CL 329: Process Control

Week 2

Sayantan Dutta
 Logistics of course
• From next week onwards, the attendance will be biometric.
• Please register for biometric attendance at the
corresponding office near LA 202.

• A couple of minutes of your time will save hours of time

from research for the TAs.
 Logistics of course
•For the Thursday Lecture, it will be good if you bring your
laptop with MATLAB installed.
 Learning objective for week 2
• Mathematical models, their classification, and building
them.

• Degree of Freedom analysis

• Refresher on MATLAB
 Mathematical models: Observation to Prediction
observation quantitative models prediction

5 Wikimedia
 Mathematical models: Observation to Prediction
observation quantitative models prediction

m1m2
F=G 2
r

6 Wikimedia
 Mathematical models: Observation to Prediction
observation quantitative models prediction

m1m2
F=G 2
r

7 Wikimedia
Some quotes on process modeling

Slide from Prof. Bharat Suthar

Some quotes on process modeling

Slide from Prof. Bharat Suthar

 Mathematical models in process control

Disturbance Unmeasured
variable Output

Process

Manipulated
variable Measured
Output

Controller

Mathematical models are necessary for establishing the relationship between

input and output of a process, essential to design a control system.
Example, we already discussed in class

Fout = Fh + Fc
Fout FhTh + FcTc
Tout Tout =
Fh + Fc
Fh Fc
Th Tc

Fh, Fc, Th, Tc Process Fout, Tout

 Types of Mathematical Models
• First Principle (White box) models: Models with geometric,
physical or chemical basis

• Data driven (Black box) models: Mathematical

representation of the observed data.

• Gray Box models.

 First Principle Models
•Conservation Equation:
Rate of
accumulation = Rate of
in flow
− Rate of
out flow
+ Generation-
Consumption
 First Principle Models
•Conservation Equation:
Rate of
accumulation = Rate of
in flow
− Rate of
out flow
+ Generation-
Consumption

i
Fi, CA

A→B

F, CA
 First Principle Models
•Conservation Equation:
Rate of
accumulation = Rate of
in flow
− Rate of
out flow
+ Generation-
Consumption

Conservation of Mass:
i
Fi, CA dV
= Fi − F
dt
A→B Conservation of species:
d(VCA) i
= FiCA − FCA − VrA(CA)
F, CA
dt
 First Principle Models
•Constitutive Equations: Physical/Chemical properties of
material
 First Principle Models
•Constitutive Equations: Physical/Chemical properties of
material
· dT ·
• Transport rates: Q = kA ; Q = UA(T − Tamb)
dx
 First Principle Models
•Constitutive Equations: Physical/Chemical properties of
material
· dT ·
• Transport rates: Q = kA ; Q = UA(T − Tamb)
dx
• Reaction rates: rA = kCA
 First Principle Models
•Constitutive Equations: Physical/Chemical properties of
material
· dT ·
• Transport rates: Q = kA ; Q = UA(T − Tamb)
dx
• Reaction rates: rA = kCA
• Phase equillibrium: PA = P0xA
 First Principle Models
•Constitutive Equations: Physical/Chemical properties of
material
· dT ·
• Transport rates: Q = kA ; Q = UA(T − Tamb)
dx
• Reaction rates: rA = kCA
• Phase equilibrium: PA = P0xA
2
an
• Equation of states : PV = nRT; (P + )(V − nb) = nRT
V2
 First Principle Models
•Constitutive Equations: Physical/Chemical properties of
material
· dT ·
• Transport rates: Q = kA ; Q = UA(T − Tamb)
dx
• Reaction rates: rA = kCA
• Phase equilibrium: PA = P0xA
2
an
• Equation of states : PV = nRT; (P + )(V − nb) = nRT
V2
• Material properties: σxx = Eϵxx
Is this a conservation or constitutive equation?

x component of Navier-Stokes equation.

Is this a conservation or constitutive equation?

x component of Navier-Stokes equation.

Both!!!!
Is this a conservation or constitutive equation?

Momentum conservation equation

Is this a conservation or constitutive equation?

Momentum conservation equation

Property of a incompressible
dvx dvy
[ dy dx ]
τyx = − μ + & . . . , ρ = const . Newtonian fluid:
Constitutive equation
First principle models: Disasdvantange

x y
Input Process Output

Often the physics/chemistry/biology of the process is not

known with precise quantitative details.
Data driven or Black box models
x y
Input Process Output
Data driven or Black box models
x y
Input Process Output
Data driven or Black box models
x y
Input Process Output

y = 2x + 0.1
Problem with black box models

The data looks linear

Problem with black box models

y = 1.2x − 0.28

The data with a linear fit

 Problem with black box models

Actual model:
2
y = x + 0.05

y = 1.2x − 0.28

The fit performs badly in the range outside where the model was fit.
 Problem with black box models

Actual model:
2
y = x + 0.05

y = 1.2x − 0.28

You shouldn’t design for a situation, which you haven’t measured with a black box model.
Neural networks are black box models

Dog
Cat
Lion
Bird

Set of pixels
 Degrees of freedom
x+y=4 x+y=4 x+y=4
x−y=2 x−y=2 2 variables,
1 equation,
2 variables, 2x − 2y = 4 Infinitely many solutions
2 equations, Repetitive equation
Solutions exactly known x = 3,y = 1
x+y=4
x−y=2
2x − 2y = 6
No possible values of x, y

2 variables,
3 equations,
 Degrees of freedom
x+y=4 x+y=4 x+y=4
x−y=2 x−y=2 2 variables,
1 equation,
2 variables, 2x − 2y = 4 Infinitely many solutions
2 equations, Repetitive equation
Solutions exactly known x = 3,y = 1
x+y=4
x−y=2
2x − 2y = 6
No possible values of x, y

2 variables,
3 equations,
 Degrees of freedom
Degrees of
freedom = Number of
variables
− Number of
Equations
 Degrees of freedom
x+y=4 x+y=4 x+y=4
x−y=2 x−y=2 2 variables,
1 equation,
2 variables, 2x − 2y = 4 Infinitely many solutions
2 equations, Repetitive equation
Solutions exactly known x = 3,y = 1 DOF=1
x+y=4
x−y=2
DOF=0

2x − 2y = 6
No possible values of x, y
DOF=-1
 Degrees of freedom

i
Fi, CA

A→B

F, CA

Consider the reaction to be first order.

Calculate the degrees of freedom.
 Degrees of freedom

dV
= Fi − F
dt
i
Fi, CA
d(VCA) i
= FiCA − FCA − kVCA
dt
A→B
i
Variables: Fi, CA, F, V, CA
Equations: 2
F, CA DOF: 5-2=3

Consider the reaction to be first order.

Calculate the degrees of freedom.
 Controller reduces degrees of freedom

Each controller brings one equation and reduces degrees of freedom by 1.

 Controller reduces degrees of freedom

dV
= Fi − F
dt
i
Fi, CA
d(VCA) i
= FiCA − FCA − kVCA
dt
A→B
F = kc(V − Vsp)

i
Variables: Fi, CA, F, V, CA
F, CA Equations: 3
DOF: 5-3=2

Consider you have a level controller.

Each controller brings one equation and reduces degrees of freedom by 1.
 Controller reduces degrees of freedom
dV
= Fi − F
dt
d(VCA) i
= FiCA − FCA − kVCA
dt
F = kc(V − Vsp)
i
Variables: Fi, CA, F, V, CA,Vsp
Equations: 3
DOF: 6-3=3

Each controller brings one equation and reduces degrees of freedom by 1.

However, the degree of freedom is taken back if in your design you plan to
change the set point.
 Why time is not considered in DOF analysis ?

Space (i.e. x, y, z) and time (t) are part of the reference frame.
They don’t vary as the part of the process.
Consider degree of freedom in other context (such as Thermodynamics/Stat Mech)
Practice problems
 Problem 1

• What are the state variables ?

FR
• What are the balances you should consider ?
• Develop the state model of the system.
• Perform a degree of freedom analysis.

• System of 2 CSTRs
• First order reaction with constant rate k.
• Flow rates are volumetric.
• F1 and F2 are proportional to respective
height of the tank
• Density of the liquid constant.
 Problem 1

• Propose a feedback control loop

FR
• What is the control
objective?
• Identify the CV, MV, DV
• What is the qualitative
control action ?
You want to control the quality of product in final • Perform a degree of freedom analysis.
stream. You have a gas chromatograph to measure the
conc of B in the product stream and a valve to control
the return flow rate .
 Problem 2

• Develop the state model of the system.

• Perform a degree of freedom analysis.

Consider a liquid ow system consisting of a sealed tank with non-

condensible gas above the liquid. The gas obeys the ideal gas law. A
constant amount of n moles of gas is present in the tank. Consider
that the temperature within the tank doesn’t uctuate. The outlet
ow through the pipe is proportional to the square root of the
pressure difference across pipes. The ambient pressure may
uctuate.
fl
fl
fl
fl
 Problem 2

• Consider qi and PA are constant.

• What is the degree of freedom of the
system ?
• Would it be possible to implement a
level controller anymore ?
Consider a liquid ow system consisting of a sealed tank with non-
condensible gas above the liquid. The gas obeys the ideal gas law. A • What does it physically mean ?
constant amount of n moles of gas is present in the tank. Consider
that the temperature within the tank doesn’t uctuate. The outlet
ow through the pipe is proportional to the square root of the
pressure difference across pipes. The ambient pressure may
uctuate.
fl
fl
fl
fl