Process Control: Chapter 4 – Transfer Functions

4.0 Introduction:

- In chapter 3 we discussed how Laplace transform techniques could be used to determine the
response from different ODEs.
- However, the full solving procedure should be repeated for every model, even when changing
only the initial conditions or the type of forcing function.
- Transfer function: Algebraic expression for the dynamic relation between a selected input and
output for the process model.
- It’s independent on the: initial condition and type of forcing function.
- The properties of the dynamic system become apparent when the transfer function is written in
its standard form.
- TFs are derived only from linear ODEs; therefore, the non-linear ODEs have to be linearized first.

4.1 Development of Transfer Functions:

The following model describes a simple blending system, that blends two streams, one with pure A
(𝑥𝐴,2 = 1), and the other is diluted (𝑥𝐴,1 << 1), with a stream that leaves the tank with a small
concentration of A (𝑥𝐴 << 1).
𝑑𝑥 𝑤1 𝑤2
= (𝑥1 − 𝑥) + (𝑥 − 𝑥) (4.1)
𝑑𝑡 𝑉𝜌 𝑉𝜌 2
Assuming a constant flow for the first inlet stream: 𝑤1 = 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 = 𝑤
̅1

Assuming that the process at the beginning was at steady-state:

𝑤
̅1 𝑤
̅2
0= (𝑥̅1 − 𝑥̅ ) + (𝑥̅ − 𝑥̅ ) (4.2)
𝑉𝜌 𝑉𝜌 2
Simplifying and applying assumptions:

0=𝑤
̅1 𝑥̅1 − 𝑤
̅1 𝑥̅ + 𝑤
̅ 2 𝑥̅2 − 𝑤
̅ 2 𝑥̅ = 𝑤
̅1 𝑥̅1 − 𝒘 ̅−𝒘
̅ 𝟏𝒙 ̅+𝑤
̅ 𝟐𝒙 ̅2
0=𝑤 ̅̅
̅1 𝑥̅1 − 𝒘 𝒙+𝑤
̅2 (4.3)
Since 𝑥1 and 𝑥 are very small, and 𝑤2 is pure A, we would need small amount of mass flow rate of 𝑤2 ;
therefore, 𝑤
̅2 ≪ 𝑤̅, hence:

𝑤
̅ ≈𝑤
̅1 = 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 (4.4)
Therefore, for unsteady-state process:
𝑑𝑥
𝑉𝜌 =𝑤
̅𝑥1 + 𝑤2 − 𝑤
̅𝑥 (4.5)
𝑑𝑡
To get rid of the dependence of the process model on the first steady-state, subtract the steady-state
model from the unsteady-state model (4.5) - (4.3):
𝑑(𝑥 − 𝑥̅ )
𝑉𝜌 ̅(𝑥1 − 𝑥̅1 ) + (𝑤2 − 𝑤
=𝑤 ̅2 ) − 𝑤
̅(𝑥 − 𝑥̅ ) (4.6)
𝑑𝑡
Divide by 𝑤
̅ on both sides:
𝑉𝜌 𝑑(𝑥 − 𝑥̅ ) (𝑤2 − 𝑤
̅2 )
= (𝑥1 − 𝑥̅1 ) + − (𝑥 − 𝑥̅ ) (4.7)
𝑤̅ 𝑑𝑡 𝑤
̅
𝑉𝜌
Since, ̅
is a constant value, assume that:
𝑤

𝑉𝜌
𝜏= (4.8)
𝑤̅
And:
1
𝐾= (4.9)
𝑤
̅
And introducing new concept called deviation variable:

𝑥 ′ = 𝑥 − 𝑥̅
𝑥1′ = 𝑥1 − 𝑥̅1
Etc.

Therefore, equation 4.7 becomes:

𝑑𝑥′
𝜏 = 𝑥1′ + 𝐾𝑤2′ − 𝑥 ′ (4.10)
𝑑𝑡
Small 𝜏 → Faster response

Large 𝜏 → Slower response

K is process gain.

In case of steady-state, (4.10) becomes:

𝑥 ′ = 𝑥1′ + 𝐾𝑤2′ (4.11)

Therefore, if 𝑥1 was constant and has a value that is similar to the 𝑥1 at the first steady-state (𝑥1′ = 0),
(4.11) will be as follows:

𝑥 ′ = 𝐾𝑤2′ (4.12)
Therefore, any deviation of 𝑤2 from steady-state will lead to a deviation in 𝑥 K times larger than the
deviation in 𝑤2 .

However, since the process gain of 𝑥1 is 1; therefore, any deviation of 𝑥1 will cause 𝑥 to deviate with
same factor.

4.1.1 Transfer Functions

Apply Laplace on transform on (4.10) knowing that 𝑥 ′ ,𝑥1′ , and 𝑤2′ are functions of time:

𝑑𝑥 ′
ℒ (𝜏 ) = ℒ(𝑥1′ ) + ℒ(𝐾𝑤2′ ) − ℒ(𝑥 ′ )
𝑑𝑡
 𝜏𝑠𝑋 ′ (𝑠) − 𝑥 ′ (0) = 𝑋1′ (𝑠) + 𝐾𝑊2′ (𝑠) − 𝑋 ′ (𝑠) (4.13)

Since the system was initially at its first steady-state; therefore, 𝑥 ′ (0) = 𝑥̅ − 𝑥̅ = 0:

𝜏𝑠𝑋 ′ (𝑠) = 𝑋1′ (𝑠) + 𝐾𝑊2′ (𝑠) − 𝑋 ′ (𝑠) (4.14)

Rearranging,

(𝜏𝑠 + 1)𝑋 ′ (𝑠) = 𝑋1′ (𝑠) + 𝐾𝑊2′ (𝑠)

1 𝐾
𝑋 ′ (𝑠) = ( ) 𝑋1′ (𝑠) + ( ) 𝑊 ′ (𝑠)
𝜏𝑠 + 1 𝜏𝑠 + 1 2
𝑋 ′ (𝑠) = 𝐺1 (𝑠)𝑋1′ (𝑠) + 𝐺2 (𝑠)𝑊2′ (𝑠) (4.15)
Therefore, the transfer functions are:
1 𝐾
𝐺1 (𝑠) = , 𝐺2 (𝑠) =
𝜏𝑠 + 1 𝜏𝑠 + 1
Hence, in the following cases:
𝑋 ′ (𝑠) 1
𝐺1 (𝑠) = ′ (𝑠) = , 𝑖𝑓 𝑎𝑛𝑑 𝑜𝑛𝑙𝑦 𝑖𝑓 𝑊2′ (𝑠) = 0
𝑋1 𝜏𝑠 + 1

Similarly,
𝑋 ′ (𝑠) 𝐾
𝐺2 (𝑠) = ′ (𝑠) = , 𝑖𝑓 𝑎𝑛𝑑 𝑜𝑛𝑙𝑦 𝑖𝑓 𝑋1′ (𝑠) = 0
𝑊2 𝜏𝑠 + 1

Notice that the transfer functions 𝐺1 (𝑠) and 𝐺2 (𝑠) do NOT depend on the initial conditions. They’re only
dependent on the operating conditions.

How to use equation 4.15?

Assume a process where 𝑥1 is not changing and it is kept as its initial steady-state value; therefore,
𝑥 ′ (𝑡) = 0. While the mass flow rate of pure A changed from 𝑤 ̅ 2 + 𝑀; therefore, 𝑤2′ (𝑡) = 𝑀:
̅ 2 to 𝑤

Hence,
𝑀
ℒ(𝑤2′ (𝑡)) =
𝑠
Substituting in equation 4.15:
𝐾𝑀
𝑋 ′ (𝑠) =
𝜏𝑠 2 + 𝑠

𝐾𝑀 𝐾𝑀 𝜏𝐾𝑀 1 𝜏 1 1
= − = 𝐾𝑀 ( − ) = 𝐾𝑀 ( − ) = 𝐾𝑀(1 − 𝑒 −𝑡/𝜏 )
𝑠(𝜏𝑠 + 1) 𝑠 𝜏𝑠 + 1 𝑠 𝜏𝑠 + 1 𝑠 𝑠+1
𝜏
𝑡
𝑥 ′ (𝑡) = 𝐾𝑀 (1 − 𝑒 −𝜏 ) (4.16)
*it’s mentioned that transfer functions do not depend on initial conditions, however, G1(s) contains w (bar) and G2(s) contains w2(bar) which
are the flow rates of the initial steady-state conditions.

From equation 4.16:

0
𝑥 ′ (𝑡 = 0) = 𝐾𝑀 (1 − 𝑒 −𝜏 ) = 𝐾𝑀(1 − 1) = 0

And,
𝑡
lim (𝑥 ′ (𝑡)) = lim (𝐾𝑀 (1 − 𝑒 −𝜏 )) = 𝐾𝑀(1 − 0) = 𝐾𝑀
𝑡→∞ 𝑛→∞

Approximate plot:

x'(t) vs. t
1.2

0.8

0.6

0.4

0.2

0
0 2 4 6 8 10 12