International Conference on Control, Automation and Systems 2010

Oct. 27-30, 2010 in KINTEX, Gyeonggi-do, Korea

Apply Adaptive Fuzzy Sliding Mode Control to SMA Actuator

Nguyen Trong Tai1, Kyoung Kwan Ahn2
1
Department of Mechanical and Automotive Engineering, Graduate school , Ulsan University, Ulsan, Korea
(Tel : +82-52-259-1501; E-mail: [email protected] )
2
Department of Mechanical and Automotive Engineering, Ulsan University, Ulsan, Korea
(Tel : +82-52-259-2282; E-mail: [email protected])

Abstract: In this paper, an adaptive fuzzy sliding-mode controller (AFSMC) is proposed to control SMA actuator. An
approach of self-tuning fuzzy sliding-mode control which combines fuzzy control with the sliding-mode control (SMC),
is presented for the tracking control of a class of nonlinear systems with the parameter uncertainties. Based on Lyapunov
theory, the asymptotic stability of the overall systems is proved. The controller is applied to compensate the hysteresis
phenomenon of SMA. The control results show that the controller is applied successfully to SMA.
Keywords: SMA Actuator, smart material, Adaptive Control, Sliding Mode Control, Adaptive Fuzzy Control.
Presentation style: Oral presentation
presents the AFSMC design for general nonlinear system
1. INTRODUCTION and its application to SMA Actuator. The control results
Shape Memory Alloy actuators, which have ability to are shown in section 4.
return to their original shape at a preset temperature,
possess interesting properties in term of force generation 2. EXPERIMENTAL SETUP
capacity, possibility of miniaturization, and power Figure 1 shows the experimental setup for
consumption. The unusual characteristic of SMA material measurement and control of the SMA actuator. In this
has significant potential applications in control systems. model, the SMA actuator is DM01 Linear Memory Metal
The nonlinearity hysteresis effect existing in the SMA Actuator (Miga Motor Company). The steel spring
actuators introduces delays and leads to inaccuracy in attached at the end of SMA Actuator takes a role of bias
control of these actuators. The SMA characteristics and spring and load. The displacement of SMA actuator is
working principles can be seen in [1].
measured by a high precision potentiometer (Copal
In order to obtain better performance of the control Electronics). These feedback signals are fed to the
system, hysteresis phenomena must be compensated. computer through an A/D Advantech PCI-1711 card. The
There are some studies investigating to compensate the control current applied to the SMA Actuator was obtained
hysteresis effect for SMA. In [2] the SMC is applied to from a D/A card and a V/I converter. This system is
control SMA arm. Or in [3], the internal model controller controlled in real-time with Real-time Windows Target
is applied to compensate the SMA hysteresis phenomena. Toolbox of Matlab. The sampling time was set to 0.01s in
Sliding mode control (SMC) is a better robust control all experiments.
method. Once the states of controlled systems enter the
sliding mode, the dynamic characteristics of overall
system are determined by the designed sliding surface and
independent of uncertainties. However, in the practical
control, the chattering phenomena in the sliding mode due
to switching operating and high gain of the SMC influence
the tracking accuracy and limit its application. The Fuzzy
Controller (FC) combine with SMC to eliminate the
chattering of SMC is presented in [4-5].Or adaptive sliding
mode control was also investigated in [6]. A method of
designing AFSMC was presented was applied to control
the invert pendulum in simulation in [7].
This paper proposes a simple structure of controller
means of fuzzy sliding mode control with self-learning
capability where the fuzzy logic system is employed to Figure 1: Experimental Setup
approximate the equivalent control of sliding mode in a
compact set. In this controller, the sugeno fuzzy controller 3. DESIGN OF ADAPTIVE FUZZY SLIDING
is designed in which the input membership functions are MODE CONTROLLER
designed by experience and the output membership 3.1 Description of problem
functions are capable of changing. The output membership
Consider the following nonlinear system:
functions are tuned in Lyapunov sense; thus, the system
stability is always satisfied. Then the proposed method is
used to control SMA
The remainder of the paper is organized as follows. In
Section 2, the experimental setup is presented. Section 3

978-89-93215-02-1 98560/10/$15 ©ICROS 433

Authorized licensed use limited to: Carleton University. Downloaded on November 02,2020 at 03:39:04 UTC from IEEE Xplore. Restrictions apply.
 ⎧ x1 = x2 Fuzzy rules: The fuzzy rules are expressed as:
⎪ If e is A1i (m1i , σ 1i ) and e is A2 j (m2 j , σ 2 j ) then u if, j = pi , j ,
⎪ x2 = x3
⎨
⎪# (1) i = 1, 2,..., m , j = 1, 2,..., n
⎪⎩ xn = f ( x) + b( x)u + d (t ) Defuzzification:
y = x1 Using minimum method for fuzzy intersection, the rule
T firing strength is defined as:
where x = [ x1 , x2 , ..., xn ] = ⎡⎣ x, x,..., x ( n −1) ⎤⎦ is system's
T

wk = μ A1i ∩ A 2 j = min ( μ A1i (e), μ A2 j (e) )

state vectors, b( x) > 0 , f ( x) are nonlinear scalar
functions or either one is nonlinear function, and their Using the weighted average defuzzification method, the
value isn't exactly known, u is the control input of system final output of the fuzzy controller can be obtained:
m *n
and y is the output of system, and d (t ) is a bounded
disturbance. The control goal for the systems is that the
∑w p i i
1
uf = i =1
= wT p (7)
output y can precisely track the desired output xd when m *n

∑w i
∑ i
w
b( x) , f ( x) , d ( x) contain some uncertainties. i =1

3.2 Sliding mode controller where w = [ w1 ,..., wm*n ] , p = [ p1 ,... pm*n ] , p is an

T

The tracking error between desired reference and system unknown consequent vector, which can be obtained by the
response is taken as: learning control method in next part.
e = xd − y or e = xd − x1 (2) 3.4 Stability, Learning law and compensator control
T signal
e = ⎡⎣e e ... e( n −1) ⎤⎦ is defined as the vector of error.
In order to derive the learning laws, at first, the following
The Sliding mode surface is defined as: assumptions should be used:
s = cT e = c1e + c2 e + ...e( n −1) , (ci > 0) (3) Assumption 1: In the formulation (1) b( x) and its
estimated value bˆ( x) satisfy the following conditions:
where c = [ c1 c2 ... cn −1 1]
T

1 b( x )
According to the sliding mode control theory, when s = 0 , b( x) > 0 , and ≤ ≤ β , ( β > 0) (8)
the equivalent control input of sliding mode can be β bˆ( x)
obtained as follows: In fact, equation (12) can be obtained from a number of
1⎛ n −1
⎞ mechantronics systems [8-9].
ueq = u s = 0 = ⎜ − f − d + xd( n ) + ∑ ci e (i ) ⎟ (4)
b⎝ i =1 ⎠ Generally, bˆ( x) can be selected as follows:
If the accurate values of b , f , d have been known, ueq bˆ( x) = bmax .bmin = constant (9)
can be gotten directly from (4). Based on the theory of
Assumption 2: The vector set T = [ e, e] is a compact set
T
sliding mode control, control law is chosen as:
u = ueq + us (5) and T ∈ R 2 , exist a group of consequent parameter vector
T
where us is a nonlinear control input signal, which p = p* = ⎡⎣ p1* , p2* ,..., pm* *n ⎤⎦ , which satisfy the following
produce sliding modes. condition:
In practical control, there often exist some uncertainty Sup ueq − u *f ≤ ε 0 , ( ε 0 ≥ 0 ) (10)
factors in b , f , d . Thus, a fuzzy system controller u f is T ∈R n

employed to approximate sliding mode equivalent control where:

signal ueq . Then (5) becomes: 1
u*f = wT p* (11)
u = u f + us (6) ∑ wj
3.3 Design of fuzzy controller In fact, this assumption is introduced in [10].
From (4), a sugeno fuzzy controller is designed with 2 From two above assumption, the system stability and
inputs e , e and 1 output u f as follows: control signal us are derived as follows:

Input, output: Let p − p* = pˆ

Then
1 1
u f − u *f = wT ( p − p* ) = wT pˆ (12)
∑ wj ∑ wj

Differentiating s(t ) with respect to t along the trajectory,

we have:
Figure 2: Fuzzy controller structure

978-89-93215-02-1 98560/10/$15 ©ICROS 434

Authorized licensed use limited to: Carleton University. Downloaded on November 02,2020 at 03:39:04 UTC from IEEE Xplore. Restrictions apply.
 n −1 then, to produce sliding mode, one compensator is added
s = ∑ ci e( i ) + e( n ) to modify the nonlinear control signal ( us ) to drive the
i =1
system moving to the equivalent point.
Using (1),(4) and (6) the following equation can be
obtained: The block diagram of AFSMC is shown as below:
s = b ( ueq − u f − us ) (13)
e
Thus
d
ss = sb ( ueq − u f − us ) (14) uf
dt
Then a candidate Lyapunov function can be chosen as Plant
follows: us
1 2 1 T s = cT e k0 s + k1sign(s)
V= s + pˆ pˆ (15)
2bˆ 2
where b̂ is selected according to (9). Differentiating the Figure 3: Block diagram of AFSMC
equation (15) in two sides, then
1 1 To reducing the chattering phenomenon, the sign function
V = ss + pˆ T pˆ = sb ( ueq − u f − us ) + pˆ T pˆ in (16) is replaced by the sat function at follows:
bˆ bˆ
⎧u
≤ s β ( ueq − u f ) − s β ( u f − u*f ) − s β us + pˆ T pˆ
*
(16) ⎪ 0
if s >u
ϕ 0

Then, the compensator control signal us is selected as: ( )ϕ

⎪
k1 .sat s = ⎨−u0
⎪
if s <u
ϕ 0 (22)
us = k0 s + k1 .sign ( s ) (17) ⎪s others
⎩ ϕ
⎧k if s>0
k1 .sign ( s ) = ⎨ 1 (18)
⎩ −k1 if s<0 4. EXPERIMENTAL RESULTS
where k0 > 0 and k1 > 0 . From (10), (12), (18), the This part presents the experimental results of the real
time position control of the proposed SMA actuator using
following equation can be obtained:
AFSMC. The experiments were implemented at room
1 temperature (250C). Sampling time was set to 0.01s in all
V ≤ s βε 0 − s β wT pˆ − s β k0 s − s β k1 sign( s ) + pˆ T pˆ
∑ wj ϕ experiments. The experiments results showed that the
AFSMC is applied successfully to control SMA actuator.
⎛ 1 ⎞ The control parameters are set as:
≤ s βε 0 − s β k1 − β k0 s 2 + ⎜ pˆ T − s β wT ⎟ pˆ
⎜
⎝ ∑ w j ⎟⎠ Sliding surface pole : −λ = 5
(19) Adaptive updating factor : β = 0.01
From (19), p̂ and k1 is chosen as: Gausian var iance :σ = 0.5
⎧ 1 Initial fuzzy output membership function :
⎪ pˆ = p = s β w
⎨ ∑ wj (20) p = [ p1,0 , p2,0 ..., p25,0 ] = 0.1 ⋅ ones (25,1)
⎪k > ε
⎩ 1 0
Compensator Controller parameters :
Then k0 = 1, ϕ = 0.4, u0 = 0.7
V < − β k0 s 2 < 0 (21) Figure 4 shows the system response with step reference
The formula (21) indicates that the stability of SMC using AFSMC. The control error is shown in Figure 6. It
systems can be guaranteed when u f , output of fuzzy can be seen that the system can archive the small error
using AFSMC. The control signal is shown in Figure 5.
control, is used to approximate sliding mode equivalent
control ueq and sliding mode state is asymptotically stable
in terms of Lyapunov theory. When t → ∞ , the following
dynamic characteristic can be achieved:
s = c1e + c2 e + ... + e( n −1) = 0
Consequently, tracking error approaches zero along sliding
surface.
The design block diagram of the AFSMC structure is
shown in Fig. 6. In the diagram, the fuzzy controller is
used to generate control signal at the equivalent point
( u f = ueq ). The fuzzy output is updated by the sliding
mode surface due to the non-linearity of the system. And

978-89-93215-02-1 98560/10/$15 ©ICROS 435

Authorized licensed use limited to: Carleton University. Downloaded on November 02,2020 at 03:39:04 UTC from IEEE Xplore. Restrictions apply.
 8
9

7 6

Dispacement [mm]
Displacement [mm]

5
4
4 Reference
AFSMC Response
3
2
2 Reference
AFSMC Response
1
0
0 0 10 20time
0 20 40 60 80 100 [sec]30 40 50
time [sec]

Figure 4: Position control with respect step reference Figure 7: Position control with respect sinusoidal
reference
0.9 0.8
AFSMC Control Input AFSMC Control Signal
0.8 0.7

0.7
0.6
0.6
Current [A]

Current [A]
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2

0.1 0.1

0.0 0.0
0 20 40 60 80 100 0 10 20 30 40 50
time (sec) time [s]

Figure 5: Control signal with respect to step reference Figure 8: Control signal with respect to sinusoidal
0.20 reference
AFSMC Error
0.15
0.10
0.10 AFSMC Control Error
error [mm]

0.05
0.05
0.00
Error [mm]

-0.05
0.00
-0.10

-0.15
-0.05
-0.20
0 20 40 60 80 100
time [s]
-0.10
Figure 6: Position error with respect step reference 0 10 20 30 40 50
time [s]
The AFSMC is also applied to control the SMA position Figure 9: Position error with respect sinusoidal
with response to sinusoidal reference. The control results reference
are shown in Figures 7-9. It can be seen that the SMA can
track well to the reference signal with small error (0.07
mm).

978-89-93215-02-1 98560/10/$15 ©ICROS 436

Authorized licensed use limited to: Carleton University. Downloaded on November 02,2020 at 03:39:04 UTC from IEEE Xplore. Restrictions apply.
 of a Shape Memory Alloy Actuated Manipulator”
Hysteries curve before applying Controller Trans. of the ASME, Vol.124, pp.566-575, 2002
8
[3] K.K. Ahn and N.B. Kha, “Internal Model Control For
Shape Memory Alloy Actuators Using Fuzzy Based
Preisach Model”, Sensors and Actuators A: Physical,
Displacement [mm]

6
Vol.136, No.2, pp.730-741, 2007.
[4] R. Swiniarski, Neuromorphic fuzzy variable structure
4 controller, Proc. of IEEE on Industrial Electronics
Conference, Pacific Grove, CA, USA,
pp.1221-1226,1990.
2 [5] R. Palm, “Sliding mode fuzzy control”, Proc.1 st
IEEE International Conference on Fuzzy Systems, San
0.0 0.2 0.4
Current input [A]
0.6 0.8 Diego, pp. 519-526, 1992.
8
Hysteresis curve after apllying Controller
[6] G. Cheng and P. Shuangxia, “Adaptive sliding mode
control of electro-hydraulic system with nonlinear
7
unknown parameters”, Control Engineering Practice,
Displacement [mm]

Vol.16, No.11, pp. 1275-1284, 2008.

6
[7] S.C. Lin & Y.Y. Chen, “Design of Adaptive Fuzzy
5 Sliding Mode for Nonlinear System Control”, Proc. of
IEEE International Conf. on Fuzzy System, Orando,
4 pp. 35-39, 1994.
[8] D. Suolin, “Research on robust sliding tracking
3
control for mechanelectronic systems with
uncertainties and its applications”, Doctoral thesis,
2
2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 Xi’an Jiaotong University，Xi’an, 1999.
Desired Signal [mm]
[9] H. Abid, M. Chtourou & A. Toumi, “Fuzzy Indirect
Adaptive Control Scheme for Nonlinear Systems
Figure 10: Hysteresis curve before and after applying Based on Lyapunov Approach and Sliding Mode”,
AFSMC International Journal Of Computational Cognition, 5,
pp. 36-43, 2007.
The hysteresis curve between input and output [10] J.L. Castro, “Fuzzy Logic Controllers Are Universal
relation of a system including the SMA actuator and Approximators”, IEEE Transactions On Systems, Man,
controller before (open loop control) and after applying the And Cybernetics, 25, pp.629-635, 1995.
AFSMC control algorithm (closed loop control) is shown
in Fig. 10. The hysteresis curve is obtained from the
second cycle of the system response with respect to the
sine reference signal shown in Fig. 7. It shows nearly
linear relationship between desired displacement (input)
and the system output. Consequently, the AFSMC
algorithm proves to be effective for the position control
and the hysteresis compensation of SMA actuator.

5. CONCLUSION
In this paper, the simple structure of adaptive fuzzy
sliding-mode control was proposed to control SMA
actuator. Based on the Lyapunov theory, the stability of
controller was proved. The controller was applied to the
position control of SMA Actuator tracking to the desired
signals. The results showed that the SMA position could be
tracked well to references, and the SMA hysteresis
phenomenon was almost compensated.

ACKNOWLEDGEMENTS
This work was supported by the BK21, Korea Research
Foundation.

REFERENCES
[1] C. L. Dimitris, Shape Memory Alloys: Modeling and
Engineering Applications, Springer, 2008.
[2] H. E. Mohammad,and A. Hashem, “Nonlinear Control

978-89-93215-02-1 98560/10/$15 ©ICROS 437

Authorized licensed use limited to: Carleton University. Downloaded on November 02,2020 at 03:39:04 UTC from IEEE Xplore. Restrictions apply.