CHE 456

CHE 456
Process Dynamics and
Control
Topic 8
Dynamic response of more
complex processes
Prof. Zoltan K. Nagy

Background:
Last time we saw
First order systems (time domain and s domain)
Second order systems (time domain and s domain)
Linear/locally linear systems

CHE 456

Lecture Overview
Review (background) Transfer functions
Poles and Zeros and process response
Integrating process
Processes with zeros (nominator
dynamics)
Time delay systems
Approximation of higher order systems
(process approximation)
Noninteracting versus interacting systems
3

Background
Review/look ahead
Transfer Functions - Properties
The TF enables us to determine the output
response to any input signal

G s

Y s
X s

TF determines intrinsic dynamic features

(response/natural modes)

Principle of Superposition

simple block
diagram algebra to combine simple linear
dynamic systems into complex linear systems.
Two important properties:

A. Multiplicative Rule

B. Additive Rule

CHE 456

Transfer function representations

Transfer Functions
Generic representations:
G s

Y s

bm s m

b0

U s

a0

an s

bm ( s z1 ) ( s zm )
an ( s p1 ) ( s pn )

Polynomial form

( 1s 1) (
( 1s 1) (

m
n

1)
1)

Gain-time constant form

Pole-zero form

Order of the TF = order of the denominator polynomial (n)

Poles response modes; zeros
weights of response modes
(found by partial fraction expansion)
Hint: You can use the Matlab function residue to perform the partial
fraction expansion
5

Poles and Zeros and their effect on

Process Response
G s

Imaginary
Part

bm ( s z1 ) ( s zm )
an ( s p1 ) ( s pn )

Natural modes/natural modes

determined by poles not zeros
(except pole-zero cancellation)

faster
response

unstable

Zero in RHP

inverse response
number of

output

input

No of zeros in RHP
inversions

time

Real
Part

Inverse
response
(zeros)

Pole in RHP

Unstable
(poles)

more oscillatory

1 zero
2 zeros
time

Is this
system
easy or
difficult to
control?
6

CHE 456

Poles of poles on Process Response

Negative real root

x
time

Positive real root

time

Complex roots (negative

real part)

x
time
x

Complex roots (positive

real parts)

x
time
7

Special case of first order process

integrating system
Remember tank example with pump at the outlet
(both qi and q are independent):
dh
A
dt

(deviation, Laplace etc.)

qi

1 #
qi
As

Ah # s

qi#

h#
qi#

1 #
q
As

q#

1/ A
s

and

Ah # s

qi#

q#

h#
q#

1/ A
s

Integrating systems: G(s )

h( t )

dh
h0

t0

1
[qi (t ) q(t )]dt
A

h(t ) h0
t0

1
[qi (t ) q(t )]dt
A

No steady-state gain (gain is the slope of the change)

Operating at steady state any small disturbance leads
to linear increase/decrease of output

K
s

pole

s=0

CHE 456

Special case of first order process

integrating system & pure gain system
K
s 1

G(s )

K/
s

Integrating
systems: G(s )
Pure gain:

G(s )

1.2

Example: response to unit step

for the following TF:

y(t)

0.8

Pure gain with K = 1

1
2s 1

G(s )

0.6
0.4

Integrating system with

slope = K/ = = 0.5

0.2
0
0

time

10
9

Poles and natural modes of response

K
s( 1s 1)( s

Example:

2 2
2

e-t/

constant

0
2

1)
2s

t
2

Function
with term:

sin( 1

2 t

cos( 1

2 t

t
2

Natural modes (plus some modes determined by the input (forcing function)
s1

integrating

s2

x s3

s2

s3 , s 4

j
2

s1

1
2

s4
10

CHE 456

Examples of more complex TFs

TFs with zeros
Transfer function discussed so far can be extended to represent
more complex process dynamics by simply adding numerator
dynamics.
E.g. some control systems contain lead-lag element

dy
dt

dy
dt

K (u

K u

du
)
dt

G( s )

K ( a s 1)
s 1

u(t )dt

G( s )

K ( as 1)
1)
a s( s

a 0

Generally m < n
physically realizable systems
If m = n the output response in discontinuous after a step input
change (most real system exhibit some level of inertia)
11

First order process with numerator

dynamics (m = n)
Calculate the step response of a lead-lag element to a step change of
magnitude M in the input (u# = M/s)

K ( as 1)
s 1

y (t )

MK 1 (1

y#
a

)e

MK ( as 1)
s( s 1)

t/

if < a
y jumps to new value and
then decays exponentially to MK
if > a
y jumps to new value and
then increases exponentially to MK
if a < 0
y jumps to negative value and
then increases exponentially to MK (inverse

response)
if

= a
pole-zero cancellation
G(s)=K
pure gain (step change in output)

y#

MK

1
s

s 1

if t = 0
y(0) = MK a /
y changes abruptly
New steady state y( ) = MK

y(t)/(MK)

G( s )

1
0
time
12

CHE 456

Second order process with numerator

dynamics (m = n)
Single zero and overdamped second order TF:

G s

1s

as

1
2s

The response of this system to a step change in input is:

AK

y#

as

1s 1

1
2s 1

y t

KM 1

a
1

1 e t/
2

a
2

2 e t/
1

if t = 0
y(0) = 0
no abrupt change
New steady state y( ) = MK
Including single zero does not change initial or final value
Response is the sum of two exponential modes
Zero does not affect basic characteristics of the process
(initial & final value & stability) but affects how the natural
modes are weighted in the response
13

Second order process with numerator

dynamics
G s

K
1s 1

as

1
2s 1

y t

KM 1

a
1

1 e t/
2

a
1

2 e t/
2

Adding zero to a second order

overdamped system may yield
overshoot and inverse response
Inverse response is observed
when the zeros lie in the right
half complex plane, Re(z)>0
Overshoot is observed when
the zero is dominant a > 1
Pole-zero cancelation leads to
first order process behaviour
In physical systems, overshoot and inverse response are a result of
two process with different time constants, acting in opposite directions
14

CHE 456

Second order process with numerator

dynamics
Two first order process elements acting in parallel (in
opposite directions)
K1
1s 1
U(s)

Y(s)

K2
2s 1

G ( s)

( a s 1)
( 1s 1)( 2 s 1)

K1 K2

K1 2 K2 1
K1 K2

If gains are of opposite signs and time constants are

different then a right half plane zero occurs
15

Systems with time delay

Time delays (dead time) occur due to:

Chapter 6

1. Fluid flow in a pipe

2. Transport of solid material (e.g., conveyor belt)
3. Chemical analysis
-

Sampling line delay

Time required to do the analysis (e.g., on-line gas

chromatograph)

Mathematical description:
A time delay, , between an input u and an output y results in
the following expression:
y t

0
u t

for t
for t
16

CHE 456

Systems with time delay

Fi
Control loop

Time required for the fluid to

reach the valve usually
approximated as dead time
Manipulation of valve does not lead to immediate change in level
17

Systems with time delay

Input (u)

Output (y)

Typical example: plug flow in a pipe

= time delay (dead time)

Length of pipe
fluid velocity

volume of pipe
volumetric flowrate

y(t )

Ku(t

G (s )

y(s )
u(s )

Ke

Note that time delay can be accounted for if velocity (flowrate) is

constant and known (time delay compensation we will see later)

18

CHE 456

Delayed transfer functions

U(s)

Y(s)

G ( s)

e ds

Y ( s) e d sG ( s)U ( s)
Example first order plus time
delay (FOPTD)
1

0.9
0.8
0.7

FOPTD response
to a unit step
change at t = 0

0.6

s 1

y/KM

G ( s)

e d sK p

Second order plus time delay

0.5
0.4
0.3
0.2

G ( s)

2 s2

ds

KP

0.1
0

2 s 1

0.5

1.5

2.5
3
t/tau

3.5

4.5

Time delay

5.5

19

Approximation of time delay

e d sK p

G ( s)

Problem

s 1

use of the dead time makes analysis (poles & zeros) more difficult

Approximate dead-time by a rational (polynomial) function

Taylor series expansion (generally truncated after first few terms):

2 2

3 3

s
2!

s
3!

4 4

s
4!

Most common is Pade approximation:

1
G1(s)

2
2

s
e

First order Pade approximation

1
G2 (s )
1

2
2

12
2

12

s2
s2

Second order Pade approximation

20

10

CHE 456

Approximation of time delay

Pure time delay

In general Pade approximations

do not approximate dead-time
very well
Pade approximations are better
when one approximates a first
order plus dead time process
(in particular when << )

FOPTD

G ( s)

e sK p
s 1

s K
p
s s 1

Pade approximations introduce

inverse response (right half
plane zeros) in the transfer
function
21

Process approximation
Higher order TFs can often be approximated by lower order TFs plus
time delay which have similar dynamic and steady-state characteristics
FOPTD model is often used to approximate complex higher order
(overdamped) processes
1.2

Real Process

0.8

0.6

- FOPTD
o Second Order

-- First order plus dead time

- Second order overdamped
o Actual process

0.4

0.2

-0.2

Second order overdamped or first order plus dead time?

Second order process model may be more difficult to
identify
22

11

CHE 456

Process approximation
Higher order TFs can often be approximated by lower order TFs plus
time delay which have similar dynamic and steady-state characteristics
Taylor series approximation (for small values of s)
s

RHP zero at s =

An alternative first-order approximation consists of the transfer function

1
e

1
1

These were derived to approximate time-delay terms.

However, these expressions can also be used to approximate the pole or
zero term on the right-hand side of the equation by the time-delay term
on the left side.
23

Process approximation
Higher order process (example N tanks in series)

U(s)

K P1K P 2 K PN
( 1s 1)( 2 s 1) ( N s 1)

For two dominant time constants

G ( s)

e sK p
( 1s 1)( 2 s 1)

e sKp
( 1s 1)

and

>

i 3
1

>

>>

process well approximated by

For one dominant time constant

G ( s)

Y(s)

1
1

process well approximated by

N
i 2

i
24

12

CHE 456

Process approximation - Example

1

G ( s)

(10s 1)(25s 1)( s 1) 2

1.2

0.8

e 12 s
25s 1
e 2s
G2 ( s)
(10s 1)(25s 1)

G1( s)

0.6

0.4

0.2

Model identification is to
estimate parameters (e.g.
Kp, and ) so that model
describes experimental
data
Least squares estimation
Process reaction curve
method

-0.2

20

40

60

80

100

120

140

160

180

200

25

Process approximation
Skogestads half rule
Skogestad (2002) has proposed a related approximation
method for higher-order models that contain multiple time
constants.
He approximates the largest neglected time constant in the
following manner.
One half of its value is added to the existing time delay (if
any) and the other half is added to the smallest retained
time constant.
Time constants that are smaller than the largest neglected
time constant are approximated as time delays.

26

13

CHE 456

Example
Consider a transfer function:
0.1s 1

G s

5s 1 3s 1 0.5s 1

Derive an approximate first-order-plus-time-delay model,

Ke s
s 1
using two methods:
G s

(a) The Taylor series expansions

(b) Skogestads half rule
Compare the normalized responses of G(s) and the approximate
models for a unit step input.
27

Solution
K

G s

0.1s 1

5s 1 3s 1 0.5s 1

(a) The dominant time constant (5) is retained. Applying

the approximations :
0.1s 1 e

1
e
3s 1

0.1s

3s

1
e
0.5s 1

0.5 s

Substitution into the TF gives the Taylor series approximation:

GTS s

Ke

0.1s

e 3s e
5s 1

0.5 s

Ke 3.6 s
5s 1
28

14

CHE 456

Solution
G s

0.1s 1

5s 1 3s 1 0.5s 1

(b) To use Skogestads method, we note that the largest neglected

time constant has a value of 3.
According to his half rule, half of this value is added to the
next largest time constant to generate a new time constant
5 0.5(3) 6.5.
The other half provides a new time delay of 0.5(3) = 1.5.
The approximation of the RHP zero provides an additional time
delay of 0.1.
Approximating the smallest time constant of 0.5 produces an
additional time delay of 0.5.
Thus the total time delay is,
Ke 2.1s
GSk s
6.5s 1
1.5 0.1 0.5 2.1
29

Solution
G s

0.1s 1

5s 1 3s 1 0.5s 1

GSk s

Ke 2.1s
6.5s 1

GTS s

Ke 3.6 s
5s 1

30

15

CHE 456

Interacting vs. Noninteracting Systems

Noninteracting: changes in downstream unit have no
effect on upstream units
Interacting: downstream units affect upstream units
and vice versa (often processes with recycle)
In general, transfer functions for interacting processes
are more complicated than those for noninteracting
processes.
As an example, we will consider the two liquid-level
storage systems

31

Interacting vs. Noninteracting Systems

A noninteracting system:
two surge tanks in series.

Interacting: Two tanks in series whose liquid levels interact.

32

16

CHE 456

Noninteracting Systems
1
h1
R1

q1

A1

dh1
dt

qi

H1 s
Qi s
Q1 s
H1 s

H2 s
Q2 s

R2
A2 R2 s 1

K2
2s 1

q1

A1

dh1
dt

R1
A1R1s 1
1
R1

Q2 s
H2 s

qi

1
h1
R1

K1
1s 1

1
K1

1
R2

1
K2
33

Noninteracting Systems
Q2 s

Q2 s H 2 s Q1 s H1 s

Qi s

H 2 s Q1 s H1 s Qi s

Q2 s
Qi s

1
K2

K 2 1 K1
2 s 1 K1 1s 1

Q2 s
Qi s

1
1s

Steady state gain is 1. Why?

2s

34

17

CHE 456

Dynamic Model of An Interacting Process

Two tanks in series whose liquid levels interact.

Flow from tank 1 to tank 2 is proportional with the pressure
difference (hydrostatic pressures in the two tanks)
1
h1 h2
R1

q1

35

Dynamic Model of an Interacting Process

The transfer functions for the interacting system are:

where

1 2

H1 s

K1

as

Qi s

2 2

2 s 1

R2 A1

, and

R1R2 A2 / R1

R2

1 2

It is possible to show that >1 by analyzing the denominator hence, the

transfer function is overdamped, second order, and has a negative zero.
H2 s
H1 s

R2
R1 R2
R1R2 A2
s 1
R1 R2

H2 s
Qi s

R2
2 2

2 s 1
36

18

CHE 456

Model Comparison
Noninteracting system

Q2 s

Qi s
where

Interacting system

1s

A1 R1 and

Q2 s

2s

A2 R 2 .

Qi s

2 2

where

1 and

2 s 1
1 2

General Conclusions:
1. The interacting system has a slower response.
(Example: consider the special case where = 1=
2. Which two-tank system provides the best damping
of inlet flow disturbances?

2.)

37

Summary
Dynamic response can be analyzed by looking at the poles
and zeros of the TF
Poles determine natural modes, zeros weights between them
Poles - stability, oscillations
zeros - inverse response, overshoot for overdamped system

Special cases of first order system integrating and pure gain

Processes with zeros (nominator dynamics)
Time delay systems - very important for control (main reasons
for instability)
Approximation of higher order systems (process
approximation) often with time delay
stability margin
Noninteracting versus interacting systems
38

19