15

Lecture 4 – Dynamic Models - Review

Motivation
• Improve understanding of the process =⇒ steady-state and transient
conditions.

• Train plant operating personnel =⇒ run complex units and deal

with emergencies.

• Develop a control strategy for a new process =⇒ discovering con-

trolled/manipulated variables, controller tuning.

• Optimize operating conditions =⇒ maximize profit.

Models Classification

Model Type Description Advantages Disadvantages

Theoretical mod- Based on first princi- Provide physical in- expensive and time-
els ples of physics sight, applicable over consuming to develop,
wide ranges of condi- complex processes can
tions include parameters
that are not readily
available
Empirical models Based on fitting ex- Easier to develop and Typically do not ex-
perimental data to use trapolate well
Semi-empirical Based on theoretical incorporate theoretical
models models, some pa- knowledge, applies
rameters calculated over a wider range of
from experimental operating conditions,
data require less develop-
ment effort

A Simple Example
We will start with the general material balance equation:
( ) ( ) ( ) S1 S2
rate of mass rate of rate of x1 x2
= −
accumulation mass in mass out w1 w2

Applying to total mass:

d(ρV )
= w1 + w2 − w (2)
dt x
Applying to component A:3 w
3
Assuming well-mixed conditions (low
d(ρVx ) viscosity, adequate agitation)
= x1 w1 + x2 w2 − xw (3)
dt
Solving this model gives us connections between the different vari-
ables: x vs. V, x vs. (x1 , x2 ), and so on.

ChE 551: Process Control Dr. Yazan Hussain

16

A Simple Example: Steady state

At steady state, there is no accumulation:

0 = w̄1 + w̄2 − w̄

0 = x̄1 w̄1 + x̄2 w̄2 − x̄ w̄

The bar indicates nominal steady-state values.

x̄

x
Some Principles
All models are wrong, but some are useful.

Approach for Developing Dynamic Models

• State the modeling objectives and the end use of the model. Then
determine the required levels of model detail and model accuracy.

• Draw a schematic diagram of the process and label all process

variables.

• List all of the assumptions involved in developing the model. Try

to be parsimonious: the model should be no more complicated
than necessary to meet the modeling objectives.

• Determine whether spatial variations of process variables are im-

portant. If so, a partial differential equation model will be re-
quired.

• Write appropriate conservation equations (mass, component, en-

ergy, and so forth).

• Introduce equilibrium relations and other algebraic equations (from

thermodynamics, transport phenomena, chemical kinetics, equip-
ment geometry, etc.).

• Perform a degrees of freedom analysis to ensure that the model equa-

tions can be solved.

• Simplify the model. It is often possible to arrange the equations so

that the output variables appear on the left side and the input
variables appear on the right side. This model form is convenient
for computer simulation and subsequent analysis.

• Classify inputs as disturbance variables or as manipulated vari-

ables.

Conversation Laws: Mass

( ) ( ) ( )
rate of mass rate of rate of
= − (4)
accumulation mass in mass out

ChE 551: Process Control Dr. Yazan Hussain

17

Conversation Laws: Component i

 
( ) ( ) ( )  rate of 
rate of component i rate of rate of  
= − + component i
accumulation component i in component i out 
 produced 


(5)

Conversation Laws: Energy

( ) ( ) ( )
rate of energy rate of energy in rate of energy out
= −
accumulation by convection by convection
   

 net rate of heat 
   net rate of work 

+ addition to the system + performed on the system
   
 from the surroundings   by the surroundings 

(6)

The total energy is given by:

Utot = Uint + UKE + UPE (7)

For typical chemical processes, the kinetic energy, potential energy,

and net work terms are small and can be neglected. This leads to:

dUint
= −∆ w Ĥ + Q


(8)
dt

Mixing Tank Example Revisited

Looking back at the example in slide 3, we can simplify the solution
by making the following assumptions:

• The tank is well-mixed.

• The density of the liquid is constant.

Using these two assumptions: The total balance becomes:

dV 1
= ( w1 + w2 − w ) (9)
dt ρ

And the component balance becomes:4 4

Try to derive this relation

dx w w
= 1 ( x1 − x ) + 2 ( x2 − x ) (10)
dt Vρ Vρ

Example Blending Tank

If V is 2 m3 , ρ is constant at 900 kg · m−3 :

ChE 551: Process Control Dr. Yazan Hussain

18

a) Find the steady state value of x if w1 = 500 kg · min−1 ,

w2 = 200 kg · min−1 , x1 = 0.4 and x2 = 0.75.
Using the steady state form of Eq. 9:
w̄1 w̄
0= ( x̄ − x̄ ) + 2 ( x̄2 − x̄ )
Vρ 1 Vρ

0 = 500 × (0.4 − x̄ ) + 200 × (.75 − x̄ ) −→ x̄ = 0.5

b) If w1 changes suddenly to from 500 to 400 kg · min−1 ,

determine x (t) (composition as a function of time)?
Using Eq. 9:

dx w w
= 1 ( x1 − x ) + 2 ( x2 − x )
dt Vρ Vρ
400 200
= (0.4 − x ) + (0.75 − x )
2 × 900 2 × 900
dx
3 = 3.1 − x
dt
This is a first-order linear ODE. The initial condition
is given in the previous part (x̄ = 0.5). Solving the
equation:

x (t) = 0.5e−t/3 + 0.517(1 − e−t/3 )

c) Repeat the previous part if w2 changed to 100 kg · min−1

while w1 remained at 500 kg · min−1 .

x (t) = 0.5e−t/3 + 0.458(1 − e−t/3 )

d) What if both x1 and w2 changed to 0.6 and 100 kg ·

min−1 respectively?

x (t) = 0.5e−t/3 + 0.625(1 − e−t/3 )

ChE 551: Process Control Dr. Yazan Hussain

19

e) For all the above cases, plot x (t) and the normalized
response for x defined as:

x ( t ) − x (0)
x (t) =
x ( ∞ ) − x (0)

where x (∞) is the new steady state value, and x (0) is

the initial value of x.
0.7

d
0.6
x (t)

b
0.5
c

0.4
0 5 10 15 20 25
Time (min)
The normazlized curves can be found by calculating
x (∞) for each case. For example, for (b):
∞ ∞
e−
x (t → ∞) = 0.5 e−
+ 0.517(1 −  ) = 0.517

The plots for all three curves are shown below:

1
Normalized response

0.5

0
0 5 10 15 20 25
Time (min)
The interesting observation here is that the normalized
response has the same time behavior for all cases. This
is because the characteristic time (ρV/w) was kept
constant for all cases (= 3 s).

ChE 551: Process Control Dr. Yazan Hussain