Linear Induction Motor Demo-Track : a Tool for Renewed Motivation in

Electrical Engineering

STRUBIN, J.-M., VEENSTRA, M., RUFER, A.

Laboratoire d’électronique industrielle
EPFL, Ecole Polytechnique Fédérale de Lausanne
CH 1015 Lausanne
Phone: +4121 693 46 76
E-Mail: alfred.rufer@epfl.ch

Abstract been used on this drive are presented. The first of these
strategies is running using a speed sensor, the second one
Technical education in classical fields like electrical without.
engineering is confronted in western countries with on one
side a high demand for engineers coming from industry and 2 Variable voltage and variable
on the other side by the reduced interest from younger frequency LIM system
generation, whose interest is highly polarised by
information and communication technologies. A linear
motor demo-track has been set up, in order to reinforce the 2.1 Introduction
demonstrative aspects of modern technology as a
contribution to the motivation of students for the industrial Early developments in LIM have been done at EPFL
world. It is also an alternative example to the too classical during the 60’s and 70’s. [4]. Practical results have been
rotating asynchronous motor in the field of education on verified without use of power electronics converters and
variable speed drive. control. Similarly to the newer LIM applications [5], the
existing motor is reused with variable voltage and variable
Keywords: demonstration drive, linear asynchronous frequency, this achieved with a modern IGBT PWM-
motor, sensorless. Inverter fed through a diode rectifier from the AC Network.
The scheme of the power circuitry is given in Fig. 1.
Depending on the control strategy implemented, the
1 Introduction regulation algorithm may need to precisely know the stator
voltage. Instead of measuring each output phase voltage the
As Electrical engineering is not really popular for the complete alimentation circuit is modelled in function of the
new students, revisited methods and tools for renewed DC-link voltage. This way, the output of the PWM-Inverter
motivation are currently a discussion theme in many can be controlled measuring only the DC voltage, which
technical universities [1]. needs only one voltage sensor.
As simulations and computers are intensively used in the
classes and labs in relation with teaching activities, most 2.2 Motor (stator)
faculties now omit to demonstrate, with visible and moving
attractive high technology devices, what can be realised The used LIM is a three-phase Linear Induction Motor
using recent research and development. with double inductor. The windings and connections on the
At EPFL, an existing linear induction motor (LIM) has stator are represented on Fig. 2.
been reused, on a newly installed V-shaped test track, as In this utilisation, the neutral point isn’t connected, but a
training and demo feature for students in the department of connector exists so it can be done if desired in the future.
electrical engineering. Asymmetries and other non-typical
behaviour specifics to the LIM can now be studied and
presented.
As field oriented, sensorless or direct torque control of
asynchronous motors have been in the centre of research
during the past twenty years in many universities and
industries, the characteristic properties of linear motors due
to asymmetries lead to complex modelisation and complex
control strategies [2]. Simpler approaches are easily
possible, because the asynchronous linear induction
machine with massive rail is highly tolerant regarding
orientation because of its dissipative rotor characteristics.
Figure 1 : Supply of the LIM
In this paper the power and control systems are first
described. Then two different control strategies that have
 N Control board
Bottom view
phase 1 phase 2 phase 3
Figure 2 : Windings of the LIM
Its main characteristics are:
· Sn= 36[kVA] : nominal apparent power
· In=41[A] : nominal current
· Un=165[V] : nominal phase voltage
· p=1 : number of pole pairs
· tp=129[mm] : pole length
· vs=12 [m/s] : synchronous speed
· Ns=58[turns/inductor]
· Zn=12: number of slots
· Star coupling, free neutral point
· d=2[mm] : air gap
2.3 Rail (rotor)
The motor rail, corresponding to the massive rotor of a
conventional induction machine, is a simple V-shaped
aluminium I-profile, carrying and guiding the moving stator
by rolls. In Fig. 3, the picture represents the track mounted
on the wall of a high power laboratory at EPFL. Steep and
reduced power climbing segments simulate conditions
similar to a roller coaster, allowing high accelerations and
fast movements but also stand-still under load conditions,
as test-conditions for the control performances.
Figure 3 : Rail of the LIM (rotor)
DSP SHARC
XILINX
User commands Interface
board Inverter
motor
Figure 4 : Control system
2.4 Control system
The control system is made of a floating-point
architecture Digital Signal Processor, which facilitates the
practical implementation of functionality by students
without any needed experience in assembler programming.
The DSP is completed by devices like FPGA for digital
modulation of the PWM Inverter, and by fast AD
converters for fast real time data acquisition.
Because of its quickness, its large number of AD’s (14)
and its multiple graphical tools, this dedicated card
developed at EPFL/LEI is a very powerful and efficient
development tool [7]. Especially the easy up and download
facilities for parameter setting and representation of records
are particularly suited for education purpose.
2.4 Complete system
Fig. 5 shows the complete system. Depending on the
control strategy implemented, each of the sensors may or
not be used. But they are anyway useful to dimension all
the regulation parameters because they give interesting
informations on the system behaviour.
2.6 Characteristic curves
Fig. 6 shows the characteristic curves of the regulation.
They are the same for the both regulation strategies
described next. These curves characterise the motor when it
is operated with constant stator flux. To assure this, the
stator frequency must rise with the stator voltage [3]. Along
with these curves, the nominal stator voltage is set to
165[V] corresponding to the nominal stator frequency of
50[Hz].
Figure 5 : Complete system

Figure 6 : Characteristic curves of the LIM
The two important points on these curves are the nominal
point (wrn=80[rad/s], In=41[A]) and the breakdown point
(wrk=151[rad/s], Ik=52[A])). In comparison with classical
rotating induction motors, the values of the slip during
operation may be very high.
Another particularity is the value of the magnetising
current (wr=0) which is quite important (I0=29[A]).
3 Position regulation using a speed
sensor.
3.1 Regulation strategy :
Even if sophisticated control for the LIM is currently
under way [2], a simpler strategy is followed for the set up
of the test track, which is particularly easy to implement
and simple to be understood by younger college students
[3]. First the drive is regulated in speed, then an additional
loop for regulating the position will be added.
The external loop is a speed regulation. Depending on
the speed error, the regulation system sets the motor force
by acting on the slip frequency (block 9).
Figure 7 : Regulation strategy
The stator frequency is obtained by summing the desired
slip frequency and the measured mechanical speed
(converted in electrical Herz).
Each phase current, corresponding to the slip frequency
as in Fig. 6, is then set (block 8) and regulated with a PI
regulator (block 4). The instantaneous phase of each current
(block 6) is obtained by integrating the stator frequency
(block 7). The phase difference between the three phase
currents is 2p/3.
3.2 Realisation
Current regulation (block 4) is one of the keypoint of this
strategy. Especially when the frequency is high, the
regulation must be very efficient. The regulation sampling
period is set to 150[ms].
The output voltage command signal is represented on the
top of Fig. 8. If symmetric current is imposed, more voltage
is needed in the central phase. The reason is that the others
phases surround this one. This confers to it a greater self-
inductance.
In this case the stator frequency is 19[Hz]. As we can see
on the bottom of Fig. 8, the current regulation is almost
perfect. Because the current regulation is working well, we
can expect to use the drive efficiently, with good dynamic
performances.
Despite the size of the motor which weights about 100
[kg] and the steep slopes (up to 84[%]), we set the
acceleration/deceleration to 4.5[m/s2].
Even if the rail is short, as the motor can be slowed down
very well, a high set value of 3[m/s] can be chosen for the
constant speed.
Figure 8 : Command voltage and phase currents

Figure 9 : Speed profile
To be able to present impressive demonstrations like
stopping a few centimetres before the end of the rail, we
make sure that there will not be any overshot.
Fig. 9 shows the chosen speed profile depending on the
position error. It means that as long as the motor is far
enough from the desired position (>1.2[m]), the regulation
sets the speed reference to a constant value of 3[m/s]
towards the desired position. As soon as the motor is within
1.2 meter from the final position, the speed reference is
diminishing linearly with the position error.
Several strategies for the position regulation are
admissible. In fact the chosen strategy is a proportional
regulation with limitation. The main concern is that we do
not want any overshot. Because the system behaviour is
good an integrator is not added for the position regulation.
If it is done in the future, an antireset windup is mandatory
to avoid integrator saturation and overshoots.
Fig. 10 shows the results recorded on the real track.
Position 0 corresponds to the lowest part of the track in the
hollow of the rail in V (Fig. 3). The speed reference is
ramped. In fact, an infinite acceleration can of course not be
obtained. Ramping the reference avoids the integrator
regulating the slip frequency to saturate and disturb the
dynamic.
Both speed and position regulations are working well and
there is no position overshoot. The excellent current
regulation makes it possible to use the motor in an efficient
manner. The demonstrative skills of the drive are in
consequence.
Figure 10 : Speed and position regulation
4 Sensorless speed regulation
4.1 Control strategy:
Because the motor model is not accurate in every case,
speed estimation will not be perfect, especially during
transient behaviour. Consequently the same strategy than
above will not be used because its efficiency and its
stability depend on a precise measuring of the instantaneous
speed.
Rather than setting the currents, the regulation sets the
voltage and always imposes constant and rated stator flux.
This way the motor constantly has a maximum of force. To
achieve that, the magnitude of the phase voltage must vary
proportionally with the frequency [3]. The phase difference
between the 3 phase voltages is constant and equals 2p/3.
The regulation strategy is represented on Fig. 13.
The stator currents are measured to evaluate the field loss
due to the stator resistance. This loss is compensated by
acting on the voltage (designated by comp. Rs).
To obtain a given mechanical speed wm the slip wr is
estimated. Adding this two results in the necessary stator
frequency ( ωs =ω m +ω r ). This operation is designated by
comp. slip on Fig. 13. This is then a regulation with
positive feedback. We then must be cautious to avoid
destructive behaviour of the drive. In fact the currents have
to be kept under rated values in any case.
Because of the positive feedback, it is mandatory to
filter the slip estimation. Consequently, the user must not
expect immediate reaction of the regulation system in case
of load or speed transient.
4.2 Resistive Field losses compensation
A typical model of the asynchronous motor in steady
state (Fig. 11) is used. This representation does not to take
into account the difference between the three phases nor the
specific behaviours due to the linearity of the motor.
The complex representation of the vectors associated is
on Fig. 12.
Figure 11 : One phase electric scheme of the asynchronous motor.
Figure 12 : Complex representation of the vectors associated with
Figure 11.
 Figure 13 : Control strategy

Using Fig. 12, the magnitude of Em can be written: An estimation strategy based on this principle was
presented in [6]. The same principles are used here, with
2 2 2
E m =U s +(I s × R s ) -2 × U s × I s × R s × cos(Φ) (4.1) modifications.
A first relation between slip and torque is given in [8]:
2 × Tk
Let Em0 be the value of Em at rated stator frequency fsn. TL = (4.4a)
Because we would like a constant flux, the magnitude of Em s sk
+
must vary linearly with the stator frequency (fs): sk s
f This relation is obtained by using the equivalent scheme
E m =E m0 × s (4.2) of Fig. 11. It is still accurate to describe a linear motor. In
fsn
this case, the ratio TL/Tk is switched to FL/Fk. The following
Substituting (4.2) in (4.1) yields the value of the phase formula is then valid for a linear motor:
voltage (Us): 2 × Fk
2 FL = (4.4b)
æ E m0 × fs ö 2
s sk
+
Us =Is × R s × cos(Φ)+ ç ÷ -(Is × R s × sin(Φ)) (4.3)
f sk s
è sn ø
As visible on figure 13, only the compensation is filtered. If s is replaced by wr/ ws a second order equation which
In fact, the filtering is only applied to (4.3) minus (4.2). relates slip pulsation and load force can be obtained, where
Then (4.2) is added again after the filtering. wrk [rad/s] is the breakdown slip pulsation and Fk the
breakdown force.
4.2 Slip compensation Fk ± Fk2 -FL2
ωr12 = × ωrk (4.5)
FL
Because the speed is not measured, the slip frequency
Only one of the two solutions has a physical meaning. The
can not be precisely known. The following strategy
sign preceding the square root has to be determined.
estimates the slip by measuring only the currents.
Simulations showed that the sign should always be
negative. In other hand, this formula for wr might cause
Let:
some numerical problems. In fact, when working at low
ws = stator voltage pulsation [Hz]
stator frequency (FL»0) there might be a division by zero,
ws = slip pulsation [Hz]
which is a cause of failure for the calculator.
s = fr/fs = relative slip
In formula (4.5), if s1 is a solution, the s2 = sk2/s1 is also a
sk = breakdown slip
solution. The solution with positive sign corresponds to
TL = load torque [Nm]
that. Another expression for w r is then:
Tk = breakdown torque
Ps = air gap power [W] FL
ωr = × ωrk (4.6)
p = number of poles pairs [-] Fk + Fk2 -FL2
Wm = mechanical speed [rad/s] This solution can be used as will in the control algorithm,
because there is no indetermination (FL £ Fk).
 The electromagnetic force FL can be evaluated by In the motor case, Ps can then be written:
considering the air gap power. The relation between 2
U
mechanical power, speed and force is reported underneath. P = P -P -P = 3 × U × I × cos(j)-3 × R × I2 - m (4.12)
δ el cus m s s s s R
P m
FL = mec (4.7) with U = U - (R +j × ω × L ) × I
vm m s s s fs s
For an asynchronous motor, mechanical power is related
to air gap power as follow (Fig. 14): By looking at (4.3), (4.6), (4.10) and (4.12), it appears
that the only necessary machine parameters are Rs, Rm, Lfs,
Pmec =Pd × (1 - s ) (4.8) Fk and wrk.
If the parameters of the motor are unknown, the model
In other hand, the mechanical speed of the motor can be can be simplified by choosing Rm=0, which is equivalent to
expressed using the stator pulsation and the slip. not consider any iron losses. In this case Pm=0. The
τ p × ωs parameters still needed for the estimation algorithm are
vm = × (1-s) (4.9) then Rs, Fk et wrk. These three parameters can be directly
π
measured. Of course the estimation is then less precise, but
Replacing (4.7) and (4.8) into (4.9) yields a new
still gives good results.
expression force the electromagnetic force (tp = pole length,
p = number of pole pairs):
4.3 Results and comments
π × Pδ
FL = (4.10) First the speed response in three different cases is
τ p × ωs showed. At the beginning of each test, the motor is in the
hollow of the rail. Then the speed reference is set to a
This few formulas are directly deductible from the constant value (successively 2.46, 1.64 and 0.82 [m/s]).
formulas concerning a rotating motor by using the When the motor has travelled over 8 meters, the reference
following transformation: is set back to zero, so that the motor stands still on the steep
π rail.
p® (4.11)
τp
The air gap power has to be evaluated. Underneath is the
representation of the power flux in the motor (Fig. 14).

With:

PEL [W] = Stator electric Power

PCUS [W] = Stator resistive losses
Pm [W] = Iron losses
PCUR [W] = Rotor resistive losses
Ps [W] = Air gap power
PMEC [W] = mechanical power
PF+V [W] = friction + fanning losses
PSHAFT [W] = Power at the shaft

Figure 14 : Power flux in the motor Figure 15 : Speed regulation

 The speed results are presented in [rad/s] (Fig. 15). The
upper value of the speed reference is also indicated in [m/s].
The first thing to mention is that the resistant force acting
on the drive is not constant. At the beginning the gravity
force is null and raises rapidly to a constant value (because
of the constant slope). But the resistant force due to the
supply cable is variable and depends much of the position
and speed.
Because the resistant force is highly variable, there are
some speed variations even when the reference is constant.
These variations are in this case quite important. The reason
is that the motor is never in steady state. But it is very
pleasing that despite these hard test conditions the system
behaves in a very stable and proper manner.
Also the measured speed is not exactly equal to the
reference. As mentioned earlier, the model of the motor is
not perfect, which means that identifying the mechanical
speed by measuring only electrical quantities will not be
accurate in every operating case. The regulation is here
optimised to ensure that the motor can be totally stopped,
which is the more important in the case of a sensorless
drive. In fact, a sensorless drive with 90[%] of speed
precision might well be good enough for most applications,
but 99[%] of speed precision could be useless if it is
impossible to stop the motor.
To show the behaviour of the motor in more general
case, the speed reference is varied with a potentiometer.
Fig. 16 shows that when the reference is not varied to
quickly and for lower commanded speeds, the speed
response is quite good, with only little oscillations.
For the same reasons as mentioned before the regulation
delay and a slight offset can still be observed. But despite
that, the precision is sufficient for a large number of
applications.
Figure 16 : Reference and measured speed
4.4 Synthesis
The thoughts and measures above make it possible to
draw the following conclusions concerning a sensorless
regulation:
Speed estimation and regulation algorithm are stable.
This makes it possible to guaranty a stable behaviour in any
operating case.
Most of the speed oscillations are due to the varying
force due to the supply cable. In the future, it will be
avoided by using an integrated supply from on board
batteries.
This strategy does not allow very good dynamic
behaviour. Indeed the filtering of the speed estimation
slows down the regulation response.
Speed estimation depends strongly of the rail
temperature. Indeed the slip estimation depends strongly of
the rotor resistance (Rr). If the rail temperature raises, the
rotor resistance raises too. Consequently the breakdown slip
(wrk) raises too (linearly with Rr). Because the strategy
directly uses wrk to estimate the slip, it is quite sensitive to
the variations of Rr.
Because the rotor (rail) is linear, a constant rotor
resistance for every position can not be expected. In case a
very precise regulation is demanded, a way of evaluating Rr
has to be implemented. Some methods have been presented
[13]-[15], they usually use the difference between the
measured current and the modelled current.
Evaluating Rr will cause the strategy to be more complex
and conduct to more precise modelling which would reduce
the robustness. In the case of a sensorless drive, if a slight
offset and regulation delay is tolerable, it’s better not to
complicate the strategy.
In other hand, the proposed regulation suits well the
demonstration purposes. The motor can accelerate, run at
steady speed and break in a very proper manner. It is able to
climb steep segment and completely stop on them. The
electrical and mechanical phenomenons can be observed
using the control board facilities, making it possible to
show the particular behaviours of the LIM and to compare
it with conventional rotating motors.
5 Conclusion
A drive using a linear asynchronous motor has been
realised. First the system is regulated in a “classical” way
by using a speed sensor. The current regulation is excellent
and allows use to use the drive in a very demonstrative
manner.
In the next step, the same motor is regulated without any
speed sensor. The regulation reacts well at any change of
reference or resistant force and the system is stable. But the
speed estimation is not perfect. Indeed the varying electrical
characteristics of the rail with temperature and position
have not been taken into account. In the case that a high
precision is demanded, it will have to be done.
It is important to remind that the first regulation (with
sensor) is totally unstable if the sensor breakdowns. In case
it is desired to guaranty a greater availability than dynamic
precision, a sensorless regulation might be more indicated.
Also in case very high dynamic performances are
desired, a better model will be needed in order to take care
of the characteristic properties of linear motors.
 In other hand, the drive can be well used as a
demonstrative feature. The graphical procedures included in
the control system allow showing easily the electrical and
mechanical phenomenons occurring in the system. The two
regulation strategies make it possible to show the
advantages/inconvenients of a sensorless control, since the
performances of the drive in each case can be easily
compared.
Today, spectacular demos can be achieved for visitors
and during information seminars for young college
students. Actually and in the future, visiting students or
other verification of research can also use the demo test
track.
6 Bibliography
[1] E=TeM2, “Tomorrow’s education in electrical
technologies: revisited methods & tools for reneved
motivation”, Liège, Belgium 14-16 March 2001.
[2] Morizane T., Rufer A., Tanigucci K., “Sensorless
control of linear induction motor considering its
asymmetric parameter”, MAGLEV 2000, 16th
international conference on magnetically levitated
systems and drives, Rio de Janeiro, Brasil.
[11] F. Gardiol, Electromagnétisme, traité d’électricité
(Vol. III), presses polytechniques et universitaires
romandes, Lausanne, Suisse, 1996.
[12] C. Debrunner, ”Propulsion par moteur linéaire
asynchrone”, travail de diplôme, EPF-Lausanne,
Suisse, 1991.
[13] G. Garcia Soto, E. Mendes & A. Razec,”Adaptative
variable structure rotor flux observer for an induction
motor”, Conference publication N°456, IEE, 1998.
[14] T. -H. Chin, I.Miyashita & T.Koga, “Sensorless
induction motor drive: an innovative component for
advanced motion control”, control. Eng Practice,
Vol5, N°12,pp. 1653-1659, Elsevier Science Ltd,
1997.
[15] S.A. Shirsavar & M.D. McCulloch, ”Speed sensorless
vector control of induction motors with parameters
estimation”, Conference publication N°429, IEE,
1996.

[3] H. Bühler, Convertisseurs statiques, Presse

Polytechniques et universitaires romandes, Lausanne,
Suisse, 1991.

[4] N. Wavre, “Etudes harmoniques tridimensionnelles

des moteurs linéaires asynchrones à bobinages
polyphasés quelconques“, Thèse EPFL, Lausanne,
Suisse, 1975.

[5] T. Seki, ”The development of HSST-100L 14th

International Conference on magnetically Levitated
systems, Hotel Maritim, Bremen, 1995.

[6] A. Munoz-Garcia, T.A. Lipo & D.W. Novotny, “A

new induction motor open-loop speed control capable
of low frequency operation”, IEEE Industry
Applications Society, New Orleans, Louisiana, USA,
1997

[7] http://leiwww.epfl.ch/sharc/sld003.htm

[8] M. Jufer, Electromécanique, traité d’électricité (Vol.

IX), presses polytechniques et universitaires
romandes, Lausanne, Suisse, 1995.

[9] H. Bühler, Théorie du réglage de l’électronique de

puissance, notes de cours EPF-Lausanne, Suisse,
1994.

[10] B. Kawkabani, Contrôle vectoriel des machines

asynchrones, notes de cours, EPF-Lausanne, Suisse,
1995.