Direct cascaded MSMR Proposed MSMR higher torque ripple when compared to the proposed MSMR
(Halventional)
machine. Although the 5th harmonic in the proposed design
Power Density (Watt/Kg) Torque Ripple (%) is higher than the conventional design, it cannot affect the tor-
que ripple of the proposed design as the overall torque ripple
Fig. 21 Performance comparison between direct cascaded reduction happened due to the lower 3rd and 7th harmonics.
MSMR and Proposed fully coreless MSMR AFPM generator. Table 3 shows the performance comparisons between the pro-
posed MSMR with direct cascade MSMR which are further
5. Results and performance comparison illustrated in Fig. 21. The proposed fully coreless MSMR Hal-
ventional AFPM generator offers higher power density, lower
The results and performance of the proposed design are ana- torque ripple and lower axial length of the whole generator.
lyzed based on four important parameters, i.e. average torque, The higher power density comes basically from two important
power, power density, and torque ripple. For analyzing the modification of the generator. The first one is the reduction of
above parameters some fundamental parameters like magnetic the internal rotor as well as one side of magnets and the second
flux density, voltage, current, FFT of voltage and magnetic one is the combination of the Halbach and conventional mag-
flux are considered and present here. The simulated results nets. The lower torque ripple comes for lower amplitude of the
for both magneto-static and transient (ANSYS Maxwell) are 3rd and 7th harmonic, which has less interaction with the fun-
provided here for the better understanding and analyzing of damental harmonics. The reduced axial length of the machine
the proposed design. From the magneto-static simulation, is the cause of internal rotor modification as well as the one
the magnetic flux density of the machine shows underneath side magnet array elimination.
in various point of view. Figs. 12–15 is showing all the mag-
netic flux density analysis from simulation results. Fig. 12 6. Conclusion
shows mid-air gap flux density for both upper and lower side
of the Proposed design, Fig. 13 shows the flux density for This paper demonstrated the possibility to implement a fully
the whole machine for both conventional and proposed coreless MSMR AFPM by implementing a novel structure
MSMR, whereas Figs. 14 and 15 are showing the magnetic flux instead of mere cascading the AFPMs. The proposed design
density of the stator coils, where Fig. 14 is for the conventional combines the use of Halbach array and conventional magnet
(direct cascade) MSMR and Fig. 15 is for the proposed (Hal- arrays to achieve desirable characteristics: Halbach arrays
ventional) MSMR. Since the two machines utilizes the same are used for the external rotors to eliminate rotor back iron,
stator coils, both figures look similar apart from the separation while the internal rotors uses simple conventional magnet
between two sets of stator coils. It is worth highlighting that arrays for better flux density and lower complexity. Detailed
even though the overall volume of the proposed MSMR is sig- design considerations were discussed to obtain the desired elec-
nificantly smaller than the conventional MSMR generator, the tromagnetic performance of the proposed machine. In addi-
peak flux densities of the two machines are comparable. This is tion, von Mises stress analysis was conducted to determine
the reason why the proposed MSMR generator is able to the required rotor thickness in the proposed coreless structure
achieve better torque density than the conventional MSMR to compensate for the reduction in structural strength. Analy-
generator. Figs. 16–18 show the voltage, current, average tor- sis showed that even though 80% increase in rotor thickness is
que, and torque ripple (TR) respectively for the both proposed necessary there is still a net reduction in the overall machine’s
and conventional MSMR generator. The proposed generator weight due to the substitution of iron rotor with epoxy rotor.
is able to deliver sinusoidal voltage and current similar to The proposed structure was validated using ANSYS Max-
the conventional MSMR which is required for a three-phase wellÒ 3D FEA software and compared with direct cascaded
Multi-Stator Multi-Rotor (MSMR) AFPM generator 599

MSMR AFPM. Simulation results showed that even though [11] P. Virtic, P. Pisek, T. Marcic, M. Hadziselimovic, B.
eliminating rotor back iron reduces the air gap flux density Stumberger, Analytical analysis of magnetic field and back
slightly the proposed structure still provide numerous advan- electromotive force calculation of an axial-flux permanent
tages over direct cascaded MSMR AFPM like: magnet synchronous generator with coreless stator, IEEE
Trans. Magn. 44 (11) (2008) 4333–4336.
[12] S.M. Hosseini, M. Agha-Mirsalim, M. Mirzaei, Design,
1. increasing power density by 6% prototyping, and analysis of a low cost axial-flux coreless
2. reducing torque ripple by 5.4% permanent-magnet generator, IEEE Trans. Magn. 44 (1) (2008)
3. reducing axial length by 24% 75–80.
[13] W. Fei, P.C.-K. Luk, K. Jinupun, Design and analysis of high-
Furthermore, the fully coreless structure provides interest- speed coreless axial flux permanent magnet generator with
ing potential for cost saving, not only in terms of the material circular magnets and coils, IET Electr. Power Appl. 4 (9) (2010)
(substituting iron with plastic), but also manufacturing cost via 739–747.
the use of additive manufacturing (3D printing). [14] J.F. Eastham, F. Profumo, A. Tenconi, R. Hill-Cottingham, P.
Coles, G. Gianolio, Novel axial flux machine for aircraft drive:
Design and modeling, IEEE Trans. Magn. 38 (5) (2002) 3003–
Declaration of Competing Interest 3005.
[15] Z. Zhu, D. Howe, Halbach permanent magnet machines and
The authors declared that there is no conflict of interest. applications: a review, IEE Proc.-Electr. Power Appl. 148 (4)
(2001) 299–308.
Acknowledgment [16] L.X. Qi, X.Y. Wang, J.J. Du, Z. Fang, R.Y. Tang, Magnetic
field analysis of disk coreless permanent magnet synchronous
motor based on thickness variation Halbach array, J. Iron Steel
The authors thank the technical and financial assistance of Res. Int. 13 (2006) 439–443.
Ministry of Education Malaysia, University Malaya and UM [17] M.G. Lee, S.Q. Lee, D.-G. Gweon, Analysis of Halbach magnet
Power Energy Dedicated Advanced Centre (UMPEDAC). This array and its application to linear motor, Mechatronics 14 (1)
project is funded by the Higher Institution Centre of Excellence (2004) 115–128.
(HICoE) Program Research Grant, Fundamental Research [18] R. Praveen, M. Ravichandran, V.S. Achari, V.J. Raj, G.
Grant Scheme MO013-2016, UMPEDAC-2018 (MOHE Madhu, G. Bindu, A novel slotless Halbach-array permanent-
HICoE-UMPEDAC), RU007-2018 and RU012-2019. magnet brushless dc motor for spacecraft applications, IEEE
Trans. Ind. Electron. 59 (9) (2012) 3553–3560.
[19] D. Zhu, S. Beeby, J. Tudor, N. Harris, Increasing output power
References of electromagnetic vibration energy harvesters using improved
Halbach arrays, Sens. Actuators, A 203 (2013) 11–19.
[1] M. Aydin, M. Gulec, A new coreless axial flux interior [20] K. Sakai, Y. Tabuchi, T. Washizu, Structure and characteristics
permanent magnet synchronous motor with sinusoidal rotor of new high speed machines with two or three rotor discs, in:
segments, IEEE Trans. Magn. 52 (7) (2016) 1–4. Conference Record of the 1993 IEEE Industry Applications
[2] S. Kahourzade, A. Mahmoudi, H.W. Ping, M.N. Uddin, A Society Annual Meeting, 1993, IEEE, 1993, pp. 19–26.
comprehensive review of axial-flux permanent-magnet [21] R. Hill-Cottingham, P. Coles, J. Eastham, F. Profumo, A.
machines, Can. J. Electr. Comput. Eng. 37 (1) (2014) 19–33. Tenconi, G. Gianolio, Multi-disc axial flux stratospheric aircraft
[3] A. Mahmoudi, S. Kahourzade, N.A. Rahim, W.P. Hew, Design, propeller drive, Thirty-Sixth IAS Annual Meeting. Conference
analysis, and prototyping of an axial-flux permanent magnet Record of the 2001 IEEE Industry Applications Conference,
motor based on genetic algorithm and finite-element analysis, 2001, vol. 3, IEEE, 2001, pp. 1634–1639.
IEEE Trans. Magn. 49 (4) (2013) 1479–1492. [22] M. Aydin, S. Huang, T. Lipo, Axial flux permanent magnet disc
[4] F. Caricchi, F. Crescimbini, E. Santini, Basic principle and machines: a review, in: Conf. Record of SPEEDAM, 2004, pp.
design criteria of axial-flux PM machines having counter- 61–71.
rotating rotors, Conference Record of the 1994 IEEE Industry [23] A. Chen, R. Nilssen, A. Nysveen, Performance comparisons
Applications Society Annual Meeting, 1994, vol. 2, IEEE, 1994, among radial-flux, multistage axial-flux, and three-phase
pp. 247–253. transverse-flux PM machines for downhole applications, IEEE
[5] F. Caricchi, F. Crescimbini, F. Mezzetti, E. Santini, Multistage Trans. Ind. Appl. 46 (2) (2010) 779–789.
axial-flux PM machine for wheel direct drive, IEEE Trans. Ind. [24] A. Mahmoudi, N. Rahim, W. Hew, Axial-flux permanent-
Appl. 32 (4) (1996) 882–888. magnet machine modeling, design, simulation, and analysis, Sci.
[6] S. Javadi, M. Mirsalim, A coreless axial-flux permanent-magnet Res. Essays 6 (12) (2011) 2525–2549.
generator for automotive applications, IEEE Trans. Magn. 44 [25] J.H. Kim, B. Sarlioglu, Closed-form method for multi-stage
(12) (2008) 4591–4598. axial flux permanent magnet machine: design and analysis,
[7] S. Javadi, M. Mirsalim, Design and analysis of 42-V coreless Electr. Power Compon. Syst. 45 (7) (2017) 785–797.
axial-flux permanent-magnet generators for automotive [26] J.F. Gieras, R.-J. Wang, M.J. Kamper, Axial Flux Permanent
applications, IEEE Trans. Magn. 46 (4) (2010) 1015–1023. Magnet Brushless Machines, Springer Science & Business
[8] Y. Tan, J.-X. Shen, M.-J. Jin, H. Hao, Study of magnet fatigue Media, 2008.
in hybrid PM memory motors, in: 2015 International [27] S. Huang, J. Luo, F. Leonardi, T.A. Lipo, A comparison of power
Conference on Sustainable Mobility Applications, Renewables density for axial flux machines based on general purpose sizing
and Technology (SMART), IEEE, 2015, pp. 1–5. equations, IEEE Trans. Energy Convers. 14 (2) (1999) 185–192.
[9] M.R. Minaz, M. Çelebi, Design and analysis of a new axial flux [28] P.R. Upadhyay, K. Rajagopal, FE analysis and computer-aided
coreless PMSG with three rotors and double stators, Results design of a sandwiched axial-flux permanent magnet brushless
Phys. 7 (2017) 183–188. DC motor, IEEE Trans. Magn. 42 (10) (2006) 3401–3403.
[10] M. Sadeghierad, A. Darabi, H. Lesani, H. Monsef, Rotor yoke [29] T. Chan, L. Lai, An axial-flux permanent-magnet synchronous
thickness of coreless high-speed axial-flux permanent magnet generator for a direct-coupled wind-turbine system, IEEE
generator, IEEE Trans. Magn. 45 (4) (2009) 2032–2037. Trans. Energy Convers. 22 (1) (2007) 86–94.
600 A. Habib et al.

[30] A. Daghigh, H. Javadi, H. Torkaman, Optimal design of [33] F.G. Rossouw, Analysis and Design of Axial Flux Permanent
coreless axial flux permanent magnet synchronous generator Magnet Wind Generator System for Direct Battery Charging
with reduced cost considering improved PM leakage flux model, Applications, University of Stellenbosch, Stellenbosch, 2009.
Electr. Power Compon. Syst. 45 (3) (2017) 264–278. [34] K. Latoufis, G. Messinis, P. Kotsampopoulos, N.
[31] K.-C. Kim, D.-H. Koo, J.-P. Hong, J. Lee, A study on the Hatziargyriou, Axial flux permanent magnet generator design
characteristics due to pole-arc to pole-pitch ratio and saliency to for low cost manufacturing of small wind turbines, Wind Eng.
improve torque performance of IPMSM, IEEE Trans. Magn. 43 36 (4) (2012) 411–431.
(6) (2007) 2516–2518. [35] M.J. Kamper, R.-J. Wang, F.G. Rossouw, Analysis and
[32] P. Virtič, M. Vražić, G. Papa, Design of an axial flux permanent performance of axial flux permanent-magnet machine with air-
magnet synchronous machine using analytical method and cored nonoverlapping concentrated stator windings, IEEE
evolutionary optimization, IEEE Trans. Energy Convers. 31 Trans. Ind. Appl. 44 (5) (2008) 1495–1504.
(1) (2016) 150–158.