Module Code: EE4001

Module Title: Control System 1

Name of the Lecturer : Eng. Achintha Madusanka
Date of Submission: 29/10/2017

MAGNETIC LEVITATION CONTROL

EXPERIMENTS

Name: K.A.L.P.Abeykoon
Student ID No.2014093CL1
Advanced Diploma Stage 2
INTRODUCTION

The Maglev set up serves as a simple model of devices, which are becoming more and more popular in
recent years i.e. Maglev trains and magnetic bearings. Maglev trains are recently tested and some lines are
already available as for example in Shanghai. Magnetic bearings are used in turbines for the same reason as
Maglev trains are being built, which is low friction is the bearing itself. Already many turbines are used
commercially where the rotating shaft is levitated with magnetic flux. Some other magnetic bearing
applications include pumps, fans and other rotating machines. The magnetic levitation systems are
appealing for their additional possibility of active vibration damping. This can be done various control
algorithms implementations and without any modification to the mechanical parts of the whole system.
The maglev unit allows for the design of different controllers and tests in real time using Matlab and
Simulink environment.

AIMS
1) To understand the behavior of a SISO system (such as Maglev.)

2) To design a PD controller and a PID controller to ball position commands to the Maglev.

3) To explain how PD and PID designs are going to affect overall dynamics.

4) To understand the real-time application of control laws in a SISO environment.

OBJECTIVES

1) To understand the dynamic properties of the system.

2) Obtain response plots from the nonlinear simulation.

3) To obtain a model to describe the relations between the control voltage and the ball position.

4) Obtain a root locus plot with a PD controller closing the loop around the ball position response and record
performance parameters for different PD values.

5) Obtain a root locus plot with a PID controller closing the loop around the ball position response and record
performance parameters for different PID values.
2
APPARATURS

1) Magnetic Levitation mechanical unit

2) Analog to Digital control unit
3) PID controlling software.
4) Magnetic levitation system and remote power control box.
5) PC with MATLAB software
6) Control system interfacing circuit

THEORY
Magnetic levitation is a method by which an object is suspended with no support other than magnetic
fields. Magnetic pressure is used to counteract the effects of the gravitational and any other accelerations.
Magnetic materials and systems are able to attract or press each other apart or together with a force
dependent on the magnetic field and the area of the magnets, and a magnetic pressure can then be
defined.

The Maglev unit consists of a connection - Interface Panel with a Mechanical Unit on which a coil is
mounted. An infra - red sensor is attached to the Mechanical unit. Two steel spheres are included in the
package apart from the mechanical units play an important role in Maglev control. They allow measured
signals to be transferred to the PC via and I/O card. The analog control interface is used to transfer control
signals from the PC to Maglev and back. The mechanical and electrical units provide a complete control
system.

3
 The Magnetic
levitation model
we use shown in
below figure.

Every control project has plant modeling, Maglev is also presented as a Mechanical-Electrical model. The model is
shown in below figure

4
Usually Phenomenological models are nonlinear, in order to present nonlinear model as transfer function
model has to be linearized. According to the above figure the nonlinear model equations can be derived.

The simplest nonlinear model of the magnetic levitation system relating the ball position and the coil
current і is the following.

Where’ k’ is a constant depending on the coil parameters.

To present the full phenomenological model a relation between the constant voltage u and the coil current
would have to be introduced analyzing the whole maglev circuitry. However maglev is equipped with an
inner control loop providing a current proportional to the control voltage that is generated for control
purpose.

The Maglev is a SISO Plant - single input signal output. Position is the model out put and voltage
is the control signal. X

5
 U

U: control signal voltage

X: Ball position

EXERCISE 1- NONLINEAR MODEL

TESTING
Procedure
1. The response of the magnetic
levitation model
(Maglev_NonlinearModel.mld)
was checked at the beginning.
2. The test was run with other value of control signal by changing the input value.
3. The output signals were identified and its sing convention

Response of the maglev model

6
Discussion
In this case ball keep falling because the sensors didn’t applied at that moment, maglev can’t hold the ball
positions without sensors feedback signal. without feedback signal control signal won’t apply. This system is
unstable it is because system cannot hold the ball at one position in here either ball will magnetized and
falling.

EXERCISE 2 -MODEL IDENTIFICATION

A closed loop system describes the relation between signal r and y are identified.

𝐺(𝑍 −1 )
𝑇(𝑍) =
1 + 𝐶(𝑍 −1 ). 𝐺(𝑍 −1 )

The above relation can be transformed to yield the pendulum model.

𝑇(𝑍 −1 )
𝐺(𝑍 −1 ) =
1 − 𝐶(𝑍 −1 ). 𝑇(𝑍 −1 )

7
Procedure

1) The model called MaglevIdent.mld has taken using the identification experiment. This model
controls the ball levitation and adds an extra excitation signal to the controller output.
2) In the work place typed ‘ident’ and performed the identification.
3) The model structure set to the oe331.

4) The maglev transfer function 𝑇(𝑧 −1 ) was extracted and the following matlab codes were used to
transform the model into the continuous from (use maglev.m file)

𝑠𝑦𝑠𝑑 = 𝑓𝑖𝑙𝑡(𝑜𝑒331. 𝑏, 𝑜𝑒331. 𝑓, 0.01);

𝑠𝑦𝑠𝑐 = 𝑑2𝑐(𝑠𝑦𝑠𝑑,′ 𝑧𝑜ℎ′ );

4) The following cording were used to perform the transformation from 𝑇(𝑠)𝑡𝑜𝐺(𝑠)

𝑃𝐼𝐷 = 𝑡𝑓([0.2 4 2], [1 0]);

𝐺_ 𝐶𝑜𝑛𝑡 = 𝑠𝑦𝑠𝑐/ (1 − 𝑃𝐼𝐷 ∗ 𝑠𝑦𝑠𝑐);

6) Open loop poles were determined by obtaining a root locus plots.

Maglev continuous root locus plot

8
Maglev ident response

Discussion
When consider the root locus plot we can see one pole is gone to positive side of the real axis, hence the
system is unstable.
When consider the maglev ident response plot we can see that within first 5 second ball tried to comes to
its steady state.

EXERCISE 3-PD CONTROL OF BOLL POSITIONS.

1) The maglev rot locus was opened.

2) The PD controllers in the workspace were defined and import that in to the root locus as the
controller.
3) Different PID values, such as(P=4,D=2)(P=8,D=0.2),(P=4,D=3)were tried. (Here the damping ratio
should have within the 0.7 and 0.8, and the frequency should have be reasonably low from
oscillatory poles. (Use MagLev.m file))

9
4) Changed the each parameter and the closed loop damping was observed (observed how it will
influence in over all dynamics of the system)
5) By the pen loop system bode plots and the PD controllers were obtained.
6) The controller into workspace was imported and test it on the model by means of the maglev with
Maglev_PD_Model.mld.

10
PD root locus (4, 0.2)

PD bode plots (4, 0.2)

11
PD root locus (8, 0.2)

12
PD bode plots (8, 0.2)

PD root locus (4, 3)

13
PD bode plots (4, 3)

Discussion

14
According to the root locus plot of (P=4, D=0.2) system is unstable one pole is on the right side (positive real
side) of the root locus plot. In here pole value in the real side is 76.8

According to the root locus plot of (p=8, D=0.2) system is also unstable. pole value of the positive real side is
82.7. When comparing the (P=4, D=0.2) and (p=8, D=0.2) we can’t see much differences of the stability of
PD value sets however when changing P value system has become more unstable. Bode plots also doesn’t
show much differences from each value sets.

According to the root locus plot of (P=4, D=3) system is unstable one pole is on the right side (positive real
side) of the root locus plot. In here pole value in the real side is 169. Here the only changes were done to the
system was changing D value from 0.2 to 3. When comparing with other two graphs poles are closer to the
imaginary axis in PD values (P=4, D=3).

Even if the system is unstable for the those three PD value sets, when comparing with other values (P=4,
D=0.2) is the most stable value set.

EXERCISE 4- REAL TIME PD CONTROL OF BOLL POSITION.

1) The sinusoidal or rectangular input signal frequency was changed and observed the response of the
system.
2) The controller P and D values were changed and observed the input signal was tracked by the system
output.
3) Changed the step response properties with different P and D values were observed.

15
PD real time control step signal (4,0.2) Graph 1.1

6PD real time control sin signal (4,0.2) Graph 1.2

16
PD real time control square signal (4,0.2) Graph 1.3

PD real time control step signal (8,0.2) Graph 2.1

PD real time control sin signal (8,0.2) Graph 2.2

17
PD real time control square signal (8,0.2) Graph 2.3

PD real time control step signal (4,3) Graph 3.1

PD real time control sin signal (4,3) Graph 3.2

18
PD real time control square signal (4,3) Graph 3.3

Discussion

(P=4 , D=0.2) -Graph (1.1 ,1.2 ,1.3)

When we look at the graph 1.1 (step signal) we can see that the system comes to the stable state within 2
seconds and the steady state error is also less. After the 15 seconds when the next step is started steady state
error is very low than the beginning.

Graph 1.2 (sin signal) shows that the system comes to steady state within 3 seconds and the steady state error
is low however the desired position wasn’t obtained

Graph 1.3 (square signal) shows that the system has come to the steady state within 1second and the steady
state error is low.

(P=8 , D=0.2) -Graph (2.1 ,2.2 ,2.3)

Graph 2.1 (step signal) shows that the position has comes to steady state within 3 seconds, but until the 15Th
second there were some variations occurred in ball position after the 15th second steady state error has been
increased.

Graph 2.2 (sin signal), this graph is similar to the graph 1.2 steady state error is low.

Graph 2.3 (square signal) in here steady state error is very high because the system is not stable.

The P value has changed to the 8. According to the graphs only difference done to the system was changing
the p value only. Changing the P value has affect to the stability of the system

19
(P=4 , D=3) -Graph (3.1 ,3.2 ,3.3)
From graph 3.1 (step signal) shows that it takes nearly 6 seconds to comes to the steady state and the steady
state error is higher than the first PD values. After the 15th second steady state error has risen again and also
there was variations in the ball position.

Graph 3.2 (sin signal) shows that the it takes nearly 5 seconds to come to the steady state, even if the ball is
in steady state there were some variations on ball position and the steady state error is high.

Graph 3.3 (square signal) it takes nearly 3 seconds to comes to the steady state but the variation of ball
position are continues with the high steady state error.

After analyzing all the details from three PD value set first PD value set (P=4 , D=0.2) is the stable values
for this system

When comparing the realtime results and the simulations results, realtime result matched with the simulation
results. System was most stable for the (P=4 , D=0.2) values in realtime and the simulations.

EXERCISE 5- REAL TIME PID CONTROL OF BALL POSITION

1) The maglev unit was tested by the PID controller. (Used the Maglev_PID.mld).
2) The error integration was turned on after 15 second in this exercise.
3) Each parameter of the PID controller was changed and observed the change in responses.

20
PID real time step signal (P=4 I=1,D=0.2) Graph 4.1

PID real time step signal (P=4 I=1,D=0.2) Graph 4.2

21
PID real time step signal (P=4 I=1,D=0.2) Graph 4.3

PID real time step signal (P=4 I=5,D=0.2) Graph 5.1

22
PID real time step signal (P=4 I=5,D=0.2) Graph 5.2
PID real time step signal (P=4 I=5,D=0.2) Graph 5.3

Discussion
Here we selected the most stable PD vales and use it to PID controller by adding values to I those values are
1 and 5.

When considering the graphs of PID (P=4 I=1,D=0.2), for the step, sine and square signal we can see the it
take nearly 20 second to come to the steady state and the steady state error is very low when compare with
the PD values.

23
From the graphs of PID (P=4 I=5,D=0.2) we can see that the settling time is nearly 15seconds.Steady state
error of the sine wave graph 5.2 and steady state error of square wave graph 5.3 is very less when comparing
with the graph 4.2 and graph 4.3.

When compare the graph 4.1 and 5.1 (step wave) we can see in the graph 5.1 some variances of the ball
positions are occurred while graph 4.1 doesn’t show much of position variance.

OVERALL DISCUSSION
From this practical we got the idea about magnetic levitation. Some countries widely using this technology
for their transportation and the power generation. By using this technology minimize the power loss during
mechanical transitions, as for the examples magnetic bearings and trains are using this technology now a
days.

In the beginning of this practical we got the nonlinear model and we convert it to liner model and obtained
the transfer function and checked the system stability.

During this practical we got the good idea about the PD controllers and the PID controllers and the behavior
of the P, I, D components. In order to stabilize a control system P,I,D value must be find in other words PID
tune must be done.

24