Second International Conference on Computational Intelligence, Modelling and Simulation

Control of Magnetic Levitation System Using Fuzzy Logic Control

A.K. Ahmad, Z. Saad, M.K Osman, I.S Isa, S. S. S. Abdullah

Sadimin, Department of Control and Instrumentation
Faculty of Electrical Engineering, Faculty of Electrical Engineering, UTM
Universiti Teknologi MARA (UiTM) Malaysia, Skudai, Malaysia
Kampus Pulau Pinang, 13500 Permatang Pauh, [email protected]
Malaysia
[email protected]

Abstract— This paper presents the investigation on a on feedback linearization where a nonlinear state-space
system model for the stabilisation of a Magnetic transformation along with nonlinear state feedback is used to
Levitation System (Maglev's). Furthermore, the linearize the system. From this paper shows that the position
investigation on Proportional Integrated Derivative tracking error of the system was oscillation about ± 0.45 mm
Controller (PID) also reported here. In this paper shows [1].
to design both PID and Fuzzy Logic Control (FLC) based From the author Dan Cho Et al come out the idea of
on the system model. Maglev's give the contribution in sliding mode control (SMC) to overcome the parameter
industry and this system has reduce the power uncertainties and reject disturbances to achieve robust
consumption, has increase the power efficiency and performance. In this paper present the SMC was applied to
reduce the cost maintenance. The common applications a magnetic levitation system. It is found that the
for Maglev's are Maglev's Power Generation, Maglev's performance of the SMC is better than that of the classical
trains and Maglev's ball bearingless system. In this controllers[2].
study, it has also been observed that the basic design of The conventional controller such as PID controller is
Maglev's is an arrangement of electromagnets placed on very reliable and simple controller to design. This controller
top of the plant and makes the ball levitated in the air. used the method based on a linearization of the systems
The focus of this study is that to design the controller dynamics and compensates the effects of the non-modelled
that can cope with Maglev's which highly nonlinear and nonlinearity. Using this approach certain systems can be
inherently unstable. The modeling system is simulated stabilized close to their nominal operating point. From the
using MATLAB simulink. This paper presents the author, Wenbai Chen et al said, PID controller can be a
comparison output for both PID Controller and Fuzzy robust and reliable system if the PID parameter can be
controller to control the ball levitate on the air. The ISE determined or tuned that make the system very stable. The
performance index is shown to compare the performance author proposed the method chaos optimization that can give
both controller. contribution in PID parameter setting. The author also said
Index Terms—Magnetic Levitation System (Maglev’s), with chaos optimization applied to PID parameter, the
Fuzzy Logic Control, PID control. performance of PID controller was increase [3].
The fuzzy set theory was introduced by Zadeh et al [7]
I. INTRODUCTION has become as a powerful modeling tool that can work with
Nowdays, Magnetic Levitation System or in short name the unstable system and highly nonlinearities of modern
Maglev’s is very usable system that can be applied in many control. It was intelligent control and the good thing about
application area such as in magnetic bearings, high speed fuzzy logic control, the parameter of fuzzy logic is very easy
to tune by non expert person if compare with PID controller
trains, vibration isolation, levitation of wind power
that need experience person to tune the parameter. The
generation, levitation of molten metal in induction furnaces,
author, Tzuu-Hseng S Li et al said the Fuzzy Sliding Mode
and levitation of metal slabs during manufacturing. This Controller (FMSC) can achieve the asymptotic stability of the
Maglev’s can be categorized as a repulsive systems and this system. This FMSC controller very helpful because with this
system based on the source of levitate forces. These type of controller, we need not to know in detail [8]. From the
systems are normally is unstable and it is described as highly author, Chao-lin Kuo proposed the Novel Fuzzy Sliding-
nonlinear where it is difficult to control the system. It is very Mode Control (NFSMC) and from his paper presents the
challenging in order to construct the high performance comparison effect of uncertainty in the ball mass between
feedback controllers to regulate the position of the levitation Sliding-Mode controller, Fuzzy Sliding-Mode controller and
ball. NFSMC. NFSMC show the minimum IAE and ISV
In 1996, Walter Bariet et al come out with idea linear and performance[9].
nonlinear state space controllers for magnetic levitation From the study, Fuzzy Logic Controller has good
system. From the author said, the two state-space controllers potential to stabilise the ball levitation in this research.
are compared together in terms of their performance in
controlling the hall’s position. The first controller is based

978-0-7695-4262-1/10 $26.00 © 2010 IEEE 47

51
DOI 10.1109/CIMSiM.2010.99
Authorized licensed use limited to: ULAKBIM UASL ISTANBUL TEKNIK UNIV. Downloaded on December 25,2024 at 17:00:15 UTC from IEEE Xplore. Restrictions apply.
 II RESEARCH METHODOLOGY deviation variables, the small differences from the operating
point are normalized over operating spaces (G, D, I max, U max)
A Dynamic Model Analysis
and they are defined as follow:
The magnetic levitation system experiment is a magnetic
wF wF
ball suspension system which is used to levitate a steel ball F F0  .( X  X 0 )  .( I  I 0 ) (4)
on air by the electromagnetic force generated by an wX I wI X
0 0
electromagnet.
Where f is the normalized resultant force, x is the
normalized air gap, i is the normalized current and u is the
normalized voltage. Xo, Io, and Uo are steady state values.

F G X  X0 I  I0 V  V0
f x i u (5)
G D I max Vmax
d 2x D d 2x D d 2x
Figure 1 Maglev Plant Equipment f m. D . 2 m  (6)
G dt m.g dt 2 g dt 2
The magnetic ball suspension system consists of an
electromagnet, a ball rest, a ball position sensor, and a steel D wF I max wF
ball. The magnetic ball suspension system can be f .x  . .i (7)
G wX I0 G wI X0
categorized into two systems: a mechanical system and an
electrical system. The ball position in the mechanical L di
u i (8)
system can be controlled by adjusting the current through the R dt
electromagnet where the current through the electromagnet Let set the gains and time constant as
in the electrical system can be controlled by applying
I max wF L D wF D
controlled voltage across the electromagnet terminals. Ke , Te , Km  , Tm (9)
G wI X0 R G wX I0 g
Then the equation (5), (6), and (7) can be rewritten as

d 2x
f Tm 2 . (10)
dt 2
f  K m .x  K e .i (11)
di
u i  Te . (12)
dt
Figure 2 Schematic of Magnetic Levitation System
The voltage equation of the electromagnetic coil is given Then the block diagram of the linearised model as shown in
in equation 1. Figure 3.
di
U IR  L (1)
dt
The magnetic force applied by the electromagnet is opposite
direction compare to gravity force and it maintains the
suspended steel ball levitated. The magnetic force F
depends on the electromagnet current I, electromagnet
characteristics and air gap x between the steel ball and the
Figure 3 Block Diagram of Linearization System
electromagnet. The motion of the steel ball in the magnetic
filed is expressed as
B Fuzzy Logic Control
d 2x
F  mg m (2) Fuzzy logic controller design is based on the linguistic
dt 2 description of the control strategy. There are specific
E i2 components characteristic of a fuzzy controller to support a
F (3) design procedure. In the block diagram in Figure 4, the
h
In the equation 3, shows that the current flows in the coil controller is between a preprocessing block and a
is nonlinear. The steady state of the operating point air gap postprocessing block. The following explains the diagram
between mass and the electromagnet is maintained by block by block. There are three sources of nonlinearity in a
generating on magnetic force, which is adjusted so that the fuzzy controller.
gravitational force of the steel ball is balanced. Using • The Rule Base - The position, shape and number of
fuzzy sets as well as nonlinear input scaling cause

52
48

Authorized licensed use limited to: ULAKBIM UASL ISTANBUL TEKNIK UNIV. Downloaded on December 25,2024 at 17:00:15 UTC from IEEE Xplore. Restrictions apply.
 nonlinear transformations. The rules often express a
nonlinear control strategy.
• The Inference Engine - If the connectives and and or
are implemented as for example Min and Max
respectively, they are nonlinear.
• The Defuzzification - Several defuzzification
methods are nonlinear.

Figure 7. Membership Function for Input, Rate of Error

Figure 4 Block Diagram of Fuzzy Logic Controller

The first block inside the controller is fuzzification,

which converts each piece of input data to degrees of Figure 8. Membership Function for Output
membership by a lookup in one or several membership The rules may use several variables both in the condition
functions. and the conclusion of the rules. The controllers can
There is a degree of membership for each linguistic term therefore be applied to both multi-input-multi-output
that applies to that input variable. The degree of (MIMO) problems and single-input-single-output (SISO)
membership is the fuzzified input and this proceeds to the problems. The typical SISO problem is to regulate a control
rulebase. A comparison to a pretuned membership function, signal based on an error signal. The rule base utilizes if then
such as that given in Figure 5, Figure 6, Figure 7and Figure condition statements to alter the controlled variable. The
8 must be carried out for this purpose. ‘inference engine’ is part of the rule base. Simple
A possible of the membership functions for the syllogisms are used to ‘infer’ a decision from one or several
three mentioned variables of the magnetic levitation conditions. The rule base used in the magnetic levitation
system represented by a fuzzy set is as follows: system can be represented by the following Table 1 with
fuzzy terms derived by modeling the designer’s knowledge
and experience.

Table 1 Rule base

Delta e
Output = Voltage
NS Z PS
NS NL NS ZE
e Z NS ZE PS
PS ZE PS PL

Figure 5. Membership Function for Input, Error

The resulting fuzzy set must be converted to a number
that can be sent to the process as a control signal. This
operation is called defuzzification. Defuzzification is the
inverse process by which the decision taken on the input is
transformed into a crisp output.
C. Simulation using MATLAB
This software is able to send data (desired value of
current) to fuzzy controller. Firstly plot a graph of
detected current versus time to monitor the performance of
Figure 6. Membership Function for Input, Rate of Error the system. Sample of the MATLAB simulation diagram is
shown in Figure 9, Figure 10 and Figure 11.

53
49

Authorized licensed use limited to: ULAKBIM UASL ISTANBUL TEKNIK UNIV. Downloaded on December 25,2024 at 17:00:15 UTC from IEEE Xplore. Restrictions apply.
 Figure 9 Different shapes like bell-shaped, triangular,
trapezoidal and singleton.

Figure 13 Performance of all controllers using ISE method

From the Figure 13 shows:

i) ISE performance index for fuzzy : 6.399e-7

ii) ISE performance index for PID : 8.26 e-7
iii) ISE performance index for Matlab Controller :4.735 e-6

The ITAE provides the best selectively of the

performance indices; that is, the minimum value of the
integral is readily discernible as the system parameters are
Figure 10 Fuzzy toolbox. varied. Here only shown the performance index for ISE
because it almost give satisfaction because minimization of
ISE is often of practical significance. Performance indices
are useful for analysis and design of control systems. Below
show the performance of controller using rise time,
maximum overshoot and settling time. Performance of
controller show as

Fuzzy Logic Controller:

Max overshoot = 0.0513
Figure 11 Membership functions Rise time = 0.0523 s, at 0.005117.
Settling time = 0.9898s.
III. RESULT AND DISCUSSION Percent overshoot = 2.6%.

A. Simulation result with the MATLAB controller, PID PID Controller:

controller and Fuzzy Logic Controller Max overshoot = 0.0542
The input gains to PID controllers are; Kp = 3000, Td Rise time = 0.0335 s, at 0.005378.
= 100 and Ti = 50000. The step input is set as: step time = Settling time = 0.3021s.
0.3s, initial value = 0.05, final value = 0.05. Percent overshoot = 8.4%.
The output of PID controller, MATLAB design controller
and fuzzy logic controller show in figure 12. MATLAB Controller
Max overshoot = 0.0567
Rise time = 0.0523 s, at 0.005603.
Settling time = 0.5024s.
Percent overshoot = 13.4%.

From all data shows that least error is coming from Fuzzy
controller. The biggest errors come from the controller
designed by Matlab.

Figure 12 The output response with PID controller, Kp = 3000 , Td = 100

and Ti = 50000, output response with Fuzzy Logic controller, Error gain =
5.8, Change of error gain = 0.03 and Offset gain = 3.3 and MATLAB
design controller.

54
50

Authorized licensed use limited to: ULAKBIM UASL ISTANBUL TEKNIK UNIV. Downloaded on December 25,2024 at 17:00:15 UTC from IEEE Xplore. Restrictions apply.
 Figure 17 Error gain = 2, dif error = 0.02, output gain = 400,
integral error gain = 0.
In Figure 16, shows that only error applied in
membership function in Fuzzy Logic Controller, the output
Figure 14 Simulink of Magnetic Levitation System from the MAGLEV tends to unstable. Then when applied
the change of error in the membership function show in
Figure 14 show the simulink block in the MATLAB tool. Figure 17, the output now to be steady state value. But still
From figure, it has the design of PID controller, Fuzzy logic now good because have steady state error.
controller and MATLAB controller.

B Discussion for Simulation result with the Fuzzy Logic

Controller.
In Figure 15, shows the system is unstable system.
When no controller applied to the system the ball will fall
down or attract to magnetic.

Figure 18 Error gain = 8, dif error = 0.05, output gain = 900,

integration error gain = 4
When applied the integration of error or offset in
membership function, the output now tend to be steady state
value with zero steady state error. It shows in Figure 18.

Figure 15 No controller apply to the Magnetic Levitation System

Figure 19 Comparison output result from all controller

In Figure 19 shows the comparison output from all

controllers. From observation, it shows that the fuzzy
Figure 16 Only Error applied in membership
controller has best performance compared to other
controller. It can seen from ISE figure that shown in Figure
13.

55
51

Authorized licensed use limited to: ULAKBIM UASL ISTANBUL TEKNIK UNIV. Downloaded on December 25,2024 at 17:00:15 UTC from IEEE Xplore. Restrictions apply.
 IV CONCLUSION AND FUTURE WORK Propulsion Conference, 2008 , Page 1-3.
[6] Yang, Zhenyu, Pedersen, Gerulf, “Automatic tuning of PID
The output of the magnetic levitation system is observed controller for a 1-D levitation system using a genetic
and analyzed. Apart from that, comparisons are made to see algorithm – a real case study”, IEEE international
which systems give better performances by considering the Conference on Control Application, 2006, Page 3098 – 3103.
time settling, time peak, steady state error and how much [7] John Yen, Reza Langari, Lotfi A.Zadeh.(1994),Industria
oscillations occurs. Application of Fuzzy Logic and Intelligent Systems. New
From the figures, we can conclude that the PID Jersey: IEEE PRESS.
controller gain is proportional with time integral. This [8] Tzu-Hseng S Li, Ming-Yuan Shieh, “Switching-type fuzzy
means when the Kp value is bigger, the output will have sliding mode control of a cart±pole system”, Mechatronic
Elsevier, March 2000, Page 91-109.
smaller value of offset but the more oscillatory the process
[9] CHAO-LIN KUO, TZUU-HSENG S. LI_ and NAI REN
becomes. Where else, if the Ti is bigger, the offset is bigger GUO (2005). “Design of a Novel Fuzzy Sliding-Mode
but the oscillation is less. Control for Magnetic Ball Levitation System “Journal of
As for the fuzzy controller, the response is slower than Intelligent and Robotic Systems 42: 295–316 © Springer
the PID controller. However, fuzzy controller shows the 2005.
best performance in terms of lowest overshoot among three [10] Mamdani, E.H., and S.Assilian (1975).An experiment in
controller and no steady state error. It reaches the desired linguistic synthesis with a fuzzy logic controller.
setpoint at t = 0.98 sec. The PID controller has steady state International Journal of Machine Studies 7(1).
error and with the best adjustment, it reaches to the [11]Franklin. G.,D.Powell, and A.Emami-Naeini (1991).
Feedback control of dynamic systems.
desired set point. Scaling factors are most important with
Reading,Mass.:Addison-Wesley.
respect to fuzzy controller performance and provide a [12] John Gerry, Michel Ruel (2002). Control Solutions.
guideline for tuning. It was shown that the scaling Simulation on The Road to Optimization.
factors play a role similar to that of the gain coefficients for rd
conventional controllers. [13] Nise S.N.(2000), Control System Engineering (3 ed).New
York: Wiley.
In the fuzzy controller, if there not provide the
[14] Jantzen, Jan.(2007),Foundation of Fuzzy Control. British:
integration of error, the steady error cannot be eliminates. Wiley.
That is why here integration of error need to apply in the [15] Smith, A.C. (2006), Principles and Practice of Automatic
membership function. Process Control (3rd edition). New Jersey: Wiley.
The extensions of this research lie primarily in the areas
of implementation and applications of the FLC. It is highly
recommended that a faster Digital Signal Processing (DSP)
be used for the implementation of the FLC.
In the future, it can be implemented the PID-Fuzzy
controller or Neural Network controller to control the
MAGLEV system.
Lastly, this simulation can be implemented to magnetic
levitation system, CE152 model. Then it can be compared
the result between simulation value and the actual plant
value

REFERENCES
[1] Walter Barie; John Chiasson , Linear and nonlinear state-space
controllers for magnetic levitation , International Journal of
System Science, Volume 27, Issue 11 November 1996, Pages
1153-1163.
[2] Cho, D.; Kato, Y.; Spilman, D., “Sliding-Mode Controller and
Classical Controller on Magnetic Levitation System”, Control
System Magazine, IEEE, 1993,Volume 13, Page 42-48.
[3] Chen Wenbai, Meng Xuan, Li Jinao, “PID Controller Design of
Maglev Ball System Based on Chaos Parameter Optimization”,
International Conference on Machine Vision and Human-
Machine Interface(MVHI), 2010, Page 772-775.
[4] Wai R, Lee J, Chuang K, “Real-Time PID Control Strategy for
Maglev Transportation SystemVia Particle Swarm
Optimization”, IEEE Transaction on Industrial Electronics,
2010, Page 1-1.
[5] Ding Zhang, Ning Wang, Yungang Li, Wensen Chang,
“Parameter self-adaptive PID algorithm for magnetic power
supply on maglev train”, IEEE Vehicle Power and

56
52

Authorized licensed use limited to: ULAKBIM UASL ISTANBUL TEKNIK UNIV. Downloaded on December 25,2024 at 17:00:15 UTC from IEEE Xplore. Restrictions apply.