Você está na página 1de 9

Analytical modelling of variable-reluctance

machine magnetisation characteristics

D.A. Torrey
X.-M. Niu
E.J. Unkauf

Indexing tenns: Magnetisation, Modelling

f3 = factor for third dimension correction


A h c t : The paper presents models that c1,c j = contants
c o w the fundamental design of the variable- error = percentage error between model and flux tube
reluctance machine (VRM) to an analytical analysis
expression which summarises the terminal B = flux density, T
magnetisation characteristics. The modelling pro- 1 = phase flux linkage, Wb
cedure takes simple piece-wise linear models 8 = mechanical rotor position, rad
which are based on the geometry and turns per a = number of electrical cycles per revolution
phase of the VRM and maps them into an analy- = position where rotor and stator reach maximum
tic terminal expression through a simple and fast overlap, rad
algorithm. The piece-wise linear and analytical el = a position between O1 and 02,rad
expressions are based on physical reasoning. The O2 =position where rotor and stator reach zero
modelling of three existing VRMs is presented, overlap, rad
and the model predictions are compared with
experiment. The simple models are shown to do Subscripts
an effective job of capturing the spatial and mag- a = aligned position
netic nonlinearities of the machine, thereby pro- U = unaligned position
viding useful estimates of net and instantaneous max = maximum value
machine performance. The analytic model is corner = value at O2
shown to compare favourably with another docu- min = minimum value
mented approach for the rapid description of s =stator pole
VRM magnetisation characteristics. r =rotor pole
sfi =shaft
rbi = rotor back iron
List of symbols rph = rotor pole height
= p h m current, A rpb = rotor pole base
= magnetisation coefficient, Wb sbi = stator back iron
= magnetisation coettlcient, A - o =outside
= magnetisation coefficient, H A = corresponding to point A, Fig. 1
p" = incremental inductance, H B = correspondingto point B, Fig. 1
L = linear p h e inductance, H
1 Introduction
= k e m e n t d inductance at high current, H
= number of turns per phase The variable-reluctance machine (VRM) is a doubly
= number of rotor poles salient synchronous machine which finds applications as
= effective airgap area, m2 aerospace motors and generators, in wind energy systems
= radial airgap length, m and in other applications, ranging from fractional horse-
d = effective axial length, m power up to several hundred kilowatts. The design of the
R = radius, m machine is complicated because of its strong spatial and
9 = normalised permeance, pH/m magnetic nonlinearities, combined with a large number of
kl =ratio of rotor interpolar arc length to stator
pole arc length
k, = ratio of rotor pole undercut to rotor interpolar This work has been supported in part by the
arc length United States National Science Foundation
through grant ECS-9209123 and through the
0 IEE, 1995 Niagara Mohawk Power Electronics Research
Paper 1567B (Pl), first received 11th April and in revised form 16th Chair. The authors gratefully acknowledge the
August 1994
constructive criticisms of the reviewers, which
D.A. Torrey and EJ. Unkauf are with the Department of Electric
Power Engineering, Renssclaer Polytechnic Institute, Troy, NY 12180- have helped to strengthen this paper significantly.
3594 USA Prof. T.J.E. Miller is gratefully acknowledged for
X . M . Niu is with the Department of Electrical Engineering, Southwest providing the construction and operational data
Jiaotong University, Chcngdy Sichuan 610031, People's Republic of for the OULTON 8/6 VRM quoted here.
China

14 I E E Proc.-Electr. Power Appl., Vol. 142, No. I , January 1995


degrees of freedom. Some important works on the VRM magnetisation data from the geometry and turns per
are reb- 1 and,2 awl the r e f e m therein. phase.
This paper is concereed with connecting the funda-
mental geometry and turns per phase of the VRM with 2.1 Analytical terminal model
an analytic terminal description. The terminal description The analytical equation which is used to summarise the
can then be used to predict the performance of the VRM terminal magnetisation characteristicsis
machine. The motivation for pursuing such a model is in
the physical insight it provides into the design and oper- n(i, 0) = a i ( W - exp Caz(elill+ a,(@ (1)
ation of the machine. Furthermore, it adds to the dis- where 1 is the phase flux linkage, i is the phase current
cussion of VRM models to motivate models which are and 0 is the mechanical rotor position. The spatially
still more insightful and accurate. periodic structure of the machine allows one to express
The work presented here connects the physical struc- the magnetisation coefficients al& as a truncated
ture of the VRM to its terminal characteristics, thereby Fourier cosine series
supporting the rapid and accurate design of the machine. "
The inclusion of optimal control and nonlinear system
theory in the VRM design process could lead to a deter-
a, = 1A,,
k=O
cos a0 (2)
ministic design process, rather than the traditional iter- where a represents the number of electrical cycles in each
ative process. This is an area of active research. mechanical revolution. The Fourier coefficients are easily
This paper describes the VRM differently than the determined through the Marquardt gradientexpansion
work documented by Reference 3, wherein the authors algorithm for a,-,(e) [6]. The series represeatation of the
develop an interesting approach to the modelling and magnetisation coefficients is a natural one, but it intro-
design of the VRM. Their goal is the rapid computer- duces some ripple in the static torque characteristic
aided design of the machine. Instead of representing the because of the higher-order harmonics, especially when a
magnetisation data of the machine by flux linkage as a is large. Eqns. 1 and 2 summarise the magnetisation char-
function of current, parameterised by rotor position, they acteristics over all rotor positions and currents. Further,
consider flux linkage to be a function of rotor position, they provide a closed-form expression for the instant-
parameterised by current. This brings out the spatial aneous phase torques.
structure of the machine, especially the motional voltage. Eqn. 1 was suggested in Reference 7 as a simpler
Based on the structure of the magnetisation data, the version of the expression used in Reference 8, and applied
authors develop three magnetisation and torque models, to a 60 kW VRM in Reference 9. There, it was shown to
each valid over a limited range of rotor positions. The do an excellent job of predicting the instantaneous and
accurate modelling results reported by the authors reflect average phase currents and torque of the VRM. The
some degree of empiricismin the model form. 60 kW VRM described in Reference 9 is one of the three
Another paper on the computer-aided design of the VRMs used in Section 4 to evaluate the models devel-
VRM [4] describes yet another approach to the model- oped in this paper.
ling. In Reference 4 the authors indicate that they base In eqn. 1 al(0) can be thought of as the flux linkage
their performance analysis on a magnetic model which is when the phase makes the transition from linear to satur-
carefully constructed for the extreme rotor positions of ated operation. The coefficient a,@) controls the curva-
maximum and minimum alignment between the stator ture of the magnetisation characteristic. The coefficient
and rotor poles. At all other rotor positions, the magnet- a,(@ is the incremental phase inductance at high cur-
isation data are interpolated using a biased cosinusoidal rents. The connection between the magnetisation coeffi-
variation with rotor position. This approach is con- cients and the magnetisation curve for 0 = 0, (rotor and
sidered adequate when only net energy conversion and stator poles in maximum alignment) is shown schematie
bounds on torque ripple are of interest. The accurate ally in Fig. 1; additional comments concerning the mod-
models at the extreme rotor positions are determined
using flux tube analysis in the spirit of Corda and Step-
henson [5]. The work presented here also takes advant- 4
age of the important work of corda and Stephenson,
though to a lesser degree.
The approach taken here is motivated by two con-
siderations. The first is to gain insight into the fundamen-
tal behaviour of the VRM. The second is to take the
basic design parameters for the machine and map them
into an analytic equation that summarises the terminal
characteristics of the machine. These goals are accom-
plished through the use of
(a) conceptually simple piece-wise linear models to Fig. 1 Magnetisation curw for rotor and stator alignment, and its
describe the spatial nonlinearities of the machine relationship to the magnetisationco&cients
(b) a description of the E H data for the magnetic
material elling of the 0 = 0. magnetisation characteristic are made
(c) a physically motivated analytic summary equation. in the next Subsection. Based on the physical significance
which can be attributed to the magnetisation coefficients,
it is reasonable to expect that the VRM geometry, B-H
2 VRMmodels data and number of turns per phase dictate the shape of
al-,(0). Indeed, this is shown below to be the case.
This Section presents the model used to summarise ana- The determination of a,-,(@) from the. basic design
lytically the terminal behaviour of the VRM. This motiv- relies heavily the incremental inductance of the general
ates the simple models used to construct the terminal VRM phase in the aligned position. The incremental
IEE Pm.-Electr. Power Appl., Vol. 142, No. I , January 1995 15
inductance is given by The magnetisation coefient a,(@) is not easily
described through a piece-wise linear model becaw a,(!)
aqi
ti, e) = -
e)
= -al(e)a,(e) exp [a2(e)il + a3(e) (3) is used to i n e o p r a t e the nonlinearity of the magnetic
ai
where the symbol I is used to denote the incremental
inductance to avoid confusion with the conventional
interpretations of inductance. In the limit as i + O the
incremental inductance is reduced to the linear phase
inductance 40)
49)= -a1(8)a,(0) + a,(@ (4)
In the high current h i t i > -4/a,(e) the inmemental
inductance is reduced to reveal the physical significance
of a3
1, = a,(@ (5)
The next Subsection motivates the form of the simple I 1 I I
piece-wise linear models that are used to relate the mag- lj lj lj
netisation &mts
ture of the VRM.
to the geometry and phase struc- n n stator
rotor
nu
Fig. 3 Piecewise-linear model assumed for the unsaturated phose
inductance as afunction $rotor position
2 2 Relationship between VRM geometry and
magnetisation coefficients
The geometric structure assumed for the general VRM is
shown in Fig. 2 tor a three-phase VRM with six stator

nU Un stator U
rotor
nu
Fig. 4 Piecewise-linear model assumed for the magnetisation CO@-
cient a , as a function of rotor position

F(l. 2 Assumed geomhic structurefor a 614 VRM

poles and four rotor poles, a conventional 6/4 VRM. In


Fig. 2 the numbers refer to the different segments of the
machine to which a flux-tube analysis is applied. The
rotor position is assumed to be 0, for the general phase
when two rotor poles are aligned with the excited stator
poles. The rotor position is assumed to be 0, for the
general phase when the rotor interpolar slot is aligned
with the excited stator poles.
Based on the geometry shown in Fig. 2 and the phys-
ical significance associated with each of the magne-
tisation codficents of eqn. 1, it is reasonable to expect I I I I
I I I 1
that a, will have a maximum at 0 = 0, and a minimum at
0 = e,. In addition it should be expected that a3 will
have a minimum at 0 = 0, and a maximum at 0 = 0;. the nU ljn stator
rotor
lj lj
n u
inmmental inductance at high currents is quite small Fig. 5 Piecewise-linear model assumed for the magnetisation
when t& rotor and stator poles are in alignment, and it cient a j as a f i t i o n of rotor position
grows as the severity of saturation is eased by moving the
rotor toward e,. Further, the unsaturated phase induc- steel which forms the magnetic circuit. The coefficient a,
tance 4 0 ) should exhibit a spatial dependence similar to can, however, be. described through I@), a@) and a,(@)
that of al. The piecewise linear models assumed for 4 0 ) , by rearranging eqn. 4
ai(@)and a@) are shown in Figs. 3-5, respectively. The
rotor and stator pole orientations are also shown in Figs.
3-5 for reference.
16 IEE Proc.-Ekctr. Power Appl., Vol. 142, No. I , January 1995
The next Section provides the specific details of how the where R, and R, are the radii at the airgap for the rotor
values for the magnetisation coefficients are determined. and stator, respectively. For the purposes of the model
presented here, eqn. 12 provides more consistent results
than the Carter coefficient used in Reference 5.
3 Modelling process The inductances LC,,,., and L,,,h are found through
look-up tables according to
The modelling process involves determining each value
referred to in Figs. 3-5. Of the eleven parameters which L,", =W@C,,", df3 (13)
form the VRM model (L-, Lm,,,, Lmh,a,(O,), al(e,,), = fN2@m, dh (14)
a,(@,), a,(&), e,, P,, e2 and e,,), seven are based on simple
Lmin

calculations, two are based on values from tables, and where I,,",is the normalised airgap permeance in the
two are found iteratively. This Section outlines how each comer position, is the normalised minimum airgap
model parameter is determined. permeance, and f3 is a factor which takes third dimension
effects into account. The normalised airgap permeances
are derived from finite-element analyses, as described in
3.1 Angular breakpoints 0,' e;, e2and 8, Reference 10. The data for and 8,, are given in
The angles 0,. e2 and 0, are found directly from the Tables 1 and 2, where k , is the ratio of rotor interpolar
VRM design parameters. The angle 0, denotes the rotor
position where the rotor pole is about to begin pulling
Table 1 : Normalised permeance data for the corner p a i t i o n
out from under the stator pole. ea a function of k. and k.. .@.-- is aiven in units of B H h

k2 k,
(7)
1.05 1.20 1.50 2.00 250
where 0, and 0, are the rotor and stator pole arcs, 0.125 19.15 16.66 14.55 12.80 11.77
respectively. In eqn. 7,it is assumed that 0,2 e,, which is 0.1875 16.57 14.16 12.68 11.19 10.49
0.25 15.22 12.89 11.72 10.46 9.83
usual practice in VRM design. The angle 0, denotes the 0.375 14.11 11.98 10.85 9.91 9.53
rotor position where the rotor pole has completely 0.5 13.68 11.61 10.77 9.57 9.18
uncovered the stator pole. 0.75 13.42 11.27 10.38 9.42 8.89
1.0 13.39 11.21 10.19 9.41 8.88
e, =2e, + 0 s
Table 2: Normalised minimum airgap permeance ~ la lfunc-
The angle 0, denotes the unaligned position; this is where tion of k , and k,. @& is given in units of pH/m
the rotor interpolar slot is aligned with the stator pole. k2 k,
It 1.05 1.20 1.5 2.0 2.5
e =- (9)
" N, 0.125 17.09 12.98 10.22 6.14 7.00
0.1875 14.48 10.40 7.99 6.29 5.4.4
where N , is the number of rotor poles. 0.25 13.22 9.27 6.93 5.42 4.68
The angle (7, shown in Figs. 4 and 5 has been observed 0.375 12.23 8.30 6.03 4.65 4.01
to be important in reducing the modelling error over 0.50 11.86 7.91 5.69 4.36 3.75
0.75 11.56 7.70 5.48 4.18 3.60
rotor positions between approximately 0, and 1.00 11.54 7.62 5.43 4.14 3.57
+
(e, 0,)/2. The angle & is taken as
arc length to stator pole arc length and k, is the ratio of
the rotor pole undercut to the rotor interpolar arc length.
In terms of the measured inductance profile for a VRM, Fig. 6 serves to define k , and k, in terms of VRM
Plappears to correlate closely with the angle at which geometry.
the iron permeance has only a second-order impact on
the linear inductance profile.

32 Inductance values L,,,, Lcorns,andLmin


The maximum inductance value assmated with the
piece-wise linear model for 4 e ) is given by

where N is the number of turns per phase, A, is the effect-


ive airgap area, G is the radial airgap length and d is the
effective active axial length of the VRM. The effective
airgap area takes fringing into account by adding the
radial airgap to each end of the VRM. In addition, fring-
ing flux at the sides of the teeth is considered by averag-
ing the stator and rotor arc lengths, which are generally Fig. 6 VRM magnetic circuit with k , = I5 and k, = 05
unequal. That is,
Third-dimension effects are taken into account
through a factor which depends on the aspect ratio of the
VRM. For all of the VRMs considered in Section 4, the
I E E Roe.-Ele~tr.Pow- Appl, Val. 142,No. I , January 1995 17
correction factor is taken as error is given by

f3 = (1.06 + 1.16 exp (- 1.44 ;)y error,=,. ,= al[l - exp (a2 iJ]
NE, A,
+ a3i, -1 (19)

As suggested in eqn. 15, third-dimension effects are less where it is understood that i, is the phase current which
sigmiicant for long VRMs which have a small radius corresponds to the stator pole flux density E,.
than for short VRMs with a large radius. The constants (f)Adjust B. based on error,. If error, > 0, increase
in eqn. 15 were chosen from experience with existing E,. Otherwise, decrease E,.
VRMs, and include a nominal correction even as d/R, +
(8) Repeat Steps (b)-(f) until I error, I 1 error, I has
tends to infinity. been mitumised.
The value of a3(0,,) is taken to be
3 3 The magnetisation coefficient values a, (ea), a3(U = c3 L,," (20)
a,(@,).a,(@,) and a,(@,) where c3 is a constant, nominally chosen as c3 = 0.95.
The magnetisation coefficient values al(Oa) and a,(@,,) are
The value of al(e,) is taken to be
found iteratively, as described below. The values for
al(O,,) and a3(0,) are then taken as constant fractions of ado") = clal(eJ (21)
al(O,,) and L,, respectively.
The value of a,@#) represents an average incremental where c1 is a constant, nominally chosen as c1 = 0.1. The
inductance. In Fig. 1 the incremental inductances at values of c1 and c3 have been chosen to minimise the
points A and B are significantly different. The value of modelling error.
a,(&) is selected to minimise the error between the ana- With the piece-wise linear models of Figs. 3-5, the
lytic model and flux tube analysis at points A and B,and angular breakpoints given by eqns. 7-9, the inductances
therefore represents an average of aA/aiI, and aA/ai(,. of eqns. 11-14, the values of al(OJ and a3(OJ given by the
The point A is selected to correspond to a flux density at iterative procedure, and the values of a3(0,,) and al(O,,)
which there is significant curvature in the magnetisation given by eqns. 20 and 21, a complete model of the VRM
characteristic; the point B is selected to correspond to a has been developed which connects the terminal magnet-
flux density which represents significant saturation. isation characteristics to the underlying geometry and
The determination of a,(O,) and al(OJ is performed number of turns per phase. The next Section demon-
iteratively using flux tube analysis in the spirit of Refer- strates the accuracy of this analytic model for three
ence 5 as the metric. The flux tube analysis is based on VRMs of widely varying size and structure.
imposing a flux density E, in the stator pole, determining
the corresponding flux density in each segment of the 4 Experimental verification
VRM, determining the corresponding reluctance in each
segment, then solving for the ampere-turns which are The VRM models of Section 2 were applied to three
VRMs of different sizes and structures using the pro-
consistent with E,. The VRM segmentation for flux tube cedure outlined in Section 3. These VRMs include a
analysis is shown in Fig. 2.
In computing the incremental inductance, the incre- 60 kW 6/4 VRM which is documented in Reference 9, a
mental reluctance of each section of the VRM is based on 7.5 kW VRM which is documented in References 3 and
11, and a 1.1 kW 6/4 VRM which is documented in
the incremental permeability,which is defined as
Reference 12. These VRMs are chosen for the experimen-
tal verification of the analytic model because of their
varying size and previous documentation. This Section
addresses both static and dynamic experimental verifica-
where E, is the flux density in the place where the incre- tion.
mental permeability is found. In computing phase cur-
rents the effective reluctance of each section of the VRM 4.1 Verification of magnetisation characteristics
is based on the effective permeability, which is defined as The geometry of the 60 kW 6/4 VRM is described in
Table 3. Table 4 summarises the model parameters of this
VRM, as described by the analytical model of Section 2.
Fig. 7 shows the flux linkage as a function of current and
rotor position trajectories described by the data of Table
The iterative procedure for finding al(O,) and as(&)con- 4. Experimentally measured flux linkage information is
sists of the following steps. also shown in Fig. 7 as a basis of comparison. Fig. 7
(a) Initialise E, to be the flux density that corresponds shows that although the model of Section 2 does not
to point B, Fig. 1 exactly reproduce the measured data, the fit is quite
(b) Compute a3 using flux tube analysis based on good. The next Subsection examines the impact of the
incremental permeabilities modelling error on dynamic performance predictions.
(c) Determine al using flux tube analysis based on The errors present in the magnetisation characteristics
effective permeabilitiesand of Fig. 7 raise two issues. The first issue is the fundamen-
al(OJ = NE,A, - a3(OJi tal ability of eqn. 1 to represent the VRM magnetisation
(18) data. Given the spatial and magnetic nonlinearities
where the current i corresponds to the phase current associated with the VRM and the number of parameters
required to support E , , and is obtained from the flux which govern its design, it is unreasonable to expect that
tube analysis. eqn. 1 should fit the data at all rotor positions and cur-
(d)Compute a,(@,,) using eqn. 6. rents with only three degrees of freedom. This becomes
(e) Determine the error between the model and flux something of a philosophical question with regard to the
tube analysis at points A and B, Fig. 1. At each point the trade-offs between model simplicity and its associated
18 IEE Proc.-Ekctr. Power Appl., Vol. 142, No. I, January 1995
Table 3: Geometry and turn information for the 60 k W 614 As an indication of the fundamental ability of eqn. 1 to
VRM. All lengths are in metres; all angles are in degrees represent the VRM magnetisation characteristics, Figs.
ParametEI Svmbol Value 7-9 indicate the flux linkage derived from flux tube
Shaft radius 4, 0.0143
Rotor back iron radius R,,i 0.034 1.2r 1
Radius behind rotor pole R,,,, 0.041
Width of rotor pole base w,~,, 0.032
Rotor radius at airgap R, 0.055
Radial airgap G 0.0002
Stator radius at airgap R, 0.0552
Radius at top of stator pole 0.082
Outside radius Ro 0.1
Active length d 0.3
Rotor pole arc 0, 45
Stator pole arc 0, 32
Number of rotor poles N, 4
Turns per pole N 13
Magnetic steel M19
Point A. Fig. 1 B. 1.65 T
Point B. Fig. 1 B, 2.025T

Table 4: Model Darameten for the 60 k W 6/4 V R M phase current, A


~ ~~

Paramer Value Fig. 8 Measured and modelled magnetisation characteristics for the
7 5 kW 816 VRM. The measured data are given by the solid lines. Points
e, 6.5" A and B were deriuedfiom theflux tube analysis
e; 12.83"
e, 38.5"
0: 45.0"
k, 1.40
k2 0.486
3' 1.125
Lm,, 23.62 mH
L,,,"., 1.19 mH
Lmi" 0.67 mH
8, (e,) 439.6 mWb
a,&) 43.96 mWb
a,(@,) 117.7 JJH
a,(e,) 636.5 JJH
Iterations 2

insight, and a model with greater accuracy but less phys-


ical insight; there is certainly value in both approaches.
The second issue involves the selection of control
phase current, A
points A and B in Fig. 1. These control points dictate, in
Fig. 9 Measured and modelled magnetisation characteristics for the
1.1 kW 614 VRM. The measured data are giuen by the solid lines. Point
0 A was deriuedfiom theflux tube analysis
10
20 analysis to drive minimisation of the modelling error.
30 Only the lower points are given in Figs. 7 and 9, because
the inclusion of the higher point would go well beyond
40 the measured data. To the extent that the analytic model
passes through the points generated by the flux tube
analysis, eqn. 1 does a good job of fitting the magnet-
isation characteristics of the VRM. Significant error
between eqn. 1 and the flux tube points suggests that eqn.
1 is limited in its ability to model the VRM magne-
tisation characteristics. Inspection of Figs. 7-9 indicates
that eqn. 1 actually does a very good job of modelling the
magnetisation characteristics, and that the majority of
the modelling error is actually introduced through the
phase current, A flux tube analysis.
Fig. 7 Measured and modelled magnetisation characteristics for the The geometry of the 7.5 kW 8/6 VRM is described in
60 kW 614 VRM. The meanaed data are given by the solid lines. Point A Table 5. Table 6 summarises the model parameters of this
was d e r i d f i o m the@ tube analysis VRM as described by the anlytic model of Section 2. Fig.
8 shows the flux linkage as a function of current and
essence, the range of currents over which eqn. 1 faithfully rotor position trajectories described by the data of Table
represents the magnetisation data. Since most VRMs are 6. Experimentally measured flux linkage information is
intended to operate well into saturation, all of the VRMs also shown in Fig. 8 as a basis of comparison. As with
considered here are modelled with the same control Fig. 7, Fig. 8 shows that although the model of Section 2
points, with emphasis on the saturated region of the mag- does not exactly reproduce the measured data, the fit is
netisation characteristics. A different selection of control quite good. The next Subsection examines the impact of
points would place the emphasis elsewhere. the modelling error on dynamic performance predictions.
IEE Proc-Ekctr. Power Appl., Vol. 142, No. I , January I995 19
Table 5 : Geometry and turn information f o r the 7.6 kW 8/6 4 2 Dynamic performance accuracy
VRM. All lengthsare in metres; all angles are in degrees As discussed in the preceding Subsection, the simple
paanaet Svmbol Value I analytic model of Section 2 does a good job of modelling
Shaft radius R,, 0.0183
Table 8: Model Darameters f o r t h e 1.1 kW 6/4 VRM
Rotor back iron radius R,,, 0.0364
Rotor radius at airgap R, 0.0582 Parameter Value
Radial airgap G 0.0003
Stator radius at airgap R, 0.0585 e, 0"
Radius at top of stator pole R,, 0.0879 ffl 13.33"
Outside radius R, 0.102 62 40.0"
A c t i i length d 0.17 e. 45.0"
Rotor pole arc 6, 23.6 k, 1.39
Stator pole arc e, 20.1 k2 0.40
Number of rotor polas N, 6 f3 1.243
Turns per pole N 70 LW. 71.28 mH
Magnetic stwl 3% Si Lcorner 5.85 mH
Point k Fig. 1 BA 1.65 T Lon," 3.42 mH
Point B. Fia. 1 B. 2.025 T Ji(ea) 419.2 mWb
a, (e,) 41.92 mWb
a,(e,) 955.6 PH
a m 3.249 mH
Table 6: The model Darametenr f o r the 7.5 kW 8/6 VRM Iterations 2
Parameter Value
01 1.75" the magnetisation characteristics of the VRM. However,
e; 7.28"
some errors are present, and it is important to under-
02 21.85"
0" 30.0" stand how these errors impact the prediction of
kl 1.80 instantaneous and average quantities. This Subsection
k2 0.59 addresses these issues through comparison of simulated
3' 1.161 and measured data for the 60 kW 6/4 VRM and the
Lm.. 145.9 mH
Le,,,, 19.54 mH
7.5 kW 8/6 VRM. The simulated performance of the 8/6
L,,, 9.15 mH VRM is also contrasted with the modelling approach
a,(@,) 873.6 mWb documented in Reference 3.
~,(e,) 87.36 mWb Fig. 10 compares the simulated and measured phase
a,(@,) 2.599 mH current for the 60 kW 6/4 VRM under conditions where
a&) 8.697 mH
Iterations 2

The geometry of the 1.1 kW 6/4 VRM is described in


Table 7. Table 8 summarises the model parameters of this
VRM as described by the analytic model of Section 2.

Table 7: Geometry and turn information f o r the 1.1 kW 614


VRM. All length.are in metres; all anales are in demrees
Parsmeter Svmbol Value
SMt radius Ret 0.01 27
Rotor back iron radius R,,, 0.0274
Rotm radius at airgap R, 0.0420
Radii1 airgap G 0.0002
Swa radius at airgap R, 0.0422
Radius at top of stator pole R,, 0.0603
Outside radius Ro 0.0762 rotor position, deg
Active length d 0.089 Fig. 10 Measured and simulated phase currentsfor the 60 kW VRM
Rotor pole arc 0, 40 for a rotor speed of 783 r.p.m., a turn-on angle of 45" a turn-off angle of
ststor pole arc e, 40 74" and a DC supply voltage of 231 V . The solid curve corresponds to
Numbe~of rotor polas N, 4 measured data
T m psr pole N 50
Magnetic steel 3% Si the phase currents are regulated. Table 9 compares some
Point A. Fig. 1 BA 1.65 T
Point E. Fig. 1 BB 2.025 T important dynamic and average phase quantities for the
current waveforms of Fig. 10. The operating conditions
Table 9: Comparison between some dynamic and average
Fig. 9 shows the flux linkage as a function of current and quantities for the phase currents of Fig. t O
rotor position trajectories described by the data of Tabk
8. Experimentally measured flux linkage information is Quantity Measured Simulated
also shown in Fig. 9 as a basis of comparison. Fig. 9 Peak current, A 323.4 314.9
shows that, although the model of Section 2 does not RMS current. A 188.0 191.8
exactly reproduce the measured data, the trends are Peak diode current, A 323.4 314.9
Average toraue. Nm 133.6 116.5
correct. Of the three VRMs considered, the accuracy of
the model appears to be poorest with this VRM, because
this VRM is not intended to be operated with heavy include a rotor speed of 783 r.p.m., a turn-on angle of
saturation. As such, this VRM provides a good example 4S0, a turn-off angle of 74" and a DC suppJy voltage of
of the model sensitivityto E , and B,, . 231 V. Fig. 10 shows that the VRM model of Section 2,
20 IEE hoc.-Electr. Power Appl., Vol. 142, No. I , January 199s
as summarised in Table 4, does a good job of predicting cycle. Over the range 25-35 A when the current and flux
the general and specific behaviour of the VRM, even linkage are increasing, the movement between two paral-
though there are some modelling errors in the magnet- lel trajectories can be traced to the discontinuities in
isation characteristic.
Fig. 11 compares the simulated and measured phase

4120-
c
C
E -
a 80-
%
I
Q
40 -

i
O*O
rotor position, deg
phose current, A
Fig. 12 Measured and calculated energy c o n w s w n cycles for a
7 5 kW OULTON VRM
Measureddata
Fig. 11 Measured and simulated phose currentsfor the 60 kW VRM ~

. . .. . . . Model summarised in Table 6


for a rotor speed of4704 r . p n , a turn-on angle of 31". a turn-off angle of ____ Model documented in Referen= 3
7225"and a DC supply voltage of 231 V . The solid curve corresponds to
measured data
Table 11 : Measured and calculated performance of a 7.5 kW
OULTON VRM
current for the 60 kW 614 VRM under conditions where
the phase currents are not regulated. Table 10 compares Measured PC-SRD [3] Model of
Section 2
some important dynamic and average phase quantities
DC supply voltage, V 280 280 280
Table 10: Comparison between some dynamic and average Speed, r.p.rn. 1500 1500 1500
quantitiea for the phasa currents of Fig. 11 Turn on angle, degrees 22.73 22.73 22.73
Turn off angle. degrees 45 45 45
Measured Simulated
Shaft torque, Nm 50.5 46.9 51.7
Peak current, A 183.5 174.0 Peak phase current, A 35.5 36.5 41.1
RMS current, A 79.9 80.7 RMS phase current, A 18.1 17.8 21.8
Peak diode current, A 60.0 46.2 Peak diode current, A 33.0 29.9 34.0
Averaaatoraue. Nm 53.5 53.2
slope at the breakpoints of the piece-wise linear models
for the current waveforms of Fig. 11. The operating con- of Figs. 3-5.
ditions include a rotor speed of 4704 r.p.m., a turn-on
angle of 31", a turn-off angle of 72.25" and a DC supply 5 Summary and conclusions
voltage of 231 V. Fig. 11 shows that the VRM model of
Section 2 does an excellent job of predicting the shape of This paper has presented an analytic summary equation
the dynamic phase current, but there are almost exactly for the terminal magnetisation characteristics of VRMs.
4" of error over the entire current waveform. That is, if It has connected the terminal model to the fundamental
the simulated waveform were to be shifted by 4", then geometry and number of turns per phase of the VRM
there would be almost perfect correspondence between through simple piece-wise linear models. The magnet-
the simulated and measured currents. This error can be isation coefficients in the analytic model are physically
traced to the magnetisation data in Fig. 7 where the motivated, and so are the shapes of the piece-wise linear
model deviates from the measured curves as the VRM models for the magnetisation coefficients.
enters into saturation in the 50-100 A range. The process of taking the basic VRM geometry and
In consideration of the OULTON 7.5 kW VRM turns per phase and producing the terminal magne-
which was used for experimental verification in Reference tisation characteristics was carried out for three VRMs
3, the models of Section 2 were used to simulate the per- which have been previously documented. Comparisons
formance of the 7.5 kW VRM under the conditions docu- between the experimentally measured magnetisation data
mented in Reference 3; the details of the model are with those computed using the models of this paper
summarised in Table 6. Fig. 12 summarises the trajec- reflect some modelling error, though the overall agree-
tories through flux linkage-current plane for the meas- ment is quite good. In a direct comparison with the mod-
ured waveform, the model of Reference 3 and the model elling approach of Reference 3, it was shown that the
of Table 6. Table 11 compares some of the dynamic and models presented here offer a similar performance. In
average quantities associated with the waveforms of Fig. addition, they are structurally simpler and should be
12. Although the model of Reference 3 offers greater computationally competitive with the modelling
accuracy with regard to peak and r.m.s. phase currents, approach of Reference 3.
the model of Table 6 offers greater accuracy with regard The models presented here are intuitive, simple and
to average torque and peak diode current. fast. Further, the terminal summary model is applicable
As shown in Fig. 12, there is one region where the over all rotor positions, and it leads to a closed-form
energy conversion cycle predicted by the model of expression for the instantaneous torque produced by
Section 2 deviates from the actual energy conversion each VRM phase. They are amenable to incorporation in
IEE Proc.-Ek?ctr.Power Appl? Vol. 142,No. I , January 1995 21
computer-aided V R M design packages and general 6 BEVINGTON. P.R.: Data reduction and error analysis for the
circuit-based simulators, such as Saber. physical sciences (John Wiley & Sons, 1969)
7 ILIC-SPONG, M, MARINO, R., PERESADA, S.M., and
TAYLOR. D.G.: Feedback linearizing control of switched reluct-
ance motors, IEEE Trans. Autom. Conirol, 19%7,32,pp. 371-379
6 References 8 BYRNE, J.V., and ODWYER, J.B.: Saturable variable reluctance
machine simulation using exponential functions. Proceedings of the
1 LAWRENSON, P.I., STEPHENSON, J.M., BLENKINSOP, P.T., International Conference on Stepping Motors and Systems, Uni-
CORDA. J., and FULTON, N.N.: Variable-speed switched reluct- versity of Leeds, UK, 1976,pp. 11-16. (This paper also appears in
ance motors, IEE Proc. B, 1980,127,pp. 253-265 MILLER, TJ.E. (Ed.): Switched reluctance motor drives, a refer-
2 MILLER, TJ.E.: Brushless permanent-magnet and reluctance ence book of collected papers (Intertec Communications, Ventura,
motor drives (Clarendon Press, 1989) CA, 1988),pp. 154-159)
3 MILLER, T.J.E.: Nonlinear theory of the switched reluctance 9 TORREY, D.A., and LANG, I.H.: Modelling a nonlinear variable
motor for rapid computer-aided design, IEE Proc. B, 1990, 137, reluctance motor drive, IEE Proc. B, lY90,137,pp. 314-326
pp. 337-347 10 TORMEY, D.P., TORREY, D.A., and LEVIN, P.L.: Minimum-
4 CAMERON, D.E., LANG, I.H., and BELANGER, D.: The airgappermeance data for the doubly-slotted pole structures
computer-aided design of variablereluctance generators. Pro- common in variablereluctance motors. Proceedings of the lEEE
d i n a s of the Auulied Power Electronics Conference. Boston. Industry Applications Society Annual Meeting, Seattle, USA, 1990,
us4 iw,pp. 1 i i z o pp. 19~L200
5 CORDA, J., and STEPHENSON, J.M.: Analytical estimation of 11 HARRIS, M.R., and MILLER, TJ.E.: Comparison of design and
the minimum and maximum inductances of a double-salient motor. performance parameters in switched reluctance and induction
Proceedings of the International Conference on Stepping Motors motors. IEE Conference on Electrical Machines and Drives,
and Systems, University of Leeds, UK, 1979,pp. 50-59. (This paper London, UK, 1989,pp. 303-307
also appears in MILLER, T.J.E. (Ed.): Switched reluctance motor 12 HARRIS, W.D., and LANG, J.H.: A simple motion estimator for
drives, a reference book of collected papers (Intertec Communica- variable-reluctance motors, IEEE Trans. Ind. Appl., 1990, 26, pp.
tions, Vmtura, CA, 1988). pp. 144153) 237-243

22 IEE Proc.-Electr. Power Appl., Vol. 142, No. I , January 1995

Você também pode gostar