Sie sind auf Seite 1von 6

Proceedings of the 18th World Congress

The International Federation of Automatic Control


Milano (Italy) August 28 - September 2, 2011

Advanced control for the XY-table driven by piezo-actuators


Xinkai Chen*, Ying Feng** and Chun-Yi Su***
*Department of Electronic and Information Systems, Shibaura Institute of Technology,
Saitama 337-8570, Japan (Tel: 81-48-687-5805; e-mail: chen@shibaura-it.ac.jp)
**College of Automation Science and Engineering, South China University of Technology,
Tianhe, Guangzhou 510641, China (e-mail: zhdfengying@gmail.com)
***Department of Mechanical and Industrial Engineering, Concordia University,
Montreal, Quebec, H3G 1M8 Canada (e-mail: cysu@alcor.concordia.ca)
Abstract: The XY table is composed of two piezo electric actuators (PEA) and a positioning mechanism
(PM). Due to existence of hysteretic nonlinearity in the PEA and the friction in the PM, the high precision
control for the XY table is a challenging task. This paper discusses the high precision adaptive control for
the XY talbe, where the hysteresis is described by Prandtl-Ishlinskii model. The proposed control law
ensures the global stability of the controlled stage, and the position error can be controlled to approach to
zero asymptotically. Experimental results show the effectiveness of the proposed method.

1. INTRODUCTION
Recently, piezo-actuated stage has many effective
applications in ultra-high precision positioning systems
(Chen and Hisayama, 2008; Fleming and Moheimani, 2007;
Krejci and Kuhnen, 2001; Kuhnen and Krejci, 2009;
Moheimani and Goodwin, 2001; Shieh and Hsu, 2007). The
piezo electric actuator (PEA) is used to meet the requirement
of nanometer resolution in displacement, high stiffness and
rapid response. However, the main disadvantage of PEA is
the hysteresis phenomenon between the applied electric
voltage and the displacement. Due to the undifferentiable and
nonmemoryless character of the hysteresis, it causes position
errors which limit the operating speed and precision of the
PEA. The development of control techniques to mitigate the
effects of hysteresis has been studied for decades and has
recently re-attracted significant attention (Moheimani and
Goodwin, 2001). Interest in studying dynamic systems with
actuator hysteresis is motivated by the fact that they are
nonlinear system with nonsmooth nonlinearities for which
traditional control methods are insufficient and thus require
development of alternate effective approaches. Development
of a control frame for the piezo-actuated stage is quite a
challenging task.
About the challenge of controlling the piezo-actuated stage,
the thorough characterization of the hysteresis forms the
foremost task (Brokate and Sprekels, 1996). Appropriate
hysteresis models may then be applied to describe the
nonsmooth nonlinearities for their potential usage in
formulating the control algorithms. It is reported in the
authors previous work (Chen and Hisayama, 2008) that the
Prandtl-Ishlinskii (PI) model can describe the hysteretic
nonlinear relation between the applied electric voltage and
the displacement in the piezo electric actuator. The basic idea
of PI hysteresis model consists of the weighted aggregate
effect of all possible so-called elementary hysteresis
978-3-902661-93-7/11/$20.00 2011 IFAC

operators. Elementary hysteresis operators are noncomplex


hysteretic nonlinearities with a simple mathematical structure,
where the stop operator which is parameterized by a single
threshold variable is employed (Brokate and Sprekels, 1996)
In order to effectively drive the piezo-actuated stage, it is by
nature to seek means to fuse the PI hysteresis models with the
available control techniques to mitigate the effects of
hysteresis, especially when the hysteresis is unknown, which
is a typical case in many practical applications. However, the
results on the fusion of the available hysteresis models with
the available control techniques is surprisingly spare in the
literature (Chen and Hisayama, 2008; Tao and Kokotovic,
1995; Tan and Baras, 2004). The most common approach in
coping with hysteresis in the literature is to construct an
inverse operator, which is pioneered by Tao and Kokotovic
(1995), and the reader may refer to, for instance, Iyer et al,
(2005), Natale et al (2001), Tan and Baras (2004) and the
references therein. Essentially, the inversion problem
depends on the modeling methods of the hysteresis. Due to
multi-valued and non-smooth features of hysteresis, the
inversion always generates certain errors and possesses
strong sensitivity to the model parameters. These errors
directly make the stability analysis of the closed-loop system
very difficult except for certain special cases.
This paper tries to fuse of the adaptive control techniques
with the Prandtl-Ishlinskii hysteresis model for the XY-table,
which is composed of two PEAs and a positioning
mechanism. The advantage is that only the parameters in the
formulation of the controller need to be adaptively estimated,
and the real values of the parameters of the stage need to be
neither identified nor measured. The proposed control law
ensures the global stability of the adaptive system, and the
position error of the XY-table can be controlled to approach
to zero asymptotically. Experimental results confirm the
effectiveness of the proposed method.

5255

10.3182/20110828-6-IT-1002.01750

18th IFAC World Congress (IFAC'11)


Milano (Italy) August 28 - September 2, 2011

2. PROBLEM STATEMENT

Lemma 1. If T << 0.1 , then a (q 1 ) and b(q 1 ) are coprime,

2.1 System Description


In this paper, the XY-table is composed of two peizoactuators and a positioning mechanism (PM). The scheme of
the XY table is shown in Figure 1. First, let us consider one
direction of the table, say y-axis. The positioning mechanism
can be modeled as a mass-spring-damper mechanic system.
The PEA can be regarded as a force generator which
generates force due to the applied voltage. The dynamic
equation of the piezo-actuated stage in y-axis can be
formulated as follows
m y y(t ) + y1 y (t ) + y 2 y (t ) = u y (t ) ,
(1)

b(q 1 ) is a Hurwitz polynomial, and b0 > 0 .


Proof. By using the fact that y 2 >> y1 , the lemma can be
easily proved by some elementary calculations.
Now, consider the discrete time model of the XY-table. By
referring (3) and considering the interference between the
two axes, the model of the XY-table can be expressed by the
following relation

A(q 1 ) z (k ) = q 1 B(q 1 )u (k ) ,
z (k ) = [ x(k ), y (k )] ,
T

u (k ) = [u x (k ), u y (k )]
with

A(q 1 ) = 1 + A1 q 1 + A2 q 2 ,
1

B (q ) = B0 + B1 q ,
(i )
a (i ) a12
(for i = 1,2 ),
Ai = 11
i)
i)
(
(
a 21 a 22
b (0 ) 0
B0 = 11 (0 ) ,
0 b22
(1)
b (1) b12
B1 = 11
(1) (1) ,
b21 b22

Probe

(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)

where x(k ) represents the x-axis displacement of the table,


Fig. 1 The experimental setup.
where y (t ) represents the y-axis displacement of the table,

u y (t ) is the force generated by the PEA in y-axis; m y is the


mass of the y-axis mover, y1 is the viscous friction
coefficient (which is very small) of the PM in y-axis and y 2
is stiffness factor satisfying y 2 >> y1 .

y (t ) 0
+ 1

y (t )
m y
my

y1

express the relation between v x (k ) and u x (k ) and the


relation between v y (k ) and u y (k ) as

u x (k ) = H x [v x ](k ) , u y (k ) = H y [v y ](k ) ,

(14)

where H x [] and H y [] are the hysteresis operators which


will be given later.

Now, express (1) as

0
d y (t ) y 2

=
dt y (t ) m

u x (k ) is the force generated by the PEA in x-axis. Let v x (k )


and v y (k ) be the voltages applied to the actuators, and

u y (t ) .

The control purpose is to drive the position


z (k ) = [ x(k ), y (k )]T of the XY-table to track a uniformly
(2)

bounded signal z d (k ) = [ x d (k ), y d (k )]T for the system (6)


together with (14).

Let T be the sampling period and suppose u y (t ) is constant

Remark 1: It should be noted that the signal u (k ) is not


available.

during the sampling instants. By discretizing system (2)


based on ZOH input, the input-output discrete time
expression of system (1) can be given by
(3)
a (q 1 ) y (k ) = q 1b(q 1 )u y (k ) ,
where q 1 is the delay operator, a (q 1 ) and b(q 1 ) are
polynomials defined by
(4)
a (q 1 ) = 1 + a1 q 1 + a 2 q 2 ,

b(q 1 ) = b0 + b1 q 1 .
The parameters a1 , a 2 , b0 and b1 are unknown.

(5)

2.2 Hysteresis Model


In this paper, we adopt the Prandtl-Ishlinskii (PI) model in
discrete time. The hysteresis is denoted by the operator
w(k ) = H [v](k ) , where v(k ) is the input (voltage), w(k ) is
the output (generated force) of the PEA. The basic element of
the PI operator is the so-called stop operator
(k ) = Er [v; w1 ](k ) with threshold r. For arbitrary piecewise monotone function v(k ) , define er : R R as
er (v) = min(r , max(r , v)) .

5256

(15)

18th IFAC World Congress (IFAC'11)


Milano (Italy) August 28 - September 2, 2011

For any initial value w1 R and r 0 , the stop operator


E r [ ; w1 ](k ) is defined as

E r [v; w1 ](0) = er (v(0) w1 ) ,

(16)

E r [v; w1 ](k ) = er (v(k ) v(k i ) + E r [v; w1 ](k i )) , (17)


for k i < k k i +1 , where the function v(k ) is monotone for
k i k k i +1 . The stop operator is mainly characterized by
the threshold parameter r 0 which determines the height of
the hysteresis region in the (v, w) plane. For simplicity,
denote E r [v; w1 ](k ) by E r [v](k ) in the following of this
paper. It should be noted that the stop operator E r [v](k ) is
rate-independent. The PI hysteresis model is defined by
R

w(k ) = p (r ) Er [v](k )dr ,


0

(18)

3. ADAPTIVE CONTROL DESIGN


3.1 Some Preliminaries
To begin with, define the variable
s (k ) = C (q 1 )(z (k ) zd (k ) ) ,
where C (q ) is in the form

C (q 1 )
0
1
2
C (q 1 ) = 1
= I + C1 q + C 2 q , (20)
C 2 (q 1 )
0

C1 (q 1 ) and C2 (q 1 ) are Schur polynomials, and


c (i )
C i = 11
0

satisfying p (r ) 0 with

rp (r )dr < , R is a positive

constant which is sufficient large.


For the piezo electric actuator, the constant R depends on the
saturation input voltage. Figure 2 shows the relation between
v(k ) and w(k ) given by model (18).

i = 1, 2

for

0
(i )
c 22

Clearly,

lim (z (k ) zd (k ) ) = 0 .

where p (r ) is the density function which is usually unknown,

(19)

(21)

lim s (k ) = 0

implies

Now, consider the polynomial matrix equation

C (q 1 ) = A(q 1 ) + q 1 F (q 1 ) ,

(22)

where F (q ) is in the following form

F (q 1 ) = F0 + F1 q 1 ,

(23)

F0 and F1 are 2 2 matrices. Thus, if the parameters A1


and A2 are known, then the parameters in F (q 1 ) can be
determined uniquely.
Multiplying (22) with z(k) and employing (6) gives
(24)
C (q 1 ) z (k + 1) = B(q 1 )u (k ) + F (q 1 ) z (k ) ,
where u (k ) is the force generated by the PEAs defined as

u (k ) = [u x (k ), u y (k )]T

[ ]

Figure 3 shows the experimental relation between the input


voltage and the displacement for the PEA (PFT-1110, Nihon
Ceratec Corp.). The relation between the input voltage and
the output force is similar to that shown in Figure 3. It can be
seen that the PI model (18) indeed generates the hysteresis
curves and can be considered to be well-suited to describe the
hysteretic behavior in PEA.

R
R
= p x (r )E r [v x ](k )dr , p y (r )E r v y (k )dr , (25)
0

0
p x (r ) and p y (r ) are respectively the density functions of

Fig. 2 Hysteresis curves given by model (18).

of PEAs in x and y axes. Substituting (25) into (24) yields


px (r )Er [vx ](k )
R
dr
C (q 1 ) z (k + 1) = T (k ) + B0
0
p y (r )Er v y (k )

[ ]

p x (r )E r [v x ](k 1)
R
dr
+ B1
0
p y (r )E r v y (k 1)

[ ]

with

(k ) = z T (k ), z T (k 1) ,
= [F0 , F1 ]T .

(26)

(27)

(28)
Relation (26) will be used to formulate the adaptive control.
3.2 Parameter Estimation
Since the parameters A1 and A2 are unknown in practice,
Fig. 3 Relation between the input voltage and the
displacement of PEA.

the parameters in F (q 1 ) can not be obtained. Furthermore,


the density functions p x (r ) and p y (r ) are also unknown in

5257

18th IFAC World Congress (IFAC'11)


Milano (Italy) August 28 - September 2, 2011

practice. In the following, we will try to estimate the


unknown parameters. Let
T
(29)
(k ) = F0 (k ), F1 (k )
denote the estimate of the unknown parameter at the k -th
step and let
b (i ) (r , k ) b (i ) (r , k )
11
12

(30)
Bi (r , k ) =
(i )

(
)
i

b21 (r , k ) b22 (r , k )

lim (k ) (k ) = 0 ,

lim

k 0

(B (r, k ) B (r, k ))dr = 0


i

Proof. By considering the Lyapunov function candidate


T
L(k ) = trace (k ) (k )

)(

+ trace

i =0

(B (r, k ) B P(r )) (B (r, k ) B P(r ))dr


T

( 0)
with b12
(r , k ) = b21(0) (r , k ) = 0 be the estimate of Bi P(r ) for

and taking the difference of L(k ) , the lemma can be proved.

a fixed r at the k -th step, where P (r ) is defined as

3.3 Control Input Design

p x (r ) 0
(31)
P(r ) =
.
0 p y (r )
By observing (26) and replacing the parameters and
Bi P(r ) in the right hand side with their corresponding
estimates, the estimation error can be defined as
e(k ) = [e1 (k ), e 2 (k )]T
= C q 1 z (k ) T (k 1) (k 1)
1

R
B i (r , k 1)E r [v ](k 1 i )dr

Define
and

(32)

[ ] ]

E r [v ](k ) = E r [v x ](k ), E r v y (k ) T

(33)

E rT
0

i =0

(34)

B i (r , k ) and (k ) are updated by the following adaptation


laws with constraints

(k 1) e T (k )
1 + (k 1)

(35)

E [v ](k 2 ) e (k )
B1 (r , k ) = B1 (r , k 1) + r
1 + (k 1)

(r , k ) =

( 0)
b11

( 0)
b22

(r , k ) =

( 0)
b22

(36)

[ ]

E r v y (k 1) e 2 (k )
1 + (k 1)

(37)
.

(38)

The initial conditions should be chosen such that

bii(0 ) (r ,0 ) > 0 and rbii(0) (r , 0)dr < . The adaptation gain


0

should be chosen as 0 < < 2 .

(P2). lim

e(k + 1)

( 0)

(r , k ) E r [Vi* ](k )dr = Wi (k ) .

(40)

Let v1, min , v1, max and v 2, min , v 2, max be the corresponding
practical input ranges to the PEAs in x-axis and y-axis, and
define
R
(41)
W (k ) = b (0) (r , k ) E [v
](k )dr ,
i , sat

ii

i , max

R
W i , sat (k ) = bii(0) (r , k ) E r [v i , min ](k )dr .

(42)

R
Thus, it yields W i , sat (k ) bii(0) (r , k ) Er [ ](k )dr Wi , sat (k )

for any v i , min (k ) v i , max .


For simplicity, we give the algorithm of deriving Vi* (k ) for
the case i = 1 . Without loss of generality, suppose W1 (k ) is
monotonically increasing on k i < k k i +1 .
For each k i < k k i +1 , define new variables V (k ) and

W (k ) , where is a parameter varying in the range

Lemma 2. For the adaptation algorithm in (35)-(38), the


following properties hold.
R
(P1). (k ) and B i (r , k ) dr are bounded for all k > 0 .
0

0 bii

E [v ](k 1) e1 (k )
(r , k 1) + r x
1 + (k 1)

(r , k 1) +

(39)

where is defined as = 1
with 0 < i < 1 for
0

i = 1, 2 . It is obvious that W (k ) is an available signal at


instant k.

( 0)
b11

[v](k 1 i )E r [v](k 1 i )dr .

(k ) = (k 1) +

R
B1 (r , k ) E r [v ](k 1)dr

In the following, we try to derive a signal Vi* (k ) such that

(k 1) = T (k 1) (k 1)
1

w (k )
W (k ) = 1 = T (k ) (k )
w2 (k )

+ C (q 1 ) z d (k + 1) + s (k ) ,

( )

i =0

The control input is determined so that lim s (k ) = 0 . Define

2
2

=0
1 + (k )
(P3). For any positive finite integer ,

[0, v1, max v1, min )

V (k ) = V1* (k 1) + ,
R ( 0)
(r , k ) E r [V ](k )dr .
W (k ) = b11
0

If W1 (k ) > W1, sat (k ) , let V1* (k ) = v1, max ;


If W1 (k ) < W 1, sat (k ) , let V1* (k ) = v1, min ;

5258

(43)
(44)

18th IFAC World Congress (IFAC'11)


Milano (Italy) August 28 - September 2, 2011

If W 1, sat (k ) W1 (k ) W1, sat (k ) , the value of V1* (k ) is

Wi , sat (k ) for i = 1, 2 , then all the signals in the closed-loop

derived from the following algorithm.


Step 1: Let increase from 0.
Step 2: Calculate V (k ) and W (k ) . If W (k ) < W1 (k ) , then
let increase continuously and go to Step 2;
Otherwise, go to Step 3.
Step 3: Stop the increasing of , memorize it as 0 and
define V1* (k ) = V 0 (k ) .

remain bounded and lim e(k ) = 0 . Furthermore,


k

lim (z (k ) z d (k ) ) = 0 .

(52)

Proof. By applying the Key Technical Lemma (Lemma 6.2.1


in Goodwin and Sin (1984)) to property (P2) in Lemma 2, it
gives lim e(k ) = 0 . Thus, by (51), it can be seen that all the
k

signals in the loop remain uniformly bounded. Then, from


(49), it can be concluded that lim s (k ) = 0 . Therefore, the

Remark 2: V0 (0) can be defined as V0 (0) = v1, min .

Similarly, V 2* (k ) can be derived. In this paper, the adaptive


control input is considered as

v x (k ) = V1* (k ) , v y (k ) = V 2* (k ) .

output tracking error approaches to zero.

(45)

3.4 Stability Analysis


In the following, suppose
W i , sat (k ) Wi (k ) Wi , sat (k ) .

+ B 1 (r , k ) E r [v](k 1)dr
0

= C (q 1 ) zd (k + 1) + s (k ) .
By (32) and (47), we have
R

(47)

T (k ) + B0 P(r ) Er [v](k )dr + B1 P(r ) E r [v](k 1)dr


(48)
= C (q 1 ) z d (k + 1) + s (k ) + e(k + 1) .
By substituting (48) into (26) and using (19), the dynamics of
s (k ) is given as
(49)
s (k + 1) = s (k ) + e(k + 1) .
The closed-loop system equation can be obtained by
combining (6), (48) and (49)
A(q 1 ) q 1B(q 1 )
0 z (k )
1

1
1
1
1
q F (q ) q B(q ) q u (k )

0
0
1 q 1 s (k )

0
0

(50)
= I e(k ) + I C (q 1 ) zd (k )
I
0
The closed-loop system equation (50) will be employed to
analyze the stability of the system (6) together with (25)
controlled by the derived input (45).
Lemma 3. If W i , sat (k ) Wi (k ) Wi , sat (k ) , then there exist
positive constants C1 and C 2 such that

(k ) C1 + C2 max e( + 1) 2 .

Remark 3: In the implementation of the proposed algorithm,


by discretizing [0, R ] uniformly into L sub-intervals, the
integrals can be calculated by the well-known Simpson
Method. Thus, a good calculation of the integrals can be
obtained even though L is not very large.
4. EXPERIMENTAL RESULTS

(46)

From (39), (40) and (45), it yields


R
T (k ) (k ) + B 0 (r , k ) E r [v](k )dr

1 k

controlled by the derived input (45). If W i , sat (k ) Wi (k )

(51)

Proof. The lemma can be proved based on (50) by referring


Chen and Hisayama (2008).
Theorem 1. Consider the system (6) together with (25)

For the experiment setup shown in Figure 1, the piezo


electric actuators used in the experiment are PFT-1110
(Nihon Ceratec Corp.). The generated maximum force is not
less than 80kgf. The maximum displacement is not smaller
than 83m . The applied voltage range is -10V~150V. The
non-contact capacitive displacement (NCCD) sensors (PS-1A,
Nanotex Corp.) with 2nm resolution are used to measure the
displacement of the stage. Since the adaptive method is used
in this paper, the real values of the parameters A1 , A2 , B0 ,

B1 and the density functions in P(r ) are needed neither to


be measured nor to be identified. The sampling period is
chosen as 0.004 second. The experiments are conducted for
xd = 5 sin( 2 (k 103 )) m , y d = 5 cos(2 (k 10 3 )) m .
The offset of the driver is set to 30V. Since the piezo driver
can enlarge the inputted signal 20 times, the value of R is
chosen as R = 10 , v i , min and v i , max are respectively set as

vi , min = 0.5 , and v i , max = 7.5 . The parameter L is chosen as

L = 1000 . The polynomial C (q 1 ) is chosen as C (q 1 ) =

(1 + q 1 + 0.25q 2 ) I . The design parameters are set to


= 0.65 , 1 = 2 = 0.45 . The initial values of the estimated
parameters are chosen as B (r ,0) = B (r ,0 ) = 0.3I and
F0 (0) = F1 (0) = 0.8 I .

Figure 4 shows the control inputs to the piezo actuators in xaxis and y-axis. Figure 5 shows the displacement error in xaxis, where the maximum error in the steady state is about
0.02m . Figure 6 shows the displacement error in y-axis,
where the maximum error in the steady state is about
0.02m . Figure 7 shows the outputs in the XY plane, where
a circle tracking can be confirmed. It can be seen that
relatively good results have been obtained. Furthermore, the
convergence of the estimated parameters has also been
confirmed.

5259

18th IFAC World Congress (IFAC'11)


Milano (Italy) August 28 - September 2, 2011

5. CONCLUSIONS
This paper has discussed the adaptive control for the piezoactuated XY-table. The hysteresis existing in the piezo
electric actuator is described by Prandtl-Ishlinskii model.
Only the parameters directly needed in the formulation of the
controller are adaptively estimated online, where the
parameters of the table need not to be identified or measured.
The proposed control law ensures the global stability of the
controlled piezo-actuated XY-table, and the position error
can be controlled to be relatively small. Experimental results
have shown the effectiveness of the proposed method.
Further researches on tracking control problems for high
frequency desired signals would be expected.
ACKNOLEDGMENTS

Shieh, H.-J. and C.-H. Hsu (2007). An IntegratorBackstepping-Based Dynamic Surface Control Method
for a Two-Axis Piezoelectric Micropositioning Stage.
IEEE Trans. on Control Systems Technology, vol. 15,
no.5, pp. 916926.
Su, C.Y., Wang, Q., Chen, X. and Rakheja, S. (2005).
Adaptive variable structure control of a class of nonlinear
systems with unknown Prandtl-Ishlinskii hysteresis. IEEE
Transactions on Automatic Control, 50, 2069-2074.
Tao, G. and Kokotovic, P.V. (1995). Adaptive control of
plants with unknown hysteresis. IEEE Trans. on
Automatic Control, 40, 200-212.
Tan, X. and Baras, J.S. (2004). Modeling and control of
hysteresis in magnetostrictive actuators. Automatica, 40,
1469-1480.

The work was partially supported by the Grants-in-Aid for


Scientific Research of Japan Society for the Promotion of
Science (JSPS) (No. 21560474), the National Fundamental
Research Program of China under Grant 2009CB320601, the
Funds for Natural Science Foundation of China under Grant
61020106003 and 61074097, the 111 Project under Grant
B08015, and the Fundamental Research Funds for the Central
Universities, SCUT under Grant 2009ZZ0005.
REFERENCES
Brokate, M. and Sprekels, J. (1996). Hysteresis and Phase
Transitions. New York: Springer-Verlag.
Chen, X. and T. Hisayama (2008). Adaptive sliding mode
position control for piezo-actuated stage. IEEE
Transactions on Industrial Electronics, vol. 55, no. 11, pp.
3927-3934.
Fleming, A.J. and S.O.R. Moheimani (2007). Sensorless
vibration suppression and scan compensation for
piezoelectric tube nanopositioners. IEEE Trans. on
Control Systems Technology, vol. 14, no.1, pp. 33-44.
Goodwin, G.C. and Sin, K.S. (1984). Adaptive Filtering,
Prediction and Control. Prentice-Hall, Inc., Englewood
Cliffs, NJ.
Iyer, R.V., X. Tan, and P.S. Krishnaprasad (2005).
Approximate Inversion of the Preisach Hysteresis
Operator with Application to Control of Smart Actuators.
IEEE Transactions on Automatic Control, vol. 50, no. 6,
pp. 798-810.
Krejci, P. and Kuhnen, K. (2001). Inverse control of systems
with hysteresis and creap. Proc. Inst. Elect. Eng. Control
Theory Appl., vol. 148, pp. 185-192.
Kuhnen, K. and P. Krejci (2009). Compensation of complex
hysteresis and creep effects in piezoelectrically actuated
systems A new Preisach modeling approach. IEEE
Transactions on Automatic Control, vol. 54, no. 3, pp.
537-550.
Moheimani, S.O.R. and Goodwin, G.C. (2001). Guest
editorial introduction to the special issue on dynamics and
control of smart structures. IEEE Trans. Control Systems
Technology, 9, 3-4.
Natale, C., Velardi, F. and Visone, C. (2001). Identification
and compensation of preisach hysteresis models for
magnetostrictive actuators. Physica B, vol. 306, pp. 161165.

Fig. 4 Control inputs

Fig. 5 Output tracking error in x-axis.

Fig. 6 Output tracking error in y-axis.

Fig. 7 Outputs in XY plane.

5260

Das könnte Ihnen auch gefallen