Process Control CL 302

Prof. Bharat Suthar

Department of Chemical Engineering

IIT Bombay, Mumbai, India, 400076

Lecture 38
Chapter#18

bharat.k.suthar[at]iitb.ac.in
Landline: +91 (22) 2576 7243

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 1

Admin

Average 18.3
Quiz 5 Min 2
Max 49

Best Grades: Utkarsh Daga

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 2

Recap: Smith’s Predictor:

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 3

Recap: Smith’s Predictor Comparison

No time delay: (Black dash line)

1
❖ 𝐺𝑂𝐿 =
(5𝑠+1)(3𝑠+1)
❖ 𝐾𝑐 = 3.02, 𝜏𝐼 = 6.5 min

When process has time delay

𝑒 −2𝑠
𝐺𝑂𝐿 =
(5𝑠 + 1)(3𝑠 + 1)
❑ Conventional PI control
𝑒 −2𝑠
❖ 𝐺𝑂𝐿 =
(5𝑠+1)(3𝑠+1)
❖ 𝐾𝑐 = 1.23, 𝜏𝐼 = 7 min

❑ Smith predictor
❖ No plant model mismatch
➢ i.e. 𝐺 = 𝐺෨
❖ 𝐾𝑐 = 3.02, 𝜏𝐼 = 6.5 min
❖ We can use aggressive controller
setting as if the delay is not present.

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 4

Outline:

Multiloop and Multivariable Control

❑ Process Interactions
❖ SISO vs MIMO
❑ Control Loop Interactions
❖ Block diagram and stability

❑ Pairing of Controlled and Manipulated Variables

❑ Singular Value Analysis

❑ Tuning of Multiloop PID Control Systems

❑ Decoupling and Multivariable Control Strategies

❑ Strategies for Reducing Control Loop Interactions

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 5

Context:

❑ Up to now we have taken problems with

❖ 1 CV and 1 MV→ Single input single output (SISO)

❑ This chapter→ Multiple input multiple output (MIMO)

❖ 2 MV and 2 CV
❖ issues:
➢ Process interactions
➢ How to pair MV and CV?

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 6

Example: MIMO Processes
𝑅 𝑥𝐷
❑ Distillation column:
❖ Both reflux flow rate R or steam
flow rate S
❖ will affect both distillate
composition 𝑥𝐷 and bottoms
composition 𝑥𝐵 .
𝑆 𝑥𝐵

𝑥𝐷 𝐺𝑝11 𝐺𝑝12 𝑅
𝑥𝐵 =
𝐺𝑝21 𝐺𝑝22 𝑆

Wood and Berry Model

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 7
Example: MIMO Processes

❑ Flash tank: 𝐺 𝑃
❖ If 𝐺 ↑
➢ Direct effect 𝑃 ↓
➢ Indirect (slow effect) ℎ ↓
❖ If 𝐿 ↑
➢ Direct effect ℎ ↓
➢ Relatively small effect and
indirect effect 𝑃 ↓ ℎ
𝐿

𝑃 𝐺𝑝11 𝐺𝑝12 𝐺
=
ℎ 𝐺𝑝21 𝐺𝑝22 𝐿

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 8

Several Issues with MIMO Control

❑ If we use multiloop control then how to pair MV to CV?

❑ What MV-CV pairing will lead to best control?

❑ Will control loop interactions generated by the process variables cause

problems?
❖ Will it affect stability?
❖ If the process interactions are significant, even the best multiloop control
system may not provide satisfactory control.
➢ Then we will have to move to Multivariable control.

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 9

SISO vs MIMO

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 10

Outline

❑ Process Interactions
❖ SISO vs MIMO
❑ Control Loop Interactions
❖ Block diagram and stability
❑ Pairing of Controlled and Manipulated Variables
❑ Singular Value Analysis
❑ Tuning of Multiloop PID Control Systems
❑ Decoupling and Multivariable Control Strategies
❑ Strategies for Reducing Control Loop Interactions

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 11

Block Diagram Analysis of MIMO Process

❑ 2 × 2 control problem:
❖ 2 MV and 2 CV
❖ 4 TFs are expected.

Compact Vector matrix notation

𝐺𝑝 (𝑠) is the process transfer function matrix

𝐺𝑝 |𝑠=0 = 𝑲 process steady state gain
matrix

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 12

Example

❑ 2 × 2 control problem:
❖ 2 MV and 2 CV
❖ 4 TFs are expected.

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 13

Example Overall mass and component balances:
1. 𝑤 = 𝑤𝐴 + 𝑤𝐵
2. 𝑤𝑥 = 𝑤𝐴 𝑥𝐴ҧ + 𝑤𝐵 𝑥ҧ𝐵
❑ Expressing Eq1 in deviation variables:
❑ Whiteboard exercise
❖ 𝑤′ = 𝑤𝐴′ + 𝑤𝐵′
𝑊′ 𝑊′
❖ Hence = 1, = 1 (after LT)
𝑊𝐴′ 𝑊𝐵′
❑ Taking derivative of second eqn.
ഥ = 𝑤𝐴′ 𝑥𝐴ҧ + 𝑤𝐵′ 𝑥ҧ 𝐵
➢ 𝑤 ′ 𝑥ҧ + 𝑥 ′ 𝑤
❖ Replacing 𝑤 ′ = 𝑤𝐴′ + 𝑤𝐵′
➢ 𝑤𝐴′ + 𝑤𝐵′ 𝑥ҧ + 𝑥 ′ 𝑤
ഥ = 𝑤𝐴′ 𝑥𝐴ҧ + 𝑤𝐵′ 𝑥ҧ 𝐵
Variables: 1
➢ 𝑥′ = 𝑤𝐴′ 𝑥𝐴ҧ + 𝑤𝐵′ 𝑥ҧ 𝐵 − 𝑤𝐴′ + 𝑤𝐵′ 𝑥ҧ
𝑤, 𝑥, 𝑤𝐴 , 𝑤𝐵 ഥ
𝑤
𝑥ҧ 𝐴 −𝑥ҧ 𝑥ҧ 𝐵 −𝑥ҧ
➢ 𝑥 ′ = 𝑤𝐴′ ഥ
+ 𝑤𝐵′ ഥ
𝑤 𝑤
Deviation variables:
❖ Hence (after LT)
𝑤 ′ , 𝑥 ′ , 𝑤𝐴′ , 𝑤𝐵′
𝑋′ 𝑥ҧ 𝐴 −𝑥ҧ 𝑋 ′ 𝑥ҧ 𝐵 −𝑥ҧ
➢ = , ′ =
𝑊𝐴′ 𝑤ഥ 𝑊𝐵 𝑤ഥ
Four TFs are needed:
𝑤′ 𝑤′ 𝑥′ 𝑥′
1 1 𝑊 ′
𝑤𝐴′ 𝑤𝐵′ 𝑤𝐴′ 𝑤𝐵′ 𝑊′ = 𝑥𝐴ҧ − 𝑥ҧ 𝑥ҧ𝐵 − 𝑥ҧ 𝐴
𝑋′ 𝑊𝐵′
𝑤
ഥ 𝑤
ഥ
CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 14
Example

❑ Whiteboard exercise

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 15

Block Diagram: 1-1/2-2 controller pairing

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 16

Block Diagram: 1-2/2-1 controller pairing

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 17

 Process interactions can induce
undesirable interactions between the
control loops

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 18

Example: MIMO Processes

❑ Flash tank:

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 19

Hidden Feedback loop

Note that the initial change in U1 has two effects on Y1:

(1) a direct effect and
(2) an indirect effect via the control loop interactions.
Although it is instructive to view this dynamic behavior as a sequence of events, in practice
the process variables would change continuously and simultaneously.
CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 20
 Process Control CL 302
Prof. Bharat Suthar

Department of Chemical Engineering

IIT Bombay, Mumbai, India, 400076

Lecture 39
Chapter#18

bharat.k.suthar[at]iitb.ac.in
Landline: +91 (22) 2576 7243

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 21

Admin

DX Grade assigned for the the 9 students having attendance less than 40.

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 22

Outline

❑ Process Interactions
❖ SISO vs MIMO
❑ Control Loop Interactions
❖ Block diagram and stability
❑ Pairing of Controlled and Manipulated Variables
❑ Singular Value Analysis
❑ Tuning of Multiloop PID Control Systems
❑ Decoupling and Multivariable Control Strategies
❑ Strategies for Reducing Control Loop Interactions

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 23

 Example: MIMO Processes

❑ Distillation column:
❑ Wood and Berry Model:

❑ MV:
❖ reflux flow rate R
❖ steam flow rate S
❑ CV:
❖ Distillate composition 𝑥𝐷
➢ (methanol mol% in distillate)
❖ Bottoms composition 𝑥𝐵 .
➢(water mol% in bottom)

RK Wood, MW Berry: (Wood & Berry) Distillation column Model

https://era.library.ualberta.ca/items/73a4af83-1cb8-4e9c-9eae-7499869bf632/view/32791d45-79df-416e-b520-8ec02dc70fcf/MK17452.pdf

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 24

Recap

1
0.604 1 +
16.37𝑠

1
−0.127 1 +
14.46𝑠

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 25

Simulink Exercise

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 26

Closed Loop Stability

𝑌1
❑ Lets derive
𝑌𝑠𝑝1
❑ Onenote exercise

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 27

Derive 𝑌1 /𝑌𝑠𝑝1

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 28

Closed Loop Stability

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 29

Example:
❑ Q1: What will be the limits if the
interactions are absent:

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 30

Example:
❑ Q1: What will be the limits if the
interactions are absent:

❑ Char eq.
❖ 1 + 𝐺𝐶1 𝐺𝑝11 = 0
2
❖ 1 + 𝐾𝑐1 10𝑠+1 = 0
❖ 10𝑠 + (1 + 2𝐾𝑐1 ) = 0
❑ For stability coeff. >0
1
❖ 𝐾𝑐1 > − 2
❖ Similarly for other loop
1
❖ 𝐾𝑐2 > − 2
CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 31
Stability

❑ Stability limit when interactions are present?

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 32

Maple Exercise:

❑ Char. Eqn

❑ Coefficient of 𝑠 𝑛 where n=0,1,2,3,and 4 which should be all positive.

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 33

 Hyperbola Eqn.

http://jwilson.coe.uga.edu/EMT668/EMT668.Folders.F97/Wynne/exam%20part%201%28equation%29/hyperbolas.html

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 34

Stability requirement: coefficients>0

p1:=inequal(cf1>0, Kc2 = -11 .. 5, Kc1 = -11 .. 5)

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 35

Overlapping area where all coefficients>0

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 36

Stability of 1-2/2-1 system

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 37

Outline:

Multiloop and Multivariable Control

❑ Process Interactions
❖ SISO vs MIMO
❑ Control Loop Interactions
❖ Block diagram and stability

❑ Pairing of Controlled and Manipulated Variables

❖ Relative Gain Array (RGA)

❑ Tuning of Multiloop PID Control Systems

❑ Decoupling and Multivariable Control Strategies

❑ Strategies for Reducing Control Loop Interactions

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 38

❑ Syllabus Ends Here

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 39

Pairing of Controlled and Manipulated Variables

❑ Question to address:
❖ How to choose 1-1/2-2 vs 1-2/2-1 pairing?
❖ What happens when it becomes n×n system where physical insight is not
sufficient.

❑ Relative Gain Array

❑ Provides two types of useful information:

❖ Measure of process interactions
❖ Recommendation about best pairing of controlled and manipulated variables.

❑ Requires knowledge of steady-state gains but not process dynamics.

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 40

Relative Gain Array

❑ The relative gain λij between a controlled variable yi and a manipulated

variable uj is defined to be the dimensionless ratio of two steady-state
gains:

(∂yi/∂uj)y is evaluated with all of

the controlled variables except yi
held constant.

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 41

RGA Calculation:

❑ RGA for 2x2 process:

y1 = K11u1 + K12u2
y2 = K 21u1 + K 22u2

❑ For nxn process: where⊗denotes the Schur product (element by

element multiplication):
Λ = 𝐾⊗(𝐾 −1 )𝑇

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 42

Pairing recommendation:

❑ Recommendation: Pair the controlled and manipulated variables so

that corresponding relative gains are positive and as close to one as
possible.
❑ In general:
❖ Pairings which correspond to negative pairings should not be selected.
❖ Otherwise, choose the pairing which has 𝜆𝑖𝑗 closest to one.

❑ Examples:

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 43

RGA Examples

❑ Examples:  K11 K12   2 1.5

K =  =
 K 21 K 22  1.5 2 

 2.29 −1.29 
 Λ=  
 −1.29 2.29 
Recommended pairing is Y1
and U1, Y2 and U2.

❑ Examples:
 −2 1.5 0.64 0.36 
K =   Λ=  
1.5 2   0.36 0.64 
Recommended pairing is Y1 with U1 and Y2 with U2.

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 44

RGA Example:

u1 u2 u3 u4
y1  0.931 0.150 0.080 −0.164 
y2  −0.011 −0.429 0.286 1.154 
Λ =  
y3  −0.135 3.314 −0.270 −1.910 
 
y4  0.215 −2.030 0.900 1.919 

❑ u1→ y1
❑ u2 → y3
❑ u3 → y4
❑ u4 → y2

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 45

Outline:

Multiloop and Multivariable Control

❑ Process Interactions
❖ SISO vs MIMO
❑ Control Loop Interactions
❖ Block diagram and stability

❑ Pairing of Controlled and Manipulated Variables

❖ Relative Gain Array (RGA)

❑ Tuning of Multiloop PID Control Systems

❑ Decoupling and Multivariable Control Strategies

❑ Strategies for Reducing Control Loop Interactions

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 46

Syllabus ends here:

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 47

 Hyperbola Eqn.

http://jwilson.coe.uga.edu/EMT668/EMT668.Folders.F97/Wynne/exam%20part%201%28equation%29/hyperbolas.html

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 48

Stability: Recap

❑ Stability limit when interactions are present?

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 49

 Hyperbola Eqn.

http://jwilson.coe.uga.edu/EMT668/EMT668.Folders.F97/Wynne/exam%20part%201%28equation%29/hyperbolas.html

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 50

Relative Gain Array

❑ The relative gain λij between a controlled variable yi and a manipulated

variable uj is defined to be the dimensionless ratio of two steady-state
gains:

(∂yi/∂uj)y is evaluated with all of

the controlled variables except yi
held constant.

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 51

Pairing of Controlled and Manipulated Variables

❑ Question to address:
❖ How to choose 1-1/2-2 vs 1-2/2-1 pairing?
❖ What happens when it becomes n×n system where physical insight is not
sufficient.

❑ Relative Gain Array

❑ Provides two types of useful information:

❖ Measure of process interactions
❖ Recommendation about best pairing of controlled and manipulated variables.

❑ Requires knowledge of steady-state gains but not process dynamics.

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 52

Outline:

Multiloop and Multivariable Control

❑ Process Interactions
❖ SISO vs MIMO
❑ Control Loop Interactions
❖ Block diagram and stability

❑ Pairing of Controlled and Manipulated Variables

❖ Relative Gain Array (RGA)
➢ Dynamic considerations

❑ Tuning of Multiloop PID Control Systems

❑ Decoupling and Multivariable Control Strategies

❑ Strategies for Reducing Control Loop Interactions

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 53

Dynamic considerations

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 54

Undesirable Control Loop Interactions

❑ Dealing with Undesirable Control Loop Interactions

❖ "Detune" one or more FB controllers.
➢ Make it more conservative
❖ Select different manipulated or controlled variables.
➢ e.g., nonlinear functions of original variables
❖ Use a decoupling control scheme.
➢ ?
❖ Use some other type of multivariable control scheme.
➢ Next topic

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 55

Decoupling Control Systems

❑ Basic Idea: Use additional controllers to compensate for process

interactions and thus reduce control loop interactions

❑ Ideally, decoupling control allows setpoint changes to affect only the

desired controlled variables.

❑ Typically, decoupling controllers are designed using a simple

process model (e.g., a steady-state model or transfer function model)

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 56

Basic Idea of Decoupling Control Systems

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 57

Idea of FF+FB control system

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 58

Decoupling control systems

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 59

Decoupling control systems

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 60

Outline:

Multiloop and Multivariable Control

❑ Process Interactions
❖ SISO vs MIMO
❑ Control Loop Interactions
❖ Block diagram and stability

❑ Pairing of Controlled and Manipulated Variables

❖ Relative Gain Array (RGA)
➢ Dynamic considerations

❑ Tuning of Multiloop PID Control Systems

❑ Decoupling and Multivariable Control Strategies

❑ Strategies for Reducing Control Loop Interactions

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 61

Reference:

❑ Process Dynamics and Control (Seborg, Edgar, Mellichamp, Doyle III)

❑ Process Control Designing Processes and Control Systems for Dynamic
Performance: Thomas E. Marlin

CL 302 | Bharat Suthar | Process Control | IIT Bombay, Mumbai, India 62