IEEE TRANSACTIONS ON MAGNETICS, VOL. 46, NO.

7, JULY 2010 2675

Unbalanced Magnetic Pull Due to Asymmetry and Low-Level Static Rotor

Eccentricity in Fractional-Slot Brushless Permanent-Magnet Motors
With Surface-Magnet and Consequent-Pole Rotors
David G. Dorrell1 , Mircea Popescu2 , and Dan M. Ionel3
School of Electrical, Mechanical and Mechatronic Systems, University of Technology Sydney, Sydney NSW 2007, Australia
Motor Design Ltd., Ellesmere, Shropshire, SY12 0EG, U.K.
AO Smith Corp., Corporate Technology Center, Milwaukee, WI 53224 USA

We report on an investigation into the unbalanced magnetic pull (UMP) in ferrite-magnet fractional-slot brushless permanent-magnet
motors due to either magnetic asymmetry or static rotor eccentricity. We justify the work in terms of the establishment of total indicated
runout (TIR) tolerance and use 10% eccentricity as the nominal tolerance. The UMP will generate force and vibration, which causes
wear. We investigated several machines with different pole numbers, slot numbers, and winding arrangements. Some of the windings
contain sub-harmonics, and we found these to be more susceptible to UMP when there is rotor eccentricity and also to produce vibrating
UMP. We found that consequent rotor poles produce high UMP under centered and rotor eccentricity conditions. We conducted detailed
finite-element studies to calculate the UMP.
Index Terms—Brushless permanent-magnet motors, fractional slot, unbalanced magnetic pull.

I. INTRODUCTION layouts for single- or double-layer windings. Ten percent static

NBALANCED MAGNETIC PULL (UMP) is usually as-
U sociated with larger induction and synchronous machines
under winding faults or rotor eccentricity conditions [1]–[3] and
there is a large amount of literature on this subject. However,
smaller machines are still subject to unbalanced magnetic pull
[4], [5] even without rotor anomalies. Unbalanced magnetic pull
is important because it affects the wear on the bearings [6] as
well as noise and vibration [7]. This is particularly the case in
brushless servo motors where fractional slots are used [8]. This
type of machine is studied here for the generalized case and sev-
eral examples are used to highlight aspects of UMP generation
in brushless PM machines. An analytical model is developed
using rotating field theory and this is used to identify possible
sources of UMP and the associated vibration frequencies. The
example machines are analyzed use finite element analysis to
obtain the magnitudes of the UMP and small rotor eccentricity
is explored.
In this paper, example machines are put forward. These de-
signs are typical of industrial permanent-magnet servo motors
and were furnished by a manufacturer. They consist of an 8-pole
9-slot machine and a 6-pole 9-slot machine; the former is an ex-
ample of a machine that can have substantial UMP even when
the rotor is centered. Results are obtained from the FEA to verify
the method. The analysis is then extended to include rotor ec-
centricity; a 12-pole 18-slot machine and a 16-pole 18-slot ma-
chine are studied. These machines can have different winding
Manuscript received August 05, 2009; revised November 23, 2009 and Jan-
uary 28, 2010; accepted February 11, 2010. First published March 08, 2010;
current version published June 23, 2010. Corresponding author: D. G. Dorrell
(e-mail: david.dorrell@uts.edu.au).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TMAG.2010.2044582
0018-9464/$26.00 © 2010 IEEE
rotor eccentricity is put into these machines to assess the UMP
under conditions that would produce UMP but not negate oper-
ation. This is often called the total indicated runout (TIR) tol-
erance if it was set as the eccentricity level that is acceptable
after manufacture. If the eccentricity is high then the UMP will
be excessive generating high audible noise (which will flag up
a developing fault when inspected) or complete rotor pull-over
in the air gap which will cause drive failure. These machines
are often servo drives and under ac (sine wave current) control.
However, brushless dc (trapezoidal current) control will also be
investigated.
While there is much literature on rare-earth PM motors, this
paper addresses the operation of ferrite machines, which are still
manufactured in very large numbers since they are cost effec-
tive to manufacture and straightforward to magnetize. UMP in
larger rare-earth magnet machines with only 4 poles was inves-
tigated in [13]. This also studied the UMP at higher levels in per-
manent-magnet machines; it illustrated that the UMP is almost
linear with a degree of eccentricity up 80% (which is in contrast
with UMP in cage induction motors [9]). In this paper, a TIR tol-
erance level of eccentricity (10%) is used since this will cause
wear and vibration over prolonged time rather then sudden rotor
pull-over. This is the rationale behind the work and is very rele-
vant to the industry in terms of specifying the correct grade and
type of bearing and predicting vibration levels. The focus of the
simulation work will be on static eccentricity (rotor rotating on
its own axis but displaced from the stator bore center)—this is
more likely to exist in a machine since it is a function of bearing
misplacement in assembly. Accurate turning of the rotor on the
shaft should ensure low dynamic eccentricity (where the rotor
is not turning on its own axis but is turning on the stator bore
center) unless the bearing is faulty or worn. Obviously it is pos-
sible for both to exist simultaneously.
The motors selected, as already mentioned, are typical frac-
tional-slot permanent-magnet motor arrangements. There are
actually few studies on UMP in these types of machines, even
2676 IEEE TRANSACTIONS ON MAGNETICS, VOL. 46, NO. 7, JULY 2010

though surface-magnet rotor machines are extremely common. where are the phasors of the complex Fourier components
The consequent-pole rotor machine is also studied as an ex- (or waves) of the air-gap flux density with general harmonic, or
ample of an alternative which may first appear a reasonable pole-pair, and . These waves rotate with velocities and
substitute, but exhibits much higher levels of UMP which may . Rather than use the inverse of the average air-gap radius ,
cause higher vibration levels and bearing wear if a direct substi- then is used where . The forces in two perpendicular
tution is carried out. radial directions can be found. The force on the axis is
There have been studies that have measured the UMP di-
rectly. Larger induction machine UMP was measured in [14].
If magnetic bearings are used then it is possible to use feedback
to derive the UMP [15]. UMP calculation in smaller machines
has been attempted [16] but it is still a difficult measurement
problem to attempt. In this paper, UMP is addressed by careful
simulation using finite-element analysis. When conducting the
UMP simulation in this environment it is necessary to take care
to set the air-gap mesh correctly and with a dense mesh. At least
four layers of nodes are required in the air gap.
(3)
II. ANALYTICAL ANALYSIS
This then becomes
In this section an analytical model is put forward to help iden-
tify the sources of UMP and quantify the frequency of any vibra-
tions. Fractional-slot brushless permanent-magnet motors have
a complex spectrum of flux waves in the air gap. In this sec-
tion, the flux wave terms and calculation of UMP will be briefly
investigated.

A. Rotating Air-Gap Flux Waves and Unbalanced

Magnetic Pull (4)
UMP is generally calculated using the approximation for the
normal stress around a closed contour in the air gap. If is the And for the force on the (vertical) axis
force in the direction, then

(1)

where is the radius of the contour, is the permeability of free

space, is the axial length, is a general instantaneous
radial flux density value (at time ) of harmonic (or pole-pair
number) at point , and is the linearized distance around the
contour (which is considered as being at the center of the air gap (5)
in a rotating machine). It is commonly known that UMP will be
generated when there are two air-gap flux waves with pole-pair
numbers differing by one [9]. The relative rotational velocities
of the waves will dictate as to whether the force is a steady pull
or a pulsating force. Multiplying two flux waves together with
pole-pairs differing by one gives
(6)

This illustrates that UMP is generated by pole-pair numbers

varying by one. This can be observed by the limits in (4) and (6),
while (2) shows the flux waves are of general pole-pair numbers
and . If they are different by more than one then their in-
teraction does not contribute to the UMP. The signing of and
simply indicates the direction of rotation. The sum or differ-
ence of and gives the frequency of vibration. Indeed, if
(2) it is zero then this shows a steady pull in one radial direction.
DORRELL et al.: UNBALANCED MAGNETIC PULL DUE TO ASYMMETRY AND LOW-LEVEL STATIC ROTOR ECCENTRICITY
These expressions for the UMP can be used to identify the pole
number combinations that result in UMP.
B. Rotor Air-Gap Flux Waves
The magnets will produce a trapezoidal shaped rotating flux
and this can be put into a complex form [9], where
(7)
where 1, 3, 5, etc. This time the angular position around
the air gap is used where . A general phase angle is
set by . The synchronous rotational velocity of the rotor is
and this is matched to the supply frequency (in rad/s) by
the equation where is the pole-pair number of the
machine. If stator slotting is included and restricted to the
fundamental harmonic of the stator slot number then
(8)
where is the coefficient of the slot permeance fundamental
wave.
C. Stator Air-Gap Flux Waves
In many machines it can be assumed that the winding is a bal-
anced 3-phase winding. However, in a fractional slot machine it
should not be assumed so that the winding MMF is made with a
fundamental one pole-pair harmonic with 2nd, 3rd, 4th, 5th, 6th,
7th, etc. windings. The fundamental harmonic has to be taken
as two for the general case and eliminate harmonics if they are
zero. Hence
(9)
where is the winding harmonic of the non symmetrical wind-
ings (shown below). Equations (8) and (9) can be used with and
(6) to identify UMP components and frequencies.
2677
D. Rotor Eccentricity
Rotor eccentricity can be denoted using the method as de-
scribed in [8] using permeance modulation so that the magnet
flux (i.e., the rotor flux) waves can be represented as
(10)
where is the slot number. This equation is for static eccen-
tricity and includes slotting. It can be seen that eccentricity mod-
ulates that MMF to produce air-gap flux waves that vary by one
pole-pair, i.e., the criterion for UMP is met. This assumes that
the eccentricity is restricted to the first permeance harmonic (in
a similar way to the slot permeance modulation approximation).
Eccentricity can be defined as existing as two types. Static
eccentricity describes a condition whereby the rotor is rotates on
its own axis but is not centered on the stator bore axis. This could
be caused by out of tolerance, misplaced or worn mountings or
bearings. Dynamic eccentricity describes a condition whereby
the rotor does not rotate on its own axis but does rotate on the
stator axis so that the point of minimum air gap rotates with rotor
speed. This could be caused by a bent shaft or out of tolerance
manufacturing.
A similar equation exists for the stator air-gap fluxes:
(11)
where, for a balanced and evenly distributed 3-phase machine
winding, 1, 5, 7, etc, in the usual sequence. However, if
it is a fractional and non-evenly distributed winding, which are
commonly used in fractional-slot permanent-magnet machines,
then the fundamental pole number of has to be assumed.
If the winding is not evenly distributed then the general case
will be 1, 2, 3, 4, 5, etc. A balanced 3-phase
winding may eliminate some MMF harmonics (such as the 3rd
harmonics)—individual machine components need studied by
a Fourier breakdown of the winding. Winding harmonics for
the machines under study will be examined in a later section.
Also note that the time component is now in (11) rather than
2678
in (10). This will lead to different vibration components
when the machine is open-circuit (when there are only rotor flux
wave components) and loaded (when there are both stator and
rotor flux wave components) as discussed in Section II-A. In
an induction motor the general solution for UMP gives a steady
pull and twice-supply frequency vibration by the solution of (4)
and (6) using the flux density waves described by (11) and as-
suming the eccentricity is static. However, a permanent-magnet
motor has a more complex set of air-gap waves that can give a
more complex set of vibration terms due to the sub harmonics
and addition rotor flux waves. From (4) and (6), together with
(10) and (11), it can be seen that it is important to identify the
different MMF wave components generated by the stator wind-
ings and rotor magnets. Eccentricity will modulate the the MMF
waves to produce UMP components. The next section looks at
the different winding MMFs in the machine under study.
E. Consequent Pole Rotor Machine
A consequent pole rotor has alternate magnet and steel poles.
Hence all the magnets will face in one direction and be of one
pole orientation, while the other poles will be steel. Therefore
there is one magnet per pole-pair. This will give a particular type
of rotor saliency that will modulate the stator MMF wave.
(12)
where 1, 3, 5, etc. and is the harmonic perma-
nence coefficient for the consequence rotor. This assumes a gen-
eral stator winding starting at one pole-pair. This will have im-
plications in terms of the interaction of a fractional slot winding
with the consequent pole rotor—this can be a source of UMP as
demonstrated later.
III. MACHINE EXAMPLES—DETAILED WINDING ANALYSIS
Several machine examples are put forward in the paper. This
section will study the detailed analysis of the windings together
with the machine dimensions. Complete cross sections are given
in Sections IV and V. The focus will be on two 9-slot ma-
chines—one with 8 poles and one with 6 poles (surface mag-
nets) and three 18 slot machines—two with 12 poles (one with
surface magnets and one with a consequent-magnet rotor) and
one with 16 surface-magnet poles. All the machines have sur-
face ferrite magnet rotors. Ferrite was purposely chosen for this
study in order to eliminate potential unknowns related to losses
in the magnets and to provide a relatively low magnetic loading
that would limit saturation and enable a more straightforward
comparison of numerical and analytical results. Internal PM ro-
tors with NdFeB magnets will be the focus of future work. The
machines are modeled in SPEED PC-BDC [11] to examine the
winding layout harmonics then passed through to the finite ele-
ment package PC-FEA [12] to obtain the UMP under different
IEEE TRANSACTIONS ON MAGNETICS, VOL. 46, NO. 7, JULY 2010
TABLE I
MACHINE PARAMETERS AND RATINGS
loading conditions. The teeth are bifurcated. The machines have
the same air-gap length, axial length, electric loading, and diam-
eter and the nominal specifications are given in Table I. The ma-
chines will be simulated at a speed on 500 rpm and the full-load
torques are given at 4.5 A (on the q-axis). The amp-turns and
the basic geometry (including the axial length) are identical for
all machines. This allows direct comparison.
A. 9-Slot Machines
In Fig. 1, the 9-slot machines are shown. The top left machine
(8 pole) shows the that winding a nonsymmetrical compared to
the 6-pole machine and this is reflected in the winding harmonic
distribution as shown, where the 8-pole machine has many har-
monics, including sub-harmonics (below 4). Single phase and
total winding harmonics are shown (sum of all three phases and
assuming a balanced 3-phase current set); some harmonics are
eliminated by the summation. The 8-pole machine was shown
to exhibit high UMP under loading in [7] and it is obvious from
the winding layout that there will be a pull towards the excited
winding due to asymmetry and a large number of winding har-
monics with pole-pair numbers varying by one. The 6-pole ma-
chine has a symmetrical winding; however, the rotor pole–stator
slot combination is asymmetrical. Symmetry here is defined as
having diametrically identical geometry. This can be identified
by the fact that rotor pole faces are diametrically opposite each
other whereas a tooth and a slot opening are diametrically op-
posite. This will produce UMP due to asymmetry; [17] explains
this slotting UMP. There are no pole-pair winding harmonics
that vary by one so there should be no additional load UMP.
An exercise to identify the UMP components can be con-
ducted for the 9-slot 8-pole machine since this will produce high
UMP even without rotor eccentricity. In Table II the air-gap flux
waves are tabulated to identify the pole numbers and associated
vibration frequencies. Table III gives an example of matching
pole-pair numbers and the resulting vibration components. In
[8], it was shown that the asymmetrical winding dominates the
UMP and this has a vibration of which is twice the supply
frequency. Similar tables can be developed for all the machines
when eccentricity is included in the table.
In (10) and (11), it is seen that the slotting of the stator will
also interact with the magnets to possibly generate additional
waves which may produce some UMP.
B. 18-Slot Machines
Figs. 2 and 3 show the 18-slot 12-pole machines with double-
and single-layer windings while Figs. 4 and 5 repeat this for
the 16-pole arrangement. While none of these machines have
DORRELL et al.: UNBALANCED MAGNETIC PULL DUE TO ASYMMETRY AND LOW-LEVEL STATIC ROTOR ECCENTRICITY
Fig. 1. Winding distributions for 9-slot 8-pole machine—(a) and (c), and 9-slot
6-pole machine—(b) and (d); showing one phase of the 3-phase winding and
phase winding harmonic.
TABLE II
ROTOR AND STATOR AIR-GAP FLUX WAVES FOR 9-SLOT 8-POLE MACHINE
TABLE III
VIBRATION COMPONENTS FOR MATCHING POLE-PAIR NUMBERS (NOT A
COMPLETE LIST) FOR 9-SLOT 8-POLE MACHINE. NOTE: FOR 8-POLE MACHINE
THE SUPPLY FREQUENCY (rad/s) IS 4 TIMES THE ROTATIONAL VELOCITY
an asymmetrical windings (as with the 9-slot 8-pole machine),
the 16-pole machine has a series of winding harmonics starting
from a 4-pole (double layer) or 2-pole (single layer) harmonic
2679
Fig. 2. Phase winding distribution and stator lamination topology for
18-slot 12-pole machine. (a) Double-layer configuration and (b) single-layer
configuration.
Fig. 3. Mechanical MMF harmonics for 18-slot 12-pole configuration. (a)
Double-layer configuration and (b) single-layer configuration.
that may produce higher UMP when there is some eccentricity.
The harmonic series in Fig. 3 for the 12-pole motor does not
show any winding harmonics with pole-pair numbers varying
by one so there should not be UMP with the surface-magnet
rotor when the rotor is centered.
For the 16-pole machine the winding harmonic division in
Fig. 5(a) is two, so there are no winding harmonic pairs with
pole-pairs varying by two for the double layer winding so no
centered-state UMP. However, in Fig. 15(b), the division is one
and there are several instances of pairs of harmonics varying by
one (most notably 1 and 2 together with 6 and 8) which may
produce UMP. This should be at twice-supply frequency (as il-
lustrated in the 9-slot 8-pole machine). This is examined in the
FEA analysis. However, when using a consequent pole rotor,
there may be interactions between flux waves due to the conse-
quent pole modulation illustrated by (12), even when the rotor is
centered. This is again investigated in the FEA analysis section.
2680
Fig. 4. Phase winding distribution and stator lamination topology for
18-slot 16-pole machine. (a) Double-layer configuration and (b) single-layer
configuration.
Fig. 5. Mechanical MMF harmonics for 18-slot 16-pole machine. (a) Double-
layer configuration and (b) single-layer configuration.
IV. FINITE-ELEMENT MODELING
A. Centered Rotor UMP in 9-Slot 8-Pole Machine
The flux wave harmonics are complex in a fractional slot ma-
chine so that to obtain accurate values for the UMP, finite-ele-
ment analysis is used to calculate it. Once a solution is obtained
then a closed contour of the radial air-gap flux density can be
obtained and the UMP calculated by implementing (4) and (6)
in an elemental form.
The UMP calculations for the 9-slot 8-pole machine when
the rotor is centered, and either open circuit or fully loaded,
are put forward in Fig. 6. On the left are flux plots while the
UMP results are shown on the right. It can be seen that the UMP
oscillates twice with 90 mechanical degrees of movement which
represents one cycle of the current (which is sinusoidal; Phase
A is shown), i.e., it is twice-supply frequency vibration. This
is predicted in Table III under the stator-stator column. On the
right is the UMP on open circuit.
IEEE TRANSACTIONS ON MAGNETICS, VOL. 46, NO. 7, JULY 2010
Fig. 6. 9-slot 8-pole machine (a) flux plot when fully loaded, (b) UMP in x and
y directions over 90 mech deg rotor movement at full loading, (c) flux density
distribution under open-circuit operation and (d) UMP in x and y directions
under open-circuit conditions over 360 mech deg.
There are UMP vibrations as predicted in Table III, which
is a twice-supply frequency vibration for an 8-pole machine,
and this matches the FEA results. This vibration frequency is
derived from (4) and (6) and is a result of two flux waves with
pole-pairs differing by one and rotating with a velocity of in
opposite directions. What should be borne in mind is that UMP
vibrations are net radial forces acting on the rotor (or stator in
the opposite direction). These are not the sole source of vibra-
tion. There are additional vibrations that can occur due to nodal
stator deformity [18] and other tooth forces, especially in other
machines such as the switched reluctance machine [19]. It can
be seen that the open-circuit UMP is much less than the full load
UMP due to the winding asymmetry but magnet symmetry. The
magnets are ferrite magnets and the slot openings are narrow.
The UMP under full load suggests that the machine is liable
to experience high bearing wear during full-load operation (al-
though these machines are commonly used in small production
motors). There does look to be some asymmetry in the flux plot
in Fig. 6(c) on open-circuit, however, this is in the stator yoke
and teeth; the imbalance of flux in the air gap is much less due
to ripple flux crossing the stator tooth tips rather than flowing
around the yoke and teeth. This can be observed by the uniform
flux distribution in the magnets.
B. Centered Rotor UMP in 9-Slot 6-Pole Machine
This machine is likely to have much less UMP than the 8-pole
machine due to the fact that the stator winding is symmetrical.
There is some stator/rotor slotting asymmetry as was discussed
in Section III-A; teeth and slot openings are diametrically op-
posite each other causing a slotting UMP effect [17]. The open
circuit UMP is shown in Fig. 7. What can be observed is that the
DORRELL et al.: UNBALANCED MAGNETIC PULL DUE TO ASYMMETRY AND LOW-LEVEL STATIC ROTOR ECCENTRICITY 2681

Fig. 7. 9-slot 6-pole machine flux density distribution under open-circuit oper-
ation and UMP in x and y directions.

rotor magnets are fully pitched and the slot openings are very
narrow. There are also bifurcated stator teeth (center notches).
This has the effect of making the effective geometry now virtu-
ally symmetrical. Hence, it can be seen that there is very little
UMP due to the slotting in the (horizontal) direction and neg-
ligible in the direction. However, in a different geometry, the
UMP may not be negligible, and it also needed addressing here
for completeness. When loaded, the balanced winding leads
to similar negligible results so they are omitted here. The fre- Fig. 8. 18-slot 16-pole machine with double-layer winding and 10% eccen-
quency of the low-level vibration can be observed to be equal to tricity (a) flux plot when fully loaded, (b) UMP in x and y directions over 45
mech deg rotor movement at full loading, (c) flux density distribution under
the pole number. open-circuit operation and (d) UMP in x and y directions under open-circuit
conditions over 45 mech deg. Centered-rotor F also shown for completeness
C. Centered and 10% Eccentricity in 18-Slot 16-Pole and shows no UMP.
Surface-Magnet Machines
The 9-slot 8-pole machine has inherent asymmetry. The other
machines do not have this obvious asymmetry when the rotor
is centered in the stator bore but the winding breakdown in
Fig. 5(b) shows that the single-layer winding does have winding
harmonics with pole-pairs differing by one. There is always
some degree of eccentricity due to tolerance variation and also
wear and here it is examined using the variation of static eccen-
tricity. The rotor is displaced in the direction and the simu-
lation carried out. Ten percent static eccentricity will be inves-
tigated in the 18-slot machines to highlight the effect. It is also
compared to the UMP when the rotor is centered. Fig. 8 gives the
UMP under these conditions under open-circuit (d) and full load Fig. 9. 18-slot 16-pole machine with single-layer winding and 10% eccen-
conditions (b), for the double-layer winding. The UMP during tricity, flux plot when fully loaded and UMP in x and y directions over 45 mech
deg rotor movement at full loading. Centered-rotor F UMP included to illus-
open circuit operation is almost constant but when fully loaded trate twice-supply frequency vibration is mostly due to windings.
there is a vibration component. The rotation is 45 mechanical
degrees which is movement of 2 rotor poles or one cycle of the
current. The UMP has a constant pull of 25 to 30 N and a 10 N variation of is also shown in Fig. 8(b) under load and this
peak-to-peak oscillation. The oscillation is times the very low, which highlights that the UMP is almost entirely due
rotational frequency which is again equal to the pole number (as to the rotor eccentricity.
with the 9/8 and 9/6 machines) and represents twice-supply fre- Since there is 10% static rotor eccentricity then even with a
quency vibration (Phase A current is given for reference). As symmetrical rotor/stator arrangement there is still UMP when
already stated, this is due to two counter-rotating flux waves open-circuit and this can be seen in Fig. 8(d). There is a steady
with pole-pairs differing by one which have supply-frequency pull in the direction on minimum air-gap length as would be
time variation. In Fig. 5(a), it can be seen that there are sev- expected since there will be high flux density around this point.
eral winding harmonics that vary in pole-pair number by two, In Fig. 9, the 18/16 machine is investigated with the single
and these will be counter rotating. If the rotor is eccentric then layer winding in Fig. 4(b). It can be seen that there is now a
there will now be flux waves with pole-pair numbers varying by much higher vibration component in the UMP when using the
one due to permeance modulation, which will generate a steady single layer winding due to the high winding harmonic compo-
pull, plus the twice-supply frequency vibration when loaded due nents—particularly in the single-layer winding where there are
to additional winding-harmonic flux waves. The centered-rotor a 2-pole subharmonics [Fig. 5(b)]. It should also be appreciated
2682
Fig. 10. UMP in 18-slot 12-pole machine with 10% eccentricity over 60 mech.
deg.; (a) open-circuit, (b) loaded double-layer winding, and (c) loaded single-
layer winding.
that there are winding harmonics that have pole-pairs differing
by one in the single-layer winding so that even when the rotor
is centered then there is reasonably high twice-supply vibration
as shown in Fig. 9 (centered —there is a similar character-
istic for with a 90 deg elec. phase shift; Phase A current is
also included for reference). The eccentricity does amplify the
vibration by about 15%.
D. 10% Eccentricity in 18-Slot 12-Pole Machine
The 18-slot 12-pole UMP results are shown in Fig. 10. These
are for open-circuit and loaded conditions (double and single
layer). It can be seen that the UMP for the 12-pole machine is
just under 25 N—this is a steady pull. This rotor movement is
over 60 mechanical degrees and it shows a minor 36-times rota-
tional frequency vibration. The UMP appears to be dominated
by the rotor magnetic forces interacting with the eccentricity.
The winding harmonics are more sparse for the 12-pole ma-
chine (Fig. 3) and the lower sub-harmonic is 3 (6-pole). There
are also no winding harmonics [Fig. 3(a) and (b)] in either
winding that have pole-pair numbers varying by either one or
two. Hence, there is no twice-supply frequency vibrating UMP
at the centered position (winding pole-pair differences of one)
or at 10% eccentricity (winding pole-pair differences of two),
as was found in the 18-slot 16-pole machine.
V. CONSEQUENT ROTOR MACHINE STUDIES
In this section, the consequent pole rotor arrangement is in-
vestigated in terms of its UMP generation when used with ma-
chines with high winding harmonic content. The saliency in the
rotor will modulate the stator MMF to generate further air-gap
IEEE TRANSACTIONS ON MAGNETICS, VOL. 46, NO. 7, JULY 2010
Fig. 11. UMP in 18-slot 12-consequent-pole machine with 10% eccentricity
over 60 mech. deg.; (a) under open-circuit and loaded with double-layer winding
(b) and single-layer winding (c).
flux waves as illustrated by (12) and this is somewhat compli-
cated in terms of the frequency of the UMP vibrations.
A. 10% Eccentricity in 18-Slot 12-Consequent-Pole Machine
It was mentioned earlier that other machine arrangements will
be investigated. A simple comparison here can be made to a
consequent pole machine, where every second pole is steel so
that the number of magnets is equal to the number of pole-pairs.
In Table I it can be seen that this machine actually produces less
torque since the amount of magnet material is halved. Often,
thicker magnets are used in a consequent-pole machine but here
constant thickness and strength will be maintained.
For direct comparison, alternate magnets in the 18-slot
12-pole machine (with double- and single-layer winding) were
changed to steel in the FEA model. The UMP for these arrange-
ments is shown in Fig. 11. It can be seen that the UMP is now
much higher, up to four times, for the consequent pole machine
and the peak is about 100 N. This is because steel rotor poles
considerably reduce the air-gap length and modifies the air-gap
permeance so that there is a much higher imbalance in air-gap
flux due to the 10% eccentricity. This also has an advantage
when applied to a bearingless machine [10]. There are also
several vibration components in the UMP waveform due to the
loading combined with the rotor saliency. These are examined
in the 18-slot 16-pole machine where these vibrations are far
more pronounced.
A flux plot for the consequent pole machine is shown in
Fig. 12 when loaded using the single-layer winding. It is still
difficult to observe the flux imbalance even at these higher
DORRELL et al.: UNBALANCED MAGNETIC PULL DUE TO ASYMMETRY AND LOW-LEVEL STATIC ROTOR ECCENTRICITY 2683

Fig. 13. UMP in 18 slot 16 consequent pole machine with 10% eccentricity
over 45 mech. deg.; (a) under open-circuit and (b) loaded single-layer winding.

Fig. 12. Flux plot in 18-slot 12-consequent-pole machine with 10% eccen-
tricity and loaded with single-layer winding.

UMP values. This also shows the consequent arrangement of

the magnets in the rotor.

B. Centered and 10% Eccentricity in 18-Slot

16-Consequent-Pole Machine

It can be seen in Sections IV-D and V-A that moving from

surface-magnet to consequent-poles in the 18/12 machine
increased the UMP and also increased the UMP due to loading
(loading had a minimal effect in the surface-magnet arrange-
ment). Therefore, is it is worthwhile investigating the UMP in
a consequent-pole 18/16 machine on open-circuit and when
loaded using the single-layer winding. Fig. 13 gives the UMP
under these situations. On open-circuit the UMP is almost
doubled compared to the surface magnet arrangement. When
the machine is loaded using the single-layer winding the UMP
increases many fold. The -axis (direction of eccentricity)
UMP has an 8 times rotational frequency (which corresponds
to supply frequency vibration—Phase A current included for
reference) with a peak-to-peak force of over 400 N and a steady
pull of about 150 N. There is also a peak-to-peak vibration of
400 N in the perpendicular direction. Since the vibrating UMP
is much increased for the single layer winding, then the UMP
was assessed under load with the rotor centered [as shown in
Fig. 13(b)] and it can be clearly seen that the UMP vibration
is due to the salient consequent-pole arrangement. In fact, the
net UMP at 10% eccentricity in the x direction appears to be
the open circuit UMP in Fig. 13(a) and the centered-rotor UMP
in Fig. 13(b). In Fig. 5(b), the main winding harmonics are
the 1 pole-pair and 8 pole-pair (main motor pole-pair number)
harmonics. If the we consider (9) and only the fundamental
motor pole-pair then
(13)
Fig. 14. Flux plot in 18-slot 16-consequent-pole machine with 10% eccen-
tricity and loaded with single-layer winding.
which is the main rotating 16-pole flux wave. Now consider the
large 1-pole-pair winding harmonic and the air-gap MMF set up
by this. The consequent pole rotor will modulate it so that
(14)
Combining (13) and (14), which represent sizable air-gap flux
waves, will produce UMP vibration with the sum of the time
varying terms equal to the supply frequency . Hence, this
motor arrangement (consequent pole rotor, single-layer, 16 pole
fractional-slot winding) appears to generate large pulsating
UMP when loaded, even when the rotor is centered.
A flux plot is shown in Fig. 14 for the loaded condition. This
time it is possible to see and imbalance in flux and at this point
there is a 200 N force in the vertical direction.
It can be concluded that under some circumstances conse-
quent-pole machines can generate high UMP with even a cen-
tered rotor or a low-level eccentricity when the windings have a
high harmonic content.
2684
Fig. 15. Three-phase open-circuit back-EMF waveforms for 18/12 machine
with double-layer winding (taken from PC-BDC).
Fig. 16. UMP in 18-slot 12-pole machine under dc control with 10% eccen-
tricity over 60 mech. deg. and double-layer winding.
VI. ADDITIONAL ASPECTS OF OPERATION
A. DC Operation in 18-Slot 12-Pole Machine With
10% Eccentricity
These machines examples used in this paper are suitable for
brushless AC (sinewave current) control. However, analysis of
the back-EMF waveform for the 18-slot 12-pole machine with a
double-layer winding suggests it may also be possible to control
this machine effectively using dc control (trapezoidal current
waveforms). The back-EMF waveform is shown in Fig. 15.
If brushless DC control is used then the UMP is shown in
Fig. 16. This can be compared to Fig. 10(b). It can be seen
that the brushless DC control produces UMP vibration. This is
a saw-tooth waveform therefore it will consist of a harmonic
series of vibration frequencies that correspond to the 3-phase
current switching—Phase A trapezoidal current is given for il-
lustration.
B. Effect of 10% Eccentricity on the Torque of an 18-Slot
16-Pole Machine With Single-Layer Winding
Space constrains prevent a full analysis of the effect of ec-
centricity on the torque; however, it can be seen in Fig. 8 that
the 18/16 machine with double layer produces UMP vibration.
Therefore, here the change in torque will also be investigated
using the current–flux linkage (I-Psi) loops. This is an effective
way to calculate the mean torque and the argument will be re-
stricted to this. In Fig. 17, the I-Psi loops are shown when the
IEEE TRANSACTIONS ON MAGNETICS, VOL. 46, NO. 7, JULY 2010
Fig. 17. I-Psi loops for phase 1 when rotor is centered and 10% eccentric (from
PC-FEA).
rotor is eccentric and also when centered. From PC-FEA (rather
than PC-BDC as given in Table I) the centered-rotor torque is
4.3 Nm and the 10% eccentric value is 3.9 Nm. This is a 10%
decrease in torque—this needs further investigation through ex-
periment. The loops for phase 1 are given in Fig. 17 and the
other phases loops do follow the same respective orbits.
VII. CONCLUSION
This paper has addressed the issue of UMP in fractional slot
brushless permanent-magnet machines. An analytical algorithm
is first developed which illustrates how it is possible to identify
the different components in the UMP. After this a more detailed
approach is taken via the use of finite-element analysis and this
was used under many different machine arrangements.
The work here is couched in terms of ac operations with some
dc control briefly addressed (which produced more UMP vibra-
tion due to switching effects). The work illustrates that some
machine arrangements will be very susceptible to UMP when
eccentricity is present and also they may be a source of noise and
vibration. Different winding configurations are inspected and
these can produce larger values of UMP. Consequent-pole ro-
tors are also addressed and these generally produce higher UMP
compared to their surface-magnet counterpart. Some winding
arrangements generate vibrating UMP even when the rotor is
centered and an explanation is put forward to account for this.
Some of the results indicate that for certain motor designs ec-
centricity can cause, apart from UMP, a reduction in the average
output torque.
There is little in the literature that details the effects of dif-
ferent winding arrangements to the depth put forward here, and
assesses several arrangements in comparison. This is impor-
tant when specifying bearings and mechanical structure together
with assessing motor vibrations.
ACKNOWLEDGMENT
The authors are grateful for the kind co-operation of AO
Smith Corp. in the preparation of this paper.
DORRELL et al.: UNBALANCED MAGNETIC PULL DUE TO ASYMMETRY AND LOW-LEVEL STATIC ROTOR ECCENTRICITY
REFERENCES
[1] A. Burakov and A. Arkkio, “Comparison of the unbalanced magnetic
pull mitigation by the parallel paths in the stator and rotor windings,”
IEEE Trans. Magn., vol. 43, no. 12, pp. 4083–4088, Dec. 2007.
[2] D. G. Dorrell, “Experimental behaviour of unbalanced magnetic pull
in 3-phase induction motors with eccentric rotors and the relationship
with tooth saturation,” IEEE Trans. Energy Convers., vol. 14, no. 3, pp.
304–309, Sep. 1999.
[3] R. Perers, U. Lundin, and M. Leijon, “Saturation effects on unbalanced
magnetic pull in a hydroelectric generator with an eccentric rotor,”
IEEE Trans. Magn., vol. 43, no. 10, pp. 3884–3890, Oct. 2007.
[4] C. Bi, N. L. H. Aung, H. N. Phyu, Q. Jiang, and S. Lin, “Unbalanced
magnetic pull induced by drive current in PM-BLDC motor operation,”
in Int. Conf. Electrical Machines and Systems, Seoul, Korea, Oct. 8–11,
2007.
[5] C. Bi, Z. J. Liu, and T. S. Low, “Effects of unbalanced magnetic pull
in spindle motors,” IEEE Trans. Magn., vol. 33, no. 5, pp. 4080–4082,
Sep. 1997.
[6] D. G. Dorrell, “The sources and characteristics of unbalanced mag-
netic pull in cage induction motors with either static or dynamic ec-
centricity,” in IEEE Int. Symp. Electric Power Engineering, Stockholm,
Sweden, Jun. 18–22, 1995, pp. 229–234, Volume on Electrical Ma-
chines and Drives.
[7] P. Vijayraghavan and R. Krishnan, “Noise in electric machines: A re-
view,” IEEE Trans. Ind. Appl., vol. 35, no. 5, pp. 1007–1013, Sep./Oct.
1999.
[8] Z. Q. Zhu, D. Ishak, D. Howe, and J. Chen, “Unbalanced magnetic
forces in permanent-magnet brushless machines with diametrically
asymmetric phase windings,” IEEE Trans. Ind, Appl,, vol. 43, no. 6,
Nov./Dec. 2007.
[9] A. C. Smith and D. G. Dorrell, “The calculation and measurement of
unbalanced magnetic pull in cage induction motors with eccentric ro-
tors. Part 1: Analytical model,” 1996 Proc. IEE Electric Power Appli-
cations, vol. 143, no. 3, pp. 193–201.
[10] D. G. Dorrell, J. Amemiya, A. Chiba, and T. Takenaga, “Analytical
modelling of a consequent-pole bearingless permanent magnet motor,”
in IEEE Int. Conf. Power Electronics and Drive Systems, Singapore,
Nov. 17–19, 2003.
[11] T. J. E. Miller and M. I. McGilp, PC-BDC 8.0 for Windows—Software.
SPEED Laboratory, University of Glasgow, Glasgow, U.K., 2008.
[12] M. Olaru, T. J. E. Miller, and M. I. McGilp, PC-FEA 5.5 for
Windows—Software. SPEED Laboratory, University of Glasgow,
Glasgow, U.K., 2007.
[13] D. G. Dorrell, M.-F. Hsieh, and Y. Guo, “Unbalanced magnet pull in
large brushless rare-earth permanent magnet motors with rotor eccen-
tricity,” IEEE Trans. Magn., vol. 45, no. 10, pp. 4586–4589, Oct. 2009.
[14] D. G. Dorrell and A. C. Smith, “The calculation and measurement
of unbalanced magnetic pull in cage induction motors with eccentric
rotors. Part 2: Experimental investigation,” 1996 Proc. IEE Electric
Power Applications, vol. 143, no. 3, pp. 202–210.
[15] A. Arkkio, M. Antila, K. Pokki, A. Simon, and E. Lantto, “Electromag-
netic force on a whirling cage rotor,” Proc. IEE Electric Power Appl.,
vol. 147, pp. 353–360, 2000.
[16] H. M. Kelk, A. Eghbali, and H. A. Toliyat, “Modeling and analysis
of cage induction motors under rotor misalignment and air gap eccen-
tricity,” in IEEE IAS Annual Meeting, 2005, vol. 2, pp. 1324–1328.
2685
[17] B. Heller and V. Hamata, Harmonic Field Effects in Induction Ma-
chines. Amsterdam, The Netherlands: Elsevier Scientific Publishing,
1977.
[18] P. J. Tavner, L. Ran, J. Penman, and H. Sedding, Condition Moni-
toring of Rotating Electrical Machines. U.K.: The Institution of En-
gineering and Technology, 2008.
[19] D. G. Dorrell and C. Cossar, “A vibration-based condition monitoring
system for switched reluctance machine rotor eccentricity detection,”
IEEE Trans. Magn., vol. 44, no. 9, pp. 2204–2214, Sep. 2008.
David G. Dorrell (M’95–SM’08) received the B.Eng. (Hons) degree in elec-
trical and electronic engineering from The University of Leeds in 1988, the
M.Sc. degree in power electronics engineering from The University of Bradford
in 1989, and the Ph.D. degree in engineering from The University of Cambridge
in 1993.
He has held lecturing positions with The Robert Gordon University and
The University of Reading. He was a Senior Lecturer with The University of
Glasgow, U.K., for several years. In 2008 he took up a post as Senior Lecturer
with The University of Technology Sydney, Australia, and he was promoted to
Associate Professor in 2009. He is also an Adjunct Associate Professor with
The National Cheng Kung University, Taiwan. He research interests cover the
design and analysis of various electrical machines and also renewable energy
systems with over 100 technical publications to his name.
Dr. Dorrell is a Chartered Engineer in the U.K. and a Fellow of the Institution
of Engineering and Technology.
Mircea Popescu (M’98–SM’04) received the Doctor of Science in electrical
engineering from Helsinki University of Technology, Helsinki, Finland, in 2004.
He has more than 25 years experience in electrical motor design and analysis.
He worked for Research Institute for Electrical Machines, Bucharest, Romania,
Helsinki University of Technology, Finland and SPEED Laboratory, Glasgow
University, Glasgow, U.K. In 2008, he joined Motor Design Ltd., Ellesmere,
U.K. as Engineering Manager. He has published over 100 papers in conferences
and peer reviewed journals.
Dr. Popescu received the first prize, best paper award from IEEE IAS EMC
in 2002, 2006 and 2008.
Dan M. Ionel (M’91–SM’01) received the M.Eng. and Ph.D. degrees in elec-
trical engineering from the University “Politehnica” of Bucharest, Romania. His
doctoral program included a Leverhulme visiting fellowship at University of
Bath, U.K. He was a post-doctoral researcher in the SPEED Laboratory, Uni-
versity of Glasgow, U.K.
Currently, he is the Electromagnetics Engineering Fellow of the AO Smith
Corp., Milwaukee, WI. He is also an Adjunct Professor with the Department of
Electrical and Computer Engineering, Marquette University, Milwaukee. Ear-
lier in his career he was with the Research Institute for Electrical Machines
(ICPE-ME) in Bucharest, Romania and with the Invensys Brook Crompton Co.
in Huddersfield, U.K. He has published more than 50 journal and conference
papers, including a first prize winner of the IEEE Industry Application Society
Electric Machine Committee. He holds 18 U.S. patents and three European
patents, including a medal winner at the Geneva Invention Fair.