You are on page 1of 103

Abstract

This study seeks to improve the feedback control strategies of a twin rotor multi-input
and multi-output system (TRMS) by changing the existing control scheme. The exit
TRMS is maintained by the combination of two PID controllers, a tail rotor controller
and a main rotor controller. More than 90% of industrial controllers are still
implemented based around PID algorithms and ease of use offered by the PID
controller. However, performance will be influenced heavily by the tuning algorithm.
In the first place, we develop the mathematical models for the TRMS system in this
study. The system contain two main problems, first one in DC-motor that can be
consider as a nonlinear system. Secondly, angular momentum and reaction turning
moment are the two main effects can be regard as a disturbance. A disturbance signal is
an unwanted input signal that affects the system’s out signal. Many control systems are
subject to extraneous disturbance signals that cause the system to provide an inaccurate
output. We wish to reduce the effect of unwanted input signal, disturbances, on the
output signal. We will show how we may design a control system to reduce the impact
of disturbance signals. Then simulations will be made used of developing control
schemes. Finally a suitable deadbeat robust schemes has been designed that could be
applied to the existing control system, a deadbeat robust with decoupling technique was
proposed. The result will be a significant improvement for the system overshoot and
settling time. From this design procedure the system that will be very robust.
Acknowledgments

I am grateful to Dr. Paul for his valuable comments and suggestions throughout the
duration of the research project and preparation of this dissertation.

I would like to take this opportunity to thank my parent for all their care and support.

David Lu

University of Southern Queensland

January 2006
Certification

I certify that the ideas, designs and experimental work, results, analyses and
conclusions set out in this dissertation are entirely my own effort, except where
otherwise indicated and acknowledged.

I further certify that the work is original and has not been previously submitted for
assessment in any other course or institution, except where specifically stated.

TE-WEI LU

Student Number: W0038289

Signature

Date
Notations

α h Horizontal position (azimuth position) of TRMS beam [rad]

Ωh Angular velocity (azimuth position) of TRMS beam [rad/s]

Uh Horizontal DC-motor voltage control input [V]

Gh Linear transfer function of tail rotor Dc-motor

h Non-linear part of Dc-motor with tail rotor: h (Uh) = ω h [rad/s]

ω h Rotational speed of tail rotor [rad/s]

Fh Non-linear function (quadratic) of aerodynamic force from

Tail rotor Fh = Fh (ω h ) [N]

lh effective arm of aerodynamic force from tail rotor lh = lh (α v) [m]

Jh non-linear function of moment of inertial with respect to

vertical axis Jh = Jh (α v) [kg m2]

Mh horizontal turning torque [Nm]

Kh horizontal angular momentum [N m s]


List of Figures v

Fh moment of friction force in vertical axis [N m]

α v vertical position (pitch position) of TRMS beam [rad]

Ωv angular velocity (pitch position) of TRMS beam [rad/s]

Uv vertical Dc-motor voltage control input [V]

Gv linear transfer function of main rotor Dc-motor

v non-linear part of DC-motor with main rotor v (Uv ) = ω v [rad/s]

ω v rotational speed of main rotor [rad/s]

Fv non-linear function (quadratic) aerodynamic force from

tail rotor Fv = Fv (ω v) [N]

lv arm of aerodynamic force from main rotor [m]

Jv moment of inertia with respect to horizontal axis [kg m2]

Mv vertical turning moment [Nm]

Kv vertical angular momentum [N m s]

fv moment of friction force in horizontal axis [N m]

f vertical turning moment from counterbalance f = f (α v) [Nm]

Jhv vertical angular momentum from tail rotor [N m s]

Jvh horizontal angular momentum from main rotor [N m s]


List of Figures vi

gvh non-linear function (quadratic) of reaction turning moment

gvh = gvh (ω v) [N m]

gh non-linear function (quadratic) of reaction turning moment

ghv = ghv (ω h) [N m]

t time [s]

1/s transfer function of an integrator


Content

ABSTRACT...............................................................................................I

ACKNOWLEDGMENTS..........................................................................II

CERTIFICATION.....................................................................................III
Te-Wei Lu................................................................................................................................iii

NOTATIONS...........................................................................................IV

CONTENT..............................................................................................VII

LIST OF FIGURES.................................................................................IX

LIST OF TABLES................................................................................XIV

Introduction......................................................................................................1
1. Introduction................................................................................................1
2. Project Aim.................................................................................................4
3. Thesis Structure.........................................................................................6

SYSTEM MODELING........................................................................................7
1. Introduction................................................................................................7
List of Figures viii

2. TRMS System Description.........................................................................8


3. Mathematical Model and State Equation....................................................9
4. Characteristics of Main and Tail Motor.....................................................19
5. System Simulation...................................................................................24

PROBLEM DEFINITION AND APPROACH....................................................33


1. Introduction..............................................................................................33
2. Nonlinear DC Motors................................................................................37
3. Cross-coupling Effects.............................................................................39
4. Summary..................................................................................................45

PID Controller Study......................................................................................46


1. Introduction..............................................................................................46
2. Review of PID Controller..........................................................................47
3. Optimization Controller.............................................................................50
4. Simulation Results...................................................................................52

Deadbeat Robust Scheme.............................................................................58


1. Introduction..............................................................................................58
2. Review of Deadbeat Controller.................................................................58
3. Design Method and Procedures...............................................................63
4. Performance Evaluation...........................................................................69

Conclusions and Future Development........................................................74


1. Conclusions..............................................................................................74
2. Recommendations for future developmen................................................76

REFERENCES.......................................................................................78

APPENDIX A.........................................................................................83
the Procedure of Optimization................................................................................................83

APPENDIX B.........................................................................................84
The Procedure to Determine Settling Time.............................................................................84
List of Figures

FIGURE 1-1 PID CONTROL SCHEME....................................................4

FIGURE 1-2 DEADBEAT ROBUST CONTROL SCHEME......................5

FIGURE 2-3 THE LABORATORY SET-UP TRMS SYSTEM..................8

FIGURE 2-4 SCHEMATIC DIAGRAM OF TRMS....................................9

FIGURE 2-5 GRAVITY FORCES IN TRMS CORRESPONDING TO THE


RETURN TORQUE.................................................................................11

FIGURE 2-6 PROPULSIVE FORCE MOMENT AND FRICTION


MOMENT.................................................................................................13

FIGURE 2-7 MOMENTS OF FORCES IN HORIZONTAL PLANE........16

FIGURE 2-8 BLOCK DIAGRAM OF EQUATION 2-15 AND 2-16.........19

FIGURE 2-9 BLOCK DIAGRAM OF TRMS MODEL.............................25

FIGURE 2-10 BLOCK DIAGRAM OF THE TAIL ROTOR.....................26


List of Figures x

FIGURE 2-11 1-DOF MODEL OF THE HORIZONTAL PART OF TRMS


.................................................................................................................27

FIGURE 2-12 THE CONTENTS OF THE GROUPED MODEL BLOCK


(HORIZONTAL)......................................................................................27

FIGURE 2-13 BLOCK DIAGRAM OF ROTATIONAL SPEED OF TAIL


ROTOR....................................................................................................27

FIGURE 2-14 BLOCK DIAGRAM OF DRIVING TORQUE OF TAIL


ROTOR....................................................................................................27

FIGURE 2-15 BLOCK DIAGRAM OF AERO FORCE (TAIL ROTOR). 28

FIGURE 2-16 ROTATIONAL SPEED OF TAIL ROTOR.......................28

FIGURE 2-17 BLOCK DIAGRAM OF THE MAIN ROTOR....................29

FIGURE 2-18 1-DOF MODEL OF THE VERTICAL PART OF TRMS...29

FIGURE 2-19 THE CONTENTS OF THE GROUPED MODEL BLOCK


(VERTICAL)............................................................................................30

FIGURE 2-20 BLOCK DIAGRAM OF ROTATIONAL SPEED OF MAIN


ROTOR....................................................................................................30

FIGURE 2-21 BLOCK DIAGRAM OF DRIVING TORQUE OF MAIN


ROTOR....................................................................................................30

FIGURE 2-22 BLOCK DIAGRAM OF AERO FORCE (MAIN ROTOR) 30

FIGURE 2-23 ROTATIONAL SPEED OF MAIN ROTOR......................31

FIGURE 2-24 2-DOF COMPLEX MODEL OF TRMS............................31


List of Figures xi

FIGURE 2-25 DETAILED 2-DOF MODEL OF TRMS............................32

FIGURE 3-26 BLOCK DIAGRAM OF THE TRMS SYSTEM.................35

FIGURE 3-27 DETAIL OF MODEL INCLUDE CROSS-COUPLING.....35

FIGURE 3-28 THE DIFFERENCE BETWEEN DIFFERENTIAL


EQUATION AND TRANSFER FUNCTION IN MAIN ROTOR...............36

FIGURE 3-29 THE DIFFERENT BETWEEN DIFFERENTIAL


EQUATION AND TRANSFER FUNCTION IN TAIL ROTOR................36

FIGURE 3-30 BLOCK DIAGRAM OF MAIN AND TAIL PROPELLER


SYSTEM..................................................................................................38

FIGURE 3-31 MEASURED CHARACTERISTICS OF THE MAIN


ROTOR....................................................................................................39

FIGURE 3-32 POLYNOMIAL APPROXIMATION OF THE MAIN


ROTOR CHARACTERISTICS................................................................39

FIGURE 3-33 MEASURED CHARACTERISTICS OF TAIL ROTOR....39

FIGURE 3-34 POLYNOMIAL APPROXIMATION OF THE TAIL ROTOR


CHARACTERISTICS..............................................................................39

FIGURE 3-35 THE INTERACTION FRAMES OF TRMS.......................41

FIGURE 3-36 THE FINGNAL FLOW GRAPH OF TRMS......................41

FIGURE 4-37 STRUCTURE OF PID CONTROLLER............................48

FIGURE 4-38 SIMULINK MODEL OF PID CONTROLLER...................48

FIGURE 4-39 SCHEMATIC DIAGRAM OF TUNING OF PID


List of Figures xii

PARAMETERS FOR TRMS...................................................................50

FIGURE 4-40 SEARCHING PATH OF STEEPEST DESCENT............51

FIGURE 4-41 CONTROL SYSTEM DEVELOPMENT FLOW DIAGRAM


.................................................................................................................53

FIGURE 4-42 SIMULINK MODEL IN HORIZONTAL AXIS...................54

FIGURE 4-43 SYSTEM SIMULATION RESPONSE (HORIZONTAL)...54

FIGURE 4-44 SIMULINK MODEL IN VERTICAL AXIS.........................55

FIGURE 4-45 SYSTEM SIMULATION RESPONSE (HORIZONTAL)...55

FIGURE 4-46 SIMULINK MODEL OF TRMS WITH PID CONTROLLER


.................................................................................................................56

FIGURE 4-47 THE SYSTEM SIMULATION WITH PID CONTROL


SCHEME ................................................................................................56

FIGURE 5-48 THE CHARACTERISTICS OF DEADBEAT RESPONSE


.................................................................................................................59

FIGURE 5-49 THE PERFORMANCE OF DEADBEAT CONTROLLER


.................................................................................................................60

FIGURE 5-50 CONTROL SYSTEM WITH FEEDBACK........................62

FIGURE 5-51 RESPONSE OF 4TH ORDER SYSTEM.........................62

FIGURE 5-52 THE STRUCTURE OF THE ROBUST SYSTEM DESIGN


(HORIZONTAL)......................................................................................63

FIGURE 5-53 THE STRUCTURE OF THE ROBUST SYSTEM DESIGN


List of Figures xiii

(VERTICAL)............................................................................................64

FIGURE 5-54 THE BASIC STRUCTURE OF THE ROBUST SYSTEM 64

FIGURE 5-55 MODIFY OF BASIC STRUCTURE OF ROBUST


SYSTEM..................................................................................................65

FIGURE 5-56 THE CONTROL STRUCTURE OF THE 2-D SYSTEM...69

FIGURE 5-57 THE SIMULINK MODEL OF VERTICAL AXIS WITH


DEADBEAT ROBUST............................................................................70

FIGURE 5-58 THE RESPONSE OF MAIN ROTOR (K=10)..................71

FIGURE 5-59 SIMULINK MODEL OF HORIZONTAL AXIS WITH


DEADBEAT ROBUST SCHEME............................................................71

FIGURE 5-60 THE RESPONSE OF TAIL ROTOR (K=7)......................72

FIGURE 5-61 SIMULINK MODEL OF TRMS WITH DEADBEAT


ROBUST ALGORITHM..........................................................................72

FIGURE 5-62 THE RESPONSE OF TAIL AND MAIN ROTOR.............73

FIGURE 6-63 BLOCK DIAGRAM OF IDENTIFICATION PROCEDUE77

FIGURE 6-64 BLOCK DIAGRAM OF TRMS SYSTEM WITH


DECOUPLERS.......................................................................................77
List of Tables

TABLE 1 EFFECT OF INDEPENDENT P, I AND D TUNING...............48

TABLE 2 DEADBEAT COEFFICIENTS AND RESPONSE TIMES. ALL


TIMES NORMALIZED BY .....................................................................63
CHAPTER 1

INTRODUCTION

1. INTRODUCTION

Recent advances in aircraft technology have led to the development of many new
concepts in aircraft design which are strikingly different from their predecessors. The
differences are in both aircraft configuration and control paradigms. This trend can be
attributed to the increasing emphasis on the aircraft to be agile and multi-purpose.
These new generation air vehicles have presented a challenges and opportunities to the
aerodynamics and control engineers.

In order to reduce money and time spend, computer simulation has become variable
asset to control engineering. Simulations are often quite cheap and simple to use
compared to testing designs on real hardware, especially when that hardware is a
helicopter or an aircraft. While designing new controls it is much easier to test the
designs on a simulation first. If there is any problem they can be cheaply and quickly
corrected without damaging any equipment. Also, it allows the control engineers a
chance to try new methods of controllers safely. In addition, there are also growing
literatures on laboratory platforms simulating aircraft manoeuvres and also a number of
publications that deal with the problem of PID controller in TRMS system[1-6],
2.3 Mathematical Model and State Equation 2

proportional-integral-derivative (PID) control offers the simplest and most efficient


solution to many real-world control problems. The TRMS is an aero-dynamical system
similar to a helicopter[7]. The main difficulties in designing controllers for them follow
from non-linearity and coupling. Due to the flight mechanics equations are not always
easy to establish. Some of modeling details of the vehicles are reported in [4, 5, 8-10]
[4, 5, 8-11].A simpler approach, decoupling technique, used to design and analyze the
controller

Over the past 50 years, several methods for determining PID controller parameters
have been developed for stable processed that are suitable for auto-tuning and optimal
control.[11, 12][12, 13] However, these tuning methods use only a small amount of
information about the dynamic behavior of the system and often do not provide good
tuning. Some employ information about robustness of PID controller has been
discussed. PID controllers that are widely applicable and can be set up easily.
Optimization methods are the one of tuning techniques, the steepest descent method;
steepest gradient decent algorithm optimization is used in this work to tune the
parameters of feedback compensators. The performance of the proposed control
scheme is assessed in terms of input tracking. it will usually converge even for poor
starting approximations.

Elementary or introductory control course in control engineer is almost entirely based


on linear systems; this is what we all start with the reason for this is twofold. First,
there are relatively simple closed analytical solutions to many control problems, so the
linear theory is nice, transparent and feasible. On the other hand, practical applications
are also based on linear or linearized models in most cases and handle nonlinearities
only when it is absolutely unavoidable. The TRMS which contain two DC motor, it can
be considered as a nonlinear system. To control and modelling nonlinear system might
be a difficult task, here, we are doing a simple linear approximation to obtain the model
of DC-motor.

In our study of automatic control, we usually have considered only control with a
single control objective or controlled variable. However, we encounter platform in
which more than one variable must be controlled, That is, multiple control objectives.
In such platform, we can still consider each control objective separately from the others
as long as they do not interact with each other. In the later chapter, we will study and
2.3 Mathematical Model and State Equation 3

design control systems for platform in which the various control objectives interact
with each other. We refer to these systems as multivariable control systems or as
multiple-input, multiple-output (MIMO) control systems. The problem we will be
addressing is that of loop interaction. In many complex industrial control problems, the
coupling among control loops often invalidates conventional single-loop controllers.
How to achieve decoupling control has become a topic of considerable importance in
the field of control engineering. Decoupling control was initially developed for
deterministic linear systems. Typical approaches include design of pre-compensator
that transforms the controlled transfer function matrix into a diagonal matrix or
diagonal dominance[13][14], and design of state feedback to reach decoupling of state
equation[14][15], decoupling in frequency domain through inverse Nyquist array[15]
[16], and decoupling method of Bristol-Shinskey[16][17]. These approaches separate
the controlled multivariable system into several SISO subsystems through a suitable
decoupler that depends on accurate model of system before controller design.

In order to control system here, design deadbeat robust system will be introduced[17]
[18]. This design method includes PID controller and deadbeat control design. It
provides the system ‘s response will remain almost unchanged when all the plant
parameters vary by as much as 50%, which mean, suppose our model of nonlinear
rotor is inaccurate but our design result still accomplish the real system. Also, we
consider the cross coupling as a disturbances it might affect the system model under
50% changes so the deadbeat robust technique can tolerate it.

This dissertation focused on the PID controller design based on the deadbeat robust
scheme specification for a given multi-input and multi-output plant. We attempt to
present the basic ideas, techniques, and results are presented in language and notation
familiar. Because the twin rotor system is highly nonlinear and cross-coupling, an
analytical tuning or modeling methods are not yet available. Approximation simplified
approach has been adopted to treat this problem. It is important to introduce the
steepest decent method; it is used to automatically tune the PID controller parameters.
In addition, the deadbeat robust scheme with PID controller is definition and presented
by the de-couple approach. Finally, a novel control scheme with PID is firstly proposed
to obtain the better performance. Simulation results are also presented to show the
effectiveness.
2.3 Mathematical Model and State Equation 4

2. PROJECT AIM

The general control scheme show in Figure 1-1 the aim is to design the controller that
enables us to command a desired helicopter pitch and yaw angles. Controlling the
system consists in stabilizing the TRMS beam in an arbitrary, within practical limits,
desired position (pitch and azimuth) or making it track a desired trajectory.

The compensators based on PID are designed and used as feedback controllers.
Steepest gradient decent algorithm optimization is also used in this work to tune the
parameters of feedback compensators. The performance of the proposed control
scheme is assessed in terms of input tracking. This is accomplished by comparing the
system response to open loop system performance without the feed-forward
components.

Steepest gradient decent algorithm is demonstrated in tuning the parameters of the


feedback controllers to improver the system response in the time domain. An objective
function is created to tune the PID controller within the augmented strategy that gives
the smallest overshoot, fastest rise time, quickest settling time and very small steady
state error.

Finally, the controller will be implemented on a PC-based nonlinear system, called twin
rotor multi-input multi-output system, to practically test the performance of the
proposed control scheme. The experimental results do illustrate its outstanding tracking
performance and good robustness against parameter variations and output disturbances.

Steepest Decent

r(t) c(t)
∑ PID Plant

FIGURE 1-1 PID CONTROL SCHEME


2.3 Mathematical Model and State Equation 5

Deadbeat Robust
Scheme

r(t) c(t)

Gain
PID Σ Plant

Gain

Feedback

FIGURE 1-2 DEADBEAT ROBUST CONTROL SCHEME

Below, is the results which we expect to obtain:

 Establishing the mathematical model and state equation of twin rotor


system. In this section, we classify into three categories. The forces
around the horizontal axis: considering the rotation of the beam in the
vertical plane (around the horizontal axis). Form Newton’s second law of
motion, we can obtain (1)The moments of gravity forces. (2)The moments
of propulsive forces. (3)The moment of centrifugal forces around the
vertical axis. (4)The moment of friction around the horizontal axis. The
forces around the vertical axis: similarly, we can describe the motion of
the beam around the vertical axis as: (1) The thrust of tail rotor. (2) The
moment of friction around the vertical axis. State equation: using the
equations as shown above, we can write the state equation describing the
motion of the system.
 Obtaining the values of model parameters by making some measurements:
the angular velocities are non-linear functions of the input voltage of the
DC-motor. Thus we need to identify the non-linear functions. The non-
linear input characteristics determining dependence of DC-motor
rotational speed on input voltage and the non-linear characteristics
determining dependence of propeller thrust on DC-motor rotational
speeds.
 Building the Simulink model accord to our mathematical model with PID
controller: Software package Matlab/Simulink are using in this project.
The simulation models of the dynamics of the TRMS system and
controller will be created by Simulink. It will be divided into four groups.
2.3 Mathematical Model and State Equation 6

(1) 1 degree of freedom in horizontal motion (2) 1 degree of freedom in


vertical motion (3)(4) 2 degree of freedom with/without cross-coupling
effect.
 Obtaining the simulation result by using steepest decent tuning algorithm:
To find the optimised PID controller I am going to use Simulink and
attach Matlab code to find the optimum controller.
 Obtaining the simulation result by using our deadbeat robust control
scheme.
 Comparing the software simulation result between the performance of PID
control and deadbeat robust control to demonstrate the effectiveness of our
control scheme. By comparing the result, it will verify our model
accuracy.

3. THESIS STRUCTURE

This section briefs the organization in the thesis as follows. In chapter 2, we show the
system modeling and gives the description about the twin rotor multi-input and multi-
output system which implemented within our control environment. Also, It gives detail
about the assumption and equation for modeling use. In order to obtain the values of
the model parameter some measurement also be investigated. Chapter 3, we present the
problem definition. Chapter 4 introduces the designing the PID controller by changing
the parameter, the method for finding the parameters, here, we introduce the steepest
decent algorithm. This method is simple and straightforward. Chapter 5, for obtaining
better-response performance, a novel deadbeat robust control scheme is proposed.
Conclusion and future developments are contained in Chapter 6
CHAPTER 2

SYSTEM MODELING

1. INTRODUCTION

Like most flight vehicles, the helicopter body is connected to several elastic bodies
such as rotor, engine and control surfaces. The physical nature of this system is very
complex, a simple mathematical modeling seems not to be very precise. Nonlinear
aerodynamic forces and gravity act on the vehicle, and flexible structures increase
complexity and make a realistic analysis difficult. Several assumptions can be made to
reduce this complexity to formulate and solve relevant problems. For application in
helicopter controls, where the main objective is to control the dynamic behavior of the
helicopter, it is necessary to find a representative model that shows the same dynamic
characteristics as the real aircraft. The Two Rotor Multi-Input and Multi-Output
System (TRMS) [7]is a laboratory set-up designed for control experiments. The
schematic diagram of the laboratory setup is shown in Figure 2-1, in certain aspects it
behaves like a helicopter. This chapter describes assumptions necessary for a
satisfactory modeling of the helicopter motion and introduces the fundamental motion
of the flight vehicle in general. Some features for the helicopter case are emphasized
and explained with respect to experiment measurement as needed. Also, the
2.3 Mathematical Model and State Equation 8

modification of this simulation model will be obtained and used to Simulink which is
software to give the user a graphical based system for the further implementation and
development.

PC+PCI 1711 Power Interface

FIGURE 2-3 THE LABORATORY SET-UP TRMS SYSTEM

2. TRMS SYSTEM DESCRIPTION

The TRMS is an aero-dynamical system similar to a helicopter as shown in Figure 2-2.


It consists of a beam pivoted on its base in such a way that it can rotate freely both in
its horizontal and vertical planes. At both end of a beam, there are two propellers
driven by DC motors. The TRMS system has main and tail rotors for generating
vertical and horizontal propeller thrust. The main rotor produces a lifting force
allowing the beam to rise vertically making a rotation around the pitch axis. While, the
tail rotor is used to make the beam turn left or right around the yaw axis.

In a normal helicopter the aerodynamic force is controlled by changing the angle of


attack. However the laboratory set-up is constructed such that the angle of attack of its
blades is fixed. The aerodynamic force is controlled by varying the speed of the motors.
Therefore, the control inputs are supply voltages of the DC motors. A change in the
voltage value results in a change of the rotational speed of the propeller, which results
in a change of the corresponding position of the beam.

The state of the beam is described by four process variables: horizontal and vertical
angles measured by optical encoders fitted at the pivot, and two corresponding angular
velocities. Two additional state variables are the angular velocities of the rotors,
measured by tacho-generators coupled to the driving DC motors.
2.3 Mathematical Model and State Equation 9

The pivot point allows the helicopter to move simultaneously in both the horizontal and
vertical planes. It is said to have two degrees of freedom (DOF). Either the horizontal
or the vertical degree of freedom can be restricted to 1 degree of freedom using the
screws.

Tail Rotor Tail Shield


Main shield
Main Rotor

free free beam DC-motor +


tachometer
DC-motor + Pivot
tachometer

Counterbalance

FIGURE 2-4 SCHEMATIC DIAGRAM OF TRMS

3. MATHEMATICAL MODEL AND STATE EQUATION

Modern methods of design and adaptation of real time controller require high quality
mathematical models for the system. In addition, there are some studies available to
model TRMS system[4, 6, 9, 10][4, 6, 9-11]. From the control point of view, TRMS is
a high-order nonlinear system with significant cross coupling. Mathematical models
and some assumptions used to support the physical law. To obtain dynamic equations,
the mathematical model of the TRMS helicopter system is developed under some
simplifying assumption.

 It is assumed that the dynamics of the propeller subsystem can be


described b first order differential equations.
 It is assumed that the friction in the system is of the viscous type.
 It is assumed also that the propeller- air subsystem could be described in
accordance with the postulates of the flow theory.
The mechanical system of TRMS is simplified by a four point-mass system, includes
main rotor, tail rotor, balance-weight and counter-weight. Based on Lagrange’s
equations, in modeling twin rotor system, we are going to classify into three categories,
2.3 Mathematical Model and State Equation 10

the forces around the horizontal axis, the forces around the vertical axis and state
equation and the above assumption will be used into each section.

3.1 Forces around the Horizontal Axis

Consider the rotation of the beam in the vertical plane (around the horizontal axis). The
driving torques are produced by the propellers, the rotation can be described in
principle as the motion of a pendulum. From Newton’s second law of motion we
obtain:

d 2α v
M v = Jv Equation 2-1
dt 2

Where:

M v is the total moment of forces in the vertical plan

4
M v = ∑ M vi
i =1

J v is the sum of moments of inertial relative to the horizontal axis.

4
M v = ∑ M vi
i =1

α v is the pitch angle of the beam.

The forces around the horizontal axis can be organized into four parts:

 The moments of gravity forces.


 The moments of propulsive forces.
 The moment of centrifugal forces around the vertical axis.
 The moment of friction around the horizontal axis.
Consider the situation shown in Figure 2-3.
2.3 Mathematical Model and State Equation 11

FIGURE 2-5 GRAVITY FORCES IN TRMS CORRESPONDING TO THE

RETURN TORQUE

Where each of parameters shown below:

M v1 is the return torque corresponding to the force of gravity

m mr is the mass of the main SC-motor with main rotor

mm is the mass of main part of the beam

mt is the mass of the tail motor with tail rotor

mcb is the mass of the counter-weight

mb is the mass of the counter-weight beam

m ms is the mass of the main shield

mts is the mass of the tail shield


2.3 Mathematical Model and State Equation 12

Im is the length of main part of the beam

It is the length of the tail part of the beam

Ib is the length of the counter-weight beam

I cb is the distance between the counter-weight and the joint

g is the gravitational acceleration.

To determine the moments of gravity forces applied to the beam and making it rotate
around the horizontal axis. The total moment of forces can be describe as equation.

  m  m   m  
M v1 = g   t + mtr + mts l t −  m + mmr + m ms l m  cosα v −  b l b + mcbl cb  sinα v 
  2   2    2  
Equation 2-2

This can be expressed as

M v1 = g { [ A − B ] cosα v − C sin α v } Equation 2-3

Where:

m 
A =  t + mtr + mts l t
 2 

m 
B =  m + mmr + mms l m
 2 

m 
C =  b lb + mcb l cb 
 2 

Consider the situation given in Figure 2-4.


2.3 Mathematical Model and State Equation 13

FIGURE 2-6 PROPULSIVE FORCE MOMENT AND FRICTION MOMENT

To determine the moments of propulsive forces applied to the beam

M v 2 = I m Fv (ω m ) Equation 2-4

Where:

M v 2 is the moment of the propulsive force produced by the main rotor and ω m is
angular velocity of the main rotor. Fv (ωm ) denotes the dependence of the propulsive
force on the angular velocity of the rotor. It should be measured experimentally.

To determine the moment of centrifugal forces around the vertical axis:

 m  m  m 
M v 3 = −Ω 2h  t + mtr + mts lt +  m + mmr + mms l m +  b lb + mcb l cb   sin α v cosα v Equatio
 2   2   2 

n 2-5

Where:

M v 3 = −Ωh2 ( A + B + C ) sin αv cos αv

dα h
Ωh =
dt

m 
C =  b lb 2 + mcb lcb 2 
 2 

M v 3 is the moment of centrifugal forces corresponding to the motion of the beam


2.3 Mathematical Model and State Equation 14

around the vertical axis.

Ω h is the angular velocity of the beam around the vertical axis and α h is the azimuth
angle of the beam.

To determine the moment of friction around the horizontal axis:

M v 4 = −Ωv k v Equation 2-6

Where:

M v 4 is the moment of friction depending on the angular velocity of the beam around
the horizontal axis.

Ωv is the angular velocity around the horizontal axis.

k v is a constant.

According to figure we can determine components of the moment of inertia relative to


the horizontal axis.

I m2
J v1 = mmr I m2 ; J v 2 = mm
3

I b2
J v 3 = mcb I cb2 ; J v 4 = mb
3

I t2
J v 5 = mtr I t2 ; J v 6 = mt
3

mms 2
J v7 = rms + mms I m2 ; rms is the radius of the main shield.
2

J v8 = mts rts2 + mts I t2 ; rts is the radius of the tail shield.


2.3 Mathematical Model and State Equation 15

3.2 Forces around the Vertical Axis

Similarly, we can describe the motion of the beam around the vertical axis. The driving
torques are produced by the rotors and that the moment of inertia depends on the pitch
angle of the beam. From Newton’s second law of motion we obtain:

d 2α h
Mh = Jh Equation 2-7
dt 2

Where:

2 8
M h = ∑ M hi ; J h = ∑ J hi
i =1 i =1

M h is the sum of moment of force acting in the horizontal plane

J h is the sum of moments of inertia relative to the vertical axis.

The forces around the vertical axis can be organized into two parts:

 The thrust of tail rotor.


 The moment of friction around the vertical axis.
Consider the situation shown in Figure 2-5
2.3 Mathematical Model and State Equation 16

FIGURE 2-7 MOMENTS OF FORCES IN HORIZONTAL PLANE

To determine the moment of forces applied to the beam and making it rotate around the
vertical axis. It can be expressed as:

M h1 = I t Fh (ω t ) cos α v Equation 2-8

Where:

M h1 is the thrust of tail rotor

ω t is the rotational velocity of tail rotor

Fh ( ω t ) denotes the dependence of the propulsive force on the angular velocity of the
tail rotor, which should be determined experimentally.

To determine the moment of friction, it can be expressed as:

M h 2 = −Ωh k h Equation 2-9

Where:

M h 2 is the moment of friction depending on the angular velocity of the beam around
the vertical axis.

Ω h angular velocity around the vertical axis.

k h is a constant.

According to Figure we can determine components of the moment of inertia relative to


the vertical axis:

mm mt
J h1 = ( I m cos α v ) 2 ; J h2 = ( I t cos α v ) 2
3 3

mb
( I b sin α v ) 2 ; J h 4 = mtr ( I t cos α v )
2
J h3 =
3
2.3 Mathematical Model and State Equation 17

J h 5 = mmr ( I m cos α v ) ; J h 6 = mcb ( I cb sin α v )


2 2

mts 2
rts + mts ( I t cos α v ) ; rts is the radius of the tail shield.
2
J h7 =
2

J h8 = mms rms2 + mms ( I m cos α v ) ; rms is the radius of the main shield.
2

As the description above, the moment of inertia can rewrite as below:

J h = D cos 2 α v + E sin 2 α v + F Equation 2-10

Where:

mb 2
D= I b + mcb I cb2 ,
3

m  m 
E =  m + m mr + m ms  I m2 +  t + mtr + mts  I t2 ,
 3   3 

mts 2
F = mms rms2 + rts
2

3.3 State Equation

From Equation 2-1 to Equation 2-10, we can write the equations describing the motion
of the system as follows:

 The Main Rotor Model:


dSv 1
= I m S f Fv ( ωm ) − Ω v k v + g ( ( A− B ) cosα v − C sinα v )− Ω 2
h ( A+ B+ C ) sin 2α v Equation 2-11
dt 2

dα v J tr ω t
= Ω v ; Ωv = S v + Equation 2-12
dt Jv
2.3 Mathematical Model and State Equation 18

dSv
Mv =
dt

 The Tail Rotor Model:


dS h
= I t S f Fh ( ωt ) cos αv − Ωh kh Equation 2-13
dt

dα h S + J mr ωm cos αv Sh + Jmr ωm cos αv


= Ω h , Ωh = h = Equation 2-14
dt Jh D sin 2 αv + E cos2 αv + F

dS h
Mh =
dt

Where:

J mr is the moment of inertia in DC-motor-main propeller subsystem.

J tr is the moment of inertia in DC-motor-tail propeller subsystem.

S h is the angular momentum in the horizontal plane of the beam.

S v is the angular momentum in the vertical plane of the beam.

S f is the balance scale.

Furthermore, the angular velocities ( ωm , ωt ) are non-linear functions of the input


voltage of the DC-motor ( uv , ut ), the model of the motor-propeller dynamics is
obtained by substituting the non-linear system by a serial connection of a linear
dynamics system and static non-linearity. The system block diagram shown in Figure
2-6, its system equation can express as

duvv 1
= ( −uvv + uv ) ; ωm = Pv ( uvv ) Equation 2-15
dt Tmr

duhh 1
= ( −uhh + uh ) ; ωt = Ph ( uhh ) Equation 2-16
dt Ttr
2.3 Mathematical Model and State Equation 19

Where:

Tmr is the time constant of the main rotor-propeller system.

Ttr is the time constant of the tail motor-propeller system.

FIGURE 2-8 BLOCK DIAGRAM OF EQUATION 2-15 AND 2-16

4. CHARACTERISTICS OF MAIN AND TAIL MOTOR

In order to obtain the values of the model parameter it is necessary to make some
measurements. The relation between rotor speed and force is too complex to calculate
but may be measure using an electronic balance connected to the rotor[7]. First, the
geometrical dimensions and moving masses of TRMS should be measured. The
notations are explained in Figure 2-3, Figure 2-4 and Figure 2-5.

I t = 0.2 5 [ m] I m = 0.24 [ m] I b = 0.26 [ m]

I cb = 0.13 [ m] rms = 0.155 [ m] rts = 0.10 [ m]

mtr = 0.206 [ kg ] mmr = 0.228 [ kg ] mcb = 0.068 [ kg ]

mt = 0.0155 [ kg ] mm = 0.0145 [ kg ] mb = 0.022 [ kg]

mts = 0.165 [ kg] m ms = 0.225 [ kg ]


2.3 Mathematical Model and State Equation 20

4.1 The moment of inertia about the horizontal axis

Using the above measurements the moment of inertia about the horizontal axis can be
calculated as:

[ ]
8
J v = ∑ J iv kg m 2
i

The terms of the sum are calculated from elementary physics laws:

J v1 = mtr I t2 = 0.20 6× 0.2 52 = 0.12 8 7 5 [ kg m ] 2

J v 2 = mcb I cb2 = 0.068× 0.132 = 0.001149 [ kg m ] 2

J v 3 = mmr I m2 = 0.228× 0.242 = 0.013132 [ kg m ] 2

I t2 2
J v 4 = mt = 0.0155 × 0.25 = 0.000322  kg m2 
3 3

J v 5 = mm
I m2
3
= 0.0 1 4 5× 0.2 4
2

3
= 0.0 0 0 2 7 8 [k g m ] 2

J v 6 = mb
I b2
3 3
2
= 0.0 2 2× 0.2 6 = 0.0 0 0 4 9 5 [k g m ] 2

r2
 2


2
(
J v 7 = mms  ms + I m2  = 0.225 0.155 + 0.242 = 0.015622
2
) [ kg m ]2

J v8 = mts ( rts2 + It2 ) = 0.165 ( 0.102 + 0.252 ) = 0.011962  kg m2 

Giving finally:
2.3 Mathematical Model and State Equation 21

8
J v = ∑ J iv = 0.055846  kg m2 
i

4.2 Moment of inertia about vertical axis

The calculated moment of inertia about the vertical axis is:

8
J h = ∑ J hi
i

Where the terms of the sum are:

J h 2 = mt ( It cos αv )
2
3 = 0.0003229 cos 2 αv  kg m2 

J h1 = mm ( I m cos αv )
2
3 = 0.0002784 cos 2 αv  kg m2 

J h3 = mb ( Ib sin αv )
2
3 = 0.0004595sin 2 αv  kg m2 

J h5 = mmr ( I m cos αv ) = 0.013132 cos 2 αv


2
 kg m2 

J h 4 = mtr ( It cos αv ) = 0.012875cos 2 αv


2
 kg m2 

J h 6 = mcb ( Icb sin αv ) = 0.0011492sin 2 αv


2
 kg m2 

J h 7 = mts = ( rts2 2 + It2 cos2 αv ) = 0.000825 + 0.010312 cos2 αv  kg m2 

J h8 = mms = ( rms2 + It2 cos2 αv ) = 0.00540 + 0.0129611cos2 αv  kg m2 


2.3 Mathematical Model and State Equation 22

Hence;

8
J h = ∑ J hi =cos
D 2α v E+ 2 α
sin v + = cos2
F 0.049881 v α + sin2
0.0016449 v α
0.0062306 +
i

4.3 Returning torque

The returning torque from gravity forces is given by the equation

M v1 = g { [ A − B ] cosα v − C sin α v }

Where

m  m  m 
A =  t + mtr + mts l t ; B =  m + mmr + mms l m ; C =  b lb + mcb l cb 
 2   2   2 

Hence;

 0.0155 
A = + 0.206 + 0.165 0.25 ; A = 0.0946875
 2 

 0.0145 
B = + 0.228 + 0.225 0.24 ; B = 0.11046
 2 

 0.022 
C = 0.26 + 0.068 × 0.13  ; C = 0.0117
 2 

  m  mm   mb   
M v1 = g   t +mtr + ts lt
m − m mr+ m ms+ αl m 
cos  v sincbl cb α+
l b− m v  
  2  2   2   

Substituting A, B and C in equation

M v1 = g { [ A − B ] cosα v − C sin α v }
2.3 Mathematical Model and State Equation 23

M v1 = g { [ 0.0948675 − 0.11046] cos α v − 0.0117 sin α v }

Giving:

M v1 = 9.81(−0.0155925 cos α v + 0.0117016 sin α v ) [ N m]

4.4 Moment of centrifugal force

The moment of the centrifugal forces is:

6
M v 3 = ∑ M v 3,i
i

Where:

M v 3,1 = ( mtr + m)ts 2It 2 cos


Ωh α sin
v α v =
0.0231875 2
h Ω α
cos sin
v αv [ ] Nm

M v 3,2 = mt It 2Ωh 2 cos α v / 2 = 0.0002421Ω 2h cos α v sin α v [ Nm]

M v 3,3 = mb I b 2Ωh 2 cos αv / 2 = 0.0003718Ωh2 cos αv sin αv 2 [ Nm ]

M v 3,4 = mcb Icb2 Ωh2 cos αv sin αv = 0.0011492Ωh2 cos αv sin αv [ Nm]

M v 3,5 = mm I2m Ω2 hcosα vsin α v 2 =


0.0002018 2

cos
h αvsin α2
v [ ] Nm

M v 3,6 = ( mmr + mms) 2Im 2 Ωh α sin


cos v α v =
0.02523028 2
h Ω α
cos sin
v αv [ ] Nm

Giving finally:
2.3 Mathematical Model and State Equation 24

6
M v 3 = ∑ M v 3,icf = 0.05038268Ωh 2 cos α v sin α v [ Nm ]
i

4.5 Static characteristics

The static characteristics of the propellers are measured using a proper electronic
balance with voltage output[7]. Thus we can identify the following non-linear
functions: Two non linear input characteristics determining dependence of DC-motor
rotational speed on input voltage:

ω m = Pv ( u vv ) ω t = Ph ( u hh )

Two non-linear characteristics determining dependence of propeller thrust on DC-


motor rotational speeds:

Fh = Fh (ωt ) Fv = Fv (ω m )

5. SYSTEM SIMULATION

Based on block diagram representation of the system is very suitable for use in the
Simulink environment. A block diagram of the TRMS shown in Figure 2-7.
2.3 Mathematical Model and State Equation 25

FIGURE 2-9 BLOCK DIAGRAM OF TRMS MODEL

It can consider as a high order, non-linear, cross-coupled systems. However a simpler


approach, decoupling technique, used to create two 1-DOF separate models for
horizontal and vertical. This section presents Simulink models of TRMS. The models
are based on non-linear equation given in previously section.

5.1 Simulink model of horizontal part of TRMS

In order to simulate system, by decouple technique the dynamic equation of TRMS can
be described as follows:

The Tail Rotor Model:

dS h
= I t S f Fh ( ωt ) cos αv − Ωh kh
dt

dα h S + J mr ωm cos αv Sh + Jmr ωm cos αv


= Ω h , Ωh = h =
dt Jh D sin 2 αv + E cos2 αv + F

duhh 1
= ( −uhh + uh ) ; ωt = Ph ( uhh )
dt Ttr

Suppose that main rotor is independent the equation above can rewrite as below:
2.3 Mathematical Model and State Equation 26

dS h
= I t S f Fh ( ωt ) cos αv − Ωh kh
dt

dα h
= Ω h , Ωh = 90Sh
dt

duhh 1
= ( −uhh + uh ) ; ωt = Ph ( uhh )
dt Ttr

The block diagram of the tail rotor model can be represented as below:

FIGURE 2-10 BLOCK DIAGRAM OF THE TAIL ROTOR

The Simulink 1-DOF model of the horizontal part of TRMS is shown in Figure 2-9
which shows the grouped model with scopes for the visualization of input, position and
velocity. It can be used to observe the behavior of the open loop system. Figure 2-10
shows the contents of the grouped model block it includes detail for the speed of tail
rotor, the driving torque of tail rotor, rotational speed of tail rotor and aero force
characteristic. Those are developed by block diagram and can be modified them if
necessary.
2.3 Mathematical Model and State Equation 27

FIGURE 2-11 1-DOF MODEL OF THE HORIZONTAL PART OF TRMS

FIGURE 2-12 THE CONTENTS OF THE GROUPED MODEL BLOCK

(HORIZONTAL)

FIGURE 2-13 BLOCK DIAGRAM OF ROTATIONAL SPEED OF TAIL ROTOR

FIGURE 2-14 BLOCK DIAGRAM OF DRIVING TORQUE OF TAIL ROTOR


2.3 Mathematical Model and State Equation 28

FIGURE 2-15 BLOCK DIAGRAM OF AERO FORCE (TAIL ROTOR)

FIGURE 2-16 ROTATIONAL SPEED OF TAIL ROTOR

5.2 Simulink model of vertical part of TRMS

The Main Rotor Model:

dSv 1
= I m S f Fv ( ωm ) − Ω v k v + g ( ( A− B ) cosα v − C sinα v )− Ω 2
h ( A+ B+ C ) sin 2α v
dt 2

dα v J tr ω t
= Ω v ; Ωv = S v +
dt Jv

duvv 1
= ( −uvv + uv ) ; ωm = Pv ( uvv )
dt Tmr

Suppose that main rotor is independent the equation above can rewrite as below:

dSv
= I m S f Fv ( ωm ) − Ω v kv + g ( ( A− B ) cosα v − C sinα v )
dt
2.3 Mathematical Model and State Equation 29

dα v
= Ω v ; Ω v = 9.1Sv
dt

duvv 1
= ( −uvv + uv ) ; ωm = Pv ( uvv )
dt Tmr

The block diagram of the main rotor model can be represented as below:

FIGURE 2-17 BLOCK DIAGRAM OF THE MAIN ROTOR

The Simulink 1-DOF model of the vertical part of TRMS is shown in Figure 2-16
which shows the grouped model with scopes for the visualization of input, position and
velocity. It can be used to observe the behavior of the open loop system. Figure 2-17
shows the contents of the grouped model block it includes detail for the speed of main
rotor, the driving torque of main rotor, rotational speed of main rotor and aero force
characteristic. Those are developed by block diagram and can be modified them if
necessary.

FIGURE 2-18 1-DOF MODEL OF THE VERTICAL PART OF TRMS


2.3 Mathematical Model and State Equation 30

FIGURE 2-19 THE CONTENTS OF THE GROUPED MODEL BLOCK (VERTICAL)

FIGURE 2-20 BLOCK DIAGRAM OF ROTATIONAL SPEED OF MAIN ROTOR

FIGURE 2-21 BLOCK DIAGRAM OF DRIVING TORQUE OF MAIN ROTOR

FIGURE 2-22 BLOCK DIAGRAM OF AERO FORCE (MAIN ROTOR)


2.3 Mathematical Model and State Equation 31

FIGURE 2-23 ROTATIONAL SPEED OF MAIN ROTOR

5.3 Simulink model of TRMS in 2-DOF

The Simulink 2-DOF model of TRMS is shown in Figure 2-22. This model can be used
for observation of all the state variables in the open loop mode. Also, it can be used for
developing closed-loop control systems as described in the following chapter.

FIGURE 2-24 2-DOF COMPLEX MODEL OF TRMS


2.3 Mathematical Model and State Equation 32

FIGURE 2-25 DETAILED 2-DOF MODEL OF TRMS


CHAPTER 3

PROBLEM DEFINITION AND


APPROACH

1. INTRODUCTION

Modeling and control of the twin rotor multi-input and multi-output system (TRMS)
have been studied for many years. The behaviour of the TRMS can be resembled as a
helicopter. Also TRMS can be an excellent platform which be used to prove any new
theorems in simulation environment or real-time experiment situation. The block
diagram of Twin Rotor MIMO System (TRMS) can be shown in Figure 3-1, for a
control system the achievable performance is typically limited by four main features:

 Process dynamics, TRMS is a air vehicle with complex dynamics.


 Nonlinearities, there are two non-linear inputs which are DC-motors.
 Uncertainties, Modelling between mathematical model and real equipment
there might have much uncertain situations which have been ignored.
 Disturbances, Angular momentum and reaction turning moment are the
two main effects from cross-coupling which be considered as a
2.3 Mathematical Model and State Equation 34

disturbance
The main problem with this TRMS system is that the tail and main rotor interact badly.
Initially TRMS system contain two PID control both compensate tail and main rotor
individually. PID is the control algorithm which have been successfully used for many
years, the simple structure and the well know Ziegler and Nichols tuning algorithms
has been used since 1942[18]. The major drawback of PID controller is strong affected
by tuning tools. Some works are developed by the appropriate tuning tools for TRMS.
Wang [19] investigated the effect of the simplified genetic algorithm (GA) on
controller tuning for improving system performance. Ahmad [20] employed his model
in the design of a feedback linear quadratic Gaussian compensator (LQR) this has a
good tracking capability. Islam, Liu and Juang [1, 21, 22] these articles are developed
by fuzzy compensator and presented a improvement of the tracking performance.

As described as above, modelling non-linear rotor is a difficult task. It is hard to find


the exact model of the dynamic system. For modelling system Ahmad is the first
researcher who used TRMS as the platform [4, 8, 9, 20, 23-27] by doing system
identification technique. Radial basis function networks are shown to be suitable for
modelling complex engineering systems in cases where the dynamics are not well
understood or are not simple to establish from first principles.[9] Black-Box also a
good start to parametric model to the actual plant dynamics.[25]

Even if we get the system model, however it might not exactly represent the real-
system for the entire input range. If we apply PID controllers for the system for both
main and tail rotor, we would have six parameters to tuning [28][24]. The final result
would be influenced heavily by the tuning algorithm and the performance is hard to
predict [11, 29][11, 25].

The TRMS can consider as a non-linear, cross-coupled system which is very


complicated. The problem with this system is that the controller of Tail and Main rotor
interact badly. Assume we are using PID controller for the system for both main and
tail rotor, it will include six parameters, and the final result will be influenced heavily
by the tuning algorithm and the performance of computing. Also, modelling non-linear
rotor is difficult task; moreover suppose we have the transfer function of non-linear
rotor however it might not exactly represent the real-system.
2.3 Mathematical Model and State Equation 35

This chapter will present the design and tuning of multivariable feedback control
systems. We first explained the effect of interaction and nonlinear behaviour then
introduced an approach technique. First, decouple technique are used to minimize the
effect of interaction. Then, a simple nonlinear approximation, use Matlab to simplify
the problems which be occurred in DC-motor. Finally, a procedure for tuning nonlinear
and interacting system will be discussed.

Cross-Coupling

Nonlinear Rotor

FIGURE 3-26 BLOCK DIAGRAM OF THE TRMS SYSTEM

FIGURE 3-27 DETAIL OF MODEL INCLUDE CROSS-COUPLING


2.3 Mathematical Model and State Equation 36

FIGURE 3-28 THE DIFFERENCE BETWEEN DIFFERENTIAL EQUATION AND

TRANSFER FUNCTION IN MAIN ROTOR

FIGURE 3-29 THE DIFFERENT BETWEEN DIFFERENTIAL EQUATION AND

TRANSFER FUNCTION IN TAIL ROTOR


2.3 Mathematical Model and State Equation 37

2. NONLINEAR DC MOTORS

Many physical relationships are often represented by linear equations, in most cases
actual relationships are not quite linear. In fact, a careful study of physical systems
reveals that even so-called “linear systems: are really linear only in limited operating
ranges. In control engineering a normal operation of the system may be around a
equilibrium point[30][21]. However, if the system operates around an equilibrium point
then it is possible to approximate the nonlinear system by a linear system. Such a linear
system is equivalent to the nonlinear system considered within a limited operating
range. Linearized model, this is very important in control engineering. Later, we are
going to do a linear approximation of nonlinear mathematical models.

Modeling is an indispensable step to the synthesis of high performance control systems.


The model must represent the most relevant characteristics of the system for the
purposed application. The modeling of a DC motor is well accepted and discussed in
some research paper. [30-32][21-23].DC motors, as a components of electromechanical
systems, are widely used as actuating elements in industrial applications for their
advantages of easy speed and position control and wide adjustability range[33]. The
general approach is to neglect the nonlinear effects and build a linear transfer function
representation for the input-output relationship of the DC motor[34]. In this paper, it
should be noted that angular velocities are non-linear functions of the input voltage of
the DC-motor. The block diagram shows in Figure 3-5, thus we have two equations:

duvv 1
= (−uvv + uv ) ; ωm = Pv (uvv )
dt Tmr

duhh 1
= (−uhh + uh ) ; ωt = Ph (uhh )
dt Ttr

Where

Tmr is the time constant of main propeller system.

Ttr is the time constant of tail propeller system.


2.3 Mathematical Model and State Equation 38

uv 1 uvv ωm
Pv (uvv )
Tmr s + 1

uh 1 uhh ωt
Ph (u hh )
Ttr s + 1

FIGURE 3-30 BLOCK DIAGRAM OF MAIN AND TAIL PROPELLER SYSTEM

The above model of the motor-propeller dynamics can be obtained by substituting the
non-linear system by a serial connection of a linear dynamic system and static non-
linearity. For this purposed, one can use the Matlab polyfit.m function which can
provide a polynomial curve fitting and fits the data in a least squared sense. Figure 3-4
to Figure 3-7 show the approximation of each tail and main rotor and also the
polynomials can be given as below:

For the main rotor:

ωm = 90.99uvv 6 + 599.73uvv 5 − 129.26uvv 4 − 1238.64uvv 3 + 63.45uvv 2 + 1283.4uvv

× −12ω
Fv = −3.48 10 5
m + 109× − ω4 m
1.09 + 106
4.123 ×3−m ω 1.632 −10 4 2
m × −9.544
ω 10+2 m ×−

For the tail rotor:

ωt = 2020uhh 5 − 194.69uhh 4 − 4283.15uhh 3 + 262.27uhh 2 + 3796.83uhh

Fh = −3 ×10−14ω 5
t −
1.595 10×11 − 4ωt + 10 7 ×3 t − ω1.808 −10 4
2.511 2
t

ω
× 0.0801 + 2
10 t × −ω
2.3 Mathematical Model and State Equation 39

FIGURE 3-31 MEASURED CHARACTERISTICS OF THE MAIN ROTOR

FIGURE 3-32 POLYNOMIAL APPROXIMATION OF THE MAIN ROTOR

CHARACTERISTICS

FIGURE 3-33 MEASURED CHARACTERISTICS OF TAIL ROTOR

FIGURE 3-34 POLYNOMIAL APPROXIMATION OF THE TAIL ROTOR

CHARACTERISTICS

3. CROSS-COUPLING EFFECTS

The TRMS can consider as a high order, non-linear cross-coupled systems which is
2.3 Mathematical Model and State Equation 40

often very complicated. However a simpler approach, decoupling technique, used to


design the control scheme. Controlling a single-variable process is comparatively easy
even if the dynamics in the loop are unfavorable. There is only one way to close the
loop. When a second pair of variables appears, however, the picture is entirely
different, not only must a choice be made between pairs used for control, but coupling
can exist. And if there is coupling, the ease of control that was found with independent
loops disappears. This facility can be restored, however, by decoupling the variables
through a computing system.[16, 35][17, 23]

Interaction among control loop in a multivariable system has been the subject of much
research over the last 30 years. All of this work is based on the premise that interaction
is undesirable. This is true for setpoint disturbances. We would like to change a
setpoint in one loop without affecting the other loops. And if the loops do not interact,
each individual loop can be tuned by itself, and whole system should be stable if each
individual loop is stable.[36][24]

Unfortunately, much of this interaction analysis work has clouded the issue of how to
design an effective control scheme for a multivariable process. In most control
application the problem is not setpoint response but load response. We want a system
that holds the position at the desired values in the face of load disturbances. Interaction
is therefore not necessarily bad; in fact, in some systems it helps in rejecting the effects
of load disturbances.

This section going to discuss the Cross-coupling behaviors and also provide a decouple
example. Figure 3-3 presents the block diagram for an 2 × 2 interacting system which
be applied into our research. [37][25] This block diagram shows graphically that the
interaction between the two loop is caused by the “cross” blocks with transfer functions
Ghv and Gvh where:

Gh transfer function of tail rotor; Gv transfer function of main rotor

Gvv transfer function of individual vertical part

Gvh transfer function for vertical effect affect to horizontal part

Ghv transfer function for horizontal effect affect to vertical part


2.3 Mathematical Model and State Equation 41

Ghh transfer function of individual horizontal part

Uh +
- + αh
Gh Ghh
+

Ghv

Gvh
+
Uv + αv
Gv Gvv
+

FIGURE 3-35 THE INTERACTION FRAMES OF TRMS

-1

Uh αh αh
1 Gh Ghh 1

Ghv

Gvh
Uv αv αv
1 Gv Gvv 1

-1

FIGURE 3-36 THE FINGNAL FLOW GRAPH OF TRMS

To obtain the closed-loop transfer for the diagram, we first draw the corresponding
signal flow graph, Figure 3-4 the graph has three loops, two of which do not touch each
other

L11 = −GhGhh

L12 = −Gv Gvv

L13 = Gh Ghv Gv Gvh


2.3 Mathematical Model and State Equation 42

Loops L1 and L2 are the familiar feedback loops. L3 is more complex and goes
through both controllers and the “cross” transfer function. The determinant of the graph
is then

∆ = 1 − ∑ L1 + ∑ L2 + ....

∆ = 1 + GhGhh + Gv Gvv − Gh Ghv Gv Gvh + Gh Ghh Gv Gvv

Where the last term is the product of the two no touching loops. There are two paths
between U h and α h :

P1 = GhGhh

P2 = −Gh Ghv Gv Gvh

The first of these paths does not touch the bottom loop, and the other one touches all
three loops.

∆1 = 1 + Gv Gvv

∆ =1

The Mason’s Gain Formula provides a compact guide to the development of the
transfer functions of a complex graph where

∑ P∆i i
EQUATION 3-17
T= i

T = transfer function between input and output nodes

Pi = product of the transfer functions in the ith forward path between input and out
nodes

From Equation 3-1, the transfer function is


2.3 Mathematical Model and State Equation 43

α h Gh Ghh [1 + Gv Gvv ] − Gh Ghv Gv Gvh


= EQUATION 3-18
Uh ∆

There is only on path between U h and α v and it touches all three loops in the graph.
Thus the transfer function can be obtained as:

α v GhGhv
=
Uh ∆

By the same procedure, we can obtain the transfer functions between U v and the two
controlled variables. They are:

α h Gv Gvh
=
Uv ∆

α v Gv Gvv [1 + Gh Ghh ] − Gv Gvh Gh Ghv


=
Uv ∆

As with any dynamic system, the response is determined by the location of the roots of
the denominator polynomial or characteristic equation. To obtain the characteristic
equation, just set the determinant of the graph equal to zero. ∆ = 0

It is enlightening to rearrange the determinant, Equation 3-2, into the following form

∆ = [1 + GhGhh ][1 + Gv Gvv ] − Gh Gvh Gv Ghv = 0 EQUATION 3-19

The roots of this equation determine the stability and response of the interacting 2 × 2
system. Equation 3-3 also gives us following features:

 The tuning of each controller affects the response of both controlled


variables, because it affects the roots of the common characteristic
equation.
 The effect of interaction on one loop may be eliminated by interrupting
the other loop.
 For interaction to affect the response of the loops, it must act both ways.
That is, each manipulated variable must affect the controlled variable of
the other loop.
2.3 Mathematical Model and State Equation 44

By apply decoupling technique both vertical and horizontal model can simplify as
below:

The Main Rotor Model:

dSv
= I m S f Fv ( ωm ) − Ω v kv + g ( ( A− B ) cosα v − C sinα v )
dt

dα v
= Ω v ; Ω v = 9.1Sv
dt

duvv 1
= ( −uvv + uv ) ; ωm = Pv ( uvv )
dt Tmr

The Tail Rotor Model:

dS h
= I t S f Fh ( ωt ) cos αv − Ωh kh
dt

dα h
= Ω h , Ωh = 90Sh
dt

duhh 1
= ( −uhh + uh ) ; ωt = Ph ( uhh )
dt Ttr

For the further design controller for TRMS system, the transfer function of horizontal
and vertical part is necessary. Consider the block diagram of vertical and horizontal
model of TRMS which is shown in Figure 2-8 and Figure 2-15. The transfer function
can be known either by block reduction method or Matlab. Here the following result
was executed by Matlab. The extracted continuous transfer function of the parametric
model that represents the TRMS in vertical and horizontal movement is given as:

1.519
Gm ( s ) = Equation 3-20
s +0.748 s 2 +1.533s +1.046
3

15.02
Gt ( s )= Equation 3-21
s +3.458 s 2 + 2.225 s
3
2.3 Mathematical Model and State Equation 45

Where Gm ( s ) represents the transfer function of main rotor and Gt ( s ) represents the
transfer function of tail rotor. These transfer functions will be utilized throughout this
work.

4. SUMMARY

Figure 3-3 and 3-4 show the difference between differential equation and transfer
function which be obtained by doing some approximation, it has be discussed above.
PID controller is one of the solutions which robustness enough to control the platform
however it has dramatic influence on tuning algorithm, these will be discussed later.
The other solution is to design a robustness control system with model-base design
procedure. The disadvantage of model base design procedure that need high accurate
transfer function, to avoid the problem, in later chapter we are introducing one
procedure that can handle system by changing the exit control scheme to achieve even
the platform contain disturbance or be modeled inaccurate, the scheme maintain the
system in the desired value.
CHAPTER 4

PID CONTROLLER STUDY

1. INTRODUCTION

Even though control theory has been developed significantly, the proportional-integral-
derivative (PID) controllers are used for a wide range of process control, motor drives,
magnetic and optic memories, automotive, fight control, instrumentation, etc. More
than 90% of industrial controllers are still implemented based around PID algorithms
and ease of use offered by the PID controller[11, 29][12, 20]. Optimization methods are
the one of tuning techniques[12, 38, 39][13, 26, 27], the steepest descent method; it
will usually converge even for poor starting approximations. As a consequence, this
method is used to find sufficiently accurate starting approximations for other
techniques. The method is valuable quite apart from the application as a starting
method for solving nonlinear systems.

In this chapter the design of the PID controller to control the helicopter position is
discussed. The controller designed in this chapter uses the steepest decent algorithm
that will derive in the later section, a discussion of the implementation for a controller
which achieves desired position will be given in the section 4.3.Based on the non-linear
equation that is presented in chapter 2 the simulations implementation data for both
horizontal and vertical controller implementations will be proposed. The program
Matlab was used to perform most of the calculation of optimization. Simulation data
was obtained by using Simulink to simulate to controller. The final result can be a good
reference for future using.

2. REVIEW OF PID CONTROLLER

With its three-term functionality covering treatment to both transient and steady-state
responses, proportional-integral-derivative (PID) control offers the simplest and most
efficient solution to many real-world control problems. The PID controllers are usually
standard building blocks for industrial automation. The most basic PID controller has
the form:

d
u ( t ) = K p e( t ) + K i ∫ e(τ ) dτ + K d ( e( t ) ) EQUATION 4-22
t

0 dt

Where:

u (t ) is the control output and the error

e(t ) is defined as e(t ) = desired value – measured value of quantity being controlled.

K p , K i , and K d are the control gains.

Diagrammatically, the PID controller can be represented as Figure 4-1; also it can
convert into Simulink model as shown in Figure 4-2

K p e(t )

e(t ) u (t )
K i ∫ e (t )dt ∑

de (t )
Kd
dt
FIGURE 4-37 STRUCTURE OF PID CONTROLLER

du
Kd
dt

in Kp out
Umax

1
Ki
s
Isat

FIGURE 4-38 SIMULINK MODEL OF PID CONTROLLER

Determine the weight of the contribution of the error, the integral of the error, and the
derivative of the error to the control output. These gains will dictate the response of the
closed-loop system to initial conditions and inputs. Some features of PID controller
was collected in Table 4-1.The “three-term” functionalities are also can be highlighted
by the following:

response Rise time Overshoot Settling Steady


time state error

Increasing Small
Decrease Increase Decrease
Kp increase

Increasing Small Large


Increase Increase
Ki decrease decrease

Increasing Small Minor


Decrease Decrease
Kd decrease change

TABLE 1 EFFECT OF INDEPENDENT P, I AND D TUNING

 The proportional term providing an overall control action proportional to


the error signal through the all-pass gain factor
 The integral term reducing steady-state errors through low-frequency
compensation by an integrator.
 The derivative term improving transient response through high-frequency
compensation by a differentiator.
There are a number of tuning methods for PID controllers. The controller parameters
are tuned such that the closed-loop control system would be stable and would meet
given objectives associated with the following:

 Stability robustness
 Set-point following and tracking performance at transient, including rise-
time, overshoot, and settling time
 Regulation performance at steady-state, including load disturbance
rejection.
 Robustness against plant modeling uncertainty.
 Noise attenuation and robustness against environmental uncertainty.
With give objectives, tuning methods for PID controllers can be grouped according to
their nature and usage, as follow:

 Analytical methods-PID parameters are calculated from analytical or


algebraic relations between a plant model and an objective such as internal
model control (IMC).
 Heuristic methods-These are evolved from practical experience in manual
tuning (such as the Ziegler-Nichols tuning rule).
 Frequency response methods-Frequency characteristics of the controlled
process are used to tune the PID controller such as loop-shaping.
 Optimization methods-These can be regarded as a special type of optimal
control, where PID parameters are obtained using an offline numerical
optimization method for a single composite objective.
 Adaptive tuning methods-These are for automated online tuning, using
one or a combination of the previous methods based on real-time
identification.
Optimization based methods are often applied offline or on very slow processed using a
conventional (such as least mean squares) or and unconventional (genetic algorithms)
search method. Formula based tuning methods are still the most actively developed
however most does not yield global or multi-objective optimal performance, hence,
often limited. In this work, we are using steepest descent method which is the simplest
procedure which will be discussed in the later section.

3. OPTIMIZATION CONTROLLER

Optimization is one of the tuning mechanisms for tuning PID parameter. In this work
the Steepest Gradient Decent optimization process, depicted in Figure 4-3, is initialized
with a company default setting. After calculating the PID coefficients, the PID
parameters are applied to a Simulink model. Then we can study the behavior of the
modeled closed-loop system. On completion of the simulation, the response due to step
or square input is stored and error is assessed taking the difference between the desired
and actual response. Then, the error signal is processed based on performance criteria.

Steepest Decent

r(t) c(t)
∑ PID Plant

Procedure
CRITERION
Min (P,I,D)

FIGURE 4-39 SCHEMATIC DIAGRAM OF TUNING OF PID PARAMETERS FOR


TRMS

3.1 Steepest Gradient Method

Gradient methods use information about the slope of the function to dictate a direction
of search where the minimum is thought to lie. The simplest of these is the method of
steepest descent in which a search is performed in a direction, −∇f ( x) where ∇f ( x)
is the gradient of the objective function. In Figure We can see that the search is in the
direction opposite to the gradient, where the search started with an arbitrary initial
weight w(0) , then modify w(0) proportionally to the negative of the gradient, change
the operating point to w(1) , and applying the same procedure iteratively, we can get
the equation

w(k + 1) = w(k ) − η∇J (k )

Where η is called the learning rate, which is a small constant to maintain stability in
the search by ensuring that operating point doesn’t move too far along the performance
surface, and ∇J (k ) denotes the gradient of the performance surface at the kth iteration.
The method will work as illustrated in Figure 4-4.

FIGURE 4-40 SEARCHING PATH OF STEEPEST DESCENT

3.2 Performance Criteria

Performance criterion can be calculated or measured and used to evaluate the system’s
performance. A system is considered an optimum control system when the system
parameters are adjusted so that the index reaches an extreme value, usually a minimum
value. ISE is easily adapted for practical measurements; the squared error is
mathematically convenient for analytical and computational purposed. The integral of
the square of the error, ISE, which is defined as below:
T
ISE = ∫ e 2 (t )dt
0

Where

e(t ) = c(t ) − r (t )

The r (t ) represents the reference input and c (t ) represents the system response. The
upper limit T is a finite time chosen somewhat arbitrarily so that the integral
approaches a steady-state value.

As described as above, to obtain optimal values of PID controller parameters the


following steps should be performed:

 Invoke Simulink model


 Setting PID initial values
 Simulation
 Change parameter of PID controller according to steepest decent
algorithm.
 Observe value of criterion
 If this value is minimal finish tuning
 Otherwise go back
The procedure of programming discusses in Appendix B which give the details about
the algorithm apply into our work.

4. SIMULATION RESULTS

The proposed control schemes were implemented and tested within the simulation
environment of the TRMS. The relationship of element shows as Figure 4-5. The
system proposed uses a personal computer, MATLAB and associated toolboxes which
act as an application host environment. The system model is developed by using
Simulink which gives the user a graphical based system for modeling and control. The
algorithm created in Matlab command then passed to the Matlab workspace. Via
Matlab workspace, the program or algorithm interfaces to the Simulink model. The
simulation executed in Simulink then returned the result to Workspace until program
search one data which satisfy our requirement.

Steepest Decent
Algrothim

Initial Matlab
Command
Return

Matlab Simulink
Workspace Model
Return

FIGURE 4-41 CONTROL SYSTEM DEVELOPMENT FLOW DIAGRAM

Their performances have been thoroughly investigated and corresponding results in


time domain are presented in this section. The time domain specifications such as
overshoot, rise time, settling time, steady state error are compared for each case.

The horizontal Simulink model with a fix structure shows in Figure 4-5 using the
simplified steepest decent algorithm for off-line tuning the parameters of PID
controllers, the tracking output of tail rotor system is shown in Figure 4-6. Therefore,
we find the optimal PID parameters as kp=1.2811 ki=0.63003 kd=0.61756. By using
our control scheme, the tuning mechanism has resulted in a signification reduction of
overshoot in comparison to the system with PID controller only. The oscillation in the
system response has been significantly reduced due to the steepest decent method. This
can be observed by comparing the system performance after steepest decent tuning
process is shown in Figure 4-6.
+
-
Step
PID Model of
horizontal part
TRMS

Horizontal
Desire+Actual

ISE
Criterion

FIGURE 4-42 SIMULINK MODEL IN HORIZONTAL AXIS

FIGURE 4-43 SYSTEM SIMULATION RESPONSE (HORIZONTAL)

The vertical Simulink model show in Figure 4-7 the tracking output of main rotor for
off-line tuning PID parameters is shown in Figure 4-8, the optimal parameters of PID
controller are kp=1.7673 ki=2.7565 kd=4.0901. By using our control scheme, the
tuning mechanism has resulted in a signification reduction of overshoot in comparison
to the system with PID controller only. The oscillation in the system response has been
significantly reduced due to the steepest decent method. This can be observed by
comparing the system response shown in Figure 4-8.
+
-
Step
PID Model of
vertical part
TRMS

Vertical
Desire+Actual

ISE
Criterion

FIGURE 4-44 SIMULINK MODEL IN VERTICAL AXIS

FIGURE 4-45 SYSTEM SIMULATION RESPONSE (HORIZONTAL)

The TRMS Simulink model show in Figure 4-9 the tracking output of TRMS for off-
line tuning PID parameters is shown in Figure 4-10, the optimal parameters of PID
controller in Tail rotor are kp=1.1002 ki=1.57 kd=2.87 and Main rotor are kp=0.077
ki=0.385 kd=1.186. Figure 4-10 shows the response provided by steepest decent PID
regulator which developed in this article. This figure demonstrates that the controller
which the system had was not the most adequate one, so an improvement was carried
out. In tail rotor, comparing Figure 4-6 with Figure 4-10, the settling time has been
significantly increased due cross-coupling effect change the model of the system. The
effect has also affect the system to increase rise time in system response by amount of 5
second. On the other hand, at main rotor, Figure 4-6 and Figure 4-8 show the when the
cross-coupling effect was added, the level of rise time and settling time was significant.
However, the oscillation of system had an big improvement.

-
+
Tail rotor
Desire+Actual
Step
PID

+
-
Step
PID Main rotor
Desire+Actual

ISE
Criterion

FIGURE 4-46 SIMULINK MODEL OF TRMS WITH PID CONTROLLER

FIGURE 4-47 THE SYSTEM SIMULATION WITH PID CONTROL SCHEME

Finally, we designed the PID controllers with steepest decent algorithm both in 1-DOF
and 2-DOF. The analysis of the optimized control scheme was performed by
comparing its response to that of the original system, Figure 4-6, Figure 4-8 and Figure
4-10 show the system response based on PID compensators. The initial PID parameters
use company default setting and apply steepest decent algorithm to optimize controller.
The system response shows with unit step wave. This result can be an excellent
reference for comparing.
CHAPTER 5

DEADBEAT ROBUST SCHEME

1. INTRODUCTION

In order to control system here, we use the technique proposed in [17, 40][18, 28]
which includes a PID controller and a deadbeat controller. In [17][18] Dawes claims
that “response will remain almost unchanged when all the plant parameters vary by as
much as 50%”. We are going to decouple the system into two SISO systems. We will
design a controller for each of the SISO systems using the above method. This time
optimal controllers designed are robust to system parameter changes. When we join the
two SISO systems together, the coupling effects are considered as system parameter
changes, and can be handled the controller well. In directly, we have achieved the time
optimal control for this MIMO TRMS system.

2. REVIEW OF DEADBEAT CONTROLLER

The goal of a deadbeat controller is to drive a system to a desired state in a finite


number of time steps. This is accomplished by having an accurate model of the plant.
Often the goal for a control system is to achieve a fast response to a step command with
minimal overshoot. We define a deadbeat response as a response that proceeds rapidly
to the desired level and holds at that level with minimal overshoot. We use the ±2%
band at the desired level as the acceptable range of variation from the desired response.
Then if the response enters the band at time Ts , it has satisfied the settling time Ts
upon entry to the band, as illustrated in Figure The deadbeat response has the following
features:

 Steady-state error=0
 Fast response →minimum rise time and settling time
 0.1% ≤ percent overshoot < 2%

1.4

1.2
Less oscillatory No steady-state error
No ripples
1 ±2%
Settling down in the shortest time
Response

0.8
Faster initial response
0.6

0.4

0.2

0
0 1 2 3 4 5 6 7
Time

FIGURE 5-48 THE CHARACTERISTICS OF DEADBEAT RESPONSE

To control a system which achieves a fast response with minimum possible settling
time and zero steady-state error, the system met the above is called deadbeat control
system. Figure 5-2 illustrates the operation of how deadbeat controllers affect the
system to satisfy the above requirement.
1.2 1.2

1
Deadbeat 1

0.8
Controller 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0

-0.2 -0.2
0 5 10 15 20 25 0 5 10 15 20 25

FIGURE 5-49 THE PERFORMANCE OF DEADBEAT CONTROLLER

To design a system with deadbeat response, we consider the transfer function of a


closed-loop system, T ( s ) , is a third-order system:

ωn 3
T (s) = EQUATION 5-23
s 3 + αωn s 2 + βωn 2 s + ωn3

First, normalized the system by dividing the numerator and denominator by ωn


3

ωn 3
T (s) =
s 3 ωn3 + α ( s2 ωn2 ) s2 + β ( s ωn ) s + 1

s
Let S = to obtain
ωn

3 2
T ( s ) = 1 ( S + α S + β S + 1) EQUATION 5-24

Equation 5-1 is the normalized, third-order, closed-loop transfer function. For a higher-
order system, the same method is used to deriver the normalized equation. The
coefficients of the equation α , β , γ and so on which were selected by looking up
Table2, for example, if we have a 4th order system with a required settling time of 0.95
seconds, we note from Table 2 that the normalized settling time is

ωnTs = 4.81

Therefore we require
4.81 4.81
ωn = = = 5.063
Ts 0.95

The characteristic equation of the closed loop transfer function equal to:

s 4 + αωn s3 + βωn 2 s2 + γωn3 s + ωn 4

Where

α = 2.20 ; β = 3.50 ; γ = 2.80

Once ωn is chosen, the complete closed-loop transfer function is know form Equation
5-1. Hence, the transfer function is equal to:

5.063
T (s) =
s + 11.1386s + 89.71889 s2 + 363.397 s + 657.1
4 3

Let us consider a feedback system as shows in Figure 5-3.

Where

K [ K 3 ( s 2 + Xs + Y )] G ( s ) = 1
Gc ( s) = ;
s [ ( s + 1)( s + 2)( s + 4)]
H1 ( s) = (1 + K b s) ; H 2 ( s) = K a

The closed-loop transfer function is

C (s ) K [ K 3 ( s 2 + Xs + Y )]
= 4
R( s ) s + { 7 + K b KK 3 } s + {14 + KK3 + KKb K3 X + } s2 + {8+ Ka + KK3 X + KKb K3 Y } s+ {KK3 Y
3
}

The required closed-loop system has the characteristic equation

s 4 + 11.1386 s3 + 89.71889s2 + 363.397 s + 657.1

Then we determine that


{ K3Y } = 657.1 ; { 8 + K a + K3 X + Kb K3 Y} = 363.397

{ 14 + K3 + Kb K3 X } = 89.71889 ; { 7 + K b K3 } = 11.1386

Then

K 3 = 17 , Y = 38.6529 , K b = 0.243 , X = 14.21 , K a = −45.848

To sum up, the step response of this system are shown in Figure 5-4 which have
overshoot under 2% and the final settling time is about 0.97 seconds

R(s) C(s)
Σ Gc Σ G (s )

- -
H2

H1

FIGURE 5-50 CONTROL SYSTEM WITH FEEDBACK

FIGURE 5-51 RESPONSE OF 4TH ORDER SYSTEM


Order(np) α β γ δ Ts’
2nd 1.82 4.82
3rd 1.90 2.20 4.04
4th 2.20 3.50 2.80 4.81
5th 2.70 4.90 5.40 3.40 5.43

TABLE 2 DEADBEAT COEFFICIENTS AND RESPONSE TIMES. ALL TIMES

NORMALIZED BY ωn

3. DESIGN METHOD AND PROCEDURES

First, the de-couple techniques are used to separate the system into two SISO one.
Without angular momentum and reaction turning moment, the TRMS system will be
modelled into two 1-D systems as below:

Decouple the TRMS to become two separate systems

Horizontal part (Tail Rotor)

H2(s)

Ka2
- -
Rt(s) Σ K2 PID2 Σ Horizontal C (s)
t

FIGURE 5-52 THE STRUCTURE OF THE ROBUST SYSTEM DESIGN

(HORIZONTAL)

Vertical part (Main Rotor)


Rm(s) Cm(s)
Σ K1 PID1 Σ Vertical
- -
Ka1

H1(s)

FIGURE 5-53 THE STRUCTURE OF THE ROBUST SYSTEM DESIGN

(VERTICAL)

The Figure 5-3 is the basic structure of the robust system design, Richard Dorf and Jay
Dawes created in 1994. The proportional-integral-derivative (PID) controller enables a
system to achieve robustness; however, it will only work for lower order plants. As a
result, there is a need for more variable gain when higher order systems are analysed.
This design method has been tested which will result in systems that are insensitive to
plant parameter variations of up to ±50% .

R(s) C(s)
Σ K Gc (s ) =
K s (s 2 + Xs + Y )
s Σ Plant G(s)
- -
Ka
H(s)

FIGURE 5-54 THE BASIC STRUCTURE OF THE ROBUST SYSTEM

3.1 Case study-Main rotor

Here, an example were created to illustrate the procedure; a third order plant which is
the transfer function of tail rotor in TRMS system. Use PID controller as Gc ( s ) . Refer
to figure of the basic structure: it can be simplify as Figure 5-4.
R(s) C(s)
Σ K Gc ( s ) =
K s ( s 2 + Xs + Y )
s Σ Plant G(s)
- - G2 ( s)
G1 ( s)
Ka H 2 ( )s
H(s)
H1 ( s)

FIGURE 5-55 MODIFY OF BASIC STRUCTURE OF ROBUST SYSTEM

Determine the close loop transfer function

C (s) G1 ( s )G2 ( s )
=
R ( s ) 1+G2 ( s ) H 2 ( s )+ G1 ( s )G2 ( s ) H1 ( s )

Where

K [ K3 ( s 2 + Xs +Y )]
G1 ( s )= =Gc ( s )
s

15.02 15.02
G2 ( s ) = = =G ( s )
s3 +3.458s 2 + 2.225s s ( s +2.603)( s +0.8547)

H1 ( s ) =(1+ Kb s )

H 2 ( s )= K a

The close loop transfer function can be drawn as:

C (s) 15.02 K [ K 3( s 2 + Xs +Y )]
=
R ( s ) s 4 +{ 3.458+15.02 K KK } s 3+ { 2.225+ 15.02 KK + 15.02 KK K X+ } s 2
b 3 3 b 3
+{ 15.02 K a +15.02 KK3 X+ 15.02 KKb K3Y } s+ {15.02 KK3Y }

By Richard’s design deadbeat response method; the characteristic equation of the


transfer function is equal to the characteristic equation of the deadbeat transfer
function. To obtain the characteristic equation of the deadbeat transfer function, we set
the characteristic equation of the closed loop transfer function equal to:
s 4 + αωn s3 + βωn 2 s2 + γωn3 s + ωn 4

By looking up table 2 to select the coefficients, to determine n p for Gc ( s )G ( s ) , where

n p equals the number of poles in Gc ( s )G ( s )

Set

α = 2.20 ; β = 3.50 ; γ = 2.80

ωn = Ts′ /(80% of the desired settling time Ts )

Ts′ 4.81
ωn = = =3.00625
Ts ×80% 1.6

Therefore, the characteristic equation of the deadbeat transfers function is:

s 4 + 6.6138s3 + 31.6314s2 + 76.0735s + 81.6771

Comparison of the characteristic equation and Set K equal to 1 then

{ 7+ Kb K3} = 11.1386

{ 14+ K3 + Kb K3 X } = 89.71889

{ 8+ K a + K3 X + Kb K3Y } = 363.397

{ K3Y } = 657.1

Hence

K b = 0.243; Ka = −45.848

K 3 = 17; X = 14.21; Y = 38.6529

Select K until system meet deadbeat requirement


3.2 Case study-Tail rotor

One more example to demonstrate the procedure of our design method, a third order
plant which is the transfer function of main rotor in TRMS system.

Determine the close loop transfer function:

C ( s) G1 ( s)G2 ( s )
=
R ( s ) 1 + G2 ( s) H 2 ( s) + G1 ( s)G2 ( s ) H1 ( s)

Where

K [ K 3 ( s 2 + Xs + Y )]
G1 ( s ) = = Gc ( s) ;
s

1.519 1.519
G2 ( s) = = = G(s)
s + 0.748s + 1.533s + 1.046 ( s + 0.6982)( s + 0.04983s + 1.498)
3 2 2

H1 ( s) = (1 + K b s) ; H 2 ( s) = K a

The close loop transfer function can be drawn as:

C (s) K [ K 3 (s 2 + Xs + Y )]
= 4
R( s ) s + { 0.748 + 1.519 KK 3K b } s + {1.533 + 0.1549 KK 3 + 1.519 KK 3 XK b } s 2 + {1.046 + 1.519K a + 0. 1549KK 3 X + 1.519KK 3YK b } s + {1.519KK 3Y
3
}

The characteristic equation of the closed loop transfer function equal to:

s 4 + αωn s3 + βωn 2 s2 + γωn3 s + ωn 4

By looking up table, Where:

α = 2.20 ; β = 3.50 ; γ = 2.80

Ts′ 4.81
Ts = 2sec; ωn = = = 3.00625
Ts × 80% 1.6
Therefore

s 4 + 6.6138s3 + 31.6314s2 + 76.0735s + 81.6771

Comparison of the characteristic equation and Set K equal to 1 then

{ 0.748 + 1.519 K3 Kb } s3 = 6.6138

{ 1.533 + 1.519 K3 + 1.519 K3 XKb } s 2 = 31.6314

{ 1.046 + 1.519 K a + 1.519 K3 X + 1.519 K3YKb } s = 76.0735

{ 1.519 K3Y } = 81.6771

Hence

K b = 0.5; K a = −2.5453

K 3 = 7.723; X = 3.131; Y = 6.963

3.3 Apply in 2-DOF situation

When the individual of each parameter obtained, it was carried out and apply into 2-
DOF model which show in Figure 5-9.
H2(s)

Ka2
- -
Rt(s) Σ K2 PID 2 Σ Horizontal C (s)
t
angular
momentum
&
reaction turning
moment
Rm(s) Cm(s)
Σ K1 PID1 Σ Vertical
- -
Ka1

H1(s)

FIGURE 5-56 THE CONTROL STRUCTURE OF THE 2-D SYSTEM

4. PERFORMANCE EVALUATION

The proposed control schemes were implemented and tested within the simulation
environment of the TRMS. Their performances have been thoroughly investigated and
corresponding results in time domain. The time domain specifications such as
overshoot, settling time, steady state error are compared for each case.

Figure 5-10, Figure 5-12 and Figure 5-14 is a Simulink model and also can be consider
as a block diagram representation of the TRMS system both in vertical, horizontal and
2-DOF respectively.

Figures 5-11, 5-13 and 5-15 illustrate that the basic performance of the TRMS position
and control effort. It can be also seen that the system really settles except Figures 5-15
and this is due to the introduction of disturbance. But even with the introduction of
disturbance, it can be seen that both tail and main rotor position reach the desire
position faster than previous study.

Overall, in this chapter a discussion of the simulation and implementation of the


deadbeat robust scheme for TRMS position control was given. Form the simulation
results it can be seen that the control scheme moves the TRMS system to the desired
location which the system exhibit cross-coupling effects. The control strategy is shown
to be robust in the presence of disturbance; even there is some oscillation occurred that
the system response still meets all the deadbeat control requirements. This is clearly
evident to show the scheme is robust against parameter uncertainties.

For 1-DOF vertical plant where:

1.519
G (s) = 3
m s + 0.748s + 1.533s + 1.046
2

The settling time is desired to be 2 seconds. The gains K a =−2.5453 and K3 =7.723 are
arbitrarily set. This result in X =3.131 and Y =6.963 . K = 10 is found to produce the desired
response; thus, the system is now complete, and the response can be determined in
Figure 5-11.

+ K +
- -
PID
Step controller
Model of
Ka vertical part
TRMS
Vertical
Desire+Actual

H(s)
workspace

FIGURE 5-57 THE SIMULINK MODEL OF VERTICAL AXIS WITH DEADBEAT

ROBUST
FIGURE 5-58 THE RESPONSE OF MAIN ROTOR (K=10)

For 1-DOF Horizontal plant with a third order system where:

15.02
G ( s) = 3
t s + 3.458s 2 + 2.225s

It also has a settling time of 2seconds. To find ωn , Ts′ is divided by 80% of the desired

settling time. Therefore, choosing Kb =0.5 and X =7.323 result in Y =12.95 and K a =−0.73 .
Setting K = 7 gives the response.

+ K +
- -
PID
Step controller
Model of
Ka Horizontal
part TRMS
Horizontal
Desire+Actual

H(s)
workspace

FIGURE 5-59 SIMULINK MODEL OF HORIZONTAL AXIS WITH DEADBEAT

ROBUST SCHEME
FIGURE 5-60 THE RESPONSE OF TAIL ROTOR (K=7)

Apply study into 2-DOF, once obtain the result in 1-DOF system, these study results
can be carried to apply into 2-DOF system. Figure 5-7 shows the block diagram of
TRMS system combine with deadbeat robust algorithm. Figure 5-8 reports the final
result of our system. The settling time of both tail and main rotor set to 2 seconds. By
tuning each k both in horizontal and vertical plant until the system response meet the
requirement of deadbeat response.

FIGURE 5-61 SIMULINK MODEL OF TRMS WITH DEADBEAT ROBUST

ALGORITHM
FIGURE 5-62 THE RESPONSE OF TAIL AND MAIN ROTOR

.
CHAPTER 6

CONCLUSIONS AND FUTURE


DEVELOPMENT

1. CONCLUSIONS

A TRMS model, whose dynamics resemble that of a helicopter, has been successfully
identified. System identification is an ideal tool to model no-standard aircraft
configurations, whose flight mechanics are not well understood. The extracted model
has predicted the system behavior well. High fidelity system model is an important first
step in control system design and analyses.

This project also described how the control scheme reduces oscillation and settling time
between PID control and deadbeat robust control. Simulation results for off-line tuning
the parameters of PID controller have been illustrated to show the effectiveness of the
optimization–base design and model base design. However, model base design
procedure need accurate transfer function, in simulation result especially in main rotor,
the system response is out of our prediction which might need to re-modify the system
or use other techniques to reduce the effect from the modeling the system. Careful
selection of excitation signal is also an important part of nonlinear system
identification. Without due consideration to his issue, the obtained model would not be
able to capture the system dynamics, resulting in a poor model.

Simulation results for off-line tuning the parameters of PID controllers have been
illustrated to show the effectiveness of the steepest decent algorithm base design. This
work may provide a design guideline for design the controller. We have successfully
applied the time optimal robust controller design technique to out MIMO TRMS
system. Comparing with the system obtained using PID controllers; the system
performance has been improved dramatically. For example, the settling time has been
shortened 20 seconds and the overshoot has been reduced about 20%.

This control scheme does not include many complicated math and calculation. It is
generally base on the deadbeat controller design procedure, and the tuning procedure of
a PID controller. It is easy to be accepted by industrial designers. Further more, we
only change the control scheme for the system without any new investment for
controller. In PID controller design in 2 degree of freedom, at least, it includes 6
parameters. However we can reduce it to 2 parameters in multi-input and multi-output
with cross-coupling system.

Comparing the responses, we can clearly see the following:

In the two SISO systems

 The settling time after start-up has been reduced from approximately six
and twenty sec in tail and main rotor respectively.
 The amount of overshoot has been reduced.
In the joined 2-D system

 he settling time has been reduced up to 20 second.


 The amount of overshoot has been reduced as well
2. RECOMMENDATIONS FOR FUTURE DEVELOPMEN

In general, tasks of a TRMS control system can be listed as follows:

 Stabilisation. Design a controller so that the state vector of the closed-loop


system is stabilised around a desired point of the state space.
 Tracking. Design a controller so that the closed-loop system output
follows a given trajectory.
PID control systems are widely used as a basic control technology in today’s industrial
control systems. However, the tuning of PID control systems is not always easy. In
order to improve several advanced PID control technologies can apply into this project.
When obtain an excellent 1-DOF model on software simulation, this system could be
tested with different kind of control algorithms. Different control scheme also can be
implemented. In order to apply into real platform using other tuning method should be
an interesting work for the future.

The further work can be done on improving the system on improving the transfer
function. As long as decrease the inaccuracy of transfer function, the control scheme
can accept more disturbances or any other influence that we overlook. As shown in
chapter 5, to design a deadbeat robust control system which need higher accurate
transfer function to represent the system. System identification is a well established
technique for modelling of complex systems whose dynamics is not well understood.
During identification the parameter of the mathematical model are tuned to obtain a
satisfactory degree of conformity of the model with the real system. The point is to tune
the parameters of the model in such a way, that the outputs of the model fit the
experimental data in the sense of a criterion function. Figure 6-1 illustrates the idea of
tuning the coefficients of the model using relate algorithm by trial and error. By
changing model parameters, using an appropriate minimisation method, we find the
minimum value of the objective function Q and the corresponding values of model
parameters.
FIGURE 6-63 BLOCK DIAGRAM OF IDENTIFICATION PROCEDUE

For the cross-coupling system, one solution is designing a decoupler. Recalling to


Figure 3-3, it can be modified as the following system, Figure 6-2, the purpose of the
decouplers is to cancel the effects of the cross-coupling blocks. In other words,
decoupler Dvh cancels the effect from vertical part, and Dhv cancels the effect from
horizontal part.

Decoupler

Uh - αh
+
Gch
+
Gh Ghh
+

+ +

Dhv Ghv

Dvh Gvh
+
Uv + αv
+
Gcv +
Gv Gvv
+

FIGURE 6-64 BLOCK DIAGRAM OF TRMS SYSTEM WITH DECOUPLERS


References

[1] B. U. Islam, N. Ahmed, D. L. Bhatti, and S. Khan, "Controller design using


fuzzy logic for a twin rotor MIMO system," 2003, pp. 264-268.

[2] W.-Y. Wang, T.-T. Lee, and H.-C. Huang, "Evolutionary design of PID
controller for twin rotor multi-input multi-output," Shanghai, China, 2002,
pp. 913-917.

[3] F. M. Aldebrez, M. S. Alam, and M. O. Tokhi, "Input-shaping with GA-


tuned PID for target tracking and vibration reduction," Limassol, Cyprus,
2005, pp. 485-490.

[4] S. M. Ahmad, A. J. Chipperfield, and M. O. Tokhi, "Modelling and control


of a twin rotor multi-input multi-output system," 2000, pp. 1720-1724
vol.3.

[5] I. Z. Mat Darus, F. M. Aldebrez, and M. O. Tokhi, "Parametric modelling of


a twin rotor system using genetic algorithms," Hammamet, Tunisia,
2004, pp. 115-118.

[6] S. Juhng-Perng, L. Chi-Ying, and C. Hung-Ming, "Robust control of a class


of nonlinear systems and its application to a twin rotor MIMO system,"
2002, pp. 1272-1277 vol.2.

[7] Manual, Twin Rotor MIMO System Manual. UK: Feedback Instruments
79

Ltd., 1996.

[8] S. M. Ahmad, A. J. Chipperfield, and O. Tokhi, "Dynamic modeling and


optimal control of a twin rotor MIMO system," 2000, pp. 391-398.

[9] S. M. Ahmad, M. H. Shaheed, A. J. Chipperfield, and M. O. Tokhi,


"Nonlinear modelling of a twin rotor MIMO system using radial basis
function networks," 2000, pp. 313-320.

[10] F. M. Aldebrez, I. Z. M. Darus, and M. O. Tokhi, "Dynamic modelling of a


twin rotor system in hovering position," 2004, pp. 823-826.

[11] K. H. Ang, G. Chong, and Y. Li, "PID control system analysis, design, and
technology," IEEE Transactions on Control Systems Technology, vol.
13, pp. 559-576, 2005.

[12] M. Yukitomo, T. Shigemasa, Y. Baba, and F. Kojima, "A two degrees of


freedom PID control system, its features and applications," Melbourne,
Australia, 2004, pp. 456-459.

[13] Kavangh, "Noninteracting Controls in Linear Multivariable Systems,"


Transactions of the American Institute of Electronic Engineers, 1957.

[14] P. Falb and W. Wolovich, "Decoupling in the design and synthesis of


multivariable control systems," Automatic Control, IEEE Transactions
on, vol. 12, pp. 651-659, 1967.

[15] Rosenbrock, "Design of multivariable control systems using inverse


Nyquist array," IEEE Transactions on Automatic Conyrol, pp. 651-659,
1969.

[16] F. G. Shinskey, Process-control systems : application, design, adjustment,


2nd ed. New York: : McGraw-Hill, 1979.

[17] J. Dawes, L. Ng, R. Dorf, and C. Tam, "Design of deadbeat robust


systems," Glasgow, UK, 1994, pp. 1597-1598.

[18] J. G. Ziegler and N. B. Nichols, "Optimum Settings for Automatic


80

Controllers," Transactions of the ASME, pp. 759-768, 1942.

[19] W. Wei-Yen, L. Tsu-Tian, and H. Hung-Chih, "Evolutionary design of PID


controller for twin rotor multi-input multi-output system," 2002, pp. 913-
917 vol.2.

[20] S. M. Ahmad, A. J. Chipperfield, and M. O. Tokhi, "Dynamic modelling and


linear quadratic Gaussian control of a twin-rotor multi-input multi-output
system," Proceedings of the Institution of Mechanical Engineers Part I-
Journal of Systems and Control Engineering, vol. 217, pp. 203-227,
2003.

[21] L. Chuan-Sheng, C. Liang-Rui, L. Bing-Ze, C. Shih-Kai, and Z. Zhao-


Syong, "Improvement of the Twin Rotor MIMO System Tracking and
Transient Response Using Fuzzy Control Technology," 2006, pp. 1-6.

[22] J. G. Juang and W. K. Liu, "Fuzzy compensator using RGA for TRMS
control," in Computational Intelligence, Pt 2, Proceedings. vol. 4114,
2006, pp. 120-126.

[23] S. M. Ahmad, A. J. Chipperfield, and M. Tokhi, "Dynamic modelling and


open-loop control of a twin rotor multi-input multi-output system,"
Proceedings of the Institution of Mechanical Engineers Part I-Journal of
Systems and Control Engineering, vol. 216, pp. 477-496, 2002.

[24] S. M. Ahmad, A. J. Chipperfield, and M. O. Tokhi, "Dynamic modelling and


control of a 2-DOF twin rotor multi-input multi-output system," 2000, pp.
1451-1456 vol.2.

[25] S. M. Ahmad, A. J. Chipperfield, and M. O. Tokhi, "Parametric modelling


and dynamic characterization of a two-degree-of-freedom twin-rotor
multi-input multi-output system," Proceedings of the Institution of
Mechanical Engineers Part G-Journal of Aerospace Engineering, vol.
215, pp. 63-78, 2001.

[26] S. M. Ahmad, A. J. Chipperfield, and M. O. Tokhi, "Dynamic modelling and


open-loop control of a two-degree-of-freedom twin-rotor multi-input
81

multi-output system," Proceedings of the Institution of Mechanical


Engineers Part I-Journal of Systems and Control Engineering, vol. 218,
pp. 451-463, Sep 2004.

[27] S. M. Ahmad, M. H. Shaheed, A. J. Chipperfield, and M. O. Tokhi, "Non-


linear modelling of a one-degree-of-freedom twin-rotor multi-input multi-
output system using radial basis function networks," Proceedings of the
Institution of Mechanical Engineers Part G-Journal of Aerospace
Engineering, vol. 216, pp. 197-208, 2002.

[28] L. TE-WEI, "The 2-D Optimal Control of a Twin Rotor MIMO System," in
Faculty of Engineering & Surveying. vol. Master of Engineering
Technology Toowoomba, Australia: University of Southern Queensland,
2005.

[29] K. J. Astrom and T. Hagglund, "The future of PID control," Control


Engineering Practice, vol. 9, pp. 1163-1175, 2001.

[30] K. Ogata, Modern control engineering, 4th ed. Upper Saddle River:
Prentice Hall, 2002.

[31] M. di Bernardo, Analysis and control of nonlinear process systems, K.M.


Hangos, J. Bokor, G. Szederkenyi; Springer-Verlag London Limited
2004; ISBN: 1-85233-600-5 vol. 42: Elsevier Ltd, Oxford, OX5 1GB,
United Kingdom, 2006.

[32] T. Kara and I. Eker, "Nonlinear modeling and identification of a DC motor


for bidirectional operation with real time experiments," Energy
Conversion and Management, vol. 45, pp. 1087-1106, 2004.

[33] J.-H. Horng, "Neural adaptive tracking control of a DC motor," Information


Sciences, vol. 118, pp. 1-13, 1999.

[34] S. E. Lyshevski, "Nonlinear control of mechatronic systems with


permanent-magnet DC motors," Mechatronics, vol. 9, pp. 539-552,
1999.

[35] E. Bristol, "On a new measure of interaction for multivariable process


82

control," Automatic Control, IEEE Transactions on, vol. 11, pp. 133-134,
1966.

[36] M. L. Luyben and W. L. Luyben, Essentials of process control. New York:


McGraw-Hill, 1997.

[37] C. A. Smith and A. B. Corripio, Principles and practice of automatic


process control, 2nd ed. New York: J. Wiley, 1997.

[38] R. L. Fox, Optimization methods for engineering design. Reading, Mass. b


Addison-Wesley Pub. Co., 1971.

[39] R. L. Burden and J. D. Faires, Numerical analysis, 7th ed. Australia ;


Pacific Grove: Brooks/Cole, 2001.

[40] R. C. Dorf and R. H. Bishop, Modern control systems, 9th ed. Upper
Saddle River: Prentice Hall International, 2001.
Appendix A

THE PROCEDURE OF OPTIMIZATION


Appendix B

THE PROCEDURE TO DETERMINE SETTLING TIME


85

Steepest descent has been used with moderate success on a wide variety of problem.
Through the middle 1950s, it was one of the most popular methods. The general flow
chart shown below:

k p , ki , kd Are the control gain


T
S = ISE = ∫ e 2 (t ) dt
0

For (kp, ki, kd) we have S0

For (kp+δ kp, ki, kd) we have S1

For (kp, ki+δ ki, kd we have S2

For (kp, ki, kd +δ kd we have S3


86

Parameters update

γ ∂S
k p ' = kp −
∆ ∂k p
γ ∂S
ki ' = ki −
∆ ∂ki
γ ∂S
k d ' = kd −
∆ ∂kd

Repeat until convergence


γ stepsize
α* = ≈
∇ 2
 ∂S   ∂S   ∂S 
2 2

  +   + 
 ∂k p   ∂ki   ∂kd 
2 2 2
 ∂S   ∂S   ∂S 
∇=   +   + 
 ∂k ∂
 p   ki   ∂kd 

∂S S −S ∂S S 2 − S0
Sk p = ≈ 1 0 S ki = ≈
∂k p k p′ − k p ∂ki ki′ − ki

∂S S −S
S kd = ≈ 3 0
∂kd kd ′ − kd

We made the Matlab program to minimize the objective function by using the steepest
descent method. The source code show as below:
87
88

You might also like