Process Control Engineering

Part II
Prof. Mahmoud A. El-Rifai
Prof. Reem S. Ettouney
 Revision of Frequency Response

Obtaining the frequency response of a system, consists in

expressing the amplitude ratio (sometimes called magnitude
ratio) (B/A) and the phase angle () as a function of frequency.

It is also usual to plot the amplitude ratio and the phase shift
versus frequency so that the complete response may be readily
visualized.
Several methods of presenting frequency response data have
been developed each showing its relative merits.
Frequency response is widely used in control systems design
analysis. It can be obtained by simple manipulations of the
differential equation.
It may be determined experimentally in the absence of a
mathematical model.
The frequency response of a general 1st order system:
^ 1 ( t / T) ^
( t / T )
o (t )  e  e i ( t )dt  Be ( t / T )
T
^
 oP.I  (A / T)e ( t / T )  e ( t / T ) sin( t ) dt
^ A
o P.I ( t )  . sin( t  tan 1 (T))
^
1   2 2
T
Since i  A sin( t )
For a first-order system of time constant T, the amplitude ratio
1
is given by A.R 
1  2T 2
And the phase angle is given by    tan 1 (T)
 Bode Diagram 1
A.R 
1  2T 2
1
It makes use of two Slope = -1

A.R
separate graphs, the first of
which is a plot of the Log-Log
logarithm of the A.R. on 0
1/T , rad/min

Phase Shift
the ordinate versus 𝝋 = −𝒕𝒂𝒏−𝟏 𝝎𝑻
logarithm of the applied -45
frequency on the abscissa. Semi-Log
-90
0.001 100

The second graph is a plot of the phase shift in degrees versus

the logarithm of the applied frequency. The two graphs are
usually plotted on the same sheet of paper with a common
abscissa.
 Amplitude Ratio, AR
𝑑𝑦
0.4 + 𝑦 = 0.2𝑥
𝑑𝑡
𝟎.𝟐 Slope =-1
𝑨. 𝑹 = and -1
𝟏+𝟎.𝟒𝟐 𝝎𝟐
𝝋= −𝒕𝒂𝒏−𝟏 (𝟎. 𝟒 × 𝝎)

w AR Fie w AR Fie
0.00001 0.2 -0.00023 5 0.04 -63.4349
𝟏
0.00005 0.2 -0.00115 6 0.029586 -67.3801 𝑻
0.0001 0.2 -0.00229 7 0.022624 -70.3462
0.0005 0.2 -0.01146 8 0.017794 -72.646
0.001 0.2 -0.02292 9 0.014327 -74.4759
0.005 0.199999 -0.11459 10 0.011765 -75.9638
0.01 0.199997 -0.22918 20 0.003077 -82.875
0.05 0.19992 -1.14576 30 0.001379 -85.2364
0.1 0.199681 -2.29061 40 0.000778 -86.4237
Phase Angle
0.15 0.199283 -3.43363 50 0.000499 -87.1376
0.2 0.198728 -4.57392 60 0.000347 -87.6141
0.3 0.197161 -6.84277 100 0.000125 -88.5679
0.4 0.195008 -9.09028 200 3.12E-05 -89.2838 −𝟒𝟓
0.5 0.192308 -11.3099 500 5E-06 -89.7135
0.6 0.189107 -13.4957 1000 1.25E-06 -89.8568
0.7 0.18546 -15.6422 2000 3.12E-07 -89.9284
0.8 0.181422 -17.7447 5000 5E-08 -89.9714
0.9 0.177054 -19.7989 10000 1.25E-08 -89.9857
1 0.172414 -21.8014
2 0.121951 -38.6598
2.5 0.1 -45
3 0.081967 -50.1944
4 0.05618 -57.9946
 -1
Example -2

-1
-1
-1
-3

Sketch the Bode

Diagrams of the
following
transfer function:
10
𝐺 𝑠 = 1 1 1
(1 + 40𝑠)(1 + 20𝑠)(1 + 10𝑠)
40 20 10
 -1
+1
Example
-1

𝐺 𝑠 = 1 + 5𝑠
𝐺 𝑗𝜔 = 1 + 𝑗 5𝜔
-1
𝐴𝑅 = 1 + 25𝜔 2
-1
𝜔→∞
𝐴𝑅 = 5𝜔
log 𝐴𝑅 = log 5 + log 𝜔

𝜑 = 𝑡𝑎𝑛−1 5𝜔

Sketch the Bode

Diagrams of the
following
transfer function:

10(1 + 5𝑠) 1 1
𝐺 𝑠 = 1
(1 + 7𝑠)(1 + 2𝑠) 7 5 2
 Example

Sketch the Bode Diagram of the following transfer functions: 𝐺 𝑠 = 𝑒 −10𝑠

𝐺 𝑗𝜔 = 𝑒 −10𝑗𝜔 = 𝑒 −𝑗 10𝜔 = cos 10𝜔 − 𝑗 sin 10𝜔

𝑒 𝑗𝜃 − 𝑒 −𝑗𝜃 𝑒 𝑗𝜃 + 𝑒 −𝑗𝜃
sin 𝜃 = cos 𝜃 = cos 𝜃 − 𝑗 sin 𝜃 = 𝑒 −𝑗𝜃
2𝑗 2
−sin 10𝜔
𝐴𝑅 = cos2 10𝜔 + sin2 10𝜔 = 1 𝜑 = 𝑡𝑎𝑛−1 = −10𝜔
cos 10𝜔 𝑨𝑹

𝝎 𝑨𝑹 𝝋, rad 𝝋, deg 1
0 1 0 𝟎
𝝅/𝟐𝟎 1 −𝝅/𝟐 −𝟗𝟎 𝝎
𝝅/𝟏𝟎 1 −𝝅 −𝟏𝟖𝟎
𝟑𝝅/𝟐𝟎 1 −𝟏. 𝟓𝝅 −𝟐𝟕𝟎
𝟒𝝅/𝟐𝟎 1 −𝟐𝝅 −𝟑𝟔𝟎

𝝋
 Dynamics of
Simple Distributed Parameter Systems
Spatial variation of the intensive variables characterizes the
behaviour of a wide variety of process equipment. This is
particularly prominent in tubular equipment e.g. homogeneous
and catalytic tubular reactors, heat exchangers, packed mass-
transfer equipment, etc. Strictly speaking, the assumption of
complete mixing, yielding a lumped parameter description of
stirred tank reactors is only an approximation to the actual flow
pattern inside such vessels.
3.1. Time Delay N M
L
Consider the shown system where
water is heated by the injection of
live steam. The temperature of the
water is raised at the nozzle N and
measured at point M.
Steam
It is obvious that any temperature change at N will not be measured
at M until the water at the new temperature travels from N to M. If
points N and M are L meters apart and the fluid is moving at a
velocity (v) m/s, the time delay before the new temperature is
indicated is (d = L/v) sec.

Temperature
If the temperature change at N is x = f(t)
expressible as: ^
x  f (t) (1)
the temperature at M will be:
^
y  f ( t  d )(2) y = f(t - d)
t
The time delay is sometimes referred to as distance/velocity lag,
transportation lag or dead time. It may be defined as the time
interval between an alteration in the value of a signal and its arrival
unchanged at another part of the system, interval arising solely
from the transmission speed of the signal.
When Eqs. (1) and (2) which are expressed in terms of deviation
variables are Laplace transformed we get:
X (s)  F(s) (3)
X(s) Y(s)
 d s e-ds
Y (s)  e .F(s) (4)
Hence the transfer function of a pure time Imaginary
delay is given by:
Y(s)  d s
e (5)
-1.0 1.0
Real
X(s) Nyquist Diagram of pure 
time delay
 A.R
Bodé Diagram of pure time delay
1.0
It is noticed that unlike the previously
discussed lumped parameter transfer 

functions, the phase angle of the pure time 0
delay transfer function does not have a
high frequency asymptote. 
Since the transportation lag is an inherent characteristic of
distributed parameter systems, the frequency response of such
systems will not show asymptotic behaviour of the phase angle in
the high frequency range.
In some automatic control applications, the function e-ds is
approximated to a ratio of two polynomials in s. The simplest
form of which is:
Y(s) 2  d s

X(s) 2  d s
 Sheet 1, Problem 1
For each of the following transfer functions:
A. Identify the poles and zeros of the transfer function.
B. Plot the response to a unit step change.
𝟏𝟎 𝟓
𝑮 𝒔 = −
𝟏 + 𝟎. 𝟏𝒔 𝟏 + 𝟎. 𝟎𝟒 𝒔
0.4 𝑠 + 10 − 0.5 𝑠 − 5 5 − 0.1 𝑠
𝐺 𝑠 = =
(1 + 0.1𝑠)(1 + 0.04 𝑠) (1 + 0.1𝑠)(1 + 0.04 𝑠)

Poles: 𝑠 = −10, −25 Zeros: 𝑠 = +50

𝑌(𝑠) 5 − 0.1 𝑠 1
𝐺 𝑠 = = ,𝑋 𝑠 =
𝑋(𝑠) (1 + 0.1𝑠)(1 + 0.04 𝑠)
𝑠
1 10 1 5
𝑌 𝑠 = −
𝑠 1 + 0.1𝑠 𝑠 1 + 0.04 𝑠
𝑡 𝑡
−0.1 −0.04
𝑦 𝑡 = 10(1 − 𝑒 ) − 5(1 − 𝑒 )
C. Determine the steady state response to a sinusoidal input
sin 2t.
𝟏𝟎 𝟓
𝑮 𝒔 = −
𝟏 + 𝟎. 𝟏𝒔 𝟏 + 𝟎. 𝟎𝟒 𝒔
𝑥 𝑡 = sin 2 𝑡 y 𝑡 = 𝐵1 sin 2 𝑡 + 𝜑1 + 𝐵2 sin(2 𝑡 + 𝜑2 )

𝐵1 𝐾 10
𝐴𝑅1 = = = = 9.8058
1 2
1+𝜔 𝑇 2 1 + 4 × 0.12

𝐵2 𝐾 5
𝐴𝑅2 = = = = 4.98407
1 2
1+𝜔 𝑇 2 1 + 4 × 0.04 2

𝜑1 = −𝑡𝑎𝑛−1 𝜔𝑇 = −𝑡𝑎𝑛−1 2 × 0.1 = −11.31° = −0.197 𝑟𝑎𝑑

𝜑2 = −𝑡𝑎𝑛−1 𝜔𝑇 = −𝑡𝑎𝑛−1 2 × 0.04 = −4.574° = −0.0798 𝑟𝑎𝑑

y 𝑡 = 9.8058 sin 2t − 0.197 − 4.98407 sin (2 t − 0.0798)

 Sheet 1, Problem 1
For each of the following transfer functions:
A. Identify the poles and zeros of the transfer function.
B. Plot the response to a unit step change.
𝟏𝟎 𝒆− 𝒔 𝟓
𝑮 𝒔 = − 𝟐 − 𝝉𝒅 𝒔 𝟏 − 𝟎. 𝟓 𝒔
𝟏 + 𝟎. 𝟏𝒔 𝟏 + 𝟎. 𝟎𝟒 𝒔 𝒆− 𝒔 = =
𝟐 + 𝝉𝒅 𝒔 𝟏 + 𝟎. 𝟓 𝒔
10 𝟏 − 𝟎. 𝟓 𝒔 5 −0.45𝑠 2 − 7.6 𝑠 + 5
𝐺 𝑠 = × − =
1 + 0.1𝑠 𝟏 + 𝟎. 𝟓 𝒔 1 + 0.04 𝑠 (1 + 0.1𝑠)(1 + 0.04 𝑠)(1 + 0.5𝑠)

Poles: 𝑠 = −10, −25, −2 Zeros: 𝑠 = −17.5, +0.63

1 10 𝒆− 𝒔 1 5
𝑌 𝑠 = −
𝑠 1 + 0.1𝑠 𝑠 1 + 0.04 𝑠
C. Determine the steady state response to a sinusoidal input
sin 2t.
𝟏𝟎 𝒆− 𝒔 𝟓
𝑮 𝒔 = −
𝟏 + 𝟎. 𝟏𝒔 𝟏 + 𝟎. 𝟎𝟒 𝒔
𝑥 𝑡 = sin 2 𝑡 y 𝑡 = 𝐵1 sin 2 𝑡 + 𝜑1 + 𝐵2 sin(2 𝑡 + 𝜑2 )

𝐵1 𝐾 10
𝐴𝑅1 = = ×1= = 9.8058
1 2
1+𝜔 𝑇 2 1 + 4 × 0.12

𝐵2 𝐾 5
𝐴𝑅2 = = = = 4.98407
1 2
1+𝜔 𝑇 2 1 + 4 × 0.04 2

𝜑1 = −𝑡𝑎𝑛−1 𝜔𝑇 − 𝜔𝜏𝑑 = −𝑡𝑎𝑛−1 2 × 0.1 − 2 = −2.197 𝑟𝑎𝑑

𝜑2 = −𝑡𝑎𝑛−1 𝜔𝑇 = −𝑡𝑎𝑛−1 2 × 0.04 = −4.574° = −0.0798 𝑟𝑎𝑑

y 𝑡 = 9.8058 sin 2t − 2.197 − 4.98407 sin (2 t − 0.0798)

1. Review of Elementary Control System Design
Control
Action:
Error Signal Controller Output p
e Controller
Input Signal

3-Mode Control Action:

1 de
p  K p e   edt  TD  po
TI dt
1 de
p  e  f  edt  r  po
b dt
Controller tuning involves adjusting the values of Kp,
TI, and TD to obtain a satisfactory dynamic response.
 D

CV

Sluggish
Response

Offset

Optimum Oscillatory
Response Response
The optimal values of Kp, TI, and TD depend on process
dynamic parameters. If they change during operation the
response is no longer optimum till the controller is re-
tuned.
Proportional Control
P
The output signal of a proportional E
controller regulates the position of the 10 KP
final control element in proportion to 1.0
the magnitude of the error. The
equation of action of proportional 0.1
1 10 100

Phase Shift
controllers is: Frequency

p  K p e  p o (1)
dp de 0

 Kp (2)
dt dt Frequency

Equation (2) shows that the output signal speed is zero when the
error is constant, i.e. with proportional control the system may
accommodate itself in a situation where equilibrium is established
between the error and the controller output signal.
 p  K p e  p o (1)
P
E
Laplace transforming equation (1)
10 KP
we get:
1.0

P  K pE 0.1
1 10 100

Phase Shift
Frequency

P
 K p  G c (3)
E 0

Frequency
It is clear from equation (3) that the frequency response
characteristics of proportional controllers are a constant
magnitude ratio equal to the gain (Kp) of the proportional
controller and a zero phase shift regardless of the input frequency.
The gain of the proportional controller is the reciprocal of the
“Proportional Band, b”.
Integral Control
The equation of action of an integral controller is:
1 dp 1
p   e dt  p o (4)  e (5)
TI dt TI
Laplace transforming equation (4) we get:
1
P E (6)
TIs
Floating speed or reset rate
P f 1 and its reciprocal TI is the
 Gc   (7) integral time constant.
E s TIs
It is clear from Eq. (5) that the integral control action provides a
continuous increase of correction as long as an error exists. This
means that there will be no offset.
The frequency response
characteristics of integral
controllers may be obtained from P
Eq. (7) by replacing (s) by (j). E
Slope = -1
10
P 1
 (7) 1.0
E TIs 0.1
j
G ( j)   (8)  1
Frequency
TI  0
TI

The magnitude ratio curve is a

straight line of slope –1 -90
corresponding to infinite amplitude
-180
ratio at zero frequency and to an Frequency
amplitude ratio of (1) at  = (1/TI).
The phase shift is independent of
frequency being constant at -90.
Proportional + Integral Control
The combination of the proportional and integral modes, retains
the advantages of eliminating offset by reset action and of
immediately responding to load changes through proportional
action. The equation of action is:
1
p  e  f  e dt  p o (9)
b
Laplace transforming equation (9) we get:
1 f 1
P  (  )E  (K p  )E (10)
b s TIs
The frequency response is thus given by:
1
G ( j)  K p (1  j ) (11)
K p TI 
 1
G ( j)  K p (1  j ) (11)
K p TI 
P
1 1
A.R   2 2 E Slope = -1
K TI (12)
 2 p
TI KP

1
1
  tan  (13)  1
K p TI  0
K p TI Frequency

High values of (KP) and (1/TI or f) -45

increase tendency to cycling. On the
other hand low values of (KP) and (1/TI -90
or f) give rise to sluggish response. Frequency
Proportional + Rate Control
The rate control action is expressed by:
de
p  r  p o (14)
dt
Derivative Time Constant

It is obvious that the controller is not active unless the error is

changing and has thus no effect in overcoming offset.

In order to illustrate the effect of addition of the derivative

mode to a proportional controller, consider a proportional
controller with a gain (KP), the equation of action of which is
expressed by:
p  p o  K p e (15)
If a derivative controller with a rate time constant of (r or TD) is
added to the above controller, the equation of action of the
combined controllers will be:
de
p  po  K pe  r (16)
dt
e 1 r de 1
 (1  ) (17)
p  po K p p  p o dt K p apparent
1
K p apparent  Kp (18)
r de
1
p  p o dt
Eq. (18) may be used to compare the behaviour of a (P + D)
controller with that of a (P) controller. If the error increases with
time, the (P + D) controller behaves apparently as a proportional
controller but with a larger gain than the original proportional
controller and vice versa.
 de
p  po  K pe  r (16)
dt
𝑃 = 𝐾𝑝 𝐸 + 𝑟𝑠𝐸 P
E
𝑃
= 𝐾𝑝 + 𝑟𝑠 Slope = +1
𝐸 KP

𝑃
= 𝐾𝑝2 + 𝑟 2 𝜔 2  KP
𝐸 r Frequency
90
−1
𝑟𝜔
𝜑 = 𝑡𝑎𝑛
𝐾𝑝 45

0
Frequency
The effect of rate mode is not visible until the error changes.
Since the derivative action is not effective in overcoming offset,
(P+D) control is usually combined with integral control to give
3-mode control.
Three Mode Control
The equation of action of a three-mode controller is:

1 de
p  e  f  edt r  p o (19)
b dt
Laplace transforming equation (9) we get:
P 1 f
G     rs
E b s
1 rs
G  K p (1   ) (20)
TI K ps K p
The frequency response is thus given by:
 r 1 
G( j)  K p (1  j  ) (21)
K K T  
 p p I 
P Slope = -1
E
At low frequencies the effect of
integral action dominates. At high
frequencies the rate mode is KP
Slope = +1
predominant and at some
intermediate frequency   1 / TI r  Frequency
the integral and derivative actions 90
cancel each other and the amplitude
ratio curve has a minimum point 0
corresponding to an amplitude ratio
of (KP) and a phase shift of (zero). -90
1 Kp
TI K p r
Lead Compensation
A disadvantage of the rate action is the extreme sensitivity at high
frequencies. This means that the controller will respond by
generating unduly large corrective signals in response to small
magnitude high frequency noise in the error. This drawback of
derivative action may be overcome by the introduction of a first
order lag in order to limit the sensitivity of the rate action at high
frequency. Such an arrangement is known as a lead compensator.
The transfer function of a lead compensator is:
sa
G Lead  (b  a )
sb
T2 1  T1s
G Lead  (T1  T2 ) (22)
T1 1  T2s
The factor (T2/T1) is less than unity, so additional amplification
must be provided to overcome the effect of this attenuation. In
practice, lead compensating networks are made so that the value
of (T /T ) is between (1/10) and (1/20).
The Bode diagram for various P
E
values of T2/T1 of the transfer 100
function Slope = +1
1  T1s
T2 = T1/

G Lead  10
T2 = T1/10

1  T2s T2 = T1/2
2
The upper curves corresponding to 1
(T2=0) are those of a (P+D)
controller with KP=1. The phase  Frequency
shift curves exhibit a maximum 90
phase lead of:
45
1 T1 / T2  T2 / T1
  tan (23)
0
2
1 2 10 20
and this occurs at a frequency of
TI TI TI TI
max  1 / T1T2
When a lead compensator is inserted in series with a (P+I)
controller, the resulting transfer function will have the form:
P 1 1  T1s
G   K p (1  ) (24)
E K p TIs 1  T2s
Exercise:
Draw the Bode plot for the above transfer function.
2.a. Write down the transfer function of a lead compensator
inserted in series with a (P+I) controller.
b. Sketch the Bode diagram of the system.
c. What is the advantage of this system compared to the
conventional 3-mode controller?
𝟏 𝟏 + 𝑻𝟏 𝒔
𝑮 𝒔 = 𝑲𝒑 (𝟏 + )
𝑲𝒑 𝑻𝑰 𝒔 𝟏 + 𝑻𝟐 𝒔
P
P 100 P
E Slope = +1 Slope = +1
Slope = -1 E E

10
KP
2
1 1
 1
Frequency
K p TI
 Frequency Frequenc
0
90 90

-45 0 0

-90 -90 -90

Frequency 𝟏 𝟏 𝟏 𝟏 𝟏
𝑻𝟏 𝑻𝟐 𝑲𝒑 𝑻𝟏 𝑻𝟏 𝑻𝟐
Lag Compensation
Another widely used control action is lag compensation. The
transfer function of a lag compensator is given by:

1  T2s
G Lag  (T1  T2 ) (25)
1  T1s
The minimum phase lag P
occurs at a frequency of: E
T2 = T1/2
min  1 / T1T2 1
0.5
T2 T2 = T1/10
T1
The lower curves 0.1
corresponding to T2 = 0 are T2 = 0
those of a first order system
having a time constant of T1. 0.01
Frequency

0
It is clear that the amplitude
ratio and phase angle T2 = T1/10
characteristics of a lag -45
compensator network are the T2 = 0
mirror image of those of a -90
lead compensator. 1 2 10 20
TI TI TI TI
Lag-Lead Compensation A.R.
A still more versatile control action 1
may be used to approximate more
elaborate control actions called for T2
by conditions to be discussed later. T1
It has the transfer function:
1  cT2s 1  T1s
G Lag Lead   Frequency
1  cT1s 1  T2s 90
Where T1>T2 and c>1 is a constant
so that cT1>cT2>T1>T2 0

The Bode diagram for a Lag-Lead -90

compensator is readily comparable 1 1 1 1
with the three-mode controller cTI cT2 T1 T2
frequency response diagram.
The electrical generation of the above three control actions is
illustrated in the circuits below.
Lead Lag Lag-Lead
Compensator Compensator Compensator
R1 R1
R1
+ + + + + +
R2 R2
i C o i o C1 o
R2 C C2
- - - - - -