Chapter Four

The Transfer function

 Disturbances
• Disturbances can be classified and defined in several w
ays:
1- Unit step function:

0 t  0
u (t ) = 
1 t  0

The response of a system to a step disturbance is called

step response or the transient response.
 2- Pulse. A pulse function occurs when the forcing variable
changes from its initial condition at time zero and then at
subsequent time returns to initial condition.

3- Impulse. The impulse is

defined as the delta function,
an infinitely high pulse whose
width is zero and whose area
is unity.

 t =0 
 (t ) = 
0 t0
, 0
 (t )dt = 1
 4- Ramp: Ramp inputs are functions that change
linearly with time.

5- Sinusoidal: Pure periodic sine or cosine inputs seldom

occur in real chemical engineering systems. The response
of systems to this kind of forcing function called the
frequency response of the system.
 The Transfer function
• What is Transfer Function (TF)?
an algebraic expression for the dynamic relation between the input
and output of the process model.
It is defined so as to be independent of the initial conditions and of the
particular choice of forcing function.

– Properties; It can be derived only for a linear differential equation

model because Laplace transform can be applied only to linear
equations. If model is non-linear, then it must be linearized first.
– Advantage; It is easy to interpret and use in calculating output responses
for particular input changes.
▪ Let G(s) denote the transfer function between an input, u,
and an output, y. Then, by definition:
 Deviation Variables
It used to remove dependence on initial condition
 4.1 Development of Transfer Functions
• Example ; Stirred-tank heating system

Figure 4.1. Continuous stirred-tank heater.

– Energy balance equation.

dT
VC = wC (Ti − T ) + Q (4.1)
dt
– Assumption ; constant liquid holdup and constant inflow(w is
constant), a linear model result.
– If the process is at steady-state, dT dt = 0 .
T (0) = T , Ti (0) = Ti , Q(0) = Q ( 4 .2 )

0 = wC (Ti − T ) + Q ( 4.3)
– Subtract (4-3) from (4.1).
dT d (T − T )
VC = VC = wC[(Ti − Ti ) − (T − T )] + (Q − Q) (4.4)
dt dt
– Define some important new variables(Deviation variables).
T ' = T − T , Ti ' = Ti − Ti , Q' = Q − Q (4.5)

T ' (0) = T (0) − T = 0, Ti ' (0) = Ti (0) − Ti = 0, Q' (0) = Q(0) − Q = 0 (4.6)

By substituting deviation variables for variables, the transfer

functions are not related to initial conditions.
 dT '
VC = wC (T 'i −T ' ) + Q' (4.7)
dt
– Let k = 1 wC , = V w .
dT '
 = (T 'i −T ' ) + kQ' (4.8)
dt
– Apply Laplace Transform.
sT ' ( s ) = (T 'i ( s ) − T ' ( s )) + kQ' ( s ) (4.9)
(s + 1)T ' ( s ) = T 'i ( s ) + kQ' ( s ) (4.10)
1 k
T ' (s) = T 'i ( s ) + Q' ( s) (4.11)
s + 1 s + 1
– If Ti ' ( s ) = 0 .
T ' (s) k
G1 ( s) = = (4.12)
Q' ( s) s + 1
– If Q ' ( s ) = 0 .
T ' (s) 1
G2 ( s) = = (4.13)
T 'i ( s) s + 1
 1 k
T ' (s) = T 'i ( s ) + Q' ( s) (4.11)
s + 1 s + 1
Let T(s)= Y(s),,,, Ti= Y(s),,,,,,, Q(s)= U(s)
 4.2 Property of Transfer Functions (TF)

1. The steady-state output change for a sustained change in input can be

calculated directly; we can easily obtain steady state gain K. - Final
value theorem.
Steady State gain; the ratio of the output variable change to an input variable
change when the input is adjusted to a new value and held there, thus allowing
the process to reach a new steady state.
y2 − y1
K= (4.14)
x2 − x1 example:

-For a unit step change in input, the gain is the change in output - From the final
value theorem, unit step change in input with zero initial condition gives

Example. for the previous example (slide 9) find the static gain with unit step change
in input:
 4.2 Property of Transfer Functions (TF)

- Some TF models do not have steady-state gain as for integrating

processes and processes with sustaining oscillation in output.

2- The order of the transfer function(TF): The transfer function

has the order of the denominator polynomial (in s) as same as the
order of the equivalent differential equation.
General n-th order differential equation.
dny d n −1 y dy
an n + an −1 n −1 +  + a1 + a0 y
dt dt dt
d mx d m −1 x dx
= bm m + bm −1 m −1 +  + b1 + b0 x (4.15)
dt dt dt
Apply Laplace transform to (4.15).
m

Y ( s) i
b s i

bm s m + bm −1 s m −1 +  + b1 s + b0
G ( s) = = i =0
= (4.16)
n
an s n + an −1 s n −1 + + a1 s + a0
a s
X ( s) i
i
i =0
For step input function, we can obtain the gain K.

b0
lim y (t ) = lim sY ( s) = lim s  G ( s)  X ( s) = lim G ( s) = G (0) = (4.17)
t → s →0 s →0 s →0 a0
3- Physical Realizability. ( n  m )
- if n  m, n = 0, m = 1.
Y ( s) b1s + b0 dx
= , a0 y = b1 + b0 x (4.18)
X ( s) a0 xt
for step input change,( X ( s ) = 1 s ), dx dt =  (t ) which has infinite size,
y should be infinite size at t=0. Therefore, assumption is contradictive
(unrealizable), and n  m .
Note that the step change here is an impulse at t=0, and the ability to
respond infinitely fast to a sudden change in input is impossible to
achieve with real (physical) process
3. Additive property of the transfer function in parallel processes.
X 3 ( s ) = G1 ( s ) X 1 ( s ) + G2 ( s ) X 2 ( s ) (4.19)

Figure 4.2. Block diagram of additive transfer function model.

4. Multiplicative property of the transfer function in series
processes. X 3 ( s ) = G2 ( s ) X 2 ( s ) = G2 ( s )G1 ( s ) X 1 ( s ) (4.20)

Figure 4.3. Block diagram of Multiplicative(series) transfer function model.

 • Example; Two liquid surge tanks in series.

Assumption; The outflow from each tank

is linearly related to the height of the
liquid(head) in that tank. The two tanks
have different cross-sectional areas A1 and
A,2 and that the outflow valve resistance are
fixed at R1 and R 2 .

Find the transfer function relating changes in

outflow from the second tank, Q2 ' ( s,) to
changes in inflow to the first tank, Qi ' ( s ) .
Show how this transfer function is related to
Figure 4.4. Schematic diagram of two the individual transfer functions, H 1 ' ( s ) Qi ' ( s )
liquid surge tanks in series. , Q1 ' ( s ) H 1 ' ( s ), H 2 ' ( s ) Q1 ' ( s )and Q2 ' H 2 ' .
H 1 ' ( s ) and H 2 ' ( s ) denote the deviations in
Tank 1 and Tank 2 levels, respectively.
• For tank 1.
dh 1 1
A1 1 = qi − q1 = qi − h1 (4.21) (  q1 = h1 )
dt R1 R1
– Putting (4-21) into deviation variable form gives
dh1 ' 1
A1 = qi '− h1 ' (4.22)
dt R1
– Apply Laplace transform and rearrange to obtain transfer function.
H1 ' ( s) R1 K1
= = (4.23)
Qi ' ( s) A1 R1 s + 1  1 s + 1
– Similarly, other transfer functions are obtained as follows.
Q1 ' ( s) 1 1
= = (4.24)
H 1 ' ( s ) R K1
H 2 ' ( s) R2 K2
= = (4.25)
Q1 ' ( s) A2 R2 s + 1  2 s + 1
Q2 ' ( s) 1 1
= = (4.26)
H 2 ' ( s) R2 K 2
• The desired transfer function relating the outflow from Tank2 to
the inflow to Tank 1 can be derived by forming the product of
(4-23) through (4-26).
Q2 ' ( s) Q2 ' ( s) H 2 ' ( s) Q1 ' ( s) H1 ' ( s)
=
Qi ' ( s) H 2 ' ( s) Q1 ' ( s) H1 ' ( s) Qi ' ( s)
1 K 2 1 K1
= (4.27)
K 2  2 s + 1 K1  1 s + 1
Q2 ' ( s) 1
= (4.28)
Qi ' ( s) ( 1s + 1)( 2 s + 1)

Figure 4.5. Input-output model for two liquid surge tanks in series.
Common Transfer Functions
 4.3 Linearization of Non-linear Models
• Necessity
There is a wide variety of processes for which the dynamic
behavior on the process variables in a non-linear fashion.
Laplace transform cannot be applied!
• How? Taylor series expansion.
For dy dt = f ( y, x) around the nominal steady-state operating
point [ y , x] . Where y is the output and x is input.
dy f f
= f ( y, x)  f ( y, x) + ( y − y) + ( x − x) (4.29)
dt y y,x
x y,x

At steady state, f ( y, x) = 0 .
Put (4-29) into deviation variable form.
dy' f f
= y '+ x' (4.30)
dt y y,x
x y,x
•Example; A liquid-level storage system.
Assume that the valve discharge rate is related to the square root of
liquid level; q = Cv h.
Derive an approximate dynamic model for this process.
– Material balance.
dh
A = qi − q = qi − Cv h (4.31)
dt
– Deviation variables.
h' = h − h, qi ' = qi − qi (4.32)
– Linearize about the steady-state
conditions (h, qi .) Figure 4.6. A liquid-level storage
dh' C system.
A = qi '− v h' (4.33)
dt 2 h
• Useful results of Taylor series expansion.
(1) (1 + x) n = 1 + nx , where n is real.
Ex) 1
=1− x (4.34)
1+ x
1
=1+ x (4.35)
1− x
1
(1 + x)
1/ 2
=1+ x (4.36)
2

(2) ln(1 + x) = x − 1 x 2 +  (4.37)

2
1
(3) exp( x) = 1 + x + x 2 +  (4.38)
2
1 3
(4) sin( x) = x − x + (4.39)
3!
• Problem 4.2, 4.5,4.7, 4.12