You are on page 1of 8

Mechatronics 24 (2014) 328335

Contents lists available at ScienceDirect

Mechatronics
journal homepage: www.elsevier.com/locate/mechatronics

Fuzzy control of nonlinear electromagnetic suspension systems q


Xiaojie Su a, Xiaozhan Yang a, Peng Shi b,c,, Ligang Wu a
a

Space Control and Inertial Technology Research Center, Harbin Institute of Technology, Harbin 150001, China
School of Electrical and Electronic Engineering, University of Adelaide, SA 5005, Australia
c
College of Engineering and Science, Victoria University, Melbourne, VIC 8001, Australia
b

a r t i c l e

i n f o

Article history:
Received 27 December 2012
Revised 1 July 2013
Accepted 20 August 2013
Available online 23 September 2013
Keywords:
Electromagnetic suspension systems
Fuzzy controller design
TakagiSugeno fuzzy model
21 Performance

a b s t r a c t
This paper presents a TS model-based fuzzy controller design approach for electromagnetic suspension
systems. The TS fuzzy model is rstly applied to represent the nonlinear electromagnetic suspension
systems. Then, based on the obtained TS fuzzy model, a fuzzy state feedback controller is used to ensure
the required mixed 21 performance of original electromagnetic suspension system to be achieved.
This controller is designed in a nonparallel-distributed compensation scheme. And sufcient conditions
for the existence of such a controller are derived in terms of linear matrix inequalities. Finally, numerical
simulation on an electromagnetic suspension system is performed to validate the effectiveness of the
proposed approach.
Crown Copyright 2013 Published by Elsevier Ltd. All rights reserved.

1. Introduction
Owing to its environmental, commercial and technological
attractions, the electromagnetic suspension (EMS) system [1] has
been widely adopted in many real world applications. Systems,
such as that in high-speed maglev passenger trains [24], levitation of wind tunnel models, levitation of molten metal in induction
furnaces, vibration isolation and frictionless bearings, are all based
on EMS systems. The main objective of EMS system is to overcome
the inuence of gravity by means of electromagnetic forces. Thus,
in this view, EMS system can be regarded as repulsive system or
attractive system which is based on the source of electromagnetic
levitation forces. Due to the involvement of magnetic force, this
kind of system is described by highly nonlinear differential equations and it is usually open-loop unstable. The nonlinear property
of EMS system presents additional difculties in its manipulation.
Then constructing the high-performance feedback controllers will
be an important task for the regulation of EMS system. Over the
past few years, various controllers have been considered to manipulate the EMS system. To mention a few, nonlinear control methods are proposed in [5,6]; model reference adaptive control has
been studied in [7]; sliding mode control has been adopted by
q
This work was supported in part by the National Natural Science Foundation of
China (61174126 and 61222301), and the Fundamental Research Funds for the
Central Universities (HIT.BRETIV.201303), the National Key Basic Research Program
(2012CB215202), and the 111 Project (B12018).
Corresponding author at: School of Electrical and Electronic Engineering,
University of Adelaide, SA 5005, Australia. Tel.: +61 8 83136424.
E-mail addresses: suxiaojie1985@gmail.com (X. Su), peng.shi@adelaide.edu.au
(P. Shi), ligangwu@hit.edu.cn (L. Wu).

[8]; fault tolerance control is implemented in [9]; and analysis


about active suspension system is given in [1012].
On the other hand, it is well known now that fuzzy logic based
controller design is an effective approach for the manipulation of
complex nonlinear systems. Recent years have witnessed the rapidly growing popularity of fuzzy control systems in practical applications. Usually fuzzy controller is designed in accordance with the
fuzzy-type model of original nonlinear system. Among various
kinds of fuzzy models, the TakagiSugeno (TS) fuzzy model [13]
is viewed as one of the most popular. This kind of model is described by a group of fuzzy IF-THEN rules which describe the local
inputoutput relationships of original nonlinear system. Constructing a group of weight functions from the membership functions in each rule, we can obtain the weighted sum of all local
linear systems. This weighted sum, which is equivalent to the original nonlinear system, can be regarded as a special kind of timevarying linear system. Thus, most of the classical linear control
methods will be applicable. There are numerous cases in this approach. For example, the application to permanent magnet synchronous motor was analyzed in [14]; the guaranteed cost
controller design was illustrated in [15,16]; state estimation problem was solved in [17,18], and passivity-based output feedback
control problem was widely studied in [19,20].
Over the past few decades, a framework of robust control has
been developed to investigate the problems of stability and system
performance in presence of uncertainties and disturbance. Desired
system performances are achieved by minimizing the appropriate
norms (such as H1 (or 22, for continuous-time cases, should
be replaced by L), 21 and 11) of a transfer function. Since
the 2-norm of a signal describes its energy and the 1-norm

0957-4158/$ - see front matter Crown Copyright 2013 Published by Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.mechatronics.2013.08.002

329

X. Su et al. / Mechatronics 24 (2014) 328335

represents its peak value, the above three kinds of norms can be
also described as energy to energy, energy to peak, and peak to peak
performances. Many works are analyzed based on those special
norms of the transfer function, namely, the energy to energy robust
control problem in [2124]; energy to peak optimization problem
in [2527]. In some practical cases, the 21 control is more reasonable in achieving the desired system performance. Such a control procedure ensures that, the controlled output is less than a
prescribed level in case of energy-bounded noise input.
In this work, we are motivated to design a 21 fuzzy controller
for the EMS system of maglev train. This controller is designed such
that, in presence of energy bounded disturbance, the position offset
of maglev train is within an allowable scale. Consequently the comfort level of high speed maglev train will be greatly improved. To
construct such a controller, rstly the nonlinear dynamic equations
of EMS system are obtained from physical laws. Then, discrete-time
TS fuzzy model [28,29] is applied to approximate the original nonlinear dynamic equations. Further, using the LyapunovKrasovskii
technique, sufcient 21 performance conditions in terms of linear matrix inequalities (LMIs) are obtained, thus the desired controller is designed. Finally, simulation results are provided to verify the
feasibility and effectiveness of our controller design method.
The remaining parts of this paper are organized as follows. The
dynamic analysis of EMS system is illustrated in Section 2, and the
TS fuzzy modeling is performed in Section 3. Sections 4 and 5 provide the 21 performance analysis and the controller design
method respectively. And in Section 6, simulation results are represented to illustrate the effectiveness of proposed methods. Finally, conclusions are drawn in Section 7.
1.1. Notations
The notations used throughout this paper are fairly standard.
The superscript T stands for matrix transposition; Rn denotes
the n-dimensional Euclidean space; the notation P > 0(P0) means
that P is real symmetric and positive denite (semi-denite); In and
0mn represent n  n identity matrix and m  n zero matrix,
respectively; diag{  } stands for a block-diagonal matrix; k  k2 denotes the Euclidean norm (2-norm) of a vector; k  k1 indicates the
1-norm of a vector; In symmetric block matrices or long matrix
expressions, we use an asterisk () to represent a term that is induced by symmetry. Matrices, if their dimensions are not explicitly
stated, are assumed to be compatible for algebraic operations.
2. Dynamic analysis of the EMS system
The maglev train is a kind of high-tech transportation
means. In the normal working state, this kind of train is

levitated by the EMS system. A single module of the EMS system is presented in Fig. 1.
The parameters of the EMS system module in Fig. 1 are presented as follows,
gross mass of carriage and electromagnet
coil resistance
valid pole area of the coil
gravitational acceleration
number of turns in the coil
vertical disturbance force
suspension airgap
current of the electromagnet coil
voltage of the electromagnet coil

m
Rm
am
g
N
fd(t)
d(t)
i(t)
u(t)

By Newtons law and Kirchhoffs law, we know, the dynamic


motion of the system can be described by

8
2
>
m d dtdt
>
2 Fit; dt fd t mg;
>
>
>
h i
>
>
< Fit; dt l0 N2 am it 2 ;
4

dt

2:1

>
dWit;dt
>
>
ut  Rm it;
>
dt
>
>
>
: Wit; dt l0 N2 am it ;
2dt

where F(i(t), d(t)) is the suspension force produced by electromagnet; W(i(t), d(t)) is the magnetic potential of the electromagnetic
system; l0 is the permeability of air. By some simple manipulations,
dynamics in (2.1) can be changed into the following equivalent
equations

8
h i
2
>
l0 N2 am it 2
< m d dt
fd t mg;
2 
4

dt

2dt
l0 N2 am

dt

>
: dit
dt

it ddt
dt dt

2:2

ut  Rm it:

_
and
Assuming that the system states are z1(t) = d(t), z2 t dt
z3(t) = i(t), the state-space equations in (2.2) are expressed as

8
z_ 1 t z2 t;
>
>
>
h i2
<
N 2 am z3 t
g m1 fd t;
z_ 2 t  l04m
z1 t
>
>
>
: z_ 3 t 2R2 m z1 tz3 t z2 tz3 t 2z12t ut:
z1 t
l N a
l N a
0

2:3

Since the electromagnet system is required to be kept at the stable


position z1e = dref, the corresponding equilibrium of the system will
p
p
T
be ze , z1e ; z2e ; z3e T dref ; 0; jdref and ue jRm dref , where

j,

4mg

l0 N 2 am

. For convenience of the following analysis, we consider

the following change of coordinates

3 2
3
x1 t
z1 t  z1e
xt , 4 x2 t 5 4 z2 t 5:
x3 t
z3 t  z3e

Guideway

Area am

(t )

F (i (t ), (t ))

Turns N

2:4

With v(t) = u(t)  ue and the above transformation of coordinates, it


is easy to obtain the following equivalent state-space equations for
the system in (2.3),

8
x_ 1 t x2 t;
>
>
p
>
<
2 jd x3 tgx3 t
2dref x1 tgx1 t
 x
m1 fd t;
x_ 2 t jxreftd
2
2
1
1 tdref
ref
>
p

>
x t jd
jx td
>_
:
x3 t 3x1 td ref x2 t  12mg ref Rm x3 t  v t:

2:5

ref

u ( t)

Electromagnet

mg

f d (t )

i (t )

Fig. 1. Conguration of the EMS System module.

It is obvious that, the obtained state-space equations are highly


nonlinear. Then the conventional linear control scheme will be
not applicable to regulate this nonlinear system. To solve this problem, we will adopt the fuzzy control scheme which is mainly based
on the approximated TS fuzzy model of (2.5).

330

X. Su et al. / Mechatronics 24 (2014) 328335

(3) When h1(t) is near h1min and h2(t) is near h2max, the nonlinear
equations can be simplied as

3. TS fuzzy modeling
Before the design of fuzzy controller, rstly we should get the
TS fuzzy model of original nonlinear system. Let us dene h1(t) , x1(t) and h2(t) , x3(t). With the denition

8
>
h1 t;
< h1 max , max
xt

h1 min , minh1 t;

>
: h2 max , maxh2 t;

h2 min , minh2 t;

xt

xt

3:1

xt

(4) When h1(t) is near h1min and h2(t) is near h2min, the nonlinear
equations can be simplied as

h1(t) and h2(t) can be expressed by

h1 t n11 h1 t  h1 max n21 h1 t  h1 min ;

3:2

h2 t n12 h2 t  h2 max n22 h2 t  h2 min :


where nij(hj(t)) 2 [0, 1] for all (i = 1, 2; j = 1, 2), and

n11 h1 t n21 h1 t 1;

8_
x1 t x2 t;
>
>
p
>
>
2 jdref h2 max gx3 t
2dref h1 min gx1 t
>
< x_ 2 t
 h
m1 fd t;
2
jh1 min dref 2
1 min dref
p
h2 max jdref
jh1 min dref
>
>
Rm x3 t  v t;
>
2mg
> x_ 3 t h1 min dref x2 t 
>
:
yt x1 t:

n12 h2 t n22 h2 t 1:

3:3

Then, based on 3.1, 3.2 and 3.3, the membership functions can be
calculated as

8
< n11 h1 t

h1 th1 min
h1 max h1 min

max h1 t
n21 h1 t hh11max
;
h1 min

: n12 h2 t

h2 th2 min
h2 max h2 min

max h2 t
n22 h2 t hh22max
;
h

3:4

8_
x1 t x2 t;
>
>
p
>
>
2 jdref h2 min gx3 t
2dref h1 min gx1 t
>
< x_ 2 t
 h
m1 fd t;
2
jh1 min dref 2
1 min dref
p
h
jdref
jh
dref
>
>
x2 t  1 min
Rm x3 t  v t;
x_ 3 t 2hmin
>
2mg
>
1 min dref
>
:
yt x1 t:
Then, employing the Euler rst-order approximation, we will
get the following discrete-time TS fuzzy model.
Plant Form:

2 min

which have been shown in Figs. 2 and 3.


In this work, fd(t) is regarded as the exogenous noise input, and
we are interested in the inuence of fd(t) on d(t). Thus, the output
of this system is chosen as y(t) = x1(t). By using the local approximation in fuzzy partition spaces [30], the corresponding four-rules
fuzzy model can be obtained for the EMS system.
(1) When h1(t) is near h1max and h2(t) is near h2max, the nonlinear
equations can be simplied as

8
x_ 1 t x2 t;
>
>
>
p
>
2 jdref h2 max gx3 t
2dref h1 max gx1 t
>
1
>
>
< x_ 2 t jh1 max dref 2  h1 max dref 2 m fd t;
p
>
>
h
jdref
jh
dref
>
>
x2 t  1 max
Rm x3 t  v t;
x_ 3 t 2hmax
>
2mg
1 max dref
>
>
:
yt x1 t:

(2) When h1(t) is near h1max and h2(t) is near h2min, the nonlinear
equations can be simplied as

8_
x1 t x2 t;
>
>
p
>
>
2 jdref h2 min gx3 t
2dref h1 max gx1 t
>
< x_ 2 t
 h
m1 fd t;
2
jh1 max dref 2
1 max dref
p
h2 min jdref
jh1 max dref
>
>
Rm x3 t  v t;
>
2mg
> x_ 3 t h1 max dref x2 t 
>
:
yt x1 t:

Rule 1: IF h1(k) is M11 and h2(k) is M12 , THEN

xk 1 A1 xk B11 xk B21 v k;
yk C 1 xk:

Rule 2: IF h1(k) is M21 and h2(k) is M22 , THEN

xk 1 A2 xk B12 xk B22 v k;
yk C 2 xk:

Rule 3: IF h1(k) is M31 and h2(k) is M32 , THEN

xk 1 A3 xk B13 xk B23 v k;
yk C 3 xk:

Rule 4: IF h1(k) is M41 and h2(k) is M42 , THEN

xk 1 A4 xk B14 xk B24 v k;
yk C 4 xk:

where w(k) is the discrete form of fd(t); M11 and M21 represent
about h1max; M31 and M41 represent about h1min; M12 and M32
represent about h2max; M22 and M42 represent about h2min.
Correspondingly, their membership functions are as follows

M11 h1 k M21 h1 k n11 h1 k;

M31 h1 k M41 h1 k

n21 h1 k;
M12 h2 k M32 h2 k n12 h2 k;

M22 h2 k M42 h2 k

n22 h2 k:

21

( 1 (t ))

11

( 1 (t ))

22

( 2 (t ))

12

( 2 (t ))

1
1min

1max

Fig. 2. n11(h1(t)) and n21(h1(t)).

(t )

2
2min

2max

Fig. 3. n12(h2(t)) and n22(h2(t)).

(t )

331

X. Su et al. / Mechatronics 24 (2014) 328335

The system matrices are expressed as

1
6 T2dref h1 max g
6  h d 2
A1 6
1 max
ref
4
0

B11

2 3
0
6 7
4 mT 5;

T
1

6
B21 4

7
5;

jTh1 max dref

jdref h2 max g

jh1 max dref 2

1

jTRm h1 max dref

2 3
0
6 7
4 mT 5;

6
B22 4

7
5;

jTh1 max dref

1

2mg

p
T2 jdref h2 max g

1
p
Th2 max jdref
h1 min dref

jh1 min dref 2

1

jTRm h1 min dref

i1

v k K i xk;

2 3
1
6 7
C T2 4 0 5;

i 1; 2; 3; 4;

3:7

where Ki is the gain matrix of the state-feedback controller. Thus,


the controller in (3.7) can also be represented by
P
v k ri1 hi hkK i xk, with its compact form:

6 T2dref h1 min g
6
A3 6 h1 min dref 2
4
0

7
7
7;
5

jTRm h1 max dref

2mg

i1

Assume that the premise variable of the fuzzy model h(k) is available for feedback, which implies that hi(h(k)) is available for feedback. Suppose that the controllers premise variables are the same
as those in the plant. The parallel distributed compensation strategy
(PDC) is utilized and the fuzzy state feedback controller obeys the
following rules,
 Controller Form:
Rule i: IF h1(k) is Mi1 and h2(k) is Mi2 , THEN

jh1 max dref 2

4
X
hi hkB1i ;

4
X
B2 k ,
hi hkB2i :

i1

p
Th2 min jdref
h1 max dref

B1 k ,

i1

2 3
1
6 7
C T1 4 0 5;

p
T2 jdref h2 min g

4
X
hi hkAi ;

4
X
Ck ,
hi hkC i ;

6 T2dref h1 max g
6
A2 6 h1 max dref 2
4
0

Ak ,

7
7
7;
5

2mg

2mg

B12

T2

p
Th2 max jdref
h1 max dref

where

v k Kkxk;

3
7
7
7;
5

3:8

where

Kk

2mg

4
X
hi hkK i :

3:9

i1

B13

2 3
0
6 7
4 mT 5;

6
B23 4

7
5;

0
1

6 T2dref h1 min g
6
A4 6 h1 min dref 2
4
0
2 3
0
6 7
4 mT 5;

p
Th2 min jdref
h1 min dref

6
B24 4

7
5;

jTh1 min dref

8
4 X
4
X
>
>
>
hi hkAi hj hkB2i K j xk B1i xk;
> xk 1
>
<

1

i1 j1

jdref h2 min g
jh1 min dref 2

T2

Therefore, the closed-loop system can be obtained as

2mg

B14

jTh1 min dref

2 3
1
6 7
C T3 4 0 5;

jTRm h1 min dref

4
>
X
>
>
>
hi hkC i xk:
>
: yk

7
7
7;
5

i1

3:10

2mg

and its compact form is expressed as

2 3
1
6 7
C T4 4 0 5;

yk Ckxk:

2mg

hi hk ,

3:11

where

with T the sampling time. Noting that,


2
Y

b
xk 1 Akxk
B1 kxk;

b
Ak
, Ak B2 kKk:
Mij hj k

j1
4 Y
2
X

i 1; 2; 3; 4;

Mij hj k

i1 j1

we will further get the following fuzzy basis functions,

h1 hk n11 h1 k  n12 h2 k;

h2 hk n11 h1 k  n22 h2 k;

h3 hk n21 h1 k  n12 h2 k;

h4 hk n21 h1 k  n22 h2 k:

It is assumed that the premise variables do not depend on the input


variable v(k) explicitly. Then the defuzzied TS fuzzy model can be
given by

8
4
X
>
>
>
xk 1
hi hkAi xk B1i xk B2i v k;
>
<
i1

4
>
X
>
>
>
hi hkC i xk:
: yk

Remark 1. The availability of the system states is previously


required in the assumption, so we can easily construct the
regulator by the state-feedback controller. But we should also
note that, in some practical cases where not all the system states
are available, additional state observer or dynamic output controller is needed to reproduce the unavailable system states.
Before presenting the main objective of this paper, we rst
introduce the following denitions for the closed-loop system in
(3.11), which will be essential for our main results.
Denition 1. The closed-loop system in (3.11) is said to be
asymptotically stable if under x(k) = 0,

lim jxkj 0;

k!1

3:5

where jxkj ,

p
xT kxk.

i1

A more compact presentation of the TS fuzzy model is given by

xk 1 Akxk B1 kxk B2 kv k;
yk Ckxk;

3:6

Denition 2. Given a scalar c > 0, the closed-loop system in (3.11)


is said to be asymptotically stable with an 21 performance c, if it
is asymptotically stable under x(k) = 0, and satises

kykk1 < ckxkk2 ;

80 x 2 2 0; 1;

332

X. Su et al. / Mechatronics 24 (2014) 328335

where ky(k)k1 is the 1-norm of y(k) and kx(k)k2 is the 2-norm of

x(k), that is,

6
4

q
kykk1 , sup yT kyk;
k

qX

1
kxkk2 ,
xT kxk:
k0

"

Remark 2. The 2-norm of a signal stands for its energy, and the
1-norm stands for the peak of its value, so the 21 performance
is also called the guaranteed energy-to-peak performance.
Our objective in this work is to determine the fuzzy controller
matrices in (3.8) such that the closed-loop system in (3.11) is
asymptotically stable with a guaranteed 21 performance level c.

8
T
1
b
>
>
< Pk , G kPkG k;
P4
Pk , i1 hi hkPi ;
>
>
P
:
Gk , 4i1 hi hkGi ;

4:1

b
Vk , xT k Pkxk:

4:2

Then, based on the fuzzy LKF (4.2), we can obtain the following
result.
Theorem 1. The closed-loop system in (3.11) is asymptotically stable
with a guaranteed 21 performance level c if there exist matrices
Pi > 0 and Gi, i = 1, 2, 3, 4, which are dened in (4.1), such that for any
integer k, the following matrix inequalities hold for any integer k,

Gk 1  GT k 1 Pk 1

c2 I CkGk
Pk

c I Ck
b
Pk

I

> 0:

4:8

By Schur complement, it follows that (4.7) and (4.8) are equivalent


respectively to

"

b T k Pk
b
b 1 Ak
b
A
 Pk

b T k Pk
b 1B1 k
A
b 1B1 k  I
BT k Pk

b
B1 k
AkGk

Pk

I

3
7
5 < 0;

!k ,

J k , Vk 

xT ixi:

i0

For any nonzero x(k) 2 2[0,1) and k > 0, we have

T
k1 
X
xi

xi

k1
X

k1
X
DVi  xT ixi

i0

i0

xT ixi



xi
:
Xi
xi

4:12

where

"

b T i Pi
b  Pi
b 1 Ai
b
A


#
b T i Pi
b 1B1 i
A
:
b 1B1 i  I
BT i Pi
1

By (4.9), we have J k < 0. It further gives rise to

Pk 1  Gk 1P 1 k 1Pk 1  Gk 1 P 0;

b
<
xT k Pkxk

we have

b 1 k 1 6 Gk 1  GT k 1 Pk 1:
P

k1
X

J k Vk  V0 

4:4

#
b T k Pk
b 1B1 k
A
:
b 1B1 k
BT k Pk

b T k Pk
b
b 1 Ak
b
A
 Pk

i0

> 0:

4:10

By (4.9) and (4.11), it follows that DV(k) < 0, thus the closed-loop
system in (3.11) is asymptotically stable.
In the following, we shall investigate the 21 performance.
Under the zero initial condition, we have V(0) = 0. Dene an index
as

4:3

k1
X

xT ixi:

4:5

Then, based on (4.10) we have

b
yT kyk < c2 xT k Pkxk:

Combining (4.13) and (4.14) together, we have

Pk

I

3
7
5 < 0:

4:13

i0

Thus, it follows from (4.3) and (4.5) that the following inequality
holds

b
B1 k
AkGk

4:9

where

Proof 1. From the fact

b 1 k 1
P

< 0;

b 1xk 1  xT k Pkxk
b
DVk , Vk 1  Vk xT k 1 Pk

T


xk
xk
;
4:11
!k

xk
xk

Xi ,

6
4

Considering the fuzzy LKF (4.2), and along the trajectories of the
closed-loop system in (3.11), we have

"

where Pi > 0 and Gi, i = 1, 2, 3, 4 are n  n matrices. We construct the


following fuzzy LyapunovKrasovskii function:

"


2

4:7

b
< 0:
C T kCk  c2 Pk

In this section, a new LyapunovKrasovskii function (LKF) for


the EMS systems is introduced. Based on this LyapunovKrasovskii
functional candidate, a new stability criterion with the 21 performance level c is derived.
To this end, let

6
4

4. Performance analysis

3
b
B1 k
Ak
7
b
 Pk
0 5 < 0;

b 1 k 1
P

4:6

Dene the following matrices:

yT kyk < c2

k1
X

xT ixi:

4:14

4:15

i0

T k , diagfI

G1 k Ig;

Taking the supremun of yT(k)y(k) over k and the limit of


Pk1 T
i0 x ixi with k ? 1, we obtain

Sk , diagfI

G1 kg:

supfyT kykg < c2

Performing congruence transformations to (4.4) and (4.6) by


fS T k; Skg and fT T k; T kg, respectively, and considering (4.1),
we have

1
X

xT ixi:

4:16

i0

Obviously ky(k)k1 < ckx(k)k2 holds for all nonzero x(k) 2 2[0,1),
and the proof is completed. h

333

X. Su et al. / Mechatronics 24 (2014) 328335

Remark 3. In Theorem 1, the proposed stability condition for the


obtain TS fuzzy system is fuzzy-rule-dependent. By constructing
the quadratic LyapunovKrasovskii function V(k) , xT(k)Px(k), we
can obtain the fuzzy-rule-independent stability condition directly,
which is a special case of Theorem 1 and has much large conservativeness for the concerned TS fuzzy systems. Moreover, the 21
fuzzy-rule-dependent controller is obtained directly by this fuzzyrule-dependent method.

In this section, Theorem 1 will be utilized to the design a state


feedback controller for the EMS system in (3.11). Assume that all of
the states are available for feedback control. As in [31], the following nonparallel distributed compensation (non-PDC) control law is
considered for the closed-loop system (3.11),

v k KkG1 kxk;

5:1

where Kk and Gk are dened in (3.9) and (4.1), respectively.


Obviously, if we take Gi = G, (5.1) becomes a PDC controller. Substituting (5.1) into (3.6), we obtain the following closed-loop system

e c kxk B1 kxk;
xk 1 A

5:2

e c k Ak B2 kKkG1 k.
where A
Theorem 2. The closed-loop system in (5.2) is asymptotically
stable with a guaranteed 21 performance level c if there exist
matrices Pi > 0 and Gi,i = 1, 2, 3, 4, which are dened in (4.1), such
that inequality in (4.4) and the following inequality hold for any
integer k,
6
4

Proof 3. The inequalities of (4.4) and (5.3) can be respectively


written as
4 X
4 X
4
X
hs hk 1hj hkhi hkPsij < 0;

5:8

s1 j1 i1
4 X
4
X
hj hkhi hkUij < 0:

5:9

j1 i1

5. Fuzzy controller design

Meanwhile, there exists a fuzzy controller of the form (5.1).

3
Gk 1  GT k 1 Pk 1 AkGk B2 kKk B1 k
7
0 5 < 0:

Pk


I

5:3

Proof 2. In the view of the fuzzy closed-loop system (5.2), we


b
e c k. Following the same line of the
replace Ak
in (3.11) with A
proof of the Theorem 1, the result can be easily derived. The proof
is completed. h
It is noted that Theorem 2, expressed in the form of fuzzy-basisdependent matrix inequalities, cannot be directly implemented for
the fuzzy controller design with 21 performance level c. Our
next objective is to convert the inequalities to some nite LMIs,
which can be readily solved by using standard numerical software.
We have the following result.

By using [32], if conditions (5.4)(5.6) and (5.7) hold, then (5.8) and
(5.9) are fullled. Therefore, it follows from Theorem 2 that the EMS
system in (3.11) is asymptotically stable with a guaranteed 21
performance level c. Thus, the proof is completed. h

6. Simulation results
In this part, the proposed method is applied to design a fuzzy
feedback controller for the afore mentioned EMS system. The
parameter values of the EMS system are given in Table 1 (other sets
of parameter values can be found in [2,5,6]), and the allowable
bounds of h(k) are listed in Table 2.
With the parameter value given in Tables 1 and 2, we will get
the exact expressions of original nonlinear EMS system and its
approximated TS fuzzy model. Firstly let us analyze the stability
property of the original system. We set the initial state of the
EMS system as x(0) = [0.0001, 0, 0]T. And then system states of the
original open-loop system are plotted in Fig. 4. Obviously the origi-

Table 1
Parameter values of the EMS system.
Parameter
Unit

m
kg

Rm
X

Value

150

1.1

g
m/
s2
9.8

am
m2

N
kilo

l0
H/m

dref
m

T
ms

1.024  102

4p  107

0.004

0.5

Table 2
Allowable bounds of h(k).
Parameter
Unit
Value

h1max
m
0.001

h1min
m
0.001

h2max
A
1

h2min
A
1

100

150

200

0.2

Theorem 3. The closed-loop system in (3.11) is asymptotically stable


with a guaranteed 21 performance level c if there exist matrices
Pi > 0, Gi and Ki, i = 1, 2, 3, 4, which are dened in (4.1), such that for
any s, i, j = 1, 2, 3, 4, the following inequalities hold,
1
1
Psii Psij Psji < 0; i j;
3
2
Psii < 0;
1
1
Uii Uij Uji > 0; i j;
3
2
Uii > 0;

5:4
5:5

Psij , 4

Gs 

GTs

Ps




Ai Gj B2i K j
P i


0.2
0.4
0.6

5:6
5:7

where

B1i
0 5;
I

0.8
1
1.2
0

50

c2 I C i Gj :
Uij ,


Pi

Fig. 4. States of the original open-loop system.

334

X. Su et al. / Mechatronics 24 (2014) 328335

nal open-loop EMS system is unstable. By using the LMI toolbox of


matlab and setting c = 0.005, we can nd the following matrices
which satisfy inequalities (5.4)(5.6) and (5.7).
5

0.05

K 1 10  0:0001 0:0058 1:4860 ; K 2 10  0:0001 0:0031 1:4994 ;


K 3 105  0:0001 0:0159 2:1868 ; K 4 105  0:0001 0:0100 2:1925 ;
3
2
3
0:0000 0:0011 0:0001
0:0000 0:0011 0:0001
6
7
6
7
G1 4 0:0011 0:2835 0:0820 5; G2 4 0:0011 0:2780 0:0325 5;
0:0052 3:0222 57:9373
0:0053 3:0567 58:3998

3
2
3
0:0000 0:0011 0:0002
0:0000 0:0011 0:0001
6
7
6
7
G3 4 0:0010 0:2969 0:5271 5; G4 4 0:0010 0:2892 0:3933 5;
0:0051 2:8855 55:9006
0:0051 2:9619 55:9910

Constructing a non-PDC fuzzy controller by (5.1), we will get the


system states of the closed-loop system. State response of the controlled TS fuzzy system is expressed in Fig. 5. It can be seen from
Fig. 5 that the approximated TS fuzzy system has been stabilized
by our designed fuzzy controller. Also note that approximation errors have been inevitably introduced in the process of fuzzication,
so the approximated fuzzy model is not completely equivalent to
our original nonlinear dynamic model. To achieve the prescribed
21 performance of the original system, we should also make sure
that the constructed TS fuzzy controller is robust enough to overcome the encountered approximation errors and possible parameter uncertainty. Then the obtained fuzzy control method is
applied to regulate the original nonlinear model. The control result
of original nonlinear system is plotted in Fig. 6, from which we can
nd that, compared with the fuzzy model, the controlled nonlinear
model just takes a bit longer time to converge to its equilibrium.
Obviously, the constructed controller can be well applied to manipulate the original nonlinear system, even there exist approximation
errors when we get the fuzzy-type model of the original nonlinear
system. And in this manner, the robustness of the illustrated
21 fuzzy control strategy is thus conrmed.
Furthermore, let us analyze the 21 performance of the controlled EMS system. Here, we set x(0) = [0, 0, 0]T and choose the following exogenous disturbance,

xk

0.1

100

50

200

150

250

Fig. 6. States of the controlled nonlinear system.

x 10

3
2.5
2
1.5
1
0.5
0
0.5
1

90; when 0 < k < 20


0;

0.05

when k 6 0 or k P 20:

p
P1
T
By calculation, we have kxkk2
k0 x kxk 392:3. Corresponding output of the controlled nonlinear system is shown in
Fig. 7, where the peak value of y(k) is about 3  105. So the actual

50

100

150

200

250

300

350

Fig. 7. Zero-state response of the controlled nonlinear system.

10

0.05

0.05
10

0.1

15

50

100

150

200

Fig. 5. States of the controlled fuzzy model.

250

20

50

100

150

200

250

Fig. 8. The fuzzy control input.

300

350

X. Su et al. / Mechatronics 24 (2014) 328335


1
value of kkykk
xkk in this situation is well below c = 0.005, which exactly
2

veries that our designed fuzzy controller has ensured the prescript
21 performance of the given EMS system. The fuzzy control input
u(k) is presented in Fig. 8.
7. Conclusions
In this paper, the TS fuzzy model based 21 fuzzy control
method has been proposed for the nonlinear EMS system. The
standard nonlinear dynamic equations of the EMS system have
been rstly obtained from some physical laws. Then the original
nonlinear system has been presented by a TS fuzzy model. Further, 21 performance analysis of the EMS system has been performed based on its approximated TS fuzzy model, and
corresponding fuzzy controller design method has been derived
under a non-PDC scheme. Finally, simulation results have been
presented to demonstrate the effectiveness of the proposed controller design method.
References
[1] Hurley WG, Wole WH. Electromagnetic design of magnetic suspension
system. IEEE Trans Edu 1997;40(2):12430.
[2] Xu J, Zhou Y. A nonlinear control method for the electromagnetic suspension
system of the maglev train. J Mod Transport 2011;19(3):17680.
[3] Kang BJ, Hung LS, Kuo SK, Lin SC, Liaw CM. H1 2DOF control for the motion of a
magnetic suspension positioning stage driven by inverter-fed linear motor.
Mechatronics 2003;13(7):67796.
[4] Lim TM, Cheng S. Magnetic levitation of a one DOF system using simultaneous
actuation
and
displacement
sensing
technique.
Mechatronics
2011;21(3):54859.
[5] Joo S, Seo JH. Design and analysis of the nonlinear feedback linearizing control
for an electromagnetic suspension system. IEEE Trans Control Syst Technol
1997;5(1):13544.
[6] Sinha PK, Pechev AN. Nonlinear H1 controllers for electromagnetic suspension
systems. IEEE Trans Autom Control 2004;49(4):5638.
[7] Sinha PK, Pechev AN. Model reference adaptive control of a maglev system
with stable maximum descent criterion. Automatica 1999;35(8):145765.
[8] Fei J, Xin M. Robust adaptive sliding mode controller for semi-active vehicle
suspension system. Int J Innovative Comp Inf Control 2012;8(1B):691700.
[9] Sung HK, Lee SH, Bien Z. Design and implementation of a fault tolerant
controller for EMS systems. Mechatronics 2005;15(10):125372.
[10] Du H, Zhang N. Fuzzy control for nonlinear uncertain electrohydraulic active
suspensions with input constraint. IEEE Trans Fuzzy Syst 2009;17(2):34356.
[11] Gysen B, Paulides J, Janssen J, Lomonova E. Active eletromagnetic suspension
system for improved vehicle dynamics. IEEE Trans Veh Technol
2010;59(3):115663.
[12] van der Sande TPJ, Gysen BLJ, Besselink IJM, Paulides JJH, Lomonova EA,
Nijmeijer H. Robust control of an electromagnetic active suspension system:
simulations and measurements. Mechatronics 2013;23(2):20412.

335

[13] Takagi T, Sugeno M. Fuzzy identication of systems and its applications to


modeling and control. IEEE Trans Syst Man Cyber 1985;SMC-15(1):11632.
[14] Choi HH, Jung JW. TakagiSugeno fuzzy speed controller design for a
permanent magnet synchronous motor. Mechatronics 2011;21(8):131728.
[15] Nguang SK, Shi P. H1 fuzzy output feedback control design for nonlinear
systems: an LMI approach. IEEE Trans Fuzzy Syst 2003;11(3):33140.
[16] Ting CS. An output-feedback fuzzy approach to guaranteed cost control of
vehicle lateral motion. Mechatronics 2009;19(3):30412.
[17] Nguang SK, Shi P, Ding S. Fault detection for uncertain fuzzy systems: an LMI
approach. IEEE Trans Fuzzy Syst 2007;15(6):125162.
[18] Ho W-H, Chen S-H, Chen I-T, Chou J-H, Shu C-C. Design of stable and quadraticoptimal static output feedback controllers for TS-fuzzy-model-based control
systems: an integrative computational approach. Int J Innovative Comp Inf
Control 2012;8(1A):40318.
[19] Karimi HR. Passivity-based output feedback control of Markovian jump
systems with discrete and distributed time-varying delays. Int J Syst Sci
2013;44(7):1290300.
[20] Chang W-J, Ku C-C, Huang P-H. Robust fuzzy control via observer feedback for
passive stochastic fuzzy systems with time-delay and multiplicative noise. Int
J Innovative Comput Inf Control 2011;7(1):34564.
[21] Gao H, Zhao Y, Chen T. H1 fuzzy control of nonlinear systems under unreliable
communication links. IEEE Trans Fuzzy Syst 2009;17(2):26578.
[22] Karimi HR. Robust delay-dependent H1 control of uncertain time-delay
systems with mixed neutral, discrete and distributed time-delays and
Markovian switching parameters. IEEE Trans Circuits Syst I: Regular Papers
2011;58(8):191023.
[23] Karimi HR, Dufe NA, Dashkovskiy S. Local capacity H1 control for production
networks of autonomous work systems with time-varying delays. IEEE Trans
Autom Sci Eng 2010;7(4):84957.
[24] Wu L, Su X, Shi P, Qiu J. A new approach to stability analysis and stabilization
of discrete-time TS fuzzy time-varying delay systems. IEEE Trans Syst Man
CyberPart B: Cyber 2011;41(1):27386.
[25] Wu L, Su X, Shi P, Qiu J. Model approximation for discrete-time state-delay
systems in the TS fuzzy framework. IEEE Trans Fuzzy Syst
2011;19(2):36678.
[26] Mahmoud MS, Xia Y. Design of reduce-order 21 lter design for singular
discrete-time systems using strict linear matrix inequalities. IET Control
Theory Appl 2010;4(3):50919.
[27] Zhou Y, Li J. Reduced-order L2 -L1 ltering for singular systems: a linear
matrix inequlity approach. IET Control Theory Appl 2008;2(3):22838.
[28] Su X, Shi P, Wu L, Song Y-D. A novel control design on discrete-time TakagiSugeno fuzzy systems with time-varying delays. IEEE Trans Fuzzy Syst
2013;21(4):65571.
[29] Su X, Shi P, Wu L, Song Y-D. A novel approach to lter design for TS fuzzy
discrete-time systems with time-varying delay. IEEE Trans Fuzzy Syst
2012;20(6):111429.
[30] Tanaka K, Wang HO. Fuzzy control systems design and analysis: a linear
matrix inequality approach. New York: John Wiley and Sons; 2001.
[31] Guerra TM, Vermeiren L. LMI-based relaxed nonquadratic stabilization
conditions for nonlinear systems in the TakagiSugenos form. Automatica
2004;40(5):18239.
[32] Tuan HD, Apkarian P, Narikiyo T, Yamamoto Y. Parameterized linear matrix
inequality techniques in fuzzy control system design. IEEE Trans Fuzzy Syst
2001;9(2):32432.

You might also like