Sunteți pe pagina 1din 12

Table of Contents

1. INTRODUCTION

2. TYPES OF SYSTEM REPRESENTATIONS

2.1. TRANSFER FUNCTION REPRESENTATION


2.2. BLOCK DIAGRAM REPRESENTATION
2.3.
STATE SPACE REPRESENTATION

3
5
5

3. NEGATIVE FEEDBACK

4. PID CONTROLLER

5. FULL STATE FEEDBACK

10

6. CONCLUSTION

11

7. REFERENCES

12

1.Introduction
Use of feedback systems in industrial applications dates back to the 18 th
century, when James Watt invented centrifugal governor to sense the speed of an
engine, and then use it as a negative feedback to control the steam valve of the
engine, which regulates the speed of the steam engine. A century later, James
Clerk Maxwell made a mathematical model of this system, and presented it to the
Royal Society in 1868. In recent times, negative feedback systems can be found in
virtually any industrial application. These systems are called closed loop control
systems [1].
In contrast to the closed loop control systems, there are open loop control
systems. This type of system is cheaper than closed loop system, because there is
no need for sensors. These systems are lot less robust, and can be only used in
non-critical applications that dont have any disturbances, because they dont have
any knowledge on error they are making.
Error of system can be found by subtracting desired output value with
actual output value. This error is then fed back to the input of the system, which
then can correct error.
There are several types of control systems, such as PID controller or full
state feedback. There will be words on both control systems in later chapters.
Open loop systems are naturally stable, which is not the case with closed
loop control systems. Engineers and mathematicians invented several ways to
determine stability of closed loop control systems. When designing closed loop
control system, stability of the system must be taken into account.
Today, most controllers are digital, which removes some issues found in
their analog counterparts such as noise and price, while bringing some new issues
like finite precision and non-continuous operation.
To design some closed loop system, engineer must model it first. Models
can be represented in numerous ways, like:
- Transfer function representation
- State Space representation
- Block diagram
First chapter of this article is dedicated to various types of representations of a
system, second will be dedicated to negative feedback. In third chapter there will
be words on PID controllers, and fourth one will try to explain full state feedback.
To simplify matter covered by the article, we will focus only on Single input Single output (SISO) systems.

2.Types of system representations


2.1. Transfer function representation
One of the most common and useful methods of representing a system is
by its transfer function. The transfer function is easily determined once the system
has been described as a single differential equation with zero initial conditions. To
find the transfer function, first take the Laplace transform of the differential
equation [2]. Laplace transform transforms function from time domain to complex
domain. It can translate differential equations to algebraic problem, which is easier
to solve. Laplace transform is defined by equation 1.

L ( f ( t ) ) = f ( t ) est dt (1)
0

Example:
Consider f (t)=1, Find Laplace transform of this function.

L ( f ( t ) ) = f ( t ) est dt
0

L ( 1 ) = 1est dt
0

est
L ( 1 ) =lim
s
T
L(1)=

1
s

Some Laplace transforms can be found in Table 1.


Function/Operator
Unit impulse

Table 1. Some Laplace transforms


Time domain
Laplace (s) domain
(t)
1

Unit step

h(t)

Delay

t-

1
s
s
e

f ( t ) dt

1
F (s)
s

Integral

t
0

Derivative

d f (t)
dt n

s n F ( s ) sn i f (i1) (0)
i=1

Consider the model of the system represented by differential equation (2),


where x(t) is input, and y(t) is output of the system, and an, bm are coefficients:
''
'
'
y ( t ) + a0 y ( t ) + a1 y ( t )=b0 x ( t )+ b1 x ( t ) (2)

Laplace transform of this differential equation is (3):


s 2 Y ( s ) +a 0 sY ( s ) +a1 Y ( s )=b0 sX ( s )+ b1 X ( s )( 3)

Transfer function is defined as the ratio between the input and the output of the
system represented with Laplace transform (4).
H ( s )=

b0 s +b1
Y ( s)
= 2
(4)
X (s) s +a 0 s+ a1

Equation s 2 +a0 s +a1 =0 is called the characteristic equation of the system, and
it determines the behavior of the system.
The numerator of transfer function contains zeros of the system, and denominator
contains poles of the system. Poles and zeros can be real or conjugated-complex.
Performance of a system largely depends on the type of the poles. Figure 1 shows
the response of the system with different types of poles.

Figure 1. Unit step response of second order system with


different types of poles:
a) Two real poles, system is overdamped, b) Imaginary pair of
poles, system is undamped, c) Conjugated-complex pair, system
is underdamped, d) Double real pole, system is critically
damped

Note that all poles are in the left half plane. If all poles are in the left half plane,
then the system is stable, otherwise, the system is unstable and response of the
system diverges to infinity. Poles that are lying on imaginary axis make system
critically stable, and then system oscillates within some range.

2.2. Block diagram representation


Block diagram representation is away of showing relationships between
parts of a system, where every block contains one transfer function. Whole block
diagram can be reduced to single transfer function, and single transfer function can
be represented by several smaller transfer functions in a block diagram. Transfer
function (4) is represented as a block diagram in figure 2.

Figure 2. Block diagram representation of the


transfer function (4)

Although not common, above block diagram representation can be


decomposed into several blocks, as shown on figure 3.

Figure 3. Decomposed block diagram of the transfer function (4)

In the example above, it is shown that block diagrams can have summers.
Blocks b0s and b1 are summed together, and then multiplied by the block on the
right. This block diagram representation is equivalent to the one shown in figure 2.
Block diagrams and transfer functions are interchangeable. Apart from summing
and multiplying, there are more block diagram operations, which are not on the
topic of this article.

2.3. State space representation


Transfer functions are part of the classical control system theory, where
differential equations describing a system are transformed from time domain to
Laplace domain. In modern control theory, model of a system is described by a set
of equations in time domain named State space representation. This representation
takes n-th order differential equation and translates it to the set of n first order
equations.
State variables x (t) can be reconstructed from the measured input-output data, but
states are not directly measured during an experiment. This representation is good

because it is easy to estimate a model of the system knowing only inputs, outputs
and order of the system [3]. General form of a State space representation is given
by the two equations shown in (5) and (6).
q ( t )= Aq ( t ) + Bu ( t ) (5)
y (t )=Cq ( t )+ Du ( t ) (6)

Where:
-

A is the state matrix, which doesnt have to have any physical meaning
B is the input matrix
U(t) is input, which is function of time
C is the output matrix
D is direct transition (feedtrough) matrix [4]

For example, consider differential equation (7):


y (t ) +a1 y ( t ) +a 0 y ( t )=u ( t ) (7)

We can substitute:
x 1= y
x 2= y = x1 (8)

So now, we have:
x 2 ( t )+ a1 x 2 ( t ) +a 0 x 1 (t )=u ( t )
x 2 ( t )=u ( t )a1 x2 ( t )a0 x 1 ( t ) (9)

State space representation is then:

][ ] [ ]

1 x 1 + 0 u ( t ) (10)
x ( t )= 0
a 0 a1 x 2 1
y ( t )= [ 1 0 ]

[]

x1
(11)
x2

State space representation of a system is not unique, meaning that one system can
have many different state space representations. This is good, because we can
choose State space representation which best suits our needs.

3.Negative feedback
Feedback occurs when the output of a system is fed back to the input. Negative
feedback is subtracting the output of a system with the input, reducing fluctuations
in the output. Negative feedback tends to increase robustness of the system,
linearity of the system, and decreases error of the system. Negative feedback also
removes the influence of disturbances on a system. Block diagram of a typical
negative feedback closed loop system is shown in figure 4.

Figure 4. Block diagram of negative feedback closed loop system

Transfer function of this system is defined as (12).


Y (s)
G(S)
=
(12)
U (S) 1+G( S)

More general example is given in figure 5, and its transfer function is given by
(13).

Figure 5. Block diagram of negative feedback with transfer


function in feedback branch

Y (s)
G(s)
=
(13)
U (s) 1+G ( S ) H ( S)

Block G(S) receives error that system made. u(t) is input of the system, also called
set point. Error of the system is defined as: e(t) = y(t)-u(t).

4.PID controller

Proportional-Integral-Derivative controller is the type of controller that tends


to reduce error made by a system. It also removes the disturbances. It has 3
different actions, each one doing different operation. Proportional component
depends only on the error of the system being controlled. It amplifies the error
with proportional gain KP, and feeds it to the system. Bigger the error, faster the
speed of error compensation will be. Bigger proportional gain will speed up
system response, but choosing too big proportional gain will make the system
unstable. Proportional component cant bring error to zero, but will reduce it. The
integral component sums the error over time. Integral response will grow unless
the error is zero, so its purpose is to cancel steady state error made by a
proportional component of PID controller. If controller cant bring down the error
to zero (for example, a motor is blocked in a servomechanism), integral action will
saturate controller. This phenomenon is called integral windup. The solution to this
problem is to turn off integral action if the error is too big, and only activate it to
cancel steady state error. This is called anti-windup mechanism. Integral gain is K I,
but is often used as integral time TI. Third action is called derivative action, and it
tracks changes of error during time. The output of the derivative action is
proportional to the derivative of the error. If derivative time T D is big, the system
will react more strongly and quickly to the changes, but choosing too big
derivative time will destabilize the system, because derivative response is very
sensitive to noise [4]. Ideal PID controlled system is shown in figure 6.

Figure 6. Ideal PID controller including process to be controlled

Transfer function of an ideal PID controller is given by:


1
H ( s )=K P + K I +K D s (14)
s

Standard form is more common, and its given by:


1
+T s (15)
TI s D
Most popular method of tuning PID controller parameters is Zeiger-Nichlos
method. Integral and derivative actions are turned off, and proportional gain is
increasing until the system becomes critically stable. Then, parameters are set
according to table 2 [5].
Table 2.
Rule name
KP
TI
TD
Classic ZN
0.6 KC
0.5 TC
0.125 TC
Some overshoot
0.33 KC
0.5 TC
0.33 TC
No overshoot
0.2 KC
0.5 TC
0.33 TC

H ( s )=K P 1+

KC is the proportional gain of critically stable system, and T C is the period of


oscillation of critically stable system.
Response of a system with P, PI, and PID controller is shown in figure 7.

Figure 7. Response of the system with various types of controllers (Setpoint is set to
10)
a) No controller, b) P controller system is unstable, c) PI controller system is
unstable, d) PID controller stable system without steady state error

In some cases, there is no need for all three actions of PID controller. In this
particular case, PD controller is more desireable than a full PID controller.
Response of the PD controller is given in figure 8.

5.Full

Figure 8. PD control system doesnt


overshoot

state
feedback

Full state feedback is used to tune closed loop system, so it can perform in the
desired manner. This type of feedback can make the system to be more responsive, or
to be less oscillative. Method employed in this feedback type is called Pole
placement, because the idea of the full state feedback is to choose desired poles that
dictates dynamics of the system being controlled. To design a full state feedback, all
states of the system must be measured. This is rarely done, because some states may
not have any physical meaning, there is no instrument that can measure some states,
or sensors are expensive. This can be solved with state observer. State observer is a
computer program that estimates states of a system from the input and the output of
the system. Full state feedback is shown in figure 9.

Figure 9. Full state feedback

Full state feedback is given by the equation:


x ( t )=(ABK ) x ( t ) + Bu ( t ) (16)
Point is to find values of K, so the system has desired poles.
First, make the desired characteristic polynomial by choosing the desired poles:
n

( s pi ) =0(17)
i=1

10

Then, make the characteristic equation of the controlled system:


det ( sI A+ BK )=0(18)
Solving these two equations will yield coefficients K needed to accomplish desired
dynamical behavior of the system [6].

6.Conclustion
This article shows an importance of the closed loop control systems that are
used in the various industries across the world. It explains what is feedback, and
two most commonly used types of controllers. It shows methods of determinating
coefficients needed for the proper operation of controllers, and shows how to use
these controllers. Also, this article covers theory and mathematics needed to
understand the closed loop control systems.

11

7.References
[1] Bohdan T. Kulakowski, John F. Gardner, J. Lowen Shearer, Dynamic modeling
and control of engineering systems, Cambridge university press, 2007
[2] Web site, http://lpsa.swarthmore.edu/Representations/SysRepTF.html, accessed
06.12.2014
[3] Web site, http://www.mathworks.com/help/ident/ug/what-are-state-spacemodels.html, accessed 13.12.2014
[4] Web site, http://www.ni.com/white-paper/3782/en/, accessed 22.12.2014
[5] Web site, http://www.mstarlabs.com/control/znrule.html, accessed 23.12.2014
[6] Web site, https://noppa.aalto.fi/noppa/kurssi/as-74.2112/luennot/AS74_2112_chap_4__print_.pdf, accessed 23.12.2014

12

S-ar putea să vă placă și