Permanent magnet fixation concepts for

linear generator
Oskar Danielsson, Karin Thorburn, Mikael Eriksson and Mats Leijon
Division for Electricity and Lightning Research
Department of Engineering Sciences
Uppsala University, Box 534, S-751 21 UPPSALA

Abstract - Lately, the field of linear generators

for wave energy conversion has attracted large
attention, [1], [2]. One encouraging way to
convert wave energy into electrical through
direct drive is by using a linear, synchronous,
longitudinal flux permanent magnet (PM)
generator where the rotor piston is driven by,
for example, a point absorber. As prices
decrease, it has only recently become possible
to use high remanence NdFeB magnets in
technical applications. This opens up for new
possibilities for linear generators, as high
magnetic excitation can be achieved with
smaller magnets.

converters (WEC), [1][2]. With a linear

generator it is possible to couple the motion of
a point absorber directly to the generator. Such
configuration eliminates the need of complex
power take-off schemes. This, in turn, reduces
the complexity of the plant and the need for
maintenance.
Direct driven linear generators for WEC have
been investigated by M.A. Muller in a number
of articles [3,4]. It is mainly the transverse flux
permanent magnet machine (TFM) that has
been investigated. The studies have shown that
the TFM exhibits very good power to mass
ratios but also that it suffers from high inherent
synchronous reactance. Another disadvantage
of the TFM is the structure of the stator, which
demands significant supporting structures to
keep the cores in place.

In the present work two concepts of magnet

fixation, surface mounted magnets and
magnets buried between pole shoes have been
studied with NdFeB and ferrite magnets. While
the former alternative demands stronger
magnets, which are more difficult to handle,
the latter yields a somewhat heavier rotor
piston. Pole shoes provides better possibilities
to shape the magnetic flux curve in the
periphery of the air gap, and the magnets are
more protected from transients. The surface
mounted concept allows for a more efficient
use of the magnets, as the leakage is lower than
when the pole shoe approach is used. This
means that it is possible to reduce the length of
the rotor with surface mounted magnets.

In this paper the electromagnetic properties of

a longitudinal flux PM machine (LFM) are
investigated. The LFM has a simple and robust
stator construction with inherently low
synchronous reactance. The stator of a LFM
also provides a large option of winding
configurations: single phase, multiple phase,
different slots per pole and phase ratios, etc.
Although the properties of LFM have been
well investigated and a number of solutions
has been proposed for conventional, highspeed rotational generators, the relatively low
speeds and the reciprocal motion of a direct
coupled linear generator changes the
prerequisites
fundamentally.
Thorough
analyses of the magnetic circuit are needed to
optimize the use of expensive permanent
magnet and other materials, increase the
efficiency to a maximum, reduce the ripple in
the electromagnetic forces, etc. This paper is
based on full physics simulations revealing the
time dependent behaviour of the machine. Four
different rotor concepts are investigated and
the electromagnetic performances of the
machines are compared.

The different configurations are compared with

respect to total magnet mass, efficiency and
load angle. Thorough analysis of the
generators has been carried out using full
physics simulations using a finite element
method (FEM) of the time dependent magnetic
field. Magnetic intensity distributions, time
dependent currents and voltages are also
among the parameters that have been studied.
1

INTRODUCTION

Linear generators have lately been suggested

as suitable energy converters in wave energy

1.2 Stator

MACHINE CONFIGURATION
1.1 General description

The stator is made of laminated electrical steel,

piled into one solid unit, se Figure (2). The
conductors are power cables with a circular
cross-section and a conducting area of 16 mm2,
insulated with a 1.1 mm PVC-layer, which
adds up to an outer diameter of 7.2 mm. The
coil winding is a three-phase winding with a
slot per pole and phase ratio of 5/4. This
winding configuration aims at minimizing the
fluctuation in the output power caused by
cogging. A three-phase LFM with a slot per
pole and phase ratio equal to one is proposed
as generator in the Archimedes Wave Swing
[1].

A possible WEC concept with a linear

generator as power take-off is shown in Figure
(1). The WEC consists of a buoy coupled
directly to the rotor of a linear generator by a
rope. The tension of the rope is maintained
with a spring pulling the rotor downwards. The
rotor will move up and down at approximately
the same speed as the waves and the maximum
speed will be in the order of 1 m/s. The
relatively low speed implies that the reaction
force developed between the rotor and stator to
be very high. For example, a 10 kW generator
needs a reaction force in the order of 10 kN
with a rotor speed of 1 m/s. This implies that a
directly driven generator must be larger than a
conventional high-speed generator.
(a)

Buoy

Rope

Linear
generator

Figure (2): Tilted side view of a section of the

stator of LFM linear generator.
1.3 Rotor
Two types of magnet fixation methods, surface
mounting and burying magnets between pole
shoes, are tested with two different types of
permanent magnets. The two fixation methods
are illustrated in Figure (3). In both
configurations adjacent magnets have opposite
polarity and a movement of the rotor creates an
altering magnetic field in the stator coils.

Stators
Rotor

Spring
b
)
Figure (1): (on top) The principles of a wave
energy plant with a linear generator as power
take-off. (bottom) Cross-section of the linear
generator.

Figure (4): Magnetic circuit for a rotor with

buried magnets. The magnetic flux is marked
by closed curves and arrows indicate the
direction of the flux. Leakage flux through the
aluminium plate is marked with a dashed line.

Figure (3): Tilted side view of rotor for a LFM

linear generator. Buried magnets to the left
and surface mounted magnets to the right. The
direction of the magnetic flux is indicated with
arrows.
Figure (4) shows the magnetic circuit of a rotor
with buried magnets. The flux is led from the
magnets through bars of magnetic steel, called
pole shoes. The pole shoe enables control of
the magnetic flux distributions in the periphery
of the air gap and it also protects the magnets
from transient magnetic fields generated by
short circuit in the outer circuit. The
aluminium plate on the backside of the rotor
serves as a barrier for the magnetic flux to pass
through the backside of the rotor. A portion of
the magnetic flux will unavoidably pass
through the back. That flux will not contribute
to the magnetic coupling and can be
considered as lost.

Figure (5): Magnetic circuit for a rotor with

surface mounted magnets. The magnetic flux is
marked by closed curves and arrows indicate
the direction of the flux.

The magnetic circuit of a generator with

surface mounted magnets are illustrated in
Figure (5). Surface mounted magnets are more
exposed to transients and face a larger risk of
demagnetization. On the other hand the
magnetic circuit has no obvious shortcuts, as is
the case for the buried magnets.

The major difference between the two concepts

is that the width of surface mounted magnets,
wmag, is limited of by the pole width, wpole.
Buried magnets have no theoretical limit of the
width of the magnet, instead it is the height of
the magnet hmag that must be smaller than the
pole width. Normally the magnet height should
be considerably smaller than the magnet width
and a limitation of the height is a minor
problem.
Two permanent magnets have been examined:
ordinary ferrite (Fe) magnets and high-energy
Neo-dymium-Iron-Boron (NdFeB) magnets.
The basic properties of the magnets are
presented in Table (1). NeFeB magnets are
considerably more expensive than Fe magnets
and the relation of the kilo price is assumed to
be 10:1.

Table (1): Properties of the magnets

Remanence (T)

1,32

0,3663

volume and cable length is calculated from the

geometry and vertical length of the rotor and
stator. A fix stator design is used in all
simulations to minimize the number of
variables affecting the results.

Relative
permeability

1,06

1,06

Table (2): Fixed parameters for case linear

generators

Density (kg/m3)

7700

4700

Material

Neodymium-Iron-Boron
Ferrite Magnet
(Vacodym 633 PT)
(Oe MagnetY30 BH)

METHOD
2.1 Simulations

This work is based on full-physics simulations

with a starting point in the electromagnetic
field equations. The simulations are based on a
2-dimensional model of a segment of the
generator. Design symmetry is used to
minimize the size of the simulated segment. In
this case, with a slot per pole and phase ratio of
5/6, six poles need to be simulated to achieve
symmetry. Maxwells equations are solved in
two dimensions with nodal finite element
method. The grid is automatically generated
and the number of nodes depends of the size
and geometry of the problem. Material
properties such as resistivity, permeability,
coercivity, sheet thickness etc. are taken into
account in the model. Three-dimensional
effects, such as coil end reactance are
approximated by empirical factors. Hysteresis
losses and eddy current losses are calculated
from the maximum magnetic field intensity
distribution.

Parameter

Value

Voltage

200 V

Power

10 kW

Number of stator sides

Number of cables in each slot

Cable cross-section

16 mm2

Slots per pole, phase

5/4

Pole width

55 mm

Airgap width

3 mm

Numerical results along with magnetic

intensity distribution plots have been used
throughout the work to analyze the results. The
maximum magnetic intensity in the stator is
kept below 1.8 T in order to avoid saturation
and excessive iron losses.
3

RESULTS

To more easily compare the magnet mass the

Fe magnet mass is divided by a factor of ten in
the graphs. This gives approximately the same
values and relates the magnet mass to the
relative cost of 10:1 between NdFeB and Fe
magnets.

2.2 Evaluation criteria

The main factors of interest in this work are
the apparent power, the electromagnetic
efficiency, the load angle and the amount of
material needed. A high electromagnetic
efficiency and low material usage is desired.
The electromagnetic efficiency includes
hysteresis losses, eddy current losses and
resistive losses. Furthermore, a low load angle
is desired. A machine with inherently low load
angle has better performance at both normal
and transient conditions and is less affected by
changing loads and varying frequencies.

3.1 Magnet dimensions

Different heights to width relations of the
magnets have been investigated in order to see
if there is an optimum. In Figure (6) the
magnet volume of single magnet is kept
constant and the electromagnetic efficiency
and total magnet weight is plotted for different
height to width relations of the magnets. As
can be seen the electromagnetic efficiency is
steadily increasing and the total magnet weight
is decreasing with increasing magnet width for
the surface mounted magnets. The pole width
limits the magnet width and no optimum is
reach for the surface mounted magnets. The
width of the buried magnets can be chosen

The four rotor concepts are simulated for

different magnet dimensions and various pole
widths. Output power, voltage and stator width
are held constant in the simulation and the
vertical length of the rotor is iterated to fulfil
these conditions, see Table (2). The material
usage such as total magnet volume, stator steel

freely and an optimal width to height ratio is

somewhere around 5 for both Fe and NdFeB
magnets.

electromagnetic efficiency for different total

magnet mass. This is illustrated in Figure (7).
It should be noted again that the mass of the Fe
magnets are divided by a factor of ten. The
buried NdFeB magnet rotor reaches about the
same electromagnetic efficiencies but demands
approximately the double total magnetic mass
compared with the surface mounted NdFeB
magnet rotor. The surface mounted Fe magnet
rotor has a slowly increasing electromagnetic
efficiency that never exceeds 70 %, even for
relatively large magnets.
In Figure (8) the load angle of the different
machines is plotted for different total magnet
weight. The surface mounted NdFeB magnet
rotor has lower load angle than the buried Fe
magnet rotor irrespective of the total magnet
weight. Again the surface mounted Fe magnet
rotor and the buried NdFeB magnet rotor
shows to be poorer than the other two options.

Figure (7): Electromagnetic efficiency for the

different rotor concepts plotted against the
total magnet weight. The total mass of the Fe
magnets is divided by a factor of ten.
Figure (6): Total magnet weight and
electromagnetic efficiency for different magnet
height to with relations. The total mass of the
Fe magnets is divided by a factor of ten.
3.2 Comparison of the concepts
Two of the alternatives, surface mounted
NdFeB magnet rotor and the buried Fe magnet
rotor, show
stunning
similarities
in

relatively good electromagnetic properties. It

should, however, be mentioned that the ferrite
magnets needed are about ten times heavier
than the NdFeB magnets, which will result in a
much heavier rotor. A large and heavy
machine will be more expensive than a
smaller. Therefore it is obvious that the overall
most advantageous is the machine with surface
mounted NdFeB magnets, when the magnet
cost ratio of 1:10 is used. If the cost of ferrites
decrease substantially, the pole shoe mounted
ferrite option will be more economical.
The surface mounted Fe magnet rotor shows
very poor electromagnetic properties. It is clear
that Fe magnets are not suitable for surface
mounting in machines with small pole widths,
which demands high magnetic excitation of the
stator. The buried NdFeB magnet rotor shows
similar characteristics as the surface mounted
NdFeB magnet but demands the double
amount of magnet. On explanation can be that
a part of the magnetic flux is passing through
the back of the rotor when the magnets are
buried between pole shoes, se Figure (9).

Figure (8): Load angle for the different rotor

concepts plotted against the total magnet
weight. The total mass of the Fe magnets is
divided by a factor of ten.
3.3 Material usage
The material usage is estimated for one
machine of each of the four generator types,
and is presented in Table (3) with efficiency
and load angle. For each type the machine with
the highest efficiency in combination with
lowest material consumption is chosen.
Table (3): Material utilization, efficiency and
load angle for simulated generators
Surface
mounted
ferrite

Surface
mounted
NdFeB

Pole shoe
ferrite

Pole shoe
NdFeB

2275 kg

N/A

1240 kg

N/A

N/A

120 kg

N/A

242

Core material

3852 kg

893 kg

1020 kg

1334 kg

Cable

3779 m

1272 m

1234 m

1299 m

Efficiency

68.5

85.2

85.1

83.5

Load angle

6.2

4.5

4.8

5.2

Fe magnet
NdFeB magnet

Figure (9): Magnetic flux lines in a cross

section of a linear generator with NdFeB
magnet rotor revealing the leakage flux in
thick lines.
Another circumstance worth mentioning is that
the basic properties, i.e. voltage, output power,
stator width etc., of the generators have been
constant for all simulations. It is possible that
an option other than surface mounted NdFeB
magnets is most advantageous for another
machine configuration. This study, however,
lies outside the scope of this paper.

DISCUSSION

The results, presented in the previous section,

reveal two concepts, surface mounted NdFeB
and buried ferrites, which show similar and

CONCLUSION

Two simulated generators exhibit similar high

efficiencies and low load angles, and are
regarded as good options. These two have
rotors with surface mounted NdFeB magnets
and ferrite magnets buried between pole shoes,
respectively. It can be argued, that when the
whole generator is examined, the surface
mounted NdFeB may be a better option, as it
renders a less heavy rotor machine. The cost
relation between ferrite and NdFeB magnets is
an important evaluation parameter.
6

ACKNOWLEDMENTS

The work in this paper is supported by the

Swedish Energy Agency, the ngpannefreningen research foundation, the J Gust
Richert foundation, Vargn Alloys AB,
Gteborg Energi and Draka Kabel.
7

REFERENCES

[1]

H. Polinder, F. Gardner and B.

Vriesema, Linear PM generator for
wave energy conversion in the AWS,
Proceeding of ICEM 2000, pages
309-313, August 2000, Espoo,
Finland

[2]

M.A. Mueller, Electrical generators

for direct drive wave energy, IEE
Proceedings, Gener. Transmission
and Distribution, Vol. 149 Issue
4:446-456, July 2002

[3]

M.A. Mueller, N.J. Baker and E.

Spooner, Electrical Aspects of
Direct Drive Energy Converters,
Proceedings of the 4th European Wave
Energy Conference, 2000, Aalborg,
Denmark

[4]

M.A. Mueller and N.J. Baker, A low

speed reciprocating permanent
magnet generator for direct drive
energy converters, Conference
publication, Power Electronics,
Machines and Drives, 16-18 April