Design study of magnet shapes for axial Halbach

arrays using 3D Finite Element Analyses

Oliver Winter, Christian Kral, Erich Schmidt

Abstract—Magnet material prices has become an uncertain possible efficiency and minimum weight, the ironless axial-
factor for electric machine development. Most of all, the output of flux topology may be the best choice [6]. Today, the most
ironless axial flux motors equipped with Halbach magnet arrays widely-used commercially available axial-flux ironless motor
depend on the elaborated magnetic flux. Therefore, possibilities
to reduce the manufacturing cost without negatively affecting was presented by Lovatt [7] and is in use for solar race cars [8].
the performance are studied in this paper. Both magnetostatic Even though the recommendation given by [9] to use Halbach
and transient 3D finite element analyses are applied to compare magnet arrays was considered during the design phase, the
flux density distribution, elaborated output torque and induced available motor from CSIRO is equipped with standard iron-
back EMF. It is shown, that the proposed magnet shapes and backed magnet rings. A comprehensive overview on Halbach
magnetization pattern meet the requirements. Together with the
assembly and measurements of functional linear Halbach magnet permanent magnet applications is given by [10]. The combina-
arrays, the prerequisite for the manufacturing of axial magnet tion of latest light-weight manufacturing techniques, air cored
arrays for an ironless in-wheel hub motor are given. 1 winding arrangement and Halbach array magnet rings may
lead to a new benchmark when it comes to highest efficiency
N OMENCLATURE
and nominal torque to active weight ratio [11]. This paper
AFPM Axial flux permanent magnet machine is focused on the magnet shape selection and forces between
CSIRO Commonwealth Scientific and Industrial Re- magnets in Halbach arrangements for axial flux ironless in-
search Organisation wheel motors.
CFRP Carbon fibre reinforced plastic
EDM Electric discharge machining II. H ALBACH ARRAYS - H ISTORY AND ANALYTICAL
EM Electromagnetic calculation CALCULATION
FEA Finite element analysis
GRP Glass fibre reinforced plastic If a certain changing magnetization pattern is applied to
PM Permanent magnet a planar structure such as recording tapes, flux concentra-
tion occurs on one side of the specimen. First regarded as
I. I NTRODUCTION ”magnetic curiosity” [12] this effect was further elaborated
Axial flux in-wheel hub motors were often regarded as by K. Halbach, a physicist and eponym for the so called
promising solution for electric car propulsion [1]. However Halbach effect [13]. The first applications were multipole
direct-driven in-wheel vehicles are nowadays still mostly arrangements and undulators for synchrotron storage rings
restricted to scooters and motor show presentation concept [14]. Planar Halbach magnet arrays are today applied in planar
cars. Even though axial flux machines provide high torque to actuators for nanolithography used in semiconductor industry
volume ratio [2] the number of realized direct drive prototypes [15].
is marginal. The argument of reduced comfort or safety due The main advantages of the Halbach magnet arrangement
to added unsprung mass was also rebutted by [3] and [4]. are a
Mechanical requirements to limit the overall weight were • stronger fundamental field compared to conventional PM
addressed in [2] and the application of novel materials were arrays,
recommend. A water cooled two-stage 25 kW axial-flux motor • heavy-weight backing steel elements can be omitted and
was presented in 1996 [5]. However, considering both highest • therefore no iron losses occur,
1
• the magnetic flux density is more sinusoidal and,
The authors gratefully acknowledge the support of the Austrian Research
Promotion Agency (Oesterreichische Forschungsfoerderungsgesellschaft • there are very low back-side fields [16].
mbH, Klima- und Energiefonds, Neue Energien 2020) for the research The general description of the Halbach magnetization pat-
project 829727 HeAL - High efficient ironless drive for lightweight vehicles.
tern was given by [12], a simple superposition of two trigono-
Oliver Winter and Christian Kral are with the AIT Austrian Institute metric functions
of Technology GmbH, Mobility Department, Electric Drive Technologies,
Giefinggasse 2, 1210 Vienna, Austria. Telephone: +43(5)0550 6559; fax:  
+43(5)05506595; e-mail: oliver.winter@ait.ac.at; web: http://www.ait.ac.at 2πx
Erich Schmidt is with the Vienna University of Technology, Institute of Mx = M̂ sin (1a)
Energy Systems and Electrical Drives, Gusshausstrasse 25-29/370-2, 1040 λ
 
Vienna, Austria 2πx
My = M̂ cos (1b)
λ

978-1-4673-0142-8/12/$26.00 ©2012 IEEE 2660

 wm z
Halbach magnet array

hm
Airgap winding
x
z
Bz Halbach Magnet array

x (i)

Spatial period λ

Fig. 1. Halbach array with 4 single magnets per wavelength, nm = 4,

magnetization pattern and magnetic flux density characteristics

where λ denotes the wavelength and M̂ the magnetization

amplitude. However, this approach cannot be applied to rare-
earth anisotropic magnet material with their so called easy axis (ii)
characteristics. The predominated direction elaborated during
the manufacturing process requires a description on magnet Fig. 2. Double sided Halbach magnet array arrangement without back iron,
(i) magnetization pattern, (ii) resulting field pattern
segment basis, as depicted in Fig. 1. The magnetic flux density
peak value at the active surface is
−2πhm
 
 
π
B̂ = Br 1 − e λ  si . (2) (v)
nm

The example given in Fig. 1 is defined with the magnet

height hm and the number of magnets per wavelength nm = 4. (iv)

The rectangular shape yields from the wavelength λ = 8 · hm ,

Br represents the remanent magnetic flux density. For the (vi)
distance δ between two facing Halbach arrays, as depicted
in Fig. 2 (i), the flux density in normal direction is [16]:
 
2πz (iii) (iii)
  cosh
2πx λ
Bz (x, z) = B̂ sin   (3)
λ πδ
cosh
λ (ii) (ii)

The magnets with vertical magnetization are considered as

pole magnets and the horizontal magnets as gap magnets, (i)
respectively. For the equtations, the coordinate system is
placed in the center of the air-gap. Augmenting the number Fig. 3. Motor main parts cross section: (i) light metal hub, (ii) bearings,
of magnet segments from the minimum number nm = 4 to 6 (iii) two equal CFRP half rims, (iv) air cored winding with GRP connection
segments leads to a slight increase of the peak flux density by to the hub, (v) two Halbach magnet array rings, (vi) brake disk, [11]
6 percent. However, two practical problems arise in this case. TABLE I
First, the production of magnets with an easy axis of 60◦ M OTOR DESIGN SPECIFICATION
is very cost-intensive and second, the handling effort during
Description Symbol Nominal value
assembly is increased as well. Therefore four magnet pieces
per wavelength are considered for the proposed design. Continuous output power Pnom 3500 W
The effect of different magnet width to height ratios on the Peak power (10 s) Ppeak 5·Pnom
Speed at 80 km/h nn 660 rpm
peak value according equation (2) and correlation to the air-
Continuous torque Tnom 55 Nm
gap distance was studied both analytically and by 2D FEA.
1 Peak torque (10 s) Tpeak 5·Tnom
An example is given in Fig. 2 (ii).
Nominal torque/active weight 4.8 Nm/kg
III. H ALBACH ARRAY DESIGN FOR AN AFPM Peak torque/active weight 24 Nm/kg

The main electro mechanic parts within an ironless AFPM,

as shown in Fig. 3, are the air cored winding and two Halbach

2661
 TABLE II wtro

M AIN DIMENSIONS

Description Symbol Nominal value

Outer diameter Do 416 mm
Inner diameter Di 306 mm ltr
Mean diameter Dm 306 mm wm wm Do

Magnet length lm 55 mm
Magnet height hm 8.6 mm
Magnet width wm equ. (4a)-(4e)

wtto Dm

wtr Di
i
αp

ltt
wm wm Do Fig. 5. Magnet shape trapezoid-rectangular - TR

rectangular shaped magnet. This arrangement as depicted in

Fig. 5 facilitates two options to apply either the rectangular
Dm or the trapezoid magnet as pole. These three options, TT, TR
wtt Di Rpole (rectangular pole magnet) and TR Tpole (trapezoid pole
i
magnet) were evaluated.
αp
Based on the parameters,
α 
p
Fig. 4. Magnet shape trapezoid-trapezoid array - TT tan
wm = 2
 α  Dm (4a)
p
1 + cos
 2α 
magnet array rings. Considering the geometric limitations, wtti = tan
p
Di (4b)
the winding concept and the magnet array were analytically  α2 
p
designed to meet the motor specification given in Table I. The wtto = tan Do (4c)
winding axial width was set to 5 mm with 1.5 mm air-gap on α  2
p wm
both sides. For the given diameter and magnet height, an ideal wtri = tan Di − α  (4d)
2 cos
p
quadratic magnet would lead to around 65 Poles, therefore an 2
approximately 20 percent decrease on peak flux density was α  wm
p
wtro = tan Do − α  (4e)
accepted in order to reduce the number of poles. The magnet 2 cos
p
width to height ratio γm = wm /hm was set to 1.57. The 2
magnet ring dimensions are summarized in Table II. the model for each arrangement was designed and applied to
a 3D EM FEA software. Taking advantage of the periodicity,
IV. H ALBACH MAGNET ARRAY - S HAPE VARIATION the arrangement was reduced to one pole pair. The section
The manufacturing of classic ring segment magnets on model including the winding is shown in Fig. 6.
prototype level, would require EDM for an precise but quite
expensive result. Additionally for high pole machines, the dif- V. 3D E LECTROMAGNETIC F INITE E LEMENT A NALYSES
ference between narrow ring segment and trapezoid magnets A. Magnetostatic solver
is considered to be marginal. But even if the magnet array
In the magnetostatic domain, Maxwell’s equations
consists of trapezoid magnets, due to the small production
volume in the prototype phase, the manufacturing costs per
magnet are an important factor. To meet this challenge, three ∇×H = J (5)
different magnet shapes and arrangements, respectively, were ∇·B = 0 (6)
developed and evaluated. The already mentioned trapezoid
magnet ring, where both pole and gap magnet are of the are solved together with the constitutive equation
same dimension, as shown in Fig. 4, subsequently regarded B = µ · H = µ0 · µr · H. H is the magnetic field strength and
as TT (trapezoid - trapezoid). Actually, the trapezoid shape B the magnetic flux density. With the permeability of vacuum
has also to be manufactured based on the requirements from µ0 = 4 · π · 10−7 H/m, the relative permeability µr for each
mechanical and electromagnetic view, therefore the measure material, appropriate excitations and boundary conditions,
to reduce cost is to replace every other magnet by a standard the finite element analysis furnishes all field quantities as

2662
 Magnitude magnetic flux density Bz
0.9

0.8

0.7

0.6
TT shape, B¯z = 0.758 T
0.5 TR Rpole shape, B¯z = 0.758 T
TR Tpole shape, B¯z = 0.761 T
0.4

Di DT eval Do
Center pole radial distance

Fig. 7. No load FEA result, axial magnetic flux density distribution at the
Fig. 6. 3D motor section for electromagnetic FEA, magnet array and winding pole
encapsulation transparent

0.9 TT shape
functions of the Cartesian coordinates (x, y, z) throughout
TR Rpole shape
the specimen volume. The following investigations are made: 0.7 TR Tpole shape
Magnetic flux density Bz

1) No load: The first assessment was focused on the

0.5
magnetic flux distribution in the air-gap at no load condition.
The winding was masked out and multiple evaluation surfaces 0.25
and lines were established to compare the results. The model
consisted of approximately 160000 second order tetrahedral 0
elements. The most crucial factor for the elaborated output
−0.25
torque is the axial magnetic field in the pole area, therefore
the magnitudes of the magnetic flux density were calculated in −0.5
the middle of the air-gap at the pole center, as shown in Fig. 7.
−0.7
As indicated in the picture, the values at the mean diameter are
almost identical because the magnet width at this level is equal −0.9
for all arrangements. However, close to the inner and outer 0 αp /2 αp /2
diameter, the differences between gap and pole magnet width
Angluar distance
of both TB Tpole and TB Rpole are visible. To review the field
solution not only at one point along the circumference, circular Fig. 8. No load FEA result, tangential magnetic flux density distribution at
arcs were placed in the air-gap center. At the mean diameter, DT eval
there are no apparent differences, however Fig. 8 shows the
characteristics at the diameter DT eval = Di + (Do − Di ) /6
(1/6 of the magnet length starting at the inner diameter Di ).
Only minor deviations are visible. Therefore, from the field on single magnet basis showed an interesting detail. Obviously,
waveform point of view, there is no elaborated preference the pole magnets which share the same magnetization direction
among the considered magnet shapes. attract each other, the gap magnet on the other hand does not
2) No load - forces: The same no load field solution is used serve as simple flux bridge without reaction, but the specimen
to calculate the forces on the individual magnets by using is stressed with a force in the opposite direction towards the
the virtual work method. The surface tetrahedral elements outer side of the ring. Especially for the assembly of the
are virtually distorted and from the change in the virtual magnet ring itself, this detail is a valuable finding and is also
work and the small deviation, the forces are derived by the discussed in chapter VI regarding manufacturing of Halbach
software. The three values are in good correlation and no magnet arrays.
further comparison for the magnet arrangement quality can
be drawn from this result. The motor overall attracting force 3) Full Load: Due to almost the same field characteristics,
between the magnet rings were calculated to take measure for the results for various excitations up to nominal current show
the future assembly apparatus. Also this result coincides quite only negligible difference in the elaborate output torque among
well among the three different arrangements. The evaluation the three arangements for each excitation.

2663
 100
BackEMF Amplitude / V

80 TT shape
TR Rpole shape
TR Tpole shape
60

40
Fig. 10. Functional model, 3 piece natural balanced Halbach array, two pole
magnets and one gap magnet
20

0
1 2 3 4 5 6 7 8 9 10
BackEMF FFT / Order

Fig. 9. Transient FEA result, FFT analyses of the back EMF

(i) (ii)

B. Transient solver Fig. 11. Functional models, double sided N40 Halbach magnet arrays (2×9
pieces 10×10×45 mm each), (i) 5 mm air-gap, (ii) 10 mm air-gap
The induced voltage at nominal speed nn = 660 rpm was
calculated by the transient solver with a time step of 40 µs.
The FFT spectra of the three arrangements are compared in and tangential magnetized. The ironless in-wheel hub motor
Fig. 9. The TR Rpole shape results in a slightly reduced first prototype will be finished in August 2012.
and both higher third and fifth order harmonic components.
The single noteworthy deviation between TT and TR Tpole VII. C ONCLUSION
shape is given at the third order harmonic, therefore the TR Halbach magnet arrays are a beneficial choice to reduce both
Tpole shape is considered to be the better choice and will be weight and increase the magnetic flux density and therefore
applied to the prototype machine. the torque output of ironless axial flux motors. Especially
VI. R EALIZATION AND MEASUREMENTS on prototype level, the manufacturing of custom tailored
magnets is an important factor. The decrease the overall cost,
As mentioned in the previous chapter, during the assem- every second magnet is replaced by an rectangular block
bly and in the final configuration within the Halbach array, magnet. The impact is studied in comparison to a magnet ring
distracting magnet forces occur within the specimen. Without consisting of trapezoid magnets. The comparison of magnetic
further measures, the magnets form the staircase arrangement
shown in Fig. 10. The shear stress analysis yields to the
conclusion that without the presence of a backing material, 1.2 single N35
Magnitude magnetic flux density Bz

the shear stress maximum reaches up to 6 MPA, which is 1 single N35 FEA
quite challenging for some adhesives especially at elevated 0.8 delta5 N40
temperatures. According to this result, a adequate adhesive was 0.6 delta5 N40 FEA
chosen to manufacture small scale prototypes to validate the 0.4 delta10 N40
FEA results. During measurements under various temperature delta10 N40 FEA
0.2
conditions, it was observed and later confirmed by 2D and
0
3D FEA calculations, that rare-earth magnets exhibit an aug-
mented temperature degrading effect compared to standard ap- −0.2
plications [17]. Samples from different magnet materials were −0.4
used to manufacture double-sided linear Halbach arrays, as −0.6
depicted in Fig. 11. The magnetic flux density characteristics −0.8
were measured along the air-gap distance, an example at 20 −1
◦
C for three different arrangements, single sided, double sided −1.2
with an air-gap of 5 mm and 10 mm, is shown in Fig. 12 and 0 λ/4 λ/2 3λ/4 λ
compared to 2D FEA. Spatial period λ
According to the results from the previous section, the
TR Tpole magnet ring configuration was manufactured Fig. 12. Measurement and FEA results, magnetic flux density distribution
(cf. Fig. 12) by using trapezoid axial magnetized pole magnets at 2 mm distance from the surface, Temperature 20 ◦ C

2664
 R EFERENCES
[1] Z. Zhang, F. Profumo, and A. Tonconi, “Axial flux interior PM syn-
chronous motors for electric vehicle drives,” Symposium on Power
Electronics Electrical Drives Automation and Advanced Electric Motors,
(SPEEDAM’94), pp. 323–328, Jun. 1994.
[2] F. Profumo, F. Eastham, A. Tenconi, and G. Gianolio, “”Plastic” electric
motors: a viable solution for axial flux machines,” in Proceedings
of the 2002 IEEE International Symposium on Industrial Electronics,
(ISIE’02), vol. 1, 2002, pp. 1–10.
[3] M. Anderson and D. Harty, “Unsprung mass with in-wheel motors-myths
and realities,” in 10th International Symposium on Advanced Vehicle
Control, (AVEC’10), Aug. 2010, pp. 261–266.
[4] D. van Schalkwyk and M. Kamper, “Effect of hub motor mass on
stability and comfort of electric vehicles,” in IEEE Vehicle Power and
Propulsion Conference, (VPPC’06), Sept. 2006, pp. 1–6.
[5] F. Caricchi, F. Crescimbini, and A. Di Napoli, “Prototype of innovative
wheel direct drive with water-cooled axial-flux pm motor for electric
Fig. 13. Manufactured Halbach array with trapezoid poles and rectangular vehicle applications,” in Eleventh Annual Applied Power Electronics
gap magnets Conference and Exposition, (APEC’96), vol. 2, Mar 1996, pp. 764–770.
[6] J. Gieras and I. Gieras, “Performance analysis of a coreless permanent
magnet brushless motor,” in Industry Applications Conference, 2002.
37th IAS Annual Meeting. Conference Record of the, vol. 4, oct. 2002,
flux density and elaborated torque for a given motor design pp. 2477 –2482 vol.4.
showed only minor differences. Based on FFT analyses of the [7] H. Lovatt, V. Ramsden, and B. Mecrow, “Design of an in-wheel motor
for a solar-powered electric vehicle,” IEE Proceedings Electric Power
induced voltage at nominal speed, the most suitable choice Applications, vol. 145, no. 5, pp. 402–408, Sept. 1998.
is the arrangement with trapezoid magnets serving as pole [8] R. Al Zaher, S. de Groot, H. Polinder, and P. Wieringa, “Comparison
magnet and rectangular magnets in between. Finally, results of an axial flux and a radial flux permanent magnet motor for solar
race cars,” in XIX International Conference on Electrical Machines,
from both functional linear and ring Halbach magnet arrays (ICEM’2010), Sept. 2010, pp. 1–6.
are presented, whereby forces and the assembly itself were [9] J. Ofori-Tenkorrang and J. Lang, “A comparative analysis of torque
studied to be able to manufacture the prototype of an ironless production in Halbach and conventional surface-mounted permanent-
magnet synchronous motors,” in Conference Record of the 1995 IEEE
in-wheel hub motor. 30th IAS Annual Meeting Industry Applications Conference, (IAS’95),
vol. 1, Oct. 1995, pp. 657–663.
B IOGRAPHIES [10] Z. Zhu, “Recent development of halbach permanent magnet machines
and applications,” in Power Conversion Conference, (PCC’07), Apr.
2007, pp. K–9–K–16.
[11] O. Winter, S. Ucsnik, M. Rudolph, C. Kral, and E. Schmidt, “Ironless in-
wheel hub motor design by using multi-domain finite element analyses,”
Oliver Winter (S’11) received his Dipl.-Ing. degree in Mechatronics from the in International Symposium on Power Electronics, Electrical Drives,
University of Linz, Austria, in 2006. After two years in industry he joined Automation and Motion, (SPEEDAM’12), Jun. 2012.
the AIT Austrian Institute of Technology in 2009. As Junior Scientist within [12] J. Mallinson, “One-sided fluxes – A magnetic curiosity?” IEEE Trans-
the business unit Electric Drive Technologies, he is working towards his PhD actions on Magnetics, vol. 9, no. 4, pp. 678–682, Dec. 1973.
in electrical engineering. His research interests include modeling of dynamic [13] K. Halbach, “Design of permanent multipole magnets with oriented rare
electromechanic and thermal behavior of mechatronic systems by means of earth cobalt material,” Nuclear Instruments and Methods, vol. 169, no. 1,
coupled FEA. pp. 1–10, 1980.
[14] Klaus and Halbach, “Application of permanent magnets in accelerators
and electron storage rings (invited),” Journal of Applied Physics, vol. 57,
no. 8, pp. 3605–3608, Apr. 1985.
[15] J. Rovers, J. Jansen, J. Compter, and E. Lomonova, “Analysis method
Christian Kral (M’00, SM’05) received the diploma and doctoral degrees of the dynamic force and torque distribution in the magnet array of a
from the Vienna University of Technology, Vienna, Austria, in 1997 and 1999, commutated magnetically levitated planar actuator,” IEEE Transactions
respectively. From 1997 to 2000, he was a Scientific Assistant in the Institute on Industrial Electronics, vol. PP, no. 99, p. 1, 2011.
of Electrical Drives and Machines, Vienna University of Technology. Since [16] J. F. Gieras, R.-J. Wang, and M. J. Kamper, Axial Flux Permanent
2001, he has been with the AIT Austrian Institute of Technology GmbH (the Magnet Brushless Machines, 1st ed. Dordrecht, The Neatherlands:
former Arsenal Research) in Vienna. From January 2002 until April 2003, he Kluwer Academic Publishers, 2004.
was a Visiting Professor at the Georgia Institute of Technology, Atlanta. His [17] O. Winter, C. Kral, and E. Schmidt, “Augmented temperature degrading
current research interests include diagnostics and monitoring techniques and effect of rare earth magnets arranged in segmented Halbach arrays,” in
the modeling and simulation of electric machines and drives with a particular IEEE International Magnetics Conference, (INTERMAG’12), no. 4, May
focus on nonlinear effects, thermal behavior and faulty machine conditions. 2012.

Erich Schmidt (M’98) was born in Vienna, Austria, in 1959. He received his
MSc and PhD degrees in Electrical Engineering from the Vienna University
of Technology, Austria, in 1985 and 1993, respectively. Currently, he is an
Associate Professor of Electrical Machines at the Institute of Energy Systems
and Electric Drives of the Vienna University of Technology. His research and
teaching activities are on numerical field computation techniques and design
optimization of electrical machines and transformers. He has authored more
than 100 technical publications mainly in the fields of electrical machines and
numerical field calculation.

2665

Powered by TCPDF (www.tcpdf.org)