IOP Conference Series: Materials Science and Engineering

PAPER • OPEN ACCESS

Modelling and state estimation for control of magnetic levitation system

via a state feedback based full order observer approach
To cite this article: Deepak Sharma et al 2018 IOP Conf. Ser.: Mater. Sci. Eng. 377 012156

View the article online for updates and enhancements.

This content was downloaded from IP address 139.81.188.106 on 13/07/2018 at 01:49

International Conference on Mechanical, Materials and Renewable Energy IOP Publishing
IOP Conf. Series: Materials Science and Engineering 377 (2018) 012156 doi:10.1088/1757-899X/377/1/012156
1234567890‘’“”

Modelling and state estimation for control of magnetic

levitation system via a state feedback based full order
observer approach

Deepak Sharma1*, S B Shukla2 and S K GHOSAL3

1, 2, 3
Indian Institute of Technology (Indian School of Mines), Dhanbad, Jharkhand -
826004, India.

E-mail address: [email protected]

Abstract. This work proposes to develop mathematical model and event trigger control using a
state observer, which precisely tracks position of the solid steel ball of magnetically actuated
levitation system. The lumped uncertainties i.e. non-linarites exists in the magnetic levitation
system, overwhelmed by the state observer. State space model was derived by the system
equation. Based on desired specification required poles are placed using pole placement
method to stabilize the closed loop system in response to a small size step input then full order
observer estimate the state variables. Here ball position and coil current is directly measurable,
but velocity is not directly measurable, which has been investigated in this work.

1. Introduction
Magnetic levitation is an electromechanical coupling concept which enables the contactless motion
and levitation of an object with the help of magnetic field. The magnetic force overwhelms the
gravitational force. The magnetic field is created by the electromagnetic coil and current in the coil
controls the magnitude of the magnetic force. This force can be attractive or repulsive type. Magnetic
levitation (Maglev) is a non-contact technology; it has a wide area of application such as high speed
maglev train [1], magnetic bearing, launching rocket, electromagnetic aircraft system, maglev wind
turbine and maglev micro robot etc. It offers many advantages such as friction less movement, high
speed, isolation of environment and low maintenance. The modelling and control of this system is
difficult due to the system behaviour, which is highly non-linear in nature. Consequently, the system
becomes unstable with fast dynamics. A lot of control techniques [2] and methods discussed to
overcome nonlinearity such as feedback linearization [3], Gain Scheduling linear controller design,
feed forward control, linear and nonlinear state space control [4-5].
In this paper, a single axis (1DOF) magnetic levitation system is considered as shown in Figure 1.
The system consists of an electromagnetic coil which is an upper part of apparatus, analogue and
digital interface, power amplifier, one inch diameter solid steel ball and position sensor. This study
develops non-linear dynamic model of the system which is linearized about the operating point by
Taylor’s series approximation method [6]. The state variables considered for this system are ball
position, velocity of the ball and current through the electromagnet. Mathematical modelling of the
system is developed by the study of electrical and mechanical subsystem. According to design
specification, a state observer [7] is proposed, which estimate the state variables of the system.
Position sensor of the apparatus tracks the ball position directly but there is no device to measure the

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
International Conference on Mechanical, Materials and Renewable Energy IOP Publishing
IOP Conf. Series: Materials Science and Engineering 377 (2018) 012156 doi:10.1088/1757-899X/377/1/012156
1234567890‘’“”

velocity of the ball so that the full order observer [8] is considered to estimate velocity of the ball and
the other state variables.

Figure 1. Magnetic levitation system set up

2. Dynamic modelling of maglev system

The schematic diagram of the maglev system is shown in Figure 2. The ball is levitating in free space
has a six degree of freedom but only vertical axis of the system is control. The system used L-R circuit
for control action which is divided into two following systems i.e. an electrical and an
electromechanical [9].

Figure 2. Schematic view of maglev plant

2
International Conference on Mechanical, Materials and Renewable Energy IOP Publishing
IOP Conf. Series: Materials Science and Engineering 377 (2018) 012156 doi:10.1088/1757-899X/377/1/012156
1234567890‘’“”

Table 1. System parameters

Notations Description Numerical Notations Description Numerical
value value
Lc Inductance of the 412.5mH rb Radius of the 0.0127m
coil ball
Rc Coil resistance 10Ω lc Length of the 0.0825
coil
g Acceleration due 9.81 Mb Mass of the ball 0.068kg
to gravity
Km Force constant of 6.5308×10-5 Kb Ball position 0.0028
electromagnet sensor
sensitivity
Rs Current sense 1Ω Nc Coil wire turns 2450
resistance
rc Coil core radius 0.008m

The voltage sense (Vs) is used to measure the current in the coil. The coil current can be computed
using the following relationship
( ) (1)
The first-order differential equation is obtained by Kirchhoff’s voltage law,
( ( ) ( ))
( ) ( ) ( ) (2)
Where, is applied coiled voltage and ic is the coil current. The force induced by the electromagnet
acting on the ball,
( )
(3)
Where, xb>0 is the air gap between the ball and the face of the electromagnet and Km is the
electromagnetic force constant. The force due to gravity on the ball is given by,
(4)
Applying the Newton's second law of motion on the ball,
( )
̈ ( )
(5)
Assuming xb=x1; v=x2; ic=x3; u=Vc.
Non-linear state-space model,
̇
( )
[ ̇] [ ] (6)
̇
[ ]

[ ] (7)

Where, v is the ball velocity and position of ball (xb) is taken as output.
This nonlinear model is linearized about an operating point (equilibrium position). By Taylor’s
approximation about xb=x10=xbo, ic=ic0= x30.
The associated state vector X0=[x10 0 x30]. Equilibrium position of ball xb0=6 mm and equilibrium
current ic0=.86 A.

Linearized model in state space form as under,

̇
[ ̇] [ ] [ ] (8)
̇
[ ]

3
International Conference on Mechanical, Materials and Renewable Energy IOP Publishing
IOP Conf. Series: Materials Science and Engineering 377 (2018) 012156 doi:10.1088/1757-899X/377/1/012156
1234567890‘’“”

A= B=[ ] C= (9)
[ ]

3. Stability
Maglev open loop system is unstable because one of the Eigen values of det (sI - A) = 0,
[p1= 57.1839, p2= -57.1839, p3= -26.6700] lies in positive half of the s-plane. The response of open
loop for small amount of step is shown below,

Figure 3. Open loop response ball position

The response shows that the ball is going outside the range of linearization validity that means the
ball hits the electromagnet or the pedestal.
According to design specification, the ball position behaviour should satisfy the following criteria:
The settling time and the overshoot should be less than 0.5 sec and 5% respectively, based on that
following 2nd order parameter ζ=0.7, ωn=20 rad/s are needed, the two dominant poles at -14±14i and
the third pole is placed at -80 and feedback gain matrix K is calculated in matlab.
K = [6569 106.9 -33.6]
For a small amount of step input, the value of ̅ =568.1139 is calculated using matlab commands,
run the simulation to check whether it is able to track the trajectory or not.

Figure 4. Block diagram of feedback control system Figure 5. Closed loop response ball position

The closed loop response of ball position is shown in Figure 5.If all the states variables are not
measurable, design an observer to estimate them. In maglev system only ball position and coil current
can be measurable directly but velocity cannot be measurable directly, so full order observer is used
and three new estimated states added.

4
International Conference on Mechanical, Materials and Renewable Energy IOP Publishing
IOP Conf. Series: Materials Science and Engineering 377 (2018) 012156 doi:10.1088/1757-899X/377/1/012156
1234567890‘’“”

Figure 6. Block diagram of full order state observer

The observed matrices for the dynamics of the system is given below,

̅
A(t) = [ ], B(t) = [ ], C(t) = [ C zeros(size(C)) ]
( ( )) ( ( ))

The poles (Eigen value of matrices) of (A-LC) give the error dynamics of the observer [8]. The
behaviour of the observer should be faster than that of the system. In order to get this, the poles are
placed much far away from the dominant poles in the left half of s-plane [7].
Here observer poles are placed at op1= -100, op2= -105, op3= -110. By using full state feedback
observer, the error dynamics of the state is illustrated as:
e = x – x_est.
(10)
̇ = (A-LC) e
Assuming the initial condition of the observer to be zero as a result of which the state’s initial
condition and the error’s initial condition will same. If matrix is an asymptotically stable matrix, the
error vector will converge to zero for any initial error vector e(0). The error dynamics tends to zero as
t→∞ [5] [8]. Plot the responses for all the states x and respective estimated states x_est. Using matlab,
simulation responses are plotted. All the state variables (x) are represented by solid line and estimated
states (x_est) are represented by dotted line. The behaviour of the state can be estimated by the
observer quiet perfectly and faster.

Figure 7. Response of state observed variable

5
International Conference on Mechanical, Materials and Renewable Energy IOP Publishing
IOP Conf. Series: Materials Science and Engineering 377 (2018) 012156 doi:10.1088/1757-899X/377/1/012156
1234567890‘’“”

Figure 8. Actual velocity and estimated velocity of ball

Figure 9. Actual current and estimated current

4. Conclusions
The modelling and control of 1DOF magnetic levitation system has been investigated in this paper.
The mathematical models of the plant are derived by the general physical laws of the system and the
non-linearity is linearized by Taylor's approach about an operating range. Based on design criteria
desired poles are placed to obtain a stable response of the system. A full order state observer is
designed for the systems, the result is able to track the reference step input quite accurately.
Simulation results illustrates the stabilisation of the ball and estimate the unknown states effectively.

References
[1] Lee H W, Kim K C and Lee J 2006 Review of maglev train technologies IEEE transactions on
Magnetics 42 1917-1925.
[2] Nakagawa T, Hama M and Furukawa T 2000 Study of magnetic levitation technique applied to
steel plate production line IEEE Transactions on Magnetics 36 5 3686-3689.
[3] Baranowski J and Piatek P 2012 Observer-based feedback for the magnetic levitation system
Transactions of the Institute of Measurement and Control 34 4 422–435.
[4] Hajjaji A E and Ouladsine M 2001 Modelling and nonlinear control of magnetic levitation
systems IEEE Transaction on Industrial Electronic 48 4 831- 838.
[5] Barie W and Ckiasoson J 1996 Linear and nonlinear state- space controllers for magnetic
levitation International Journal of Systems Science 27 11 1153-1163.
[6] Salim T and Karsli V 2013 Control of single axis magnetic levitation system using fuzzy logic
control International Journal of Advanced Computer Science and Applications 4 11 83-88.
[7] Chen WH 2004 Disturbance observer based control for nonlinear systems IEEE/ASME
Transactions on Mechatronics 9 4 706–710.
[8] K.Ogata 2010 Modern Control Engineering 5th ed. Prentice Hall.
[9] Quanser Inc. 2012 Magnetic Levitation Plant User Manual. Ontario Canada Quanser Inc.