Sensorless Control of AC Machines at Low Speed

and Standstill Based on the "INFORM'' Method

Manfred Schroedl, Member, IEEE
Institute of Electrical Machines and Drives
Vienna University of Technology
Vienna,A USTRIA

Abstract - The increasing use of AC machines compared to DC is due to a stator voltage SP ys during a short time interval (P.u. time
motors in electrical drive applications has several reasons. A 7) and repeating this measurement in different SP directions. The
very important advantage of AC machines is their simple complex ratio (defined at standstill in order to eliminate the
construction. However, AC drives often need mechanical sensors influence of the EMF)
(tachometers, position encoders) for field orientation. In many
applications these sensors reduce robustness and increase costs of
a drive considerably. The main objective of this paper is to
present the "INFORM" method based on real-time inductance
measurements for sensorless control of AC drives. is called "complex INFORM reactance" (P.u. value) and contains the
desired information about the flux (or rotor) angular position. (In the
I. INTRODUCTION paper, p.u. reactances x are defined at reference angular frequency
and are hence proportional to the respective inductances). In
Sensorless control of AC drives has been a field of research for practice, the voltage test phasor y, is applied to the machine by an
many years [1]-[4].A lot of flux-detecting methods based on inverter (usually three-phase machines, possible SP angles arg(ys) =
evaluation of the EMF have been developed. By using improved yv = k d 3 , k=0,1,2,..). Since geometric and saturation effects are
algorithms, the operating range of EMF-based models can be 18O0-periodic,gWom is a 180O-periodicfunction, too:
extended to the low-speed range [5]-[7].However, EMF evaluation
does not work in the range of flux standstill since EMF vanishes. In
XmFORM = XlMFORM (2Y, - 2 Yv 1 (2)
this paper, the "INFORM" method ("IJdirect Flux detection by
On-line Reactance Measurement") [8], based on real-time
For implementation on a microprocessor, it is advantageous to work
inductance measurements for sensorless control of induction motors with the inverse (P.u.) INFORM reactance,
(IM), permanent magnet synchronous motors (PM) including "EC-"
and "brushless DC motors" and synchronous reluctance motors (RM)
will be presented. The idea and first measurements using saliency
and saturation effects for flux detection in AC machines has been
published by the author in 1988 [9].A related evaluation using high- since divisions can be avoided. Assuming a reference frame E,q with
frequency signals has been described in [lo]. real axis 5 in the direction of maximum magnetic conductivity
1Ms are widely used in industry. They have a simple and robust rotor ( W O R M reference frame, angular position ymOw relative to the
construction and offer field-weakening capability. A comparison of stator-oriented (a@)reference frame) yields the following real and
different flux models has been described e.g. in [l 11. Application of imaginary part ofymoRM:
the TNFORM method enables torque production at zero speed with
very good static and dynamic properties. As an example, a traction
drive will be presented [12]. PMs are frequently used in servo
applications and show excellent efficiency. A lot of publications deal
with sensorless control, e.g. [13].Applying the INFORM method, a
highly dynamic sensorless drive for speed- and torque-controlled or in complex notation:
operation as well as simple position control with limited accuracy
can be realized. RMs offer advantages at drives with exact speed
control and field-weakening capability [ 141.
For illustration, assume a repeated measuring procedure according to
11. BASIC IDEA OF THE INFORM METHOD eq. (1) with a voltage test SP in the real stator axis (inverter
switching state "1,0,0", yv= 0, Fig. 1). The rotor is turned (slowly) in
A. The Complex INFORM Reactance mathematical positive direction, starting with yWo+ =O (Posl in
Fig. 1). In Fig. 1 , y at Position 2 (INFORM axis 45") is shown.
All types of AC machines have magnetic properties which are not
exactly independent of the angular position of the rotor or the flux,
respectively. This is either due to geometric construction (salient
pole character) as in the case of reluctance machines and permanent
machines with buried magnets or due to saturation effects as given in
the case of induction machines and permanent magnet machines
with surface-mounted magnets. This magnetic anisotropy gives ___I)
information about either the direct (d-) and quadrature (9-) axis (or VOLTAGE SP
INVERSE' DIRECTION
rotor angular position, respectively) of RMs and PMs or about the INFORM REACTANCE-
flux axis (IMs, PMs). The magnetic anisotropy is detected by PlRECTloN OF CURRENT CHANGE] Ipos21
measuring the change of the complex (P.u.) stator current space
phasor (subsequently abbreviated SP, complex quantities underlined) Fig. 1 . Locus of djs/dz and y in the complex plane due to a voltage SP

0-7803-3544-9/96$5.00 0 1996 IEEE 270

Authorized licensed use limited to: Technical University Kosice. Downloaded on October 04,2024 at 14:21:24 UTC from IEEE Xplore. Restrictions apply.
The reactances in the INFORM-axes ("Posl" and "Pos3" in Fig. 1) It is straightforward to show that
are obviously real quantities
arg[GlHFoM,REa] = 2 Y I N F " M , R m + 7L (15)
<-axis: X g r N F o n = [Yo - Ayy1-l
q-axis: xRINFORM = [yo +Ay]-'
The parameter yo is a zero-sequence quantity when applying
definition (14) to eqs. (11)-(13) and vanishes. The parameter Ay
represents the radius of the circle G ~ in the complex
~ plane. ~ ,
B. Calculating yIHFom from INFORM Measurements The argument of the circle is the desired quantity 2ywFo,,+n which
is independent of Ay. Hence, this strategy provides the desired
It is advantageous to use at least two measurements as it is then not
INFORM angular position without explicit knowledge of any
necessary to know the INFORM reactance explicitely. Performing
inductance parameter (Fig, 3).
two or three measurements, a lot of mathematical evaluations for
calculating ymoRM are possible. Each A&-measurement at given
2) Using two measurements and imaginay parts yhwINFOwonly:
measuring time AT and given voltage SP v, (measuring direction yv) While example 1 utilized the change of the magnetic conductivity
yields two equations which determine the real and imagmary part of evoked by the INFORM effect, this example uses the phase shift
the inverse INFORM reactance, yE,moM(ym,M,yv) and between current change and voltage due to the variable magnetic
ym,wFoRM(ymoRM,yv). In the following, two examples of evaluations properties. Consider two measurements 1,2 with measurement
are given. directions yV,,=2a/3 (phase axis B), yVz=4d3 (phase axis C),
represented by the following equations:
I ) Using three measurements and real parts yIIEINFOmon&:
Consider three W O R M measurements carried out within a
relatively short time interval with "measuring directions" yv, 1-0
(phase axis A), y =2d3 (phase axis B), yv,3 =47d3 (phase axis C)
(inverter states y?,O,O", "0,1,0" and "0,0,1"). The machine (at
standstill) reacts with current changes accordmg to eqs. (1) and (3): Transforming again the current change SPs into a voltage SP-
reference frame by multiplying eqs. (16) and (17) with exp(-jy,,) and
exp(-j yVJ, respectively, and taking the imaginary parts yields

Im{ .exp( - j 2 z/
3)) = AP I.[-Ay. sin(2ylNFoM- 4 n / 3)] (18)
Multiplying eq. (9) with exp(-jyKk) yields the current change in a
voltage SP - oriented reference frame: Im{ 1.
.exp(-j4 n/3)) = Az. (vs [-Ay. sin(2 ylwon - 2 a/3)] (19)
Expressing the imaginary parts of the left-hand sides by phase
quantities using basic space phasor relations
The real part of the left-hand side is the projection of the current
change SP onto the respective measuring voltage SP direction, which 1
Im{ Ais,, .exp(-j.2n/ 3)} = -.(Aisc,l -AisA,,)
is identical to the regarded phase axis. Hence, the mentioned J5
part is the phase current itself. The magnitude of all voltage SPs is 1
Im{Als,2 .exp(-j.4n/ 3))
equal to 2/3 of DC link voltage. Comparing the real parts of left- and
right-hand side of eq. (10) yields with eq. (4)
we obtain after trigonometric calculations from eqs. (1 8) and (
AiSA,I = ~z-~vsI.[Yo-~Y~-~s(2~,,,n)l (11)
~ z = A~ ~ + ~~ I . [, - ~ A, ~ ~ w ~ , ~ -4Ki3)i
~ , ~ (12)
Aisc,3 = Az.lvsI.[yo -Ay~cos(2yl,,,,,,, -2n/3)] (13)

Note that the different phase current changes are taken from This expression is again independent of any inductance parameter.
different INFORM measurements 1,2,3. Only the phase which Of course, by using Kirchhoffs law, it is sufficient to measure only
coincides with the test voltage SP is considered. This enables a low- two currents and calculate the third one.
cost realization of the INFORM method without full current space
phasor measurement (Fig. 2). 3) Combining evaluations of real and imaginay parts: Since the
A complex linear combination of eqs. (11)-(13) yields a complex measuring noise of real and imaginary parts of ywow is
quantity c,,,,:
cINmM,REar.:=
AisA,, +AisB,z ~exp(j4n/3)+Aisc,,.exp(j2z/3) (14)

DC link current =
phase current E

shunt yoitage
proportiona to
phase current E
14 -_/
AC motor
0
-IT 0 U

Fig. 3. Modulated real parts of current change SP due to saturation and

calculated argument 2y = 2yw,RM,EAL+7t over angular position y
Fig. 2. INFORM realization with DC link current measurement only (measurement on a NdFeB - PM used in [ZO])
271

Authorized licensed use limited to: Technical University Kosice. Downloaded on October 04,2024 at 14:21:24 UTC from IEEE Xplore. Restrictions apply.
statistically independent, a certain improvement of the measurement For distinguishing between north and south poles, the influence of
result can be achieved by stochastically optimal combination of real the stator current on the flux-parallel saturation level in the machine
and imaginary part evaluations (weighting factor k ) is utilized. For this purpose, after having detected the flux axis (but
not knowing the sign of the flux yet) as described above, a flux-
2 YIAF" = k . 2Y,wom* REAL + (1 - k ) .2 YIN.FoRM,IMac (23) parallel stator current component (produces no torque!) is applied.
Now, an W O R M measurement is camed out. This procedure is
Let the variances V of the measurement errors be repeated with changed sign of the stator current. Again, an INFORM
measurement is performed. Comparing the magnitudes of the current
changes during the INFORM measurements shows which case was
the flux-increasing (larger current change per time in flux direction)
or flux-decreasing (smaller current change per time since saturation
i s reduced). Hence, the flux SP i s fully detected now (Fig. 5).
at which the actual INFORM angular position 2yINFORM,act has been
obtained by an encoder (PM, RM) or sensor-based flux model (M).
The probability density function ~(2~,,,,-~-2y,,~~~~)
can be assumed Gaussian as measurements verify (Fig. 4).
Minimizing variance V of the combined measurement with respect
to k yields

k = vk4G
> Vl, =
VREAl ' V M G
(26)
OPf
V, +VIA,, V m . +VIM,

The variance of the measurement error i s used for calculating

feedback coefficients when constructing a state model with output
error feedback (Kalman filter, chapter IV).

C. Eliminating the 180a-ambigui?vof the INFORM angle yINFORM

As already stated, only double the desired angle, 2y, (+2kn!)

can be detected from the above described measurement procedure. Fig 5 Changmg the magnebc set-point of a PM by a
flux-parallel stator current component
Calculating yrNFoRMyields two solutions ywoRMJ and ymoRMJI with
3) Reluctance machines: After detecting the INFORM axis, which
~INFORM,II = YINFORn4,I + 7 l '
(27) coincides with the direct (d-) axis, the flux-producing current
component for normal operation is applied to either positive or
To get proper information ywoRM, different strategies for PMs, LMs negative INFORM axis direction In both cases, the flux in the
and RMs are used. machine is well-defined and torque-producing operation can start

1) Induction machines: Before operation, no flux is in the machine. n1. TNFORM METHOD FOR TRANSIENT OPERATION
Hence, the machine is fully symmetric (besides slotting effects). No
W O R M axis can be detected. If (magnetizing) current is applied to When changing from standstill to arbitrary transient operation, the
the machine at no load (e.g. current SP in phase axis +A) then the dynamic equations of the machines have to be considered.
flux SP will have the same direction (phase axis +A). Starting an
INFORM measurement after magnetization will provide flux axis A. General Equations Valid For IM, PM, RM
information due to saturation effects. The sign of the flux has
already been determined by the magnetizing stator current. Hence, 1) Stator voltage equation: A general relation is the coherence
initial conditions for starting machine state calculation (Kalman between stator voltage SP y s and stator flux linkage SP 2,since it
filter) and normal operation are well-defined. does not depend on geometric parameters and inductances (p.u
stator resistance rs ,arbitrary reference frame rotating at oK):
2) Permanent magnet machines: The detected INFORM axis
(direction of maximum inductance) coincides with the flux axis in
the case of flux-concentrating machines with buried magnets
(reluctance effect dominates) or i s perpenticular to the flux axis if air
gap magnets are used (saturation effect dominates). 2) Mechanical equation:

dco
z;-=tt,+t,
dr

The resulting torque (sum of motor torque t, and load torque tL)
accelerates the rotor ( p u angular velocity a) The start-up time
constant 7, IS the time for accelerating from standstill to rated speed
at rated resulting torque
Expressing the produced torque by stator flux linkage SP and stator
current SP yields again a relation valid for IM, PM and RM
(* complex conjugate, arbitrary reference frame <,q)

272

Authorized licensed use limited to: Technical University Kosice. Downloaded on October 04,2024 at 14:21:24 UTC from IEEE Xplore. Restrictions apply.
According to the control strategy, modified expressions of eq. (30) miminum leakage reactance (INFORM reference frame <,q with
will be used. IMs are usually controlled based on rotor flux SP dR, angular velocity am). Using two axis theory, the saturation effect in
PMs use the permanent magnet flux linked with the stator (& ), the leakage path due to main flux saturation is described by
RMs may be controlled based on given d-axis flux linkage. More
details are given after introducing flux linkage equations.

3) Flux linkage equations: They describe the relations between flux

linkages and evoking MMFs (due to winding currents and permanent
magnets) by inductance parameters. If there is no significant angular
dependence of the magnetic properties, then the flux linkage Obviously, the EMF. and the stator resistance have influence on the
equations can be written in complex SP-form. This is possible for IM change of current. In order to eliminate these effects, assume a
and PM with surface-mounted magnets as far as saturation effects combination of two INFORM measurements with the following
are not taken into account. However, for evaluating the INFORM properties:
effect, a generalization is necessary as will be shown subsequently. - Test voltage SPs are different (e.g. opposite directions zs,m= - Y ~ , ~ )
Machines with a considerable magnetic anisotropy as RM and PM - Tests within a short period (ARJI =
with buried magnets are usually described in components.
amoRM;II = aINFoRM,I))
- Current changes during INFORM tests are small = is,I).
B. Equations of the ZM with Respect to INFORM Effect
Subtracting measurement equations I and I1 (eqs. (39) and(40) with
indices "i'and "II"r respectively), yields with the defined measuring
The flux linkage is produced by stator and rotor currents:
properties

Introducing a leakage parameter o instead of the mutual inductance

2
1-* (33)
' R 'xS
The effects of stator resistance and EMF are eliminated when
subtracting both equations since they are more or less identical in
and choosing (without loosing generality) a proper "winding ratio" both measurements. This means that W O R M measurements are
between stator and rotor in order to obtain possible at arbitrary speed by substituting a simple measurement by
a pair of EMF-compensating measurements and replacing voltages
'R = xSR (34) and current changes by appropriate differences according to eqs. (41)
and (42).
Introducing eqs. (7) and (8) in eqs. (41) and (42) yields

(35) (43)
(36) (44)

The torque equation (30) can hence be expressed using the rotor flux With the inverse complex INFORM reactance
linkage by (x,y-reference frame; x-axis coinciding with rotor flux SP
direction)

ti = -Im{is *.A.,}= IARI.iSy. (37) Eqs. (41) and (42) can be expressed in the compact complex
notation
For short time intervals (essentialy shorter than the rotor circuit time
constant ,z = xR/r,) the magnitude of the rotor flux linkage 1 & 1
can be assumed constant. Assuming linear relations lead with eqs.
(28) and (36) to
with

When considering saturation effects in the motor, eq. (38) has to be

modified. The saturation in the main path leads to variable Obviously, eq. (45) is valid not only in the INFORM reference frame
reactances with respect to flux angular position. In the IM case, this 5,q but in arbritrary reference frames, since the complex parameter
variable main reactance cannot be detected by a current ymoRMdepends on the angular difference between INFORM axis
measurement, since the current's bahaviour is defined by leakage and test voltage SP axis.
path relations, represented by (oxs)in eq. (38). However, there is a
certain influence of the main path saturation on the leakage path A necessary condition for proper working of the INFORM
saturation. Hence, it is possible to get indirect information about the evaluation applied to IMs is that there is a sufficiently measurable
flux axis by measuring the magnetic properties of the lcakcige path effect of leakage path modulation due to main path saturation. This
by analyzing the stator current. (Specially constructed rotors with depends on the one hand on geometric and material parameters and
geometric leakage reactance modulation were presented in [151. on the other hand on the flux level in the machine. Usually, the
However, in this paper, only standard rotors are taken into account). higher the main path saturation, the larger the desired effect.
A mathematical description of the saturation effect is obtained by However, "INFORM-friendly" motors show good effects at rated
defining a reference frame coinciding with the maximum and saturation level [ 121. There exist four types of motors:

Authorized licensed use limited to: Technical University Kosice. Downloaded on October 04,2024 at 14:21:24 UTC from IEEE Xplore. Restrictions apply.
Type A: INFORM axis 5 perpendicular to flux SP direction
Type B: INFORM axis 5 parallel to flux SP direction
Type C: Motor changing from type A to B when changing flux level
Type D: INFORM effect too small for evaluation

The INFORM method has been applied to "type A - motors" (e g 80

kW traction motors of new Vienna low-floor tram) for test purposes The reactances x ~ and, xS,- ~ are ~ &fferential
~ quadrature- and
direct-axis reactances x ~ , and
~ , x~ ~ in ,the ~magnetic
~ set point
and to "type B" - motors (30 kW Volkswagen VW T 4 [12], Fig 6),
defined by the permanent magnets (linearized system in the
which has been running in the Vienna traffic for a few years A "type
magnetic set point). For eliminating EMF and stator resistance, two
C" - motor is under test at Vienna University of Technology
INFORM measurements are combined in the same way as shown in
the IM case:, leading to

and

Hence, the INFORM axis t is uerpenticular to the flux (d-) axis.

Fig 6 Measured complex INFORM curve ymOm of the IM
used in the VW T4 [ 121 (set point rated flux, rated torque) 2).PM with flw-concentration and buried magnets: The flux of
two neighboured buried magnets is lead to the air gap (Fig. 8).
C. Equations of the PM with Respect to INFORM Eflect Hence, the direct axis has the maximum stator inductance.In the
quadrature axis, the magnets act like air gap for the stator winding.
I ) PM with air gap magnets and slotted stator: The flux that links Obviously, the INFORM axis E, is in parallel with the flux axis.
the stator is produced by the stator current itself and the permanent Due to the considerable inductance difference in direct (d-) and
magnetization quadrature (9-) axis the equations are described in two axis
representation. Furthermore, the relatively large direct axis reactance
enables certain field-weakening and hence change of the d-axis
saturation level, represented by a changing d-axis reactance. This
at which ifict,M represents a (fictitious) MMF producing the same fact is taken into account by introducing set-point reactances xdoand
stator flux linkage AM as the permanent magnet. xqoand differential reactances x ~and, xq,dle,
~ the~ latter
~ responsible
The general torque expression (30) can be specialized for PMs with for the current changes [ 161. Neglecting cross-coupling effects, the
constant air-gap (saturation effects neglected for torque production): flux linkages can be written as

Eq. (47) yields with eq.(28) in a reference frame rotating at oK:

The torque equation (48) gets an auxiliary reluctance term which has
(49) no effect if d-axis current is controlled to zero:

Since the permanent magnetization evokes certain saturation mainly

in the stator teeth (Fig 7) it IS obvious from INFORM point of view
With flux linkage equations (56) and (57) the voltage equations
to model this fact again by space-dependent reactances The
become
saturation effect is modeled in two-axis theory by defining a
reference frame coinciding with the maximum and minimum stator
reactance (INFORM reference frame E,,q w t h angular velocity ow)

Combining again two INFORM measurements yields

xis

01 !E a 2.u E

Fig. 7. A x gap flux density BMand permeability pe in the stator teeth

depending on stator angular position a (at fixed rotor position y) Hence, the INFORM axis coincides with the flux (d-) axis

274

Authorized licensed use limited to: Technical University Kosice. Downloaded on October 04,2024 at 14:21:24 UTC from IEEE Xplore. Restrictions apply.
 ~

IV. OPTIMAL STATE ESTIMATION

After correcting the systematic deviations stated above, an offset-

free noisy INFORM-based angular position yINF,corr= y has been
measured (y.. preferrably rotor flux angular position in IM case, rotor
angular position in PM and RM case). Since the statistical properties
(see Fig. 4) can simply be detected, a stochastically optimal state
calculator can be constructed (Kalman filter, [ 181, [191).

A. State Model

Fig. 8. PM with buried magnets and flux concentration A very advantageous strategy, which needs only little mathematical
expense is modelling only the mechanical part of the machine. A
D. Equations of the Synchronous Reluctance Machine general disctete-time structure with sampling time 7, = r(k+ 1)-7(k),
valid for IM, PM and RM, can be given by introducing the state
The RM equations are obtained by setting the permanent variables "angular position y" as defined before, "mechanical angular
magnetization in the PM with buried magnets zero. Furthermore, velocity w" and "load torque tL"
cross-coupling effects are neglected. The flux linkages are hence

The torque equation consists only of a reluctance term: This means that the load torque is modeled as a first order Gauss-
Markov process with Gaussian noise input w2(k), representing the
expected change of load within a sampling interval 7,. wZ(k)depends
on the load and hence on the application. The difference between
Obviously, the stator current must have both a d- and a q-component actually produced machine torque and reference value ti(k) is
for producing torque. An INFORM measuring procedure yields the modeled in the stochastic term w,(k) (white Gaussian sequence with
same results as given in the PM case with buried magnets (eqs (61) mean zero), which is mainly caused by the non-ideal behaviour of
and (62)). Again, the INFORM axis 5 coincides with the d- axis. the switching current control loops.
In a realized drive (4 pole 1 kW RM) Fig. (yy) shows that good
INFORM results are achieved as long as the d-axis current does not While the coefficients
produce high d-axis saturation. In a realized drive (4 pole 1 kW RM
[17]), d-axis current should not exceed 0.6 (Fig. 9) since the
differential reactances in both axes get equal.

are the same for all machine types, the coefficient b, has different
meaning in the IM and PMiRM case:

q -axis
The additional term in the IM coefficient is due to the slip angular
velocity (aFLux.-0).

In the measurement equation (69), wmnM(k) with variance V (refer

0 0.6 0.8 1.2
0.2 0.4 1
. .
to subsection II.B.3) is the stochastic INFORM measurement noise:
'Sd 'Sa
Y ( k )= 2.y,,,,(k)+w,,,,,(k) (69)
Fig. 9. Measured magnetizing curves in d- and q-axis of a RM
B. Kalman Filter
E. Influence of the Stator Current on the INFORM Angle
In the prediction step, the state (k+l) is predicted from state (k)
without measurement information:
As shown, the INFORM anmlar position yINFnRM,which is closely
related to the flux axis, is obtained without any machine parameter
by appropriate compensating measuring algorithms, even at
changing reactances due to transiently changing magnetic set points. (70)
However, for accurate measurement results at loaded machines, the
stator current should be considered. Each machine type shows more
or less the same behaviour. When applying a torque-producing
current component, the INFORM angular position gets a certain After an INFORM measurement, the predicted state is corrected by
weighting the difference between measurement and expected
angular displacement, which is usually taken into account by a
measurement according to the Kalman gain coefficients K,.- K3.
simple correction function.
Since the filter reaches its steady-state parameters within a few
tenths of seconds, it is sufficient to use the (constant) steady-state
Kalman gain matrix K,:
275

Authorized licensed use limited to: Technical University Kosice. Downloaded on October 04,2024 at 14:21:24 UTC from IEEE Xplore. Restrictions apply.
[;L
i j ( k + 1/ k + 1)
p(k+l/k+l) = p(k+l/k)
i j ( k + 1/ k )

( k + 1/ k + l j [ & ( k + I / k j
+
[z]
K2, . [ 2 y ~ ~ , , , - 2 p ( k + l / k ) ]

K,,, can be calculated off-line via Matrix Riccati equations. This

reduces real-time calculating expense considerably.
(71)
inverter with hard DC link, fixed-point DSP).The car has been
running for a few years (20.000 km) in Vienna traffic without
problems. In case of torque control as usual in traction, the speed
controller is not implemented and a direct torque reference
command t,,refis used.
The behaviour of the traction drive in low-speed range at the test
stand is presented in Fig. 10. The load machine was p-controlled to
zero speed so that low speed (about 100 rpm) occurs at rated torque.
V. MEASUREMENT RESULTS A torque reversing process from -120% to +120% of rated torque
within 300 ms is shown. Furthermore, angular position (from witness
A lot of machines (IM, PM, RM) have been tested using INFORM encoder, INFORM-calculated flux angular position (different due to
flux/position detection and torqueispeedlposition control (e.g. 181). slip!) and angular speed are shown. In the real application, it is no
In the higher-speed range, EMF-based models have been used [SI A problem to accelerate at steep roads from rolling back to positive
few examples of measurements will be given subsequently. The speed with full torque and to keep the car at constant position
control structure which can be applied to all types of machines is without using a brake.
given in Fig. 12 (The meaning of y is different for IM and PM/RM
B. Permanent Magnet Motor
and has previously been defined).

A. Induction Motor PM motors with INFORM-based control have been tested in traction
applications (2 kW motorbike 1201) and for speed- and rough
The INFORM method was implemented in traction applications like position control (standard 5 kW servomotor with SmCo5 magnets).
electric vehicles (VW T4, 30 kW induction motor, speed 100 k d h , Figs. 11 and 13 show the transient behaviour in speed- and position-
rated torque 100 Nm, peak torque 130 Nm, motor speed 0 - 2250 controlled mode. In position-controlled mode, the limits of the
(base speed) - 6000 rpm (field weakening), 220 V lead battery, IGBT INFORM method concerning angular accuracy are shown. The
reference curve y is obtained by a witness encoder, which is not used
in the control loop.
Presently, a low-cost PM application with structure of Fig. 2 is being
tested with good results.

Fig. 10. Low-speed txansient behaviour of INFORM flux detection (IM) Fig. 11. Step response of the speed-controlled PM
Torque change from -120% to 120% ofrated value withu~300 ms

1 7
I

! SIGNAL PROCESSING SYSTEM

I IY- ~ I INVERTER
I

I
I

I
I
I

I ..................................................................

I
I Fig. 12. INFORM-based control structure of AC machines

276
Authorized licensed use limited to: Technical University Kosice. Downloaded on October 04,2024 at 14:21:24 UTC from IEEE Xplore. Restrictions apply.
 [4] Arefeen, M.S., Ehsani, M., Lipo, T.A., "Sensorless position
measurement in synchronous reluctance motor", IEEE Trans. on
Power Electronics, Vol. 9, No. 6, pp. 624-630, 1994.

[5] Bausch, H., Kanelis, K., Lange, B., Zeng, W., "Torque control of
synchronous and asynchronous drives without mechanical sensors",
Proc. Int. Conf. on Electrical Machines (ICEM) Paris, 1994, pp. 324-
328.

[6] Tajima, H. Hori, Y., "Speed sensorless field-orientation control o f

the induction machine", IEEE Trans. Ind. Appl., Vol. 29, No. 1, pp.
175-180, 1993.

[7] Yang, G., Chin, T.-H., "Adaptive-speed identification scheme for a

vector-controlled speed sensorless inverter-induction motor drive",
IEEE Trans. Ind. Appl., Vol. 29, No. 4, pp. 820-825, 1993.

[8] Schroedl, M., "Sensorless control of A.C. machines", VDI

Fortschrittberichte2 1, No. 117. VDI-Verlag Diisseldorf (Germany),
Fig.13. Response to a 45"-reference step in angular position of a PM 1992.

C. Reluctance Motor [9] Schroedl, M., "Detection of the rotor position of a permanent magnet
synchronous machme at standstill", Proc. Int. Conf. on Electrical
Using the structure of Fig. 12,a 4 pole, 1 kW synchronous reluctance Machines (ICEM), Pisa (Italy), 1988, pp. 195-197.
machine was controlled [I 71. The Kalman-filtered values of speed
and angular position are shown during a reversing process in Fig. 14. [ 101 Jansen, P.L., Corley, M.J., Lorenz, R.D., "Flux, position and velocity
The drive is well-suited to speed-controlled applications with field- estimation in AC machines at zero and low speed via tracking of high
weakening capability fiequency saliencies", Int. Conf. on Power Electronics and
Applications (EPE), Seville (Spain), 1995, pp. 3.154-3.159

[I 11 Jansen, P.L., Lorenz, R.D., "A physically insightful approach to the

design of flux observers for field oriented induction machine drives",
IEEE Trans. Ind. Appl., Vol. 30, No. 1, pp. 101-110, 1994.

[12] Schroedl, M., Hennerbichler, T., Wolbank, T.M., "Induction motor

drive for electric vehicles without speed- and position sensors". 5 .
Int. Conf on Power Electronics and Applications (EPE), Brighton
(U.K.), 1993, Vol. 5, pp. 271-275.

[ 131 Ertugrul, N., Acamley, P., "A new algorithm for sensorless operation
of permanent magnet motors", IEEE Trans. Ind. Appl., Vol. 30, No.
Fig. 14. Reversing process of an INFORM-based speed-controlled RM 1, pp. 126-133, 1994.

VI. CONCLUSIONS [ 141 Lipo, T.A., "Synchronous reluctance machines: A viable altemative
for ac drives?" El. Machines and Power Systems, Vol. 19, pp. 659-
The INFORM method for detecting flux andor rotor position in AC 671, 1991.
drives especially at low speed and standstill has been presented. The
basic idea is to utilize inductance differences due to saturation or [IS] Jansen, P.L., Lorenz, R.D., "Transducerless position and velocity
geometric construction. A very general machine description with Estimation in induction and salient AC machines", IEEE Trans. Ind.
respect to INFORM evaluation was used, valid for induction, Appl., Vol. 3 1, NO.2, pp. 240-247, 1995.
synchronous and reluctance motors. It was shown how to eliminate
'[16] Schroedl, M., Stefan, T., "New rotor position detector for permanent
the EMF by a special measuring sequence and how to improve the magnet synchronous machines using the "INFORM"-method,
result by constructing a simple Kalman filter for state estimation. European Transactions on Electrical Power Engineering (ETEP),
Measurement examples at different machine types show the static Vol. I, Nr. 1, 1991, pp. 47-53, VDE-Verlag (Germany).
and dynamical properties of the method.
[ 171 Schroedl, M, Weinmeier, P., "Sensorless control of reluctance
REFERENCES machines at arbitrary operating conditions including standstill", IEEE
Trans. on Power Electronics, Vol. 9, No. 2, 1994, pp. 225-231.
[ 11 Rajashekara, K., Kawamura, A., Matsuse, K., "Sensorless control of
AC motor dnves", IEEE Press, 1996. [18] Kahnan, R.E., "A new approach to linear filtering and prediction
problems", Trans. ASME, J. Basic Eng., Vol. 82, 1960, pp. 35-44.
[2] Sepe, R.B., Lang, J.H., "Real-time observer-based (adaptive) control
of a permanent-magnet synchronous motor without mechanical [19] Jazwinski, A.H., "Stochastic processes and filtering theory",
sensor", IEEE Trans. Ind. Appl., Vol. 28, No. 6, pp. 1345-1352, Mathematics in Science and Engineering, Vol. 64, Academic Press,
1992. New York-London, 1970.

[3] Henneberger, G., Brunsbach, B.J., Klepsch, T., "Sensorless operation [20] Schroedl, M., Colle, T., "Electric motorbike with sensorless
of induction machines", Proc. Symp. Power Electronics, Electrical controlled permanent magnet synchronous motor", Proc. 11th Int.
Drives, Advanced Elecbical Motors (SPEEDAM) Italy, May 1992, Electric Vehicle Symposium (EVS), Florence (Italy), 1992, pp.
pp. 263-268. 13.07/1-11.

277

Authorized licensed use limited to: Technical University Kosice. Downloaded on October 04,2024 at 14:21:24 UTC from IEEE Xplore. Restrictions apply.