Computer Aided Design of a Permanent Magnet Synchronous

Motor
Muhtar Hanif Alhassan
Nigerian Turkish Nile University, Abuja

1 Introduction
This paper contains the description of a design method, based largely on the finite-element magnetic
field analysis, used to design a permanent magnet synchronous motor PMSM for variable-speed drive
applications.[1] Since the stator of the PMSM is identical to that of any three phase AC machine, it was
decided to use the stator frame of an induction machine for the design. Thus the stator dimensions and
current rating are already given reducing the design to the optimisation of the magnetic loading of the
motor. In particular it is desirable to obtain a machine capable of producing sinusoidal back emf and
thus suitable for smooth torque generation when supplied with sinusoidal currents. As the motor was
to be fed from a variable frequency variable amplitude current-regulated inverter there was no need to
facilitate direct-on-line start capability in the form of cage bars, as was the practice with earlier designs
of permanent-magnet synchronous motors. [3]

2 Design Considerations
The design of a permanent-magnet synchronous motor PMSM is best carried out with the help of
magnetic field analysis as suggested by a number of authors: [8, 9, 10, 11]. The traditional method of
design which relies purely on empirical formulae and certain rules of thumb, is unable to cope with the
need for accurate computation of field values in a machine with complex cross-sectional area and having
heterogeneous nature. Furthermore the presence of materials with complex geometries and nonlinear
properties means that the simple linearised models, that mostly neglect saturation, employed in the
traditional method are not acceptable in this case. Thus in order to evaluate accurately the e↵ects of
design changes aimed at achieving desirable performance-specifications, it is necessary to perform detailed
magnetic analysis. As a result of such analysis, performance characteristics, winding inductance, flux
plots, localised flux densities etc., are easily obtained. Another reason for the shift to design by analysis
is the need to optimise the overall drive rather than the electric motor alone, especially in cases where
operation is in variable-speed mode that usually involves costly power conditioning.

2.1 Internal-Magnet Rotors

The degree of complexity in the design and construction of the permanent-magnet synchronous motor
varies with the shape, size and location of the permanent magnets. Designs di↵er by the location of the
magnets, internal, surface or inset, and by the orientation of the magnet axes; radial or circumferential.
Figure 1 shows various types PM rotor configurations and how the permanent magnets are placed with
respect to the rotor. The choice of rotor topology is very important as it directly a↵ects such factors as
magnetic flux distribution in the machine, the homogeneity of the rotor, the windage loss, etc. From the
construction point of view, the most complicated case is when the magnets are embedded in the rotor
to produce as narrow airgap as possible and thus ensure flux weakening capability. Another reason for
burying the magnets in the rotor is that it enhances robustness, and o↵ers protection for the magnets
from exposure to the airgap. This makes the internal-magnet construction (figure 1a,b), suitable for
operation at high speeds. Furthermore, the circumferentially-oriented magnet-axes design of Figure 1b
is used for flux concentration especially in motors that utilise low-energy magnets. A draw-back for this
type of construction however is that the shaft has to be made from nonmagnetic materials so as to avoid
magnetic short-circuiting of the magnets.

1
 The internal-magnet machines possess magnetic saliency and, due to their relatively short airgap
lengths in the q-axes, display noticeable armature reaction e↵ects, which have to be taken into account
when choosing the optimum operating point. [7, 8, 2] The situation is almost identical in motors that
have inset magnets with narrow gaps as they too are characterised by magnetic saliency and appreciable
armature reaction.
The application of the direct-design approach in the design of a polyphase PM excited motor with
inset magnets has been carried out using a computer program -DESIGN [7]. A general magnetic model
of the machine was established, and it was shown that the electric and magnetic loadings of the machine
depend on the angle between the armature and magnet excitations, and that there exists an optimal
value of that yields maximum torque for maximum allowable flux density in the stator teeth and
another one that yields the minimum magnet volume. Both of these values have been shown to be in
the same neighborhood of | | ⇠ = 70 electrical degrees. [7, 8]

1
2

3
4

a) b)
1 - Nonmagnetic filler

2 - Rotor iron

3 - Shaft

4 - Magnet

1
2
3

c) d)

Figure 1: Typical PM-Rotor Designs: a)-Internal-magnet type with radial flux; b)-Internal-magnet with
circumferential flux; c)-Surface-magnet type; d)-Inset-magnet type

2.2 Surface-Magnet Rotor

Motors with surface-mounted radially-oriented permanent magnets have only become of practical im-
portance with the developments of newer PM materials having high residual-flux densities. The surface-
magnet rotor has a better utilisation of magnetic flux than is achievable with the flux-concentration
designs using materials of lower residual-flux densities. The surface-magnet motor can be designed to
be magnetically nonsalient by making sure that the inter-magnet gaps, if filled, are iron free. Thus the
low reluctance in the d-axis due to the presence of the magnets is equally present in the q-axis so that
magnetic saliency is virtually absent. The e↵ects of armature reaction are weakened by the large e↵ective
airgap in both the d- and q- axes. The design, analysis and control of such a machine become relatively
easier and the optimum value of the angle is simply zero, when the maximum torque is developed. On
the other hand, operation beyond base speed is not possible since flux-weakening cannot be implemented.
Furthermore the absence of reluctance torque means that these machines have less torque capabilities

2
than the those with internalmagnet construction. Another source of worry for this type of construction
is how to retain the magnets and prevent them from moving relative to the rotor in any direction. Since
these problems are directly a↵ected by rotor speed, the surface magnet motor is mainly a low speed
machine, except in cases when the magnets are retained by special cylinders.

2.3 Inset-Magnet Rotor

When the rotor magnets are placed in slots on the rotor surface, the inset-magnet machine results.
Depending on the depth of the slot into which the magnets are placed, the machine can be either
magnetically salient or display very little saliency. Machines that have shallow slots e↵ectively perform
as surface-magnet types. On the other hand if the slots are deep, and the magnets are almost entirely
inset the result is a magnetically salient motor, with the characteristics similar to those of the buried- or
internal-magnet machine.

2.4 Choice of PM materials

The choice of magnet materials for motor applications is based not only on their magnetic characteristics
but also on their mechanical and thermal properties, coupled with the cost e↵ectiveness.
The demagnetization curve of some permanent-magnet materials are shown in Fig.2. Alnico has
the a very high residual flux density but its relatively low coercivity means it is unsuitable for motor
application as it can easily be demagnetized by the stator mmf. Ferrite materials are, on the other
hand have high values of coercive force, but relatively low values of residual flux density. Their high
resistance to demagnetization coupled with the low cost has made them relatively attractive for motor
applications. To enhance torque production, however, special flux-focussing arrangements leading to
complicated rotor topologies are usually needed. Samarium Cobalt materials have higher residual flux
density and coercivity than ferrites but are considerably more expensive.
A relatively new addition to the list of magnet materials for motor applications is Neodymium-
Boron-Iron, NdBFe, which, at room temperature, has the highest energy product of all commercially
available magnets [10, 4]. However both Ferrite and Neodymium magnets are sensitive to temperature
rise and care must be taken where working temperatures exceed 100 C. Thus for very high temperature
applications, Alnico or Rare-earth Cobalt magnets are more suitable.
Most magnet materials are brittle especially the ferrite types, and care must be taken when handling
them.

1
1.2

1 - ALNICO

2 - FERRITE

3 - SAMARIUM COBALT
0.8

B M -[T]

4 - NEODYMIUM

4
3
0.4

2
0.00

1.00 0.75 0.50 0.25 0.00

HM -[MA/m]
Figure 2: Demagnetization Curves for some PM materials.

It is very important that the permanent magnets be secured to the rotor in such a way as to ensure
they do not vibrate excessively. For high-speed applications the magnets should be bonded to the rotor
by special stainless steel retaining shells or using a Kevlar banding. [10]

3
2.5 Levi’s Direct-Design Approach
[7, 8] This is an approach that aims at incorporating systems aspect into the design of the AC motor.
Specifically it aims to satisfy the following requirements:

1. be applicable to complex topologies such as those found in the PMSM.

2. take account of proximity, skin and saturation e↵ects.
3. be capable of accurately predicting performance.

4. enable the computation of lumped parameters for both systems and dynamic analysis.
5. provide a ”feel” for the significance of various e↵ects and changes in dimensions.
6. be simple but capable of optimising the entire drive system.
7. provide sufficient insight to inspire further innovation.

The direct-design approach proceeds along two converging paths, one beginning from the performance
specifications, and the other from basic physical principles. The performance specifications are translated
into terminal characteristics and circuit parameters, while physical laws are used to determine limits for
the electric, magnetic, thermal and mechanical stresses based on material properties. The two paths
are then joined as a second step by replacing the terminal variables with various loadings from which
the main dimensions of the machine are obtained. These dimensions are then expressed as simplified
functions of critical stresses so as to expose any weak links, thus guiding the designer toward innovation
in shape, size and topology. The initial part of the design is completed by calculating the the losses,
weight and cost of the motor, its power supply and mechanical transmission. The final part of the design
involves the use of flux mapping techniques in assessing proximity, skin, and iron saturation e↵ects.

3 Design Specifications
The starting point for an electric machine design is to obtain and analyse the performance specifications
based upon which the main machine dimensions and parameters are computed. In the particular case
where the motor is to be operated at constant frequency and voltage the performance specifications
amount to the definition of a single-rated operating point. For such a case the design input data would
include the power level and speed as required by the load on the one hand, and the voltage, frequency
and number of phases as dictated by the available power source on the other. Corresponding to these
specifications a rated torque is defined and the design would aim at producing a machine capable of
developing the necessary starting, synchronous pull-in and pull-out torques and maintaining the rated
torque without over heating. In the case, however, where the motor is to be designed for variable-
speed application, there is no single point of operation that may serve to define a nominal power and
speed. Thus the behaviour of the motor and its load must be analysed over a duty-cycle. Variable-speed
operation of the permanent-magnet synchronous motor is achieved using the closed-loop control of the
inverter firing-angle.
The angle between the magnet flux density vector B¯M and the stator surface current density
vector K̄s plays an important role in the performance of the PMSM. If the motor is to be a radial-flux
surface-magnet type, as is the case here, magnetic saliency can be neglected. Furthermore the e↵ects
of armature reaction are minimal due to the large e↵ective airgap. For such a machine an optimum
operating condition will be when the angle is set to zero corresponding to maximum developed torque.
In other cases when the PMSM has pronounced magnetic saliency and narrow airgap the optimum value
of for maximum torque and minimal machine dimensions has to be determined analytically. Levi has
shown that the optimum value of for minimal machine dimensions is in the same neighborhood as
that for minimum volume of magnets and depends very much on the e↵ective reactance-coefficient in the
d-axis.[7, 8]

3.1 Stator Dimensions

The usual input to a motor-design programme include specified values of such operating-point parameters
as the rated power, rated speed, efficiency, power-factor and so on. These values would normally be used
to determine the main dimensions of the motor such as the stator bore-diameter, axial length airgap

4
 lp

BM

Bs
Ks

K s
Bs

BM

Figure 3: Relation between Magnet and Stator fields

length and so on which are linked to the operating-point parameters by empirical formulae based largely
on thermal considerations. In [7] and [8] Levi has used the idea of an ideal power (defined as the rated
power divided by the product of efficiency and power factor) to determine the bore diameter and axial
length of the machine given the number of poles and the operating frequency.
The input data for the current design included the dimensions and winding parameters of the three-
phase stator of an induction motor shown in table 1 which was to be fitted with the permanent-magnet
rotor. Since the stator dimensions were already given the normal approach of determining them from
performance specifications was not required. Instead, the stator dimensions were used as a guide in
setting the performance specifications in form of the maximum ideal power obtainable without any
electrical, magnetic or mechanical over stressing of the machine.

Quantity Symbol Value Unit

Rated Power PR 370 Watts
Number of phases m 3 –
Phase Voltage Vph 240 Volts
Rated Speed n 700 rpm
Frequency f 50 Hz
Full load current IR 1.2 A
Rated torque TR 5.85 Nm
Efficiency at full load ⌘ 0.91 –
Power factor cos ' 0.8 –
Stator bore diameter D 0.083 m
Active stack length L 0.083 m
Stacking factor sf 0.96 –
Number of stator slots Q 36 –
Number of conductors per slot u 120 –

Table 1: Induction Motor Data

3.2 Ideal Power and Torque

The stator bore diameter D is related to the ideal power Pi approximately by the expression [7]

D = 0.097⇥w 0.385 Pi0.29 n0.55

pp f
0.23
(1)
where

• ⇥w is the winding temperature rise above 40 C;

• Pi is the ideal power of the motor;

5
 • npp is the number of pole pairs;
• f is the frequency of stator currents.
Taking ⇥w as 70 C so as to limit the maximum working temperature of the magnets to about 110 C
due to the sensitivity of permanent-magnet materials to high temperature rise [7], equation 1 becomes

D = 0.0189Pi0.29 n0.55
pp f
0.23

which, after rearranging to make Pi the subject becomes

Pi ⇠
= 9 ⇥ 105 D3.45 npp1.9 f 0.8 (2)
Thus equation 2 gives an approximate estimate of the ideal power available based on thermal
considerations from a machine having stator bore diameter D, number of pole pairs npp and operating
with stator-current frequency f . The maximum rated operating frequency of the motor was set at 600
rpm so that for a six-pole machine the maximum operating frequency becomes 30Hz. Substituting these
values in equation 2 and the value of D = 0.083m, an ideal power of about 320 Watts was found.
Corresponding to the power an ideal torque Ti was computed as

Ti = 60Pi /2⇡n (3)

Substituting the value of Pi above into equation 3 yielded an ideal-torque value of 5Nm.

3.3 Rated Current

The stator frame used had 36 slots as shown in the induction machine data of table 1. It was fitted
with an 8-pole concentric winding which was replaced by a 6-pole distributed fractional-pitch type so
as to obtain the required speed and at the same time improve the mmf wave shape of the machine by
distributing the winding. In rewinding the stator the number of conductors per slot and the conductor
size were left unchanged, so that the space factor of the slots and the winding resistance remained
approximately unchanged. Furthermore the insulation class (F) was maintained since the temperature
rise would not exceed the limit mentioned above, determined by the presence of the permanent-magnet
material. Thus it was reasonable to assume that the upper limits of the electric loading, and hence the
current ratings, of both machines are practically identical. Thus since the stator dimensions are fixed
and the three-phase stator winding limits the choice of peak-current loading, the design is reduced to
the iterative choice of the magnetic loading for the PMSM so as to satisfy the ideal-torque requirement
established above. Based on the above approach a set of design input data was specified as shown in
table 2.

Quantity Symbol Value Unit

Rated Power PR 320 Watts
Number of phases m 3 –
Phase Voltage Vph 220 Volts
Rated Speed n 600 rpm
Frequency f 30 Hz
Full load current IR 1.2 A
Rated torque TR 5.0 Nm
Power factor cos ' 0.9 –
Efficiency ⌘ 0.9 –

Table 2: PMSM Specifications

4 Design Approach
The approach used involved two main stages. First empirical formulae based largely on thermal consider-
ations were used to set initial values to some of the machine parameters. The input data for this stage of
the design are the design specifications which are first transformed into terminal characteristics and then

6
into the respective forms of loading (i.e. electrical, magnetic and mechanical). In this particular case the
main motor dimensions and the peak value of the electric loading are already known, hence the motor’s
parameters are iteratively chosen so as to obtain optimum values for the magnetic and mechanical load-
ings. The second stage in the design involved the use of a finite-element based computer-aided-design
method to refine the results obtained in the first stage of the design. A key element in the design of
permanent-magnet machines is the choice of magnet size and shape and location within the rotor. This
aspect of the design is very much simplified by the use of finite-element magnetic field analysis by which
various options can be modelled and simulated so as to arrive at an optimum variant.

Parameter Symbol Value Unit

Operating frequency f 30 Hz
Number of phases m 3 –
Pole pairs npp 3 –
Pole arc – 40.2 mech. degrees
Pole arc -to- pole pitch ratio ↵p 0.75 –
Stator bore diameter D 0.083 m
Stator Stack length L 0.083 m
Stator Slot pitch ⌧s 7.24 mm
Stator Slot opening wo 2 mm
Stator slot height hs 17.1 mm
Physical airgap length lg 0.6 mm
Carter coefficient kc 1.12 –
E↵ective airgap g 0.67 mm
Number of turns per phase Nph 720 –
Fundamental winding factor kw 0.933 –
Operating flux density in magnet and gap BM 0.93 T
Remanent flux density Br 1.12 T
Recoil Permeability µr 1.05 –
Magnet arc length wM 29 mm
Magnet length lM 3.5 mm

Table 3: PMSM Parameters

5 Preliminary Design
As mentioned earlier this section of the design aims at the approximate estimation of the PM motor
parameters from the performance specifications and the known dimensions of the stator. The electric and
magnetic loadings of the motor corresponding to the ideal power Pi are determined iteratively making
sure that their respective limits are not exceeded. The electric loading is limited by factors like insulation
class, copper fill-factor, slot depth and the required current density. The magnetic loading on the other
hand is limited by iron saturation especially noticeable in the stator tooth area where the peak value of
the magnetic flux density should not exceed 2.1 T. In most designs the flux density is normally limited
to a peak of about 0.9 T.

5.1 Electric Loading

The electric loading is defined in this work as the rms value of the surface current density distribution
and is given by the equation
K = 2mNph kw Is /⇡D (4)
where m is the number of phases, Nph is the number of series turns per phase, kw is the fundamental
winding factor and Is is the rms value of the stator current. The choice of electric loading K for
a machine is based mainly on thermal considerations as it represents the copper loss and hence the
winding temperature-rise in the machine. In a situation where the winding parameters m, Nph and
kw together with the stator-bore diameter D are already fixed, K is uniquely determined by the stator

7
current. The value of K corresponding to the rated stator current is given by

KR = 2mNph kw IR /⇡D

Thus for a rated current IR of 1.2A, a winding factor kw = 0.933 and a number of turns per phase
Nph = 720, the surface current density is about 18.55kA/m.
The sensitivity of permanent-magnet materials to rise in temperature, means that the current loading
estimated above might have to be lowered. The cross-section of the winding conductor was estimated
at 2.463 ⇥ 10 7 m2 . For a stator rms current of 1.2A, this corresponds to a current density J of about
4.872 ⇥ 106 A/m2 . The equivalent current-sheet thickness is given by
K
dcu = (5)
kw J
For the above current rating the equivalent thickness of the stator current-sheet has a value of about
4mm. The heat transfer coefficient, h, is given by
KJ
h= (6)
kw ⇥ w
Here is the electrical conductivity of the conductor material, and ⇥w is the maximum allowable
temperature rise. For copper, has a value of about 4.24 ⇥ 107 S/m. ⇥w was taken as 70 C according
to considerations given in section 3.2. The result is a heat transfer coefficient of about 32.7WK 1 m 2 ,
which is rather high for totally-enclosed-fan cooling TEFC design. [7]
In order to reduce h to an acceptable value, a lower current rating of 1A was chosen, which results
in the following values for the above parameters

K = 15.46 ⇥ 103 A/m

J = 4.06 ⇥ 106 A/m2
dcu = 4mm
1 2
h = 22.7W K m

The power dissipated as heat per unit surface of the conductor is given by

Pdis = h⇥w (7)

Thus copper loss in the machine is given by

lco
Pcu = Pdis · ⇡DLf e (8)
Lf e

Here Lf e is the stack length of the stator and Llco

fe
is the ratio of the conductor length to the stack length,
which is taken to be approximately 2.
After substituting the relevant values into equation 8, the copper loss is found to be about 68.8 W.

5.1.1 Magnetic Loading

The magnetic loading should be high enough to satisfy the torque requirement as represented by the
ideal torque Ti without exceeding the saturation level. The idea of torque per unit-rotor-volume
(TRV) was used to link the motor torque with the magnetic and electric loadings. By definition TRV is
given by
T RV = T /⇡r2 L (9)
where T is the developed torque, r is the rotor radius and L is the rotor length. Neglecting the
physical airgap length of the machine, T RV can be approximated to

T RV ⇠
= 4T /⇡D2 L (10)

Thus given the dimensions of the stator D = L = 83 mm, and an ideal torque value of 5 Nm, a TRV of
about 11.13kN m/m3 is obtained.
The electric and magnetic loadings of the machine are linked to the TRV by the equation

T RV = K̂ B̂ cos (11)

8
where K̂ and B̂ are the peak values of the electric and magnetic loadings respectively, and is the
angle between the axes of K and B in space. If the rotor flux is assumed to be orthogonal to the stator
ampere-conductor distribution, which is the condition for maximum motor torque for a motor with no
saliency, then angle vanishes and after rearranging we have an expression for the peak value of the
magnetic flux density in the airgap as
p
Bˆg = T RV / 2K (12)
The average value of the flux density in the airgap is then given by

Bave = 2Bˆg /⇡ (13)

Thus after substituting the values of K and T RV into equation 12 a peak airgap flux density of
.51T was obtained. This corresponds to an average gap flux density Bave of about 0.324 T. For a magnet
material like Neodymium, this flux value is rather too low, since it would require a very small magnet
length. Thus the choice of a final value for the peak airgap flux density B̂g is put forward to the next
section.

5.2 Choice of Air gap and Magnet Lengths

It is generally accepted that the physical gap between the rotor surface and the stator teeth should be as
small as mechanically possible for good magnet utilisation [7, ?, 3]. The gap should however be adequate
to provide enough clearance for easy rotor assembly and subsequent rotation. Based on such mechanical
considerations Levi, [7, 8], suggests a gap-to-bore-diameter ratio of

g/D ⇠
= 10 2
Pi0.01 p 0.5
(14)

The application of equation 14 in our case yields an e↵ective physical airgap length g of about
0.5mm. The final decision about the choice of airgap length is however based also on the required
magnet flux density in the gap, the type of magnet material and the magnet length. [?]
For high-energy magnet materials like Neodymium, the demagnetization characteristics are linear
over the quadrant, and can be represented by the equation

B M = B r + µ r µ 0 HM (15)

where
• BM = flux density in magnet

• Br = remanent flux density of magnet

• µr = relative permeability of magnet
• µ0 = permeability of free space.

Furthermore for such materials the relative permeability, µr , is very nearly unity. If the rotor and
stator iron cores are assumed to be infinitely permeable, the application of Ampere’s Law yields the
following relation for the magnet length and the airgap.

HM lM + Hg g = 0 (16)

Here HM and Hg are the magnetic field intensities in the magnet and in the airgap respectively, lM is
the magnet radial length, and g is the e↵ective length of the airgap. Finally using the principle of flux
continuity, and assuming the ratio of the magnet pole area to the area of the airgap, AM /Ag , to be unity,
we have
Bg = BM = µ0 HM lM /g (17)
If equations 15 and 17 are combined and HM is eliminated, the operating flux density in the
magnet, and hence in the air gap, is found as
Br
BM = (18)
1 + µr g/lM

9
 Finally an expression for the magnet length is found by rearranging equation 18 to obtain
µr gBM
lM = (19)
Br BM
The flux density in the magnet is related to the peak value of the airgap flux density by the equation
⇡ ˆ ↵p ⇡
BM = Bg / sin (20)
4 2
It should be noted that the magnet flux density BM in equations 15, 17, 18 19 and 20 is a level
value which is identical to Bg and has Bˆg as its fundamental component as shown in figure 18. The
value of BM corresponding to the peak value of the airgap flux density Bˆg of 0.51T found earlier, is about
0.44T for a pole arc -to- pole pitch ratio (↵p ) of 0.75. If this value of BM is substituted in equation 19
and g is taken as 0.5mm as suggested by equation 14, the magnet length lM for Neodymium with a
remanent flux density Br of 1.12T and relative permeability µr of 1.05, is found to be about 1mm which
is too small for all practical purposes.
Sebastian et al have shown in reference [?], that the ratio of the magnet length to the e↵ective airgap
length lies in the range 5-10 for Neodymium. The value of the magnet flux density BM was raised to
0.8T. This meant that the flux density peak B̂g rose to a value of 0.94T which is still within acceptable
bounds.
The physical airgap length was raised to 0.6mm and its e↵ective value determined as

g = k c lg (21)

where lg is the physical length of the airgap and kc is Carter’s coefficient given by [7]
 1
wo
kc = 1 (22)
⌧s (5lg + wo )

Here wo is the stator slot opening, ⌧s is the slot pitch and lg is the physical gap length as all in meters.
The parameters ⌧s and wo are given in table 3. Carter’s coefficient for the chosen physical gap length
of 0.6mm was found to be about 1.12 giving an e↵ective airgap g of 0.67mm. The new values of BM and
g were substituted in equation 19 and the Neodymium magnet length required was found to be about
2.63mm. A final magnet length of 3.5mm was chosen to ensure mechanical robustness. The magnet flux
density BM correspondingly rose to about 0.93T, with the peak of the fundamental reaching 1.1T. With
this choice of magnet material and its radial length together with the airgap length, the preliminary
stage of the design was completed. Its results are shown in table 3.

6 Finite-Element Analysis
6.1 FEM in Machine Design
The finite-element method (FEM) is increasingly becoming a very powerful tool in the computation and
analysis of electro magnetic fields. The method has found application in the design, and performance
analysis of electrical machines due largely to its reliability and relative simplicity which have both been
enhanced by the availability of powerful computing facilities in the modern design environment. Thus
FEM has opened up a whole new approach to the design of electric machinery by making it possible
to model, simulate and analyse various designs without building costly prototypes. Thus necessary
improvements to a design are carried out thoroughly before a prototype is built, which saves time on
repetitive building and testing of physical models. Furthermore computer simulation often provides
information which would have been impossible to obtain by conventional testing methods.
It must be pointed out here however,that if strictly accurate results are to be sought, full three-
dimensional finite -element analysis would be required. Hysteresis and material anisotropy must also be
taken into account for such a standard of accuracy to be attained. Such an approach would be rather
difficult and is as yet not cost e↵ective. Thus the usual approach is to employ two-dimensional analysis
and reduce the complexity of the analysis either by successive problem refinement, or by using
subproblem analysis. [9]
The method of successive refinement involves the initial creation of several simple models and
progressively improving the accuracy by reducing their number and increasing the model complexity

10
until finally the most accurate model is arrived at. In subproblem analysis detailed portions of the
problem are replaced by simpler ones which may be quite wrong locally, but yield the right external
(net) result.
The region under consideration is partitioned into discrete elements and the field problem is solved
by a variational approach. The problem is formulated in terms of an expression, the energy functional,
which in engineering applications is identified with the stored energy in the system. For electrical machine
analysis, under the assumption of infinite axial length, a two-dimensional formulation using first-order
triangular elements is often applied. First-order triangular elements are characterised by geometric
flexibility which permits easy modelling of complicated electric machine cross-sections. Furthermore
their relative simplicity in computation when compared with higher order elements adds considerably to
their popularity in machine analysis.
In this work finite-element analysis was used to carry out the following:
1. Simulation and analysis of various initial models of the PMSM all having the same main dimensions
but di↵ering in rotor configuration, magnet size and/or shape and physical length of the airgap.
2. Simulation and thorough analysis of a final model that was being considered for construction, so
as to predict and optimise its performance.
3. Computing the parameters of the built motor for use in its dynamic and steady-state modelling
and for comparison with measured and theoretically determined values.

6.2 Mesh Generation

The analysis was carried out using MagNet2D, a two-dimensional, finite-element, magnetic-field-analysis
software, which makes intensive use of graphics. Models were created by graphical preprocessing using a
sub-program, Mesh2D, which facilitates the generation of meshes interactively. For reasons of symmetry
only one pole pitch was modelled in each case. The rotor and stator models were built separately and
joined together to produce the motor model so that di↵erent rotor positions could be examined without
rebuilding each model. The airgap region was finely meshed so as to improve the local accuracy there
since this is where the torque computations are carried out. Appropriate parts of the machine were
modelled as individual regions to be assigned material names and current source values in a variety of
ways so that di↵erent problems could be created without much remeshing.

MAGNET

COPPER

MILD STEEL

AIR

ALUMINIUM

Figure 4: A single-pole representation of a PMSM showing material boundaries

Models were created having all the four main rotor configurations shown in 1, and each housed in
a 36 slot 6-pole stator. For reasons of symmetry only one pole-pitch was modelled in each case, and,
at the preliminary stage, the number of meshes was limited to about 2000. Figure 4 shows the single-
pole model of the surface-magnet PMSM with the di↵erent material regions shaded accordingly. The
stator slots bearing the armature-winding conductors were each discretised separate from the airgap and
surrounding iron, so that appropriate current-density values could be assigned to each slot according to
the current-density distribution along the airgap periphery.
In those cases where the slot contains conductors from coil sides belonging to two di↵erent phases,
the average current-density value was assigned. The interpolar gaps were modelled as separate regions
too, so that various materials could be assigned to them for comparisons.
The magnet and airgap regions were modelled in layers, so both airgap and magnet sizes could be
varied by simply reassigning material properties to the defined layers, without having to create other
models. In the airgap region odd numbers of layers were used so that the centre of the airgap did not

11
 Figure 5: Finite element model outline for the prototype PMSM

Figure 6: Finite element model of the prototype PMSM showing the triangular meshes

constitute a boundary. This was necessary in order to allow the definition of a contour at the gap centre
later in the post processing, when the values of forces and inductances are to be calculated.

6.3 Problem Definition and Editing

The magnetic field in an electric machine can be described in terms of the magnetic vector potential Ā,
which satisfies the equation
CurlĀ = B̄ (23)
where B̄ is the flux density vector. In the two-dimensional finite-element analysis of the field, the
magnetic vector potential is computed over a pole pitch by solving the equation
1 @ 2 Az @ 2 Az
( + ) = Js (24)
µ @x2 @y 2
which is the two-dimensional quasi-Poissonian equation in terms of the normal component Az of the
magnetic vector potential Ā. The current density Js and the vector potential are thus both assumed to
be directed along the z-axis only. Homogeneous Dirichlet boundary conditions, given by
Az = 0 (25)

12
 Figure 7: Meshes forming half a tooth.

were imposed at the external surface of the stator and at the centre of the shaft. At the interpolar
boundaries, binary conditions for anti-symmetry given by
⇡
Az (⇥) = Az (⇥ + ) (26)
2
were imposed.
In problem definition, the various regions created at the mesh-creating stage, are assigned material
properties and current-density (source ) values. All nonmagnetic materials are treated as air. These
include the stator-winding material which was copper, nonmagnetic wedges, and in some cases the
material of the permanent magnet itself.

6.4 Modelling the Magnets

The permanent magnets were modelled by shifting the manufacturer-supplied demagnetization curve of
the magnet material so that it passes through the origin, and recording the amount of shift in terms of
either the coercive force Hc or the remanent flux Br . The permanent-magnet material was thus treated
as a soft material encompassed by fictitious current sheets. The shift of the B-H curve is equivalent to
dividing the the flux density in the magnet material into two parts: a passive part which has a zero
value at H = 0, and a magnetomotive part given by the coercive force Hc . Figure10 shows the result of
shifting the B-H curve.
The total magnetic flux density B̄ in the permanent-magnet material can be represented as

B̄ = µo (µr H̄ + M̄o ) (27)

where M̄o is the remanent intrinsic magnetization of the material, H̄ is the magnetic field intensity, µo is
the permeability of free space and µr is the relative permeability of the magnet material. If the vectors
B̄, H̄, and M̄ are resolved along two mutually orthogonal axes p and n, such that p is the direction of
intrinsic magnetization, equation 27 becomes

Bp = µo (µrp Hp + Mo ) (28)
Bn = µo µrn Hn (29)

By shifting the B-H curve of the magnet as in figure 10, the above equations are transformed into

Bp = µo µ̂rp (Hp Hc ) (30)

Bn = µo µ̂rn Hn (31)

13
 Figure 8: Mesh formation in half of a slot

where µo µ̂rp ia associated with the shifted curve in the figure, and the magnetomotive term Hc , is
independent of the field. In two-dimensional analysis this is equivalent to a non-hysteretic magnet with
the permeability of µo µ̂rp wrapped in a current sheet of strength equal to Hc amperes per unit length
of the magnet. If equations 30 and 31 are rearranged and the curl of each side of the resulting equation
are taken the following equation in terms of the vector potential Ā is obtained.

@ 1 @ Ā @ 1 @ Ā
+ = (J¯z + J¯m ) (32)
@n µp @n @p µn @p

where µp = µo µ̂rp , µn = µo µ̂rn ,

J¯m = curlHc 1̄p (33)
and J¯z is the actual current density in the system. 1̄p is the unit vector in the direction of intrinsic
magnetization. Finally Stoke’s theorem is applied to equation 33 to obtain an expression for the
equivalent current sheet as follows
Z Z I
¯ =
J¯m .ds ¯ =
curlHc 1̄p .ds Hc 1̄p .dl = IM (34)
R C

Here IM is the equivalent current sheet, R is the open-surface region containing the permanent magnet
and C is its periphery. For a uniformly-magnetized magnet the current sheet exists only at the boundary
between the magnet and the surrounding materials.
In a two-dimensional approach, first-order, two-dimensional elements are used for simplicity. The
magnet equivalent current density J¯M then contributes to the energy functional in the form
Z
l
ai dl = (35)
2

where the a’s are approximating polynomials [11].

The equivalent current sheet for two such elements lying side-by-side as in Figure 11 is obtained as

l[Hc (i) cos Hc (i + 1) cos ] = IM (36)

Here Hc (i) is the coercive force in element i in the p̄ direction and l is the length along the common
side of the two neighboring elements. Thus the equivalent current-sheet density representing the magnet
JM is obtained by dividing the value of the equivalent current sheet IM , obtained in equation 34 by
the length l. The finite-element program calculates the magnetic vector potential in permanent-magnet
excited systems based on equations 32 and 36.

14
 Figure 9: A zoomed view of the airgap region

Bp

Br

Hp
Hc

Figure 10: Shifted Magnetization Curve

6.5 Modelling the stator slots

Each of the models was assumed to have the same 36-slot, six-pole stator, with a distributed two-layer
winding. Since only one pole-pitch is required for modelling, this means any situation can be represented
with six consecutive slots as shown in Figure 12.
The total current in a slot is obtained from the instantaneous value of the stator current in each phase
at a given moment. The current load-angle is taken to be the electrical angle between the d-axis and the
centre of the slot having a net total current of zero, as shown in Figure 12.Thus for the situation shown
in the figure, a current load-angle of 90 electrical degrees would result if the current in the a-phase is at
its peak. The current distribution in the slots for load-angle values of between zero and 90 degrees is as
shown in Table 4.
The armature current in an m-phase distributed fractional-pitch winding sets up an mmf in the airgap
with a fundamental peak value given by
4 m Nph kw p
F̂s = 2Is (37)
⇡ 2 2npp

Thus taking the d-axis as the reference, the armature mmf wave at an instant of time can be expressed
as
Fs (✓e ) = F̂s sin (✓e + ) (38)
where ✓e is an angle from the d-axis in electrical degrees, and is the angle between the d-axis and the
peak of the fundamental of the armature mmf wave. This angle is in e↵ect the same as the current angle

15
 Y

i
p

i +1

Figure 11: 1st order 2D elements along the surface of a permanent magnet

i Phase Currents Slot Currents [A]

el.deg ia ib ic s1 s2 s3 s4 s5 s6 s7
0 0 0.866 -0.866 -51.96 -103.92 -51.96 0 51.96 103.92 51.96
10 -0.174 0.94 -0.766 -51.18 -91.92 -35.54 20.84 66.82 112.80 51.18
20 -0.342 0.985 -0.643 -48.84 -77.16 -18.06 41.04 79.62 118.2 48.84
30 -0.500 1.00 -0.500 -45 -60 0 60 90 120 45
40 -0.643 0.985 -0.342 -39.81 -41.04 18.06 77.16 97.68 118.2 39.81
50 -0.766 0.94 -0.1736 -33.41 -20.84 35.54 91.92 102.36 112.8 33.41
60 -0.866 0.866 0 -25.98 0 51.96 103.92 103.92 103.92 25.98
70 -0.94 0.766 0.174 -17.77 20.84 66.82 112.8 102.36 91.92 17.77
80 -0.985 0.643 0.342 -9.03 41.04 79.62 118.20 97.68 77.16 9.03
90 -1 0.5 0.5 0 60 90 120 90 60 0

Table 4: Current distribution in slots for various current angles i and a stator current of 1A p-p

i defined earlier. The slot current distribution can be deduced by computing the equivalent armature
surface current wave from the mmf wave.
Z
Ks = F̂s sin (✓e + )d✓e (39)

Thus each slot can be assumed to be carrying a current equal to the equivalent surface current over
a slot pitch, given by
Z ✓i+1
Isi = F̂s sin (✓e + )d✓e (40)
✓i

where ✓i is the electrical angular distance of the ith slot from the d-axis and Isi is the current in the ith
slot. Thus for simulating the stator current at any given time, the slots are each loaded with current
equal to
Isi = F̂s [cos (✓i + ) cos (✓i+1 + )] (41)
The above scheme was compared with another scheme in which the current in each slot was determined
simply as the net total of ampere conductors in the slot, for a particular current angle and at a particular
moment, taking into consideration the direction of flow of current in each coil side.No noticeable change
in computation results were observed.

6.6 Solving the Magnetic Field Problems

The MagNet2D package comes with several ”solvers” which constitute program modules for solving the
magnetic field problems that have been defined earlier using the problem editing module. Each solver is
capable of solving a particular class of problems which are classified according to the following criteria:
1. Coordinate system :
(a) Cartesian or (x-y) solver

16
 d-axis
stator core
c b b a a c c
b b a a c c b

magnet

Figure 12: Distribution of coil sides in slots

(b) Axisymmetric (r-z) solver

2. Problem type:
(a) Non-linear without hard materials
(b) Non-linear with hard materials
(c) Time-harmonic linear without hard materials
(d) Time-harmonic non-linear without hard materials
Based on combinations of the above criteria eight di↵erent solver programs were available for DOS
systems. For this particular work the xypm solver was used. This solver is designed for solving non-
linear problems with hard materials in the cartesian coordinates. It accepts several problems as its input
and then proceeds to solve each problem individually until no problem remains to be solved. For each
problem the solver executes conjugate-gradient iterations within Newton-Raphson ones and continually
reports on the iteration progress. Whenever a problem is being solved for the first time on the solver, it
starts by reordering the mesh and issuing a progress report as it does so. Subsequent solution of problems
defined on the same model does not require any reordering of meshes and hence proceeds relatively faster
than the first solution. The solution time for each problem depends very much on the number of nodes
in the model which in turn is a function of the number of elements or meshes generated.

6.7 Post-processing
The solution to the magnetic-field problem is used in this stage of the finite-element method to carry out
torque calculations under di↵erent operating conditions, and to compute such parameters of the motor
as the d- and q- axes inductances Ld , Lq and the magnet flux per pole M , etc. The postprocessor
program-module, Post2D, uses a number of stack based operations to compute various parameters from
the magnetic field solution. A very useful device is the possibility of creating user-defined verbs that are
used to carry out specific series of operations on the stacks so as to obtain desired parameters.
Specifically the following values were computed in the postprocessing stage of the PMSM design:
1. Average flux density per pole in the airgap due to the magnets alone for various magnet sizes
2. Radial component of airgap flux density due to the magnet alone also for di↵erent magnet sizes
3. Peak magnetic flux density in the most-stressed tooth of the stator under varying load conditions
4. Electromagnetic torque developed by the motor for di↵erent current angles and/or stator current
values
5. The q- and d-axis inductances of the machines
Post-processing is a very important stage in the finite-element analysis of the machine, and care
should be taken in operating on the field solution so as to obtain reasonable interpretation of the result.
In particular the definition of contours of integration for obtaining energy values and other data, must be

17
carried out with care so as to obtain results that are not too localised. The basic post-processor commands
are extremely simple operations on the three stacks of the processor. Thus a simple computation usually
involves a number of these basic commands which makes the postprocessing procedure seem rather
cumbersome. This problem is solved by the possibility of creating user-defined verbs that are made up
of a series of post-processor commands compiled together and invoked by a single name. Listings for
some of the user-defined verbs used in the post-processing stage of the finite-element analysis are shown
in Appendix ??.

Figure 13: Computed radial flux density in the magnet

The normal component of flux density, Bn , for a contour is computed as the derivative of magnetic
vector potential Ā in a direction tangential to the contour on each of a set of 128 points. For a rotary
machine the contour constitutes an arc the length of which corresponds to one pole pitch. The normal
component in this case is radially directed either towards or away from the rotor surface. Computing
the normal flux density either in the airgap or in the magnet itself involves several basic post-processor
commands which have been conveniently arranged and saved as a single system-defined verb. Simi-
larly, system-defined verbs are available for computing the flux density magnitude, Bmag , the tangential
component of flux density, Bt , the X- and Y- components of flux density, Bx and By in the cartesian
coordinates, and so on. Figure 13 shows the radial component of flux density inside the magnet on load,
ie with the stator unexcited.

Figure 14: Computed radial airgap flux density due to magnet alone

Figure 14 shows the airgap radial flux density again with no stator current. The e↵ects of slots on
the no-load flux are clearly visible. Fourier analysis of these waves was carried out using a Modula-2
program written by J.D.Edwards which takes as its input the 128 values of the flux density generated
by the post-processor, and outputs the amplitudes of the fundamental and several higher harmonics in

18
a neat form. Figures 15 and 16 are the graphic outputs of the program, showing the actual flux density
values, and the corresponding fundamental harmonic values.

Figure 15: Radial flux density in the magnet and the corresponding fundamental harmonic: Y-axis range
is -1.25 to 1.25 T

Figure 16: Radial airgap flux density due to magnet with the corresponding fundamental harmonic: Y-axis
range is between -1.25 to 1.25 T

For torque computations, contours were defined in the central area of the airgap. The torque value
for each contour was computed and the average value of these torques was then taken as the torque
developed. There are several possible approaches to the computation of torques and forces from the
magnetic field solution. These include
1. Volume integration of force density
2. Maxwell stress formulae
3. Virtual-work approach
4. Energy derivative method
The first two methods rely on material and field data values in one single relative position of the movable
versus the fixed parts of the machine, while the last two require the determination of energy values in
several positions of the movable parts in order to compute the torque. The Maxwell stress method sums
up local stress values at all points of a bounding surface by means of a surface integral to find the net
electromagnetic force. This was the method used to compute the torque values for all the models.
The inductance values were computed by exciting the stator while representing the magnet as air.
Here too the computations were carried out along various contours from which an average value was
obtained. The magnetic energy per pole of the machine was used for inductance computations. Listings
of user-defined verbs used for computing some of these values are shown in Appendix ??

7 Loss Analysis
The total no-load loss in the machine is made up of mechanical loss Pm and armature iron loss. The
analysis of no-load armature iron loss was carried out based on a method presented by Dote and Kinoshita

19
in [4]. The method relies on quasi-linearised mechanical and magnetic modelling of the surface-magnet
synchronous motor and makes extensive use of empirical formulae for the computation of both mechanical
and iron losses.

7.1 No-load Mechanical Loss

This is made up of the bearing friction and windage losses. The bearing friction loss in turn is given by

2
Pbf = 1.5 ⇥ vsh ⇥ 10 W (42)
D
Here is the force applied on bearing’s outer ring in kg, D is the diameter of the steel ball in cm, and
vsh is the peripheral speed of the shaft in m/s.
The windage loss on the other hand is related to the air capacity V of the ventilating fan and the
wind velocity, vw as given by the equation

Pw ⇠ 2
= 1.1 ⇥ V ⇥ vw W (43)

Thus the no-load mechanical loss becomes

Pm = Pbf + Pw (44)

The above method for computing the no-load mechanical loss was not used because of the difficulty in
obtaining the required data. Instead, the rundown test was used to measure these losses. [1]

7.2 Armature Iron Loss

No load iron loss in the armature is made up of hysteresis, eddy current and skin e↵ect losses. Hysteresis
and eddy current losses are generated as a direct result of the alternating of magnet flux with respect
to the stator iron. In order to estimate them the machine magnetic path is divided into two sections,
the core section and the slot section. Each part has its cross section and its weight and its own value of
peak flux density. Thus the hysteresis loss is given by
f 2
Ph = " B ⇥ Gi (45)
10 im
while eddy current loss is given by
✓ ◆2
f 2
Pec = k0 Bim ⇥ Gi (46)
100

ho

tw
hs

Ro

Figure 17: Dimensions of stator magnetic paths

In these equations, Bim represents the maximum flux density in the ith part of the magnetic path,
and Gi is the corresponding weight of the path in kg, and f is the operating frequency in Hz. The
factors " and k0 are chosen from table 5 for the material of the stator iron core. These values are used
to calculate Ph and Pec for each section of the magnetic path. The values of Bim and Gi for each
part of the magnetic path section are calculated using figure 17 as follows.
The core section (see figure 17) has a height of ho and a magnetic path cross sectional area given by

a c = ho L f e (47)

20
where, Lf e is the stack length of the stator. The magnetic flux density in the core section is given by
1
2 g g
Bc = = (48)
ac 2ho Lf e

The weight of iron in the core section is computed as

Gc = ⇡(2Ro ho )ho ⇢ (49)

where Ro is the outer radius of the stator core and ⇢ is the density of the stator steel. The slot section

Core Material Thickness[mm] " k0

Ordinary steel 1.0 4.4 22.4
0.50 4.4 5.6
0.35 4.7 3.2
High silicon steel 0.5 3.0 1.2
0.35 2.4 0.6

Table 5: Iron-loss constants for steel

has a magnetic path cross-sectional area given by

Zs
a s = tw L f e ⇥ (50)
npp

Here Zs is the number of stator slots, tw is the width of a stator tooth and npp is the number of pole
pairs. The weight of iron in the slot area is given by

G s = h s tw L f e Z s ⇢ (51)

where hs is the slot height (see Fig. 17. The magnetic flux density in the slot area is given by

g g
Bs = = (52)
as tw Lf e nZpp
s

Results of iron-loss computations are shown in Table 6.

Section Cross section Flux Density Weight of Iron Ph Pec

of magnetic path [m2 ] [T] [kg] [W] [W]
Core Section 8.22 ⇥ 10 4 1.55 2.57 82.6 3.11
Slot Section 5.1 ⇥ 10 3 1.004 2.04 27 1.03

Table 6: Iron-loss computations

8 Rotor Construction
To ensure magnetic neutrality of both the shaft and the frame of the PMSM it is necessary to maintain
symmetry of the magnetic circuit. Net shaft flux, for example causes bearing currents leading to bearing
damage. Thus all poles should be equal with regards to magnet flux and reluctance of magnetic path.
The following harmful imperfections were avoided while constructing the rotor;
• stationary and/or rotating eccentricity of the rotor in the stator bore
• reluctance variations in the core due to anisotropic e↵ects
• unequal current distribution between winding parts connected in parallel but located on di↵erent
poles
• mechanical imbalance leading to vibrations.

21
Another important requirement for the smooth operation of the permanent-magnet motor is that the
rotor magnets be securely fixed to the rotor, without any risk of the magnets vibrating. To this e↵ect
very many ingenious methods of retaining the magnets have been suggested and/or used. The internal-
magnet construction, where the magnets are placed in slots inside the rotor iron, is the most popular
form of construction, as this gives not only a robust machine, but also a machine that is capable of
developing reluctance torque as a result of the magnetic saliency. The disadvantages of the internal-
magnet construction include the heterogeneous nature of the rotor and the costly processes of cutting
the magnet slots.

Figure 18: A neodymium magnet piece and two aluminium wedges used in the rotor construction.

Figure 19: PM rotor in the process of assembly

In the surface-magnet machine, the task of keeping the magnets attached to the rotor surface is not
simple. In particular, if the machine is to operate at high speeds, there is the danger that the magnets
might get thrown o↵ the rotor and get crushed. Thus some non-magnetic retaining shells might be needed
to keep the magnets from moving in any direction relative to the rotor iron. For low-speed machines,
however, the magnets are simply glued to the rotor using some sort of bonding.
The method used to keep the magnets attached to the rotor in this design employs aluminum retaining
bars wedged between neighbouring magnets as shown in figure 20. Each pole of the machine was built
with 3 arc-shaped Neodymium magnet-pieces stacked one after the other along the axis of the rotor.
To ensure minimum torque ripple, the magnet pieces were staggered thereby creating a skew of one
slot-pitch. The interpolar region was filled with aluminium bars acting as wedges to the magnets and
held to the mild-steel rotor cylinder by aluminium screws. Finally both ends of the rotor were fitted with
aluminium rings to prevent any movement of the magnet pieces in the axial direction. The constructional
details of the permanent-magnet rotor are shown in figures 18 19 and 20

22
 Figure 20: The completed rotor

9 Finite Element Results

In the design stage a variety of rotor topologies were modelled so as to asses the e↵ects of such things as
magnet shape and location, airgap width, orientation of the magnet axis etc on the developed torque,
saliency, and armature reaction. It was also possible to study the e↵ects of using various materials
as wedges, and flux barriers in order to concentrate the flux density into the airgap. Thus at this
stage of the analysis the solutions to the magnetic field problems were essentially used to determine
the torque-production capacity of the various models developed. The second part of the finite element
analysis involved the rigorous study of a chosen rotor configuration, with a view to predicting its torque
capability, inductances and optimum current load-angle.

Figure 21: No-load flux plot of the prototype PMSM

9.1 Model Comparison

Flux plots of the 4 models chosen for comparison are shown in Fig. 27. These are classified as follows;

1. PMSM1 - an internal-magnet motor with circumferential orientation of the magnet axes

2. PMSM2 - an inset-magnet motor with arc-shaped magnets and large pole-arc angle
3. PMSM3 - an inset-magnet motor with arc-shaped magnets and small pole-arc angle

23
Figure 22: Flux plot of the PMSM for di↵erent current-angles with the rotor unmagnetized: a) = 0 el.
deg.; b) = 30 el. deg; c) = 60 el. deg; d) = 90 el. deg.

4. PMSM4 - a surface-magnet motor with tapered arc-shaped magnets

All the models di↵er only in their rotor configurations. The stator is in each case of identical physical
dimensions, as given in the preliminary part of the motor design. Furthermore, the magnet volume for
all the models has been maintained constant.

Model Current Angle Magnet Flux Peak Torque (Iˆa = 1.7A)

Name imax [degrees] M [mWb] [Nm]
PMSM1 98 2.32 10.01
PMSM2 118 2.23 8.99
PMSM3 125 2.22 9.51
PMSM4 90 2.55 8.7

Table 7: Peak torque devloped, the corresponding current-angles, and the airgap flux due to magnets for
various models

10 Results of tests on the prototype motor

Results of measurements carried out on the prototype motor are presented here. First a no-load test was
carried out with the machine operating as a generator, driven by a DC motor. This test clearly showed
that the emf of the designed machine is sinusoidal in shape. Furthermore a linear relationship between
the speed and the open circuit voltage was evident, as can be seen in Fig. 29.
The machine was not capable of self-starting, which means that there is very little or no damper
e↵ects. Care had been taken so that the aluminium retaining bars did not form an equivalent squirrel-
cage together with the end-rings. This was avoided by shortening the bars so that they did not make
electrical contact with one another.
The torque/stator current characteristic of the motor was measured for various current load-angles
using a DC-machine -based dynamometer. The current load-angle was varied with the help of a rotary
digital switch which generates codes corresponding to rotor-shaft position steps that could be either added
or subtracted from the shaft-position signal coming from the counter-bank connected to the incremental
encoder.

24
 Figure 23: Radial component of magnet flux with a q-axis-oriented model

Current angle i Torque (Nm)

in degrees PMSM1 PMSM2 PMSM2 PMSM1
0 0 0 0 0
15 1.51 0.7 -0.1 1.42
30 2.72 1.6 0.3 2.65
45 3.83 2.5 1.0 3.70
60 4.71 3.4 1.94 4.52
75 5.41 4.15 3.0 5.01
90 5.82 4.75 4.1 5.13
105 5.75 5.1 5.09 5.01
120 5.26 5.3 5.5 4.52
135 4.61 5.15 5.3 3.71
150 3.41 4.15 3.85 2.63
165 1.72 2.2 1.85 1.43
180 0 0 0 0

Table 8: Torque/angle Characteristics of various models for stator current ia =1.0A

11 Conclusion
A brief description has been given of the design and construction of a 6-pole surface-mounted PM
synchronous motor using computer-aided-design techniques. The motor was designed to operate from a
variable-frequency, variable-amplitude inverter with a sinusoidal current output. The method of binding
the rotor magnets to the rotor does not involve the use of banding or glue, so that the e↵ective airgap is
very small. This has produced a machine that has a very high utilization of magnet, and a considerably
light rotor. The torque capability was higher than projected, due to the deliberate increase of the required
magnet length so as to ensure a robust mechanical design.
The motor was subjected to a variety of tests,[1], so as to determine its actual paremeters and torque
capability. The relatively high number of stator winding turns, has meant that the machine back-emf is
comparatively high. This limits the speed range, and calls for a high value of inverter DC link voltage.
In the next chapter an analytical method is developed for predicting the torque capability of PM
synchronous machines. Results of computations of torques and forces carried out both on the finite-
element model of the PMSM and on the prototype motor itself are compared with those obtained in the
analytical method developed.

25
 Figure 24: Tangential component of magnet flux

Figure 25: Finite-element flux plot of the PMSM for di↵erent load angles: a) 0 deg.; b) 30 deg.; c) 60
deg.; d) 90 deg.

26
Figure 26: Flux plot of one of the various models tested, showing current-allocation to slots for a current-
angle of 90 el. deg.

Figure 27: Flux plot of 4 di↵erent models of permanent-magnet synchronous machines that were analysed:
a) Internal-magnet type; b)Inset-magnet type with large pole-arc angle; c) Inset-magnet type with small
pole-arc angle; d) Surface-magnet type

27
 Figure 28: Torque/current angle characteristics for various models of the PMSM

250
*

200 *
Open circuit emf per phase [v]

150 *

*
100
*

*
50
*

0*
0 100 200 300 400 500 600

Rotor speed [rpm]

Figure 29: Measured open circuit characteristics of the PMSM

16
1o
2x
14
1-current load-angle=90 deg
o 3 *
2-current load-angle=75 deg x
12
3-current load-angle=60 deg
*
4-current load-angle=30 deg
10 xo
Torque [Nm]

*
8 o 4+
x

* +
6
o +
x
*
4 +

xo +
2 *
+

0 +x*o
0 0.5 1 1.5 2 2.5 3

Stator current [A]

Figure 30: Measured torque/stator current of the surface PMSM for various load angles

28
 6

5
Torque developed [Nm]

0
0 20 40 60 80 100 120 140 160 180

Current load angle [degrees]

Figure 31: Measured torque/stator current angle characteristic of the surface PMSM

29
References
[1] ALHASSAN, M.H.: ’Permanent Magnet Synchronous Motors in Position Control Systems,’ D.
Phil Thesis, University of Sussex, Brighton, 1993.
[2] BOSE, B.K.: ’A High-Performance Inverter-Fed Drive System of an Interior Permanent Magnet
Synchronous Machine,’ IEEE Trans. on Industry Applications, 1988, 24, (6), pp. 987-997

[3] CORNELL, E.P.: ’Permanent Magnet AC Motors,’ Conf. Proc. Drives/Motors/Controls, 1983,
pp.102-107
[4] DOTE, Y. and KINOSHITA, S. (-editors): ’Brushless servomotors: fundamentals and applications,’
Clarendon P. Oxford 1990.

[5] GANGLA, V. and DE LA REE, J.: ’Electromechanical Forces and Torque in Brushless Permanent
Magnet Machines,’ IEEE Trans. on Energy Conversion, 1991, 6, (3), pp. 546-552
[6] HAGUE, B.: ’The Principles of Electromagnetism Applied to Electrical Machines,’ New York:
Dover 1962

[7] LEVI, E.: ’Polyphase motors: a direct approach to their design,’ Wiley New York 1984.
[8] LEVRAN, A.and LEVI, E.: ’Design of Polyphase Motors with PM Excitation,’ IEEE Trans. on
Magnetics, 1984, MAG-20, (3), pp. 507-515
[9] LOWTHER, D.A. and SILVESTER, P.P.: ’Computer-Aided Design in Magnetics,’ Springer-Verlag
N.Y.1986.

[10] MILLER, T.J.E.: ’Brushless Permanent Magnet and Reluctance Motor Drives,’ Oxford University
Press, New York, 1989.
[11] PAVLIK, D., GARG, V.K., REPP, J.R. and WEISS, J.: ’A Finite Element Technique for Calcu-
lating the Magnet Sizes and Inductances of Permanent Magnet Machines,’ IEEE Trans. on Energy
Conversion, Vol. 3, No. 1, March 1988, 3, (1), pp. 116-122
[12] ŠTAFL, M.: ’Electrodynamics of Electrical Machines,’ Academia, Prague, 1967.

30