Beruflich Dokumente
Kultur Dokumente
ABSTRACT
Growing consumer and legislative pressures are increasing the demand for improved efficiency household appliances.
Since improving the efficiency of single phase motors is one
method of improving the overall efficiency of many appliances,
there is an increasing pressure for improved efficiency single
phase motors. The single-phase line-start permanent magnet
motor has been shown to be capable of very desirable steady
state performance, but is known to have undesirable starting performance. In this paper, a time-stepped finite element method
of simulating dynamic behaviour of these motors is presented.
Comparisons between simulations and measured performance
are presented and used to illustrate some of the difficulties associated with modeling these motors.
I. INTRODUCTION
As awareness of the need for improved energy efficiency increases, there is a growing pressure for motor with increased
efficiency. While increasing the efficiency of industrial motors
will provide the greatest energy saving, it is also important to
improve the efficiency of household appliances. In recognition
of this need Canadian energy efficiency regulations , [l], set
minimum standards for a range of household appliences that use
electric motors. Although other countries may not legislate this
efficiency, there is still a growing consumer demand for higher
efficiency appliances, creating in turn, a demand for high efficiency single-phase motors.
The single-phase line-start permanent magnet (LSPM) rnotor has a number of features which make it an attractive proposition for this type of application. Starting asynchronously by
means of an induction cage, it operates in steady- state as a synchronous motor. This combination provides the steady state performance of a PM motor without the need for an expensive drive
system. The work done in the mid 1980s by by Miller [2] and
Rahrnan and Osheiba [3] showed that these motors are capable
of very high efficiency, near unity power factor performance.
Perhaps more importantly, this performance may be acheived
with a motor of the same external dimensions as an induction
motor. More recent work by Knight and Williamson [4] has
shown that it may be possible to acheive as much as a 40% re-
0-7803-5589-X/99/$10.00
0 1999 IEEE
duction in losses when comparing a single phase line-start permanent magnet motor to a same size induction motor.
Desirable steady state line-start permanent magnet motor performance is, however, tempered by undesirable aspects to the
starting performance. There is usually a large ripple torque
present, due to a combination of the asynchronous interaction
of the permanent magnet and stator fields and reluctance torque
due to rotor saliency. In addition to this ripple, the permanent
magnets can create a net drag torque below half speed, with the
result that the motor just does not start. This undesirable performance is not limited to single phase line start permanent magnet
motors. Honsinger [ 5 ] ,Miller [6] and Stephens et a1 [7] have all
investigated the dynamic performance of three-phase line start
permanent magnet machines. Dynamic simulation of singlephase LSPM motors has received less attention. The interaction of fields in a single phase motor and introduction of stator
backwards rotating fields makes starting a single phase motor
significantly more complicated than starting a three phase motor. Consequently starting behaviour is more difficult to predict.
Despite these difficulties Carlson et al., [8], have produced dynamic simulation results for this type of motor using a limited
finite element model.
In recent years, the computation speed ofaffordable computers has rapidly increased. It is continually becoming possible to
carry out simulations that were not viable only two or three years
previously. In this paper the authors apply a non-linear timestepped eddy current finite element formulation in order to simulate dynamic performance. Simulation times for this modeling
technique are found to be acceptable when used on 350MHz
Pentium PC. Use of this model allows effects such as current
redistribution, saturation and demagnetization to be accounted
for during starting. A number of measured starting curves are
presented, and by comparison between measured, expected and
simulated performance, the difficulties associated with accurate
dynamic predictions are explored.
11. MOTORDESIGN
Prior to describing the simulation technique used it is important to examine the motor which we are trying to simulate.
Figure 1 shows a design for a single phase line-start perma-
2582
111. M O D E L I N G
TECHNIQUE
The eddy current method used in this paper is based on that
described by Arkkio [9]. In this technique, the non-linear field
equations are solved simultaneously with the rotor and stator circuit equations, allowing current redistribution in the rotor bars
to be properly accounted for. The equations for a single phase
LSPM motor may be written as follows:
\ ,
(3)
The clear adaptations to the standard eddy current technique
required for this type of motor are the inclusion of magnet terms
in equation 1 and the capacitor related term in equation 2.
Rotation is taken into account by using the air gap element
as described with negative perioidicity by Flack and Volschenk
[lo]. Use of this method allows accurate torque calculations to
be made without the need to mesh the air gap. This method
does have the drawback that a relatively dense block is created
in the matrix equations, increasing the matrix simulation time.
However,this increase is offset to some extent by the removal of
any need to remesh the airgap or renumber the matrix equations.
A. Demagnetisation
vs
Lm
La
The excitation terms due to the magnets and the magnet reluctivity terms are directional, with their values dependant not
only on the residual induction, B, but also the direction of magnetisation. If magnet demagnetisation is not taken into account,
the magnet excitation terms and reluctivity can be determined in
advance, and do not need to be updated during the course of the
time-stepped solution. Unfortunately, demagnetisation is known
to be a problem with some designs of LSPM motor and should
be included in the finite element model. Dealing with the possibility of demagnetisation during each iteration of a non-linear
finite element solution would be quite daunting, requiring a nonlinear magnetisation curve for the permanent magnet material.
However, it is possible to take demagnetisation into account at
the end of each time-step, outside the non-linear solution. This
approximation to the demagnetisation process may be described
using figure 3 and considering a single magnet element of the
finite element mesh.
1. At the end of each time-step, the flux density in the direction
of magnetisation is calculated.
2583
dA
13
(4)
The elements forming the deep part of the bar are not, however,
included in the rotor cage terms in equations 1 and 3.
Before carrying out dynamic simulations using the method
described above, it is important to know the mechanical parameters of the motor and the conditions under which testing was
carried out.
ir
IV. MECHANICAL
CONDITIONS A N D TEST RIG
dB
Jmech
(5)
where
w, =
Tern =
Fr =
2. If the flux density in the direction of magnetisation is positive, demagnetisation will not have occured and the remanent
flux density for that element is unchanged, at B,.
3. If the flux density in the direction of magnetisation is negative and decreasing, it is allowed to continue to decrease along
the existing recoil line defined by B,.
4. If the flux density is negative and starts to increase, the remanent flux density for that element is reduced by an amout equal
to the minimum flux density, d B and recoil occurs along the line
defined by a remanent flux density of B, - d B .
5. The permanent magnet excitation terms for that element are
recalculated using the reduced value of B, .
The method described above is clearly an approximation to
the demagnetisation process. However, in the absence of detailed non-linear material properties for the magnets, it is felt
that this method gives a good indication of the extent of any
demagnetisation during starting.
.
B. Deep Bars
As stated earljer, the deep bars in the motor are only connected to the end rings at the top of bar. This creates another
difficulty which must be overcome before it is possible to model
the motor accurately. Those elements which are positioned so
that their radius is greater than the inside radius of the end ring
are treated in a similar fashion to elements making up other,
standard, bars. The remaining elements cannot be treated as
part of the normal bar arrangement.When we consider equation
3, the applied bar voltages and resulting currents are in part defined by an end ring connection matix. Since the deeper section
of the deep bar is not connected to the end ring, the equations
for the applied voltage are invalid, as are equations for current
flowing out of the bar into the end ring. In order to account for
some induced current in the deep-part of the bar, those elements
forming the deep deep part of the bar are included in the eddy
2584
UnloadedSkewed Motor
4000.0
-E
4000.0
3000.0
2000.0
a
1000.0
0.0
1000.0
00
1 /
4000.0 1
3000.0
2000.0
00
20
10
Measured
1000.0
30
Time (s)
0.0
1 .o
0.0
2.0
3.0
Time (s)
V. SIMULATION
RESULTS
With the mechanical data for the machine obtained as described above, simulations of dynamic behaviour for the test
unskewed motor were carried out.
The first simulation presented is a prediction of no-load
starting performance. A comparison between measured speed
against time is presented in figure 5. The simulated starting
performance is shown in the top graph, with measured performance in the second graph. At first sight, the simulated performance bears little resemblance to the measured performance.
There is considerable speed ripple predicted throughout starting
and the simulation predicts synchronous speed will be reached
faster than was observed with measurement. Before considering performance under other conditions it is important that we
understand any reasons for the apparent discrepancies found so
far.
The first point to consider is the lack of ripple on the measured starting curve. As stated earlier, this is not what one
would normally expect from observing the behaviour of the motor during starting and may be accounted for by a smoothing
effect due to the frequency to voltage converter. If the simulated starting curve is averaged out, representing the average
al
1000.0
''
0.0
0.0
Simulation:Smoothed]
1 .o
2.0
3.0
Time (s)
2585
B,,,,
+0 +a)
+ Bs,cos((l - 2sw,) + 0 + p)
+ Bs,cos(-(2 - s)w, + 0-+ y)
= islcos(sw,
(6)
(7)
(8)
5.0
3000.0
0.5 Nm Load
2000.0
0.0
1000.0
-5.0
I-
1.-/I--'
4000.0
-10.0
L . . , . , . . . . ' ' . . . '.''' ' " " ~ ' ' " ~ " ' ' ' '
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
'.'.."''...'
0.8
0.9
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0
4000.0
1.0
3000.0
0.1 Nm Load
2000.0
1000.0
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0
15
10
05
4000.0
3000.0
00
2ooo'o
-05
1000.0
-10
00
01
02
03
04
05 06
Time (5)
07
08
09
'"'/-----i
1
1
- 0.15 Nm load
0.0 tA-Y--
'
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0
Time (s)
10
30000 -
(9)
Unloaded]
where:
is the stator forwards rotating field
BSI
is
the stator backwards rotating field
Asz
due to saliency
6 s g is the usual single phase stator backwards
rotating field
Inspection of equation 9 shows that both backwards rotating
fields rotate at the same frequency when s = 0.6611. This corresponds to a rotor speed of 1220 rpm. At this slip, the interaction of the backwards rotating fields creates additional drag
torque components, resulting in the predicted reduction in available torque around one third speed.
The simulations so far are encouraging, offering results in line
wth expectations, and within an acceptable time frame. Running
the simulations on a PC, modeling 2 seconds of real time takes
approximately 12 hours. While this is still quite long for everyday design work, having information overnight on whether or
not a design is viable is certainly preferable to building numerous prototypes.
The time averaged torque plot indicates that the available
torque around one third speed is only slightly more than 0.1 Nm.
If the simulation is valid, the motor is likely to have difficulty
starting loads greater than 0.1Nm
The plots in figure 7 show that the motor can start loads of up
to 0.1 Nm with relatively little difficulty. However, as expected,
when the load is increased to 0.15 Nm the motor has difficulty
starting.
The simulation results presented in figure 8, like the results
from the measured performance, indicate that the motor can start
a 0.1 Nm load. However, the simulations indicate a greater difficulty in accelerating past 1200rpm than is observed in the measured results.
A. Adjustments
After comparing the measured performance with the simulations so far, three points begin to become clear:
00
02
04
06
08
10
'
12
'
'
14
16
18
'
1
20
4000.0
3000.0
2000.0
1000.0
n
-."n
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
-I
4000.0
3000.0
2000.0
1000.0
0.0
0.0
0.2
0.4
0.6
0.8
%-dAL-LJ
1.0 1.2
Time (s)
1.4
1.6
1.8
2.0
25 86
I
,
1
4000.0
-E
4000.0
3000.0
2000.0
3000 0
Unloaded
U)
1000.0
0.0
0.0
4000.0
0.1
0.2
0.3
0.4 0.5
0.6 0.7
0.8 0.9
20000
1.0
(0
Simulation Load=O 1 Nm
10000
00
00
0.0
0.1
0.2
0.8 0.9
05
10
15
20
25
Time (s)
1.0
'
'
"
'
'
5.0
2
P
'r
a,
z$
s
!!
0.0
-5.0
20.0
0.0
-20.0
10.0
5.0
0.0
-5.0
-10.0
0.0
0.2
0.4
0.6
0.8
1.0
Time (s)
1.2
1.4
1.6
1.15 s. This is evidenced by the sharp decrease in auxilliary current at this time. Note, however, that the magnitude of the main
winding current is not significantly decreased until the motor is
almost synchronised. At the same time, the positive peak of the
torque ripple is reduced, while magnitude of the negative peak
increases slightly.
The change in predicted starting time shown in figures 9 and
10 indicates the performance of line-start permanent magnet
motors can be very sensitive to mechanical parameters. A sixty
percent increase in inertia results in a predicted increase in starting time of almost a factor of 4. It is clearly very important to
have accurate mechanical data to produce reliable simulations.
However,with the above adjustments made to the input param-
2587
REFERENCES
Natural Resources Canada, Energy efficiency regulations, Canada
Gazette, Partll, vol. 128, no. 22, 1994.
T.J.E. Miller, Single-phase permanent-magnet motor analysis, IEEE
Transactions on Industry Applications, vol. IA-21, no. 4, pp. 651-658,
May 1985.
M.A. Rahman and A.M. Osheiba, Performance analysis of single-phase
permanent magnet motors., Electric Energy Conference 1987.Adelaide.,
October 1987.
A.M. Knight and S. Williamson, Influence of magnet dimensions on the
performance of a single-phase line-start permanent magnet motor, Proceedings of the IEEE International Elecric Machines and Drives Coqference, Seattle, pp. 710-772,1999.
V.B. Honsinger, Permanent magnet machines: Asynchronous operation,
IEEE Transactions on Power Apparatus and Systems, vol. PAS-99, no. 4,
pp. 1503-1509, July 1980.
T.J.E. Miller, Synchronization of line-start permanent magnet ac motors,
IEEE Transactionson Power Apparatusand Systeins, vol. PAS-103, no. 7,
pp. 1822-1828, July 1984.
C.M. Stephens, G.B. Kliman, and J. Boyd, A line-start permanent magnet
motor with gentle starting behaviour, IEEE Industry Applications Society
Annual Conferece. St Louis., vol. 1, pp. 311-379,1998.
R. Carlson, N.Sadowski, S.R. Arruda, C.A. da Silva, and L. Von Donkal,
Single-phase line-started permanent magnet motor analysis using finite
element method, IEEE Industrial Applicarions Society Annual Conference. Denver, vol. 1, pp. 227-233,October 1994.
A. Arkkio, Finite element analysis of cage induction motors fed by static
frequency converters, IEEE Transactions on Magnetics, vol. 26, no. 2,
pp. 551-554,1990.
[IO] T.J. Flack and A.F. Volschenk, Computational aspects of time-stepping
finite-elememt analysis using an air gap element, ICEM, pp. 158-163,
1994.
[ I I] A.M. Knight and J.C. Salmon, A comparison between finite element
techniques when modelling single phase line-start permanent magent motors, To be presented at the 9th international conference on electrical
nuchines arid drives, (EMD 99). Canterbury, September 1999.
2588
Session 59
INDUSTRIAL DRIVES