You are on page 1of 6

SPEEDAM 2006

International Symposium on Power Electronics,


Electrical Drives, Automation and Motion

Control design aspects for an


EDS levitation system
M. Andriollo*, S. Carmeli**, F. Castelli Dezza**, M. Mauri***
* Department of Electrical Engineering, University of Padova, Via Gradenigo 6/a, 35131 Padova, (Italy)
** Electrotecnical Department, Politecnico di Milano, P.zza Leonardo da Vinci 32, 20132 Milano, (Italy)
*** Mechanical Engineering Department, Politecnico di Milano, Via La Masa 34 20156 Milano, (Italy)

Abstract—EDS systems based on high frequency supplied z


windings are convenient candidates in standstill and low Rm Rm
motion applications where strict mechanical tolerances are a
not practicable. In this paper a simple EDS system, based on real coil
a single levitation coil is presented. Images approach has b
been used to tune up a simplified model of the levitation coil
in order to design levitation control. Simulations and h
conducting slab y
preliminary experimental results are presented in order to
x
validate the EDS system model and control. h

Index Terms—EDS, magnetic levitation, image modeling.


image coil
I. INTRODUCTION Fig. 2. Coil sizes (a, b: cross section width and height, Rm: mean coil
radius)
In some industrial application magnetic levitation may be z
a desirable option, provided that it is achieved with a
relatively inexpensive apparatus, using robust control
system and when mechanical tolerances are not too dz
severe. To such purpose, EDS systems based on high dz
frequency supplied windings can be convenient h
conducting slab y
candidates, especially in standstill and low motion
x
applications, since levitation is independent from the h
relative motion between the load and the active or passive dz
guideway. In its simplest configuration, an EDS image coil
levitation system is represented by a coil faced to an dz

horizontal slab with high electrical conductivity V and


Fig. 3. Application of the virtual work principle related to a vertical
magnetic permeability ȝ§ȝ0 (Fig.1). displacement dz to evaluate the levitation force Fz

z The levitation force can be evaluated applying the virtual


real coil work principle: a vertical coil displacement dz results in
an opposite displacement –dz of the image coil, so the
relative displacement is 2dz (Figure 3) and the variation
h of magnetic energy can be obtained as twice the
difference between the values of the surface integral of
h y the tangential component Aș(r,z) of the magnetic potential
vector generated by the image coil on the upper and on
the lower face of the coil.
x Many analytical and pseudo-analytical formulations
image coil
conducting slab related to such configuration can be retrieved in literature,
Fig. 1. Parallel perspective of the coil and of its mirror image (h: mainly concerned with the evaluation of coil self and
levitation height) mutual inductances. The electrodynamics effects can be
derived from such formulations taking the derivative of
Supplying the coil with an alternating current of the magnetic energy with respect to the displacement
relatively high frequency f and provided that the variable or by the direct integration of the expression of
conducting plate thickness is larger than the skin depth Lorentz’ force all over the coils volume.
į=1/ SfPV, the electrodynamic effect can be reproduced The levitation force can be therefore expressed in per unit
by replacing the plate itself with a fictitious “image” coil value, assuming the levitation height h as the base
placed symmetrically to the real coil with respect to the quantity ( * superscript denotes p.u. value), as indicate in
plate surface (Figure 2). (1):

1-4244-0194-1/06/$20.00 ©2006 IEEE


S11 - 17

Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on October 09,2021 at 20:41:12 UTC from IEEE Xplore. Restrictions apply.
a* di d[ dL([ )
Rc i (t )  L [
*
Rm 
2
§ *§ * *
· § * ·
· v(t )  i (t ) (4)
Fz h 2 4 ¨ AT ¨ r ,2  3b ¸  AT* ¨ r * ,2  b ¸ ¸r *dr * (1) dt dt d[
2J ˜ h ³ ¨ ¨
© 2 ¹¸ ¨
© 2 ¸¹ ¸¹
* a* ©
Rm 
2 evidencing the coil current i:
with
a* b*
*
Rm  t
S 2 2
U * cos T ˜ dz * ˜ dU * ˜ dT 1 § § d[ dL([ ) · ·
*
AT r , z * *
³ ³ ³
0 r * *
*
U  z
* 2 *
] * 2
* *
 2r U cos T
(2) i (t ) i (0) 
L([ ) ³
0
¨ v(t )  ¨¨ Rc 
¨
© ©
¸i (t ) ¸dt
dt d[ ¸¹ ¸¹
(5)
* a b
Rm  
2 2

where Rc is the coil resistance, taking into account


and J is the current density. (1) shows that the force the shield losses due to the induced currents but
depends on the coil extension raised to the fourth power, neglecting variation effects due to the vertical
according to a well-known law scale [1]. displacement. Eq. (5) can be numerically integrated
since the expression of the inductance derivative can
II. LEVITATING COIL MODEL be straightforwardly obtained from (3).
x expression of the levitation force f(t) given by:
A. Simulink model
In order to design and test the control system, a 1 dL([ ) 2
SimulinkTM simulation of levitating coil has been f (t ) i (t ) (6)
2 d[
implemented according to the following equations
(Fig.4):
x dependence of the coil inductance on the vertical B. Stability analysis
displacement [=z–z0 with respect to the reference Neglecting the aerodynamic drag due to the low coil
levitation height z0, defined using a simplified speed, the equilibrium point stability is analysed
exponential expression [1]: according to the mechanical dynamic equation related to
the vertical motion:
[

L [ Lf  L0  Lf e J
(3) d 2[
f (t )  mg m (7)
dt 2
with J characteristic height and L’ and L0 inductance
values related to the actual coil self inductance (i.e., Assuming little variations '[ of the vertical displacement
unaffected by the shield effect) and to the reference
and 'I of the r.m.s. current value around the steady-state
levitation height z0, respectively. These values are
obtained by fitting the results obtained by the above equilibrium point {0, I0= 2mgJ / Lf  L0 } [2], (7) can
presented analytical approach or by a sequence of be linearized as:
FEM analyses.
x representation of the electrical behavior of the coil '[ D'I  E'[ (8)
controlled current source depending on the applied
voltage v, the vertical displacement [ and the height
where, according to (3) and (6), D, E, K can be expressed
rate d[/dt rearranging the voltage equation:
as:

+ v –

d[/dt
× ––+ ×
1/s i
y 1/s

Rcc
R
[
 n
L [ Lf  L0 Lf e J

2
[ u
u2 i
[ ×
 0.5
1/2 f
L' [ k L0 Lf e J

Fig. 4. Model block diagram

S11 - 18

Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on October 09,2021 at 20:41:12 UTC from IEEE Xplore. Restrictions apply.
2 g ( Lf  L0 ) g
D E (9)
mJ J

Considering the Laplace transforms X(s) and I(s) of '[


and 'I, respectively, (8) can be rearranged to define the
transfer function linking the vertical displacement to the
current variation:

X (s) D
F (s) 2
(10)
I ( s) s E

Since F(s) has only imaginary poles, the steady-state


equilibrium point {0, I0} is not asymptotically stable. Fig.7.Coil position vs velocity
Therefore, an external damper system (mechanical or
electrical [3]) is necessary to prevent unwanted
oscillation around the equilibrium point.

III. OPEN LOOP SIMULATION RESULTS


In order to verify the theoretical considerations, some
numerical simulations were performed assuming an open-
loop arrangement of the supply system according to the
electric circuit diagram of Fig.5 and carried out in the
experimental set-up (data of Tab.I and II). Figure 6 shows
the levitation height vs. time: the undamped oscillations
around the equilibrium height confirm the stability
analysis. The persistence of the oscillations is confirmed
by the progress of the velocity vs. position (Fig.7). The
main benefit of this circuit arrangement is represented by
the low value of the ratio inverter current/coil current, Fig. 8. Inverter and coil current
confirmed by Fig. 8.
IV. EXPERIMENTAL RESULTS
variable frequency\
voltage converter An experimental set-up Preliminary was carried out for a
levitating coil preliminary feasibility test and allowed to check the
Ls model parameters and to verify the results of the open
loop simulations. The test configuration includes (see
Cd data of Table I):
x a square voltage single phase inverter with an open
loop voltage and frequency control;
Fig. 5. Circuit diagram of the supply system.
x an output filter inductance;
x a capacitor bank.
TABLE I
EXPERIMENTAL SET-UP PARAMETERS
Description Value
Inverter frequency range 0 – 5 kHz

Inverter voltage range 0 – 400 V

Inverter rated current 20 A

Filter Inductance Ls 0.512 mH

Bank capacitance Cd 77 PF

As concerns the coil, its parameters are reported in Table


II (inductance values obtained by a sequence of FEM
Fig. 6. Levitation height
analyses).

S11 - 19

Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on October 09,2021 at 20:41:12 UTC from IEEE Xplore. Restrictions apply.
TABLE II resonance frequency: in fact, when the resonance
COIL PARAMETERS
condition is achieved, the LC voltage is in phase with
Symbol Description Value
the inverter current;
n Number of turns 52 - the levitation height and coil current are managed by
two PID controllers, to keep the current at the value
m Coil mass 2.82 kg
necessary to balance the coil weight, at the actual
J Characteristic height 17.3 mm height (the contribution of the height PID controller is
limited to allow slight parameters mismatching);
R 20°C mean resistance 0.252 : - the feed forward terms are introduced to improve the
control response.
L0 Coil inductance at z=z0=20 mm 0.345 mH

Lf Coil inductance at z=f 0.436 mH

Rm Mean coil radius 70 mm

a Coil cross-section width 40 mm


Force
b Coil cross-section height 16 mm
Gauge
Figures 9a-b shows the coil under test suspended by
nylon wires to a force gauge allowing to measure the
difference between the coil weight and the levitation
force to verify the achievement of the full levitation.
Different values of levitation height were tested. In
particular, for a levitation height z0 of about 20 mm, the
resonance condition, which allowed to obtain the
levitation consistently with the inverter rated current (20
A), was achieved for a frequency of about 1.1 kHz.
Under this operating conditions, the current peak value in
the LC loop is nearly 120 A. Experimental results have
confirmed the presence of undamped oscillations around a)
the steady-state equilibrium point.

V. THE CONTROL SYSTEM Nylon


Wire
To improve the system stability, eliminating the
persistent oscillations, a closed-loop control architecture
is devised (Fig.10), based on the following features:
- an estimator of the levitation height based on the coil b)
inductance value is implemented, allowing to spare a
height sensor; Fig. 9. Experimental coil: a) Force gauge b) wire details
- a factor displacement regulator is used to follow the

f ref
coil mass I ref I base V ref
L
h ref estimation estimation

+ V co mp +
+ PID + V ref
h ref PI Reg
_ Reg I co mp _ +

I
f ref
L h est
V LC L estimation h(L) V LC Factor
I f comp f ref
I inv displacement
f ref regulator + +
f base
sqrt(1/4S²CL)

Fig. 10. The proposed control architecture (on the left the estimator; on the right the compensator)

S11 - 20

Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on October 09,2021 at 20:41:12 UTC from IEEE Xplore. Restrictions apply.
Figure 13 shows the evolution of the levitation height
VI. CLOSED LOOP SIMULATIONS imposing a linear variation of the reference height from
Some simulations were performed to test the control 20 mm to 30 mm. Afterwards, for t=0.5 s, a sudden mass
effectiveness and to evaluate its performances. increment of 1 kg is simulated. The effectiveness of the
Figure 11 refers to a closed loop simulation starting from control system response is confirmed for both the
the reference height of 20 mm with a null coil current and operative conditions.
a 200 V capacitor voltage. It is possible to see that the The influence of the compensating term, allowing to
undamped oscillations are absent. The progress of the balance the mass variation, is evidenced by Figure 14: the
velocity vs. position (Fig.12) confirms the achievement current reference follows the height reference, while the
of the asymptotic stability. compensating current, ruled by the height controller,
allows to compensate the mass increment.

Fig. 11. Levitation height with a closed loop control.


Fig. 14. Current reference and compensating term (peak values)

The resonance follower allows to limit the inverter


current during the overall transient; its effectiveness is
shown by Figure 14, related to the final transient phase: it
can be recognized the low value of the inverter current in
comparison with the coil current, even with a relevant
height variation and mass increase.

coil current

inverter current

Fig. 12. Position vs. velocity

Fig. 14. . Inverter and coil current after the mass increment.

VII. CONCLUSIONS
The presented preliminary analysis has been
underlined that it is possible to obtain a simple magnetic
levitation system using a coil and an industrial supply
converter. The presence of a resonant LC stage allows to
supply the required high coil current keeping the inverter
output current under its rated value.
Fig. 12. Levitation height with a mass variation.

S11 - 21

Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on October 09,2021 at 20:41:12 UTC from IEEE Xplore. Restrictions apply.
A simplified model of the system, obtained by a IEEE Transactions on Magnetics, vol. 32. no. 5, September
suitable fitting of the FEM results, has been presented, 1996, pp. 5055-5057
allowing a fast simulation of the whole system and [3] B.T. Ooi and M.H. Banaka, Passive and Active Damper
Winding for the Repulsive Magnetic Levitation System,
therefore simplifying control design. A preliminary
IEEE Transactions on Magnetics, vol. MAG-13, no. 5,
stability analysis, confirmed by the open loop September 1977, pp. 1418-1420.
simulations, evidenced the presence of undamped [4] H.Kifune, T.Yamaguchi, D.Yoshida, Y.Hatanaka,
oscillations (steady-state equilibrium point without M.Nakaoka, Novel load adaptive frequency tracking
asymptotic stability). In order to avoid mechanical or control scheme for high frequency inverter without PLL
electromagnetic dampers, a sensorless closed loop control scheme, PESC04, 35th Annual IEEE Power Electronics
architecture is presented, allowing to eliminate Specialists Conference, Aachen, Germany, June 20-25,
oscillations, to control the levitation height and to 2004
compensate mass variations. Moreover, a frequency [5] H.Kifune, Y.Hatanaka, M.Nakaoka, Cost effective phase
resonance follower allows to limit the inverter current shifted pulse modulation soft switching high frequency
inverter for induction heating applications, PESC04, IEEE
with respect to the coil current. The simulation results Proceedings on Electric Power Applications, Vol.151,
confirmed the effectiveness of the control strategy. No.1, pp.19-25, January 2004
[6] N.S.Bayindir, O.Kukrer, M.Yakup DSP-based PLL-
controlled 50-100 kHz 20 kW high frequency induction
REFERENCES heating system for surface hardening and welding
applications, IEEE Proceedings on Electric Power
Applications,Vol.150, No.3, pp.365-371, May 2003
[1] Marc T. Thompson: Electrodynamic magnetic suspension
[7] .Kaczmarczyc, B.Grzesik, 800W, 16 MHz Laboratory class
– Models, scaling laws and experimental results, IEEE
E inverter, 10° European conference on Power Electronics
Trans. on Education, vol. 43, n. 3, August 2000, pp. 336-
and applications, Toulouse, France, September 2003
341
[8] M.S.Carmeli, F.Castelli Dezza, G.Superti Furga, Smart
[2] Benito Brunelli, Domenico Casadei, Giovanni Serra,
modulation: a new approach to power converter control,
Angelo Tani, Active Damping Control for Electrodynamic
EPE01, 9th European Conference on Power Electronics
Suspension Systems without Mechanical Transducers,
and Applications, Graz, Austria, August 27-29, 2001

S11 - 22

Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on October 09,2021 at 20:41:12 UTC from IEEE Xplore. Restrictions apply.

You might also like